Datasets:

timaeus

/

pile-pubmed_central

like 0

Follow

Timaeus 9

Introduction {#sec1-1} ============ Infliximab (IFX), a chimeric anti-TNFα antibody, is effective in inducing and maintaining remission in a considerable proportion of IBD patients refractory to any other treatments \[[@ref1],[@ref2]\]. However, 8-12% of adult and/or pediatric patients fail to respond to the induction regimen (known as primary non responders) and approximately 40% of patients who respond initially and achieve clinical remission inevitably lose response over time\[[@ref3],[@ref7]\]. Lack of response to IFX is a stable trait and suggests that the differences in response might be in part genetically determined. Considering the high cost and safety profile of this drug, genetic targeting of patients responding to this therapy is certainly of great interest \[[@ref8]\]. So far, limited candidate gene association studies with response to IFX have been reported \[[@ref9]-[@ref11]\]. Recently, a genome-wide association study (GWAS) in paediatric IBD patients has revealed that the 21q22.2/BRWDI loci were associated with primary non response \[[@ref12]\]. Furthermore, although TNFa gene is of great interest as a candidate gene for pharmacogenetic approaches few studies have been performed to date and some have led to contradictory results \[[@ref10],[@ref11],[@ref13]-[@ref15]\]. All anti-TNF agents share an IgG1 Fc fragment, but the contribution of the Fc portion to the response to treatment among currently used TNF blockers remains unknown. Receptors for IgG-Fc portion (FcR) are important regulatory molecules of inflammatory responses. FcR polymorphisms alter receptor function by enhancing or diminishing the affinity for immunoglobulins \[[@ref16]\]. Three major classes of FcR that are capable of binding IgG antibodies are recognised: FcγRΙ (CD64), FcγRΙΙ (CD32), and FcγRΙΙΙ (CD16). FcγRΙΙ and FcγRΙΙΙ have multiple isoforms (FcγRΙΙΙA/C and B; FcγRΙΙΙA and B) \[[@ref16]\]. The most frequent polymorphism of *FcγRΙΙΙA* is a point mutation affecting amino acids in codon 158 in the extracellular domain. This results in either a valine (V158) or a phenylalanine (F158) at this position. Recently, it has been reported that CD patients with *FcγRΙΙΙA* -158V/V genotype had a better biological and possibly better clinical response to IFX \[[@ref17]\]. However, further studies did not confirm this observation \[[@ref18]\]. The aim of this study was to assess whether the *TNF* and/ or *FcγRΙΙΙA* gene polymorphisms are genetic predictors of response to IFX, in a cohort of Greek patients with adult or paediatric onset of CD. Patients - Methods {#sec1-2} ================== Patients {#sec2-1} -------- We enrolled 106 consecutive patients with newly diagnosed CD attending the outpatient IBD Clinic at the 1^st^ Department of Gastroenterology, "Evangelismos" Hospital (79 adults) or the 1^st^ Department of Pediatrics, University Hospital of Athens "Aghia Sophia"(27 children). The diagnosis of CD was based on standard clinical, endoscopic, radiological, and histological criteria \[[@ref1],[@ref19]\]. Eligible patients should have inflammatory (luminal) disease and be naive to IFX. IFX was administered intravenously at a dose of 5mg/kg at weeks 0, 2, 6 and then every 8 weeks. Clinical and serological responses were assessed using the Harvey-Bradshaw Index (HBI) \[[@ref20]\] and the serum levels of C-reactive protein (CRP), respectively, at baseline (before the 1st infusion of IFX), the day before each subsequent IFX infusion and after 12 weeks of treatment. Ileocolonoscopy was performed by a single endoscopist (GJM) at baseline and after 12-20 weeks of therapy to assess mucosal healing. Any changes in endoscopic appearance compared to baseline endoscopy were classified in four categories \[[@ref21],[@ref22]\] \[[Table 1](#T1){ref-type="table"}\]. Patients were classified in accordance to response to IFX therapy as shown in [table 2](#T2){ref-type="table"}. The ethical committee of the participating hospitals approved the study. Research was carried out according to Helsinki Convention (1975) and written inform consent was obtained in advance from each patient. ###### Grading of endoscopic mucosal lesions \[[@ref21],[@ref22]\]  ###### Classification of the study population due to response to infliximab therapy  Genotyping {#sec2-2} ---------- Genomic DNA from whole blood containing EDTA was extracted using standard techniques (NucleoSpin Blood kit, Macherey-Nagel, Germany). All polymerase chain reactions (PCRs) were run under conditions previously described \[[@ref23]\]. Primer sequences for the gene polymorphism at --308 were forward 5′-GGG ACA CAC AAG CAT CAA GG-3′ and reverse 5′-GGG ACA CAC AAG CAT CAA GG-3′, for the polymorphism at −238 forward 5′-ATC TGG AGG AAG CGG TAG TG-3′ and reverse 5′-AGA AGA CCC CCC TCG GAA CC-3′. The PCR products were digested at 37 °C with NcoI to detect the SNP in the −308 gene allele and MspI to detect the polymorphism of the −238 nucleotide. The -857 C/T polymorphism was analyzed by allele-specific PCR method24 using the primers TNF857-C: 5′-aag gat aag ggc tca gag ag-3′, TNF857-N: 5′-cta cat ggc cct gtc ttc g-3′ and TNF857-M: 5′-t cta cat ggc cct gtc ttc a-3′. The --158V/F polymorphism of FcγRΙΙΙA gene was detected as described by Leppers-van de Straat et al \[[@ref25]\] using the primers 5′-CTG AAG ACA CAT TTT TACT CC CAA (A/C)-3′ and 5′-TCC AAA AGC CAC ACT CAA AGA C-3′. The PCR products were then subjected to 3% agarose-gel electrophoresis. "No target" controls were included in each PCR batch to ensure that reagents had not been contaminated. Statistical Analysis {#sec2-3} -------------------- Genotype frequencies were compared with the chi-square with Yate's correction using S-Plus (v. 6.2Insightful, Seattle, WA). Odds ratios (ORs) and 95 confidence intervals (CIs) were obtained with GraphPad (v. 3.00, GraphPad Software, San Diego, CA). The p values are all two-sided. Correction for multiple testing was not applied in this study. *P* values of \< 0.05 were considered to be significant. Results {#sec1-3} ======= Patient demographic and clinical characteristics are given in [Table 3](#T3){ref-type="table"}. There were 68 (64.15%) complete responders, 25 (23.58%) partial responders and 13 (12.26%) non responders to IFX in this study. There were no statistical differences in the mean age, gender, disease duration, location and behavior and smoking habits between complete or partial responders and primary non-responders. There was no disagreement between HBI scores and serum CRP levels. Although, the post-treatment CRP levels were significantly lower in complete responders compared to partial and non-responders, the decrease in CRP levels did not differ significantly between the three groups. Post-treatment CRP levels and mean HBI score were significantly lower in complete responders compared to pre-treatment values in contrast to partial and/or non-responders where the CRP levels and the mean HBI score did not differ significantly. ###### Demographic, clinical and biological characteristics of the study population  The -238 G/A, -308 G/A, and -857 C/T polymorphisms of the TNF gene and the -158 V/F polymorphism in the *FcγRΙΙΙA* gene were successfully determined in all subjects. The genotype distribution in complete, partial and non-responders were presented in [Table 4](#T4){ref-type="table"}. No significant difference was observed for the polymorphism tested. In addition, although there may be genetic differences in early (paediatric)-onset and late (adult)-onset CD we were unable to detect any such differences although the number of paediatric patients included in the current study did not allow firm conclusions. ###### Genotype frequency in complete responders, partial responders and non responders  In the present study, we could not correlate the decrease in serum CRP levels with the genotypes tested in any particular group of patients since in most of the cases serum CRP levels dropped by more than 25% after 12 weeks of treatment. However, no significant decrease in CRP was observed between the TNF genotypes tested. Regarding the -158 V/F polymorphism in the *FcγRΙΙΙA* gene, the relative decrease in serum CRP levels was greatest in VV homozygotes (78.15 ± 33.68%) and lowest in FF homozygotes (69.84 ± 28.7%) but this difference was not significant. Due to the small number of cases we did not stratify the genotype frequencies according to age. Discussion {#sec1-4} ========== The mechanism of IFX action in IBD seems to be multifactorial and the response to IFX is a complex phenomenon influenced by several parameters \[[@ref1]\]. Interestingly, a certain proportion of patients do not respond to IFX at all whereas a significant proportion will lose response over time \[[@ref3]-[@ref7]\]. This is the first Greek study aiming at identifying any significant associationbetween the -238 G/A, -308 G/A, and -857 C/T polymorphisms in the promoter region of the TNF gene and the -158V/F polymorphism in *FcγRΙΙΙA* gene and response to IFX in a cohort of adult and paediatric patients with CD and it was negative. Efficacy of IFX was assessed by clinical, serological and endoscopic parameters. Clinical response to IFX was evaluated using the HBI, which has been used in many clinical trials, is simple to use and has shown good correlation with the Crohn's Disease Activity Index (CDAI) \[[@ref26]\]. Serological evaluation of response to IFX was based on changes in serum levels of CRP, which has shown a good correlation with clinical activity and to a certain degree with endoscopic activity of CD \[[@ref27]\]. Finally, endoscopic activity of disease was assessed before and after IFX therapy using a simple description of healing of ulcerative and non ulcerative lesions \[[Table 1](#T1){ref-type="table"}\] as has been previously described \[[@ref21],[@ref22]\]. Endoscopic healing was assessed after 12-20 weeks of IFX treatment. It is conceivable that 12 weeks may be early to assess mucosal healing induced by biologic therapies \[[@ref27]\] but the vast majority of patients underwent endoscopy at least 16 weeks after initiation of IFX therapy (average time 17.6 weeks) and therefore it is unlikely that we have not obtained an objective view of the intestinal mucosal at follow up ileocolonoscopy. Regarding the *TNF* genotypes, our results are in agreement with Louis et al \[[@ref11]\] who did not find any significant difference between response groups when genotyped CD patients for the TNF -308G/A polymorphism and compared response rates after IFX treatment. The same results were reported by Mascheretti et al \[[@ref10]\] and Dideberg et al \[[@ref13]\]. Moreover, our results are in agreement with Tomita et al \[[@ref28]\] who reported no significant difference on *TNFa*, *FcgammaRIIA* and *FcgammaRIIIA* between responders and non responders 8 weeks after IFX treatment as well as with results of ACCENT I study where the relative decrease in serum CRP levels after IFX treatment was greatest in -158 VV homozygotes and lowest in FF homozygotes \[[@ref18]\]. In contrast, Louis et al \[[@ref17]\] observed a significant association between the -158V/F polymorphism in *FcγRΙΙΙA* and both the proportion of patients who had a drop in serum CRP levels after IFX treatment and the magnitude in decrease of serum CRP levels. This may account for the relatively small population of patients in our study, genetic differences in the studied populations and/or methodological differences between studies. Although it would be useful to genetically differentiate 'responders' from 'non-responders', there are not enough data on TNF polymorphisms in IBD and often only selected polymorphisms are genotyped. Small studies have shown possible associations between poor response to IFX and increasing mucosal levels of activated NF-kappaB, homozygosity for the polymorphism in exon 6 of TNFR2 (genotype Arg196Arg), positivity for perinuclear antineutrophil cytoplasmic antibodies and with the presence of increased numbers of activated lamina propia mononuclear cells producing interferon-gamma and TNFa \[[@ref29]\]. In conclusion, our study did not detect any associations between three TNFα gene polymorphisms or the -158 V/F polymorphism in the *FcγRΙΙΙA* gene and response to IFX in CD. However, in view of discrepant results in the literature large-scale pharmacogenetic studies in different populations, with similar baseline disease phenotypes and treatment protocols are needed to adequately estimate associations between genetic polymorphisms and treatment outcomes. Conflict of interest: None ^a^Evangelismos Hospital, ^b^Laboratory of Biology, School of Medicine, ^c^1^st^ Department of Pediatrics, School of Medicine, University of Athens, Greece

INTRODUCTION {#s1} ============ Hepatitis B virus (HBV) is still a major global health problem, with an estimated 257 million people worldwide that are chronically infected with HBV ([@B1]). HBV, together with duck hepatitis B virus (DHBV) and several other related animal viruses, belongs to the *Hepadnaviridae* family ([@B2]). The HBV virion is comprised of an outer envelope and an inner icosahedral nucleocapsid (NC) assembled by 240 copies of core protein (HBc) and packaged with a 3.2-kb partially double-stranded circular DNA genome ([@B3][@B4][@B8]). In addition to DNA-containing virions, a large amount of incomplete viral particles, such as hepatitis B surface antigen (HBsAg) particles, empty virions, and naked capsids, can also be released from cells in the process of virus replication ([@B9]). Subviral HBsAg particles are spherical or rodlike and are present in vast excess over virions in sera of CHB patients ([@B2]). Empty virions share the same structure as DNA-containing virions but are devoid of nucleic acids ([@B10][@B11][@B14]). Naked capsids, which exit cells via a route different from that of virions ([@B15][@B16][@B17]), have the same structure as NCs but are either empty or filled with viral RNA and immature viral DNA ([@B7], [@B11], [@B18][@B19][@B20]). In NC, pgRNA undergoes reverse transcription into minus-strand DNA, followed by plus-strand DNA synthesis ([@B2], [@B21][@B22][@B24]). Intracellular NCs can be packaged with viral nucleic acids at all levels of maturation, including pgRNA, nascent minus-strand DNA, minus-strand DNA-RNA hybrids, and relaxed circular DNA (RC DNA) or double-stranded linear DNA (DSL DNA) ([@B5], [@B7]). Only the NCs with relatively mature viral DNA (RC or DSL DNA) are enveloped and secreted as virions. HBV replicating cells can release empty core particles assembled from HBc proteins and NCs that contain various species of replicative intermediate nucleic acids into the culture supernatant. However, while free naked capsids could be readily detected *in vitro* ([@B7], [@B11], [@B18][@B19][@B20]), they are hardly found in the blood of HBV-infected patients ([@B17], [@B25], [@B26]). Although extracellular HBV RNA was detected in both *in vitro* cell culture systems and in clinical serum samples, its origin and composition remain controversial. It was proposed that extracellular HBV RNA represents pgRNA localized in virions ([@B27]). However, HBV spliced RNA and HBx RNA were also detected in culture supernatant of HBV stably replicating cells as well as in sera of CHB patients ([@B28], [@B29]). In addition, extracellular HBV RNA was also suggested to originate from damaged liver cells ([@B30]), naked capsids, or exosomes ([@B11], [@B29]). Hence, these extracellular RNA molecules have never been conclusively characterized. Here, we demonstrate that extracellular HBV RNAs are heterogeneous in length, ranging from full-length pgRNA (3.5 kilonucleotides \[knt\]) to RNA fragments with merely several hundred nucleotides. These RNA molecules represent 3′ receding pgRNA fragments that have not been completely reverse transcribed to DNA and pgRNA fragments hydrolyzed by the RNase H domain of polymerase in the process of viral replication. More importantly, extracellular HBV RNAs are localized in naked capsids and in virions in culture supernatants of HBV replicating cells and also circulate as CACs and virions in blood of hepatitis B patients. RESULTS {#s2} ======= Extracellular HBV RNAs are heterogeneous in length and predominantly integral to naked capsids instead of virions in HepAD38 cell culture supernatant. {#s2.1} ------------------------------------------------------------------------------------------------------------------------------------------------------ To ascertain the origin of extracellular HBV RNA, we first examined viral particles prepared from culture medium of an *in vitro* HBV stably transduced cell line. A human hepatoma HepAD38 cell line was used in this study, as it sustains vigorous HBV replication under the control of a tetracycline-repressible cytomegalovirus (CMV) promoter ([@B31]). Total viral particles were concentrated and centrifuged over a 10% to 60% (wt/wt) sucrose gradient. Most of the subviral HBsAg particles, virions, and empty virions were detected between fractions 9 to 14 ([Fig. 1A](#F1){ref-type="fig"}, upper and middle). Naked capsids, detected only by anti-HBcAg and not by anti-HBsAg antibodies, settled in fractions 5 to 8 ([Fig. 1A](#F1){ref-type="fig"}, middle and lower). The majority of viral nucleic acids were detected in fractions between 4 and 11 ([Fig. 1B](#F1){ref-type="fig"}, upper), which coincided with the fractions containing virions (fractions 9 to 11), naked capsids (fractions 4 to 7), and the mixture of these particles (fraction 8). Consistent with previous observations, HBV virions are packed with mature viral DNA (RC or DSL DNA), while naked capsids contain both immature single-stranded DNA (SS DNA) and mature viral DNA ([Fig. 1B](#F1){ref-type="fig"}, upper). Moreover, Northern blot results showed that most of the HBV RNA was detected in the naked capsids ([Fig. 1B](#F1){ref-type="fig"}, lower, fractions 4 to 7), whereas only a very small amount was associated with virions ([Fig. 1B](#F1){ref-type="fig"}, lower, fractions 9 to 11). HBV RNA detected in naked capsids ranged from the full length of pgRNA down to a few hundred nucleotides (shorter than the HBx mRNA \[0.7 knt\]). Moreover, RNA molecules within virions were much shorter than those within naked capsids. We excluded the possibility of artifacts generated by the SDS-proteinase K extraction method, as a similar RNA blot pattern was obtained using a TRIzol reagent to extract both intracellular nucleocapsid-associated and extracellular HBV RNA (not shown). Furthermore, quantification of viral RNA extracted by either the SDS-proteinase K method or TRIzol reagent produced a very similar copy number, except that the TRIzol reagent is known to preferentially extract RNA rather than DNA (not shown). Moreover, the RNA signal detected by Northern blotting could not be attributed to DNA fragments generated by DNase I treatment, which would reduce DNA to below the detection limit of the hybridization method (not shown). Furthermore, the RNA signal could be completely removed by an additional RNase A treatment (not shown). {#F1} To confirm the above-described results and to better separate naked capsids from HBV virions, isopycnic CsCl gradient ultracentrifugation was employed. Naked capsids were observed mainly in fractions 5 to 7, with densities ranging from 1.33 to 1.34 g/cm^3^ ([Fig. 2A](#F2){ref-type="fig"}). The smearing bands of naked capsids were likely caused by high concentrations of CsCl salt, as fractionation of naked capsids in a 1.18-g/cm^3^ CsCl solution produced single bands. Virions, detected by both anti-HBcAg and anti-HBsAg antibodies ([Fig. 2A](#F2){ref-type="fig"}, upper and middle), were packaged with viral DNA ([Fig. 2A](#F2){ref-type="fig"}, lower) and settled in fractions 13 to 15, with densities ranging from 1.23 to 1.25 g/cm^3^. In agreement with the results shown in [Fig. 1](#F1){ref-type="fig"}, HBV virions contained only the mature viral DNA (RC or DSL DNA), while naked capsids contained viral DNA replicative intermediates that ranged from the nascent minus-strand DNA to mature viral DNA ([Fig. 2B](#F2){ref-type="fig"} and [C](#F2){ref-type="fig"}). The lengths of viral minus- and plus-strand DNA in naked capsids and virions were determined by alkaline agarose gel electrophoresis analysis, a condition where denatured single-stranded DNA molecules migrate according to their lengths. In contrast to the complete minus- and mostly complete plus-strand DNA (closed to 3.2 knt) in virions, in naked capsids the minus-strand DNA and the plus-strand DNA can be both complete and incomplete (shorter than 3.2 knt) ([Fig. 2D](#F2){ref-type="fig"} and [E](#F2){ref-type="fig"}). Moreover, the length of HBV RNAs within naked capsids still ranged from 3.5 knt of pgRNA to shorter than the 0.7 knt of HBx mRNA. Full-length pgRNA accounted for only 10% of total RNA signal detected by Northern blotting (quantified from gray density of bands shown in [Fig. 2F](#F2){ref-type="fig"}). In contrast, HBV RNA species in virions are relatively shorter and barely detectable. In addition, we also determined viral DNA and RNA copy numbers in pooled naked capsids (fractions 3 to 7) and virions (fractions 10 to 21) by quantitative PCR. Quantification results showed that viral DNA in naked capsids and in virions accounted for about 60% and 40%, respectively, of total viral DNA signal in the HepAD38 cell culture supernatant ([Fig. 2G](#F2){ref-type="fig"}). More importantly, 84% of the HBV RNA was associated with naked capsids, while merely 16% was detected within virions ([Fig. 2G](#F2){ref-type="fig"}). Additionally, the DNA/RNA ratio was 11 in virions and 3 in naked capsids ([Fig. 2H](#F2){ref-type="fig"}), suggesting that more HBV RNA is present in naked capsids. {ref-type="fig"}).](zjv0241840640002){#F2} Extracellular HBV RNAs and immature viral DNA are detected in sera from CHB patients. {#s2.2} ------------------------------------------------------------------------------------- Employing the HepAD38 cell culture system, we demonstrated the presence of extracellular HBV RNAs and immature and mature viral DNA packaged in both the naked capsids and virions. Interestingly, Southern blot analyses showed that SS DNA could also be observed in serum samples from some CHB patients. We speculated that SS DNA in circulation would be carried by capsid particles that were released by HBV-infected hepatocytes into patients' bloodstreams. However, we reasoned that due to strong immunogenicity of naked capsids ([@B32], [@B33]), it would be difficult to detect them as free particles; rather, they would form complexes with specific anti-HBcAg antibodies and therefore circulate as antigen-antibody complexes ([@B25], [@B32][@B33][@B34]). To entertain this possibility, we then used protein A/G agarose beads to pull down the immune complexes. Forty-five serum samples obtained from CHB patients, with HBV DNA titers higher than 10^7^ IU per ml, were examined for the presence of particles containing SS DNA by a combination of protein A/G agarose bead pulldown assay and Southern blot analysis ([Fig. 3A](#F3){ref-type="fig"} and [B](#F3){ref-type="fig"}). SS DNA was detected, albeit to a different extent, in 34 serum samples ([Fig. 3A](#F3){ref-type="fig"} and [B](#F3){ref-type="fig"}, upper). The particles containing SS DNA were pulled down by protein A/G agarose beads from 11 out of the 34 samples ([Fig. 3A](#F3){ref-type="fig"} and [B](#F3){ref-type="fig"}, lower). Patient sera negative for SS DNA (patients 37, 38, 14, and 35) or positive for SS DNA (patients 17, 21, 42, and 44), as determined by the protein A/G agarose bead pulldown experiments, were selected for further studies ([Fig. 3C](#F3){ref-type="fig"}). {#F3} Northern blot analyses showed that HBV RNA was only detected in serum samples from patients 17, 21, and 42 ([Fig. 3D](#F3){ref-type="fig"}). Moreover, total viral DNA was analyzed by Southern blotting, and SS DNA was readily observed in serum samples from patients 17, 21, and 42 ([Fig. 3E](#F3){ref-type="fig"}). We also analyzed the lengths of DNA minus and plus strands in patients' sera. Despite the finding that most minus-strand DNA was complete, a small amount of viral DNA (that of patients 38, 35, 17, 21, and 42) was shorter than 3.2 knt ([Fig. 3F](#F3){ref-type="fig"}). Compared with viral minus-strand DNA, the length of plus-strand DNA, particularly in sera from patients 17, 21, and 42, was more variable, ranging from shorter than 2 knt to ∼3.2 knt ([Fig. 3G](#F3){ref-type="fig"}). Naked capsids form CACs with anti-HBcAg antibody in blood of CHB patients. {#s2.3} -------------------------------------------------------------------------- We showed that particles containing SS DNA were present in CHB patients' sera. To further examine these particles, we used CsCl density gradient centrifugation to fractionate a serum mixture from patients 37, 38, 14, and 35. In agreement with our earlier results ([Fig. 2A](#F2){ref-type="fig"}, lower, fractions 13 to 15, and B) and previous reports, HBV virions, with the characteristic mature viral DNA (RC or DSL DNA), were detected in fractions 12 to 14 with densities between 1.26 and 1.29 g/cm^3^ ([Fig. 4A](#F4){ref-type="fig"}) ([@B2]). Careful inspection of the blots revealed that SS DNA could be detected, albeit at very low level, in fractions 8 and 9, with densities from 1.33 to 1.34 g/cm^3^, and in fractions 18 to 21, with densities from 1.20 to 1.23 g/cm^3^ ([Fig. 4A](#F4){ref-type="fig"}). In contrast, CsCl density gradient separation of viral particles from serum of patient 17 showed a mixture of mature and immature viral DNA species. As SS DNA was detected at densities ranging from 1.37 to 1.20 g/cm^3^ ([Fig. 4B](#F4){ref-type="fig"}), no distinct viral DNA (mature RC or DSL DNA) specific to virions could be identified at densities between 1.27 and 1.29 g/cm^3^. Similar results were obtained using CsCl density gradient fractionation of sera from patient 21 (not shown) and patient 46 ([Fig. 4E](#F4){ref-type="fig"}). {#F4} We hypothesized that naked capsids could be released into blood circulation of CHB patients but were bound to specific antibodies. As SS DNA was detected in both high- and lower-density regions in CsCl gradient ([Fig. 4B](#F4){ref-type="fig"} and [E](#F4){ref-type="fig"}), we envisaged that the binding with specific antibodies led to a change of capsids' buoyant density. To test this, anti-HBcAg antibody was mixed with HepAD38 cell culture supernatant to mimic the postulated CACs in serum samples. The results demonstrated that in contrast to SS DNA from naked capsids, distributed to three fractions at densities between 1.33 and 1.34 g/cm^3^ ([Fig. 2A](#F2){ref-type="fig"}, lower, and B), the mixture of naked capsids and CACs (SS DNA) was distributed more widely and could be detected in the lower density region (1.25 to 1.32 g/cm^3^) ([Fig. 4C](#F4){ref-type="fig"}, fractions 11 to 16). Similarly, intracellular capsids from HepAD38 cells were incubated with anti-HBcAg antibody, and a density shift of CACs to a lower-density region was also observed (not shown). To further confirm the lower density of CACs, NCs in virions secreted to HepAD38 cell culture supernatant were treated with NP-40 and mixed with anti-HBcAg antibody. CsCl fractionation showed that naked capsids and virion-derived NCs have become a homogenous mixture banding at densities from 1.37 to 1.27 g/cm^3^ ([Fig. 4D](#F4){ref-type="fig"}). Likewise, virion-derived NCs, obtained by treatment of serum sample from patient 46 with NP-40 bound with antibody, further formed new homogeneous CACs that settled at densities between 1.23 and 1.27 g/cm^3^ ([Fig. 4E](#F4){ref-type="fig"} versus F). However, NP-40 treatment alone did not produce a homogeneous mixture of naked capsids and virion-derived NCs, as these two particles still settled at distinct density regions with their characteristic viral DNA content ([Fig. 4G](#F4){ref-type="fig"}). On the other hand, DNA molecules in the two types of capsids still banded at densities between 1.38 and 1.31 g/cm^3^, further confirming that CACs have relatively lighter density ([Fig. 4G](#F4){ref-type="fig"}). Alternatively, the appearance of a homogenous mixture of virion-derived NCs and naked capsids ([Fig. 4D](#F4){ref-type="fig"} and [F](#F4){ref-type="fig"}) suggests the formation of higher-order antibody-mediated complexes of capsids. For instance, the complexes might not represent individual antibody-coated capsid particles but rather big CACs consisting of several capsid particles interconnected by antibodies. To verify whether intercapsid immune complexes exist, anti-HBcAg antibody was added to the purified HBV capsids expressed by Escherichia coli, and this mixture was examined by an electron microscope. E. coli-derived capsids were scattered as separate, distinct particles ([Fig. 5A](#F5){ref-type="fig"}). However, addition of antibody caused capsids to aggregate into clusters, making them too thick to be properly stained ([Fig. 5B](#F5){ref-type="fig"}). Despite this, a few capsids, which might not have been bound by antibodies or might have been associated with antibodies but did not form intercapsid antibody complexes, could be observed by electron microscopy (EM) ([Fig. 5B](#F5){ref-type="fig"}). {#F5} We then examined CACs in serum samples from CHB patients by EM. Sera from patients 11, 17, 21, 22, 23, 27, 28, 30, and 41, positive for SS DNA, were combined. Serum mixtures, with diminished HBsAg particles by centrifugation through a 20% and 45% (wt/wt) sucrose cushion, were examined by EM. The 27-nm capsid particles or CACs were visible ([Fig. 5C](#F5){ref-type="fig"}, arrow) along with the 42-nm HBV virions ([Fig. 5C](#F5){ref-type="fig"}, arrowheads) and the 22-nm spheres and rods of residual HBsAg particles (not indicated). However, the picture was not clear enough for us to conclusively determine if capsids were connected by or bound with antibodies, as described for unrelated virus in *in vitro* experiments ([@B35]). In addition, it is possible that some of the CACs are not visible by EM, as the complexes maybe too thick to gain clear contrast between lightly and heavily stained areas ([Fig. 5B](#F5){ref-type="fig"}). Lastly, CACs might be heterogeneous, having different molecular sizes and isoelectric points (pI) in hepatitis B patients' blood circulation. *In vitro* binding of naked capsids derived from HepAD38 cell culture supernatant with anti-HBcAg antibody changed their electrophoretic behavior and made them unable to enter the TAE-agarose gel ([Fig. 6A](#F6){ref-type="fig"}). Moreover, viral particles from sera of patients 0, 37, 38, 14, 35, 17, 21, 42, and 44 could not enter agarose gels prepared in TAE buffer. However, in buffer with higher pH value (10 mM NaCHO~3~, 3 mM Na~2~CO~3~, pH 9.4), they appeared as smearing bands on blots ([Fig. 6B](#F6){ref-type="fig"} and [C](#F6){ref-type="fig"}). Hence, the irregular electrophoretic behavior of these viral particles may result from changes in molecular size and/or pI value of capsid particles (pI  4.4) following their association with specific immunoglobulin G (or other types of antibodies) having different pI values (pI of human IgG may range from 6.5 to 9.5) ([@B36][@B37][@B39]). {#F6} Circulating HBV RNAs are of heterogeneous lengths and associated with CACs and virions in hepatitis B patient's plasma. {#s2.4} ----------------------------------------------------------------------------------------------------------------------- To characterize HBV RNAs circulating in CHB patients' sera, a plasma sample from patient 0 was studied. Similar to results obtained for patients 17, 21, and 46 ([Fig. 4B](#F4){ref-type="fig"} and [E](#F4){ref-type="fig"} and not shown), viral DNA in the plasma sample of patient 0 was detected in a broad density range in CsCl gradient and no distinct bands specific to HBV virions or naked capsids could be identified, indicating the presence of a mixture of virions and CACs ([Fig. 7A](#F7){ref-type="fig"}). {#F7} Furthermore, viral particles were pelleted through a 20% sucrose cushion and separated in a sucrose gradient. HBsAg was detected in fractions 5 to 14, peaking at fraction 11. The PreS1 antigen was found in fractions 5 to 12 with the peak at fractions 7 and 10, indicating its presence in HBsAg particles and HBV virions ([Fig. 7B](#F7){ref-type="fig"}, upper). Viral DNA, representing a combination of both mature and immature viral DNA, was detected in fractions 4 to 9 ([Fig. 7B](#F7){ref-type="fig"}, middle), suggesting the localization of CACs and virions in these fractions. HBV RNA was detected between fractions 5 and 7 and appeared in the same peak as viral DNA ([Fig. 7B](#F7){ref-type="fig"}, lower), indicating that HBV RNA is incorporated in the same viral particles as viral DNA. Therefore, circulating HBV RNA may be localized within CACs and/or virions. To better characterize HBV RNA in CACs and virions, plasma sample from patient 0 was centrifuged through a 20% sucrose cushion and pellets were fractionated in a homogenous CsCl solution (1.18 g/cm^3^) as previously described ([@B8]). However, possibly due to a tendency of capsid particles to aggregate and stick to the wall of the centrifugation tube and the low density of the initial CsCl solution ([@B8], [@B40]), only mature DNA species from virions were detected in densities ranging from 1.22 to 1.24 g/cm^3^ ([Fig. 7C](#F7){ref-type="fig"}, upper). Northern blot analyses demonstrated that the lengths of virion-associated HBV RNAs were approximately several hundred nucleotides ([Fig. 7C](#F7){ref-type="fig"}, lower). Virion-associated RNAs were unlikely to be contaminated by CAC-associated HBV RNAs, since the immature SS DNA could not be observed even after a long exposure of X ray film. Moreover, RNA molecules would have been longer if there were CAC contamination ([Fig. 7D](#F7){ref-type="fig"}, lower). Viral nucleic acids in pellets recovered from the centrifugation tube sidewalls could be readily detected on Northern ([Fig. 7C](#F7){ref-type="fig"}, lower, lane P) or Southern ([Fig. 7C](#F7){ref-type="fig"}, upper, lane P) blots using plus-strand-specific rather than minus-strand-specific riboprobe. To analyze viral nucleic acids in CACs, concentrated plasma sample was separated in a higher CsCl density gradient (1.18 g/cm^3^ and 1.25 g/cm^3^). Both mature and immature viral DNA species were only detected in fractions with densities from 1.21 to 1.26 g/cm^3^ ([Fig. 7D](#F7){ref-type="fig"}, upper), indicating the presence of a mixture of HBV virions and CACs. Viral RNAs were detected and ranged in length from a little shorter than the full-length pgRNA to a few hundred nucleotides ([Fig. 7D](#F7){ref-type="fig"}, lower). Compared to virion-associated RNAs ([Fig. 7C](#F7){ref-type="fig"}, lower), HBV RNA species detected in the mixture of CACs and virions were longer, with the longer RNA molecules possibly being associated with CACs. Extracellular HBV RNAs could serve as templates for synthesis of viral DNA. {#s2.5} --------------------------------------------------------------------------- Intracellular NCs are known to contain viral nucleic acids in all steps of DHBV DNA synthesis, including pgRNA, nascent minus-strand DNA, SS DNA, and RC DNA or DSL DNA ([@B5]). Our results showed that naked capsids contained almost the same DNA replicative intermediates as intracellular NCs ([Fig. 1B](#F1){ref-type="fig"} and [2B](#F2){ref-type="fig"}) ([@B7], [@B11]). We also demonstrated that extracellular HBV RNAs within the naked capsids, CACs, and virions were heterogeneous in length ([Fig. 1B](#F1){ref-type="fig"}, lower, [2F](#F2){ref-type="fig"}, and [7C](#F7){ref-type="fig"} and [D](#F7){ref-type="fig"}). In the presence of deoxynucleoside triphosphates (dNTPs), viral RNA could be degraded and reverse transcribed into minus-strand DNA by the endogenous polymerase *in vitro* ([@B5], [@B41], [@B42]). Also, incomplete plus-strand DNA with a gap of about 600 to 2,100 bases could be extended by endogenous polymerase ([@B43], [@B44]). Based on these results, we wished to examine whether extracellular HBV RNAs could serve as RNA templates for viral DNA synthesis and be degraded by polymerase in the process. As shown in [Fig. 8](#F8){ref-type="fig"}, endogenous polymerase assay (EPA) treatment of extracellular viral particles from either culture supernatant of HepAD38 cells or plasma sample from patients led to DNA minus ([Fig. 8A](#F8){ref-type="fig"} and [C](#F8){ref-type="fig"})- and plus ([Fig. 8B](#F8){ref-type="fig"} and [D](#F8){ref-type="fig"})-strand extension and, more importantly, HBV RNA signal reduction ([Fig. 8E](#F8){ref-type="fig"}, lane 4 versus 6 and lane 8 versus 10). The apparent low efficiency of EPA reaction might have been due to our hybridization method, which detected both extended and unextended DNA strands rather than detecting only newly extended DNA. {#F8} In the process of HBV DNA replication, prior to minus-strand DNA synthesis, capsid-associated RNA is the full-length pgRNA. Upon transfer of viral polymerase-DNA primer to the 3′ DR1 region of pgRNA and cleavage of the 3′ epsilon loop RNA (a 3.2-knt pgRNA fragment remained), minus-strand DNA synthesis initiates and the pgRNA template is continuously cleaved from 3′ to 5′ by RNase H activity of viral polymerase. Consequently, from the initiation to the completion of minus-strand DNA synthesis, there will be a series of pgRNA fragments with receding 3′ ends ranging from 3.2 knt to 18 nt of the 5′ cap RNA primer ([@B2], [@B21][@B22][@B24]), representing the RNA templates that have not yet been reverse transcribed into minus-strand DNA. In addition to pgRNA with receding 3′ ends, there are also short RNA fragments arising from intermittent nicks by the RNase H domain of polymerase. Therefore, we used RNA probes spanning the HBV genome to map whether these RNA molecules are present in extracellular naked capsids and virions. Five probes that spanned the HBV genome, except for the overlapping region between the 5′ end of pgRNA and the RNA cleavage site (nt 1818 to 1930), were prepared to map the extracellular HBV RNAs from HepAD38 cell culture supernatant ([Fig. 9A](#F9){ref-type="fig"}). Intracellular nucleocapsid-associated HBV RNA from HepAD38 cells was used as a reference. As the probes moved from the 5′ end to 3′ end of pgRNA, especially for probes 1 to 4, RNA bands shifted from a wider range, including both short and long RNA species, to a narrower range, close to full-length pgRNA, with fewer RNA species detected ([Fig. 9A](#F9){ref-type="fig"}, upper, lanes 2, 5, 8, 11, 14, and 17). Similarly, with the probes moving from the 5′ end to the 3′ end of pgRNA, a stronger intensity band representing extracellular HBV RNAs detected by each probe, especially for probes 1 to 4, was also shifting toward a longer RNA migration region ([Fig. 9A](#F9){ref-type="fig"}, upper, lanes 3, 6, 9, 12, 15, and 18). It should be noted that the shifting pattern was more apparent when RNAs were detected with probes 1 to 4 but not with probe 5. It is possible that the reverse transcription speed is relatively quicker in the initial step (from the 3′ end of pgRNA, which overlaps the probe 5 sequence), and as a result, fewer pgRNA fragments will harbor RNA sequence for probe 5. Also, a short RNA species from either intracellular nucelocapsids or naked capsids and virions migrated faster than 0.7 knt and could be detected by all probes ([Fig. 9A](#F9){ref-type="fig"}, upper, lanes 2, 3, 5, 6, 8, 9, 11, 12, 14, 15, 17, and 18). These RNA molecules likely represent the pgRNA fragments that have been hydrolyzed by the RNase H domain of viral polymerase (including the 3′ epsilon loop RNA cleaved by polymerase in the reverse transcription step) ([@B24]). Collectively, as predicted, longer extracellular HBV RNA species that migrated slower and closer to the position of pgRNA had longer 3′ ends, the shorter viral RNA molecules that migrated faster had relatively shorter 3′ ends, and the RNA species detected by all probes may represent products of pgRNA hydrolysis. ). (B) Identification of 3′ ends of extracellular HBV RNAs. 3′ Ends of extracellular HBV RNAs were identified by the 3′ RACE method using different HBV-specific anchor primers (the same 5′ primers used for generating templates for producing riboprobes used in panel A, lower). Identified 3′ ends were numbered as described above, and numbers in parentheses indicate the amount of clones with the same 3′ ends. The asterisk indicates unknown nucleic acid copurified with intracellular capsid-associated viral RNA by TRIzol reagent. FL, full-length; Cap, 5′ cap of pregenomic RNA; pA, the polyadenylation site; An, poly(A) tail.](zjv0241840640009){#F9} These results were further confirmed by employing a 3′ rapid amplification of cDNA ends (RACE) method. Various 3′ ends spanning the HBV genome were identified ([Fig. 9B](#F9){ref-type="fig"}), validating the presence of 3′ receding RNA and the heterogeneous nature of extracellular HBV RNAs. EPA treatment clearly demonstrated that extracellular HBV RNAs could be used as templates for DNA synthesis, and the presence of 3′ receding-end pgRNA fragments further confirmed not only the existence but also the use of such molecules as templates for viral DNA synthesis. Therefore, just like the viral RNA counterpart within intracellular NCs, extracellular HBV RNA molecules represent the RNA molecules generated in the process of viral DNA replication. ETV reduces viral DNA level but increases extracellular HBV RNA level in naked capsids and virions *in vitro*. {#s2.6} -------------------------------------------------------------------------------------------------------------- Entecavir (ETV), widely used in anti-HBV therapy, is a deoxyguanosine analog that blocks the reverse transcription and plus-strand DNA synthesis steps in the HBV DNA replication process ([@B45][@B46][@B47]). Treatment of CHB patients with nucleos(t)ide analogs (NAs), including entecavir, efficiently reduces the level of serum viral DNA but at the same time increases circulating HBV RNA levels ([@B28], [@B48][@B49][@B52]). We examined the effect of entecavir on the levels of both intracellular and extracellular viral nucleic acids in HepAD38 cell culture. Total viral RNA level remained unchanged or marginally increased upon entecavir treatment ([Fig. 10A](#F10){ref-type="fig"}), and the intracellular capsid-associated viral RNA level was increased ([Fig. 10B](#F10){ref-type="fig"}, upper). In contrast and as expected, the intracellular capsid-associated viral DNA level was decreased ([Fig. 10B](#F10){ref-type="fig"}, lower). Similarly, extracellular viral DNA synthesis was significantly inhibited, while viral RNA was increased ([Fig. 10C](#F10){ref-type="fig"} and [D](#F10){ref-type="fig"}). Quantitative results showed that entecavir suppressed extracellular viral DNA to about one-tenth but at the same time increased viral RNA by about twofold the level for the untreated group ([Fig. 10E](#F10){ref-type="fig"}). {ref-type="fig"}. Viral particles in each fraction were separated by native agarose gel electrophoresis, followed by immunoblotting with anti-HBcAg antibody. Viral DNA and RNA were extracted and subjected to Southern or Northern blot analyses.](zjv0241840640010){#F10} Since viral DNA and RNA were enclosed in both naked capsids and virions, CsCl density gradient was used to separate these particles and to further study the antiviral effect of entecavir. As shown in [Fig. 10](#F10){ref-type="fig"}, DNA-containing naked capsids were detected in fractions 6 to 11 and virions in fractions 15 to 24 ([Fig. 10F](#F10){ref-type="fig"}). Entecavir effectively reduced viral DNA ([Fig. 10G](#F10){ref-type="fig"}, fractions 6 to 10 and 15 to 17; this was also seen in a longer exposure of [Fig. 10G](#F10){ref-type="fig"} \[not shown\]) but increased viral RNA content mainly in naked capsids ([Fig. 10H](#F10){ref-type="fig"}, fractions 6 to 9). Moreover, the increase in RNA content within naked capsids led to an increased density of naked capsids ([Fig. 10F](#F10){ref-type="fig"}, fractions 6 and 11, lower, versus fractions 6 and 11, upper). Interestingly, entecavir seemed to reduce HBcAg signal within virions (i.e., empty virions) ([Fig. 10F](#F10){ref-type="fig"}, fractions 15 to 21, upper, versus fractions 15 to 21, lower) while increasing the egress of naked capsids from HepAD38 cells (data not shown). DISCUSSION {#s3} ========== The RNA molecules in either intracellular NCs or extracellular virions were reported more than three decades ago ([@B5], [@B41], [@B42]), and naked capsids were shown to carry pgRNA *in vitro* ([@B9], [@B11]). Recently, it was suggested that the extracellular or circulating HBV RNA could serve as a surrogate marker to evaluate the endpoint of hepatitis B treatment ([@B27], [@B30], [@B48][@B49][@B53]). With this in mind and to facilitate its application as a novel biomarker for viral persistence, we studied the origin and characteristics of extracellular HBV RNA. In the present study, we extensively characterized extracellular HBV RNAs and demonstrated that extracellular HBV RNAs were mainly enclosed in naked capsids rather than complete virions in supernatant of HepAD38 cells ([Fig. 1B](#F1){ref-type="fig"} and [2F](#F2){ref-type="fig"}). These RNAs were of heterogeneous lengths, ranging from full-length pgRNA (3.5 knt) to a few hundred nucleotides. Furthermore, circulating HBV RNAs, also heterogeneous in length, were detected in blood of hepatitis B patients ([Fig. 3D](#F3){ref-type="fig"} and [7C](#F7){ref-type="fig"} and [D](#F7){ref-type="fig"}). Interestingly, the detection of HBV RNAs coincided with the presence of immature HBV DNA ([Fig. 3D](#F3){ref-type="fig"} and [E](#F3){ref-type="fig"}). Isopycnic CsCl gradient ultracentrifugation of RNA positive serum samples exhibited a broad range of distribution of immature HBV DNA, which contrasted with the results obtained in HepAD38 cells ([Fig. 2B](#F2){ref-type="fig"} versus [@B4]B and E, [@B7]A). For the first time, we provided convincing evidence that unenveloped capsids containing the full spectrum of HBV replication intermediates and RNA species that are heterogeneous in length could be detected in the circulation of chronic hepatitis B patients. In view of our results and literature reports ([@B2], [@B21][@B22][@B24]), the presence of extracellular HBV RNAs could easily be interpreted in the context of the HBV DNA replication model ([Fig. 11A](#F11){ref-type="fig"}). Since naked capsids contain viral DNA at all maturation levels, they will also carry HBV RNA molecules originating from pgRNA, including full-length pgRNA prior to minus-strand DNA synthesis, pgRNA with 3′ receding ends, and the pgRNA hydrolysis fragments. On the other hand, virions that contain only mature forms of viral DNA species would likely bear only the hydrolyzed short RNA fragments remaining in the nucleocapsid ([@B43]). Likewise, the HBV RNA species found in CACs are longer than those in virions in sera of hepatitis B patients ([Fig. 7D](#F7){ref-type="fig"}, lower, versus C, lower). In line with this reasoning, treatment of HepAD38 cells with entecavir reduced viral DNA in naked capsids and virions ([Fig. 10C](#F10){ref-type="fig"}, [E](#F10){ref-type="fig"}, and [G](#F10){ref-type="fig"}) but at the same time increased HBV RNA content within naked capsids ([Fig. 10H](#F10){ref-type="fig"}). This may be a result of the stalled activity of viral RT with concomitant shutdown of RNA hydrolysis ([@B46], [@B54]). {#F11} Contrary to a recent report claiming that the pgRNA-containing NCs can be enveloped and secreted as virions ([@B27]), we clearly demonstrated that secreted naked capsids carry the majority of HBV RNAs ([Fig. 1B](#F1){ref-type="fig"} and [2F](#F2){ref-type="fig"}) and that virion-associated RNAs are approximately several hundred nucleotides long ([Fig. 1B](#F1){ref-type="fig"} and [7C](#F7){ref-type="fig"}). Our results are consistent with earlier reports demonstrating that only mature nucleocapsids with RC/DSL DNA are enveloped and secreted as virions ([@B6][@B7][@B8], [@B11]), and under this condition, virions carry only short RNase H-cleaved pgRNA ([Fig. 11A](#F11){ref-type="fig"}, step 3). In this research, we were unable to separate hydrolyzed pgRNA fragments from the pgRNA and pgRNA with 3′ receding ends. Thus, the length of these RNA molecules could not be determined. The existence of hydrolyzed RNA products during reverse transcription is not without precedent. In some retroviruses, DNA polymerization speed of RT is greater than the RNA hydrolysis speed of RNase H, thus hydrolysis of RNA template is often incomplete ([@B55], [@B56]). For example, RT of avian myeloblastosis virus (AMV) hydrolyzed RNA template once for every 100 to 200 nt, while cleavage frequency of RTs of human immunodeficiency virus type 1 (HIV-1) and Moloney murine leukemia virus (MoMLV) appeared to be around 100 to 120 nt ([@B57]). Moreover, RNA secondary structures, such as hairpins, may stall the RT activity promoting RNase H cleavage, producing shorter RNA fragments ([@B55], [@B56]). Furthermore, the cleaved RNA fragments may not disassociate but anneal to the nascent minus-strand DNA forming the DNA-RNA hybrids until they are displaced by plus-strand DNA synthesis ([@B55], [@B56]). Although similar studies on HBV replication were hampered by lack of fully functional viral polymerase *in vitro* ([@B58][@B59][@B61]), the reported presence of DNA-RNA hybrid molecules clearly indicated the existence of degraded pgRNA fragments that still annealed to the minus-strand DNA ([@B5], [@B41], [@B42], [@B62]). Consistent with a previous study, our results also showed that at least part of the SS DNA is associated with RNA molecules as the DNA-RNA hybrid molecules, as detected by either RNase H digestion or the cesium sulfate density gradient separation method ([@B5] and data not shown). Given the fact that HBV RNA and immature HBV DNA are packaged in naked capsids ([Fig. 1B](#F1){ref-type="fig"} and [2B](#F2){ref-type="fig"} and [F](#F2){ref-type="fig"}) ([@B11]), we postulated that, in CHB patients, unenveloped capsids are released into circulation, where they rapidly form CACs with anti-HBcAg antibodies ([Fig. 11B](#F11){ref-type="fig"}) ([@B25], [@B33], [@B34]). In support of this notion, we showed that protein A/G agarose beads could specifically pull down particles with mature and immature HBV DNA from sera of CHB patients, implying the involvement of antibody. Addition of anti-HBcAg antibody to HepAD38 cell culture supernatant led to a shift of naked capsids' buoyant density to lower-density regions ([Fig. 4C](#F4){ref-type="fig"} and [D](#F4){ref-type="fig"}), a pattern similar to that obtained in HBV RNA-positive serum samples ([Fig. 4B](#F4){ref-type="fig"} and [E](#F4){ref-type="fig"}, and [7A](#F7){ref-type="fig"}). These particles exhibited heterogeneous electrophoretic behavior that differed from that of particles in HepAD38 culture supernatant, suggesting that they are not individual naked capsid particles but are associated with antibodies and have nonuniform compositions ([Fig. 6](#F6){ref-type="fig"} and [11B](#F11){ref-type="fig"}) ([@B36][@B37][@B38]). In CHB patients, the high titers of anti-HBcAg antibodies, which exceed 10,000 IU/ml, preclude circulation of antibody-unbound naked capsids ([@B63]). Indeed, the excessive amounts of anti-HBcAg antibodies present in the plasma sample of patient 0 were able to pull down naked capsids from the culture supernatant of HepAD38 cells (not shown). We have demonstrated the presence of circulating CACs as the new form of naked capsids in CHB patients. It is known that naked capsid particles can be secreted either by the natural endosomal sorting complex required for transport (ESCRT) pathway ([@B15][@B16][@B17]) or possibly by cell lysis consequent to liver inflammation. Our preliminary clinical data (not shown) are in agreement with a recent study showing an association of circulating HBV RNA with serum ALT level ([@B64]). However, this connection can be interpreted in a different manner, as the capsid-antibody complexes might constitute a danger signal triggering inflammation. Interestingly, the release of naked capsids seems to be an intrinsic property of hepadnaviruses preserved through evolution. Recent studies by Lauber et al. provided evidence as to the ancient origin of HBV descending from nonenveloped progenitors in fish, with their envelope protein gene emerging *de novo* much later ([@B65]). Thus, it is reasonable to propose that the active release of HBV capsid particles should be deemed a natural course of viral egress. Apart from HBV particles, it was also reported that exosomes could serve as HBV DNA or RNA carriers ([@B29], [@B66], [@B67]). However, HBV DNA and RNA was detected in naked capsids or CACs and virion fractions rather than in lower-density regions where membrane vesicles like HBsAg particles (density of 1.18 g/cm^3^) and exosomes (density of 1.10 to 1.18 g/cm^3^) would likely settle ([@B2], [@B27], [@B48], [@B68], [@B69]) ([Fig. 1](#F1){ref-type="fig"} and [7B](#F7){ref-type="fig"}). As a result, it is not likely that exosomes serve as the main vehicles carrying HBV DNA or RNA molecules. Numerous pieces of data showed that HBV spliced RNAs also represent a species of extracellular HBV RNAs ([@B28], [@B70], [@B71]). However, in HepAD38 cells, as most of the RNAs are transcribed from the integrated HBV sequence other than the cccDNA template, pgRNA packaged into nucleocapsids is the predominant RNA molecule ([Fig. 9A](#F9){ref-type="fig"} and [10D](#F10){ref-type="fig"}), and viral DNA derived from pgRNA is the dominant DNA form ([Fig. 2D](#F2){ref-type="fig"} and [E](#F2){ref-type="fig"} and data not shown). For the same reason, it would be difficult for us to estimate the amount of spliced HBV RNAs in clinical samples. Although we could not completely rule out the possibility that HBV RNAs are released into blood circulation by association with other vehicles or other pathways, it is possible that the spliced HBV RNAs also egress out of cells in naked capsids and virions like the pgRNA. In summary, we demonstrated that extracellular HBV RNA molecules are pgRNA and degraded pgRNA fragments generated in the HBV replication process *in vitro*. Moreover, we provided evidence that HBV RNAs exist in the form of CACs in hepatitis B patients' blood circulation. More importantly, the association of circulating HBV RNAs with CACs or virions in hepatitis B patients suggests their pgRNA origin. Hence, our results here suggest the circulating HBV RNAs within CACs or virions in hepatitis B patients could serve as novel biomarkers to assess efficacy of treatment. MATERIALS AND METHODS {#s4} ===================== Cell culture. {#s4.1} ------------- HepAD38 cells that replicate HBV in a tetracycline-repressible manner were maintained in Dulbecco's modified Eagle's medium (DMEM)-F12 medium supplemented with 10% fetal bovine serum, and doxycycline was withdrawn to allow virus replication ([@B31]). Patients and samples. {#s4.2} --------------------- Serum samples from 45 chronic hepatitis B patients with HBV DNA titer higher than 10^7^ IU per ml were randomly selected. Detailed medical records of these patients are included in [Table 1](#T1){ref-type="table"}. ###### Medical records of hepatitis B patients used in this research[^*a*^](#T1F1){ref-type="table-fn"} Patient no. Sex Age (yr) HBV DNA titer (IU/ml) HBeAg (IU/ml) HBsAg (IU/ml) ALT (IU/liter) SS DNA result ------------- ----- ---------- ----------------------- --------------- --------------- ---------------- --------------- 0 NA NA 2.67E + 06 4,932 396 \+ 1 M 54 1.24E + 07 25 \>250 69 \+ 2 F 32 1.20E + 07 1,067 69,384 38 \+ 3 F 21 1.36E + 07 1,712 200 149 \+ 4 M 33 \>5.00E + 07 4,812 113,933 133 \+ 5 NA NA 1.25E + 07 3,423 33 − 6 M 26 1.17E + 07 545 2,759 22 − 7 M 36 1.77E + 07 4,332 19,541 136 **+** 8 M 35 \>5.00E + 07 1,199 \>250 104 **+** 9 M 26 2.20E + 07 \>250 143 − 10 M 30 \>5.00E + 07 2 4,265 123 − 11 F 23 \>5.00E + 07 20 5,757 120 **+** 12 M 37 2.07E + 07 2,315 16,128 177 **+** 13 M 28 \>5.00E + 07 3,495 60,676 58 NA 14 F 28 \>5.00E + 07 16,515 89,575 78 \+ 15 M 37 1.62E + 07 574 +, ND 112 \+ 16 M NA \>5.00E + 07 1,601 \>250 22 NA 17 M 15 2.28E + 07 2,038 32,739 180 \+ 18 M 41 2.71E + 07 694 \>250 313 \+ 19 M 34 2.35E + 07 80 32,514 148 \+ 20 F 44 \>5.00E + 07 1,596 4,306 172 − 21 M NA 3.48E + 07 107 \>250 103 \+ 22 NA NA \>5.00E + 07 2024 45,873 147 \+ 23 M 20 1.32E + 07 13,411 12,387 344 \+ 24 M 48 \>5.00E + 07 5,511 76,914 33 − 25 M NA 3.15E + 07 15,984 366 − 26 M 31 4.16E + 07 10,251 50,469 442 \+ 27 M 60 1.35E + 07 749 \>250 105 \+ 28 F 41 \>5.00E + 07 4,173 \>52,000 194 \+ 29 NA NA \>5.00E + 07 4,233 49,125 39 \+ 30 M 29 1.42E + 07 25 5,800 940 \+ 31 M 27 2.34E + 07 1,117 22,412 129 \+ 32 M 37 2.65E + 07 70 109 NA 33 NA NA 2.03E + 07 4,902 111 \+ 34 M 32 \>5.00E + 07 993 43,582 249 \+ 35 NA NA 2.94E + 07 4,641 93,336 12 \+ 36 NA NA \>5.00E + 07 10,956 2,496 108 \+ 37 F 43 \>5.00E + 07 1,021 \>250 74 \+ 38 F 28 \>5.00E + 07 215 446 26 \+ 39 M 31 \>5.00E + 07 +, ND 38,165 194 \+ 40 NA NA \>5.00E + 07 25 \>250 69 \+ 41 M 26 1.52E + 07 +, ND +, ND 95 \+ 42 M 25 \>5.00E + 07 6,300 43,151 373 \+ 43 M 22 \>5.00E + 07 3,844 23,620 329 \+ 44 M 27 1.36E + 07 1,185 11,106 149 \+ 45 M 44 1.28E + 07 663 23,330 425 − 46 F 29 \>5.00E + 07 +, ND +, ND 667 \+ NA, not available; ND, not determined; M, male; F, female; sera from patients 0 and 46 were not included with sera from other patients for SS DNA screening. Plasma sample was the plasma exchange product obtained from an HBeAg-negative hepatitis B patient (patient 0) (HBV genotype B with A1762T, G1764A, and G1869A mutation) who died of fulminant hepatitis as a consequence of reactivation of hepatitis B ([Table 1](#T1){ref-type="table"}). Ethics statement. {#s4.3} ----------------- All samples from HBV-infected patients used in this study were from an already-existing collection supported by the National Science and Technology Major Project of China (grant no. 2012ZX10002007-001). Written informed consent was received from participants prior to collection of clinical samples ([@B72]). Samples used in this study were anonymized before analysis. This study was conducted in compliance with the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the ethics committee of the Shanghai Public Health Clinical Center. Preparation of viral particles. {#s4.4} ------------------------------- HepAD38 cell culture supernatant was mixed with polyethylene glycol 8000 (PEG 8000) to a final concentration of 10% (wt/vol) and incubated on ice for at least 1 h, followed by centrifugation at 925 × *g* for 20 min. Pellets were suspended in TNE buffer (10 mM Tris-Cl \[pH 7.5\], 100 mM NaCl, and 1 mM EDTA) containing 0.05% β-mercaptoethanol to 1/150 of the original volume, followed by a brief sonication ([@B73], [@B74]). Alternatively, viral particles in HepAD38 cell culture supernatant were concentrated 50- to 100-fold by ultrafiltration using a filter unit (Amicon Ultra-15, 100 kDa). Plasma samples from patient 0 were centrifuged through a 20% (wt/vol) sucrose cushion at 26,000 rpm for 16 h in an SW 32 Ti rotor (Beckman), and pellets were resuspended in 1/200 the original volume of TNE buffer and sonicated briefly ([@B75]). Samples prepared using methods described above were either used immediately or aliquoted and stored at −80°C for later use. Sucrose density gradient centrifugation. {#s4.5} ---------------------------------------- HepAD38 cells culture supernatant concentrated by PEG 8000 was centrifugation at 500 × *g* for 5 min to remove aggregates. Ten percent, 20%, 30%, 40%, 50%, and 60% (wt/wt) sucrose gradients were prepared by underlayering and incubated for 4 h in a water bath at room temperature to allow gradient to become continuous. Five hundred microliters of concentrated sample was layered over the gradient and centrifuged at 34,100 rpm for 14 h at 4°C in a Beckman SW 41 Ti rotor. Fractions were collected from top to bottom, and the density of each fraction was determined by refractometry ([@B10]). Fractions containing viral particles were subjected to native agarose gel analysis, and HBsAg level was determined by enzyme-linked immunosorbent assay (ELISA) (Shanghai Kehua). Cesium chloride density gradient centrifugation. {#s4.6} ------------------------------------------------ HepAD38 cell culture supernatant (1.5 ml), concentrated by ultrafiltration, or serum samples from chronic hepatitis patients diluted with TNE buffer to 1.5 ml were mixed with equal volumes of 37% (wt/wt) CsCl-TNE buffer (1.377 g/cm^3^) and underlayered with 1.9 ml 34% (wt/wt) CsCl-TNE buffer (1.336 g/cm^3^), followed by centrifugation at 90,000 rpm at 4°C for 12 h (Beckman VTi 90 rotor) ([@B8]). The tube was punctured from the bottom, and every six to seven drops were collected as one fraction. Densities of separated fractions were determined by weighing. Each fraction was then desalted against TNE buffer by ultrafiltration, followed by native agarose gel separation or nucleic acid extraction. All of the CsCl density gradient centrifugation experiments were carried out at 90,000 rpm at 4°C for 12 h in a Beckman VTi 90 rotor. Native agarose gel analysis of viral particles and capsid-associated DNA. {#s4.7} ------------------------------------------------------------------------- Viral particles were resolved by native agarose gel (0.8% agarose gel prepared in Tris-acetate-EDTA \[TAE\] buffer) electrophoresis and transferred in TNE buffer to either a nitrocellulose membrane (0.45 μM) for detection of viral antigens with specific antibodies or a nylon membrane for Southern blot analysis of viral DNA. For viral antigens detection, the membrane was first fixed as previously described ([@B74]), and HBV core antigen was detected by anti-HBcAg antibody (Dako) (1:5,000). The same membrane then was soaked in stripping buffer (200 mM glycine, 0.1% SDS, 1% Tween 20, pH 2.2) and reprobed with anti-HBsAg antibody (Shanghai Kehua) (1:5,000). For Southern blot analysis of viral DNA, the membrane was dipped in denaturing buffer (0.5 N NaOH, 1.5 M NaCl) for 10 s and immediately neutralized in 1 M Tris-Cl (pH 7.0)--1.5 M NaCl for 1 min, followed by hybridization with minus-strand-specific riboprobe ([@B76]). Viral nucleic acid extraction, separation, and detection. {#s4.8} --------------------------------------------------------- **(I) Nucleic acid extraction.** To extract total viral nucleic acids (DNA and RNA), the SDS-proteinase K method was used ([@B77]). Samples were digested in solution containing 1% SDS, 15 mM EDTA, and 0.5 mg/ml proteinase K at 37°C for 15 min. The digestion mixture was extracted twice with phenol and once with chloroform. Aqueous supernatant were added with 1/9 volume of 3 M sodium acetate (pH 5.2) and 40 μg of glycogen and precipitated with 2.5 volumes of ethanol. In addition to the SDS-proteinase K method, viral RNA was also extracted with TRIzol LS reagent according to the manufacturer's instructions (Thermo Fisher Scientific). To isolate intracellular capsid-associated viral RNA, HepAD38 cells were lysed in NP-40 lysis buffer (50 mM Tris-Cl \[pH 7.8\], 1 mM EDTA, 1% NP-40), and cytoplasmic lysates were incubated with CaCl~2~ (final concentration, 5 mM) and micrococcal nuclease (MNase) (Roche) (final concentration, 15 U/ml) at 37°C for 1 h to remove nucleic acids outside nucleocapsids. The reaction was terminated by addition of EDTA (final concentration, 15 mM), and then proteinase K (0.5 mg/ml without SDS) was added to the mixture, followed by incubation at 37°C for 30 min to inactivate MNase. Viral nucleic acids were released by addition of SDS to a final concentration of 1% and extracted as described above. **II. Separation. (i) TAE agarose gel.** Viral DNA was resolved by electrophoresis through a 1.5% agarose gel in 1× TAE buffer, followed by denaturation in 0.5 M NaOH--1.5 M NaCl for 30 min and neutralization with 1 M Tris-Cl (pH 7.0)--1.5 M NaCl for 30 min. **(ii) Alkaline agarose gel.** Viral DNA was denatured with a 0.1 volume of solution containing 0.5 M NaOH and 10 mM EDTA and resolved overnight at 1.5 V/cm in a 1.5% agarose gel with 50 mM NaOH and 1 mM EDTA. After electrophoresis, the gel was neutralized with 1 M Tris-Cl (pH 7.0)--1.5 M NaCl for 45 min ([@B78]). **(iii) Formaldehyde-MOPS agarose gel.** Viral RNA was obtained by treatment of total nucleic acids extracted using the above-described SDS-proteinase K method with RNase free DNase I (Roche) for 15 min at 37°C. The reaction was stopped by addition of equal amounts of 2× RNA loading buffer (95% formamide, 0.025% SDS, 0.025% bromophenol blue, 0.025% xylene cyanol FF, and 1 mM EDTA) supplemented with extra EDTA (20 mM), followed by denaturing at 65°C for 10 min. Viral RNA extracted by TRIzol LS reagent was mixed with 2× RNA loading buffer and denatured. Denatured mixtures were separated by electrophoresis through a 1.5% agarose gel containing 2% (vol/vol) formaldehyde solution (37%) and 1× MOPS (3-\[N-morpholino\]propanesulfonic acid) buffer. The gels described above were balanced in 20× SSC solution (1× SSC is 0.15 M NaCl and 0.015 M sodium citrate, pH 7.0) for 20 min, and viral nucleic acids were transferred onto nylon membranes overnight with 20× SSC buffer. III. Detection. {#s4.10} --------------- Digoxigenin-labeled riboprobes used for detection of HBV DNA and RNA were prepared by *in vitro* transcription of a pcDNA3 plasmid that harbors 3,215 bp of HBV DNA (nt 1814 to 1813) by following the vendor's suggestions (12039672910; Roche). Riboprobes used for HBV RNA mapping were transcribed from DNA templates generated by PCR by incorporating T7 promoter into the 5′ end of reversed primers ([Fig. 9A](#F9){ref-type="fig"}). Hybridization was carried out at 50°C overnight, followed by two 5-min washes in 2× SSC--0.1% SDS at room temperature and two additional 15-min washes in 0.1× SSC--0.1% SDS at 50°C. The membrane was sequentially incubated with blocking buffer and anti-digoxigenin-AP Fab fragment (Roche) at 20°C for 30 min. Subsequently, the membrane was washed twice with washing buffer (100 mM maleic acid, 150 mM NaCl, and 0.3% Tween 20, pH 7.5) for 15 min, followed by detection with diluted CDP-Star substrate (ABI) and exposure to X-ray film. Protein A/G agarose bead pulldown of antibody-antigen complexes. {#s4.11} ---------------------------------------------------------------- Two hundred microliters of serum sample was first mixed with 300 μl of TNE buffer, and then 15 μl of protein A/G agarose bead slurry (Santa Cruz) was added to the mixture, followed by incubation overnight at 4°C in a sample mixer. Subsequently, protein A/G agarose beads were washed three times with TNE buffer, and viral DNA in input serum samples (40 μl) and agarose bead pulldown mixtures were extracted and subjected to Southern blot analysis. EM. {#s4.12} --- Serum samples from patients 11, 17, 21 22, 23, 27, 28, 30, and 41 were pooled (200 μl each) and mixed with 200 μl of 20% (wt/wt) sucrose. Serum mixtures were centrifuged through 2 ml of 20% (wt/wt) and 2 ml of 45% (wt/wt) (1.203 g/cm^3^) sucrose cushions at 34,100 rpm for 8 h at 4°C in an SW 41 Ti rotor (Beckman) to remove HBsAg particles. Supernatants were decanted and the centrifugation tube was placed upside down for 20 s, and residue sucrose was wiped out. One milliliter of phosphate buffer (10 mM Na~2~HPO~4~, 1.8 mM KH~2~PO~4~, and no NaCl) (pH 7.4) was added, and the bottom of the tube was gently washed without disturbing the pellet. A volume of 11.5 ml of phosphate buffer then was added into the tube and centrifuged again at 34,100 rpm for 3 h at 4°C. The pellet was resuspended in a drop of distilled water and dropped onto a carbon-coated copper grid, followed by staining with 2% phosphotungstic acid (pH 6.1) and examining in an electron microscope (Philip CM120) ([@B13], [@B79]). Viral DNA and RNA quantification. {#s4.13} --------------------------------- Viral DNA used for quantification was extracted using the SDS-proteinase K method as described above. Viral RNAs were extracted by TRIzol LS reagent, and DNase I was used to remove the remaining DNA, followed by phenol and chloroform extraction and ethanol precipitation. Reverse transcription was carried out using Maxima H minus reverse transcriptase (Thermo Fisher Scientific) with a specific primer (AGATCTTCKGCGACGCGG \[nt 2428 to 2411\]) according to the manufacturer's guidelines, except the 65°C incubation step was skipped to avoid RNA degradation. To ensure removal of viral DNA signal (below 1,000 copies per reaction), a mock reverse transcription, without addition of reverse transcriptase, was carried out. Quantitative real-time PCR (qPCR) was carried out using Thunderbird SYBR qPCR mix (Toyobo) in a StepOnePlus real-time PCR system (ABI). Primer pairs (F, GGRGTGTGGATTCGCAC \[nt 2267 to 2283\]; R, AGATCTTCKGCGACGCGG \[nt 2428 to 2411\]) conserved among all HBV genotypes and close to the 5′ end but not in the overlap region between the start codon and the poly(A) cleavage site of pgRNA were chosen. The cycling conditions were 95°C for 5 min, followed by 40 cycles of 95°C for 5 s, 57°C for 20 s, and 72°C for 30 s. DNA fragment containing 3,215 bp of full-length HBV DNA was released from plasmid by restriction enzymes, and DNA standards were prepared according to a formula in which 1 pg of DNA equals 3 × 10^5^ copies of viral DNA. EPA. {#s4.14} ---- HepAD38 cell culture supernatant or plasma from patient 0 were concentrated as described above and mixed with equal volumes of 2× EPA buffer (100 mM Tris-Cl, pH 7.5, 80 mM NH~4~Cl, 40 mM MgCl~2~, 2% NP-40, and 0.6% β-mercaptoethanol) with or without dNTPs (dATP, dCTP, dGTP, and dTTP, each at a final concentration of 100 μM) ([@B80]). The reaction mixtures were incubated at 37°C for 2 h and stopped by addition of EDTA to a final concentration of 15 mM. 3′ RACE. {#s4.15} -------- Concentrated HepAD38 cell culture supernatant (by ultrafiltration) was digested with MNase in the presence of NP-40 (final concentration, 1%) for 30 min at 37°C. EDTA (final concentration, 15 mM) and proteinase K (final concentration, 0.5 mg/ml) were then added and incubated for another 30 min at 37°C. Viral nucleic acids were extracted with TRIzol LS reagent followed by DNase I treatment to remove residue viral DNA. Poly(A) tails were added to the 3′ end of HBV RNA by E. coli poly(A) polymerase (NEB). The preincubation step at 65°C for 5 min was omitted to reduce potential RNA degradation, and reverse transcription was carried out with Maxima H minus reverse transcriptase (Thermo Scientific) using an oligo-dT(29)-SfiI(A)-adaptor primer (5′-AAGCAGTGGTATCAACGCAGAGTGGCCATTACGGCCTTTTTTTTTTTTTTTTTTTTTTTTTTTTT-3′) in reverse transcription buffer \[1× RT buffer, RNase inhibitor, 1 M betanine, 0.5 mM each dNTP, and 5 μM of oligo-dT(29)-SfiI(A)-adaptor primer\] at 50°C for 90 min, followed by heating at 85°C for 5 min and treatment with RNase H at 37°C for 15 min. PCR amplification of cDNA fragments was then performed with 5′ HBV-specific primers \[the same sequences of forward primers used for riboprobe preparation ([Fig. 9A](#F9){ref-type="fig"}), except each primer containing a flanking sequence plus a SfiI(B) site (5′-AGTGATGGCCGAGGCGGCC-3′)\] and 3′ adaptor primer (5′-AAGCAGTGGTATCAACGCAGAGTG-3′). The reaction was carried out with PrimeSTAR HS DNA polymerase (TaKaRa) at 95°C for 5 min, followed by 5 cycles of 98°C for 5 s, 50°C for 10 s, and 72°C for 210 s, 35 cycles of 98°C for 5 s, 55°C for 10 s, and 72°C for 210 s, and a final extension step at 72°C for 10 min. PCR amplicons were digested with SfiI enzyme and cloned into pV1-Blasticidin vector (kind gift from Zhigang Yi, Shanghai Medical College, Fudan University). Positive clones were identified by sequencing, and only clones with 3′ poly(dA) sequence were considered authentic viral RNA 3′ ends. We thank Zhuying Chen and Xiurong Peng for handling serum samples and compiling the clinical data used in this research. This research was supported by the National Natural Science Foundation of China (NSFC) (81671998, 91542207), National Key Research and Development Program (2016YFC0100604), National Science and Technology Major Project of China (2017ZX10302201001005), Shanghai Science and Technology Commission (16411960100), and Innovation Program of Shanghai Municipal Education Commission (2017-01-07-00-07-E00057). [^1]: **Citation** Bai L, Zhang X, Kozlowski M, Li W, Wu M, Liu J, Chen L, Zhang J, Huang Y, Yuan Z. 2018. Extracellular hepatitis B virus RNAs are heterogeneous in length and circulate as capsid-antibody complexes in addition to virions in chronic hepatitis B patients. J Virol 92:e00798-18. <https://doi.org/10.1128/JVI.00798-18>.

Koolstra K, Beenakker J‐WM, Koken P, Webb A, Börnert P. Cartesian MR fingerprinting in the eye at 7T using compressed sensing and matrix completion‐based reconstructions. Magn Reson Med. 2019;81:2551--2565. 10.1002/mrm.27594 30421448 **Funding information** This project was partially funded by the European Research Council Advanced Grant 670629 NOMA MRI. 1. INTRODUCTION {#mrm27594-sec-0005} =============== Ophthalmologic disease diagnosis conventionally relies mainly on ultrasound and optical imaging techniques such as fundus photography and fluorescent angiography (FAG), MRI is increasingly being used in the radiological community.[1](#mrm27594-bib-0001){ref-type="ref"}, [2](#mrm27594-bib-0002){ref-type="ref"}, [3](#mrm27594-bib-0003){ref-type="ref"} One of the main advantages of MRI is its capability to assess nontransparent tissues such as ocular tumors or structures behind the globe such as the eye muscles. Currently, however, these applications are mainly based on qualitative MRI methods using the large number of tissue contrasts addressable by MR. As an example, in Graves' ophthalmopathy fat‐suppressed T~2~‐weighted MRI is the standard to detect inflammation in the eye muscles,[4](#mrm27594-bib-0004){ref-type="ref"}, [5](#mrm27594-bib-0005){ref-type="ref"} whereas in the diagnosis of retinoblastoma, a rare intraocular cancer in children, standard T~1~‐ and T~2~‐weighted MRI is often performed to confirm the presence of the tumor and to screen for potential optic nerve involvement.[2](#mrm27594-bib-0002){ref-type="ref"} In more recent ophthalmologic applications of MRI, such as uveal melanoma (the most common primary intraocular tumor), quantitative MRI techniques including DWI[6](#mrm27594-bib-0006){ref-type="ref"} and DCE imaging[7](#mrm27594-bib-0007){ref-type="ref"} have been shown, but currently diagnosis is still based on qualitative methods.[3](#mrm27594-bib-0003){ref-type="ref"} To personalize treatment plans quantitative parameters of the tissues involved, as can be acquired invasively for example by performing biopsies,[8](#mrm27594-bib-0008){ref-type="ref"} are highly desirable. However, quantitative parameter mapping by means of MRI requires long examination times, which would result in significant eye‐motion artifacts, as well as patient discomfort.[9](#mrm27594-bib-0009){ref-type="ref"} MR fingerprinting (MRF) is a recently introduced method for rapid quantitation of tissue relaxation times and other MR‐related parameters.[10](#mrm27594-bib-0010){ref-type="ref"} It uses a flip angle sweep to induce a unique signal evolution for each tissue type. Incoherent undersampling can be applied during sampling of the MRF train, enabling acceleration of the MRF scans.[10](#mrm27594-bib-0010){ref-type="ref"} Together with its ability to measure simultaneously T~1~ and T~2~, MRF offers a solution to the problem of obtaining quantitative measures in an efficient manner and in relatively short scanning times. One of the main challenges in ocular imaging is in‐plane and through‐plane eye motion, often associated with eye blinking.[11](#mrm27594-bib-0011){ref-type="ref"}, [12](#mrm27594-bib-0012){ref-type="ref"}, [13](#mrm27594-bib-0013){ref-type="ref"} The motion results in corrupted k‐space data that introduces artifacts and blurring throughout the entire image. Shortening the scans would reduce motion‐related artifacts, but standard acceleration techniques are not optimal for the current eye application due to the following 3 reasons. First, a cued‐blinking protocol is typically used to control and reduce the eye motion.[3](#mrm27594-bib-0003){ref-type="ref"}, [11](#mrm27594-bib-0011){ref-type="ref"} This requires an instruction screen placed at the end of the MR tunnel to be visible to the patient which complicates the use of small phased array receive coils in front of the eye, blocking the view. Instead, a custom‐built single‐element eye loop coil is used, which provides a high local SNR[3](#mrm27594-bib-0003){ref-type="ref"} and screen visibility, but which clearly excludes the possibility of scan acceleration by means of parallel imaging.[14](#mrm27594-bib-0014){ref-type="ref"} Second, the gel‐like vitreous body has an extremely long T~1~, particularly at high field.[15](#mrm27594-bib-0015){ref-type="ref"} Its value of 3 to 5 s requires a long duration of the MRF sequence to encode the MR parameters (T~1~,T~2~) sufficiently. Thus, using a flip angle train with a small number of RF pulses is not feasible, hindering scan time reduction. Finally, a time‐efficient spiral sampling scheme, usually applied in MRF,[10](#mrm27594-bib-0010){ref-type="ref"}, [16](#mrm27594-bib-0016){ref-type="ref"}, [17](#mrm27594-bib-0017){ref-type="ref"}, [18](#mrm27594-bib-0018){ref-type="ref"}, [19](#mrm27594-bib-0019){ref-type="ref"} introduces off‐resonance effects in each of the individual MRF images.[20](#mrm27594-bib-0020){ref-type="ref"} This occurs even when combined with unbalanced sequences such as fast imaging with steady state precession,[16](#mrm27594-bib-0016){ref-type="ref"} which are in themselves robust to off‐resonance effects.[21](#mrm27594-bib-0021){ref-type="ref"} The off‐resonance effects present in spiral sampling schemes are much stronger at high field, where they result in blurring,[22](#mrm27594-bib-0022){ref-type="ref"} caused by strong main field inhomogeneities (particularly in the eye region due to many air‐tissue‐bone interfaces), as well as the presence of significant amounts of off‐resonant orbital fat around the eye. In this work, a Cartesian sampling scheme is used, which is more robust than spiral sampling to off‐resonance effects, but which is significantly less time‐efficient.[23](#mrm27594-bib-0023){ref-type="ref"} With such a Cartesian sampling scheme, undersampling artifacts have a more structured nature compared with spiral sampling, which increases the temporal coherence of the artifacts in the MRF image series.[10](#mrm27594-bib-0010){ref-type="ref"}, [20](#mrm27594-bib-0020){ref-type="ref"} In this case, direct matching of the measured MRF signal reconstructed by plain Fourier transformations, to the simulated dictionary elements is not sufficiently accurate for high undersampling factors.[24](#mrm27594-bib-0024){ref-type="ref"}, [25](#mrm27594-bib-0025){ref-type="ref"} Therefore, the quality of the reconstructed MRF data has to be improved before the matching process. Compressed sensing (CS) has been introduced as a technique to reconstruct images from randomly undersampled data by enforcing signal sparsity (in the spatial dimension only or both in spatial and temporal dimensions),[26](#mrm27594-bib-0026){ref-type="ref"}, [27](#mrm27594-bib-0027){ref-type="ref"} allowing a scan time reduction in many applications.[28](#mrm27594-bib-0028){ref-type="ref"}, [29](#mrm27594-bib-0029){ref-type="ref"}, [30](#mrm27594-bib-0030){ref-type="ref"} The flexibility of MRF toward different sampling schemes and undersampling factors makes it possible to reconstruct the source images by means of CS.[27](#mrm27594-bib-0027){ref-type="ref"}, [31](#mrm27594-bib-0031){ref-type="ref"}, [32](#mrm27594-bib-0032){ref-type="ref"} Higher acceleration factors might be feasible if the correlation in the temporal dimension is better used.[33](#mrm27594-bib-0033){ref-type="ref"} Examples of such reconstructions specifically tailored to MRF are given in Davies et al, Pierre et al, and Zhao et al[34](#mrm27594-bib-0034){ref-type="ref"}, [35](#mrm27594-bib-0035){ref-type="ref"}, [36](#mrm27594-bib-0036){ref-type="ref"} which take into account the simulated dictionary atoms in the image reconstruction process. Recent work has shown that the temporal correlation in the MRF data can be exploited even further by incorporating the low rank structure of the data into the cost function,[37](#mrm27594-bib-0037){ref-type="ref"} a technique which was introduced into MR in Liang[38](#mrm27594-bib-0038){ref-type="ref"} and in MRF in Zhao[39](#mrm27594-bib-0039){ref-type="ref"} and used by many others[40](#mrm27594-bib-0040){ref-type="ref"}, [41](#mrm27594-bib-0041){ref-type="ref"}, [42](#mrm27594-bib-0042){ref-type="ref"}: these techniques can also be combined with sparsity constraints.[43](#mrm27594-bib-0043){ref-type="ref"}, [44](#mrm27594-bib-0044){ref-type="ref"} Most of the aforementioned techniques involve Fourier transformations in each iteration, making the reconstruction process time‐consuming. In this application, the single‐element receive coil allows us to perform the reconstruction process entirely in k‐space when exploiting the low rank structure of the MRF data as is performed in matrix completion (MC)‐based reconstructions.[42](#mrm27594-bib-0042){ref-type="ref"}, [45](#mrm27594-bib-0045){ref-type="ref"} In this work, undersampled Cartesian ocular MRF is investigated using CS and MC‐based reconstructions. Simulations and experiments performed in 6 healthy volunteers for confirmation are compared with fully sampled MRF in terms of the quality of the parameter maps, and mean relaxation times were derived for different ocular structures at 7T. Finally, parameter maps after an MC‐based reconstruction are included for a uveal melanoma patient, showing the feasibility of ocular MRF in eye tumor patients. 2. METHODS {#mrm27594-sec-0006} ========== 2.1. Fingerprinting definition {#mrm27594-sec-0007} ------------------------------ The MRF encoding principle is based on a variable flip angle train with relatively short TRs, so that the magnetization after each RF pulse is influenced by the spin history. Following closely the implementation of the sinusoidal MRF pattern described in Jiang et al,[16](#mrm27594-bib-0016){ref-type="ref"} a flip angle pattern of 240 RF excitation pulses ranging from 0° to 60° (see Figure [1](#mrm27594-fig-0001){ref-type="fig"}A) was defined by the function$$FA\left( x \right) = \left\{ \begin{matrix} {20\,\text{sin}(\frac{\pi}{110}x)\,\text{for}\, 1 \leq x \leq 110} \\ {60\,\text{sin}(\frac{\pi}{130}\left( {x - 110} \right))\,\text{for}\, 110 < x \leq 240} \\ \end{matrix} \right.$$ {#mrm27594-fig-0001} preceded by an inversion pulse (16). A fast imaging with steady state precession sequence was used,[16](#mrm27594-bib-0016){ref-type="ref"}, [19](#mrm27594-bib-0019){ref-type="ref"} in which the TE was chosen as 3.5 ms and 4.0 ms for low resolution scans and high resolution scans, respectively. The selected excitation RF pulse had a time‐bandwidth product of 10, resulting in a reasonably sharp slice profile. The RF pulse phase was fixed to 0°. To simplify dictionary calculations, because of the simplification of the magnetization coherence pathways,[46](#mrm27594-bib-0046){ref-type="ref"} the TR was set to a constant value of 11 ms. A 3D dictionary was calculated following the extended phase graph formalism,[21](#mrm27594-bib-0021){ref-type="ref"}, [46](#mrm27594-bib-0046){ref-type="ref"} based on the Bloch equations,[47](#mrm27594-bib-0047){ref-type="ref"}, [48](#mrm27594-bib-0048){ref-type="ref"} incorporating 27,885 signal evolutions.[46](#mrm27594-bib-0046){ref-type="ref"} T~1~ values ranged from 10 to 1000 ms in steps of 10 ms, and from 1000 to 5000 ms in steps of 100 ms. T~2~ values ranged from 10 to 100 ms in steps of 10 ms and from 100 to 300 ms in steps of 20 ms. A B~1~ ^+^ fraction ranging from 0.5 to 1.0 in steps of 0.05 was incorporated into the dictionary calculation. To shorten the scan time, we used a short waiting time between repetitions of the MRF train (called the repetition delay) of 2.5 s. Therefore, each MRF scan was preceded by 3 dummy trains to establish steady state magnetization,[19](#mrm27594-bib-0019){ref-type="ref"} which was considered in the dictionary calculation. The longitudinal magnetization after the 3 dummy trains, required for correction of the M~0~ maps, was calculated for each T~1~/T~2~ combination. The repetition delay of 2.5 s was efficiently used as the blink time.[3](#mrm27594-bib-0003){ref-type="ref"}, [11](#mrm27594-bib-0011){ref-type="ref"} 2.2. Experimental setup {#mrm27594-sec-0008} ----------------------- All experiments were approved by the local medical ethics committee, and all volunteers and patients signed an appropriate informed consent form. The experiments in this study were performed on 6 healthy volunteers and 1 uveal melanoma patient using a 7T MR system (Philips Healthcare) equipped with a quadrature head volume coil (Nova Medical) for transmission and a custom‐built single‐element eye coil for reception, with a diameter of approximately 4 cm.[3](#mrm27594-bib-0003){ref-type="ref"}, [49](#mrm27594-bib-0049){ref-type="ref"} A cued‐blinking protocol was followed, which means that all subjects were instructed to focus on a fixation target shown on a screen during data acquisition and to blink in the 2.5 s repetition delay. This was performed using a small mirror integrated into the eye coil, allowing visualization of a screen placed outside the magnet through 1 eye, while the eye to be imaged was closed and covered by a wet gauze to reduce susceptibility artifacts in the eye lid.[50](#mrm27594-bib-0050){ref-type="ref"} This setup is shown schematically in Figure [1](#mrm27594-fig-0001){ref-type="fig"}B. 2.3. MR data acquisition {#mrm27594-sec-0009} ------------------------ Because of the presence of significant orbital fat around the eye, and the sensitivity of the spiral to off‐resonance resulting in blurring,[22](#mrm27594-bib-0022){ref-type="ref"} a Cartesian sampling scheme was used to acquire all data. The fingerprinting scans were acquired as a single slice at 2 different spatial resolutions: 1.0 × 1.0 × 5.0 mm^3^ and 0.5 × 0.5 × 5.0 mm^3^. The lower resolution scan was performed twice, the first fully sampled to serve as a reference, and the second one undersampled. The scan time of the fully sampled scan was 7:02 min, while the scan time of the undersampled scan, in which 15% of the data was acquired, was 1:16 min. The high resolution scan was only acquired as an undersampled data set, in which 12.5% of the data was acquired, resulting in a scan time of 1:57 min. In the undersampled scans a simple variable density k‐space sampling was applied, schematically shown in Figure [1](#mrm27594-fig-0001){ref-type="fig"}C, supporting both CS and MC‐based reconstructions. A fully sampled center of k‐space was acquired for each time point consisting of 6/8 k‐space lines for the low resolution/high resolution scans, respectively. For all scans, the FOV was set to 80 × 80 mm^2^, resulting in an acquisition matrix of 80 × 80 and 160 × 160 for the low and the high resolution scans, respectively. The phase encoding direction was set from left‐to‐right to minimize contamination by any residual motion artifacts in the eye lens, and the read out direction was set to the anterior‐posterior direction. B~1~ ^+^ maps were acquired using the dual refocusing echo acquisition mode method[51](#mrm27594-bib-0051){ref-type="ref"} with the following scan parameters: FOV = 80 × 80 mm^2^, in‐plane resolution 1 mm^2^, slice thickness 5 mm, 1 slice, TE~1~/TE~2~ = 2.38/1.54 ms, TR = 3.7 ms, FA = α:60°/ß:10°: the scan time for a single slice was less than 1 s. 2.4. Reconstruction {#mrm27594-sec-0010} ------------------- For each time point, the corresponding images were reconstructed from the available data, using custom software written in MATLAB (Mathworks, Inc) and run on a Windows 64‐bit machine with an Intel i3‐4160 CPI @ 3.6 GHz and 16 GB internal memory. Different reconstructions were performed: (i) a fast Fourier transform (FFT) of the fully sampled data and of the zero‐filled undersampled data; (ii) a CS reconstruction with total variation regularization in the spatial dimension (2D CS), and with total variation in both spatial and temporal dimensions (3D CS) of the undersampled data; (iii) an MC‐based reconstruction of the undersampled data. ### 2.4.1. CS reconstruction {#mrm27594-sec-0011} In this reconstruction, the complete image series is reconstructed by iteratively solving the nonlinear problem$$\hat{\mathbf{x}} = \text{argmin}_{\mathbf{x}}TV\left( \mathbf{x} \right)s.t.\, RF\mathbf{x} = \mathbf{y}_{u}$$ through the unconstrained version$$\hat{\mathbf{x}} = \text{argmin}_{\mathbf{x}}{\frac{\mu}{2}{|RF\mathbf{x} - \mathbf{y}_{u}|}}_{2}^{2} + \frac{\lambda}{2}TV{(\mathbf{x})}$$ In this formulation, $F \in \mathbb{C}^{Nt \times Nt}$ is a block diagonal matrix with the 2D Fourier transform matrix in each diagonal block, $R \in \mathbb{C}^{Nt \times Nt}$ is a diagonal matrix incorporating the sampling locations, $\mathbf{y}_{u} \in \mathbb{C}^{Nt \times 1}$ is the undersampled k‐t space data, $\hat{\mathbf{x}} \in \mathbb{C}^{Nt \times 1}$ is an estimate of the true image series and $\mathit{TV}$ is a total variation operator which is used to enforce sparsity in the reconstruction.[52](#mrm27594-bib-0052){ref-type="ref"}, [53](#mrm27594-bib-0053){ref-type="ref"} Here, $N$ is the number of k‐space locations per image frame and $t$ is the number of measured time points (or flip angles in the MRF train). The regularization parameters $\mu$ and $\lambda$ in Equation [\[Link\]](#mrm27594-disp-0001){ref-type="disp-formula"} were determined empirically and set to $\mu = 0.1\, and\,\lambda = 0.2.$ Two basic versions of the total variation operator,$$TV\left( \mathbf{x} \right) = {|\nabla_{x}{\mathbf{x}|}}_{1} + {|\nabla_{y}{\mathbf{x}|}}_{1}$$ $$TV\left( \mathbf{x} \right) = {|\nabla_{\mathbf{x}}{\mathbf{x}|}}_{1} + {|\nabla_{\mathbf{y}}{\mathbf{x}|}}_{1} + {|\nabla_{\mathbf{t}}{\mathbf{x}|}}_{1}$$ were implemented to investigate the effect of promoting sparsity either only in the spatial dimension (2D CS) or in both the spatial and temporal dimensions (3D CS). In these expressions, $\nabla_{x},\nabla_{y}$ and $\nabla_{t}$ are the first derivative operators acting on the spatial $x$ and $y$ dimensions and the time dimension, respectively. Solving the problem given in Equation [\[Link\]](#mrm27594-disp-0001){ref-type="disp-formula"} is done in this work using Split Bregman. For details on this algorithm the reader is referred to Goldstein and Osher.[54](#mrm27594-bib-0054){ref-type="ref"} ### 2.4.2. MC reconstruction {#mrm27594-sec-0012} Similar to CS with the TV operator acting in 3 dimensions (see Equation [(1)](#mrm27594-disp-0003){ref-type="disp-formula"}), MC uses the information from the temporal dimension.[45](#mrm27594-bib-0045){ref-type="ref"}, [55](#mrm27594-bib-0055){ref-type="ref"} A main difference between CS and MC, however, is that sparsity of singular values, which is a priori information in the MC reconstruction, can be observed both in image space and in k‐space. This allows one to complete the entire reconstruction in k‐space, which is computationally efficient, especially if only a single receiver coil is used.[42](#mrm27594-bib-0042){ref-type="ref"} The MC‐based reconstruction iteratively solves$$\hat{M} = \mathit{argmin}_{M}{|M|}_{\ast}\, s.t.\,\mathcal{P}_{\Omega}M = M_{u}$$ with ${| \bullet |}_{\ast}$ being the nuclear norm, $\mathcal{P}_{\Omega}$ the sampling operator selecting the measured k‐t space locations, $M_{u} \in {}^{t \times N}$ the undersampled k‐t space data and $\hat{M} \in \mathbb{C}^{t \times N}$ an estimate of the true k‐t space. The nuclear norm of *M* sums the singular values of *M*, and can thus be written as ${|\sigma(M)|}_{1}$, where $\sigma$ transforms $M$ into a vector containing the singular values of $M$. The central k‐t space is used as calibration data, of which the rank can be used as a priori information in the reconstruction of undersampled data. In this process, a projection matrix $\mathcal{P}_{U_{n}} \in \mathbb{C}^{t \times t}$ projects in each iteration $i$ the undersampled data matrix $M^{i}$ onto a low‐rank subspace spanned by the columns of $U_{n} \in \mathbb{C}^{t \times n}$, such that$$\overset{\mspace{600mu}}{M^{i}} = \mathcal{P}_{U_{n}}M^{i}$$ with$$\mathcal{P}_{U_{n}} = U_{n}U_{n}^{H}.$$ Here, $U_{n}$ contains the $n$ most significant left singular vectors of the calibration matrix $M_{c} \in \mathbb{C}^{t \times p}$ and is constructed from the full singular value decomposition $M_{c} = U\Sigma V^{H}$, $U \in \mathbb{C}^{t \times t}$, $\Sigma \in \mathbb{R}^{t \times p}$, $V \in \mathbb{C}^{p \times p}$, which is performed once at the beginning of the algorithm. In the second step of each iteration, the data are updated according to$$M^{i + 1} = M_{u} + {(I - \mathcal{P}_{\Omega})}{\overset{\sim}{M}}^{i}.$$ The value $n$ was determined empirically from the singular value plots (shown in Figure [1](#mrm27594-fig-0001){ref-type="fig"}D for 1 volunteer) and set to 4 for all MC‐based reconstructions. Further details of the adopted algorithm to solve Equation [(2)](#mrm27594-disp-0004){ref-type="disp-formula"}, and its implementation can be found in Doneva et al.[42](#mrm27594-bib-0042){ref-type="ref"} To ensure convergence of the iterative CS and MC‐based reconstructions, 40 Split Bregman iterations (1 inner loop) were used for the CS reconstructions and 100 iterations were used for all MC‐based reconstructions. To judge the performance of the reconstruction methods, relative error measures are defined throughout the manuscript as$$\mathit{RelativeError}\left( \mathbf{u} \right) = \frac{{{|\mathbf{u} -}\mathbf{u}_{\mathbf{r}\mathbf{e}\mathbf{f}}|}_{2}}{{|\mathbf{u}_{\mathbf{r}\mathbf{e}\mathbf{f}}|}_{2}},$$ where $\mathbf{u}_{\mathit{ref}}$ is the fully sampled image series and both $\mathbf{u}$ and $\mathbf{u}_{\mathit{ref}}$ are vectorized. 2.5. Dictionary matching process {#mrm27594-sec-0013} -------------------------------- For each subject, the measured B~1~ ^+^ map was used to calculate an average B~1~ ^+^ value in the eye. Based on this value, a 2D subdictionary was chosen that matches the drop in B~1~ ^+^ for each volunteer. Each voxel signal in the reconstructed MRF image series was then matched to an element of the subdictionary. In this process, the best match between the measured signal and the dictionary elements was found for each voxel by solving$$m = \mathit{argmax}_{\mathbf{i} \in {\{{1,..,\mathbf{M}}\}}}\left\{ {\mathbf{d}_{\mathbf{i}} \bullet \mathbf{s}} \right\}$$ where $\mathbf{d}_{i} \in \mathbb{C}^{t \times 1}$ is the ith normalized dictionary element and $\mathbf{s} \in \mathbb{C}^{t \times 1}$ is the normalized measured signal. The index $m$ that maximizes the inner product describes the dictionary element $\mathbf{d}_{m}$ (with corresponding T~1~ and T~2~ values) that gives the best match with the measured signal. Finally, the scalar proton density per voxel was determined from the model$$\mathbf{S}{= rM}_{0}\mathbf{D}_{\mathbf{m}},$$ where $\mathbf{S} \in \mathbb{C}^{t \times 1}$ is the nonnormalized signal per voxel and $\mathbf{D}_{m} \in {}^{t \times 1}$ the nonnormalized dictionary element corresponding to the best match $\mathbf{d}_{m}$, such that$$M_{0} = \frac{1}{r}\frac{(\mathbf{D}_{m} \bullet \mathbf{S})}{(\mathbf{D}_{m} \bullet \mathbf{D}_{m})}$$ *r* is a value between 0 and 1, describing the fraction of the initial longitudinal magnetization that is left after the dummy trains, also depending on T~1~ and T~2~, which takes into account the short repetition delay in between the MRF trains. M~0~ maps are all shown on a log‐scale due to the high dynamic range of the respective proton densities, with that of the vitreous body being more than an order of magnitude larger than other structures. The processed T~1~, T~2~, and M~0~ maps were compared for different reconstruction methods (FFT, 2D CS, 3D CS, and MC) and for different acquisitions (low spatial resolution, high spatial resolution). T~1~ and T~2~ values were averaged in different regions of interest, annotated in Figure [1](#mrm27594-fig-0001){ref-type="fig"}E for each volunteer. These values were used to determine mean ± SD values over all volunteers for the different reconstructions. 3. RESULTS {#mrm27594-sec-0014} ========== 3.1. Simulation results {#mrm27594-sec-0015} ----------------------- Figure [2](#mrm27594-fig-0002){ref-type="fig"} shows the parameter maps (T~1~, T~2~, and M~0~) obtained for different reconstruction methods, after subsampling the fully sampled k‐space data of 1 healthy volunteer. Even though an incoherent sampling scheme was used, a zero‐filled FFT reconstruction does not lead to accurate parameter maps. The CS reconstruction with total variation regularization in the spatial domain leads to only minor improvement for the high undersampling factor that was chosen. The results show that including the sparsity constraint in the temporal dimension on top of the spatial dimension improves the CS reconstruction, with the largest improvement in the optic nerve and the lens nucleus, indicated by the white arrows. The total undersampling factor of 6.7, however, in combination with the low resolution reconstruction matrix and the single channel signal, results in loss of detail in the CS approach. {#mrm27594-fig-0002} This is not the case for the MC‐based reconstructions. The parameter maps resulting from the MC‐based approach are very close to the parameter maps obtained from the fully sampled scan, enabling visualization of the extraocular muscles and the orbital fat, indicated by the white circles. The error maps in Figure [2](#mrm27594-fig-0002){ref-type="fig"}, defined as the relative difference with the parameter maps from the fully sampled scan, given in percentages, confirm these findings. The error has a more noise‐like behavior for the MC‐based reconstruction compared with the CS reconstruction, and is much lower in the sensitive region of the eye coil. The error maps for T~1~ show larger percentage improvements compared with T~2~. These general trends were also true for different undersampling factors (see Supporting Information Figure [S1](#mrm27594-sup-0001){ref-type="supplementary-material"}, which is available online). 3.2. Experimental results {#mrm27594-sec-0016} ------------------------- Parameter maps obtained in an undersampled experiment are shown in Figure [3](#mrm27594-fig-0003){ref-type="fig"} for low spatial resolution images. The experimental results confirm the findings from the simulation study. The parameter maps obtained from the undersampled MRF scan with a 3D CS reconstruction show loss of detail compared with the parameter maps obtained with an MC‐based reconstruction. This is especially visible in the M~0~ maps. For the MC‐based reconstruction, the parameter maps are similar quality to those obtained from the fully sampled scans, showing the feasibility of accelerating MRF in the eye using a Cartesian sampling scheme. It should be noted that the full k‐space data and the undersampled k‐space data originate from different scans, which is why residual motion artifacts are different between the resulting parameter maps. The parameter maps at high resolution in Figure [4](#mrm27594-fig-0004){ref-type="fig"} show more detail compared with the parameter maps at low resolution in Figure [3](#mrm27594-fig-0003){ref-type="fig"}, indicated by the white circle. For the high resolution case, however, the 3D CS reconstruction gives larger improvements compared with the low resolution case. {ref-type="fig"}). The parameter maps obtained from a CS reconstruction show loss of detail. The quality of the maps obtained from the undersampled scan after an MC‐based reconstruction is comparable to the quality of the maps from a fully sampled scan. Inhomogeneities are visible in the vitreous body, which is very hard to accurately encode due to the low sensitivity of the MRF train for very long T~1~ values](MRM-81-2551-g003){#mrm27594-fig-0003} {ref-type="fig"} show more structural detail, indicated by the white circle. Note that Figure [3](#mrm27594-fig-0003){ref-type="fig"} and Figure [4](#mrm27594-fig-0004){ref-type="fig"} were different scans, in which motion artifacts are also different. Fully sampled data sets were not acquired for the high resolution case due to the prohibitively long scanning times required](MRM-81-2551-g004){#mrm27594-fig-0004} Parameter maps obtained in the 6 different volunteers for the low resolution scans are shown in Figure [5](#mrm27594-fig-0005){ref-type="fig"}. In all volunteers, some inhomogeneities are visible in the vitreous body, which is a region that is very sensitive to any type of motion or system imperfections because of the low sensitivity of the MRF sequence for very long T~1~ compared with short T~1~. This effect is illustrated in Figure [6](#mrm27594-fig-0006){ref-type="fig"}, where differences in short T~1~ values (500‐1000 ms) result in more distinguishable dictionary elements compared with the same absolute differences in long T~1~ values, (3500‐4000 ms) especially in the first half of the MRF train. These inhomogeneities differ slightly between successive scans in the same volunteer, and are more visible in the scans of volunteer 3 (Figure [5](#mrm27594-fig-0005){ref-type="fig"}C) and volunteer 5 (Figure [5](#mrm27594-fig-0005){ref-type="fig"}E). Overall, the shortened scan time reduces the risk of motion artifacts, which is clearly visible in volunteers 5 and 6 (Figure [5](#mrm27594-fig-0005){ref-type="fig"}E,F). The high resolution parameter maps for the same volunteers are shown in Supporting Information Figure [S2](#mrm27594-sup-0001){ref-type="supplementary-material"}A‐F, with several regions of improved structural detail indicated by the white circles. {ref-type="table"}. In some volunteers the inhomogeneities in the vitreous body appear stronger than in others, which probably correspond with cases of more motion. This can also be seen in (E,F), where the quality of the maps is better for the shorter scans (MC) compared with the fully sampled ones](MRM-81-2551-g005){#mrm27594-fig-0005} {#mrm27594-fig-0006} Average T~1~ and T~2~ values in the lens nucleus, the vitreous body, the orbital fat, and the extraocular muscles are reported in Table [1](#mrm27594-tbl-0001){ref-type="table"} for the different low resolution scans and reconstruction methods. The relaxation times obtained with a CS reconstruction are relatively close to those of the MC‐based reconstruction, but differences are observed in small anatomical structures such as the extraocular muscles and the eye lens. Differences between the relaxation times from the MC‐based reconstructions and the FFT of the fully sampled data can in part be explained by the fact that motion artifacts differ from scan to scan. Average relaxation times obtained from high resolution scans (not reported) follow the results for the low resolution scans. Reference T~1~ values at 7T reported in Richdale et al[15](#mrm27594-bib-0015){ref-type="ref"} are included in Table [1](#mrm27594-tbl-0001){ref-type="table"}; it should be noted that these reported values show large differences in relaxation times between different measurement techniques. ###### T~1~ and T~2~ values for different ocular structures (annotated in Figure [1](#mrm27594-fig-0001){ref-type="fig"}C), averaged within the structure and over 6 volunteers[a](#mrm27594-note-0002){ref-type="fn"} CS 3D MC Full 7T Richdale et al. -------------------- --------------- ---------- ---------- -------------------- Lens nucleus 1403±178 1037±220 996±248 1520/1020 Vitreous body 3632±375 3614±444 3599±334 5000/4250 Orbital fat 93±23 100±29 95±26 -- Extraocular muscle 731±342 1736±346 1545±191 -- **T~2~ (ms)** Lens nucleus 29±9 29±12 21±10 -- Vitreous body 139±14 147±20 145±12 -- Orbital fat 55±12 51±16 51±19 -- Extraocular muscle 67±26 50±12 55±25 -- Values, given in milliseconds, were averaged in different regions of interest (lens nucleus, vitreous body, orbital fat, and extraocular muscle) from the different scans at low resolution, using different reconstruction methods, for each of the 6 healthy volunteers. The resulting values were used to determine mean ± SD values over all volunteers. The CS reconstruction produced different relaxation times in small anatomical regions such as the lens nucleus and the extraocular muscles. The TRs for the MC‐based reconstructions are close to the values for the fully sampled scans. Remaining differences can be explained by motion artifacts that differ from scan to scan. Reference values at 7T (variable flip angle gradient echo/inversion recovery) from previous literature were reported in the last 2 columns, showing large differences in T~1~ values between different techniques. John Wiley & Sons, Ltd Parameter maps in a uveal melanoma patient are shown in Figure [7](#mrm27594-fig-0007){ref-type="fig"}, together with a T~2~‐weighted, fat‐suppressed, TSE image for anatomical reference. The tumor and the detached retina are characterized in the MRF maps by much lower T~1~, T~2~, and M~0~ values compared with the vitreous body, which allows for clear discrimination between tumor and healthy tissue. Dictionary matches and measured signals (both normalized) in the detached retina, the lens nucleus, the eye tumor, and the fat are also shown. The average values in regions of interest are reported in Table [2](#mrm27594-tbl-0002){ref-type="table"}. {#mrm27594-fig-0007} ###### T~1~ and T~2~ values for different ocular structures in a uveal melanoma patient[a](#mrm27594-note-0003){ref-type="fn"} T~1~(ms) T~2~(ms) ------------------------------- ---------- ---------- Lens nucleus 916 24 Vitreous body 4218 209 Orbital fat 112 84 Extraocular muscle 1282 56 Eye tumor 883 36 Liquid behind detached retina 1814 64 T~1~ and T~2~ values in milliseconds were averaged over drawn regions of interest. The eye tumor shows different relaxation times (both T~1~ and T~2~) compared with the vitreous body and with the liquid behind the detached retina, which allows for discrimination between tumor and healthy tissue. John Wiley & Sons, Ltd Reconstruction times for the different reconstruction methods were averaged over 6 healthy volunteers and reported in Table [3](#mrm27594-tbl-0003){ref-type="table"}. The iterative nature of CS and MC increases the reconstruction times compared with the direct FFT reconstruction, but the MC‐based reconstruction is much more time‐efficient because it is performed entirely in k‐space, and uses only fast matrix vector multiplications.[42](#mrm27594-bib-0042){ref-type="ref"} ###### Reconstruction times[a](#mrm27594-note-0004){ref-type="fn"} Computation time (s) -------------------------- ---------------------- ------ CS 3D (40 SB iterations) 584 2734 MC (100 iterations) 12 44 FFT 0.1 0.5 Mean values of reconstruction times in seconds calculated over 6 healthy volunteers for CS 3D, MC, and the direct FFT. The reconstruction times for both CS and MC take longer compared to the direct FFT due to the iterative process, but the MC‐based reconstruction is much more time‐efficient than the CS reconstruction because it is performed entirely in k‐space. John Wiley & Sons, Ltd 4. DISCUSSION {#mrm27594-sec-0017} ============= The results in the simulation study clearly show the benefit of using the temporal dimension in the reconstruction of MRF data, as is performed using MC. The low rank property of the signal evolutions allows higher undersampling factors than in a CS reconstruction, in which the TV operator was used to enforce sparsity in the temporal as well as in the spatial dimensions. The experimental results confirmed these findings, and showed the feasibility of reducing the MRF scan time with the proposed MC‐based reconstruction from 7:02 min to 1:16 min. Using MC, high resolution parameter maps can be obtained, which was out of practical reach for full sampling due to the long scan time. The technique was also demonstrated in a uveal melanoma patient, in which relaxation times showed a clear difference between tumor and healthy tissue. The CS reconstruction resulted in smoothed parameter maps, which averages out motion artifacts, but also reduces the amount of visible detail. One reason why the CS reconstruction did not perform as well as the MC‐based reconstruction might be that the TV operator is not the optimal sparsifying transform for transforming the measured data along the temporal domain. Other sparsifying transforms, such as the Wavelet transform or even learned transforms or dictionaries,[56](#mrm27594-bib-0056){ref-type="ref"}, [57](#mrm27594-bib-0057){ref-type="ref"} might result in improvements of the parameter maps after a CS reconstruction. For the high resolution data, however, the 3D CS reconstruction seemed to perform better compared with the low resolution case, while the MC‐based reconstruction performed well in both the low and the high resolution cases. This suggests that the CS reconstruction is more dependent on the resolution of the acquired data than MC, which might be explained by the fact that MC, as implemented here, does not incorporate any spatial correlation into the reconstruction process. Furthermore, reducing the resolution might reduce the sparsity of the images in appropriate transform domains, while this is one of the key ingredients for CS to work. Images from undersampled scans were reconstructed with MC, in which the chosen rank of the projection matrix influences the error. Here, the number of incorporated singular values was determined empirically in a simulation study: 4 singular values resulted in the smallest error after 100 iterations of the algorithm. Other sampling patterns, flip angle trains or anatomies will likely require new optimization of the projection matrix. In the current acquisition, 15% or 12.5% of the data was acquired with 6 or 8 fully sampled central k‐space lines for each image frame. Further tuning of the sampling pattern might improve the accuracy of the reconstructions or allow even shorter scan times. One should keep in mind, however, that the sampled k‐t lines are used to reconstruct the missing k‐t lines. Because higher undersampling factors result in shorter scan times, this reduces the risk of motion‐corrupted k‐space lines, but if there is still significant motion, this affects a larger percent of the acquired data. Therefore, care should be taken to find a balance between the scan time and the robustness of the reconstruction algorithm to motion. In this work, the projection matrix was constructed from the central k‐t lines of the measurement data. In Doneva et al,[42](#mrm27594-bib-0042){ref-type="ref"} it was shown that this type of projection matrix results in a more accurate reconstruction compared with a projection matrix constructed from randomly selected k‐t lines due to the lower SNR in the latter case. Other works have used the simulated MRF dictionary as calibration data, which would eliminate the need to fully sample the centers of k‐space.[41](#mrm27594-bib-0041){ref-type="ref"} Such an approach will probably show a steeper decay in normalized singular values due to the absence of noise and motion in the simulations (see Supporting Information Figure [S3](#mrm27594-sup-0001){ref-type="supplementary-material"}). The central k‐space based projection matrix, however, results in a smaller reconstruction error, indicating that the central k‐space approximates the rank of the measurement data better. Further work should investigate whether this approach could be advantageous in terms of mitigating motion artifacts. As an alternative approach to the method used in our work, in which a low‐rank constraint is added as a penalty term to the cost function, the low‐rank property of the unknown image series can be incorporated directly in the data fidelity term, transforming the minimization problem into a linear one, which may be beneficial in terms of computational costs.[41](#mrm27594-bib-0041){ref-type="ref"} It would be interesting to compare the accuracy of the 2 methods in future work. Although this study has shown the feasibility of using MR fingerprinting to characterize the relaxation times of different anatomical structures in the eye, eye motion can still be a limiting factor. The parameter maps presented in the results section show inhomogeneities in the vitreous body, which can be a result of different types of motion in the eye (see Supporting Information Figure [S4](#mrm27594-sup-0001){ref-type="supplementary-material"}). The presence of motion in combination with the long T~1~ of the vitreous body and the low sensitivity of the MRF train to these long values, make it challenging to accurately map the relaxation times in the vitreous body itself, as was shown in Figure [6](#mrm27594-fig-0006){ref-type="fig"}. Adopting a longer MRF train, as well as pattern optimization of the MRF train, might help to increase the encoding capability, but a longer time between the cued‐blinks will strongly increase the chance of blink‐induced artifacts. However, one should recognize from a clinical point‐of‐view that for almost all ocular conditions the vitreous body is not affected and, therefore, an accurate quantification of its T~1~ is clinically not relevant. Outer volume suppression pulses, applied immediately before the inversion pulse or during 0 flip angle phases in the MRF train, might offer a way to reduce the flow of fresh magnetization (caused by motion) coming from slices above and below the imaging slice or from the left and the right of the imaging field of view, during repetitions of the flip angle train. However, such an approach and its effect on the quality of the parameter maps has to be investigated further. The parameter maps corresponding to patient data showed a very large difference between tumor tissue and healthy vitreous body, suggesting that fully homogeneous regions of T~1~ in the vitreous body are not necessary for disease quantification and classification. Future work should investigate the extension of the current single slice approach to a 3D approach, such that the entire eye can be efficiently quantified from 1 scan. The measured relaxation times are different between volunteers, potentially explained by anatomical or other volunteer‐specific differences. Small differences in relaxation times were observed for different scans in the same volunteer, caused by motion artifacts that change from scan to scan, but overall they are consistent within each volunteer, which is important for the use of this technique in practice. Considering the large deviations in measured relaxation times between different studies, it will be interesting to compare the MRF technique to standard T~1~ and T~2~ mapping techniques on a patient‐specific basis, and in this way investigate the origin of deviations from mean values as well as compare the robustness to motion for the different techniques. It should be noted, however, that in Ma et al,[58](#mrm27594-bib-0058){ref-type="ref"} it was already observed that MRF values do not always agree perfectly with reference values from other techniques, and potential reasons for this need to be investigated. Parameter maps in the current study were not corrected for slice profile effects, but all experiments were performed using an RF pulse with a very high time‐bandwidth product, minimizing the effects as demonstrated in Ma et al.[58](#mrm27594-bib-0058){ref-type="ref"} The flip angle map, which is used as an input in the matching process, was produced with DREAM, in which the B~1~ ^+^ encoding slice thickness was set to be double the acquisition slice thickness to eliminate the slice profile effect.[51](#mrm27594-bib-0051){ref-type="ref"} Values for the optic nerve were not reported in this study because the optic nerve was not visible in all scans due to small differences in planning and anatomy, and the slice thickness of 5 mm makes the measured values in the optic nerve very sensitive to partial volume effects. These partial volume effects also complicate quantification of heterogeneous tumors. In particular, tumor relaxation values could become inaccurate due to averaging with the strong signal coming from the surrounding vitreous body. Planning the imaging slice through the tumor as well as through the center of the vitreous body, such that the imaging plane is perpendicular to the tangent along the retina, would help to reduce these effects. One limitation of the current study is the rather high slice thickness used (which is limited by the gradient strengths). With small changes in the sequence such as using a slightly longer echo time, acquisition and reconstruction of a 2‐mm‐thick slice is feasible (see Supporting Information Figure [S5](#mrm27594-sup-0001){ref-type="supplementary-material"}). The in‐plane resolution of 0.5 mm is satisfactory for tumor quantification and classification, as well as visualizing small structures such as the sclera and the ciliary body. The results in this study show the potential to perform ocular MRF in tumor patients. To adopt ocular MRF in clinics, the technique could be further tailored to quantify specifically the relevant T~1~ and T~2~ values of tumors. Extensions to multislice or 3D acquisitions could be developed such that the whole tumor volume can be covered and quantified. Further studies should investigate which clinical applications will benefit from ocular MRF and in that way explore the clinical relevance of the technique. In conclusion, the high undersampling factors used for this Cartesian, nonparallel imaging‐based approach shorten scan time and in this way reduce the risk of motion artifacts, which is most relevant for elderly patients, who typically experience difficulties focusing on a fixation target. Supporting information ====================== ###### **FIGURE S1** The effect of the undersampling factor on the performance of different reconstruction methods. Undersampled data sets were obtained by subsampling a fully sampled data set, while fixing the number of central k‐space lines to six for all undersampling factors. For larger undersampling factors, MC outperforms 2D and 3D CS. For undersampling factors smaller than three, MC has a slightly higher error compared to 3D CS. Overall, the error appears to be less affected by the undersampling factor for MC compared to the other reconstruction methods. Error measures are defined according to Equation 5 **FIGURE S2** The parameter maps in all healthy volunteers for high resolution scans. Parameter maps obtained in six healthy volunteers are shown in (a)‐(f), respectively. The CS 3D reconstruction performs better for the high resolution scans than for the low resolution scans, but the parameter maps still show loss of detail compared to the maps obtained from the undersampled scan after an MC‐based reconstruction, with examples indicated by the white circles. Fully sampled reference scans were not obtained due to the long scan time required. A zoomed‐in version of the MC result in volunteer 1 is shown in (g), and repeated in (h) with a different color scale **FIGURE S3** Comparison of 2 different projection matrices. (a) The normalized singular value vector of the simulated MRF dictionary shows a steeper decay compared to the normalized singular vector of the central k‐space data. (b) The reconstruction error (defined as in Equation 5) as a function of the n most significant left singular values, is smaller when using the central k‐space as calibration data. A rank 3‐4 projection matrix results in the smallest reconstruction error when using the central k‐space data **FIGURE S4** The effect of motion on the parameter maps. (a) Motion was simulated by randomly replacing 1 of the 12 acquired k‐space lines in each MRF frame by (type 1) its phase‐modulated version with a random phase shift between 0 and 2π, mimicking in‐plane rigid body motion and (type 2) white gaussian noise (matching the maximum intensity of the replaced k‐space line), representing the worst case scenario of a completely corrupted signal. For motion type 1 larger differences are visible in the vitreous body. Motion type 2 results in noise break‐through in the parameter maps. For both types of motion, less than 6% change in T~1~ was observed in the vitreous body, while the T~2~ of the eye lens was changed by more than 20%, underlining the nonlinear effect of motion on the parameter maps. (b) The singular values of the calibration data show a less steep decay when k‐space lines are corrupted by motion **FIGURE S5** Parameter maps obtained from a thinner slice. By increasing the echo time from 3.5 ms to 4.6 ms, a slice of 2 mm can be acquired, spatial resolution 1×1×2 mm^3^. With this slice thickness the resulting parameter maps are less susceptible to partial volume effects, but slightly more noise is present in the maps due to the reduced SNR in the MRF images ###### Click here for additional data file. The authors thank Mariya Doneva for helpful discussions on reconstruction, and Thomas O'Reilly and Luc van Vught for useful insights during data acquisition.

Historical conceptualizations of depression =========================================== There is a long tradition in phenomenologlcal psychopathology that stresses basic bodily alterations as core features of depressive states. Thus, Wernicke used the term "vital feelings" to describe certain somatic symptoms occurring in affective psychoses.^[@ref1]^ Vital feelings refer to the close relationship of the body to the awareness of self. They determine the way we experience our body and the impression we assume our physical presence makes on other people. Vital feelings are somatic affects localized In different parts of the body. Whereas vital feelings constitute the bodily background of our normal experiences, they may move to the fore In a depressive mood. For example, depressed patients very often complain of a headache which is described not exactly as an ordinary pain, but more as an unbearable pressure "like a band around the head." Other disturbed vital feelings affect the chest or the abdomen, and mediate unpleasant sensations of weight, tension, heaviness, or Inhibition, totally absorbing the focus of attention. In quite a similar way Dupré speaks of "coenestopathic states" which mean a distressing, qualitative change of normal physical feeling In certain areas of the body during an episode of depressive mood. It Is a global loss of vitality In which all bodily parts and functions may be altered, and all their performances depressed.^[@ref2]^ Kurt Schneider considered these disturbances of vital feelings to be the core of cyclothymic depression. In his psychopathologlcal assessment they were of paramount diagnostic significance In depressive Illness, more or less equivalent to the first-rank symptoms In schizophrenia.^[@ref3]^ Huber discriminated between vital disturbances on the one hand and vegetative symptoms In depression on the other.^[@ref4]^ Vital disturbances refer to the vital feelings just mentioned. They comprise a loss of general vital tone of the body, a prevailing fatigue or exhaustibility, and various forms of somatic dysesthesia, typically of a static, more localized character affecting head, chest, heart region, or abdomen. All-pervasive sensations of anesthesia, stiffness, and alienation of the total body may characterize a somatopsychic depersonalization in depression which may appear as a Cotard\'s syndrome in its extreme form. If the vital disturbances take on a peculiar form that is difficult to describe in ordinary everyday words, Huber speaks of a "coenesthetic" depression which must be typologically differentiated from the bizarre states of coenesthetic schizophrenia. Vegetative symptoms are closely associated with these vital disturbances and coenesthesias in depression. Disturbances of sleep, appetite, and digestion are most frequent. However, there may be many other vegetative symptoms in depression such as disordered salivation, transpiration and lacrimation, cardiac arrhythmias and dyspnea, loss of libido and various sexual dysfunctions, dys- or amen? orrhea, loss of or increase in body weight, decreased turgor of the skin, loss of hair, decrease in body temperature, nausea, vomiting, meteorism, dizziness, sweating, or sensations of coldness. Both vital disturbances, coenesthesias and vegetative symptoms, are typically coexistent with the well-known affective, behavioral, and cognitive symptoms of depression. With respect to the different settings of medical care, however, these psychological symptoms of depression may be masked by a dominant reporting of somatic symptoms. M. Bleuler addressed the point in his book *Depressions in Primary Care,* in 1943: *"It is a common and frequent observation that depressive patients with single somatic complaints come to the consulting room of the general practitioner, internal specialist, and even the surgeon, gynecologist, ophthalmologist, urologist and other medical specialists, and spontaneously, they only speak of somatic phenomena while concealing their state of depressive mood. They report palpitations, tightness of the chest, loss of appetite, obstipation, pollakiuria, amenorrhea and many others. Only when one looks at their psychic state does one discover that they report numerous hypochondriac ideas also in other areas, that in addition they produce depressive ideas of impoverishment and sin, that beyond that their whole stream of thoughts is inhibited, that the depression manifests itself not only in the somatic complaints reported, but in various other bodily expressions."^[@ref5]^* In spite of this long-standing psychopathological view on the somatic foundation of depressive mood, at least in moderate and severe clinical states, it is bewildering that the official psychiatric classification systems of the *Diagnostic and Statistical Manual of Mental Disorders,* 4th edition *(DSM-IV)* and the *ICD-10* *Classification of Mental and Behavioral Disorders. Clinical Descriptions and Diagnostic guidelines (ICD-10)* only marginally appreciate somatic symptoms as diagnostic criteria for depressive disorders while focussing on the psychological symptoms of affect and cognition. So, *DSM-IV* lists only three criteria of somatic symptoms for major depressive disorder: sleep disturbance, appetite disturbance, and fatigue or loss of energy. And correspondingly, in *ICD-10,* disturbances of sleep and appetite, loss of libido, and amenorrhea are the only somatic symptoms considered to be of diagnostic significance for major depression. Beyond this short list of predominantly vegetative symptoms, no painful physical symptoms are mentioned in either the *DSM-IV* or *ICD-10.* There seems to be a major shift In diagnostic practice, however; the second version of the *Diagnostic and Statistical Manual of Mental Disorders,* 4th edition, Text Revision *(DSM-IV TR)* now Includes new criteria referring to "excessive worry over physical health and complaints of pain (eg, headaches or joint, abdominal, or other pains)."^[@ref6]^ This supplement of diagnostic criteria Is Indicative of an againIncreasing awareness of the importance of somatic symptoms in depression. What is meant by "somatic" in somatic symptoms of depression? ============================================================= In the literature there are many terms used to describe somatic symptoms in depression: somatic, somatlzed, physical, bodily, somatoform, painful, psychosomatic, vegetative, medically unexplained, masked, etc.^[@ref7]^ These diverse terms refer to different theoretical or diagnostic concepts. For states of depressive mood the neutral term "somatic" is preferred, comprising various bodily sensations that a depressed individual perceives as unpleasant or worrisome. These dysesthesias are very often localized In certain body parts or organs, or may affect the whole body In Its vital condition, as In the case of fatigue or loss of energy. Several basic physical dysfunctions, such as those of sleep, appetite, or digestion, are also to be included in the term "somatic." In addition, It may be clinically relevant to differentiate between painful and nonpalnful somatic symptoms of depression. From a diagnostic perspective one has to keep in mind that somatic symptoms play a significant role both in primary psychiatric disorders, first and foremost depressive and anxiety disorders, and in somatoform disorders. And In differential diagnosis, somatic symptoms must be considered as possibly even Indicative of underlying somatic diseases. A diagnostic challenge may be seen In the well-known fact that depressive, anxiety, somatoform disorders, and medical conditions are frequently coexistent, or Interact In the Individual patient.^[@ref8]-[@ref10]^ Regarding the assessment of somatic symptoms, Kroenke correctly points out that diagnosis very often is more approximative than precise. Presented somatic symptoms may be either clearly attributed to a distinct medical disorder or be placed into one of the following heuristic categories: somatoform disorder, another primary psychiatric disorder (often depression and/or anxiety), functional somatic syndrome (eg, irritable bowel syndrome, fibromyalgia, chronic fatigue syndrome), "symptom-only" diagnosis (eg, low back pain, idiopathic dizziness) or only partially explained by a defined medical disorder (eg, many states of chronic pain).^[@ref11]^ Epidemiological studies may provide an illuminating survey of the prevalence of somatic symptoms in depressive disorders, especially those encountered in primary care, and the prognostic value of somatic symptoms regarding their development in the further course of illness. Somatic symptoms of depressive disorders in inpatient care and primary care =========================================================================== In a clinical study, Hamilton reported that somatic symptoms prevailed in a great majority of depressed patients.^[@ref12]^ Somatic symptoms, particularly somatic anxiety and fatigue, were documented in up to 80% of a sample of 260 women and 239 men suffering from major depression. These somatic symptoms very frequently had an underlying psychopathologically relevant hypochondriasis, both in women and men. This study confirmed earlier studies showing that depressive disorders with predominantly somatic presentation were likely to be the most common form of depression, both in inpatient and outpatient care.^[@ref13],[@ref14]^ Hagnell and Rorsman stressed the Indicative significance of somatic symptoms in depressed primary care patients regarding their risk of suicide.^[@ref15]^ Epidemiological studies designed to establish prevalence figures for depressive disorders In primary care during recent years have uniformly demonstrated that depressive disorders are highly prevalent at this level of medical care.^[@ref16]-[@ref19]^ For the great majority of depressed patients seeking professional help in the official health care system, general practitioners and internists are the decisive interface for diagnosis and treatment of depression.^[@ref20]^ Primary-care patients with depression very often present with somatic complaints. This seems to be more the rule than the exception worldwide.^[@ref21],[@ref22]^ Two of the three most common symptoms reported during a current depressive episode were somatic (tlred/no energy/listless: 73%, broken sleep/decreased sleep: 63%) as shown by the European Study Society study (DEPRES II).^[@ref23]^ This study, however, also underlined that 65% of the depressed primary care patients suffered from a concomitant medical condition pointing to some likely difficulties In differential diagnosis. The multlcenter International study (n =1146) conducted by the World Health Organization (WHO) confirmed that two thirds of the patients presented their depressive mood with somatic symptoms exclusively, and more than half complained of multiple medically unexplained somatic symptoms.^[@ref24]^ In another primary care study, Kirmayer et al arrived at a similar finding of patients presenting their depressive or anxiety disorders exclusively with somatic symptoms in an overwhelming majority (73%). The identified somatic symptoms were the main reason for the initial visit to the primary care physician.^[@ref25]^ In a US study in 573 patients with the diagnosis of major depression, two thirds (69%) complained of general aches and pains, hinting at a close relationship between pain symptoms and depression.^[@ref26]^ The diagnostic situation In primary care frequently manifests Itself, however, as somewhat more complicated. Many patients present only with a single or a few somatic symptoms which remain medically unexplained and do not fulfill the affective and cognitive criteria for a discrete depressive or anxiety disorder at the end of the clinical interview. Single somatic symptoms are the primary reason for more than 50% of patients visiting a general practitioner or an outpatient clinic. In some 20% to 25%, these somatic symptoms are recurrent or chronic. Somatic symptoms that remain unexplained after a careful medical assessment generally bear a high risk for psychiatric morbidity, regardless of the type of symptoms.^[@ref27]-[@ref29]^ Up to two thirds of these patients develop a depressive disorder in the medium term, and between 40% to 50% fulfill the criteria for an anxiety disorder.^[@ref30]-[@ref33]^ In a cross-sectional study in 1042 primary care patients, Gerber et al investigated the differential relationship between specific somatic complaints and underlying depressive symptoms. Some somatic symptoms showed a high positive predictive value (PPV) for depression: Sleep disturbances (PPV: 61%), fatigue (PPV: 60%), three or more complaints (PPV: 56%), nonspecific musculoskeletal complaints (PPV: 43%), back pain (PPV: 39%), amplified complaints (PPV: 39%), vaguely stated complaints (PPV: 37 %).^[@ref34]^ Some somatic symptoms are typically covarlant In the patients\' complaints without having received the nosological status of a discrete medical condition. These clusters of symptoms are instead considered as functional somatic syndromes and termed according to the diagnostic standards of the various medical disciplines, eg, fibromyalgia, chronic fatigue syndrome, and irritable bowel syndrome, etc. For some authors in psychiatry these functional somatic syndromes represent typical variants of somatoform disorders. There is still a controversial dispute in the medical literature, however, as to whether to assemble all these functional somatic syndromes within one general category of somatization,^[@ref35],[@ref36]^ or to split them up into separate clinical entities.^[@ref37]^ From an empirical standpoint, it is remarkable that among these syndromes there is a significant overlap on the level of symptoms and a strong association with depressive and anxiety disorders.^[@ref38]-[@ref41]^ A close relationship between states of depressive mood and symptoms of pain, especially of chronic pain, has been impressively established in many empirical studies.^[@ref26],[@ref42]-[@ref44]^ Depression and painful symptoms commonly occur together. As both conditions are highly prevalent in the general population, their frequent co-occurrence might be due to mere statistical coincidence.^[@ref45],[@ref46]^ From an empirical standpoint, however, the prevalence figures of coexistence are far beyond statistical expectation. In a meta-analytical survey, Bair et al demonstrated that around two thirds of all depressed patients treated in primary, secondary, and tertiary centers, both in outpatient and inpatient settings, report distressing painful somatic symptoms.^[@ref26]^ Conversely, the prevalence rate of major depression in patients with various pain syndromes is about 50%. There seem to be higher rates in clinical states characterized by multiple diffuse pain symptoms than by more defined types of pain. The risk of major depression is considered to be dependent on the severity, frequency, persistence, and number of pain symptoms.^[@ref47],[@ref48]^ From the perspective of primary care an epidemiological study assessing the predictive power of chronic pain for depressive morbidity showed that the prevalence rate of at least one chronic painful physical condition (CPPC) in the general population was 17.1%. At least one depressive symptom was present in 16.5% of subjects; 27.6% of these subjects had at least one CPPC. Major depression was diagnosed in 4% of subjects, and 43.4% of these subjects had at least one CPPC, which was 4 times more often than in subjects without depressive disorder.^[@ref49]^ This significant Interrelationship of CPPC and depression confirmed the earlier clinical advice of Katon, suggesting that if all patients with painful physical conditions were systematically assessed regarding a possible underlying depression, some 60% of all states of depression could be detected in primary care.^[@ref50]^ Generally, one has to keep in mind that, both from a cross-sectional and a longitudinal perspective, there is a relevant overlap of depressive, anxiety, and somatoform disorders, especially chronic painful physical conditions, among primary care patients presenting with medically unexplained symptoms.^[@ref51]-[@ref58]^ It is an important clinical finding that, with an increasing number of medically unexplained symptoms, the risk of an underlying depressive disorder increases in an impressive dose-response relationship. In a study which included 1000 adults and another study comprising 500 patients with a chief complaint of somatic symptoms, the presence of any somatic symptom increased the likelihood of a mood or anxiety disorder by two- or threefold. Only 2% of patients with no or only one somatic symptom had a mood disorder, but 60% of those patients presented nine or more somatic symptoms.^[@ref31],[@ref59]^ Patients with multiple medically unexplained somatic symptoms also show a greater amount of associated other psychiatric comorbidity.^[@ref60],[@ref61]^ Somatic symptoms in depression and rates of diagnostic recognition within primary care ====================================================================================== The typical form of presenting a depression In primary care Is via somatization. This form of somatic presentation, however, is considered to be one of the main reasons for low rates of recognition of depression In this sector of the medical care system.^[@ref20],[@ref62]^ It must be acknowledged that the alarmingly low figures of diagnosed and consecutively treated depressive disorders in only 25% to 33% of affected patients found in epidemiological studies during the early 1990s have increased up to some 60%. ^[@ref17],[@ref19]^ From a perspective of primary care, general practitioners are consulted by two groups of depressed patients who may pose a diagnostic challenge. Patients suffering from a medical condition have a frequent depressive comorbidity^[@ref23],[@ref63]^ These associated depressions often remain undetected, as the diagnostic focus of the primary care physicians is led by a dominant model of somatic disease.^[@ref64]^ Indeed, certain somatic symptoms such as sleep disturbances, diffuse bodily pains and aches, fatigue, changes of appetite, etc, may characterize both the pathophysiological process of a discrete medical condition and a depressive disorder as well. The differential diagnosis may be difficult. The role and significance of somatic symptoms for the diagnosis of depression in medically ill patients have been a controversial issue in the scientific literature. Meanwhile, a clinically reasonable consensus has been arrived at that the *DSM-IV* criteria for major depression do not require significant modification for patients with medical comorbidities.^[@ref65]-[@ref67]^ Somatic symptoms can positively contribute to a diagnosis if they are assessed in line with typical concomitant affective, behavioral, and cognitive symptoms of depression.^[@ref9]^ For a primary care physician It Is Important to know that at least 20% to 30% of patients with chronic medical conditions suffer from a coexisting depression.^[@ref68]^ It must be assumed that, even In those patients being diagnosed with an acute somatic disease for the first time, depression coexists In a significant percentage.^[@ref69]^ All In all, patients with medical conditions are to be considered as a risk group for nonrecognitlon of concomitant depression.^[@ref70]^ This especially applies to elderly medically ill patients.^[@ref71]^ In the other major group of depressed primary care patients, the somatic symptoms complained of very often remain medically unexplained. If one focuses on the mode of presentation, about 50% of the patients report somatic symptoms exclusively, and a minor percentage of some 20% present their depressive disorder with prevailing psychological, ie, affective and cognitive symptoms.^[@ref7],[@ref21],[@ref72],[@ref73]^ There is not, however, a categorical split between a somatic mode of presentation on the one hand and a psychological mode on the other. Rather, a broad spectrum of transition must be assumed, and the grading of somatization has an impact on the probability of recognition of an underlying depression.^[@ref25]^ As a rule, primary care physicians do not recognize a depression with an individual patient better when he or she is complaining of many actual medically unexplained somatic symptoms (here they rather prefer a diagnostic standpoint of wait and see), but when the patient returns again and again to consult because of these symptoms.^[@ref74]^ In addition, the extent of hypochondriacal worries and health anxieties facilitate, a correct diagnosis of depression.^[@ref75],[@ref76]^ Patients with somatic complaints that are not explained medically in an adequate way, however, do not represent a uniform group regarding diagnostic categorization. Besides depressive disorders, which in primary care manifest themselves according to the traditional concept of an endogenous type only in minority but instead show many atypical features,^[@ref77]-[@ref79]^ one must consider various anxiety and somatoform disorders in differential diagnosis.^[@ref60],[@ref61],[@ref80]-[@ref82]^ Again as a rule, there exists an Impressive overlap on the level of symptoms among all these diagnostic categories.^[@ref10]^ Aspects facilitating somatic symptoms in depression =================================================== Many factors may contribute to the form and extent to which a depression is presented in somatic symptoms. Female gender has been confirmed to be closely associated with somatization in many studies covering differential aspects on various theoretical levels.^[@ref83]^ In a gender differential analysis, Sllversteln draws some Interesting conclusions from the epidemiological data of the National Comorblty Survey.^[@ref84],[@ref85]^ By dividing respondents Into those who met overall criteria for major depression and exhibited fatigue, appetite, and sleep disturbances ("somatic depression") and those who met overall criteria without these somatic symptoms ("pure depression") she demonstrated gender differences only for "somatic depression" but not for "pure depression." The higher prevalence of "somatic depression" In females was strongly associated with a high frequency of anxiety disorders. Interestingly, this type of "somatic depression" among female patients already had Its onset during early adolescent years with predominantly bodily pains and aches. Wenzel et al attributed the higher prevalence of "somatic depression" in women largely to changes in appetite.^[@ref86]^ Gender differences can also be found in primary care. Women consistently reported most typical somatic symptoms at least 50% more often than men. Although mental disorders, above all depressive and anxiety disorders, were found to be correlated with this mode of somatic presentation, there was also an independent female gender effect on somatic symptom reporting.^[@ref87]^ In a later study Jackson et al found that among primary care patients with somatic symptoms, on the whole, women were younger, more likely to report stress, endorsed more "other, currently bothersome" symptoms, were more likely to have a mental disorder, and were less likely to be satisfied with the care.^[@ref88]^ A greater susceptibility of women, both to psychosocial stress and somatic illness stress, was held responsible for this higher prevalence of depressive and anxiety disorders in female patients.^[@ref89]^ A greater vulnerability to depressive and anxiety disorders on the one hand, and a strong neurobiological association to defined functional somatic syndromes (eg, fibromyalgia, irritable bowel syndrome, chronic fatigue syndrome) on the other may further increase the extent of this gender difference.^[@ref40],[@ref90]^ The disposition both to somatization and to depressive and anxiety disorder may be intermingled in various ways. Thus, a depressive mood may trigger the immediate illness behavior to enter the medical care system and to report somatized problems caused otherwise.^[@ref91]^ The very high frequency of somatic anxiety symptoms in patients with major depression may be interpreted by the idea that anxiety appears to be a major source of bodily distress and consecutive hypochondriasis, thus fostering somatization behavior.^[@ref12]^ Indeed, specific effects of depression, panic, and somatic symptoms on illness behavior must be considered.^[@ref92]^ Various causal illness interpretations, a tendency to amplify somatic distress, and difficulties In Identifying and communicating emotional distress, all have an impact on the form and extent of a somatic mode of presentation.^[@ref93]-[@ref95]^ Again, regarding the course of Illness, depressive and anxiety disorders following somatoform disorders may significantly contribute to the chronlflcatlon and complication of the latter.^[@ref39],[@ref96]^ From a perspective of etiologically relevant risk factors It Is a well-established epidemiological finding that the extent and severity of early adverse events, especially manifold traumatic experiences, are tightly connected with the mental and somatic state of adults. This general disposition may be detected In a series of psychiatric disorders, as In conversion and somatization syndromes,^[@ref97]-[@ref103]^ several chronic pain conditions,^[@ref104]-[@ref106]^ hypochondriacal attitudes,^[@ref107]^ factitious disorders,^[@ref98]^ and depressive, anxiety, and substance disorders.^[@ref108]-[@ref110]^ One can draw a basic conclusion from many epldemiologlcally designed longitudinal studies that the more a person has been exposed to severe and early trauma, the higher the risk will be that she/he will suffering from recurrent or chronic depression with pronounced suicidality, multiple medically unexplained somatic symptoms, especially chronic physical pain conditions with an onset already during adolescence or young adulthood, the more her/his psychic and somatic state as a whole will be negatively affected, and the more she/he will demonstrate abnormal illness behavior.^[@ref61],[@ref111]^ Culture and society are other factors that may have an important impact on the way a depressive mood is presented in a predominantly somatic way.^[@ref25]^ Interestingly, the comprehensive international WHO study on depression in primary care, conducted in 12 countries on different continents, was not able to identify clear cultural influences on the somatic mode of presenting a depression. A somatic presentation was much more common at centers where patients lacked an ongoing relationship with a primary care physician than at centers where most patients had a personal physician. This variable had a robustly differentiating effect beyond the various cultural settings.^[@ref24]^ Besides gender, culture, and type of patient-physician relationship, there may be many other factors influencing a more somatic mode of presentation, such as different ages in life cycle, association with medical conditions, earning a lower income, and imprisonment.^[@ref7],[@ref112]^ Burden of somatic symptoms in depression ======================================== Most patients who are psychopharmacologically treated for depression fail to reach full remission.^[@ref113]-[@ref114]^ A majority of patients may respond to antidepressants (by definition a reduction of symptoms by some 50% or more), but still suffer from residual symptoms. These residual symptoms are often somatic in nature. Symptoms of somatic anxiety and various painful conditions seem be especially common in states of incomplete remission.^[@ref115]^ Residual symptoms which are not treated must effectively be considered as a negative risk factor with respect to earlier relapse, and a more severe and chronic future course of illness.^[@ref116]-[@ref119]^ The clinical significance of somatic symptoms in depression may best be illustrated with the relationship between depression and painful physical conditions. In general, the worse the painful somatic symptoms, the more severe and the longer a depressive episode persists. In their general population-based study, Ohayon and Schatzberg found that depressed patients with chronic pain symptoms reported a longer duration of depressive mood (19.0 months) than those without chronic pain (13.3 months). In addition, a chronic physical pain condition in persons with at least one key symptom of depression was associated with an elevated rate of suicidal thoughts.^[@ref49]^ Fishbain considered chronic pain as a major suicide risk factor in depression.^[@ref120]^ Von Korff and Simon demonstrated a significant correlation between the intensity of pain symptoms and a worse outcome of depressive disorders. This worse outcome included more pain-related functional impairments, a worse state of general health, higher rates of unemployment, use of more opiates, more frequent polypharmacy, and more intensive utilization of medical services due to pain complaints. ^[@ref121]^ Although both painful and nonpainful somatic symptoms improve with antidepressant treatment, It Is the Intensity and extent of pain symptoms at baseline that significantly contribute to a less favorable response to medication, and to a longer duration of treatment necessary for a satisfying result, if at all.^[@ref122]-[@ref124]^ If one asssembles painful and nonpainful somatic symptoms of depression into a single dimension of somatization, It is this factor that must be correlated with an impressively increased overall use of health care services,^[@ref125]-[@ref127]^ to significant treatment nonadherence and a resulting higher risk of relapse and more chronic course of illness.^[@ref128]^ Again, a recurrent or chronic depression includes a higher risk of suicide^[@ref129]^ and an increased morbidity and mortality due to Illness-inherent factors or associated natural causes.^[@ref130]-[@ref132]^ All in all, it must be concluded that: when somatic symptoms, above all painful physical conditions, accompany the already debilitating psychiatric and behavioral symptoms of depression, the economic burden that ensues for patients and their employers increases considerably,^[@ref133]-[@ref134]^ the functional status may be hampered signifiacantly,^[@ref135]^ and the health-related quality of life is lowered dramatically^[@ref136]^ Neurobiological underpinnings of somatic symptoms in depression =============================================================== Various psychosocial and biological stressors may trigger a depression. Neurobiological processes underlying any depressive illness are manifold; this applies to the different somatic symptoms in particular. A strong heritable disposition, polygenetic in nature, seems to be established, but maladaptive neurobiological stress response systems already acquired by stressful and traumatic experiences during early development may play a major role in the pathophysiology of depression as well.^[@ref137]^ Dysfunctions in the serotonergic, noradrenergic, and dopaminergic neurotransmitter systems have been considered as relevant for quite a long time. Drawing from the neuroanatomical serotonergic tracts, starting in the midbrain raphe cell bodies and projecting to the frontal cortex, basal ganglia, limbic system, and hypothalamus on the one hand, of noradrenergic pathways originating in the locus ceruleus of the brain stem and projecting again to the same regions of the frontal cortex, limbic areas, and hypothalamus, but also uniquely to other parts of the frontal cortex and to the cerebellum on the other, Stahl stressed that deficiencies in the activity of specific pathways of serotonin and norepinephrine might account for the differential clinical phenomenology in depression. This seems to be true both for the typical psychological and somatic symptoms. Regarding somatic symptoms, especially vegetative symptoms such as changes in appetite or weight, lack of pleasure and sexual appetence, and sleep abnormalities, dysfunctional hypothalamic and sleep centers may be of paramount importance, all influenced by both serotonin and norepinephrine.^[@ref138]^ Fatigue, exhaustibility, or loss of energy, common distressing symptoms during a depressive episode, but also obstinate residual symptoms, may be mediated by different malfunctioning neuronal circuits that are regulated by multiple neurotransmitters.^[@ref139]^ Fatigue can be experienced as reduction in either mental or more physical vital feeling. Likely candidates for the neuronal structures that may mediate physical fatigue refer to brain areas regulating motor functions, such as striatum or cerebellum, but also to certain spinal pathways transferring sensory input from the body and thus modulating the perception of physical tiredness. In addition to serotonin and norepinephrine, dopamine may be involved in this process. Mental tiredness, on the other hand, may be mediated by diffuse cortical circuits and be influenced by cholinergic, histaminergic, noradrenergic, and dopaminergic neurotransmitters. The various painful somatic symptoms in depression may essentially be associated with serotonergic and noradrenergic pathways descending from brain stem centers to the spinal cord. An imbalance in these neurotransmitters, normally serving to inhibit the sensory input from the intestines, musculoskeletal system, and other body regions, may accentuate pain sensitivity.^[@ref26],[@ref140]^ As a matter of course, neither psychological nor somatic symptoms in depression can be explained by dysfunctional neurotransmitters exclusively. Many other neurobiological processes are involved in the pathophysiology of depression, such as an abnormal HPA axis with a disordered feedback mechanism of the corticotropin-releasing factor (CRF) -adrenocorticotropic hormone (ACTH) - Cortisol stress response, a reduced secretion of the neuropeptide hypocretin thus contributing to a desynchronization of the sleep-wake cycle, various abnormalities in the inflammatory system with an increased production of certain proinflammatory cytokines, a resulting depletion of the serotonin system, sickness behavior and depressive mood, reduced concentrations of various neurotrophlns such as brain-derived neurotropic factor (BDNF) causing Impaired neuroplastlcity, cell resistance, and neurogenesis.^[@ref137],[@ref141]-[@ref147]^ The intricate pathophysiological interplay of neuroendocrine stress response, inflammation, and neurotransmitter systems, both centrally and peripherally, may perhaps best be illustrated by the relationship between chronic pain conditions and depressive mood states (succinctly summarized in refs 148-150). In short, chronic stress evoked by chronic pain leads to a loss of negative glucocorticoid feedback in the (hypothalamic-pituitary-adrenocortical (HPA) axis and downregulation of the glucocorticoid receptors within the brain and the body periphery. Inflammation and nerve injury stimulate nociresponsive neurons within the dorsal horn of the spinal cord, and the relay of the nociceptive information ascends to the brain stem to be gated within the thalamus, prior to its cognitive appraisal within the somatosensory cortex. Monoamlnergic neurons In the brain stem normally descend to the spinal cord to act as a "brake" on nociceptive transmission. During chronic pain, loss of serotonergic and noradrenergic tone In response to glucocortlcold-lnduced monoamlnergic depletion may lead to descending Inhibitory Impulses to the spinal cord to effect an enhancement of pain sensation. Loss of glucocorticoid Inhibition of proinflammatory cytokines leads to proliferation of peripheral inflammatory events, contributing to pain sensitization. Although acute stress may be analgesic, implying an inhibitory circuitry between the limbic and somatosensory cortices, chronic stress evoked by chronic pain, leads to downregulation of glucocorticoid-mediated activity of this inhibitory connection, causing enhanced pain perception. Similarly, although acute pain may be mood-enhancing via both sympathetic and glucocorticoid routes (implying an excitatory reciprocal link between the somatosensory and limbic cortices), chronic pain-Induced downregulation of glucocorticoid modulation of this link may lead to depressed mood. Psychopharmacological implications for the treatment of somatic symptoms in depression ====================================================================================== Numerous trials with antidepressants have demonstrated that full remission of the psychological, and especially of the somatic, symptoms in depression can be achieved only by a minority group of depressed patients within a usual 6- to 8-week treatment period.^[@ref62],[@ref151],[@ref152]^ These sobering facts are reflected by a higher risk of relapse, a worse course of illness with many associated psychosocial disabilities, and a hampered health-related quality of life. Therefore, achieving a state of symptomatic remission must be a treatment goal of utmost clinical importance. Targeting both serotonin and norepinephrine in those neuronal circuits that mediate somatic symptoms is the most widely employed strategy to reduce painful and nonpainful somatic symptoms in depression.^[@ref90]^ In comparison with selective serotonin reuptake inhibitors, antidepressants with a dual action on both the serotonin and norepinephrine system were significantly superior in alleviating these somatic symptoms and achieving full symptomatic remission of depression. This may be a promising approach, even to treating chronic pain conditions, eg, fibromyalgia, without prevailing depressive symptoms.^[@ref153],[@ref154]^ This seems to have been well established In clinical trials with venlafaxlne,^[@ref155]-[@ref159]^ duloxetlne,^[@ref160]-[@ref163]^ mllnaclpran,^[@ref164]^ or mlrtazaplne.^[@ref165]^ In order to Improve distressing symptoms of fatigue, the use of psychostimulants, modafinil, bupropion, or selective norepinephrine reuptake inhibitors such as reboxetine or atomoxetine may be recommended.^[@ref166]^ As a rule, psychopharmacological efforts to treat severe states of depression or states of depression with prominent somatic symptoms effectively must be guided by a perspective of a longer duration than usual. Higher dosages of a selected antidepressant have to be used very often. Sometimes shifts within or between pharmacological classes of antidepressants or an augmentation with, eg, lithium or tri-iodthyronine, are necessary to arrive at the desired aim. From a pragmatic standpoint, clinically rational algorithms may favorably guide this endeavor.^[@ref167]^ Finally, it must be stressed that a reasonable combination of pharmacological and psychotherapeutic approaches can improve the treatment results in many depressed patients.^[@ref168],[@ref169]^

{#sp1 .466} {#sp2 .467} {#sp3 .468}

Irinotecan (CPT-11, Campto®) -- a semisynthetic, water-soluble derivative of the plant alkaloid camptothecin -- is the standard of care in the treatment of advanced colorectal cancer when 5-fluorouracil (5-FU)-based therapy has failed ([Cunningham *et al*, 2001](#bib5){ref-type="other"}). Phase II trials have demonstrated objective response rates of 16--27% in pretreated patients, with stabilisation of disease in a further 40--60% of patients ([Rougier *et al*, 1997](#bib13){ref-type="other"}; [Van Cutsem *et al*, 1999](#bib17){ref-type="other"}). Median overall survival rates of up to 10 months are achievable when irinotecan is used in relapsed/refractory colorectal cancer ([Shimada *et al*, 1993](#bib15){ref-type="other"}; [Rothenberg *et al*, 1996](#bib12){ref-type="other"}, [1999](#bib11){ref-type="other"}; [Pitot *et al*, 1997](#bib10){ref-type="other"}; [Rougier *et al*, 1997](#bib13){ref-type="other"}; [Van Cutsem *et al*, 1999](#bib17){ref-type="other"}). Two European phase III trials investigating the efficacy and safety of irinotecan, following 5-FU failure in advanced colorectal cancer, have demonstrated significant improvements in survival compared with best supportive care and 5-FU ([Cunningham *et al*, 1998](#bib6){ref-type="other"}; [Rougier *et al*, 1998](#bib14){ref-type="other"}). The main adverse events accompanying treatment with irinotecan in these trials were diarrhoea, neutropenia, fatigue, nausea and vomiting. Although 350 mg m^−2^ as an intravenous infusion every 3 weeks is the standard recommended dosage of irinotecan, pharmacokinetic parameters of irinotecan-lactone and the active metabolite SN-38-lactone vary between individuals ([Xie *et al*, 2002](#bib18){ref-type="other"}). This may be attributed to differences in the levels of the enzymes that metabolise irinotecan, notably carboxylesterase for SN-38. Furthermore, the variable interindividual patient exposure to SN-38 has been identified as an important determinant of toxicity ([Mathijssen *et al*, 2002](#bib8){ref-type="other"}). At the same time, there is convincing evidence of a dose--response relationship, and therefore a rationale for increasing doses when possible. In a phase I trial by [Abigerges *et al* (1995)](#bib1){ref-type="other"}, there were two recommended doses: 350 mg m^−2^ without high-dose loperamide and 600 mg m^−2^ with high-dose loperamide. With the exception of one responder treated at 260 mg m^−2^, all objective responses were observed at dose levels above 350 mg m^−2^. [Merrouche *et al* (1997)](#bib9){ref-type="other"} provided further support for this from a phase I trial in which an increased tumour response was seen at an irinotecan dose level of 500 mg m^−2^. Thus, these data suggest that a fixed-dose strategy for administration of irinotecan may not be optimal for all patients, thereby comprising treatment. The interindividual variability in pharmacokinetic parameters and dose--response relationship provided the rationale for investigating a dose optimisation strategy for irinotecan ([Chabot *et al*, 1995](#bib4){ref-type="other"}). The present study investigated different strategies, using doses of irinotecan up to 500 mg m^−2^, as single-agent therapy in the treatment of patients with metastatic colorectal cancer resistant to 5-FU. METHODS ======= Patients -------- Eligibility criteria included metastatic, histologically proven adenocarcinoma of the colon or rectum progressing on 5-FU-based chemotherapy (adjuvant and/or palliative); administration of ⩽2 5-FU-based regimens in the adjuvant setting or ⩽1 in the palliative setting; World Health Organization (WHO) performance status (PS) of ⩽2; adequate haematological, renal and hepatic function. Exclusion criteria included prior treatment with topoisomerase-I inhibitors; evidence of central nervous system metastases; prior history of chronic diarrhoea; current infection; or any other serious illness or medical condition. Study design and conduct ------------------------ This was a prospective, randomised, multicentre, open-label, phase II study. The study was conducted in accordance with the Declaration of Helsinki (Hong Kong revision, 1989) and with the approval of the Ethics Committee (Institutional Review Board) at each participating centre. Written informed consent was obtained from each patient prior to his or her enrolment into the trial. An independent Monitoring Committee regularly assessed the safety and efficacy issues and reviewed the conduct of the study if needed. An External Response Review Committee (ERRC) assessed tumour responses without knowledge of the randomisation arm. The aim of the study was to determine the optimal dosing strategy in terms of efficacy and safety of single-agent irinotecan (by individual dose optimisation based on patient tolerance to treatment, or optimisation based on specific baseline risk factors) in the treatment of 5-FU-resistant patients with metastatic colorectal cancer. The primary efficacy endpoint was the overall response rate. ### Dosing scenarios Patients were randomised to one of three groups (A, B and C (outlined below)), each group receiving irinotecan as a 30 min intravenous infusion scheduled every 21 days. This dosing interval could be extended to a maximum of 35 days in the event of persistent toxicity to allow satisfactory recovery from the previous cycle. Doses \<250 mg m^−2^ or \>500 mg m^−2^ were not used in this study; patients who exhibited significant toxicity at 250 mg m^−2^ were withdrawn from the study. Group A was the reference group in which a fixed dose of 350 mg m^−2^ of irinotecan was administered on Day 1. In subsequent cycles, the dose of irinotecan could be decreased (but not increased) according to the presence of significant toxicity at this dose. Groups B and C investigated dosing scenarios to select patients for whom the higher dose of irinotecan (500 mg m^−2^) could be optimally used. Patients randomised to Group B received irinotecan at a starting dose of 250 mg m^−2^ followed by increasing doses (350 and 500 mg m^−2^) depending on the tolerance observed in the preceding cycle. In the event of significant toxicity, dose reductions were implemented. In Group C, the irinotecan dose was based on protocol-defined toxicity risk factors identified at baseline: grade 3--4 neutropenia (bilirubin \>70% upper limit of normal (UNL), haemoglobin \<12 g dl^−1^, \>3 organs involved) and/or grade 3--4 diarrhoea (PS⩾1, creatinine \>70% UNL ([Freyer *et al*, 2000](#bib7){ref-type="other"})). Patients could be started at an irinotecan dose of 500 mg m^−2^ in the absence of toxicity risk factors. The starting dose of irinotecan was 350 mg m^−2^ in patients with one risk factor or one factor from each group, and 250 mg m^−2^ for patients with \>2 risk factors or two factors from the same group. The dose was not escalated, but could be reduced to 250 mg m^−2^ in the event of significant treatment-emergent toxicity. Concomitant treatments and follow-up ------------------------------------ Antiemetic drugs were administered as premedication to irinotecan infusions. Atropine was permitted for acute anticholinergic symptoms and loperamide (or similar) for delayed diarrhoea. In addition, preventative oral antibiotic therapy (e.g. an oral fluoroquinolone) was administered to patients with persistent (\>48 h) grade 4 diarrhoea or for diarrhoea associated with grade 3--4 neutropenia or fever. No granulocyte-colony-stimulating factor (G-CSF) support was allowed. All patients were followed until disease progression, unacceptable toxicity or death occurred, or the patient chose to withdraw from the trial. In all cases, in each group where toxicity necessitated a dose reduction, delay or study treatment termination, the patient was followed up until the event had resolved. Efficacy, safety and pharmacokinetic evaluations ------------------------------------------------ Tumour response rate, the primary efficacy end point, was measured according to WHO criteria and evaluated by the ERRC. Response was defined as complete (CR) plus partial (PR) response and as tumour growth control in terms of stabilisation of disease (PR plus no change/stable disease). Secondary efficacy variables were the duration of response and disease stabilisation, time to progression (TTP), time to treatment failure (TTF) and overall survival. The duration of response was measured from the first day of infusion of irinotecan to the first date that disease progression was noted or to the date of death for any reason. Time to progression was calculated from the date of randomisation to the first documented date of progression or the date of death for any reason. Time to treatment failure was the period between the date of randomisation and the date of tumour progression or treatment discontinuation for any reason. Survival was defined as the time between randomisation and death. Efficacy evaluations were performed using intent-to-treat (ITT) and per-protocol (eligible and evaluable) patient populations. The safety population comprised all patients who had started at least one infusion of study treatment. Safety was assessed according to the National Cancer Institute Common Toxicity Criteria or, if this was not applicable, graded as mild, moderate, severe or life threatening. The safety analysis was based on the worst grade by patient and by cycle. Deaths during the trial and up to 30 days from the last infusion were recorded. Pharmacokinetic evaluations were performed using a population approach ([Chabot *et al*, 1995](#bib4){ref-type="other"}; [Canal *et al*, 1996](#bib3){ref-type="other"}). At 30 min prior to infusion, and at 5 min and 3--4 h postinfusion (an additional sample was collected at 24 h postinfusion in some cases), three 5 ml blood samples (plus one predrug sample) were taken for analysis at the first cycle of chemotherapy for Groups A and C, and at the first, second and third cycles for Group B. Plasma levels of irinotecan and SN-38 were measured using reverse-phase high-performance liquid chromatography with camptothecin as an internal standard. Peak plasma concentration (*C*~max~) and the area under the plasma concentration--time curve (AUC) were calculated for both irinotecan and SN-38. In addition, total body clearance was calculated for irinotecan, and the time to reach *C*~max~ (*t*~max~) as well as the AUC normalised to l mg of irinotecan (AUC~N~) were calculated for SN-38. A three- and two-compartment model was used for irinotecan and SN-38, respectively. Statistical considerations -------------------------- Using the hypothesis that the response rate in Groups B and C would be 20%, a total of 64 patients in each of these groups were required to yield a confidence interval (CI) band of ⩽20%. For the reference group (Group A), the number of subjects randomised was half that of Groups B and C. The 95% CIs were estimated for response, using the exact method. Confidence intervals on median values were estimated using the method described by Brookmeyer and Crowley ([Simon *et al*, 1985](#bib16){ref-type="other"}). Descriptive statistics only were used for the pharmacokinetic parameters in each group. RESULTS ======= Patients -------- A total of 164 patients entered the study: 36 in Group A, 62 in Group B and 66 in Group C ([Table 1](#tbl1){ref-type="table"}). The majority of patients (⩾97%) had received surgery and 20--30% had received radiotherapy and/or prior adjuvant chemotherapy. Based on the assessment of baseline risk factors previously described, 23 (35%) patients in Group C were assigned to receive a starting dose of 250 mg m^−2^ irinotecan, 37 (56%) patients to 350 mg m^−2^ and six (9%) patients to 500 mg m^−2^. A total of 144 (88%) patients (31, 51 and 62 in Groups A, B and C, respectively) were eligible and evaluable for the efficacy analyses. Nine patients were ineligible due to major protocol violations (\>1 line of palliative chemotherapy, and past or concurrent history of neoplasm other than colorectal adenocarcinoma in one patient) and 12 patients (not mutually exclusive) were nonevaluable for response. Early discontinuation because of adverse events rendered eight patients nonevaluable. Extent of exposure to irinotecan -------------------------------- The median dose intensity of irinotecan was similar in the three arms: 114.21 mg m^−2^ week^−1^ (95% CI 76.14--119.21) in Group A, 101.36 mg m^−2^ week^−1^ (95% CI 68.22--158.17) in Group B and 106.69 mg m^−2^ week^−1^ (95% CI 67.11--170.93) in Group C. However, the median cumulative dose was higher in Group A (1948.80 mg m^−2^) than in Groups B (1564.26 mg m^−2^) and C (1326.77 mg m^−2^), possibly due to the longer median treatment time in this group (18 weeks, compared with 16 and 13 weeks in Groups B and C, respectively). The percentage of cycles delivered at doses of 250, 350 and 500 mg m^−2^ were as follows: 3, 92 and 0% (as this was not an option) in Group A; 41, 30 and 27% in Group B; and 33, 51 and 8% in Group C. A few cycles in each group were given at intermediate doses or at doses above 525 mg m^−2^. In Group B, the only dose escalation option, 63% of patients had at least one dose escalation from the 250 mg m^−2^ start dose. More than 80% of patients in each group did not require dose reduction. A total of 36--40% of patients experienced a cycle delay ([Table 2](#tbl2){ref-type="table"}). Although the majority of dose reductions in each group were made for treatment-related reasons (mostly nonhaematological adverse events across all arms), the majority of cycle delays occurred for reasons unrelated to treatment. Efficacy -------- ### Response rate In the total (ITT) patient population (*n*=164), the overall response rates were 8, 13 and 9% in Groups A, B and C, respectively ([Table 3](#tbl3){ref-type="table"}). There were no CRs. Tumour growth control rates were higher in Groups A and B and the rates of progressive disease were lower, compared with Group C ([Table 3](#tbl3){ref-type="table"}). The pattern of response across the groups was maintained in the per-protocol (eligible and evaluable) patient population (*n*=144), with overall response rates (no CR) of 10, 16 and 10% in Groups A, B and C, respectively. Corresponding tumour growth control rates were 61, 65 and 53%. Responses occurred at all dose levels ([Table 3](#tbl3){ref-type="table"}). However, there were only two responses at the 250 mg m^−2^ dose of irinotecan, both in Group C. Although it is difficult to interpret the data based on the small patient numbers in this study, they suggest that starting patients on a dose of 250 mg m^−2^ was not beneficial. The median duration of response and TTP were significantly longer in Groups A and B compared with Group C (*P*=0.030) ([Table 3](#tbl3){ref-type="table"}). Despite a trend towards a shorter TTF and median overall survival in Group C, there were no significant differences across the arms for these parameters. Safety and tolerability ----------------------- All patients were evaluable for safety. At least one adverse event was reported in all patients. However, grade 3--4 adverse events possibly or probably related to the study treatment were reported in less than half of the patients in each group ([Table 4](#tbl4){ref-type="table"}). Most of these were related to haematological or gastrointestinal (GI) events ([Table 4](#tbl4){ref-type="table"}). Grade 3--4 neutropenia with fever or infection was infrequent. Although anaemia was common, it was infrequently reported at grade 3--4 level of severity ([Table 4](#tbl4){ref-type="table"}). Diarrhoea was the most common GI event, occurring in 85% of patients, although grade 3--4 diarrhoea was less frequent (31, 21 and 27% in Groups A, B and C, respectively) ([Table 4](#tbl4){ref-type="table"}). There were no significant between-group differences for any of the adverse events reported. In addition, analysis of adverse events at the different dose levels showed no consistent evidence that toxicity increased with increasing dosage. There was no difference between the three treatment groups for the number of patients reporting ⩾1 grade 3--4 adverse event considered to be possibly or probably treatment-related (Group A, 42%; Group B, 48%; Group C, 49%). Overall, 74 serious adverse events considered possibly or probably related to study medication occurred in 39 patients. Treatment discontinuations -------------------------- At the designated study end date, 159 (96.95%). patients had discontinued treatment (Group A, 97%; Group B, 95%; Group C, 99%) ([Table 5](#tbl5){ref-type="table"}). Disease progression resulted in proportionately fewer discontinuations in Group B (57%) than in Groups A (72%) and C (80%), and included fatalities arising from progressive disease (one patient in each of Groups A and B, and two patients in Group C). There was also one fatality: a case of aspiration pneumonia secondary to vomiting in a patient in Group B receiving the first cycle of irinotecan 250 mg m^−2^. Five (42%) of the patients who discontinued treatment from Group B were receiving the 250 mg m^−2^ dose option during cycle 1 at the time of withdrawal. Adverse events leading to discontinuations are listed in [Table 5](#tbl5){ref-type="table"}. Pharmacokinetic parameters -------------------------- The principal pharmacokinetic parameters for irinotecan and SN-38 measured in 29 assessable patients are presented in [Table 6](#tbl6){ref-type="table"}. The mean total body clearance values of irinotecan were similar across all three groups and no relevant differences in dose-normalised exposure were seen. Exposure to irinotecan and SN-38 increased proportionally over the 250--500 mg m^−2^ irinotecan dose range. In the population pharmacokinetic analysis, exposure to irinotecan appeared to be increased in patients with PS 1 or 2, and in patients with high alkaline phosphate levels. DISCUSSION ========== The results of this phase II study confirm the activity of single-agent irinotecan in patients with metastatic colorectal cancer who have failed previous therapy with 5-FU. All three treatment strategies investigated were active and demonstrated acceptable tolerability patterns. Although almost all patients in the study had ⩾1 adverse event, less than half of the patients in each treatment strategy had grade 3--4 toxicity. The main aim of this study was to determine the optimal irinotecan dosing regimen for the treatment of this population, with the primary end point being response rate. The highest overall response rate was seen in patients in Group B (13%). In this group, four (21%) of the 19 patients receiving irinotecan 500 mg m^−2^ achieved a response. There was little difference in the overall response rates in Groups A and C (8 and 9%, respectively). An interesting observation in this study was the relatively higher rate of progressive disease in Group C (44%) compared with Groups A and B (36 and 31%). None of the differences in response rate between the groups were statistically significant. It is worth mentioning that the response rate observed in Group A was unusually low, and less than that seen in published studies of similar populations of patients treated with the same schedule ([Rougier *et al*, 1997](#bib13){ref-type="other"}; [Van Cutsem *et al*, 1999](#bib17){ref-type="other"}). This may be due to changes in first-line treatment that have occurred in recent years; compared with patients treated in earlier studies, those in the present study may have been more heavily pretreated with 5-FU and oxaliplatin in the first-line setting, thus making them more chemotherapy resistant. Despite the lower response rate in Group A, it is within the CIs of previous studies and so can be considered representative. The lack of a significant difference in overall response rates between the groups may reflect the fact that the median dose intensity of irinotecan delivered was relatively constant across the three groups, despite a proportion of patients in Groups B (34%) and C (9%) receiving an irinotecan dose of 500 mg m^−2^. This finding is probably due mainly to the fact that a disproportionate number of patients (more than one-third) in each of Groups B and C never received a dose of more than 250 mg m^−2^, and so were possibly underdosed. The likelihood of underdosing in Groups B and C is supported by the observation that only 6% of patients in Group A required dose reduction from 350 to 250 mg m^−2^. There were no significant differences between Groups A and B in TTP or overall survival. This may be due to an insufficient powering of the study and/or too small a difference in response rates. A previous meta-analysis conducted in patients with advanced colorectal cancer reported that only large differences in response rate correspond to a significant difference in TTP ([Buyse *et al*, 2000](#bib2){ref-type="other"}). Both TTP and duration of response were significantly shorter in Group C than in Groups A and B, and there was also a trend for a shorter overall survival in this group. The relatively poor efficacy seen in Group C may have been due to a combination of underdosing (i.e. a significant number of patients receiving irinotecan 250 mg m^−2^) and the small number of patients who received the high dose of irinotecan (500 mg m^−2^) (six patients or 9%). There was a trend towards a better safety profile in Group B. Grade 3--4 neutropenia was 31% in Group B, 47% in A and 44% in Group B. Similarly, there was less grade 3--4 diarrhoea in Group B as compared with Groups A and C (21 *vs* 31 and 27%, respectively), despite 34% of patients receiving the highest irinotecan dose. We cannot exclude the contribution to this difference of imbalances in gender ratio (more male patients in Group B) and PS (more patients with PS=0 in Group B). However, it is possible that the results reflect the aim of the strategy adopted in Group B, which was to avoid subjecting patients to higher doses than they were able to tolerate. Indeed 10 out of 12 patients in Group B who withdrew from the study due to treatment-related adverse events received the lowest dose of irinotecan (250 mg m^−2^) and therefore would not have tolerated an increased dose of irinotecan. However, it should also be noted that severe toxicity leading to treatment discontinuation occurred more frequently in Group B despite the low dose given to all patients in the first cycle. In Group C, despite the strategy of basing the initial irinotecan dose on predetermined risk factors, patients administered the 250 mg m^−2^ dose demonstrated higher rates of grade 3--4 anaemia and diarrhoea compared with those receiving the 350 and 500 mg m^−2^ doses. This study demonstrates that intrapatient dose escalation based on toxicity in the preceding cycle dose, as practised in Group B, is feasible. Although the increase in the response rate over the whole group was modest compared with the standard irinotecan dose, clinical benefit may be seen in those patients who are able to receive 350 and 500 mg m^−2^, which, in this study, was associated with a response rate of 25 and 21%, respectively. The findings of our study in pretreated patients are in agreement with those of a nonrandomised study in previously untreated patients ([Ychou *et al*, 2002](#bib19){ref-type="other"}): the greater proportion of patients who are able to receive the higher dose and the higher response rate achieved in the latter study compared with our study is probably a reflection of interstudy differences in the starting dose, dose escalation guidelines and in the study population (previous treatment compared with no previous treatment). In contrast with the feasibility of the strategy in Group B, the use of dose optimisation according to the baseline risk characteristics identified in our study protocol (as practised in Group C) appeared not to be an appropriate approach. This may be because the risk characteristics identified were not valid in this setting or that the algorithm for dose assignment was not relevant. Further investigation is required to clarify this. In conclusion, the data from our randomised phase II study suggest that individual dose optimisation based on toxicity in the preceding cycle is feasible and merits further investigation. Increasing the dose of irinotecan to 500 mg m^−2^ can be of benefit in selected patients. It will be necessary to identify the most appropriate starting dose, as the dose of 250 mg m^−2^ used in this study was possibly too conservative. Data from pharmacogenomic research are likely to be useful in the future for identifying the most appropriate starting dose of irinotecan for individual patients. The following additional investigators contributed to this study: F Cavalli (Switzerland), A Etxeberria (Spain), C Focan (Belgium), H Honegger (Switzerland), R Mathijs (Belgium), M Pestalozzi (Switzerland), M Symann (Belgium) and A Tres (Spain). ###### Patient demographics and baseline characteristics   **Treatment group** ---------------------------------------------------------------------------- --------------------------- --------------------------- ---------------------------- Number of patients (*n*); randomised (eligible and evaluable) 36 (31) 62 (51) 66 (62) Gender; male : female (%) 50 : 50 71 : 29 62 : 38 Age in years; median (range) 60 (29--71) 59 (33--70) 60 (30--70) Weight loss at baseline in relation to usual body weight (% of population)  ⩽5% 88.9 85.5 87.9  \>5% 5.6 4.8 3.0  Unknown 5.6 9.7 9.1  Mean loss (kg) 1.1 0.9 1.0 WHO PS  Median 1 0 1  0 (%) 50.0 59.7 45.5  1 (%) 44.4 35.5 53.0  2 (%) 5.6 4.8 1.5 Primary tumour location  Colon 63.9 66.1 66.7  Rectum 36.1 33.9 33.3 Number of organs with metastatic involvement; median (range) 2 (1--3) 2 (1--3) 2 (1--4) Synchronous metastases (%) 41.7 59.7 56.1 Sites of metastatic disease (%)  Liver 69.4 79.0 80.3  Liver alone 48.0 53.1 37.7  Liver and other organs 52.0 46.9 62.3  Lung 41.7 30.6 31.8  Peritoneum 11.1 4.8 13.6  Lymph nodes 11.1 21.0 22.7  Colon 0 6.5 1.5  All others^a^ 27.8 22.6 30.3 Median (range) time to randomisation (months) from  First diagnosis 18.1 (4.7--82.3) (*n*=35) 12.7 (3.0--76.3) (*n*=61) 12.6 (3.2--160.1) (*n*=66)  First metastasis 9.1 (0.0--54.7) (*n*=35) 9.0 (0.6--42.7) (*n*=62) 8.1 (0.1--51.6) (*n*=65) Prior anticancer treatment (% of patients)  Surgery 97.2 98.4 97.0  Radiotherapy 30.6 21.0 22.7  Adjuvant chemotherapy 33.3 25.8 21.2 At least one symptom at baseline (% of patients) 72.2 62.9 77.3 At least one abnormal laboratory value at baseline (% of patients) 97.2 95.2 93.9 Soft tissue, bone, adrenal, pelvis, abdomen, pleura, retroperitoneum, spleen, mediastinum, skin. WHO, World Health Organization. ###### Extent of exposure to irinotecan   **Treatment group** ----------------------------------------------------------- --------------------------- ----------------------------- --------------------------- Number of patients exposed 36 62 66 Number of treatment cycles 216 370 333 Median (range) number of cycles 6 (1--24) 5 (1--21) 4 (1--15) Median (range) treatment duration (weeks) 18 (3--78) 16 (3--64) 13 (3--46) Cycles by dose (% of cycles)^a^  250 mg m^−2^ 3 41 33  350 mg m^−2^ 92 30 51  500 mg m^−2^ --- 27 8 Median actual dose intensity (mg m^−2^ week^−1^) (95% CI) 114.21 (76.14--119.21) 101.36 68.22--158.17) 106.69 (67.11--170.93) Median cumulative dose (mg m^−2^) (95% CI) 1948.80 (314.65--8373.08) 1564.26 (247.52--10 100.00) 1326.77 (249.73--4899.13) At least one dose increase (% of patients) --- 63 --- At least one dose reduction^b^  % of patients 17 15 17  % of cycles 4 3 5 At least one cycle delayed^b^  % of patients 36 40 36  % of cycles 19 15 15 Some cycles were administered at intermediate doses. For any reason (see text). ###### Efficacy results **Parameter** **Group A (*n*=36)** **Group B (*n*=62)** **Group C (*n*=66)** ***P*-value^a^** --------------------------------------------------- ---------------------- ---------------------- ---------------------- ------------------ Overall response rate, % (95% CI)^b^ 8 (1.8--22.5) 13 (5.7--23.9) 9 (3.4--18.7)   Overall response rate, % (95% CI)^b^ Per protocol 10 (2.0--25.8) 16 (7.0--28.6) 10 (3.6--19.9)    250 mg m^−2^ ^c^ --- (0/16) 0% (2/20) 10% NC  350 mg m^−2^ ^c^ (3/31) 10% (4/16) 25% (4/36) 11% NC  500 mg m^−2^ ^c^ --- (4/19) 21% (0/6) 0% NC Tumour growth control rate (%) 58% 60% 50% NC Progressive disease (%) 36% 31% 44% NC Median duration of response (months) 6.4 6.6 4.3 0.03 Median TTP (months) 4.1 4.2 3.0 0.019 Median TTF (months) 3.7 3.4 2.5 NS Median overall survival (months) 12.5 12.1 10.9 NS Results are presented for the ITT population, unless otherwise stated. A *vs* C and B *vs* C. There were no CRs. Response rate is expressed as a percentage of patients treated at that dose level as their highest dose in each group. CI, confidence interval; NC, not calculated; NS, not significant. ###### Adverse events **Grade 3--4 adverse events^a^** **Treatment group: *n* (% of patients)** ----------------------------------------------------------- ------------------------------------------ --------- --------- At least one grade 3--4 adverse event^a^ 15 (42) 30 (48) 32 (48) Haematological  Leukopenia 9 (25) 15 (24) 21 (32)  Neutropenia 17 (47) 19 (31) 29 (44)  Anaemia 3 (8) 1 (2) 5 (8)  Infection (grade 3--4 neutropenia present) 2 (6) 0 2 (3)  Fever without infection (grade 3--4 neutropenia present) 0 2 (3) 3 (5) Gastrointestinal (GI)  Vomiting 5 (14) 10 (16) 6 (9)  Diarrhoea 11 (31) 13 (21) 18 (27)  Nausea 4 (11) 7 (11) 7 (11)  All other GI events^b^ 5 (14) 5 (8) 4 (6) Other adverse events  Fatigue 3 (8) 7 (11) 8 (12)  Fever (grade 3--4 neutropenia absent) 0 1 (2) 3 (5)  Infection (grade 3--4 neutropenia absent) 2 (6) 1 (2) 3 (5) Possibly or probably related to study treatment. Anorexia, five (3%) cases; cholinergic syndrome, three (2%) cases; GI pain, two (1%) cases; dehydration, three (2%) cases; stomatitis, one (1%) case. ###### Patient discontinuations   **Treatment group: *n* (% of patients)** --------------------------------------------------- ------------------------------------------ --------- --------- No. of patients still on treatment at cutoff date 1 (3) 3 (5) 1 (2) Total treatment discontinuations 35 (97) 59 (95) 65 (99) *Nonfatal reasons* Progressive disease 25 (69) 34 (55) 51 (77) Treatment-related adverse event 2 (6) 12 (19) 6 (9) Adverse events leading to discontinuation^a^  Fatigue 1 (3) 3 (5) 2 (3)  Vomiting 1 (3) 3 (5) 2 (3)  Diarrhoea --- 4 (7) 2 (3)  Nausea --- 2 (3) 2 (3)  Neutropenia --- 2 (3) 1 (2)  Febrile neutropenia --- 2 (3) ---  Neutropenic infection 1 (3) --- ---  Infection --- --- 2 (3)  Fever (infection absent) --- 1 (2) --- All other nonfatal events^b^ --- 5 (8) 1 (2)  Patient refusal 1 (3) 4 (7) 1 (2)  Other 6 (17) 7 (11) 4 (6) *Fatal reasons* Death due to treatment-related adverse events --- 1 (2) --- Death due to progressive disease 1 (3) 1 (2) 2 (3) Cardio-respiratory failure --- --- 1 (2) Not mutually exclusive. Patients may have discontinued treatment for more than one adverse event reason. Group B: aggravation reaction, two (3%) cases; anorexia, one (2%) case; dehydration, one (2%) case; small bowel obstruction, one (2%) case. Group C: anorexia, one (2%) case. ###### Pharmacokinetic profiles of irinotecan and SN-38 at different doses of irinotecan   **Arm A** **Arm B** **Arm C** ------------------------ ------------------- -------------------- -------------------- ------------------- ------------------- ------------------- Cycle 1 1 2 3 1 1 No. of patients 6 13 8 5 5 5 Dose (*n*) (mg m^−2^) 350 250 350 300 (1)/500 (4) 250 350 *Irinotecan* Infusion duration (h) 0.5--1.5 0.5--1.5 0.5--1.0 0.5--1.1 0.5--1.6 0.5--1.1 *C*~max~ (mg l^−1^) 5.88 (4.79--9.18) 4.55 (3.05--5.87) 6.12 (5.33--6.70) 8.40 (4.57--8.62) 3.61 (3.26--4.56) 7.13 (5.10--7.79) AUC (mg h l^−1^) 32.7 (14.3--36.4) 20.5 (11.6--30.9) 27.8 (20.4--39.2) 44.7 (28.0--50.6) 19.7 (15.3--29.0) 33.4 (22.9--46.3) Clearance (l h m^−2^) 9.3 (8.4--21.3) 10.6 (7.0--18.9) 10.9 (8.2--14.9) 9.1 (5.9--15.6) 10.6 (7.53--14.4) 9.03 (6.55--13.3) *SN-38* Median *t*~max~ (h) 0.7 (0.6--1.5) 0.6 (0.5--1.6) 0.6 (0.5--1.0) 0.7 (0.6--1.1) 1.5 (0.6--1.6) 1.0 (0.6--1.1) *C*~max~ (μg l^−1^) 61.9 (33.5--86.7) 49.7 (24.0--138.0) 58.7 (38.0--168.0) 80.7 (34.9--97.3) 40.6 (33.9--962) 67.9 (50.2--135) AUC (μg h l^−1^) 668 (362--1110) 676 (324--1140) 960 (546--1300) 1420 (609--1610) 595 (403--903) 768 (579--1395) AUC~N~^a^ (μg h l^−1^) 1.9 (1.1--2.9) 2.4 (1.3--5.0) 2.6 (1.6--4.4) 2.6 (1.3--5.1) 2.0 (1.5--3.2) 2.5 (1.5--4.4) Normalised to 1 mg irinotecan dose. Data are expressed as median (95% CI) unless otherwise stated. AUC, area under the plasma concentration--time curve; *C*~max~, maximum plasma concentration; *t*~max~, time to reach maximum plasma concentration.

Introduction {#sec1} ============ Acute aortic dissection (AAD) is a relatively uncommon medical emergency with a high mortality after symptom onset. The mortality of acute type A aortic dissection increases by 1--2% per hour during the first 48 h if no treatment is received \[[@cit0001]\]. Meanwhile, other common causes of acute chest pain, such as acute myocardial infarction (AMI) and pulmonary embolism (PE), also require rapid differentiation from AAD due to their critical and lethal characteristics \[[@cit0002]\]. However, the misdiagnosis rate of AAD has been reported to be approximately 30% on initial evaluation \[[@cit0003], [@cit0004]\]. Currently, noninvasive imaging modalities, including enhanced computed tomography (CT), transesophageal echocardiography (TEE) and magnetic resonance imaging (MRI), have been developed to improve the diagnosis of AAD, but these imaging modalities are expensive, time-consuming and unavailable at the bedside. Therefore, a rapid, cheap, reliable and sensitive laboratory test is urgently needed to diagnose AAD. D-dimer, the degradation product of cross linked fibrin, is significantly elevated in AAD patients \[[@cit0005]--[@cit0008]\] and has been suggested for use as a complementary marker to rule out AAD \[[@cit0005]--[@cit0007], [@cit0009]--[@cit0011]\]. However, in real-world clinical practice, AAD, PE and AMI are all thrombogenic diseases with high mortality, and whether the D-dimer level is helpful for differentiating these diseases remains to be elucidated. We therefore conducted a prospective cohort study to evaluate the validity and reliability of D-dimer level for differentiating AAD from other types of acute chest pain, including PE, AMI, unstable angina (UA), and other uncertain diagnoses of chest pain. Material and methods {#sec2} ==================== Study population {#sec2.1} ---------------- A single-center, prospective cohort study was conducted in Fuwai Hospital (the National Center for Cardiovascular Diseases in China) from January 2009 to January 2010. A series of consecutive patients with acute chest pain who presented to the emergency department (ED) of Fuwai Hospital within 24 h of symptom onset were enrolled in a prospective manner. Baseline clinical characteristics such as sex, age, Stanford types of AAD, intervals from onset of symptoms to hospital admission, medical histories, baseline parameters of physical examinations and laboratory tests including C-reactive protein (CRP), imaging examinations, in-hospital managements, ED diagnosis and discharge diagnosis were recorded according to pre-designed case report forms. The study protocols were approved by the appropriate institutional review boards of Fuwai Hospital and complied with the Declaration of Helsinki. All subjects provided written informed consent. D-dimer test and diagnosis {#sec2.2} -------------------------- Plasma D-dimer levels were measured using a stago-evolution device (France) in patients with chest pain immediately following admission. The results collected are expressed in micrograms per milliliter. The effective detection range of the assay is 0.22--20 µg/ml. Diagnoses of AAD and PE were confirmed by aorta or pulmonary angiography with multi-detector CT scan. Acute myocardial infarction was confirmed by acute chest pain, elevated cardiac-enzyme levels (cardiac troponin I or T, or the MB fraction of creatine kinase exceeded the 99^th^ percentile upper reference limit), documented findings of a new ST segment elevation/depression or a new T wave inversion on electrocardiography, and/or with evidence of obstructive coronary artery on angiography. Unstable angina was confirmed by chest pain, ST segment depression or T wave changes with evidence of obstructive coronary artery on angiography, but without the elevation of cardiac enzymes. Statistical analysis {#sec2.3} -------------------- Continuous variables are presented as mean ± SD or median and interquartile range according to whether they follow Gaussian distributions. Categorical data are presented as numbers and proportions. Baseline characteristics between groups were compared using Student's *t* test or the nonparametric Mann-Whitney test for continuous data and the χ^2^ test for categorical data. Receiver-operating characteristic (ROC) curves were constructed to calculate the sensitivity for AAD. The area under the curve (AUC) was calculated. A *p-*value \< 0.05 was considered statistically significant. The statistical calculations were performed with SPSS 19.0 (SPSS Inc., Chicago, Illinois, USA). Results {#sec3} ======= A total of 790 patients were enrolled, including 202 AAD, 43 PE, 315 AMI, 136 UA, and 94 cases with other uncertain diagnoses. Of the 202 AAD patients confirmed by CT angiography, 119 (58.9%) were Stanford type A AAD cases and 83 (41.0%) were Stanford type B AAD cases. Patient demographics and baseline characteristics are shown in [Table I](#t0001){ref-type="table"}. Compared to the patients with other causes of chest pain, AAD patients were more likely to be younger and male and tended to have concomitant hypertension but rarely have diabetes mellitus (all *p* \< 0.001). ###### Baseline characteristics of AAD patients and non-AAD (PE, UA, AMI, and uncertain diagnosis) Parameter AAD (*n* = 202) Non-AAD *P*-value ------------------------------------ ----------------- ----------- ------------ ------------ ----------- ---------- Age \[years\] 51 ±12 55 ±17 61 ±12 60 ±12 54 ±17 \< 0.001 Male, *n* (%) 169 (83.7) 21 (48.8) 102 (75.0) 254 (80.6) 65 (69.1) \< 0.001 Systolic blood pressure \[mm Hg\] 141 ±31 129 ±21 138 ±23 128 ±23 133 ±23 \< 0.001 Diastolic blood pressure \[mm Hg\] 80 ±21 81 ±10 87 ±57 79 ±14 81 ±14 0.535 Heart rate \[beats per minute\] 81 ±19 87 ±17 72 ±13 76 ±18 80 ±28 \< 0.001 Body mass index \[kg/m^2^\] 24.6 ±3.2 25.7 ±3.7 26.7 ±4.2 25.5 ±3.4 26.2 ±4.9 0.450 Creatinine kinases \[U/l\] 269 ±544 85 ±61 97 ±84 497 ±688 109 ±105 \< 0.001 Fasting blood glucose \[mmol/l\] 7.5 ±1.9 6.3 ±1.6 7.4 ±3.1 8.4 ±3.4 7.1 ±2.7 \< 0.001 Hypertension, *n* (%) 133 (65.8) 13 (31.0) 86 (63.2) 161 (51.3) 42 (46.2) \< 0.001 Diabetes mellitus, *n* (%) 5 (2.5) 2 (4.8) 31 (22.8) 68 (21.7) 13 (14.3) \< 0.001 Hypercholesterolemia, *n* (%) 18 (8.9) 3 (7.1) 34 (25.0) 75 (24.0) 13 (14.3) \< 0.001 Stroke, *n* (%) 10 (5.0) 2 (4.8) 13 (9.6) 33 (10.5) 7 (7.7) 0.471 Smoker, n (%) 64 (31.7) 7 (16.7) 31 (22.8) 105 (33.5) 18 (19.8) 0.060 Drinker, *n* (%) 21 (10.4) 0 (0.0) 6 (4.4) 14 (4.5) 6 (6.6) 0.110 AAD -- acute aortic dissection, PE -- pulmonary embolism, UA -- unstable angina, AMI -- acute myocardial infarction. The D-dimer level was elevated (\> 0.50 µg/ml) in 190 (94.1%) AAD patients. The D-dimer level in AAD patients was approximately 9-fold higher than that in non-AAD patients (median: 4.19 vs. 0.45 µg/ml, *p* \< 0.05). [Figure 1](#f0001){ref-type="fig"} shows the D-dimer level in patients with different causes of chest pain. The D-dimer level was significantly higher in patients with AAD than in patients with UA (median: 0.38 µg/ml, *p* \< 0.001), AMI (median: 0.45 µg/ml, *p* \< 0.001) and other uncertain diagnoses (median: 0.44 µg/ml, *p* \< 0.001), but it was comparable with that of PE patients (median: 2.72 µg/ml, *p* = 0.065). Similarly, the D-dimer level in PE patients was significantly higher than that in patients with UA, AMI, or other uncertain diagnoses (all *p* \< 0.001). Moreover, patients with type A AAD had higher D-dimer levels than those with type B AAD (median: 4.64 vs. 4.0 µg/ml, *p* = 0.022). {#f0001} [Figure 2](#f0002){ref-type="fig"} shows the ROC for patients with AAD versus non-AAD patients. The AUC value was 0.90 (95% CI: 0.87--0.93) for patients with AAD vs. all non-AAD patients. The AUC value was 0.59 (95% CI: 0.5--0.68) vs. PE, 0.91 (95% CI: 0.88--0.94) vs. AMI, 0.95 (95% CI: 0.93--0.97) vs. UA, and 0.93 (95% CI: 0.91--0.96) vs. patients with other uncertain diagnoses. Moreover, the best cut-off value of D-dimer for predicting PE was 1.14 µg/ml by ROC analysis with an AUC of 0.79 (95% CI: 0.74--0.84). The sensitivity and specificity were 88.4% and 71.2%, respectively. {#f0002} The diagnostic performance at the cutoff level of 0.5 µg/ml was analyzed. At this cutoff level, the sensitivity was 94.0% and the specificity was 56.8% for AAD compared to non-AAD patients; the negative and positive likelihood ratio were 0.10 and 2.18, respectively with a positive predictive value of 42.6% and a negative predictive value of 96.6%. The specificity was 4% for PE, 56% for AMI, 72.9% for UA, and 65.1% for uncertain diagnostic cases ([Table II](#t0002){ref-type="table"}). ###### Diagnostic performance of D-dimer at the cutoff level of 0.5 µg/ml Variable Sensitivity (%) Specificity (%) Youden's index PPV (%) NPV (%) PLR NLR ----------- ----------------- ----------------- ---------------- --------- --------- ------ ------ AAD 94.0 Non-AAD: 56.8 0.51 42.6 96.6 2.18 0.10  PE 4.0 --0.02 81.1 14.2 0.97 1.25  AMI 56.0 0.49 57.5 93.5 2.11 0.12  UA 72.9 0.67 83.7 89.2 3.48 0.08 Uncertain 65.1 0.56 86.3 83.3 1.44 0.09 PLR -- positive likelihood ratio, NLR -- negative likelihood ratio, PPV -- positive predictive value, NPV -- negative predictive value, AAD -- acute aortic dissection, PE -- pulmonary embolism, UA -- unstable angina, AMI -- acute myocardial infarction. Discussion {#sec4} ========== The present study demonstrated a significantly higher admission D-dimer level in patients with AAD within 24 h after symptom onset than those with AMI, UA, and other uncertain diagnoses. At the widely used cutoff level of 0.5 µg/ml, a favorable negative likelihood ratio of 0.10 and negative predictive value of 96.6% were found in patients with AAD. However, the D-dimer level was not significantly different between patients with AAD and PE. Our study suggests that a plasma D-dimer test within 24 h of symptom onset may be helpful for differentiating AAD and PE from other causes of acute chest pain. Acute aortic dissection is a catastrophic medical emergency, which requires early and accurate diagnosis and treatment. Imaging modalities, including enhanced CT and MRI, can facilitate an accurate diagnosis. However, these methods are limited due to unavailability at the bedside and their time-consuming nature, and they are not cost effective for routine screening. Thus, a rapid and reliable biomarker is urgently needed. Previous studies have evaluated several biomarkers for AAD, such as the smooth muscle myosin heavy chain \[[@cit0012]--[@cit0014]\], the BB-isozyme of creatine kinase \[[@cit0015]\], and calponin \[[@cit0016]\]. However, none of these markers have been adopted into routine clinical practice due to their inability to meet the requirements of a 'gold standard' biomarker including having adequate sensitivity and specificity in addition to a favorable time course of release that covers a time window necessary for nonambiguity in the clinical setting \[[@cit0017]\]. D-dimer is a fibrin fragment seen in coagulopathic disorders, and measurements are routinely used for the exclusion of venous thromboembolic diseases and PE \[[@cit0018]--[@cit0020]\]. In recent years, multiple studies have confirmed that D-dimer is elevated in AAD, and several studies have assessed its diagnostic value for AAD. However, at a defined cutoff value, the sensitivity and specificity of D-dimer for the diagnosis of AAD have been reported to vary, possibly due to different assay methods used in different studies. Generally, when a cutoff value of 0.5 µg/ml is used, the sensitivity and negative predictive value can reach almost 100% with a specificity of 54--68.6% \[[@cit0005], [@cit0009]\], and the specificity can be increased to 73% when the cutoff value is 0.626 µg/ml \[[@cit0006]\]. Shimony *et al.* \[[@cit0021]\] recently performed a meta-analysis of D-dimer to diagnose AAD and found that at a cutoff value of 0.5 µg/ml, the sensitivity and negative predictive value were 0.97 and 0.96, respectively. However, the specificity and positive predictive value were low, 0.56 and 0.60, respectively. Moreover, the negative likelihood ratio showed an excellent discriminative ability (0.06), whereas the positive likelihood ratio did not (2.43). They concluded that a plasma D-dimer level \< 0.5 µg/ml was a useful screening tool to identify patients who do not have AAD. Therefore, the plasma D-dimer level may thus be used to identify subjects who are unlikely to benefit from further aortic imaging. Our results were consistent with this study, suggesting that the cutoff value of D-dimer \< 0.5 µg/ml, which is widely used for excluding PE \[[@cit0022]\], is also applicable for the exclusion of AAD. However, the D-dimer level in patients with AAD is not always elevated, and several studies \[[@cit0023], [@cit0024]\], including ours, have observed this phenomenon. Hazui *et al.* \[[@cit0025]\] proposed that younger patients with a short dissection length and a thrombosed false lumen without ulcer-like projections may have false-negative D-dimer results. Therefore, patients who present classic characteristics of AAD but have a negative D-dimer test should receive further aortic imaging. Due to its non-specific characteristics, an elevated D-dimer level is also seen in patients with other morbidities such as PE, AMI, UA, and other diseases. Therefore, further investigation is necessary to clarify whether D-dimer tests can differentiate AAD from other diseases that presented with elevated D-dimer levels. Suzuki *et al.* \[[@cit0026]\] reported that when the cutoff level was 1.6 µg/ml, D-dimer was a useful tool for differentiating AAD from AMI, angina or other ischemic heart diseases within the first 6 h, and when the cutoff value was 0.8 or 0.9 µg/ml, the D-dimer level could differentiate AAD from AMI \[[@cit0027]\]. Sakamoto *et al.* \[[@cit0028]\] also found that a cutoff value of 0.5 µg/ml was effective for distinguishing AAD and PE from AMI, with a sensitivity of 68% and a specificity of 90%. Although their results were mostly consistent with ours, the cutoff values used in these studies were different and the obtained D-dimer levels in various causes of acute chest pain varied greatly. One possible explanation for this variation was the different measurement equipment and the test strip used. Therefore, a standard and unified detection protocol may improve the heterogeneity of measurement, making the detection value more reliable. Additionally, the D-dimer level was elevated in both AAD and PE patients, with no significant difference in our study, consistent with the findings of Sakamoto *et al.* \[[@cit0028]\] and Eggebrecht *et al.* \[[@cit0006]\]. Given the high mortality of the two morbidities, immediate contrast CT imaging or tissue Doppler imaging \[[@cit0029]\] may be good choices to differentiate AAD from PE. In the setting of AMI/UA, rupture of atherosclerotic plaques causes thrombopoiesis and activates fibrin degradation, leading to D-dimer formation. Therefore, D-dimer is elevated in patients with AMI/UA but not in patients with stable angina and healthy controls \[[@cit0030], [@cit0031]\]. Although the D-dimer level does not directly reflect the degree of myocardial damage, it has been confirmed that an elevated D-dimer level is a strong predictor of mortality in patients with AMI/UA \[[@cit0032], [@cit0033]\]. Therefore, the D-dimer level is not only a useful tool for the differentiation of diagnoses, but it also plays an important role in the prognostic evaluation for some cardiovascular diseases. Some limitations of the present study need to be addressed. First, although our study shows good prediction for AAD with the D-dimer level at the cutoff of 0.5 µg/ml, the specificity is low (56.8%). Indeed, D-dimer as a diagnostic biomarker of AAD did have some limitations due to the relatively high false positive rate. Therefore, for patients with a D-dimer level \> 0.5 µg/ml, the D-dimer level should be combined with other diagnostic tests, especially imaging tests, for an accurate diagnosis of AAD. Second, the small sample size of PE patients may affect the statistical power. Furthermore, the difference in D-dimer levels was not evaluated between patients with ST-segment elevation AMI and non-ST-segment elevation AMI. Therefore, further large, prospective, multi-center studies are needed. In conclusion, the D-dimer level within 24 h after symptom onset might be helpful for differentiating patients with suspected AAD from other causes of chest pain. The first two authors contributed equally to this study. We wish to thank the patients for their participations in our study, and we are also grateful to other clinical doctors and nurses for their help in the study. This work was supported by a grant (81170286) from the National Natural Science Foundation of China to Dr. Fan Xiaohan. Conflict of interest ==================== The authors declare no conflict of interest.

Background ========== Polysaccharide-rich fungi and plants have been employed for centuries by cultures around the world for their dietary and medicinal benefits \[[@B1]-[@B5]\]. Often thought to merely support normal bowel function and blood glucose and lipid levels \[[@B6]-[@B8]\], certain polysaccharides have attracted growing scientific interest for their ability to exert marked effects on immune system function, inflammation and cancers \[[@B9]-[@B11]\]. Many of these chemically and structurally diverse, non- to poorly-digestible polysaccharides have been shown to beneficially affect one or more targeted cellular functions *in vitro*\[[@B11]-[@B16]\], but much of the *in vivo*literature consists of studies in which polysaccharides were injected \[[@B1],[@B2]\]. For clinicians and scientists interested in immunologic effects following dietary intake, the value of such studies is uncertain. Polysaccharides that elicit effects *in vitro*or by injection may be ineffective or have different effects when taken orally \[[@B17]\]. We thus decided to conduct a systematic review to evaluate the specific immunologic effects of dietary polysaccharide products on rodents and human subjects. Methods ======= Literature review ----------------- Studies were identified by conducting electronic searches of PubMed and Google Scholar from their inception to the end of October 2009. The reference lists of the selected articles were checked for additional studies that were not originally found in the search. Study selection and data extraction ----------------------------------- The following search terms were combined with the term polysaccharide: dietary AND immune, or oral AND immune, or dietary AND inflammation, or oral AND inflammation. When specific polysaccharides or polysaccharide-rich plants and fungi were identified, further searches were conducted using their names with the same search terms. Studies were selected based on the following inclusion criteria: 1\. Rodent or human studies 2\. The presence of test group and control group (using either placebo, crossover, sham, or normal care) 3\. Studies reporting statistically significant immunomodulatory effects 4\. English language 5\. Studies published up to October 2009. Two researchers (JER, EDN) reviewed the list of unique articles for studies that fit the inclusion criteria. Uncertainties over study inclusion were discussed between the researchers and resolved through consensus. Searches were then conducted to obtain specific polysaccharide product information: safety (using the search terms: toxicity, NOAEL, LD~50~), composition and structure, and disposition. Quality assessment ------------------ Each study was assessed as to whether or not it reported a significant outcome measure for the polysaccharide intervention group. Results ======= A total of 62 rodent publications (Tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}) and 15 human publications (Table [4](#T4){ref-type="table"}) were deemed appropriate for inclusion in this review. Available structural and compositional information for these immunomodulatory polysaccharides are provided in Table [5](#T5){ref-type="table"} and safety information is provided in Table [6](#T6){ref-type="table"}. The majority of animal studies explored models in which animals were injected or implanted with cancer cells or tumors, were healthy, or were exposed to carcinogens. Other studies investigated immunodeficient, exercise-stressed, aged animals, or animals exposed to inflammatory agents, viruses, bacterial pathogens, pathogenic protozoa, radiation or mutagens. Human studies assessed immunomodulatory effects in healthy subjects, or patients with cancers, seasonal allergic rhinitis or aphthous stomatitis. Because of the limited number of human studies, we included some promising open-label controlled trials. Human study durations ranged from four days to seven years; daily doses ranging from 100-5,400 mg were reported to be well-tolerated. ###### Immunomodulatory Glucan Extracts: Oral Animal Studies ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Source Extract Animal Dose/day Duration of study Treatment Effects Reference ------------------------------------- ---------------------------------- --------------------------------------------------------------------------- ---------------------------------------------------------------------- ------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ ------------ *Agaricus*\ α-1,6 and\ 8-week ♀ C3H/He mice (5/group) 100 mg/kg IG every 3 days 1 month Healthy animals ↑ \#s splenic T lymphocytes (Thy1.2, CD4+ and CD8+) \[[@B24]\] (*A. blazei*) *subrufescens* α-1,4 glucans Aqueous 7-9-week ♂ Balb/cByJ mice (40/group) 1 ml 0.45N, 0.6N, or 3N aqueous extract 2 months All doses ↑ serum IgG levels, CD3+ T cell populations and PML phagocytic activity \[[@B22]\] 7-9-week male Balb/cByJ mice (40/group) 1 ml 0.45N, 0.6N, or 3N aqueous extract 10 weeks IP injection of OVA at 4 weeks 0.6N and 3N ↑ levels of OVA-specific serum IgG 28 days post-immunization; all doses ↑ delayed-type hypersensitivity and TNF-α secreted from splenocytes at 10 weeks; 0.6N ↑ splenocyte proliferation at 10 weeks 5-6 -week ♀ BALB/cHsdOla mice (8/group × 2) One 200 μl extract day 1, orogastric intubation 1 week Injected IP fecal solution day 2 ↓ CFU in blood of mice with severe peritonitis & improved overall survival rate in all peritonitis groups \[[@B46]\] 6-week BALB/c nu/nu mice (7/group) 2.5 mg extract days 20-41, drinking water 41 days Injected SC Sp-2 myeloma cells day 1 ↓ tumor size & weight after 21 days treatment \[[@B65]\] Aqueous, acid treated 6-week ♀ C57BL/6 mice (10/group) 20, 100 or 500 μg/ml, drinking water 9 days Injected IP human ovarian cancer cells day 1 500 μg/ml ↓ tumor weight \[[@B66]\] 20, 100 or 500 μg/ml, drinking water 3 weeks Injected IV murine lung cancer (3LL) cells 100 & 500 μg/ml ↓ \#s metastatic tumors Aqueous, with 200 ng/day\ 6-week ♀ BALB/c mice (10/group) 200 ng days 5-21 3 weeks Injected Meth A tumor cells day 1 ↓ tumor size & weight \[[@B23]\] β-glucan 2 weeks Injected Meth A tumor cells ↑ cytotoxic T lymphocyte activity & spleen cell IFN-α protein 300 mg 5 days Healthy animals ↑ splenic NK cell activity *Avena*spp. β-glucans (particulate) 6-7 -week ♀ C57BL/6 mice (7/group) 3 mg every 48 h, days 1-3 1 month Oral *E. vermiformis*oocytes day 10 ↓ *E. vermiformis*fecal oocyte \#s; increased intestinal anti-merozoite IgA; ↓ \# of IL-4-secreting MLN cells \[[@B42]\] 3 mg on alternating days, days 1-10 22 days Injected IP *Eimeria vermiformis*day 10 ↓ *E. vermiformis*fecal oocyte \#s; ↑ anti-merozoite intestinal IgA \[[@B43]\] β-glucans (soluble) 4-week ♂ CD-1 mice (24/group) 0.6 mg/ml 68% β-glucan, drinking water 1 month Resting or exercise-stressed (days 8-10) animals administered HSV-1 IN\ ↓ morbidity in resting and exercise-stressed animals; ↓ mortality in exercise-stressed animals; pre-infection, ↑ Mø anti-viral resistance in resting and exercise-stressed animals \[[@B38]\] day 10 \~3.5 mg days\ Resting or exercise-stressed (days 5-10) animals administered HSV-1 IN\ Pre-infection, ↑ Mø antiviral resistance in resting animals \[[@B41]\] 1-10, drinking water day 10 4-week ♂ CD-1 mice (10/group) 0.6 mg/ml 68% β-glucan, drinking water 10 days Resting animals or animals exposed to a bout of fatiguing exercise days 8-10 or moderate exercise days 5-10, injected IP with thioglycollate on day 10 ↑ neutrophil mobilization in resting & moderately exercised animals; ↑ neutrophil respiratory burst activity in resting and fatiguing exercised animals \[[@B37]\] 4-week ♂ CD-1 mice (19-30/group) 0.8 mg/ml 50% β-glucan, days\ 1 month Resting or exercise-stressed (days 8-10) animals administered IN clodronate-filled liposomes to deplete Mø days 8 & 14 & infected IN with HSV-1 day 10 ↓ morbidity, mortality, symptom severity in exercise-stressed animals, without Mø depletion \[[@B40]\] 1-10, drinking water 4-week ♂ CD-1 mice (20/group) Resting or exercise-stressed (days 8-10) animals administered HSV-1 IN day 10 ↓ morbidity in exercise-stressed & resting animals; ↓ mortality in exercise-stressed animals \[[@B39]\] *Ganoderma lucidum* Aqueous 7-week ♂ CD-1 mice (26/group) 5% of diet 5 months Injected IM DMH once a week, weeks 1-10 ↓ aberrant crypt foci per colon, tumor size, cell proliferation, nuclear staining of β-catenin \[[@B69]\] 4-8-week BALB/c mice (10/group) 50, 100 or 200 mg/kg, oral 10 days Injected SD Sarcoma 180 cells ↓ of tumor weight was dose dependent: 27.7, 55.8, 66.7%, respectively \[[@B67]\] *Ganoderma lucidum*(mycelia) Aqueous 7-week ♂ F344/Du Crj rats (16/group) 1.25% or 2.5% of diet 6 months Injected SC AOM once a week, weeks2-5 Both doses ↓ colonic adenocarcinoma incidence; 2.5% ↓ total tumor incidence; both doses ↓ nuclear staining of β-catenin and cell proliferation \[[@B68]\] *Ganoderma tsugae* Aqueous 8-week ♀ BALB/cByJNarl mice (14/group) 0.2-0.4% of diet (young fungi); 0.33 or 0.66% of diet (mature fungi) 5 weeks Injected IP OVA days 7, 14, 21; aerosolized OVA twice during week 4 In splenocytes, both doses of both extracts ↑ IL-2 and IL-2/IL-4 ratios, 0.2% young extract and 0.66% mature extract ↓ IL-4; in Mø, 0.66% mature extract ↑ IL-1β, both doses of both extracts ↑ IL-6 \[[@B53]\] *Grifola frondosa* D fraction Mice: 1) ICR, 2) C3H/HeN, 3) CDF~1~(10/group) 1.5 mg every other day, beginning day 2 13 days Implanted SC: 1) Sarcoma-180, 2) MM-46 carcinoma, or 3) IMC carcinoma cells ↓ tumor weight & tumor growth rate: 1) 58%, 2) 64%, and 3) 75%, respectively \[[@B71]\] 5-week ♂ BALB/c mice (10/group) 2 mg,\ 45 days Injected in the back with 3-MCA, day 1 ↓ (62.5%) \# of animals with tumors; ↑ H~2~0~2~production by plasma Mø; ↑ cytotoxic T cell activity \[[@B72]\] days 15-30 *Hordeum vulgare* β-1,3;1,4 or β-1,3;1,6-D-glucans Athymic nu/nu mice\ 40 or 400 μg IG for 4 weeks 31 weeks Mice with human xenografts (SKMel28 melanoma, A431 epidermoid carcinoma, BT474 breast carcinoma, Daudi lymphoma, or LAN-1 neuroblastoma) ± mAb (R24, 528, Herceptin, Rituximab, or 3F8, respectively) therapy twice weekly 400 μg + mAb ↓ tumor growth & ↑ survival; higher MW ↓ tumor growth rate for both doses \[[@B75]\] (4-12/group) β-1,3;1,4-D-glucans Athymic BALB/c mice 4, 40, or 400 μg for 3-4 weeks 1 month Mice with neuroblastoma (NMB7, LAN-1, or SK-N-ER) xenografts, ± 3F8 mAb therapy twice weekly 40 and 400 μg doses + mAB ↓ tumor growth; 400 μg dose ↑ survival. Serum NK cells required for effects on tumor size \[[@B76]\] C57BL/6 WT and CR3-deficient mice (10/group) 0.4 mg for 3 weeks 100 days Injected SC RMA-S-MUC1 lymphoma cells day 1 ± IV 14.G2a or anti-MUC1 mAb every 3rd day ±mAB ↓ tumor diameter; ↑ survival \[[@B73]\] β-glucans ♀ Fox Chase ICR immune-deficient (SCID) mice (9/group) 400 μg days 1-29 50 days Mice with human (Daudi, EBV-BLCL, Hs445, or RPMI6666) lymphoma xenografts, ± Rituximab mAb therapy twice weekly +mAB ↓ tumor growth and ↑ survival \[[@B74]\] *Laminaria digitata* Laminarin ♂ ICR/HSD mice (3/group) 1 mg 1 day Healthy animals ↑ Mø expression of Dectin-1 in GALT cells; ↑ TLR2 expression in Peyer\'s patch dendritic cells \[[@B29]\] ♂ Wistar rats (7/group) 5% of diet days 1-4, 10% of diet days 5-25 26 days Injected IP *E. coli*LPS day 25 ↓ liver ALT, AST, and LDH enzyme levels; ↑ ED2-positive cells, .↓ peroxidase-positive cells in liver; ↓ serum monocytes, TNF-α, PGE2, NO~2~ \[[@B44]\] *Lentinula edodes* SME 6-week nude mice 0.1 ml water with10% SME/10 g body weight days 1-19, 33-50 50 days Injected SC prostate cancer (PC-3) cells day 1 ↓ tumor size \[[@B80]\] β-glucans ♀ 3- and 8-week BALB/c mice (15/group) 50, 100 or 250 μg 1-2 weeks Healthy animals 250 μg dose ↑ spleen cell IL-2 secretion \[[@B27]\] ♀ 3- and 8-week BALB/c mice (15/group) 50, 100 or 250 μg 1-2 weeks Injected murine mammary carcinoma (Ptas64) cells into mammary fat pads 2 weeks before treatment ↓ tumor weight Lentinan 6-week ♂ Wistar-Imamichi specific-pathogen free rats (10/group) 1 mg twice weekly 1-2 months Healthy animals ↑ T cell \#s, helper-cell \#s & helper/suppressor ratio, ↓ suppressor cell level at 4, but not 8 weeks \[[@B26]\] 5-6-week ♂\ 3 mg, days 1-7 3 weeks Injected SC K36 murine lymphoma cells day 7 ↓ tumor weight; ↑ tumor inhibition rate (94%) \[[@B82]\] pre-leukemic AKR mice (10/group) 5-6-week athymic mice (10/group) 5 weeks Injected SC colon cancer (LoVo and SW48, SW480 and SW620, or SW403 and SW1116) cells day 7 ↓ tumor weight, ↑ tumor inhibition rate (\>90%) ♂ AKR mice 3 mg 1 day Pre-leukemic mice ↑ serum IFN-α and TNF-α, peak at 4 h and then back to normal at 24 h; ↑ IL-2 and IL-1α, peak at 2 h and back to normal at 24 h; ↑ CD3+ T, CD4+ T, CD8+ T, B lymphocytes \[[@B81]\] *Phellinus linteus* Aqueous, alcohol-precipitated 6-7-week C57BL/6 mice (10-50/group) 200 mg/kg in drinking water 1 month Healthy animals ↑ production and secretion of IFN-γ by con A stimulated T cells \[[@B32]\] *Saccharomyces cerevisiae* Scleroglucan ♂ ICR/HSD mice (3/group) 1 mg one day before challenge (day 1) 6 days IV *Staphylococcus aureus*or *Candida albicans*day 2 ↑ long-term survival \[[@B29]\] β-1,3;1,6 glucans (particulate) 3 and 8-week ♀ BALB/c mice (15/group) 50, 100 or 250 μg 1-2 weeks Injected murine mammary carcinoma (Ptas64) cells into mammary fat pads 2 weeks before treatment ↓ tumor weight \[[@B27]\] β-1,3-glucan Healthy animals All 3 doses ↑ phagocytic activity of blood monocytes & neutrophils & ↑ spleen cell IL-2 secretion WT or CCD11b^-/-^C57BL/6 mice (2/group) 0.4 mg for 3 weeks 100 days Injected SC RMA-S-MUC1 lymphoma cells ± 14.G2a or anti-MUC1 mAb IV injection every 3^rd^day ↓ tumor diameter when included with mAb; ↑ survival with and without mAb \[[@B73]\] C57BL/6mice (4/group) 25 mg 1 week Healthy animals ↑ \# intestinal IELs; ↑ \# TCRαβ+, TCR γδ+, CD8+, CD4+, CD8αα+, CD8αβ+ T cells in IELs; ↑ IFN-γ mRNA expression in IELs and spleen \[[@B28]\] *Sclerotinia sclerotiorum* SSG 6-8-week specific pathogen-free ♂ CDF~1~mice (3/group) 40 or 80 mg/kg days 1-10 2 weeks Healthy animals 10 mg dose ↑ acid phosphatase activity of peritoneal Mø (day 14) \[[@B30]\] 40, 80 or 160 mg/kg days 2-6 35 days Implanted SC Metha A fibrosarcoma cells day 1 80 mg dose ↓ tumor weight 6-8-week specific pathogen-free ♂ CDF~1~mice (10/group) 40, 80 or 160 mg/kg days 2-11 Injected ID IMC carcinoma cells day 1 6-8-week specific-pathogen free ♂ mice of BDF1 and C57BL/6 mice (7/group) 0.5, 1, 2, or 4 mg days 1-10 2-3 weeks Injected IV Lewis lung carcinoma (3LL) cells 2 mg ↓ \# of 3LL surface lung nodules at 2 weeks \[[@B83]\] *Sclerotium rofsii* Glucan phosphate ♂ ICR/HSD mice (3/group) 1 mg 1 day Healthy animals ↑ systemic IL-6; ↑ Mø expression of Dectin-1 in GALT cells; ↑ TLR2 expression in dendritic cells from Peyer\'s patches \[[@B29]\] *Trametes*(*Coriolus*) *versicolor* PSP 6-8-week ♂ BALB/c mice (10/group) 35 μg days 5-29 in drinking water 29 days Implanted SC Sarcoma-180 cells day 1 ↓ tumor growth & vascular density \[[@B94]\] ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ ###### Immunomodulatory Non-Glucan Extracts: Oral Animal Studies ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Extract Source Animal Oral dose/day Duration Treatment Significant effects Reference ---------------------------------------------------------------------- ------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------- ---------- ---------------------------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------- Fucoidans *Cladosiphon okamuranus Tokida* 8-week ♀ BALB/c mice, 10/group 0.05% w/w of diet 56 days DSS-induced UC ↓ disease activity index and myeloperoxidase activity; ↓ \# of B220-positive colonic B cells; ↓ colonic MLN IFN-γ and IL-6 and ↑ IL-10 and TGF-β; ↓ colonic IgG; ↓ colonic epithelial cell IL-6, TNF-α, and TLR4 mRNA expression \[[@B49]\] *Undaria pinnatifida* 5-week ♀ BALB/c mice (10-12/group) 5 mg, days 1-14 or 7-14 2 weeks Injected HSV into cornea day 7 ↓ facial herpetic lesions; ↑ survival, particularly in pre-treated animals \[[@B45]\] 10 mg 1 week Administered\ ↑ plasma NK cell activity 5-fluorouracil Injected SC HSV ↑ cytotoxic splenic T lymphocyte activity 0.1 or 0.5 mg 3 weeks Injected IP HSV Both doses ↑ serum neutralizing Ab titers, weeks 2 and 3 6-week ♂ ddY mice (5/group) 50, 100, 200 400 or\ 3 weeks Injected with Ehrlich carcinoma in back day 14 200-500 mg/kg ↓ tumor growth \[[@B116]\] 500 mg/kg\ days 1-28 6-week ♂ BALB/c mice (8/group) 40 mg/kg alternating days\ 19 days Injected IP Meth A fibrosarcoma day 1 ↓ tumor growth 7-19 Furanose (COLD-FX^®^) *Panax quinquefolium* Weanling ♂ SD rats (10/group) 450 or\ 1 week Healthy animals Both doses ↑ spleen Il-2 and IFN-γ production following ConA or LPS stimulation; ↓ proportion of total MLN and Peyer\'s patch CD3+ cells & activated T cells; high dose ↑ spleen cell IL-1β production following 48 h ConA stimulation. \[[@B33]\] 900 mg/kg in food Galacto-mannan (partially hydrolyzed guar gum) *Cyamopsis tetragonolobus* 10-week ♀ BALB/c mice,\ 5% of diet 3 weeks DSS-induced UC at beginning of\ ↓ disease activity index scores, ↓ colonic mucosal myeloperoxidase activity & lipid peroxidation; ↓ colonic TNF-α protein levels & mRNA expression up regulated by DSS exposure \[[@B50]\] 11-15/group week 3 Galacto-mannans\ 8-month- SD rats, 5/group 5% of diet 3 weeks Older animals ↓ serum IgG; ↑ MLN lymphocyte IgA, IgM and IgG production \[[@B36]\] (guar gum) Glucomannan (KS-2) *Lentinula edodes* DD1 mice (10-20/group) 140 mg/kg days\ 50 days Injected IP Ehrlich ascites tumor cells day 1 ↑ survival \[[@B84]\] 2-13 0.1, 1, 10, or 100 mg/kg dose days 2-13 100 days Injected Sarcoma-180 tumor cells\ 1, 10, and 100 mg/kg doses ↑ survival day 1 Heteroglycan (ATOM) *A. subrufescens* Mice (10/group): 1) 5-week ♂ Swiss/NIH; 6 week- ♀ DS mice; 3) 8-week ♀ BALB/c nude; 4) 5-week C3H/HcN 100 or\ 8 weeks Implanted SC 1) Sarcoma-180, 2) Shionogi carcinoma 42, 3) Meth A fibrosarcoma, or 4) Ehrlich ascites carcinoma cells Both doses ↓ Sarcoma-180 tumor size at 4 weeks & ↑ survival; 300 mg/kg ↑ peritoneal macrophage and C3-positive cells; 300 mg/kg ↓ Shionogi and Meth A tumor sizes at 4 weeks. Both doses ↑ survival of Ehrlich ascites mice \[[@B93]\] 300 mg/kg\ days 2-11 Heteroglycan (LBP~3p~) *Lycium barbarum* ♂ Kunming mice (10/group) 5, 10 or\ 10 days Injected SC Sarcoma-180 cells 5 & 10 mg/kg ↑ thymus index; all doses ↓ weight, ↓ lipid peroxidation in serum, liver and spleen & ↑ spleen lymphocyte proliferation, cytotoxic T cell activity, IL-2 mRNA \[[@B91]\] 20 mg/kg Heteroglycan (PNPS-1) *Pholiota nameko* SD rats (5/group) 100, 200 or 400 mg/kg days 1-8 8 days Implanted SC cotton pellets in scapular region\ ↓ granuloma growth positively correlated with dose: 11%, 18% and 44%, respectively \[[@B55]\] day 1 Heteroglycan (PG101) *Lentinus lepideus* 8-10-week ♀ BALB/c mice (3/group) 10 mg 24 days 6 Gy gamma irradiation ↑ colony forming cells, granulocyte CFUs/Mø, erythroid burst-forming units, and myeloid progenitor cells in bone marrow; induced proliferation of granulocyte progenitor cells in bone marrow; ↑ serum levels of GM-CSF, IL-6, IL-1β \[[@B92]\] Mixed poly-saccharides (Ambrotose^®^or Advanced Ambrotose^®^powders) *Aloe barbadensis*, *Larix*spp, and other plant poly-saccharides ♂ SD rats (10/group) 37.7 or 377 mg/kg Ambrotose^®^powder or 57.4 or 574 mg/kg Advanced Ambrotose^®^powder 2 weeks 5% DSS in drinking water beginning day 6 574 mg/kg Advanced Ambrotose powder ↓ DAI scores; 377 mg/kg Ambrotose complex & both doses Advanced Ambrotose powder ↑ colon length and ↓ blood monocyte count \[[@B52]\] Pectin *Pyrus pyrifolia* 6-8-week ♂ BALB/c mice (11/group) 100 μg\ 22 days Injected IP OVA day 7, provoked with OVA aerosol day 21 bronchial fluid:↓ IFN-γ & ↑ IL-5; splenic cells: ↑ IFN-γ, ↓ IL-5; normalized pulmonary histopathological changes; ↓ serum IgE \[[@B54]\] days 1-7 Pectins (bupleurum 2IIc) *Bupleurum falcatum* 6-8-week ♀ specific-pathogen-free C3H/HeJ mice 250 mg/kg 1 week Healthy animals ↑ spleen cell proliferation \[[@B35]\] Pectins (highly methoxylated) *Malus*spp. 8-month- SD rats (5/group) 5% of diet vs. cellulose control 3 weeks Older animals ↑ MLN lymphocyte IgA & IgG \[[@B36]\] Pectins Citrus spp. 5-week ♀ F344 rats (30/group) 15% of diet 34 weeks Injected SC AOM once a week, weeks 4-14 ↓ colon tumor incidence \[[@B86]\] *Malus*spp. 5-week ♀ BALB/c mice (6/group) 5% of diet 2 weeks Healthy animals ↑ fecal IgA and MLN CD4+/CD8+ T lymphocyte ratio & IL-2 & IFN-γ secretion by ConA-stimulated MLN lymphocytes \[[@B51]\] 5-week ♀ BALB/c mice (6/group) 5% of diet days 5-19 vs. cellulose control 19 days DSS-induced UC days 1-5 Significantly increased MLN lymphocytes IgA, and significantly decreased IgE; significantly decreased ConA-stimulated IL-4 and IL-10 4-week ♂ Donryu rats (20-21/group) 20% of diet 32 weeks Injected SC AOM once a week,\ ↓ colon tumor incidence \[[@B85]\] weeks 2-12 4-week ♂ Donryu rats (19-20/group) 10 or 20% of diet 32 weeks Injected SC AOM once a week,\ Both doses ↓ colon tumor incidence; 20% ↓ tumor occupied area & ↓ portal blood and distal colon PGE~2~ \[[@B90]\] weeks 2-12 Pectins (modified) Citrus spp. 2-4-month BALB/c mice (9-10/group) 0.8 or 1.6 mg/ml drinking water,\ 20 days Injected SC with 2 × 2 mm section of human colon-25 tumor on day 1 Both doses ↓ tumor size \[[@B87]\] days 8-20 NCR nu/nu mice (10/group) 1% (w/v) drinking water 16 weeks Orthotopically injected human breast carcinoma cells (MDA-MB-435) into mammary fat pad on day 7 ↓ tumor growth rate & volume at 7 weeks, lung metastases at 15 weeks, \# of blood vessels/tumor at 33 days post-injection \[[@B89]\] NCR nu/nu mice (10/group) 1% (w/v) drinking water 7 weeks Injected human colon carcinoma cells (LSLiL6) into cecum on day 7 ↓ tumor weights and metastases to the lymph nodes and liver SD rats (7-8/group) 0.01%, 0.1% or 1.0% wt/vol of drinking water, days 4-30 1 month Injected SC MAT-LyLu rat prostate cancer cells 0.1% and 1.0% ↓ lung metastases; 1.0% ↓ lymph node disease incidence \[[@B88]\] ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ###### Immunomodulatory Polysaccharide-Rich Plant Powders: Oral Animal Studies ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Source Animal Oral dose/day Duration Treatment Significant effects Reference ------------------------------------------------------ ----------------------------------------------------------------- --------------------------------------------------------- ----------- ------------------------------------------------------------------------------------------------------------------------------ ---------------------------------------------------------------------------------------------------------------------------------------------------------- ------------ *Agaricus*(*A. blazei*) *subrufescens*(fruit bodies) 6-week ♂ C57BL/6, C3H/HeJ and BALB/c mice (3/group) 16, 32 or 64 mg 2 weeks Healthy animals 32 and 64 mg ↑ liver mononuclear cell cytotoxicity \[[@B25]\] *Grifola frondosa* 6-week ♀ ICR mice (10-15/group) 5% of diet 36 weeks Oral N-butyl-N\'-butanolnitrosamine daily for first 8 weeks ↓ \#s of animals with bladder tumors; ↓ tumor weight; ↑ peritoneal Mø chemotactic activity, splenic lymphocyte blastogenic response & cytotoxic activity \[[@B70]\] *Laminaria angustata* Weanling SD rats (58/group) 5% of diet 26 weeks IG DMBA, beginning of week 5 ↑ time to tumor development and ↓ \# of adenocarcinomas in adenocarcinoma-bearing animals \[[@B77]\] *Lentinula*(*Lentinus*) *edodes* 6-week ♀ ICR mice (10-17/group) 5% of diet 36 weeks Oral BBN daily for first 8 weeks ↓ \# of animals with bladder tumors; ↓ tumor weight; ↑ Mø chemotactic activity, splenic lymphocyte blastogenic response, cytotoxic activity \[[@B70]\] 7-8 -week ♂ Swiss mice (10/group) 1%, 5% or 10% of diet of 4 different lineages days 1-15 16 days Injected IP N-ethyl-N-nitrosourea day 15 All 3 doses of one lineage and the 5% dose of two other lineages ↓ \#s of micronucleated bone marrow polychromatic erythrocytes \[[@B79]\] *Lentinula edodes*(fruit bodies) 5-week ♀ ICR mice\ 10%, 20% or 30% of diet 25 days Injected IP Sarcoma-180 ascites All 3 doses ↓ Sarcoma-180 tumor weight \[[@B78]\] (14/group × 2) Mice: 1) CDF~1~; 2) C3H; 3) BALB/c; 4,5) C57BL/6N (9/group × 3) 20% of diet 25 days Injected SC 1) IMC carcinoma, 2) MM-46 carcinoma, 3) Meth-A fibrosarcoma, 4) B-16 melanoma, or 5) Lewis lung carcinoma cells ↓ growth of MM-46, B-16, Lewis lung, and IMC tumors; ↑ lifespan in Lewis lung and MM-46 animals ICR mice (14/group × 2) 20% of diet days 1-7, days 7-31 or days 14-31 31 days Injected IP Sarcoma-180 ascites ↓ tumor weight & growth when fed days 7-31 or 14-31 Mice: 1) CDF~1~; 2) C3 H (5/group × 4) 20% of diet 7-12 days Injected SC: 1) IMC carcinoma or 2) MM-46 carcinoma cells ↑ spreading rate of activated Mø ↑ phagocytic activity *Phellinus linteus* 4-week ♂ ICR mice (10/group) 2 mg 1 month Healthy animals ↓ serum & splenocyte IgE production; ↑ proportion of splenic CD4^+^T cells & splenocyte IFN-γ production \[[@B31]\] *Pleurotus ostreatus* 6-week ♀ ICR mice\ 5% of diet 36 weeks Oral BBN daily for first 8 weeks ↓ \#s of animals with bladder tumors; ↓ tumor weight; ↑ plasma Mø chemotactic activity, splenic lymphocyte blastogenic response, cytotoxic activity \[[@B70]\] (10-20/group) ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ###### Immunomodulatory Polysaccharide Products: Oral Human Studies ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Extract Source Study design Population N (experimental/control) Dose/day Dura-tion Significant effects Reference -------------------------------- ---------------------------------- ---------------------------------------------- ---------------------------------------------------------------------------------- -------------------------------------- ---------------------------------------------------------------------- ------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------ Arabino-galactans *Larix occidentalis* Randomized, double-blind, placebo-controlled Healthy adults 8/15 4 g 6 weeks ↑ % CD8+ lymphocytes & blood lymphocyte proliferation \[[@B18]\] Arabino-galactans (ResistAid™) Healthy adults given pneumococcal vaccinations day 30 21/24 4.5 g 72 days ↑ plasma IgG subtypes \[[@B19]\] Fucoidans *Undaria pinnatifida*sporophylls Randomized, single-blind, placebo-controlled Healthy adults 25 (75% fucoidan, 6 (10% fucoidan)/6 3 g 12 days 75% fucoidan: ↓ \#s blood leukocytes, lymphocytes\' ↑ plasma stromal derived factor-1, IFN-γ, CD34+ cells; ↑ % CXCR4-expressing CD34+ cells \[[@B21]\] Furanose extract (Cold-FX^®^) *Panax quinque-folium* Randomized, double-blind, placebo-controlled Healthy older adults given influenza immunization at the end of week 4 22/21 400 mg 4 months During weeks 9-16, ↓ incidence of acute respiratory illness, symptom duration \[[@B20]\] Glucans *Agaricus subru-fescens* Randomized, double-blind, placebo-controlled Cervical, ovarian or endometrial cancer patients receiving 3 chemotherapy cycles 39/61 5.4 g (estimated) 6 weeks ↑ NK cell activity, ↓ chemotherapy side effects \[[@B64]\] Glucans\ Not identified Placebo-controlled Recurrent aphthous stomatitis patients 31/42 20 mg 20 days ↑ PBL lymphocyte proliferation,↓ Ulcer Severity Scores \[[@B48]\] (β-1,3;1,6) Glucans\ *S. cerevisiae* Randomized, double-blind, placebo-controlled Adults with seasonal allergic rhinitis 12/12 20 mg 12 weeks 30 minutes after nasal allergen provocation test, nasal lavage fluid: ↓ IL-4, IL-5, % eosinophils, ↑ IL-12 \[[@B47]\] (β-1,3;1-6) Glucans (PSK) *Trametes versicolor* Randomized, controlled Patients with curatively resected colorectal cancer receiving chemotherapy 221/227 200 mg 3-5 years ↑ disease-free survival and overall survival \[[@B56]\] Controlled Post-surgical colon cancer patients receiving chemotherapy 123/121 3 g for 4 weeks, alternating with 10 4-week courses of chemo-therapy 7 years ↑ survival from cancer deaths; no difference in disease-free or overall survival \[[@B57]\] Post-surgical colorectal cancer patients receiving chemotherapy 137/68 3 g daily 2 years ↑survival in stage III patients; ↓ recurrence in stage II & III patients \[[@B58]\] Post-surgical gastric cancer patients receiving chemotherapy 124/129 3 g for 4 weeks, alternating with 10 4-week courses of chemo-therapy 5-7 years ↑ 5-year disease-free survival rate, overall 5-year survival \[[@B59]\] Pre-surgical gastric or colorectal cancer patients 16 daily; 17 every other day/13 3 g daily or on alternate days before surgery \<14 days or 14-36 days ≥14 day treatment: ↑ peripheral blood NK cell activity, PBL cytotoxicity, proportion of PBL helper cells; ↓ proportion of PBL inducer cells; \<14 day treatment: ↑ PBL response to PSK and Con A, proportion of regional node lymphocyte suppressor cells \[[@B62]\] Randomized, double-blind, placebo-controlled Post-surgical stage III-IV colorectal cancer patients 56/55 3 g for 2 months, 2 g for 22 months, 1 g thereafter 8-10 years ↑ remission & survival rates \[[@B61]\] Controlled Post-surgical stage III gastric cancer patients receiving chemotherapy 32/21 3 g 1 year ↑ survival time \[[@B60]\] Glucans (PSP) *Trametes versicolor* Randomized, double-blind, placebo-controlled Conventionally-treated stage III-IV non-small cell lung cancer patients 34/34 3.06 g 1 month ↑ blood IgG & IgM, total leukocyte and neutrophil counts, % body fat; ↓ patient withdrawal due to disease progression \[[@B63]\] ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ###### Immunomodulatory Polysaccharide Products: Composition and Structure -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Source Category Features MW Monosaccharide composition Reference ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------- *Agaricus subrufescens*(*A. blazei*) Extract β-1,6-D-glucan 10,000 NA \[[@B66]\] *Agaricus subrufescens*(fruit body) Extract α-1,6- and α-1,4 glucans with β-1,6-glucopyranosyl backbone (629.2 mcg/mg polysaccharides, 43.5 mcg/mg protein) 170,000 glucose \[[@B24]\] α-1,4 glucans & β-1,6 glucans with β-1,3 side branches; α-1,6 glucans; β-1,6; 1-3 glucans, β-1,4 glucans; β-1,3 glucans; β-1,6; α-1,3 glucans; riboglucans, galactoglucomannans, β-1,2; β-1,3 glucomannans NA glucose, mannose, galactose, ribose \[[@B25],[@B117],[@B118]\] *Agaricus subrufescens*(mycelia) Extract (ATOM) β-1,6-D-glucan, protein complex, 5% protein 100,000-1,000,000 glucose, mannose, galactose, ribose \[[@B93]\] *Aloe barbadensis*(leaf gel) Whole tissue Dry weight: 10% polysaccharides; acemannan, aloemannan, aloeride, pectic acid, galactans, arabinans, glucomannans average 2,000,000 mannose, glucose, galactose, arabinose, xylose, rhamnose \[[@B119],[@B120]\] Extract (aloemannan) neutral partially acetylated glucomannan, mainly β-1,4-mannans \>200,000 mannose, glucose \[[@B121]\] Extract (aloeride) NA 4,000,000-7,000,000 37% glucose, 23.9% galactose, 19.5% mannose, 10.3% arabinose \[[@B122]\] Extract (acemannan) β-1,4 acetylated mannan 80,000 mannose \[[@B123]\] *Aloe barbadensis*, (leaf gel), *Larix*sp. (bark), *Anogeissus latifolia*(bark), *Astragalus gummifer*(stem), *Oryza sativa*(seed), *glucosamine* Extracts (Ambrotose^®^powder) β-1,4 acetylated mannan, arabinogalactans, polysaccharide gums, rice starch, 5.4% protein 57.3% ≥ 950,000; 26.4% \< 950,000 and ≥80,000; 16.3% ≤ 10,000 mannose, galactose, arabinose, glucose, galacturonic acid, rhamnose, xylose, fructose, fucose, glucosamine, galacturonic acid (unpublished data, Mannatech Incorporated) *Aloe barbadensis*(leaf gel), *Larix*sp. (bark), *Undaria pinnatifida*(frond), *Anogeissus latifolia*(bark), *Astragalus gummifer*(stem), *Oryza sativa*(seed), *glucosamine* Extracts (Advanced Ambrotose^®^powder) β-1,4 acetylated mannan, arabinogalactans, polysaccharide gums, fucoidans, rice starch, 6% protein, 1% fatty acids 13% = 1,686,667; 46% = 960,000 30% \<950,000 and ≥70,000; 11% ≤ 10,000 *Avena*spp. (seed endosperm) Extract β-1,3;1,4 particulate (1-3 μ) glucans 1,100,000 glucose \[[@B43]\] *Avena*spp. (seed) Extract β-1,4,1,3 particulate glucans (linear chains of β-D-glycopyranosyl units; 70% β 1-4 linked) 2,000,000 NA \[[@B41],[@B124]\] *Buplerum falcatum*(root) Extract (bupleuran 2IIc) 6 linked galactosyl chains with terminal glucuronic acid substituted to β-galactosyl chains NA galactose, glucuronic acid, rhamnose \[[@B35]\] Citrus spp. (fruit) Extract α-1,4-linked partially esterified D-anhydrogalacturonic acid units interrupted periodically with 1,2-rhamnose 70,000-100,000 galactose, galacturonic acid, arabinose, glucose, xylose, rhamnose \[[@B125]\] *Cladosiphon okamuranus*(frond) Extract α-1,3-fucopyranose sulfate 56,000 fucose:glucuronic acid (6.1:1.0) \[[@B126]\] *Cordyceps sinensis*(mycelia) Extract β-1,3-D-glucan with 1,6-branched chains NA NA \[[@B127]\] *Cyamopsis tetragonolobus*(seed) Extract (guar gum) Main chain of β-1,4-mannopyranosyl units with α-galactopyranosyl units 220,000 mannose, galactose \[[@B36],[@B128]\] Extract (partially-hydrolyzed guar gum) NA 20,000 mannose, galactose \[[@B50]\] *Flammulina velutipes* Extract NA NA glucose, mannose, galactose \[[@B117]\] *Flammulina velutipes*(fruit body) Extract β-1,3 glucan NA glucose \[[@B129]\] *Ganoderma lucidum* Whole tissue Linear β-1,3-glucans with varying degrees of\ 400,000-1,000,000 glucose, galactose, mannose, xylose, uronic acid \[[@B130]\] D-glucopyranosyl branching, β-glucan/protein complexes, heteropolysaccharides Extract NA 7,000-9,000 NA \[[@B67]\] *Ganoderma lucidum*(fruit body) Extract NA 7,000-9,000 NA β-linked heteroglycan peptide 513,000 fructose, galactose, glucose, rhamnose, xylose (3.167:\ \[[@B15]\] 0.556:6.89:0.549:3.61) *Ganoderma tsugae* Extract 55.6% carbohydrates (12.5% polysaccharides); 12% triterpenes, 1.7% sodium, 0.28% protein, 0% lipid NA NA \[[@B53]\] *Ginkgo biloba*(seed) Extract 89.7% polysaccharides NA glucose, fructose, galactose, rhamnose \[[@B131]\] *Grifola frondosa* Whole tissue β-1,3; 1, 6-glucans, α-glucans, mannoxyloglucans, xyloglucans, mannogalactofucans NA glucose, fucose, xylose, mannose, galactose \[[@B117]\] *Grifola frondosa*(fruit body) Extract\ β-1,6-glucan with β-1,3 branches, 30% protein NA glucose \[[@B132]\] (D fraction) Extract\ β-1,6-D-glucan with α-1,4 branches, 35% protein 550,000-558,000 glucose (X fraction) *Hordeum*spp. (seed) Extract β-1,3;1,4-and β-1,3;1,6-D-glucans 45,000-404,000 glucose \[[@B75]\] Primarily linear β-1,3;1,4- glucans NA glucose \[[@B124]\] *Laminaria*spp.\ Extract (laminarin) β-1,3;1-6 glucan 7,700 glucose \[[@B29]\] (frond) β-1,3 glucan with some β-1,6 branches and a small amount of protein 4,500-5,500 glucose \[[@B44]\] Extract Fucoidan NA NA \[[@B133]\] *Larix occidentalis*(bark) Extract β-1,3;1,6-D-galactans with arabinofuranosyl and arabinopyranosyl side chains 19,000-40,000 galactose:arabinose (6:1), uronic acid \[[@B128],[@B134]\] *Lentinula edodes* Extract (SME) β-1,3-glucans (4-5%), α-1,4-glucan (8-10%), protein (11-14%) NA glucose \[[@B80]\] Extract β-glucan 1,000 glucose \[[@B27]\] Whole tissue Linear β-1,3-glucans, β-1,4;1,6-glucans, heterogalactan NA glucose, galactose, mannose, fucose, xylose \[[@B135]\] Extract (lentinan) β-1,3-glucan with 2 β-1,6 glucopyranoside branchings for every 5 β-1,3-glucopyranoside linear linkages 500,000 glucose \[[@B136]\] *Lentinula edodes*(fruit body)\ Extract (lentinan) Neutral β-1,3-D glucan with two β-1,6 glucoside branches for every five β-1,3 units 400,000-800,000 glucose \[[@B137]\] *Lentinula edodes* Extract\ Peptide units and mannan connected by α-glycosidic bonds 60,000-90,000 mannose, glucose (KS-2) *Lentinula edodes*(mycelia or fruit body) Extract Triple helical β-1,3-D glucan with β-1,6 glucoside branches 1,000,000 glucose \[[@B3]\] *Lentinula edodes*(mycelia) Extract\ 44% sugars, 24.6% protein \~1,000,000 xylose, arabinose, glucose, galactose, mannose, fructose \[[@B3]\] (LEM) Extract (PG101) 72.4% polysaccharides, 26.2% protein, 1.4% hexosamine NA 55.6% glucose, 25.9% galactose, 18.5% mannose \[[@B138]\] *Lycium barbarum* Whole tissue α-1,4;1,6-D-glucans, lentinan, β-1,3;1,6 heteroglucans, heterogalactans, heteromannans, xyloglucans NA glucose, galactose, mannose, xylose \[[@B139]\] *Lycium barbarum*(fruit body) Extract\ 88.36% sugars, 7.63% protein 157,000 galactose, glucose, rhamnose, arabinose, mannose, xylose (molar ratio of 1:2.12:1.25:1.10:1.95:1.76) \[[@B91]\] (LBP~3p~) *Panax quinquefolium*(root) Extract Poly-furanosyl-pyranosyl saccharides NA arabinose, galactose, rhamnose, galacturonic acid, glucuronic acid \[[@B33]\] NA NA glucose, mannose, xylose \[[@B140]\] Extract\ 90% poly-furanosyl-pyranosyl-saccharides NA furanose \[[@B20]\] (Cold-fX^®^) *Phellinus linteus*(fruit body) Extract α- and β-linked 1,3 acidic proteoglycan with 1,6 branches 150,000 glucose, mannose, arabinose, xylose \[[@B141]\] *Phellinus linteus*(mycelia) Extract 83.2% polysaccharide (4.4% β-glucan), 6.4% protein, 0.1% fat NA glucose \[[@B142]\] *Pholiota nameko*(fruit body) Extract (PNPS-1) NA 114,000 mannose, glucose, galactose, arabinose, xylose (molar ratio of 1:8.4:13.6:29.6:6.2) \[[@B55]\] *Pleurotus ostreatus*(mycelia) Extract β-1,3;1,6-D-glucans 316,260 glucose \[[@B143]\] *Saccharomyces cerevisiae* Extract (WGP) Particulate β-1,3;1,6-D-glucan NA glucose \[[@B144]\] Extract β-glucans with β-1,6 branches with a β-1,3 regions NA glucose \[[@B124]\] Extract\ soluble β-1,3-D-glucan with β-1,3 side chains attached with β-1,6 linkages 20,000 glucose \[[@B145]\] (SBG) *Sclerotinia sclerotiorum*(mycelia) Extract\ β-1,3-D-glucan, \<1% protein (\>98% polysaccharide) NA glucose \[[@B83]\] (SSG) *Sclerotium rofsii* Extract (scleroglucan) β-1,3;1,6 glucan 1,000,000 glucose \[[@B29]\] *Trametes versicolor*(fruit body) Extract\ α-1,4, β-1,3 glucans, 10% peptides 100,000 glucose, arabinose, mannose, rhamnose \[[@B146]\] (PSP) *Trametes versicolor*(mycelia) Extract\ β-1,4;1,3;1,6-D-glucans, protein 94,000 glucose (74.6%), mannose (15.5%), xylose (4.8%), galactose (2.7%), fucose (2.4%) \[[@B137],[@B147]\] (PSK) *Undaria pinnatifida*(sporophyll) Extract Galactofucan sulfate 9,000 fucose:galactose 1.0:1.1 \[[@B148]\] Galactofucan sulfate 63,000 fucose:galactose:gluc-uronic acid (1.0:1.0:0.04) \[[@B149]\] β-1,3-galactofucan sulphate 38,000 fucose, galactose \[[@B150]\] Unidentified source Extract (modified citrus pectin) NA 10,000 galactose, rhamnose, uronic acid \[[@B125]\] Extract (highly methoxylated pectin) NA 200,000 NA \[[@B36]\] -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ###### Safety of Immunomodulatory Polysaccharide Products Following Oral Intake ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Category Source Test group Test Design Results Equivalent human dose\* Reference ------------------- ----------------------------------------------------------- ---------------------------------------------------- --------------------------------- ------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------- ------------------------------------ Arabino-galactans *Argemone mexicana*(arabinogalactan protein) Pregnant rats Develop-mental toxicity 250, 500, or 1,00 mg/kg, gestational days 5-19 No developmental toxicity: NOAEL = 1 g/kg 68 g \[[@B151]\] ♀ and ♂ rats Fertility 250, 500, or 1,00 mg/kg, 1 month No effects on reproduction: NOAEL = 1 g/kg Fucoidans *Undaria pinnatifi*da Rats Subchronic toxicity 1.35 g/kg, 1 month No evidence of toxicity 91.8 g \[[@B152]\] Galacto-mannans *Cyamopsis tetragonolobus* Adolescent and adult ♂ rats Subchronic and chronic toxicity 8% of diet, 6-67 weeks No evidence of toxicity 8% of diet \[[@B153]\] Rats Acute toxicity One 7.06 g/kg dose: observed 2 weeks LD~50~= 7.06 g/kg 480 g \[[@B96]\] Subchronic and chronic toxicity 1, 2, 4, 7.5 or 15% of diet, 3 months All doses ↓ ♀ BW; 7.5-15% ↓ ♂ BW; 15% ↓ bone marrow cellularity; ↓ kidney and liver weights 1-15% of diet 19 adults with hypercholesterol-emia 18 g/day, 1 year Short-term gastric bloating/loose stools, in 8 subjects, resolved in 7-10 days; 2 withdrew because of diarrhea. No toxicity for 13 subjects completing study 18 g \[[@B154]\] 16 Type II diabetics 26.4-39.6 g/day, 6 months No effects on hematologic, hepatic, or renal function 39.9 g \[[@B155]\] 18 Type II diabetics 30 g/day, 4 months 30 g *Cyamopsis tetragonolobus*(partially hydrolyzed guar gum) Mice & rats Acute toxicity One 6 g/kg dose; observed\ LD~50~\> 6 g/kg \>408 g \[[@B156]\] 2 weeks Rats Subchronic toxicity 0.2, 1.0 or 5% of diet, 13 weeks No evidence of toxicity 5% of diet 0.5 or 2.5 g/kg, 1 month NOAEL \> 2.5 g/kg \>170 g \[[@B157]\] *S. typhimurium* Mutagenicity Ames test Not mutagenic NA Glucans *Agaricus subrufescens*(aqueous extract) Rats Subchronic toxicity 0.63, 1.25, 2.5 or 5% of diet, 3 months NOAEL = 5% of diet 5% of diet \[[@B158]\] 3 women with advanced cancers Case reports Specific identity of products, doses, and durations of intake unknown Severe hepatotoxicity; two patients died NA \[[@B97]\] *Agaricus subrufescens*(freeze dried powder) 24 normal adults and 24 adults with liver problems Subchronic toxicity 3 g, 4 months No evidence of toxicity 3 g \[[@B159]\] *Ganoderma lucidum*\ Elderly woman Case report 1 year *G. lucidum*(and another unidentified product, initiated one month previous) Elevated liver enzymes and liver tissue damage NA \[[@B98]\] (supplement) *Grifola frondosa*(powder) Rats Acute toxicity One 2 g/kg dose No evidence of toxicity 136 g \[[@B160]\] *Lentinula edodes*(powder) 10 adults Safety 4 g/day for 10 weeks; repeated\ 50% of subjects experienced blood eosinophilia, ↑ eosinophil granule proteins in serum and stool, ↑GI symptoms 4 g \[[@B99]\] 3-6 months later *Lentinula edodes*\ Nude mice Safety 10% of diet days 1-18, 33-50 No adverse events 10% of diet \[[@B80]\] (SME) 61 men with prostate cancer 0.1 g/kg, 6 months No adverse events 6.8 g *Lentinus lepideus*(PG101) Female mice Subchronic toxicity 0.5 g/kg, 24 days No evidence of toxicity 34 g \[[@B92]\] *Phellinus linteus*\ Rats Acute toxicity One 5 g/kg dose; observed\ LD~50~\> 5 g/kg 349 g \[[@B161]\] (crude extract) 2 weeks *Pleurotus ostreatus*(aqueous extract) Mice Acute toxicity One 3 g/kg dose; observed\ LD~50~\> 3 g/kg \>204.g \[[@B100]\] 1 day Subacute toxicity 319 mg/kg, 1 month Hemorrhages in intestine, liver, lung, kidney; inflammation and microabscesses in liver 21.7 g *Saccharomyces cerevisiae*(particulate glucan \[WGP\]) Rats Acute toxicity One 2 g/kg, observed 2 weeks LD~50~\> 2 g/kg \>136 g \[[@B144]\] Subchronic toxicity 2, 33.3 or 100 mg/kg, 3 months NOAEL = 100 mg/kg 6.80 g Heteroglycans *Trametes versicolor*\ Rats Subchronic toxicity 1.5, 3.0 or 6.0 mg/kg, 2 months No evidence of toxicity 408 mg \[[@B162]\] (PSP) Rats & monkeys Subchronic and chronic toxicity 100-200X equivalent human dose, 6 months No evidence of toxicity NA *Trametes versicolor*\ Humans with colon cancer Safety 3 g/day, up to 7 years No significant adverse events 3 g \[[@B57]\] (PSK) Humans with colorectal cancer 3 g/day, 2 years 3 g \[[@B58]\] Mannans *Aloe vera*gel Dogs Acute toxicity Fed one 32 g/kg; observed 2 weeks LD50 \> 32 g/kg \>2,176 g Bill Pine, personal communi-cation Rats One 21.5 g/kg; observed 2 weeks LD50 \> 10 g/kg \>680 g ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ \*150 lb adult A number of studies in healthy human adults demonstrated immune stimulating effects of oral polysaccharides. Arabinogalactans from *Larix occidentalis*(Western larch) were shown in RCTs to increase lymphocyte proliferation and the number of CD8+ lymphocytes \[[@B18]\] and to increase the IgG subtype response to pneumococcal vaccination \[[@B19]\]. A furanose extract from *Panax quiquefolium*(North American ginseng) was shown in an RCT of healthy older adults to decrease the incidence of acute respiratory illness and symptom duration \[[@B20]\]. Finally, an RCT of healthy adults consuming *Undaria pinnatifida*(wakame) fucoidans found both immune stimulating and suppressing effects, including increased stromal-derived factor-1, IFN-g, CD34+ cells and CXCR4-expressing CD34+ cells and decreased blood leukocytes and lymphocytes \[[@B21]\]. Studies in healthy animals showed a number of immune stimulating effects of various glucan products from *Agaricus subrufescens (A. blazei)*(aqueous extracts \[[@B22]\], aqueous extracts with standardized β-glucans \[[@B23]\], α-1,6 and α-1,4 glucans \[[@B24]\], and whole plant powders \[[@B25]\]); *Lentinula edodes*(shiitake) (lentinan \[[@B26]\] and β-glucans \[[@B27]\]); *Saccharomyces cerevisiae*(β-1,3-glucans \[[@B27],[@B28]\]); *Laminaria digitata*(laminarin \[[@B29]\]); *Sclerotium rofsii*(glucan phosphate \[[@B29]\]); *Sclerotinia sclerotiorum*(SSG \[[@B30]\]); and *Phellinus linteus*(powder \[[@B31]\] and aqueous, alcohol-precipitated extract \[[@B32]\]). A furanose extract from *P. quiquefolium*and pectins from *Buplerum falcatum*and *Malus*(apple) spp. have also been shown to enhance immune function in healthy young animals \[[@B33]-[@B35]\]. *Cyamopsis tetragonolobus*galactomannan (guar gum) or highly methoxylated pectin feeding exerted numerous stimulating effects on antibody production in older animals \[[@B36]\]. Evidence for the effectiveness of oral polysaccharides against infection and immune challenges has been mainly demonstrated in animals. Immune stimulating effects have been shown in resting and exercise-stressed animals with thioglycollate, clodronate, or HSV-1 injections fed *Avena*(oat) spp. soluble glucans \[[@B37]-[@B41]\]; animals injected with or fed *E. vermiformis*and fed *Avena*spp. particulate glucans \[[@B42],[@B43]\]; animals with *E. coli*injections fed *L. digitata*glucans (laminarin) \[[@B44]\]; animals with HSV injections fed *U. pinnatifida*fucoidans \[[@B45]\]; animals with *Staphylococcus aureus*or *Candida albicans*injections fed *S. cerevisiae*glucans (scleroglucan) \[[@B29]\]; and animals with fecal solution injections fed an aqueous extract of *A. subrufescens*(*A. blazei*Murrill) \[[@B46]\]. Additional controlled human and animal studies have shown anti-inflammatory and anti-allergy effects of some polysaccharide products. In an RCT of adults with seasonal allergic rhinitis, *S. cerevisiae*β-1,3;1-6 glucans decreased IL-4, IL-5 and percent eosinophils, and increased IL-12 in nasal fluid \[[@B47]\], while a placebo-controlled study of patients with recurrent aphthous stomatitis (canker sores) consuming β-1,3;1-6 glucans found increased lymphocyte proliferation and decreased Ulcer Severity Scores \[[@B48]\]. Animal models of inflammatory bowel disease have shown anti-inflammatory effects of *Cladosiphon okamuranus*Tokida fucoidans \[[@B49]\], *Cyamopsis tetragonolobus*galactomannans \[[@B50]\], *Malus*spp. pectins \[[@B51]\], and mixed polysaccharide supplements \[[@B52]\]. Animals challenged with ovalbumin have demonstrated anti-inflammatory/allergy effects of *A. subrufescens*aqueous extracts \[[@B22]\], an aqueous extract *of Ganoderma tsugae*\[[@B53]\], and *Pyrus pyrifolia*pectins \[[@B54]\]. Anti-inflammatory effects have also been seen in animals with cotton pellet implantations fed a *Pholiota nameko*heteroglycan (PNPS-1) \[[@B55]\]. *Trametes versicolor*glucans have demonstrated anti-cancer effects in humans. In two RCTs and five controlled trials, PSK from *T. versicolor*mycelia increased survival of advanced stage gastric, colon and colorectal cancer patients \[[@B56]-[@B62]\] with one study showing increased immune parameters (including blood NK cell activity, leukocyte cytotoxicity, proportion of helper cells and lymphocyte suppressor cells) \[[@B62]\]. An RCT of advanced stage lung cancer patients consuming PSP from *T. versicolor*fruit bodies found increased IgG and IgM antibodies and total leukocyte and neutrophil counts, along with a decrease in the number of patients withdrawing from the study due to disease progression \[[@B63]\]. An RCT of ovarian or endometrial cancer patients consuming *A. subrufescens*glucans showed increased NK cell activity and fewer chemotherapy side effects \[[@B64]\]. In numerous animal models of cancer, a wide range of polysaccharides have shown anti-tumorogenic effects. Glucan products sourced from *A. subrufescens*demonstrating anti-cancer activities in animal models include an aqueous extract \[[@B65]\], an aqueous, acid-treated extract \[[@B66]\], and an aqueous extract with standardized levels of β-glucans \[[@B23]\]. Anti-cancer effects have been reported following intake of aqueous extracts of *G. lucidum*\[[@B67]-[@B69]\]; the powder and D fraction of *G. frondosa*\[[@B70]-[@B72]\]; *Hordeum vulgare*β-glucans \[[@B73]-[@B76]\]; *Laminaria angustata*powder \[[@B77]\]; *Lentinula edodes products*(powders \[[@B70],[@B78],[@B79]\], SME \[[@B80]\], β-glucans \[[@B27]\], and lentinan \[[@B81],[@B82]\]); *Pleurotus ostreatus*powder \[[@B70]\], *Saccharomyces cerevisiae*particulate β-1,3;1,6 and β-1,3glucans\[[@B27],[@B73]\]; and a glucan from *Sclerotinia sclerotiorum*(SSG) \[[@B30],[@B83]\]. A glucomannan from *L. edodes*(KS-2) improved survival of animals with cancer cell injections \[[@B84]\]; apple and citrus pectins have exerted anti-cancer effects, including decreased tumor incidence \[[@B85]-[@B90]\]. Finally, heteroglycans from *Lycium barbarum*(LBP~3p~), *Lentinus lepidus*(PG101) and A. *subrufescens*(ATOM) demonstrated a number of immune stimulating effects in animal cancer models \[[@B91]-[@B93]\]. Interestingly, only one animal study has been performed using glucans from *T. versicolor*(PSP): animals with cancer cell implantations showed decreased tumor growth and vascular density \[[@B94]\]. Most polysaccharide products appear to be safe, based on NOAEL, acute and/or chronic toxicity testing in rodents (Table [6](#T6){ref-type="table"}). As would be expected, powders, extracts and products that have not been fully characterized pose the most concerns. Other than for aloe vera gel, which was shown in a small human trial to increase the plasma bioavailability of vitamins C and E \[[@B95]\], the impact of polysaccharide intake on the absorption of nutrients and medications is not known. While one rat toxicity study raised concerns when guar gum comprised 15% of the daily diet \[[@B96]\], the product was safe in humans studies when 18-39.6 g/day was consumed for up to a year (Table [4](#T4){ref-type="table"}). Product contamination may explain three case reports of hepatotoxicity and/or death following intake of an *A. subrufescens*aqueous extract \[[@B97]\]. Seven animal studies reporting positive immunologic effects of *A. subrufescens*extracts in healthy animals or animals with cancers found no evidence of toxicity (Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}). In humans, six weeks of *A. subrufescens*glucans intake was safe for cancer patients, and four months of 3 g/day intake by 24 healthy adults and 24 adults with liver disease reported no evidence of toxicity (Table [4](#T4){ref-type="table"}). Another case report associated liver toxicity with *G. lucidum*intake, but the elderly subject also took an unidentified product a month previous to her admission for testing \[[@B98]\]. Three animal studies reported immunologic benefits and no adverse effects following intake of *G. lucidum*aqueous extracts; in one study intake was 5% of the diet for 5 months (Table [1](#T1){ref-type="table"}). While adverse effects were also reported in a study in which 10 adults consumed 4 g/day *L. edodes*powder for 10 weeks \[[@B99]\], immunologic animal studies reported no ill effects of either *L. edodes*powder (5 studies, up to 5% of the diet up to nine months) or extract (7 studies, up to 40 days intake) (Tables [1](#T1){ref-type="table"} and [3](#T3){ref-type="table"}). Finally, while intake of 319 mg/kg of an aqueous extract of *P. ostreatus*by mice for 1 month caused hemorrhages in multiple tissues \[[@B100]\], there was no reported toxicity when mice consumed the mushroom powder as 5% of their diet for nine months (Table [3](#T3){ref-type="table"}). While ≥1 gram/day of *T. versicolor*glucan products were safely consumed by cancer patients for up to 10 years, the long-term effects of ingestion of the other polysaccharide products discussed in this review is also not known. Discussion ========== The majority of studies that qualified for inclusion in this review employed models investigating immune stimulation; fewer explored anti-inflammatory effects. Animal studies reported immune system effects in the gut, spleen, bone marrow, liver, blood, thymus, lungs, and saliva; controlled human studies reported evidence of immune stimulation in the blood, anti-inflammatory effects in nasal lavage fluid and improved survival in cancer patients. The literature is highly heterogenous and is not sufficient to support broad structure/function generalizations. For the limited number of studies that investigated well-characterized, isolated products (primarily glucan products), effects can be unequivocally attributed to polysaccharides. Such associations are certainly more tenuous when considering product powders or products obtained by extraction methods designed to isolate polysaccharides, but without complete compositional analyses. Dietary polysaccharides are known to impact gut microbial ecology \[[@B101],[@B102]\], and advances in microbial ecology, immunology and metabolomics indicate that gut microbiota can impact host nutrition, immune modulation, resistance to pathogens, intestinal epithelial development and activity, and energy metabolism \[[@B103]-[@B107]\]. Other than fucoidans, the polysaccharides discussed in this review appear to be at least partially degraded by bacterial enzymes in the human digestive tract (Table [7](#T7){ref-type="table"}). Arabinogalactans, galactomannans, a glucan (laminarin), glucomannans, and mixed polysaccharide products (Ambrotose^®^products) have been shown to be metabolized by human colonic bacteria. Orally ingested fucoidans, glucans and mannans (or their fragments) have been detected in numerous tissues and organs throughout the body \[[@B73],[@B108],[@B109]\], (Carrington Laboratories, personal communication). We know of no study that has determined the specific identity of orally-ingested polysaccharide end products in animal or human tissues. ###### Fate of Immunomodulatory Polysaccharide Products Following Oral Intake -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Category Product Metabol-ized by human gut bacteria? Study type Fate\ References (method: tissues detected) ------------------------------- --------------------------------------------------------------------------------- ------------------------------------- ------------ ------------------------------------------------------------------------------------------------------------------- --------------------------------------------------- Arabinogalactans *Larix*spp. yes *in vitro* NA \[[@B163]-[@B169]\] Fucoidans *Undaria pinnatifida* no *in vitro* Ab: human plasma \[[@B108],[@B170]\] Galactomannans *Cyamopsis tetragonolobus*(partially hydrolyzed guar gum) yes *in vivo* NA \[[@B171]\] *Cyamopsis tetragonolobus*(guar gum) yes *in vitro* NA \[[@B167]\] Glucans *Hordeum vulgare* NA *in vivo* Fluorescein-labeled: mouse Mø in the spleen, bone marrow, lymph nodes \[[@B73]\] *Laminaria digitata*(laminarin) yes *in vitro* NA \[[@B29],[@B170],[@B172]\] *Sclerotium rofsii*(scleroglucan) glucan phosphate, *Laminaria*spp. (laminarin) NA *in vivo* Alexa Fluor 488-labeled: mouse intestinal epithelial cells, plasma, GALT \[[@B29]\] *Saccharomyces cervisiae*(particulate) NA *in vivo* Fluorescein-labeled: mouse macrophage in the spleen, bone marrow, lymph nodes \[[@B73]\] *Trametes versicolor*\ NA *in vivo* ^14^C-labeled: rat and rabbit serum; mouse GI tract, bone marrow, salivary glands, liver, brain, spleen, pancreas \[[@B173]\] (PSK) Mannans *Aloe barbadensis*(aloemannan) yes *in vitro* FITC-labeled: mouse, GI tract \[[@B121],[@B174]\] *Aloe barbadensis*\ yes *in vitro* NA \[[@B163]\] (gel powder) *Aloe barbadensis*(acemannan) NA *in vivo* ^14^C-labeled: dog systemic, particularly liver, bone marrow, gut, kidney, thymus, spleen (Carrington Laboratories, personal communication) Mixed polysaccharide products Ambrotose complex^®^, Advanced Ambrotose^®^powder yes *in vitro* NA \[[@B163],[@B175]\] Pectins NA yes *in vitro* NA \[[@B165]-[@B167],[@B176]\] *Buplerum falcatum*(bupleuran 2IIc) NA *in vivo* Ab bound: mouse Peyer\'s patch, liver \[[@B109]\] -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- One can only speculate upon the mechanisms by which the polysaccharides discussed in this review influence immunologic function, particularly when one considers the exceedingly complex environment of the GI tract. It is possible that fragments of polysaccharides partially hydrolyzed by gut bacteria may either bind to gut epithelia and exert localized and/or systemic immune system effects, or be absorbed into the bloodstream, with the potential to exert systemic effects. Current studies investigating the link between the bioconversion of dietary polysaccharides, their bioavailability and their downstream effects on the host metabolism and physiology are utilizing metabolomic and metagenomic approaches that can detect and track diverse microbial metabolites from immunomodulatory polysaccharides \[[@B103]\]. These and other innovative approaches in the field of colonic fermentation are providing novel insights into gut microbial-human mutualism \[[@B110],[@B111]\], its impact on regulating human health and disease, and the importance of dietary modulation \[[@B112]-[@B115]\]. Additional RCTs of well-characterized products are needed to more completely understand the immunomodulatory effects and specific applications of oral polysaccharides. Such studies will need to better investigate the optimal timing and duration for polysaccharide ingestion. That is, should they be consumed continuously, before, at the time of, or after exposure to a pathogen or environmental insult? Only a few studies have actually investigated the impact of timing of polysaccharide intake to achieve optimal benefits. Daily feeding with some polysaccharides appears to result in tolerance (and diminished benefits); this has been demonstrated for some mushroom β-glucans \[[@B3],[@B26]\]. For those polysaccharides whose immunologic effects are dependent on their prebiotic activities, regular feeding would be presumed necessary. Conclusions =========== The dietary polysaccharides included in this review have been shown to elicit diverse immunomodulatory effects in animal tissues, including the blood, GI tract, and spleen. In controlled human trials, polysaccharide intake stimulated the immune system in the blood of healthy adults, dampened the allergic response to a respiratory inflammatory agent, and improved survival in cancer patients. Additional RCTs of well-characterized products are needed to more completely understand the immunomodulatory effects and specific applications of oral polysaccharides List of abbreviations ===================== ♀: female; ♂: male; Ab: antibody; AIDS: autoimmune deficiency syndrome; AOM: azoxymethane; BBN: N-butyl-N\'-butanolnitrosamine; BLCL: Burkitt\'s Lymphoma Cell Line; BW: body weight; CBC: complete blood count; CD: cluster of differentiation; CFU: colony forming unit; ConA: concanavalin A; CXCR: CXC chemokine receptor; DMBA: 7,12-dimethylbenz*(a)*anthracene; DMH: N-N\'-dimethylhydrazine; DMN: dimethylhydrazine; DSS: dextran sulfate sodium; EBV: Epstein-Barr virus; GALT: gut-associated lymphoid tissue; GI: gastrointestinal; H~2~0~2:~hydrogen peroxide; HSV: herpes simplex virus; ICR: imprinting control region; ID: intradermal; IEL: intraepithelial lymphocytes; IFN-λ: interferon gamma; IG: intragastric; IgA: immunoglobulin A; IgE: immunoglobulin E; IgG: immunoglobulin G; IgM: immunoglobulin M; IL: interleukin; IMC: invasive micropapillary carcinoma; IN: intranasally; IP: intraperitoneal; IV: intravenous; LPS: lipopolysaccharide; Mø: macrophage; mAb: monoclonal antibody; 3-MCA: methylcholanthrene; MLN: mesenteric lymph nodes; MM-46 carcinoma: mouse mammary carcinoma; MW: molecular weight; NK: natural killer; NOAEL: no observable adverse effect level; OVA: ovalbumin; PBL: peripheral blood leukocytes; PBMC: peripheral blood mononuclear cells; PHA: phytohaemagglutinin; PMA: phorbol 12-myristate 13-acetate; PML: polymorphonuclear lymphocyte; RCT: randomized, controlled trial; RNA: ribonucleic acid; SC: subcutaneous; SD rats: Sprague Dawley; TCR: T cell receptor; TLR: toll like receptor; TNF-α: tumor necrosis factor alpha; UC: ulcerative colitis; WT: wild type. Competing interests =================== The authors are employees of the Research & Development Department at Mannatech, Incorporated, which sells two of the polysaccharide products (Ambrotose^®^powder and Advanced Ambrotose^®^powder) discussed in this review. Authors\' contributions ======================= JER and EDN conducted literature searches and wrote the manuscript. RAS provided technical guidance. All authors read and approved the final manuscript. Acknowledgements ================ The authors would like to thank Barbara K. Kinsey, Ward Moore and Mrs. Jennifer Aponte for their assistance with the preparation of this manuscript, and Dr. Azita Alavi and Mrs. Christy Duncan for their editorial assistance.

Introduction ============ Phytogenic feed additives (PFA) have attracted a lot of attention in recent years. When used in diets, these substances exhibit antioxidative properties and antimicrobial activity, improve nutrient absorption, and could ultimately improve animal performance ([@bib26]; [@bib30]; [@bib37]). PFA are less toxic, have fewer side effects, and are residue-free compared with synthetic feed additives, and are thought to be ideal feed additives in animal production ([@bib25]). Therefore, many phytogenic compounds have been recommended for use as feed additives. Fenugreek (*Trigonella foenum-graecum* L) belongs to the Leguminosae family and is cultivated predominantly in India, West Asia, the Mediterranean, North Africa, and Canada ([@bib29]). The seeds are generally used for condiments in various food preparations, are regarded to have great nutritive value and restorative properties, and have been used in folk medicine for centuries for their hypoglycemic, anthelmintic, antibacterial, antiinflammatory, antipyretic, and antimicrobial properties ([@bib49]; [@bib7]; [@bib50]). The major active components of fenugreek seeds are 4-hydroxyisoleucine, trigonelline, galactomannan with flavonoids, carotenoids, coumarins, and saponins, which confer pharmacological activity and beneficial effects ([@bib46]; [@bib7]). Recently, the modes of PFA action in vivo, including aromatic plants, plant extracts, their single active components, or their blended additives, have been investigated in several studies ([@bib54]; [@bib6]; [@bib37]; [@bib38]). However, in laying hens, a comparatively small number of studies have investigated the effects of fenugreek seed on production performance and egg quality. Therefore, in this study, an attempt was made to evaluate the fenugreek seed as natural growth promoter in laying hen diets. The evaluation included biochemical changes in egg production, egg quality, blood profiles, cecal microflora, and excreta noxious gas emission. Materials and Methods ===================== Ethical Considerations ---------------------- The experimental protocols describing the management and care of animals were reviewed and approved by the Animal Care and Use Committee of Dankook University. Experimental Design, Animals, and Housing ----------------------------------------- A total of 384, 26-week-old (Hyline-brown) laying hens were used in this 6-week trial. Laying hens were randomly assigned to one of three treatments with eight replicates (16 hens/replicate). The experimental treatments were: control, basal diet (FSE 0%); fenugreek seed extract (FSE) 0.05%, basal diet + 0.05% FSE; FSE 0.1%, basal diet + 0.1% FSE. The FSE used in our study was Nutrifen^®^ (Emerald Seed Products Ltd., Avonlea, Canada), which contains ≥ 15.0 mg/g saponin. Nutrifen^®^ was composed of 3584 kcal/kg ME, 10.4% moisture, 28.0% CP, 9.9% crude fiber, 10.4% crude fat, and 4.6% crude ash. It also contained macrominerals as follows; 0.34% calcium, 0.28% sulfur, 0.74% phosphorus, 0.26% magnesium, 36.2 mg/kg zinc, 21.4 mg/kg manganese, and 36.2 mg/kg iron. Laying hens were provided ad libitum access to water and feed. All the diets were formulated in mash form to meet or exceed the [@bib36] nutrition requirement ([Table 1](#T1){ref-type="table"}). Treatment additives were included in the diet by replacing the same amount of corn. Laying hens were allowed to adjust to the environment for 5 days prior to beginning the feeding trial, during which they were fed a basal diet. They were raised in an ambient-regulated house, in which the temperature was maintained below 23°C and the light regime was set on a 16:8-light: dark cycle throughout the entire experiment. Birds were individually reared in adjacent steel cages fitted with a nipple drinker, feeder, and an egg collecting plate. ###### Formula and chemical composition of basal diet (as-fed basis) Item (%) -------------------------------------------------------- ------------- Ingredients   Corn 50.40   Soybean meal (CP 46%) 18.70   Wheat grain 10.00   Corn gluten meal   2.00   Wheat bran   5.00   Animal fat   4.40   Limestone   7.50   Dicalcium phosphate (P 18%)   1.40   Salt   0.30   [dl]{.smallcaps}-Met (50%)   0.10   Vitamin premix[^1^](#tf1){ref-type="table-fn"}   0.10   Trace mineral premix[^2^](#tf2){ref-type="table-fn"}   0.10   Total 100 Calculated values   ME (kcal/kg) 2,904         CP (%) 15.02   Lys (%)   0.78   Met + Cys (%)   0.65   Ca (%)   3.25   P (%)   0.61 Provided per kilogram of premix: 125,000 IU vitamin A; 2,500 IU vitamin D~3~; 10 mg vitamin E; 2 mg vitamin K~3~; 1 mg vitamin B~1~; 5 mg vitamin B~2~; 1 mg vitamin B~6~; 15 mg vitamin B~12~; 500 mg folic acid; 35,000 mg niacin; 10,000 mg Ca-Pantothenate and 50 mg biotin. Provided per Kg of diet: 8 mg Mn (as MnO~2~); 60 mg Zn (as ZnSO~4~); 5mg Cu (as CuSO~4~·5H~2~O); 40mg Fe (as FeSO~4~·7H~2~O); 0.3 mg Co (as CoSO~4~·5H~2~O); 1.5 mg I (as KI), and 0.15 mg Se (as Na~2~SeO~3~·5H~2~O). Laying Production, Performance, and Egg Quality ----------------------------------------------- The hen-day egg production and egg weights were recorded daily, while feed consumption was measured weekly. The feed conversion ratio was calculated as the feed consumption per hen divided by egg weight per day per hen. For each treatment, 40 normal eggs (five eggs/cage) were collected randomly at 32 weeks and used to determine the egg quality. Eggshell color scores were determined using an eggshell color fan on a 1--15 scale (1=light to 15=dark brown) by a single trained evaluator. Haugh units, albumen height, and yolk color were determined, using an egg multi tester (Touhoku Rhythm Co., Ltd., Tokyo, Japan). Eggshell breaking strength was evaluated, using an Eggshell force gauge, model II (Robotmation Co., Ltd., Tokyo, Japan), and eggshell thickness was measured, using a dial pipe gauge (Ozaki Mfg. Co., Ltd., Tokyo, Japan). Blood Profile ------------- At the end of the experiment, 16 laying hens were randomly selected from each treatment (two hens/replication) and blood samples were taken from the jugular vein by a sterilized syringe with needle. Then, the samples were transferred to either a vacuum or K3EDTA vacuum tube (Becton Dickinson Vacutainer Systems, FranklinLakes, NJ, USA). The blood samples were centrifuged at 2000×g at 4°C for 15 min to separate the serum. High-density lipoprotein (HDL), low-density lipoprotein (LDL), and total cholesterol, and immunoglobulin G (IgG) concentrations in the serum were then analyzed using an automatic biochemistry blood analyzer (HITACHI747, Tokyo, Japan). Whole blood samples from the K~3~EDTA vacuum tube were analyzed immediately to determine the white blood cells (WBC), red blood cells (RBC), and lymphocyte concentrations using an automatic blood analyzer (ADVIA 120, Bayer, Tarrytown, NY, USA). Cecal Microflora ---------------- At the end of the experiment, samples of cecal contents were collected from 16 laying hens randomly selected from each treatment, then placed on ice for transportation to the laboratory, where analyses were immediately performed using the method described by [@bib51]. One-gram of pooled cecal content sample was diluted 1:9 (wt/vol) with phosphate buffer saline solution (PBS; 0.1M, pH 7.0). Then, 10-fold serial dilutions (10^−3^ to 10^−6^) of cecal content samples were generated with PBS and placed onto Mac-Conkey (Difco Laboratories, Detroit, MI, USA) and *Lactobacillus*-Rogosa agar plates (Difco Laboratories) to isolate the *Escherichia coli* and *Lactobacillus*, respectively. The *Lactobacillus*-Rogosa and MacConkey agar plates were then incubated for 24 h at 37°C under anaerobic and aerobic conditions, respectively. After incubation, *Lactobacillus* and *E. coli* colonies were counted immediately using a colony counter, and are reported as colony-forming units (CFU) log~10~ per g. Excreta Noxious Gas Emission ---------------------------- Fresh excreta from laying hens were collected from eight cages per treatment to determine excreta noxious gas emission according to the method described by [@bib11]. Excreta samples (300 g) were stored in 2-L plastic boxes. The samples were allowed to ferment for 1 day at room temperature (28°C). After the fermentation period, the gases that formed were determined using a Gastec (model GV-100) gas sampling pump (Gastec Corp., Gastec detector tube No. 3L and 3La for ammonia; No. 4LL and 4LK for hydrogen sulfide; No. 70 and 70L for total mercaptan, Gastec Corp, detector tube, Ayase, Japan) from approximately 5 cm above the excreta samples. Statistical Analysis -------------------- Data were statistically analyzed via ANOVA using the GLM procedure of SAS (SAS Inst. Inc., Cary, NC) for a randomized complete block design. The linear and quadratic effects of FSE among treatments were analyzed using a contrast statement. The mean values and standard error of means (SEM) were reported. Probability values less than 0.05 were considered significant. Results ======= Egg Production Performance -------------------------- Laying hens fed diets supplemented with FSE during weeks 27--32 showed significant differences in egg weight compared with those fed the control treatment, as dietary FSE increased from 0.05 to 0.1% (linear, *P*=0.012). However, throughout the experimental period, there was no significant difference in egg production, feed intake, and feed conversion in laying hens fed different levels of FSE in their diet ([Table 2](#T2){ref-type="table"}). ###### The effects of dietary fenugreek seed extract (FSE) on productivity in laying hens[^1^](#tf3){ref-type="table-fn"} FSE, % SEM[^2^](#tf4){ref-type="table-fn"} *P*-value ---------------------- -------- ------------------------------------- ----------- ------ ------- ------- Egg production, % 96.0 97.9 96.5 1.21 0.677 0.115 Egg weight, g 63.1 64.2 66.4 0.89 0.012 0.642 Feed intake, g 115 115 117 1.93 0.659 0.698 Feed conversion, g:g 1.833 1.797 1.757 0.02 0.052 0.952 Each treatment mean represents 8 replicates (16 hens/replicate). Standard error of mean Egg Quality ----------- Eggshell breaking strength and eggshell thickness were found to be increased in the FSE groups compared with the control group (linear *P*\<0.05). As the dietary levels of FSE increased, a linear increase in the intensity of yolk color was observed (*P*=0.001). There was no significant difference in eggshell color, Haugh units, or albumen height of laying hens fed different levels of FSE in their diet ([Table 3](#T3){ref-type="table"}). ###### The effects of dietary fenugreekseed extract (FSE) on egg quality in laying hens[^1^](#tf5){ref-type="table-fn"} FSE, % SEM[^2^](#tf6){ref-type="table-fn"} *P*-value -------------------------------------- -------- ------------------------------------- ----------- ------ ------- ------- Eggshell color 11.25 11.35 11.3 0.19 0.855 0.752 Eggshell breaking strength, kg/cm^2^ 4.390 4.483 4.661 0.06 0.012 0.588 Eggshell thickness, mm^−2^ 35.41 36.36 36.73 0.43 0.038 0.304 Haugh unit 82.52 83.03 82.62 1.41 0.959 0.788 Albumen height, mm 7.55 7.84 7.68 0.19 0.635 0.355 Yolk color 7.0 7.1 7.8 0.14 0.001 0.091 Each treatment mean represents 40 replicates (5 eggs/replicate). Standard error of mean Blood Profiles -------------- Serum levels of HDL-cholesterol were elevated in the FSE treatment group compared with the control group (linear, *P*=0.017). The total cholesterol concentration decreased as dietary FSE increased in laying hens compared with laying hens fed the control diet (linear, *P*=0.042). However, serum LDL-cholesterol and IgG levels were not affected by the addition of FSE. Furthermore, FSE treatment had no significant effect on WBC, RBC, or lymphocytes ([Table 4](#T4){ref-type="table"}). ###### The effect of dietary fenugreekseed extract (FSE) on blood profiles in laying hens[^1^](#tf7){ref-type="table-fn"} FSE, % SEM[^2^](#tf8){ref-type="table-fn"} *P*-value ------------------------------- -------- ------------------------------------- ----------- -------- ------- ------- White blood cells, 10^3^/*µl* 27.0 27.3 31.5 3.92 0.365 0.636 Red blood cells, 10^6^/*µl* 2.3 2.6 2.3 0.15 0.906 0.141 Lymphocyte, % 70.4 60.7 66.2 9.92 0.745 0.495 HDL-cholesterol, mg/dL 33 41 44 2.05 0.017 0.503 LDL-cholesterol, mg/dL 63 62 61 6.24 0.848 0.987 Total cholesterol, mg/dL 222 172 184 11.14 0.042 0.067 Immunoglobulin G, mg/dL 446 469 452 106.45 0.968 0.883 Each treatment mean represents 16 replicates (2 hens/replicate). Standard error of mean Cecal Microflora ---------------- FSE supplementation linearly increased *Lactobacillus* numbers in the cecum, compared with the control treatment (*P*=0.012). Cecal *E. coli* decreased quadratically as dietary FSE increased (*P*=0.010) ([Table 5](#T5){ref-type="table"}). ###### The effects of dietary fenugreekseed extract (FSE) on cecal microbiota in laying hens[^1^](#tf9){ref-type="table-fn"} FSE, % SEM[^2^](#tf10){ref-type="table-fn"} *P*-value -------------------------------- -------- -------------------------------------- ----------- ------ ------- ------- *Lactobacillus*, log~10~ cfu/g 6.98 7.33 7.42 0.10 0.012 0.298 *E. coil*, log~10~ cfu/g 6.57 6.14 6.37 0.08 0.151 0.010 Each treatment mean represents 16 replicates (2 hens/replicate). Standard error of mean Excreta Noxious Gas Emissions ----------------------------- Excreta ammonia emissions were decreased as dietary FSE supplementation increased (linear, *P*=0.020). However, FSE supplementation did not affect total mercaptan and hydrogen sulfide emissions, compared with the control treatment ([Table 6](#T6){ref-type="table"}). ###### The effects of dietary fenugreekseed extract (FSE) on excreta gas emission in laying hens[^1^](#tf11){ref-type="table-fn"} FSE, % SEM[^2^](#tf12){ref-type="table-fn"} *P*-value ----------------------- -------- -------------------------------------- ----------- ------ ------- ------- Total mercaptan, ppm 1.4 0.9 1.0 0.17 0.058 0.134 Ammonia, ppm 48 33 30 4.40 0.020 0.277 Hydrogen sulfide, ppm 2.4 2.4 2.4 0.07 0.987 0.892 Each treatment mean represents 8 replicates (16 hens/replicate). Standard error of mean Discussion ========== Laying performance of the FSE-treated groups did not significantly differ over 27--32 weeks in our study. PFA has been investigated in numerous studies on poultry with varying responses, depending upon the plant species, disease challenges, and different environmental conditions used. Results have also varied depending on the specific plant-derived compound used, or stress type ([@bib15]; [@bib22]; [@bib10]; [@bib12]). Several studies have reported that dietary supplementation with extracts of some plants increased egg production, feed conversion ratio, and egg weight in laying hens ([@bib48]; [@bib35]; [@bib41]). Recent reports have also suggested an improvement in the early growth performance of broilers and pigs, when supplemented with FSE ([@bib28]; [@bib39]). In contrast, previous studies have shown that neither PFA nor FSE exhibited significant effects on feed intake, feed efficiency, and/or egg production of laying hens ([@bib8]; [@bib16]; [@bib3]). Growth responses of broilers, or the egg production performance of laying hens supplemented with PFA remain controversial, and the mechanism of FSE action on poultry performance has not yet been clearly established. Our finding suggests that FSE had no direct beneficial effect on egg production, feed intake, or feed conversion. However, a significant improvement in egg weight, eggshell strength, and eggshell thickness was observed when FSE was included in the diet. The results of our study were similar to those reported recently by [@bib19], who showed significant effects on the eggshell thickness, eggshell weight, and specific gravity between control and fenugreek powder groups. [@bib40] reported that fenugreek seeds in the diet may enhance nutrient absorption due to greater villus height and crypt depth, thereby improving intestinal health. Similarly, [@bib5] found that the apparent ileal digestibility of crude ash, calcium, and phosphorus increased linearly with increasing dose of PFA in the diet. Moreover, fenugreek seeds naturally contain high levels of minerals, including calcium, iron, and zinc ([@bib27]). The beneficial effect on egg weight, eggshell breaking strength, and eggshell thickness might be attributable to favorable alterations in the intestinal environment and function, which may have increased intestinal calcium absorption. Therefore, FSE may help calcium utilization for egg-shell formation, although no significant differences were observed in egg production performance between the FSE groups and the control. In our study, the addition of FSE increased the intensity of the yolk color. Fenugreek seeds are also widely used as a spice and food colorant in food preparations because of their light yellow-brown color ([@bib13]). An increase in the intensity of yolk color is likely to be due to the action of the yellow color agent present in FSE on the yolk following the addition of FSE. Thus, FSE could be supplied as a component of pigment in the diets in order to influence egg yolk yellowness. Serum lipid metabolites in relation to PFA and their extracts have been extensively studied. The addition of PFA could affect serum triglyceride and cholesterol concentrations and effectively prevent the progress of hypercholesterolemia and cholesterol accumulation in the liver induced by a high cholesterol diet in rats ([@bib20]). PFA also significantly decreased blood lipid fractions and increased high-density lipoprotein in poultry ([@bib1]). Results obtained in the present study show that FSE increases serum HDL-cholesterol, and decreases total cholesterol, regardless of LDL-cholesterol levels. Some studies have indicated that fenugreek seeds have hypocholesterolemic activity, which may reduce the risk of heart disease ([@bib42]; [@bib31]). Saponins, which are a major active constituent of the fenugreek seeds, have beneficial effects on blood cholesterol and triglyceride concentrations in rats, rabbits, and humans ([@bib34]; [@bib43]; [@bib2]). Additionally, non-starch polysaccharides, which provide the major soluble fiber content in fenugreek seeds, include galactomannan ([@bib9]), which has been reported to reduce blood lipids ([@bib53]; [@bib44]). The presence of these compounds in particular could increase the viscosity of the digested food and decrease serum total cholesterol levels by inhibiting bile salt reabsorption in the small intestine ([@bib32]). Therefore, the decreased concentration of blood lipid metabolites observed in the present study due to the inclusion of dietary FSE may be related to the saponins or galactomannan fiber content in FSE. Several PFAs have long been recognized for their antimicrobial actions against major pathogens including coccidium and fungi ([@bib21]; [@bib52]). An in vitro study by [@bib18] found that carvacrol and thymol inhibited the growth of *E. coli, Clostridium perfringens*, and *Salmonella* strains. [@bib24] reported that in vitro, essential oils including basil, lemon balm, marjoram, oregano, rosemary, sage, and thyme, strongly inhibited the growth of *B. cereus, P. aeruginosa, E. coli*, and *L. monocytogenes*. In an in vivo study, [@bib23] demonstrated that the herb extracts were largely associated with reduced cecal *Bacteroides* spp., enterococci, and *E. coli* numbers, but increased numbers of bifidobacteria and lactobacilli, relative to the control and antibiotic groups in chickens. In addition, [@bib55] noted that a fenugreek diet caused higher *Lactobacillus* and *Clostridium* concentrations and lower *Escherichia, Hafnia*, and *Shigella* concentrations in the small intestine of piglets. Many investigators have reported an antibacterial effect of fenugreek seeds ([@bib14]; [@bib4]; [@bib33]). Such an effect seems to be related to the presence of certain molecular compounds, usually in the form of secondary metabolites, such as flavonoids, saponins, and phenolic compounds ([@bib14]). In our study, laying hens fed the FSE diets had a higher population of cecal *Lactobacillus* and a lower population of *E. coli*. Therefore, decreased cecal *E. coli* counts in laying hens fed dietary FSE may be explained by the antibacterial activity of FSE against different pathogenic bacteria. These effects of FSE on cecal microflora have led to a reduction in ammonia gas emission in laying hen excreta. In other words, a positive effect of FSE is exerted by suppressing harmful cecal *E. coli*, and favoring beneficial *Lactobacillus*, which may be reflected in the ammonia gas emissions of the excreta. This result is in line with the reports of [@bib47] and [@bib45], who determined that supplemental PFA decreased noxious gas emission, including ammonia concentrations, in pigs and laying hens. Furthermore, decreased levels of excreta ammonia gas found in the FSE groups may be associated with intestinal pH. [@bib17] observed that supplementation of quail diets with PFA (thyme and black seed essential oils) significantly decreased intestinal pH. Similarly, [@bib55] indicated that fenugreek seed reduced the pH in the cecum and colon of piglets, due to an increase in the number of lactate-producing bacteria or their relevant metabolic activity. Therefore, the decrease in ammonia gas emission observed in excreta following the addition of FSE to laying hen diets is thought to be induced by the maintenance of beneficial microbiota or a reduction in intestinal pH. In conclusion, based on these findings, FSEs are a valuable natural feed additive for laying hens, particularly with regard to egg weight, egg quality, serum cholesterol, cecal lactobacilli number and excreta ammonia gas emission.

Updated to correct an editorial error in the element named in the deck. The largest industrial application of olefin metathesis today is the synthesis of propylene from ethylene and butenes^[@ref1]^ employing WO~3~ on SiO~2~, a relatively long-lived and regenerable catalyst that operates at 350--400 °C. It is widely proposed that high temperatures are required because the percentage of metal sites actually involved in the metathesis reaction is extremely low, or the reaction that generates alkylidenes is not a high yield reaction, or both. A recent paper by Copéret, Mashima, and co-workers^[@ref2]^ tackles head-on the question concerning how in WO~3~/SiO~2~ catalysts the alkylidene is formed from an olefin alone. Hundreds of papers have attempted to answer this question, although one has to admit that there may not be a single answer for all supported oxide catalysts or all olefins. Copéret and Mashima employ Me~4~BTDP to reduce four-coordinate (SurfO)~2~WO~2~ sites on silica in the absence of olefins to give 2,3,5,6-tetramethylpyrazine, hexamethyldisiloxane, and M(IV) sites ([eq [1](#eq1){ref-type="disp-formula"}](#eq1){ref-type="disp-formula"}). Analogously, five-coordinate (SurfO)~4~WO sites are also reduced to (SurfO)~4~W(IV) sites. When the purple solid containing a high percentage of W(IV) sites produced in this manner is then exposed to *cis*-4-nonene and heated to 70 °C, 1000 equiv of the alkene are metathesized in 6 h. When instead ethylene is added to the purple solid, solid-state NMR studies reveal that propene is formed along with unsubstituted square pyramidal metallacyclobutane and metallacyclopentane complexes. A variety of experiments led the authors to conclude that above 70 °C, metathesis activity can be ascribed to a relatively efficient contraction of a metallacyclopentane ring to a metallacyclobutane ring, from which loss of propylene generates an initial methylidene complex (eq 2). Ultimately, rearrangement of a metallacyclobutane complex to an olefin results in reduction to W(IV) and reformation of a metallacyclopentane and subsequently another methylidene. It is not yet known whether only TBP (SurfO)~2~W(O)(C~4~H~8~) sites undergo this "ring-contraction" to give a methylidene. "Ring-contraction" was discovered in the process of exploring reactions between tantalum(III) olefin complexes and terminal olefins to give two dimers of the terminal olefins, not metathesis products. This reaction turned out to be a good model for nonmetathetical steps in alkylidene/metallacycle chemistry of Mo and W.^[@ref3],[@ref4]^ It was recognized at the time that "the MC~4~ to MC~3~ ring contraction is a straightforward and reasonable way of forming an alkylidene ligand from olefins---assuming that some MC~3~ complexes which form in this manner will cleave to give metathesis-type products instead of rearranging."^[@ref3]^ Although unsubstituted d^0^ metallacyclopentane (MC~4~) complexes of Mo and W (especially) have been observed as the end products of a decomposition "cascade" in the presence of ethylene,^[@ref5],[@ref6]^ there is little hard evidence in homogeneous systems that alkylidenes arise from M(IV) olefin complexes^[@ref7]^ through ring-contraction of metallacyclopentanes in homogeneous metathesis reactions at 22 °C. Virtually the only exception in Mo-based or W-based olefin metathesis systems is the catalytic homologation of vinyltributylstannane to allyltributylstannane in the presence of ethylene,^[@ref8]^ which can so far only be explained through a ring-contraction mechanism. An alternative to ring-contraction as a mechanism of forming an alkylidene is a mechanism in which an allyl hydride is formed through allylic CH activation in an olefin. Allyl hydrides are intermediates in rearrangement of a metallacyclobutane to an olefin and consequent reduction of a d^0^ complex to a d^2^ olefin complex with loss of metathesis activity, so formation (to some degree) of a metallacyclobutane from an alkenyl hydride also seems feasible. The work by Copéret and Mashima may revolutionize the synthesis and use of inexpensive supported metathesis catalysts for hydrocarbons on an industrial scale by allowing the use of much lower temperatures than currently employed. It also may open up opportunities for regenerating catalysts in flow systems. However, it remains to be seen to what extent functional groups are tolerated as metathesis substrates or whether C=C bond isomerization^[@ref7]^ becomes a complication at the temperatures employed. Finally, it also must be noted that the level of selectivity found in homogeneous catalysts today^[@ref9]^ may be difficult to match in a heterogeneous catalyst since the latter are unlikely to contain true (100%) "single sites" that can be tuned with the high level of molecular precision as soluble catalysts.

1. Introduction {#sec0005} =============== Cytokines are a broad and loose group of small cell-signaling proteins that play important roles in cell growth, proliferation, differentiation, apoptosis, angiogenesis, immunity and inflammatory response. They include interferons (IFN), interleukins (IL), chemokines, lymphokines, and tumor necrosis factors (TNF). They are produced by a variety of cells, including macrophages, B and T lymphocytes, mast cells, endothelial cells, epithelial cells, fibroblasts, and various stromal cells. Cytokines that are produced at the site of infection stimulate and coordinate innate and adaptive immune responses against invading pathogens ([@bib0135], [@bib0185], [@bib0365]). They act through their matching receptors on the surface of target cells, followed by cascades of intracellular signaling. One such frequently activated intracellular signaling is the JAK-STAT pathway, which is indispensable and pivotal in many biological processes including immunity and inflammatory response ([@bib0275], [@bib0340]). Dysregulation of JAK-STAT signaling results in immunodeficiency and immune-mediated disorders ([@bib0270], [@bib0275]). Mutations in the components of JAK-STAT pathway cause immunodeficient and autoimmune disorders ([@bib0050], [@bib0190]). Due to the importance of JAK-STAT signaling in the host immune response, it is often targeted by pathogens, including PRRSV ([@bib0290], [@bib0400], [@bib0405]). PRRSV causes a contagious disease that is characterized by reproductive failure in sows and respiratory disease of variable severity in pigs of all ages ([@bib0235]). PRRS has caused substantial economic losses to the swine industry and remains one of the most economically important diseases in pigs since it was first reported in 1987 ([@bib0165], [@bib0265]). A typical feature of the immune response to PRRSV infection in pigs is delayed production and low titer of virus neutralizing antibodies, and weak cell-mediated immune response ([@bib0210], [@bib0235], [@bib0435]). PRRSV infection is also characterized by prolonged viremia followed by the persistent presence in regional lymph nodes for as long as 250 days ([@bib0425]). One of the possible reasons for the weak protective immune response is that PRRSV interferes with innate immunity, including type I interferons (IFNs), and cytokine-mediated JAK-STAT signaling ([@bib0010], [@bib0045], [@bib0290], [@bib0350], [@bib0405], [@bib0395]). 1.1. JAKs and STATs {#sec0010} ------------------- In mammals, there are four JAKs: JAK1, JAK2, JAK3, and Tyk2, ranging in size from 120 to 140 kDa ([@bib0140]). JAK1, JAK2, and Tyk2 are ubiquitously expressed, whereas JAK3 expression is restricted to cells of the hematopoietic system. The JAK protein is pre-associated with cytokine receptors in the cytoplasmic side and is an important determinant of their levels and signaling potential ([@bib0140]). Upon cytokine binding, the receptor chains are brought into close proximity, leading to the juxtaposition of two JAK kinase domains and consequent trans-phosphorylation. Once activated, JAKs phosphorylate STAT proteins via Src homology 2 (SH2) domain interaction ([@bib0155]). Even though responding to different cytokines, JAKs selected by different receptors activate specific STAT members for defined functions ([@bib0140]). There are seven mammalian STAT proteins: STAT1, STAT2, STAT3, STAT4, STAT5A, STAT5B, and STAT6, which range from 750 to 950 amino acids in polypeptide length and feature several conserved domains ([@bib0330], [@bib0340]). STATs are latent transcription factors located in the cytoplasm until activated. Each STAT member responds to a defined set of cytokines ([@bib0205], [@bib0270], [@bib0275]). The seven STATs go through similar activation processes and exhibit global conservation in function ([@bib0340]). In brief, ligand-mediated receptor multimerization leads to trans-phosphorylation of JAKs, which then create docking sites in the receptor for STATs and phosphorylate them. Phosphorylated STATs form homodimer or heterodimer complexes, followed by translocation into the nucleus by importins and binding to response element in DNA to activate or repress transcription of a defined set of genes ([@bib0340]). 1.2. STAT signaling and functions {#sec0015} --------------------------------- Among the seven STATs, STAT1 and STAT2 mainly mediate the IFN-activated signaling ([@bib0270]). STAT1 is involved in signaling by type I, type II, and type III IFNs. In response to type I IFNs (IFN-α or IFN-β), STAT1 and STAT2 are phosphorylated, followed by heterodimer formation and then interaction with interferon regulatory factor 9 (IRF9) to form a heterotrimer known as interferon-stimulated gene factor 3 (ISGF3) ([@bib0090]) ([Fig. 1](#fig0005){ref-type="fig"} A). The ISGF3 is translocated into nucleus and binds to interferon-stimulated response element (ISRE) in DNA to activate the expression of interferon-stimulated genes (ISGs). Upon IFN-γ stimulation, activated STAT1 forms homodimers, followed by nuclear translocation and activation of gene expression via binding to interferon-gamma-activated-sequence (GAS) in DNA ([@bib0270]). Type III IFNs also activate STAT1 and STAT2 for ISRE transactivation like type I IFNs ([@bib0470]).Fig. 1PRRSV interference with type I IFN-activated JAK-STAT signaling. A. Canonical signaling. IFN-α/β binds to their receptors IFNAR-1 and IFNAR-2 on the cell membrane and activates the JAK-STAT1/STAT2 pathway. The phosphorylated STAT1 and STAT2 form heterodimer, followed by interaction with IRF9 to form interferon-stimulated gene factor 3 (ISGF3). Karyopherin α1 (KPNA1), an adaptor protein binding ISGF3, is essential to mediate the nuclear import of ISGF3 via interaction with karyopherin β1 (KPNB1). The ISGF3 binds to interferon-stimulated response element (ISRE) in DNA to activate transcription of interferon-stimulated genes (ISGs). "P" besides STATs indicates phosphorylation. PRRSV nsp1β inhibits ISGF3 nuclear translocation via inducing degradation of KPNA1. PRRSV N protein also inhibits ISGF3 nuclear translocation. PRRSV nsp2 reduces ISG15 production and conjugation via its deubiquitination activity. PRRSV induces elevation of miRNA miR-30c to downregulate JAK1 and SOCS1 to inhibit JAKs. PRRSV inhibits PKR during its early infection of pulmonary alveolar macrophages. B. STAT1-independent signaling. Type I IFNs activate alternative JAK-STAT2 signaling without STAT1. The ISGF3-like complex binds to ISRE and interferon-gamma-activated sequence (GAS) to activate alternative sets of ISGs. PRRSV reduces STAT2 protein to inhibit this pathway.Fig. 1 In addition to the canonical signaling described above, IFNα signaling occurs through alternative complexes containing STAT2 and IRF9 without STAT1 ([@bib0030], [@bib0110]) ([Fig. 1](#fig0005){ref-type="fig"}B). Moreover, STAT2 can form heterodimer with other STATs, like STAT3 and STAT6, followed by binding to diverse sequences, like GAS. Further studies demonstrate the existence of a STAT1-independent IFN signaling pathway, in which STAT2/IRF9 directs a prolonged antiviral activity ([@bib0025], [@bib0035]). STAT3 is activated by many cytokines and had multiple functions including differentiation of T helper 17 (Th17) and generation of CD8^+^ T cell memory response ([@bib0055], [@bib0275], [@bib0380]). Numerous cytokines including IL-5, IL-6, IL-9, IL-10, IL-11, IL-12, IL-21, IL-22, IL-27, oncostatin M (OSM), IFN-γ, TNF-α and leukemia inhibitory factor (LIF), trigger STAT3 activation ([@bib0125], [@bib0205]). The IL-6 family of cytokines including IL-6, OSM, and LIF bind to the receptor complex containing the common glycoprotein 130 (gp130) and activate STAT3, known as gp130/JAK-STAT3 signaling ([Fig. 2](#fig0010){ref-type="fig"} ). STAT3 is needed for differentiation of follicular T helper and Th1 cells ([@bib0295]), as well as activation and maturation of dendritic cells (DCs) ([@bib0285]). Mutations in STAT3 cause autosomal dominant hyper-IgE syndrome, a rare multisystem primary immunodeficiency characterized by recurrent bacterial infections in skin and lung and with abnormally high levels of IgE ([@bib0160], [@bib0245]). STAT3 is indispensable for promoting host defense against virus infections. For instance, during Herpes simplex virus-1 (HSV-1) infection, STAT3 promotes the activation of CD8+ T cells response ([@bib0460]). It has been shown that gp130-STAT3 signaling is critical for the innate immune response against coxsackievirus B3 virus (CVB3) infection ([@bib0450]). In addition, STAT3 plays a protective role in regulating virus-induced proinflammatory response, as shown in STAT3 knock-out studies ([@bib0100], [@bib0195], [@bib0240]). Highly pathogenic avian influenza (HPAI) H5N1 virus induces strong proinflammatory response in chickens via inhibition of STAT3 phosphorylation ([@bib0200]).Fig. 2PRRSV interference with JAK-STAT3 signaling activated by IL-6 family cytokines. IL-6 binds to its receptor IL-6R and gp130, leading to JAK phosphorylation of STAT3, followed by STAT3 homodimer formation. KPNA6 is the adaptor protein to bind to STAT3 and interact with KPNB1 for the nuclear translocation. The STAT3 homodimer binds to STAT3 response element (RE) in DNA to activate transcription of target genes. PRRSV nsp5 induces STAT3 degradation to inhibit JAK-STAT3 signaling.Fig. 2 STAT4 is activated by IL-12 and is essential for Th1 cell differentiation ([@bib0325]). Even though its distribution is restrained in myeloid cells, testis and thymus, STAT4 is critical for the host immunity. It is also activated by IL-23 to induce expansion of Th17 cells and the associated autoimmunity ([@bib0325]). Moreover, STAT4 is crucial for the biological effects of macrophage, natural killer cell, mast cell, and dendritic cell as well as IFN-γ production ([@bib0225]). STAT5 is activated by IL-2, IL-3, IL-5 and granulocyte macrophage colony-stimulating factor (GM-CSF), and is essential for regulatory T cell (Treg) differentiation ([@bib0270]). STAT5 can also be activated by IL-7 and IL-15, contributing to the generation of CD8^+^ memory cells and B lymphopoiesis ([@bib0270], [@bib0275]). STAT5A and STAT5 B are two highly related proteins and have indispensable roles, especially to the effector and Treg response, for which STAT5 B is dominant ([@bib0385]). Tregs takes charge of maintaining homeostasis and controlling the immune response by restraining immunocompetent effector cells ([@bib0020]). STAT6 is activated by IL-4 and IL-13 and is pivotal for Th2 and Th9 lymphocyte differentiation ([@bib0275], [@bib0390]) ([Fig. 3](#fig0015){ref-type="fig"} A). STAT6 has been demonstrated to regulate lung inflammatory responses in animal models ([@bib0390]). Moreover, STAT6 contributes to alternative activation of macrophages and lung anti-viral responses in a JAK-independent manner ([@bib0065]). Viral or cytoplasmic nucleic acids trigger STING (stimulator of interferon genes) or MAVS (mitochondrial antiviral-signaling protein) to recruit STAT6 to the endoplasmic reticulum, followed by TBK1 (TANK-binding kinase 1) phosphorylation of STAT6 and nuclear translocation ([Fig. 3](#fig0015){ref-type="fig"}B). Expression of chemokines, including CCL2, CCL20, and CCL26, are then activated to recruit immune cells to combat viral infection ([@bib0065]).Fig. 3PRRSV interferes with STAT6 signaling. A. Canonical JAK-STAT6 signaling. IL-4 and IL-13 bind to receptors, leading to JAK phosphorylation of STAT6, followed by homodimer formation and nuclear translocation to activate target genes. PRRSV reduces STAT6 protein level to inhibit the signaling. B. JAK-independent STAT6 signaling. Viral nucleic acids (dsRNA or dsDNA) activate MAVS and STING on mitochondria and endoplasmic reticulum (ER), respectively, leading to TBK1 (TANK-binding kinase 1) phosphorylation of STAT6, followed by homodimer formation and nuclear translocation to activate alternative target genes, including antiviral chemokines. PRRSV reduces STAT6 protein level to inhibit the signaling.Fig. 3 1.3. PRRSV {#sec0020} ---------- PRRSV is a small, enveloped RNA virus of the genus *Arterivirus*, family *Arteriviridae*, order *Nidovirales* ([@bib0105]). There are two PRRSV genotypes: Type 1 (European PRRSV) and Type 2 (North American PRRSV), which have around 60% genomic nucleotide identity ([@bib0105]). In a new proposal to the International Committee on Taxonomy of Viruses (ICTV), PRRSV Type 1 and Type 2 viruses are classified as two separate species: *Porcine reproductive and respiratory syndrome virus 1* and *Porcine reproductive and respiratory syndrome virus 2,* respectively, due to their big genomic difference ([@bib0040]). PRRSV genome is a linear, non-segmented, single-stranded, and positive-sense RNA in a size of approximately 15 kb ([@bib0075], [@bib0105]). Over ten open reading frames (ORFs) have been identified in the PRRSV genome. About four-fifths of the genome encode polyproteins that are cleaved into 14 non-structural proteins (nsps) and the 3′-terminal one-fifth encodes eight structural proteins ([@bib0235]). The main target cells for PRRSV infection *in vivo* are some monocyte/macrophage lineages, mainly pulmonary alveolar macrophages (PAMs) ([@bib0235], [@bib0320], [@bib0420]). PAMs are important effector immune cells of the innate immune system against pathogens in the lung ([@bib0115]). Cytokines like IFN-γ and GM-CSF activate JAK-STAT signaling to regulate macrophage phenotype and activation ([@bib0170]). Activated macrophages secrete immune regulatory cytokines including IL-1β, IL-6, IL-12, TNF-α, and so on ([@bib0115]). 2. PRRSV interference with JAK-STAT signaling {#sec0025} ============================================= To evade the host antiviral response, viruses have evolved numerous strategies including dysregulating JAK-STAT pathway. For example, Epstein-Barr virus (EBV) suppresses IFN signaling by inhibiting the expression of the IFN-γ receptor ([@bib0255]). Paramyxovirus V protein induces STAT protein degradation to evade IFN response ([@bib0280]). Ebola virus VP24 blocks pSTAT1 (phosphorylated STAT1) nuclear translocation by binding KPNA1 ([@bib0300], [@bib0305]). The ORF6 product of severe acute respiratory syndrome coronavirus disrupts nuclear import of pSTAT1 by tethering KPNA2 to the endoplasmic reticulum/Golgi membrane ([@bib0120]). Highly pathogenic avian influenza (HPAI) H5N1 virus inducing a strong proinflammatory response in chickens by inhibiting STAT3 phosphorylation ([@bib0200]). PRRSV is known to inhibit IFN-activated JAK-STAT signaling by blocking the ISGF3 nuclear translocation ([@bib0290], [@bib0405]). 2.1. PRRSV inhibits IFN-mediated JAK-STAT signaling {#sec0030} --------------------------------------------------- PRRSV inhibits the IFN-activated JAK-STAT signal transduction and ISG expression in both MARC-l45 and PAM cells ([@bib0290], [@bib0400], [@bib0405]). PRRSV proliferation in MARC-145 cells suppresses JAK-STAT signaling stimulated by IFN-α. The transcripts of ISG15 and ISG56 and protein level of STAT2 in PRRSV-infected cells were much lower than mock-infected cells upon IFN stimulation. PRRSV blocks the nuclear translocation of the IFN-induced ISGF3 complex via nsp1β ([@bib0060], [@bib0290]). Avirulent Ingelvac^®^ PRRS MLV has no effect on the IFN-activated JAK-STAT signaling in PAMs ([@bib0290]). Further studies demonstrate that nsp1β inhibits the JAK-STAT signaling via inducing the degradation of KPNA1, which is a critical adaptor protein to mediate the nuclear import of ISGF3 ([@bib0405]) ([Fig. 1](#fig0005){ref-type="fig"}A). Infection of MARC-145 cells by moderate virulent PRRSV strains VR-2332 and VR-2385 also result in KPNA1 reduction, whereas the Ingelvac^®^ PRRS MLV does not. Nsp1β of VR-2385 induces elevation of KPNA1 ubiquitination and shortening of its half-life. Analysis of nsp1β deletion constructs identifies its N-terminal domain to be involved in the ubiquitin-proteasomal degradation of KPNA1 ([@bib0405]). Sequence analysis of nsp1β from VR-2332 and MLV indicates there are only two different nucleotides, leading to two variable amino acids at residue 19 and 151. Substitution of the N-terminal nucleotide resulting in alteration of residue 19 from valine to isoleucine abolishes the ability of VR-2385 nsp1β to induce KPNA1 degradation and to inhibit IFN-mediated signaling. In contrast, MLV nsp1β has no effect on KPNA1, however, a mutant MLV nsp1β with residue 19 alternation from isoleucine to valine gains the ability to induce KPNA1 degradation ([@bib0405]). These data demonstrate that nsp1β blocks ISGF3 nuclear translocation to inhibit JAK-STAT signaling via inducing KPNA1 degradation and that the residue valine-19 in nsp1β correlates with the inhibition. Besides nsp1β, other PRRSV proteins including nsp7, nsp12, GP3 and N also inhibit IFN-induced downstream signaling, albeit at a smaller scale ([@bib0400]). The N protein inhibits IFN-activated signaling by blocking STAT1 nuclear translocation ([@bib0400]). Among PRRSV strains, there are variable effects on the IFN-activated JAK-STAT signaling. In MARC-145 cells, PRRSV strains VR-2385, VR-2332, NVSL97-7895, and Lelystad, but not MN184, block the activity of exogenous IFN-α ([@bib0400]). In primary PAMs, strain VR-2385, VR-2332, MN184, and Lelystad, but not NVSL, inhibit the activity of IFN-α. For NVSL97-7895 and MN184, the same virus infection in MARC-145 and PAM cells has variable effects on the IFN-activated signaling. This is not totally unexpected as NVSL strain differs from other PRRSV strains in its failure to induce IL-10 expression *in vivo* ([@bib0345]). These two strains might have alternative interacting mechanisms with the JAK-STAT signaling in the two types of cells. 2.2. PRRSV inhibits JAK-STAT3 signaling {#sec0035} --------------------------------------- Among all the STAT proteins, STAT3 is known as highly pleiotropic in mediating the expression of a variety of genes in response to both cytokines and growth factors, and thus plays a pivotal role in numerous cellular processes including cell survival, proliferation, embryogenesis, and immune responses ([@bib0205], [@bib0445]). Thus STAT3 has been found to be the target of some viral pathogens. Measles virus V protein inhibits IL-6 mediated STAT3 signaling ([@bib0370]). V protein of mumps virus prevents responses to IL-6 and v-Src by inducing STAT3 ubiquitination and degradation ([@bib0375]). Rabies virus interferon antagonist P protein inhibits gp130 receptor signaling by interacting with activated STAT3 ([@bib0230]). Inhibition of STAT3 signaling by these viruses and PRRSV can lead to inhibition of a broad spectrum of cytokines and growth factors to thwart host antiviral responses and allow virus replication and spread *in vivo*. OSM, a member of the IL-6 family activating JAK-STAT3 signaling, enhances the antiviral effects of IFN-α and plays a role in the induction of an adaptive immune response to pathogens ([@bib0175], [@bib0215]). Treatment of MARC-145 cells with OSM alone leads to inhibition of PRRSV replication ([@bib0455]). To overcome the inhibition, PRRSV uses a strategy to reduce STAT3 protein level. PRRSV infection of MARC-145 cells and primary PAMs leads to significant reduction of STAT3, whereas it has minimum effect on STAT1 protein level ([@bib0455]). Several PRRSV strains tested induce a similar reduction of STAT3 protein level but have no effect on its transcript level. Treatment of the PRRSV-infected cells with MG-132, a proteasomal inhibitor, restores the STAT3 level. The further study identifies that PRRSV nsp5 induces the STAT3 degradation, shown by increased polyubiquitination level and shortened half-life ([@bib0455]). As a result, nsp5 inhibited STAT3 signaling. Further study is being undertaken to elucidate the mechanism. 2.3. PRRSV interference with other STATs and JAKs {#sec0040} ------------------------------------------------- As aforementioned, other STATs also have important roles in the host immune response. Besides inhibiting JAK-STAT1/STAT2 and JAK-STAT3 signaling, PRRSV might interfere with other STATs. Our preliminary study shows that PRRSV reduces STAT2 and STAT6 protein level (unpublished result). Further study is being undertaken to delineate the mechanisms. A recent study shows that PRRSV upregulates a host microRNA, miR-30c, which is a negative regulator by targeting JAK1 ([@bib0465]). PRRSV reduces JAK1 expression in infected cells and is expected to affect the phosphorylation of both STAT1 and STAT2. But in our studies, IFN-induced phosphorylation of both STAT1 and STAT2 in PRRSV-infected MARC-145 cells was not affected ([@bib0290]). It may play a role *in vivo* as PRRSV infection in pigs leads to elevation of miR-30c in lungs and PAMs and its level corresponds to the viral load ([@bib0465]). This indicates that PRRSV has multiple strategies to block host IFN-signaling. 3. Perspective {#sec0045} ============== PRRSV uses multiple strategies to evade the host innate and adaptive immunity. By interfering with the JAK-STAT signaling, PRRSV may perturb the function of cytokines in the regulation of the host immune response. 3.1. PRRSV and IFN-mediated JAK-STAT signaling {#sec0050} ---------------------------------------------- IFNs are essential for antiviral response and targeted by PRRSV at multiple levels ranging from the induction of IFNs, IFN-activated JAK-STAT signaling, to ISGs. PRRSV interference of IFN induction has been reviewed elsewhere ([@bib0350], [@bib0395]). The modulation of innate immunity by nsp1 is conserved in all members of *Arteriviridae* ([@bib0150]). PRRSV also interferes with ISGs, for example, nsp2 inhibits ISG15 by reducing its production and counteracting its conjugation to cellular proteins ([@bib0355]). PRRSV downregulates interferon-induced double-strand RNA-activated protein kinase (PKR) at its early infection of PAMs ([@bib0440]). PRRSV counteracts interferon-induced transmembrane protein 1 (IFITM1) and Tetherin in MARC-145 cells by nsp3 and E proteins, respectively, ([@bib0415]). It is expected that PRRSV may antagonize other ISGs that restrict the virus entry, replication, and spread. The JAK-STAT signaling is modulated by host suppressive signals, such as suppressors of cytokine signaling (SOCS) proteins, ubiquitin carboxy-terminal hydrolase 18 (USP18), and miRNAs ([@bib0080], [@bib0180]). PRRSV upregulates the suppressive signals to antagonize JAK-STAT pathway. For example, PRRSV infection leads to upregulation of SOCS1 ([@bib0430]); PRRSV upregulates miRNA miR-30c against JAK1 ([@bib0465]). But overexpression of USP18 leads to a reduction of PRRSV replication ([@bib0005]). Further study needs to be done to determine the mechanism. The combined effect of multiple perturbations by PRRSV results in efficient PRRSV replication and invasion, and consequently, contributes to the poor induction of protective immune response. Minimizing the interference is expected to improve the host immune response to PRRSV infection. Indeed, PRRSV strain A2MC2, which induces type I IFNs in infected cells *in vitro* and has no inhibitory effect on IFN signaling ([@bib0260]), elicits higher virus-neutralizing antibodies than the MLV in pigs ([@bib0410]). Site-directed mutagenesis of R128 and R129 of nsp1β reduced its inhibition of IFN induction and leads to improvement of innate and adaptive immune responses ([@bib0220]). 3.2. PRRSV antagonizes JAK-STAT3 signaling pathway {#sec0055} -------------------------------------------------- As mentioned above, STAT3 has pleiotropic activity and plays important roles in many biological processes. STAT3 is a central regulator of lymphocyte differentiation and function ([@bib0190]). STAT3 is needed for activation and maturation of dendritic cells (DCs) and plasmacytoid DCs are considered to be the major source of IFN-α production during viral infection. STAT3 deficiency affects the generation of memory CD8^+^ T cells ([@bib0085], [@bib0335]) and memory B cells ([@bib0015], [@bib0095]). PRRSV infection induces a weak cell-mediated immune response, in which PRRSV-specific T cells transiently appears two weeks after infection without a change in frequencies of CD4^+^ and CD8^+^ T-cells ([@bib0435]). The STAT3 antagonizing may be one of the reasons for PRRSV interference with the development of protective immune response. PRRSV inhibits STAT3 signaling by inducing degradation of STAT3 but has minimum effect on STAT1 protein level ([@bib0455]). This reduction appears to be an intrinsic property of PRRSV as both Type 1 and 2 PRRSV strains are able to downregulate STAT3. PRRSV infection in pigs leads to elevation of IL-10 ([@bib0070], [@bib0360]) and induces lung lesions with inflammatory cell infiltration ([@bib0145]). IL-10 signaling via mediator STAT3 results in the generation of regulatory macrophages, which have an anti-inflammatory activity to dampen immunopathology. PRRSV antagonizing STAT3 signaling could interfere with the IL-10 regulatory function and leads to dysregulation of inflammation. PRRSV nsp5 is responsible for the STAT3 reduction. Nsp5 is a hydrophobic transmembrane protein and can possibly form a membranous structure in the cytoplasm that could be the site for PRRSV replication ([@bib0235]). No direct interaction between nsp5 and STAT3 is identified, which suggests that nsp5 might activate an E3 ligase of STAT3. Further study will reveal the mechanism of the induction of STAT3 degradation. 3.3. PRRSV's effect on STAT2/5/6 signaling {#sec0060} ------------------------------------------ As aforementioned, other than STAT1 and STAT3, other STAT proteins have also been shown to be significant for the host immune response. STAT2 is involved in STAT1-independent signaling by interacting with IRF9 to drive the expression of ISRE-containing genes ([@bib0035], [@bib0110]). As PRRSV reduces STAT2 protein, we expect it inhibits the STAT1-independent antiviral signaling. PRRSV's effect on the signaling of STAT4, STAT5 and STAT6 has not been determined yet. The main target cells for PRRSV infection are certain lineages of monocytes/macrophages *in vivo*. The PRRSV effect on JAK-STAT signaling in T cells would be indirect, such as by exosomes from infected cells ([@bib0130], [@bib0250]). But its interference of the JAK-STAT signaling in infected macrophages would have significant consequence as the response of PAMs against viral or bacterial pathogens is critical in determining the outcome of infection in the host. Cytokines like IFN-γ and GM-CSF activate JAK-STAT signaling to regulate macrophage phenotype and activation ([@bib0170]). Activated macrophages secrete immune regulatory cytokines including IL-1β, IL-6, IL-12, TNF-α and so on ([@bib0115]). STAT6 has been demonstrated to have JAK-independent antiviral signaling ([@bib0065]). PRRSV is expected to inhibit this alternative STAT6 signaling, which may contribute to the PRRSV evasion. Indeed, PRRSV reduces STAT6 protein level in infected MARC-145 cells (unpublished result). Further study will examine the mechanisms of the interference of STAT6 signaling. 4. Conclusion {#sec0065} ============= In conclusion, PRRSV perturbs JAK-STAT signaling pathways by disturbing STATs protein level and their nuclear translocation. PRRSV inhibits IFN-activated JAK-STAT signaling by blocking nuclear translocation of ISGF3 ([@bib0290], [@bib0405]). Recently, we discovered that PRRSV inhibits JAK-STAT3 signaling via inducing degradation of the STAT3 protein ([@bib0455]). We also noticed that STAT2 and STAT6 protein levels were much lower in PRRSV-infected cells than uninfected controls (unpublished results). PRRSV interaction with the JAK-STAT signaling pathways is complex and consequences would be possibly depending on the context of the milieu during infection. Despite substantial efforts to study and control PRRS, no production or vaccination regimen has demonstrated sustaining success ([@bib0310], [@bib0315]). This is likely in part due to the PRRSV poor induction of protective immune response, allowing for PRRSV replication, spread, and persistence in infected populations. Elucidation of the mechanisms of PRRSV evasion of JAK-STAT signaling would yield insightful information, which may facilitate the development of improved vaccines or therapeutics against PRRSV and other pathogens. Conflicts of interest {#sec0070} ===================== None. This study was partially funded by the National Pork Board and internal fund of the University of Maryland.

Introduction {#s1} ============ Evidence-based veterinary medicine (EVM) can be defined as 'the use of current best evidence in making clinical decisions' ([@R4]). Additionally, when making evidence-based decisions, the circumstances of the patient alongside the circumstances and values of the owner must also be taken into consideration ([@R3]). Although EVM was first mentioned in 1998 ([@R19]), it is less advanced than in the medical field in relation to the availability of synthesised evidence and the support available for the integration of evidence by clinicians into their practice ([@R10]). The first step in EVM is to identify relevant answerable questions ([@R29]), and veterinary clinicians have a crucial role in highlighting these ([@R25], [@R12]). By identifying what common species and conditions clinicians experience in practice, researchers can prioritise studies so that a large proportion of the profession will gain from future studies. To our knowledge, few published studies describe the entire veterinary population (including both practising and non-practising members) and what species and conditions practitioners commonly encounter. A comprehensive survey of veterinarians in the UK was conducted by the Royal College of Veterinary Surgeons (RCVS) in 2010 where it was reported that the species veterinary clinicians mostly worked with were dogs, cats, horses, cattle and rabbits ([@R23]). Another study by [@R17] found that cats were more commonly seen than dogs in small and mixed animal practice in The Netherlands. Conditions seen in practice in the United States were investigated by [@R18] who found that the most common clinical finding was dental calculus followed by gingivitis from 120,000 consultations in cats and dogs. [@R16] found that the majority of clinical time in equine practice was spent on lameness and reproduction in The Netherlands. The aim of this study was to describe the UK veterinary population, and what species and conditions veterinary clinicians think they commonly encounter in practice. A second aim was to gather data relating to how much information veterinary clinicians perceived was available for these species. Materials and methods {#s2} ===================== Population of interest {#s2a} ---------------------- The target population was all members of the veterinary profession within the UK. The sampling frame was the RCVS register of members. All veterinary surgeons legally practicing in the UK must be registered with the RCVS. This register incorporates individuals, including non-practicing and retired individuals, who have consented for their details to be made available to external organisations for research or marketing purposes. A questionnaire was used to collect data from individuals on this register. As a census of all individuals on the list was conducted, a sample size calculation was not carried out. Questionnaire structure {#s2b} ----------------------- Several methods were employed to increase response rates, including a mixed-mode survey design (utilising both paper-based and online methods) ([@R8], [@R26], [@R6]). The questionnaire was made up of 36 questions and had four main sections; a copy of the questionnaire is available on request. The questions in the first section concerned the collection of demographic information about respondents. The second section was made up of open questions requiring clinicians to nominate up to four species they most frequently encountered, and the three main conditions or complaints they thought they saw most commonly in those species with associated perceived information levels ([Fig 1](#VETREC2013101745F1){ref-type="fig"}). The other two sections are not discussed here and will appear in a separate manuscript. Questions were constructed using recommendations from several resources to optimise clarity, minimise ambiguity and to avoid leading terminology ([@R7], [@R13], [@R31], [@R14], [@R9], [@R28], [@R2], [@R6]). {#VETREC2013101745F1} Questionnaire development and distribution {#s2c} ------------------------------------------ Pretesting of the survey questions was carried out by researchers within the Centre for Evidence-based Veterinary Medicine (CEVM). Piloting of the survey was carried out three times (24 and 25 people, respectively, for paper version and once transferred to the online format, 8 people for online version) with a combination of private veterinarians, academic veterinarians, veterinary specialists and government veterinarians. Formatting of the questionnaire was carried out using TeleForm V.10.5.2 (Verity Inc. 2010), an automated content capture system. This programme enables scanning of completed questionnaires to facilitate entry of closed question data (open question data was manually entered) into a Microsoft Office Access V.14.0.6 (2010 Microsoft Corporation) database automatically. The software of Cvent (2011 Cvent Inc.), an online survey company, was used to construct the online version of the finalised paper questionnaire. The questionnaires were printed on magnolia coloured paper to make them easily identifiable against white paper. White envelopes were printed with the CEVM logo and the words 'THIS IS A SCIENTIFIC RESEARCH STUDY. THIS IS NOT JUNK MAIL, AN APPEAL FOR DONATIONS OR MARKET RESEARCH' to make it distinguishable from marketing mailings. A pen, chocolate and a return postage paid envelope were included and a prize incentive was offered (£500 towards the continuing professional development course/s of choice). If participants filled in the online version, they had an extra chance of winning £50 worth of department store vouchers. The RCVS mailing list was obtained in October 2010. An initial mailing was posted to all individuals on this list between 1st and 5th November 2010; a link to Cvent was included allowing participants to choose to complete either an online or paper version of the questionnaire. A first reminder was sent six weeks later to non-responders followed by a second copy of the questionnaire 10 weeks later for those still not responding. Data entry {#s2d} ---------- Returned paper-based questionnaires were scanned using Teleform, with the system set to check 10 per cent of questionnaires to enable the detection of scanning errors. Questionnaires were accepted from respondents until scanning was completed (November 2011); coding of the common conditions and complaints was completed in May 2012. Responses received electronically were downloaded into a Microsoft Excel V.14.0.6 (2010 Microsoft Corporation) document from Cvent and integrated into a Microsoft Access V.14.0.6 (2010 Microsoft Corporation) database with the paper responses. Data coding {#s2e} ----------- Data relating to the common conditions or complaints nominated by veterinary clinicians were classified according to species and type of condition. Classification definitions were primarily based on those created by N. J. [@R24], with some modifications for suitability across all species. Species were coded according to animal or production type (see online supplementary Appendix 1). The type of condition or complaint was coded according to the category it was most relevant to in relation to either body system (eg, musculoskeletal) or topic (eg, behaviour) (see online supplementary Appendix 2). This was further broken down to another level of classification which more specifically described the nature of the problem (see online supplementary Appendix 3), resulting in two levels of classification for each condition or complaint (eg, Musculoskeletal-ligament). Additionally, the condition or complaint was coded into a 'type' according to whether it was a disease, a clinical sign the animal might be presented for, or was deemed unclassifiable (see online supplementary Appendix 4). One researcher (MLB) coded all conditions. If conditions were unknown to the coder or required clarification, the online resource Merck Veterinary Manual ([@R20]) was used. A second veterinary resource (eg, textbook, online veterinary resource, colleagues, Google 2012) was used if the condition was not found in the first resource. A Microsoft Excel V.14.0.6 (2010 Microsoft Corporation) spreadsheet of coding was created to maintain consistency for the same complaints or conditions. At the end of the coding process, a second researcher (TDN) identified any discrepancies between similar conditions, and conferred with the first researcher (MLB). Data management and analysis {#s2f} ---------------------------- The dataset was transferred to a Microsoft Excel V.14.0.6 (2010 Microsoft Corporation) document for data management. Frequency tables and graphs were generated in Excel and RStudio (R Core Team 2011). A posthoc sample size analysis was performed using Raosoft ([www.raosoft.com/samplesize.html](www.raosoft.com/samplesize.html)). There was a high degree of correlation between observations for perceived information level within clinician and species. In order to account for this clustering, the median perceived information level within species for each veterinarian was calculated. A χ^2^ test (excluding 'don\'t know' observations) was then used to determine if perceived information level was different between species. The level of statistical significance was set at P\<0.05. Some questions were left unanswered by participants, therefore, the number of responses per question could be less than the total number of respondents; the number of respondents per question is identified where appropriate. This project received ethical approval from the ethics research committee at the School of Veterinary Medicine and Science at The University of Nottingham. Results {#s3} ======= Response rate {#s3a} ------------- Of the 14,532 questionnaires distributed, 5407 (37 per cent) were returned. Of these: 259 were return to sender, 230 were retired veterinarians, 72 were returned blank, 3 stated that the veterinarian was deceased and 1 was blank except for one comment box. Therefore, 4842 responses (33%; CI 32% to 35%) could be used in the analysis. Posthoc analysis revealed with a population size of 14,532, working on a margin of error of 2 per cent and a confidence level of 99 per cent (with a response distribution of 50 per cent), the sample size required in order to draw meaningful conclusions from the data was 3227. The majority of responses were returned in the form of paper-based questionnaires (3775/4842; 78 per cent). General respondent information {#s3b} ------------------------------ Of the respondents, 59 per cent (2856/4828) were female. The median age of all respondents was 37 years; for female respondents median age was 33 years (IQR 29--42), and for males it was 46 years (IQR 34--56). Approximately 78 per cent of respondents graduated in the UK (3759/4797; [Table 1](#VETREC2013101745TB1){ref-type="table"}) with 22 per cent (1038/4797) graduating abroad. Of the 1038 overseas graduates, the largest groups graduated in Ireland (154; 15 per cent) and Australia (114; 11 per cent). ###### ​Distribution of respondents graduating from UK vet schools and those training overseas (n=4797) Vet school Bristol Cambridge Edinburgh Glasgow Liverpool London Overseas ------------ --------- ----------- ----------- --------- ----------- -------- ---------- Number 582 499 626 583 608 861 1038 \% 12 10 13 12 13 18 22 In total, 1914 (40 per cent of 4835) of the respondents declared that they had one or more postgraduate qualification. The highest proportion (778; 16 per cent) had an academic degree (eg, BSc, MSc, PhD). Furthermore, 738 (15 per cent) had a RCVS certificate, a General Practitioner certificate or fellowship; 271 (6 per cent) had a Master of Business and 270 (6 per cent) had a diploma (either RCVS or European). The majority of respondents undertook clinical work (3982/4835; 82 per cent) with 76 per cent (3674/4835) of respondents working in private practice ([Table 2](#VETREC2013101745TB2){ref-type="table"}). Most veterinary clinicians reported they worked with small animals, with the second largest group working with small animals, equine and production animals ([Table 3](#VETREC2013101745TB3){ref-type="table"}). ###### ​Type of workplace for respondents (n=4835 respondents)\* Type of workplace Number Per cent ------------------------------------------------------------------------- -------- ---------- Private practice 3674 76 University practice or education 277 6 Government (including Veterinary Laboratories Agency and Animal Health) 248 5 Charity 221 5 Research (university or institute) 171 4 Career break 153 3 Other 147 3 Meat inspection 139 3 Outside profession 117 2 Industry (eg, pharmaceutical or feed company) 91 2 Pathology/clinical pathology laboratory 70 1 Army 15 \<1 \*Respondents could nominate more than one category ###### ​Type of animal practice undertaken by veterinary clinicians (n=3921 respondents\*) Species Number Per cent ------------------------------------------------------------------------------------------- -------- ---------- Small animal (including rabbits and exotics) 2266 58 Small animal and production animal† and equine 507 13 Equine 311 8 Small animal or production animal† or equine and laboratory animal or zoo animal or other 290 7 Production animal† 178 5 Small animal and production animal† 172 4 Small animal and equine 90 2 Equine and production animal† 64 2 Laboratory animal or zoo animal or other 43 1 \*Only 3921/3982 respondents who did clinical work stated what type of animal practice they undertook †Ruminants/pigs/poultry Common species and conditions or complaints {#s3c} ------------------------------------------- Overall, 36,504 conditions or complaints were mentioned by 3982 respondents undertaking clinical work. Conditions in dogs and cats were most frequently mentioned ([Fig 2](#VETREC2013101745F2){ref-type="fig"}). Skin was a commonly mentioned body system, as well as the gastrointestinal and musculoskeletal systems ([Table 4](#VETREC2013101745TB4){ref-type="table"}). ###### ​The seven most common species and the three main body systems or topics mentioned by veterinary clinicians performing clinical work (n=3982)\* Species Per cent of veterinary clinicians (3982) Number of conditions per species Body system/Topic Number of conditions per body system Per cent of conditions per species ------------ ------------------------------------------ ---------------------------------- ------------------- -------------------------------------- ------------------------------------ Dog 81 9606 Skin 3064 32 Gastrointestinal 2735 28 Musculoskeletal 1742 18 Cat 78 9251 Skin 2467 27 Non-specific 1272 14 Urinary 1123 12 Rabbit 54 6410 Dental 1918 30 Skin 1619 25 Gastrointestinal 974 15 Equine 21 2498 Musculoskeletal 899 36 Gastrointestinal 512 20 Respiratory 336 13 Cattle 21 2463 Reproduction 1085 44 Respiratory 379 15 Non-specific 344 14 Guinea pig 17 1974 Skin 743 38 Non-specific 365 18 Dental 346 18 Sheep 11 1228 Reproduction 360 29 Non-specific 282 23 Musculoskeletal 229 19 Non-specific body systems or topics related to conditions that were either too general to belong to one category (eg, nutrition problems) or could fit into more than one category (eg, bacterial infection) \*Veterinary clinicians were asked to mention up to four species that they worked with and three conditions for each species. See online supplementary Appendices 1, 2 and 3 for further details on condition classification {#VETREC2013101745F2} The most commonly mentioned body systems or topics for the most frequently mentioned species were analysed further ([Fig 3](#VETREC2013101745F3){ref-type="fig"}a,b). The 'Skin-skin' category for dogs and cats contained responses such as 'dermatitis' and 'pruritus'; for dogs it also commonly contained 'atopy', and for cats, 'cat bite abscess'. The 'Skin-non-specific' category for both species included responses such as 'skin disease', 'skin problems' or just 'skin'. The 'Dental-dental' category for rabbits commonly included the responses 'dental disease' and 'dental malocclusion', the 'Dental-non-specific' category included 'teeth' or 'dental problems' and 'Dental-oral' included 'oral' or 'mouth disease'. The 'Skin-non-specific' category in guinea pigs commonly contained responses such as 'skin disease', 'skin problems' and 'skin'. The 'Skin-skin' category in guinea pigs contained more specific responses such as 'pruritus', 'dermatitis' and 'abscesses'. For cattle, the response 'mastitis' relating to the category 'Reproduction-mammary' was the most commonly nominated for reproductive conditions. 'Reproduction-non-specific' in cattle included responses such as 'fertility problems' and 'infertility'; 'Reproduction-reproduction' contained complaints such as 'dystocia' and 'calvings'. In sheep, the category 'Reproduction-reproduction' incorporated responses such as 'lambing', 'dystocia', 'abortion' and 'twin lamb disease'. The 'Reproduction-non-specific' sheep category contained less specific terms such as 'obstetrics', 'fertility' and 'parturition problems'. For equines, most responses were classified into 'Musculoskeletal-musculoskeletal', and were exclusively related to 'lameness'; the 'Musculoskeletal-non-specific' category contained responses such as 'foot abscess' or hoof abscess\', 'orthopaedics' and 'back pain'. {#VETREC2013101745F3} Type of common condition or complaint {#s3d} ------------------------------------- Cats had a higher proportion of responses classified as 'Disease' (eg, hypothyroidism; 20 per cent), but a lower proportion of conditions classified as 'Clinical sign' (eg, weight loss) when compared with other species ([Fig 4](#VETREC2013101745F4){ref-type="fig"}). Sheep had the highest proportion of conditions classified as 'Clinical sign' (eg, lameness; 51 per cent) while equines had the highest proportion of observations in the 'Both' category (30 per cent), which relates to conditions that could be considered a clinical sign but are used as a disease description (eg, colic). A high proportion of responses regarding dental issues in rabbits and the respiratory system in equines and cattle were classified as 'Unclassifiable'. This category was for those conditions that could not be classified (eg, zoonoses) or if the terminology was too vague to be considered either a clinical sign or a specific disease (eg, production diseases). Overall, a higher proportion of conditions in rabbits (eg, skin; 53 per cent) and guinea pigs (eg, lumps; 57 per cent) were classified into the 'Unclassifiable' group than the average for all species (41 per cent). {#VETREC2013101745F4} The type of classification was compared with the three most common body system or topic groups for each species (using the categories in [Table 4](#VETREC2013101745TB4){ref-type="table"}). Very few conditions could be classified into the 'Disease' category. Musculoskeletal conditions in sheep (eg, lameness) were almost exclusively classified as 'Clinical sign' ([Fig 5](#VETREC2013101745F5){ref-type="fig"}). Dental conditions in guinea pigs (eg, teeth problems) and rabbits (eg, dental disease) and respiratory conditions in cattle (eg, respiratory) had the highest proportion of 'Unclassifiable' conditions. {#VETREC2013101745F5} Perceived information level for conditions or complaints {#s3e} -------------------------------------------------------- Generally, respondents thought that there was a lot of information about approximately 60 per cent of the nominated conditions for cattle, equines and dogs, in contrast with rabbits and guinea pigs at 18 per cent and 5 per cent, respectively ([Fig 6](#VETREC2013101745F6){ref-type="fig"}). After investigating if the perceived information level was similar for the different species, there was significantly less perceived information available for guinea pigs and rabbits, and more for dogs, cattle and equines, and to some extent, cats (P\<0.0001; data not shown). {#VETREC2013101745F6} Discussion {#s4} ========== Our findings of the species and conditions reported to be commonly seen in veterinary practice were similar to previous reports ([@R17], [@R18], [@R11], N. J. [@R24]) as dental conditions in rabbits and guinea pigs and skin conditions in small animal practice were highlighted in these studies. However, some studies have shown that the most common presentation in small animal practice is for preventive medicine ([@R11]) which was not found in the current study. This difference could be due to the fact that preventive medicine may not be seen by veterinary clinicians as a condition or complaint, which was how these questions were phrased. Reproductive conditions, particularly mastitis in cattle, were most frequently mentioned for sheep and cattle, which is similar to results found from a Swedish study using information from a cattle database (Mörk and others 2009). Conditions relating to the musculoskeletal body system in equines have also been commonly found in a study by [@R16]. The minor differences in species and conditions in this study compared to earlier studies, could be due to different data collection methods. Results in this study are based on which species and conditions or complaints veterinary clinicians nominated as seeing commonly in practice. This could result in conditions recently encountered or those that are difficult to deal with being reported rather than what is actually seen commonly. However, the previous studies highlight similar results, suggesting that clinicians are likely to be acutely aware of the caseload they see, and therefore, are well placed to highlight areas for further research that would be of benefit to the profession. Classification of common conditions with regards to disease versus clinical sign appeared to be species and body system dependent. It is unclear why this is the case, and requires further investigation; it is possible this could be partially explained by the variability in how different types of client present their animals (eg, pet owner versus farmer). Species and body systems with high percentages of conditions in the 'unclassifiable' category could represent a number of things. It could be a reflection of the complexities of certain conditions, for example, respiratory disease complex in cattle, or could be an indication of a lack of evidence behind certain conditions in certain species, for example, dental disease in guinea pigs. It could also identify areas where least specific terminology is being used and may indicate areas of uncertainty for clinicians, potentially highlighting areas where further research is required. The perception by individual veterinary clinicians of the available levels of information available for the nominated common conditions was found to be similar within species. This suggests that some vets may be more aware of the existing evidence than others. Overall, there was perceived to be a lower amount of information available for guinea pigs and rabbits compared with other species. This could be due to a lack of access to information, lack of familiarity with the information available, or an actual lack of published information about these species. Further work is needed to quantify the amount and quality of information available for each species which is accessible to veterinarians to determine the reason for this result. Study limitations {#s4a} ----------------- As veterinarians can opt out of being contacted by third-party organisations, our sample did not include all RCVS registered veterinarians. It is unknown whether the non-responders possess particular ­characteristics which are different to the responders (non-responders were not followed up due to feasibility restrictions). Additional information would need to be gathered to assess the bias that may have occurred by the self-selection of respondents (eg, individuals interested in EVM may be more likely to reply) ([@R27], [@R30], [@R15]). However, responses were received from individuals from a variety of different age groups and occupations, and the distribution of work places and proportion of women (once retirees were excluded) was similar to that found by the RCVS ([@R23]). Similar response rates have been reported in other studies ([@R1], [@R5], [@R32]), although it has been recommended that care be taken when interpreting results with response rates less than 70 per cent ([@R28]). The posthoc calculation (and CI) indicated that the sample size obtained was likely to be adequate to draw meaningful conclusions about the data; however this calculation may not be appropriate for all questions. Clinicians were asked what percentage of their working time was spent on 1st opinion, 2nd opinion and referral cases. Because of the design of the question, it was not possible to distinguish clearly between clinicians working with 2nd opinion and referral cases. However, very few respondents stated that they spent a considerable percentage of their working time with these types of cases, therefore, the vast majority of responses were from 1st opinion clinicians. We requested that participants only nominated up to four species they most frequently saw, and three conditions per species. It is possible that if more than this number had been requested the results may have appeared differently though it is likely that the main species and conditions have been captured. The way the nominated conditions were coded could have influenced the results found here. However, attempts were made to improve consistency and repeatability as outlined. No attempt was made to define what was meant by 'published veterinary information' which could also have resulted in varying interpretations by respondents, particularly as evidence quality was not assessed as part of this study. Conclusion {#s5} ========== This study highlights specific areas in which research could be pertinent for veterinarians in the UK; skin conditions were mentioned frequently in small animals, and reproductive conditions and musculoskeletal conditions in cattle and sheep, and equines, respectively. There is a perception that little information exists for certain species; further research is required to identify whether this information exists and if it is accessible to veterinarians to aid decision making in practice. Supplementary Material ====================== ###### Web supplement ###### Web supplement ###### Web supplement ###### Web supplement **Correction notice:** This article has been corrected since it was published Online First. The first sentence of the discussion was amended for clarity. **Acknowledgements:** The authors wish to thank the veterinarians who participated in the survey. The Centre for Evidence-based Veterinary Medicine is supported by an unrestricted grant from The University of Nottingham and Novartis Animal Health. [^1]: Provenance: Not commissioned; externally peer reviewed

**Zhang X, Bao S, Lai D, Rapkins RW, Gillies MC. Intravitreal triamcinolone acetonide inhibits breakdown of the blood-retinal barrier through differential regulation of VEGF-A and its receptors in early diabetic rat retinas. Diabetes 2008;57:1026--1033** In the print version of the article listed above, the second affiliation for Xinyuan Zhang is incorrect. The correct affiliation is as follows: Beijing Institute of Ophthalmology, Beijing Tongren Eye Center, Beijing Tongren Hospital, Capital Medical University, Beijing, China. The online version reflects these changes.

1. Introduction {#s0005} =============== Interactions of integral membrane proteins with their surrounding lipid environment play a key role in stabilising their structure and influencing their activity. To obtain insight into the nature and type of interactions occurring between lipids and integral membrane proteins a range of biophysical techniques including electron spin resonance [@bb0005; @bb0010] and fluorescence [@bb0145; @bb0020] spectroscopy have been used in numerous studies. This has provided a wealth of information on the specificity of proteins for particular classes of lipids and the affinity of these interactions [@bb0025; @bb0030]. These studies have revealed that in addition to the 'bulk' lipids whose dynamic properties remain largely unchanged by the presence of integral membrane proteins within the bilayer, there exist a population of lipids whose motional freedom is constrained through their interaction with integral membrane proteins. The motionally restricted population of lipids can be segregated into 'annular lipids' which exhibit low affinity interactions with the hydrophobic surface of membrane proteins and 'non-annular lipids' which exert high affinity interactions at sites located in clefts on the protein surface or at the interface between protein subunits [@bb0025; @bb0035; @bb0040; @bb0045]. The interactions at both sites play an important role in modulating the function of integral membrane proteins, but it has recently become apparent that occupation of the non-annular binding site can be particularly critical for function [@bb0050]. Despite the importance of these interactions a detailed atomic level description of the type and nature of the interplay between lipids and integral membrane proteins is limited as the lipids are often missing from high-resolution crystal structures of membrane proteins. One of the best-studied ion channels is the potassium channel, KcsA from *Streptomyces lividans*. This potassium channel, in common with other members of this family, has an absolute requirement for anionic lipids such as phosphatidylglycerol, phosphatidylserine or cardiolipin for function; in their absence the channel exists in a non-conducting state [@bb0055; @bb0060]. The binding of lipids to KcsA has been extensively investigated by fluorescence quenching studies that clearly identify annular and non-annular populations of lipids [@bb0035; @bb0050]. Binding of lipids to the annular sites revealed marked differences between the inner and extracellular leaflets, with the extracellular side showing similar affinities for anionic and zwitterionic lipids. In contrast the intracellular side showed a twofold higher affinity for anionic lipids over phosphatidylcholine, presumably due to the clustering of charged residues on KcsA close to the bilayer surface. Fluorescence quenching studies have revealed that the non-annular binding sites show a high degree of selectivity of binding anionic lipids almost exclusively, albeit with moderate affinity. The crystal structure of KcsA in detergent has provided valuable insights into the interaction of lipids with the non-annular binding site with electron density being seen at the interface between the protein subunits indicating the presence of a diacylglycerol-like moiety. The crystal structure revealed that the *sn*-1 chain of this lipid molecule was tightly buried in the groove between the pore helix and the M2 helix whilst the *sn*-2 chain was less intimately associated with the protein [@bb0055; @bb0065]. Subsequent studies have revealed the lipid present in the KcsA crystals to be phosphatidylglycerol which co-purifies with the KcsA [@bb0055] (pdb accession number 1K4C), suggesting that the phosphatidylglycerol headgroup exhibits significant motion or disorder within the crystal, resulting in the absence of electron density. The absence of electron density in the region corresponding to the lipid headgroup has precluded a detailed understanding of how the phosphatidylglycerol is recognised. The presence of two key arginine residues (R64 and R89) in close proximity to the proposed headgroup region suggested a putative role for electrostatic interactions between the sidechains and the anionic lipids within the binding site [@bb0055]. The proposed interactions between R64 and R89 have been investigated by molecular dynamics, revealing the formation of H-bonds between the headgroups of the anionic lipids phosphatidic acid and phosphatidylglycerol and the arginine residues [@bb0070]. Notably, these interactions were absent or reduced in bilayers containing the zwitterionic lipid phosphatidylethanolamine [@bb0070]. To demonstrate the feasibility of detecting interactions between non-annular binding sites and anionic lipids we have undertaken a 31P magic-angle spinning (MAS) NMR study of KcsA reconstituted into a model lipid bilayer composed of phosphatidylcholine and the negatively charged phosphatidylglycerol. The application of MAS permits the acquisition of ^31^P NMR spectra in which the individual lipid components are resolved on the basis of their chemical shift [@bb0075]. The chemical shift observed provides information on the local electrostatic environment of the phosphate moiety of the lipid headgroup and is thus an excellent reporter on the interaction of the lipid headgroup with the protein [@bb0080; @bb0085; @bb0090]. Typically the application of NMR to study lipid/protein interactions has proved challenging as exchange between the bulk lipid and annular sites occurs too rapidly on the NMR timescale to permit the observation of the bound lipids. Furthermore, interactions between the lipids and the proteins are typically hydrophobic in nature with little difference in electrostatic environment between the bound and free lipids resulting in only minor perturbations in chemical shift. In contrast, phosphatidylglycerol bound at the non-annular binding site is predicted to experience a significantly different electrostatic environment from the annular/bulk lipids with higher-affinity interactions resulting in longer residency times within the binding site. Below we demonstrate that this results in a spectroscopically distinct species that we resolve in the ^31^P MAS-NMR spectrum of KcsA reconstituted into lipid vesicles, providing insights into the interactions involved in lipid binding. Using site directed mutagenesis we have demonstrated that the perturbation in electrostatic environment arises through the interaction of the lipid head group with the positively charged sidechain of R64 and R89. Single-channel current recordings have been measured to ascertain the role that these residues play in determining the channel gating behaviour of KcsA. 2. Materials and methods {#s0010} ======================== 2.1. Materials {#s0015} -------------- Palmitoyloleoyl-phosphatidylcholine (POPC) and palmitoyloleoyl-phosphatidylglycerol (POPG) were purchased from Avanti Polar Lipids (Alabaster, AL). The pQE32 vector and M15\[PREP\] *Escherichia coli* strain were bought from Qiagen (UK). The detergent, dodecylmaltoside (DDM) was from Anatrace (UK). The other reagents for the purification were obtained from Sigma (UK). 2.2. Cloning and mutagenesis of KcsA {#s0030} ------------------------------------ The pQE32 vector containing the KcsA gene with a hexahistidine epitope at the N-terminus, kindly donated by Professor Lee (University of Southampton, UK), was expressed in M15 cells. Three KcsA mutants were generated by site-directed mutagenesis using the Quik-change protocol from Stratagene (La Jolla, CA). Three KcsA mutants were prepared replacing the arginine with the slightly smaller but uncharged leucine at residue 64 (R64L), 89 (R89L) or at both sites (R64,89L). The mutants were generated by PCR using synthetic oligonucleotide primers containing the desired mutations (Eurofins MWG, UK). Complementary oligonuceotides, 5′-AGCTGATCAC GTATCCGTTA GCGCTGTGGT GGTCC-3′ and 5′-GACCACCACA GCGCTAACGG ATACGTGATC AGCTG-3′ were used as forward and reverse primers respectively to create the R64L mutation. Similarly, R89L was produced using the synthetic oligonucleotides 5′-GTGACTCTGT GGGGCCTGC TCGTGGCCG TGGTGGTGA T-3′ and 5′-ATCACCACCA CGGCCACGA GCAGGCCCC ACAGAGTCA C-3′ as primers. Polymerase chain reaction was used to generate the mutants and the resultant PCR product was Dpn1- treated for 1 hr at 37 °C to digest any methylated parental DNA. The DNA was used to transform competent M15 \[PREP\] *E.coli* cells that were then plated onto agar plates supplemented with ampicillin. Mutations were confirmed by sequencing. 2.3. Over expression and purification of KcsA {#s0035} --------------------------------------------- M15 *E.coli* cells transformed with KcsA or one of the KcsA mutants were used to inoculate 10 mL of Luria Broth (LB) medium containing 100 μg/mL of ampicillin. The overnight culture was then used to inoculate 1 L of LB containing 100 μg/mL of ampicillin and grown to an OD~600~ of 0.8 at 37 °C. Over-expression of KcsA wild type and mutant protein was induced by the addition of IPTG to a final concentration of 1 mM and the culture was grown for a further 4 h at 37 °C. The cells were harvested at 4 °C by centrifugation at 12,000 *g* for 20 min. The cell pellet was resuspended in buffer A (50 mM Tris, 150 mM NaCl, 150 mM KCl, pH 7.4) with 1 mM PMSF and sonicated on ice for 5 min: 15 s on; 20 s off at power level 7 (Misonix sonicator). The membrane fraction was clarified by ultracentrifugation at 420,000 *g* for 40 min at 5 °C. The membrane-containing pellet was homogenised and then gently stirred in solubilisation buffer (buffer A, 40 mM imidazole, 1 mM DDM, pH 7.4) for 1 h at room temperature. Insoluble material was removed by centrifugation at 21,000 *g* for 20 min at 5 °C and the supernatant was loaded onto a Ni^2+^ affinity column (GE Healthcare, UK) pre-equilibrated with buffer A. The column was washed with 20 bed volumes of wash buffer (Buffer A, 40 mM imidazole, 1 mM DDM, and pH 7.4) and KcsA was eluted in Buffer A, 400 mM imidazole, 1 mM DDM, pH 7.4). 2.4. Reconstitution of KcsA {#s0040} --------------------------- KcsA wild type and mutants were reconstituted in lipid vesicles composed of POPC and POPG at a molar ratio of 70:30 respectively. Briefly, 10 mg of lipids in chloroform were dried in a vacuum desiccator and resuspended in hydration buffer (10 mM Tris, 100 mM KCl, 1 mM EDTA pH 7.4) containing 40 mM octyl glucoside. Following mixing and sonication to clarity, KcsA protein was gently added and mixed to give a lipid to KcsA tetramer molar ratio of 100:1. The vesicles were generated by detergent removal using Bio-beads. For NMR studies the vesicles were pelleted by ultracentrifugation at 100,000 *g* for 30 min at 5 °C and loaded into an MAS rotor for analysis by NMR. 2.5. Static and magic angle spinning solid-state ^31^P NMR studies {#s0045} ------------------------------------------------------------------ All ^31^P NMR measurements were performed on a 300 MHz Infinity+ spectrometer (Varian, USA) at a 121.37 MHz using a 4 mm triple resonance MAS probe (Varian, USA). ^31^P NMR spectra were acquired with a 5 μs π/2 pulse for excitation, 70 kHz continuous wave proton decoupling during acquisition and a recycle delay of 3 s. All spectra were obtained at 25 °C with a magic angle spinning frequency of 6 kHz. All spectra were externally referenced to phosphoric acid (85%) with a chemical shift of 0 ppm. Typically magic-angle spinning (MAS) and static NMR spectra were obtained by averaging 4096 transients or 8192 transients respectively and 5 Hz linebroadening was added prior to Fourier Transform. Data processing and analysis were performed in matNMR [@bb0095]. 2.6. Single channel electrophysiology of wild-type, R64L, R89L and R64L/R89L in planar lipid bilayers {#s0050} ----------------------------------------------------------------------------------------------------- Single-channel current recordings were acquired as detailed previously by Marius et al.[@bb0050]. Briefly, POPC/POPG at a molar ratio of 70 to 30 was first dried from chloroform and then dissolved in decane to a concentration of 20 mg/mL. A lipid planar bilayer was painted across a 150 μm aperture in a Delrin cuvette (Warner Instruments, CT, USA) which separated two chambers with a volume of 1 mL. The *cis* (extracellular) chamber contained buffer B (10 mM Hepes, 150 mM KCl) at pH 7.0 whilst the *trans* buffer contained buffer B at pH 4.0. Bilayer formation was verified with capacitance measurements. The reconstituted vesicle suspension (5 μL) was added to the *cis* chamber and the transmembrane current was measured using Ag\|AgCl electrodes. The electrode in the *cis* chamber was connected to the input of the headstage of an ID562 bilayer amplifier (Industrial Developments Bangor, Bangor, UK), whilst the bias voltage was applied to the *trans* chamber, as described fully in Marius et al. [@bb0050]. The bilayer was voltage-clamped at + 100 mV and electrical recordings were digitised at 5 kHz, and subsequently digitally low-pass filtered at 1 kHz for analysis and at 2 kHz for display. Current traces were analysed using Origin (OriginLab, US). Episodes of rapid channel gating, with a duration of 15 s or more, were used for analysis of the channel open probability. 3. Results and discussion {#s0020} ========================= 3.1. Expression, purification and reconstitution of wild type and mutant KcsA {#s0055} ----------------------------------------------------------------------------- The generation of the KcsA mutants, R64L, R89L and R64, 89L was confirmed by sequencing and mass spectrometry (results not shown). The overexpression of wild type and mutants of KcsA as His-tag fusion protein in M15 cells yielded cell pellets of 3 ± 1 g/L culture. The extraction of the protein from cells involved solubilisation of the membrane fraction using the non-ionic detergent DDM, followed by Ni-NTA affinity chromatography. Increasing the imidazole concentration from 40 mM to 400 mM resulted in the displacement of bound protein. The purity of the proteins was determined by SDS-PAGE ([Fig. 1](#f0010){ref-type="fig"}) with the wild type and three mutants running as a stable SDS resistant tetramer. The yield of pure wild type and single mutant KcsA protein ranged from 5 mg to 10 mg/L whilst the double mutant, R64, 89L, had a yield of 4 ± 2 mg/L. Wild type and mutants of KcsA were prepared for solid-state NMR studies by reconstitution into lipid vesicles as described. The proteins were incorporated into pre-formed lipid-detergent micelles. The addition of Bio-beads for the removal of detergent led to the formation of a cloudy suspension of proteoliposomes that were pelleted for subsequent analysis by NMR. 3.2. Phosphorous static and MAS NMR of wild type-KcsA reconstituted into PC/PG vesicles {#s0060} --------------------------------------------------------------------------------------- To ascertain the effect of KcsA on POPC and POPC/POPG (70%/30% mol/mol) vesicles ^31^P solid-state NMR spectra were recorded in the presence and absence of KcsA ([Fig. 2](#f0015){ref-type="fig"}). To confirm that the KcsA spectra had been successfully reconstituted into lipid bilayers and that this had not altered the phase behaviour of the lipids, static ^31^P proton decouple NMR spectra were acquired ([Fig. 2](#f0015){ref-type="fig"}C and D). In each case the spectra showed the axially symmetric powder pattern expected for POPC and POPC/POPG bilayers in the L~α~-phase, with no evidence of smaller isotropic micellar contributions. Upon the introduction of acidic POPG into the vesicles a significant reduction in the chemical shielding anisotropy was observed, falling from − 31.9 ppm in pure POPC vesicles to − 21.3 ppm in vesicles composed of 70% POPC/30% POPG (mol/mol). Such a reduction has previously been observed and attributed to the change in the surface charge of the bilayer [@bb0085; @bb0090]. Comparison of the static spectra of pure lipid vesicles and those containing KcsA reveals a similar reduction in chemical shielding anisotropy in both samples indicating that the introduction of KcsA did not significantly perturb the dynamics observed in the bilayer and had no effect on the overall charge on the bilayer surface [@bb0085; @bb0090]. The MAS spectra of POPC and POPC/POPG (70%/30% mol/mol) of lipid vesicles in the absence of KcsA are shown in [Fig. 2](#f0015){ref-type="fig"}A and B (black). The spectra of POPC alone show a single resonance at − 0.8 ppm, whilst the mixture of POPC/POPG shows two well-resolved resonances at − 0.79 ppm and 0.30 ppm, which are assigned to POPC and POPG respectively on the basis of their relative intensities and earlier studies [@bb0085; @bb0100]. The ^31^P MAS-NMR spectrum of KcsA reconstituted into POPC bilayers ([Fig. 2](#f0015){ref-type="fig"}A, red) also shows a single resonance at − 0.81 ppm although the linewidth is broader than that for lipid alone, increasing from 0.34 ppm to 0.66 ppm. Such an observation is consistent with earlier studies that also detected an increase in linewidths upon the reconstitution of proteins into lipid bilayers, a behaviour that was attributed to the overall reduction of mobility of the lipids within the bilayer arising from the low lipid to protein ratio. The ^31^P MAS-NMR spectrum of KcsA reconstituted into negatively charged POPC/POPG vesicles (70%/30% mol/mol) is shown in [Fig. 2](#f0015){ref-type="fig"}B (red) and deconvoluted into its components in [Fig. 3](#f0020){ref-type="fig"}. As expected, two resonances are observed at − 0.75 ppm and 0.23 ppm, but in addition a resonance is also present at − 0.14 ppm. The two larger components at − 0.75 ppm and 0.23 ppm, we assign to the POPC and the POPG present in the bulk lipids respectively. Both resonances show significant broadening compared to the pure lipid vesicles with the linewidths of POPC and POPG increasing from 0.15 ppm to 0.61 and 0.80 ppm respectively. We attribute the presence of the third resonance at − 0.14 ppm to a population of lipid that is bound to the KcsA and therefore observes a different electrostatic environment to that seen by the bulk lipids. Deconvolution of the three spectral components ([Fig. 3](#f0020){ref-type="fig"}) indicates that this third component comprises 6.6% of the overall intensity in the ^31^P NMR spectrum and is accompanied by a slight reduction in bulk POPG intensity. On the basis of this reduction in intensity and the known presence of a POPG binding site, we assign this resonance to POPG bound at the non-annular binding site on KcsA. The observed intensity (6.6% of total phosphorous signal) for this bound component is in good agreement with the predicted intensity, which we expect to be maximally 4% of the total intensity when all four non-annular binding sites are occupied with phospholipid and the lipid/KcsA tetramer ratio is 100:1. We note however that with a POPC/POPG ratio of 70%/30% mol/mol some of the binding sites may be occupied with POPC. The appearance of this third resonance and its relatively narrow linewidth (0.21 ppm) when compared to the bulk lipids indicates that there is little exchange on the NMR timescale (microseconds) between the non-annular binding site and the annular/bulk lipid which would otherwise lead to a broadening of the resonance. Comparison of the direct excitation spectra ([Fig. 2](#f0015){ref-type="fig"}A/B) with ^1^H-^31^P cross-polarisation (data not shown) showed no significant differences in the relative intensities of the resonances, in agreement with crystallographic studies, indicating that the phosphate does not undergo a significant reduction in mobility upon binding to the non-annular binding site. The perturbation of the POPG chemical shift upon binding to the KcsA (− 0.37 ppm) indicates that upon binding the phosphate in the lipid headgroup experiences a different local electrostatic environment arising through either changes in local headgroup geometry or local environment. Although a number of factors contribute to the overall chemical shift the upfield perturbation observed is consistent with the interaction of the phosphate group with positively charged residues within the non-annular binding site. Similar upfield perturbations have previously been observed upon the binding of negatively charged lipid species to positively charged binding sites on peripheral membrane proteins [@bb0080; @bb0105]. 3.3. Phosphorous NMR of R64L, R89L and R64L/R89L KcsA mutants reconstituted into PC/PG vesicles {#s0065} ----------------------------------------------------------------------------------------------- On the basis of the crystal structure [@bb0055] and molecular dynamics simulations [@bb0070] of KcsA it had been proposed that R64 and R89 might contribute positive charge to the non-annular lipid binding site ([Fig. 6](#f0035){ref-type="fig"}). To ascertain the role of residues R64 and R89 in the binding of POPG to the non-annular binding site, we mutated these residues from arginine to leucine giving three different mutant proteins, R64L, R89L and the double mutant R64L/R89L. The ^31^P proton decoupled MAS-NMR spectra of R64L, R89L and the double mutant R64L/R89L are shown in [Fig. 4](#f0025){ref-type="fig"}. The spectra of R64L, R89L and the double mutant R64L/R89L all show two distinct resonances at − 0.75 and 0.23 ppm, corresponding to the POPC and POPG in the bulk lipid respectively. Notably the resonance at − 0.14 ppm that we have attributed to the POPG bound to the non-annular binding site is absent in the spectra of the three mutants. The disappearance of this resonance indicates that the population of lipids previously identified within the non-annular binding site no longer experiences a significant change in electrostatic environment upon binding, either through absence of the positive charges in the binding site or the abolition of lipid binding to these sites. This demonstrates experimentally the observations made in earlier molecular dynamics simulations [@bb0070] that indicate both R64 and R89 residues are in close proximity to the phosphate headgroup of the POPG within the non-annular binding site and are implicated in binding either through electrostatic or hydrogen bonding interactions. 3.4. Single-channel current recordings of wild-type and mutant KcsA {#s0070} ------------------------------------------------------------------- To ascertain whether interactions between these positive charge sidechains in the non-annular lipid-binding site and the lipid headgroup could affect channel activity, single-channel current recordings were taken of KcsA wild type and mutants reconstituted in a planar bilayer composed of 70% POPC and 30% POPG ([Fig. 5](#f0030){ref-type="fig"}). The bilayer was painted across a 150 μm aperture separating a *cis* chamber, equivalent to the extracellular side of the KcsA channel containing pH 7 buffer, from the *trans* chamber (pH 4) representing the intracellular or cytoplasmic side of the channel. In this configuration, KcsA channels oriented with their protonation sites on the *trans* side will be functional [@bb0110]. At 100 mV holding potential, pure-lipid bilayer traces displayed a peak-to-peak noise of \~ 2--4 pA (data not shown). In the presence of wild-type KcsA, the single-channel traces ([Fig. 5](#f0030){ref-type="fig"}A) revealed short bursts of activity followed by long periods of inactivity leading to low overall open probability as reported previously [@bb0050; @bb0115]. Histogram analysis of the activity bursts gave an open probability for the wild-type KcsA of 4.8% with open current amplitudes distributed around 6 pA in agreement with earlier studies [@bb0050; @bb0115]. The single-channel recordings of the R64L and R89L mutants ([Fig. 5](#f0030){ref-type="fig"}B and C respectively) displayed fewer channel openings, however bursts of channel activity were still observed. Using a current amplitude of 4 pA as a threshold, an open probability (P~0~) of 1.6% and 2.0% was measured for the R64L and R89L mutants respectively. In contrast to the wild-type behaviour however, a distribution of current amplitudes is observed ranging from \< 4 pA (closed channels) up to a maximum of 10 pA in both cases. We are unable from this data to determine whether the apparent disappearance of a well-defined conductance state is due to channel opening events that are occurring on timescales faster or comparable to our sampling rates or due to an inherent property of the mutant channels, e.g. the existence of additional sub-conductance states. Analysis of the single channel recordings of the R64L/R89L double mutant ([Fig. 5](#f0030){ref-type="fig"}D) showed a further decrease in channel opening, with an open probability (P~0~) of 0.8%. As for the single mutants discrete conductance states were not observed. Despite the open probability of the wild type KcsA being low, the single-channel current recordings obtained for the R64L, R89L and R64L/R89L double mutant clearly indicate that the replacement of the charged arginine residues at position 64 and 89 with leucine alters the gating of KcsA, reducing both its open probability and conductance. Care must be taken in the interpretation of the electrophysiology recordings as lipid protein interactions have been reported to play a significant role in the folding and stability of the tetrameric channel which would clearly have implication on the conductance behaviour [@bb0140]. However, the appearance of stable tetramers on the SDS-PAGE of the wild-type and mutants ([Fig. 1](#f0005){ref-type="fig"}) suggests that the mutant proteins are both correctly folded and assembled into stable tetramer and thus the changes observed are a result in different gating behaviour. Several patch-clamping studies have previously been reported on the R64A mutant of KcsA reconstituted by mixing bacterial membranes with asolectin vesicles [@bb0120; @bb0125]. There are significant differences in the responses of the R64A mutation between these studies with Perozo and co-workers reporting a significantly higher open probability of 65% [@bb0120] whilst Kremer *et al.* find that the R64A mutant has a higher tendency to close than wild type KcsA [@bb0125]. Interestingly, channel-gating behaviour was also reported to be dependent on the nature of the substitution with a R64D mutation giving rise to a higher open probability than R64A, which Kremer et al. suggested to arise from the formation of a salt bridge across the non-annular lipid binding site which may stabilise the open state thereby mimicking the presence of an anionic lipid [@bb0125]. The low open probabilities, compared to wild type KcsA, that we observe in the R64L, R89L and R64L/R89L double mutant would be consistent with such a hypothesis, suggesting that the bulky, uncharged sidechains are unable to form the necessary interactions necessary to stabilise the open state. 4. Conclusion {#s0025} ============= Utilising ^31^P proton-decoupled MAS-NMR we have studied the interaction between KcsA and its surrounding lipid bilayer. No specific interaction is observed between KcsA and the zwitterionic POPC membrane constituent, however upon reconstitution into a lipid bilayer composed of POPC/POPG (70%/30% mol/mol), we observe the appearance of a distinct spectral component at − 0.14 ppm that we have assigned to the binding of POPG to the non-annular lipid-binding site on KcsA. The upfield perturbation in chemical shift observed upon the binding of POPG to the non-annular lipid-binding site on KcsA is consistent with the interaction of the anionic lipid with positive charges within the binding site. Based on earlier computational studies and crystallographic studies that proposed a role for R64 and R89 in the recognition of the negatively charged sites in the lipid headgroup we mutated R64 and R89 to a slightly smaller leucine residue. Subsequent ^31^P proton-decoupled MAS-NMR studies of the R64L, R89L and the R64L/R89L double mutant showed the disappearance of the resonance attributed to the bound POPG at − 0.14 ppm. Single channel recordings have revealed that the positive charges on the side chains of R64 and R89 act synergistically to further reduce the low channel open probability observed for KcsA under these conditions. Finally we note that the slow exchange between the lipids at the non-annular binding site compared to the faster exchange observed for annular lipids results in the presence of a distinct spectral component as opposed to broadening. This observation demonstrates the feasibility of observing lipids whilst bound to non-annular binding sites in other, potentially less well characterised systems, providing an alternative biophysical technique to study how lipids and other lipophilic molecules may interact with these sites and the role this plays in regulating the function of integral membrane proteins. Furthermore, the detection of non-annular lipids by NMR paves the way to more complex magic-angle spinning experiments which will enable the accurate structural characterisation of the lipid within its binding site and the interactions responsible for its binding. PTFW and PM are funded by a Wellcome Trust Career Development Fellowship to PTFW. Prof. M.H. Levitt is kindly thanked for providing support and instrument time to conduct these measurements. We would like to acknowledge Prof. G. Gröbner and Prof. A.G. Lee for many interesting discussions. {#f0010} {#f0015} {#f0020} {#f0025} {#f0030} ![Crystal structure of wild type KcsA showing the location of the resolved DAG buried at the interface between the two subunits. Residues R64 and R89 are both in close proximity to the glycerol backbone of the DAG and well positioned to interact with the negatively charged phosphate group of the anionic lipids which although not resolved in this structure are known to occupy the non-annular binding site. Figure generated from pdb file 1K4C.pdb [@bb0130] and visualised in Chimera [@bb0135].](gr6){#f0035}

(J Am Heart Assoc. 2017;6:e007026 DOI: 10.1161/JAHA.117.007026.)29042422 Clinical PerspectiveWhat Is New?In this nonrandomized, retrospective, observational study of patients undergoing cardiac resynchronization therapy (CRT) with quadripolar (QUAD) and non‐QUAD left ventricular leads, programmed to biventricular, single‐site left ventricular pacing, QUAD was associated with a lower total mortality, cardiac mortality, and heart failure hospitalization.These benefits were observed after both CRT‐defibrillation and CRT‐pacing, after adjustment for heart failure etiology.Re‐interventions for left ventricular displacement or phrenic nerve stimulation, which were lower with QUAD, were associated with worse outcomes.What Are the Clinical Implications?The markedly better outcomes after CRT observed with QUAD supports their preferential use over non‐QUAD in clinical practice.The relative benefits of CRT‐defibrillation over CRT‐pacing requires further evaluation in the QUAD era. Introduction {#jah32587-sec-0008} ============ Cardiac resynchronization therapy (CRT), with CRT‐defibrillation (CRT‐D) or without (CRT‐pacing \[CRT‐P\]) defibrillation, is a standard treatment for selected patients with heart failure (HF) with severe left ventricular (LV) dysfunction and a wide QRS complex.[1](#jah32587-bib-0001){ref-type="ref"} Since the first transvenous CRT implantations were undertaken in the 1990s,[2](#jah32587-bib-0002){ref-type="ref"}, [3](#jah32587-bib-0003){ref-type="ref"} improvements in delivery catheter and LV lead design, as well as implantation techniques using venoplasty and snaring, have helped to improve implantation success. Prominent among the challenges still encountered at implantation and thereafter is achieving acceptable LV pacing thresholds without phrenic nerve stimulation (PNS).[4](#jah32587-bib-0004){ref-type="ref"} Deactivation of the LV lead mainly occurs as a result of LV lead displacement which, in studies using unipolar and bipolar leads, occurs more frequently than with atrial or right ventricular leads.[5](#jah32587-bib-0005){ref-type="ref"}, [6](#jah32587-bib-0006){ref-type="ref"}, [7](#jah32587-bib-0007){ref-type="ref"}, [8](#jah32587-bib-0008){ref-type="ref"}, [9](#jah32587-bib-0009){ref-type="ref"} Since their launch in 2010, quadripolar LV leads (QUAD) have been considered by implanters as a "game‐changer," even before robust clinical evidence emerged in their favor. Observational studies and a randomized, controlled trial[10](#jah32587-bib-0010){ref-type="ref"} have since shown that QUAD is associated with higher implant success rates and lower rates of re‐interventions for LV lead displacement or PNS.[11](#jah32587-bib-0011){ref-type="ref"}, [12](#jah32587-bib-0012){ref-type="ref"} Some observational studies have suggested that CRT‐D using QUAD programmed to single‐site LV pacing also improves survival.[12](#jah32587-bib-0012){ref-type="ref"}, [13](#jah32587-bib-0013){ref-type="ref"}, [14](#jah32587-bib-0014){ref-type="ref"} These findings, however, are not consistent,[15](#jah32587-bib-0015){ref-type="ref"} and there is uncertainty as to whether they also apply to CRT‐P. Moreover, the possible influence of HF etiology and the effects of QUAD on HF hospitalization and mode of death remain largely unexplored. Methods {#jah32587-sec-0009} ------- This is a nonrandomized, retrospective, observational study comparing clinical outcomes of patients undergoing CRT‐D and CRT‐P device implantation using unipolar, bipolar, and quadripolar leads in a single center (Queen Elizabeth Hospital, Birmingham, United Kingdom) from February 2010 to January 2017. The study was approved by the Clinical Audit Department at the Queen Elizabeth Hospital, which does not require informed consent for audit of clinical care delivery. The study conforms to the Declaration of Helsinki. Implantation {#jah32587-sec-0010} ------------ Device implantation was undertaken using standard techniques with patients under local anesthesia and intravenous sedation. Access was gained via subclavian, axillary, and cephalic veins. The LV pacing site was chosen by the implanter on the basis of lead stability, absence of PNS, and adequate pacing parameters. An implant was considered a failure in the event of failure to deploy all desired leads and device at the index procedure. The first QUAD was implanted in February 2010. The following QUAD leads were used: Quartet 1458Q (St. Jude Medical, Sylmar, CA), Attain Performa (Medtronic Inc, Minneapolis, MN), and Acuity X4 (Boston Scientific, Marlborough, MA). The choice of vector was made at implantation and was made on the basis of presence or absence of PNS. Follow‐Up {#jah32587-sec-0011} --------- Patients were followed up in dedicated device therapy clinics. Before 2013, patients underwent systematic echocardiographic optimization. To this end, patients in sinus rhythm underwent transmitral Doppler‐directed optimization of atrioventricular delay using an iterative technique before discharge and at every scheduled visit thereafter. In patients with sinus rhythm, atrial pacing was set at 60 beats/min, and the pacing mode was set to DDDR with an interventricular delay of 0 to 4 ms, according to the manufacturer. In patients with permanent atrial fibrillation, right ventricular and LV leads were implanted and a CRT generator was used, plugging the atrial port and programming the generator to a ventricular triggered mode. In patients with uncontrolled atrial fibrillation despite medical therapy with suboptimal biventricular pacing capture (\<98%), atrioventricular junction ablation was undertaken, according to the individual clinician\'s decision. After 2013, echocardiographic optimization was only undertaken in symptomatic nonresponders. End Points {#jah32587-sec-0012} ---------- The primary end point was total mortality, which included cardiac transplantation. Secondary end points included cardiac mortality and unplanned HF hospitalization. The first event was included in the analysis. With respect to mode of death, sudden cardiac death was defined as a "natural, unexpected death due to cardiac causes, heralded by an abrupt loss of consciousness within 1 hour of the onset of acute symptoms,"[16](#jah32587-bib-0016){ref-type="ref"} whereas death from pump failure was defined as "death after a period of clinical deterioration in signs and symptoms of heart failure despite medical treatment"[17](#jah32587-bib-0017){ref-type="ref"} or cardiac transplantation. Mortality data were collected through medical records every 3 months by investigators who were blinded to all other patient data. Mortality and event data were collected by separate investigators who were blinded to all other data, except patient identifiers. Statistical Analysis {#jah32587-sec-0013} -------------------- In preliminary analyses, no differences in outcomes emerged between unipolar and bipolar LV leads (data not shown). On this basis, the latter were classified as "non‐QUAD" in statistical analyses. Normality was tested using the Shapiro--Wilk test. Continuous variables are expressed as mean (±SD) and compared using the Student *t* test. Categorical variables were compared using the χ^2^ tests. Kaplan--Meier curves and the log‐rank tests were used to assess observed cumulative survival and to test for differences in survival, respectively. Cox proportional hazard models were used to compare hazard rates of subgroups. Variables reaching a *P*\<0.10 on univariable analyses were entered in multivariable models, and further backward elimination was applied for the final multivariable models. Confounders included in final models were the following: quadripolar lead, sex (male), age at implantation, New York Heart Association (NYHA) class, creatinine, QRS duration, and medication of angiotensin‐converting enzyme inhibitors/angiotensin receptor blockers. Proportionality hypotheses were verified by visual examination of log (survival) graphs to ensure parallel slopes, and by examining Schoenfeld residuals. Statistical analyses were undertaken using Stata 14 (StataCorp, Houston, TX). A 2‐sided *P*≤0.05 was considered statistically significant. Results {#jah32587-sec-0014} ======= Baseline Characteristics {#jah32587-sec-0015} ------------------------ Over the study period of 6.9 years, 847 patients underwent CRT (CRT‐D: 436 \[51.5%\]; CRT‐P: 411 \[48.5%\]), using QUAD (287 \[33.9%\]), unipolar (63 \[7.43%\]), or bipolar (497 \[58.7%\]) leads. Implantations using unipolar and bipolar leads were classified as non‐QUAD. As shown in Table [1](#jah32587-tbl-0001){ref-type="table"}, the groups were well matched for age, sex, cause of cardiomyopathy, comorbidities, proportion of upgrades from pacemaker, atrial rhythm (sinus rhythm or atrial fibrillation), QRS morphology, QRS duration, and left ventricular ejection fraction. Compared with the non‐QUAD group, QUAD were more likely to be in NYHA class I and II (*P*\<0.001) and to undergo CRT‐D (62.0% versus 46.1%, *P*\<0.001). In addition, QUAD had a lower uptake of loop diuretics (*P*=0.003) and a higher uptake of β‐blockers (*P*=0.004). ###### Characteristics of the Study Group QUAD Non‐QUAD *P* Value[a](#jah32587-note-0002){ref-type="fn"} ------------------------------------------------------------------- --------------- --------------- -------------------------------------------------- N 287 560 Sex (male), n (%) 209 (72.8) 398 (71.1) 0.592 Age, y 72.5±12.2 73.2±11.3 0.517 NYHA class, n (%) I 40 (14.3) 32 (5.8) \<0.001 II 87 (31.1) 66 (11.9) III 148 (52.9) 419 (75.5) IV 5 (1.79) 38 (6.9) Cause of cardiomyopathy, n (%) Ischemic 151 (52.6) 279 (49.8) 0.442 Nonischemic 136 (47.4) 281 (50.2) Device type, n (%) CRT‐D 178 (62.0) 258 (46.1) \<0.001 CRT‐P 109 (38.0) 302 (53.9) Comorbidities, n (%) Diabetes mellitus 80 (28.1) 127 (22.9) 0.099 Hypertension 87 (30.5) 165 (29.7) 0.811 CABG 48 (16.7) 102 (18.2) 0.591 Upgrade from pacemaker 56 (19.5) 119 (21.6) 0.554 ECG variables Sinus rhythm, n (%) 195 (67.9) 356 (63.6) 0.206 Atrial fibrillation, n (%)[b](#jah32587-note-0003){ref-type="fn"} 92 (32.1) 204 (36.4) QRS morphology (LBBB), n (%) 228 (79.4) 438 (78.2) 0.680 QRS duration, ms[c](#jah32587-note-0004){ref-type="fn"} 152.6±24.0 152.4±25.0 0.896 Medication, n (%) Loop diuretics 274 (95.5) 553 (98.8) 0.003 ACEIs/ARA 257 (89.6) 489 (87.3) 0.344  β‐Blockers 227 (79.1) 391 (69.2) 0.004 MRA 124 (43.2) 244 (43.6) 0.919 LVEF, % 24.9±9.1 25.8±11.2 0.249 Creatinine, μmol/L[d](#jah32587-note-0005){ref-type="fn"} 106 (89--129) 104 (87--132) 0.727 Variables are expressed as mean±SD, unless indicated otherwise. ACEI indicates angiotensin‐converting enzyme inhibitors; ARA, angiotensin receptor blockers; CABG, coronary artery bypass grafting; CRT‐D, cardiac resynchronization therapy‐defibrillation; CRT‐P, cardiac resynchronization therapy‐pacing; LBBB, left bundle branch block; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonists; NYHA, New York Heart Association; QUAD, quadripolar left ventricular lead. Refers to differences between the groups from ANOVA for continuous variables and from χ^2^ tests for categorical variables. Includes permanent, persistent, and paroxysmal atrial fibrillation. Excludes upgrades to pacemaker. Log‐transformed for statistical analyses. Total Mortality {#jah32587-sec-0016} --------------- Over a follow‐up period of 3.2 years (median \[interquartile range,1.90 to 5.0; 1.8 years \[interquartile range, 1.0--2.6\] for QUAD; 4.7 years \[interquartile range, 3.4--5.7\] for non‐QUAD), QUAD was associated with a lower total mortality in Kaplan--Meier survival analyses (log rank *P*\<0.001, Figure [1](#jah32587-fig-0001){ref-type="fig"}). The annualized total mortality rate was 3.6% (n=19) for QUAD and 10.9% (n=218) for non‐QUAD. Event rates are shown in Table [2](#jah32587-tbl-0002){ref-type="table"}. Univariable Cox proportional hazards analyses are shown in Table [3](#jah32587-tbl-0003){ref-type="table"}. In multivariable analyses (Table [4](#jah32587-tbl-0004){ref-type="table"}), QUAD was associated with a lower mortality (adjusted hazard ratio \[aHR\]: 0.32, 95% confidence interval \[CI\], 0.20--0.52), after adjustment for age, sex, NYHA class, and creatinine. Other confounders failed to reach significance in multivariable models. In order to exclude a possible time‐related bias, we explored whether date of implant emerged as a predictor of total mortality. In multivariable Cox proportional hazards analysis, date of implant did not predict total mortality (HR: 1.46, 95% CI, 0.95--2.26). {#jah32587-fig-0001} ###### Event Rates According to Lead Type QUAD (n=287) Non‐QUAD (n=560) ------------------------- -------------- ------------------ ----- ------ Total mortality 19 3.6 218 10.9 Cardiac mortality 13 2.4 136 6.0 HF hospitalization 22 4.4 104 5.6 Death from pump failure 11 2.1 122 5.4 SCD 2 0.4 13 0.6 HF indicates heart failure; QUAD, quadripolar left ventricular lead; SCD, sudden cardiac death. Data are expressed in terms of annualized event rates. ###### Univariable Cox Proportional Hazards Analyses of Baseline Variables in Relation to Clinical Outcomes   Total Mortality Cardiac Mortality HF Hospitalization ------------------------ ----------------- ------------------- -------------------- --------- ------ ------ ------- --------- ------ ------ ------ --------- Lead type (QUAD) 0.31 0.19 0.49 \<0.001 0.35 0.20 0.63 \<0.001 0.60 0.37 0.95 0.030 Sex (male) 1.76 1.28 2.42 0.001 1.99 1.30 3.04 0.001 1.23 0.82 1.84 0.309 Age 1.04 1.02 1.05 \<0.001 1.02 1.00 1.03 0.029 1.03 1.01 1.04 0.004 NYHA class III 1.50 1.04 2.18 0.031 1.41 0.90 2.21 0.138 1.25 0.80 1.94 0.329 IV 3.74 2.25 6.21 \<0.001 2.40 1.25 4.61 0.009 1.24 0.51 3.03 0.634 Cause (ischemic) 1.23 0.95 1.59 0.112 1.21 0.88 1.67 0.249 1.31 0.92 1.87 0.130 Device type (CRT‐D) 0.81 0.63 1.04 0.100 0.96 0.70 1.33 0.810 0.98 0.69 1.39 0.922 Comorbidities Diabetes mellitus 1.33 1.00 1.77 0.050 1.57 1.11 2.22 0.010 1.48 1.02 2.16 0.042 Hypertension 1.10 0.83 1.45 0.509 0.91 0.63 1.31 0.620 0.96 0.65 1.41 0.820 CABG 1.16 0.85 1.60 0.352 1.05 0.70 1.59 0.801 1.04 0.66 1.63 0.872 ECG variables Atrial fibrillation 1.36 1.05 1.76 0.019 1.26 0.91 1.74 0.171 0.94 0.65 1.36 0.741 QRS morphology (LBBB) 0.81 0.60 1.09 0.161 0.71 0.50 1.02 0.064 0.57 0.39 0.84 0.004 QRS duration, ms 1.00 0.99 1.00 0.115 0.99 0.99 1.00 0.041 0.99 0.98 0.99 \<0.001 Medication Loop diuretics 1.38 0.44 4.32 0.578 2.73 0.38 19.51 0.317 ACEIs/ARA 0.54 0.39 0.76 \<0.001 0.52 0.35 0.79 0.002 1.28 0.69 2.39 0.428 β‐Blockers 0.87 0.66 1.14 0.309 0.81 0.57 1.13 0.215 0.91 0.62 1.33 0.612 MRA 0.90 0.69 1.16 0.410 1.01 0.73 1.40 0.949 1.25 0.88 1.77 0.210 LVEF, % 1.00 0.98 1.01 0.462 0.99 0.97 1.01 0.211 1.01 0.99 1.02 0.404 Creatinine, log μmol/L 2.20 1.72 2.82 \<0.001 1.92 1.38 2.67 \<0.001 1.93 1.34 2.76 \<0.001 ACEIs indicates angiotensin‐converting enzyme inhibitors; ARA, angiotensin receptor blockers; CABG, coronary artery bypass grafting; CRT‐D, cardiac resynchronization therapy‐defibrillation; HF, heart failure; LBBB, left bundle branch block; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonists; NYHA, New York Heart Association; QUAD, quadripolar left ventricular lead. Results are expressed as hazard ratios and 95% confidence intervals (CI) from Cox proportional hazards analyses. ###### Multivariable Analyses of Baseline Variables in Relation to Clinical Outcomes Total Mortality Cardiac Mortality HF Hospitalization ------------------------ ----------------- ------------------- -------------------- --------- ------ ------ ------ ------- ------ ------ ------ --------- Lead type (QUAD) 0.32 0.20 0.52 \<0.001 0.36 0.20 0.65 0.001 0.62 0.39 0.99 0.047 Sex, male 1.65 1.18 2.31 0.003 1.74 1.13 2.70 0.013 ··· ··· ··· ··· Age, y 1.03 1.02 1.05 \<0.001 1.02 1.00 1.03 0.043 1.03 1.01 1.05 0.001 NYHA class (IV) 1.89 1.25 2.86 0.003 ··· ··· ··· ··· ··· ··· ··· ··· QRS duration, ms ··· ··· ··· ··· 0.99 0.99 1.00 0.019 0.99 0.98 0.99 \<0.001 ACEIs/ARAs ··· ··· ··· ··· 0.64 0.42 0.99 0.044 ··· ··· ··· ··· Creatinine, log μmol/L 1.68 1.25 2.25 0.001 1.50 1.04 2.16 0.030 1.91 1.30 2.80 0.001 Only variables with *P*\<0.10 on univariable analyses were included in multivariable models. ACEI, angiotensin‐converting enzyme inhibitors; ARA, angiotensin receptor blockers; HF, heart failure; NYHA, New York Heart Association; QUAD, quadripolar left ventricular lead. Results are expressed as hazard ratios and 95% confidence intervals (CI) from Cox proportional hazards analyses. Cardiac Mortality {#jah32587-sec-0017} ----------------- The annualized cardiac mortality rate was 2.40% (n=13) for QUAD and 6.0% (n=136) for non‐QUAD (Table [2](#jah32587-tbl-0002){ref-type="table"}). In Kaplan--Meier survival analyses, QUAD was associated with a lower cardiac mortality (log rank *P*\<0.001, Figure [1](#jah32587-fig-0001){ref-type="fig"}). In multivariable analyses (Table [4](#jah32587-tbl-0004){ref-type="table"}), QUAD was associated with a lower cardiac mortality (aHR: 0.36, 95% CI, 0.20--0.65), after adjustment for known confounders. HF Hospitalization {#jah32587-sec-0018} ------------------ The annualized HF hospitalization rate was 4.40% for QUAD (n=22) and 5.6% (n=104) for non‐QUAD (Table [2](#jah32587-tbl-0002){ref-type="table"}). In Kaplan--Meier survival analyses, QUAD was associated with a lower risk of HF hospitalization (log rank *P*=0.028, Figure [1](#jah32587-fig-0001){ref-type="fig"}). In multivariable analyses (Table [4](#jah32587-tbl-0004){ref-type="table"}), QUAD was associated with a lower risk of HF hospitalization (aHR: 0.62, 95% CI, 0.39--0.99), after adjustment for potential confounders. QUAD and Device Type {#jah32587-sec-0019} -------------------- In univariable (Table [3](#jah32587-tbl-0003){ref-type="table"}) and mutivariable (Table [4](#jah32587-tbl-0004){ref-type="table"}) analyses, CRT‐D did not emerge as a predictor of any end point. Separate analyses according to device type were also undertaken. As shown in Figure [2](#jah32587-fig-0002){ref-type="fig"}, QUAD was superior to non‐QUAD with respect to all end points in Kaplan--Meier survival analyses. In univariable analyses including CRT‐D patients only, QUAD was superior to non‐QUAD with respect to total mortality (HR: 0.44, 95% CI, 0.25--0.76) and cardiac mortality (HR: 0.43, 95% CI, 0.22--0.85). A lower, albeit nonsignificant reduction with QUAD was observed in HF hospitalization (HR: 0.54, 95% CI, 0.29--1.01). In univariable analyses including CRT‐P patients only, QUAD was superior to non‐QUAD with respect to total mortality (HR: 0.16, 95% CI, 0.06--0.45) and cardiac mortality (HR: 0.23, 95% CI, 0.07--0.73), but no differences emerged in HF hospitalization (HR: 0.67, 95% CI, 0.32--1.37). {#jah32587-fig-0002} Mode of Death {#jah32587-sec-0020} ------------- Over the follow‐up period, there were 11/287 (3.83%) deaths because of pump failure with QUAD and 122/558 (21.8%) with non‐QUAD. There were 2/287 (0.70%) sudden cardiac deaths with QUAD and 13/560 (2.32%) with non‐QUAD. Noncardiac deaths accounted for 3/287 (1.04%) deaths with QUAD and 42/560 (7.5%) deaths with non‐QUAD. The cause and mode of death was unknown in 3 (1.04%) patients with QUAD and in 40 (7.14%) patients with non‐QUAD. Excluding these patients, QUAD was associated with a lower mortality from pump failure (log rank *P*\<0.001; aHR: 0.33; 95% CI, 0.18--0.62), but no differences emerged with respect to sudden cardiac death (aHR: 0.58; 95% CI, 0.13--2.68, Figure [3](#jah32587-fig-0003){ref-type="fig"}). {#jah32587-fig-0003} Implantation {#jah32587-sec-0021} ------------ There were 856 first attempts at CRT device implantation, 833 (97.3%) of which were successful at the first attempt and 847 (98.9%) after ≥1 attempts. Re‐interventions for LV displacement or PNS were lower with QUAD than with non‐QUAD (Table [5](#jah32587-tbl-0005){ref-type="table"}, Figure [4](#jah32587-fig-0004){ref-type="fig"}). In univariable analyses, re‐interventions for LV displacement or PNS predicted total mortality (aHR: 1.68, 95% CI, 1.11--2.54), cardiac mortality (aHR: 2.61, 95% CI, 1.66--4.11), and HF hospitalization (aHR: 2.09, 95% CI, 1.22--3.58). ###### Implant‐Related Complications and Re‐Interventions All QUAD Non‐QUAD *P* Value[a](#jah32587-note-0013){ref-type="fn"} ------------------------------------------------------------------ ----------- ----------- ----------- -------------------------------------------------- Implant‐related complications, n (%) Hematoma treated conservatively 23 (2.72) 10 (3.48) 14 (2.50) 0.390 Hematoma requiring evacuation 4 (0.47) 0 4 (0.71) Pneumothorax treated conservatively 5 (0.59) 2 (0.70) 3 (0.54) Pneumothorax requiring drainage 1 (0.12) 0 1 (0.18) Perforation by RV lead 2 (0.24) 1 (0.35) 1 (0.18) Coronary sinus dissection[b](#jah32587-note-0014){ref-type="fn"} 5 (0.59) 4 (1.39) 3 (0.54) Subclavian artery aneurysm 1 (0.12) 1 (0.35) 0 Arrhythmia requiring cardioversion 1 (0.12) 1 (0.35) 0 Anemia postprocedure 1 (0.12) 0 1 (0.18) Pulmonary edema 1 (0.12) 0 1 (0.18) Total, n (%) 44 (5.19) 19 (6.62) 28 (5.00) Extractions for infection Within 1 y 8 (1.43) 3 (1.05) 5 (0.89) 0.297 After 1 y 3 (0.53) 0 3 (0.54) Total, n (%) 11 (1.96) 3 (1.05) 8 (1.43) LV lead re‐interventions LV lead displacement 34 (4.01) 6 (2.09) 28 (5.0) 0.007 Phrenic nerve stimulation 19 (2.24) 3 (1.05) 16 (2.86) Total 53 (6.26) 9 (3.14) 44 (7.86) RV indicates right ventricular. Refers to χ^2^ tests of quadripolar (QUAD) compared with non‐QUAD left ventricular (LV) leads. No coronary sinus dissections required re‐interventions. {#jah32587-fig-0004} Other Complications {#jah32587-sec-0022} ------------------- As shown in Table [5](#jah32587-tbl-0005){ref-type="table"}, implant‐related complications were similar for QUAD and non‐QUAD (odds ratio: 1.30, 95% CI, 0.71--2.36). A total of 8 extractions for system infection were undertaken within 1 year of implantation (QUAD: 3 (1.05%); non‐QUAD: 5 (0.90%; *P*=0.828) and 3 after 1 year (QUAD: 0); non‐QUAD: 3 (0.54%; *P*=0.214). No device‐related infection or subsequent extraction led to death. Lead Design {#jah32587-sec-0023} ----------- Three LV lead families from 3 manufacturers were used, namely, Quartet (n=189, St. Jude Medical, Sylmar, CA), Attain Performa (n=87, Medtronic Inc, Minneapolis, MN), and Acuity X4 (n=11, Boston Scientific, Marlborough, MA). Compared with non‐QUAD leads, Quartet leads (aHR: 0.36, 95% CI, 0.21--0.6; sample size: 560 non‐QUAD and 189 Quartet leads) as well as the Attain Performa leads (aHR: 0.11, 95% CI, 0.03--0.45; sample size: 560 non‐QUAD and 87 Attain Performa) were associated with lower total mortality. Comparison of Quartet (n=189) with Attain Performa (n=87) revealed no difference in total mortality (Quartet HR: 3.06, 95% CI, 0.70--13.38). Boston Scientific Acuity X4 leads were excluded from these analysis because of the small numbers involved (n=11). Discussion {#jah32587-sec-0024} ========== In this study, we have compared clinical outcomes after CRT using QUAD and non‐QUAD, programmed to biventricular, single‐site LV pacing. Several findings have emerged. First, QUAD was associated with a 68% lower total mortality. Second, QUAD was associated with a marked reduction in cardiac mortality (by 64%) and in HF hospitalization (by 38%). Third, QUAD was associated with a lower mortality from pump failure, while no differences emerged in sudden cardiac death. Fourth, HF cause did not impact on the superior outcomes of QUAD over non‐QUAD. Fifth, QUAD was superior to non‐QUAD after both CRT‐D and CRT‐P. Sixth, no group differences emerged in implant complications, but QUAD was associated with fewer re‐interventions for LV lead displacement or PNS. Seventh, re‐interventions for LV displacement or PNS predicted total mortality, cardiac mortality, and HF hospitalization. Mortality {#jah32587-sec-0025} --------- A recent retrospective study comparing QUAD with bipolar leads showed no difference in survival at 12 months (mean follow‐up 256 days for QUAD).[15](#jah32587-bib-0015){ref-type="ref"} In contrast, a US‐wide study based on data from device implant records and telemonitoring showed that CRT‐D using QUAD was associated with a better survival than CRT‐D using bipolar leads.[13](#jah32587-bib-0013){ref-type="ref"} Observational data from 3 centers in the United Kingdom showed similar findings.[12](#jah32587-bib-0012){ref-type="ref"} Our annualized total mortality rate for non‐QUAD (10.9%) is comparable to that found in randomized, controlled trials using non‐QUAD, which amounted to 9.7% in CARE‐HF (Cardiac Resynchronization Heart Failure)[18](#jah32587-bib-0018){ref-type="ref"} and 15% in COMPANION (Comparison of Medical Therapy, Pacing and Defibrillation in heart failure)[19](#jah32587-bib-0019){ref-type="ref"}, [20](#jah32587-bib-0020){ref-type="ref"} after CRT‐P. In the CRT‐D arm of COMPANION, the annualized total mortality rate was 12%.[20](#jah32587-bib-0020){ref-type="ref"} In contrast, the annualized total mortality rate in the present study was 3.6% with QUAD. HF Hospitalization {#jah32587-sec-0026} ------------------ This is the first study to explore HF hospitalization after CRT using QUAD. Survival free from cardiovascular hospitalization at 1 and 2 years with QUAD was 94% and 91%, respectively. Among the few studies to address the long‐term effects of CRT on HF hospitalizations in the non‐QUAD LV lead era, van Bommel et al found survival free from cardiovascular hospitalization at 1 and 2 years was 80% and 70%, respectively.[21](#jah32587-bib-0021){ref-type="ref"} The reasons for reduced HF hospitalizations with QUAD are not entirely clear. Several cofounders, however, could potentially explain our findings. In the telemonitoring study of Turakhia et al, potential confounders for a benefit of QUAD was limited to age, sex, remote monitoring enrollment, and socioeconomic status.[13](#jah32587-bib-0013){ref-type="ref"} Behar et al did not adjust for NYHA class, QRS duration, left ventricular ejection fraction, HF medication, or comorbidities.[12](#jah32587-bib-0012){ref-type="ref"} In the present study, which comprises a longer follow‐up period, the survival advantage of QUAD versus bipolar LV leads was observed after adjustment for age, sex, device type (CRT‐P or CRT‐D), NYHA class, QRS duration, QRS morphology, left ventricular ejection fraction, HF cause, medication, or history of hypertension, coronary artery bypass grafting, or diabetes mellitus. Lower rates of LV lead re‐interventions may also be relevant. In this respect, we observed that no LV lead revision led to death and that LV lead re‐interventions were lower for QUAD than for non‐QUAD. On the other hand, LV lead re‐interventions were associated with an increased risk of total mortality, cardiac mortality, and HF hospitalization. These findings suggest that LV lead deactivation and the associated re‐intervention contributed to a higher risk of HF hospitalization. This, however, does not explain the lower risk of total mortality observed with QUAD. It would appear that the survival benefit of QUAD relates to the lead itself. CRT‐D and CRT‐P {#jah32587-sec-0027} --------------- Previous studies on QUAD[12](#jah32587-bib-0012){ref-type="ref"}, [13](#jah32587-bib-0013){ref-type="ref"} have exclusively focused on CRT‐D. We have shown better outcomes for QUAD after both CRT‐D and CRT‐P. In fact, the magnitude of the survival benefit of QUAD over non‐QUAD after CRT‐P (by 84%) was higher than after CRT‐D (by 56%). While we should be careful with overinterpreting the results of a retrospective study, such marked differences raise the possibility that the benefit of QUAD is proportionally higher after CRT‐P than after CRT‐D. If that is the case, we should reconsider the findings of COMPANION, which was underpowered to compare CRT‐D and CRT‐P. At the low event rates observed in the present study, proof of superiority of CRT‐D over CRT‐P may require much higher numbers of patients than those included in COMPANION. Meta‐analyses of CRT‐D versus CRT‐P may need to be revisited in the QUAD era. LV Lead Re‐Interventions {#jah32587-sec-0028} ------------------------ In the MORE‐CRT (More Options Available With a Quadripolar LV Lead Provide In‐Clinic Solutions to CRT Challenges) trial, 1074 patients undergoing CRT‐D were randomized in 1:2 ratio to bipolar leads or QUAD. Freedom from the composite end point of intraoperative and postoperative LV lead--related events at 6 months was greater with QUAD than with bipolar leads (83.0% versus 74.4%, *P*=0.0002), but this was because of differences in the intraoperative rather than postoperative events.[10](#jah32587-bib-0010){ref-type="ref"} In the present study, which involved a longer follow‐up period, QUAD was associated with a lower incidence of PNS and LV lead displacement. This might be expected in view of the fact that vector optimization almost invariably eliminates PNS in patients who initially have PNS with QUAD.[12](#jah32587-bib-0012){ref-type="ref"} HF Cause {#jah32587-sec-0029} -------- In the QUAD era, Forleo et al[22](#jah32587-bib-0022){ref-type="ref"} showed that the cause of cardiomyopathy did not influence the LV reverse remodeling response to CRT‐D using QUAD. Similarly, Behar et al found no interaction between HF cause and the survival benefit of CRT‐D using QUAD, compared with bipolar leads.[12](#jah32587-bib-0012){ref-type="ref"} We have also found that HF cause has no bearing on survival benefit of QUAD over non‐QUAD. Importantly, however, the event rate with QUAD is much lower than with non‐QUAD. It is possible that with QUAD, higher numbers of patients are needed to show a HF cause‐specific difference in outcomes after CRT. Single‐Site and Multipoint Pacing {#jah32587-sec-0030} --------------------------------- In this study, multipoint pacing was not activated in any patient, suggesting that the survival advantage of QUAD is simply because of lead design or availability of multiple pacing vector configurations. We should also consider that electrical stimulation over a dipole in a QUAD could depolarize the myocardium at the anodal pole if this has a comparatively low threshold.[23](#jah32587-bib-0023){ref-type="ref"} We cannot determine whether anodal capture (and effectively multipoint pacing) could account for some of the observed effects of QUAD. These questions could not be addressed in the present study. Limitations {#jah32587-sec-0031} ----------- This study has the typical limitations of a single‐center, nonrandomized, retrospective study and therefore we cannot discount the possible influence of unobserved variables. While we adjusted for potential confounders, only randomization could fully correct for their biological effects. In particular, we should stress that, despite covariate adjustment, the better outcomes observed with QUAD might be because, at least in part, of different patient characteristics towards the end of the recruitment period, rather than primarily or uniquely to the utilization of QUAD. The greater proportion of patients in NYHA class I and II and in sinus rhythm as well as higher uptake of angiotensin‐converting enzyme inhibitors/angiotensin receptor blockers and β‐blockers may still contribute to better outcomes with QUAD. Moreover, we cannot exclude the possibility that the time interval from actual LV lead displacement (deactivation) to re‐intervention may have adversely influenced outcomes. As we did not use telemonitoring, we cannot quantify the duration of LV lead deactivation before re‐intervention. A further possibility is that allowing programming over a wider range of vectors, QUAD could have converted nonresponders to responders.[24](#jah32587-bib-0024){ref-type="ref"} Unfortunately, the present study does not address vector locations or how vector configurations changed during follow‐up. It is hoped that ongoing prospective studies[25](#jah32587-bib-0025){ref-type="ref"} may shed further light on this issue. We did not collect data as on Q‐LV as an aid for targeting LV lead positions, but it is possible that this approach could influence outcomes. Conclusions {#jah32587-sec-0032} =========== In this study of real‐world clinical practice, we have shown that CRT using QUAD, programmed to biventricular, single‐site LV pacing, was associated with a dramatic reduction in total mortality, cardiac mortality, and HF hospitalization, compared with non‐QUAD. These findings emerged after both CRT‐D and CRT‐P, after adjustment for HF etiology and other potential confounders. The remarkably low event rate observed with QUAD in this study has implications for clinical practice and the design of future CRT trials. Disclosures {#jah32587-sec-0033} =========== Leyva is a consultant and has received research support from Medtronic Inc, St Jude Medical, Boston Scientific, and LivaNova. Marshall is a consultant for Spectranetics. The other authors report no conflicts of interest.

Background ========== Diamond holds a variety of extraordinary physical and chemical properties, facilitating its possible applications in novel functional devices \[[@B1]-[@B7]\]. As a semiconductor with a wide bandgap of 5.47 eV, it is a promising candidate for short-wavelength optoelectronic devices such as ultraviolet light-emitting diodes. The extreme mechanical hardness of diamond endows it with potential applications in nanomechanical devices. When doped with boron, it was found to display superconductivity around liquid helium temperature. To utilize the qualities of diamond, it is imperative to grow high-quality materials. Chemical vapor deposition is an efficient and versatile technique for the growth of diamond. A large body of experiments and theories are dedicated to understanding the growth process \[[@B8]\]. Graphitic-like surface reconstructions on stepped C(111) surfaces are predicated by first-principles calculations \[[@B9]\]. Surface graphitization of diamond nanoparticles is investigated from an experimental viewpoint \[[@B10]\]. A unique character of diamond growth is the existence of *sp*^2^-hybridized bonds in the graphitic-like layer of diamond surfaces, in contrast to other group IV element semiconductors (Si and Ge), which do not exhibit energetically favorable *sp*^2^ bonding configurations. This may account for different surface reconstructions on Si and diamond surfaces \[[@B11]\]. Besides low-index surfaces, high-index Si surfaces are extensively investigated to unveil their atomic and electronic structures \[[@B12],[@B13]\], whereas less attention has been paid to the study of high-index diamond surfaces. The graphite-like *sp*^2^ bonding is expected to give rise to the significant difference between high-index diamond and Si surfaces. Graphene, a two-dimensional atomic crystal with graphite-like *sp*^2^ bonding, has attracted considerable interests due to its novel physical and chemical properties and its potential applications in nanoelectronics and optoelectronics \[[@B14]\]. Large-scale graphenes are grown on metal substrates \[[@B15]\]. Here, we explore the formation of graphene-like stripes on a reconstructed high-index diamond C(331) surface using first-principles density functional theory (DFT) calculations. During the structural relaxation of the bulk-terminated surface, the terrace C atoms in the first layer delaminate from the second layer, leading to local *sp*^3^ to *sp*^2^ rehybridization and the formation of graphene-like stripes on the surface. The driving force for the graphitic-like reconstruction is the presence of high-density dangling bonds on the surface, which gives rise to the rebonding of top-layer atoms. The comparison of the calculated absolute surface energies of C(331), C(111), and C(110) demonstrates the relative stability of the C(331) surface with the graphitic-like reconstruction. Local density of electronic states (LDOS) analysis reveals the occurrence of localized electronic states near the Fermi level (FL), which may play an essential role in determining the surface conductivity \[[@B16],[@B17]\]. Methods ======= The calculations are conducted in the framework of the DFT method by DMol^3^ codes \[[@B18]\]. We use the Perdew-Burke-Ernzerhof generalized gradient approximation \[[@B19]\]. A double numeric basis set including *d*-polarization function, all electron treatment, and an 8 × 2 × 1 Monkhorst-Pack *k-*point mesh for the Brillouin zone sampling \[[@B20]\] are employed to carry out geometry optimization and electronic band structure calculations. Spin-unpolarized self-consistent field calculations are performed with a convergence criterion of 2.0 × 10^−5^ hartree (1 hartree = 27.2114 eV) for total energies. The maximum force tolerance is 0.004 hartree Å^−1^, and the maximum displacement tolerance is 0.005 Å. The periodically repeated slabs separated by approximately 10 Å of vacuum are used to represent the surface structures. Each slab of C(331) surface is composed of 11 atomic layers with 40 C atoms and 6 H atoms per unit cell. The H atoms are used to passivate the surface C atoms at the bottom of the slabs to make the calculation more efficient. The dashed lines in Figure [1](#F1){ref-type="fig"}a and the dashed box in Figure [1](#F1){ref-type="fig"}b indicate the supercell used for the calculation. Each slab of H-passivated C(331) surface is composed of 12 atomic layers with 40 C atoms and 12 H atoms per unit cell. The dashed lines in Figure [2](#F2){ref-type="fig"} indicate the supercell used for the calculations. ![**Calculated atomic structure of diamond C(331) surface with graphene-like stripes.** (**a**) The dashed lines indicate the supercell viewed from the $\left\lbrack {0\overline{1}1} \right\rbrack$ direction. The large circles denote the C atoms, and the small circles denote the H atoms. (**b**) The dashed box indicates the supercell viewed from the \[331\] direction, and the bottom is viewed from the $\left\lbrack {0\overline{1}1} \right\rbrack$ direction. The large circles denote the C atoms of the graphitic layer, and the smaller circles indicate the *sp*^3^-bonded C atoms in the outmost surface. The other C and H atoms are represented by the smallest circles.](1556-276X-7-460-1){#F1} {#F2} Results and discussion ====================== Figure [1](#F1){ref-type="fig"} shows the atomic structure of the graphene-like stripes formed on the reconstructed diamond C(331) surface calculated after the structural relaxation of the bulk-terminated surface. We allow this surface to relax using a steepest descent algorithm. The top-layer C atoms exhibit the *sp*^2^ bonding configuration in the graphene-like structure, as shown in Figure [1](#F1){ref-type="fig"}b. Upon structural relaxation, the terrace C atoms (see 4 and 10 C atoms in Figure [3](#F3){ref-type="fig"}) delaminate from the subsurface diamond and form the graphene-like stripes along the $\left\lbrack {0\overline{1}1} \right\rbrack$ direction. The energetically favorable hexagonal rings are found to emerge in the graphitic layer on the reconstructed surface. The driving force for the graphitic-like reconstruction on the surface is the presence of high-density dangling bonds which have unpaired electrons. This situation is similar to the reconstruction of the C(111) surface, where the top-layer C atoms are rearranged to make the dangling bonds become the nearest neighbors and form the π bonding \[[@B21]\]. For the C(331) surface, the delamination of the terrace C atoms can lead to the formation of graphite-like *sp*^2^ bonds, thereby reducing the energetically unfavorable dangling bonds. {#F3} The representative C-C bond lengths for the graphitic-like reconstructed C(331) surface are shown in Figure [3](#F3){ref-type="fig"}. The distance between the delaminated C atom and the subsurface C atom increases to approximately 2.51 Å, much larger than the bond length of diamond (1.54 Å). The bond lengths for the C atoms in the graphitic structure decrease to 1.44 and 1.46 Å. These values are quite close to the bond length of graphite (1.42 Å), whereas much smaller than that of diamond. The C atoms with the unsatu-rated dangling bonds at the subsurface positions remain *sp*^3^-hybridized in character, although they have stretched by almost 34%. The C-C bonds are stretched to 1.62 and 1.57 Å for the outmost C atoms attached to the second-layer C atoms. The severe subsurface rebonding increases the elastic strain, which is energetically unfavorable. The competition between the favorable *sp*^2^ bonding in the graphitic layer and the unfavorable strain energy leads to the graphitic-like reconstruction of the C(331) surface. The energetic stability of the C(331) surface is studied by comparing its absolute surface energy (ASE) with those of low-index diamond C(111) and C(110) surfaces \[[@B21]-[@B23]\]. In the centrosymmetric slab used for computing the ASE, the top and bottom surfaces are physically equivalent. After full structural relaxation, the same *n* × *m* surface reconstruction is observed to occur on both sides of the slab. Therefore, it allows calculating directly the ASE. For the slab with *N* atoms at the atomic configuration $\left\{ R_{i} \right\}$, the surface energy per 1 × 1 surface cell, $E_{\text{surf}}^{n \times m}$, can be calculated from the total energy $E_{\text{tot}}\left( {N,\left\{ R_{i} \right\}} \right)$ of the slab subtracted by *N* times the bulk diamond energy μ per atom. The surface energy is expressed as $$E_{\text{surf}}^{n \times m} = \frac{1}{2nm}\left\{ {E_{\text{tot}}\left( {N,\left\{ R_{i} \right\}} \right) - N\mu} \right\}\text{.}$$ Since two equivalent surfaces are involved in the calculations for a slab, a prefactor, $\frac{1}{2}$, is added in Equation 1. For the *n* × *m* surface reconstruction, the *nm* gives the number of the 1 × 1 surface cell. The surface energy per unit area is as follows: $$\gamma^{n \times m} = \frac{E_{\text{surf}}^{n \times m}}{A}\text{,}$$ where *A* is the area of a 1 × 1 surface cell for a given surface orientation *n*. For the H-covered C(331) surface, the surface energy per 1 × 1 surface cell is given by $$E_{\text{surf}}^{\text{H}} = \frac{1}{2}\left\{ {E_{\text{tot}}\left( {N,N_{\text{H}},\left\{ R_{i} \right\}} \right) - N\mu - N_{H}\mu_{H}} \right\},$$ where $E_{\text{tot}}\left( {N,N_{\text{H}},\left\{ R_{i} \right\}} \right)$ is the total energy of the slab, N~H~ is the number of H atoms, and μ~H~ is the chemical potential of the H atom in the reservoir that is defined in \[[@B21]\]. Table [1](#T1){ref-type="table"} collected the surface energies $E_{\text{surf}}^{n \times m}$, $\gamma^{n \times m}$, and $E_{\text{surf}}^{H}$ for various orientations and reconstructions. The computed energies for low-index C(111) and C(110) surfaces agree well with the previous investigation \[[@B21]\]. The graphitic-like reconstructed C(331) surface is found to have lower $\gamma^{n \times m}$ than low-index C(111) and C(110) surfaces, indicating that the C(331) surface is one of the stable crystalline diamond surfaces. ###### **Absolute surface energies**$\mathbf{E}_{\textbf{surf}}^{\mathbf{n} \times \mathbf{m}}$**and**$\mathbf{\gamma}^{\mathbf{n} \times \mathbf{m}}$**for various orientations and reconstructions** **Orientation** **Reconstruction** ***E***~**surf**~**(eV/1 × 1 cell)** ***γ*(J/m**^**2**^**)** ----------------- -------------------- -------------------------------------- ------------------------- \(111\) 2 × 1 0.993 2.91 (1.369) (4.06) H-covered −1.903 −5.57 (−2.760) (−8.19) \(110\) 1 × 1 relaxed 1.824 3.27 (3.264) (5.93) H-covered −4.971 −8.91 (−5.496) (−9.99) \(331\) 1 × 1 graphitic 2.040 2.31 H-covered −5.808 −6.58 The values in parentheses from \[[@B21]\] are listed for comparison. The H adsorption on the graphitic-like reconstructed C(331) surface is found to give rise to the reversion of *sp*^2^ hybridization back to *sp*^3^ hybridization. Figure [2](#F2){ref-type="fig"} shows the calculated atomic structure of the H-covered C(331) (1 × 1) surface. The top-layer C atoms display *sp*^3^ bonding configuration. Thus, the H atoms can give rise to the dereconstruction of the graphitic-like C(331) surface. Figure [4](#F4){ref-type="fig"}a shows the LDOS of the H-passivated diamond (331) surface. The zero energy corresponds to the FL which is at the position of the top valence band. An energy bandgap of 4.2 eV is obtained from the calculated electronic band structure. Figure [4](#F4){ref-type="fig"}b shows the LDOS of the reconstructed C(331) surface with the graphene-like stripes. The zero energy corresponds to the FL, which lifts up to a position in the bulk bandgap. The peak near the FL in the LDOS curve is attributed to the localized electronic states at the graphitic surface and subsurface regions, which may give rise to the semimetallic or metallic conduction along the surface. Further partial electronic density of states (PDOS) analysis reveals that the localized electronic states near the FL is predominant with the *p* character for the graphitic-like reconstructed C(331) surface. {#F4} Conclusions =========== We carry out first-principles DFT calculations to study the spontaneous formation of graphene-like stripes on the reconstructed diamond C(331) surface. The *sp*^2^-hybridized bonding in the graphitic layer on the surface plays a central role in reducing the energetically unfavorable dangling bonds on the bulk-terminated surface, thereby lowering the surface free energy. A sharp peak is found to occur near the FL in the LDOS curve, which arises from the localized electronic states at the surface and subsurface regions. These states may have a significant impact on the surface conductivity. The graphene-like stripes directly formed on a semiconductor surface may be used for nanoelectronic and optoelectronic devices. Abbreviations ============= ASE: absolute surface energy; DFT: density functional theory; FL: Fermi level; LDOS: local density of electronic states; PDOS: partial electronic density of states. Competing interests =================== The authors declare that they have no competing interests. Authors' contributions ====================== MJX did the calculations and wrote the manuscript. YFZ conceived and suggested the calculations. YZZ, JZ, BJQ, JYL, DJL, LW, XSC, and HS discussed about the calculations and revised the final manuscript. All authors read and approved the final manuscript. Authors' information ==================== Dr. MJX obtained his Ph.D. from University of Tsukuba, Japan, and is currently working with Prof. YFZ as postdoctoral research fellow in Shanghai Jiao Tong University, China. Mr. YZZ, Ms. JZ, Mr. BJQ, Mr. JYL, and Mr. DJL are currently postgraduate students in Shanghai Jiao Tong University. Dr. YFZ obtained his Ph.D. from Lanzhou University, China, and is currently working as a professor in Shanghai Jiao Tong University. Dr. LW obtained his Ph.D. from Shanghai Institute of Technical Physics, Chinese Academy of Sciences, China, and is working with Prof. YFZ as postdoctoral research fellow. Dr. XSC obtained his Ph.D. from Nanjing University, China, and is currently working as a professor in Shanghai Institute of Technical Physics, Chinese Academy of Sciences, China. Dr. HS obtained his Ph.D. from Tokyo University, Japan, and is currently working as a professor in University of Tsukuba, Japan. Acknowledgments =============== This work is supported by the National High-Tech R&D Program (863 Program) of China under contract no. 2011AA050504, the National Natural Science Foundation of China (grant no. 61006002), the U-M/SJTU Collaborative Research Program and the Analytical and Testing Center of SJTU.

1. Introduction {#sec1} =============== Nuclear magnetic resonance spectroscopy (NMR, or MRS) has enormous potential for the study of biochemical and physiological changes in cancer tissues, due to its noninvasive nature and the large quantity of specific molecular information it can generate. Despite the sensitivity limitations of the technique, the inherent complexity of the spectra, and inevitable presence of overlapping resonances, there have been several successful NMR-metabonomics studies of cell tissue culture and culture extracts. The focus has been on elucidating the physiopathology of tumors and tumor cells, their drug toxicology and drug resistance, often with a view to identifying diagnostic markers \[[@B1]--[@B8]\]. A further significant complication in such studies arises from variability in the metabolite profile from sample to sample. This reflects many factors \[[@B9]\] including minor variations in growing conditions, the biochemical heterogeneity of the growing cells, the effect of different batches of sera (if used), and variations in cell and sample preparation. These additional factors may mask the inherent metabolite distribution, which may be diagnostic of the pathophysiological state of interest. Experimental complications and difficulties also compromise the extraction of critical information from *in vivo* MRS experiments. In this case, the problems arise from the use of different MR-protocols, which affect the quality of the water suppression, differences in echo time and in the baseline, and so forth. While the causes are different in origin, they have a similar effect on the application. For both forms of magnetic resonance, many of these issues can, in principle, be addressed by improved experimental design, however, it is common for additional sources of variance to be identifiable only after extensive experimentation. In addition to technical issues are the natural physiological variability and the individual treatment history of the subject. As a result, there is an ongoing requirement for the development of magnetic resonance-based diagnostics using advanced statistical-, or other data-, analysis techniques which can reduce or compensate for additional sources of variability. ^1^H NMR spectra of intact tissues or whole-cell samples are inherently complex due to the large number of contributing species which results in significantly overlapping resonance signals. Cell membranes also produce magnetic field inhomogeneity, further broadening the spectra \[[@B10]\]. In the case of cancer cells, a significant proportion of the lipids reside in a fluid environment and hence appear in the liquid-state ^1^H spectra as strong "mobile-lipid" resonances \[[@B7], [@B8], [@B11]\]. Although the identification of the major resonances in ^1^H NMR spectra can be used to characterise the metabolite profile, the complexity of the data sets usually necessitates the use of data reduction and pattern recognition techniques. These can provide information on the biochemical and physiological changes in cancer tissues, related to their physiopathology, drug toxicology, and drug resistance \[[@B12], [@B13]\]. Prominent amongst such techniques is principal component analysis (PCA), \[[@B14], [@B15]\] which involves diagonalisation of the spectral correlation or covariance matrix to identify independent sources of variance (principal components) across the set of spectra, and ranking of the components by their contribution to the overall variance. Thus, PCA is an unsupervised approach to data reprojection that can reveal the presence of classes, it has been applied to a variety of problems in biological science \[[@B16], [@B17]\]. Artificial Neural Networks (ANNs) belong to the so-called Artificial Intelligence group of methods, which were inspired by neurobiology and by the architecture of the human brain \[[@B18]\]. In recent times, these approaches have found applications in many branches of science. For example, they have been used in chemotaxonomy to classify limpets \[[@B19]\] from HPLC mass spectrometric data and in the identification of insect species from morphological measurements \[[@B20]\]. ANNs can be used to model data where the relations, or functions, are not known. There have been some reports of the use of artificial intelligence and network methods in medical diagnostics which have involved analysis of magnetic resonance spectroscopic data. El-Deredy et al. \[[@B21]\] used ANNs to achieve reasonable prediction of the measured *in vitro* chemotherapeutic response from ^1^H NMR of glioma biopsy extracts. More recently, Suna et al. \[[@B22]\] demonstrated the diagnostic potential of unsupervised approaches to classification by successfully analysing simulated ^1^H NMR spectra using self-organising maps. This approach allowed the identification of stages along a metabolic pathway ranging from "normolipidaemic" to "metabolic syndrome". Tate and coworkers \[[@B23]\] reported the trial of an automated decision support system for classification of brain tumors from *in vivo* MRS, which showed a small but significant improvement in diagnostic accuracy over spectroscopy used and interpreted on its own. In recent work \[[@B24]\], we reported PCA of   ^1^H NMR spectra recorded for a group of human lung carcinoma cell lines in culture and ^1^H NMR analysis of extracts from the same samples. The samples studied were cells of lung tumor origin with differing chemotherapy drug resistance patterns. For whole-cell samples, it was found that the statistically significant causes of spectral variation were an increase in the choline and a decrease in the methylene and mobile lipid ^1^H resonance intensities, which were correlated with our knowledge of the level of resistance displayed by the different cell lines. In this paper, we investigate the use of artificial neural network (ANN), a supervised method, to classify lung carcinoma. Two sets of whole-cell ^1^H NMR spectra will be examined. These were recorded for two groups of human lung carcinoma cell lines, these were grown in culture and characterised over two different periods by two different groups of researchers (each consisting of a biologist and a spectroscopist), who both adhered to the same experimental protocol and used the same spectrometer. The cell lines studied include (i) the parent cell line DLKP, a human squamous nonsmall cell lung carcinoma; (ii) DLKP-A; (iii) DLKP-A5F, two resistant daughter lines; (iv) A549, a human lung adenocarcinoma cell line. The study also examines the capability of supervised techniques to compensate for experimental sources of variance, which may include operator bias and the cell culture growth process and in particular provide a test case for the application of ANN architectures in the identification and monitoring of resistance states in cancer tissue by MRS. 2. Experimental {#sec2} =============== 2.1. Cell Samples {#sec2.1} ----------------- The cell lines DLKP \[[@B25], [@B26]\], DLKP-A \[[@B27]\], DLKP-A5F \[[@B28]\], and A549 were grown in culture to approximately 70--80% confluency in 175 cm^2^ tissue culture flasks. Culture conditions were as follows: DLKP, DLKP-A, and DLKP-A5F and were cultured in minimal essential medium/Hams F12 (1 : 1, v/v) supplemented with 5% fetal calf serum and 2 mM L-glutamine. A549 was cultured in Dulbecco\'s modified Eagle\'s medium/Hams F12 (1 : 1, v/v) supplemented with 5% fetal calf serum. Cells were cultured as monolayers in tissue culture flasks and incubated at 37°C. A cell count was performed and c. 5 × 10^7^ cells were separated and pelleted. These were then resuspended in deuterated PBS buffer and were kept in a container at 37°C before the start of the NMR measurements. The methods used were described in detail previously \[[@B24]\]. DLKP cells express a small amount of the multidrug resistance protein-1 (MRP-1) MDR drug efflux pump \[[@B25], [@B26]\]. DLKP-A \[[@B27]\] is a highly resistant clone of DLKP, which overexpresses the P-gp drug efflux pump. DLKP-A5F \[[@B28]\] was derived from DLKP by a different drug exposure profile, it is also highly drug resistant. A549 is an unrelated human lung adenocarcinoma cell line which was obtained from the American Type Culture Collection. The first group of 13 samples, G1_13_21, were grown by a biologist during a six-month period, they were analysed by a first NMR spectroscopist. G1_13_21 contained 21 spectra and so was relatively sparse, it comprised DLKP \[4 samples, 6 spectra\], DLKP-A \[[@B4], [@B6]\], DLKP-A5F \[[@B3], [@B5]\], and A549 \[[@B2], [@B4]\]. The second group of 17 samples, G2_17_33, was grown independently, by a second biologist during a later six-month period and was analysed by a second spectroscopist \[[@B24]\]. G2_17_33 contained 33 spectra, it comprised DLKP \[[@B3], [@B6]\], DLKP-A \[[@B5], [@B10]\], DLKP-A5F \[[@B5], [@B9]\], and A549 \[[@B4], [@B8]\]. Thus for the integrated study presented here, a total of 30 samples were prepared and 54 ^1^H spectra was recorded. The same protocols and methods were used by all the researchers for cell growth and NMR spectroscopy. The biologist and spectroscopist who produced G1_13_21 will be collectively referred to as R1, and the biologist and spectroscopist who produced G2_17_33 will be referred to as R2. Due to the significant work involved in producing the large number of cells required for each spectrum, the number of samples in the study is inevitably somewhat limited. However, the total data set is larger than those usually reported in the analysis of NMR data by pattern recognition methods \[[@B16], [@B17], [@B29]\]. 2.2. ^1^H NMR Spectroscopy of Intact Cells {#sec2.2} ------------------------------------------ NMR spectra of the intact cell samples were recorded in deuterated PBS buffer on a Bruker DPX 400 spectrometer operating at 400.13 MHz for ^1^H. Before all NMR experiments, the sample temperature was calibrated and controlled at 36.4 ± 0.2°C using an internal ethylene glycol thermometer (80% solution of ethane-1,2-diol in dimethyl sulfoxide-d^6^). ^1^H NMR spectra were acquired, without spinning, using WET \[[@B30]\] solvent suppression, with two Carr-Purcell-Meiboom-Gill (CPMG) echoes appended, using an echo delay of 1 ms \[[@B10]\]. Chemical shifts were referenced to an external 0.1% solution of sodium trimethylsilyl-\[2,2,3,3-d~4~\]-propionate (TSP) in D~2~O. All experiments were performed with a spectral width of 5200 Hz, an acquisition time of 3.15 s, and relaxation delay of 2 s. Three acquisition schemes were used to record the one-dimensional ^1^H NMR spectra, all amounting to 128 scans. The first scheme (I) employed cycles of 16 dummy scans followed by four acquisition scans, (16,4)~32~, giving an acquisition time of 3/4 hour. In the second scheme (II), 16 dummy scans were applied once prior to acquisition 16((0,16)~8~), giving an acquisition time of 13 minutes. In the third scheme (III), 16 dummy scans and 128 acquisition scans were collected into 32 K data points, giving an acquisition time of 15 minutes. The time taken from resuspension to the start of data acquisition was typically less than 3/4 hour, and never more than 1 hour. All the data presented were recorded within 1 hour. For the first group of 13 samples (G1_13_21) in the study, the acquisition schemes (I) and (II) were used for each sample. For the second group of 17 samples (G2_17_33), all three schemes were tested for each sample. Hence, the greater number of repeat spectra is for the second group. The inclusion of multiple spectra in the analysis from the same sample tests the stability of the samples over the time of the analysis. The insensitivity of the spectra to the sampling scheme used demonstrates that the samples do not change, for example, due to sedimentation, over the timescale that a single spectrum is acquired. 2.3. PCA Analysis {#sec2.3} ----------------- In the spectral region from 1.08 to 1.20 ppm, ethanol was observed, which was probably the result of endogenous processes. However, its intensity was highly variable, even within the same cell line, so this region was excluded from the analysis. The region containing the residual water resonance signal (3.56--6.05 ppm) was also excluded. The region above 6.05 ppm contained no features of sufficient intensity for reliable quantification, given the linewidth. For this study, we chose, as descriptors, the integrals over chemical shift regions (bins) of size 0.04 ppm \[[@B12]\] which was found to produce the clearest separation of the cell types in the scores plots and the least noise in the corresponding loadings plots. Thus, the NMR spectra were reduced to 71 descriptors, with bin centres in the range 0.60--1.04, 1.24--3.56 ppm. We adopted the conventional approach \[[@B31]\] of normalisation relative to the total sum of the bin intensities in the region of interest. All the measures were implemented through writing an MATLAB (version 6.5.1, The Mathworks Inc.) code making use of the built in eigensolver. 2.4. ANN Analysis {#sec2.4} ----------------- ANNs are a sophisticated computational modelling tool, which can be used to solve a wide variety of complex problems. The attractiveness of ANNs comes from their capability to "learn" and/or model very complex systems and from the possibility of using them in classification. An ANN is a computational model formed from a certain number of single units, artificial neurons, or nodes, connected with coefficients (weights), *w*~*ij*~, which constitute the neural structure. Many different neural network architectures can be used. One of the most common is the feed forward neural network of multilayer perceptions. The network is conventionally constructed with three or more layers, that is, input, output, and hidden layers, [Figure 1](#fig1){ref-type="fig"}. Each layer has a different number of nodes. The input layer receives the information about the system (the nodes of this layer are simple distributive nodes, which do not alter the input value at all). The hidden layer processes the information initiated at the input, while the output layer is the observable response or behaviour. The inputs, input~i~, multiplied by connection weights *w*~*ij*~ are first summed and then passed through a transfer function to produce the output, out~i~. The determination of the appropriate number of hidden layers and number of hidden nodes in each layer is one of the most critical tasks in ANN design. Unlike the input and output layers, one starts with no prior knowledge of the number and size of hidden layers. The use of ANN consists of two steps: "*Training*" and "Prediction". The "*Training*" consists first of selecting input and output data for the network. This data is referred to as the training set. In the training phase, where actual data must be used, the optimum structure, weight coefficients and biases of the network are identified. Training is considered complete when the neural networks achieve the desired statistical accuracy, that is, when they produce the required outputs for a given sequence of inputs. A good criterion to find the correct network structure and therefore to stop the learning process is to minimise the root mean square (RMS) error as follows: $$\text{RMS} = \sqrt{\frac{\sum_{i = 1}^{N}{\sum_{j - 1}^{M}\left( {y_{ij} - \text{out}_{ij}} \right)^{2}}}{N \times M}},$$ where *y*~*ij*~ is the element of the matrix (*N* × *M*) for the training set or test set, and out~*ij*~ is the element of the output matrix (*N* × *M*) of the neural network, where *N* is the number of variables in the pattern, and *M* is the number of samples. RMS gives a single number, which summarises the overall error. After a supervised network performs well on training data, it is important to check its performance with data that has not been used in training. This process is called *verification*. This testing is critical to insure that the network has not simply memorised the training set but has learned the general patterns involved within an application. At this stage, other input data are submitted to the network in order to evaluate if it can predict the outputs. In this case, the outputs are already known, but they are not shown to the network. The predicted value is compared to the experimental one to see how well the network is performing. If the system does not give reasonable outputs for this test set, the training period is not over or the network is able to model the data but cannot predict them. In this work, ANN was used as a supervised method where a training data set was created from the library of NMR spectra, and the lung carcinoma classification of this training data set was known. The backpropagation method was used throughout. Firstly, the optimal ANN architecture was searched for and when the correct classification in the training phase was obtained, the usefulness of the created database and the prediction power of the networks were validated using an independent verification set. For the ANN analysis, we used 72 inputs; the 71 binned NMR intensities and the identity of the pairs of researchers (R1 and R2) as numbers 1 and 2. For output 4, nominal values were used, these identify the four cell lines, DLKP, DLKPA, DLKP-A5F, and A549, for which there were 12, 16, 14, and 12 spectra, respectively. All calculations were performed using the software Trajan Neural Network Simulator, Release 3.0 D. (Trajan Software Ltd 1996--1998, UK), on a standard PC computer running Microsoft Windows Professional XP 2000. 3. Results {#sec3} ========== 3.1. ^1^H NMR Spectroscopy of Whole Cells {#sec3.1} ----------------------------------------- A typical ^1^H NMR spectrum of intact DLKP cells is shown in [Figure 2](#fig2){ref-type="fig"}. The appearance of the spectra and the assignment suggested below are broadly similar for all the cell samples analysed. A tentative assignment which is consistent with the literature \[[@B2], [@B4], [@B32], [@B33]\] is included in the figure \[[@B24]\]. Direct quantitative analysis of the whole-cell spectra is hampered by the potential multiple contributions from different metabolites to any given resonance line by the nonlorentzian lineshapes and by the broadness of the resonance lines. The resonances in the downfield region arise from species that are at low concentration, so quantification is precluded by the sensitivity limitations of the NMR measurement. 3.2. PCA Visualization of Whole-Cell Spectra {#sec3.2} -------------------------------------------- The binned NMR spectra of the intact cells were analysed using PCA. The scores plots are shown in [Figure 3](#fig3){ref-type="fig"}. Separation of the four cell types, within each of the two data sets, is apparent using the first two PCs, demonstrating that resistance type can be classified by PCA. It also demonstrates that the samples were stable over the course of the experiment and that the spectra are insensitive to the NMR sampling scheme. Loadings analysis shows that, for each data set, the spectral regions that contribute significantly to the first two principal components are from 1.24 to 1.50 ppm, corresponding to overlapped resonances from lipid methylenes and lactate methyls, and from 2.90 to 3.40 ppm, corresponding to overlapped resonances from N-methyl signals in the choline moieties of phosphatidylcholine, phosphocholine, and glycerophosphocholine. The contribution from other spectral regions to these two principal components is marginal. Despite the fact that the same spectral regions allow separation within each data set, separation using PCA fails when the two sets of spectra are combined into one; see Supplementary Material available at doi:10.1155/2011/158094. It is apparent that, in addition to the metabolite differences of biological interest, there are subtle differences between G1_13_21 and G2_17_33 in the distribution of metabolites, which prevent classification of the entire (54 spectra) data set. The loadings analysis indicates contributions from across the spectral range, which may suggest variations in more than one metabolite. These spectral differences arise despite stringent efforts of the second group of researchers to adhere to the original experimental protocols and are reflected in the fact that there is not a simple correspondence between the orientation of the first two principal components between the two sets of spectra, [Figure 3](#fig3){ref-type="fig"}. 3.3. ANN Analysis of Whole-Cell Spectra {#sec3.3} --------------------------------------- ANN analysis consists of separate training and verification steps. For this study, we adopted the strategy of choosing multiple verification sets of spectra at random from the 54 spectra available. In training, the first aim is to find an optimal ANN architecture to enable classification of the training data set. Several architectures of three up to four layered structures were examined for this purpose. 3.4. 3-Layers Architecture {#sec3.4} -------------------------- Initially we adopted the simplest 3 layers architecture, in which case the search of the optimal architecture consists of optimising the number of nodes in the single hidden layer, effectively determining the corresponding weights, *w*~*ij*~, to minimize the RMS (root mean square error) value according to ([1](#EEq1){ref-type="disp-formula"}). For our analysis, the RMS value ceases to decrease significantly above 5 to 6 nodes, [Figure 4](#fig4){ref-type="fig"}, we therefore used networks with 6 hidden nodes for verification. This optimal architecture will be labelled (72, 6, 4), with it we obtained an RMS = 1.38 × 10^−3^. [Figure 4](#fig4){ref-type="fig"}illustrates the process of searching for the optimal network architecture. In spite of the fact that very low values for the residual mean squares were achieved using the (72, 6, 4) architecture, the appropriateness of the architecture and of the training set was then tested with various verification sets, that is, a "cross-validation" procedure was undertaken. Initially, five spectra were randomly chosen and excluded from the training set and used then as the verification set. From 10 combinations and 10 independent networks trained, in only two cases were any of the 5 spectra classified as unknown, [Table 1](#tab1){ref-type="table"}. These results are encouraging; two cases represent \~4% of the total, so for (72, 6, 4) the classification was verified as 96% successful. The failures may have arisen due to an insufficient number of spectra in the training set or because networks with three layers have insufficient complexity for 100% prediction accuracy, in this case. 3.5. 4-Layers Architecture {#sec3.5} -------------------------- We then examined networks with four layers (2 hidden). From several cases examined, it was found that four-layer ANN architectures performed similarly to simpler three layers architectures. Networks of the form (72, 4, 3, 4) or (72, 5, 4, 4) were investigated, note that the numbers in brackets refer to the number of inputs, the number of nodes in the first and in the second hidden layers, and the number of outputs. Acceptable RMS values, of 1.22 × 10^−3^ and 1.41 × 10^−3^ were obtained for (72, 4, 3, 4) and (72, 5, 4, 4), respectively, which are similar to the values obtained using the optimal three-layer architecture. Networks with the architecture (72, 4, 3, 4) performed very similarly to (72, 5, 4, 4) and require fewer unknowns (or weights, *w*~*ij*~), 312 as opposed to 396. As a result, (72, 4, 3, 4) was found to converge faster and to be less sensitive to the number of spectra excluded from training to form the verification set. In fact, we found that 5 to 10 samples could be used for verification with 100% correct classification of the spectra, see [Table 1](#tab1){ref-type="table"}. So in summary, the optimal 3- and 4-layer architectures were found to be (72, 6, 4) and (72, 4, 3, 4), respectively, [Figure 5](#fig5){ref-type="fig"}. 4. Discussion {#sec4} ============= The ^1^H NMR spectra of intact cells for both G1_13_21 and G2_17_33 have similar general appearance with severe signal overlap and line broadening. Reprojection of either data set, using PCA, demonstrates that separation by cell types is possible due to systematic differences in the lipid methylene and lactate methyl resonances and the overlapped N-methyl ^1^H nuclei of the choline-containing species \[[@B24]\]. Alterations in signal intensity and chemical shift from such cellular metabolites and biochemical intermediates have been described by other researchers in the area \[[@B6], [@B11]\]. However, because of the complex biochemical role played by these substances, we cannot ascribe a particular functional role to the findings, what is more the alterations appear to correlate and associate with particular phenotypic changes, for example, drug resistance. On the basis of the principal component analysis of either group, one could speculate that metabolite profiling by *in vivo* MRS has potential applications in monitoring the development of resistance in a given cancerous tissue. However, for the full data set such a possibility is effectively prevented by other influences on the metabolite distribution, which are comparable to, and nonorthogonal with, the "relevant" biochemical variation. We have shown that this significant obstacle can be eliminated, at least for *in vitro* studies of cell culture, by using a suitable ANN architecture. The most successful network was a four-layer structure with two hidden layers. After appropriate training, the (72, 4, 3, 4) architecture enabled 100% successful classification. Our approach may, in time, be expanded to the classification of larger data sets of spectra which have been recorded with less stringent control over sources of variance unrelated to the classification of interest. This result is encouraging and it is, to our knowledge, the first reported application of the use of ANNs specifically to correctly classify ^1^H NMR spectra in a data set when additional "nonrelevant" sources of variance are included. Other related examples of the combination of supervised and unsupervised methods include a report by Griffiths and coworkers \[[@B34]\], who obtained 85% accurate classification of meningiomas from nonmeningiomas, by initially using PCA to reduce the dimensionality of ^1^H NMR spectra recorded for tumor biopsy extracts. The first thirty PCs from this first stage of analysis were then classified using a network. More recently, the performance of lineshape fitting and quantitative ANN analyses were compared by Hiltunen et al. \[[@B35]\] for both *in vivo* and simulated ^1^H spectra. The good correlation obtained with these two approaches, for simulated data at least, suggested that ANNs have potential for quantification of *in vivo* MRS long echo time spectra. A further advantage of ANNs in the development of analysis methods for *in vivo* MRS is that they require less processing time than line fitting or other computational approaches \[[@B36]\]. Thus, our study adds to the growing number of applications of supervised techniques for exploiting the diagnostic potential of ^1^H NMR spectra for biomedical purposes. 5. Conclusions {#sec5} ============== We have found that NMR data recorded for human lung carcinoma whole-cell culture samples can be used for analysis and classification. When sources of variation not directly related to the biological state of interest (drug resistance) are minimised or kept constant, visual separation of the cell type can be achieved using unsupervised pattern recognition techniques, such as PCA. On the other hand, when this condition is not met, in our case when different researchers were responsible for cell culture and spectroscopy, successful classification of the cell type could be achieved using artificial neural networks. The experimental and ANN methodology developed are a step towards the goal of robust and reliable diagnostics based on magnetic resonance spectral data. Furthermore, as similar experimental problems may be encountered in metabolomics applications using other spectroscopic techniques, biological classification using ANNs of data sets that include "nonbiological" sources of variance may be generally possible. Supplementary Material {#supplementary-material-sec} ====================== ###### Supplementary figure 1: Scores plot for the entire data set; group 1 solid markers, group 2 open markers. ###### Click here for additional data file. The authors acknowledge the support of the Higher Education Authority of Ireland, under the Programme for Research in Third Level Institutions (PRTLI3). D. Brougham, M. Gottschalk, and G. Ivanova acknowledge the financial support of the National Institute for Cellular Biotechnology, at DCU. They would like to thank the School of Chemical Sciences for its provision of spectrometer time. J. Havel would like to acknowledge the support of the EU Erasmus/Socrates exchange program between DCU and Masaryk University and to thank the Ministry of Education and Sports of the Czech Republic, Project LC 0635. {#fig1} {#fig2} ![PCA scores plots for A549, DLKP, DLKPA, and DLKP-A5F, whole-cell data. Analysis is shown for G1_13_21 (a), G2_17_33 (b). The right hand panel is reproduced from \[[@B24]\] with permission.](JBB2011-158094.003){#fig3} {#fig4} {#fig5} ###### Results of cross-validation verification process for the three- and four-layer ANN networks. Architecture (72, 6, 4)\* ---------------------------- --------------------------------------- --------------------------------  1 2, 13, 17, 27, 38 all correct  2 21, 24, 31, 35, 51 all correct  3 4, 12, 22, 35, 44 spec. 35 classified as unknown  4 16, 17, 22, 25, 52 all correct  5 15, 16, 17, 23, 54 all correct  6 9, 15, 20, 24, 43 spec. 9 classified as unknown  7 3, 12, 15, 25, 51 all correct  8 19, 21, 43, 47, 54 all correct  9 16, 36, 37, 47, 48 all correct  10 12, 42, 44, 48, 50 all correct Architecture (72, 4, 3, 4)  1 5, 13, 20, 21, 22, 23, 24, 31, 51, 54 all correct  2 5, 8, 12, 15, 16, 29, 35, 36, 42, 49 all correct  3 8, 10, 13, 18, 23, 28, 33, 39, 40, 53 all correct  4 3, 5, 7, 9, 17, 27, 41, 45, 50, 52 all correct  5 5, 11, 16, 14, 20, 22, 24, 26, 44, 50 all correct \*where (72, 6, 4) refers to (the no. of inputs, the number of nodes in the hidden layer(s), the number of outputs). [^1]: Academic Editor: Mika Ala-Korpela

Diet is believed to be the single most important contributor to colonic carcinogenesis ([Tomatis *et al*, 1990](#bib25){ref-type="other"}). Experimental data have shown that saturated fatty acids (SFAs) and n-6 polyunsaturated fatty acids (PUFAs) have tumour-enhancing properties in the colon ([Reddy and Maeura, 1984](#bib18){ref-type="other"}; [Zhao *et al*, 1991](#bib28){ref-type="other"}, [Woutersen *et al*, 1999](#bib26){ref-type="other"}). Epidemiological data suggest that increased consumption of all meat or red meat, which contains high levels of SFAs, is strongly associated with colorectal cancer ([Giovannucci and Goldin, 1997](#bib10){ref-type="other"}; [Sandhu *et al*, 2001](#bib19){ref-type="other"}), but there is only limited evidence on the role of dietary n-6 PUFAs ([Zock and Katan, 1998](#bib29){ref-type="other"}; [Flood *et al*, 2003](#bib9){ref-type="other"}). The putative mechanism through which dietary n-6 PUFAs may enhance colonic carcinogenesis is the increased formation of prostaglandins, with the rate-limiting and committal step being mediated by the cyclooxygenase (COX)-2 enzyme ([Dubois *et al*, 1998](#bib7){ref-type="other"}). Prostaglandins possess a wide spectrum of procarcinogenic properties ([Handler *et al*, 1990](#bib11){ref-type="other"}; [Cowlen and Eling, 1993](#bib5){ref-type="other"}; [Coffey *et al*, 1997](#bib4){ref-type="other"}; [Dermott *et al*, 1999](#bib6){ref-type="other"}). We therefore hypothesised that functional *COX-2* gene polymorphisms may impact on the conversion of n-6 PUFAs into prostaglandins, with consequent change in level of cancer risk. A single nucleotide polymorphism (−*765G*\>*C)* in the promoter region of the *COX-2* gene was recently described ([Papafili *et al*, 2002](#bib16){ref-type="other"}). We therefore investigated whether this *COX-2* gene polymorphism was related to colorectal cancer risk within a population-based, prospective cohort of middle-aged and older Chinese men and women in Singapore. MATERIALS AND METHODS ===================== Study subjects -------------- The study design and subject recruitment of the Singapore Chinese Health Study have been described ([Hankin *et al*, 2001](#bib12){ref-type="other"}). Briefly, 63 257 Chinese women and men aged 45--74 years belonging to the Hokkien or Cantonese dialect group were enrolled in the study between April 1993 and December 1998. At recruitment, information on lifestyle factors and usual diet over the last year was obtained through in-person interviews. The dietary component of the questionnaire was validated through a series of 24-h food recalls ([Hankin *et al*, 2001](#bib12){ref-type="other"}). Respondents were asked to choose from predefined frequency and portion size categories for each of the 165 listed food/beverage items that he/she consumed during the past 12 months. We used the Singapore Food Composition Table to estimate average daily intake of 96 nutrient and non-nutrient compounds for each study subject ([Hankin *et al*, 2001](#bib12){ref-type="other"}). The Institutional Review Boards at the University of Southern California and the National University of Singapore had approved this study. We identified incident colorectal cancer cases through the population-based cancer registry in Singapore ([Chia *et al*, 2000](#bib3){ref-type="other"}). As of 30 April 2002, 592 colorectal cancer cases had occurred among cohort participants. All cases (including one carcinoid tumour and two *in situ* cancers) were histologically confirmed except three (ascertained by death records and clinical evidence). Details of the biospecimen collection, processing and storage procedures have been described ([Koh *et al*, 2003](#bib14){ref-type="other"}). Briefly, we attempted to collect blood and single-void urine specimens from a random 3% sample of cohort enrollees. If the subject refused to donate blood, he/she was asked to donate buccal cells. We collected blood/buccal cell samples from 1194 subjects during April 1994--July 1999. Of these subjects, 13 developed colorectal cancer by 30 April 2002, and the remaining 1181 subjects constituted the referent group for the present study. We also attempted to collect blood/buccal cell and urine samples from all incident colorectal cancer cases. Of the 592 colorectal cancer cases, 312 (53%) donated blood/buccal cell samples. COX-2 genotyping ---------------- Genomic DNA was extracted from buffy coats (228 cases and 895 controls) and buccal cell samples (84 cases and 286 controls) using a QIAamp 96 DNA Blood Kit (Qiagen, Valencia, CA, USA). A TaqMan assay for the −*765G*\>*C COX-2* polymorphism was developed using a TaqMan PCR Core Reagent kit (Applied Biosystems Inc., Foster City, CA, USA). The oligonucleotide primers for amplification of the polymorphic region of *COX-2* were GC093 for (5′-CATTAACTATTTACAGGGTAACTGCTTAGG-3′) and GC093rev (5′-CCCCCTCCTTGTTTCTTGGA-3′). In addition, the fluorogenic oligonucleotide probes (TaqMan MGB Probes; ABI) used to detect each of the alleles were GC093F (5′-CTTTCCCGCCTCTCT-3′) labelled with 6-FAM to detect the *G* allele and GC093V (5′-CTTTCCCCCCTCTCT-3′) labelled with VIC to detect the *C* allele. Experimental samples were compared to 12 controls to identify the three genotypes at each locus (*GG, GC, CC*). All samples were processed without knowledge of their case/control status. Any samples that were outside the parameters defined by the controls were identified as noninformative and were retested. Four controls and two cases had noninformative *COX-2* genotypes and were excluded from the present analysis. Statistical analysis -------------------- Data were analysed by standard methods for unmatched case--control studies ([Breslow and Day, 1980](#bib1){ref-type="other"}). Unconditional logistic regression models were used to examine the associations between *COX-2* genotypes and risk of colorectal cancer, and their possible modification by n-6 PUFA intake. The associations were measured by odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) and *P*-values (two-sided). Limited by the very low frequency of the *CC* genotype (0.003), the *GC* and *CC* genotypes were combined when compared with the *GG* genotype. All ORs were adjusted for age (year) at recruitment, year of recruitment, gender, dialect group (Cantonese, Hokkien), level of education (no formal schooling, primary school, secondary school and higher), body mass index (\<20, 20 to \<24, 24 to \<28, 28+ kg m^−2^), smoking status (never, exsmoker, current smoker), frequency of alcohol consumption (nondrinker, monthly drinker, weekly drinker, daily drinker), and familial history of colorectal cancer (yes, no). RESULTS ======= Of the 592 incident colorectal cancer cases, 282 were excluded from the present analysis due to unavailable blood/buccal cell samples (*n*=280) or noninformative *COX-2* genotype (*n*=2). Cases included in the present study (*n*=310) were comparable to those excluded in terms of age (mean: 65.4 *vs* 66.1 years), but slightly different in gender (57 *vs* 49% male), dialect group (45 *vs* 37% Cantonese) and level of education (69 *vs* 60% attaining primary school education or higher). In total, 180 (58%) cases had colon cancer, and the remaining cases had either rectal or rectosigmoid cancers. [Table 1](#tbl1){ref-type="table"} Table 1Selected characteristics of colorectal cancer cases and controls, the Singapore Chinese Health Study**CharacteristicsControls (*n*=1177)Cases (*n*=310)*P*-value^a^** Mean age±s.d.^b^ (years)56.5±8.161.3±7.5\<0.001     Number (%) *Sex*    Males509 (43.2)178 (57.4)\<0.001 Females668 (56.8)132 (42.6)     *Dialect group*    Cantonese571 (48.5)138 (44.5)0.23 Hokkien606 (51.5)172 (55.5)     *Level of education*    No formal schooling318 (27.0)95 (30.6)0.01 Primary school504 (42.8)151 (48.7)  Secondary school288 (24.5)54 (17.4)  A level/university67 (5.7)10 (3.2)     *Body mass index* (*kg* *m*^−*2*^)    ⩽20187 (15.9)46 (14.8)0.04 20−\<24659 (56.0)151 (48.7)  24−\<28267 (22.7)92 (29.7)  28+64 (5.4)21 (6.8)     *Cigarette smoking*    Never smokers853 (72.5)184 (59.4)\<0.001 Former smokers131 (11.1)52 (16.8)  Current smokers193 (16.4)74 (23.9)     *Frequency of alcohol consumption*    Nondrinkers964 (81.9)243 (78.4)0.14 Monthly85 (7.2)21 (6.8)  Weekly93 (7.9)29 (9.4)  Daily35 (3.0)17 (5.5)     *Familial history of colorectal cancer*    No1149 (97.6)297 (95.8)0.12 Yes^c^28 (2.4)13 (4.2)     *Median (5th percentile, 95th percentile)*Total calories (kcal day^−1^)1483.5 (829.9, 2477.5)1494.0 (819.3, 2589.2)0.54Total fat (g day^−1^)40.5 (19.6, 81.7)39.8 (19.0, 77.1)0.66SFAs (g day^−1^)14.0 (5.9, 30.0)13.7 (5.8, 29.2)0.62MUFAs (g day^−1^)13.6 (6.3, 28.0)13.4 (6.1, 26.5)0.88PUFAs (g day^−1^)8.1 (3.3, 18.6)7.9 (3.4, 16.5)0.74N-3 PUFAs (g day^−1^)0.8 (0.4, 1.7)0.8 (0.4, 1.7)0.94N-6 PUFAs (g day^−1^)7.2 (2.9, 16.8)7.0 (3.0, 14.9)0.75Fiber (g day^−1^)12.2 (5.0, 23.1)11.9 (4.3, 22.5)0.83Calcium (mg day^−1^)373.5 (156.7, 853.5)363.4 (158.8, 773.6)0.56Folate (*μ*g day^−1^)146.0 (64.1, 277.7)147.1 (57.9, 286.9)0.65[^1][^2][^3] shows the distributions of selected characteristics of colorectal cases and controls. Cases were older, less educated, more obese and more likely to smoke cigarettes than controls. Intakes of total calories, total fat, SFAs, monounsaturated fatty acids (MUFAs), PUFAs, n-3 PUFAs, n-6 PUFAs, fibre, calcium or folate were comparable between cases and controls. Among control subjects, the *G* and *C* allele frequencies of the *COX-2* genotype were 0.952 and 0.048, respectively, and the *GG*, *GC* and *CC* genotype frequencies were 0.907, 0.090 and 0.003, respectively. These genotypic distributions were in Hardy--Weinberg equilibrium (*P*=0.43). Overall, there was no association between colorectal cancer risk and *COX-2* −*765G*\>*C* genotype or n-6 PUFA intake ([Table 2](#tbl2){ref-type="table"} Table 2*COX-2* −*765G*\>*C* genotype and dietary intake of n-6 PUFAs in relation to risk of colorectal cancer, the Singapore Chinese Health Study  **Colorectal cancerColon cancerRectal cancer** **Controls**align=\"center\"**CasesOR (95% CI)^a^CasesOR (95% CI)^a^CasesOR (95% CI)^a^***COX-2 genotype*        GG10672731.001551.001181.00 GC/CC110371.18 (0.77--1.79)251.50 (0.92--2.47)120.87 (0.45--1.66)        *Dietary n-6 PUFAs in quartile*^b^        1st (low)260701.00421.00281.00 2nd297831.08 (0.72--1.60)481.15 (0.70--1.87)350.97 (0.55--1.71) 3rd312721.00 (0.65--1.53)451.09 (0.64--1.83)270.82 (0.44--1.54) 4th (high)308851.04 (0.63--1.70)451.04 (0.56--1.92)400.99 (0.49--1.99)[^4][^5]). When subjects were stratified into high (above median) *vs* low (below median) intake levels of n-6 PUFAs, a borderline statistically significant association between genotype and risk was observed among high consumers of n-6 PUFAs (OR=1.65, 95% CI=0.95--2.87), which was mainly confined to colon cancer (OR=2.38, 95% CI=1.23--4.59) ([Table 3](#tbl3){ref-type="table"} Table 3*COX-2* −*765G*\>*C* genotype in relation to risk of colorectal cancer stratified by level of dietary n-6 PUFAs, the Singapore Chinese Health Study  **Colorectal cancerColon cancerRectal cancer** **ControlsCasesOR (95% CI)^a^CasesOR (95% CI)^a^CasesOR (95% CI)^a^***Low dietary n-6 PUFAs*^b^        GG5081391.00811.00581.00 GC/CC49140.82 (0.42--1.59)90.95 (0.43--2.09)50.68 (0.25--1.89)        *High dietary n-6 PUFAs*^b^        GG5591341.00741.00601.00 GC/CC61231.65 (0.95--2.87)162.38 (1.23--4.59)71.09 (0.46--2.59)[^6][^7]). There was no association between genotype and rectal cancer risk regardless of dietary n-6 PUFA intake levels. There was indication of an interaction effect between *COX-2* genotype and dietary n-6 PUFAs in colon cancer (*P*=0.07), which was absent in rectal cancer (*P*=0.51). The corresponding *P*-value for the gene--diet interaction effect in colorectal cancer combined was 0.10. DISCUSSION ========== In this cohort of Singapore Chinese, we reported a statistically significant effect of the *COX-2* −*765G*\>*C* gene polymorphism on colon cancer risk among subjects with high intake of dietary n-6 PUFAs. Our data support the hypothesis that COX-2 exerts its effects on colon carcinogenesis through its influence on prostaglandin synthesis from n-6 PUFAs. The current study has several strengths. (1) Our prospective study design precludes the possibility of recall bias. Furthermore, reliable dietary nutrient estimates including n-6 PUFAs were assessed using a validated food frequency questionnaire. (2) The nationwide cancer registry has been in place in Singapore since 1968 ([Parkin *et al*, 2002](#bib17){ref-type="other"}), and migration out of Singapore has been negligible since inception of the study. This relatively complete ascertainment of cancer and death outcomes eliminates a concern for potential selection bias. (3) Study subjects originated from two contiguous regions in Southern China, resulting in a genetically homogeneous study population that facilitated the investigation of gene--disease associations. (4) Exposure information on other known/suspected risk factors for colorectal cancer was collected and accounted for in all analyses of gene--diet--cancer risk associations. The chief limitation of our study is the lack of information on use of COX-2 inhibitors, which may bias the effect of *COX-2* genotype on risk. However, if use of COX-2 inhibitors were to exert a confounding effect on the observed COX-2 genotype/colon cancer association, our inability to control for such confounding is likely to lead to an underestimation, rather than an overestimation, of the risk associated with the putative high-activity genotype. This is because use of COX-2 inhibitors is likely to be more common among subjects with more severe symptoms of inflammation, possibly due to the possession of the high activity *COX-2* genotype. Another weakness of the present study is our relatively small number of cancer cases, which may result in less precise estimation of risk factor--disease associations. The major n-6 PUFA in most diet is linoleic acid, the precursor of arachidonic acid. The latter is consequently converted to various prostaglandins, and COX is the crucial and rate-limiting enzyme for this conversion. There is compelling evidence that prostaglandins play important roles in colorectal carcinogenesis by enhancing cell proliferation and growth, promoting angiogenesis and inhibiting apoptosis ([Cao and Prescott, 2002](#bib2){ref-type="other"}; [Stoehlmacher and Lenz, 2003](#bib23){ref-type="other"}). *COX-2* gene expression and its mRNA and protein levels were markedly elevated in most human colorectal cancers relative to adjacent normal mucosa ([Kargman *et al*, 1995](#bib13){ref-type="other"}; [Sano *et al*, 1995](#bib20){ref-type="other"}). It is hypothesised that the COX-2-associated effect on colorectal carcinogenesis is due to the increased production of prostaglandins from dietary n-6 PUFAs ([Eberhart and Dubois, 1995](#bib8){ref-type="other"}). In support of this hypothesis, high dietary n-6 PUFAs has been shown to promote colon tumorigenesis by upregulating COX-2 expression in animal studies ([Singh *et al*, 1997](#bib22){ref-type="other"}). The human *COX-2* gene is mapped to chromosome 1q25.2--q25.3 and spans about 8.3 kb pairs with 10 exons ([Kosaka *et al*, 1994](#bib15){ref-type="other"}). Previous studies on the 5′ flanking region of the human *COX-2* gene show that this region contains a canonical TATA box as well as several putative factor elements that are critical in inducing *COX-2* gene transcription, such as Sp1, NF-*κ*B, GRE (glucocorticoid) and IRE (insulin) elements ([Tazawa *et al*, 1994](#bib24){ref-type="other"}; [Yang *et al*, 1997](#bib27){ref-type="other"}). The region from nucleotide −827 to −454 has been described as a negative region since deletion of this region led to increased luciferase activity in reporter expression studies. The −*765G*\>*C* mutation lies within this region, and is also within one of the five putative Sp1 elements ([Yang *et al*, 1997](#bib27){ref-type="other"}). At present, data on the functionality of the −*765G*\>*C* polymorphism and the direction/magnitude of change in protein expression/activity between the *C* and *G* alleles are limited and mixed ([Papafili *et al*, 2002](#bib16){ref-type="other"}; [Schneider *et al*, 2003](#bib21){ref-type="other"}). In summary, the present study provides the first epidemiological evidence for a possible cause-and-effect connection between the production of prostaglandins from n-6 PUFAs through the enzymatic activity of COX-2, and increased risk of tumour development in the colon. Our novel findings require confirmation in larger studies with varying levels of substrate intake and genotype frequency. We thank Ms Siew-Hong Low of the National University of Singapore for supervising the field work of the Singapore Chinese Health Study, and Ms Kazuko Arakawa of the University of Southern California for the development and management of the cohort study database. The Singapore Chinese Health Study has been supported by Grants R01 CA55069, R35 CA53890 and R01 CA80205 from the National Cancer Institute, Bethesda, MD. [^1]: Two-sided *P*-values derived from *t*-test (for age), *χ*^2^ test (for categorical variables) or Wilcoxon rank sum test (for nutrient intakes). [^2]: Mean age at recruitment into the cohort, s.d.=standard deviation. [^3]: Any of first-degree relatives had colorectal cancer. [^4]: ORs were adjusted for age at recruitment, year of recruitment, gender, dialect group, level of education, body mass index, smoking status, frequency of alcohol consumption, and familial history of colorectal cancer; CI=confidence interval. [^5]: In addition to all variables listed above, OR were adjusted for total energy intake. [^6]: ORs were adjusted for age at recruitment, year of recruitment, gender, dialect group, level of education, body mass index, smoking status, frequency of alcohol consumption, and familial history of colorectal cancer; CI=confidence interval. [^7]: Defined as less than or equal to the median ('low') and greater than the median ('high') intake level (6.96 g day^−1^) of dietary n-6 PUFAs among all cohort members.

1. Introduction {#sec1} =============== Stroke is the second most common cause of death in developed countries and thus is a major health problem \[[@B1]\]. According to the 2009 Annual Public Health Report by the Korean National Statistical Office, cerebrovascular disease, or stroke, was the second-leading cause of disease-related deaths in Korea, after cancer \[[@B2]\]. In Korea, many stroke patients receive traditional medical care because the country has its own system of traditional alternative medicine called Traditional korean medicine (TKM), the role of which has been emphasized in stroke management \[[@B3]\]. The Korean medical diagnosis system has unique characteristics similar to the traditional Chinese medical diagnosis system. One such feature is pattern identification (PI), which is based on information obtained from four diagnostic processes including inspection, listening and smelling, inquiry, and palpation \[[@B4]\]. PI is a diagnostic system that entails a comprehensive analysis of symptoms and signs, with implications for determining the cause, nature, and location of the illness, the patient\'s physical condition, and the patient\'s treatment \[[@B3], [@B5]\]. The inspection process involves the examination of the patient\'s symptoms or disease by observing his or her shape, expression, and tongue \[[@B6]\], among others. Observation of the tongue, also known as tongue diagnosis, is an important procedure in diagnosis by inspection in TKM. The status of the tongue is an important indicator in the diagnosis of one\'s condition, including the physiological and clinicopathological changes of internal organs in the body \[[@B7]\]. A number of studies have shown that tongue diagnosis plays an important role in clinical prognosis and treatment \[[@B8]--[@B15]\], specifically in patients with a history of stroke. However, the clinical competence of tongue diagnosis was determined by the experience and knowledge of the clinicians who used tongue diagnosis. Environmental factors, such as differences between light sources and levels of brightness also had significant influences on clinicians and their diagnostic decisions using the tongue. Unfortunately, much of the experiences in traditional tongue diagnosis have not been verified scientifically or quantitatively. Therefore, it is necessary to build an objective diagnostic standard for tongue diagnosis \[[@B7]\]. We investigated the reliability of TKM tongue diagnosis in stroke patients by evaluating interobserver reliability regarding tongue indicators as achieved by TKM practitioners. 2. Methods {#sec2} ========== 2.1. Study Subjects {#sec2.1} ------------------- The data for this analysis were collected as part of the project named the Fundamental Study for the Standardization and Objectification of Pattern Identification in TKM for Stroke (SOPI-Stroke). Stroke patients admitted to the following oriental medical university hospitals, Kyung Hee Oriental Medical Center (Seoul), Kyung Hee East-West Neo Medical Center (Seoul), Dong Guk International Hospital (Kyunggi-do), Kyung Won Oriental Medical Hospital (Incheon), Dae Jeon Oriental Medical Hospital (Daejeon), Dong Sin Oriental Medical Hospital (Gwangju), Won Kwang Oriental Medical Hospital (Jeollabuk-do), Dae Gu Hanny University Medical Center (Daegu), and Sang Ji Oriental Medical Hospital (Gangwon-do), participated in this study between February 2010 and December 2010 ([Figure 1](#fig1){ref-type="fig"}). All patients provided written informed consent under procedures approved by institutional review boards (IRB). Eligibility inclusion criteria were that participants had to be enrolled within 30 days of the onset of their symptoms as confirmed by imaging diagnosis, such as computerized tomography (CT) or magnetic resonance imaging (MRI). Exclusion criteria were traumatic stroke patients, such as those with subarachnoid, subdural, or epidural hemorrhage. 2.2. Data Processing and Analysis {#sec2.2} --------------------------------- All patients were seen by two experts from the same department in each site, who were well trained in standard operation procedures (SOPs) \[Appendix\] and were subjected to an examination of the status of the tongue, tongue color (pale, red, and bluish purple), fur color (white fur, yellow fur), fur quality (thick fur, thin fur, moist fur, and dry fur), and special tongue appearance (teeth marked, enlarged, mirror, and spotted). The examination parameters were extracted from parts of a case report form (CRF) for the standardization of stroke diagnosis that had been developed by an expert committee organized by the Korean Institute of Oriental Medicine (KIOM). These assessments were given individually without discussions among the clinicians. Descriptions of the grading severity for each variable were scored as the following: 1 = very much so, 2 = Much so, and 3 = Not so much. In particular, the clinicians had to measure the stroke PI of each patient following the fire-heat pattern, the phlegm-dampness pattern, the blood stasis pattern, the qi deficiency pattern, and the Yin deficiency pattern, as suggested by the KIOM \[[@B3], [@B16]--[@B18]\]. Interobserver reliability was measured in three ways, using simple percentage agreements, Cohen\'s kappa coefficient and Gwet\'s AC~1~ statistic, as well as via the corresponding confidence intervals (CI). Kappa, the preferred measure of rater reliability for nominal data, measures the reliability of agreement between two or more independent raters using a rating scheme with mutually exclusive categories. In general, definitive kappa interpretations have been proposed \[[@B19]--[@B24]\]. For most purposes, however, values ≤0.40 represent poor agreement, values between 0.40 and 0.75 represent moderate to good agreement, and values ≥0.75 indicate excellent agreement \[[@B24]\]. The AC~1~ statistic is not vulnerable to the well-known paradoxes that make Kappa appear ineffective \[[@B25]--[@B27]\]. First, interobserver reliability for the tongue indicator among all subjects was calculated via simple percentage agreements, Cohen\'s kappa coefficient, and Gwet\'s AC~1~ statistic. Later, interobserver reliability regarding PI with same opinions between the raters was calculated in the same way. The blood stasis pattern was omitted because the sample size was too small (*n* = 1). The data were statistically analyzed with SAS software, version 9.1.3 (SAS Institute Inc., Cary, NC). 2.3. Ethical Approval {#sec2.3} --------------------- This study was approved by institutional review board of the KIOM and by each of the oriental medical university hospitals. 3. Results {#sec3} ========== A total of 658 stroke patients were enrolled in the study. Thirty patients were excluded from analysis due to PI omitted by any one of 2 TKM clinicians. The interobserver reliability results regarding tongue indicators for all subjects (*n* = 628) are shown in [Table 1](#tab1){ref-type="table"}. The kappa measure of agreement between the two experts was generally moderate to good for the tongue indicators, ranging from 0.42 to 0.69, except for moist fur (*κ* = 0.29) and spotted (*κ* = 0.37). Moreover, the AC~1~ measure of agreement between the two experts was generally high for the tongue indicators, ranging from "moderate" (AC~1~ = 0.43) to "excellent" (AC~1~ = 0.97). Agreement, as assessed by the kappa values, was considerably lower than the AC~1~ values in most cases. The results of interobserver reliability for subjects of PI with the same opinion between the raters are shown in [Table 2](#tab2){ref-type="table"}. A total of 451 stroke patients received PI with the following same resulting opinions by the raters: Fire-Heat Pattern (*n* = 147), Phlegm-Dampness Pattern (*n* = 158), Yin Deficiency Pattern (*n* = 80), and Qi Deficiency Pattern (*n* = 66). The blood stasis pattern was excluded because the sample size was too small (*n* = 1). The kappa measure of agreement for the subjects of PI was generally moderate to good for the tongue indicators, ranging from 0.40 to 0.72, except for moist fur (*κ* = 0.31). Moreover, the AC~1~ measure of agreement between the two experts was generally high for the tongue indicators, ranging from "moderate" (AC~1~ = 0.5) to "excellent" (AC~1~ = 0.98) ([Table 2](#tab2){ref-type="table"}). 4. Discussion {#sec4} ============= Inspection of the tongue in TKM diagnosis, as well as in western medicine \[[@B28]\], is one of the most important approaches for obtaining significant evidence in diagnosing the patient\'s health conditions \[[@B7]\]. It is used to observe the color, coating, and body of the tongue, among other features, in rendering a disease diagnosis. Also, as tongue diagnosis has played a prominent role in the diagnosis and subsequent treatment of stroke patients, it has attracted an increasing amount of attention in oriental medicine \[[@B8]--[@B15]\]. Park et al. \[[@B12]\] analyzed markers that classified tongue body color, fur, fur quality, dryness, and shape to standardize tongue diagnosis and PI for stroke patients. Choi et al. \[[@B14]\], to assess the usefulness of tongue diagnosis in evaluating PI, observed the coating of the tongue and compared it with PI in acute stroke stage patients within 72 hours from the onset of stroke. In his study, a red tongue was significantly related to the fire-heat pattern and the yin deficiency pattern, while a faint white tongue was related to the phlegm-dampness pattern. Thin fur was related to the Wind and fire-heat pattern, and thick fur was related to the phlegm-dampness and blood stasis patterns. Another study \[[@B15]\] by the same author found that a stroke patient\'s motor recovery might be related to tongue diagnosis. In Kim et al.\'s recent study \[[@B3]\], the authors attempted to standardize the oriental medical PI for stroke patients using logistic regressions. An interesting finding was that all of the patterns in their study basically included tongue and pulse diagnoses in their final equations. This result shows that TKM clinicians tongue and pulse diagnosis are seriously considered in their patient management. However, traditional tongue diagnosis does have its inevitable limitations because the clinical skill involved in tongue diagnosis depends on the clinician\'s experience and knowledge as well as on environmental factors that can exert a significant influence on the diagnostic results. Therefore, it is necessary to build an objective diagnostic standard for tongue diagnosis. To date, only a few studies have reached wide consensus among TKM clinicians, while many studies have investigated agreement measures for western medical diagnosis \[[@B29]\]. An evaluation of interobserver reliability is important when one is interested in the "true" differences among observers that often report different values for the same quantity. In other words, interobserver reliability, rather than the total observer reliability, should be used to explore the causes of the disagreements among observers. The total observer reliability masks these sources of disagreement because it contains both interobserver reliability (true differences) and intraobserver reliability (random error among the observations) made by the same observer for the same subject \[[@B30]\]. Li et al. \[[@B32]\] used the kappa value to evaluate the consistency of tongue and pulse signs for 55 patients as observed by traditional Chinese medicine (TCM) clinicians. Zhang et al. \[[@B33]\] analyzed the effect of training on improved agreement in TCM diagnosis among its practitioners based on a sample of 42 patients with rheumatoid arthritis (RA). Mist et al. \[[@B34]\], using interrater correlations and kappa values, assessed whether a training process that focused on a questionnaire-based diagnosis would improve agreement in traditional Chinese medicine TCM diagnosis. Finally, Kim et al. \[[@B35]\] examined the reliability of TCM tongue inspection by evaluating inter- and intrapractitioner agreement levels for specific tongue characteristics. The data for this analysis were collected as parts of a multicenter study of standardization of stroke diagnosis in Korea. In this study, the evaluation of interobserver reliability in tongue status in stroke patients, as achieved by TKM clinicians, as well as interobserver reliability in all subjects (or subjects of PI with same the opinions between the raters), was calculated as simple percentage agreements, kappa values and AC~1~ measures. When investigating agreement between observers, clinicians have long used kappa and other chance-adjusted measures. A commonly used scale used to interpret kappa derives from the work of Landis and Koch in 1977 \[[@B20]\]. However, the appropriateness of kappa as a measure of agreement has recently been debated \[[@B26], [@B27]\]. A relatively new statistic, the AC~1~, has been suggested by Gwet to adjust for chance in agreement studies \[[@B25], [@B31]\]. According to our results, interobserver agreement in tongue diagnosis between the raters was generally moderate to good. The AC~1~ measure of agreement between the two experts was generally moderate to good for the tongue indicators, ranging from 0.43 to 0.97. In particular, the AC~1~ measure of agreement was nearly perfect in mirror, spotted, and bluish purple tongue. These tongue indicators are certain signs of special tongue appearance. Mirror tongue means that the surface of tongue is smooth and shiny like a mirror, without fur. Spotted tongue means that there are purple spots on the whole tongue and bluish purple tongue means that the color of tongue body is bluish purple, or bluish purple spots appear on the surface of the tongue. It is thought that a description of these indicators is relatively objective, that agreement is very high. Whereas the AC~1~ of pale, thin fur, moist fur is lower than 0.5. The reason why the AC~1~ of these indicators is low is that perception of quality and color of tongue vary markedly from person to person. Therefore, it is necessary to improve the validity and reliability of tongue diagnosis through the development of detail-oriented criteria and enhanced training of clinicians. One limitation of our study is the fact that we did not analyze the impact of the each site participated in this study. All patients were allocated into two experts among the eighteen clinicians in each site. While the large number of clinicians who participated in the study increased the generalizability of the results, it is possible that the variety of experiences offered by these clinicians was another limitation to the study. All clinicians who had at least more than three years of clinical experience in the field took regular SOPs training twice a year, therefore we assumed all clinicians have equal ability to take information from the patients. But, in reality it is certain that this assumption is not true. We will consider the subject carefully in the future work. Furthermore, this study has a limitation in that the actual diagnosis process in TKM is carried out not only through tongue diagnosis but also through other three diagnostic processes. Further studies may be necessary to conduct a comprehensive analysis considering all the four diagnostic processes. Tongue diagnosis is a very important diagnostic procedure in TKM, despite its inevitable limitations associated with clinician experience and knowledge. However, this study shows that interobserver reliability in tongue status in stroke patients between the raters was considerably high. This may help to alleviate the lack of objectivity and reproducibility in tongue diagnosis in TKM. We expect that future studies will help to further establish tongue diagnosis as a useful oriental medicine diagnostic tool in the clinical management of stroke patients. M. M. Ko, and J. A. Lee contributed equally to this work. This research was supported by a Grant from the Korea Institute of Oriental Medicine (K11131). {#fig1} ###### Agreement between raters for all subjects. Variables \% Agreement Kappa (*κ*) CI of *κ* AC~1~ CI of AC~1~ ---------------------------- -------------- ------------- -------------- ------- -------------- Tongue color:  Pale 71.38 0.42 (0.35, 0.49) 0.43 (0.36, 0.51)  Red 75.84 0.51 (0.44, 0.58) 0.52 (0.45, 0.59)  Bluish purple 91.99 0.42 (0.35, 0.49) 0.9 (0.87, 0.93) Fur color:  White fur 75.17 0.49 (0.42, 0.56) 0.51 (0.44, 0.59)  Yellow fur 85.29 0.69 (0.63, 0.75) 0.71 (0.66, 0.78) Fur quality:  Thick fur 81.15 0.60 (0.54, 0.67) 0.63 (0.57, 0.70)  Thin fur 74.61 0.49 (0.41, 0.56) 0.49 (0.42, 0.57)  Moist fur 70.27 0.29 (0.21, 0.37) 0.49 (0.42, 0.57)  Dry fur 80.5 0.48 (0.40, 0.56) 0.68 (0.63, 0.75) Special tongue appearance:  Teeth marked 87.68 0.46 (0.36, 0.56) 0.84 (0.80, 0.88)  Enlarged 89.63 0.51 (0.41, 0.61) 0.86 (0.83, 0.90)  Mirror 97.44 0.60 (0.42, 0.78) 0.97 (0.95, 0.99)  Spotted 96.96 0.37 (0.15, 0.59) 0.96 (0.95, 0.98) CI: confidence interval. ###### Agreement measures in PI with the same opinions between the raters. Variables \% Agreement Kappa (*κ*) CI of *κ* AC~1~ CI of AC~1~ ---------------------------- -------------- ------------- -------------- ------- -------------- Tongue color:  Pale 75.00 0.49 (0.41, 0.58) 0.51 (0.42, 0.59)  Red 76.67 0.53 (0.45, 0.61) 0.54 (0.46, 0.62)  bluish purple 94.09 0.57 (0.42, 0.72) 0.93 (0.90, 0.96) Fur color:  White fur 77.05 0.52 (0.44, 0.60) 0.56 (0.48, 0.64)  Yellow fur 86.37 0.71 (0.64, 0.78) 0.74 (0.68, 0.81) Fur quality:  Thick fur 83.25 0.65 (0.58, 0.73) 0.68 (0.61, 0.75)  Thin fur 75.18 0.50 (0.42, 0.58) 0.51 (0.42, 0.59)  Moist fur 70.53 0.31 (0.22, 0.40) 0.5 (0.41, 0.58)  Dry fur 82.46 0.52 (0.42, 0.61) 0.72 (0.66, 0.79) Special tongue appearance:  Teeth marked 89.06 0.53 (0.42, 0.64) 0.86 (0.82, 0.90)  Enlarged 90.67 0.57 (0.46, 0.69) 0.88 (0.84, 0.92)  Mirror 97.99 0.72 (0.54, 0.89) 0.98 (0.96, 0.99)  Spotted 96.88 0.40 (0.15, 0.65) 0.97 (0.95, 0.99) PI: pattern identification; CI: confidence interval. [^1]: Academic Editor: Wolfgang Weidenhammer

Introduction {#Sec1} ============ Small heat-shock proteins (sHSPs) are a diverse family of proteins that share a conserved ≈ 90-residue α-crystallin domain (ACD) that is flanked by variable N- and C-terminal regions (Basha et al. [@CR3]; Hilton et al. [@CR17]; McHaourab et al. [@CR28]). Although sHSPs are relatively small as monomers (12 to 42 kDa), the majority assemble into large oligomers. These range in size from 12 to \> 40 subunits, with some family members being monodisperse and others forming polydisperse ensembles (Basha et al. [@CR3]; Hilton et al. [@CR17]; McHaourab et al. [@CR28]). Found in all kingdoms of life, many sHSPs have been demonstrated in vitro to act as ATP-independent molecular chaperones with the ability to capture denaturing proteins in a partially unfolded form such that they can be reactivated by the cell's ATP-dependent chaperones. Recent reviews have described models for this canonical mechanism of sHSP chaperone action; however, details are derived primarily from in vitro studies with recombinant proteins and model interactors from non-homologous organisms (Haslbeck and Vierling [@CR16]; Treweek et al. [@CR40]). Thus, a major gap in our understanding of sHSP mechanism is the considerable lack of information about which substrates they protect in the cell. In order to investigate the properties of proteins that are sHSP interactors, we identified HSP16.6 from the single-celled cyanobacterium *Synechocystis* sp. PCC 6803 (hereafter *Synechocystis*) as an ideal system to interrogate. HSP16.6 is the only sHSP in *Synechocystis* (Giese and Vierling [@CR12]; Lee et al. [@CR25]). It is strongly induced at high temperature, and cells deleted for HSP16.6 (Δ16.6) grow normally at optimal growth temperature but are sensitive to heat stress (Giese and Vierling [@CR12], [@CR13]). The temperature-sensitivity phenotype of Δ16.6 cells has enabled studies of sHSP properties required for activity in vivo in a homologous system. Crucially, point mutations in the N-terminal domain were found to decrease heat tolerance in vivo, but to have no effect on the efficiency of chaperone function in assays with model substrates in vitro (Giese et al. [@CR14]). This observation emphasizes the need to identify native interactors of sHSPs and renders *Synechocystis* an excellent system with which to do so. We previously used immunoprecipitation and mass spectrometry (MS)-based proteomics to identify 13 proteins associated in vivo with HSP16.6 from *Synechocystis* cells that had been heat-stressed prior to cell lysis (Basha et al. [@CR2]). Notably, these 13 proteins were not detected in equivalent pull-downs from cells that had not been heat-stressed, or when recombinant HSP16.6 was added to heat-stressed Δ16.6 cells before lysis (to control for sHSP-protein interactions that might occur in the lysate, as opposed to during heat stress in vivo). Although these proteins were associated with the sHSP in the soluble cell fraction, they were also found in the insoluble cell fraction after heat stress (Basha et al. [@CR2]). All of these proteins, whose functions span a variety of cellular processes, including translation, transcription, secondary metabolism, and cell signaling, could be released from the immunoprecipitate by addition of DnaK, co-chaperones, and ATP (Basha et al. [@CR2]). In addition, one of these interactors, a serine esterase, when purified, was shown to be heat sensitive and to associate with HSP16.6 and thereby be protected from insolubilization (Basha et al. [@CR2]). While these data identified 13 proteins as potential interactors for canonical sHSP chaperone function, their relatively small number meant it was not possible to derive any common protein features that might dictate interaction with the sHSP. Here, we have extended the identification of HSP16.6-interactors to a total of 83 proteins by performing an affinity pull-down from heat-stressed *Synechocystis*. By performing rigorous bioinformatic analyses, we provide new insights into the primary and secondary structural properties of proteins that interact with sHSPs in the soluble cell fraction during stress. We also catalogue the functions of the interactors and compare these to sHSP interactors previously identified in two other prokaryotes, *Escherichia coli* and *Deinococcus radiodurans* (Bepperling et al. [@CR4]; Fu et al. [@CR10]). Our combined results indicate that sHSPs protect a specific yet diverse set of proteins from aggregation in the cell. Methods {#Sec2} ======= Affinity isolation of HSP16.6-interacting proteins {#Sec3} -------------------------------------------------- Isogenic *Synechocystis* strains were used in which the wild-type HSP16.6 gene had been replaced with a spectinomycin resistance gene (*aadA* gene) (ΔHSP16.6 strain) or with the spectinomycin gene and HSP16.6 carrying a Strep-tag II affinity tag (WSHPQFEK) on the C-terminus (HSP16.6-Strep strain) (Basha et al. [@CR2]). This HSP16.6-Strep strain had been shown previously to behave like wild type in assays of heat tolerance (Basha et al. [@CR2]), and recombinant HSP16.6-strep protein was equivalent to untagged protein in assays of chaperone activity in vitro (Friedrich et al. [@CR9]). Cells were grown in 50-mL cultures at 30 °C as described previously to *A*~730~ ≈ 0.2 (Basha et al. [@CR2]) and then subjected to treatment at 42 °C for 2 h followed by 1 h recovery at 30 °C, to allow accumulation of HSP16.6-Strep protein. Control samples were prepared directly after this treatment, while heat-stressed samples were treated for an additional 30 min at 46 °C. To control for interaction of HSP16.6-Strep protein during sample processing, recombinant HSP16.6-Strep protein was added to heat-stressed samples of the ΔHSP16.6 strain directly after heat treatment at a concentration matching that in heat-stressed cells. Cells were harvested, suspended in 1.5 mL lysis buffer (25 mM HEPES-KOH, 0.2 M NaCl, 0.5% Triton X-100, 5 mM ϵ-aminocaproic acid, 1 mM benzamidine, 1 μg mL^−1^ leupeptin, and 1 mM EDTA, pH 7.5), and opened as described previously (Basha et al. [@CR2]). The soluble fraction was mixed with 30 μL of Strep-Tactin resin (Sigma) at 4 °C for 2 h. Resin was washed six times in lysis buffer, and bound proteins were eluted using either sample buffer (for SDS-PAGE) or isoelectric focusing (IEF) rehydration buffer (for 2D gels) (7.0 M urea, 2.5 M thiourea, 2% CHAPS, 2% IPG buffer pH 3--10 NL (Amersham Biotech), and 3 mg mL^−1^ dithiothreitol). For 2D gel analysis, pH 3--10 NL first dimension strips (18 cm; Amersham Biotech) were rehydrated overnight at room temperature using 600 μL of sample in IEF rehydration buffer. IEF was carried out for 2 h at 150 V, 2 h at 300 V, 5 h at 500 V, and 7 h at 3500 V. The second dimension was separated by 11--17% SDS-PAGE for 30 min at 15 mA and then for 7 h at 25 mA. Samples were also separated by SDS-PAGE according to standard protocols, using 8% acrylamide gels in order to afford good separation of proteins above 100 kDa, which are typically not well resolved on the 2D system. Gels were silver stained according to a previous protocol (Rabilloud [@CR33])*.* Protein identification by means of mass spectrometry {#Sec4} ---------------------------------------------------- Proteins unique to the heat-stressed HSP16.6-Strep sample were excised from 1D or 2D gels and digested with trypsin, and peptides were prepared for MS as described previously (Basha et al. [@CR2]). Peptide extracts were introduced onto a 100-μm I.D. × 5-cm C18 column using an autosampler and separated with a 25-min gradient of 2--100% acetonitrile in 0.5% formic acid. The column eluate was directed into a Thermo Finnigan LCQ Deca ion trap mass spectrometer. The mass range scanned was 400 to 1500 *m*/*z*, and data-dependent scanning was used to select the three most abundant ions in each parent scan for tandem MS. Peptides were searched using SEQUEST and allowed for static modification of Cys (57 Da; iodoacetamidation), and differential modification of Met (16 Da; oxidation) was considered. X correlation cutoffs of 2.0 for 2+ ions, 3.0 for 3+ ions, and delta Xcorr \> 0.05 were applied, and data were sorted using DTASelect (Tabb et al. [@CR39]). The complete list of 83 proteins identified as HSP16.6 interaction partners from these and our previous experiments (Basha et al. [@CR2]) is given in Supplemental Table [1](#MOESM1){ref-type="media"}. For the purpose of comparisons and calculations, this set is considered to represent sHSP interactors and denoted *I*, where \|*I*\| = 83. Known protein-protein interactions (PPIs) from yeast-2-hybrid experiments are available for *Synechocystis* (Sato et al. [@CR38]). We identified all PPIs made by members of *I* (Supplemental Table [1](#MOESM1){ref-type="media"}), excluding PPIs that were not identified with multiple positive prey clones, in order to avoid false positives. Bioinformatic analyses {#Sec5} ---------------------- The *Synechocystis* sp. PCC 6803 genome (Kaneko et al. [@CR22]; Kotani et al. [@CR23]) was obtained from CyanoBase, [http://genome.microbedb.jp/cyanobase/](http://genome.microbedb.jp/cyanobase) (Nakamura et al. [@CR31]). A set *G* representing the genome, containing all proteins such that *I* ⊆ *G*, was created from the protein-coding sequences in the genome. Only proteins with estimated isoelectric point (pI) within the range 4--9.5 and mass *m* between 10 and 200 kDa, corresponding to the range of proteins that could be identified in either the 1D or 2D gels, were included (see [Supporting Information](#Sec13){ref-type="sec"}). This filtering resulted in *G* comprising 3021 proteins (i.e., \|*G*\| = 3021), which amounts to \> 80% of the proteins encoded in the genome. The mass, sequence length *n*~aa~, and abundance (absolute numbers *n*~*F*~ and frequencies *f*~*F*~ = *n*~*F*~/*n*~aa~) of various sequence features *F* were determined for every protein. These were DnaK-binding motifs; VQL, IXI, and \[I/L/V\]X\[I/L/V\] motifs (where X refers to any amino acid); charged (D,E,H,K,R), positive (H,K,R), negative (D,E), and hydrophobic (C,F,I,L,M,V,W) residues. DnaK-binding motifs were identified using a previously described algorithm (Van Durme et al. [@CR41]), and the other motifs were found through *regexp* pattern-matching using the Python Standard Library. Long-range disorder was predicted with IUPred (Dosztanyi et al. [@CR6], [@CR7]) using default parameters, and residues with a score \> 0.5 were considered unstructured. For the remainder, secondary structure was predicted from the sequences using the EMBOSS (Rice et al. [@CR34]) implementation of the GOR method. β-strands and β-turns were pooled together into "β-structures." Average abundances were calculated separately for *I* and *G*. Statistical significance testing and representation {#Sec6} --------------------------------------------------- A bootstrapping approach was employed to assess the statistical significance of any differences between *I* and *G*. First, a random subset, *R*, was taken from *G* by arbitrarily picking, with replacement, of 83 proteins (i.e., *R* ⊆ *G* and \|*R*\| = \|*I*\|). The mean, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\overline{Q}}_R $$\end{document}$, was then calculated for the given quantity of interest *Q*, to allow comparison with $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\overline{Q}}_I $$\end{document}$, the mean calculated from *I* for the same quantity. This was repeated *N* times, after which the *p* value was calculated as the frequency by which $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\overline{Q}}_R\ge {\overline{Q}}_I $$\end{document}$ or $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\overline{Q}}_R\le {\overline{Q}}_I $$\end{document}$, in the respective cases of $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\overline{Q}}_I>{\overline{Q}}_G $$\end{document}$ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\overline{Q}}_I<{\overline{Q}}_G $$\end{document}$. For each quantity, a total of *N* = 100,000 iterations was run, and the statistical significance was tested at the 0.01 level. Kernel density estimates were plotted for all quantities where a statistically significant difference was found. A Gaussian kernel with a bandwidth equal to 2% of the visible range was used in all cases and the amplitude was set such that the integrated density was equal to the number of proteins in each set. As such, the amplitudes are inversely proportional to the ranges along the *x*-axis, and their heights can thus differ substantially between distributions. Moreover, the *y*-axes' ranges were chosen to make the *I* and *G* distributions occupy the same visible area in the resulting plot. Biological function analysis {#Sec7} ---------------------------- A PANTHER Overrepresentation Test (release 20170413) against the GO Ontology database (release 20170926) was made for all proteins in *I*, using the *Synechocystis* reference list and the "GO biological process complete" annotation data set. Bonferroni correction was applied for multiple testing, and a *p* value cut-off of 0.05 was used to filter the results. Proteins that were not mapped to any entry in the reference list were added to the set of "unclassified" proteins. Enrichment was defined as *n*~*p*~/E(*n*~*p*~), where *n*~*p*~ is the number of proteins in *I* being ascribed to biological process *p*, and E(*n*~*p*~) is the expected number of such proteins based on their frequency in *G* and the size of *I*. Proteins that were assigned to the GO-class "biological process" but not to any of its subclasses were given the collective label "other biological process." Since a single protein can have multiple classifications, the sum of proteins in the different classes exceeds \|*I*\|. Protein BLAST (Altschul et al. [@CR1]) was used to find orthologs among the interactors identified for HSP16.6 in *Synechocystis*, IbpB in *E. coli* (Fu et al. [@CR10]), and HSP20.2 in *D. radiodurans* (Bepperling et al. [@CR4]). Three pairwise comparisons were made to define the overlap between the sets of interactors, where the list of interactors from one organism was used as the "database" and the list of interactors from the other as the "query." Using *E. coli* as the database yielded poorly annotated hits; hence, the primary database was set to be *Synechocystis* and the secondary database to be *D. radiodurans*. An *E* value cut-off of 10^−10^ was used for all BLAST searches, and whenever a protein in the query yielded several matches the one with the lowest *E* value was chosen. Lastly, the overlap between *E. coli* and *D. radiodurans* was used as a query against *Synechocystis* in order to find the overlap between the interactors in all three organisms. The triply overlapping set of proteins were also analyzed for an overrepresentation test in *Synechocystis* as described above, but without imposing a *p* value because of the small number of proteins in the query. Results {#Sec8} ======= Identification of proteins associated with HSP16.6 during heat stress in vivo {#Sec9} ----------------------------------------------------------------------------- To identify a larger number of HSP16.6-associated proteins than we did previously (Basha et al. [@CR2]), we developed a *Synechocystis* strain in which the wild-type *HSP16.6* gene was replaced with an *HSP16.6* gene modified to encode a Strep-tag II at the C-terminus. HSP16.6-Strep was shown to complement HSP16.6 in vivo in thermotolerance assays (Basha et al. [@CR2]), as well as functioning in vitro to protect model interactors from irreversible heat-denaturation (Friedrich et al. [@CR9]). The HSP16.6-Strep strain and an isogenic strain carrying wild-type HSP16.6 were subjected to mild heat stress to allow accumulation of the sHSP and then to a short, more severe heat stress to maximize association of thermally unstable proteins with the sHSP. The soluble cell fraction from control and heat-stressed cells of the HSP16.6-Strep and HSP16.6 strains was subjected to Strep-Tactin affinity chromatography and the recovered proteins compared by means of 2D electrophoresis (or, to examine high molecular mass proteins, by using 1D electrophoresis) (Fig. [1](#Fig1){ref-type="fig"}). Individual spots or bands unique to proteins affinity-purified with HSP16.6-Strep from the heat stress samples were excised and subjected to MS analysis.Fig. 1Identification of HSP16.6 interactors. **a** SDS-PAGE separation of proteins recovered in association with HSP16.6-Strep in cells grown at 30 °C and treated at 42 °C for 2 h plus 1 h recovery at 30 °C to allow sHSP accumulation (control sample, C) or further treated with an additional 30 min at 46 °C (heat-stressed sample, HS). To recover proteins in the high molecular mass range, separation was performed using an 8% acrylamide gel, and the position of molecular mass markers is indicated. Bands that were excised for analysis are annotated with red dashes. Double-width dashes indicate bands that gave hits for proteins associated with protein-folding processes. **b** 2D gel separation of samples prepared as described in **a**. The position of molecular mass markers and the acidic (+) and basic (−) sides of the silver-stained 2D gels are indicated. Spots that were excised and yielded the reported data are annotated with red circles (right panel). The ellipse in each panel indicates the spots due to HSP16.6 We identified a total of 72 proteins in these experiments which, when combined with others we had identified previously (Basha et al. [@CR2]), expanded to a total of 83. Notably, the proteins were recovered from the soluble fraction, so they do not represent those that underwent excessive aggregation, or associations with membranes and cytoskeletal elements that may have led to partitioning into the pellet. As such, these proteins represent potential sHSP interactors that have been prevented from insolubilization by interaction with HSP16.6. We denote this set of interactors *I*, representing a subset of the genome *G* detectable in our experiments. This allows us to test hypotheses about the features of these interactors to shed light on what distinguishes them from the other proteins in *Synechocystis*. Though many of the interactors have known PPIs, based on cross-referencing to genome-wide yeast-2-hybrid data (Sato et al. [@CR38]) (Supplementary Table [1](#MOESM1){ref-type="media"}), notably there are only three described pairwise PPIs within *I*, and all three of these are self-associations. To see if this low count was an artifact from our conservative approach of excluding PPIs that were identified with only one prey clone, we also tested including the latter, which presumably yields more false positives. This increased the number of pairwise PPIs within *I* to 12, including six self-interactions, which is still a small subset of *I*. Consequently, the proteins in *I* appear largely independent of each other in their interaction with HSP16.6, consistent with our affinity-isolate methodology being sensitive to stable interactors. Primary- and secondary-structure features of HSP16.6 interactors {#Sec10} ---------------------------------------------------------------- We first compared the average mass and sequence lengths of the interactors to the genome. We found that these were very different, with the interactors being about 60% larger on average (Table [1](#Tab1){ref-type="table"}, Fig. [2](#Fig2){ref-type="fig"}a, b). While this is informative about the interactor profile of HSP16.6, it also means that the absolute number, *n*~*F*~ of any feature *F*, is likely to be larger for the interactors. To account for this, all subsequent analyses are consequently focused on fractional quantities, *f*~*F*~, which are normalized by sequence length in order to reveal distinctive features for the proteins associated with HSP16.6.Table 1Comparison of various primary- and secondary-structure features between interactors of HSP16.6 in *Synechocystis* with the wider genome. Mean values obtained for the proteins in *I* and *G*, along with *p* values for the differences between themQuantityInteractors, *I*Genome, *Gp* value*m*/Da**57,86036,561\< 10** ^**−5**^*n* ~aa~**525336\< 10** ^**−5**^*f* ~DNAK~**0.01980.285\< 10** ^**−5**^*f* ~VQL~0.0003350.0002740.27*f* ~IXI~0.003490.003050.15*f* ~\[ILV\]X\[ILV\]~**0.03780.04260.002**pI5.225.630.036*f* ~Charged~**0.2520.2306.0∙10** ^**−5**^*f* ~+~0.1180.1150.24*f* ~−~**0.1340.114\< 10** ^**−5**^*f* ~H-phobic~**0.3090.3311.0∙10** ^**−5**^*f* ~d~0.0860.0580.015*f* ~β~**0.3550.415\< 10** ^**−5**^*f* ~α~**0.3830.3383.1∙10** ^*−5*^Bold text indicates statistically significant differences, defined as *p* \< 0.01Fig. 2Probability distributions of the statistically significant differences identified in Table [1](#Tab1){ref-type="table"}. **a**, **b** The distributions of protein mass (**a**) and sequence length (**b**) for *I* and *G*. The proteins in *I* are on average approximately 60% larger than those in *G*, both in terms of mass and sequence length. **c**, **d** Distributions of frequencies of \[I/L/V\]X\[I/L/V\] motifs (**c**) and DnaK-binding motifs (**d**). Both sequence features are less frequent and more narrowly distributed in *I*. **e**--**g** The fraction of hydrophobic (**e**), charged (**f**), and negative (**g**) residues. Charged residues are more frequent in *I*, which can be attributed to a higher fraction of negatively charged residues and a lower fraction of hydrophobic residues. **h**, **i** Fraction of residues with predominately helical (α and 3~10~, **h**) propensity and β-structure (sheet and turn, **i**). The helix content is higher in *I* than in *G*, and conversely, the β-structure content is lower in *I*. The distributions were normalized such that their integral equals the number of proteins in each set. Consequently, the amplitudes are inversely proportional to the width of the distributions, and the amplitudes of the two distributions in each panel reflect the different sizes of the two sets We judged that certain sequence motifs might be implicated in the association of interactors with sHSPs. To develop hypotheses for testing, we considered a model in which interfaces that allow the sHSP to self-assemble might be the same as interactor binding sites (Jacobs et al. [@CR20]). In this context, the inter-monomer contact made between the highly conserved "IXI" motif in the C-terminal region and the β4--β8 groove of the ACD has been proposed as an auto-inhibitory interface (Jehle et al. [@CR21]; van Montfort et al. [@CR42]). Theorizing that IXI motifs might mediate contacts with the sHSPs, we therefore asked whether they were differentially represented in the interactors. We also posed this question in a more general form, by searching for motifs matching the requirement \[I/L/V\]X\[I/L/V\], which is more encompassing across the breadth of sHSPs (Poulain et al. [@CR32]). Furthermore, we searched for VQL motifs, as this corresponds to the specific manifestation of the "IXI" in HSP16.6. Comparing the fractional abundance of these motifs (*f*~IXI~, *f*~\[ILV\]X\[ILV\]~, *f*~VQL~, respectively) between the interactors and the genome, we found there to be no meaningful difference for IXI and VQL, but the general form \[I/L/V\]X\[I/L/V\] was significantly under-represented in the interactors (Table [1](#Tab1){ref-type="table"}, Fig. [2](#Fig2){ref-type="fig"}c). sHSPs are thought to transfer interactors to the DnaK (HSP70 in eukaryotes) system for ATP-dependent refolding (Haslbeck and Vierling [@CR16]). We therefore hypothesized that the presence of DnaK-binding motifs (Rudiger et al. [@CR36]), which mediate association with this downstream chaperone, might be different between the interactors and the genome. We found the fractional abundance of DnaK motifs (*f*~DnaK~) to be \> 30% lower in the interactors (Table [1](#Tab1){ref-type="table"}, Fig. [2](#Fig2){ref-type="fig"}d). We next considered electrochemical properties of the proteins. The difference in pI between the interactors and genome was just outside our significance criterion (*p* = 0.036 \> 0.01). However, when examining the fraction of charged residues (*f*~Charged~), we discovered it to be higher in the interactors. By investigating negatively and positively charged residues separately (*f*~−~ and *f*~+~, respectively), we found this difference to be due to the former, with negatively charged residues \> 16% more abundant in the interactors. Conversely, the genome contains a higher fractional abundance of hydrophobic residues (*f*~H-phobic~) (Table [1](#Tab1){ref-type="table"}, Fig. [2](#Fig2){ref-type="fig"}e--g). Lastly, we asked whether predicted secondary structure differed between the two sets. The fraction of residues in disordered regions (*f*~*d*~) is insignificantly higher in the interactors, albeit very near our threshold (*p* = 0.015 ≈ 0.01). For the structured regions, on average, the interactors had a higher fraction of residues in helices (*f*~α~) and lower fraction in β-structures (*f*~β~), compared to the proteins in the wider genome (Table [1](#Tab1){ref-type="table"}, Fig. [2](#Fig2){ref-type="fig"}h, i). Functional classification of HSP16.6-associated proteins {#Sec11} -------------------------------------------------------- Where possible, interactors were classified according to their gene-ontology annotation into either "metabolic process," "cellular process," or "other biological process." Many proteins were assigned to multiple classes, and 15 proteins could not be matched to the reference list and were added to the set of unclassified proteins, which then comprised 24 proteins. This classification yielded different distributions of processes in *I* and *G* (Fig. [3](#Fig3){ref-type="fig"}a), indicating that HSP16.6 has an interaction profile that reflects the biological function of its interactors. To quantify the differences, we calculated the overrepresentation of proteins involved in the various biological processes (Fig. [3](#Fig3){ref-type="fig"}b). The data reveal statistically significant enrichment of proteins ascribed to certain biological processes in the interactors, suggesting that HSP16.6 makes function-specific interactions. The most striking association was for proteins involved in protein folding, with 6 out of the 19 known such proteins being found in *I* (Table [2](#Tab2){ref-type="table"}), corresponding to a thirteen-fold enrichment.Fig. 3Classification of proteins involved in different gene-ontology annotations of biological processes. **a** Pie charts show the extent of different classes in *I* and *G*. The most fundamental classes have labels in bold face. Note that "cellular metabolic process" belongs to both "metabolic process" and "cellular process" and is therefore represented by two colors. **b** Enrichment within *I* of proteins taking part in the various biological processes. Circle areas reflect the number of proteins in *I*, and numbers indicate proteins in *I* and *G*. *I* contains a smaller fraction of unclassified proteins than *G*, and all classes are somewhat enriched in *I*. Proteins involved in protein folding are enriched thirteen-fold, with 6 of the 19 such proteins known being found among the interactors. Inset: Same analysis performed for the 10 overlapping proteins from the analysis in (**c**). In all featured classes, the fold-enrichment is higher. **c** Venn diagram showing the overlap of sHSP interactor ranges from *Synechocystis*, *E. coli*, and *D. radiodurans*. Note that, with the exception of the intersection of the three sets, all areas of the diagram reflect the number of elements withinTable 2The six interactors of *Synechocystis* HSP16.6 annotated as belonging to the "protein folding" categoryGeneUniProt IDNamesll0058Q55154DnaK 1sll0170P22358DnaK 2sll1932P73098DnaK 3slr2076Q0597260 kDa chaperonin 1sll0533Q55511Trigger factor (TF)slr1251P73789Peptidyl-prolyl cis-trans isomerase To compare HSP16.6-interactors with those identified in other prokaryotes, we cross-referenced our list with those reported as IbpB interactors in *E*. *coli* (Fu et al. [@CR10]), and HSP20.2 in *D. radiodurans* (Bepperling et al. [@CR4]) (Fig. [3](#Fig3){ref-type="fig"}c). There were unique orthologs for 17 HSP16.6 interactors among the 113 IbpB interactors and 17 for the 101 HSP20.2 interactors. The overlap between IbpB and HSP20.2 interactors was larger still, comprising 36 unique orthologs. A total of 10 proteins were found in all three sets of interactors. Notably, these overlaps are much larger than one would expect by chance (approximately 3 for each pairwise overlap, and fewer than 1 for the triple overlap). Interestingly, these proteins were also diverse, spanning multiple biological processes, with only one eluding classification (Table [3](#Tab3){ref-type="table"}, Fig. [3](#Fig3){ref-type="fig"}b inset). With the exception of the "protein folding" and "other biological process," which were not represented at all in this subset, all categories were even more overrepresented than in the complete list of HSP16.6 interactors. We note that the small number of proteins precluded low *p* values for the levels of enrichment for the individual categories. Taken together, they nonetheless indicate that the enrichment pattern seen for the *Synechocystis* interactors is particularly prominent for the interactors that are common for all three sHSPs, with the striking exception of the protein-folding interactors, which might be a species- or sHSP-specific phenomenon, or the result of differences in the methods used for recovering interacting proteins.Table 3Proteins that we associated to all three of HSP16.6 (*Synechocystis*), IbpB (*E. coli*), and HSP20.2 (*D. radiodurans*). The GO annotations for biological processes are coded as follows: metabolic process (MP), cellular process (CP), nitrogen-compound metabolic process (NCMP), primary metabolic process (PMP), biosynthetic process (BP), organic substance metabolic process (OSMP), cellular metabolic process (CMP), and unclassified (U). In some cases, two distinct IbpB or HSP20.2 interactors would correspond to an HSP16.6 interactor, in which case, both UniProt IDs were included in the table*Synechocystis* geneUniProd ID\ *Synechocystis*\ *E. coli*\ *D. radiodurans*NameGO biological processsll0018Q55664\ G64976n\ NP_295312.1Fructose-bisphosphate aldolase, class IIMP, CP, NCMP, PMP, OSMP, CMPsll1099P74227\ NP_289744.1, pdb\|1EFC\|A\ NP_295522.1Elongation factor TuMP, CP, NCMP, PMP, BP, OSMP, CMPsll1180P74176\ NP_287490.1\ NP_295291.1Toxin secretion ABC transporter ATP-binding proteinCP, NCMP, PMP, OSMPsll1326P27179\ CAA23519.1\ NP_294424.1ATP synthase alpha chainMP, CP, NCMP, PMP, BP, OSMP, CMPsll1787P77965\ AAC43085.1\ NP_294636.1RNA polymerase beta subunitMP, CP, NCMP, PMP, BP, OSMP, CMPsll1789P73334\ NP_290619.1\ NP_294635.1RNA polymerase beta' subunitMP, CP, NCMP, PMP, BP, OSMP, CMPsll1818P73297\ CAA37838.1\ NP_295851.1RNA polymerase alpha subunitMP, CP, NCMP, PMP, BP, OSMP, CMPsll1841P74510\ NP_285811.1, NP_286443.1\ NP_293809.1, NP_293979.1Pyruvate dehydrogenase dihydrolipoamide acetyltransferase component (E2)MPslr0542P54416\ NP_286179.1\ NP_295695.1ATP-dependent protease ClpPMP, NCMP, PMP, OSMPslr1105P72749\ NP_289127.1\ NP_294922.1GTP-binding protein TypA/BipA homologU Discussion {#Sec12} ========== Here, we have examined the properties of 83 proteins that associate in vivo with HSP16.6 under conditions of heat stress. Given that the proteins were obtained from the soluble supernatant after centrifugation, they are likely to under-represent membrane- and cytoskeleton-associated proteins. Furthermore, as our experiment involves affinity pull-downs, these interactors are inevitably restricted to those that form interactions that are stable on the timescale of the experiment. In the context of the model proposed for sHSPs wherein they display both a low-affinity mode with high capacity, and a high-affinity mode with low capacity (McHaourab et al. [@CR28]), our interactors are likely representative of the latter. Notwithstanding these potential biases of the experiment, we have shown that the interactors were on average larger than the proteins in the genome, have a distinct electrochemical profile, an increased fraction of helical secondary structure, and a lower fraction of \[I/L/V\]X\[I/L/V\] and DnaK-binding motifs. We observed that HSP16.6 preferentially binds longer, more massive, proteins. This is in agreement with analysis of sHSP interactors *E. coli* and *D. radiodurans* (Fu et al. [@CR11]) and is interesting in light of recent data noting that thermally unstable proteins in cells are typically longer than those that are stable (Leuenberger et al. [@CR26]). Longer proteins might therefore be overrepresented in the interactors by virtue of being more likely to be destabilized by the heat-shock condition assayed here. Alternatively, or in addition, it is possible that longer proteins, by virtue of having more binding sites, might be held tighter by the sHSPs. This would stem from avidity effects resulting from the multivalency of sHSP oligomers (Hilton et al. [@CR17]), similar to observations made for other molecular chaperones (Huang et al. [@CR19]; Saio et al. [@CR37]). Upon considering amino acid motifs and composition, we found a lower fraction of \[I/L/V\]X\[I/L/V\] motifs in the interactors. This suggests that the β4--β8 groove, which binds this motif intra-molecularly in sHSP oligomers (Basha et al. [@CR3]; Hilton et al. [@CR17]), is not the binding site for these stable interactors. However, this does not preclude the β4--β8 groove being a site for low-affinity, or transient, interactions. This is consistent with the observation that the excised ACD can display potent chaperone activity (Cox et al. [@CR5]; Hochberg et al. [@CR18]). We also identified an overabundance of charged and, in particular negatively charged, residues in the interactors. A preponderance of charged residues was also observed for sHSP interactors in *E. coli* and *D. radiodurans* (Fu et al. [@CR11]). Notably, aspartates have been shown to be enriched in thermally unstable proteins (Leuenberger et al. [@CR26]), again hinting that thermal stability could be a key attribute for recognition by sHSPs. It is also interesting to consider the electrochemical profile of the sHSPs themselves, which have an overabundance of charged residues in the ACD and C-terminal region (Kriehuber et al. [@CR24]). As such, it is possible that there may be charge-complementarity aspects to binding. The depletion of DnaK-binding motifs in the HSP16.6 interactors is striking, particularly when considering that DnaK is able to release interactors from the complexes made with HSP16.6. This suggests that the DnaK-binding motif is not responsible for the recognition events that mediate interactor transfer between the chaperones. Instead, the DnaK-binding motif may be more reflective of DnaK's holdase, rather than refoldase activity. In this way, proteins that are not protected by the sHSPs are captured by HSP70 instead (Mayer and Bukau [@CR27]). The interactors are also enriched in α-helical propensity and depleted in β-structure. It is possible that, based on the observation that there is little cooperativity in the folding of β-sheets (Wu and Zhao [@CR43]), this may be reflective of physico-chemical differences in re- or unfolding. Gene-ontology analysis demonstrates that, while capable of associating with many interactors, HSP16.6 nonetheless does so with statistically significant specificity, evidenced by varying enrichments for different biological processes. This observation is validated by the overlap between *Synechocystis*, *E. coli*, and *D. radiodurans* sHSP interactors. The notion that sHSPs have specific interactors in the cell also extends to eukaryotes, where different sHSPs found in the same cellular compartment have differing interactor profiles (Fleckenstein et al. [@CR8]; McLoughlin et al. [@CR29]; Mymrikov et al. [@CR30]). The most enriched groups of proteins associated with HSP16.6 were other components of the protein folding machinery. We interpret this as due to HSP16.6 being part of a tightly linked molecular chaperone network (Gong et al. [@CR15]), collaborating to prevent and reverse improper protein interactions in the wider heat-shock response of the cell (Richter et al. [@CR35]). Possibly, these interactions are indirect, captured due to HSP16.6 and other protein-folding components acting on the same substrates. An indirect interaction with protein-folding components could also explain the lack of equivalent proteins in the *E. coli* sHSP interactors (Fu et al. [@CR10]), as the previous report employed covalent-crosslinking and urea solubilization prior to immunoprecipitation. The *D. radiodurans* interactors were identified by a different method, employing ex vivo addition of purified HSP20.2 to cell lysates, prior to heat stress and immunoprecipitation. Given the differences in methodology between these studies, we suggest that those proteins comprising common interactors are highly significant (Table [3](#Tab3){ref-type="table"}). In sum, our study provides an initial view of the functional interactome of prokaryotic sHSPs and of *Synechocystis* in particular. In addition, the statistical framework we have implemented for examining sequence determinants can be applied to the analysis of the likely future profusion of proteomic data identifying molecular chaperone interactors in cells. Electronic supplementary material ================================= {#Sec13} ESM 1(DOCX 28 kb) The original version of this article was revised: Table 1 needed corrections. The DOI of the Erratum is: 10.1007/s12192-018-0901-6 **Electronic supplementary material** The online version of this article (10.1007/s12192-018-0884-3) contains supplementary material, which is available to authorized users. A correction to this article is available online at <https://doi.org/10.1007/s12192-018-0901-6>. **Change history** 5/3/2018 Table 1 in the original publication has been corrected. We thank Linda Breci (University of Arizona) for performing MS experiments and Georg Hochberg (University of Chicago) for helpful discussions. This work was supported by the Swedish Research Council and the European Commission for a Marie Skodowska Curie International Career Grant (2015-00559) to EGM, the Biotechnology and Biological Sciences Research Council (BB/K004247/1) to JLPB, and the National Institutes of Health (RO1 GM42762) to EV.

Introduction {#Sec1} ============ Alcohol use disorders (AUD) are common and associated with significant morbidity and mortality \[[@CR1]--[@CR3]\], but are substantially undertreated. In 2013, 16.6 million U.S. adults met diagnostic criteria for an AUD, but research suggests only 7.8% received any formal treatment \[[@CR4]\]. One of the major gaps in treatment for AUD is the significant under-utilization of medications that are effective for treating AUD \[[@CR1], [@CR5], [@CR6]\]. Three medications---*disulfiram*, *acamprosate*, and *naltrexone* (both oral and injectable)---have FDA approval specifically for the treatment of AUD, and *topiramate* has strong meta-analytic support \[[@CR7]\]. Efforts to increase treatment of AUD with medications is motivated in part because the modality may address many reported barriers to receiving any formal AUD treatment \[[@CR4], [@CR8]\]. For instance, psychosocial treatments are often offered in group settings, heightening stigma-related issues for some patients, whereas medications can be provided on an individual basis \[[@CR9]\]. In addition, patients may not be ready to abstain \[[@CR8], [@CR10]\]. Further, though this may be shifting over time \[[@CR11], [@CR12]\], many treatment programs view abstinence as the ultimate goal \[[@CR8]\], whereas abstinence is not required with all medications and reduced drinking can be a goal of medication treatment \[[@CR9]\]. Finally, AUD medications can be offered across healthcare settings, including primary care, which has been highlighted as an optimal setting for expansion of care for AUD \[[@CR8], [@CR13], [@CR14]\]. Despite the promise of medication treatment for addressing several known barriers to AUD treatment and national recommendations encouraging medications be made available to all patients with AUD \[[@CR15], [@CR16]\], rates of pharmacotherapy for AUD remain extremely low. Among patients with AUD, 4-12% are treated pharmacologically \[[@CR1], [@CR6], [@CR17]--[@CR21]\]. Among subsets of patients with AUD and co-occurring schizophrenic, bipolar, posttraumatic stress or major depressive disorder, receipt of medications for AUD ranged from 7 to 11%, whereas receipt of medications for the comorbid disorder ranged from 69 to 82% \[[@CR19]\]. This gap in the quality of AUD treatment is well known, and the substantial barriers to provision of AUD medications in diverse contexts have been described \[[@CR22]--[@CR27]\]. However, the optimal strategies for addressing these barriers and increasing use of medications for AUD treatment remain elusive. In recent years, two related lines of research have contributed to knowledge regarding strategies to increase use of medications to treat AUD: evaluations of care delivery interventions and evaluations of implementation interventions. Care delivery interventions typically focus on improving patient-level clinical outcomes (e.g., reduction in heavy drinking days or abstinence from alcohol use), but often secondarily assess patient- or clinician-level process outcomes focused on treatment receipt (e.g., engagement in pharmacotherapy for AUD). Implementation interventions are typically designed to improve patient- or clinician-level process outcomes, but sometimes secondarily include patient-level clinical outcomes when the evidence for the effects of the underlying practice is weak (so called Hybrid I studies) \[[@CR28]\]. Other key differences exist between these types of research that may influence both clinical and process outcomes. Most importantly, care delivery interventions typically involve recruitment of patients who are willing to be randomized to the treatment arms contained within the new care delivery model. Thus, these trials may be restricted to patients who are at least open to, if not actively interested in, treatment for AUD. On the other hand, evaluations of implementation interventions typically recruit and intervene on clinical entities (e.g., providers, clinics, hospitals) who serve large groups of patients who likely have more variable interest in treatment. Further, evaluations of care delivery interventions are typically designed to establish the effectiveness (or lack thereof) of particular care delivery models. Thus, these studies generally put significant effort and resources into ensuring fidelity to the care delivery model. On the other hand, implementation evaluations are often trying to establish the effectiveness of bundles of strategies (interventions) to increase uptake of practices that do not depend on external research resources. Thus, evaluations of implementation interventions may measure fidelity as a process outcome but typically exert less direct control \[[@CR29]\]. Even though care delivery and implementation interventions differ in terms of methodology, patient inclusion criteria, and primary outcomes, they may evaluate the effectiveness of the same underlying implementation strategies, such as reorganizing, supplementing, or intervening on existing models of care \[[@CR29]\]. The fact that the same component implementation strategies (e.g., audit and feedback) have been evaluated by these different research designs with very different patient populations affords an opportunity to take stock of the effectiveness of these interventions, and to distill insights into which designs, contexts, and component strategies appear to drive outcomes. Therefore, our goal was to conduct a structured review of published evaluations of care delivery and implementation interventions that have either primarily or secondarily aimed to increase use of pharmacotherapy for patients with AUD, with the goal of identifying component strategies that may be effective in increasing pharmacologic treatment of AUD. Our review was guided by an existing taxonomy of implementation strategies and terms identified via a three-round modified-Delphi process \[[@CR30]\]. The purpose of our review was to learn which components have been tried most commonly and which strategies might be associated with larger effects. Also, due to the fact that evaluations of care delivery interventions exert greater efforts to ensure fidelity and include patients willing to be randomized, we hypothesized that higher adoption of medications for AUD will be observed in those contexts compared to implementation interventions, which typically aim to intervene on clinician and patient populations with greater variability in treatment motivation, knowledge, and preferences. Methods {#Sec2} ======= For this structured literature review, we sought to identify published evaluations of care delivery and implementation interventions reporting effects on receipt of medication treatments for patients with AUD. We reviewed literature through May 2018. Studies were identified via searching PubMed, Google Scholar, and PsychInfo with relevant search terms (e.g., pharmacotherapy, alcohol use disorder medications, AUD medications, naltrexone, Acamprosate, disulfiram, medication-assisted treatment). We also reviewed reference lists from identified studies to identify additional studies that may have been missed by our search. Finally, because we have personally conducted and/or served as co-investigators on related studies, additional studies were also identified via networking. Once identified, each individual article was coded for implementation strategies used, as guided by Powell et al.'s refined compilation of implementation strategies resulting from the Expert Recommendations for Implementing Change (ERIC) project \[[@CR30]\]. All articles were independently reviewed and coded by two investigators (EW and TM). When multiple articles and/or published protocols or commentaries were identified that described a single intervention and/or implementation effort, these articles were aggregated to the level of the intervention (e.g., three studies had adjoining published protocol papers, which were coded under the umbrella of a single study). Once coded, all authors met to review coding discrepancies, discuss interpretation of codes, arrive at consensus, and revise individual codes based on consensus. After reaching internal consensus on coding, we reached out to the lead or senior author of each study to ask whether our codes aligned with their understanding/interpretation of their study and associated report. We shared Powell et al's description of strategies and asked them to review our coding to see if they thought we had missed or miscoded anything. Finally, process (e.g., rates of prescribed AUD pharmacotherapy) and alcohol use outcome data were extracted from each study and described. All authors reviewed the coding of implementation strategies against study outcomes data to qualitatively identify sets of implementation strategies that might have been be most effective for increasing provision of AUD medications and report whether interventions that increased AUD pharmacotherapy also improved alcohol use outcomes. Results {#Sec3} ======= Our literature review identified nine studies that evaluated interventions to primarily or secondarily increase utilization of pharmacotherapy for AUD. Four were randomized clinical trials of care delivery interventions designed to improve alcohol-related outcomes \[[@CR31]--[@CR38]\]. Four were quasi-experimental evaluations of large-scale implementation interventions designed to increase medication receipt \[[@CR39]--[@CR43]\], and one was a quasi-experimental evaluation of targeted implementation intervention in a single-site \[[@CR44]\]. Two additional studies were identified but not included. The first reported on a large-scale implementation intervention designed to increase screening and brief intervention for unhealthy alcohol use and secondarily assessed whether the implementation was associated with increased receipt of AUD medications among those who screened positive \[[@CR45]\]. However, it was not clear how many of the patients who screened positive met diagnostic criteria for AUD and thus would have been eligible for medication treatment, and, though findings regarding medication use were summarized, detailed data were not reported. The second report was a description of a demonstration project to implement extended release naltrexone in Los Angeles County, but no evaluation of the program's effect on receipt of medication treatment among patients with AUD was reported \[[@CR46]\]. Table [1](#Tab1){ref-type="table"} presents implementation strategies identified by our internal coding process across each identified study (labelled with X). All lead or senior authors of studies responded to our request for review of the codes and added additional codes (labelled with an O). Implementation strategies used were variable across the studies, and no strategy was used across all studies (Table [1](#Tab1){ref-type="table"}). The most frequently used strategies were assessing readiness and identifying barriers and facilitators, distributing educational materials, facilitating relay of clinical data to providers (audit and feedback), and providing ongoing consultation. Strategies less frequently used involved payment and/or incentives or changes in laws and/or credentialing and licensing.Table 1Implementation strategies identified in published evaluations of care delivery and implementation interventions that have aimed to increase medication treatment for patients with alcohol use disorderStrategySaitz AHEAD CCM \[[@CR32]\]Oslin\ Alcohol Care Management \[[@CR31]\]Watkins\ SUMMIT \[[@CR35]--[@CR38]\]Bradley CHOICE \[[@CR33], [@CR34]\]Robinson Group Manage \[[@CR44]\]Harris\ VA Academic Detailing Program \[[@CR40]\]Hagedorn\ ADaPT--PC \[[@CR39], [@CR42]\]Ford\ Medication Research Partnership \[[@CR43]\]Ornstein\ PPRNet-TRIP \[[@CR41]\]Row totalAssess readiness and identify barriers/facilitatorsOXXXXXX7Distribute educational materialsXXXXXXX7Facilitate relay of clinical data to providersXXXXXXX7Provide ongoing consultationOXXXXXO7Intervene with patients/consumers to enhance uptake and adherenceOXOXXX6Conduct ongoing trainingOXXXXX6Create new clinical teamsXXXXXO6Identify and prepare championsOXXXXX6Provide local technical assistanceOOXXXX6Conduct educational meetingsXXXXXX6Develop and implement tools for quality monitoringXXXXXX6Develop/organize quality monitoring systemsXXXXX5Conduct educational outreach visitsOXXXX5Audit and provide feedbackOOXXX5Develop educational materialsXXXXO5Organize clinician implementation team meetingsOXXXX5FacilitationOXXXX5Obtain formal commitmentsXXXXX5Remind cliniciansOXOXX5Revise professional rolesOXXXO5Provide clinical supervisionXXXX4Develop academic partnershipsXXOO4Promote adaptabilityOXXX4Centralize technical assistanceOXXO4Conduct cyclical small tests of changeXOXO4Create a learning collaborativeXXXO4Make training dynamicXXOO4Purposely reexamine the implementationOOXX4Tailor strategiesXXXX4Use an implementation advisorXOOO4Use data warehousing techniquesOXXX4Conduct local needs assessmentXXX3Change record systemsOOX3Promote network weavingXXO3Build a coalitionOXO3Conduct local consensus discussionsOXO3Develop a formal implementation blueprintXXX3Recruit, designate, and train for leadershipXOX3Access new fundingOX2Change service sitesXX2Increase demandXX2Involve executive boardsOO2Involve patients/consumers and family membersXX2Prepare patients to be active participantsOX2Use advisory boards and workgroupsOO2Use data expertsX?2Capture and share local knowledgeXX2Identify early adoptersOO2Make billing easierXO2Visit other sitesXX2Alter incentive/allowance structuresX1Alter patient/consumer feesX1Change physical structure and equipmentO1Inform local opinion leadersX1Mandate changeX1Model and simulate changeX1Use train-the-trainer strategiesO1Fund and contract for the clinical innovation0Work with educational institutions0Develop resource sharing agreements0Change accreditation or membership requirements0Change liability laws0Create/change credentialing and/or licensure standards0Develop an implementation glossary0Develop disincentives0Obtain/use patient/consumer/family feedback0Place innovation on fee for service lists/formularies0Shadow other experts0Stage implementation scale up0Start a dissemination organization0Use capitated payments0Use mass media0Use other payment schemes0 The effects of the interventions on receipt of AUD pharmacotherapy were also variable across studies (Table [2](#Tab2){ref-type="table"}). In three of the four randomized evaluations of care delivery models \[[@CR31]--[@CR33]\], the interventions were associated with varying magnitude of increased receipt of AUD medications. At follow-up, treatment group rates of medication receipt ranged from 13 \[[@CR36]\] to almost 70% \[[@CR31]\]. The latter study, Oslin's Alcohol Care Management model \[[@CR31]\], was the only approach to significantly increase receipt of AUD medications and improve patient-level alcohol use outcomes (Table [2](#Tab2){ref-type="table"}). Two of the four implementation interventions \[[@CR40], [@CR41]\] were associated with increased AUD medication receipt. While Ornstein's Practice Partner Research Network-Translating Research Into Practice (PPRNet-TRIP) intervention appeared to have small early effects, proportions of patients receiving medications were so low that continued evaluation over time was not possible \[[@CR41]\]. The Veterans Health Administration's (VA) Academic Detailing Program appeared to increase rates of AUD medication receipt from 4.6 to 8.3% among patients with AUD \[[@CR40]\]. Receipt of AUD medications also appeared to increase in in a single VA facility after implementation of a group medication management program attended by patients taking and considering medication treatment \[[@CR44]\].Table 2Study designs and intervention effects on AUD medication receiptStudy\ (Author, abbreviated name, and reference)Sample size\ (Patients/sites)% Receiving AUD medicationsMeasure of AUD medication receiptIntervention and intervention effectSAITZ, AHEAD CCM \[[@CR32]\]563/1*BASELINE*:\ Intervention 4%\ Control 8%Receipt of addiction medication (buprenorphine, methadone, naltrexone, Acamprosate, disulfiram)*Program Name and Brief Description*: The Addiction Health Evaluation and Disease (AHEAD) Management Chronic Care Management (CCM) model "included longitudinal care coordinated with a primary care clinician; motivational enhancement therapy; relapse prevention counseling; and on-site medical, addiction, and psychiatric treatment, social work assistance, and referrals (including mutual help). The control group received a primary care appointment and a list of treatment resources including a telephone number to arrange counseling." AHEAD CCM was delivered by a multidisciplinary team (nurse care manager, social worker, internists, psychiatrist with addiction expertise)\ *Setting*: Hospital-based primary care practice (patients recruited from residential detoxification unit and referrals from urban teaching hospital) in Boston, MA\ *Goal*: Harm reduction\ *Key Components*: Use of registry to track and proactively reach out to patients, longitudinal care coordinated with primary care clinician and facilitated by shared electronic health record (EHR), motivational enhancement therapy, relapse prevention counseling, on-site medical, addiction and psychiatric treatment, social work assistance, and referrals (including to mutual help)\ *Effect on Medication Receipt*: OR = 1.88 (95% CI 1.28--2.75) *p* = 0.001\ *Effect on Alcohol Use Outcomes*: Not significant*FOLLOW-UP*:\ Intervention 21%\ Control 15%OSLIN Alcohol Care Management \[[@CR31]\]163/3*BASELINE*:\ Not reportedReceipt of naltrexone*Program Name and Brief Description*: Alcohol Care Management "focused on the use of pharmacotherapy and psychosocial support. Alcohol Care Management was delivered in-person or by telephone within the primary care clinic. The control group received standard treatment in a specialty outpatient addiction treatment program" Delivered by a behavioral health provider in-person or over-the-phone with primary care provider recommendation and support. Behavioral health providers were trained in motivational interviewing\ *Setting*: Veteran Affairs (VA) primary care in New York and Philadelphia\ *Goal*: Abstinence\ *Key Components*: Weekly 30 min visits, individualized patient education, pharmacotherapy and psychosocial support, repeated assessment of alcohol use, encouraged treatment adherence, monitoring of problems and management of potential side effects, use of shared EHR for communication with primary care provider\ *Effect on Medication Receipt*: Naltrexone prescribed in 65.9% of the Alcohol Care Management group relative to 11.5% in control; Chi^2^ 50.10, *p* \< 0.001\ *Effect on Alcohol Use Outcomes*: The Alcohol Care Management group was more likely to refrain from heavy drinking than the control (OR = 2.16, 96% CI 1.27--3.66) but no effect on any alcohol use (OR = 1.40, 95% CI 0.75--2.59)*FOLLOW-UP*:\ Intervention 65.9%\ Control 11.5%WATKINS SUMMIT \[[@CR35]--[@CR38]\]377/2*BASELINE*:\ Not reportedReceipt of any "medication assisted treatment" with either long-acting injectable naltrexone or buprenorphine/naloxone.*Program Name and Brief Description*: Collaborative care "was a system-level intervention, designed to increase the delivery of either a 6-session brief psychotherapy treatment and/or medication-assisted treatment with either sublingual buprenorphine/naloxone for opioid use disorders (OUDs) or long-acting injectable naltrexone for alcohol use disorders (AUDs). The control group was told that the clinic provided opioid and/or alcohol use disorder treatment and given a number for appointment scheduling and list of community referrals." Delivered by care coordinators and therapists with a social work degree\ *Setting*: Primary care at Federally Qualified Health Center in L.A., CA\ *Goal*: Increase screening and brief intervention for unhealthy alcohol use\ *Key components*: Intended 6 sessions of brief psychotherapy and/or med-assisted treatment (buprenorphine/naloxone for OUD and naltrexone for AUDs), repeated assessments of substance use, use of registry to track and proactively reach out to patients, motivation and encouragement of engagement in therapy\ *Effect on Medication Receipt*: OR comparing intervention to control at 6-months follow-up for patients with AUD and/or OUD = 1.23 (95% CI 0.60--2.40) *p* = 0.53. Published commentary from SUMMIT investigators \[[@CR37]\] suggests similar non-significant findings among patients with AUD only\ *Effect on Alcohol Use Outcomes*: Among patients with AUD only (54% of the sample) the SUMMIT intervention was significantly associated with abstinence from any alcohol use and all opioids at follow-up and was borderline significant for no heavy drinking in the past 30 days.*FOLLOW-UP*:\ Intervention 13.4%\ Control 12.6%BRADLEY CHOICE \[[@CR33], [@CR34]\]304/3*BASELINE*:\ Intervention 1% versus Control 2%Receipt of naltrexone, Acamprosate or disulfiram*Program Name and Brief Description*: Choosing Healthier Options in Primary Care (CHOICE) was a care management intervention in which "nurse care managers offered outreach and engagement, repeated brief counseling using motivational interviewing and shared decision making about treatment options, and nurse practitioner--prescribed AUD medications (if desired), supported by an interdisciplinary team (CHOICE intervention). The control group received usual primary care."\ *Setting*: VA Primary care in Washington State\ *Goal*: Harm reduction\ *Key components*: Proactive outreach and engagement, repeated brief counseling using MI and shared decision-making about treatment options (AUD medication, biomarker assessment if abnormal baseline, behavioral goal-setting and skills development for reducing drinking, encouragement of mutual help and/or specialty addictions treatment, self-monitoring)\ *Effect on Medication Receipt*: OR = 6.3 (95% CI 3.4--11.8) *p* \< 0.0001\ *Effect on Alcohol Use Outcomes*: Not significant*FOLLOW-UP*:\ Intervention 32% versus Control 8%ROBINSON, Group Management \[[@CR44]\]1600/1*BASELINE*:\ Increasing 0.08%/month in pre-implementation periodReceipt of naltrexone or Acamprosate, or extended-release naltrexone*Program Name and Brief Description*: Group Management of pharmacotherapy initially implemented to provide continued access during a staffing shortage, sought to provide psychosocial education on medication management for alcohol dependence. Delivered by an addiction psychiatrist in collaboration with either an Addiction Therapist or a Certified Nurse Specialist\ *Setting*: VA San Diego Health Care System\ *Goal*: Increase adoption of pharmacotherapy for AUD\ *Key components*: Group participants capped at 8, review of naltrexone and Acamprosate, discussion of side effects or benefits, discussion of barriers to sobriety in group format. Sessions lasted 1 h\ *Effect on Medication Receipt*: The rate of increase in the percent of patients treated pharmacologically for alcohol dependence increased 0.08% per month in the pre-implementation period to 0.21% per month after group visits were implemented\ *Effect on Alcohol Use Outcomes*: Not measured*FOLLOW-UP*:\ Increasing 0.21%/month in post-implementation periodHARRIS, VA Academic Detailing Program \[[@CR40]\]NA/37*BASELINE*:\ Intervention 4.56%\ Control 6.01%Monthly rates of receipt of naltrexone (oral or injectable), Acamprosate, disulfiram, or topiramate*Program Name and Brief Description*: VA Academic Detailing Program in which "The academic detailers strove to educate, motivate, and enable key health care providers to identify and address the spectrum of hazardous alcohol use, especially to facilitate more active consideration of pharmacological treatment options for AUD." Academic detailers were clinical pharmacy specialists\ *Setting*: VA medical centers and outpatient clinics in California, Nevada and the Pacific Islands\ *Goal*: Increase adoption of pharmacotherapy for AUD\ *Key components*: Local champions and leadership buy-in, dashboard for identifying patient candidates for AUD medication, repeated in-person visits to educate and build rapport with priority providers, problem-solve barriers and address knowledge gaps/misunderstanding about AUD meds, additional educational resources (e.g., patient education tools and pocket cards), integrated audit and feedback tools into EHR for identifying AUD patients, commitment from providers to increase prescribing patterns for AUD medication, monitoring and follow-up\ *Effect on Medication Receipt*: Slope of intervention sites increased more steeply than slope at control sites (*p* \< 0.001). From immediately pre-intervention to the end of the observation period (Month 16--Month 36), the percent of patients with AUD who received medication increased 3.36% in absolute terms and 67.77% in relative terms\ *Effect on Alcohol Use Outcomes*: Not measured*FOLLOW-UP*:\ Intervention 8.32%\ Control 6.90%HAGEDORN, ADaPT--PC \[[@CR39], [@CR42]\]NA/3*BASELINE*:\ Intervention 3.8%\ at end of pre-implementation period\ Control 3.7%Monthly rates of filled prescription for AUD medication (oral/injectable naltrexone, Acamprosate, disulfiram, topiramate) within 30 days after PC visit*Program Name and Brief Description*: Alcohol Use Disorder Pharmacotherapy and Treatment in Primary Care settings (ADaPT--PC) "targets stakeholder groups with tailored strategies based on implementation theory and prior research identifying barriers to implementation of AUD pharmacotherapy. Local SUD providers and primary care mental health integration (PCMHI) providers are trained to serve as local implementation/clinical champions and receive external facilitation. Primary care providers receive access to consultation from local and national clinical champions, educational materials, and a dashboard of patients with AUD on their caseloads for case identification. Veterans with AUD diagnoses receive educational information in the mail just prior to a scheduled PC visit." Delivered by site champions and external facilitators\ *Setting*: VA primary care\ *Goal*: Increase adoption of pharmacotherapy for AUD\ *Key components*: Training local champions, website with educational materials for primary care providers, a case-finding dashboard, technical assistance from local and national experts\ *Effect on Medication Receipt*: Rate of change (slope) increased significantly in the implementation period (*p* = 0.0023). Immediate post-implementation change not significant (*p* = 0.3401). Change over 12-month post-implementation relative to pre-implementation change significant (0.0033). No difference between intervention and control sites in immediate post-implementation change (p-0.8508). No difference between intervention and control sites in post-implementation slope (*p* = 0.4793)\ *Effect on Alcohol Use Outcomes*: Not measured*FOLLOW-UP*:\ Intervention 5.2%\ at end of implementation period\ Control 5.8%FORD\ Medication Research Partnership \[[@CR43]\]3887/9*BASELINE*:\ Intervention 9.0%\ Control 11.4%Receipt of AUD medication during an episode of care*Program Name and Brief Description*: Medication Research Partnership, "a collaboration between a national commercial health plan and nine addiction treatment centers, implemented organizational and system changes to promote use of federally approved medications for treatment of alcohol and opioid use disorders." Delivered by commercial health plan, "nationally recognized experts in the substance abuse field," and "change leaders."\ *Setting*: Specialty addiction treatment centers located on Northeastern seaboard of the U.S.\ *Goal*: Promote use of federally approved medications for AUD/OUD\ *Key components*: "Change leaders" and "change teams," external coaches, rapid change cycles to test strategies to promote medication use, provider training and technical assistance\ *Effect on Medication Receipt*: Difference in differences at Year 3:\ Unadjusted: 5.8%; Adjusted: 5.2% (95% CI − 4.1 to 14.5) *p* = 0.27\ *Effect on Alcohol Use Outcomes*: Not measured*3-YEAR FOLLOW-UP*:\ Intervention 26.5%\ Control 23.1%ORNSTEIN PPRNet-TRIP \[[@CR41]\]15053/19*EARLY INTERVENTION CLINICS:*\ Phase 1: 2.6%\ Phase 2: 5.5%Prescription for disulfiram, naltrexone (oral or injectable), Acamprosate, or topiramate*Program Name and Brief Description*: Practice Partner Research Network-Translating Research Into Practice (PPRNet-TRIP) involved "practice site visits for academic detailing and participatory planning and network meetings for 'best practice' dissemination"\ *Setting*: Primary care practices from 15 U.S. States\ *Goal*: Increased prescription for disulfiram, naltrexone (oral or injectable), Acamprosate, or topiramate for those with an AUD\ *Key components*: EHR template, performance reports, provider education, and development of an implementation plan\ *Effect on Medication Receipt*: Due to small proportions of subjects receiving medications, pre-post (phase 1 versus phase 2) comparison of medication receipt was only estimable in the early intervention clinics. The adjusted OR for phase 1 versus phase 2 in the early intervention clinics was 2.24 (95% CI 1.03 -4.88) *p* \< 0.05\ *Effect on Alcohol Use Outcomes*: Not measured*DELAYED INTERVENTION CLINICS*:\ Phase 1: 0%\ Phase 2: 2.4% Patterns of implementation strategies did not clearly distinguish studies that successfully increased use of pharmacotherapy versus those that did not. Discussion {#Sec4} ========== Nine studies have evaluated the effects of care delivery or implementation interventions designed to increase active consideration and use of pharmacologic treatment options for patients with AUD. The interventions varied widely in context, intensity, target populations, and the underlying strategies, though many strategies were shared across studies, regardless of design (care delivery or implementation intervention). As hypothesized, care delivery interventions, targeted on patients willing to be randomized, were associated with much larger and more consistent improvements in rates of medication receipt compared to implementation interventions targeted at the overall population of patients with AUD. Among the care delivery interventions evaluated, three out of four increased use of medications. However, of these three, only Oslin's Alcohol Care Management intervention improved initiation of medications for AUD with more than one third of enrolled patients (69%) and improved in patient-level alcohol use. This trial may have been distinct from the others in its recruitment approaches---patients were recruited with the knowledge that the intervention aimed to provide pharmacologic treatment \[[@CR31]\]. Among the implementation interventions evaluated, only the VA Academic Detailing Program \[[@CR40]\] showed significant promise in increasing rates of medication receipt. It may be noteworthy that, compared to the other implementation interventions, the VA Academic Detailing Program was very labor intensive and targeted to diverse clinical settings with a high density of patients with AUD, not just primary care. The study of group medication management visits, intended as a means to increase prescribing capacity and educate patients who were considering medication treatment, \[[@CR44]\] showed signals of effectiveness in one VA facility with a highly motivated champion. Interestingly, group settings have previously been identified as a barrier to receiving treatment for AUD, but appeared to facilitate increased treatment receipt among persons already seeking treatment. This intervention should be more rigorously evaluated in contexts where the primary barrier is low capacity to provide medication management. A major goal of this review was to identify the underlying implementation strategies that were positively associated with larger effects. We categorized strategies based on published reports, but then solicited feedback from the intervention designers. There was substantial heterogeneity of strategies and some heterogeneity of effects, but no clear mapping of strategies to effectiveness was apparent. This process nonetheless proved informative by highlighting potential limitations of using of Powell et al.'s taxonomy to classify implementation strategies \[[@CR30]\]. Specifically, strategies listed in the taxonomy appeared not be hermeneutically distinct, causing frequent difficulty classifying strategies as one or another. Relatedly, strategy definitions are somewhat inexplicit and hard to confidently map onto what was done in the interventions, resulting in different decisions being made by our two independent coders and between our coders and the lead or senior authors of publications. This discordance was greater when our team was not involved with the study and therefore had to rely on the published report to garner information. In all but one case, intervention developers added strategies to those identified by our 2-expert Delphi process. In some cases, the additional strategies were not fully described in the published reports. These findings suggest that an improved compilation of implementation strategies may be needed to enable accurate and reliable identification of distinct strategies. Efforts to refine such a compilation should consider designating umbrella strategies and sub-categories within them or providing a list of strategies that are similar but variable with regard to naming or minor procedural variants. Findings from our study also make clear the importance of comprehensive reporting of strategies used. While providing full descriptions of multi-faceted implementation strategies can be difficult in a single outcomes paper, authors should be encouraged to publish more detailed study protocols (as several did in the present study \[[@CR34], [@CR35], [@CR38], [@CR39]\]), and reviewers may, nonetheless, need to query intervention developers as a final validity check. Perhaps more importantly, no method has been developed to characterize the intensity of strategies or cross-classify strategies with targets. Oslin's Alcohol Care Management used many of the same strategies as other care delivery models but was targeted on patients willing to participate in an intervention focused on pharmacologic treatment. VA's Academic Detailing Program did not differ from other implementation interventions in terms of component strategies so much as intensity and diversity of targets. Developing methods to more fully characterize interventions beyond component strategies may lead to insights that have greater utility for creating generalizable knowledge. In addition, because effectiveness of implementation interventions and strategies often depends on context, methods to cross-classify strategies with context and/or setting should be developed. Beyond the aforementioned limitations of the existing implementation science tools used in this study, other limitations are worth noting. Although we searched multiple data sources and used reference lists from identified studies and networking to ensure comprehensive capture of existing studies, it is possible we missed intervention studies that aimed at increasing pharmacologic treatment of AUD. Second, our review identified only a small number of studies that reported receipt of AUD medication as a primary or secondary outcome. The small number of studies to date may limit the ability to identify generalizable information about the effectiveness of specific strategies. Moreover, of the nine studies that met inclusion criteria for this review, four were care delivery interventions tested in trials that were powered on main (clinical) outcomes. These studies may have been underpowered to detect differences in secondary outcomes, such as medication receipt. Despite these limitations, this is the first review to our knowledge conducted with the goal of understanding strategies that may be effective for implementing medication treatment for AUD---a substantially underutilized treatment. Unfortunately, our review did not reveal which strategies are most effective for implementing AUD medications. However, we cataloged the use of specific strategies, perhaps suggesting candidates for future study. Further work is needed to understand why rates of medication treatment of AUD continue to be so low, even after patients are enrolled in care management interventions and/or receiving care in a healthcare setting that has been targeted by a multifaceted intervention. It is entirely possible that previous examinations of barriers, and interventions designed to overcome them have missed the mark. To further assess this, research will be needed to better understand patient-level perspective, preferences and barriers to receipt of medications. VA : U.S. Veterans Health Administration AUD : alcohol use disorder ERIC : Expert Recommendations for Implementing Change project PPRNet-TRIP : Practice Partner Research Network-Translating Research Into Practice AHEAD : Addiction Health Evaluation and Disease CCM : Chronic Care Model EHR : electronic health record SUMMIT : Substance Use Motivation and Medication Integrated Treatment CHOICE : Choosing Healthier Drinking Options in Primary Care ADaPT--PC : Alcohol Use Disorder Pharmacotherapy and Treatment in Primary Care Settings OUD : opioid use disorder PCMHI : primary care mental health integration AHSH and ECW collaborated on the conception of the manuscript. All authors reviewed the literature, coded implementation strategies, and participated in drafting the manuscript. All authors read and approved the final manuscript. Acknowledgements {#FPar1} ================ The authors would like to acknowledge the lead and/or senior investigators of publications included in this review for coding additional implementation strategies that may not have been apparent in the published article. Competing interests {#FPar2} =================== The authors declare that they have no competing interests. Availability of data and materials {#FPar3} ================================== Data sharing is not applicable to this article as no datasets were generated or analysed during the current study. Consent for publication {#FPar4} ======================= Not applicable. Ethics approval {#FPar5} =============== Not applicable. Funding {#FPar6} ======= This study was supported in part by a VA HSR&D Research Career Scientist award (RCS 14-132) to Dr. Harris and a VA HSR&D Career Development award (CDA 12-276) to Dr. Williams. The views expressed in this article are those of the authors and do not necessarily reflect the position nor policy of the Department of Veterans Affairs (VA) or the United States government. Publisher's Note {#FPar7} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

INTRODUCTION {#s1} ============ Hepatocellular carcinoma (HCC) is one of the most common malignant tumors and ranks the third highest cause of cancer-related deaths worldwide \[[@R1], [@R2]\]. The resection rate of HCC has increased over decades due to the improvements in early diagnostic methods and surgical techniques. However, the postoperative recurrence rate and overall survival (OS) are not optimistic due to limited response to various adjunctive therapies and aggressive behaviors in advanced stages of HCC \[[@R3]\]. Thus, an accurate understanding of the biological behavior of therioma is critical in predicting the prognosis of HCC patients. Traditional prognostic factors related to the clinicopathological characteristics of the neoplasm after hepatic resection such as tumor size, vascular invasion, tumor-node-metastasis(TNM) stage, functional liver reserve and Child-Pugh class have a limited clinical value for outcome prediction \[[@R4]\]. Therefore, a variety of other potential molecular predictive markers need to be further identified. A sequential process, including escape from the primary tumor site, local invasion, systemic transport through vascular or lymphatic vessels, adhesion to distant organs, re-colonization, and expansion, is believed to be involved in hepatic carcinogenesis. Epithelial to mesenchymal transition (EMT) is known to play a pivotal role in the diffusion of cancer cells and the growth of tumors, in which epithelial cells lose their polarity and cell-cell contacts due to repressed expression of E-cadherin and up-regulated expression of mesenchymal markers such as N-cadherin, vimentin and α--smooth muscle actin (α--SMA) \[[@R5]\]. EMT could enhance not only the capacity of invasion and migration but resistance to apoptosis and chemoresistance in cancer. EMT may alter the gene expression of epithelial cells due to the activation of EMT-inducing transcription factors (EMT-TFs). In this meta-analysis, we focused on the most prominent inducers of EMT such as the zinc finger E-box binding homeobox (ZEB1 and ZEB2), the zinc-finger transcriptional repressor (Snail and Slug), and the basic helix-loop-helix transcription factor (Twist1) through searching the published literature \[[@R6], [@R7]\], knowing that EMT-TFs are directly or indirectly involved in metastasis of malignant cells through a series of signaling cascades, including the wingless-related integration site(Wnt), serine/threonine-specific protein kinase (Akt), mitogen-activated protein kinase (MAPK) and signal transducer and activator of transcription 3 (STAT3) pathways \[[@R8], [@R9]\]. During the past decade, much research has begun noticing the correlation between the expression of EMT-TFs and the prognosis of HCC. However, the results are often unconvincing due to the limited sample sizes. Here, we sought to perform a meta-analysis to evaluate clinicopathological and prognostic significance of EMT-TFs overexpression in HCC patients, especially those with high incidences of recurrence after curative resection. RESULTS {#s2} ======= Study selection and patient characteristics {#s2_1} ------------------------------------------- The initial search identified 418 potentially relevant studies. After screening, 10 published studies including 1,334 patients were selected for this pooled analysis \[[@R10]--[@R19]\]. A flowchart depicting the selection of the eligible literature is shown in Figure [1](#F1){ref-type="fig"}. {#F1} All the included studies were retrospectively analyzed, with the sample size ranging from 40 to 323 (median 133). The overexpression of EMT-TFs was reported in 662 (49.6%) of the 1,334 included patients. The highest positive expression rate was Twist1, accounting for 60.3%, followed by Snail (51.9%), ZEB2 (50.3%), ZEB1 (43.6%) and Slug (29.4%). These studies were published between 2007 and 2015. Among all cohorts, Asia was the only source region of the 10 included studies, including 9 studies from China \[[@R11]--[@R19]\] and one from Japan \[[@R10]\]. Newcastle-Ottawa Quality Assessment Scale was applied to assess these studies. The result showed that the quality scores ranged from 5 to 8 (median 6.5), indicating a relatively high study quality. Characteristics of the included studies are listed in Table [1](#T1){ref-type="table"}. All the studies focused on OS, with a median follow-up period of at least 48 (48--80) months. The definition of EMT-TFs positive expression was based on immunohistochemistry (IHC) or western blot analysis (WB) evaluation in all the eligible articles, as expressed as the percentage of positive cells or/and staining intensity. Hazard ratios (HRs) and 95% confidence intervals (CIs) were directly recorded in 8 studies \[[@R10]--[@R12], [@R15]--[@R19]\] and could be inferred from two other studies using the Tierney\'s methods described above, among which one \[[@R14]\] were calculated by variance and *P* value, and the other \[[@R13]\] was estimated only by Kaplan-Meier survival curves. ###### Characteristics of the included studies EMT-TFs Author Year Country Case EMT-TFs Positive(%) Treatment Antibody method Outcome MFu time (months) NOS score --------- ---------- ------- --------- ----------- --------------------- ----------- ------------ ------------ ---------------------- ---------------------- ----------- ---- ---- --- ZEB1 Motoyuki 2013 Japan 108 23 (21.3) Surgery goat polyclonal 1:100 SantaCruz, CA, USA IHC OS 60 8 Zhou 2011 China 110 72 (65.5) Surgery NA NA NA SantaCruz, CA, USA WB OS 60 7 ZEB2 Cai 2012 China 248 150 (60.5) Surgery rabbit polyclonal 1:100 Sigma, St.Louis, USA IHC OS 80 8 Yang 2015 China 92 21 (22.8) Surgery rabbit polyclonal 1:100 Abcam, Cambridge, UK IHC OS 60 5 Snail1 Zhou 2014 China 323 161 (49.8) Surgery NA NA NA Novus, USA IHC OS 60 6 Zhao 2012 China 97 57 (58.8) Surgery NA NA 1:250 SantaCruz, CA, USA IHC OS 60 7 Slug Zhang 2013 China 119 35 (29.4) Surgery NA NA NA Danvers, MA, USA IHC OS 60 8 Twist1 Zhang 2010 China 100 70 (70.0) Surgery rabbit polyclonal 1:50 SantaCruz, CA, USA IHC OS 76 7 Zhao 2011 China 97 51 (52.6) Surgery NA NA 1:100 SantaCruz, CA, USA IHC OS 60 7 Niu 2007 China 40 22 (55.0) Surgery rabbit monoclonal 1:50 SantaCruz, CA, USA IHC OS 48 6 Abbreviations: EMT-TFs = epithelial to mesenchymal transition-inducing transcription factors; NA = not available; IHC = Immunohistochemistry; WB = Western Blot; MFu = median Follow-up; OS = overall survival; NOS = Newcastle-Ottawa Scale. Evidence synthesis {#s2_2} ------------------ EMT-TFs and OS in HCC {#s2_3} --------------------- The pooled HR for OS indicated that EMT-TF positive expression was associated with poor OS \[HR = 1.71; 95% CI: 1.40--2.08; *p* \< 0.00001\] in HCC with a statistically significant 71% increase in the risk for mortality (Figure [2](#F2){ref-type="fig"}). Seeing that the heterogeneity test showed a *P* value of 0.08 and an I^2^ statistic index of 41%, we considered that there may be relatively substantial heterogeneity between these studies, and therefore we used the random-effects model. {#F2} Figure [3](#F3){ref-type="fig"} shows the impact of various individual EMT-TFs on the survival of HCC patients. The significantly higher HRs for OS was Slug \[HR = 2.12; 95% CI: 1.16--3.86; *p* = 0.01\]. But as the transcription factor was reported in only one study, the result should be considered with caution. In addition to ZEB2 \[HR = 1.23; 95% CI: 0.59--2.57; *p* = 0.58\], HRs for Twist1 \[HR = 2.04; 95% CI: 1.50--2.78; *p* \< 0.00001\], Snail1 \[HR = 1.87; 95% CI: 1.41--2.48; *p* \< 0.0001\], and ZEB1 \[HR = 1.61; 95% CI: 1.12--2.31; *p* = 0.01\] suggested that their positive expression correlated with poor OS. {#F3} EMT-TFs and clinicopathological features in HCC {#s2_4} ----------------------------------------------- In the meta-analysis, the pooled data revealed that the associations between EMT-TFs and the following clinicopathological features were significant: TNM stage \[III+IV vs. I+II; OR = 2.18; 95% CI: 1.08--4.38; *p* = 0.03\], histological differentiation \[poor vs. well+moderate; OR = 1.96; 95% CI: 1.22--3.17; *p* = 0.006\], intrahepatic metastasis \[pos vs. neg; OR = 2.94; 95% CI: 1.56--5.54; *p* = 0.0009\] and vascular invasion \[pos vs. neg; OR = 3.09; 95% CI: 1.67--5.73; *p* = 0.0003\]. Therefore, the findings from the subgroup analysis were consistent with the conclusion that EMT-TFs as a poor prognostic factor. However, no significant association between EMT-TFs overexpression and age (\> 55 vs. ≤ 55), gender (male vs. female), tumor size (\> 5 cm vs. ≤ 5 cm), cirrhosis (yes vs. no), hepatitis B surface antigen (pos vs. neg) or AFP (\> 20 ng/ml vs. ≤ 20 ng/ml) was found. The details of the subgroup analysis results are summarized in Table [2](#T2){ref-type="table"}. ###### Correlation of EMT-TFs overexpression and clinicopathological features in HCC variable No.of No.of OR (95% CI) *P* Heterogeneity Model used ------------------------------------------- ------- ------- ------------------- -------- --------------- ------------ -------- TNM stage (III+IV vs. I+II) 6 766 2.18 (1.08--4.38) 0.03 76 0.0008 random Differentiation (poor vs. well+ moderate) 5 435 1.96 (1.22--3.17) 0.006 0 0.58 fixed Intrahepatic metastasis (pos vs. neg) 3 258 2.94 (1.56--5.54) 0.0009 0 0.86 fixed Vascular invasion (pos vs. neg) 2 218 3.09 (1.67--5.73) 0.0003 0 0.98 fixed Age (\> 55 vs. ≤ 55) 2 219 1.27 (0.71--2.26) 0.42 48 0.16 fixed Gender (male vs. female) 7 817 1.32 (0.89--1.96) 0.17 0 0.80 fixed Tumor size (\> 5 cm vs. ≤ 5 cm) 6 687 1.09 (0.62--1.91) 0.76 63 0.02 random Cirrhosis (yes vs. no) 3 458 0.83 (0.55--1.25) 0.38 50 0.13 fixed HBSAg (pos vs. neg) 4 498 1.30 (0.75--2.26) 0.35 0 0.89 fixed AFP (\> 20 ng/ml vs. ≤ 20 ng/ml ) 4 483 0.90 (0.42--1.96) 0.80 68 0.02 random Abbreviations: HBSAg = hepatitis B surface antigen; AFP = alpha fetoprotein; pos = positive; neg = negative. Assessment of possible publication bias and sensitivity analysis {#s2_5} ---------------------------------------------------------------- The possible publication bias among these eligible studies was evaluated by applying the Begg\'s funnel plot and the Egger\'s test. As illustrated in Figure [4](#F4){ref-type="fig"}, visual assessment of the funnel plots shapes revealed no obvious publication bias for OS, and evaluation using Egger\'s test also failed to discover solid evidence for significant publication bias (*t* = 1.33; *p* = 0.220, Figure not shown). When the number of studies was smaller than 10, publication bias was not investigated because of the low sensitivity of the quantitative and qualitative tests \[[@R20]\]. In such cases, we performed the sensitivity analysis by removing one study at each time. The result demonstrated that not a single study had remarkable impact on the overall HRs. Thus, the above results further verified that the general conclusions of this current meta-analysis were credible. {#F4} DISCUSSION {#s3} ========== Early diagnostic and surgical techniques of HCC have been improved greatly within the past decade. However, recurrence and metastasis remains one of the major threats and the most critical aspect of HCC, because it is the key event causing most cancer-related deaths \[[@R1], [@R21]\]. EMT is considered to be one of the key initial steps in cancer development, progression and metastasis, knowing that EMT can induce dissemination of malignant cells, thereby increasing cell migration and invasion \[[@R22], [@R23]\]. The concept of EMT and its reverse process, mesenchymal to epithelial transition (MET), was first recognized in the field of embryology, and is now known to play diverse roles in embryonic development and a series of physiological processes such as gastrulation, neural tube formation, tissue homeostasis, wound healing, stem cell plasticity, and organ fibrosis \[[@R24], [@R25]\]. There are growing studies reporting that EMT is involved not only in tumor metastasis and progression but in cancer recurrence and resistance to conventional adjuvant therapies \[[@R26], [@R27]\]. Recent studies found that EMT-TFs were overexpressed in cancer patients, suggesting that numerous EMT-inducing transcription factors may act as primary molecular switches to induce the EMT process by activating or inhibiting the known signaling pathways \[[@R8], [@R28]\]. A meta-analysis of 3218 patients from 14 studies was published in 2016 demonstrated that the overexpression of EMT-TFs was a poor prognostic factor for metastatic breast cancer \[HR = 1.72; 95% CI: 1.53--1.93; *p* = 0.001\] \[[@R29]\]. Accordingly, there has been great interest to confirm whether EMT-TFs high expression could be used as a potential prognostic biomarker for HCC to help guide surveillance and clinical decision-making regarding adjunctive therapies. However, there is no comprehensive analysis to draw a generally accepted conclusion. In the current meta-analysis, we collected all data available from published articles to assess the correlation between EMT-TFs expression and HCC prognosis after resection for the first time. The pooled HR results suggest that the up-regulated expression of EMT-TFs (ZEB1, Snail, Slug, and Twist1) may contribute to the adverse prognosis of HCC. In addition, our study also indicates the predictive value of EMT-TFs high expression for HCC metastasis and progression. According to the results of evidence synthesis, we consider EMT-TFs high expression as a new biomarker and a risk factor for the prediction of the HCC outcome after resection. There are some possible explanations for the close association of EMT-TF high expression with poor prognosis in HCC. First of all, EMT-TFs together with other factors can specifically bind to the E-box DNA sequences within the E-cadherin promoter, recruit transcriptional corepressors and histone deacetylases, thereby repressing E-cadherin expression and acquiring the expression of mesenchymal markers, such as N-cadherin, Vimentin, and α--SMA \[[@R30], [@R31]\]. Afterwards, it regulates the EMT process directly or indirectly by activating or inactivating the known signaling pathways. Second, recent evidence has supported the discovery that EMT-TFs overexpression is closely linked to the induction of cancer stem cell (CSC) phenotype that possesses self-renewal properties in various types of human cancers, thus enhancing tumorigenesis and helping resistance to chemo/radiation therapy associated with CSC characteristics \[[@R32]--[@R34]\]. Third, the up-regulated expression of EMT-TFs induces tumor invasion and metastasis. For instance, Snail was found to induce cancer cell invasion through regulating the expression of MMP proteins in HCC \[[@R35]\]. In addition, EMT-TFs also can regulate angiogenic factors and hypoxia-inducible factor-1 alpha (HIF-1α) to promote tumor angiogenesis in HCC \[[@R36]\]. Finally, several studies have also suggested that EMT-TFs play a critical role in the regulation of anti-apoptosis and anti-cancer drug resistance \[[@R37], [@R38]\]. Consequently, EMT, CSC generation, tumor invasion and metastasis, and angiogenesis are closely associated with the transformation of cancer cells to more aggressive behavior. These roles of EMT-TFs may help partially explain why HCC patients with EMT-TF overexpression had significantly shorter OS than those with EMT-TF low expression. During EMT, tumor cells gradually lose the epithelial markers (E-cadherin, tight junction protein-1, laminin and cytokeratin) and obtain the expression of mesenchymal markers (N-cadherin, Vimentin and α--SMA). Among them, one of the essential hallmarks of EMT is the loss of E-cadherin function, which is really important to adequately understand the whole regulation mechanism of EMT-TFs as the upstream molecules of E-cadherin \[[@R39]\]. A variety of signaling pathways are triggered by EMT-TFs, including the Akt, MAPK, STAT3, transforming growth factor beta (TGFβ), β-catenin, Wnt, Ras, and Notch pathways. In addition to the classical triggering signaling pathways, some signaling molecules such as epidermal growth factor (EGF), nuclear factor kappa B (NF-κB), fibroblast growth factor (FGF), neurotrophic receptor tyrosine kinase B (Tr-kB), hepatocyte growth factor (HGF), steroid receptor co-activator (SRC)-3 protein, necrosis factor alpha (TNF-α), and HIF-1α are all activated \[[@R8], [@R9]\]. The coordination of these factors results in the repression of E-cadherin expression. Thus, success in targeting EMT-TFs via RNA interference (RNAi) technology or specific chemotherapeutic drugs will provide a new approach for the control of cancer metastasis. There are several limitations in this meta-analysis, even though efforts have been made to comprehensively evaluate clinicopathological and prognostic significance of EMT-TFs overexpression in HCC. First, different antibodies, dilution solubility and cut-off values will impact the accuracy of assessment that the positive expression of EMT-TFs. Hence, a large multicenter clinical study using the same antibody and cut-off values may be helpful to gain more credible results. Second, there may be potential language bias in this meta-analysis, because the search strategy was limited to studies published in English only. In addition, the eligible articles included only Asian populations, thus lacking the homogeneity of the population distribution. Third, not all the studies directly provided HRs and 95% CIs, so the data extracted by using Tierney\'s methods may also cause the imprecision of the original data. Despite these limitations, the results of our meta-analysis initially support the hypothesis that EMT-TF overexpression is associated with malignant phenotype features and poor postoperative OS of HCC patients in Asian populations. More investigations are needed in order to fully understand the pivotal role of each individual EMT-TF so as to provide new insights into tumor metastasis and progression, and lay a theoretical foundation for innovating target-specific drug therapies and molecular prognostic biomarkers of HCC after resection. MATERIALS AND METHODS {#s4} ===================== Literature search strategy {#s4_1} -------------------------- A comprehensive systematic literature search in the PubMed, Web of Science database and Cochrane Library was performed to retrieve all the relevant articles (deadline until December 31, 2016 ), with the limit to "human" and papers published in English. The initial electronic search strategies included using the random combination of following Medical Subject Heading (MeSH) search terms: "ZEB, Snail, Slug or Twist1 ", "hepatocellular carcinoma", and "prognosis". In addition, reference lists from identified primary articles were then once again manual cross-searched to identify any studies that were omitted by the search strategies. In the situation when multiple studies overlapped patient cohorts, only the published research with the largest sample size was included in the analysis. Data extraction {#s4_2} --------------- The titles and abstracts of all candidate articles were read independently by two reviewers (TW. and TZ.), and irrelevant ones were subsequently excluded according to the PICO principle \[[@R40]\]. Then, articles that could not be classified based on the abstracts alone were required for full-text scrutinization. Finally, eligible studies were carefully selected according to the following inclusion criteria. If any disagreement or discrepancy occurred in the eligibility of studies, the two reviewers would conduct a debate or consult the third reviewer (YZ.) until a consensus was reached. Quality assessment was conducted for each of the acceptable studies by two reviewers independently (YZ. and TZ.) using the Newcastle--Ottawa Quality Assessment Scale (NOS) \[[@R41]\]. Parameters were extracted from each included paper, including the first author\'s name, publication year, country, number of total patients, cases with positive expression rates of EMT-TFs, TNM stage, follow-up period, HRs and 95% CIs and *P*-values for OS. OS was defined as the period from the time of confirmed diagnosis of HCC to death, regardless of the patients receiving treatment or not. If the HRs were not directly shown in the article, we tried to contact the authors for additional data. If the authors did not reply, we extracted data from Kaplan-Meier survival curves by applied the Engauge Digitizer V4.1, and then the Tierney\'s methods was utilized to calculate the HRs and 95% CIs \[[@R42]\]. Criteria for inclusion and exclusion {#s4_3} ------------------------------------ To be eligible for selection of this meta-analysis, studies were required to fulfill the following criteria: (1) patients were histologically confirmed as HCC; (2) the expression of EMT-TFs (ZEB1, ZEB2, Snail, Slug, Twist1) was measured by IHC or WB; (3) studies provided the correlation between EMT-TFs and OS; (4) studies reported HRs with 95% CIs, or calculation of these statistics from the data and survival curves presented; and (5) articles were published as papers in English. Letters, reviews, editorials, abstracts, expert opinions, experiments that were performed on cell lines or animals, and articles that had inadequate original survival data for further analysis were excluded from this meta-analysis Statistical analysis {#s4_4} -------------------- All the statistical analyses were performed via Review Manager 5.3 (The Cochrane Collaboration, Oxford, UK) and Stata 12 (Stata Corporation, College Station, TX, USA) in the meta-analysis. For the pooled analysis of the correlation between EMT-TFs expression and the clinical prognosis, HRs and 95% CIs for OS were combined to calculate the effective value (logHR and SE). As for the impact of EMT-TFs on clinicopathologic parameters of HCC, the pooled ORs and 95% CIs were used. Statistical heterogeneity was evaluated through the chi-squared test and *I*^2^ test. A chi-squared *P* value \< 0.10 indicated the presence of statistically significant heterogeneity \[[@R43]\]. Pooled effects were calculated using either a fixed-effect or random-effects model \[[@R44]\]. A pooled HR \> 1 indicated a higher risk of poor survival. The potential publication bias was analyzed by the Egger\'s test and Begg\'s Funnel plots \[[@R45]\]. Sensitivity analysis was also tested by excluding each study individually. Two-tailed *P* values \< 0.05 were considered to denote statistical significance. **Authors' contributions** Conception/Design: Yanming Zhou. Provision of study materials: Tao Wan, Tianwei Zhang, Xiaoying Si. Collection and/or extract data: Xiaoying Si, Yanming Zhou. Data analysis and statistical guidance: Tao Wan, Tianwei Zhang, Yanming Zhou. Final approval of the manuscript: Tao Wan, Tianwei Zhang, Xiaoying Si, Yanming Zhou. **CONFLICTS OF INTEREST** The authors indicated no financial relationships. **FUNDING** The study was supported by Foundation of Health and Family Planning Commission of Fujian Province of China (Project no.2013-ZQN-JC-31). HCC : hepatocellular carcinoma EMT-TFs : epithelial to mesenchymal transition -inducing transcription factors ZEB : zinc finger E-box binding homeobox SNAI : zinc-finger transcriptional repressor Twist : basic helix-loop-helix transcription factor OR : odds ratios HR : hazard ratios OS : overall survival CI : 95% confidence interval IHC : immunohistochemistry WB : western blot analysis CSC : cancer stem cell MeSH : Medical Subject Heading NOS : Newcastle--Ottawa Scale

Introduction ============ Better adjuvant therapy, improved metal implants, and innovative surgical techniques have led surgeons to consider limb salvage surgery as an alternative treatment for malignant bone tumour other than amputation. Orthopaedic oncology patients have a chance for an active, disease-free life after limb salvage surgery. In the first evidence-based study, Simon *et al.* had reported the benefits of limb-salvaging procedures for bone tumours.[@b1-rado-46-03-189] Their multicentre study reported the rates of local recurrence, metastasis and survival in 227 patients with osteosarcoma in the distal femur and suggested that the Kaplan-Meier curves of the patients without recurrence were not statistically different between limb-salvaging surgery and amputation patients during a 5.5-year follow-up. Limb-salvage surgery was considered as safe as an amputation in the management of patients with high-grade osteosarcoma. The goal of limb-salvaging surgery is to preserve the function of limbs, prevent tumour recurrence, and enable the rapid administration of chemotherapy or radiotherapy.[@b2-rado-46-03-189] It can be reached with meticulous technique, detailed operative planning, and the use of endoprosthetic replacements and/or bone grafting. For a successful limb-salvage surgery in high-grade malignant tumour, such as sarcomas, a wide margin is necessary to obtain a local control.[@b3-rado-46-03-189]--[@b5-rado-46-03-189] Since marginal and intralesional margins are related to local recurrence, the reconstruction with limb-salvaging options should be carefully considered. The clinical outcome of the limb-salvage surgery with arthroplasty is closely related to the accuracy of the surgical procedure. To improve the final outcome, one must take into account the length of the osteotomy plane, as well as the alignment of the prosthesis with respect to the mechanical axis in order to keep the balance of the soft tissues. Furthermore, the parameters measured with the 3D imagine must be used during the individual manufacture of implant in order to reconstruct the skeletal structure accurately. Therefore, geometric data (such as length of leg, offset) and morphologic data are required. Magnetic resonance imaging (MRI) was beneficial for tumour detection and consequently staging of musculoskeletal neoplasia. MRI became an ideal imaging modality for musculoskeletal neoplasia because of superior soft-tissue resolution and multiplanar imaging capabilities and had a significant impact on the ability to appropriately stage lesions and adequately plan for limb-salvage surgery.[@b6-rado-46-03-189],[@b7-rado-46-03-189] In contrast, multi-slice spiral computed tomography (CT) could provide super three-dimensional morphological delineation of the diseased bone. Theoretically, the complimentary use of these two imaging modalities could give the surgeon a more accurate way to implement preoperative planning than the conventional application of 2D images. The purpose of this prospective study was to report our initial experience with limb salvage surgery for orthopaedic oncology patients by using both MR imaging and multi-slice spiral CT for preoperative planning. Patients and methods ==================== Patients and preparation ------------------------ The study protocol has complied with all relevant national regulations and institutional policies and has been approved by the local institutional ethics committee. Informed consent was obtained from all patients before the procedure. Patients with malignant bone tumours of lower/upper limb were enrolled in the study. Preoperative work-up consisted of history and clinical examination, routine laboratory tests and an aesthetic assessment, plain radiography of the limb, 64-slice spiral CT scan of the limb and chest, Technetium-99m bone scan, and, in all of the cases, MRI of the affected limb. Antibiotics were administrated before the surgery. Biopsy was performed for pathological examination. Chemotherapy was commenced 6 weeks before the surgery in those cases which were diagnosed as osteosarcoma and dedifferentiated chondrosarcoma. Patients were classified according to the Enneking staging system.[@b8-rado-46-03-189],[@b9-rado-46-03-189] The patients received a detailed narrative of conventional, surgical and amputation options after the limb salvage surgery at their own request. Nine consecutive patients with lower/upper limb malignant tumour of bone (5 women, 4 man, mean age 28.6 years, range: 19--52 years) were treated with limb-salvaging procedures. Lesion size (longitudinal direction), location and histology are summarized in [Table 1](#t1-rado-46-03-189){ref-type="table"}. MR imaging ---------- MR images were performed at a 1.5-T superconductive unit (Gyroscan Intera, Philips Medical Systems, Netherlands) and a synergy surface coil was used. The sequences included transverse, sagittal and coronal turbo spin echo T1- and fat-suppressed T2-weighted images. The parameters of these sequences were TR/TE=400 /20 ms for T1-weighted imaging, TR/TE=3500 /120 ms for T2-weighted imaging and a field of view of 480 mm×480 mm for sagittal imaging and 40 mm×40 mm for transverse imaging and 480 mm×480 mm for coronal imaging with a matrix of 512×512, 4--6 signals acquisition and a slice thickness/gap = 5/0.5 mm. Contrast enhanced sagittal, coronal and transverse T1-weighted imaging were obtained after the intravenous injection of gadopentetate dimeglumine (Magnevist, Schering, Berlin, Germany) with a dosage of 0.2 mmol/kg of body weight. Multi-slice spiral CT --------------------- CT scan was performed by using a 64-slice spiral CT (Sensation 64, Siemens Medical Systems, Germany). The raw data obtained using an axial collimation of 64×0.6 mm, a pitch of 1.0, a tubular voltage of 120 KV and a tubular current of 360 mAs, were reconstructed into contiguous 1-mm thick slices with an increment of 0.5 mm and a field of view of 376 mm × 376 mm and a matrix of 512 × 512 by using the standard soft tissue and bone algorithm. These thin-slice images were postprocessed by using the techniques of multiplanar reformation (MPR) and volume rendering (VR) to demonstrate the lesion details and perform related measurements. Preoperative planning --------------------- All preoperative radiographs were evaluated by one radiologist and two consultant orthopaedic surgeons, who were members of the surgical team performing the operations. First, the osteotomy plane was determined separately on CT and MRI. On orthogonal coronal enhanced MR images and CT MPR images, the bulk margin of the tumour in the medullary cavity was defined according to the different signal characteristics or attenuation of the tumour itself and the marrow oedema around the tumour. Then, the maximum distance from the top of the greater trochanter to this tumour margin was measured on orthogonal coronal T1-weigthted MRI images, if the tumour was located in the proximal part of femur. The maximum distance from the knee joint line to the tumour margin was measured for the tumours located in the distal part of femur. The maximum distance was defined as the intramedullary extension of the primary tumour and subsequently was used as a reference for the CT measurement. The osteotomy plane on CT MPR images was defined 30 mm distal from the margin of tumour. This distance was also used to determine the length of the extra medullary part of the prosthesis. After the osteotomy plane had been determined, the detailed shape of the medullary cavity of the preserved part of the femur was assessed using the orthogonal MPR technique for determining the diameter and length of the intra medullary part of prosthesis. Diameters of the medullary cavity at the level of the osteotomy plane and the level of the narrowest plane were measured to determine the diameter of the intra medullary stem of the prosthesis. The length of the intra- medullary stem of the prosthesis should be well matched to the length of medullary cavity of the preserved part of femur, which would be optimal if it had an equal length to the extra medullary part of prosthesis. Finally, the centre axis of the femoral shaft measured on CT was used as a reference. Offset, the distance from the central axis of the femoral shaft and the rotation centre of the femoral head, was the index used to determine the neck length of the prosthesis. Surgery ------- All patients underwent en bloc resection and customized prosthetic reconstruction. An anterolateral incision encircling the biopsy scar was used. Limb-salvage surgery consisted of intentional marginal excision, preserving important structures such as major neurovascular bundles, tendons, and ligaments. The osteotomy plane, 30 mm distal from the primary tumour was confirmed based on MRI for all patients. For patients with lesion in the proximal part of femur/humerus, the customized prosthesis was secured using methylmethacrylate cement after the resection. For patients with the tumour in the distal part of the femur, en bloc resection including the tibial plateau was performed and the customized prosthesis was secured using methylmethacrylate cement in both the tibia and femur after the resection. The extensor mechanism was reconstructed by reattachment of the patellar tendon to the slot on the tibial component. After surgery, functional rehabilitation and neoadjuvant chemotherapy were performed. Postoperative measurement ------------------------- After surgery, the patients were followed with a mean of 13 months (range, 9 to 20 months). The postoperative assessment of prosthesis was performed on plain radiography. The central axis of the femoral or humeral shaft and offset were defined. The vertical distance from the line between the top of bilateral ischial tuberosities to the femoral condylar plane was assessed to evaluate the change of the length of the lower limbs. The change of the length of the upper limbs was not assessed for those humeral tumour cases. Functional evaluation was performed in all patients using the 30-point functional classification system of the Musculoskeletal Tumour Society.[@b8-rado-46-03-189] Statistical analysis -------------------- Data were expressed as mean ± SD. All measured values were normally distributed (Kolmogorov-Smirnov test). A paired Student's *t* test was used to evaluate the differences between preoperative planning and post-operative measurements. Values for *p* \< 0.05 were considered statistically significant. The statistical analysis was done with SPSS, version 12.0 (SPSS, Inc.). Results ======= The mean postoperative functional evaluation score was 23.3 ± 2.7 (range, 15--27) according to Enneking's evaluation. Excellent or good function was achieved in all patients and all patients had preserved stable joint ([Table 2](#t2-rado-46-03-189){ref-type="table"}). There were no local recurrences, metastases or aseptic loosening determined by bone scan, CT scan, ultrasonic examination and laboratory tests in all patients until the end of the follow-up. Accuracy of determination for tumour's boundary ----------------------------------------------- To determine the accuracy of tumour boundary defined by MRT and CT, the specimens were collected from 1cm, 2cm proximal to the tumour plane and 1cm, 2cm distal as determined by MRI and CT and were examined for histopathology ([Figure 1](#f1-rado-46-03-189){ref-type="fig"},[2](#f2-rado-46-03-189){ref-type="fig"}). There was significant difference in tumour extension between MRI and CT measurements (P\<0.05). The tumour extension measured on MRI was not statistically different from the actual extension (P\>0.05), while the extension measured on CT was less than the actual extension ([Table 3](#t3-rado-46-03-189){ref-type="table"}). Accuracy of reconstruction of the limb length --------------------------------------------- Before and after operation, there was no significant difference in the length and offset of affected lower limb ([Table 4](#t4-rado-46-03-189){ref-type="table"}, [Figure 3](#f3-rado-46-03-189){ref-type="fig"}, [4](#f4-rado-46-03-189){ref-type="fig"}). Discussion ========== The effect of CT combined with MRI on the determination of invasiveness range of malignant bone tumour ------------------------------------------------------------------------------------------------------ Preoperative imaging plays an important role in determining the stage of bone tumours and then an appropriate choice of therapy for affected patients. An appropriate imaging protocol should always begin with plain radiography. If an aggressive or malignant lesion was suspected, further evaluation with cross-sectional imaging such as CT or MR imaging was needed. CT and MRI are imaging methods, often combined in diagnostic procedures of many oncology tumours.[@b10-rado-46-03-189],[@b11-rado-46-03-189] CT is useful for a detailed assessment of subtle bony lesions and anatomically complex bones. MRI is particularly useful for determining the tumour extension within medullary compartments and is able to detect tumour involvement of the adjacent muscle compartments, neurovascular structures, and joints. Fat-suppressed T2-weighted imaging proton-density weighted imaging, and contrast-enhanced T1-weighted sequences were frequently used to evaluate neurovascular bundle involvement.[@b12-rado-46-03-189]--[@b13-rado-46-03-189] Currently, MR imaging has become the modality of choice in the local staging of the primary bone tumour. Many studies have investigated the accuracy of MRI in determining the infiltration range of osteosarcoma. Sundaram *et al.* first reported that MRI would not overestimate the range of osteosarcoma, compared with histology.[@b14-rado-46-03-189] Compared with gross and microscopy examination, MRI did not overestimate or underestimate the extent of the tumour, and the false positive and false negative rate were zero. Later, O'Flanagan *et al.* found that MRI could determine the aggression radius of osteosarcoma within the accuracy of 1cm.[@b15-rado-46-03-189] For high-grade sarcomas, a wide margin is essential to obtain the local control in order to achieve a successful limb-salvage surgery.[@b16-rado-46-03-189]--[@b17-rado-46-03-189] Meyer *et al.* designed the osteotomy plane according to MRI and found that osteotomy plane could be successfully determined by MRI.[@b18-rado-46-03-189] In the present study, the aggression radius of the tumour determined by MRI and the postoperative histological examination was comparable and MRI is superior to CT for determining the tumour extension. Moreover, we found that the result of MRI was slightly larger than the actual extent. The reasons might be that the low signals of peri-tumour oedema was also assigned to the radius of the tumour, resulting in overestimation of tumour size or the preoperative chemotherapy further reduced the aggression radius of the tumour. This result was consistent with the report of O'Flanagan, who found that the aggression radius of the tumour could be evaluated accurately in coronal and sagittal views of T1-weighted images. In contrast it would be overestimated on T2-weighted or fat-suppressed T2-weighted images because of the presence of the peri-tumour oedema. We suggest that MRI was better to demonstrate peri-tumour oedema in comparison to the histological findings. Since this study does not include a long-term follow-up and a large number of patients, a further study is necessary to determine the eventual effect of MRI osteotomy plane on the long-term survival rate. The value of three-dimensional CT in the reconstruction of limbs ---------------------------------------------------------------- There is a huge variety in the human skeleton structure as to the size and shape. Therefore, an implant needs to be custom-made to be more suitable for the patient's bone structure and mechanical requirements. One major challenge is to restore the leg length adequately after the operation.[@b19-rado-46-03-189] The leg length discrepancy can affect the joint stability, can cause sciatica and low back pain, and inequable stress on the hip.[@b20-rado-46-03-189] Anja *et al* reported that in 1171 cases of total hip replacement most patients with the length of the difference less than 1 cm walked without limp, while 1/4 patients with more than 2cm difference suffered from claudication.[@b21-rado-46-03-189] Morrey found that inappropriate eccentricity was one of the factors that could induce dislocation of prosthesis.[@b22-rado-46-03-189] Therefore, reducing the eccentricity would increase the risk of dislocation. Dorr *et al.* found that both lack of strength of abductor muscles and impingement of the hip, were the important reasons for dislocation.[@b23-rado-46-03-189] Clinically, many factors could lead to hip dislocation. In the presence of the release of soft tissue around the hip and lack of strength of abductor, the decreased offset would significantly increase the incidence of hip impingement syndrome and dislocation, which would increase the instability of the hip joint and may lead to dislocation after slight changes in posture. A smaller offset might lead to excessive loads on prosthesis, and increase the incidence of proximal femoral osteolysis, prosthetic loosening and revision. Theoretically, increasing the offset can reduce the joint reaction force and then may reduce wearing of polyethylene.[@b24-rado-46-03-189] Each additional 10 mm of the offset can reduce 10% of the abductor force and 10% less force for the acetabular cup. But if the offset is too large, it can easily lead to malposition of the implant, trochanter projections, local bursitis and pain, and also can affect the transfer of stress and lead to the unequal length of limb. With the advent of multi-slice spiral CT, the development of an individualized prosthesis became realistic. High accuracy of CT provides a reliable basis for designing the individual prostheses. In this study, the three-dimensional reconstruction of CT images was performed. After the osteotomy plane was initially determined on MRI, the detailed morphological parameters were measured on MPR othorgonal planes. The prosthesis was accordingly designed. This combined use of MRI and CT measurement provided high precision for the fit of the prosthesis and excellent functional results.[@b25-rado-46-03-189] Conclusions =========== Preoperative evaluation and planning, meticulous surgical technique, and adequate postoperative management are essential for the bone tumour management. In the present study, MRI was found to be superior to CT for determining the tumour extension; the combined use of MRI and CT measurement provided high precision for the fit of the prosthesis and excellent functional results. {#f1-rado-46-03-189} {#f2-rado-46-03-189} {#f3-rado-46-03-189} {#f4-rado-46-03-189} ###### Lesion features in six patients **NO.** **Primary tumor** **Sex** **Age(y)** **Tumor characteristics** **Tumor edge disparity between CT and MR(cm)** --------- ------------------- --------- ------------ --------------------------- ------------------------------------------------ ------ ----- 1 Osteosarcoma M 20 Proximal Femur 6.5 6.0 0.5 2 Chondrosarcoma F 29 Proximal Femur 7.1 5.5 1.6 3 Osteosarcoma F 21 Distal Femur 15.5 13.3 2.2 4 Chondrosarcoma M 31 Proximal Femur 10.0 4.2 5.8 5 Osteosarcoma F 19 Proximal Femur 8.5 7.0 1.5 6 Osteosarcoma F 52 Distal Femur 9.0 7.6 1.4 7 Osteosarcoma M 41 Distal Femur 12.9 11.0 1.9 8 Osteosarcoma F 22 Proximal humerus 14.2 12.3 1.9 9 Osteosarcoma M 22 Proximal humerus 13.0 11.5 1.5 measured on MRI; measured on CT imaging. ###### Functional evaluation according to the 30-point functional classification system of the Musculoskeletal Tumour Society **Classification(points)** **Patients Score** -------------------------- ---------------------------- -------------------- -------------------------- -------------- ---------------------------------------- ---------------------------------------- --------- **Pain** None Intermediate Modest Intermediate Moderate Severe 4.5±0.7 **Emotional Acceptance** Enthusiastic Intermediate Satisfied Intermediate Accepts Dislikes 4.4±0.5 **Function** NO Restriction Intermediate Recreational Restriction Intermediate Partial Disability Total Disability 4.6±0.5 **Supports** None Intermediate Brace Intermediate One cane, One crutch Two canes, two crutches 3.8±0.5 **Walking Ability** Unlimited Intermediate Limited Intermediate Inside only Unable 4.3±0.7 **Gait** Normal Intermediate Minor Cosmetic problem Intermediate Major cosmetic problem. Minor handicap Major cosmetic problem. Major handicap 3.7±1.0 The score of postoperative functional evaluation was given as the mean and the standard deviation, which showed that excellent or good function was achieved in all patients. ###### Accuracy of CT and MRI for determining the tumour extension **Tumor margin on CT** **Tumor margin on MRI** **Position from tumor margins on CT** **Position from tumor margins on MRI** --------------------------------------------------------------- ------------------------ ------------------------- --------------------------------------- ---------------------------------------- --- --- --- --- --- --- **Positive result determined by histopathologic examination** 9 7 1 9 9 9 0 1 9 9 **Negative result determined by histopathologic examination** 0 2 8 0 0 0 9 7 0 0 The specimens were collected from 1cm, 2cm proximal to the tumours and 1cm, 2cm distal as determined by MRI and CT and were examined for histopathology (which were simplified to 1cm, 2cm, −1cm, −2cm respectively). In 9 cases, which were underestimated by CT, positive result of histopathology was determined on 1-cm-point which was distal from CT-determined boundary. In 2 cases, which were overestimated by MRI, negative result of histopathology was determined on MR-determined boundary (overestimate). ###### Preoperative and postoperative measurements of leg length and offset **No.** **Contraleral side** **Preoperative planning** **Postoperative measurement** **Disparity between preoperative and postoperative measurement** --------- ---------------------- --------------------------- ------------------------------- ------------------------------------------------------------------ ------ ----- ----- ----- **1** 39.2 4.1 38.7 4.2 39.4 4.0 0.7 0.2 **2** 36.0 4.0 37.1 4.2 36.6 4.4 0.5 0.2 **3** 38.0 3.6 38.0 3.6 38.0 3.6 0.5 0 **4** 37.3 3.4 37.0 3.4 36.5 3.2 0.5 0.2 **5** 36.5 3.5 36.0 3.6 35.5 4.0 0.5 0.4 **6** 37.5 3.7 37.0 3.7 37.2 3.7 0.2 0 **7** 37.9 3.9 37.7 3.9 37.4 3.9 0.3 0 [^1]: Jie Xu and Jun Shen contrebuted equally to this work. [^2]: Disclosure: No potential conflicts of interest were disclosed.

Building upon the pioneering work of Vicsek *et al.*[@b1], physicists, mathematicians and biologists have contemplated the self-organization of living-organism groups into flocks as an emergent process stemming from simple interaction rules at the individual level[@b2][@b3][@b4]. This idea has been supported by quantitative trajectory analysis in animal groups[@b5][@b6][@b7], together with a vast number of numerical and theoretical models[@b3][@b4], and more recently by the observations of flocking behaviour in ensembles of non-living motile particles such as shaken grains, active colloids, and mixtures of biofilaments and molecular motors[@b8][@b9][@b10][@b11][@b12]. From a physicist\'s perspective, these various systems are considered as different instances of polar active matter, which encompasses any ensemble of motile bodies endowed with local velocity--alignment interactions. The current paradigm for flocking physics is the following. Active particles are persistent random walkers, which when dilute form a homogeneous isotropic gas. Upon increasing density, collective motion emerges in the form of spatially localized swarms that may cruise in a sea of randomly moving particles; further increasing density, a homogeneous polar liquid forms and spontaneously flows along a well-defined direction[@b1][@b13][@b14]. This picture is the outcome of experiments, simulations and theories mostly performed in unbounded or periodic domains. Beyond this picture, significant attention has been devoted over the last five years to confined active matter[@b3][@b12][@b15][@b16][@b17][@b18][@b19][@b20][@b21][@b22][@b23][@b24][@b25][@b26]. Confined active particles have consistently, yet not systematically, been reported to self-organize into vortex-like structures. However, unlike for our understanding of flocking, we are still lacking a unified picture to account for the emergence and structure of such vortex patterns. This situation is mostly due to the extreme diversity in the nature and symmetries of the interactions between the active particles that have been hitherto considered. Do active vortices exist only in finite-size systems as in the case of bacterial suspensions[@b17], which lose this beautiful order and display intermittent turbulent dynamics[@b27] when unconfined? What are the necessary interactions required to observe and/or engineer bona fide stationary swirling states of active matter? In this paper, we answer these questions by considering the impact of geometrical boundaries on the collective behaviour of motile particles endowed with velocity--alignment interactions. Combining quantitative experiments on motile colloids, numerical simulations and analytical theory, we elucidate the phase behaviour of *polar* active matter restrained by geometrical boundaries. We use colloidal rollers, which, unlike most of the available biological self-propelled bodies, interact via well-established dynamical interactions[@b11]. We first exploit this unique model system to show that above a critical concentration populations of motile colloids undergo a non-equilibrium phase transition from an isotropic gaseous state to a novel ordered state where the entire population self-organizes into a single heterogeneous steadily rotating vortex. This self-organization is *not* the consequence of the finite system size. Rather, this emergent vortex is a genuine state of polar active matter lying on the verge of a macroscopic phase separation. This novel state is the only ordered phase found when unidirectional directed motion is hindered by convex isotropic boundaries. We then demonstrate theoretically that a competition between alignment, repulsive interactions and confinement is necessary to yield large-scale vortical motion in ensembles of motile particles interacting via alignment interactions, thereby extending the relevance of our findings to a broad class of active materials. Results ======= Experiments ----------- The experimental setup is fully described in the *Methods* section and in [Fig. 1a,b](#f1){ref-type="fig"}. Briefly, we use colloidal rollers powered by the Quincke electrorotation mechanism as thoroughly explained in ref. [@b11]. An electric field **E**~**0**~ is applied to insulating colloidal beads immersed in a conducting fluid. Above a critical field amplitude *E*~Q~, the symmetry of the electric charge distribution at the bead surface is spontaneously broken. As a result, a net electric torque acts on the beads causing them to rotate at a constant rate around a random axis transverse to the electric field[@b28][@b29][@b30]. When the colloids sediment, or are electrophoretically driven, onto one of the two electrodes, rotation is converted into a net rolling motion along a random direction. Here, we use poly(methyl methacrylate) (PMMA) spheres of radius *a*=2.4 μm immersed in a hexadecane solution. As sketched in [Fig. 1a](#f1){ref-type="fig"}, the colloids are handled and observed in a microfluidic device made of double-sided scotch tape and of two glass slides coated with an indium-tin-oxide layer. The ITO layers are used to apply a uniform DC field in the *z*-direction, with *E*~0~=1.6 V μm^−1^ (*E*~0~=1.1*E*~Q~). Importantly, the electric current is nonzero solely in a disc-shaped chamber at the centre of the main channel. As exemplified by the trajectories shown in [Fig. 1b](#f1){ref-type="fig"} and in [Supplementary Movie 1](#S1){ref-type="supplementary-material"}, Quincke rotation is hence restrained to this circular region in which the rollers are trapped. We henceforth characterize the collective dynamics of the roller population for increasing values of the colloid packing fraction *φ*~0~. Individual self-propulsion -------------------------- For area fractions smaller than , the ensemble of rollers uniformly explores the circular confinement as illustrated by the flat profile of the local packing fraction averaged along the azimuthal direction *φ*(*r*) in [Fig. 2a](#f2){ref-type="fig"}. The rollers undergo uncorrelated persistent random walks as demonstrated in [Fig. 2b,c](#f2){ref-type="fig"}. The probability distribution of the roller velocities is isotropic and sharply peaked on the typical speed *v*~0~=493±17 μm s^−1^. In addition, the velocity autocorrelation function decays exponentially at short time as expected from a simple model of self-propelled particles having a constant speed *v*~0~ and undergoing rotational diffusion with a rotational diffusivity *D*^−1^=0.31±0.02 s that hardly depends on the area fraction (see [Supplementary Note 1](#S1){ref-type="supplementary-material"}). These quantities correspond to a persistence length of that is about a decade smaller than the confinement radius *R*~c~ used in our experiments: 0.9 mm\<*R*~c~\<1.8 mm. At long time, because of the collisions on the disc boundary, the velocity autocorrelation function sharply drops to 0 as seen in [Fig. 2c](#f2){ref-type="fig"}. Unlike swimming cells[@b26][@b31], self-propelled grains[@b8][@b22][@b23] or autophoretic colloids[@b32], dilute ensembles of rollers do not accumulate at the boundary. Instead, they bounce off the walls of this virtual box as shown in a close-up of a typical roller trajectory in [Fig. 2d](#f2){ref-type="fig"}, and in the [Supplementary Movie 1](#S1){ref-type="supplementary-material"}. As a result, the outer region of the circular chamber is depleted, and the local packing fraction vanishes as *r* goes to *R*~c~, [Fig. 2a](#f2){ref-type="fig"}. The repulsion from the edges of the circular hole in the microchannel stems from another electrohydrodynamic phenomenon[@b33]. When an electric field is applied, a toroidal flow sketched in [Fig. 1a](#f1){ref-type="fig"} is osmotically induced by the transport of the electric charges at the surface of the insulating adhesive films. Consequently, a net inward flow sets in at the vicinity of the bottom electrode. As the colloidal rollers are prone to reorient in the direction of the local fluid velocity[@b11], this vortical flow repels the rollers at a distance typically set by the channel height *H* while leaving unchanged the colloid trajectories in the centre of the disc. This electrokinetic flow will be thoroughly characterized elsewhere. Collective motion in confinement -------------------------------- As the area fraction is increased above , collective motion emerges spontaneously at the entire population level. When the electric field is applied, large groups of rollers akin to the band-shaped swarms reported in[@b11] form and collide. However, unlike what was observed in periodic geometries, the colloidal swarms are merely transient and ultimately self-organize into a single vortex pattern spanning the entire confining disc as shown in [Fig. 3a](#f3){ref-type="fig"} and [Supplementary Movie 2](#S1){ref-type="supplementary-material"}. Once formed, the vortex is very robust, rotates steadily and retains an axisymmetric shape. To go beyond this qualitative picture, we measured the local colloid velocity field **v**(**r**, *t*) and use it to define the polarization field **Π**(**r**, *t*)≡**v**/*v*~0~, which quantifies local orientational ordering. The spatial average of **Π** vanishes when a coherent vortex forms, therefore we use its projection along the azimuthal direction as a macroscopic order parameter to probe the transition from an isotropic gas to a polar-vortex state. As illustrated in [Fig. 3b](#f3){ref-type="fig"}, displays a sharp bifurcation from an isotropic state with to a globally ordered state with equal probability for left- and right-handed vortices above . Furthermore, [Fig. 3b](#f3){ref-type="fig"} demonstrates that this bifurcation curve does not depend on the confinement radius *R*~c~. The vortex pattern is spatially heterogeneous. The order parameter and density fields averaged over time are displayed in [Fig. 3c,d](#f3){ref-type="fig"}, respectively. At first glance, the system looks phase-separated: a dense and ordered polar-liquid ring where all the colloids cruise along the azimuthal direction encloses a dilute and weakly ordered core at the centre of the disc. We shall also stress that regardless of the average packing fraction, the packing fraction in the vortex core is measured to be very close to , the average concentration below which the population is in a gaseous state, see [Fig. 3e](#f3){ref-type="fig"}. This phase-separation picture is consistent with the variations of the area occupied by the ordered outer ring, *A*~ring~, for different confinement radii *R*~c~, as shown in [Fig. 3e](#f3){ref-type="fig"}. We define *A*~ring~ as the area of the region where the order parameter exceeds 0.5, and none of the results reported below depend on this arbitrary choice for the definition of the outer-ring region. *A*~ring~ also bifurcates as *φ*~0~ exceeds , and increases with *R*~c~. Remarkably, all the bifurcation curves collapse on a single master curve when *A*~ring~ is rescaled by the overall confinement area , [Fig. 3f](#f3){ref-type="fig"}. In other words, the strongly polarized outer ring always occupies the same area fraction irrespective of the system size, as would a molecular liquid coexisting with a vapour phase at equilibrium. However, if the system were genuinely phase-separated, one should be able to define an interface between the dense outer ring and the dilute inner core, and this interface should have a constant width. This requirement is not borne out by our measurements. The shape of the radial density profiles of the rollers in [Fig. 3g](#f3){ref-type="fig"} indeed makes it difficult to unambiguously define two homogeneous phases separated by a clear interface. Repeating the same experiments in discs of increasing radii, we found that the density profiles are self-similar, [Fig. 3h](#f3){ref-type="fig"}. The width of the region separating the strongly polarized outer ring from the inner core scales with the system size, which is the only characteristic scale of the vortex patterns. The colloidal vortices therefore correspond to a monophasic yet spatially heterogeneous liquid state. To elucidate the physical mechanisms responsible for this intriguing structure, we now introduce a theoretical model that we solve both numerically and analytically. Numerical simulations --------------------- The Quincke rollers are electrically powered and move in a viscous fluid, and hence interact at a distance both hydrodynamically and electrostatically. In ref. [@b11], starting from the Stokes and Maxwell equations, we established the equations of motion of a dilute ensemble of Quincke rollers within a pairwise additive approximation. When isolated, the *i*th roller located at **r**~*i*~ moves at a speed *v*~0~ along the direction opposite to the in-plane component of the electrostatic dipole responsible for Quincke rotation[@b11]. When interacting via contact and electrostatic repulsive forces, the roller velocity and orientation are related by: Inertia is obviously ignored, and for the sake of simplicity we model all the central forces acting on the colloids as an effective hard-disc exclusion of range *b*. In addition, *θ*~*i*~ follows an overdamped dynamics in an effective angular potential capturing both the electrostatic and hydrodynamic torques acting on the colloids[@b11]: The *ξ*~*i*~\'s account for rotational diffusion of the rollers. They are uncorrelated white noise variables with zero mean and variance 〈*ξ*~*i*~(*t*)*ξ*~*j*~(*t*′)〉=2*Dδ*(*t*−*t*′)*δ*~*ij*~. The effective potential in [equation 2](#eq15){ref-type="disp-formula"} is composed of three terms with clear physical interpretations: where and **I** is the identity matrix. The symmetry of these interactions is not specific to colloidal rollers and could have been anticipated phenomenologically exploiting both the translational invariance and the polar symmetry of the surface-charge distribution of the colloids[@b34]. The first term promotes alignment and is such that the effective potential is minimized when interacting rollers propel along the same direction. *A*(*r*) is positive, decays exponentially with *r*/*H*, and results both from hydrodynamic and electrostatic interactions. The second term gives rise to repulsive *torques*, and is minimized when the roller orientation points away from its interacting neighbour. *B*(*r*) also decays exponentially with *r*/*H* but solely stems from electrostatics. The third term has a less intuitive meaning, and promotes the alignment of along a dipolar field oriented along . This term is a combination of hydrodynamic and electrostatic interactions, and includes a long-ranged contribution. The functions *A*(*r*), *B*(*r*) and *C*(*r*) are provided in the [Supplementary Note 2](#S1){ref-type="supplementary-material"}. As it turns out, all the physical parameters (roller velocity, field amplitude, fluid viscosity, etc.) that are needed to compute their exact expressions have been measured, or estimated up to logarithmic corrections, see [Supplementary Note 2](#S1){ref-type="supplementary-material"}. We are then left with a model having a single free parameter that is the range, *b*, of the repulsive *forces* between colloids. We numerically solved this model in circular simulation boxes of radius *R*~c~ with reflecting boundary conditions using an explicit Euler scheme with adaptive time-stepping. All the numerical results are discussed using the same units as in the experiments to facilitate quantitative comparisons. The simulations revealed a richer phenomenology than the experiments, as captured by the phase diagram in [Fig. 4a](#f4){ref-type="fig"} corresponding to *R*~c~=0.5 mm. By systematically varying the range of the repulsive forces and the particle concentration, we found that the (*φ*~0~, *b*) plane is typically divided into three regions. At small packing fractions, the particles hardly interact and form an isotropic gaseous phase. At high fractions, after a transient dynamics strikingly similar to that observed in the experiments, the rollers self-organize into a macroscopic vortex pattern, [Fig. 4b](#f4){ref-type="fig"} and [Supplementary Movie 3](#S1){ref-type="supplementary-material"}. However, at intermediate densities, we found that collective motion emerges in the form of a macroscopic swarm cruising around the circular box through an ensemble of randomly moving particles, [Fig. 4c](#f4){ref-type="fig"} and [Supplementary Movie 4](#S1){ref-type="supplementary-material"}. These swarms are akin to the band patterns consistently reported for polar active particles at the onset of collective motion in periodic domains[@b11][@b14]. This seeming conflict between our experimental and numerical findings is solved by looking at the variations of the swarm length *ξ*~s~ with the confinement radius *R*~c~ in [Fig. 4d](#f4){ref-type="fig"}. We define *ξ*~s~ as the correlation length of the density fluctuations in the azimuthal direction. The angular extension of the swarms *ξ*~s~/*R*~c~ increases linearly with the box radius. Therefore, for a given value of the interaction parameters, there exists a critical box size above which the population undergoes a direct transition from a gaseous to an axisymmetric vortex state. For *b*=3*a*, which was measured to be the typical interparticle distance in the polar liquid state[@b11], this critical confinement is *R*~c~=1 mm. This value is close to the smallest radius accessible in our experiments where localized swarms were never observed, thereby solving the apparent discrepancy with the experimental phenomenology. More quantitatively, we systematically compare our numerical and experimental measurements in [Fig. 3b,c](#f3){ref-type="fig"} for *R*~c~=1 mm. Even though a number of simplifications were needed to establish [equations 1](#eq13){ref-type="disp-formula"}, [2](#eq15){ref-type="disp-formula"} and [3](#eq16){ref-type="disp-formula"} (ref. [@b11]), the simulations account very well for the sharp bifurcation yielding the vortex patterns as well as their self-similar structure. This last point is proven quantitatively in [Fig. 3h](#f3){ref-type="fig"}, which demonstrates that the concentration increases away from the vortex core, where , over a scale that is solely set by the confinement radius. We shall note however that the numerical simulations underestimate the critical packing fraction at which collective motion occurs, which is not really surprising given the number of approximations required to establish the interaction parameters in the equations of motion [equation 3](#eq16){ref-type="disp-formula"}. We unambiguously conclude from this set of results that [equations 1](#eq13){ref-type="disp-formula"}, [2](#eq15){ref-type="disp-formula"} and [3](#eq16){ref-type="disp-formula"} include all the physical ingredients that chiefly dictate the collective dynamics of the colloidal rollers. We now exploit the opportunity offered by the numerics to turn on and off the four roller-roller interactions one at a time, namely the alignment torque, *A*, the repulsion torque *B* and force *b*, and the dipolar coupling *C*. Snapshots of the resulting particle distributions are reported in [Fig. 4e](#f4){ref-type="fig"}. None of these four interactions alone yields a coherent macroscopic vortex. We stress that when the particles solely interact via pairwise-additive alignment torques, *B*=*C*=*b*=0, the population condenses into a single compact polarized swarm. Potential velocity-alignment interactions are *not* sufficient to yield macroscopic vortical motion. We evidence in [Fig. 4e](#f4){ref-type="fig"} (top-right and bottom-left panels) that the combination of alignment (*A*≠0) and of repulsive interactions (*B*≠0 and/or *b*≠0) is necessary and sufficient to observe spontaneously flowing vortices. Analytical theory ----------------- Having identified the very ingredients necessary to account for our observations, we can now gain more detailed physical insight into the spatial structure of the vortices by constructing a minimal hydrodynamic theory. We start from [equations 1](#eq13){ref-type="disp-formula"}, [2](#eq15){ref-type="disp-formula"} and [3](#eq16){ref-type="disp-formula"}, ignoring the *C* term in [equation 3](#eq16){ref-type="disp-formula"}. The model can be further simplified by inspecting the experimental variations of the individual roller velocity with the local packing fraction, see [Supplementary Fig. 1](#S1){ref-type="supplementary-material"}. The roller speed only displays variations of 10% as *φ*(**r**) increases from 10^−2^ to 4 × 10^−2^. These minute variations suggest ignoring the contributions of the repulsive forces in [equation 1](#eq13){ref-type="disp-formula"}, and solely considering the interplay between the alignment and repulsion torques on the orientational dynamics of [equation 2](#eq15){ref-type="disp-formula"}. These simplified equations of motion are coarse-grained following a conventional kinetic-theory framework reviewed in[@b4] to establish the equivalent to the Navier-Stokes equations for this two-dimensional active fluid. In a nutshell, the two observables we need to describe are the local area fraction φ and the local momentum field *φ***Π**. They are related to the first two angular moments of the one-particle distribution function , which evolves according to a Fokker-Plank equation derived from the conservation of and [equations 1](#eq13){ref-type="disp-formula"} and [2](#eq15){ref-type="disp-formula"}. This equation is then recast into an infinite hierarchy of equations for the angular moments of . The two first equations of this hierarchy, corresponding to the mass conservation equation and to the momentum dynamics, are akin to the continuous theory first introduced phenomenologically by Toner and Tu[@b2][@b4]: where **Q** is the usual nematic order parameter. The meaning of the first equation is straightforward, while the second calls for some clarifications. The divergence term on the left-hand side of [equation 5](#eq26){ref-type="disp-formula"} is a convective kinematic term associated with the self-propulsion of the particles. The force field **F** on the right-hand side would vanish for non-interacting particles. Here, at first order in a gradient expansion, **F** is given by: This force field has a clear physical interpretation. The first term reflects the damping of the polarization by the rotational diffusion of the rollers. The second term, defined by the time rate *α*=(∫~*r*\>2*a*~*rA*(*r*)d*r*)/*a*^2^, echoes the alignment rule at the microscopic level and promotes a nonzero local polarization. The third term, involving *β*=(∫~*r*\>2*a*~*r*^2^*B*(*r*)d*r*)/(2*a*^2^), is an anisotropic pressure reflecting the repulsive interactions between rollers at the microscopic level. [equations 4](#eq25){ref-type="disp-formula"} and [5](#eq26){ref-type="disp-formula"} are usually complemented by a dynamical equation for **Q** and a closure relation. This additional approximation, however, is not needed to demonstrate the existence of vortex patterns and to rationalize their spatial structure. Looking for axisymmetric steady states, it readily follows from mass conservation, [equation 4](#eq25){ref-type="disp-formula"}, that the local fields must take the simple forms: *φ*=*φ*(*r*), and where *Q*(*r*)\>0. We also infer the relation from the projection of the momentum equation, [equation 5](#eq26){ref-type="disp-formula"}, on the azimuthal direction. This relation tells us that the competition between rotational diffusion and local alignment results in a mean-field transition from an isotropic state with to a polarized vortex state with and *Q*=(1)/(2)(1−*D*/(*αφ*)). This transition occurs when φ exceeds , the ratio of the rotational diffusivity to the alignment strength at the hydrodynamic level. In addition, the projection of [equation 5](#eq26){ref-type="disp-formula"} on the radial direction sets the spatial structure of the ordered phase: with again in the ordered polar phase. This equation has a clear physical meaning and expresses the balance between the centrifugal force arising from the advection of momentum along a circular trajectory and the anisotropic pressure induced by the repulsive interactions between rollers. It has an implicit solution given by *φ*(*r*) is therefore parametrized by the dimensionless number reflecting the interplay between self-propulsion and repulsive interactions. Given the experimental values of the microscopic parameters, Λ is much smaller that unity . An asymptotic analysis reveals that is the typical core radius of the vortex. For , the density increases slowly as for all *φ*~0~ and *R*~c~. As *r* reaches , it increases significantly and then grows logarithmically as away from the vortex core. However, is an integration constant, which is solely defined via the mass conservation relation: and therefore only depends on *φ*~0~ and *R*~c~. does not provide any intrinsic structural scale, and the vortex patterns formed in different confinements are predicted to be self-similar in agreement with our experiments and simulations despite the simplification made in the model, [Fig. 3e](#f3){ref-type="fig"}. In addition, [equation 8](#eq36){ref-type="disp-formula"} implies that the rollers self-organize by reducing their density at the centre of the vortex down to , the mean area fraction at the onset of collective motion, again in excellent agreement with our measurements in [Fig. 3e](#f3){ref-type="fig"}. To characterize the orientational structure of the vortices, an additional closure relation is now required. The simplest possible choice consists in neglecting correlations of the orientational fluctuations, and therefore assuming . This choice implies that [Equations 8](#eq36){ref-type="disp-formula"} and [9](#eq49){ref-type="disp-formula"} provide a very nice fit of the experimental polarization curve as shown in [Fig. 3b](#f3){ref-type="fig"}, and therefore capture both the pitchfork bifurcation scenario at the onset of collective motion and the saturation of the polarization at high packing fractions. The best fit is obtained for values of and *β*, respectively, five and two times larger than those deduced from the microscopic parameters. Given the number of simplifications needed to establish both the microscopic and hydrodynamic models, the agreement is very convincing. We are then left with a hydrodynamic theory with no free fitting parameter, which we use to compute the area fraction of the outer polarized ring where . The comparison with the experimental data in [Fig. 3f](#f3){ref-type="fig"} is excellent. Furthermore, [equations 8](#eq36){ref-type="disp-formula"} and [9](#eq49){ref-type="disp-formula"} predict that the rollers are on the verge of a phase separation. If the roller fraction in the vortex core were smaller , orientational order could not be supported and an isotropic bubble would nucleate in a polar liquid. This phase separation is avoided by the self-regulation of *φ*(*r*=0) at . Discussion ========== Altogether our theoretical results confirm that the vortex patterns stem from the interplay between self-propulsion, alignment, repulsion and confinement. Self-propulsion and alignment interactions promote a global azimuthal flow. The repulsive interactions prevent condensation of the population on the geometrical boundary and allow for extended vortex patterns. If the rollers were not confined, the population would evaporate as self-propulsion induces a centrifugal force despite the absence of inertia. We close this discussion by stressing on the generality of this scenario. First, the vortex patterns do not rely on the perfect rotational symmetry of the boundaries. As illustrated in [Supplementary Fig. 2,](#S1){ref-type="supplementary-material"} the same spatial organization is observed for a variety of convex polygonal geometries. However, strongly anisotropic, and/or strongly non-convex confinements can yield other self-organized states such as vortex arrays, which we will characterize elsewhere. Second, neither the nature of the repulsive couplings nor the symmetry of the interactions yielding collective motion are crucial, thereby making the above results relevant to a much broader class of experimental systems. For instance, self-propelled particles endowed with nematic alignment rules are expected to display the same large-scale phenomenology. The existence of a centrifugal force does not rely on the direction of the individual trajectories. Shaken rods, concentrated suspensions of bacteria or motile biofilaments, among other possible realizations, are expected to have a similar phase behaviour. Quantitative local analysis of their spatial patterns[@b10][@b12][@b15][@b16][@b17] would make it possible to further test and elaborate our understanding of the structure of confined active matter. For instance, the polar order found in confined bacteria is destroyed upon increasing the size of the confinement. The analysis of the spacial distribution of the bacteria could be used to gain insight on the symmetries and the magnitude of the additional interactions mediated by the host fluid, which are responsible for the emergence of bacterial turbulence[@b17]. In conclusion, we take advantage of a model experimental system where ensembles of self-propelled colloids with well-established interactions self-organize into macrosopic vortices when confined by circular geometric boundaries. We identify the physical mechanism that chiefly dictates this emergent behaviour. Thanks to a combination of numerical simulations and analytical theory, we demonstrate that orientational couplings alone cannot account for collective circular motion. Repulsion between the motile individuals is necessary to balance the centrifugal flow intrinsic to any ordered active fluid and to stabilize heterogeneous yet monophasic states in a broad class of active fluids. A natural challenge is to extend this description to the compact vortices observed in the wild, for example, in shoals of fish. In the absence of confining boundaries, the centrifugal force has to be balanced by additional density-regulation mechanisms[@b35][@b36]. A structural investigation akin to the one introduced here for roller vortices could be a powerful tool to shed light on density regulation in natural flocks, which remains to be elucidated. Methods ======= Experiments ----------- We use fluorescent PMMA colloids (Thermo scientific G0500, 2.4 μm radius), dispersed in a 0.15 mol l^−1^ AOT/hexadecane solution. The suspension is injected in a wide microfluidic chamber made of double-sided scotch tapes. The tape is sandwiched between two ITO-coated glass slides (Solems, ITOSOL30, 80 nm thick). An additional layer of scotch tape including a hole having the desired confinement geometry is added to the upper ITO-coated slide. The holes are made with a precision plotting cutter (Graphtec robo CE 6,000). The gap between the two ITO electrodes is constant over the entire chamber *H*=220 μm. The electric field is applied by means of a voltage amplifier (Stanford Research Systems, PS350/5000 V-25 W). All the measurements were performed 5 min after the beginning of the rolling motion, when a steady state was reached for all the observables. The colloids are observed with a × 4 microscope objective for particle tracking, particle imaging velocimetry (PIV) and number-density measurements. High-speed movies are recorded with a CMOS camera (Basler ACE) at a frame rate of 190 fps. All images are 2,000 × 2,000 8-bit pictures. The particles are detected to sub-pixel accuracy and the particle trajectories are reconstructed using a MATLAB version of a conventional tracking code[@b37]. The PIV analysis was performed with the mpiv MATLAB code. A block size of 44 μm was used. Numerical simulations --------------------- The simulations are performed by numerically integrating the equations of motion ([equations 1](#eq13){ref-type="disp-formula"} and [2)](#eq15){ref-type="disp-formula"}. Particle positions and rolling directions are initialized randomly inside a circular domain. Integration is done using an Euler scheme with an adaptive time step *δt*, and the diffusive term in the equation for the rotational dynamics is modelled as a Gaussian variable with zero mean and with variance 2*D*/*δt*. Steric exclusion between particles is captured by correcting particle positions after each time step so as to prevent overlaps. Bouncing off of particles at the confining boundary is captured using a phenomenological torque that reorients the particles towards the centre of the disc; the form of the torque was chosen so at the reproduce the bouncing trajectories observed in the experiments. Additional information ====================== **How to cite this article:** Bricard, A. *et al.* Emergent vortices in populations of colloidal rollers. *Nat. Commun.* 6:7470 doi: 10.1038/ncomms8470 (2015). Supplementary Material {#S1} ====================== ###### Supplementary Information Supplementary Figures 1-3, Supplementary Notes 1-2 and Supplementary References ###### Supplementary Movie 1 Epifluorescence movie of a dilute ensemble of colloidal rollers exploring a circular chamber. Rc=1 mm. Packing Fraction: 0.3%. Movie recorded at 100 fps, played at 25 fps. ###### Supplementary Movie 2 Emergence of a macroscopic vortex pattern. Packing fraction: 3.6%. Rc=1 mm. Epifluorescence movie recorded at 100 fps, played at 11 fps. At t=3 s, the electric field is turned on and the rollers start propelling. ###### Supplementary Movie 3 Numerical simulation of a population of rollers showing the formation of an axisymmetric vortex. Packing fraction: 10%, range of repulsive forces: b=5a. ###### Supplementary Movie 4 Numerical simulation of a population of rollers showing the formation of a finitesized swarm. Packing fraction: 4.5%, range of repulsive forces: b=2a. We benefited from valuable discussions with Hugues Chaté, Nicolas Desreumaux, Olivier Dauchot, Cristina Marchetti, Julien Tailleur and John Toner. This work was partly funded by the ANR program MiTra, and Institut Universitaire de France. D.S. acknowledges partial support from the Donors of the American Chemical Society Petroleum Research Fund and from NSF CAREER Grant No. CBET-1151590. K.S. was supported by the JSPS Core-to-Core Program 'Non-equilibrium dynamics of soft matter and information\'. **Author contributions** A.B. and V.C. carried out the experiments and processed the data. D.D., C.S., O.C., F.P. and D.S. carried out the the numerical simulations. J.-B.C., K.S. and D.B. established the analytical model. All the authors discussed and interpreted results. D.B., J.-B.C. and D.S. wrote the manuscript. D.B. conceived the project. A.B. and J.-B.C. have equally contributed to this work. {#f1} {#f2} {#f3} {ref-type="disp-formula"} (*A*+*C*). Polarized vortices emerge solely when repulsive couplings exist (*A*+*B* and *A*+*b*).](ncomms8470-f4){#f4} [^1]: These authors equally contributed to this work

{#sp1 .253} {#sp2 .254}

1. Introduction {#sec1} =============== Health care is changing dynamically in the 2010s. The economic recession and problems with recruiting professionals \[[@B1], [@B2]\], staff retention \[[@B3]\], creating healthy work environments \[[@B4], [@B5]\], and a growing demand for customer orientation \[[@B6]\] pose challenges for nurse managers\' work. More expertise in management is needed to respond to these issues. One essential area of nurse manager\'s management skills is the use of different leadership styles \[[@B7]\]. Leadership styles can be seen as different combinations of tasks and transaction behaviours that influence people in achieving goals \[[@B8]\]. Earlier studies indicate that nurse manager\'s effective leadership style is affiliated to staff retention \[[@B5]\], work unit climate \[[@B4]\], nurses\' job satisfaction \[[@B9], [@B10]\], nurses\' commitment \[[@B11]\], and patient satisfaction \[[@B12]\]. Transformational leadership style \[[@B5], [@B6], [@B13], [@B14]\] and transactional leadership \[[@B7]\] help to respond to these issues. Transformational leadership refers to the leader\'s skills to influence others towards achieving goals by changing the followers\' beliefs, values, and needs \[[@B7]\]. Transactional leadership complements and enhances the effects of transformational leadership outcomes \[[@B15]\]. There are certain skills required from nurse managers so as to be able to use these effective leadership styles. The skills include the ability to create an organization culture that combines high-quality health care and patient/employee safety and highly developed collaborative and team-building skills \[[@B1]\]. Nurse managers also need to have the readiness to observe their own behaviour \[[@B16]\] and its effects on the work unit; as a result, employees can adjust to a better leadership style. These kinds of skills are related to manager\'s emotional intelligence (EI). EI is an ability to lead ourselves and our relationships effectively \[[@B17]\]. It has been defined as the ability to observe one\'s own and others\' feelings and emotions, to discriminate among them and to use this information to direct one\'s thinking and actions \[[@B18]\]. EI is composed of personal competence and social competence. Self-awareness and self-management are reflections of personal competence, influencing the way the leader manages him/herself. Social awareness and relationship management reflect social competence, which affects how the leader manages relationships with others \[[@B17]\]. Nurse managers with that skill can easily form relationships with others, read employees\' feelings and responses accurately, and lead successfully \[[@B19]--[@B21]\]. Emotionally intelligent leaders\' behaviour also stimulates the creativity of their employees \[[@B22]\]. Goleman et al. \[[@B23]\] have identified visionary, coaching, affiliate, and democratic styles as resonant, and pacesetting and commanding styles as dissonant leadership styles. Most leaders use both resonant and dissonant leadership styles. The leadership styles of Goleman et al. are applied as the basis of this study because earlier studies refer to the significance of these styles, especially that of EI in manager\'s work. In addition, these leadership styles are one way of aiming to carry out transformational leadership. Especially visionary, coaching, affiliate, and democratic styles include elements that promote transformational leadership. Such elements are for example the leader being visionary and empowering staff \[[@B4]\]. This paper focuses on Finnish nurse managers\' leadership styles. The Finnish health care system is a strong institution where health care services are offered to all citizens and funded by taxes \[[@B24]\]. It has widely recognized that health care services in Finland are of high-quality Despite recent concerns about equity issues, Finns are in general very satisfied with their health care services. \[[@B25]\]. Consequently it is important to explore nurse managers\' leadership styles especially in this context. 2. Materials and Methods {#sec2} ======================== 2.1. Aim of the Study {#sec2.1} --------------------- The intention of this study was to explore nurses\' and supervisors\' perceptions of nurse leaders\' leadership styles. The research questions were as follows: what kind of leadership styles do nurse managers use and what are the factors affected by their leadership styles. 2.2. Participants {#sec2.2} ----------------- To achieve the aim of this study data were collected through open interviews. The majority of Finnish nurse managers, nurses, and supervisors work in hospitals or long-term facilities. Selection of participants was performed in convenience sampling \[[@B26]\]. Participants were selected paying attention to the fact they were of different ages, working in different wards and units (e.g., psychiatry, internal diseases, gerontology) in either hospitals or long-term facilities, and had worked with more than one nurse manager. The researcher contacted the participants and asked whether they were interested in taking part in the study. The participants were informed about the aim of the study. Participation was voluntary. Prior to the interviews each participant signed a form where they gave their consent to participate in the study. A total of 11 nurses and 10 supervisors, 20 women and one man, from eight Finnish hospitals and five long-term care facilities participated in the study. The age of the nurses varied between 30 and 53 and their experience in health care between 7 and 25 years. The age of the supervisors varied between 38 and 59 and their experience as supervisors between 5 and 21 years. Both nurses and supervisors had worked with many nurse leaders and they were interviewed about nurse managers in general. They thus had experience of different nurse managers on different wards and they were able to describe leadership styles from various aspects. 2.3. Data Collection and Analysis {#sec2.3} --------------------------------- Semistructured interviews were used to gather data on the perceptions of nurse managers\' leadership styles and factors affected by leadership styles. Interviews were usually carried out in the office in the participants\' workplace. All interviews were recorded with individual consent. Participants were initially asked to describe their work and earlier study and work history. They were subsequently asked about their perception of leadership styles and asked to describe the leadership styles used by their nurse managers. After that they were asked about factors affected by leadership styles. Each interview was approached individually, guided by participants\' responses. The interview sessions lasted between 30 and 85 minutes. Every interview was transcribed word for word from the recordings. Interviewing was continued until saturation of the data was achieved \[[@B27]\]. Because nurses and supervisors might have differed in their perceptions of leadership styles, the data were first analysed separately in two separate groups, following the same process for each group. Content analysis was chosen because it is a research method for making valid inferences from data to the contexts of their use \[[@B28]\]. The interview texts were read through multiple times, based on the author\'s empirical and theoretical preunderstanding of the professional area of the participating nurses and nurse managers. A structured categorization matrix of leadership styles was developed based on the primal leadership model \[[@B23]\] and research of Vesterinen et al. \[[@B29]\]. When using a structured matrix of analysis, an item of the data that does not fit the categorization frame is used to create its own concept, based on the principles of inductive content analysis \[[@B30]\]. When both the data of nurses and superiors were analysed, the results were compared. The categories and subthemes were congruent and therefore the results are presented together, albeit paying attention to differences and similarities of the perceptions of nurses and superiors. The data analysis of the factors affected by leadership styles was inductive. All the data of nurses and supervisors were analysed together. This process included open coding, creating categories, and abstraction. A classification framework of the factors was formed inductively by defining categories and sub-themes. The criteria for allocating a unit to a category were formed by asking questions if the unit was suitable to the category. The sub-themes were named using descriptive concepts and classified as "belonging" to a particular category. After that, the categories were given names \[[@B31]\]. 2.4. Trustworthiness {#sec2.4} -------------------- The trustworthiness of this study has been ensured by confirming truth value, consistency, neutrality, and transferability of this study \[[@B32]\]. When considering this study from the viewpoint of trustworthiness, there are some threats that should be taken into consideration. The researcher collected the data and performed the analysis alone and the interpretation could have been affected by her professional history \[[@B33]\]. With interviews there is a risk that respondents try to please the interviewer by reporting things they assume s/he wants to hear. The researcher confirmed the truth value of the study by selecting participants in convenience sampling. The respondents\' age distribution was wide and they worked in different units. Their perspectives and descriptions were broad and gave a diverse picture. The truth value of this study was also confirmed by analysing data as they emerged based on the interviews. To ensure the trustworthiness of the study quotes from interviews are included in the results. In view of consistency, the research process is described so that it can be repeated if necessary. This gives a possibility to understand the limitations of the process of data collection and analysis. To ensure neutrality in this study, interpretations were based on original data. This is confirmed by citations from the interview data. In this study the sample was small, consisting of Finnish nurses and supervisors, and the results only reflect their perceptions of leadership styles. As a result, transferability of results is limited. However, when considering the main objective in this study, it was not transferability of research results, but it was to enhance understanding of leadership styles and use it for future studies. 2.5. Ethical Considerations {#sec2.5} --------------------------- The data for this study were collected following approval from the administrations of the organizations. All participants were informed of the purpose of the study. They were told that their participation was voluntary and would be treated with confidentiality. Participants were asked to sign a form where they gave their consent to take part in the study. 3. Results and Discussion {#sec3} ========================= 3.1. Results {#sec3.1} ------------ Data analysis identified visionary, coaching, affiliate, democratic, commanding, and isolating leadership styles ([Figure 1](#fig1){ref-type="fig"}). Job satisfaction and commitment as well as operation and development work, cooperation and organizational climate in the work unit were the factors affected by leadership styles. ### 3.1.1. Leadership Styles {#sec3.1.1} Visionary Leadership StyleSupervisors were of the opinion that today, nurse managers use a more visionary leadership style than previously. In the past, many organizations lacked a vision of their own and had fewer possibilities to engage in development for the future. Even now, the skills of nurse managers to lead visionary development work varied. Both nurses and supervisors reported that it was characteristic of the visionary nurse manager to emphasize and discuss the vision and provide information to employees. When establishing their vision, some nurse managers provided guidelines for attaining the work unit\'s goals. These nurse managers had a systematic and purposeful leadership style, based on the knowledge of nursing science and practice. They generally worked in organizations with strategies and vision. They had clear goals and rules on how to work. Nurse managers had so-called performance development discussions with every employee once a year. During the discussion, the nurse manager explained and revised the goals and discussed the purpose of the employee\'s work together with each employee. At the same time, they agreed on the goals of the employee for the next year. Visionary nurse managers were described as being assertive and persistent in their attempts to get the work units to achieve their goals. Nurse managers with more recent education were better equipped to search for information than nurse managers with older education. In addition, they often had a clear picture of the development needs in nursing practice. Supervisors said that sometimes the fact that the organization did not have visions or direction for the future was an obstacle to a visionary leadership style. This was emphasized in cases where changes were introduced to the organization. Some nurse managers worked more on the basis of operation up until the present. The managers were guided by various situations and there were no plans for the future."*"This manager had visions and we had long-term plans, but these plans often changed."*" Nurses emphasized the importance of making the vision understandable by giving information about current issues of the work unit. Nurse manager\'s skills to provide information objectively and positively influenced the way the personnel reacted to topical issues. It was also important to explain the motivation behind decisions. Coaching Leadership StyleNurses as well as supervisors felt that nurse managers with a coaching leadership style took into consideration both the professional development of the employees and delegation of work. The employees had resources and were seen as experts and the nurse manager delegated tasks to them. The skills of employees to work independently varied. Some employees needed more coaching while others were satisfied with using their own professional skills independently. The success of delegation was affected by common instructions. They guided employees so that every employee knew his/her tasks. The employees worked and made decisions independently within the bounds agreed. The nurse managers had a significant role in supporting the employees to cope with the problems at work. They were also responsible for coordinating and organizing work in the unit as a whole."*"A nurse manager draws plans for nursing practice so that there are these areas of responsibility and everybody knows what is their area and they answer for that."*" The nurse manager paid attention to employees\' professional skills and encouraged them to study further. Both personnel\'s competence and leaders\' skills to lead influenced the development work in the unit. It was useful to clarify what kind of needs the work unit and employees had for additional education and to draw up an education plan. This plan was a meaningful basis to guide the employees to necessary training. It was each employee\'s duty to share the new knowledge with other employees. The nurse manager encouraged the employees to collect information without prompting and to think independently. S/he also gave feedback about the professional development of the employee. Affiliate Leadership StyleNurses as well as supervisors described an affiliate leadership style. Nurse managers with affiliate leadership style emphasized harmony and acceptance of difference. The employees and their best interests were the most important value to the nurse manager. They knew the rules and guidelines of the organization, but they considered the hopes and needs of employees in a flexible manner. The nurse manager had skills to understand the feelings of another person and supported him/her by listening sensitively. Both s/he and his/her personnel trusted each other. Nurses reported that this encouraged the employees to discuss their personal concerns with the nurse manager. "*"The way to act, pay attention to the employee, do you listen to her or not, that is the basic question."*" On the other hand, supervisors reported that leading could be too solicitous, in a completely motherly way. The basis of the leadership style could be supporting the well-being and job satisfaction of the employees; this might be more important than the development of nursing practise. The purpose---a harmonious atmosphere without conflicts---can be an obstacle to planned changes. "*"When there are big changes in the work unit, nurse manager is present to the employees and listen \[sic\] to them. She tries to support and say \[sic\]: there is no problem and we manage of this."*" "*"When a new employee begins to work, she leads in \[sic\] more paternalistic way and takes care of them all the time."*" Both nurses and supervisors deemed it important that the nurse manager respects differences and personal characteristics of the employees, not forgetting employees\' equality. A nurse manager who respects and accepts the employees as individuals was easy to approach. On the other hand, nurse manager\'s close friendship with employees could make it more difficult to examine the work unit and its functions objectively."*"There are managers who are very permissive and let the employees behave each in their own way, it is typical that new small managers rise beside them."*" According to the findings nurse managers sometimes behaved in a manner the employees felt to be unequal."*"It seems that if you are a strong-willed person, you are more likely to get what you want than a person who is adaptable."*" Democratic Leadership StyleBoth nurses and supervisors reported that it was typical for the democratic nurse manager to emphasize teamwork and commitment to work. All employees\' participation was important to him/her. The nurse manager worked and discussed work together with the personnel. The employees had a possibility to voice their opinions and take part in problem-solving and decision-making. However, the nurse manager was ultimately expected to be a decision-maker. "*"... and find and make the decision by thinking together and listening to opinions of the employees and discussing together; however, she is in some cases the final decision-maker."*" There were different perceptions of the nurse managers\' positions in this leadership style. On the one hand, they were deemed to be responsible for the work unit and to make reasonable decisions after discussing with the employees. On the other hand, some supervisors felt that some nurse managers did not stand out as managers, but as team members. This meant that the nurse manager\'s own tasks could be of secondary importance."*"... she is working a lot with us and she has difficulties performing her own duties as a nurse manager."*" Supervisors said that a nurse manager had an important role in cooperation and its development with the members of different professional groups and between work units. His/her skills to get the employees to commit to the common goals were deemed as significant. Planning together with the personnel formed a basis for employees\' commitment to work. That was essential for the development of the operation of the work unit."*"... leadership style influences operation as a whole, for example, how a manager gets employees to commit to common decisions"*" Commanding Leadership StyleBoth nurses and supervisors identified a commanding leadership style, characterized by an emphasis on compliancy and control. Nurses as well as supervisors reported that it was important to the nurse managers with a commanding leadership style to follow clear directions and advice which they expected to get from others, for example, their own superiors. The employees were expected to obey these orders. The nurse manager could ask employees\' opinions on how to find a solution to a problem in the work unit; usually s/he had already made a decision and it was not changed by the opinions of the employees. The nurse manager did not think it necessary to explain his/her decisions. The leadership style was described as authoritarian, hierarchical, and inflexible."*"Nurse managers who do not have the latest knowledge of leadership, they demand that there should be clear rules and laws for everything and there is no flexibility."*" Commanding leadership style was more common in the 1970s and 1980s and it was now considered traditional and out-of-date. It was, however, described as a convenient leadership style when employees are inexperienced or when there are big changes in the work unit. Nurse managers were described as controlling the behaviour of the personnel, although observations of that kind have diminished considerably. Isolating Leadership StyleBoth nurses and supervisors described that nurse managers could isolate themselves from the work unit and retire to their own room where they worked alone without active communication with the employees. In that case the employees felt that they had been left without a leader. Problematic situations like conflicts between employees often arose and they were difficult to repair. Neither the nurse manager nor the employees got the information they needed in their work. "*"The nurse manager is quite isolated, she works alone in her room, we visit her when we have something to discuss with her."*" ### 3.1.2. Factors Affected by Leadership Style {#sec3.1.2} Both nurses and supervisors reported that nurse manager\'s leadership style affects employees\' job satisfaction and commitment to work. It is felt that nurse manager\'s fairness and trust in the employees promotes their motivation and participation in work. It is important that the employees have a possibility to develop their professional skills. Leadership style contributes to job satisfaction when the nurse manager has skills to prevent and solve conflicts. All the participants reported that nurse manager\'s skills to lead the work unit and motivate people affect the success of the work unit. Often s/he has to ask for adequate resources. It is important that there are enough trained employees and the employees know and are in charge of their areas of responsibilities. Supervisors remarked on the influence of leadership style on efficiency and economy, because the fluency of operation has an impact on how much money is spent. The nurse manager\'s influence in developing and changing operation is very important. It is important that the employees have a possibility to take part in development work as well. Nurse manager\'s leadership style can promote or hinder development in the work unit. Supervisors emphasized that nurse managers have a significant role in cooperation within the work unit and outside it. Some nurse managers want to work only inside their own unit, while others take a larger view of the matter. Nurse manager\'s leadership style has an influence on how externally orientated the staff are and whether they have connections outside the work unit. A nurse manager can promote the continuity of patient care by cooperation with other units. S/he is a role model in how to treat nurse students. Both nurses and supervisors felt that problems in the organizational climate, such as conflicts between the employees or dissatisfaction with the nurse leader are reflected in patient care. The activity or passivity of the nurse manager affects the image of the work unit."*"If there is patient mistreatment, it is the nurse manager whose responsibility it is to decide how to react, for example, "in our unit we treat patients well" or "we do not react at all to this complaint"."*" All in all, organizational climate, personnel\'s job satisfaction and commitment, work unit\'s operation and development work, and cooperation influence the way patient care succeeds and how a patient experiences the care he/she gets. Leadership style has an effect on patient satisfaction and quality of care. If the nurse manager\'s basic value is good patient care, it influences in many ways his/her leadership style and how s/he organizes things in the work unit. 3.2. Discussion {#sec3.2} --------------- The discussion is structured around the findings identified above. An isolating leadership style was identified as distinct from the leadership styles that Goleman \[[@B23]\] presented, whereas pace-setting leadership style was not reported. The participants reported that nurse managers used many leadership styles, but normally they had one which they used more than others. Nurses who worked for leaders with resonant leadership styles were more satisfied with supervision and their jobs \[[@B34]\]. Furthermore, visionary leadership style, coaching leadership style, affiliate leadership style, and democratic leadership style seem to promote transformational leadership because they motivate and involve staff. That is why nurse managers should develop themselves in the use of these leadership styles. Nurse managers\' leadership style depends on many issues, such as organization, situation, and employees. Reynolds and Rogers \[[@B35]\] argue that employees have variable levels of competence depending on the situation. That requires managers to adapt their leadership style. It is important that nurse managers have skills to reflect on their own leadership style and receive feedback about it. That gives them tools to use different leadership styles in different situations. Health care is meeting ever-increasing new challenges where it has to react rapidly. It is important to the health care organizations to make long-term plans and prepare for the future by paying attention to the needs of inhabitants and the resources needed. The vision is the basis of the goals of the work unit, too. Having knowledge of nursing science and practice gives nurse managers the tools to use a visionary leadership style and make plans for the future. Morjikian et al. \[[@B36]\] argued that communication of future plans, goals, and strategies is important between the nurse manager and the employees. It is important to give information of the vision and explain it regularly to the employees, because sometimes the employees forget the purpose of their work and their working style is not appropriate. When nurse managers work like this they are also carrying out transformational leadership \[[@B4], [@B5]\]. In the future, securing skilled employees will be a big challenge in health care. Vesterinen et al. \[[@B29]\] found that nurse managers with a coaching leadership style appreciated employees\' professional skills and encouraged them to study further. Nurse manager\'s consideration of employees\' profession and educational needs influenced nurse retention positively. Kenmore \[[@B37]\] argued that a coaching leadership style works when the employees are keen to develop and make use of possibilities to do so. Education gave employees tools to work and make decisions independently. Although the nurse manager organizes the work unit as a whole and is responsible for the development work in the unit, his/her support has a significant role in helping employees to cope with the problems they meet in their work. This is also an important part in nurse managers\' role as emotionally intelligent leaders \[[@B10]\]. As a consequence of globalization, both employees and patients come from many different cultures. Their behaviour and habits to express their needs vary. An affiliate leadership style with acceptance of difference could be suited for the multicultural work unit. It is a challenge for the manager to listen sensitively and consider to employees\' personal needs individually and at the same time objectively, not forgetting employees\' equality. This requires an emotionally intelligent nurse manager \[[@B10]\]. The basis of the leadership style could be supporting the well-being and job satisfaction of the employees. As Kenmore \[[@B37]\] argues, if a nurse manager is too concerned with creating harmony, it can lead to evasion of problems. Because of shortage of employees, nurses have many possibilities to choose and change their workplaces. Democratic leadership style promoted employees\' commitment to work \[[@B29]\]. It is important that the employees can express their opinions and take part in decision-making. A commanding leadership style prevents the empowerment of the nurses, because they do not have possibilities to participate in work planning \[[@B38]\]. However, there are situations where a commanding leadership style is appropriate. The majority of Finnish nurses will retire in the next few years and there are many nurses with less work experience in the work units. Employees with less work experience may need clear directions, for example, in acute situations when a patient\'s life is in danger. According to Huston \[[@B1]\], essential nurse manager competencies for the future include the ability to create an organization culture that combines high-quality health care and patient/employee safety and highly developed collaborative and team building skills. As a result of this study, an isolating leadership style was found: the nurse manager worked alone without active communication with the employees. The employees have to work without a leader, and that could cause anxiety for the employees who need support from their leader. A good question in this case is who is really leading the work unit. If the leader does not show consideration towards the employees, it could affect their health and well-being negatively \[[@B39]\]. Nurse managers need support to develop the leadership style. Leaders and their supervisors should be considered collectively to understand how leadership influences employee performance \[[@B40]\]. A nurse manager has an important role in leading the work unit as a whole. A work unit is seen as a reflection of the nurse manager. According to Rosengren et al. \[[@B41]\], nurses reported that nursing leadership was considered "being present and available in daily work," "facilitating professional acknowledgement," "supporting nursing practice" and "improving care both as a team and as individuals." A nurse manager with an emotionally intelligent leadership style creates a favourable work climate characterized by innovation, resilience, and change \[[@B42]\]. Nurse managers have to be flexible in the changes they have directly initiated or by which they have been indirectly affected \[[@B43]\]. Leadership style affects the organizational climate and the ways how information is given and communicated and how questions of the day are discussed. The nurse manager creates the basis for how different opinions are handled and problems solved in the work unit. Nurse manager\'s leadership style affects the personnel\'s job satisfaction and commitment. It is perceived that nurse manager\'s trust in the employees promotes their motivation and participation in work. Way et al. \[[@B44]\] found that trust and job satisfaction are strong links with greater commitment and intent to stay on at work. Nurse managers create basic preconditions for the operation and for development work. Leader encourages the employees to develop goals and plan to achieve them. In this way he/she influences the professional development of the personnel \[[@B45]\]. Their skills to build bonds and seek out mutually beneficial relationships affect cooperation in the work unit and around it. On the other hand, there is no one and only correct leadership style; the same result can be achieved in many ways. A manager who has the ability to reflect on his/her own behaviour, that is, who has high EI, is better able to regulate and estimate his/her leadership style with different employees in different situations. Leadership style influences patient care and its quality at least indirectly. A nurse manager has a significant role in using a leadership style that promotes good patient care. 4. Conclusions {#sec4} ============== Nurse managers had many leadership styles, but normally they had one that they used more than the others. The nurse managers should consider their leadership style from the point of view of employees, situation factors, and goals of the organization. Leadership styles where employees are seen in a participative, active role have become more common. Together with health care organizations, nursing education programmes should include education of nurse managers to improve their self reflection, through which they are better able to vary their leadership style. {#fig1} [^1]: Academic Editor: Linda Moneyham

1. Introduction {#sec1} =============== Soil organic C (SOC) and total N (TN) are very important C and N pools in the terrestrial ecosystems \[[@B1], [@B2]\]. As the components of labile C and N pools in soils, dissolved organic C (DOC) and N (DON) and soil ammonium and nitrate N (NH~4~ ^+^-N and NO~3~ ^−^-N) play crucial roles in the biogeochemistry of C and N and in the nutrient transformation \[[@B3]--[@B5]\]. With the context of climatic warming, how SOC, TN, DOC, DON, NH~4~ ^+^-N, and NO~3~ ^−^-N respond is vital to global C and N cycling \[[@B1], [@B2]\]. However, inconsistent results on the responses of these C and N pools to climatic warming have been observed with respect to vegetation types and initial soil characteristics \[[@B2], [@B3], [@B6]--[@B14]\]. For example, He et al. \[[@B2]\] demonstrated that six-year warming (\~1.4°C increase of 10 cm soil temperature) significantly decreased soil C by 129.3 g m^−2^ in a temperate steppe of Inner Mongolia. In contrast, Li et al. \[[@B7]\] found that two-year warming significantly increased SOC in an alpine meadow (\~2.1°C increase of air temperature) but significantly reduced TN in an alpine swamp meadow (\~2.3°C increase of air temperature) on the Tibetan Plateau. Hagedorn et al. \[[@B13]\] indicated that one-growing-season warming (\~4°C increase of 5 cm soil temperature) did not significantly influence DOC. Song et al. \[[@B1]\] pointed out that six-year warming (\~1.2°C increase of 10 cm soil temperature) significantly reduced DOC in a temperate steppe in Inner Mongolia. Biasi et al. \[[@B15]\] indicated that two-year warming (\~0.9°C increase of 5 cm soil temperature) did not have obvious effects on DON, NH~4~ ^+^-N, NO~3~ ^−^-N, and N~min⁡~ in a lichen-rich dwarf shrub tundra in Siberia. Bai et al. \[[@B14]\] stated that experimental warming (\~0.6--6.7°C in soil temperature) had a significant positive effect on N~min⁡~ but not on TN across all biomes. Therefore, how climatic warming acts on C and N cycling still remains unclear. More than 70% of the Tibetan Plateau is covered with grasslands \[[@B16]\]. The alpine grasslands of this Plateau are one of the systems most sensitive to global change \[[@B17], [@B18]\]. In alpine grasslands, understanding the responses of SOC, DOC, TN, DON, NH~4~ ^+^-N, and NO~3~ ^−^-N to climatic warming are crucial for predicting future changes in soil fertility and C sequestration. The alpine meadow is one of the most typical grasslands types on the Tibetan Plateau being subjected to climatic warming \[[@B19]\]. Information on how these C and N pools along an elevation gradient respond to climatic warming is scarce on the Tibetan Plateau. Here we set up a warming experiment in an alpine meadow at three elevations (i.e., 4313 m, 4513 m, and 4693 m) on the Northern Tibetan Plateau. The main objective was to investigate the effects of short-term experimental warming on SOC, TN, DOC, DON, NH~4~ ^+^-N, and NO~3~ ^−^-N. Our previous study indicated that short-term experimental warming could not affect soil microbial biomass \[[@B20]\] and soil microbial activity regulated the balances of soil C and N pools in the alpine meadow \[[@B21]\]. We hypothesized that experimental warming may not affect these C and N pools in this study. 2. Materials and Methods {#sec2} ======================== 2.1. Study Area, Experimental Design, and Soil Sampling {#sec2.1} ------------------------------------------------------- A detailed description of the study area, the warming experimental design, the measurements of microclimate factors (including soil temperature and soil moisture), and the soil sampling are given in Fu et al. \[[@B20], [@B22]\]. Briefly, three alpine meadow sites were established at three elevations (i.e., a low (30°30′N, 91°04′E, and 4313 m), mid- (30°31′N, 91°04′E, and 4513 m), and high (30°32′N, 91°03′E, and 4693 m) elevation) at Damxung Grassland Observation Station of Tibet Autonomous Region in China in May 2010. Annual mean air temperature and precipitation is 1.3°C and \~476.8 mm, respectively \[[@B20], [@B21]\]. The vegetation is*Kobresia*-dominated alpine meadow and roots are mainly concentrated in the topsoil layer (0--20 cm) \[[@B21], [@B22]\]. The soil is classified as sandy loam, with pH of 6.0--6.7, organic matter of 0.3--11.2%, and total N of 0.03--0.49% \[[@B20], [@B22]\]. Open top chambers (OTCs, 3 mm thick polycarbonate) were used to enhance temperature \[[@B22], [@B23]\]. The bottom and top diameters and the height of OTCs were 1.45 m and 1.00 m and 0.40 m, respectively \[[@B20], [@B22]\]. For each site, four OTCs and their paired control plots (1 m × 1 m) were randomly established in May 2010. There was \~3 m distance between plots. Daily mean soil temperature (*T* ~*s*~) during the study period of July-September in 2011 inside the OTCs increased by 1.26°C, 0.98°C, and 1.37°C at the low, mid-, and high elevation, respectively, compared to control plots \[[@B20]\]. In contrast, experimental warming decreased daily mean soil moisture (SM) by 0.04 m^3^ m^−3^ in all sites \[[@B20]\]. Daily mean *T* ~*s*~ decreased with increasing elevation from the low to high elevation \[[@B20]\]. We collected topsoil samples (0--20 cm depth) inside each plot using a probe 3.0 cm in diameter on July 7, August 9, and September 10, 2011 \[[@B20]\]. Five soil subsamples were randomly sampled and composited into one soil sample for each plot \[[@B20]\]. Subsamples of the fresh soil were used to measure DOC, DON, NH~4~ ^+^-N, and NO~3~ ^−^-N and other subsamples of the fresh soil were air-dried for the measurements of SOC and TN. 2.2. Soil Analysis {#sec2.2} ------------------ A more detailed description of measurements of soil inorganic N (N~min⁡~, i.e., sum of NH~4~ ^+^-N and NO~3~ ^−^-N), DON, and DOC can be found in Fu et al. \[[@B21]\]. Briefly, soil inorganic N in 20 g fresh soil sample was extracted with 100 mL K~2~SO~4~, filtered through 0.45 *μ*m membrane, and analyzed on a LACHAT Quikchem Automated Ion Analyzer. Dissolved organic C and TN (DTN) in another 20 g fresh soil sample was extracted with 100 mL ultrapure water and filtered through 0.45 membrane. The extractable SOC and TN concentrations in the ultrapure water extracts were measured using a Liqui TOC II elementar analyzer (Elementar Liqui TOC, Elementar Co., Hanau, Germany) and a UV-1700 PharmaSpec visible spectrophotometer (220 nm and 275 nm), respectively. We also analyzed dissolved inorganic N (DIN) in the ultrapure water extracts on a LACHAT Quikchem Automated Ion Analyzer. Then DON was calculated as the difference between DTN and DIN. The potassium dichromate method was used to determine SOC \[[@B24]\]. Soil TN was measured on a CN analyzer (Elementar Variomax CN). Soil microbial biomass (MBC) and N (MBN) data were obtained from Fu et al. \[[@B20]\]. 2.3. Statistical Analysis {#sec2.3} ------------------------- In order to examine the elevation effect, repeated-measures ANOVA with experimental warming and elevation as the between subject factors and with sampling date as the within subject factor was performed for a specific soil property (i.e., SOC, TN, DOC, DON, ratio of DOC to DON (DOC/DON), NH~4~ ^+^-N, NO~3~ ^−^-N, ratio of NH~4~ ^+^-N to NO~3~ ^−^-N(NH~4~ ^+^-N/NO~3~ ^−^-N), and N~min⁡~). At each site, repeated-measures ANOVA with experimental warming (i.e., OTCs versus control) as the between subject factor and with sampling date as the within subject factor was conducted for each soil property. Single factor linear regressions were performed between soil properties and *T* ~*s*~, SM, MBC, and MBN. In addition, multiple stepwise regression analyses were conducted for soil properties to examine the relative importance of *T* ~*s*~, SM, MBC, and MBN in affecting the variations of soil properties. All data were examined for normality and homogeneity before analysis and natural logarithm transformations were made if necessary. The level of significance was *P* \< 0.05. All the statistical tests were performed using the SPSS software (version 16.0; SPSS Inc., Chicago, IL). 3. Results {#sec3} ========== 3.1. Effects of Experimental Warming on Soil Properties {#sec3.1} ------------------------------------------------------- Regardless of experimental warming, elevation had significant effects on SOC (*F* = 183.19, *P* \< 0.001), TN (*F* = 126.38, *P* \< 0.001), DOC (*F* = 26.42, *P* \< 0.001), DON (*F* = 7.08, *P* \< 0.01), NH~4~ ^+^-N(*F* = 71.98, *P* \< 0.001), NH~4~ ^+^-N/NO~3~ ^−^-N(*F* = 14.01, *P* \< 0.001), and N~min⁡~(*F* = 56.29, *P* \< 0.001) across the three sampling dates. In contrast, there were no significant effects of elevation on NO~3~ ^−^-N and DOC/DON. These C and N pools showed similar seasonal dynamics regardless of experimental warming among the three elevations ([Figure 1](#fig1){ref-type="fig"}). In line with our initial hypothesis, experimental warming had little effects on SOC, TN, DOC, DON, DOC/DON, and NH~4~ ^+^-N/NO~3~ ^−^-N ([Table 1](#tab1){ref-type="table"}). In contrast, the sensitivity of N~min⁡~ to experimental warming increased with increasing elevation ([Table 1](#tab1){ref-type="table"}). In detail, experimental warming significantly decreased N~min⁡~ by 29.2% and 23.5% at the low and mid-elevation, NO~3~ ^−^-N by 36.4%, 29.5% at the low and mid-elevation, and NH~4~ ^+^-N by 16.7% at the mid-elevation across all the three sampling dates, respectively. In contrast, experimental warming had little effects on NO~3~ ^−^-N and N~min⁡~ at the high elevation. 3.2. Relationships between Soil Properties and Environmental Variables and Soil Microbial Biomass {#sec3.2} ------------------------------------------------------------------------------------------------- Soil organic C, TN, DOC, NH~4~ ^+^-N, NO~3~ ^−^-N, NH~4~ ^+^-N/NO~3~ ^−^-N, and N~min⁡~ were significantly and positively correlated with SM ([Figure 2](#fig2){ref-type="fig"}). In contrast, SOC, TN, DOC, NH~4~ ^+^-N, and NH~4~ ^+^-N/NO~3~ ^−^-N declined with increasing *T* ~*s*~ ([Table 2](#tab2){ref-type="table"}). The negative correlations of *T* ~*s*~ with DON and N~min⁡~ were relatively lower ([Table 2](#tab2){ref-type="table"}). Soil organic C, TN, DOC, DON, NH~4~ ^+^-N, NH~4~ ^+^-N/NO~3~ ^−^-N, and N~min⁡~ increased significantly with increasing MBC and MBN, while NO~3~ ^−^--N only increased significantly with increasing MBN ([Table 2](#tab2){ref-type="table"}). Nitrate N was not related to MBC and *T* ~*s*~ ([Table 2](#tab2){ref-type="table"}), while DON was not correlated with SM (data not shown). In addition, DOC/DON was not correlated with *T* ~*s*~, SM, MBC, and MBN (data not shown). The multiple stepwise regression analyses were listed in [Table 3](#tab3){ref-type="table"}. Both SOC and TN were simultaneously affected by MBC and *T* ~*s*~, whereas MBC explained more variation of the two soil properties than *T* ~*s*~. Only MBC was included in the multiple regression equations for DOC, DON, and NH~4~ ^+^-N/NO~3~ ^−^-N, while only MBN was included in the regression equation for NO~3~ ^−^-N. Soil microbial biomass C explained the variation of NH~4~ ^+^-N more than SM. Both MBC and MBN were simultaneously and positively correlated with N~min⁡~. In addition, all the five concerned variables were excluded for DOC/DON. 4. Discussion {#sec4} ============= 4.1. Effects of Experimental Warming on SOC, TN, DOC, and DON {#sec4.1} ------------------------------------------------------------- Recently, some studies showed that short-term (\<3 years) experimental warming had little effects on SOC, TN, DOC, and/or DON in a tallgrass prairie with a silt loam soil (\~2°C increase of 5 cm soil temperature) in USA \[[@B25]\], in a dragon spruce plantation with a mountain brown soil (\~0.6°C increase of 5 cm soil temperature) on the Tibetan Plateau \[[@B8]\], in an alpine treeline with a sandy Ranker and Podzols soil (\~4°C increase of 5 cm soil temperature) in Switzerland \[[@B13]\], and in a lichen-rich dwarf shrub tundra with Gleyic Cryosols soils (\~0.9°C increase of 5 cm soil temperature) in Siberia \[[@B15]\]. However, other studies with long-term (\>3 years) experimental warming indicated that warming significantly increased or decreased SOC, TN, DOC, and/or DON in a temperate steppe with a Calcic Kastanozems soil in Inner Mongolia (\~1.4°C increase of 10 cm soil temperature) \[[@B2]\], in an alpine meadow (\~3°C increase of 5 cm soil temperature) on the Tibetan Plateau \[[@B3]\], and in a temperate steppe with chestnut soil in Inner Mongolia (\~1.2°C increase of 10 cm soil temperature) \[[@B1]\]. Therefore, the insignificant responses of SOC, TN, DOC, and DON to warming ([Table 1](#tab1){ref-type="table"}) may be due to the short period of warming treatment (14--16 months). A meta-analysis showed that the effects of experimental warming on N~min⁡~, net N mineralization, and nitrification were significantly and positively correlated with raised soil temperature (\~0.6--6.7°C for N~min⁡~, \~0.6--5.5°C for net mineralization, and \~1.3--5.5°C for net nitrification) across all biomes \[[@B14]\]. Similarly, we found that experimental warming-induced change of soil temperature tended to be negatively correlated with that of TN (*R* ^2^ = 0.43, *P* = 0.057) and positively correlated with that of MBN (*R* ^2^ = 0.43, *P* = 0.056) \[[@B20]\]. In addition, MBN was significantly correlated with SOC, TN, DOC, and DON ([Table 2](#tab2){ref-type="table"}). Therefore, the negligible responses of soil C and N pools to experimental warming ([Table 1](#tab1){ref-type="table"}) may be also due to lower warming magnitude in this alpine meadow. Microbial activity regulates the production of dissolved organic matter \[[@B5], [@B8], [@B26]\] and experimental warming-induced decline in soil moisture may suppress soil microbial activity \[[@B20], [@B27]\]. Similarly, we also found that soil C and N pools increased with increasing soil microbial biomass and soil moisture ([Figure 2](#fig2){ref-type="fig"}, [Table 2](#tab2){ref-type="table"}). Moreover, short-term experimental warming had little effect on soil microbial biomass in this system \[[@B20]\]. Therefore, the negligible responses of SOC, TN, DOC, and DON to short-term experimental warming may be also related to that of soil microbial biomass \[[@B8], [@B20]\]. Moreover, experimental warming-induced soil drying may also suppress the production of DOC and DON \[[@B8], [@B20]\]. 4.2. Effects of Experimental Warming on Soil Inorganic N {#sec4.2} -------------------------------------------------------- Bai et al. \[[@B14]\] demonstrated that experimental warming did not significantly increase net N nitrification in grasslands. Similarly, experimental warming did not increase net N mineralization in an alpine meadow on the Tibetan Plateau \[[@B28]\]. In the same alpine meadow as this study, the finding that experimental warming did not increase ecosystem photosynthesis and aboveground plant biomass \[[@B22]\] also indirectly supported that experimental warming may not increase soil N availability because it has been observed that plant productivity is positively correlated with net N mineralization \[[@B29]\]. Therefore, the negligible or negative effect of experimental warming on soil inorganic N ([Figure 1](#fig1){ref-type="fig"}, [Table 1](#tab1){ref-type="table"}) may result from the suppression of net N mineralization and nitrification under warming. The suppression of net N mineralization and nitrification may be owing to decreases in soil moisture and microbial activity because N~min⁡~, NH~4~ ^+^-N, and NO~3~ ^−^-N increased significantly with increasing soil moisture and microbial biomass ([Figure 2](#fig2){ref-type="fig"}, [Table 2](#tab2){ref-type="table"}). Similarly, the experimental warming-induced significant reductions or insignificant changes of inorganic N ([Figure 1](#fig1){ref-type="fig"}, [Table 1](#tab1){ref-type="table"}) were also partly attributed to experimental warming-induced decline in soil microbial biomass \[[@B20]\] and soil drying \[[@B10], [@B29], [@B30]\]. This was in line with the finding that the effect of experimental warming on soil moisture was significantly correlated with that on soil nitrification \[[@B14]\]. On the other hand, microbial biomass was more closely related to soil inorganic N than soil moisture ([Table 3](#tab3){ref-type="table"}). This implied that microbial biomass may dominate the variation of soil inorganic N in this study. However, our previous study showed that short-term experimental warming tended to reduce microbial biomass due to soil drying in the same alpine meadow as this study \[[@B20]\]. Therefore, the experimental warming-induced changes of soil inorganic N, net N mineralization, and nitrification may be directly related to that of microbial activity and indirectly related to that of soil moisture. The different responses of N~min⁡~ to experimental warming among the three elevations across the sampling dates could be attributed to several probable underlying mechanisms. First, DON is high-quality N source for N mineralization \[[@B8], [@B31]\]. This was supported by the positive relationships between DON and N~min⁡~ and NH~4~ ^+^-N and NO~3~ ^−^-N ([Figure 3](#fig3){ref-type="fig"}). DON under warmed plots tended to be decreased by 10.3% at the low elevation and by 28.7% at the mid-elevation but to be increased by 4.4% at the high elevation across all the three sampling dates, compared to control plots. Second, experimental warming-induced different changes in soil microbial biomass N (MBN) among three elevations \[[@B20]\] could partly explain this phenomenon considering that the production of DON and the immobilization of soil inorganic N were regulated by MBN \[[@B3], [@B32], [@B33]\]. This viewpoint was confirmed by the positive correlations between MBN and DON, N~min⁡~, NH~4~ ^+^-N, and NO~3~ ^−^-N ([Table 2](#tab2){ref-type="table"}). Third, the response of soil N availability to warming could be strongly related to the initial conditions \[[@B8], [@B34]\]. In our system, N~min⁡~, DON, and microbial biomass at the high elevation were significantly larger compared to the low and mid-elevation, whilst there were insignificant differences between the latter two \[[@B20]\]. 5. Conclusions {#sec5} ============== In summary, short-term experimental warming had no obvious effects on topsoil organic C, total N, dissolved organic C, and N pools for the alpine meadow in this study. The insignificant responses of these C and N pools to warming may be due to short-term warming treatment, experiment warming-induced soil drying, and lower warming magnitude. In contrast, the response of soil inorganic N to experimental warming differed among the three elevations, which may be attributed to different response trends of dissolved organic N and microbial biomass and different initial soil inorganic N. This work was funded by the National Natural Science Foundation of China (no. 41171084) and the National Science and Technology Plan Project of China (no. 2011BAC09B03). Conflict of Interests ===================== The authors declare that there is no conflict of interests regarding the publication of this paper. {#fig1} {#fig2} {#fig3} ###### Repeated-measures ANOVA (*F* values) for the main and interactive effects of experimental warming (W) and sampling date (D) on soil organic C (SOC), total N (TN), dissolved organic C (DOC), N (DON), ammonium N (NH~4~ ^+^-N), nitrate N (NO~3~ ^−^-N), the ratio of NH~4~ ^+^-N to NO~3~ ^−^-N (NH~4~ ^+^-N/NO~3~ ^−^-N), and soil inorganic N (N~min~, i.e., sum of NH~4~ ^+^-N and NO~3~ ^−^-N) in an alpine meadow on the Tibetan Plateau at three elevations (*n* = 4). Elevation Model SOC TN DOC DON DOC/DON NO~3~ ^−^-N NH~4~ ^+^-N NH~4~ ^+^-N/NO~3~ ^−^-N N~min~ ----------- ------- -------- ------------- ---------- -------- ------------- ------------- ------------- ------------------------- ----------- 4313 m W 0.02 1.58 0.00 0.23 0.70 39.02\*\* 4.22 1.38 26.87\*\* D 0.31 0.26 16.70\*\*\* 3.66 2.68 55.47\*\*\* 10.98\*\* 10.04\*\* 28.71\*\*\* W × D 1.04 6.87\* 4.13\* 2.51 1.40 6.32\* 3.44 0.49 5.91\* 4513 m W 1.43 0.03 4.07 5.33 4.52 9.90\* 6.45\* 3.01 10.89\* D 2.62 0.23 94.06\*\*\* 0.99 5.23 57.26\*\*\* 31.90\*\*\* 13.69\*\*\* 51.19\*\*\* W × D 0.41 3.35 2.32 6.36\* 6.15\* 11.70\*\* 0.63 8.10\*\* 4.39\* 4693 m W 0.40 2.61 0.07 0.09 0.04 0.26 0.00 0.19 0.14 D 1.96 0.12 0.67 7.80\*\* 0.81 20.22\*\*\* 21.83\*\*\* 3.36 29.26\*\*\* W × D 0.27 3.89 0.33 3.84 0.34 0.12 2.66 0.88 1.47 \*, \*\*, and \*\*\* indicate *P* \< 0.05, *P* \< 0.01, and *P* \< 0.001, respectively, while no asterisk indicates not significant. ###### Single factor linear regressions between soil properties (soil organic C, SOC; total N, TN; dissolved organic C, DOC; dissolved organic N, DON; nitrate N, NO~3~ ^−^-N; ammonium N, NH~4~ ^+^-N; the ratio of NH~4~ ^+^-N to NO~3~ ^−^-N, NH~4~ ^+^-N/NO~3~ ^−^-N; soil inorganic N, N~min~) and soil temperature (*T* ~*s*~), soil microbial biomass C (MBC), and N (MBN) showing regression parameters (slope, constant, *R* ^2^, and *P*). MBC and MBN data were obtained from Fu et al. \[[@B20]\]. ---------------------------------------------------------------------------------------------------------------------------------- Independent\ Regression\ SOC TN DOC DON NO~3~ ^−^-N NH~4~ ^+^-N NH~4~ ^+^-N/NO~3~ ^−^-N N~min~ variable parameters -------------- ------------- --------- --------- --------- -------- ------------- ------------- ------------------------- -------- *T* ~*s*~ Slope −5.29 −0.32 −7.90 −0.53 −0.03 −1.71 −0.43 −1.74 Constant 100.76 6.81 195.37 13.28 5.73 31.38 7.66 37.11 *R* ^2^ 0.63 0.64 0.38 0.21 0.001 0.31 0.41 0.21 *P* \<0.001 \<0.001 \<0.01 0.057 0.93 \<0.05 \<0.01 0.056 MBC Slope 0.05 0.003 0.10 0.01 0.01 0.03 0.004 0.03 Constant 6.46 1.24 46.29 2.89 3.34 −3.20 0.04 0.14 *R* ^2^ 0.76 0.66 0.70 0.51 0.13 0.92 0.47 0.82 *P* \<0.001 \<0.001 \<0.001 \<0.001 0.139 \<0.001 \<0.01 \<0.001 MBN Slope 0.28 0.01 0.68 0.05 0.06 0.17 0.02 0.23 Constant 10.56 1.58 43.22 2.93 1.60 −3.28 0.41 −1.69 *R* ^2^ 0.43 0.30 0.68 0.42 0.39 0.79 0.25 0.88 *P* \<0.01 \<0.05 \<0.001 \<0.01 \<0.01 \<0.001 \<0.05 \<0.001 ---------------------------------------------------------------------------------------------------------------------------------- ###### Multiple stepwise regression analyses between soil properties and environmental variables (soil temperature, *T* ~*s*~; soil moisture, SM) and soil microbial biomass (microbial biomass C, MBC; microbial biomass N, MBN) in an alpine meadow on the Tibetan Plateau. MBC and MBN data were obtained from Fu et al. \[[@B20]\]. Soil properties Factors Coefficients *R* ^2^ *P* ------------------------- ---------- -------------- --------- --------- SOC Constant 49.31   0.003 MBC 0.04 0.76 \<0.001 *T* ~*s*~ −2.69 0.10 0.006 TN Constant 4.28   0.001 MBC 0.002 0.66 0.004 *T* ~*s*~ −0.19 0.14 0.006 DOC Constant 46.29   \<0.001 MBC 0.10 0.70 \<0.001 DON Constant 2.89   0.005 MBC 0.01 0.51 0.001 NH~4~ ^+^-N Constant −4.57   \<0.001 MBC 0.02 0.92 \<0.001 SM 22.39 0.05 \<0.001 NO~3~ ^−^-N Constant 1.60   0.22 MBN 0.06 0.39 0.005 N~min~ Constant −2.24   0.085 MBN 0.15 0.88 \<0.001 MBC 0.01 0.05 0.005 NH~4~ ^+^-N/NO~3~ ^−^-N Constant 0.95   0.036 MBC 0.004 0.47 0.002 [^1]: Academic Editor: Felipe Bastida

The authors confirm that all data underlying the findings are fully available without restriction. All relevant data are within the paper. Introduction {#s1} ============ Mechanical ventilation (MV) has been used in critical care patients for decades. In spite of its life-saving potential, it has several shortcomings. A number of experimental studies have shown that mechanical ventilation may result in the appearance of inflammatory mediators in the lung [@pone.0114247-Uhlig1] and subsequently in oedema. [@pone.0114247-Dreyfuss1] Ventilator-Induced Lung Injury (VILI) causes macro and microscopic unspecific changes[@pone.0114247-Katzenstein1] similar to those found in patients with Acute Respiratory Distress Syndrome (ARDS). As it happens with ARDS, VILI is basically the result of important changes in the permeability of the alveolar-capillary membrane. [@pone.0114247-Dreyfuss2] The potential of mechanical ventilation for triggering or worsening pulmonary damages has been shown in animal models where the application of non-physiological ventilatory parameters (mostly very high tidal volumes) aggravated the condition of animals with a previously injured lung [@pone.0114247-Corbridge1], and even caused an injury in those without a previous pulmonary pathology. [@pone.0114247-Dreyfuss1] The use of low tidal volumes has proved to be a better approach in ARDS patients, survival being improved in strategies based on its usage. [@pone.0114247-Amato1]--[@pone.0114247-Villar1] Interestingly, recent experimental and clinical work has demonstrated that MV with low tidal volume can induce similar pulmonary changes to those noticed for VILI [@pone.0114247-Cobelens1]--[@pone.0114247-Wolthuis1] and that its appearance may be related to MV exposure time. [@pone.0114247-Hegeman1] Aquaporins are a family of small transmembrane proteins that help water to move fast, selectively and bi-directionally through lipid bi-layers. [@pone.0114247-Kozono1], [@pone.0114247-Ma1] 13 different types have been identified in mammals, [@pone.0114247-Verkman1] from which the lung is known to express four: AQP-1, in the pulmonary capillary endothelium (especially alveolar), and the visceral pleura; AQP-3, in the tracheal epithelium; AQP-4, in the tracheal and bronchial epithelium; and AQP-5, on type I pneumocyte cells of the alveoli, on the membrane adjoining to the alveolar lumen. [@pone.0114247-King1] Their role in the development and resolution of pulmonary oedema gives rise to controversy, although it does seem to play a part in VILI. [@pone.0114247-Hales1] This research aimed to verify if MV with low or moderately high tidal volumes (10 ml/Kg) sustained over time results in lung injury, subsequently altering pulmonary water content and microvascular permeability, as observed in VILI, and to objectivize what happens with AQP 1 and 5 expression, both types mainly involved in the formation of lung oedema, under the same ventilation conditions. Material and Methods {#s2} ==================== 1. Ethics statement {#s2a} ------------------- The project was carried out after approval from the Ethics Committee for Animal Experimentation and Wellbeing of the Research Foundation of Valencia\'s Hospital Clínico Universitario. 2. Animal model and monitoring {#s2b} ------------------------------ A total of 30 rats were anaesthetised by intraperitoneal injection of ketamine 80 mg/kg and xylazine 5 mg/kg. 5 rats (group C or controls) were sacrificed by intravenous injection of 100 mg/Kg thiopental. The rest of the animals (n = 25) were performed a surgical tracheostomy, using a teflon cannula (Surflo, 16G). Rats were randomly allocated into two groups. 12 rats were ventilated for 2 hours (group 2H) with a Harvard Rodent Ventilator, model 683 (Harvard Apparatus) with a tidal volume of 10 ml/kg and a respiratory rate of 90 breaths/minute. 13 rats were ventilated with exactly the same parameters for 4 hours (group 4H). The cervical vascular bundle was dissected, and the right internal jugular vein and the right carotid artery were catheterized to continuously monitor heart rate (HR) and mean arterial pressure (MAP). Peak inspiratory pressure and respiratory system compliance were continuously recorded. Anaesthesia was maintained by continuous intravenous infusion of ketamine and cisatracurium using dosis of 100 mcg/Kg/min and 2--3 mcg/Kg/min, respectively (20 ml of ketamine 5%, 10 ml of cisatracurium 0,2% and 20 ml of saline solution 0,9% at a rate of approximately 0,1 ml/h in the internal jugular vein). Anesthesia was supplemented, in cases in which it was necessary, by administration of an intravenous bolus of 0.1 ml of the mixture. Gasometric samples were taken in all animals in groups 2H and 4H at the beginning of MV and 30 minutes before the end of MV. Rats were sacrificed by intravenous injection of sodium thiopental. The left lung was used for the determination of lung water content. The right lung of 6 rats from groups 2H and 4H was used for determining AQP 1 and AQP 5 expression. Lungs were either frozen in liquid nitrogen or paraffin-embedded for immunohistochemical sectioning and marking. 2 rats in group C and 4 rats from groups 2H and 4H were used to establish pulmonary macrovascular permeability. 3. Measuring lung oedema {#s2c} ------------------------ Lungs were dried with filter paper and placed on a Petri dish with known weight to obtain lung wet weight (LWW). They were then placed in a drying chamber at 80°C for 96 hours and their dry weight (LDW) was determined. Two indicators of the amount of oedema were obtained: Lung WW/DW ratio and the proportion of pulmonary water, expressed in percentages (%~water~). The latter parameter was estimated using this formula: % ~water~  =  (LWW - LDW)/LWW \* 100 4. Measuring microvascular permeability {#s2d} --------------------------------------- Microvascular permeability was quantified using Evans Blue Dye. 0.5 ml Evans Blue was injected intravenously (30 mg/Kg) 30 minutes before sacrificing the animal. Rats were sacrificed by exsanguination from the carotid artery, but saline was simultaneously infused via the jugular vein in the same amount as that of the blood extracted. After death, the right lung was separated and immersed in formamide (5 ml) and homogenised for 2 min. The resulting suspension was incubated at 37°C/18 h and then centrifuged at 5000xg/30 minutes, and the supernatant was measured. Concentration of Evans Blue in the supernatant was spectrophotometrically determined. 5. Study of aquaporin expression {#s2e} -------------------------------- ### 5.1 Western blot {#s2e1} Proteins were extracted from previously frozen lungs. A Compartmental Protein Extraction Kit (Chemicon International, Temecula CA) was used. 200--400 mg tissue was homogenised in cold buffer C (1 ml/g tissue) and Ultra Turrax (KA, Staufen, Germany). Two protein fractions were obtained for each sample: cytoplasm and membrane, and they were quantified. Proteins in each fraction (100 mg) were separately run on a Tris-HCl/SDS gel, 8% acrylamide, and they were transferred onto a nitrocellulose membrane (Hybond-ECL, Amersham). After washing the membrane with distilled water, it was blocked with a PBS/Tween solution, 0.2%, with 5% skimmed milk. It was then incubated with the primary antibody during 2 hours at room temperature. The antibodies used were Anti-Rat AQP1 (Alpha Diagnostic, San Antonio, TX) and Anti-Rat AQP5 (Alpha Diagnostic, San Antonio, TX), both with a 2 mg/ml concentration. After several washes with PBS/Tween 0.2%, it was incubated with the secondary antibody Anti-Rabbit IgG (DAKO, Glostrup, Denmark) in 1∶2000 dilution. β-actin expression was detected as an internal control, and relative protein content was analysed using the enhanced chemoluminiscence method. ### 5.2 Real-time polymerase chain reaction with reverse transcriptase {#s2e2} Lungs were cut with a microtome, three sections being obtained for each sample for total RNA extraction with TRIZOL (Reagent InvitrogenTM Life Technologies) as in the phenol extraction method described by Chomczynsky. [@pone.0114247-Chomczynski1] Microsections were added 1 ml TRIZOL and homogenized (Polytron PT 1200, Kinematica AG) and centrifuged at 10.000 rpm/10 min at 4°C. The supernatant was removed and RNA was precipitated by adding 0.1 volumes of sodium acetate 3 M, 2.5 volumes cold ethanol and 0.5 µl glycogen (20 mg/ml). RNA was centrifuged again, air-dried and resuspended in 20 µl Tris/EDTA buffer. RNA was reversely transcribed to cDNA with Superscript II (Invitrogen), by incubation with reverse transcriptase at 50°C for 30 min, followed by amplification with custom primers (Invitrogen), summarised in [Table 1](#pone-0114247-t001){ref-type="table"}. 35 amplification cycles were completed, with denaturalization at 95°C (30 sec), hybridization (30 sec) (temperatures on [Table 1](#pone-0114247-t001){ref-type="table"}) and extension at 72°C (1 min). Following amplification, RT-PCR products were separated in agarose gels at 1% and bands were viewed by ethidium bromide staining, and quantified by band density scanning using Scion Image (Beta 4.02, Scion Corporation). Results were expressed in relation to the level of β-actin mRNA in the same RNA samples. 10.1371/journal.pone.0114247.t001 ###### RT-PCR primer sequences and temperature conditions. {#pone-0114247-t001-1} Gene Primer (5′-3′) Primer, counterclockwise (5′--3) T (°C) --------- ---------------------------- ---------------------------------- -------- AQP1 TCTGGAGGCTGTGGTGGCT AAGTGAGTTCTCGAGCAGGGA 60 AQP5 TGGGTCTTCTGGGTAGGGCCTATTGT GCCGGCTTTGGCACTTGAGATACT 50 β-Actin ATCATGTTTGAGACCTTCAACA CATCTCTTGCTCGAAGTCCA 56 ### 5.3 Immunohistochemical study {#s2e3} Sections (4 µm thickness) from the paraffin-embedded lungs were obtained with the microtome. After the sections were dewaxed and hydrated, autoclave pretreatment (10 min, 121°C) for AQP1 and AQP5 antigen retrieval was performed and the sections were incubated in 1% H2O2 for 30 min at room temperature to block endogenous peroxidase activity. After being washed in PBS, the sections were then preincubated with goat serum albumin for 30 min at 37°C, and subsequently incubated with the primary antibodies against AQP1 (1∶500) and AQP5 (1∶300) for 18 h at 4°C. Then, the sections were washed with PBS and stained with Biotin-labelled goat anti-rabbit IgG for 30 min at 37°C. Intervening washes in PBS again were followed by incubation with Horseradish enzyme labelled streptavidin working solution for 30 min at 37°C. The sections were washed in PBS before application of diaminobenzidine (DAB), then were mounted under coverslip and analyzed under light microscope. 6. Statistical analysis {#s2f} ----------------------- Results were expressed as mean ± standard deviation (SD). To compare results between groups, the non-parametric Kruskall-Wallis and Mann-Whitney tests were used. For the analysis of data within each individual group, the Wilconxon test was applied. Regression analyses for the amount of aquaporins and mRNA and determination coefficients (R^2^) were performed. Values of p\<0.05 were assumed to be statistically significant in all cases. Results {#s3} ======= 1. Pulmonary oedema {#s3a} ------------------- No significant differences were found between the three groups for the wet weight-dry weight ratio (group C: 4.72±0.04 vs. group 2H: 4.90±0.33 vs. group 4H: 5.23±0.79) or the percentage of pulmonary water (group C: 78.82±0.16 vs. group 2H: 79.52±1.31 vs. group 4H: 80.55±2.50), though an increasing trend was noticed for both parameters ([Fig. 1](#pone-0114247-g001){ref-type="fig"}). {#pone-0114247-g001} 2. Ventilatory mechanics and hemodynamic parameters {#s3b} --------------------------------------------------- Peak inspiratory pressure (PIP) progressively rose in both groups (group 2H and group 4H), higher values being found 90 minutes after the start of the experiment in group 2H and at minute 60 in group 4H ([Fig. 2](#pone-0114247-g002){ref-type="fig"}). No differences were found in the cut-off points of the two groups. {#pone-0114247-g002} Pulmonary compliance was reduced gradually in both groups, with significance in respect of the initial value as from minute 30 for group 2H and minute 60 in group 4H ([Fig. 3A](#pone-0114247-g003){ref-type="fig"}). No differences were found between the groups in none of the cut-off points. This decrease in compliance correlated with the peak pressure rise, with an R^2^ value of 0.98, p\<0.01 ([Fig. 3B](#pone-0114247-g003){ref-type="fig"}). {#pone-0114247-g003} Rats in both groups were hemodynamically stable. No differences were found in mean arterial pressure (group 2H: 97.95 mmHg ±27.75 vs. group 4H: 104.53 mmHg ±27.54), and the same applies to average heart rate values (group 2H: 341.29 bpm ±66.73 vs. group 4H: 306.52 bpm ±78.11). Mean arterial pressure (MAP) dropped in group 2H progressively compared to the baseline as from minute 60, but this also happened in group 4H as from minute 45. Heart rate also decreased after 120 minutes ([Fig. 4](#pone-0114247-g004){ref-type="fig"}). {#pone-0114247-g004} 3. Gasometric parameters {#s3c} ------------------------ Gasometric results for groups 2H and 4H are summarised in [Table 2](#pone-0114247-t002){ref-type="table"}. No differences were found for pH, pCO~2~, ABE and lactate values within the study groups, and differences between the two groups were not found either. But a tendency towards mixed acidosis in relation to duration of MV was observed. Oxygenation presented a tendency to pO~2~ and pO~2~/FiO~2~ ratio reduction two hours after MV in group 2H, which was slightly more marked in group 4H after 4 hours. 10.1371/journal.pone.0114247.t002 ###### Results of arterial blood gas tests in Group 2H (ventilated for 2 hours, 10 ml/Kg) and Group 4H (ventilated for 4 hours, 10 ml/Kg). {#pone-0114247-t002-2} Group 2H (2 hours) Group 4H (4 hours) -------------- -------------------- --------------------------------------------- -------------- --------------- **pH** 7,29±0,05 7,21±0,07 7,30±0,05 7,23±0,07 **pCO2** 47,96±3,56 56,87±18,04 47,33±8,51 52,84±7,73 **ABE** −3,96±2,02 −6,70±1,61[\*](#nt102){ref-type="table-fn"} −3,70±2,72 −6,26±4,20 **Lac** 2,08±0,62 1,99±0,66 1,79±0,74 1,91±1,24 **pO2** 93,30±17,72 91,56±16,56 97,28±15,76 83,03±24,27 **pO2/FiO2** 444,38±84,62 436,86±79,17 462,78±74,61 395,22±115,30 Values expressed as mean ± standard deviation. \* p \<0.05 in relation to baseline. 4. Microvascular permeability {#s3d} ----------------------------- A significant increase in microvascular permeability was not found. Evans Blue absorbance on lung tissue was as follows: 16.08 ng/mg ±2.45 in group C; 25.96 ng/mg ±9.90 in group 2H and 20.39 ng/mg ±2.20 in group 4H. 5. Expression of aquaporins 1 and 5 {#s3e} ----------------------------------- AQP 1 steady state levels measured by Western blot in membrane and cytoplasm did not show statistically significant differences in relation to duration of MV ([Figs. 5](#pone-0114247-g005){ref-type="fig"} and [6](#pone-0114247-g006){ref-type="fig"}). {#pone-0114247-g005} {#pone-0114247-g006} AQP-5 steady state levels in cytoplasm and membranes was significantly greater in both groups (2H and 4H) vs. the controls. Besides, AQP-5 expression on cytoplasm and membranes was greater in group 4H than in group 2H (p = 0.027 and p = 0.039, respectively) ([Figs. 5](#pone-0114247-g005){ref-type="fig"} and [7](#pone-0114247-g007){ref-type="fig"}). To better characterize these differences in the expression of AQP 5, a regression analysis of both variables was conducted, which provided a coefficient of determination R^2^ = 0.80 (p = 0.008) and R^2^ = 0.90 (p = 0.001) ([Fig. 8](#pone-0114247-g008){ref-type="fig"}). {#pone-0114247-g007} {#pone-0114247-g008} RT-PCR results show a significant increase in the amount of mRNA for AQP-1 in groups 2H and 4H compared to group C. For AQP 5, an increase was found in the amount of mRNA in group 4H compared to group C ([Fig. 9](#pone-0114247-g009){ref-type="fig"}). The regression analysis of the amount of mRNA showed very significant determination coefficients for alveolar AQP 5 and AQP 1 in relation to duration of MV (R^2^ AQP-5 = 0.71 (p\<0.001) and R^2^ AQP-1 = 0.69 (p\<0.001). {#pone-0114247-g009} Immunohistochemical lung preparations of AQP-5 show the membranes of type 1 pneumocytes delimiting the alveolar network. The intensity of the staining increases with duration of MV ([Fig. 10](#pone-0114247-g010){ref-type="fig"}). AQP 1 samples show the network of the alveolar capillaries, as AQP-1 is found in endothelial cells and red blood cells but not in the pneumocytes covering the alveolus. In this case, image analysis does not show an increase in dye intensity in relation to MV time ([Fig. 11](#pone-0114247-g011){ref-type="fig"}). {#pone-0114247-g010} {#pone-0114247-g011} Discussion {#s4} ========== MV with high tidal volumes or without positive end-expiratory pressure (PEEP) may lead to the appearance of inflammatory mediators in the lung via mechanotransduction. [@pone.0114247-Uhlig1], [@pone.0114247-Tremblay1], [@pone.0114247-Bueno1] MV with tidal volumes over 12 ml/kg is associated with a bad prognosis, while \"lung-protective ventilation\" with low tidal volumes (under 10 ml/kg) and PEEP optimization reduces ventilation-induced injuries. [@pone.0114247-Noauthors1] However, ventilation with low tidal volumes may result in the appearance of an inflammatory response pattern in the lung. In rats ventilated with low pressures (12 cmH~2~O) and 4 cmH~2~O PEEP during 4 hours, increased activity of myeloperoxidase and macrophage inflammatory protein-2 and interleukin-6 has been reported. [@pone.0114247-Cobelens1] Similarly, mild pro-inflammatory changes have been found in tracheal aspirates and blood of children with healthy lungs ventilated for 2 hours for heart surgery. [@pone.0114247-Plotz1] Likewise, 6 ml/Kg tidal volume MV in Wistar rats has been reported to induce a proinflammatory and profibrogenic response in the lung. [@pone.0114247-Caruso1] And in mechanically ventilated mice at 7.5 ml/Kg for 5 hours, rising levels of IL-6, TNF-α and lung water were found. [@pone.0114247-Wolthuis1] Conversely, some trials have demonstrated MV with tidal volumes of 10 ml/kg during 6 hours not to cause an increase in cytokine expression. [@pone.0114247-Altemeier1] Similarly, no increases in cytokines or mediators were found in previously healthy patients undergoing MV after one hour of exposure, even with tidal volumes of 15 ml/Kg. [@pone.0114247-Wrigge1] Based on our results, we fail to confirm that MV with scarcely injurious parameters during 4 hours causes acute lung damage or changes in pulmonary permeability with an increase in lung water. A mild trend was observed in our experiments, however. Although levels were not measured for any inflammatory mediator, based on previous studies, inflammatory mediators are likely to be increased by MV even with low tidal volumes. General anaesthesia and the supine position are known to cause pulmonary atelectasis with predominance in the dependent region. [@pone.0114247-Hedenstierna1] These give the lung a heterogeneous appearance in which atelectasis areas co-exist with aired and even overdistended areas and their corresponding transition zones. [@pone.0114247-Dreyfuss2] Besides, the use of low tidal volumes without PEEP can result in the appearance and maintenance of atelectasis in patients under general anaesthesia and muscular relaxation. [@pone.0114247-Wolthuis1] Regular recruitment with frequent deep insufflations during low tidal volume MV has been reported to improve oxygenation without signs of lung injury in mice mechanically ventilated for hours. [@pone.0114247-Allen1] Therefore, this would explain the worsening observed in the oxygenation of our animals as exposure to MV with low tidal volumes and without PEEP increased. Our animals may have developed atelectasis in dependent areas, causing different degrees of atelectrauma that would explain the degradation in oxygenation seen in the experiment (*see* [Table 2](#pone-0114247-t002){ref-type="table"}), although it was not statistically significant, as well as the fall in pulmonary compliance and the rise in peak inspiratory pressure (*see* [Figs. 2](#pone-0114247-g002){ref-type="fig"} and [3A](#pone-0114247-g003){ref-type="fig"}). The latter parameters correlated linearly with a coefficient of determination R^2^ of 0.98 in group 4H (animals ventilated for 4 hours) ([Fig. 3B](#pone-0114247-g003){ref-type="fig"}). Although acid-base parameters are well-known reliable indicators of animal wellbeing, they have only been partially evaluated in murine mechanical ventilation models. [@pone.0114247-Cobelens1], [@pone.0114247-Belperio1], [@pone.0114247-Altemeier1], [@pone.0114247-Tremblay1] In our model, ventilated animals in groups 2H and 4H showed a trend towards mixed acidosis but without significance. As a result of the formation of atelectasis and higher pressures, ventilation might have been less effective, which could account for the rise in pCO~2~ levels, although not significantly. The metabolic component of acidosis can have several causes. Metabolic acidosis in mice can be induced by saline administration, [@pone.0114247-Zuurbier1], [@pone.0114247-Wolthuis1] which in our animals was used for maintenance at very low levels. Yet, metabolic acidosis caused by some hemodynamic failure cannot be totally excluded. This is possibly related to poor compensation for losses rather than potential deficits in venous return and secondary low cardiac output, as a result of the thoracic pressure reversal observed in positive pressure MV. [@pone.0114247-Schwarte1] Similarly, heart rate data showed a significant reduction over time from minute 120 of the experiment, although always within the physiological range of rats. [@pone.0114247-Papadimitriou1] To date, no research group has explored the expression of AQP 1 and AQP 5 jointly in mechanically ventilated rats with low tidal volumes. However, a study conducted in rats ventilated with very high tidal volumes (40 ml/Kg during 4 hours) showed a reduction in AQP 1 expression. The researchers suggested an inflammatory mediator secondary mechanism, as they managed to mitigate the reduction in AQP 1 by administering a cyclooxygenase 2 inhibitor. [@pone.0114247-Jin1] In another trial, based on an animal model of ARDS induced by smoke inhalation, animals were subjected to MV, finding an increase in mRNA for AQP 1. [@pone.0114247-Schmalstieg1] The shortcoming in these studies is the absence of a control group with non-ventilated animals to study AQP expression. Our experiments show an increase in AQP 5 expression that became more significant as MV exposure was prolonged and, similarly, an increase in mRNA of AQP 5 which was also greater in group 4H compared to non-ventilated animals. These increases were not accompanied by significant variations in lung water content or microvascular permeability. Although an increase in AQP 1 in the lungs was not found, after 4 h of MV, our study did find a significant increase in mRNA. This is likely to be the previous step to AQP 1 synthesis. It is possibly due to the fact that a longer MV period is needed for an increase in AQP 1 on cytoplasm and membranes to be seen with the Western blot. Different factors have been shown to modulate the amount of AQPs, reducing them under a number of pathological conditions and leading to an increase in pulmonary water and oedema. [@pone.0114247-Wang1]--[@pone.0114247-Jiao1] In addition, the fact that lung injury parameters improve when their expression is induced, assigns AQPs a protective mechanism in the occurrence and development of pulmonary oedema. [@pone.0114247-Cao1]--[@pone.0114247-Dong1] Although some studies with genetically modified animals debate the possible relevant role of AQPs in the reabsorption of alveolar water, they prove the importance of these channels in pulmonary permeability. [@pone.0114247-Bai1], [@pone.0114247-Ma1], [@pone.0114247-Song1] Significant changes in lung water or microvascular permeability were not objectivized in our animals. The explanation for this could be that AQPs, according to some authors, only work in situations of stress. [@pone.0114247-Hales1] Further studies are needed to determine the true role of AQP1 and AQP 5 in MV. To conclude, in our model prolonged MV and tidal volumes of 10 ml/Kg for 4 hours did not result in increased pulmonary water or changes in microvascular permeability, these being mechanisms involved in ventilation-induced lung injury. Our study found an increase in the protein expression of AQP 5 and its mRNA, correlated with exposure time and mechanical ventilation. Likewise, the amount of mRNA for AQP 1 also increased in correlation with MV time. Apparently, AQP 5 and AQP 1 can have a protective effect against MV-induced pulmonary oedema, but more studies are needed to clarify whether these proteins really play a relevant role in mechanically ventilated lungs under different conditions. The experiment was carried out in the facilities of the Medical School (University of Valencia). [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: GF JGDLA. Performed the experiments: GF MM EP. Analyzed the data: GF JGDLA BS JC JDA FJB. Contributed reagents/materials/analysis tools: JC BS. Wrote the paper: GF JGDLA BS EP.

Introduction {#s1} ============ Usher syndrome (USH) is an autosomal recessive disorder characterized by hearing loss (HL), retinitis pigmentosa (RP) and vestibular dysfunction. Three clinical subtypes can be distinguished. USH type 1 (USH1) is the most severe among them because of profound HL, absent vestibular responses, and prepubertal onset RP. USH type 2 (USH2) is characterized by congenital moderate to severe HL, with a high-frequency sloping configuration. The vestibular function is normal and onset of RP is in the first or second decade. The onset of the visual symptoms such as night blindness in USH usually occurs several years later than in USH1. USH type 3 (USH3) is characterized by variable onset of progressive HL, variable onset of RP, and variable impairment of vestibular function (normal to absent) [@pone.0090688-Kimberling1], [@pone.0090688-Yan1]. To date, nine genetic loci for USH1(*USH1B-H*, *J*, and *K*) have been mapped to chromosomes 11q13.5, 11p15.1, 10q22.1, 21q21, 10q21-q22, 17q24-q25, 15q22-q23 (*USH1H* and *J*), and 10p11.21--q21.1 [@pone.0090688-Yan1], [@pone.0090688-Jaworek1], [@pone.0090688-Riazuddin1]. Six of the corresponding genes have been identified: the actin-based motor protein myosin VIIa (*MYO7A*, USH1B) [@pone.0090688-Weil1]; two cadherin-related proteins, cadherin 23 (*CDH23*, USH1D) [@pone.0090688-Bork1] and protocadherin 15 (*PCDH15*, USH1F) [@pone.0090688-Ahmed1]; and two scaffold proteins, harmonin (*USH1C*) [@pone.0090688-Verpy1] and sans (*USH1G*) [@pone.0090688-Mustapha1]; the Ca^2+^- and integrin-binding protein (*CIB2*, USH1J) [@pone.0090688-Riazuddin1]. In Caucasian USH1 patients, previous studies showed that mutations in *MYO7A*, *USH1C*, *CDH23*, *PCDH15*, and *USH1G*, were found in 39--55%, 7--14%, 7--35%, 7--11%, and 0--7%, respectively (the frequency of *CIB2* is still unknown) [@pone.0090688-Ouyang1], [@pone.0090688-Bonnet1], [@pone.0090688-LeQuesneStabej1]. In Japanese, Nakanishi et al. showed that *MYO7A* and *CDH23* mutations are present in USH1 patients [@pone.0090688-Nakanishi1], however, the frequency is not yet known. In addition, mutations in three corresponding genes (usherin *USH2A* [@pone.0090688-Eudy1], G protein-coupled receptor 98; *GPR98* [@pone.0090688-Weston1], and deafness, autosomal recessive 31; *DFNB31* [@pone.0090688-Aller1]) have been reported so far in USH2, and USH3 is caused by mutations in the clarin 1 (*CLRN1*) [@pone.0090688-Joensuu1] gene. Comprehensive molecular diagnosis of USH has been hampered both by genetic heterogeneity and the large number of exons for most of the USH genes. The six USH1 genes collectively contain 180 coding exons [@pone.0090688-Riazuddin1], [@pone.0090688-Mustapha1], [@pone.0090688-Ouyang1] the three USH2 genes comprise 175 coding exons [@pone.0090688-Weston1], [@pone.0090688-Aller1], [@pone.0090688-Nakanishi2], and the USH3 gene has five coding exons [@pone.0090688-Joensuu1]. In addition some of these genes are alternatively spliced ([@pone.0090688-Riazuddin1], [@pone.0090688-Ahmed1], [@pone.0090688-Verpy1], [@pone.0090688-Aller1], [@pone.0090688-Joensuu1] and NCBI database: <http://www.ncbi.nlm.nih.gov/nuccore/>). Thus far, large-scale mutation screening has been performed using direct sequence analysis, but that is both time-consuming and expensive. We thought that targeted exon sequencing of selected genes using the Massively Parallel DNA Sequencing (MPS) technology would enable us to systematically tackle previously intractable monogenic disorders and improve molecular diagnosis. Therefore, in this study, we have conducted genetic analysis using MPS-based genetic screening to find mutations in nine causative USH genes (except *CIB2*) in Japanese USH1 patients. Results {#s2} ======= Mutation analysis of the nine USH genes in 17 unrelated USH1 patients revealed 19 different probable pathogenic variants, of which 14 were novel ([Table 1](#pone-0090688-t001){ref-type="table"}). 10.1371/journal.pone.0090688.t001 ###### Possible pathogenic variants found in this study. {#pone-0090688-t001-1} Gene Mutation type Nucleotide change Amino acid change exon/intron number Domain control (in 384 alleles) SIFT Score PolyPhen Score Reference ---------- --------------- ------------------- ------------------- -------------------- -------------- -------------------------- ------------ ---------------- -------------------------------- *MYO7A* Frameshift c.1623dup p.Lys542GlnfsX5 Exon 14 \- N/A \- \- Le Quesne Stabej et al. (2012) c.4482_4483insTG p.Trp1495CysfsX55 Exon 34 \- N/A \- \- This study c.6205_6206delAT p.Ile2069ProfsX6 Exon 45 \- N/A \- \- This study Nonsense c.1477C\>T p.Gln493X Exon 13 \- N/A \- \- This study c.1708C\>T p.Arg570X Exon 15 \- N/A \- \- This study c.2115C\>A p.Cys705X Exon 18 \- N/A \- \- This study c.6321G\>A p.Trp2107X Exon 46 \- N/A \- \- This study Missense c.2074G\>A p.Val692Met Exon 17 Motor domain 0 0.09 0.982 This study c.2311G\>T p.Ala771Ser Exon 20 IQ 2 0.0026 0.01 0.825 Nakanishi et al. (2010) c.6028G\>A p.Asp2010Asn Exon 44 FERM 2 0 0 0.925 Jacobson et al. (2009) *CDH23* Frameshift c.3567delG p.Arg1189ArgfsX5 Exon 30 \- N/A \- \- This study c.5780_5781delCT p.Ser1927Cysfs16 Exon 44 \- N/A \- \- This study Splicing c.5821-2A\>G ? Intron 44 \- N/A \- \- This study Nonsense c.6319C\>T p.Arg2107X Exon 48 \- N/A \- \- Nakanishi et al. (2010) *PCDH15* Splicing c.158-1G\>A ? Intron 3 \- N/A \- \- This study Nonsense c.1006C\>T p.Arg336X Exon 10 \- N/A \- \- This study c.2971C\>T p.Arg991X Exon 22 \- N/A \- \- Roux et al. (2006) c.3337G\>T p.Glu1113X Exon 25 \- N/A \- \- This study Missense c.3724G\>A p.Val1242Met Exon 28 Cadherin 11 0 0 1 This study Computer analysis to predict the effect of missense variants on MYO7A protein function was performed with sorting intolerant from tolerant (SIFT; <http://sift.jcvi.org/>), and polymorphism phenotyping (PolyPhen2; <http://genetics.bwh.harvard.edu/pph2/>). N/A: not applicable. All mutations were detected in only one patient each and sixteen of the 17 patients (94.1%) carried at least one mutation, while one patient had no mutations. Thirteen of the 16 mutation carriers each had two pathogenic mutations ([Table 2](#pone-0090688-t002){ref-type="table"}). 10.1371/journal.pone.0090688.t002 ###### Details of phenotype and genotype of 17 USH1 patients. {#pone-0090688-t002-2} Sample No. Age Sex Allele1 Allele2 Hereditary form Onset of night blindness Cataract Hearing Aid Cochlear Implant --------------------------------------------------------------------------------------------------------------- ----- ----- -------------------------------------------------- -------------------------------------------------- ----------------- -------------------------- ----------- ------------- ------------------ ***MYO7A*** 1 37 M p.Gln493X p.Trp1495CysfsX55 sporadic 13 no unilateral unilateral 2 41 W p.l2069fsX6 p.l2069fsX6 AR unknown both eyes bilateral no 5 54 M p.Val692Met p.Val692Met AR 5 both eyes no no 6 54 W p.Arg570X p.Arg570X sporadic 6 no no no 8 14 M p.Lys542GlnfsX5 p.Lys542GlnfsX5 sporadic 6 no unilateral unilateral 11 54 M p.Asp2010Asn p.Trp2107X sporadic 13 no no no 17 56 W p.Cys705X p.Cys705X sporadic unknown no no no ***CDH23*** 7 12 W p.Arg1189ArglfsX5 p.Arg1189ArglfsX5 sporadic 12 both eyes no bilateral 9 9 M p.Ser1927Cysfs16 c.5821-2A\>G sporadic 8 no unilateral unilateral 15 16 W p.Arg2107X p.Arg2107X sporadic unknown no no no ***PCDH15*** 3 47 W p.Glu1113X p.Glu1113X sporadic 5 both eyes no no 16 28 W p.Arg991X p.Arg991X AR 10 no no no 10 62 M p.Arg962Cys unknown sporadic 9 both eyes no no 12 52 M p.Arg336X unknown sporadic 3 no no no 13 51 M p.Val1242Met unknown sporadic 10 no no no ***MYO7A*** [\*](#nt103){ref-type="table-fn"} **^1^** ***/PCDH15*** [\*](#nt103){ref-type="table-fn"} **^2^** 4 21 M p.Ala771Ser[\*](#nt103){ref-type="table-fn"} ^1^ c.158-1G\>A[\*](#nt103){ref-type="table-fn"} ^2^ sporadic 10 no unilateral unilateral **unknown** 14 64 W unknown unknown sporadic 15 both eyes unilateral no \*All subjects have congenital deafness and RP. Nonsense, frame shift, and splice site mutations are all classified as pathogenic, whereas missense mutations are presumed to be probable pathogenic variants based on results of prediction software for evaluation of the pathogenicity of missense variants ([Table 1](#pone-0090688-t001){ref-type="table"}). Of the 19 probable pathogenic mutations that we found, 17 were detected by MPS. The remaining two (p.Lys542GlnfsX5 in *MYO7A* and c.5821-2A\>G in *CDH23*) were sequenced by direct sequence analysis. Of our 17 USH patients, seven had *MYO7A* mutations (41.2%), three had *CDH23* mutations (17.6%), and two had *PCDH15* mutations (11.8%). We did not find any probable pathogenic mutations in *USH1C*, *USH1G*, and USH2/3 genes. Four USH1 patients (Cases \#3, 5, 8, 15) had probable pathogenic mutations in two different USH genes, with one being a biallelic mutation ([Table 3](#pone-0090688-t003){ref-type="table"}). The other heterozygous/homozygous mutations were missense variants. Three of these patients (Cases \#3, 5, 8) presented with earlier RP onset (night blindness) than in the other patients with two pathogenic mutations (Cases \#1, 6, 7, 9, 11, 16) (*p* = 0.007) ([Fig. 1](#pone-0090688-g001){ref-type="fig"}). {#pone-0090688-g001} 10.1371/journal.pone.0090688.t003 ###### The patients with mutations in two different genes. {#pone-0090688-t003-3} Sample Genes with two pathogenic mutations Gene with one heterozygous mutation Nucleotide change Amino acid change control SIFT score PolyPhen score Referense -------- ------------------------------------- ------------------------------------- ------------------- ------------------------------------------ --------- ------------ ---------------- ------------------------- 5 *MYO7A* *CDH23* c.C719T p.P240L[\*](#nt104){ref-type="table-fn"} 0.26 0.06 0.999 Wagatsuma et al. (2007) 8 *MYO7A* *CDH23* c.2568C\>G p.Ile856Met 0 0.08 1 This study 15 *CDH15* *USH1C* c.2437T\>G p.Tyr813Asp 0 0.19 0.932 This study 3 *PCDH15* *USH1G* c.28C\>T p.Arg10Trp 0 0.19 1 This study \*homozygotes. One patient (Case \#4) had heterozygote mutations in two USH1 genes (p.Ala771Ser in *MYO7A* and c.158-1G\>A in *PCDH15*). His parents and one brother were found to also be carriers for these mutations. Another brother had no variants ([Fig. 2](#pone-0090688-g002){ref-type="fig"}). {#pone-0090688-g002} Discussion {#s3} ========== For USH1, early diagnosis has many immediate and several long-term advantages for patients and their families [@pone.0090688-Kimberling1]. However, diagnosis in childhood, based on a clinical phenotype, has been difficult because patients appear to have only non-syndromic HL in childhood and RP develops in later years. Although early diagnosis is now possible through DNA testing, performing large-scale mutation screening for USH genes in all non-syndromic HL children has been both time-consuming and expensive. Therefore, the availability of MPS, which facilitates comprehensive large-scale mutation screening [@pone.0090688-Miyagawa1] is a very welcome advance. MPS technology enabled us to detect pathogenic mutations in USH1 patients efficiently, identifying one or two pathogenic/likely pathogenic mutations in 16 of 17 (94.1%) cases. This was comparable to previous direct sequence analysis results such as Bonnet et al. who detected one or two mutations in 24 out of 27 (89%) USH1 patients [@pone.0090688-Bonnet1] and Le Quesne Stabej et al. who detected one or two mutations in 41 out of 47 (87.2%) USH1 patients [@pone.0090688-LeQuesneStabej1]. In addition, MPS assists in the analysis of disease modifiers and digenic inheritance because it simultaneously investigates many causative genes for a specific disease, such as in our case, USH. Previous reports have described several USH cases with pathogenic mutations in two or three different USH genes [@pone.0090688-Bonnet1], [@pone.0090688-LeQuesneStabej1], [@pone.0090688-Ebermann1]. In our study, four patients had two pathogenic mutations in one gene and missense variants in a different gene ([Table 3](#pone-0090688-t003){ref-type="table"}). We considered the latter to possibly be a disease modifier. For example, *USH1C*:p.Tyr813Asp, which occurred in 0/384 control chromosomes and was predicted to be "probably damaging" by the Polyphen program, was found with a homozygous *CDH23* nonsense mutation (p.Arg2107X) (Case \#15). As for what the variant "modifies", we speculate that for USH1 patients with a disease modifier, RP symptoms such as night blindness show an earlier onset. However, we think that profound HL and the absence of vestibular function in USH1 patients are not affected by modifiers as they are congenital and therefore not progressive. Ebermann et al. described a USH2 patient with "digenic inheritance." a heterozygous truncating mutation in *GPR98*, and a truncating heterozygous mutation in PDZ domain-containing 7 (*PDZD7*), which is reported to be a cause of USH [@pone.0090688-Ebermann1]. Our USH1 patient (Case \#4) had segregated *MYO7A*:p.Ala771Ser and *PCDH15*:c.158-1G\>A. Molecular analyses in mouse models have shown many interactions among the USH1 proteins [@pone.0090688-Yan1]. In particular, *MYO7A* directly binds to *PCDH15* and both proteins are expressed in an overlapping pattern in hair bundles in a mouse model [@pone.0090688-Senften1]. *PCDH15*:c.158-1G\>A, predicted to alter the splice donor site of intron 3, has been classified as pathogenic. *MYO7A*:p.Ala771Ser is a non-truncating mutation, but was previously reported as disease-causing [@pone.0090688-Nakanishi1]. So, we consider the patient to be the first reported case of *MYO7A*/*PCDH15* digenic inheritance. However, we should be aware of two limitations of MPS technology. First, the target region of MPS cannot cover all coding exons of USH genes. Actually, the coverage of the target exons was 97.0% in our study. So, it is impossible to detect a mutation in a region which is not covered using this system (Case \#9: c.5821-2A\>G). Secondarily, the MPS system used in this study, is not effective for detecting homo-polymer regions, for example poly C stretch [@pone.0090688-Loman1] (Case \#8: p.Lys542GlnfsX5). In addition, concerning pathogenecity of mutations identified, functional analysis will be necessary to draw the final conclusion in the future. In UK and US Caucasian USH1 patients, USH1B (*MYO7A*) has been reported as the most common USH1 genetic subtype [@pone.0090688-Bonnet1], [@pone.0090688-LeQuesneStabej1], while USH1F (*PCDH15*) has been reported as the most common USH1 genetic subtype in North American Ashkenazi Jews [@pone.0090688-BenYosef1]. In Japanese, our study revealed that the most common type was *MYO7A* (41.7%), which was similar to the frequency in the above Caucasian patients (46.8∼55%) [@pone.0090688-Bonnet1], [@pone.0090688-LeQuesneStabej1]. However, the small number of USH1 patients in our study might have biased the frequency and further large cohort study will be needed in the future. In addition, most of our detected mutations were novel. We have previously reported genes responsible for deafness in Japanese patients and observed differences in mutation spectrum between Japanese (who are probably representative of other Asian populations) and populations with European ancestry [@pone.0090688-Usami1]. In conclusion, our study was the first report of USH mutation analysis using MPS and the frequency of USH1 genes in Japanese. Mutation screening using MPS has the potential power to quickly identify mutations of many causative genes such as USH while maintaining cost-benefit performance. In addition, the simultaneous mutation analysis of large numbers of genes was useful for detecting mutations in different genes that are possibly disease modifiers or of digenic inheritance. Materials and Methods {#s4} ===================== Subjects {#s4a} -------- We screened 17 Japanese USH1 patients (aged 9 to 64 years): three from autosomal recessive families (non-affected parents and two or more affected siblings), and 14 from sporadic families. There were 9 males and 8 females. None of the subjects had any other noteworthy symptoms. All subjects or next of kin on the behalf of the minors/children gave prior written informed consent for participation in the project, and the Ethical Committee of Shinshu University approved the study and the consent procedure. Amplicon Library Preparation {#s4b} ---------------------------- An Amplicon library of the target exons was prepared with an Ion AmpliSeq Custom Panel (Applied Biosystems, Life Technologies, Carlsbad, CA) designed with Ion AmpliSeq Designer (<https://www.ampliseq.com/browse.action>) for nine USH genes by using Ion AmpliSeq Library Kit 2.0 (Applied Biosystems, Life Technologies) and Ion Xpress Barcode Adapter 1--16 Kit (Applied Biosystems, Life Technologies) according to the manufacturers\' procedures. In brief, DNA concentration was measured with Quant-iT dsDNA HS Assay (Invitrogen, Life Technologies) and Qubit Fluorometer (Invitrogen, Life Technologies) and DNA quality was confirmed by agarose gel electrophoresis. 10 ng of each genomic DNA sample was amplified, using Ion AmpliSeq HiFi Master Mix (Applied Biosystems, Life Technologies) and AmpliSeq Custom primer pools, for 2 min at 99°C, followed by 15 two-step cycles of 99°C for 15 sec and 60°C for 4 min, ending with a holding period at 10°C in a PCR thermal cycler (Takara, Shiga, Japan). After the Multiplex PCR amplification, amplified DNA samples were digested with FuPa enzyme at 50°C for 10 min and 55°C for 10 min and the enzyme was successively inactivated for 60°C for 20 min incubation. After digestion, diluted barcode adapter mix including Ion Xpress Barcode Adapter and Ion P1 adaptor were ligated to the end of the digested amplicons with ligase in the kit for 30 min at 22°C and the ligase was successively inactivated at 60°C for 20 min incubation. Adaptor ligated amplicon libraries were purified with the Agencourt AMPure XP system (Beckman Coulter Genomics, Danvers, MA). The amplicon libraries were quantified by using Ion Library Quantitation Kit (Applied Biosystems, Life Technologies) and the StepOne plus realtime PCR system (Applied Biosystems, Life Technologies) according to the manufacturers\' procedures. After quantification, each amplicon library was diluted to 20 pM and the same amount of the 12 libraries for 12 patients were pooled for one sequence reaction. Emulsion PCR and Sequencing {#s4c} --------------------------- The emulsion PCR was carried out with the Ion OneTouch System and Ion OneTouch 200 Template Kit v2 (Life Technologies) according to the manufacturer\'s procedure (Publication Part Number 4478371 Rev. B Revision Date 13 June 2012). After the emulsion PCR, template-positive Ion Sphere Particles were enriched with the Dynabeads MyOne Streptavidin C1 Beads (Life Technologies) and washed with Ion OneTouch Wash Solution in the kit. This process were performed using an Ion OneTouch ES system (Life Technologies). After the Ion Sphere Particle preparation, MPS was performed with an Ion Torrent Personal Genome Machine (PGM) system using the Ion PGM 200 Sequencing Kit and Ion 318 Chip (Life Technologies) according to the manufacturer\'s procedures. Base Call and Data Analysis {#s4d} --------------------------- The sequence data were processed with standard Ion Torrent Suite Software and Torrent Server successively mapped to human genome sequence (build GRCh37/hg19) with Torrent Mapping Alignment Program optimized to Ion Torrent data. The average of 562.33 Mb sequences with about 4,300,000 reads was obtained by one Ion 318 chip. The 98.0% sequences were mapped to the human genome and 94% of them were on the target region. Average coverage of depth in the target region was 314.2 and 93.8% of them were over 20 coverage. After the sequence mapping, the DNA variant regions were piled up with Torrent Variant Caller plug-in software. Selected variant candidates were filtered with the average base QV (minimum average base quality 25), variant frequency (40--60% for heterozygous mutations and 80--100% for homozygous mutations) and coverage of depth (minimum coverage of depth 10). After the filtrations, variant effects were analyzed with the wANNOVAR web site [@pone.0090688-Wang1], [@pone.0090688-Chang1] (<http://wannovar.usc.edu>) including the functional prediction software for missense variants: Sorting Intolerant from Tolerant (SIFT; <http://sift.jcvi.org/>), and Polymorphism Phenotyping (PolyPhen2; <http://genetics.bwh.harvard.edu/pph2/>). The sequencing data was available in the DNA databank of Japan (Accession number: DRA001273). Algorithm {#s4e} --------- Missense, nonsense, and splicing variants were selected among the identified variants. Variants were further selected as less than 1% of: 1) the 1000 genome database (<http://www.1000genomes.org/>), 2) the 5400 exome variants (<http://evs.gs.washington.edu/EVS/>), and 3) the in-house control. Candidate mutations were confirmed by Sanger sequencing and the responsible mutations were identified by segregation analysis using samples from family members of the patients. In addition, the cases with heterozygous or no causative mutation were fully sequenced by Sanger sequencing for USH1 genes in order to verify the MPS results. Direct Sequence Analysis {#s4f} ------------------------ Primers were designed with the Primer 3 plus web server (<http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi>). Each genomic DNA sample (40 ng) was amplified using Ampli Taq Gold (Life Technologies) for 5 min at 94°C, followed by 30 three-step cycles of 94°C for 30 sec, 60°C for 30 sec, and 72°C for 30 sec, with a final extension at 72°C for 5 min, ending with a holding period at 4°C in a PCR thermal cycler (Takara, Shiga, Japan). The PCR products were treated with ExoSAP-IT (GE Healthcare Bio, Buckinghamshire, UK) and by incubation at 37°C for 60 min, and inactivation at 80°C for 15 min. After the products were purified, we performed standard cycle sequencing reaction with ABI Big Dye terminators in an ABI 3130xl sequencer (Life Technologies). Accession numbers {#s4g} ----------------- *MYO7A*, \[NM_000260.3\]; *USH1C*, \[NM\_ 153676.3\]; *CDH23*, \[NM\_ 022124.5\]; *PCDH15*, \[NM\_ 033056.3\]; *USH1G*, \[NM\_ 173477.2\]; *USH2A*, \[NM_206933.2\]; *GPR98*, \[NM\_ 032119.3\]; *DFNB31*, \[NM\_ 015404.3\]; *CLRN1*, \[NM\_ 174878.2\]; *PDZD7*, \[NM\_ 001195263.1\]. We thank A. C. Apple-Mathews for help in preparing the manuscript. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: HY SI SN SU. Performed the experiments: HY SN. Analyzed the data: HY SN SU. Contributed reagents/materials/analysis tools: HY SI SN KK TT YK HS KN KI TI YN KF CO TK HN SU. Wrote the paper: HY SN SU.

1. Introduction {#sec1-polymers-08-00159} =============== Phenol-formaldehyde (PF) is a high-performance resin that is synthesized by the copolymerization of phenol with formaldehyde. It is widely applied for industrial uses, including adhesives, impregnating resins, and plastics. The excellent properties of PF resin include high mechanical, thermal, and weather stability \[[@B1-polymers-08-00159]\]. However, the lower curing rate and required higher curing temperature compared to other thermosetting adhesives limit the application of PF resins for use in impregnating resins or adhesives \[[@B2-polymers-08-00159],[@B3-polymers-08-00159]\]. Many attempts have been made to accelerate the curing rate or lower the curing temperature, including testing of various catalysts or additives to alter the reaction kinetics, such as carboxylic acid esters \[[@B4-polymers-08-00159],[@B5-polymers-08-00159]\], anhydrides \[[@B6-polymers-08-00159]\], amides \[[@B7-polymers-08-00159]\], carbonate \[[@B8-polymers-08-00159]\], and metallic ions \[[@B9-polymers-08-00159]\]. Additionally, the effects of the condensation condition on the PF resin structure and properties have been well studied by conventional analytical techniques. For example, various mechanisms of PF resin hardening accelerated by catalysts or additives have been reported \[[@B10-polymers-08-00159]\]. Some additives, such as sodium carbonate, act solely to accelerate the curing reaction, but other additives, such as propylene carbonate, both accelerate the reaction and also increase the average functionality of the PF reaction system to allow a tighter final network \[[@B10-polymers-08-00159]\]. The properties of basic catalysts, such as the valence and ionic radius of hydrated cations, affected the mechanisms and kinetics of PF resin condensation and, thus, the composition of the final products \[[@B9-polymers-08-00159]\]. Some studies also reported that an alkaline catalyst promoted the formation of dimethylene ethers in the polymerization reaction, and that *ortho* to *ortho* (*o*,*o*′) ethers were more stable \[[@B11-polymers-08-00159]\]. However, there has been no comprehensive study about the action of catalysts to increase or decrease the ratio of *ortho*/*para* reaction position or analysis of the corresponding physicochemical properties of the accelerated PF resins. The aromatic ring of phenol has *ortho* and *para* positions capable of reaction with formaldehyde under certain conditions, but the *para* position has higher reactivity than the *ortho* position. The presence of two *ortho* positions and one *para* position in an aromatic ring generally could lead to a PF resin containing mostly *ortho* hydroxymethyl groups \[[@B2-polymers-08-00159]\]. However, in the process of PF resin synthesis, some catalysts could make more formaldehyde or methylol toward phenol *ortho* positions to increase the ratio of *ortho*/*para* substituted positions \[[@B12-polymers-08-00159]\], leading to more reactive functional groups or more unreacted *para* positions at the curing stage, which may shorten the curing time and increase the cross-linking degree of cured PF resin. Since metal ions can accelerate the curing of PF resins, we tested the ability of barium hydroxide (Ba(OH)~2~), sodium carbonate (Na~2~CO~3~), lithium hydroxide (LiOH), and zinc acetate ((CH~3~COO)~2~Zn) to decrease curing temperature and accelerate the curing rate of PF resins. To elucidate the chemical structure of the cure-accelerated PF resins, we performed quantitative liquid ^13^C NMR to analyze the structural features. Finally, possible synthesis mechanism of metal-mediated polymerization of PF resins was proposed based on the chemical structure analysis and thermogravimetric (DTG) curve. 2. Materials and Methods {#sec2-polymers-08-00159} ======================== 2.1. Materials {#sec2dot1-polymers-08-00159} -------------- Phenol, formaldehyde (37%), Ba(OH)~2~, Na~2~CO~3~, LiOH, and (CH~3~COO)~2~Zn were obtained from Zhong'an Chemical Industries, Beijing, China and were used directly without further purification, and all other chemicals were AR grade and obtained from Beijing Chemical Industries, Beijing, China. 2.2. Preparation of PF Resins {#sec2dot2-polymers-08-00159} ----------------------------- The catalyst-accelerated PF resin was synthesized by batch polymerization with phenol and formaldehyde at a molar ratio of 1:2.2, and the additive amount of catalyst was 6% based on the total mass of PF resin. In the first step, phenol was mixed in a flask with one third of the formaldehyde and one third of the catalyst. The mixture was quickly heated to 70 °C, and then the heater was turned off. The temperature of the mixture increased to 90 °C due to the heat produced by polymerization reaction and remained at 93--95 °C for 1 h. In the second step, the remaining formaldehyde and two thirds of the catalyst were added to the flask and the mixture was heated to 90 °C and kept at that temperature for 0.5 h. Finally, the mixture was cooled to 40 °C to yield PF resin. PF resins with different catalysts were synthesized with the same procedure. 2.3. Preparation of Plywood {#sec2dot3-polymers-08-00159} --------------------------- Three-layer plywood (400 mm × 400 mm × 4.8 mm) was prepared with a single poplar veneer in the middle and two poplar veneers on the top and bottom simulating actual industrial parameters. The middle poplar veneer was coated with 125--150 g/m^2^ resin on each side. Four pieces of three-layer plywood for each catalyst-accelerated resin (including control) were hot-pressed under 1.2 MPa at 100, 110, 120, and 130 °C, respectively. The hot-press time was 7 min, including the first one minute and the last one minute to load and unload the pressure, respectively. 2.4. Characterization of PF Resins {#sec2dot4-polymers-08-00159} ---------------------------------- The solid (non-volatile) content of resol resin was determined in accordance with ASTM standard D4426-01. The viscosity of resin was measured using a Brookfield DV-II viscometer (AMETEK-BROOKFIELD Corporation, Middleboro, MA, USA) using 61\# rotor with spinning rate of 100 rpm. Gel time was defined as the time period from the immersion of the test tube into the oil bath (135 °C) to the beginning of the resin gelation (resin forming a string when a glass rod was lifted from the resin). 2.5. Characterization of the Plywood {#sec2dot5-polymers-08-00159} ------------------------------------ The shear strength was measured as per ASTM D906-98. 2.6. FT-IR Analysis of PF Resins {#sec2dot6-polymers-08-00159} -------------------------------- The resins were placed in 0.01 MPa vacuum at 60 °C for 4 h to dry to non-volatility. FT-IR spectra of vacuum-dried PF resins were performed in a Nicolet IS10 instrument (Thermo Fisher Scientific Corporation, ‎Waltham, MA, USA). Each spectrum was recorded with 32 scans in a frequency range of 600--4000 cm^−1^ at a spectral resolution of 4 cm^−1^. 2.7. Contact Angle Measurement {#sec2dot7-polymers-08-00159} ------------------------------ The contact angle measurements of the PF resins were performed on the tangential surfaces of wood samples with an optical contact angle apparatus (OCA 20 DataPhysics Instruments GmnH, Filderstadt, Germany). Sessile droplets (3 μL, measured with a microsyringe) of liquid resin were placed on the wood surface. The right and left angles of the drops on the surface were collected at intervals of 0.1 s for a total duration of 60 s, and the average angle was calculated. 2.8. Quantitative Liquid ^13^C NMR Measurement {#sec2dot8-polymers-08-00159} ---------------------------------------------- All of the resins were characterized by quantitative ^13^C NMR spectroscopy with a VARIAN INOUR-300 (JEOL Corporation, Tokyo, Japan) spectrometer with a frequency of 75.51 MHz using the inverse-gated decoupling method. All of the spectra were recorded at room temperature with a delay time of 8 s, a 13 h acquisition time and a 15.4 μs pulse width (90°). About 8000 scans were accumulated to obtain spectra for each spectrum. The chemical shifts of each spectrum were accurate to 0.1 ppm and all the resin samples were directly used for ^13^C NMR measurement. 2.9. Thermogravimetric Analysis (TG) of Resins {#sec2dot9-polymers-08-00159} ---------------------------------------------- Samples were dried at 120 °C for 2 h to evaporate the moisture and then TG was performed in a nitrogen atmosphere within a temperature range from room temperature to 700 °C, with a heating rate of 10 °C/min. 3. Results and Discussion {#sec3-polymers-08-00159} ========================= 3.1. Performance of the Catalyst-Accelerated PF Resin {#sec3dot1-polymers-08-00159} ----------------------------------------------------- [Table 1](#polymers-08-00159-t001){ref-type="table"} shows the solid content, viscosity, and gel time of the PF resins. We found that the solid contents were similar for all resins, but the viscosity varied greatly for different catalysts. The viscosity of the Na~2~CO~3~-accelerated resin was 153.00 mPa·s, higher than the viscosity of the other catalyst-accelerated PF resins, especially the control resin with a low viscosity of 25.70 mPa·s. Similarly, the gel-time for the catalyst-accelerated resins varied from 11.83 min for the Na~2~CO~3~-accelerated resin to 20.46 min for the control resin. These results indicated that these catalysts were able to accelerate the synthesis reaction to different extents. The ability of the catalysts to accelerate the reactions could be ranked as Na~2~CO~3~ \> (CH~3~COO)~2~Zn \> Ba(OH)~2~ \> LiOH. Under the same reaction conditions, the catalyst Na~2~CO~3~ can dramatically accelerate the synthesis reaction process, increase the viscosity, and decrease the gel time of PF resin. 3.2. Contact Angle of the PF Resins {#sec3dot2-polymers-08-00159} ----------------------------------- The wettability of PF resin on solid surface is usually evaluated by contact angle \[[@B13-polymers-08-00159]\], which was tested on smoothed wood surface in this study. Due to liquid penetration and spreading on the wood surface, the contact angle changed as a function of time, as shown in [Figure 1](#polymers-08-00159-f001){ref-type="fig"}. The process of adhesive wetting includes three steps \[[@B14-polymers-08-00159]\]: (1) formation of a contact angle at the solid and adhesive interface; (2) spreading of the adhesive over a solid surface; and (3) adhesive penetration into the porous solid substrate, as shown in [Figure 2](#polymers-08-00159-f002){ref-type="fig"}. As shown in [Figure 1](#polymers-08-00159-f001){ref-type="fig"}, at the initial stage of the wetting process, the contact angle of the resins decreased quickly. As time elapsed, the contact angle decreased more slowly and finally attained relative equilibrium. It was observed that the Na~2~CO~3~-accelerated PF resin showed the largest equilibrium contact angle, and the resins could be ranked as Na~2~CO~3~ \> Ba(OH)~2~ \> LiOH \> (CH~3~COO)~2~Zn \> Control. The results suggested that the Na~2~CO~3~-accelerated PF resin had the largest surface tension, which may be due to its larger viscosity. [Figure 1](#polymers-08-00159-f001){ref-type="fig"} also shows that the (CH~3~COO)~2~Zn-accelerated resin had the fastest rate of contact angle change, which meant that (CH~3~COO)~2~Zn-accelerated resin could spread and penetrate more quickly into the porous structure of wood. Both viscosity and the chemical constitution of PF resin could alter the contact angle change rate. Usually, samples with lower viscosity exhibit a faster contact angle change rate. However, from the data in [Table 1](#polymers-08-00159-t001){ref-type="table"}, (CH~3~COO)~2~Zn-accelerated resin showed higher viscosity than that of the control sample, but had a faster contact angle change rate. Thus, the chemical constitution of (CH~3~COO)~2~Zn-accelerated PF resin may be the main factor altering its contact angle change rate. The hydroxymethyl of PF resin is the main chemical group that can easily connect with the hydroxyl of wood cellulose; thus, our data suggests that (CH~3~COO)~2~Zn-accelerated PF resin may contain more hydroxymethyl. 3.3. FT-IR Spectroscopy {#sec3dot3-polymers-08-00159} ----------------------- To investigate the structural changes in the PF resins accelerated by different catalysts, FT-IR spectra ([Figure 3](#polymers-08-00159-f003){ref-type="fig"}) were obtained after vacuum-drying the samples. The spectra assignments of the PF resins are shown in [Table 2](#polymers-08-00159-t002){ref-type="table"} \[[@B15-polymers-08-00159],[@B16-polymers-08-00159],[@B17-polymers-08-00159]\]. There were no significant differences between the spectra of the catalyst-accelerated resins and the control sample, which indicated structural similarity. Bands at 1020 cm^−1^ were ascribed to C--O stretching vibration of aliphatic C--OH, aliphatic C--O, and methylol C--OH. Bands at 1600 cm^−1^ were assigned to the elongation of aromatic --C=C--, which were consistent in each reaction system and unaffected by catalyst reaction. Thus, bands at 1600 cm^−1^ could be used as an internal standard for analysis. The ratio of absorption value of 1020 cm^−1^ (variable)/1600 cm^−1^ (constant) was calculated to indicate the degree of hydroxymethyl for phenol in each catalyst accelerated-reaction system, as shown in [Table 3](#polymers-08-00159-t003){ref-type="table"}. The control of PF resins had a relatively larger ratio of 1020 cm^−1^/1600 cm^−1^, which may be explained by the fact that the methylol of catalyst-accelerated PF resin tends to undergo further condensation reactions to form methylene (--CH~2~--). Thus, the control sample had relatively more unreacted methylol. This explanation was confirmed by the fact that the control sample showed the lowest viscosity due to its relatively minimum condensation degree. 3.4. Chemical Structure Analysis {#sec3dot4-polymers-08-00159} -------------------------------- In order to identify the effect of different catalysts on the functional groups, quantitative ^13^C NMR was used to study the difference of chemical shifts between the control and catalyst-accelerated PF resins. The ^13^C NMR spectra are shown in [Figure 4](#polymers-08-00159-f004){ref-type="fig"}, and their corresponding assignments of groups' signals are shown in [Figure 5](#polymers-08-00159-f005){ref-type="fig"} \[[@B12-polymers-08-00159],[@B18-polymers-08-00159],[@B19-polymers-08-00159],[@B20-polymers-08-00159]\]. The chemical shift of 150.0--158.0 ppm was assigned to phenoxy carbons (C1--OH), which was used as an integral standard and analytical standpoint. 156.2--156.8 ppm and 153.4--156.1 ppm were assigned to *para* alkylated groups and *ortho* alkylated groups, respectively. Usually, phenolic *ortho* and *para* carbons' chemical shifts vary with the sodium hydroxide contents of the resin due to the ionization of phenoxy group and the kind of substituted groups. Substitution with methylol groups in the *para* and *ortho* carbon positions was shown at 129.0--130.4 ppm and 127.0--128.1 ppm. Unsubstituted *para* and *ortho* carbons, the main reactive sites for the methylolation reaction, occurred at 119.2--120.4 ppm and 115.0--116.6 ppm, respectively. The unsubstituted *para* and *ortho* carbon peaks were only present in the control and Ba(OH)~2~-accelerated PF resins, which indicated that Na~2~CO~3~, LiOH, and (CH~3~COO)~2~Zn facilitate the reaction of formaldehyde with phenolic *ortho* and *para* position more than what occurs in the control and Ba(OH)~2~ samples. A sharp peak of methanol was evident around 50 ppm for all resins. Industrial formaldehyde usually contains a small amount of methanol which can also be formed during resin synthesis from the Canizzaro reaction of formaldehyde. The signal peak of methylol is sharper than the methylene peak due to its higher group mobility and less variation in the environment within the polymer structure. Thus, the two peaks at 63.3--65.5 ppm and 61.1--61.5 ppm were assigned to *para* methylol and *ortho* methylol. Theoretically, condensation between two methylols can occur to form methylene ether bridges. However, the data in [Figure 4](#polymers-08-00159-f004){ref-type="fig"} shows no peak between 69 and 74 ppm, indicating that methylene ether bridges were not formed between phenolic units during the synthesis of PF resin. The methylene bridges were easily observed in the range of 34--41 ppm. In a different chemical environment, different methylene linkages showed a different chemical shift, 39.7--41.0 ppm and 34.3--35.7 ppm were assigned to *para*--*para* and *ortho*--*para* methylene bridges, respectively. In order to remove the interference of carbon in CH~3~COO^−^ for ^13^C NMR analysis, the Zn(NO~3~)~2~-accelerated resin was also tested by ^13^C NMR analysis and used as a control for the ^13^C NMR analysis of (CH~3~COO)~2~Zn-accelerated resin. Further analysis of quantitative ^13^C NMR is needed to elucidate the details of the cure-acceleration effect of different catalysts on the structure and compositions of PF resin. In this study, the ratios of integral values of the substituted position *ortho* (127.0--128.1 ppm)/*para* (129.0--130.4 ppm), *ortho* methylol (61.1--61.5 ppm)/*para* methylol (63.3--65.5 ppm), and methylene bridges *ortho-para* (34.3--35.7 ppm)/*para--para* (39.7--41.0 ppm) were calculated, as shown in [Table 4](#polymers-08-00159-t004){ref-type="table"}. PF resins supplemented with Na~2~CO~3~, LiOH, and especially (CH~3~COO)~2~Zn, possessed higher ratios of *ortho*/*para*-substituted positions than did the control or the PF resin with Ba(OH)~2~. In case of the *ortho*/*para* ratio of methylol, the values of Na~2~CO~3~-accelerated and (CH~3~COO)~2~Zn-accelerated PF resins were not calculated (NC), because their ^13^C NMR spectra showed no signal for *para* methylol. Either Na~2~CO~3~ and (CH~3~COO)~2~Zn promoted the complete *para* methylol condensation reaction or drive formaldehyde toward the phenol *ortho* position exclusively. The ratio of *ortho* methylol/*para* methylol for the Zn(NO~3~)~2~-accelerated resin was much higher than other samples, excluding the Na~2~CO~3~-accelerated and (CH~3~COO)~2~Zn-accelerated PF resins. Usually, *para* methylol groups react more easily with other *para* positions to form *para--para* linkages. However, the data in [Table 5](#polymers-08-00159-t005){ref-type="table"} indicates that *ortho-para* linkages were equal to *para--para* linkages in the Na~2~CO~3~-accelerated, Zn(NO~3~)~2~-accelerated, and (CH~3~COO)~2~Zn-accelerated PF resins, suggesting that Na~2~CO~3~, Zn(NO~3~)~2~, and (CH~3~COO)~2~Zn were able to promote the condensation reaction to form *ortho-para* linkages. These results could be further proved by [Table 1](#polymers-08-00159-t001){ref-type="table"} that Na~2~CO~3~ and Zn(NO~3~)~2~-accelerated PF resins had higher viscosity, indicating that the significant promotion of phenol *ortho* position reactivity made Na~2~CO~3~ and Zn(NO~3~)~2~-accelerated PF resins have a tighter final network. In conclusion, all the catalysts tested showed accelerating effect to promote phenol *ortho* reactivity. However, (CH~3~COO)~2~Zn and Na~2~CO~3~ were able to significantly promote the reaction activity of phenol *ortho* position and the condensation reaction of *ortho* methylol or directed formaldehyde exclusively toward the phenol *ortho* position. 3.5. Plywood Performance {#sec3dot5-polymers-08-00159} ------------------------ [Figure 6](#polymers-08-00159-f006){ref-type="fig"} shows the bonding strength of the plywood prepared with these different PF resins. Each kind of plywood was prepared at four hot-pressing temperatures, namely 100, 110, 120, and 130 °C. Higher hot-pressed temperatures allowed the resin to cure more completely and, thus, increase the bonding strength. Under the same hot-pressing temperature, the plywood prepared with catalyst-accelerated PF resins exhibited higher bonding strength than the control sample, especially the one with Na~2~CO~3~-accelerated PF resin. The data in [Figure 6](#polymers-08-00159-f006){ref-type="fig"} shows that the plywood prepared with Na~2~CO~3~-accelerated resin at 110 °C yielded almost the same bonding strength of plywood with Ba(OH)~2~-accelerated resin at 120 °C, which is higher than that of the control sample pressed at 120 °C. The reason may be that the Na~2~CO~3~-accelerated PF resin had the highest viscosity among PF resins from [Table 1](#polymers-08-00159-t001){ref-type="table"}, or that Na~2~CO~3~ significantly improved PF resin performance by promoting the reaction activity of phenol *ortho* position. 3.6. Thermal Behavior of the Cured PF Resins {#sec3dot6-polymers-08-00159} -------------------------------------------- To characterize the thermal stability of the catalyst-accelerated PF resins, TG analysis was performed next, as shown in [Figure 7](#polymers-08-00159-f007){ref-type="fig"}. The temperatures at which the maximum degradation speed took place (*T*~max~) for the different thermal events of cured catalyst-accelerated PF resins are shown in [Table 5](#polymers-08-00159-t005){ref-type="table"}. It was previously known that phenolic resin degrades in three steps: post-curing, thermal reforming, and ring stripping \[[@B21-polymers-08-00159],[@B22-polymers-08-00159]\]. The mass loss (about 5%) of the first thermal event at the lower temperature range (\<155 °C) contributed to the evaporation of free water. In the second stage, with a temperature range from 230 to 300 °C, mass loss was due to the evaporation of water formed by the condensation reaction of methylol groups. The mass loss in the third event (from 350 to 440 °C) was due to the loss of water formed by the condensation reaction of methylol and phenolic hydrogen, as well as between two hydroxyl functional groups, which could cause further structure change of the cured products to a more tightly cross-linked network. In the fourth event (\>450 °C), the mass loss was due to the loss of carbon monoxide and methane formed by degradation of the methylene linkage. As the temperature further increased, the remaining mass was from 65% to 68% at 700 °C, and that of Ba(OH)~2~, Na~2~CO~3~, and (CH~3~COO)~2~Zn-accelerated PF resins was higher than other samples, indicating that a tighter network and higher thermal stability was possessed by their molecular structure. [Figure 7](#polymers-08-00159-f007){ref-type="fig"} shows TG (a) and DTG (b) curves of the catalyst-accelerated PF resins. All of the PF resins showed similar thermal stability in the first three stages of thermal events. However, in the final event, the DTG curve of the (CH~3~COO)~2~Zn-accelerated resin showed lower degradation speed than the other PF resins, and had double peaks of degradation speed. According to Pizzi and Mohamed *et al.* \[[@B23-polymers-08-00159],[@B24-polymers-08-00159],[@B25-polymers-08-00159]\], when a benzene ring was blended with a zinc ion, a complex compound between the phenolic nuclei and the zinc ion could be formed by a metal-ligand mode, which can accelerate the initial reaction of formaldehyde toward the phenolic nuclei by forming a carbocation of strong positive charge. As shown in [Scheme 1](#polymers-08-00159-sch001){ref-type="scheme"}, the mobility of the polymer chain was restricted by ion-polymer and ion-interaction, resulting in a higher thermal stability of PF resin than the control sample. Thus, the first of double peaks at 493 °C may indicate the breakage of the metal-ligand bonding mode, and the second peak of the double peaks at 518 °C may indicate the degradation of the methylene linkage. 4. Conclusions {#sec4-polymers-08-00159} ============== The poly-condensations of PF resins with different catalysts suggested different abilities to accelerate the reaction. In general, the accelerating efficiency of the catalysts was Na~2~CO~3~ \> (CH~3~COO)~2~Zn \> Ba(OH)~2~ \> LiOH. The addition of Na~2~CO~3~ had a remarkable influence on the performance of PF resin. The viscosity of Na~2~CO~3~-accelerated PF resin increased to around 153 mPa·s quickly, five-fold greater than the viscosity of the control resin. Moreover, the gel time of PF resin decreased significantly and the bonding strength of plywood increased by the addition of Na~2~CO~3~. The quantitative ^13^C NMR analysis showed that the (CH~3~COO)~2~Zn and Na~2~CO~3~ catalysts could significantly promote the reaction activity of the phenol *ortho* position, and favor the condensation reaction of *ortho* methylol or direct formaldehyde toward the phenol *ortho* position exclusively. Compared with Na~2~CO~3~, the catalyst (CH~3~COO)~2~Zn showed a slightly weaker accelerating effect, but the contact angle analysis found that the (CH~3~COO)~2~Zn-accelerated resin showed a faster contact angle change rate, which represents a better wettability on the wood surface. Furthermore, the different peaks in the DTG curve and higher weight residue of TG data indicated that (CH~3~COO)~2~Zn has a different accelerating mechanism to improve the thermal stability of PF resin. That mechanism may include metal-ligand bonding between the benzene ring and zinc ion formed by ion-polymer and ion-interaction. In conclusion, catalysts such as Na~2~CO~3~ and (CH~3~COO)~2~Zn showed significant accelerating effects to promote the curing of PF resin at lower temperatures and to improve PF resin performance. Thus, good catalyst-accelerated PF resins have promised to overcome the shortcoming of high curing temperature and to broaden their application. This work was supported by the Chinese National Science and Technology Support Program (2015BAD14B03), the Special Fund for Forestry Research in the Public Interest (Project 201504502), and China Postdoctoral Science Foundation Funded Project (2015M570039). Zhao Yi contributed to synthesis, test, data analysis, and wrote manuscript. Jizhi Zhang, Jianzhang Li and Wei Zhang suggested and supervised the work and revised the manuscript. Shifeng Zhang, and Qiang Gao provided constructive suggestions about this work. The authors declare no conflict of interest. Figures, Scheme and Tables ========================== {#polymers-08-00159-f001} {#polymers-08-00159-f002} {#polymers-08-00159-f003} {#polymers-08-00159-f004} {#polymers-08-00159-f005} {#polymers-08-00159-f006} {#polymers-08-00159-f007} {#polymers-08-00159-sch001} polymers-08-00159-t001_Table 1 ###### PF resin characteristics. Catalyst type Performance ----------------- ------------- -------- ------- Control 43 25.70 20.46 Ba(OH)~2~ 46 73.70 15.57 Na~2~CO~3~ 44 153.00 11.83 LiOH 46 58.30 15.88 (CH~3~COO)~2~Zn 44 81.00 13.98 polymers-08-00159-t002_Table 2 ###### Assignments of FT-IR spectra of the PF resin. Wavenumbers (cm^−1^) Assignment ---------------------- ---------------------------------------------------------------------------------- 3,367 --OH stretching vibration 2,900 C--H stretching vibration of methylene 1,600, 1440 The elongation of aromatic --C=C-- 1,270 C--O stretching vibration of phenolic C--OH and phenolic C--O 1,020 C--O stretching vibration of aliphatic C--OH, aliphatic C--O, and methylol C--OH 970 C--H stretching vibration of vinyl polymers-08-00159-t003_Table 3 ###### The ratio of absorption value of 1020 cm^−1^ (variable)/1600 cm^−1^ (constant) of the PF resins with different catalysts. Wavenumbers (cm^−1^) Absorption ---------------------- ------------ ------- ------- ------- ------- 1,020 43.46 32.63 33.79 45.43 29.85 1,600 29.75 28.14 29.94 38.54 24.81 Ratio (1,020/1,600) 1.46 1.16 1.13 1.17 1.20 polymers-08-00159-t004_Table 4 ###### Liquid ^13^C NMR analysis results of PF resin formed with different catalysts. PF resin *ortho*/*para* (Substituted position) *ortho*/*para* (Methylol) *ortho-para*/*para--para* (Methylene bridges) --------------------------------- --------------------------------------- --------------------------------- -----------------------------------------------     Control    Ba(OH)~2~ Na~2~CO~3~ LiOH (CH~3~COO)~2~Zn Zn(NO~3~)~2~ polymers-08-00159-t005_Table 5 ###### Thermal properties of the cured PF resins. Catalyst type *T~max~* of Thermal event (°C) Weight residue (%) at 700 °C ----------------- -------------------------------- ------------------------------ ----- ----- ----- ------ Control 155 260 394 507 -- 65.5 Ba(OH)~2~ 156 262 390 503 -- 68.0 Na~2~CO~3~ 153 300 386 512 -- 65.5 LiOH 158 283 381 497 -- 68.0 (CH~3~COO)~2~Zn 155 273 381 493 518 68.0

Introduction {#Sec1} ============ Headache is the commonest neurological disorder in the community with variable intensity, ranging from a trivial nuisance to a severe, disabling, acute or chronic disorder, and may impose a substantial burden on sufferers and on society \[[@CR1], [@CR2]\]. It is one of the commonest reasons for visiting the neurology clinics worldwide \[[@CR3]--[@CR5]\], exerting significant burden on its sufferers and impairing daily function especially when accompanied by other symptoms, hence adversely affecting quality of life \[[@CR6]\]. According to the World Health Organisation (WHO), 1.7 -- 4% of the adult population of the world have headaches on 15 or more days every month \[[@CR7]\] and a lifetime prevalence of more than 90% has been attributed to headache disorders in most populations of the world \[[@CR8]\]. It is known that Africans have a higher threshold for pain and may not present to the clinic just for an 'ordinary headache' \[[@CR9]\]. Local experiences show that patients suffering from other chronic neurological disorders present very late to doctors and sometimes never do so \[[@CR9]\]. Chronic headaches produce individual and societal burdens, the former referring to its effect on family, social and recreational activities and the latter referring to effects on healthcare cost (direct costs) and work and function (indirect costs), including absenteeism and reduced effectiveness \[[@CR10]\]. There is limited data for headache prevalence in Africa. In 2004, the 1-year prevalence of headache from a door-to-door survey of rural south Tanzania was 23.1% (18.8% males and 26.4% females) \[[@CR11]\]. Getahu and colleagues in Ethiopia found a 1-year prevalence rate of 73.2% \[[@CR12]\]. A 1992 study from Ibadan, South West Nigeria, found the crude life-time prevalence for at least one episode of headache to be 51% \[[@CR13]\]. In Nigeria, there is a paucity of data on the national prevalence and burden of chronic headaches \[[@CR14]\] despite the fact that it is the commonest presenting neurological disorder in the authors' environment \[[@CR1], [@CR3]\], and therefore the possibility that a big headache problem exists in Nigeria. There are also no known studies of the prevalence and characterization of headache among Nigerian healthcare workers or healthcare workers in South East Nigeria hence the relevance of this study. Aim of the study {#Sec2} ---------------- The aim of this preliminary study was to determine the frequency and pattern of headaches among a population of healthcare workers in a tertiary health institution located in South East Nigeria. Methods {#Sec3} ======= This was an epidemiological sampling-based study (Figure  [1](#Fig1){ref-type="fig"}) using a semi-structured questionnaire. The questionnaire was pre-tested in another health facility at Nsukka (a local government area similar to the study area) for content validity. English language was used to reduce cross- cultural misinterpretations and wrong understanding of terms.Figure 1**Flow chart of research activities.** The questionnaire was self- administered to all various cadres of health workers in medical unit of University of Nigeria Teaching Hospital, a tertiary health institution located in Enugu, South- East Nigeria, over a 3- month period from September -- November 2013, selected by simple random method out of the various units in the hospital e.g. surgical, medical, laboratory, physiotherapy, nutrition, administrative, laundry, transport, security, and medical record. Within these are various cadres of hospital staff: physicians, nurses, pharmacists and cleaners. Out of a total of 141 only 133 gave consent and hence were studied, giving a response rate of 94.3%. To ascertain the overall prevalence of headache, subjects were asked if they have ever had a headache within the previous six months and to note any association. They were to rate the severity of headache based on a scale of mild, moderate and severe. The impact of these severe headaches on the daily activity and the number of days they occur in a month were recorded. The character of the pain, location, duration, and the total numbers of times in the 6 months preceding the date of administering the questionnaire were also noted. Statistical Package for Social Sciences version 16 was used in statistical analysis. Comparison of multiplex groups was carried out with One Way ANOVA test. On the other hand comparison of two distinct groups was carried out with student t test. Chi-square test (and/or Fisher's exact test) was used in analysis of categorical variables. The results were revealed as mean ± SD. P value \<0.05 was interpreted as statistically meaningful. Ethical approval was obtained from the hospital ethics committee. Results {#Sec4} ======= Of the 2,450 hospital employees (450 medical doctors, 630 nurses, 50 pharmacists, and 1320 laboratory and administrative staff), 141 were selected using simple random method from the employment register and eventually only 133 health workers (71 males and 62 females) gave informed consent and were studied (response rate 94.3%). More of the respondents were males (53.4%) and most were within the 25 - 34 years age group (46.6%). Most of the workers had worked for only ≤5 years (72.9%). Table  [1](#Tab1){ref-type="table"} illustrates.Table 1**Demographic distribution and work experience of health workers**VariableFrequencyPercent***Sex***Male7153.4Female6246.6Total133100.0***Age Group***15 -- 24139.825 -- 346246.635 -- 443627.145 -- 541511.355 -- 6464.565 and above10.7Total133100.0***Number of years worked***1 -- 59772.96 -- 101813.511 -- 1575.316 -- 2043.021 -- 2553.826 -- 3021.5Total133100.0 The prevalence of headache in the past 6 months was 88.0% (among males the prevalence was 87.3% while in females it was 88.7%). There was no significant difference observed between the sexes (p = 0.806). In both sexes, primary headaches were more prevalent (71.0% in males and 76.4% in females). There was also no significant difference in the prevalence of the primary headaches among the sexes (p = 0.509). See Table  [2](#Tab2){ref-type="table"}.Table 2**Prevalence of headache among the health workers**General prevalence of headache in the past 6 monthsVariablesFrequencyPercentHeadache present11788.0Headache absent1612.0**Sex prevalence of headache**Male (%)Female (%)Headache present62 (87.3)55 (88.7)Headache absent9 (12.7)7 (11.3)Total71 (100.0)62 (100.0)χ^2^ = 0.060; P value = 0.806Type of headachePrimary44 (71.0)42 (76.4)\*Secondary18 (29.0)13 (23.6)Total62 (100.0)55 (100.0)χ^2^ = 0.436; P value = 0.509\*Secondary headache is headache with a definitive and identifiable cause found for it i.e. those with pre-existing conditions that may cause the headache e.g. hypertension, cervical spondylosis, refractive error, sleep apnoea, malaria and other febrile conditions \[[@CR15]\]. Most respondents reported ≤5 episodes of headache in the last 6 months (74.4%) and these were typically of short-lasting durations, \<60 minutes (44.4%). There was no observed periodicity to the headaches in 57.3% of cases (see Table  [3](#Tab3){ref-type="table"}). Most of the headaches were not located in any particular part of the head or side-locked (71.7%); were described as mildly severe in 59.8% of cases while 88.0% of respondents did not suffer any sleep disruption. The headaches were often not significantly disabling (73.4%) and in 93.2% of respondents did not lead to absenteeism or affect productivity at work (Table  [4](#Tab4){ref-type="table"}).Table 3**Characterization of the headaches**Variable/CharacteristicsFrequency (N =117)Percent***Number of episodes in past 6 months***1 -- 58774.46 -- 102521.411 -- 1532.616 -- 2021.6**Usual duration of headaches**Seconds1916.2Minutes5244.4Hours3630.9Days108.5***Usual time of day of the headache***Morning1815.4Afternoon1512.8Night1311.1Continuous43.4No particular time6757.3***Is the headache becoming stronger, last longer or occur more frequent?***Yes2218.8No9581.2**What is the commonest nature of the headache?**Throbbing/exploding4336.8Sharp43.4Tightness54.3Dull65.1Aching2420.5Pressure in head3227.3Grinding32.6Table 4**Usual location and severity of the headache*Usual Location of headache***FrequencyPercentLeft side32.6Forehead97.7Around the head/ill-defined119.4Right side21.7Both Temples21.7Top of the head10.9Neck21.7Back of head32.6No particular side8471.7***Severity of headache***Mild7059.8Moderate4538.5severe21.7***Is the headache strong enough to wake you from sleep?***Yes1412.0No10388.0***Effect of headache on daily activities***No significant disability1613.7Mild disability8673.4Moderate32.6Severe disability1210.3***Headache --related work absenteeism or reduced productivity?***Yes86.8No10993.2 Stress (35.0%) and head trauma/illness/infection (18.8%) were the commonest predisposing conditions to the headache (Table  [5](#Tab5){ref-type="table"}). Refractive errors were present in 16.2% of respondents with headaches. In 25.6% there were headache prodromes and these included irritability (10.3%) and fatigue (5.1%). During the headaches, associated symptoms occurred in 30.8% of respondents and these included nasal congestion, redness of eyes, sinusitis or allergies (26.5%) as depicted in Table  [6](#Tab6){ref-type="table"}. In most cases, there was no known family history of migraines or other chronic headaches (Figure  [2](#Fig2){ref-type="fig"}).Table 5**Predisposing conditions to the headache**Factors preceding the headacheFrequency (N =117)PercentAccident, illness or infection2218.8Odours54.3Fatigue3429.1School21.7Hunger1714.5Noise43.4Stress4135.0Exercise10.9Family problem21.7Menstrual flow21.7Lack of sleep86.8Hot weather21.7None4941.9**Existing chronic medical conditions that may cause headache**Hypertension108.5Cervical spondylosis32.6Refractive errors1916.2Diabetes mellitus21.7Sleep apnoea10.9None9782.9Note that some respondents filled more than one option.Table 6**Headache prodromes and other features associated with the headaches**Frequency (N =117)Percent***Presence of warning signs before headache***Yes3025.6No8774.4***Warning signs***Pallor10.9Mood swing65.1Irritability1210.3Dizziness32.6Tired/sleepy65.1Rings around the eyes10.9Hyperactivity10.9Eye problems21.7None10488.9Other symptoms associated with the headaches***Presence of other symptoms during the headaches***Yes3630.8No8169.2Nasal congestion, redness of eyes, sinusitis or allergies associated with the headache3126.5Nausea.21.7Stomach pain97.7Vomiting10.9Confusion32.6Numbness in arms and legs65.1Diarrhoea10.9Dropping of the eyes10.9Fever1210.3Note that some respondents filled more than one option.Figure 2**Family member with history of headaches, migraines, sick headaches, motion sickness or had trouble taking birth control pills because of headaches.** Management of headache was varied among respondents. In most cases (47.9%) no intervention was required. However in other instances, investigations (11.1%) and eye checks (7.7%) were done. The over-the counter- available analgesic, paracetamol, (83.8%) was the commonest treatment received (Table  [7](#Tab7){ref-type="table"}).Table 7**Management received for the last headache episode**Management actionsFrequency (N =117)Percent*Headache was managed by --*Health worker1916.2Self4235.9No treatment received5647.9*A. Investigations done*Laboratory1311.1Eye check97.7*B. Treatment received*Anti-malaria76.0Ergotamine10.9Food10.9Ibuprofen54.3Other NSAIDs\*21.7Paracetamol9883.8Tramadol (narcotic analgesic)10.9Eye glasses were prescribed1613.7Relaxation10.9*Other actions that relieve the headaches*Cold compress1311.1Eating2017.1Massage32.6Moving around32.6Relaxation4740.2Sleep3126.5Vomiting10.9Others10.9\*NSAIDs = non-steroidal anti-inflammatory agents.Note that some respondents filled more than one option. The health workers' ages did not significantly affect both the presence and treatment of headache (p = 0.483 and 0.293 respectively) but significantly affected the type of headache (p = 0.005) i.e. whether it was primary or secondary headache (Table  [8](#Tab8){ref-type="table"}). Years of working in the hospital did not significantly affect the prevalence of headache (P = 0.123), type of headache (P = 0.423) or treatment of the headaches (P = 0.535) as shown in Table  [9](#Tab9){ref-type="table"}. There was no correlation between the number of headache episodes and the number of years worked in the hospital \[Pearson Correlation (r) = - 0.066\] or age of the health worker \[r = 0.001\].Table 8**Age group and management of headache**Age group in yearsVariables15 -2425 - 3435 - 4445 - 5455 - 64≤ 65***Presence of headache***Yes12(92.3)53(85.5)34(94.4)13(86.7)4(66.7)1(100.0)No1(7.7)9(14.5)2(5.6)2(23.3)2(33.3)0(0.0)Total13 (100.0)62(100.0)36(100.0)15(100.0)6(100.0)1(100.0)Likelihood-ratio χ^2^ = 4.480; P value = 0.483***Type of headache***Primary7(58.3)45(84.9)27(79.4)6(46.2)1(33.3)0(0.0)Secondary5(41.7)8(15.1)7(20.6)7(53.8)3(66.7)1(100.0)Total12(100.0)53(100.0)34(100.0)13(100.0)4(100.0)1(100.0)Likelihood-ratio χ^2^ = 16.995; P value = 0.005 (significant)***Treatment of headache***Other health worker10(83.3)33(62.3)18(52.9)10(76.9)3(66.7)1(100.0)Self2(16.7)20(37.7)16(47.1)3(23.1)1(33.3)0(0.0)Total12(100.0)53(100.0)34(100.0)13(100.0)4(100.0)1(100.0)Likelihood-ratio χ^2^ = 6.135; P value = 0.293Table 9**Number of years worked in the hospital and management of headache**Number of years worked in the hospitalVariables1 - 1011 - 2021 - 30***Presence of headache***Yes101(87.8)11(100.0)5(71.4)No14(12.2)0(0.0)2(28.6)Total115(100.0)11(100.0)7(100.0)Likelihood-ratio χ^2^ = 4.199; P value =0.123***Type of headache***Primary75(65.2)8(72.7)3(42.9)Secondary40(34.8)3(27.3)4(57.1)Total115(100.0)11(100.0)7(100.0)Likelihood-ratio χ^2^ = 1.719; P value = 0.423***Treatment of headache***Other health worker78(67.8)7(63.6)6(85.7)Self37(32.2)4(36.4)1(14.3)Total115(100.0)11(100.0)7(100.0)Likelihood-ratio χ^2^ = 1.250; P value = 0.535 Discussion {#Sec5} ========== Headache is the commonest presenting neurological disorder in most communities and clinical settings worldwide \[[@CR2], [@CR12]\]. Studies from Nigeria, including Enugu, also support this \[[@CR1], [@CR3], [@CR13]\]. The prevalence of headache in health care workers has been variously reported from Western countries \[[@CR16], [@CR17]\] but there is a paucity of similar data from Nigeria and Africa. There was an inability to assess headaches as distinctly experienced in the various cadres of hospital workers and it was also not possible in this study to ascertain distinct headache entities and their roles. Other limitations of this study were its small sample size, the possibility of recall bias arising from patients' answers to occurrences of headaches in the past 6 months, and use of a 3-point pain scale instead of the 10-point Visual Analogue Scale (VAS) which has greater scale refinement and discrimination power. VAS has been noted to be a valid instrument for measurement of pain intensity in patients with headaches \[[@CR18]\]. A prevalence of 88.0% was obtained for headaches amongst the hospital workers, with slightly higher rates in females than males. Though the study periods vary, the figure compares favourably with the rate of 84.4% obtained amongst from health workers in the United States \[[@CR16]\] but is significantly higher than the 54.7% and 27.1% prevalence rates obtained from Italian and Turkish health workers respectively \[[@CR16], [@CR19]\]. A survey of headache in Ethiopian textile workers found a prevalence of 73% \[[@CR12]\]. The headache prevalence of 88% for hospital workers in this study compares favourably with the 88.3% prevalence found in a study of medical students in the same locality \[[@CR20]\]. The prevalence is also higher than the community prevalence rates of 51% and 23.1% seen in Ibadan, South West Nigeria and rural south Tanzania respectively \[[@CR11], [@CR13]\]. It is possible that the different figures may reflect a combination of environmental challenges, durations of study and varied survey instruments used. It is well noted that females tend to have higher rates for headache prevalence across cultures and continents \[[@CR1], [@CR3], [@CR12], [@CR16], [@CR17], [@CR19]\] and while this seemed to be the case in our study, the difference was not statistically significant. Reasons adduced for the higher female prevalence include the influence of oestrogens and progesterone on headaches after menarche and the greater propensity for females to seek medical attention for headaches \[[@CR21]\]. Most of our subjects had probable primary headaches although no further attempts were made in this study to distinguish between the various different types (which include the trigeminal associated cephalalgias (TACs), migraine and tension- type headaches). Headaches were of short duration (\<60 minutes) and were not side --locked in most instances unlike the longer duration (\>6 hours) migraine headaches noted in the Turkish study \[[@CR19]\]. Migraine headache prevalence rate is uniformly low across much of Africa but was found to be significantly high in a cohort of textile mill workers in Ethiopia \[[@CR12], [@CR22]\]. Stress, probably related to challenges in the work environment, played the greatest role (35.0%) in this study and this reflected in the calming role attributed to relaxation techniques utilized by the health workers (40.2%) to manage their headaches. Besides life and work stress, personality traits such as aggression, anger and type A behaviour are factors that may aggravate stress and are frequently found in headache patients but were not sought for in this study \[[@CR22]--[@CR24]\]. There was no significant association or correlation found between the prevalence of headaches and years of working experience in this study. Non-pharmacological treatment was suitable for almost half of respondents (47.9%) while the over-the --counter medicine, paracetamol, was the most utilised drug treatment. This finding is essentially similar to that of health workers with headaches in the Unites States but contrasts with the use of NSAIDs in a Turkey study \[[@CR17], [@CR19]\]. Despite working in a health facility, self-medication was commonly practised (35.9%) but this was even more significant among Turkish health workers (54.6%) \[[@CR19]\]. The low rate of medical consultation for headache in hospital workers is of interest. In this study centre, headache ranks low among the disorders seen at both the Pain Clinic and Neurology Clinic accounting for only 2.7% of all neurological cases seen in the latter and 9^th^ of the top 10 disorders encountered \[[@CR3]\]. Some reasons adduced for the low rate of presentations to clinics for headaches as well as low rates of success in headache treatment amongst Africans include underdiagnoses or misdiagnoses due to lack of adequate knowledge by healthcare professionals, headache sufferers being ignorant of effective prophylaxis and treatment, perception of headaches as a trivial problem, and great tolerance to pain \[[@CR25]--[@CR29]\]. Other reasons include poor healthcare facilities \[[@CR30]\], low economic power \[[@CR25]\], gender/child discrimination \[[@CR28]\], and unavailability of effective medication \[[@CR28], [@CR29]\]. The authors are of the opinion that African patients' preference for/reliance on non-drug options (complementary and alternative medicine, CAM) \[[@CR25], [@CR28], [@CR30]\] for pain relief may also be contributory. Of important economic interest is the rarity of absenteeism from work or loss of productive time as reported in this study. These factors are important because many headache sufferers are at the peak of their work-productive life \[[@CR26]\]. Employers may lose an average of 12 days per year because of an employee headache syndrome \[[@CR27]\]. The authors relate reason for the rarity of work absenteeism and loss of productive time to the majority of headaches being of a mild nature with low disabling rates. A similar negligible rate of absenteeism was the outcome among Italian health workers with headache \[[@CR16]\]. Conclusion {#Sec6} ========== This preliminary study has revealed headaches to be common in this community of healthcare workers. However, the seemingly low effect of headache on health workers productivity in this study, despite its high prevalence rate and contrary to views \[[@CR28]\] from other African studies, is of notable relief in a developing economy like Nigeria where health indicators are unimpressive and medical services still face huge challenges. In addition, presentation to Pain or Neurology clinic for headache disorders by respondents in this study has been shown to be low, demonstrating the need for increased and continuous health awareness on headache disorders as well as enhanced occupational health services in Nigerian hospitals. By the findings of this work, the authors encourage more robust studies on headache disorders among healthcare workers in African countries with a view to informing better practice decisions and reducing the global headache burden. **Competing interest** The authors declare that they have no competing interests. **Authors' contributions** TO and IO conceptualised the study; TO, IO, EA, BE, OE and EO designed the questionnaire and collected data; TO, EA and IO analysed the data; all authors participated in drafting the manuscript; all authors read and approved the final manuscript. The authors are grateful to Dr. Ada Shirley for her co-operation with data collection. **Declaration** This study was not supported by a grant.

INTRODUCTION ============ Silicon (Si), in the form of dissolved silicic acid---often referred to as silicate and hereafter abbreviated as DSi---is an inorganic nutrient instrumental to ocean functioning. Its availability modulates processes as relevant as ocean primary productivity ([@R1]) and the exchange of CO~2~ with the atmosphere ([@R2]). Regional patterns of DSi availability largely result from the consumption of this nutrient by marine organisms (silicifiers) to build their skeletons of biogenic silica (BSi), with diatom utilization among the best known and quantified ([@R1], [@R3]). However, sponges, radiolarians, silicoflagellates, choanoflagellates, testate amoebae, and chrysophyceans, among others, also consume DSi in the ocean ([@R4]), but their activity remains poorly quantified and little understood from a physiological and molecular perspective. In the diatoms, the kinetics of DSi uptake have been investigated in a large variety of species, all of which were reported initially to follow a saturable Michaelis-Menten model. It is also that those saturable kinetics shift into nonsaturable uptake when DSi availability drastically increases and/or under particular physiological conditions ([@R5], [@R6]). Likewise, membrane silicon transporters (SITs) incorporating actively ambient DSi into the diatom cell have long been described and, although passive transporters have not been identified yet, a diffusion-based uptake has been described for at least some diatoms at high DSi availability \[reviewed in ([@R7])\]. In contrast, little is known about these physiological and molecular processes in other groups of marine silicifiers. Because diatoms are restricted to the photic zone of the ocean, a major gap in knowledge relative to DSi utilization in the dark ocean by nonphototrophic silicifiers persists. The discoveries that extensive aggregations of highly silicified sponges are common in the deep sea ([@R8]), that they can accumulate substantial amounts of BSi at the regional scale ([@R9], [@R10]), and that they trigger significant losses of BSi from the ocean ([@R11]) have raised considerable interest in deciphering how Si is processed in these singular, sponge-dominated, deep-sea systems. The lack of knowledge regarding these processes in sponges currently hinders the ability to understand the Si utilization in the dark ocean and makes it difficult to model adequately the role of the biological component in the marine biogeochemical cycle of silicon ([@R3], [@R11]). Information on the physiology of DSi consumption by sponges is sparse and derived exclusively from shallow-water species in the class Demospongiae. These studies indicate Michaelis-Menten kinetics but with optimal DSi consumption attained at environmental DSi concentrations of \>100 μM ([@R12]--[@R16]). Because DSi concentrations higher than 100 µM are virtually never reached in the shallow waters of the modern ocean ([@R17]), the skeletal growth of all shallow-water demosponges investigated to date is therefore chronically limited by Si availability ([@R15], [@R16], [@R18]--[@R20]). Whether this kinetic limitation also applies to sponges in the class Hexactinellida---deep-sea specialists characterized by impressive siliceous skeletons---remains unknown. The physiology of DSi consumption and the molecular pathways of DSi uptake remain largely uninvestigated in hexactinellids, hampered by impediments to conducting in situ and laboratory experimentation with such relatively large and delicate deep-sea animals. Nevertheless, the interest is enormous. Major differences in the kinetics of DSi utilization between the two major lineages of siliceous sponges (i.e., Demospongiae and Hexactinellida) cannot be discarded, as silicification in demosponges revolves around the activity of the nonsoluble silicatein enzyme ([@R21]), while the process in hexactinellids appears to be governed by a phylogenetically unrelated, soluble enzyme, glassin ([@R22]). In addition, because hexactinellids have essentially syncytial organization while demosponges have a conventional cellular organization, the membrane transporters involved in Si utilization may not be shared but may be lineage specific instead. Very little is known on molecular Si transport in demosponges ([@R23], [@R24]) and nothing in hexactinellids. This lack of knowledge obscures the understanding of the evolution of the biosilicification process in the animal kingdom and its relationships to that in other organisms. Here, we characterized experimentally the kinetics of DSi consumption in the hexactinellid sponge *Vazella pourtalesii* (Schmidt, 1870), a rosellid distributed from \~100 to 935 m in the northwest Atlantic that forms extensive monospecific aggregations on the deep continental shelf off Nova Scotia ([Fig. 1A](#F1){ref-type="fig"} and fig. S1), eastern Canada ([@R25]). We tested the laboratory-based kinetic model by comparing its predictions to both in situ determinations of DSi consumption rates using incubation chambers ([Fig. 1, B to D](#F1){ref-type="fig"}, and movies S1 and S2) and rates of BSi production derived from individuals of a known age grown in the wild on an artificial substrate ([Fig. 1, E and F](#F1){ref-type="fig"}). In combination with those physiological experiments, we conducted a quantitative large-scale assessment of gene expression as a function of DSi availability. The results of this study offer a mechanistic explanation for the kinetics of DSi utilization in hexactinellids and provide fresh insights into the molecular systems of Si transport and their evolution within sponges and across other silicifying organisms. {#F1} RESULTS ======= Modeling and testing the physiology of DSi consumption ------------------------------------------------------ Live sponges were collected using the remotely operated vehicle (ROV) Remotely Operated Platform for Ocean Sciences (ROPOS) and taken to the laboratory for incubation in progressively increasing DSi concentrations (12, 30, 60, 100, 150, 200, and 250 μM DSi; see Materials and Methods). Initially, all 11 assayed individuals increased their DSi consumption rate in response to the progressive increase of DSi availability in the seawater ([Fig. 2, A and B](#F2){ref-type="fig"}, and tables S1 and S2). As also known for demosponges, the DSi consumption rate notably varied among individuals, with an average maximum consumption of 0.106 ± 0.050 μmol Si per milliliter of sponge tissue and per hour (hereafter given as μmol Si ml^−1^ hour^−1^) at an average DSi concentration of 150.9 ± 69.3 μM. Over that concentration threshold, the consumption rate of most individuals did not increase with increasing DSi availability, revealing that the Si transport system reaches the maximum speed (i.e., optimal utilization) at about 150 μM DSi and saturates at higher concentrations. {#F2} Unlike in all demosponges studied to date, the model best fitting the average DSi consumption rate of *V. pourtalesii* in response to DSi availability did not follow Michaelis-Menten kinetics (*r*^2^ = 0.841, *P* = 0.004; [Fig. 2A](#F2){ref-type="fig"}). An exponential rise to a maximum (ERM) model showed the best fit (*r*^2^ = 0.898, *P* = 0.002; [Fig. 2, A and B](#F2){ref-type="fig"}) to the empirical data on DSi consumption as a function of DSi availability, "consumption rate = *a* (1 − *b*^\[DSi\]^)". Although the difference in statistical fit between the "Michaelis-Menten" and "ERM" models was apparently small, the ERM model was built on parameters with higher statistical significance (*a* = 0.093 ± 0.008, *P* \< 0.001; *b* = 0.978 ± 0.006, *P* \< 0.001) than those of the Michaelis-Menten model \[*V*~max~ = 0.114 ± 0.019 μmol Si ml^−1^ hour^−1^, *P* = 0.002; Michaelis constant (*K*~m~) = 44.876 ± 23.934 μM Si, *P* = 0.120\]. The "*a*" parameter of the ERM model is the exact conceptual equivalent of the *V*~max~ in the Michaelis-Menten model, indicating a maximum velocity of DSi utilization of 0.093 ± 0.008 μmol Si ml^−1^ hour^−1^. Likewise, the exact ERM equivalent of the *K*~m~ parameter (i.e., the DSi concentration at which half-saturation or half *V*~max~ is achieved) can also be calculated as "\[DSi\] = log 0.5/log *b*," after having replaced in the ERM equation "consumption rate = 0.5*a*". It yields a value of 31.16 μM, revealing comparatively low affinity for the DSi in this sponge species (see [Fig. 3](#F3){ref-type="fig"}). ![Comparative summary of the DSi consumption kinetics in sponges.\ (**A**) The kinetics of *V. pourtalesii* is compared against all other sponge species investigated to date ([@R13], [@R15], [@R16], [@R18]), which were demosponges with Michaelis-Menten kinetics. For relevant physiological comparison, DSi consumption rates were normalized to ash-free dry weight (AFDW), which represents essentially the organic component of the sponge that could be involved in silicification. The DSi consumption kinetics of *V. pourtalesii*, which does not follow a Michaelis-Menten model, is among the less efficient, except for that characterizing a group of slow-growing species in the genus *Axinella.* (**B**) A zoom on the graph within the range of natural DSi concentrations illustrates how *V. pourtalesii* is also less efficient than most demosponges at low DSi availability.](aba9322-F3){#F3} We tested the predictions of the developed kinetic model against empirical determinations of DSi consumption and BSi production rates in field conditions. Using five custom-manufactured methyl methacrylate chambers incorporating two ROV-operated seawater collectors, we incubated four sponge individuals and a control chamber under natural settings ([Fig. 1, B to D](#F1){ref-type="fig"}, and movies S1 and S2). Incubations were conducted in the densest sponge aggregations of both the Sambro Bank Sponge Conservation Area and LaHave Basin (fig. S1) at depths of \~160 to 185 m, respectively, for periods varying from 19 to 28 hours, and at an average DSi concentration of 15.56 ± 0.68 μM. Individuals of different sizes were assayed (64, 126, 323, and 492 ml in volume; table S3), so that a relatively wide range of the size spectrum in the natural population was considered (all but very small or very large sponges). In situ consumption rates ranged from 0.007 to 0.034 μmol Si ml^−1^ hour^−1^, averaging 0.024 ± 0.012 μmol Si ml^−1^ hour^−1^. This average consumption was markedly similar to the one predicted (0.027 ± 0.006 μmol Si ml^−1^ hour^−1^) by the laboratory kinetics at a DSi concentration of 15.56 μM, falling within the 95% confidence range of the model ([Fig. 2B](#F2){ref-type="fig"}). We also estimated the rate at which BSi---that is, the siliceous skeleton---was produced by the sponges under natural conditions to compare BSi production rates to DSi consumption rates obtained both from the in situ incubations and the predictions of the laboratory kinetic model. The recovery of two moorings that were immersed for 15 and 58 months brought up sponges that had settled on them, making the approach possible ([Fig. 1, G and H](#F1){ref-type="fig"}; Materials and Methods). The two largest sponges on the mooring deployed for 15 months were about 14 months old, 1.4 and 2.9 cm in height, 1 and 3 ml in body volume, and 0.03 and 0.11 g in BSi content, respectively. The three largest sponges on the mooring deployed for 58 months were about 54 months old and ranged from 10 to 13 cm in height, 100 to 158 ml in volume, and 3.5 to 8.3 g in BSi content. These data indicated that skeletal BSi was produced at a rate threefold higher (0.056 ± 0.008 μmol Si ml^−1^ hour^−1^) during the first 14 months of life than in subsequent years (0.019 ± 0.004 μmol Si ml^−1^ hour^−1^), a growth pattern also known from other aquatic invertebrates ([@R26]). When the data from all five individuals were pooled together, an average BSi production rate of 0.033 ± 0.021 μM Si ml^−1^ hour^−1^ emerged. Again, this value fell within the 95% confidence range of the model prediction (0.029 ± 0.007 μM Si ml^−1^ hour^−1^) at a DSi availability of 16.93 μM ([Fig. 2B](#F2){ref-type="fig"}), which is the average DSi concentration in the bottom water on the central Scotian Shelf (table S4), as measured during a 20-year monitoring program ([@R27]). The general agreement among the rates of Si utilization predicted by the kinetic model, those measured through in situ incubations and those derived from the BSi production in field conditions, indicates consistency in the responses of the sponges in different situations, confirming that laboratory experiments are a suitable proxy for DSi utilization. It also suggests that, approximately, all consumed DSi (at least at natural ambient concentrations) becomes BSi production. The DSi consumption kinetics of *V. pourtalesii* reaching optimal utilization at about 150 μM DSi suggests that the natural population suffers from chronic DSi limitation, as mean DSi availability in the bottom water of the central Scotian Shelf averages only 16.93 ± 8.65 μM (table S4). Concentrations larger than 50 μM have not been measured at any depth in the North Atlantic ([@R17]). A comparison of the DSi utilization kinetics known for sponges to date ([Fig. 3](#F3){ref-type="fig"}) indicates that *V. pourtalesii* is comparatively less efficient in DSi consumption than all demosponges but *Axinella* spp. This suggests that to build and maintain the highly silicified skeletons characterizing hexactinellid sponges, they may need continuous exposure to DSi levels higher than those characterizing the modern photic ocean \[\<10 μM; ([@R1])\]. The question remains, however, as to why the DSi consumption system of these sponges persists largely maladapted, unable to evolve in response to a shrinking DSi availability that started in the oceans at least 60 million years (Ma) ago ([@R19], [@R28]) or even earlier ([@R29]). Molecular insight into DSi consumption -------------------------------------- To elucidate why hexactinellid sponges are particularly inefficient when using DSi at the relatively modest concentrations of the modern ocean, we attempted to activate and identify in *V. pourtalesii* (see Materials and Methods) the Si transporters potentially involved in the process of both DSi uptake and its internal transport. To this end, we quantified gene expression in six of the sponge individuals that had been exposed to progressive DSi enrichment (from 12 to 250 μM DSi) during the kinetic experiment, as indicated in the "Ex situ incubations for kinetics of DSi consumption" section. Their gene expression was contrasted with that of six individuals not exposed to any DSi enrichment but to the natural (12 to 17 μM DSi) concentration (hereafter referred to as "control individuals"). The set of treated individuals (hereafter referred to as "DSi-enriched individuals") consisted of the sponges \#3, \#4, \#5, \#7, \#9, and \#10 used in the kinetic experiment (as indicated in [Fig. 2A](#F2){ref-type="fig"}). A de novo reference transcriptome was obtained after pooling the reads from the 12 sponge libraries, and its resulting summary metrics (table S5) showed it to be well assembled and with very high BUSCO completeness scores (95.1% of the eukaryotic cassette and 87.3% of the metazoan cassette). In the DSi-enriched individuals, 597 genes were differentially up-regulated relative to the control group, of which 269 had a BLAST hit against the RefSeq and 197 against Swiss-Prot databases (fig. S2 and data file S1). In the control, 980 genes were found up-regulated when compared with the DSi-enriched individuals (fig. S2 and data file S1). Among the genes up-regulated in the DSi-enriched individuals, only 131 (33%) had a gene ontology (GO) term annotation. Identified overexpressed genes in the DSi-enriched individuals belonged to a wide array of functional categories (fig. S3 and table S5). We identified abundant transmembrane transport and vesicle-mediated transport categories, as well as responses to stress, lipid metabolism, and mRNA processing, among others in the Biological Process category. In addition, certain molecular functions were up-regulated, such as lysosome-related genes (e.g., solute carriers and cathepsins among others), transporter activity, chitin binding, and oxidoreductase activity ([Fig. 4A](#F4){ref-type="fig"} and fig. S3). Indirect evidence indicates that silicification in sponges is a complex, energy-consuming biological process ([@R10], [@R12]). Therefore, up-regulation of multiple gene pathways not directly related to Si utilization was not unexpected. For our study, we focused only on those genes that had previously been demonstrated in the preexisting literature of other organisms as being involved either in Si transport or in Si polymerization. This circumvents the need to carry out further unrealistic heterologous expressions or knockout experiments with *V. pourtalesii*, a deep-sea animal for which any subsequent gene functionalization would require additional unaffordable economic and logistic transnational investments for additional experimental work with live individuals. {#F4} We found two homologs of the gene *glassin*, which code for the only silicifying protein identified in hexactinellids to date ([@R22]). Unexpectedly, only *glassin 1* was slightly up-regulated (see Discussion) but not differentially expressed (DE), with its expression level in the DSi-enriched group being barely twofold that of the control group ([Fig. 4](#F4){ref-type="fig"}). *Silicatein* genes, which are members of the cathepsin family of cysteine proteases and code for the silicifying enzyme of demosponges, had occasionally been reported from hexactinellids ([@R30], [@R31]). However, no *silicatein* sequences were found in *V. pourtalesii*, a result consistent with the growing suspicion that initial reports of silicatein in hexactinellids were either contamination from demosponge samples or misidentified cathepsin-like proteins not involved in the silicification ([@R22], [@R32]). Two gene groups of transmembrane proteins related to Si transport (*aquaporins* and *ArsB*) were up-regulated in the DSi-enriched individuals. Aquaporins are ancient channel proteins that facilitate bidirectional passive---but relatively selective ([@R33])---flux of water and/or small noncharged solutes across membranes and that are present in all kingdoms of life ([@R34]). The overexpressed *aquaporins* consisted of three genes ([Figs. 4](#F4){ref-type="fig"} and [5A](#F5){ref-type="fig"}). One of the *V. pourtalesii* protein sequences---aquaporin 3--like---showed high similarity to aquaglyceroporin 3 of chordates, a second one---aquaporin 9--like---was more similar to the aquaglyceroporin 9 of chordates, while a third one---aquaporin 3/9--like---had equal sequence similarity to both aquaglyceroporin 3 and 9. Only this later aquaglyceroporin 3/9 showed a much higher---and statistically significant---overexpression in the DSi-enriched individuals ([Fig. 4](#F4){ref-type="fig"}). Aquaglyceroporins 3 and 9 are known to be major intrinsic proteins that function in chordates as passive channels facilitating a Si inflow from the intercellular medium into cells ([@R35]). {#F5} In a maximum likelihood (ML), noncomprehensive, phylogenetic analysis of the aquaporin protein family, which produced a tree topology with 100% congruence to that of an alternative Bayesian approach, four main clades were obtained ([Fig. 5A](#F5){ref-type="fig"} and fig. S4): (i) one containing the aquaglyceroporins 3, 7, and 9 as well as the plant nodulin26 transporters (NOD26) and Lsi1 sequences from plants, which are all involved in passive transport of silicic acid, glycerol, and possibly water to the cells; (ii) a clade containing plasma membrane intrinsic (PIP) and tonoplast intrinsic (TIP) aquaporins, involved in water, glycerol, and ammonia transport; (iii) a clade containing aquaporin 8 (present in some sponges but not in hexactinellids) and small basic intrinsic proteins (SIP), mediating water, ammonia, and urea passive flow; and (iv) a clade of several aquaporins and PIPs involved in water transport ([Fig. 5A](#F5){ref-type="fig"}). Only the first clade, recovering aquaglyceroporins 3, 7, and 9, showed robust nodal support. The three *V. pourtalesii* aquaglyceroporin sequences involved in DSi transport also clustered (with statistical significance) with other unnamed aquaporins that we have recovered from the transcriptome of the hexactinellid *Rosella fibulata* and several demosponges, including both silicifying (i.e., *Lubomirskia baikalensis*, *Cliona varians*, and *Petrosia ficiformis*) and nonsilicifying species (i.e., *Dendrilla antarctica* and *Ircinia fasciculata*; see Discussion), collectively forming a robust "sponge clade" ([Fig. 5A](#F5){ref-type="fig"} and fig. S1). While demosponges had a single member of this aquaporin clade (3, 7, and 9), hexactinellids showed expansion into three members ([Fig. 5A](#F5){ref-type="fig"}), which could be the result of gene duplication and initial subfunctionalization for improved Si transport (see Discussion). The other gene group related to Si utilization that was differentially overexpressed in the DSi-enriched individuals were *arsB*- and/or *Lsi2*-like genes ([Figs. 4](#F4){ref-type="fig"} and [5B](#F5){ref-type="fig"}), members of a superfamily of active ion transporters (Na^+^/H^+^ antiporters) that mediate a selective efflux of metalloids from the cytoplasm. The independent discovery of the *arsB* genes---initially described from bacteria as for transmembrane transporters of arsenic ([@R36]) and other metalloids ([@R37])---and the *Lsi2* (low silicon 2) genes ([@R38], [@R39])---first identified in plants, as coding for transmembrane transport of Si and other metalloids---favored two different gene denominations (i.e., *arsB* versus *Lsi2*). This historical nomenclature still persists despite these genes showing clear sequence orthology and strong similarity in function (see Discussion). Two *arsB/Lsi2* genes (herein referred to as *arsB 1* and *arsB 2*) were found differentially overexpressed in *V. pourtalesii* ([Fig. 4](#F4){ref-type="fig"}). *ArsB* genes also occur in the demosponge *Amphimedon queenslandica* and are known in several silicifying eukaryotes ([Fig. 5B](#F5){ref-type="fig"} and fig. S4), including diatoms, radiolarians, and choanoflagellates ([@R23]). These eukaryotic versions of *Lsi2* and *arsB* genes also show sequence similarity to the prokaryotic arsenic transporters ([Fig. 5B](#F5){ref-type="fig"}). In our phylogenetic hypothesis for the evolution of arsB/Lsi2 transporter proteins, in which the ML and the Bayesian approaches showed complete tree topology congruence ([Fig. 5B](#F5){ref-type="fig"}), four main subclades were recognized among the arsB/Lsi2 proteins: (i) one consisting of the arsB homologs of bacteria, with a single arsB domain; (ii) another large clade composed by the homologs of diatoms and plants, with the diatom sequence containing a CitMHS domain \[citrate-Mg^2+^:H^+^(CitM)-citrate-Ca^2+^:H (CitH) symporter\] and the plant homologs both Nhab and arsB domains; (iii) a small clade containing the Lsi2-like homologs of choanoflagellates, which only have an arsB domain; and (iv) a large clade containing the animal homologs of arsB/Lsi2, with a single CitMHS and one or several transmembrane domains. Of note, SITs, which are sodium-coupled transmembrane proteins operating as active silicic acid--specific transporters in a variety of unicellular silicifying eukaryotes ([@R23]), such as diatoms, choanoflagellates, and haptophytes, were absent in *V. pourtalesii*, in agreement with other studies on sponge genomes ([@R23]). Related silicon transporter--like genes (*SIT-L*s), occurring in some metazoans such as annelids, copepods, and tunicates ([@R23]), were also absent. Likewise, the singular active Si transporter of vertebrates, solute carrier transporter Slc34a2 ([@R40]), was also absent. The NBC (Na^+^/HCO~3~) transporter, which was tentatively suggested to be involved in cotransporting DSi in the demosponge *Suberites domuncula (*[@R24]*)*, was present in the transcriptome of *V. pourtalesii* (TRINITY_DN38500_c0_g1_i1) but not up-regulated by DSi enrichment, a response that does not support a direct involvement in DSi transport in this hexactinellid sponge. DISCUSSION ========== Hypothesis of action mechanism for DSi transport ------------------------------------------------ The results of this study suggest that molecular DSi transport in sponges appears to function through cooperation between a passive Si inflow (mediated essentially by aquaglyceroporins 3/9) and a coupled active Si efflux (mediated by the arsB transporters). Cooperation between active and passive pathways for DSi transport---but based on different transporters---has recently been found during silicification in plants ([@R41]) and mammals ([@R40]). Aquaporins are bidirectional passive channels and, therefore, are unable to act uphill against a solute gradient. They would only be able to mediate a monodirectional Si influx effectively into the sponge cells if the DSi concentration in the seawater is high enough to build initially a steep gradient and the passive Si inflow is subsequently coupled with the active Si efflux of the arsB/Lsi2-like transporter to transport the incoming Si out of the cytoplasm. This new destination should be the mesohyl for either intercellular silicification or subsequent transport into the silicifying cells (sclerocytes) to accomplish silicification within the silica deposition vesicle of the sclerocytes ([Fig. 6](#F6){ref-type="fig"}). Logistic constraints inherent to collecting and working with living deep-sea sponges prevent further experimental work on *V. pourtalesii* in the near future to empirically resolve the exact location of the DSi membrane transporters, which is tentatively proposed in [Fig. 6](#F6){ref-type="fig"}. Nevertheless, the information gained by this study will considerably ease future attempts to locate the DSi transporters in shallow-water demosponges, which are more accessible for experimentation. The participation of aquaporin channels explains why the consumption of DSi by sponges happens efficiently only at very high DSi concentrations: because high concentrations are needed to build initially a steep DSi concentration gradient between the extracellular and the intracellular environment. The coupling of the gradient-facilitated DSi inflow and the active Si efflux maintains the gradient steepness as the arsB transporters continuously expel DSi from the cell cytoplasm to somewhere else, either to the mesohyl or to the silica deposition vesicle ([Fig. 6](#F6){ref-type="fig"}), depending on whether the DSi is required for intracellular or intercellular silicification. The saturation of this active arsB transporter is the most likely reason why the DSi consumption saturates at high DSi concentrations, despite those concentrations being better for passive aquaporins working more efficiently. The absence of SITs (i.e., active adenosine triphosphate--consuming transporters) in favor of passive aquaporin channels to mediate the Si influx into sponge cells and the deposition vesicle is likely the reason for the low efficiency of sponges when processing DSi at low concentrations, compared to diatoms, which do have SITs. {#F6} Physiological consequences -------------------------- Our physiological results suggest that modern hexactinellids may be even less adapted than their demosponge counterparts to the relatively low DSi concentrations that characterize the modern diatom-dominated photic ocean ([Fig. 3](#F3){ref-type="fig"}). In contrast, at DSi concentrations of \<10 μM, diatoms are known to reach transport rates ([@R42]--[@R44]) that are two to three orders of magnitude those of any sponge investigated in this regard ([@R10], [@R15]), clearly favored by their active DSi transport system based on SITs. Therefore, competitive exclusion of the hexactinellids by the better performance of diatoms and demosponges in DSi uptake is likely the main reason why hexactinellids predominate in deep-sea environments, where diatoms cannot grow because of light limitation and where DSi concentrations remain slightly higher than in the photic zone ([@R45]). This scenario suggests that the impressive BSi skeletons that characterize most hexactinellid sponges can only be built at slow rates (i.e., over long time periods) and when encountering DSi concentrations higher than those characterizing the upper modern ocean, which are typically less than 10 μM ([@R17]). The hexactinellid *V. pourtalesii* is the only known sponge investigated to date with DSi consumption kinetics that depart from the typical Michaelis-Menten model. Further research on other hexactinellids will be needed to elucidate whether this difference consistently applies to the level of phylogenetic class. In our experimental approach to the kinetic model, we selected for the minimum number of DSi concentration steps that would identify a kinetic model with statistical significance. The reason for such tight design relates to the well-known difficulties of maintaining live deep-sea sponges under laboratory conditions for long periods. Therefore, we decided not to risk the success of the experiment unnecessarily by avoiding intermediated DSi concentrations that would have extended the duration of the assay without adding significant resolution to the essence of the outcome. Fortunately, casualties did not occur at any time during the experiment, with all sponges remaining healthy, as also reflected in the molecular message of the transcriptomes. Molecular responses ------------------- Earlier work has demonstrated that exposure of a Mediterranean demosponge species in the laboratory to DSi concentrations much higher (30 and 100 µM) than those occurring in its natural habitat (1 µM), yields production of some types of skeletal pieces that do not occur in wild populations ([@R19]). In addition to revealing that skeletal production in natural conditions was chronically limited by DSi, the finding also suggested that increasing DSi up-regulates genes involved in its utilization that are not expressed at low DSi. On this basis and in the absence of further experimental progress since, we designed our experiment and focused our analysis on those genes that were overexpressed as the result of increased DSi availability. The fact that the optimal DSi utilization rate in *V. pourtalesii* was herein demonstrated to be attained at DSi concentrations between 100 and 150 μM ([Fig. 2](#F2){ref-type="fig"}), which are not naturally available to the sponges (table S4), further encouraged the expectations of our approximation. The opposite response can be obtained in diatoms, the main Si competitors of sponges, which reach optimum DSi transport at relatively low concentrations and may down-regulate their active DSi transporters (SITs) when exposed to abnormally high DSi concentrations ([@R46]). As a result of our DSi enrichment and subsequent RNA sequencing analysis, many genes related to the GO categories of vesicle- and membrane-mediated transport and lysosome transport were overexpressed, including several *lysosome-related*, *solute carriers*, *sorting nexin-1*, *transport Sec61 subunit gamma*, *BPC intracellular cholesterol transporter 2*, *cystinosin*, *magnesium transporter NIPA2-like*, *sodium bile acid cotransporter 7 isoform 2*, *Ras-related Rab genes*, *vacuolar-sorting protein genes*, etc. However, there is no evidence to date that any of those genes are directly involved in Si utilization. Therefore, their study and functionalization fall beyond the scope of the present work. Alternatively, we obtained up-regulation of both active and passive membrane transporters that had been related to DSi transport in previous studies on organisms other than sponges. Regarding silicifying proteins, it was interesting that the *glassin* genes ([@R22]), which code for the protein that catalyzes silica polycondensation in hexactinellid sponges, were not particularly overexpressed. This finding is plausible for several reasons. Experiments on freshwater demosponges showed that enzymatic axial filaments were automatically produced by the sponges even when no DSi was available in the environment to undertake spicule silicification ([@R47]). Likewise, high experimental DSi concentrations increased the length and thickness of spicules in a marine demosponge, but the total number of spicules (which is determined by the produced number of silicatein axial filaments) did not increase substantially ([@R19]). These two previous studies on demosponges agree with our results and support the view that expression of silicifying enzymes (either silicatein or glassin), unlike that of the membrane transporters (*aquaglyceroporin 3/9* and *ArsB*), may not be controlled directly by environmental DSi concentrations but is subject to more complex levels of regulation. Regarding the passive DSi transporters, we found that *aquaglyceroporins 3* and *9* were only slightly up-regulated in DSi-enriched individuals, while *aquaglyceroporin 3/9* was significantly overexpressed ([Fig. 4B](#F4){ref-type="fig"}). Previous evidence already has indicated that *aquaglyceroporins 3* and *9* (and also *7*) are involved in Si transport in vertebrates ([@R35]). The fact that homologous genes were up-regulated in the DSi-enriched individuals of *V. pourtalesii* provides strong additional evidence for their involvement in transmembrane Si transport. Regarding the active DSi transporters, the overexpression of *arsB 1 and arsB 2* genes in the DSi-enriched individuals strongly supports their involvement in DSi transport in sponges. Such a function had already been demonstrated for other organisms, mostly land plants ([@R41], [@R48]), but it had also tentatively been suggested---based on sequences similarity---for sponges ([@R23]). In rice, the same protein (but named Lsi2) has the dual capability of transporting both arsenic and Si ([@R41], [@R48]), probably due to molecular mimicry between these two metalloids and also others, such as boron, germanium, arsenic, antimony, and tellurium ([@R33]). Likewise, *arsB* homologs in the diatom *Thalassiosira pseudonana* show coexpression with a silicification-related gene ([@R46]), suggesting that the arsB transporters, initially related to arsenic transport, can also expel actively Si across cell and vesicle membranes. ArsB/Lsi2 transporters are sister of the transporter family including pink-eyed dilution (PED) proteins ([Fig. 5B](#F5){ref-type="fig"}), which mediate transport and processing of tyrosinase and other melanosomal proteins in melanocytes and other cells in mammals but can also transport sucrose ([@R49]). Evolutionary implications for biosilicification ----------------------------------------------- Collectively, evaluation of our results in the context of the available information supports the view that while Si membrane transporters appear to be shared by the two major siliceous lineages of sponges (Hexactinellida and Demospongiae), each lineage has evolved independently its own silicifying enzymes (glassin versus silicatein). This scenario also suggests that the mechanisms for Si transport across cell membranes in sponges predate the acquisition of the mechanisms for polymerizing DSi into BSi, a pattern that appears to apply to other silicifying organisms as well. The available evidence that aquaglyceroporins 3 and 9 and arsB/Lsi2 proteins can be used for transporting, in addition to Si, other elements fundamental for survival (e.g., glycerol, several metalloids, sucrose, seawater, etc.) decreases notably the chances that those channels can be modified through evolution to improve Si transport without fatally affecting their functionality for the other elements. This is probably the reason why the DSi consumption system of siliceous sponges remained unchanged through the global DSi decrease that started at least some 100 to 65 Ma ago with the expansion of diatoms ([@R19], [@R28]) and still persists maladapted. Yet, we have found evidence of duplication and possible subfunctionalization of aquaglyceroporins in *V. pourtalesii*. *Aquaglyceroporins 3* and *9* were only moderately up-regulated in response to the DSi treatment, while *aquaglyceroporin 3/9* was significantly overexpressed ([Fig. 4](#F4){ref-type="fig"}). This last aquaglyceroporin could be the result of duplication and further subfunctionalization to more efficiently transport Si, while the two others remain more generalist passive transporters. It remains unknown at what level aquaglyceroporin 3/9 may still be involved in the flux of other essential metalloids. Initially, a disconcerting feature was the presence of *aquaporin 3*--, *aquaporin 9*--, and *aquaporin 3/9*--like genes in transcriptomes of *D. antarctica* and *I. fasciculata* ([Fig. 5A](#F5){ref-type="fig"} and fig. S4), which are nonsilicifying demosponges characterized by protein (i.e., spongin) rather than silica skeletons. The presence of those DSi passive transporters, in addition to the absence of silicatein genes in these species, has two potential explanations: (i) Passive aquaporin channels for Si, which are also used for other fundamental metalloids, were in place before the evolution of the silicifying enzymes, a hypothesis also supported by the independent acquisition of silicateins and glassin by demosponges and hexactinellids, respectively; (ii) silicatein and aquaporin channels for Si evolved concomitantly, but silicatein would have been lost secondarily and independently in various nonsilicifying members of the several demosponge lineages, while aquaporins were retained because those pore channels are also used for elements other than Si. This second option, because of invoking multiple independent loses, is less likely. Whether the silica skeleton was the primitive skeletal condition for sponges, which was subsequently lost homoplasically in several demosponge lineages, is a hypothesis that can be definitely resolved only by direct genomic evidence of the presence or absence of genes rather than only from the evidence of gene expression captured by transcriptomes. However, embryological evidence ([@R50]) supports the hypothesis that silica skeletons were replaced in at least some demosponge lineages through parallel evolution in favor of alternative skeletal materials (spongin, collagen, etc.). This skeletal evolution could have been forced by the drastic decrease in DSi availability triggered by the evolutionary expansion and proliferation of diatoms about 100 to 65 Ma ago, which would have operated as a negative selective pressure on siliceous skeletons ([@R10], [@R19]). Molecular clocks could attempt to date whether the emergence of spongin skeletons (in the Verongimorpha and Keratosa subclasses) was coincidental with the expansion of diatoms or any other past event ([@R51]) that could have caused a drastic decrease in the availability of DSi in the upper ocean. Nevertheless, such studies have not been conducted to date. The most unexpected aspect in the phylogenetic distribution of DSi transporters across lineages of silicifying organisms is perhaps the absence of SITs in siliceous sponges, because choanoflagellates, which do have SITs ([@R23]), share a common ancestor with sponges. Likewise, SIT-L genes, which appear to be ancestral and have originated SIT genes of other silicifiers by duplication, inversion, and fusion of subunits ([@R52]), occur not only in several groups of silicifiers but also in some groups that are not essentially silicifiers (but rather calcifiers), such as foraminifera, coccolitophorid haptophytes, and some nonsponge metazoans. Why sponges secondarily lost the ancestral SIT complement and are now forced to transport Si through a less efficient aquaglyceroporin system remains unclear. We suggest that passive channeling is energetically less costly and could facilitate silicification even during periods of starvation or limited food supply. In a paleo-ocean with very high DSi concentrations ([@R53]) and moderate food supply, the replacement of active transporters by passive channeling could even be advantageous. However, when the rising activity of biosilicifiers began decreasing the environmental DSi concentration in the global paleo-ocean during the Late Mesozoic or even earlier, the lack of specialized active uptake mechanisms in favor of passive channeling became detrimental to many lineages of siliceous sponges, causing extinctions, impelling bathymetric migrations to the aphotic ocean, and forcing the skeletal evolution of siliceous sponges toward other materials to reduce their dependence on Si availability ([@R19], [@R50]). It is also remarkable that the *aquaglyceroporin* genes used for passive Si transport in sponges have been conserved through evolution across lineages of animals that never were---or have never been found to be---silicifiers, arriving still functional to the genome of vertebrates, where they facilitate a silicification step that has become incorporated into the general calcifying process of bone formation ([@R35]). This now explains why a diet rich in Si is known to favor correct bone formation in vertebrates ([@R54], [@R55]) and why treatments based on Si also improve bone regeneration ([@R56]). Several lineages of demosponges contain relict species and/or genera (collectively alluded as "sclerosponges" or "coralline sponges") in which a basal skeleton of calcium carbonate coexists with isolated siliceous spicules ([@R57], [@R58]), recalling the association between calcifying and silicifying processes recently found not only in vertebrate bone but also in cyanobacteria ([@R59]), coccolitophorid haptophytes ([@R52]), and crustaceans ([@R59], [@R60]). Our discovery that the role in Si transport of aquaporins 3 and 9 is conserved from the Porifera to the Vertebrata gives rise to the possibility that biosilicification and biocalcification are not alternative biomineralization systems but instead mechanisms that have been functionally intertwined since the early stages of animal evolution, and perhaps even earlier. MATERIALS AND METHODS ===================== Species habitat, sampling, and data collection ---------------------------------------------- *V. pourtalesii* is found at depths from 100 to 935 m along the continental margin of North America from Florida (United States) to Nova Scotia (Canada), where it forms extensive, dense aggregations in the deep basins and channels that excise the continental shelf of Nova Scotia, Canada ([@R8], [@R25]). The aggregations are considered monospecific, consisting of an abundance of vase-shaped individuals ([Fig. 1A](#F1){ref-type="fig"}) measuring up to 40 cm in height. Some areas of the Scotian Shelf where *V. pourtalesii* occurs more densely aggregated---i.e., forming "*Vazella* grounds"---have recently been closed (Conservation Areas; fig. S1) for protection from bottom fishing ([@R25]). During an oceanographic mission from 2 to 7 September 2017 to collect data from the *Vazella* grounds for the European Union--funded SponGES project, the ROV "ROPOS" was deployed from the Canadian Coast Guard Ship *Martha L Black* in the *Vazella* grounds located in the Sambro Bank Sponge Conservation Area and north of nearby LaHave Basin. ROPOS is a 40-hp Science/Work Class ROV owned and operated by the nonprofit Canadian Scientific Submersible Facility (CSSF), based in North Saanich, B.C., Canada. During deployment, forward- and downward-facing video footage of the seabed was recorded to determine the fine-scale distribution and densities of *V. pourtalesii*, and using the manipulator arms of ROPOS, live sponge individuals were collected for DSi consumption laboratory experiments and for conducting in situ incubations using benthic chambers, as described below. Ex situ incubations for kinetics of DSi consumption --------------------------------------------------- For investigating the kinetics of DSi consumption ex situ, a total of 11 sponges were collected from the Sambro Bank Sponge Conservation Area and LaHave Basin (fig. S1). The depth range of collected specimens was \~160 m in the Sambro Bank closure to 185 m north of LaHave Basin. Each sponge was collected along with the small rock on which it was attached, so that the manipulator arm of the ROV never handled the sponge tissue but just the rock (movie S1). Once on board, sponges, which were at no time exposed to air during transportation and further experimental work, were maintained for 5 days in an insulated 750-liter polyethylene holding tank inside a refrigerated container. The seawater, which was also refrigerated to 6° ± 1°C, was recirculated continuously and exchanged every 8 hours. Surface water (\<5 m) was pumped using a portable pump into a tank on deck. From there, it was distributed to holding tanks inside the refrigerated container, chilled using a portable 1/3-hp chiller, and subsequently slowly pumped via a peristaltic pump (1.2 liter/min) to the tank containing the sponge specimens. The chilled water tank was refilled twice a day, resulting in four water exchanges/day in the sponge holding tank. The sponges, which were attached to a small rock in all cases, were maintained on the bottom of the holding tank by placing them within individual compartments of a polyviniyl chloride (PVC) grid. Upon return to the Bedford Institute of Oceanography (BIO) in Dartmouth, Nova Scotia, the sponges were transferred to a 500-liter aquarium for 24 hours. The saltwater intake at BIO is 200 m from shore and at a depth of 17 m (\~3 m off bottom). The seawater was passed through a sand filter and then a 20-μm (nominal) bag filter before being delivered to the lab. The filtered seawater entered a 1000-liter aerated tank (header tank) where it was heated/chilled to \~6° to 9°C. This water was then gravity-fed into the sponge holding tank (500 liters of insulated polyethylene). Flow rates were maintained at \~3 to 5 liter/min. A small magnetic drive pump was added to the bottom of the holding tank to provide circulation and horizontal flow across the sponges. The pump was modified with a 12.5 mm--by--300 mm vertical pipe with 6-mm holes added to provide horizontal flow. Upon initiation of the ex situ experimentation, sponges attached to their respective rocky substratum were transferred to a 360-liter tank (hereafter referred to as the "preconditioning tank") and left there for 24 hours for acclimation to a refrigerated (9° ± 0.5°C) seawater system with recirculation. To characterize the kinetic pattern, DSi consumption by the sponges under increasing DSi availability was measured. The experiment ended when saturation was reached by the sponges, that is, when an increase in DSi availability did not stimulate any further increase in the rate of DSi consumption. Seven levels of DSi concentration were progressively offered to the 11 assayed sponges: 12 (approximate field values), 30, 60, 100, 150, 200, and 250 μM Si. Under each concentration, sponges were incubated separately in a polypropylene 16-liter container (hereafter referred to as the "incubation aquaria") for 24 hours. Before each 24-hour incubation, sponges were maintained in the preconditioning tank for 24 hours. Thus, the approach consisted of an alternation of 24-hour "preconditioning" and "incubation" periods for 2 weeks (from 9 to 22 September 2017). The alternation of preconditioning and incubating steps had three main purposes. The first was to facilitate the survival of the sponges across the battery of incubations in the relatively small (16 liters) incubating aquaria, an objective successfully met, as there were no casualties over the course of the 2-week experiment. Second, to render the approach conservative, during each preconditioning step, the sponges were exposed for 24 hours to the same DSi concentration that was to be assayed in the following incubation. Therefore, during the preconditioning step, the sponges were able to take up as much DSi as needed to satisfy their chronic avidity for DSi, an approach that has been empirically demonstrated to lead to slightly lower DSi consumption rates during the following 24-hour incubation period ([@R15]). Third, the sum of the duration of the preconditioning period and the incubation period collectively provided a total time of exposure to each DSi concentration long enough (48 hours) to allow the sponges to unfold a complete physiological response in the silicification process. Theoretically, such a response is expected to involve the generation of new populations of silicifying cells to deal with the increasing availability of DSi, the activation of new sets of genes, and the completion of massive production of silicifying proteins ([@R10], [@R19]). Along with the 11 assayed sponges, we used three control aquaria, each containing seawater and a rock as the one used by the sponges for attachment, but with no sponge. These controls served to correct for potential processes of either DSi release from the rocks or DSi precipitation at the rock surface. It was also difficult to obtain exactly the intended DSi concentrations in the large preconditioning tank because the transfer of sponges to it upon conclusion of each incubation involved an unavoidable transference of seawater at a lower DSi concentration, causing minor dilution of the concentration in the preconditioning tank. This logistical constraint resulted in the following concentrations during the incubations: 12.5, 30.4, 59.0, 93.0, 141.6, 191.8, and 234.0 μM Si. The seawater used for the experiments was pumped in from the Bedford Basin, Nova Scotia (at 7°C) and filtered on a 1-μm mesh, a pore size small enough to prevent the passage of planktonic DSi users (e.g., diatoms, radiolarians, etc.) but allowing, in part, the passage of natural sponge food, that is, the smallest bacterioplankton. In addition, we fed the sponges during the entire experiment by adding 35 ml of a DSi-free, concentrated culture (approximately 10^6^ cells/ml) of the haptophyte *Isochrysis galbana* to the 360-liter tank at the beginning of each preconditioning step. We assumed that, after passing repetitively throughout the mechanical water pump, part of the microphytoplankton cells would be lysed, resulting in a mix of particulate and dissolved organic matter available to the sponges. The assayed DSi concentrations were prepared by adding the corresponding volume of a buffered 0.1 M sodium metasilicate solution \[Na~2~SiO~3~ (pH 10)\] to the 360 liters of filtered seawater contained in the preconditioning tank, followed by mixing of the water for 18 hours with a submersible pump to ensure complete molecular diffusion before transferring the sponges to the tank for their corresponding preconditioning period. To determine the rate of DSi utilization by the assayed sponges during the incubations at each DSi concentration step, a 50-ml water sample was collected at the beginning and end of each 24-hour incubation period. Seawater samples, collected using acid-cleaned plastic syringes, were immediately filtered through 0.22-μm pore, polycarbonate syringe filters (Millex-GS Millipore) and stored in the fridge no longer than 2 days until analysis. Samples from the same DSi concentration step were analyzed together into a single analysis using a Technicon AutoAnalyzer 3 (AA3, SEAL Analytical), a service provided by the CERC.OCEAN research group based at the Dalhousie University (Halifax, Nova Scotia, Canada). Analyses were run in triplicate following the standard colorimetric method, with a determination accuracy (as percent error) of \<5%. Samples with a DSi concentration higher than 60 μM were diluted before analysis with artificial seawater prepared with the same salinity as the water samples (35 practical salinity unit). The rate of DSi utilization by a sponge at a given DSi availability was inferred by calculating the difference in DSi concentration between the start and end of an incubation and after correcting by the average concentration change (often negligible) that occurred in the set of control aquaria. At the end of the experiment, we measured the volume (ml) of both the assayed individuals and their rock substratum by water displacement. Sponges were subsequently wet-weighed (g), dried at 60°C to a constant dry weight (g), and combusted at 540°C for 10 hours for ash-free dry weight (AFDW; g). Rates of DSi consumption were normalized by sponge volume (ml) and/or AFDW (g), volume of seawater in the incubating aquaria (liters) after discounting sponge and rock volume, and duration of the incubation (hours). We preferentially expressed data normalized to sponge volume because it facilitates their future applicability to field sponge populations using ROV images without the need of collecting individuals. However, for correct physiological between-species comparison, we used normalization by AFDW. The relationship between normalized DSi consumption rates (in μmol Si ml^−1^ hour^−1^or g^−1^) and DSi availability (μM) was analyzed by nonlinear regression to identify the best-fitting model for the empirical observations. In situ incubations for DSi consumption --------------------------------------- We built five benthic incubation chambers using methylmethacrilate and inox steel, with an incubation volume of either 17.3 or 13.3 liters ([Fig. 1, B to D](#F1){ref-type="fig"}). Chambers incorporated a floor piece of Delrin acetal resin ([Fig. 1B](#F1){ref-type="fig"}), which allowed for the incubation of sponges in isolation from the external environment, thus avoiding interference by nutrient fluxes from sediments that may be resuspended during deployment. The chambers incorporated two external sampling bottles (120 ml) made of steel and internally folded with polytetrafluoroethylene. Through a steel capillary (20 cm long and 0.6 mm wide) that pierced the wall of the chamber, each bottle was designed to collect a water sample (under negative pressure conditions) from inside the incubation chamber while it was opened for 5 min and then closed using the ROV manipulator arms (movies S1 and S2). During the 2017 oceanographic mission on the Canadian Coast Guard Ship *Martha L Black*, four sponges and a control treatment were incubated in situ. Sponges were collected along with the small rock to which they were attached using the manipulator arms of the ROV (movie S1). Each sponge was then placed on the floor piece and covered with the chamber so that the chamber rested into a groove on the floor piece designed to seal the unit from leakages. Once the sponge was inside the chamber and the chamber was properly sealed, one of the sampling bottles was opened to collect water for 5 min and then closed again to avoid water exchange with the surrounding medium (movie S2). As a control, we incubated a rock selected from the sponge grounds but without an attached sponge. After an incubation period of 19 to 28 hours (incubation time varied because of weather and the logistics of the cruise), the ROV returned to the chambers and triggered the second sampling bottle. After this second water collection, the sponge and its attachment substrate were collected to estimate volume and biomass and to normalize DSi consumption rate, as indicated in the above section of "Ex situ incubations for kinetics of DSi consumption." Seawater samples were processed for determination of the initial and final DSi concentrations and deriving consumption rate as described in the above section. Individual BSi production in the natural habitat ------------------------------------------------ We aimed to estimate how much BSi---mass of siliceous skeleton---was produced by the sponges in their natural habitat per unit time to serve as a comparison for the predictions of DSi consumption from the laboratory-based kinetic model. Preliminary field work revealed that *V. pourtalesii* biofouled acoustic mooring arrays deployed by the Ocean Tracking Network (OTN; Dalhousie University) from Halifax to the shelf break on the Scotian Shelf ([Fig. 1, E and F](#F1){ref-type="fig"}). During routine servicing of these arrays, *V. pourtalesii* individuals were recovered from two moorings that had been immersed for 15 and 58 months, and rates of sponge growth and BSi production during those two periods were estimated. During an additional recovery made during a 2018 oceanographic mission, several substrata offered for *V. pourtalesii* settlement about \~1 year before brought out no sponge recruits (SponGES Consortium, unpublished information). It suggested that larval release is likely to occur in July to August. Therefore, the two largest sponges on the mooring immersed for 15 months were estimated to be 14 months old, and the three largest sponges on the 58-month-old mooring were estimated to be 54 months old. After determining their volume, sponges were wet-weighed (g), dried at 60°C to a constant dry weight (g), and combusted at 540°C for 10 hours for AFDW (g). The BSi content was estimated as 95% of the ash weight. However, for comparative purposes, the BSi content of some of the larger sponges was also estimated through the loss of weight before and after desilicification of the sample in 5% hydrofluoric acid. Gene expression and phylogenetic analyses ----------------------------------------- We collected tissue samples from six control individuals collected from Emerald Basin and from six DSi-enriched individuals (\#3, \#4, \#5, \#7, \#9, and \#10) used in the kinetic experiment (tables S2 and S3). Approximately 5 cm^3^ of tissue sample was collected immediately upon termination of the kinetic experiment, when concentrations were 250 μM DSi, and before any further processing of the sponges for morphometric studies. Samples were preserved in RNAlater (Ambion) at 4°C for 24 hours and then stored at −80°C until further processing. Total RNA from each sample was extracted using a TRIzol (Thermo Fisher Scientific, UK) standard extraction followed by a polyA selection of mRNA using a Dynabeads Direct mRNA purification kit (Thermo Fisher Scientific, UK) according to the manufacturer's protocols. These were used to produce an RNA library for next-generation sequencing using the ScriptSeq library prep kit v2 provided by Illumina (CA, USA). Sequencing was performed on a single run of the Illumina NextSeq 500 platform by the Natural History Museum's (London, UK) Sequencing Unit at 2 × 150 bp read length. A total of 202,315,355 reads were sequenced and remained after the removal of adaptor sequence and initial quality screening. We visualized the quality across the sequencing reads using FastQC (Babraham Bioinformatics) and performed additional trimming with Trimmomatic ([@R61]) to remove areas of sequence with low Phred scores and any residual sequences shorter than 36 bp (settings: ILLUMINACLIP:/ ScriptSeq_adapters.fa:2:30:10 LEADING:3 TRAILING:3 HEADCROP:8 SLIDINGWINDOW:4:15 MINLEN:36, where ScriptSeq_adapters.fa contained the sequences of the adaptors used in sequencing). The remaining total of 173,280,196 paired reads (unpaired reads were not retained) were then used to construct a transcriptome using Trinity 2.4.0 ([@R62]) with default options. Raw reads can be accessioned at the Short Read Archive (SRA) under BioProject number PRJNA580361. Annotation and gene expression analyses --------------------------------------- We obtained the annotations for our de novo--assembled transcriptome using "BlastX" command in DIAMOND ([@R63]) against two different databases: RefSeq and Swiss-Prot (last accessed in August to September 2018), retaining only the best hit with an *e*-value threshold of 10^−5^ in both cases. Then, we used Blast2GO ([@R64]) to obtain the GO terms associated with the blast hits obtained against Swiss-Prot for Biological Process, Molecular Function, and Cellular Component, with the GOSlim function. Completeness of the transcriptome was assessed by searching for single-copy orthologs in both eukaryotic and metazoan databases using BUSCO ([@R65]). For the gene expression analysis, we used standard mapping strategies Bowtie2 and RSEM (RNA-Seq by Expectation Maximization) as implemented in Trinity to collect the number of reads aligning with each of our genes in the reference transcriptome and then used the raw count reads against each gene to perform the differential gene expression analysis using edgeR ([@R66]) as implemented in Trinity. We only retained the genes with a corrected *P* value of false discovery rate (FDR) of 0.001 and at least fourfold expression (-P 1e-3 -C 2). The DE genes were annotated using the gene IDs from Swiss-Prot and also the GO terms (see table S2). To visualize the GO categories enriched in our DE genes, we used REVIGO ([@R67]) and plotted circos plots in the R statistical software program. The expression levels of target genes involved in silica production for spicule building in hexactinellids (polymerization and biomineralization: *glassin* and *cathepsins*; ion transporters: *transporters arsB*, *aquaporins 3* and *9*, *solute carriers*, *magnesium transporters NIPA*, *ammonia transporters*, *sodium-potassium calcium exchanger*, and *sodium bile acid cotransporter*) were collected from the normalized expression values using trimmed mean of *M*-value (TMM) matrix obtained with edgeR and plotted using pheatmap in R. Phylogenetic analyses --------------------- In our phylogenetic analysis of aquaporins and arsB transporters in *V. pourtalesii*, we included sequences of all major groups of aquaporins and arsenite-antimonite efflux protein sequences collected from GenBank and our transcriptomic databases for sponges (see accession numbers in [Fig. 5](#F5){ref-type="fig"}, fig. S4, and data file S1). Phylogenetic analyses of the genes coding for the passive and active transporters were conducted separately, with ingroup and outgroup sequences selected on previous knowledge of the protein families---aquaporins ([@R34]) and arsB/Lsi2 transporters ([@R49]). All sequences were aligned with MAFFT v5 ([@R68]), and the phylogenetic trees were built with RAxML 8.1.22 ([@R69]) with Le-Gascuel as the protein model of substitution and GAMMAI correction for rate variation among sites as obtained from PROTTEST ([@R70]) using the Akaike information criterion. Reliability of the phylogenetic trees was estimated using 100 bootstrap replicates. As a congruence test for the previous ML-based phylogenies, a subsequent Bayesian phylogenetic analysis was conducted for both protein families, using MrBayes ([@R71]) v3.2.2 x64 with the model provided by PROTTEST. The Monte Carlo Markov Chain search was run over at least 20,000,000 generations. Trees were sampled every 2500 generations, and the first 25% of trees gathered were discarded as "burn-in." Convergence was checked with Tracer 1.7 ([@R72]). Supplementary Material ====================== ###### aba9322_Movie_S2.avi ###### aba9322_SM.pdf ###### aba9322_Data_file_S1.xlsx ###### aba9322_Movie_S1.avi We thank B. MacDonald for help with logistics during sponge collection and maintenance in experimental conditions, C. Sitjà for help with sponge dry and ash weights, and M. García-Puig for video editing; F. Whoriskey and J. Pratt of the (OTN, Dalhousie University) for the collection of specimens from the OTN moorings; G. Yahel (Ruppin Academic Center) for advice when building the seawater collectors of the incubation chambers. **Funding:** This research was completed mostly by funds from the SponGES H2020 grant (BG-01-2015.2, agreement number 679849-2) to M.M. and A.R. and from Fisheries and Oceans Canada Strategic Program for Ecosystem-Based Research and Advice (SPERA) and International Governance Strategy (IGS) projects awarded to L.B. and E.K. This study also benefitted from funding by a PBS grant (MINECO CTM2015-67221-R) to M.M. This study is in memory of Hans Tore Rapp, who passed away on 7 March 2020, and who was the main coordinator of the H2020 SponGES project that has made this research possible. **Author contributions:** M.M. designed the study. The physiological experiments and nutrient analyses were performed and analyzed by M.M. and M.L.-A. E.K. and L.B. dealt with mapping, collection of organisms, and the logistics of cruise organization and laboratory gearing for in vivo experimentation with deep-sea sponges and nutrient analysis. A.R. and V.K. conducted the transcriptomic analysis, the analysis of differential gene expression, and the phylogenetic analyses, being the molecular data interpreted by M.M. and A.R. M.M. assembled the first draft of manuscript, which was further refined through invaluable contributions by all authors. **Competing interests:** The authors declare that they have no competing interests. **Data and materials availability:** All data and access numbers needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Raw transcript reads can be accessioned at the SRA under BioProject number PRJNA580361. Additional data related to this paper may be requested from the authors. Supplementary material for this article is available at <http://advances.sciencemag.org/cgi/content/full/6/28/eaba9322/DC1> [View/request a protocol for this paper from *Bio-protocol*](https://en.bio-protocol.org/cjrap.aspx?eid=10.1126/sciadv.aba9322).

When Landauer argued in 1961 that any physical realisation of erasure of information has a fundamental thermodynamic work cost he irrevocably linked thermodynamics and information theory[@b1][@b2][@b3][@b4][@b5][@b6][@b7][@b8][@b9]. A practical consequence of this insight is that all computers must dissipate a minimal amount of heat in each irreversible computing step, a threshold that is becoming a concern with future computer chips entering atomic scales. The treatment of general *quantum* information processing tasks within the wider framework of quantum thermodynamics has only recently begun[@b13]. Quantum mechanics differs from classical mechanics in at least three central aspects: the special nature of measurement, the possibility of a quantum system to be in a superposition and the existence of quantum correlations. The thermodynamic energy needed to perform a (selective) measurement has been investigated[@b10] and the total work for a closed thermodynamic measurement cycle explored[@b11]. The catalytic role of quantum superposition states when used in thermal operations has been uncovered[@b12] and it has been shown that work can be drawn from quantum correlations[@b13][@b14] in a thermodynamic setting, see [Fig. 1](#f1){ref-type="fig"}. In particular, del Rio *et al.*[@b14] showed that contrary to Landauer's principle, it is possible to *extract* work while performing erasure of a system's state when the system is correlated to a memory. This can occur if and only if the initial correlations imply a negative conditional entropy, a uniquely quantum feature. The thermodynamic process does however now require operation on degrees of freedom external to the system, i.e. the memory's. Results ======= Projections and the optimal work value of removing coherences ------------------------------------------------------------- Our motivation is here to shed light on the implications of performing a measurement on a quantum state that has coherences. We will consider this task in the thermodynamic setting of Landauer's erasure, involving a heat bath at fixed temperature *T* and operation on *N* → ∞ uncorrelated and identically prepared copies of the system (i.i.d. limit). This is of interest in the context of the quantum Jarzynski equality, for example, and will also be central for experiments testing quantum thermodynamic predictions in the future. To tackle this question we define the information-theoretic "projection" for a given initial quantum state *ρ* and a complete set of mutually orthogonal projectors . Such state transformation can be seen as analogous to the state transfer of erasure, , to a blank state . Physically, this projection can be interpreted as the result of an unread, or unselective[@b15], measurement of an observable that has eigenvector projectors . In an unselective measurement the individual measurement outcomes are not recorded and only the statistics of outcomes is known. In the literature the implementation of unselective measurements is often not specified, although it is typically thought of as measuring individual outcomes, e.g. with a Stern-Gerlach experiment, see [Fig. 2a](#f2){ref-type="fig"}, followed by mixing. The crux is that the information-theoretic projection can be implemented in many physical ways. The associated thermodynamic heat and work will differ depending on *how* the projection was done and we will refer to the various realisations as "thermodynamic projection processes". One possibility is decohering[@b16] the state in the so-called pointer basis, , a thermodynamic process where an environment removes coherences in an uncontrolled manner resulting in no associated work. In general it is possible to implement the state transfer in a finely controlled fashion achieving optimal thermodynamic heat and work values. Of particular importance in thermodynamics is the projection of the system's initial state *ρ* onto the set of energy eigenstates of the system's Hamiltonian with *E*~*k*~ the energy eigenvalues. Here the state's off-diagonals with respect to the energy eigenbasis are removed - a state transformation that is frequently employed in quantum thermodynamic derivations and referred to as "dephasing" or "measuring the energy". Our key observation is that there exists a thermodynamic projection process realising this transformation and allowing to draw from the quantum system a non-trivial *optimal average work* of Here *T* is the temperature of the heat bath with which the system is allowed to interact, see illustration [Fig. 1](#f1){ref-type="fig"}, *k*~*B*~ is the Boltzmann constant and *S* is the von Neumann entropy. Crucially, this work is strictly positive for quantum states with coherences. Extending the key observation to general projections one finds that optimal thermodynamic projection processes can be implemented that allow to draw an average work of where an additional internal energy change term appears. Physical interpretation and assumptions made to derive the optimal work ----------------------------------------------------------------------- The optimal work values stated in Eqs. [(1](#eq12){ref-type="disp-formula"}) and ([2](#eq14){ref-type="disp-formula"}) are valid for processes applied to classical and quantum states alike. While for a classical ensemble the entropy change, , will be zero this is not so in the general quantum situation, where initial non-diagonal quantum states result in a strictly positive entropy change[@b17]. We note that while the optimal work values are in principle attainable, practical implementations may be suboptimal resulting in a reduced work gain or a higher work cost. The physical meaning of can be grasped by considering a lower bound[@b18] on it, , see Supplement. Here *d* is the dimension of the system and denotes the Hilbert-Schmidt norm. The first factor quantifies the distance of the initial state from the fully mixed state, while the second factor, , quantifies the angle between the diagonal basis of *ρ* and the projection basis . These terms correspond to incoherent and coherent mixing contributions. The entropy change is non-trivially bounded only if the initial state is not an incoherent mixture with respect to that basis. The entropy bound is the largest for pure initial states whose basis is mutually unbiased with respect to . In this case the optimal entropy change is . One may wonder where the work has gone to. There are two equivalent approaches to the accounting of work. In the present analysis the focus is on the work that the system exchanges, as done in statistical physics[@b5][@b19][@b20][@b21][@b22]. In this approach it is often not explicitly mentioned where the work goes to, but the only place work can go to are the externally controlled energy sources. Similarly, the heat, i.e. the energy change minus the work, is established implicitly. For example, in the experimental realisation of classical Landauer erasure with a colloidal silica bead trapped in an optical tweezer[@b21], the dissipated heat of erasure was calculated by knowing the applied tilting forces and integrating over the bead's dynamics. The second approach is to collect work in a separate work storage system[@b23], as illustrated by the weight in [Fig. 1](#f1){ref-type="fig"} and detailed in the Supplement. Both the implicit and the explicit treatment of work are equivalent in the sense that the results obtained in one approach can be translated into the other. The thermodynamic assumptions made to prove Eq. [(2)](#eq14){ref-type="disp-formula"} are congruent with current literature[@b9][@b23][@b24][@b25]; specifically they are: (T0) an isolated system is a system that only exchanges work and not heat; (T1) the validity of the *first law* relating the internal energy change, Δ*U*, of the system during a process to its average heat absorbed and work drawn, ; (T2) the validity of the *second law* relating the system's entropy change to its average absorbed heat, , when interacting with a bath at temperature *T*, with equality attainable by an optimal process; (T3) the thermodynamic entropy to be equal to the von Neumann entropy in equilibrium as well as out-of-equilibrium, . In addition we make the following standard quantum mechanics assumptions: (Q0) an isolated system evolves unitarily; (Q1) control of a quantum system includes its coherences. Details of the proof are in the Methods Summary. We note that in the single-shot setting whole families of second laws apply[@b7][@b8] that differ from (T2) stated above. However, in the limit of infinitely many independent and identically prepared copies of the system these collapse to the standard second law, (T2), on the basis of which Eq. [(2)](#eq14){ref-type="disp-formula"} is derived. From the information-theory point of view the projections considered here constitute just one example of the larger class of trace-preserving completely positive (TPCP) maps characterising quantum dynamics. Of course, all TPCP maps can be interpreted thermodynamically with the assumptions stated above, resulting in an optimal average work given by a free energy difference. Erasure is another such map whose study forged the link between information theory and thermodynamics. The benefit of discussing "projections" here lies in the insight that this focus provides: it uncovers that coherences offer the potential to draw work making it a genuine and testable quantum thermodynamic feature. This work is non-trivial even when the thermodynamic process is operated on the system alone, not involving any side-information[@b14] stored in other degrees of freedom. Qubit example for drawing optimal work -------------------------------------- To gain a detailed understanding of thermodynamic projection processes that give the optimal work stated in Eq. [(1)](#eq12){ref-type="disp-formula"} we now detail one such process for the example of a spin-1/2 particle (qubit), see illustration in [Fig. 2b,c](#f2){ref-type="fig"}. This process consists of a unitary evolution, a quasi-static evolution and a quench[@b25], and it is optimal for any finite dimensional quantum system (proof in the Methods Summary). An experimentalist, Emmy, prepares the spin in a state ( w.l.o.g.) exposed to an external magnetic field which she controls. The Hamiltonian associated with the system is where the energy difference between the aligned ground state, , and anti-aligned excited state, , is given by with the spin's magnetic moment. Importantly, in general the spin state's basis, , are superpositions with respect to the energy eigenbasis, and with . For the optimal implementation of the projection Emmy now proceeds with the following three steps. Firstly, she isolates the spin from the bath and modifies external magnetic fields to induce a unitary rotation, , of the spin into the energy basis. In nuclear magnetic resonance (NMR)[@b26] and pulsed electron spin resonance (ESR) experiments[@b27] such rotations are routinely realised by radio-frequency and microwave pulses respectively, as evidenced by Rabi oscillations. The power, duration and phase of such a pulse would be chosen to generate the spin-rotation along the green circle until the desired unitary *V* is achieved. In the same step Emmy adjusts the strength of the external B-field such that the spin state is Boltzmann-distributed at temperature *T* with respect to the energy gap of the Hamiltonian at the end of the step, *H*^(1)^. In NMR or ESR the B-field magnitude is tuned quickly on the *T*~1~ timescale to achieve the desired energy gap. In the second step, Emmy wants to implement a quasi-static evolution of the spin that is now thermal. She brings the spin in contact with the heat bath at temperature *T* and quasi-statically adjusts the magnitude of the external B-field allowing the spin state to thermalise at all times. The final B-field, , is chosen such that the final thermal state becomes *η*^*H*^. In ESR this step can be realised by changing the external B-field slowly on the *T*~1~ timescale so that the spin continuously equilibrates with its environment. Finally, Emmy isolates the spin from the environment and quickly changes the B-field to its original magnitude while the state remains *η*^*H*^. During Step 1 and 3 the system was isolated and the average work drawn is thus just the average energy change. During Step 2 the average work is the equilibrium free energy difference between the final and initial thermal states at temperature *T*, see Supplement for details. In NMR/ESR the work contributions drawn from the spin system are done on the external B-field and the microwave mode. This could be detected by measuring the stimulated emission of photons in the microwave mode or observing current changes induced by the spins dynamics[@b26][@b27]. The overall thermodynamic process has now brought the spin from a quantum state with coherences, *ρ*, into a state without coherences, *η*^*H*^, while keeping the average energy of the spin constant. The net work drawn during the three steps adds up to showing the attainability of the optimum stated in Eq. [(1)](#eq12){ref-type="disp-formula"} for the spin-1/2 example. We note that Eq. [(1)](#eq12){ref-type="disp-formula"} is also the maximal work that can be extracted from a *qubit* state *ρ* under *any* transformation of the system that conserves its average energy, , i.e. for qubits *η*^*H*^ is the optimal final state under this condition. We emphasise that this optimal implementation involves a finely tuned and controlled operation that relies on knowledge of the initial state *ρ*. This is akin to the situation considered in[@b14] where knowledge of the initial global state of system and memory is required for optimal erasure with side-information. It is important to distinguish this situation from that of Maxwell demon's who has access to knowledge of the individual micro-states that make up the ensemble state , and who uses it to beat the second law[@b28]. In the scenario considered here there is no knowledge of the individual micro-states and the process does not violate the second law, on the contrary, it is derived from it. Comparison with single-shot work -------------------------------- The preceding discussion concerned the *average* work that can be drawn when operating on an ensemble of *N* → ∞ independent spins. This scenario contrasts with the single shot situation considered in a number of recent publications[@b7][@b14][@b29][@b30]. In particular, two major frameworks[@b29][@b30] have recently been put forward to identify optimal *single-shot* work extraction and work cost of formation in the quantum setting. These frameworks rely on a resource theory approach[@b6] and make use of min- and max-relative entropies that originate from one-shot information theory. The optimal work extraction schemes of these frameworks require non-diagonal states to be decohered first to become diagonal in the energy basis. This decoherence step is assumed to not have an associated single-shot work. However, the present analysis of energy basis projections showed that thermodynamic projection processes can yield positive average work, see Eq. [(1)](#eq12){ref-type="disp-formula"}. Therefore one may expect a positive work for removing coherences from a state *ρ* in the single-shot setting, too. Since our focus is the *N* → ∞ limit we will not aim to construct the single-shot case. Nevertheless, to establish a notion of consistency between single-shot results[@b29][@b30] and the average analysis presented here we now separate the projection into a diagonal part that can be analysed in the single-shot framework and a non-diagonal part that can be analysed in the average framework. One possible decomposition of is the split in three steps each starting and ending with Hamiltonian *H*: . Here *ρ*~1~ is the rotated state defined above and is the thermal state for the Hamiltonian *H* at temperature *T*. We can now use a single-shot analysis[@b30] for Steps *b* and *c* that involve only states diagonal in the energy basis, giving a single-shot work contribution of , see Supplement. Here *D*~min~ and *D*~max~ are the min- and max-relative quantum entropies, respectively. Taking the limit of *N* → ∞ copies for Steps *b* and *c* and adding the average work contribution for the initial non-diagonal rotation *a*, , one indeed recovers the optimal average work as stated in Eq. [(1)](#eq12){ref-type="disp-formula"}. After making public our results very recently a paper appeared[@b31] that derives the work that can be extracted when removing coherences in a single-shot setting. These results are in agreement with Eq. [(1)](#eq12){ref-type="disp-formula"} and reinforce the above conclusion that coherences are a fundamental feature distinguishing quantum from classical thermodynamics. Comparison with quantum work fluctuation relations -------------------------------------------------- The key observation was that thermodynamic projection processes can have a non-trivial work and heat. Another instance where this has interesting repercussions is the quantum Jarzynski equality[@b4][@b5]. This is a generalisation of the prominent classical fluctuation relation valid for general non-equilibrium processes, which has been used to measure the equilibrium free energy surface inside bio-molecules by performing non-equilibrium pulling experiments[@b19]. The quantum version has recently been tested for the first time in a nuclear magnetic resonance experiment[@b26]. The quantum Jarzynski relation, , links the fluctuating work, *W*, drawn from a system in individual runs of the same non-equilibrium process, with the free energy difference, Δ*F*, of the thermal states of the final and initial Hamiltonian, see Supplement. In its derivation a system initially in a thermal state *ρ*~0~ with respect to Hamiltonian *H*^(0)^ at temperature *T* is first measured in the energy basis of *H*^(0)^. The Hamiltonian is then varied in time ending in *H*^(*τ*)^ generating a unitary evolution, *V*, of the system, see [Fig. 3a](#f3){ref-type="fig"}. A second measurement, in the energy basis of *H*^(*τ*)^, is then performed to establish the final fluctuating energy. For each run the difference of the two measured energies has been associated with the fluctuating work[@b5], Δ*E* = −*W*. The experiment is repeated, each time producing a fluctuating work value. On average the work extracted from the system during the quantum non-equilibrium process turns out to be where is the ensemble's state after the unitary evolution, and similarly the average exponentiated work is calculated. The above identification was made assuming that the system undergoes a unitary process with no heat dissipation. However, the need to acquire knowledge of the system's final energies requires the second measurement. The ensemble state is thus further altered from *ρ*~*τ*~ to *η*~*τ*~, the state *ρ*~*τ*~ with any coherences in the energy basis of *H*^(*τ*)^ removed. This step is not unitary - during the projection the system may absorb heat, , indicated in [Fig. 3b](#f3){ref-type="fig"}, whose value depends on *how* the process is conducted. Thus, while the energy difference for the projection is zero, , for states *ρ*~*τ*~ with coherences the entropy difference is not trivial, . This implies that in an experimental implementation of the Jarzynski relation the work done by the system on average can be more than previously thought, . We conclude that the suitability of identifying , and hence the validity of the quantum Jarzynski *work* relation, depends on the details of the physical process that implements the second measurement. This conclusion is not at odds with previous experiments[@b26] which showed nature's agreement with , involving the average of the exponentiated measured fluctuating energy. Work from coherences of correlated quantum systems -------------------------------------------------- It is insightful to extend the thermodynamic analysis of projections to correlated systems. An experimenter may have access not only to the system *S* but also the auxiliary systems *A* with which *S* is correlated[@b14]. She can then perform a global operation, , that implements a projection locally on the system *S*, i.e. , while leaving the reduced state of the auxiliary system unchanged, i.e. . By doing so the experimenter can optimally draw the overall work , where is the entropy change for the state of system + auxiliary and is still the energy change of the system *alone*. This quantity can be re-written as the sum of two terms: , the extractable work when operating on the system alone given in Eq. [(2)](#eq14){ref-type="disp-formula"}, and , a positive term quantifying the quantum correlations between *S* and *A*, see Supplement. The latter contribution was previously identified in an inspiring paper by Zurek[@b13]. It depends on the choice of projectors and is related to, but broader than, quantum discord[@b32] which is optimised over all possible projectors. This means that even states of system and auxiliary that can be considered classically correlated (i.e. no discord) provide an advantage for drawing work contrasting with the erasure process where this only occurs for highly entangled states[@b14]. The gap between these two sets of correlated states is an intriguing fact and calls for further exploration of the link between thermodynamics and information theory in the quantum regime. Discussion of implications ========================== To conclude, erasure is not the only irreversible information processing task -- in the quantum regime a second fundamental process exists that mirrors Landauer's erasure. In contrast to the minimum heat limit of erasure, thermodynamic projection processes have a maximum work limit. While the former is non-zero for the erasure of classical *and* quantum bits, optimal thermodynamic projection processes have a non-zero work *only* when applied to quantum states with coherences. The optimal average work stated in Eqs. [(1](#eq12){ref-type="disp-formula"}) and ([2](#eq14){ref-type="disp-formula"}) constitutes an experimentally accessible quantum thermodynamic prediction. Future experiments testing this optimal work may be pursued with current setups, for instance with NMR/ESR techniques[@b26][@b27] or single atoms[@b33], and promise to be accessible with other platforms entering the quantum regime, such as single electron boxes[@b22]. Experiments will be limited by practical constraints, such as achieving a quasistatic process and obtaining the maximum work for pure states which may require, for instance, very large B-fields. The derivation of the optimal work value is mathematically straightforward, just like that of Landauer's principle. The result's significance is that it opens new avenues of thought and provides key input for the construction of a future quantum thermodynamic framework. For example, the developed approach opens the door to investigate the connection between microscopic statistical physics and macroscopic thermodynamics in the quantum regime. While it is straightforward to identify the thermodynamic work of quantum processes involving macroscopic ensembles, what is needed is a microscopic concept of work that when averaged, gives the correct macroscopic work. The microscopic work concept should be valid for general (open) quantum processes and quantum states (including coherences), and only require access to properties of the system. While single-shot approaches have discarded coherences[@b29][@b30], fluctuating work approaches cannot be applied directly to a system undergoing open quantum evolution[@b20]. The observation is also important from the experimental perspective as testing quantum thermodynamic predictions will involve measurement -- a projection process. We have argued that measurements, such as those required in establishing the Jarzynski equality, are not necessarily thermodynamically neutral. Indeed, they can be implemented in different physical ways and in general play an active role in thermodynamics, contributing a non-zero average heat and work. This new perspective gives physical meaning to the change of entropy in the debated quantum measurement process - it provides a capacity to draw work. Specifically, work can be drawn when *coherences* of a state are removed during an unselective measurement. Finally, it is apparent that optimal thermodynamic projection processes require use of knowledge of the initial state *ρ*, i.e. its basis and eigenvalues. One may be inclined to exclude use of such knowledge, particularly when considering projections in the context of measurement which is often associated with the acquisition of knowledge. Such restriction would necessarily affect the set of assumptions (T0-T3, Q0-Q1) in the quantum regime. These could be changed, for example, by dropping the possibility of saturating the second law inequality (cf. T2) or choosing a new quantum non-equilibrium entropy that only considers the state's diagonal entries (cf. T3). The latter would mean a departure from standard quantum information theory where entropies are basis-independent. Thus whichever approach one takes - not making or making a restriction - quantum coherences will contribute a new dimension to thermodynamics. They either lead to non-classical work extraction or they alter the link between information theory and thermodynamics in the quantum regime. The line drawn here between the assumptions (T0-T3, Q0-Q1) and results (Eqs. [(1](#eq12){ref-type="disp-formula"}) and ([2](#eq14){ref-type="disp-formula"})) establishes a frame for this possibility to be investigated. Methods Summary =============== Further underlying research materials can be accessed in the [supplementary information](#S1){ref-type="supplementary-material"} that accompanies this article. Proof of Eq. [(2)](#eq14){ref-type="disp-formula"} -------------------------------------------------- Using the first law (T1) the average work drawn in a thermodynamic projection process is simply , where is the average energy change for that process. Relating the average heat absorbed by the system during the process to its entropy change one then obtains (T2). Here is the difference of von Neumann entropies of the system's state before and after the projection (T3). The average work drawn is thus , where the entropy change is non-negative and the energy change can be either positive or negative. The stated *optimal* work, , is achieved when the inequality is saturated by an optimal process (T2) the implementation of which may require knowledge of the initial state and control of coherences (Q1). In the special case of a projection onto the energy eigenbasis the internal energy change is zero, , and one obtains Eq. [(1)](#eq12){ref-type="disp-formula"}. Optimality of three-step process for finite-dimensional systems --------------------------------------------------------------- It is straightforward to generalise the proof of optimality from the two-dimensional spin-1/2 example to thermodynamic projection processes in dimension *d*. Again the projectors map onto the energy eigenspaces of the Hamiltonian, , where , , are the energy eigenvalues. A general initial state can be written as where are probabilities, , are rank-1 projectors on the corresponding eigenvectors , and . A unitary operation, *V*, is now chosen such that it brings the initial configuration (*ρ*, *H*) into the new diagonal and thermal configuration where and . The new energy eigenvalues, , are adjusted such that the probabilities *a*~*k*~ are thermally distributed with respect to *H*^(1)^ for the bath temperature *T*. Adjusting the Hamiltonian eigenvalues while letting the state thermalise at all times now results in a isothermal quasi-static operation from to . Here the new energy eigenvalues, , are chosen to be thermal (at *T*) for the state's probabilities which are given by . Finally, a quench brings the thermal configuration quickly into the non-equilibrium state . The average work for this overall process is where and because the first and third steps are unitary (Q0 + T0). The quasistatic step's work is[@b25][@b29] where is the thermal equilibrium free energy for Hamiltonian *H*^(1)^, and similarly, . Summing up and using , one obtains concluding the optimality proof of the process sequence. Additional Information ====================== **How to cite this article**: Kammerlander, P. and Anders, J. Coherence and measurement in quantum thermodynamics. *Sci. Rep.* **6**, 22174; doi: 10.1038/srep22174 (2016). Supplementary Material {#S1} ====================== ###### Supplementary Information We thank T. Deesuwan, M. Wolf and R. Renner, G. Morley, R. Uzdin and D. Reeb for insightful discussions and J. Gemmer, R. Renner, S. Horsley and T. Philbin for critical reading of the manuscript. P.K. acknowledges support from the Swiss National Science Foundation (through the National Centre of Competence in Research 'Quantum Science and Technology') and the European Research Council (grant 258932). J.A. is supported by the Royal Society and EPSRC (EP/M009165/1). J.A. thanks the Isaac Newton Institute in Cambridge where part of this work was conceived for the stimulating environment and kind hospitality. This work was supported by the European COST network MP1209. **Author Contributions** J.A. provided the main idea and developed the central argument. P.K. developed the single-shot analysis. Both authors wrote the manuscript and [supplementary information](#S1){ref-type="supplementary-material"}. {#f1} {#f2} ![Dynamical steps in a quantum fluctuation experiment.\ (**a**) The quantum Jarzynski relation is described as characterising the non-equilibrium work of processes that start in a thermal state *ρ*~0~ and evolve unitarily (*V*), driven by a changing Hamiltonian, reaching the final state *ρ*~*τ*~ at time *τ*. This unitary process has no heat contribution. (**b**) Illustration of three steps that are assumed in mathematical derivations of the quantum Jarzynski relation[@b4][@b5]: initial energy measurement of *H*^(0)^ indicated by *M*~0~, unitary evolution, and final energy measurement of *H*^(*τ*)^ indicated by . The ensemble state evolves here from *ρ*~0~ to *ρ*~*τ*~ and then to *η*~*τ*~, the state *ρ*~*τ*~ with its coherences removed. The observed average energy difference encompasses both, the unitary process and the second projection process, and can in general contain a heat contribution , in contrast to (**a**).](srep22174-f3){#f3}

Background {#Sec1} ========== Congenital cataracts are diagnosed within the first year of life. These cataracts are one of the leading causes of blindness in children and are estimated to occur with a prevalence of 3--6 per 10,000 live births \[[@CR1]\]. Congenital cataracts may appear either in isolation or in association with other ocular or systemic anomalies. Up to 25 % of congenital cataracts are thought to be caused by genetic defects \[[@CR2]\]. The genetic landscape of mutations causing congenital cataract is extremely diverse; more than 40 genes and additional loci have been associated with nonsyndromic cataract \[[@CR2]--[@CR6]\]. *GCNT2* (glucosaminyl (N-acetyl) transferase 2, I-branching enzyme) was first identified in 2001 as the gene encoding for the glycosyltransferase responsible for the human blood group I antigen. Recessive mutations in *GCNT2* result in an adult i blood group phenotype, which is also associated with congenital cataracts in some cases \[[@CR7]\]. Alternative splicing of the *GCNT2* gene produces three transcripts (A, B, and C). The three transcripts share a common second and third coding exon with a unique first exon for each isoform; differing expression profiles were identified for the transcripts with only the *GCNT2B* isoform expressed in lens epithelial cells and only the *GCNT2C* isoform expressed in reticulocytes \[[@CR8]\]. To date, seven missense mutations, one nonsense mutation, and two large deletions have been reported; mutations in exon 1C, affecting only the *GCNT2C* isoform, cause the adult i blood group without cataracts while mutations/deletions affecting exons 2 and 3, shared by all isoforms, result in the adult i blood group along with congenital cataract \[[@CR7]--[@CR11]\]. Case presentation {#Sec2} ================= Patient 1(individual II:1) is an 18-month old Pakistani female affected with bilateral dense central congenital cataract (Fig. [1a,](#Fig1){ref-type="fig"} Table [1](#Tab1){ref-type="table"}) which were visually significant and required extraction at 2 months of age, mild asymmetry of the palpebral fissures, and left nasolacrimal duct obstruction; her development is normal and growth parameters are generally normal with the exception of borderline microcephaly (length 83.8 cm, 75--90th centile; weight 10.2 kg, 25--50th centile; and head circumference 44 cm (3rd centile)). Physical exam at 4 months of age identified hypotelorism (familial) and mildly widely-spaced nipples. Her younger brother, age 6 months, was similarly affected with visually significant bilateral dense central congenital cataracts requiring extraction around 2 months of age; his length (67.5 cm, 25--50th centile), weight (6.8 cm, 5--10th centile), and head circumference (42.5 cm, 10--25th centile) are all within the normal range (Table [1](#Tab1){ref-type="table"}). Family history shows unaffected second-cousin parents with additional endogamous mating within the family. A double second-cousin to the proband is affected with bilateral non-syndromic anophthalmia/ microphthalmia with no additional details available.Fig. 1Patient photographs and pedigree. **a** Photograph of Patient 1's eyes at 2 months of age showing bilateral cataract. **b** Pedigree showing both affected siblings with a homozygous deletion of 6p24.3 while the unaffected parents are heterozygous carriers. WT: wild type; black arrow indicates probandTable 1Phenotype and genotype information of the affected patientsPatientLens phenotypeOther featuresDevelopmentDeletionGenes involvedPatient 1Bilateral dense central congenital cataracts; extraction at \~2 months of ageBorderline microcephaly (3rd centile), mild asymmetry of the palpebral fissures, left nasolacrimal duct obstruction, hypotelorism, somewhat widely-spaced nipplesWNL97.9 kb homozygous deletion of 6p24.3The first coding exons of *GCNT2A* and *GCNT2B,* two 5'noncoding exons of *GCNT2A*, and a part of the region upstream of *TFAP2A*Patient 2Bilateral dense central congenital cataracts; extraction at \~2 months of ageNoneWNL97.9 kb homozygous deletion of 6p24.3The first coding exons of *GCNT2A* and *GCNT2B,* two 5'noncoding exons of *GCNT2A*, and a part of the region upstream of *TFAP2A* Materials and methods {#Sec3} ===================== Whole exome sequencing was performed by Macrogen (previously Axeq) and analyzed as previously described \[[@CR12]\]; briefly, exome data from the proband was analyzed using the SNP & Variation Suite (SVS; Golden Helix, Bozeman, MT, USA) to identify/exclude mutations in the coding and splicing regions of 40 known nonsyndromic cataract genes and 7 additional crystallins \[[@CR3]--[@CR6]\]; synonymous variants and variants with a frequency of \>1 % in the general population ([http://exac.broadinstitute.org](http://exac.broadinstitute.org/), <http://evs.gs.washington.edu/EVS/>, <http://www.1000genomes.org/>) were considered to be benign variants. Copy number variation analysis was completed by screening exome sequencing data using the Copy Number Inference From Exome Reads (CoNIFER) v0.2.2 software package as previously outlined \[[@CR13]\]; regions of interest were further verified by independent quantitative PCR reactions using DNA samples from the proband and other available familial samples with SYBR Green PCR Master Mix (Applied Biosystems/Life Technologies, Carlsbad, CA, USA). qPCR reactions utilized three region-specific probes (Additional file [1](#MOESM1){ref-type="media"}: Table S1) and were performed as follows: primers located within regions of interest were designed using Primer3Plus software (<http://sourceforge.net/projects/primer3/>) using qPCR settings. Each reaction was comprised of five nanograms of DNA in a total reaction volume of 12uL. Each primer set was run three times in triplicate using patient, parental or control DNA on a Bio-Rad CFX Connect Real-Time PCR machine (Bio-Rad, Hercules, CA, USA). A primer set for the housekeeping gene *RPPH1* (ribonuclease P RNA component H1) was used to normalize all data. A probe located in *NDP* (Norrie disease (pseudoglioma)), located on the X-chromosome, was used as a copy-loss control. All experiments included a no-template control and an unaffected human DNA sample with presumably normal copy number at each region for comparison. Copy number changes were calculated using the 2^-ΔΔCt^ method as previously described \[[@CR14]\]. Following qPCR confirmation, the size and exact breakpoints of the deletion were determined using a series of regular PCR reactions that utilized primers located on both ends of the region (as defined by CoNIFER and qPCR analysis) and standard conditions (Additional file [1](#MOESM1){ref-type="media"}: Table S1). Since the patients were apparently homozygous for the deletion, no amplification product indicated that the primer(s) are located inside of the deleted region while the presence of a PCR product indicated primers outside of the deletion. Once sequences bordering the deleted region on the centromeric and telomeric sides were determined, the corresponding primers were used to amplify a 1.5 kb region across the breakpoints. The resultant product was cloned into pCRII-TOPO® (Life Technologies, Carlsbad, CA, USA) vector using the manufacturer's protocols and sequenced bidirectionally with M13 forward and reverse primers using Big Dye Terminator v3 chemistry and an ABI 3730XL sequencer (Applied Biosystems/Life Technologies, Carlsbad, CA, USA); the obtained sequences were compared with the corresponding reference sequence using BLAST (<http://blast.ncbi.nlm.nih.gov/Blast.cgi>). Results and discussion {#Sec4} ====================== Review of the whole exome sequencing (WES) data from Patient 1 did not identify any potentially pathogenic variants (with only two synonymous variants) in known nonsyndromic cataract genes. The WES data was then analyzed for copy number variation which revealed a potential 208-kb deletion (6p24.3 chr6: 10,412,788-10,621,660) affecting *TFAP2A* and *GCNT2*. The deletion was verified using qPCR probes located in the first coding exon of *GCNT2* isoform A and the first exon of *TFAP2A*; the qPCR confirmed deletion of the *GCNT2* sequence in both unaffected parents (haploid, heterozygous) and affected children (complete loss, homozygous) while diploid copy of the *TFAP2A* sequence was identified in all family members. Further analysis of the region by a series of regular PCR reactions using affected DNA identified the centromeric breakpoint between chr6:10472330--10472606 (set 7; diploid) and chr6:10474759--10474901 (set 8; complete loss) and the telomeric breakpoint between chr6:10570580--10570905 (set 13, complete loss) and chr6:10571951--10572257 (set 14, diploid). Primer sets designed to span the deleted region produced a \~1.5 kb product from the DNA of the affected patients. Sequencing of this product identified the exact deletion breakpoint sites: their analysis revealed the presence of Alu repeats and specifically a 12-bp identical sequence at both sides of the deleted region of 97.974-kb (hg19, chr6: 10,473,864--10,571,838) (Fig. [2](#Fig2){ref-type="fig"}). The homozygous deletion encompassed four exons of *GCNT2* (the first two noncoding and one coding exons of isoform A and the first coding exon of isoform B) and extended 47.471-kb upstream of the most 5' exon of the *GCNT2* gene (Fig. [2](#Fig2){ref-type="fig"}). The distance from the telomeric end of the deletion to the nearest protein-coding gene, *TFAP2A* (transcription factor AP-2 alpha), is 54.300-kb. The distance from the centromeric end of the deletion to the fist exon of *GCNT2* isoform C is 13.922-kb. Although *GCNT2C* and *TFAP2A* were not included in the deletion, effects on their expression through possible interference with regulatory elements cannot be ruled out.Fig. 2Schematic presentation of the chromosome 6p24.3−24.2 region and the identified deletion. The UCSC Genome Browser ([http://genome.ucsc.edu](http://genome.ucsc.edu/)) view of the deleted region indicating the positions of genes is included; the deletion identified in the affected family is shown as a rectangular red box; the DNA sequence across the breakpoint for the deleted allele is shown at the bottom of the drawing with regions corresponding to the telomeric and centromeric flanks of the deletion indicated by dashed lines and a 12-nt repeat highlighted in red font Genomic deletions of *GCNT2* have been previously reported in two families with blood group i and congenital cataracts but both deletions included exons 2 and 3 which are shared by all isoforms \[[@CR7], [@CR11]\]. Borck and colleagues noted that the *GCNT2* locus is rich with Alu elements and therefore is likely a hotspot for deletions or duplications to occur \[[@CR11]\]. The *GCNT2* gene has three differentially expressed transcripts, with *GCNT2B* being the only isoform associated with lens function and *GCNT2C* being the only isoform expressed in red blood cells \[[@CR8]\]. The GCNT2 protein modifies the i antigen, a linear sphingoglycolipid present on the cell surface of most human cells as well as on glycoproteins in body fluids, into the active branched I antigen; the i/I antigens are thought to play a role in the regulation of cell growth and differentiation in the developing lens \[[@CR8], [@CR9]\]. The deletion described in this case report differs from previously reported deletions and mutations since it only affects the *GCNT2A* and *GCNT2B* isoforms and leaves the *GCNT2C* isoform intact. Previous studies demonstrated that only the *GCNT2B* isoform is expressed in lens epithelial cells and patients with mutations which specifically affect the C isoform demonstrate the adult i phenotype without congenital cataracts \[[@CR8]\]. Thus, the presence of cataract in the affected patients reported here with a clear disruption of *GCNT2A* and *B* isoforms only is consistent with the isoform-specific roles identified for this gene. Additionally, in the case reported here we were able to identify the exact sequences at the breakpoints and clearly implicate Alu-mediated non-homologous end-joining as a mechanism for this rearrangement. This mechanism has been previously reported by our and other groups \[[@CR15], [@CR16]\]). The deletion reported here also extends into the genomic region upstream of *GCNT2* and *TFAP2A* which are positioned in a head-to-head orientation. *TFAP2A* is approximately 100-kb distal to *GCNT2* and the deletion removes approximately 47-kb of genomic sequence between the two genes. It is possible that this deletion could affect *TFAP2A* function through removal/rearrangement of regulatory elements, as has been shown for other genes \[[@CR17]--[@CR19]\]. *TFAP2A* is a retinoic acid responsive transcription factor which is required for normal development of the lens and optic cup as well as for parts of the craniofacial region. Heterozygous mutations in *TFAP2A* cause Branchio-Ocular-Facial-Syndrome (BOFS) characterized by craniofacial phenotypes (distinct facial features, microcephaly, and cleft lip/palate), skin defects in the cervical region or regions around the ear, ocular defects (microphthalmia, coloboma, strabismus, cataract, or ptosis), lacrimal duct obstruction, and hearing loss \[[@CR20]--[@CR22]\]. Missense mutations account for the majority of *TFAP2A* variants, however whole gene deletions have also been reported. To date, no deletions affecting the upstream region of *TFAP2A,* but not the coding region itself, have been reported. Careful physical examination of the patients did not identify sufficient features to warrant a diagnosis of BOFS in the siblings, but Patient 1 did show borderline microcephaly, mild asymmetry of the palpebral fissures, left nasolacrimal duct obstruction, and somewhat widely spaced nipples. While the shared cataract phenotype observed in the affected siblings is consistent with the *GCNT2* deficiency alone, an effect of this deletion on the function of *TFAP2A* and the observed phenotypes cannot be completely ruled out. Interestingly, a double second-cousin to the proband has been reported to be affected with bilateral anophthalmia/ microphthalmia (A/M), an ocular condition that is more consistent with the *TFAP2A* spectrum. It is possible that the familial deletion expanded to include *TFAP2A* in this patient; alternatively, the A/M diagnosis may have an independent genetic etiology. Unfortunately, no other familial samples were available for further study. Conclusions {#Sec5} =========== We identified a \~98-kb homozygous deletion involving several exons of *GCNT2* and the region upstream of *TFAP2A* in two children affected with congenital cataracts from a consanguineous family of Pakistani decent. This cataract-causing deletion removes the first coding exons of *GCNT2* isoforms *A* and *B* but leaves the *GCNT2C* sequence intact, providing further support for the isoform-specific roles of this gene; this is the first disruption of *GCNT2* reported which does not affect isoform *C*. While the patients do not fit a diagnosis of BOFS, one sibling demonstrates mild overlap with the phenotypic spectrum, and therefore an effect of this deletion on the function of *TFAP2A* cannot be ruled out. Additional file {#Sec6} =============== Additional file 1: Table S1.Summary of PCR/qPCR reactions and copy number status in the affected family. (DOCX 21 kb) A/M : Anophthalmia/Microphthalmia BOFS : Branchio-ocular-facial-syndrome GCNT2 : Glucosaminyl (N-acetyl) transferase 2, I-branching enzyme TFAP2A : Transcription factor AP-2 alpha WES : Whole exome sequencing The authors also gratefully acknowledge the patients and their family for their participation in this research study. This work was supported by the National Institutes of Health awards R01EY015518 (EVS) and funds provided by the Children's Hospital of Wisconsin (EVS), along with 1UL1RR031973 from the Clinical and Translational Science Award (CTSA) program and the National Eye Institute of the National Institutes of Health under Award Number P30EY001931. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Funding {#FPar1} ======= This work was supported by the National Institutes of Health awards R01EY015518 (EVS) and funds provided by the Children's Hospital of Wisconsin (EVS), along with 1UL1RR031973 from the Clinical and Translational Science Award (CTSA) program and the National Eye Institute of the National Institutes of Health under Award Number P30EY001931. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Availability of data and materials {#FPar2} ================================== Data from this study that do not pertain to identifiable patient information are freely available and provided as supplemental material and/or can be obtained by contacting the corresponding author. Authors' contributions {#FPar3} ====================== EVS conceived and designed the study. DC performed ophthalmological evaluation and referred the patient to the study. LMR enrolled the family. HH and EW carried out the genetic studies. HH, EW, EVS, and LMR analyzed and interpreted the data. HH, EW, LMR, and EVS drafted the manuscript. All authors read and approved the final manuscript. Competing interests {#FPar4} =================== The authors declare that they have no competing interests. Consent to publish {#FPar5} ================== Written consent for publication was obtained from the parents of the patients. A copy of the consent is available for review by the Editor of this journal. Ethics approval and consent to participate {#FPar6} ========================================== This human study was approved by the Institutional Review Board of the Children's Hospital of Wisconsin and carried out in accordance with the Declaration of Helsinki. Written and informed consent for molecular studies was obtained from the parents of the patients. A copy of the consent is available for review by the Editor of this journal.

Introduction ============ Diabetic cardiomyopathy contributes to increased cardiovascular mortality in diabetes mellitus (DM) patients and is characterized by a progressive alteration of left ventricular (LV) function. At a preclinical stage, a decrease in systolic myocardial strain has been suggested in echocardiographic studies. MRI techniques remain the gold standard for quantification of myocardial deformation but only a single study suggested systolic abnormalities in type 2 DM patients with evidence of diastolic dysfunction. MR-tagging is the most common technique for strain calculation using CMR but is intrinsically limited in measuring transmural variations. Cine-Displacement ENcoding Imaging with Stimulated Echoes(DENSE) has been recently proposed as an alternative that benefits from an increased spatial resolution. Purpose ======= To evaluate whether cine-DENSE and MR-tagging confirm the existence of a sub-clinical myocardial dysfunction in a population of type 2 DM patients with no sign or history of heart disease and normal conventional echo and MRI parameters. Methods ======= 37 patients with type 2 DM (50.6±5.6 years, 8 females, HbA1c 7.6±1.2%) and 21 age-matched controls (49.7±8.0 years, 11 females) underwent a CMR study on a 1.5T scanner. Subjects were excluded if standard echocardiography showed significant abnormality. After a standard CMR study for conventional LV function assessment, two-dimensional cine-DENSE pulse sequence with short-echo train echo-planar imaging readout and cine-tagging with complementary spatial modulation of magnetization(CSPAMM) were acquired in short axis views at the same basal, mid and apical levels. LV volumes and ejection fraction were measured on cine-MRI images. Regional circumferential maximal systolic strain(ε~c~) was calculated from cine-DENSE and MR-tagging acquisitions on 16 LV segments. Average maximal systolic strain in each slice and a whole heart mean value(ε~c~mean) for each patient were calculated. Post-processing of cine-DENSE acquisitions included adaptive phase-unwrapping and spatial filtering. CSPAMM images were processed using *InTag* post-processing toolbox (Creatis, Lyon, France) implemented in OsiriX software (Geneva, Switzerland) with motion estimation based on the *Sine Wave Modeling* approach. Results ======= Standard cine-MRI LV function parameters were normal and comparable between groups (table [1](#T1){ref-type="table"}). Whereas LV ejection fraction was similar in the 2 groups, cine-DENSE showed a significant decrease in ε~c~ at basal, mid and apical LV level and in ε~c~mean in the DM group as compared to controls. MR-tagging confirmed a decrease in ε~c~ at the 3 LV levels and in ε~c~mean in DM patients as compared with controls. ###### Left ventricular function in type 2 diabetes mellitus patients and controls. DM patients Controls P ----------------------- ---------------- ---------------- --------- LVEDV (mL) 120 ± 26 129 ± 28 0.26 LVESV (mL) 41 ± 12 41 ± 12 0.90 LVEF (%) 66 ± 6 68 ± 6 0.30 ε~c~ base MR-tagging -0.173 ± 0.040 -0.200 ± 0.028 0.004 ε~c~ mid MR-tagging -0.177 ± 0.045 -0.220 ± 0.035 \<0.001 ε~c~ apex MR-tagging -0.189 ± 0.056 -0.232 ± 0.025 \<0.001 ε~c~ mean MR-tagging -0.179 ± 0.045 -0.216 ± 0.025 \<0.001 ε~c~ base cine-DENSE -0.134 ± 0.019 -0.155 ± 0.019 \<0.001 ε~c~ mid cine-DENSE -0.150 ± 0.021 -0.174 ± 0.020 \<0.001 ε~c~ apex cine- DENSE -0.153 ± 0.022 -0.193 ± 0.018 \<0.001 ε~c~ mean cine-DENSE -0.144 ± 0.016 -0.171 ± 0.016 \<0.001 LVEDV= left ventricular end-diastolic volume; LVESV= left ventricular end-systolic volume; LVEF= left ventricular ejection fraction; ε~c~ =Régional circumferential maximal systolic strain. Conclusions =========== Cine-DENSE and MR-tagging confirm subclinical myocardial dysfunction in asymptomatic patients with Type II DM.

Background {#Sec1} ========== Hemangioblastomas (HB) are highly vascular tumors, which account for approximately 3 % of all tumors of the central nervous system (CNS) \[[@CR1]\]. It occurs in a subset of CNS locations, including the cerebellum (37 %), brainstem (10 %), and spinal cord (50 %) \[[@CR1]\]. They are classed as grade one tumors under the World Health Organization\'s classification system. While most of these tumors are low grade and benign, some hemangioblastomas can present aggressive and occasionally malignant behavior. Hemangioblastomas occur as sporadic tumors (75 %) or as a manifestation of an autosomal dominantly inherited disorder, von Hippel-Lindau (VHL) disease (25 %) \[[@CR2]\]. VHL hemangioblastomas are most commonly caused by germline exon deletions or truncating mutations \[[@CR3]\] of the Von Hippel-Lindau *(VHL*) tumor-suppressor gene. The VHL protein, which is the critical part of a ubiquitin ligase protein complex that binds to the hypoxia-inducing factors HIF-1 and HIF-2 transcription factors and targets them for ubiquitination and proteosomal degradation. Dysregulation of this VHL-associated function causes increased expression of a variety of growth factors, including erythropoietin, PDGF, VEGF and TGF. Upregulation of these factors may lead to angiogenesis and tumorigenesis. Additional mechanisms of tumorigenesis have been described outside of the HIF pathway, including alterations in microtubule binding and stabilization, abnormal extracellular matrix composition as well as apoptosis and transcription regulation \[[@CR4]\]. For most VHL disease related hemangioblastomas, the inactivation or loss of both alleles of the VHL gene is required. In addition to the phenotypic variability associated with allelic heterogeneity, genetic modifiers may influence the phenotypic expression of VHL disease. Allelic variants in the *CCND1, MMP1* and *MMP3* genes have been reported to influence hemangioblastoma development \[[@CR5]\]. This reiterates the need for elucidating other genetic alterations specific for hemangioblastoma beside the hits of *VHL* gene. Moreover, in a subset of tumors including mostly sporadic hemangioblastomas, the genetic pathways involved in tumorigenesis have not been defined yet \[[@CR6]\]. Copy number variants (CNVs) are alterations of DNA sections in result of genomic deletions (fewer than the normal number) or duplications (more than the normal number) on certain chromosomes and are common to many human cancers. Comparative genomic hybridization (CGH) by single nucleotide polymorphism (SNP) arrays is a cutting edge technology that allows characterization of CNVs. SNP array karyotyping provides genome-wide assessment of copy number and loss of heterozygosity (LOH) in one assay. SNP array platforms, such as Affymetrix SNP 6.0 (Affymetrix, Santa Clara, CA, USA), often identify amplifications/deletions at a single gene level, which could not have been accomplished by previous methods. Thus, modern SNP arrays offer a powerful method for the discovery of oncogene and tumor suppressor gene involvement in tumors, as well as for improved cancer classification \[[@CR7]\]. In contrast to surveillance of genome wide alterations by CGH arrays it is possible to directly quantify the absolute copy number of specific DNA loci by Droplet Digital PCR (ddPCR). In ddPCR, target sequences are amplified by PCR and the reaction products are partitioned into droplets and amplified to endpoint with TaqMan probes as in qPCR, then their concentrations are determined based on the number of fluorescently positive and negative droplets in a sample well. The absolute number of target and reference DNA molecules is calculated and provides the target copy number variation (CNV) \[[@CR8]\]. In the present study we used high-resolution SNP arrays for the first time to genome wide analysis of aberrations in hemangioblastomas aiming at the identification of novel pathogenetic mechanisms and possible targets for rational therapy. We validated the main reoccurring genetic changes by ddPCR highly precise quantification. Methods {#Sec2} ======= Study population {#Sec3} ---------------- A total of 44 hemangioblastoma samples were used for the present study. Thirteen frozen samples obtained from The Sourasky Medical Center, Tel Aviv, Israel were used for the CGH analysis. Additional 32 formalin fixed paraffin embedded (FFPE) samples from Sheba Medical Center, Tel Hashomer, Israel were used as validation group. The study was approved by the ethical review boards of both Sheba and Tel Aviv Sourasky Medical Centers and was consistent with the declaration of Helsinki including informed consents. Clinical parameters, such as sex, age at diagnosis, and pathologic classification were collected from patient records. Clinical information of the patient's cohort is outlined in Table [1](#Tab1){ref-type="table"}.Table 1Cohort characteristicsCharacteristicFrozenParaffinAverage age48.553.5Median age5153Spinal samples43Brain samples829Total number1232 CGH analysis {#Sec4} ------------ DNA was purified from frozen tissues using DNeasy (Qiagen Inc., Valencia, CA). One sample of pooled normal genomic DNA, provided by Affymetrix, was used as experimental positive control. 250 ng of genomic DNA was digested with *Nsp*I (New England Biolabs, Inc) and then ligated to Nsp adaptors. The adaptor-ligated DNA fragments were amplified, fragmented using DNase I, end labelled with a biotinylated nucleotide, and hybridized to a human cytoscan HD array (Affymetrix) at 50 °C for 17 h. After hybridization, the arrays were washed, stained, and finally scanned with a GeneChip scanner 3000 (Affymetrix). All procedures were performed according to the manufacturer's protocols. Array experiments were performed using the high-resolution Affymetrix CytoScan HD microarray (Affymetrix, Inc, Santa Clara, CA) containing 2,696,550 markers of which 1,953,246 are non-polymorphic markers and 750,000 SNPs with over 99 % accuracy to detect accurate breakpoint estimation as well as loss of heterozygosity (LOH) determination. This chip covers 340 International Standards for Cytogenomic Arrays (ISCA) constitutional genes, 526 cancer genes (99.6 %) and 36,121 RefSeq genes. The chip uses marker intervals of 25 markers / 100 kb. Analysis of CEL files from the Affymetrix CytoScan HD Array or Cytogenetics Whole-Genome 2.7 M Array was done with the Chromosome Analysis Suite (ChAS) software for cytogenetic analysis. Signal processing was done by Signal Covariate Adjustment, Fragment Correction, Dual Quantile Normalization and PLIER signal summarization. Dual Quantile Normalization was done to equalize each array's intensity distribution copy number and SNP probes separately. For SNP markers, multiple probes for each allele were summarized to single values. Copy number (CN) was calculated by hidden Markov model copy number segments after log2 calculation, high pass filter image correction, log2 ratio covariate adjustment and systematic residual variables removal. The baseline for CN = 2 (normal autosomal copy number state) was established and used by the analysis software by Affymetrix company using a set of 380 phenotypically normal individuals named as reference. The reference sample includes 186 females and 194 male. For chromosome X, only females were used and for chromosome Y only males were used. Log2 ratios for each marker are calculated relative to the reference signal profile. Results of the summarized Data (CYCHP files) were viewed as chromosomal aberrations in table and graphical formats. We also added visual inspection of probe performance for altered segments. Reference intensity intended to represent the copy normal state (typically 2). Log ratios above 0 mean CN gain, log ratios below 0 mean CN loss and log ratios around 0 represent no change. Abnormal DNA copy numbers are identified automatically using 25 markers for loss/50 markers for gains. VHL sequencing {#Sec5} -------------- To screen the *VHL* gene for mutations in our cohort, we performed direct sequencing of the coding region. Exons 1, 2 and 3 of the *VHL* gene and their immediately flanking sequences were amplified by PCR as described previously \[[@CR9]\]. The PCR amplification products were purified by using the QIAquick PCR Purification Kit (Qiagen), according to the manufacturer's instructions. The amplification primers were used as primers in the sequencing reactions, except for exon 1, for which we designed a new cycle sequencing primer (5′CGAAGATACGGAGGTCGA3′). Cycle sequencing was performed using the ABI PRISM Big Dye Terminator Cycle Sequencing Ready reaction kit (Applied Biosystems, Foster City, CA, USA), followed by isopropanol precipitation. The fragments were sequenced by automated sequencing analysis on an ABI Prism 377 sequencer (Applied Biosystems). Droplet digital PCR {#Sec6} ------------------- Copy Number validation was done on all samples, frozen and paraffin embedded, hemangioblastoma biopsies. Genomic DNA was purified using QIAamp DNA mini (Qiagen). Copy number variation (CNV) test was performed by droplet digital PCR (ddPCR) as previously described \[[@CR10]\]. In short, 16 ng of genomic DNA samples were added to 2xddPCR supermix (Bio-Rad) with final concentration of 500nM of each primer and 250nM probe in duplex of the tested gene and RNaseP. RNaseP served as a CNV = 2 reference gene. Probes for the tested genes contained a FAM reporter and RNaseP contained HEX. The genomic DNA and PCR reaction mixtures were partitioned into an emulsion of approximately 20,000 droplets using the QX100 droplet generator (Bio-Rad, USA). The droplets were transferred to a 96-well PCR plate, heat sealed, and placed in a conventional thermal cycler. Thermal cycling conditions were: 95oC for 10 min followed by 40 cycles of 94oC for 30 s, 60oC for 60 s and one cycle of 98oC for 10 min and finally 4oC hold. Following PCR, the plates were loaded into QX100 droplet reader (Bio-Rad) and the CNV value was calculated using Quantasoft software (Bio-Rad, USA). Primers and probes for selected areas enlisted in Table [2](#Tab2){ref-type="table"} were designed using Primer Express software (PE Corp, USA) and specificity was verified using NCBI BLAST online tool (National Library of Medicine, USA). RNaseP primers and probes were obtained from Bio-Rad.Table 2Digital PCR primers and probes for selected areasGeneOligonucleotidesSequence (5′-3′)reporter*PTPN11*Forward primerTTAGAGACAGGGTCCCACTCTTG-Reverse primerGCTTGAGGATGCAGTAAGCTATGA-ProbeCCTGGCTGGAGTGCAGTGGCGTFAM*CHECK2*Forward primerCATTTTTCTCTTAGTATCTTTCTGGGAAT-Reverse primerCATTTCTGAGCCCAGCAATACA-ProbeTCACAATCCAGGGCTACAGTAAGACCCATGFAM*PTCH1*Forward primerGCGTGCGAAGGTGGAGACT-Reverse primerTCATTGGCCTCCCACTTGA-ProbeTGTCTTCTCCCCCATGTCGGFAM*EGFR*Forward primerAGGAGGAACAACGTGGAGACA-Reverse primerGAGACACCGGAGCCACAGA-ProbeCCCAGAGGTGGAACGTTGGCCCFAM*RNaseP*Forward primerGATTTGGACCTGCGAGCG-Reverse primerGCGGCTGTCTCCACAAGT-ProbeCTGACCTGAAGGCTCTHex Results {#Sec7} ======= SNP array profiling identifies recurrent CNVs {#Sec8} --------------------------------------------- The systematic genome-wide gain and loss segments based CNVs data detected by SNP profiling in hemagioblastoma is detailed in the karyoview (Fig. [1](#Fig1){ref-type="fig"}). We identified 94 CNVs with a median of 18 CNVs per sample. Twenty-three of them involved noncoding regions in centromeres that are known to harbor spurious CNVs, most of them were less than 100 Kb long, and 56 were found in at least two samples. The most frequently gained regions were on chromosomes 1 (p36.32) and 7 (p11.2). These regions contain the *PRDM16* and *EGFR* genes, respectively. Recurrent losses were located at chromosome 12 (q24.13) which includes the *PTPN11* gene. Pathway analysis revealed that *EGFR*, Notch and HedgeHog signaling were the most frequently altered pathways promoting angiogenesis and proliferation. *Mir 551a* (part of *PRDM16*) gain was detected in seven samples, *miR-196a-2* gain was observed in five samples and *miR-196b* gain was detected in four samples. The list of recurrent CNVs found in at least five specimens (Fisher exact test *p*-value \< 0.05), including type of alteration, involved chromosome, cytobands, and overlapping genes/miRNAs according to the RefSeq database is provided in Table [3](#Tab3){ref-type="table"}. Two samples had LOH affecting the *CHECK2* region (Fig. [2](#Fig2){ref-type="fig"}).Fig. 1Ideogram illustrating genome-wide distribution of DNA amplifications and deletions (copy number variation) on each chromosome. Each line next to the chromosome represents one sample. Gains in copy number are represented by triangles pointing upwards and losses in copy number are represented by triangles pointing downwards with indication of gene in that region. Chromosome numbers are indicated in boxesTable 3Frequent chromosome losses and gains involving single genesNumber of samplesTypeChromo someBandGenemiRNAGV7Gain7p11.2*EGFR*-7Gain1p36.32*PRDM16*hsa-mir-551ahsa-mir-551av7Loss12q24.13*PTPN11*-7Gain2q31.1*HOXD11, HOXD13 HOHOHOXD13 HOXD11HOXD13*v6Loss13q12.2*FLT3*-6Gain9q22.32*PTCH*v5Loss8p11.22*FGFR1*-5Loss3p13*FOXP1*-5GainXq26.2*GPC3*-5Gain12q13.13*HOXC13, HOXC11*hsa-mir-196a-2v5Loss22q13.1*MKL1*v5Loss22q12.1*CHEK2*vThe column "Number of samples" represents the number of patients with a particular genomic abnormality. For each variation the type of variation (gain or loss), location on chromosome and band, the gene located in this locus and miRNAs located in this locus are enlisted. Variants reported in the Database of Genomic Variants (DGV) database in genes are denoted as "v" and if none reported as "-" under the column DGVFig. 2*CHECK2* copy number state. *CHECK2* locus is labeled by centrally located data track with dashed line. The data point scores form a trinomial distribution about the values 2, 0 and −2, where values around 0 represent heterozygous SNPs, while homozygous SNPs have a value of approximately 2 or −2. LOH is located on genomic region with a scarcity of heterozygous SNP calls. **a** Sample 6 and (**b**) Sample 13 exhibit LOH of *CHECK2* region, (**c**) The control sample exhibits normal copy number state VHL status {#Sec9} ---------- We determined the *VHL* mutation status of our discovery cohort by both DNA sequencing and results from the CGH arrays. In two of the thirteen frozen samples (samples 7 and 10) *VHL* deletions were detected by our CGH analyses. The deletion in sample 7 encompassed 332 kb with 564 markers. The deletion in sample 10 was shorter (135 kb) with 344 markers. Figure [3](#Fig3){ref-type="fig"} illustrates the smooth signal obtained in chromosome band 3p25.3 for these samples. Sequencing based mutational analysis of *VHL* did not reveal any additional mutations. Samples 7 and 10 incurred CNVs similarly to the other samples. The average number of CMVs in the other samples was seven. Sample 7 had 14 of the common CNVs thus had more than the average CNVs but sample 10 had less CNVs (six) which is similar to the average.Fig. 3VHL locus (Ch.3 p25.3) deletion data display. The smooth signal derived from the raw data is displayed. The smooth signal value is calculated using the intensity values of the flanking probes and the data is displayed by a continuous line which is filled down to the x axis. Focusing on VHL locus was done by selecting the specific region of the chromosome and zooming in such that this region spans the entire x-axis. The data points containing the probe values are overlaid on the graph. Smooth signal value can be between zero or four depending on the Log(2) ratio raw values of the surrounding probes. **a** Sample 7 and (**b**) Sample ten smooth signal is one exhibiting deletion in *VHL* locus (labeled with dashed line). Probes around the area are normal (around 2) CNV validation by digital droplet PCR {#Sec10} ------------------------------------- We validated our findings in a subset of four genes (*EGFR, CHECK2, PTCH1* and *PTPN11*) in the discovery cohort used for the array CGH and in an additional independent set of 32 FFPE specimens, using copy variation detection by digital droplet PCR (ddPCR) analysis. Samples were partitioned into thousands of nanoliter-sized droplets; single template molecules were amplified on a thermocycler, and counted for fluorescent signal. Absolute copy numbers of target and reference sequences were determined by Poisson algorithms \[[@CR8]\]. The RNase P (Ribonuclease P) amplicon maps within the single exon *RPPH1* gene on 14q11.2 was used as the standard reference assay for copy number analysis \[[@CR11]\]. Validation results were as follows: *EGFR* amplification was detected in 29 of 32 patient samples (Fig. [4](#Fig4){ref-type="fig"}), *PTCH1* was amplified in 18 of 32 patients (Fig. [5](#Fig5){ref-type="fig"}) and *CHEK2* was deleted in 27 of 32 (Fig. [6](#Fig6){ref-type="fig"}). Surprisingly, *PTPN11* was deleted in 7/32 whilst it was amplified in 17/32 specimens (Fig. [7](#Fig7){ref-type="fig"}).Fig. 4*EGFR* Copy number variation (CNV) values for clinical samples and one normal genomic DNA sample. Genomic DNA copy number alterations were assessed via ddPCR. The CNV is shown as the number of copies and the Poisson distribution at 95 % confidence interval. The copy number value of normal diploid sequence has a score of two. Copy number above two means amplification in that region and copy number below two means deletion in that regionFig. 5*PTCH1* Copy number variation (CNV) values for clinical samples and one normal genomic DNA sample. Genomic DNA copy number alterations were assessed via ddPCR. The CNV is shown as the number of copies and the Poisson distribution at 95 % confidence interval. The copy number value of normal diploid sequence has a score of two. Copy number above two means amplification in that region and copy number below two means deletion in that regionFig. 6*CHEK2* Copy number variation (CNV) values for clinical samples and one normal genomic DNA sample. Genomic DNA copy number alterations were assessed via ddPCR. The CNV is shown as the number of copies and the Poisson distribution at 95 % confidence interval. The copy number value of normal diploid sequence has a score of two. Copy number above two means amplification in that region and copy number below two means deletion in that regionFig. 7*PTPN11* Copy number variation (CNV) values for clinical samples and one normal genomic DNA sample. Genomic DNA copy number alterations were assessed via ddPCR. The CNV is shown as the number of copies and the Poisson distribution at 95 % confidence interval. The copy number value of normal diploid sequence has a score of two. Copy number above two means amplification in that region and copy number below two means deletion in that region It is important to note that in result of normal cell admixture within tumor samples a high CNV means either high fraction of tumor cells with relatively high copy number in each cell or low fraction of tumor cells with very high copy number in each cell. Likewise, low CNV means either high fraction of tumor cells with relatively low copy number in each cell or low fraction of tumor cells with relatively very low copy number in each cell. In short if the fraction of tumor cells is low then the relative copy number (both high and low CNV) is even higher than what is reported. In our cohort, histopathologic assessment by a pathologist determined that 47 percent of samples had more than 95 % tumor, 7 percent had more than 90 % tumor, 29 percent of samples had more than 70--80 % tumor, 14 percent of samples had between 50--70 % tumor and 3 percent had between 40--50 % tumor. Discussion {#Sec11} ========== We report here for the first time a genome-wide, high-resolution systematic analysis of chromosomal changes in hemangioblastoma. Using the SNP array 6 (Affymetrix) we analyzed 1.8 million genetic markers genome wide to identify amplifications/deletions up to single gene level. We identified a total of 94 CNVs, 23 of them involved noncoding regions. 56 (31 gains and 25 losses) were found in at least two specimens. The most frequently gained regions were on chromosomes 1 (p36.32) and 7 (p11.2). The most frequently deleted region was on chromosome 12 (q24.13). Our findings provide the first high-resolution genome-wide view of chromosomal changes in hemangioblastoma and identify 23 common, ie found in 4 or more patients, candidate genes for hemangioblastoma pathogenesis (Table [3](#Tab3){ref-type="table"}): *EGFR, PRDM16, PTPN11, HOXD11, HOXD13, FLT3, PTCH, FGFR1, FOXP1, GPC3, HOXC13, HOXC11, MKL1, CHEK2, IRF4, GPHN, IKZF1, RB1, HOXA9, HOXA11* and several microRNA, including *hsa-mir-196a-2*. We note that some of these microalterations have been reported very rarely previously in tissue samples from healthy subjects (according to Database of Genome Variants (DGV), which are reported in Table [3](#Tab3){ref-type="table"}. However, in our tumor samples these alterations are significantly more common (p \< 0.00001). Functional annotation analysis by David \[[@CR12]\] reviled that two pathways are prominently affected in the hemangioblastoma samples: the cell proliferation and angiogenesis promoting pathways. The cell proliferation pathway includes the following genes: *CHEK2, EGFR, FGFR, FLT3* and *PTCH1*. The angiogenesis pathway includes the genes *EGFR* and *FGFR*, which are significant because they are also involved in blood vessel formation. Importantly, three of these genes were verified by ddPCR: *EGFR, CHEK2* and *PTCH1*. Furthermore, *PTPN11* was selected for verification as it was lost strikingly often (Table [3](#Tab3){ref-type="table"}). Epidermal growth factor receptor (*EGFR*) is a tyrosine kinase receptor that has been documented with increased expression in a variety of human cancers such as breast cancer, \[[@CR13]\], early stage non-small-cell lung cancers and gliomas \[[@CR14], [@CR15]\]. Our findings are in line with three prior immunohistochemical studies that found *EGFR* overexpression in hemangioblastomas \[[@CR16]\] \[[@CR17]\] \[[@CR18]\]. Fibroblast growth factor receptor 1 (*FGFR1*) micro deletions have been reported in myeloid and lymphoid neoplasms \[[@CR19]\]. Interestingly we discovered that the *FGFR1* deletion involves the FGFR1 2nd intron (chromosome 8, 38291333--38314367), which according to UCSC genome browser hg19 includes a regulatory site. Checkpoint kinase 2 (*CHEK2*) has been implicated in DNA repair, cell cycle arrest, and apoptosis in response to DNA double-strand breaks \[[@CR20]\]. Mutations in *CHEK2* have been reported to be possibly associated with breast cancer and liposarcoma development \[[@CR21], [@CR22]\]. The protein patched homolog 1 (*PTCH1*) is a sonic hedgehog receptor. Loss of function mutations in *PTCH1* are associated with development of various types of cancers, including medulloblastoma \[[@CR23], [@CR24]\], pancreatic cancer \[[@CR25]\] and colorectal cancer \[[@CR26]\]. In contrast to these observations, we find consistent *PTCH1* gains in our cohort of hemangioblastoma patients (Table [3](#Tab3){ref-type="table"}). In support of our findings there are reports of *PTCH1* also acting as an oncogene in a mouse model of skin basal cell carcinomas \[[@CR27]\]. FMS-like tyrosine kinase 3 (*FLT3*) plays a role in hematopoiesis including early hematologic differentiation and early B and T-cell development \[[@CR28]\] and dendritic cells differentiation \[[@CR29]\]. Activating *FLT3* mutations are some of the most common molecular abnormalities in acute myeloid leukemia (AML) \[[@CR30]\]. Interestingly, in our data, we find losses of *FLT3* (Table [3](#Tab3){ref-type="table"}). Consistent with our findings, others have reported deletions in *FLT3* in AML as well \[[@CR31]\] \[[@CR32]\]. Among the altered genes mapped in many of the recurrently gained regions we recognized several *HOX* genes (Table [3](#Tab3){ref-type="table"}) and microRNAs (miRNAs) residing in the same region. *HOX* genes encode master transcription factors important in development and have been reported to be commonly altered in human solid tumors \[[@CR33]\]. On the other hand, miRNAs are small regulatory RNAs that have recently been implicated in a variety of cancers \[[@CR34]\]. miR-196a-2 and miR-196b gain were the most common miRNA CNV in our patients. Interestingly, miR-196a-2 differs from miR-196b by one nucleotide \[[@CR35]\]. The miR-196 gene family is located in the regions of homeobox (*HOX*) transcription factors that are essential for embryogenesis. Up-regulation of miR-196a has been found in breast cancer, adenocarcinoma, leukemia and esophageal adenocarcinoma \[[@CR35]\]. Accordingly the relevant causative change may actually be altered miRNA expression rather than *HOX* gene expression. Other studies have reported a 12-fold increase in miR-9 and a 15-fold decrease of miR-200a in hemangioblastomas distinguishing hemangioblastomas from metastatic clear cell renal cell carcinomas in the CNS \[[@CR36]\]. *Prdm16* is preferentially expressed by stem cells throughout the nervous and haematopoietic systems and promotes stem cell maintenance \[[@CR37]\]. Megakaryoblastic leukemia protein-1 (*MKL1*), is a transcription factor that regulates many processes, including remodeling of neuronal networks and epithelial-mesenchymal transition \[[@CR38]\]. Moreover, deregulation by genetic alterations and/or altered *MKL1* transcription has been shown to have role in myeloproliferative neoplasms \[[@CR39]\]. Finally and most interestingly, protein tyrosine phosphatase, non-receptor type 11 *(PTPN11*) gene encodes the tyrosine phosphatase SHP2 protein required for RTK signaling and has a role in survival, proliferation and differentiation \[[@CR40]\]. The fact that we find the *PTPN11* gene is deleted in some hemangioblastoma patients and is amplified in others may suggest that these tumors actually originate from different cellular lineages. In fact, we found that spinal tumors were overwhelmingly deleted in the *PTPN11* region: 5 of 7 spinal tumors were deleted whilst 1 was amplified and 1 was chromosomally normal at this locus. In contrast, cerebellar hemangioblastomas were generally amplified: 16 out 37 were indeed amplified. However, 10 cerebellar tumors were deleted and 11 of 37 were normal in this genomic region. Interestingly, differential overexpression or deletion of *PTPN11* has been shown in other tumors. For example, mutations in *PTPN11* has been reported to be associated with development of Juvenile Myelomonocytic Leukemia (JMML) \[[@CR41]\], acute myeloblastic leukemia (AML) \[[@CR42]\], and acute lymphoblastic leukemia (ALL) \[[@CR43]\]. Overexpression in gastric carcinomas has been reported \[[@CR44]\]. In contrast, *PTPN11* has a tumor-suppressor function in liver \[[@CR45]\] and cartilage \[[@CR46]\]. Accordingly, decreased *PTPN11* expression was detected in a subfraction of human hepatocellular carcinoma specimens \[[@CR45]\]. Thus, in contrast to its common pro-oncogenic role in hematopoietic and epithelial cells, *PTPN11* may act as a tumor suppressor in cartilage. Accordingly, we hypothesize that the *PTPN11* gene may act in a cell-specific manner: as a tumor suppressor on one hand in the progenitor cells of spinal hemangioblastomas, whilst it acts as an oncogene in the cells of origin of cerebellar hemangioblastoma tumors. Previous analysis of six VHL-related CNS hemangioblastomas showed loss of chromosome 3p or the whole of chromosome 3 to be the most common abnormality, which is detected in 70 % and loss of 1p11-p31 in 10 % \[[@CR47]\]. More relevant to our findings, published CGH studies on 10 sporadic cerebellar hemangioblastomas detected losses of chromosomes 3 (70 %), 6 (50 %), 9 (30 %), and 18q (30 %) and a gain of chromosome 19 (30 %) \[[@CR48]\]. We indeed detected losses and gains in these areas but they were not frequent (15-20 %). Chromosome 3 losses were more abundant, mostly on p13 (5 samples *FOXP1*) and p25 (3 samples showed *PPARG* loss and 2 samples showed *VHL* loss). Interestingly, *FOXP1* transcription factor, located on chromosome 3(p13), can function as a tumor suppressor gene. Low expression in glioma \[[@CR49]\] and Hodgkin lymphoma has been shown \[[@CR50]\]. Differences in methodology, sample size and definition of aberration inclusion criteria may account for some of the apparent inconsistencies between previous studies and our findings. For example, previous results were obtained from VHL-related hemangioblastomas using techniques that identify deletions that are larger than 2 Mb. In the current study we used modern CGH microarrays which scan the DNA every 1 kb and thus we were able to identify very subtle genomic changes. One of the most striking observations in our study is that many CNVs affected single genes (Table [3](#Tab3){ref-type="table"}) and revealed candidate genes, which have not been implicated in hemangioblastomas. This represents an important outcome of this study compared with previous investigations using CGH. Some of the new genes identified here as affected by CNVs in hemangioblastoma may serve as targets for future precisely targeted anti- cancer therapy. For example, antiangiogenic therapy can be given to patients with lesions that are not resectable. Conclusions {#Sec12} =========== In this study, we have demonstrated in two different tumor cohorts and using two different techniques for copy number alteration detection, SNP and digital PCR, that *Chek2* is deleted and *EGFR*, *PTPN11, Ptch1* amplified in majority of hemangioblastoma patients. EGFR is the only gene that has been previously reported as a candidate gene with hemangioblastoma. Independent of HB tumor location *PTPN11* may act as tumor suppressor or oncogene depending on the tumor cell of origin. These findings have potentially relevant clinical value, as this the first high resolution for chromosomal alteration in HB. Future research should be dedicated to the prospective validation of these alterations and further characterization of tumors that carry the deletions/amplifications, as well as of defining the role of these genes. This may offer insights into hemangioblastoma biology, provide DNA-based markers that can be analyzed by FISH suitable for routine clinical applications and eventually lead to the development of effective targeted therapies for HB. CNVs : Copy number variations HB : Hemangioblastomas CNS : Central nervous system VHL : von Hippel-Lindau CGH : Comparative genomic hybridization SNP : Single nucleotide polymorphism LOH : Loss of heterozygosity ddPCR : Droplet Digital PCR CNV : Copy number variation FFPE : Formalin Fixed Paraffin Embedded RNase P : Ribonuclease P *EGFR* : Epidermal growth factor receptor *FGFR* : Fibroblast growth factor receptor 1 *CHEK2* : Checkpoint kinase 2 *PTCH1* : Protein patched homolog 1 *FLT3* : FMS-like tyrosine kinase 3 AML : Acute myeloid leukemia miRNA : microRNA *MKL1* : Megakaryoblastic leukemia protein-1 PTPN11 : protein tyrosine phosphatase non-receptor type 11 Michal Yalon, Shlomi Constantini and Amos Toren contributed equally to this work. **Competing interests** The authors declare that they have no competing interest. **Authors' contributions** RMS conceived, designed and coordinated the study, performed the CGH analysis and drafted the manuscript, IM carried out the droplet digital PCR experiments, IB and DN participated in collecting the FFPE samples, JJ and CD carried out the CGH experiments, JR provided advice and revised the manuscript, MY, SC and AT participated in collecting the frozen samples, designing and coordinating of the study. All authors read and approved the final manuscript. This research was funded by The Sheba Medical Research fund. We thank Sarah South, PhD, University of Utah, for insightful remarks on Affymetrix CGH analysis.

The content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. Introduction ============ Pulmonary sequestration was first described by Pryce in 1946 in the Journal of Pathology and Bacteriology \[[@REF1]\]. Pulmonary sequestration is a rare congenital malformation of the respiratory tract. It constitutes approximately 0.15%-6.4% of all congenital pulmonary malformations \[[@REF2]\]. It is defined as a non-functioning mass of parenchymal lung tissue that lacks communication with the tracheobronchial tree and is supplied by an anomalous systemic artery. Anatomically, it is classified as intralobar sequestration, where it is located within a normal lobe without its own visceral pleura, or extralobar sequestration, which is outside the normal lung with its own visceral pleura. Here, we present a case of a 45-year-old female who had a diagnosis of intrapulmonary sequestration \[[@REF3]\]. Case presentation ================= A 45-year-old Caucasian female presented with a left-sided breast mass. An excisional biopsy showed a high-grade phyllodes tumor, which was treated by resection. Preoperatively, a chest radiograph was obtained, which revealed a left lower lobe shadow suspicious of consolidation. At the time, this was considered to be a community-acquired pneumonia though she had no symptoms. She was hemodynamically stable. Laboratory investigations were within normal limits. She was subsequently treated with a seven-day course of antibiotics. Follow-up chest radiograph showed persistence of the infiltrate which led to a non-contrast computed tomography (CT) scan of the chest. It revealed a moderate consolidation of the left lower lobe and hilar lymphadenopathy (Figure [1](#FIG1){ref-type="fig"}). {#FIG1} She, again, denied any signs of infection including sputum production, hemoptysis, shortness of breath or cough. She also denied a history of frequent pulmonary infections. She was a former smoker and had a 20-pack year history. A fiberoptic bronchoscopy was conducted for her unresolving left lower lobe infiltrate. Brushing of the area of infiltration was obtained along with the lymph node biopsy, which turned out to be non-malignant. At this juncture, a chest CT scan with contrast was obtained, which showed an artery from descending aorta feeding the area of infiltration highly suggestive of pulmonary sequestration (Figures [2](#FIG2){ref-type="fig"}, [3](#FIG3){ref-type="fig"}). {#FIG2} {#FIG3} After discussion with the patient, it was decided to pursue a resection. She was referred to a cardiothoracic surgeon for a video-assisted thoracoscopic surgery (VATS) lobectomy procedure. The patient underwent left lower lobectomy. Pathology report after lobectomy confirmed the features consistent with an intralobar sequestration. Discussion ========== Pulmonary sequestration is a rare congenital malformation of dysplastic lung tissue. The sequestered lung does not communicate with the tracheobronchial tree and is supplied by an anomalous systemic arterial source, most commonly from the aorta \[[@REF4]\]. Intralobar sequestration is four times more common than the extralobar type \[[@REF5]\]. Diagnosis during adulthood is relatively uncommon as 60% of cases are diagnosed in the first decade of life and is a rarity in adults greater than 40 years of age \[[@REF6]\].  Most patients develop symptoms in infancy or early childhood. Symptoms may be non-specific, including cough, chest pain and shortness of breath. Some patients develop recurrent pneumonias and bronchiectasis \[[@REF7]\]. Approximately 15% of patients remain asymptomatic \[[@REF8]\]. Our patient was asymptomatic and had denied previous pulmonary infections. As in our patient, sequestration occurs mostly in the left hemithorax, in the posterior basal segment of the left lower lobe \[[@REF8],[@REF9]\]. In 75% of the patients, the supply to the intralobar sequestration is from the descending thoracic aorta \[[@REF9]\]. CT angiography scan is the imaging test of choice as it can show the anomalous artery feeding into the pulmonary sequestration \[[@REF4],[@REF9]\]. Non-contrasted CT imaging is sometimes adequate to aid in the diagnosis of sequestration. However, in our case, the diagnosis was not clear until CT angiography was performed. Pulmonary angiography is considered to be the gold standard; however, it is not commonly used as the diagnosis is mostly clinched on CT scan imaging. On cut section, the intralobar pulmonary sequestration represents mucus filled airways and small cysts which may be filled with purulent material. On histological examination, there is mucus stasis in the airways and a systemic artery accompanies the airways \[[@REF10]\]. Surgical resection is considered the treatment of choice for intralobar sequestration especially in symptomatic patients \[[@REF8],[@REF11],[@REF12]\]. Even in asymptomatic patients, surgical resection is often recommended due to the risk of serious future complications, including infections, massive hemoptysis and malignant transformation \[[@REF2],[@REF13]\]. The usual surgery is lobectomy either via VATS or standard thoracotomy. Our patient had a VATS lobectomy performed. Conclusions =========== Pulmonary sequestration is an uncommon finding especially in the adult population. Imaging modalities, especially contrast-enhanced CT scan, can aid in diagnosis by localizing the aberrant arterial blood supply to the sequestered lung parenchyma. Contrasted CT scan is also important for preoperative evaluation in these patients. The authors have declared that no competing interests exist. Consent was obtained by all participants in this study We would like to acknowledge that we presented a poster presentation discussing the same case under the same heading at the Chest Conference 2019.

Glioblastoma multiforme (GBM) is the most common and aggressive primary brain tumor with a median survival of 12 months. Conventional therapeutic strategies of radiotherapy and chemotherapy are inadequate to eradicate GBMs because of the diffuse nature of GBM, acquired or innate resistance to therapy and the presence of blood--brain--barrier (BBB).^[@bib1],\ [@bib2]^ To alter the *status quo* that has remained unchanged over 25 years, novel targets and therapeutic strategies need to be developed. Evasion of apoptosis is a key hallmark of cancers that exhibit resistance against therapeutics,^[@bib3]^ making the reactivation of dormant apoptotic programs a favorable approach in treatment. Activating extrinsic apoptosis using death ligands, such as tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), is a promising strategy because of its tumor specificity.^[@bib4]^ Several TRAIL-based therapies such as recombinant human TRAIL and death receptor agonists have been developed and have shown success in preclinical models.^[@bib5]^ Similarly, activating intrinsic apoptosis by inhibiting the antiapoptotic Bcl-2 proteins, Bcl-2 and Bcl-XL, with BH3 peptides has shown success in preclinical tumor models.^[@bib6],\ [@bib7]^ However, the major obstacle in proapoptotic therapies is innate or acquired resistance of tumor cells to the proapoptotic agents.^[@bib8]^ Aberrant regulation of the apoptosis pathway components could be responsible for the failure of desired response to the proapoptotic agents.^[@bib3]^ One well-characterized misregulation is the epigenetic silencing of proapoptotic genes, such as death receptor 4 (*DR4*).^[@bib9]^ However, epigenetic regulation of chemotherapy response of cancer cells has been demonstrated to be dynamic and reversible.^[@bib10]^ Numerous studies have shown that overcoming TRAIL resistance is possible by using secondary agents, such as the histone deacetylase inhibitors,^[@bib11],\ [@bib12],\ [@bib13]^ corroborating the idea that the resistance can be reversed by epigenetic reprogramming. The epigenome of cells are maintained by dynamic histone and DNA modifications throughout the chromatin by a group of chromatin-modifying enzymes (CMEs) called 'writers' such as histone acetyltransferases, histone methyltransferases (HMTs) and DNA methyltransferases (DNMTs).^[@bib14]^ These modifications can be removed by 'erasers' such as histone deacetylases (HDACs) and histone demethylases (HDMs), or can be recognized by 'readers' such as bromodomain-containing proteins.^[@bib15]^ Although the molecular mechanisms leading to aberrant cancer epigenomes are becoming better understood,^[@bib16],\ [@bib17],\ [@bib18]^ most studies in GBM merely focus on DNA hyper/hypomethylation.^[@bib19]^ Therefore, an unbiased and comprehensive assessment of roles of CMEs in apoptosis resistance is needed. In this study, we aimed to interrogate the function of CMEs regulating apoptotic response in GBM and undertook a loss-of-function approach using short-hairpin RNAs (shRNAs) that were designed against a select but diverse set of CMEs and associated proteins. We have shown that loss of KDM2B -- a H3K36-specific histone demethylase -- primed GBM cells for apoptosis through the induction of proapoptotic genes and the suppression of antiapoptotic genes. Our results suggest KDM2B as a novel central epigenetic regulator of GBM cell apoptosis and identify it as a potential proapoptotic target for future of GBM treatment. Results ======= Interrogation of the CMEs reveals KDM2B as a modulator of TRAIL-induced apoptosis in GBM cells ---------------------------------------------------------------------------------------------- To identify the key CMEs regulating the apoptotic response of GBM cells, we applied a shRNA-based loss-of-function screen in U87MG cell line ([Figure 1a](#fig1){ref-type="fig"}). We used a library of 60 shRNAs published before,^[@bib20]^ and expanded it to 125 shRNAs targeting 48 different CME genes with two or three separate shRNAs. The targeted CMEs included DNMTs, HMTs, HDMs, methylated DNA-binding proteins, polycomb-group proteins (PRCs) and a few transcription factors ([Figure 1b](#fig1){ref-type="fig"}). We first assessed the effects of shRNAs alone on U87MG cell viability, to identify shRNAs displaying intrinsic toxicity before TRAIL treatment. Two of the 125 shRNAs were eliminated from further analysis in the screen as they caused \~100% cell death after puromycin selection, possibly due to problems with viral packaging. Accordingly, 6 out of 125 shRNAs reduced cell viability more than 2 S.D. compared with shControl ([Figure 1c](#fig1){ref-type="fig"}). These were shRNAs targeting KDM5C (68±1%), KDM4C (71±1%), KDM4A (75±1%), KDM4A (75±1%), Set1A (77±1%) and KDM3B (77±1%). Five out of 125 shRNAs targeting MeCP2 (115±1%), MBD1 (115±1%), AHCY (117±5%), TCF3 (117±1%) and EZH1 (121±2%) caused minor increases in viability. However, the majority of the shRNAs did not significantly alter cell viability ([Figure 1c](#fig1){ref-type="fig"}). To identify shRNAs that modulated TRAIL sensitivity, we assessed the percent viability changes after TRAIL treatment. Accordingly, we categorized CMEs into two groups based on the phenotype observed upon their silencing compared with shControl cells: (1) suppressors of apoptosis, if their knockdown sensitized the cells to TRAIL, (2) enhancers of apoptosis, if their knockdown conferred cells more TRAIL resistant ([Figure 1d](#fig1){ref-type="fig"}). While TRAIL caused 25±4% reduction in the viability of shControl cells, there were seven shRNAs that sensitized cells to TRAIL. These were targeting five genes, namely *Suv39H2* (41±2%), *G9A* (42±2%), *NR2F2* (45±4%), *RING1A* (48±2 or 50±3%) and *KDM2B* (41±2 or 44±2%) ([Figure 1e](#fig1){ref-type="fig"}). We then focused on the genes *RING1A* and *KDM2B*, whose knockdown led to a phenotype with two independent shRNAs ([Figure 1e](#fig1){ref-type="fig"}). Taken together, our screen identified KDM2B and RING1, an H3K36-specific demethylase and an E3 ubiquitin-protein ligase of H2AK119, respectively, as novel regulators of TRAIL response. Loss of KDM2B cooperates with TRAIL to reduce GBM cell viability ---------------------------------------------------------------- To assess the function of the novel apoptosis-modulating CMEs we identified in GBMs, we first checked the knockdown efficiency of the shRNAs and observed silencing down to 17--40% ([Supplementary Figure 1](#sup1){ref-type="supplementary-material"}). We then focused on KDM2B, as its protumorigenic functions have been demonstrated in various solid and hematological malignancies;^[@bib21],\ [@bib22]^ however, its function in GBMs was not defined. Quantitative RT-PCR (qRT-PCR) analysis of *KDM2B* levels revealed that both shRNAs reduced *KDM2B* mRNA levels down to \~50% ([Figure 2a](#fig2){ref-type="fig"}). However, shKDM2B-2 led to a more robust reduction at protein levels ([Figure 2b](#fig2){ref-type="fig"}). To assess whether the shRNAs targeting KDM2B is specific, we checked the levels of other KDM family members upon KDM2B knockdown and observed no major alterations in their levels ([Supplementary Figure 2](#sup1){ref-type="supplementary-material"}). To further validate the screen results, we conducted ATP-based cell viability analysis of cells transduced with both shKDM2B vectors and verified that cells with reduced KDM2B exhibit cell death significantly more than the controls ([Figure 2c](#fig2){ref-type="fig"}). To examine these differences in cell death further, we used an assay that measures cell growth in real time, where cells' electrical impedance in a well is measured and transformed into a cell index. Accordingly, silencing of KDM2B not only augmented TRAIL response but also accelerated the process of cell death ([Figure 2d](#fig2){ref-type="fig"}). There was no observable difference between untreated cells, showing that KDM2B knockdown has no significant effects on short-term (24 h) proliferation dynamics. This phenotype was also validated by live-cell imaging, where the morphology of individual cells, as well as cell death processes, was observed in real time ([Figure 2e](#fig2){ref-type="fig"} and [Supplementary Videos 1--4](#sup1){ref-type="supplementary-material"}). As shown by automated quantification of cellular blebs that were indicative of apoptotic bodies, the process of apoptosis was accelerated in shKDM2B cells compared with controls ([Figure 2f](#fig2){ref-type="fig"}). The number of apoptotic bodies per frame reached its maxima within 6 h of TRAIL treatment in shKDM2B cells, earlier and significantly in higher numbers than shControl cells. To examine whether KDM2B effects can be recapitulated in additional GBM cell lines, we used a more TRAIL-sensitive line, T98G in parallel. There, KDM2B knockdown led to increased TRAIL sensitivity, in accordance with U87MG cells ([Supplementary Figure 3](#sup1){ref-type="supplementary-material"}). Next, as a complementary approach, we tested whether overexpression of KDM2B would result in an opposite response for TRAIL. Accordingly, KDM2B overexpression led to a modest but significant reduction in TRAIL response in U87MG cells ([Figures 2g, h](#fig2){ref-type="fig"} and [Supplementary Figure 4](#sup1){ref-type="supplementary-material"}). KDM2B overexpression conferred T98G cells more resistant to TRAIL as well ([Supplementary Figure 5](#sup1){ref-type="supplementary-material"}). Taken together, our data suggest that KDM2B expression suppresses TRAIL response and therefore its inhibition leads to increased TRAIL sensitivity in GBM cells. Silencing of KDM2B sensitizes GBM cells to TRAIL-induced apoptosis ------------------------------------------------------------------ To better understand how downregulation of KDM2B reduces viability and increases TRAIL response, we examined the components of the apoptotic machinery in shControl and shKDM2B cells. We observed a 3-fold increase in caspase-3/7 and a 2.5-fold increase in caspase-8 activity levels ([Figures 3a and b](#fig3){ref-type="fig"}). Analysis of the whole-cell lysates of shControl and shKDM2B cells revealed that caspase-7 is indeed upregulated in shKDM2B cells at the protein level ([Figure 3c](#fig3){ref-type="fig"}). Importantly, we observed that there is a significant increase in levels of cleaved PARP, caspase-3 and caspase-7, all of which are indicators of apoptotic cell death. We have also observed a very slight increase in cleaved Bid levels in shKDM2B cells upon TRAIL treatment ([Supplementary Figure 6](#sup1){ref-type="supplementary-material"}). To test the functional roles of caspases in shKDM2B-mediated apoptotic sensitization, we used specific inhibitors for caspase-8, -10 (Z-IETD-FMK) and caspase-9 (Z-LEHD-FMK). In addition, we used a general caspase inhibitor that target caspase-8, -10, -9, -3 and -7 (Z-VAD-FMK), and an inactive caspase inhibitor (Z-FA-FMK) as control. In both shControl and shKDM2B cells, inhibition of all caspases abolished the TRAIL response, suggesting that the observed augmented apoptotic response in shKDM2B cells was dependent on both intrinsic and extrinsic apoptosis pathways ([Figure 3d](#fig3){ref-type="fig"}).^[@bib23]^ To examine whether KDM2B modulates the response of GBM cells to proapoptotic agents that trigger intrinsic apoptosis, besides the extrinsic apoptosis that is triggered by TRAIL, we conducted cell viability experiments with Bcl-2/Bcl-XL inhibitors, ABT-263 and ABT-737 ([Supplementary Figures 7 and 8](#sup1){ref-type="supplementary-material"}). While ABT-263 or ABT-737 alone did not markedly decrease cell viability, combination of ABT-263 or ABT-737 with TRAIL did. We observed that knockdown or overexpression of KDM2B in U87MG cells did not affect this cell line's overall response to ABT-263 and ABT-737 individual treatments. However, knockdown of KDM2B sensitized these cells to ABT and TRAIL combination ([Supplementary Figure 7](#sup1){ref-type="supplementary-material"}), and KDM2B overexpression was able to confer GBM cells more resistant to ABT and TRAIL combination ([Supplementary Figure 8](#sup1){ref-type="supplementary-material"}). These results suggested that the interplay between KDM2B and apoptosis is likely to be valid for both the intrinsic and extrinsic arms of apoptosis. Silencing of KDM2B leads to deregulation of apoptosis-related genes ------------------------------------------------------------------- KDM2B can bind throughout the genome by its CxxC-binding domain,^[@bib24]^ and its catalytic demethylase activity on the H3K36me2/me3 residues is associated with gene repression. However, the genes that are regulated in favor of apoptosis in GBM cells are unknown. To this end, we conducted RNA-sequencing (RNA-seq) analysis on shControl and shKDM2B cells. *KDM2B* expression levels were downregulated in shKDM2B cells compared with shControl cells −3.049-fold (adjusted *P*-value 7.51E−91), attesting to the validity of RNA-seq experiments. In total, 2457 genes were differentially expressed between shControl and shKDM2B cells. Ingenuity pathway analysis of these genes revealed that 'cell death and survival' and 'cellular growth and proliferation' pathways were among the top 10 that were significantly different between shControl and shKDM2B cells with --log *P*-values of 25--30 ([Figure 4a](#fig4){ref-type="fig"}). Five out of the top 10 altered functions were categorized as 'cell death and survival' where these gene sets were predicted to be activated with activation *z*-scores of \>2.5 ([Figure 4b](#fig4){ref-type="fig"}). In addition, top 2 of these 10 categories were 'cell death and survival', suggesting that KDM2B silencing alters cell death- and survival-associated genes in GBM cells. We were particularly interested in the cell death-related pathway in the shKDM2B cells, and therefore focused our attention on individual differentially expressed genes and generated a custom list of apoptosis-related genes that, when deregulated, can affect the outcome of TRAIL-induced apoptosis ([Figure 4c](#fig4){ref-type="fig"}). When we sorted the list of apoptosis-related genes based on differential expression, the most significant change was on *HRK* gene, with striking 16-fold induction in shKDM2B cells compared with shControl cells. The expression of other key apoptosis players, including *CASP7, DAPK1, BAK1* and *TNFRSF10A (DR4)* were also induced between 1.5- to 3.7-fold ([Figure 4c](#fig4){ref-type="fig"}). To validate the RNA-seq results, we designed gene-specific primers ([Supplementary Table 1](#sup1){ref-type="supplementary-material"}), performed qRT-PCR and confirmed the differential expression of *HRK* ([Figure 4d](#fig4){ref-type="fig"}). Further, qRT-PCR analyses of apoptosis-related genes showed proapoptotic *DAPK1, CARD16, TNFAIP2, TNFRSF10c, BCL2L11, BMF, XAF1, TNFRSF14, TNFRSF11B, APAF1, CASP1, CASP7, BCL2L2* and *BAK1* were induced, while antiapoptotic *MCL1, BCL11B* and *BCL2A1* were repressed ([Supplementary Figure 9](#sup1){ref-type="supplementary-material"}). These results suggest that silencing KDM2B causes genome-wide transcriptional changes and specifically alters apoptotic machinery in favor of apoptosis in GBM cells. Because *HRK* was the top induced gene in shKDM2B cells, we further tested its functional association with KDM2B. HRK is a BH3-only Bcl-2 family member that can antagonize antiapoptotic proteins Bcl-2 and Bcl-xL.^[@bib25]^ While *HRK*'s role is mostly studied in nervous system,^[@bib26],\ [@bib27]^ its role in cancers, especially in GBM is not well defined.^[@bib25]^ To assess the increased HRK expression upon KDM2B loss, we performed western blotting and observed increased HRK protein levels in shKDM2B cells compared with controls ([Figure 4e](#fig4){ref-type="fig"}). To then test the role of HRK, we used shRNA targeting *HRK* and observed significant reduction in its mRNA expression ([Figure 4f](#fig4){ref-type="fig"}). Silencing of HRK led to significant but partial recovery from TRAIL-induced death in both shControl and shKDM2B cells ([Figure 4g](#fig4){ref-type="fig"}). These results suggest that *HRK* is potentially a downstream gene regulated by KDM2B and may facilitate TRAIL response in GBM cells. Silencing of H3K36-specific HMTs leads to the opposite phenotype of KDM2B silencing in GBM cell apoptosis --------------------------------------------------------------------------------------------------------- H3K36-specific HMTs that catalyze the opposite function of KDM2B are known.^[@bib28]^ Our shRNA library included four of these HMTs, namely SETD2, NSD1, ASH1L and SMYD2. As they carry out the opposite enzymatic activity of KDM2B, we assessed their effects on apoptotic response. We observed significant reduction in gene expression with each corresponding shRNA ([Figure 5a](#fig5){ref-type="fig"}). Individual knockdown of each enzyme did not lead to global differences in H3K36me2 or H3K36me3 levels as assessed by western blotting of histone extracts ([Supplementary Figure 10](#sup1){ref-type="supplementary-material"}), suggesting that H3K36me regulation at specific loci might be responsible for the observed phenotypic changes. While silencing of KDM2B followed by TRAIL treatment led to increased cell death, silencing of SETD2, NSD1, ASH1L, and SMYD2 individually led to significantly reduced cell death ([Figure 5b](#fig5){ref-type="fig"}). These results suggest that H3K36 methylation and demethylation at specific loci of apoptosis-related genes potentially regulate the apoptotic response of GBM cells to TRAIL. Loss of KDM2B attenuates tumor growth *in vivo* ----------------------------------------------- To examine the effects of KDM2B loss on tumor growth *in vivo*, we generated shControl and shKDM2B cells expressing both firefly luciferase (Fluc) and mCherry. The bioluminescence signals of shControl-Fluc-mCh and shKDM2B-Fluc-mCh cells were comparable *in vitro* ([Figure 6a](#fig6){ref-type="fig"}). To assess the effect of KDM2B loss on long-term GBM growth before *in vivo* implantation, we performed real-time cell analysis of shControl-Fluc-mCh and shKDM2B-Fluc-mCh cells for 200 h. We observed that knockdown of KDM2B alone reduces the proliferation rate of GBM cells in long term ([Figure 6b](#fig6){ref-type="fig"}), while within short term of 24 h the proliferation rate is comparable with control cells, confirming our previous observation ([Figure 2d](#fig2){ref-type="fig"}). Long-term analysis of tumor growth for 30 days with noninvasive bioluminescence imaging revealed that loss of KDM2B also attenuates tumor growth *in vivo* ([Figures 6c and d](#fig6){ref-type="fig"}). Immunofluorescence and histological analysis of the tumors showed that loss of KDM2B led to reduced angiogenic capacity of the tumors as assessed by regular H&E and VEGF staining. Moreover, there were significantly more annexin-V-positive tumor cells within shKDM2B tumors compared with controls ([Figures 6e and f](#fig6){ref-type="fig"}). Taken together, these results suggest that KDM2B silencing led to decreased tumor growth in long term, which may be attributable to the increased basal levels of apoptosis. Discussion ========== In this study, we have used an unbiased shRNA screen to show that the deregulation of a CME in a GBM cell line can cause sensitization to TRAIL through induction of proapoptotic machinery. Several lines of evidence support this argument. First, loss of KDM2B cooperated with TRAIL to reduce GBM cell viability and augmented the apoptotic response as assessed by the hallmarks of apoptosis upon TRAIL treatment. Further, whole-genome transcriptome analysis suggested that endogenous levels of KDM2B in GBM is correlated with suppressed state of proapoptotic machinery, whose knockdown results in induction of apoptosis by inducing proapoptotic genes and repressing antiapoptotic genes. *HRK* was the top induced gene upon KDM2B loss and its silencing partially recovered the observed sensitization phenotype, suggesting that *HRK* might be a novel and direct target of KDM2B-mediated apoptosis regulation in GBM cells. In our screen, the majority of the shRNAs (110/125) targeting CMEs did not affect TRAIL response on their own. However, some were able to prime GBM cells for apoptosis. Although our shRNA screen targeted diverse types of CMEs, such as writers (DNMTs and HMTs), erasers (HDMs), readers (methyl-DNA-binding proteins) and other chromatin regulatory proteins such as polycomb-group proteins (PRCs), it did not cover all types and members of CMEs. For example, HATs and HDACs were not included in the shRNA screen. It will be of interest to see whether shRNA-mediated knockdown of HDACs would prime tumor cells for apoptosis as their small drug inhibition were previously reported to be implicated in TRAIL sensitization.^[@bib29]^ One major hit from our screen was KDM2B. KDM2B is a H3K36me2 demethylase that is indicated as a regulator of cell growth.^[@bib22],\ [@bib30],\ [@bib31]^ KDM2B, by its CxxC-ZF domain, can bind to CpG islands and can catalyze demethylation of H3K36 or H3K4 residues through its JmjC domain.^[@bib24]^ KDM2B has been shown to be required for the tumorigenesis of acute myeloid leukemia, where it acts as an oncogene through silencing *p15/Ink4b*.^[@bib32]^ Similarly, KDM2B has been shown to regulate self-renewal capacity of cancer and cancer stem cells in hematological and pancreatic cancers, and its expression is positively correlated with advanced tumor grade.^[@bib21],\ [@bib22]^ A recent study displayed the protumorigenic role of KDM2B in a broad range of human tumors.^[@bib33]^ To our knowledge, KDM2B has not been studied previously in brain malignancies and the present work is the first study to display its novel oncogenic role in GBMs. Interestingly, we found that shRNAs targeting the H3K36-specific HMTs, namely SETD2, NSD1, SMYD2, ASH1L, known to conduct the opposite enzymatic activity of KDM2B by catalyzing the addition of methyl groups to H3K36 residues, conferred GBM cells partly resistant to TRAIL-induced apoptosis. Therefore, we believe that H3K36 methylation can be a central regulator of TRAIL-induced apoptosis in GBM cells, and it would be interesting to dissect out the exact molecular mechanism by testing the function of HMTs in combination, and conducting chromatin analysis in the future. We show that short-term KDM2B loss by itself does not cause major changes in cell survival; however, long-term effect of KDM2B suppression is observed in our *in vivo* xenograft models. This suggests that while short-term loss of KDM2B can alter the epigenetic landscape of tumor cells in favor of apoptosis, long-term consequences of the epigenome alterations can reduce proliferation and be detrimental without additional external apoptotic stimuli. Therefore, it will be important to uncouple the exact mechanisms of how KDM2B regulates proliferation rate, survival and apoptosis in GBM cells. In this study, we focused on the examination of apoptotic regulation and showed that KDM2B can alter apoptotic response of GBM cells to TRAIL-induced apoptosis. This is consistent with a previous study showing that KDM2B regulated the apoptotic response of pancreatic cancer cells.^[@bib34]^ Our consistent results with multiple GBM cell lines and multiple proapoptotic agents such as TRAIL and BH3 mimetics suggest that KDM2B could be a global regulator of cell death in GBM cells. How KDM2B might regulate apoptotic priming in GBM cells was also examined in this study by comparing the transcription profiles of shControl and shKDM2B cells. Accordingly, among many apoptosis-related genes taking place in TRAIL pathway, proapoptotic genes such as *DR4* and executioner *caspase-7* were induced. Interestingly, *HRK*, a BH3-only Bcl-2 family member, that can antagonize antiapoptotic proteins Bcl-2 and Bcl-xL,^[@bib25]^ was the top induced gene. As such, silencing of *HRK* partly recovered the apoptotic sensitization, suggesting that *HRK* is directly linked to KDM2B-mediated regulation. There were other genes that were modulated by KDM2B loss, such as *DAPK1* or *BMF*, which could also be partly responsible for apoptotic sensitization. DAPK1 is a proapoptotic tumor suppressor that has been shown to be inactivated in various cancer settings,^[@bib35],\ [@bib36]^ whereas antiapoptotic BH3-only BMF can also sensitize cells via directly interacting with Bcl-2, Bcl-xL, Bcl-w and Mcl-1.^[@bib37]^ On the other hand, knockdown of KDM2B led to suppression of antiapoptotic genes *MCL1*, *BCL11B* and *BCL2A1*. This could be through an indirect cascade of events or a direct consequence of partial loss of KDM2B.^[@bib22]^ Interestingly, BCL11B silencing has been demonstrated to induce apoptosis in malignancies.^[@bib38]^ BCL2A1 is overexpressed in various cancers and its upregulation is characterized as a major resistance factor against proapoptotic therapy in cancer.^[@bib39]^ MCL1 is one of the key gatekeepers regulating apoptosis and has been targeted by small drug inhibitors to sensitize cancer cells.^[@bib40],\ [@bib41]^ Further research is warranted to understand how KDM2B regulates the expression of these apoptosis regulators at the chromatin level. Taken together, we show that KDM2B regulates the apoptotic response of GBM cells by turning on the apoptotic machinery that is otherwise dormant. Therefore, combining epigenetic reprogramming drugs with standard chemotherapeutics or with proapoptotic therapies are promising therapeutic approaches. To this end, previously, several agents have been combined with TRAIL, including HDAC inhibitors. Interestingly, bortezomib,^[@bib42]^ MS-275,^[@bib11]^ and a recent drug identified by our group, mitoxantrone,^[@bib43]^ have all suppressed the expression levels of KDM2B in favor of cell death ([Supplementary Figure 11](#sup1){ref-type="supplementary-material"}). Therefore, it is plausible to suggest that regulators of apoptosis might be acting through regulating the central epigenetic machinery in favor of enhanced cell death ([Figure 7](#fig7){ref-type="fig"}). GBM is a complex cancer with extremely poor prognosis and limited therapeutic options. Therefore, designing new and effective approaches for GBM is necessary. Drugs targeting CMEs are emerging as anticancer therapeutics.^[@bib44]^ Our results corroborate the notion that combinatorial uses of epigenetic drugs with proapoptotic agents hold great promise. To this end, specific chemical inhibitors for CMEs such as a KDM2B inhibitor can yield effective combination therapies for subgroups of GBM patients. Materials and methods ===================== Cell lines ---------- U87MG GBM cells and 293T cells were purchased and authenticated from ATCC (Manassas, VA, USA). All cells were cultured in DMEM medium supplemented with 10% FBS and 1% penicillin--streptomycin (Gibco, Grand Island, NY, USA) at 37 °C in a humidified chamber with 5% CO~2~ as described.^[@bib11]^ Puromycin selection was applied at a final concentration of 1 *μ*g/ml for 3--7 days. Reagents -------- TRAIL was commercially supplied (Enzo Life Sciences, Farmingdale, NY, USA) or produced from 293T cells as described.^[@bib45]^ Briefly, the extracellular portion of human TRAIL (amino acids 95--281) was cloned into pBabe retroviral expression vectors to produce functionally active TRAIL,^[@bib46]^ which was quantified by TRAIL ELISA according to the manufacturer's instructions (Abcam, Cambridge, UK). Caspase inhibitors (BD Pharmingen, San Diego, CA, USA) were as follows: Z-IETD-FMK (caspase-8 inhibitor), Z-VAD-FMK (general caspase inhibitor), Z-FA-FMK (negative control), Z-LEHD-FMK (caspase-9 inhibitor) and Ac-DEVD-CHO (caspase-3/7 inhibitor). Bcl-2, Bcl-xL inhibitors ABT-263 and ABT-737 were purchased from Cayman Chemicals (Ann Arbor, MI, USA) Generation of the shRNA library ------------------------------- One hundred and twenty-five of total shRNAs targeting chromatin-modifying proteins and HRK were designed using the RNAi Codex program. Accordingly, two to three shRNAs were designed using 97-mer oligos and cloning into pSMP vector system. The designed oligos were PCR amplified using the following primer pairs: (forward: 5′-GATGGCTGCTCGAGAAGGTATATTGCTGTTGACAGTGAGCG-3′ reverse: 5′-GTCTAGAGGAATTCCGAGGCAGTAGGC-3′). PCR products were gel purified, digested with *Eco*RI and *Xho*I and ligated into the retroviral pSMP vector as described,^[@bib20]^ and verified by sequencing. shRNA targeting the Fluc was used as shControl. The sequences of selected shRNAs are listed in [Supplementary Table 2](#sup1){ref-type="supplementary-material"}; viral packaging was conducted as described.^[@bib20]^ The 293T cells were seeded as 2.5 × 10^6^ cells per plate and transfected with shRNA, VSV-G and pUMVC plasmids with Fugene transfection reagent (Promega, Madison, WI, USA). The virus containing media was collected, filtered through 0.45 *μ*m filters (Millipore, Billerica, MA, USA) and stored at −80 °C. shRNA screen ------------ U87MG cells were seeded as 2500 cells per well into black 96-well plates. Next day, cells were infected with 80 *μ*l of viral supernatants with protamine sulfate (10 *μ*g/ml) in triplicates. After 2 days, puromycin (1 *μ*g/ml) was added for 3 days. Following selection, cells were treated with TRAIL (100 ng/ml) or control solution for 24 h, followed by cell viability assays. shRNAs that caused reduction in cell viability, as well as shRNAs that augmented TRAIL response in at least two independent experiments, were identified and further characterized. Cloning ------- The 4617 bp ORF of KDM2B in pCAG-puro, including 2x Strep-tag II and 1x FLAG-tag at its N terminus, was cloned by the flanking *Eco*RI cut sites into retroviral pBabe-puro mammalian expression vector. Cell viability and caspase activity assays ------------------------------------------ Viability of the cells was measured using ATP-based Cell Titer Glo assays (Promega) according to the manufacturer's instructions using a plate reader (BioTek's Synergy H1, Winooski, VT, USA). All experiments were performed in triplicates. To further validate KDM2B phenotype, stable cell lines of shKDM2B or shControl cells were expanded for three to four passages. Accordingly, 10 000 cells per well were seeded in triplicates and treated with TRAIL (0--100 ng/ml) for 24 h. To test caspase inhibition, each caspase inhibitor was used at 20 *μ*M final concentration; cells were cotreated with inhibitors and TRAIL. For caspase activity measurements, cells were seeded in 96-well plates as 10 000 cells per well in triplicates, treated with TRAIL (0--100 ng/ml) for 3 h and caspase-8 and -3/7 activities were measured using Caspase Glo (Promega) assays according to the manufacturer's instructions. Real-time cell growth analysis ------------------------------ xCELLigence RTCA SP Station and Analyzer (ACEA Biosciences, San Diego, CA, USA) was used for real-time cell growth analysis. The 1000--5000 U87MG cells per well were seeded, and impedance of the wells were measured with 20 min intervals for 24 or 202 h. The amount of cell growth was analyzed and plotted using the RTCA Software (ACEA Biosciences). Live-cell microscopy -------------------- All live-cell imaging experiments were carried out by Olympus Xcellence Pro inverted microscope (Center Valley, PA, USA) with a × 10 air objective in a chamber at 37 °C, supplied with 5% CO~2~. Time-lapse images were captured right after TRAIL treatment for 24 h as 1440 frames per minute. Five random positions were recorded per sample per well. The number of apoptotic bodies was counted on each window using the ImageJ Software (NIH Image, Bethesda, MD, USA). Particles having size (pixel\^2) of 10 to infinity and circularity of 0.60--1.00 were denoted as apoptotic bodies on each frame. qRT-PCR analysis ---------------- RNA isolation and cDNA synthesis were performed as described.^[@bib43]^ qRT-PCRs was carried out using the primers in [Supplementary Table 1](#sup1){ref-type="supplementary-material"}. Details of qRT-PCR are given in [Supplementary Information](#sup1){ref-type="supplementary-material"}. Cell lysate preparation, histone extraction and immunoblotting -------------------------------------------------------------- Cell lysate preparations were carried out as described previously.^[@bib43]^ Details of histone extraction and immunoblotting are given in [Supplementary Information](#sup1){ref-type="supplementary-material"}. *In vivo* experiments --------------------- All *in vivo* experiments were approved by the institution review boards of Koç University (HADYEK no. 2014-22). U87MG cells stably expressing shControl or shKDM2B were used for *in vivo* experiments. Cells were supertransduced with lentiviral vectors encoding Fluc and mCherry as described.^[@bib46]^ mCherry expression was verified with red fluorescence, and Fluc activity was verified with *in vitro* luminescence assays. Accordingly, increasing number of cells (0--50 000/well) were seeded on 96-well plates and incubated with [d]{.smallcaps}-luciferin (10 *μ*g/ml) for 10 min. Bioluminescence was measured by BioTek's Synergy H1 plate reader. A 6--8-week-old SCID mice were used for tumor implantation. A total of 2.5 × 10^6^ cells that stably expressed Fluc and mCherry were injected in 50 *μ*l PBS subcutaneously into seven SCID mice. One day after injection, tumor cells were visualized by Fluc imaging using IVIS Lumina III equipment (Perkin-Elmer, Rodgau, Germany) by injecting mice with 150 *μ*g/g body weight of [d]{.smallcaps}-luciferin (Biotium, Fremont, CA, USA) intraperitoneally. Tumor progression was monitored with sequential Fluc bioluminescence activity over a period of 1 month. All imaging experiments were performed under isoflurane anesthesia. Following data acquisition, mean, standard deviation and sum of the photon counts in the regions of interest were calculated and plotted. At the end, the tumors were dissected and analyzed with immunohistochemistry. Immunohistochemistry -------------------- Sections of tumors were analyzed by immunohistochemistry as described.^[@bib46]^ Details are given in [Supplementary Information](#sup1){ref-type="supplementary-material"}. RNA-seq and analysis -------------------- Total RNAs of shControl and shKDM2B cells were isolated, RNA-seq library for each sample was prepared based on protocols on Illumina HiSeq 2500 to generate 50 bp single-end reads. The data discussed in this publication have been deposited in NCBI\'s Gene Expression Omnibus,^[@bib47]^ and are accessible through GEO series accession number GSE81043 (<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE81043>). Differentially expressed genes between shKDM2B and shControl were detected using DESeq,^[@bib48]^ the complex biological processes induced by shKDM2B as compared with shControl were examined in the context of functional groups with Ingenuity Pathways Analysis, a web-delivered commercial application. Details of library preparation, statistical and functional analysis of differentially expressed genes are explained in [Supplementary Information](#sup1){ref-type="supplementary-material"}. Statistical analysis -------------------- Data were analyzed by Student\'s *t*-test when comparing two groups. Data were plotted as mean±s.e.m. and *P*-values were calculated. \*, \*\* and \*\*\* denotes *P*\<0.05, *P*\<0.01 and *P*\<0.001 on the figures. We thank Dr Halil Bayraktar for technical help with live-cell imaging. We also thank Weill Cornell Medicine Epigenomics Core for RNA-sequencing services. This work was supported by TUBITAK-115S251 (to TBO), Marie Curie CIG-618673 (to TBO) and TUBITAK-BIDEB-2210 (to ICK). We also gratefully acknowledge use of the services and facilities of the Koç University Research Center for Translational Medicine (KUTTAM), equally funded by the Republic of Turkey Ministry of Development Research Infrastructure Support Program. Findings, opinions or points of view expressed in this article do not necessarily represent the official position or policies of the Ministry of Development. [Supplementary Information](#sup1){ref-type="supplementary-material"} accompanies this paper on Cell Death and Disease website (http://www.nature.com/cddis) Edited by T Kaufmann The authors declare no conflict of interest. Supplementary Material {#sup1} ====================== ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. {#fig1} {#fig2} {#fig3} {#fig4} {#fig5} {#fig6} {#fig7}

1. Introduction {#sec1-ijerph-17-04838} =============== Recent studies show a rise in overweight and obesity among children and adolescents to 18% worldwide in 2016 \[[@B1-ijerph-17-04838]\], with one in five school-aged children in Europe being overweight or obese \[[@B2-ijerph-17-04838],[@B3-ijerph-17-04838]\]. Also in The Netherlands, childhood obesity rates are high. In 2019, 12% of children aged 4 to 12 were overweight or obese \[[@B4-ijerph-17-04838]\]. Overweight or obesity in childhood is especially problematic, since this often continues into adolescence and adulthood and is related to an increased risk of negative health consequences, such as type II diabetes, hypertension, respiratory disease and various types of cancer \[[@B2-ijerph-17-04838],[@B3-ijerph-17-04838],[@B5-ijerph-17-04838]\]. Overweight and obesity are preventable and treatable \[[@B6-ijerph-17-04838]\], especially at a young age \[[@B5-ijerph-17-04838],[@B7-ijerph-17-04838]\], by improving healthy nutrition behaviours and increasing physical activity (PA) (also called energy balance-related behaviours (EBRBs)) \[[@B8-ijerph-17-04838],[@B9-ijerph-17-04838],[@B10-ijerph-17-04838]\]. To target children's EBRBs \[[@B8-ijerph-17-04838],[@B11-ijerph-17-04838]\], multiple settings \[[@B12-ijerph-17-04838]\], such as the school \[[@B13-ijerph-17-04838]\] and home environment should be involved in interventions \[[@B8-ijerph-17-04838],[@B13-ijerph-17-04838]\]. Since children spend a large proportion of their weekdays at school, and schools reach many children, schools have been a popular intervention setting for decades \[[@B13-ijerph-17-04838],[@B14-ijerph-17-04838],[@B15-ijerph-17-04838]\]. However, the influence of the home setting on children's EBRBs is profound, especially for young children, since parents determine the food availability at home, and influence children's nutrition and PA behaviour by practices like modelling and rule setting \[[@B16-ijerph-17-04838],[@B17-ijerph-17-04838],[@B18-ijerph-17-04838],[@B19-ijerph-17-04838]\]. Therefore it is important to also focus on the home setting when implementing school-based energy balance-related interventions \[[@B15-ijerph-17-04838],[@B17-ijerph-17-04838]\]. Most effective changes in the home setting are accomplished through interventions with direct parental involvement (e.g., parents attending educational sessions, or counselling sessions) \[[@B13-ijerph-17-04838],[@B16-ijerph-17-04838],[@B19-ijerph-17-04838],[@B20-ijerph-17-04838],[@B21-ijerph-17-04838]\]. However, directly involving parents is often resource- and labour-intensive, making these types of interventions less feasible \[[@B20-ijerph-17-04838]\]. Also, the recruitment \[[@B22-ijerph-17-04838]\] and prolonged engagement of parents in these types of interventions have proven to be challenging \[[@B3-ijerph-17-04838],[@B22-ijerph-17-04838],[@B23-ijerph-17-04838]\]. Only about one-third of invited families participate in any intervention activity \[[@B3-ijerph-17-04838],[@B24-ijerph-17-04838]\], 40 to 60% of whom drop out \[[@B24-ijerph-17-04838]\]. In addition, the participants of parental involvement interventions tend to be mostly high socioeconomic status (SES) parents \[[@B25-ijerph-17-04838]\], and it is particularly challenging to engage parents with a low SES \[[@B3-ijerph-17-04838],[@B25-ijerph-17-04838]\], a lower educational level, single parents and those of ethnic minority groups \[[@B24-ijerph-17-04838]\]. This is discouraging, since these parents/families are most in need of interventions. For example, research from a large Dutch cohort study has shown that there are socioeconomic and ethnic inequalities in child health \[[@B26-ijerph-17-04838]\]. At a young age, non-Western children were more likely to be overweight compared to Dutch children, with mother's educational level being one of the contributing factors in explaining this higher prevalence in overweight in these children \[[@B27-ijerph-17-04838]\]. Beyond socioeconomic health disparities, child characteristics and the supportiveness of the home environment to perform healthy behaviours are also important predictors of children's PA and nutrition behaviour. For instance, children's nutrition behaviour is determined by their preference for healthy and/or unhealthy foods, and their PA and sedentary behaviour is associated with their activity preferences \[[@B28-ijerph-17-04838]\]. Also, children raised in an environment in which healthy nutrition \[[@B29-ijerph-17-04838]\] and sufficient PA \[[@B29-ijerph-17-04838],[@B30-ijerph-17-04838]\] are less valued are more at risk for developing unhealthy EBRBs. Therefore, vulnerability of the population is not restricted to well-known socioeconomic differences, but also includes specific child preferences and the health climate regarding EBRBs in the home environment. To involve more parents, and particularly those of vulnerable populations, the strategy of indirect parental involvement is an alternative. Even though indirect parental involvement is assumed to be less effective in changing health behaviours compared to direct parental involvement, it can lead to greater adoption and implementation rates \[[@B21-ijerph-17-04838]\]. In indirect parental involvement, parents are engaged in a way that the intervention implementers do not communicate or engage directly (i.e., face to face, or personally) with them. Instead, parents are informed via school media, or children function as the messenger. Examples of indirect parental involvement are the provision of information (newsletters, tip sheets), invitations to participate in or attend activities (events, educational sessions) and prompts or assignments directed at the child and/or parent with the aim of involving parents \[[@B19-ijerph-17-04838]\]. Previous research has shown that some of these strategies are more promising than others, for example prompting children to engage in intervention activities together with their parent(s) seems more promising than providing newsletters or invitations for optional intervention activities \[[@B17-ijerph-17-04838],[@B19-ijerph-17-04838]\]. The difference between these strategies could be described as passive (not requiring a specific action or response on the part of the parents) indirect involvement versus active (performing activities) indirect involvement. To our knowledge, there is little to no research on the participation rates of the strategy of the latter type of indirect parental involvement in school-based interventions, while this information is crucial to be able to draw conclusions on potential strategies to involve parents in (school-based) interventions \[[@B21-ijerph-17-04838]\]. Given the fact that children of vulnerable populations are less active, more sedentary and have unhealthier diets \[[@B31-ijerph-17-04838],[@B32-ijerph-17-04838],[@B33-ijerph-17-04838]\], it is possible that these children and their parents are less interested in energy balance-related interventions. Therefore, gaining insight in participation rates of children and parents in interventions using an indirect parental involvement strategy is needed. Moreover, empirical evidence is lacking on who is engaged in this strategy and therefore, it is warranted to know whether vulnerable children and parents engage in these interventions using this strategy. In the current study, we evaluated the potential of Challenge Me to engage children and parents, especially those of vulnerable populations. Challenge Me was a parental involvement intervention in which children are challenged to perform PA and nutrition-related activities by themselves and with their parents. The main aim of the current study is to examine whether conducting challenges is a feasible intervention strategy to engage children and parents in school-based energy balance-related interventions. With engagement, we refer to participation in the intervention, i.e., performing the intervention activities, and not enrolment. Additionally, a second aim is to gain insight in whether children in need for improvements, i.e., vulnerable populations, are engaged in the intervention using this indirect-involvement strategy. For this aim, we focussed on child demographics, child characteristics (preferences) and the climate towards health behaviours in the home environment. 2. Materials and Methods {#sec2-ijerph-17-04838} ======================== 2.1. Design {#sec2dot1-ijerph-17-04838} ----------- An exploratory cross-sectional study design was conducted. The Medical Ethics Committee of the Maastricht University Medical Centre and Maastricht University provided ethical approval for this study (METC163027, national number: NL58554.068.16). 2.2. Challenge Me Intervention {#sec2dot2-ijerph-17-04838} ------------------------------ ### 2.2.1. Intervention Development {#sec2dot2dot1-ijerph-17-04838} The intervention, called Challenge Me, challenged children to perform PA and nutrition-related challenges together with their parent(s) or guardian(s). Challenge Me has been developed as part of a larger evaluation \[[@B34-ijerph-17-04838]\], and is specifically focused on improving parental involvement. In the larger evaluation study, eight primary schools, located in low SES neighbourhoods (based on scoring of The Netherlands Institute for Social Research \[[@B35-ijerph-17-04838]\]) in a city in the south of The Netherlands, participated. During the development and implementation of Challenge Me in the local context, researchers collaborated with a local youth work organisation. This local organisation developed the challenges, with input from local health, education and sports professionals. Challenge Me consisted of several easy-to-perform activities, introduced weekly at school by people locally well-known by means of instruction videos (i.e., the schools' own PE teacher, local free runners, a famous vlogger and a television chef from the region). A selection of the challenges was designed to involve parents. The local nature of the intervention enhanced a fit with the schools' local context, which we anticipated would improve programme use, recognition and commitment by the children and consequently the involvement of parent(s). The result of this design process was an intervention consisting of a total of eighteen challenges ([Table 1](#ijerph-17-04838-t001){ref-type="table"}). Challenges were categorised in 'child-only' (nine challenges performed by a child on his/her own, together with other children or in class) and 'parental involvement' (nine challenges that required parental assistance). Challenges could further be categorised as PA-focussed (e.g., 'Run as far as possible with your entire class, in 2 min'/ 'Go exercise for 30 min together with your parent(s)/guardian(s)'), or nutrition-focussed (e.g., 'Find a healthy dinner recipe and bring this with you to school'/'Make a healthy breakfast for someone in your household'). To ensure active involvement of the children and parents in the intervention, children had to provide proof of completing a challenge, which was specified per challenge (e.g., by taking a picture of their exercise). The class that performed the most challenges (corrected for the number of children in each class) was awarded a prize. The prize was a PA workshop, provided by the local youth work organisation. The local youth work organisation distributed the materials. The materials included a Challenge Me poster on which the various challenges were listed ([Appendix A](#app1-ijerph-17-04838){ref-type="app"}, [Figure A1](#ijerph-17-04838-f0A1){ref-type="fig"}), an instruction form for the teacher, stickers (used to mark a challenge as completed) and four instruction videos. The local youth organisation also appointed an employee specifically tasked with the implementation and monitoring of Challenge Me. This local youth worker also provided all necessary instructions in participating classes, to limit additional workload of teachers. ### 2.2.2. Implementation {#sec2dot2dot2-ijerph-17-04838} Recruitment for the intervention took place at school level. Intervention schools (*N* = 8) already participated in a larger evaluation study \[[@B34-ijerph-17-04838]\]. These schools were invited to implement the Challenge Me intervention. Of these, four school decided to participate. Four schools declined to participate because of time constraints. Teachers of participating schools were orally informed by the local youth worker about Challenge Me and its aims, and received instructions on how to perform the intervention. The same youth worker also informed the children about the intervention and monitored the progress of the intervention. The participating schools implemented the intervention in grade six to eight (children aged 9 to 12), except for one. This school only implemented the intervention in grade eight. Each class received a Challenge Me poster. Challenges were expected to be performed during a four-week period (January--February 2019). Every week, the teacher showed one of the provided introduction videos provided by the local youth worker to kick-start the week. Depending on the challenge, challenges were performed at home, or in the neighbourhood. Children and parents participated in the intervention voluntarily. Based on the proof provided by the child, the teacher decided whether a challenge had been completed successfully. At the end of the four weeks, the local youth worker collected all the posters and decided which class had won. 2.3. Study Participants {#sec2dot3-ijerph-17-04838} ----------------------- Children from grades six to eight (aged 9 to 12) were eligible for inclusion in the study \[[@B34-ijerph-17-04838]\]. Parents received an informed consent form via the child \[[@B34-ijerph-17-04838]\]. In short, the informed consent form stated the purpose of the study, and how collected data would be handled. It also emphasised that participation in the study was voluntary, and that participants could always withdraw without stating any reason for doing so. For children to participate in the study, both parents had to provide written consent. This was based on the regulations of the Central Committee on Research involving Human Subjects \[[@B36-ijerph-17-04838]\]. 2.4. Data Collection {#sec2dot4-ijerph-17-04838} -------------------- For this exploratory study, data from the larger evaluation study were used \[[@B34-ijerph-17-04838]\]. Data were collected in March--April 2019, after the completion of the Challenge Me intervention in all four schools. During that period, both children and their parents received a questionnaire. Children received and filled in the questionnaire in the classroom, during regular school hours, with one researcher and at least one research assistant present to provide instructions. Parent questionnaires were handed out to the children in an envelope. Parents filled in the questionnaire at home, and returned it to school via the child. Children's height and weight were measured within the same week as the child questionnaire. ### 2.4.1. Performance of Challenges {#sec2dot4dot1-ijerph-17-04838} The posters provided data on which challenges were completed by whom. Data were pseudonymised and digitalised. Challenges were grouped in various categories ([Table 1](#ijerph-17-04838-t001){ref-type="table"})---PA: PA child-only and PA parental involvement; nutrition: nutrition child-only and nutrition parental involvement. Outcome measures were the percentage of challenges performed. ### 2.4.2. Child Characteristics {#sec2dot4dot2-ijerph-17-04838} Child characteristics were assessed via a questionnaire. Children reported their date of birth, gender (boy/girl), country of birth, country of birth of their mother and father, the number of brothers and sisters and whether they were a member of a sports club (yes/no). The child's age was calculated based on the date of birth. In addition, Western or non-Western ethnicity of the child was determined, based on the country of birth of both parents. A child was considered as having a non-Western ethnicity when at least one parent was born in a non-Western country, based on the definition of Huntington \[[@B37-ijerph-17-04838]\]. The number of siblings was determined, based on the number of brothers and/or sisters. Nutrition and PA preferences were also measured via a questionnaire, using an instrument developed by Rodenburg et al. \[[@B28-ijerph-17-04838]\]. This instrument ranks food, drink and leisure-time activity preferences by means of comparison. Items were visualized by means of an infographic of a child holding a food or drink item, or portraying an activity, accompanied by the description of the food, drink or activity. The items were compared in pairs, and children were asked to indicate the food, drink or activity of their preference. The food items were fruit, vegetables, sweet snacks (e.g., candy and cookies) and savoury snacks (e.g., crisps, nuts, cheese). The drinks items included sugar-sweetened beverages, light drinks (i.e., drinks that were artificially sweetened), tea without sugar, fruit juice and water (water was added to the items at a later stage). Leisure-time activity consisted of eight items, namely cycling, using the computer, watching television, playing sports, dancing, arts and crafts, reading and playing outside. In total, the instrument consisted of 44 comparisons. Items were re-categorised into active activities (cycling, playing sports, dancing and playing outside), and healthy nutrition (fruit, vegetables, tea without sugar and water). Even though fruit juice may be perceived as healthy \[[@B8-ijerph-17-04838]\], it is high in energy density and sugar content and could contribute to weight gain and overweight \[[@B8-ijerph-17-04838],[@B38-ijerph-17-04838]\]. Because of this controversy, fruit juice was left out of analysis. Children's height and weight were measured by trained research assistants, using a measurement protocol. Children were measured during a physical education class, wearing light sports clothes and without shoes. Standing height was measured using the Seca 213 stadiometer (Seca, Hamburg, Germany), to the nearest decimal in centimetres. The Seca 803 digital weighing scale (Seca, Hamburg, Germany) was used to measure the child's weight to the nearest decimal in kilogrammes. Children's BMI z-score was calculated from their weight and height, while adjusting for their age and gender, using a Dutch reference population \[[@B39-ijerph-17-04838]\]. ### 2.4.3. Family Characteristics {#sec2dot4dot3-ijerph-17-04838} Family characteristics were measured in the parent questionnaire, and consisted of postal code, level of education, family situation (partner/single) and Family Health Climate (FHC). The postal code was used to determine residential status score \[[@B35-ijerph-17-04838]\]. The residential status score ranged between −1.56 and 0.13 in the intervention region. The national mean residential status score is 0.2 \[[@B35-ijerph-17-04838]\]. Educational level of the parent was recoded into two categories \[[@B40-ijerph-17-04838]\]: (1) low (no education, primary school, secondary school, pre-vocational school or lower vocational education); and (2) high (higher vocational education or university). For two-parent families, a combined score was made, i.e., (1) low (both parents having a low level of education); (2) mixed (one parent low educated and one parent high educated); and (3) high (both parents having a high level of education). In the parent questionnaire, parents filled in the validated Family Health Climate scale (FHC) \[[@B29-ijerph-17-04838]\], which was translated into Dutch \[[@B41-ijerph-17-04838]\]. The FHC is a 31-item questionnaire measuring shared family perceptions and cognitions concerning health behaviour, i.e., nutrition (FHC-NU) and physical activity (FHC-PA). The climate concerning PA and healthy nutrition is further divided into four and three concepts respectively, namely FHC-NU value, cohesion, communication and consensus, and FHC-PA value, cohesion and information. Items belonging to these concepts are statements, introduced by 'In our family...', and are answered on a four-point Likert-scale ranging from one (definitely false) to four (definitely true). FHC-NU value (four items) measures the importance of healthy nutrition within a family ('... a healthy diet plays an important role in our lives') (α = 0.62). FHC-NU cohesion (five items) includes the importance of eating together as a family ('...everyone enjoys having meals together') (α = 0.82). FHC-NU communication (five items) covers how normalised it is to talk about nutrition, and how family members support each other in eating healthy ('...we remind each other to pay attention to a healthy diet') (α = 0.82). FHC-NU consensus (three items) encompasses the level of agreement among family members concerning nutrition ('...we usually agree on meals and food choices') (α = 0.79). A high FHC-NU score implies a high value attached to healthy nutrition within a family. FHC-PA value (five items) comprises the value attached by all family members to being physically active ('\...it is normal to be physically active on a regular basis') (α = 0.82). FHC-PA cohesion (five items) includes joint physical activities and experienced fun during such activities ('...we enjoy exercising together') (α = 0.88). Finally, FHC-PA information (four items) measures searching, sharing and using PA-related information as a family ('...we read newspaper or magazine articles on fitness, physical activity and exercise') (α = 0.80). The higher the FHC-PA score, the more integrated PA is within a family's daily life \[[@B29-ijerph-17-04838]\]. 2.5. Statistical Analyses {#sec2dot5-ijerph-17-04838} ------------------------- All results were analysed using SPSS 24.0 (IBM Corp., Armonk, NY, USA). Missing items of the FHC-PA and FHC nutrition scales (i.e., PA value, cohesion and information, and nutrition value, cohesion, communication and consensus) were imputed with the mean score of the other items of the same concept. Data were only imputed when a maximum of 10% of items per concept were missing. No other missing values were imputed. Descriptive statistics were used to assess child and parent engagement and child and family characteristics of the intervention participants. Associations between percentage of challenges (per category) performed and child and family characteristics were assessed by conducting linear regression analysis. First, the association between child and family characteristics as predictor variables and the percentage of challenges performed as outcome measures were analysed using bivariate linear regression analysis. The predictor variables used were the age, BMI z-score, gender, ethnicity (Western/non-Western), sports membership (member/non-member), preference PA and preference healthy nutrition of the child and the residential status score, the number of siblings, the combined educational level (low/mixed/high) and the FHC-PA and FHC-NU of the family. Second, multivariate linear regression analyses were performed for the PA-related outcome measures (i.e., percentage PA child-only challenges, PA parental involvement challenges) by simultaneously using PA-related child characteristics as predictor variables in the model, i.e., age, gender, ethnicity, sports membership and preference PA. The same was done to assess the associations between the nutrition-related outcome measures (i.e., percentage nutrition child-only challenges, nutrition parental involvement challenges) and nutrition-related child characteristics, i.e., age, gender, ethnicity, sports membership and preference healthy nutrition. To assess the association between PA-related outcome measures and PA-related family characteristics, i.e., residential status score, number of siblings, parental educational level (low/mixed/high) and FHC-PA, predictor variables were entered simultaneously in this model while being corrected for all PA-related child characteristics. The same was done for the association between nutrition-related outcomes measures and nutrition-related family characteristics. A *p*-value of \<0.05 was considered statistically significant. 3. Results {#sec3-ijerph-17-04838} ========== 3.1. Child and Family Characteristics {#sec3dot1-ijerph-17-04838} ------------------------------------- In total, 406 children were eligible for inclusion in the study. Of these, 226 children (55.7%) participated in the study. Children had a mean age of 10.9 years ([Table 2](#ijerph-17-04838-t002){ref-type="table"}). Slightly more girls (55.8%) participated than boys, and more children were of Western ethnicity (62.7%). The majority of the children were members of a sports club (77.9%). Children preferred active activities slightly more (mean score: 15.7 out of 28). Regarding nutrition, there was a slight preference for unhealthy food and drink items (mean score: 7.5 out of 14). Of the parents, 186 (82.3% out of 226) returned their questionnaire. Mainly mothers filled in the questionnaire (82.9%), and most of them had a partner (84.4%). Almost half of the families consisted of two children (47.7%). In the majority of the families (50%), both parents had a low educational level and in 29.9% of the families, both parents had a high educational level. On average, families scored 2.8 on FHC PA, and 3.1 on FHC nutrition (range FHC score: 1 to 4). 3.2. Performance of Challenges {#sec3dot2-ijerph-17-04838} ------------------------------ Of the 226 children participating in this study, all children (100%) performed at least one challenge and 211 children (93%) performed at least one parental involvement challenge ([Table 3](#ijerph-17-04838-t003){ref-type="table"}). Overall, PA challenges were performed more often than nutrition challenges. Only two children (0.9%) did not perform any of the PA-related challenges, whereas 85 children (37.6%) did not perform any of the nutrition-related challenges. Of the PA child-only challenges, the 'Run as far as possible with your entire class, in 2 min' challenge was completed by most children (*N* = 171, 75.7%) ([Appendix B](#app2-ijerph-17-04838){ref-type="app"}, [Table A1](#ijerph-17-04838-t0A1){ref-type="table"}). Of the PA parental involvement challenges, the 'Come to school using a means of transport other than a car' challenge was completed by most children (*N* = 176, 77.9%). Regarding nutrition challenges, challenges involving parents were performed more often compared to child-only challenges, with 126 children (55.8%) performing at least one nutrition parental involvement challenge compared to 65 children, and 28.3%, performing at least one child-only nutrition challenge. Of the nutrition parental involvement challenges, 'Make a work of art out of a piece of fruit' was performed most often (34.5%), and of the nutrition child-only challenges 'Find a healthy dinner recipe and bring this with you to school' was performed most often (24.8%). 3.3. Associations between Child and Family Characteristics and Challenges Performed {#sec3dot3-ijerph-17-04838} ----------------------------------------------------------------------------------- Child characteristics significantly associated with the number of challenges performed were age, ethnicity and sports membership ([Table 4](#ijerph-17-04838-t004){ref-type="table"} and [Table 5](#ijerph-17-04838-t005){ref-type="table"}). Younger children performed more challenges, specifically PA child-only challenges and nutrition challenges involving parents ([Table 4](#ijerph-17-04838-t004){ref-type="table"}). Children of Western ethnicity performed more PA parental involvement. Children who were a member of a sports club performed more PA child-only challenges. Characteristics of the family significantly associated with the challenges performed by the child were residential status score, the number of siblings and a mixed parental level of education ([Table 4](#ijerph-17-04838-t004){ref-type="table"} and [Table 6](#ijerph-17-04838-t006){ref-type="table"}). Children raised in high SES neighbourhoods (higher residential status score) were more likely to perform PA child-only challenges, and PA parental involvement challenges were performed more often by children of larger families. When compared to a low level of parental education, children of families with a mixed level of parental education completed more nutrition challenges, mainly child-only. This significant association was not found when comparing a mixed level of parental education to a high level of parental education. 4. Discussion {#sec4-ijerph-17-04838} ============= This study investigated the parent and child engagement in Challenge Me, an intervention targeting parents indirectly in a school-based intervention using PA- and nutrition-related challenges for children. Many children (*N* = 406) were exposed to the intervention. Of the 226 study participants, all (100%) participated in the intervention. Of these children, 93% performed at least one parental involvement challenge. This study showed that there was no difference in engagement in the intervention activities by children with a lower or higher BMI z-score, children with a higher or lower preference concerning active activities or healthy nutrition, or boys or girls. Also, children of families with an unhealthier or healthier climate concerning PA and nutrition did not differ in engagement in the intervention. Nevertheless, other characteristics associated with the performance of certain types of challenges could be distinguished. For example, PA challenges were performed more often by children who were already a member of a sports club, and younger children. These active children might be more used to performing PA after school time and be more motivated to do so, and are thus more likely to perform the challenges compared to children who are not used to sports or PA activities that much. The fact that younger children performed more PA activities is not surprising, given the age-related decline in PA that is visible in children in high-income countries \[[@B42-ijerph-17-04838]\], which already starts at the age of seven \[[@B43-ijerph-17-04838]\]. Western children performed more PA parental involvement challenges compared to non-Western children. This is in line with previous research, showing that non-Western children perform significantly less PA (overall) than Western children \[[@B44-ijerph-17-04838],[@B45-ijerph-17-04838],[@B46-ijerph-17-04838]\]. However, additional analyses showed that this association between child ethnicity and the performance of PA parental involvement challenges was mediated by sports membership. The association between ethnicity and sports membership was significant (β = −0.23, 95% CI −0.31; −0.09, *p* \< 0.01), with a higher percentage of children of Western ethnicity being a member of a sports club compared to children of non-Western ethnicity (83.0% versus 65.5%). Potentially, cultural differences between Western and non-Western families play a role in whether or not children participate in sports. It is unknown whether such cultural differences in PA and performance of PA and sports as a family exist in The Netherlands \[[@B43-ijerph-17-04838]\]. Our results also showed that children from larger families (i.e., having more siblings) performed more PA parental involvement challenges. Potentially, multiple siblings participated in the intervention, which meant a higher exposure to the challenges within the family. However, further research to support these findings needs to be done. Even though the FHC measures family cognitions on nutrition and PA behaviour specifically, FHC did not show any associations with parental involvement in nutrition or PA challenges performed. This could indicate that the challenges would be suited for all families, regardless of their current PA and nutrition family climate. However, both on FHC PA and FHC nutrition, families scored relatively high and variance was low, potentially limiting the probability of finding an association. Lastly, children living in higher SES neighbourhoods performed more PA child-only challenges compared to children living in lower SES neighbourhoods. This could be explained as lower SES neighbourhoods being perceived as less safe by parents \[[@B47-ijerph-17-04838]\], and could therefore be seen as less suited for children to perform PA on their own in their neighbourhood. Additionally, previous research has shown that high SES home environments are more supportive of PA in terms of availability of equipment \[[@B48-ijerph-17-04838]\]. We had expected that these factors would also influence the association between parental education level and the performance of PA challenges, with a higher performance of challenges in families of a high parental educational level, as educational level is an indicator of SES. However, we did not find children of high-educated parents performing more PA challenges (data not shown). We did see that children of families with a mixed educational level performed more nutrition child-only challenges compared to children of low-educated parents (but not compared to high-educated parents). However, we are unable to explain this association. Even though earlier studies often found a difference in performance of PA between boys and girls \[[@B45-ijerph-17-04838],[@B49-ijerph-17-04838],[@B50-ijerph-17-04838],[@B51-ijerph-17-04838]\], our study did not show an association between gender and certain types of challenges performed. It can be that the gender-neutrality of the challenges played a role in this. While boys prefer activities like soccer and basketball, girls prefer activities like dancing, gymnastics and exercising to music \[[@B52-ijerph-17-04838]\]. However, none of the challenges had any of these elements incorporated in them. Gender-stratified analyses were performed for the association between child and family characteristics and challenges performed (per category), which showed a difference in key characteristics for boys and girls. For example, being a member of a sports club was significantly associated with the performance of PA child-only challenges for boys, yet this association was not significant for girls (data not shown). Interactions between variables and the performance of further sensitivity analysis were outside of the scope of this study, but should be studied further in the future to better understand who actively participated in the intervention and adjust the intervention accordingly. Comparing the type of challenges performed, we found that in general, the children performed more PA challenges than nutrition challenges. Potentially, children regarded the PA challenges as more fun or as easier to perform. Children performed more or less the same percentage of PA child-only challenges and PA parental involvement challenges. By contrast, the nutrition challenges that were performed more often were the parental involvement challenges. Especially younger children performed more nutrition parental involvement challenges. To perform nutrition challenges (e.g., 'Cook a healthy soup'), help from the parents was required: parents had to buy the ingredients or they needed to help with the preparation of the ingredients, e.g., the cutting of ingredients, which might be particularly true for the younger children in our study \[[@B16-ijerph-17-04838]\]. For the same reasons (e.g., parents buying groceries, and thus being in control of the food availability at home), nutrition behaviour of the child might be more dependent on and influenced by the parents than PA behaviour. As a lower level of PA \[[@B10-ijerph-17-04838],[@B53-ijerph-17-04838]\] and unhealthy nutrition \[[@B10-ijerph-17-04838]\] could lead to a negative energy balance, and consequently to overweight and/or obesity, an association between BMI z-score and performance of PA and healthy nutrition challenges was expected. However, BMI z-score did not affect the number of challenges performed, suggesting that Challenge Me would be suited for children of all weight statuses. Based on the results, it is recommended to take the above-mentioned child and family characteristics associated with certain types of challenges performed, i.e., child age, ethnicity, SES and family size, into account. For example, besides gender-neutral challenges, age-appropriate and culture-specific parent-child challenges might improve uptake of the intervention in the target group, including the vulnerable population. The intervention had potential to engage many children and parents, however, to achieve behaviour change, we hypothesize that a longer implementation period is needed. Intervention effects could be enhanced by promoting PA and healthy nutrition via various channels, and integrating the promotion of healthy behaviour into the standard school curriculum or daily routine \[[@B14-ijerph-17-04838]\]. Research is needed to study the effects of this type of indirect parental involvement interventions on children's EBRBs, on children's cognitive factors, like awareness and attitude, and on the home environment (e.g., parent-child play, healthier home climate). Strenghts and Limitations ------------------------- To our knowledge, this is one of the few studies specifically aimed at investigating the engagement of children and parents in a school-based indirect parental involvement intervention. The reach of the intervention was extensive, and response rates among the participants high, with all children and over 80% of parents returning their questionnaire. Besides exploring the engagement in the intervention, we also aimed to specifically report on the characteristics of the children and parents participating in the challenges, as this information is currently underreported yet much needed to further understand and develop parental involvement interventions. Even though our study specifically reports on the characteristics of the study participants, we did not collect enough data to compare these characteristics to the characteristics of the school population and draw conclusions on the sample representativeness. To improve the match between the intervention and the characteristics, interests and needs of the children and their families, we suggest involving the target audience when developing the challenges \[[@B54-ijerph-17-04838]\]. This was not done when developing Challenge Me, resulting in some challenges being performed poorly. For example, 'Make a vlog about healthy/unhealthy food in your surroundings' was expected to be in line with children's interest, but turned out to be an unpopular challenge. Also, our study did not provide insights in the order in which the challenges were performed. All participating classes were free to determine the order in which they wanted to perform the challenges. However, the order in which challenges were introduced by the locally well-known person could have influenced children's interest during the intervention, and subsequently their performance of the challenges. If children had already lost interest in the beginning, it is possible that later planned challenges could therefore have not been performed, regardless of their content and suitability to the target group. Challenge Me posters solely provided information on whether a challenge was performed or not, based on the evidence provided by a child to their teacher. However, it is unknown how strict teachers were in ticking off these challenges, and whether this varied between teachers. Also, it is hypothesised that teachers had a strong influence on the implementation of the intervention, and consequently on how enthusiastically children reacted to and engaged in the programme. For future research, it is recommended to gain more contextual insights in the implementation of the intervention, i.e., what was the influence of characteristics of teachers, whether children motivated each other to perform challenges, and what the uptake was of an indirect-parental involvement intervention in the home setting. It is particularly interesting to establish whether one parent, both parents or even the whole family (e.g., siblings and/or other caregivers) participated, and how they valued the intervention. Subsequently, future research should be done on the long-term engagement of children and parents in such interventions. 5. Conclusions {#sec5-ijerph-17-04838} ============== This study was an important first step towards gaining insight in a strategy to enhance parental involvement in energy balance-related interventions. The use of parent-child challenges has potential in increasing parental involvement in school-based interventions. All children participated in the Challenge Me intervention and 93% of the children had involved a parent, yet certain child and family characteristics should be taken into account when further developing parent-child challenges (such as the development of age-appropriate and culture-specific challenges) to ensure that all children and parents, including families of vulnerable populations, participate in these types of parental involvement interventions. We thank the schools, children and parents for participating in our intervention and study. We also want to thank Dynamo Jeugdwerk Eindhoven for their efforts in designing and implementing Challenge Me. S.P.J.K., D.H.H.V.K. and S.B.V. initiated the project and obtained funding.; A.H.V., S.R.B.V.-J., D.H.H.V.K. and S.M.P.L.G. designed the study; A.H.V. and S.R.B.V.-J. carried out the study.; A.H.V. and S.R.B.V.-J. collected and analysed the data; A.H.V. drafted and revised the manuscript; A.H.V., S.R.B.V.-J. and S.M.P.L.G. helped draft the manuscript.; S.R.B.V.-J., D.H.H.V.K., S.P.J.K., S.B.V. and S.M.P.L.G. read the manuscript and provided feedback. All authors have read and agreed to the published version of the manuscript. This research was funded by Fonds NutsOhra, grant number 101.253. The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. The data that support the findings of this study are available upon reasonable request from the corresponding authors A.H.V. or D.H.H.V.K. The data are not publicly available because they contain information that could compromise research participant privacy/consent. {#ijerph-17-04838-f0A1} ijerph-17-04838-t0A1_Table A1 ###### Number of children performing each challenge. Category Challenge *N* \% ---------------------------------------------------------------------------------------------------------------------------- ----------------------------------------------------------------- ------ ------ PA Child only Run as far as possible with your entire class, in 2 min. 171 75.7 Try to perform as many bottle flips as possible in 1 min. 150 66.4 Jump rope as long and often as you can, together with your PE teacher. 125 55.3 Go to the town-/city centre and play a game of 'the floor is lava'. 99 43.8 Participate in an activity of your local youth organisation. 37 16.4 Challenge a local youth worker for a battle. Challenge him/her in something you are good at. 69 30.5 Watch instruction video: Free Running 133 58.8 PA Parental Involvement Help your parent/guardian with the housekeeping. 118 52.5 Ask your dad/mom/guardian/grandpa/grandma/neighbour what game they liked to play when they were young, and play that game. 49 21.7 Go exercise for 30 min together with your parent(s)/guardian(s). 39 17.3 Go walking, rollerblading, or stepping, together with your parent(s)/guardian(s) 61 27.0 Do a 'bob for a job' in your neighbourhood. 27 11.9 Come to school using a means of transport other than a car. 176 77.9 Nutrition child only Find a healthy dinner recipe and bring this with you to school. 56 24.8 Make a vlog about healthy/unhealthy food in your surroundings. 13 5.8 Nutrition Parental Involvement Cook a healthy soup. 69 30.5 Make a healthy breakfast for someone in your household. 54 23.9 Make a work of art out of a piece of fruit. 78 34.5 ijerph-17-04838-t001_Table 1 ###### Overview of 'Çhallenge Me' challenges. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Child-Only Parental Involvement ----------- ----------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------- PA Run as far as possible with your entire class, in 2 min.\ Help your parent/guardian with the housekeeping.\ Try to perform as many bottle flips as possible in 1 min.\ Ask your dad/mom/guardian/grandpa/grandma/neighbour what game they liked to play when they were young, and play that game.\ Jump rope as long and often as you can, together with your PE teacher.\ Go exercise for 30 min together with your parent(s)/guardian(s).\ Go to the town-/city centre and play a game of 'the floor is lava'.\ Go walking, rollerblading, or stepping, together with your parent(s)/guardian(s).\ Participate in an activity of your local youth organisation.\ *Do a 'bob for a job' in your neighbourhood.*\ Challenge a local youth worker for a battle. Challenge him/her in something you are good at.\ *Come to school using a means of transport other than a car.* Watch instruction video: Free running. Nutrition Find a healthy dinner recipe and bring this with you to school.\ Cook a healthy soup.\ Make a vlog about healthy/unhealthy food in your surroundings. Make a healthy breakfast for someone in your household.\ *Make a work of art out of a piece of fruit.* ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Note: PA = physical activity. Challenges are numbered at random, no order of completion was specified. Challenges in Italic require parental involvement dependent on the age of the child (older children are expected to be able to perform the challenge without help from the parent). ijerph-17-04838-t002_Table 2 ###### Child and family characteristics. *N* \% Mean SD ------------------------------ --------------------- ----- ------ ------ ----- Child characteristics Age 10.9 1.0 BMI z 0.09 1.1 Gender Boy 100 44.2 Girl 126 55.8 Ethnicity Western 141 62.7 non-Western 84 37.3 Sports membership Member 173 77.9 non-member 49 22.1 Preference PA (0--28) Active 15.7 3.3 Preference nutrition (0--14) Healthy 6.2 2.8 Family characteristics Filled in by Mother 141 82.9 Father 26 15.3 Other 3 1.8 Family situation Parent with partner 157 84.4 Single parent 29 15.6 Residential status score −0.9 0.8 Siblings 0 24 10.9 1 105 47.7 2 48 21.8 3 28 12.7 4--7 15 6.9 Parental level of education Low 92 50.0 Low/High mixed 37 20.1 High 55 29.9 FHC-PA (1--4) 2.8 0.4 FHC nutrition (1--4) 3.1 0.4 Note: Numbers and percentages presented are based on complete cases. SD = standard deviation. PA = physical activity. FHC = family health climate. ijerph-17-04838-t003_Table 3 ###### Number of children who performed the challenges, per category. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 0 Challenges Performed (0%) At Least One Challenge Performed (1--99%) All Challenges Performed (100%) ----------------------------------------------- ----------------------------- ------------------------------------------- --------------------------------- ------ ---- ----- PA 2 0.9 224 99.1 0 0 PA child-only (7 challenges) 17 7.5 205 90.7 4 1.8 PA parental involvement\ 22 9.7 200 88.5 4 1.8 (6 challenges) Nutrition 85 37.6 139 61.5 2 0.9 Nutrition child-only (2 challenges) 162 71.7 59 26.1 5 2.2 Nutrition parental involvement (3 challenges) 100 44.2 111 49.2 15 6.6 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Note: *N* = number of children. PA = physical activity. ijerph-17-04838-t004_Table 4 ###### Bivariate associations between child and family characteristics and challenges performed. PA PA Child-Only PA Parental Involvement Nutrition Nutrition Child-Only Nutrition Parental Involvement ------------------------------ --------------------- --------------- --------------- ------------------------- --------------- ---------------------- -------------------------------- Characteristic Reference β β β β Β β Age **−0.153 \*** **−0.227 \*** 0.016 **−0.154 \*** −0.063 **−0.165 \*** BMI z −0.010 −0.064 0.065 0.048 0.036 0.042 Gender Boys 0.047 0.052 0.018 0.098 0.009 0.122 Ethnicity Western **−0.151 \*** −0.102 **−0.146 \*** −0.078 −0.050 −0.074 Sports membership Non-member **0.226 \*** **0.221 \*** 0.128 0.110 0.112 0.082 Preference active activities 0.091 0.073 0.073 N.A. N.A. N.A. Preference healthy nutrition N.A. N.A. N.A. 0.000 −0.055 0.030 Residential status score 0.046 0.079 −0.019 0.008 −0.041 0.032 Siblings **0.142 \*** 0.046 0.200 \* 0.065 0.040 0.063 Parental educational level (both) Low educated Mixed 0.095 0.091 0.054 0.167 0.147 0.139 (both) high educated 0.054 0.125 −0.065 0.125 0.130 0.094 FHC-PA −0.021 0.038 −0.083 N.A. N.A. N.A. FHC nutrition N.A. N.A. N.A. 0.007 0.036 −0.010 Note: β = standardised beta coefficient. PA = physical activity. FHC = family health climate. N.A. = not applicable. \* Significant result (*p* \< 0.05) shown in bold print. ijerph-17-04838-t005_Table 5 ###### Multivariate associations between child characteristics and challenges performed. PA PA Child-Only PA Parental Involvement Nutrition Nutrition Child-Only Nutrition Parental Involvement ------------------------------ ------------ --------------- --------------- ------------------------- ----------- ---------------------- -------------------------------- Characteristic Reference β β β β β β Age **−0.136 \*** **−0.208 \*** 0.019 −0.131 −0.052 **−0.142 \*** BMI z −0.008 −0.065 0.069 0.050 0.040 0.044 Gender Boys 0.030 0.047 −0.004 0.115 0.040 0.127 Ethnicity Western −0.119 −0.056 **−0.148 \*** −0.052 −0.030 −0.052 Sports membership Non-member **0.182 \*** **0.180 \*** 0.101 0.109 0.103 0.085 Preference active activities 0.095 0.070 0.084 N.A. N.A. N.A. Preference healthy nutrition N.A. N.A. N.A. −0.032 −0.064 −0.007 Note: β = standardised beta coefficient. PA = physical activity. N.A. = not applicable. \* Significant result (*p* \< 0.05) shown in bold print. ijerph-17-04838-t006_Table 6 ###### Multivariate associations between family characteristics and challenges performed (adjusted for child characteristics). PA PA Child-Only PA Parental Involvement Nutrition Nutrition Child-Only Nutrition Parental Involvement ---------------------------- --------------------- -------- --------------- ------------------------- -------------- ---------------------- -------------------------------- Characteristic Reference β β β β β β Residential status score 0.082 **0.175 \*** −0.063 −0.063 −0.097 −0.029 Siblings 0.140 −0.002 **0.247 \*** 0.091 −0.030 0.138 Parental educational level (both) Low educated Mixed 0.075 0.036 0.090 **0.214 \*** **0.205 \*** 0.169 (both) high educated −0.067 −0.020 −0.094 0.133 0.142 0.097 FHC-PA −0.002 0.047 −0.059 N.A. N.A. N.A. FHC nutrition N.A. N.A. N.A. −0.038 −0.028 −0.035 Note: β = standardised beta coefficient. PA = physical activity. FHC = family health climate. N.A. = not applicable. \* Significant result (*p* \< 0.05) shown in bold print. Multivariate regression analysis adjusted for child characteristics: age, BMI z-score, gender (boy/girl), ethnicity (Wester/non-Western), sports membership (yes/no), preference active activities and preference healthy nutrition.

All relevant data are within the paper and its Supporting Information files. Introduction {#sec001} ============ Ecosystem services are defined as the benefits that nature provides to society \[[@pone.0235320.ref001]\]. Ecosystems such as forests, grasslands, croplands, coastal zones, and urban areas offer different services to society. These include provisioning services (food, water, wood, and fibers), regulation services (which affect climate, flooding, disease, waste, and water quantity and quality), cultural services (recreational opportunities, aesthetic, and spiritual values), and support services (soil formation, photosynthesis, and nutrient cycling) \[[@pone.0235320.ref001], [@pone.0235320.ref002]\]. Regulating services are obtained directly from ecosystems without any transformational process \[[@pone.0235320.ref001]\]. Water regulation is one of such services, which has great importance to society by providing adequate quality water and maintaining the water cycle \[[@pone.0235320.ref003], [@pone.0235320.ref004]\]. Water regulation in forest ecosystems involves the processes that take place after precipitation. These include interception, evapotranspiration, surface and subsurface flow, infiltration, soil erosion control, water quality, and groundwater replenishing, among others \[[@pone.0235320.ref001]\]. These water and soil movement-related processes are affected by various climatic and topographic factors, as well as by soil and vegetation cover types. In a forest system, a portion of the rain is intercepted by the top and other layers of the canopy. The intercepted rain evaporates and returns to the atmosphere, while the non-intercepted portion reaches the ground and deep parts of the soil \[[@pone.0235320.ref005]\]. Forest cover plays an important role in intercepting, capturing, and channeling rainfall. In addition, it is one factor that can be directly manipulated by resource managers using silvicultural practices that consist of different harvesting techniques of various intensities. For the purpose of this study, we focused on the processes that regulate water movement, starting from the canopy, to the forest floor, and into the stream channels. This includes the processes of throughfall, stemflow, and surface runoff. Throughfall refers to the amount of water that passes directly through the forest canopy or drips from branches and leaves of trees \[[@pone.0235320.ref006]\]. It accounts for 60 to 90% of rainfall \[[@pone.0235320.ref007]\]. Stemflow is the fraction of the water that comes in contact with the forest canopy and runs down the trunks of trees and bushes, before being deposited on the ground \[[@pone.0235320.ref008], [@pone.0235320.ref009]\]. It is often ignored in rainfall studies because it is thought to be insignificant and expensive to measure, particularly when forests are composed by rough-barked trees \[[@pone.0235320.ref010]--[@pone.0235320.ref012]\]. Stemflow values represent between 1 to 4% of total rainfall, although some studies reported values up to 20% for certain forest types \[[@pone.0235320.ref007], [@pone.0235320.ref011], [@pone.0235320.ref012]\]. In many areas, particularly semi-arid ones, stemflow creates important islands of soil moisture and nutrients around the stem and contributes to streamflow and groundwater generation \[[@pone.0235320.ref007]\]. Surface runoff refers to the rainfall that flows over the surface of the soil directly into nearby channels and bodies of water \[[@pone.0235320.ref011]\]. It is often referred to as sheet flow, e.g., the water that resembles a braiding pattern of threads, without forming channels larger than rills and gullies \[[@pone.0235320.ref013]\]. In addition to vegetation and other surface obstructions, the rate of flow is dependent upon soil characteristics. There are numerous studies that describe the effect of soil infiltration capacity on surface runoff \[[@pone.0235320.ref014]--[@pone.0235320.ref017]\]. However, there are only a few studies that have addressed the impacts of vegetation cover on surface runoff in Mexico. Furthermore, there are only a limited number of studies in Mexico that have assessed the relationship between surface runoff and forest density, which, as we said above, can be manipulated through direct silvicultural treatments. Silvicultural treatments affect hydrological fluxes. Intensive silvicultural treatments (e.g. stand thinning from above or clear-cutting) change forest density \[[@pone.0235320.ref018]\], eventually modifying throughfall and stemflow at both stand level and individual tree level \[[@pone.0235320.ref019]\], while increasing surface runoff \[[@pone.0235320.ref020]\]. The potential impact of the increased water from the surface flow may eventually affect site productivity and the provision or regulation of other ecosystem services (e.g., plant diversity, soil erosion control, carbon sequestration, etc.) \[[@pone.0235320.ref021], [@pone.0235320.ref022]\]. Varying levels of tree density affect water cycle components (namely interception, evapotranspiration, infiltration, and surface runoff), causing variations in water soil movement and groundwater reserves \[[@pone.0235320.ref021], [@pone.0235320.ref023]\]. For example, heavy rainfall occurrences, following highly intensive vegetation cover treatments (such as clear-cuts), result in increased surface runoff causing soil erosion, flooding, and water turbidity \[[@pone.0235320.ref023]\]. For short periods, large water and soil movements can modify the quality, quantity, and distribution of water resources \[[@pone.0235320.ref024]\]. This study used hydrological models to analyze throughfall, stemflow, and surface runoff in a managed pine-oak forest in northern Mexico. Hydrological models, which relate the flow of water to some stand variables, enable an evaluation of the impact of changes in forest cover on the water resources within a watershed \[[@pone.0235320.ref025]\]. The models can help determine the best forest management scenarios in places where regulation services are combined with provisioning ecosystem services. The objectives of this study were to evaluate the effects that forests, in terms of some forest tree and stand variables, have on throughfall, stemflow, and surface runoff in a temperate area of northern Mexico. The working hypotheses are that throughfall and stemflow are different depending on tree size and genus, and that stand density affects surface runoff. Materials and methods {#sec002} ===================== The study area is located in the mountainous region of the Sierra Madre Occidental, within the municipality of Durango, which lies in the southern part of the state of Durango. The experimental site is located in a private property known as Molinillos ([Fig 1](#pone.0235320.g001){ref-type="fig"}). The owners of this 2,866-hectare property have played a leading role in promoting a healthy silvicultural management, biodiversity conservation, and ecotourism in the region \[[@pone.0235320.ref026]\]. They allowed us to conduct research and field measurements on their property. The current management plan includes the application of non-intensive tree regeneration methods (selective harvesting) as well as intensive methods (seed tree retention or clear-cuts) in different parts of the property \[[@pone.0235320.ref018], [@pone.0235320.ref026]\]. Of the total area, about 2,050 ha are under timber management, with the following treatment distribution: clear-cuts 3.5%, tree retention 14%, thinning 27%, and individual selection 55.5% \[[@pone.0235320.ref026]\]. {#pone.0235320.g001} The climate in the region is temperate and sub-humid, with moderate levels of rainfall in the summer and parts of December and January. In the coldest month (January), daily temperatures can reach anywhere between -3 °C and 18 °C. In the warmest month (June), daily temperatures vary between 15 °C and 35 °C. Historical records show that the mean annual temperature varies from 8 °C to 26 °C, while the annual average is 13.3 °C. The annual rainfall varies from 443 to 1450 mm, with an average of 917 mm \[[@pone.0235320.ref027]\]. Regional elevation ranges from 1,500 to 3,000 meters above sea level. However, the elevation in the plots is closer between 2,360 and 2,630 meters. The typical slope ranges between 20% and 60%. Runoff water flows toward the hydrographic system of the Acaponeta River basin and eventually into the Pacific Ocean. The natural, pine-oak forests include mixtures of *Pinus strobiformis*, *P*. *cooperii*, *P*. *durangensis*, *P*. *engelmannii*, *P*. *teocote*, *P*. *leiophylla*, *Quercus coccolobifolia*, *Q*. *ruogosa*, *Q*. *sideroxilla*, *Q*. *obtusata*, and *Arbutus* spp. The type of soils are Regosol, Litosol, Eutric Cambisol, and Luvisol cromico types \[[@pone.0235320.ref026]\]. Tree and stand variables {#sec003} ------------------------ The information of the tree and stand variables was obtained from eight circular plots of 1,000 m^2^. Even though there were some silvicultural treatments applied in the area, this research is not evaluating the treatment effects; instead, it is assumed that they were already applied before and their effects were observed within a range of levels of tree density. In any case, the plots were located according to the stand density levels existing in the area. Where there were not enough sites with the desired basal areas, we waited until forest managers applied the needed prescribed silvicultural treatments. Density levels used in the study ranged from 0 to 32 m^2^ ha^-1^ of residual basal area, with a mean of 13.6 m^2^ ha^-1^. In each plot, in addition to the taxonomic identification of trees, diameter at breast height (DBH), canopy radius at each cardinal direction (*e*.*g*., N, S, W, E), and total tree height were measured for all trees with a DBH equal to or larger than 7.5 cm. Age was measured for four representative pine trees that vary on the basis of their diameter class. These measurements were used to calculate total tree volume (VOL), basal area (BA), and canopy cover (CC) at the plot level ([Table 1](#pone.0235320.t001){ref-type="table"}). The goal was to determine any possible associations between throughfall (TF), stemflow (SF) or surface runoff (SR), and the stand variables. For example, the functional relationship of SR to the stand variables is described as follows: $$SR = f\left( CC,DBH,VOL,BA \right)$$ where *SR* can be replaced by SF or TF. 10.1371/journal.pone.0235320.t001 ###### Site and stand structural characteristics of the experimental plots. {#pone.0235320.t001g} Variable Mean Minimum Maximum Std. Dev. ------------------------------------------------------------------------------------------ ------- --------- --------- ----------- DBH (cm) 17.2 7.5 85.7 11.4 Height (m) 10.0 0.3 30.0 5.5 BA (m^2^ ha^-1^) 13.6 0.0 32.4 9.5 Annual growth (m^3^ ha^-1^) 2.2 1.9 3.7 0.5 Age (years) 53.5 39.0 69.0 9.2 Canopy cover (%) 54.3 40.0 73.8 9.1 Timber volume (m^3^ ha^-1^) 147.2 101.8 188.8 28.2 Pine (%) 46.9 30.1 73.9 18.3 Oak (%) 44.7 19.3 67.2 19.1 Madrone (%) 7.72 0 36.14 14.1 Elevation (m) 2451 2366 2629 69.4 Slope (%) 41.3 10.0 66.0 13.8 Incident precipitation (mm)[\*](#t001fn001){ref-type="table-fn"} 31.4 7.0 69.0 16.8 Throughfall (mm)[\*](#t001fn001){ref-type="table-fn"} 19.9 1.5 69.0 16.7 Stemflow (mm)[\*](#t001fn001){ref-type="table-fn"}[^a^](#t001fn002){ref-type="table-fn"} 0.37 0.001 9.86 1.07 Surface runoff (mm)[\*](#t001fn001){ref-type="table-fn"} 0.26 0.003 2.54 0.38 \* Average per event. ^a^ Values for all genera. Source: Present study and Perez-Verdin et al. \[[@pone.0235320.ref028]\]. Throughfall {#sec004} ----------- Throughfall (TH) is one of the three components of rainfall partitioning. Rainfall is the sum of throughfall, stemflow, and interception \[[@pone.0235320.ref006]\]. Three rain gauges with a 70-mm capacity were randomly placed in each plot under forest cover (one of them anywhere in the catchment sub-plot intended to measure surface runoff), to manually measure throughfall. Four gauges were placed in open spaces near to the plots to measure incident precipitation (Pi). The measurement period covered the months with dominant rainfalls (July to September) of 2016 and 2018. The data for 2017 were partially collected and later discarded for analysis due to logistical problems resulting from a lack of timely financial resources to continue water measurements. Stemflow {#sec005} -------- Stemflow (SF) is the other component of rainfall partitioning \[[@pone.0235320.ref029]\]. Along with throughfall, but excluding interception loss, they form the net precipitation that reaches the forest floor \[[@pone.0235320.ref030]\]. To measure and evaluate SF, four trees were selected from each 1000-m^2^ plot. The selected sample included at least one tree from each genus that consisted of pines, oaks, and madrones. If only one species was present in the plot, then all four selected trees belonged to that particular species. In total, the study used 32 sampled trees. In each tree, a rubber collar was placed around the main stem in a spiral pattern to catch and direct the water towards a 20-liter container ([Fig 2](#pone.0235320.g002){ref-type="fig"}) \[[@pone.0235320.ref029]\]. The water collected in the container was poured into a plastic beaker to measure the SF volume per tree. Measurements were recorded for every rain event or for every other event while preventing container overfill to avoid any spilling. The volume was then converted into depth (mm) using the tree crown projected area \[[@pone.0235320.ref029]\]. {#pone.0235320.g002} A useful measure that links precipitation, SF, and dasometric variables together in a single index is the stemflow funneling ratio (SF~r~) \[[@pone.0235320.ref006], [@pone.0235320.ref031]\]. While stemflow represents the quantity of rain water captured by a tree, SF~r~ represents the efficiency the same tree has in capturing rainfall and generating stemflow \[[@pone.0235320.ref032]\]. Thus, SFr measures a tree's ability to funnel precipitation at the base, relative to its stem size and different levels of precipitation \[[@pone.0235320.ref006], [@pone.0235320.ref031]\]. The relationship is described as follows: $${SF}_{r} = \frac{SF}{P_{i} \bullet B_{t}}$$ where *SF*~*r*~ is the stemflow funneling ratio (dimensionless), *SF* is the stemflow yield (L) generated by a tree (*t*), *B*~*t*~ is the basal area of the tree (m^2^), and *P*~*i*~ is incident precipitation (mm). When *SF*~*r*~ is greater than one, the funneling ratio indicates that SF is greater than the incident precipitation expected in a rain gauge occupying an area equivalent to the tree basal area \[[@pone.0235320.ref031], [@pone.0235320.ref032]\]. The funneling ratio was modeled as a function of tree variables. Surface runoff {#sec006} -------------- Surface runoff is generated by three mechanisms, infiltration excess runoff, saturation excess runoff, and return of subsurface storm flow \[[@pone.0235320.ref014], [@pone.0235320.ref033]\]. The first mechanism (often called Hortonian runoff) occurs when the infiltration capacity of the soil is exceeded by rainfall. The second (sometimes called Dunne flow) occurs when the storage capacity of the soil is reached so that the soil cannot retain any more rain, thus resulting in surface runoff \[[@pone.0235320.ref013]\]. The third mechanism refers to the case when water returns from subsurface storm flow to the surface, but it is mostly detectable in larger areas or during longer evaluation periods \[[@pone.0235320.ref033]\]. Due to the steep slopes of the terrain, this study is concerned with the infiltration excess (Hortonian) runoff. This type of runoff resembles a shallow sheet flow formed by a braiding pattern of water threads. It can be measured by isolating a small plot on a slope corralled by metal or plastic sheets at the top, sides, and a gutter at the bottom \[[@pone.0235320.ref013]\]. This method has been used to evaluate the effects of forest management practices on SR \[[@pone.0235320.ref034]--[@pone.0235320.ref036]\]. In this study, the SR experiment involved construction of 16-m^2^ sub-plots (8 x 2 m) in each of the eight 1000-m^2^ plots. The perimeter of each SR plot was surrounded by a sturdy geomembrane fence inserted into the soil to a depth of 15 cm. Another 30 cm of the fence was left above the surface to guide movement of the collected runoff into a container located on the lower part of the slope ([Fig 3](#pone.0235320.g003){ref-type="fig"}). Unlike metallic or wooden fences, the geomembrane can easily accommodate to the terrain irregularities. Surface runoff measurements were conducted after a precipitation event by collecting the runoff and pouring it into a plastic graduated beaker to estimate its volume (L). The standard measure for any component of the water balance is expressed in mm depth, which is obtained by dividing the water volume by the catchment area (16 m^2^). Each SR amount measured after a precipitation event was used as the main data input to find the best relationship between SR and tree density. This type of data collection approach allowed the consideration of more variable rainfall input rather than using average data. {#pone.0235320.g003} To reduce the heterogeneity of site conditions (e.g., soil texture, grasslands, herbaceous vegetation cover, etc.) and their effects on SR, the plots were established close to each other. The average (Euclidean) distance between plots was 560 m (minimum 80 m, maximum 2.1km). Thus, stand density (as expressed in terms of basal area), would be one of the most important factors explaining the SR variability. Statistical analysis {#sec007} -------------------- Correlation estimates and scatter graphs were used to detect any preliminary associations between forest stand variables and water flow, as described in [Eq 1](#pone.0235320.e001){ref-type="disp-formula"}. The Shapiro-Wilk test was used to determine the normality of the flow variables. Given that some variables were not normally distributed, parametric and non-parametric models were used to find the best fit model. The Kruskal-Wallis' non-parametric test was also used to determine any significant differences in the SF produced by the three genera (i.e. oak, pine, and madrone). Linear regression by quantiles models were used to find the best relationship between SR, SF, and stand variables according to [Eq 1](#pone.0235320.e001){ref-type="disp-formula"}. Unlike the ordinary least squares model, quantile regression does not assume a defined distribution for the dependent variable, nor does it assume a constant variance \[[@pone.0235320.ref037]\]. The former model uses the conditional mean, but it does not consider the conditional variance of the response factor given a predictor \[[@pone.0235320.ref038]\]. In addition, quantile regression of intercept estimates are not dependent on the typical normal error distribution that ordinary least squares regression always assumes \[[@pone.0235320.ref039]\]. The *τ* symbol (Tau) if often used to specify the quantile levels. Quantile regression is robust to response outliers and easily deals with rate parameter estimation for changes in the quantiles of the distribution of responses, given the independent variables \[[@pone.0235320.ref039]\]. This cannot be equal for all quantiles in models with heterogeneous error distributions. However, quantile regression is often criticized because it is computationally intensive and requires a fair amount of data to perform properly \[[@pone.0235320.ref038]\]. Nonetheless, its efficacy to model stochastic processes is recognized, particularly in those cases where data dispersion is frequent \[[@pone.0235320.ref037]\]. The statistical analyses were conducted using the free access software R version 3.5.1 \[[@pone.0235320.ref040]\], as well as the SAS^®^ system for plotting quantile regression estimates \[[@pone.0235320.ref038]\]. Results {#sec008} ======= Between 40 and 50 precipitation events were registered in the main rainy seasons. For the years 2016 and 2018, the average incident precipitation was 626 mm and 704 mm, respectively. The months with the highest precipitation were August (51%), followed by September (29%) and July (20%). On average, throughfall accounted for 68% of the incident precipitation while stemflow amounts only 1.3% of the rain event ([Fig 3](#pone.0235320.g003){ref-type="fig"}). Stemflow and funneling ratio {#sec009} ---------------------------- The number of valid stemflow records for pines, oaks, and madrones were 98, 127, and 110, respectively. On average, the proportion of SF to incident precipitation was 0.6% for pines, 2.3% for oaks, and 0.9% for madrone. The Kruskal-Wallis test for the sampling period revealed significant differences among the genera (*χ*^*2*^ = 7.055, *p*\<0.03). Therefore, we performed the analysis for each genus. We first attempted to analyze the relationship between SF (volume and depth) and DBH for each genus, but the results were not significant, even with transformations of variables. Interestingly, the relationships for depth-DBH and volume-DBH, were negative and positive, respectively. This contrast highlights the potential differences between volume and depth stemflow modeling. Thus, we evaluated the stemflow funneling ratio (*SF*~*r*~) as a function of DBH and obtained statistically significant results. The median *SF*~*r*~ values for pines, oaks, and madrone were 1.27, 2.38, and 3.72 (*χ*^*2*^ = 47.71, *p*\<0.01), respectively. Since these values are greater than one, they suggest that a tree funnels more water to its base than what would be expected if a rain gauge of equal cavity area had occupied the same basal area as the tree trunk \[[@pone.0235320.ref006]\]. In addition, for all species combined, the median *SF*~*r*~ value for large trees (i.e., DBH ≥30 cm) was 0.10, while for small trees was 1.37 (*χ*^*2*^ = 75.05, *p*\<0.01). A quantile regression model was eventually adjusted for each genus. The quantile regression equation for the *SF*~*r*~ was: $${SF}_{r} = \beta_{0} + \beta_{1}\left( {{\text{log}\mspace{360mu}}\left. DBH \right)} + \varepsilon \right.$$ Coefficient values for BA were negative for all species and quantiles ([Table 2](#pone.0235320.t002){ref-type="table"}). This indicates that the stemflow funneling ratio decreases as tree diameter increases. This relationship is similar to the one depicted by SF depth (mm) and DBH, though as we said earlier, it was not significant. [Table 2](#pone.0235320.t002){ref-type="table"} shows the results of quantile regression for 0.1, 0.5, and 0.9 levels; however, as in many cases of quantile regression applications \[[@pone.0235320.ref039]\], we estimated the coefficients for all quantiles calculated between these numbers. [Fig 4](#pone.0235320.g004){ref-type="fig"} shows the entire grid for the DBH parameters in the interval (0, 1). The shaded blue area represents the 95% confidence limits for the quantile regression estimates. For all three genera, the slope estimates (*β*~1~) decrease exponentially as the quantile level increases. Moreover, this figure reveals that, for all types of genera, the parameter estimates and confidence limits for DBH are negative across the majority of quantile levels, which suggests that, although they are always negative, they exhibit different rates of change. In the case of oaks, the slope estimates change from -0.32 in the 0.1 quantile (*p*\<0.01) to -6.34 in quantile 0.9 (*p*\<0.01). The same exponential relationship, but positive, occurs for the intercept estimates (*β*~0~). Both slope and intercept estimates differ across quantiles because the variance in *SF*~*r*~ changes as a function of DBH. Thus, DBH not only has an impact in the median of *SF*~*r*~, but also in its variance. This variation cannot be observed using the ordinary least squares method, since only the value for the conditional mean is obtained (dashed line, [Fig 4](#pone.0235320.g004){ref-type="fig"}). {ref-type="disp-formula"}) for each genus. The dashed line in each figure represents the ordinary least squares estimate of the conditional mean effect.](pone.0235320.g004){#pone.0235320.g004} 10.1371/journal.pone.0235320.t002 ###### Quantile regression estimates and goodness-of-fit statistics for the relationship between stemflow funneling ratio and diameter at breast height (cm) in Molinillos, Mexico. {#pone.0235320.t002g} Genus Quantile levels (*τ*) Parameter[\*](#t002fn001){ref-type="table-fn"} Estimate SE Pr(\>\|t\|) Pseudo-R^2^ --------------------- ----------------------- ------------------------------------------------ ---------- -------- ------------- ------------- Pine (*n* = 98) 0.1 *β*~1~ -0.27 0.04 \<0.01 0.66 *Β*~0~ 0.93 0.15 \<0.01 0.5 *β*~1~ -0.68 0.25 0.03 0.36 *Β*~0~ 2.66 0.89 0.01 0.9 *β*~1~ -1.86 0.97 0.05 0.65 *Β*~0~ 8.67 3.55 0.02 Oak (*n* = 127) 0.1 *β*~1~ -0.32 0.09 \<0.01 0.51 *Β*~0~ 1.08 0.32 \<0.01 0.5 *β*~1~ -1.09 0.44 0.02 0.36 *Β*~0~ 3.84 1.43 \<0.01 0.9 *β*~1~ -6.34 1.25 \<0.01 0.91 *Β*~0~ 25.16 4.16 \<0.01 Madrone (*n* = 110) 0.1 *β*~1~ -1.19 0.69 0.09 0.50 *Β*~0~ 3.58 1.79 0.05 0.50 *β*~1~ -5.03 0.93 \<0.01 0.36 *Β*~0~ 15.8 2.63 \<0.01 0.1 *β*~1~ -11.82 3.63 \<0.01 0.91 *Β*~0~ 37.04 9.67 \<0.01 \*Model parameters are based on [Eq 3](#pone.0235320.e003){ref-type="disp-formula"}. Surface runoff {#sec010} -------------- A total of 249 surface runoff (SR) records were collected during the measuring seasons (2016 and 2018), with a minimum of 0.05 l (0.003 mm) and a maximum of 40.6 l (2.54 mm). The average monthly SR was recorded in August (5.24 mm), followed by September (3.2 mm), and July (0.9 mm). Surface runoff represented about 1.9% and 0.8% of the incident precipitation for 2016 and 2018, respectively. It was mainly correlated with BA, meaning that the SR modeling was estimated with this variable alone. The quantile regression model, described by the following equation, yielded the best results: $$SR = \beta_{0} + \beta_{1}*{{\text{log}\mspace{360mu}}{BA}} + \varepsilon$$ where *SR* is expressed in mm, *BA* is the basal area per hectare (m^2^ ha^-1^), *β*~*i*~ are model parameters, and *ε* is regression error. The relationship between SR and BA was moderately strong for the 0.5 quantile, which yielded a modest adjustment (Pseudo R^2^ = 0.45), compared to the much better results obtained from the other two quantiles (0.10 and 0.90) ([Table 3](#pone.0235320.t003){ref-type="table"}). The forest vegetation in the study area is overwhelming mixed, making it difficult to separate the individual effect of each type of species on SR. Therefore, we generated an all-species model for SR. The goal was to determine the cumulative effect of managing natural, mixed forests on the amount of flow at different stand densities. 10.1371/journal.pone.0235320.t003 ###### Quantile regression estimates and goodness-of-fit statistics for the relationship between surface runoff (mm) and basal area (m^2^ ha^-1^) in Molinillos, Mexico (*n* = 249). {#pone.0235320.t003g} Quantile level (*τ*) Parameter[\*](#t003fn001){ref-type="table-fn"} Estimate SE Pr(\>\|t\|) Pseudo-R^2^ ---------------------- ------------------------------------------------ ---------- -------- ------------- ------------- 0.10 *β*~1~ -0.034 0.003 \<0.01 0.98 *Β*~0~ 0.115 0.009 \<0.01 0.50 *β*~1~ -0.081 0.01 \<0.01 0.45 *Β*~0~ 0.320 0.05 \<0.01 0.90 *β*~1~ -0.152 0.046 \<0.01 0.96 *Β*~0~ 0.818 0.138 \<0.01 \*Model parameters are based on [Eq 4](#pone.0235320.e004){ref-type="disp-formula"}. While positive across all quantile levels, the intercept estimates (*β*~0~) have different rates of change. They increase as the quantile levels also increase ([Fig 5a](#pone.0235320.g005){ref-type="fig"}). The slope estimates (*β*~1~) decrease as the proportion of quantile increases ([Fig 5b](#pone.0235320.g005){ref-type="fig"}). These findings suggest that there is greater data dispersion at lower values of BA and lower variability at higher values of BA ([Fig 5c](#pone.0235320.g005){ref-type="fig"}). Again, this is a unique advantage of quantile regression. Unlike the ordinary least squares, which coefficients are represented by horizontal, dashed lines, quantile regression allowed the estimation of parameters for the upper and lower tails of the basal area distribution. {#pone.0235320.g005} The maximum SR values were observed in the plots in which the BA was zero, while the minimum SR values were observed in BA values higher than 20 m^2^ ha^-1^. Nevertheless, the SR remained relatively constant after BA reached 15 m^2^ ha^-1^ in all quantiles. At this level of tree density, the slope of the curve approaches zero, which implies that there is a balance with SR ([Fig 5c](#pone.0235320.g005){ref-type="fig"}). Discussion {#sec011} ========== This study was developed to analyze the influence of tree and stand variables on stemflow and surface runoff in a temperate forest area in northern Mexico. The results confirm that stemflow is affected by tree diameter and it significantly varies with genera. Likewise, surface runoff is strongly influenced by levels of stand density, expressed in terms of tree basal area (m^2^ ha^-1^). Some studies agree with these findings, while others do not. These contrasting viewpoints are discussed in some detail in the next section. Stemflow and funneling ratio {#sec012} ---------------------------- Results showed that SF was higher for oak trees, followed by madrone and pine species. Similar results were presented by Pérez-Suarez, et al. \[[@pone.0235320.ref041]\] who reported that SF in oak forests was 25% higher than in pine forests, and up to 22% higher than in mixed forests in central Mexico. Cantú-Silva and González-Rodríguez \[[@pone.0235320.ref042]\] found that the proportion of SF to incident precipitation was 0.6%, 0.5%, and 0.03% for pines, oak, and mixed pine-oak stands, respectively. The difference in the SF between genus (e.g. madrone, oak, and pine) may be influenced by the characteristics of each group (e.g., bark roughness, leaf area, and tree architecture). In particular, oak trees are part of the co-dominant forest structure, have a thick, rough bark \[[@pone.0235320.ref043]\], and can direct water to the stem base in greater quantities than pines and madrone. The study did not find a clear relationship between SF and tree DBH. There is a positive, but not significant, relationship between SF volume and DBH. The relationship between SF depth and DBH is negative, but again without being significant. Theoretically, the size of the tree (expressed in terms of DBH, height), the shape and size of the canopy, and the bark and angle of insertion of the branches modify the canopy catchment area and thus affect the amount of SF produced \[[@pone.0235320.ref010], [@pone.0235320.ref044]\]. However, it seems that literature does not completely agree on how these factors affect stemflow. Martinez-Meza and Whitford \[[@pone.0235320.ref045]\] reported a direct relationship between SF and the canopy area in a Chihuahuan desert species. Chen et al. \[[@pone.0235320.ref046]\] also reported a direct relationship between SF and DBH, mostly when rainfall was less than 15 mm per hour. Above this threshold, the effect of tree size was not significant. Pérez-Suarez et al., \[[@pone.0235320.ref041]\] found a positive relationship between DBH and SF for oak species. Other studies like that of Marín et al. \[[@pone.0235320.ref005]\] did not detect a clear relationship between SF and BA. Yet, León-Peláez et al. \[[@pone.0235320.ref047]\] observed an inverse relationship between DBH and SF for pine and other species. Navar et al. \[[@pone.0235320.ref048]\] likewise, found a negative relationship between SF and DBH in a Tamaulipan thornscrub forest. These studies suggest the need to continue carrying out more research on site characteristics with varying topographic, soil, climatic, and vegetation conditions to establish their effects on SF. Germer et al., \[[@pone.0235320.ref049]\] suggest that *SF*~*r*~, unlike the typical stemflow measure, offers more reliability to compare stemflow generation in concentrated point sources of water in forests with a diversity of tree sizes and species. This is because *SF*~*r*~ is normalized for basal area and precipitation. We found significant statistical results between the stemflow funneling ratio and DBH in the quantile regression analysis. Our results showed that there is an inverse, significant relationship between the tree funneling ratio and DBH, for all species and for all quantiles. Findings suggest that the *SF*~*r*~ decreases as a tree becomes larger. Many reasons can explain this inverse relationship. On one hand, as a tree grows, its bark gets thicker \[[@pone.0235320.ref050], [@pone.0235320.ref051]\]. According to Herwitz \[[@pone.0235320.ref031]\], bark has a greater water-holding capacity than foliar surfaces. Thick-barked trees have more interception storage capacity than thin-barked trees \[[@pone.0235320.ref011], [@pone.0235320.ref012]\]. On the other hand, larger trees transfer more precipitation to throughfall than smaller trees \[[@pone.0235320.ref011], [@pone.0235320.ref046]\]. In large trees, the probability of water reaching the soil by directly dripping from branches and leaves increases. Consequently, less water is transferred to stemflow and more to throughfall \[[@pone.0235320.ref010], [@pone.0235320.ref032]\]. Other studies also agreed that *SF*~*r*~ values are greater in small trees \[[@pone.0235320.ref032], [@pone.0235320.ref046], [@pone.0235320.ref048], [@pone.0235320.ref049]\]. The fact that small trees are more efficient in capturing water at their base, and thereby creating important islands of soil moisture, brings additional insight to future stemflow studies \[[@pone.0235320.ref049]\], which should include these type of trees. The presence of small trees, along with a diversity of species, may affect subsurface flow, saturation overland flow, and groundwater recharge \[[@pone.0235320.ref032], [@pone.0235320.ref048], [@pone.0235320.ref049]\]. A greater dispersion of *SF*~*r*~ values was also observed in small trees. For instance, the statistical range of *SF*~*r*~ for small pines (i.e. with a DBH less than 15 cm) was 10 whereas the range for large trees (DBH larger than 40 cm) was only 2. Many factors can explain this variability, among others, bark thickness, angle of branches, leaf index, and length of crown \[[@pone.0235320.ref032]\]. Quantile regression was able to deal with this DBH dispersion because it not only estimated the impact of DBH in the median of *SF*~*r*~, but also in its extreme values. Surface runoff {#sec013} -------------- Overall, the SR quantile regression models were highly significant for each quantile tested ([Table 3](#pone.0235320.t003){ref-type="table"}). Surface runoff is influenced by precipitation, the more of it, the more amount of water converts to SR. High intensity, short duration rainfall events produce more SR and soil loss in many vegetation types \[[@pone.0235320.ref035]\]. Quantile regression can eventually help predict the impact of varying levels of precipitation, including extreme rain events that occur during a period of study. Pérez-Verdín et al. \[[@pone.0235320.ref028]\] generated a non-linear model of SR using BA as the independent variable. They also found a negative relationship between SR and BA. However, their model was only fitted to the mean observed values, unlike our study, in which we also modeled the upper and lower quantiles. Results indicated that the higher the tree BA, the lower the SR. Bosch and Hewlett \[[@pone.0235320.ref052]\] mentioned that SR increases significantly with heavy thinning and clear-cuts. Lack of vegetation cover, after intensive harvesting, reduces canopy interception and evapotranspiration resulting in larger amounts of runoff and downstream channel flows \[[@pone.0235320.ref053]\]. However, SR and streamflow responses to forest removals declined over time as forest and understory vegetation grew. Establishing plantations or increasing cover in areas with scarce vegetation decreases SR \[[@pone.0235320.ref054], [@pone.0235320.ref055]\]. The reason for this inverse relationship is because of the high tree density, which stimulates interception (and evapotranspiration) and decreases the amount of water reaching the forest floor. Depending on the slope and type of soil, it can further infiltrate deeper into the ground \[[@pone.0235320.ref034]\]. Conversely, excessive harvests reduce forest density, which in turn disturbs and exposes surface soil to rainfall, thus increasing erosion due to SR \[[@pone.0235320.ref020], [@pone.0235320.ref021], [@pone.0235320.ref036]\]. The consideration of other predicting variables such as soil texture, grasslands, and shrublands in future studies will help broaden the scope of SR models. Chen et al., \[[@pone.0235320.ref035]\] showed that some grass and shrub species can decrease SR up to 50% compared to forestlands with poor ground cover. The inclusion of this type of variables will also help analyze the effects of diverse forest structures not only on SR but on soil erosion as well. Surface runoff remained relatively constant after BA reached 15 m^2^ ha^-1^ in all quantiles. This may suggest that there is some compatibility with timber production. Below this range, not only will SR be impacted, but so will many other ecosystem services \[[@pone.0235320.ref028]\]. Perez-Verdin et al., using multicriteria decision-making techniques, found that the most appropriate BA for the management of some ecosystem services, including SR, was between 17 and 21 m^2^ ha^-1^ \[[@pone.0235320.ref028]\], which is within the range we considered as compatible. This information can be useful to forest managers for prescribing better silvicultural treatments in this type of ecosystem. The Hortonian SR starts when the intensity of rain begins to exceed the infiltration capacity of the soil. In our study, the lowest SR was recorded in the first days of July, which coincides with the beginning of the rainy season in the area. The SR increased with high rainfall intensities in late July, August, and September. In the beginning of the rainy season, there is a low SR due to a high infiltration capacity in response to gravity and metric potential that pulls the water down into dry soil \[[@pone.0235320.ref056]\]. But, as the rainy season progresses, the soil infiltration capacity starts to decrease and eventually reaches a constant that approaches the saturated hydraulic conductivity of the soil \[[@pone.0235320.ref016]\]. Eventually, swelling of the soil colloids and the closing of small cracks lower the infiltration capacity below the saturated hydraulic conductivity \[[@pone.0235320.ref056]\]. This process results in having more overland flow over the soil surface. Conclusions {#sec014} =========== This study was conducted to evaluate the effect of tree and stand variables on stemflow and surface runoff in temperate forests of the state of Durango, Mexico. We did not find a clear relationship between stemflow and stand variables. However, the stemflow funneling ratio was significant to tree diameter at breast height changes. The stemflow funneling ratio was greater in younger trees, suggesting the need to consider this type of trees in future stemflow-related studies. Stand density, expressed in terms of basal area, had significant effects on surface runoff. Denser stands intercept more precipitation and expose it to evaporative loss, allowing less water to turn into surface runoff. Less dense forests, on the other hand, permit precipitation to reach the ground with less impedance and produce more surface runoff. Surface runoff remained relatively constant after basal area reached 15 m^2^ ha^-1^ in all quantiles. Overall, the models developed in this study show moderate statistical fit for both stemflow funneling ratio and surface runoff. Quantile regression is useful for predicting different scenarios concerning the upper and lower tails of the predictor distribution. In addition, the generated models are simple and practical to use, since DBH or BA are the only independent variables used. These variables are important measures for selecting the most suitable silvicultural treatments. Furthermore, by knowing the interactions of these variables with hydrological fluxes of interest, they can readily be used to help determining forest management regimes compatible with the quantity and quality of other ecosystem services in this type of ecosystem. Supporting information {#sec015} ====================== ###### (XLSX) ###### Click here for additional data file. We thank the owners of the Molinillos private estate for allowing us to execute this study on their property. We are also grateful to Celina Perez and three anonymous reviewers for their invaluable inputs in an early manuscript. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.

Introduction {#Sec1} ============ Psoriasis is a chronic, immune-mediated, inflammatory skin disorder that is currently incurable. Consequently, the majority of people with psoriasis require long-term treatment to maintain disease control. Traditional immunosuppressive systemic treatments, such as acitretin, methotrexate, cyclosporine, hydroxyurea, and thioguanine, may be effective in controlling psoriasis in some patients but significant toxicity and the need to closely monitor patients limit the viability of these treatments for long-term, continuous use \[[@CR23]\]. Recently developed systemic therapies that selectively target specific pathways in the inflammatory cascade of psoriasis generally have a much improved safety profile compared with traditional therapies \[[@CR26]\]. Efalizumab (anti-CD11a; Raptiva^®^) is a recombinant humanized monoclonal IgG~1~ antibody that has been approved for the treatment of moderate-to-severe chronic plaque psoriasis. It interferes with the pathogenesis of psoriasis via multiple mechanisms, including inhibition of T-lymphocyte trafficking and T-lymphocyte activation and reactivation \[[@CR1], [@CR10], [@CR11], [@CR21], [@CR25]\]. The safety and efficacy profile of efalizumab has been established in numerous clinical trials, in which more than 3,500 patients were enrolled and treatment was assessed for up to 3 years \[[@CR4]--[@CR6], [@CR12]--[@CR17], [@CR22]\]. Although psoriasis can be associated with the co-morbidity of psoriatic arthritis, a minority of patients with psoriasis (7--30%) will develop this joint disease \[[@CR27]\]. Nevertheless, psoriatic arthritis constitutes a major consideration in patients who are receiving long-term treatment for their psoriasis. A Nordic study of more than 5,000 patients with psoriasis showed that patients with arthritis exhibited greater impairment of psoriasis-related quality of life (QoL), longer disease duration, and greater self-reported disease severity, compared with patients who had psoriasis but no co-morbid arthritis \[[@CR27]\]. A low incidence of arthropathy adverse events (AEs; any form of joint disease) associated with efalizumab treatment has been reported in both clinical studies and routine clinical practice \[[@CR8], [@CR12]\]. However, anecdotal reports of arthropathy in routine clinical practice have expressed concern that efalizumab may be associated with exacerbation of arthropathy \[[@CR8]\]. To address this concern, we conducted a large-scale pooled analysis of safety data from five Phase III clinical trials (including open-label extensions of two of these studies) and two Phase III open-label clinical trials of efalizumab to explore whether arthropathy AEs were associated with efalizumab treatment in patients with psoriasis. Methods {#Sec2} ======= The primary objective of this pooled safety analysis was to assess the incidence of arthropathy AEs in patients who had received either efalizumab or placebo. Safety data were pooled from five randomized, double-blind, placebo-controlled clinical trials (including data from two open-label extension studies of two of these trials) and two open-label clinical trials of efalizumab \[[@CR4]--[@CR6], [@CR12]--[@CR17], [@CR22]\]. Patients included in these Phase III studies were aged ≥18 years and had moderate-to-severe chronic plaque psoriasis, a psoriasis area and severity index (PASI) score of ≥12 at screening, and plaque psoriasis covering ≥10% of body surface area. All patients were candidates for either systemic anti-psoriatic therapy or had received systemic anti-psoriatic therapy. Patients included in these trials received subcutaneous injections with efalizumab, 1--4 mg/kg once weekly or 2 mg/kg once-every-other week, or placebo. Details of individual study methodologies are described in other publications \[[@CR4]--[@CR6], [@CR12]--[@CR17], [@CR22]\]. Arthropathy AEs were defined according to the Coding Symbols for Thesaurus of Adverse Reaction Terms (COSTART) \[[@CR3]\] preferred terms 'arthritis' and 'arthrosis', or the Medical Dictionary for Regulatory Activities (MedDRA, <http://www.meddramsso.com/NewWeb2003/index.htm>) preferred terms 'arthritis not otherwise specified (NOS)', 'psoriatic arthropathy', 'arthropathy NOS', 'monoarthritis', 'polyarthritis', and 'osteoarthritis NOS'. Treatment groups analyzed {#Sec3} ------------------------- Due to the variety of study designs, five analyses were considered: 'first-treatment phase', 'first exposure phase', 'extended treatment phase', 're-treatment phase', and 'long-term treatment' (see Table [1](#Tab1){ref-type="table"}). Table 1Summary of the Phase III data from five placebo-controlled clinical trials (including data from two open-label extension studies of two of these trials) and two open-label clinical trials of efalizumab included in the pooled safety analysisPublication (protocol number)Study designNumber of patients in each analysisFirst treatment (0--12 weeks)Efalizumab sc 1--4 mg/kg qw or 2 mg/kg qowFirst exposure\  Extended treatment (13--24 weeks)Long-term treatment^a^(≤36 months)Re-treatmentPlaceboEfalizumab 1 mg/kgEfalizumab 2 mg/kgLeonardi \[[@CR13]\] (ACD2058g)Randomized, double-blind, parallel-group, placebo-controlled170162166462123--55Lebwohl \[[@CR12]\] (ACD2059g)Randomized, double-blind, parallel-group, placebo-controlled122232243579289----Gordon \[[@CR4]\] (ACD2390g)Randomized, double-blind, parallel-group, placebo-controlled187368--368------Papp \[[@CR17]\] (ACD2600g)Randomized, double-blind, parallel-group, placebo-controlled236449--449--449--Sterry \[[@CR22]\] (IMP24011)Randomized, double-blind, parallel-group, placebo-controlled264529--772308--145Papp \[[@CR16]\] (ACD2062g)Open-label------34137--365Gottlieb \[[@CR5], [@CR6]\] (ACD2243g)Open-label------339^b^290339--Menter \[[@CR14]\] (ACD2391g)Open-label extension^c^of study ACD2390g \[[@CR4]\]------174342----Menter \[[@CR15]\] (ACD2601g)Open-label extension^c^of study ACD2600 g \[[@CR17]\]------217622635^d^--Pooled analysis97917404093,3942,111n.a.^e^565*qow* once-every-other week, *qw* once weekly, *sc* subcutaneous^a^Number of patients at start^b^Patients received combined therapy with fluocinolone acetate (*n* = 169) or petrolatum (*n* = 170) for weeks 9--12; for months 3--15, the dose of efalizumab could be escalated to 4 mg/kg per week for up to 4 weeks if clinically indicated^c^Some patients are included in the analyses more than once because patients in the open-label extension studies are also included in analyses of the parent studies^d^Included patients who received either efalizumab or placebo in the parent study \[[@CR17]\]^e^Not applicable because data are analyzed and reported separately for the study by Gottlieb et al. \[[@CR6]\] and the study published by Papp \[[@CR17]\] and Menter \[[@CR15]\] It is worth noting that most of the studies included in this pooled analysis were designed and conducted before efalizumab had received regulatory approval and before it was known that doses of more than 1 mg/kg once weekly (the approved dose) did not confer additional treatment benefit (EMEA, Raptiva Summary of Product Characteristics; FDA US, FDA Prescribing Information for Raptiva). For this reason, only the efalizumab 1 mg/kg once-weekly dose data are reported for the 'first treatment phase' of the analysis. Due to the wide variety of study designs included in the pooled analysis, data for patients receiving any dose of efalizumab are combined for all other treatment phases analyzed. The 'first treatment phase' analysis included 0--12-week data from patients in the five placebo-controlled studies who received either efalizumab 1 mg/kg once weekly or placebo. This analysis allows a comparison between the efalizumab and placebo treatment groups. The 'first exposure phase' included 12-week data from all studies in patients who had their first exposure to any dose of efalizumab and, thus, did not include placebo data. This analysis was conducted to include the maximum number of patients who received efalizumab for their first 12 weeks of treatment (i.e., it included those patients who first received efalizumab treatment after crossing over from a placebo group, as well as the patients who first received efalizumab during weeks 0--12). The 'extended treatment phase' analysis included 13--24-week data in patients given any dose of efalizumab who had already received efalizumab during the first treatment phase. The 'long-term treatment phase' analysis included all patients who received continuous long-term treatment (up to 36 months) with any dose of efalizumab. Data were analyzed in 12-week segments to assess change in the incidence of arthropathy AEs over time. This analysis included data from two long-term studies \[[@CR6], [@CR15], [@CR17]\], which were analyzed separately due to differences in study design. The 're-treatment phase' analysis included all patients who re-started treatment with efalizumab following a treatment-free observation period. Statistical analyses {#Sec4} -------------------- Descriptive statistics were used to explore the association between efalizumab and the occurrence of arthropathy AEs. Results are expressed as point-estimates of the incidence rates (ratio of the number of patients with an arthropathy AE to the total number of patient-years at risk of an arthropathy AE) with their 95% confidence intervals (CIs). Descriptive comparisons are provided; no formal statistical tests were performed. Analyses were also conducted to explore the relationship between onset of arthropathy AEs during efalizumab treatment and a previous history of arthropathy (reported as a narrative by patients at the baseline visit) and the incidence of arthropathy AEs and clinical response to efalizumab treatment \[measured at 12 weeks using the Physician Global Assessment (PGA) and PASI scales\]. Differences in patient and psoriasis characteristics at baseline were also compared between patients who had arthropathy AEs and those who did not. An additional analysis of data from patients included in the first treatment phase of the study by Sterry et al. \[[@CR22]\] (Table [1](#Tab1){ref-type="table"}) was conducted to assess the incidence of psoriatic arthropathy in these patients. This was the only study to define arthropathy AEs according to MedDRA; other studies used the COSTART, which did not include 'psoriatic arthropathy' specifically as a preferred term. Baseline demographics and psoriasis characteristics and the proportion of patients with a previous history of arthropathy (as reported by patients at the baseline visit) were tabulated by presence/absence of an arthropathy event. Results {#Sec5} ======= The number of patients included in each of the pooled safety analyses from each of the seven trials and two open-label extensions is summarized in Table [1](#Tab1){ref-type="table"}. Up to 3,394 patients received at least one dose of efalizumab. A total of 2,719 patients were included in the first treatment phase analysis, of whom the majority (64%; 1,740 patients) received efalizumab 1 mg/kg per week; 979 patients (36%) received placebo. Efalizumab 2 mg/kg per week regimen was given to 409 patients (15%) in two of the five studies \[[@CR12], [@CR13]\]; consequently, these patients were not included in the first treatment analysis. Patient demographics and baseline psoriasis characteristics were similar between treatment groups in the first treatment phase (Table [2](#Tab2){ref-type="table"}). Table 2Baseline demographic and disease characteristics for patients in the placebo-controlled first treatment phaseCharacteristicsPlacebo (*n* = 979)Efalizumab 1 mg/kg per week (*n* = 1,740)Efalizumab 2 mg/kg per week (*n* = 409)Mean age (years), mean (SD)45 (12)45 (12)45 (13)Weight (kg), mean (SD)90.0 (20.0)89.4 (19.6)93.6 (20.5)Mean BMI^a^ (kg/m^2^)30.4 (6.4)30.2 (6.3)31.4 (6.6)Race,*n* (%) Caucasian891 (91)1,569 (90)356 (87) Other88 (9)171 (10)53 (13)Duration of psoriasis, mean number of years (SD)19.2 (11.4)19.1 (11.4)17.6 (11.7)History of arthritis,*n* (%)286 (29.2)529 (30.4)141 (34.5)*BMI* body mass index^a^Due to missing height data, BMI was calculated for 971 patients in the placebo group, 1,719 patients in the efalizumab 1 mg//kg per week group and 404 patients in the efalizumab 2 mg/kg per week group First treatment phase (weeks 0--12) {#Sec6} ----------------------------------- During the first 12 weeks of treatment a similar proportion of patients had an arthropathy AE in the efalizumab 1 mg/kg group (3.3%) and the placebo group (3.5%; Fig. [1](#Fig1){ref-type="fig"}a). Correspondingly, the incidence of arthropathy AEs per patient-year was 0.15 in the efalizumab 1 mg/kg group (95% CI 0.11--0.19) and 0.16 in the placebo group (95% CI 0.11--0.22; Fig. [1](#Fig1){ref-type="fig"}b). The majority of the arthropathy AEs was mild-to-moderate in intensity in both the efalizumab (41/58 events; 71%; 95% CI 57--82%) and placebo groups (31/34 events; 91%; 95% CI 76--98%). Fig. 1**a** Proportion of patients who had arthropathy adverse events (AEs) during each phase of the safety analysis and **b** incidence of arthropathy AEs per patient-year for each phase The additional analysis of data from the study by Sterry et al. \[[@CR22]\] demonstrated that the incidence of psoriatic arthropathy per patient-year was lower in the group treated with efalizumab 1 mg/kg per week (0.10; 95% CI 0.05--0.18) than in the placebo group (0.17; 95% CI 0.08--0.30); the proportion of patients with psoriatic arthropathy was 2.3% (12/529 patients) in the efalizumab group and 3.8% (10/264 patients) in the placebo group. First exposure phase {#Sec7} -------------------- In total, 3,394 efalizumab-treated patients were included in this analysis. A small proportion of patients had an arthropathy AE (3.6%; Fig. [1](#Fig1){ref-type="fig"}a) and the incidence of arthropathy AEs per patient-year was also low (0.16; 95% CI 0.14--0.19; Fig. [1](#Fig1){ref-type="fig"}b). The incidence of arthropathy AEs in this group of patients was similar to that in the placebo group in the first treatment phase, as indicated by the overlap in CIs. Extended treatment phase (weeks 13--24) {#Sec8} --------------------------------------- In total, 2,111 patients were included in the extended treatment phase analysis. During this phase, a low proportion of patients had an arthropathy AE (3.8%; Fig. [1](#Fig1){ref-type="fig"}a) and the incidence of arthropathy AEs per patient-year was also low (0.17; 95% CI 0.14--0.22; Fig. [1](#Fig1){ref-type="fig"}b). Overlap in the CIs indicates that the incidence of arthropathy AEs in this group of patients was also similar to that in the placebo group in the first treatment phase. Long-term treatment phase {#Sec9} ------------------------- The results of two long-term studies were analyzed separately to assess the incidence of arthropathy AEs in patients treated with efalizumab. In both of these studies (Fig. [2](#Fig2){ref-type="fig"}), there was no overall increase in the incidence of arthropathy AEs over time. Furthermore, the incidence of arthropathy remained similar to that of the placebo group in the first treatment phase and stable between 12-week periods. Fig. 2Incidence of arthropathy AEs in long-term studies of patients treated with efalizumab **a** for up to 36 months and compared indirectly with pooled placebo data from the first treatment (FT) phase \[[@CR5], [@CR6]\] and **b** for up to 15 months and compared with the study's placebo group during month 0--12 \[[@CR15], [@CR17]\]. \*Following the first 3-month double-blind, placebo-controlled phase of this study, patients in the placebo group who continued were switched to open-label treatment with efalizumab. Consequently, the month 6, 9, 12 and 15 results included patients who had received placebo during the initial 3 months of the study In total, 339 patients were included in the analysis of the study by Gottlieb et al. \[[@CR6]\]. These patients received continuous treatment with efalizumab 2 mg/kg once weekly for weeks 1--12 (fluocinolone acetate or petrolatum was co-administered during weeks 9--12), followed by continuous maintenance treatment with efalizumab 1 mg/kg once weekly for up to 36 months in patients who had a ≥ 50% improvement in PASI score. For months 3--15, the dose of efalizumab could be escalated to 4 mg/kg per week for up to 4 weeks if clinically indicated, then maintained at 2 mg/kg per week. During the entire study period, there was little variation in the incidence of arthropathy AEs (range 0.06--0.19; Fig. [2](#Fig2){ref-type="fig"}a). Reasons for discontinuation were diverse and were representative of the overall population; refer to Gottlieb et al. \[[@CR6], [@CR7]\] for details of discontinuations. For the other long-term study, the analysis included 3-month data from 449 efalizumab-treated patients in the placebo-controlled first treatment phase of the study \[[@CR17]\] and data from 635 patients who entered the open-label extension phase and received efalizumab treatment \[[@CR15]\]; 218 of the 635 patients included in the open-label extension had switched from placebo to efalizumab after completing the first treatment phase. Patients who entered the open-label extension phase continued to receive, or initiated treatment with, efalizumab 1 mg/kg once weekly for up to 15 months continuously. As in the long-term study by Gottlieb et al. \[[@CR6]\], there was little variation in the incidence of arthropathy AEs during the entire study period (range 0.06--0.12; Fig. [2](#Fig2){ref-type="fig"}b). Re-treatment phase {#Sec10} ------------------ In total, 565 efalizumab-treated patients were included in the re-treatment phase of the analysis. In this phase, a lower proportion of patients had an arthropathy AE (2.7%; Fig. [1](#Fig1){ref-type="fig"}a) compared with the first treatment phase, and the incidence of arthropathy AEs per patient-year was also lower (0.12; 95% CI 0.07--0.19; Fig. [1](#Fig1){ref-type="fig"}b). The incidence of arthropathy AEs in this group of patients was lower than in the placebo group in the first treatment phase. Baseline characteristics and previous history of arthropathy {#Sec11} ------------------------------------------------------------ There were no differences in baseline demographics or disease characteristics between the patients who had arthropathy AEs and those who did not. Patients who experienced an arthropathy AE during treatment with efalizumab appeared to be more likely to have a history of arthropathy prior to treatment. Of the patients who never developed an arthropathy AE during efalizumab treatment, 27% reported a previous history of arthropathy compared with 59% in patients who did have an arthropathy AE. During the first treatment phase, 88% (*n* = 34) and 76% (*n* = 79) of patients who developed an arthropathy AE had a history of arthropathy prior to receiving placebo or efalizumab 1 mg/kg once weekly, respectively. Arthropathy AEs and clinical response to efalizumab {#Sec12} --------------------------------------------------- Arthropathy AEs appeared to be less likely to occur in patients who had a good clinical response to treatment (≥75% improvement in PASI score; 2.3% of patients had events) than in patients who had a partial response (50--74% improvement in PASI score; 3.5% of patients had events; Fig. [3](#Fig3){ref-type="fig"}) and non-responders (\<50% improvement in PASI score; 4.5% of patients had events). The corresponding incidences of arthropathy AEs per patient-year were 0.10 in patients with a good clinical response (95% CI 0.05--0.18), 0.17 in patients with a partial clinical response (95% CI 0.11--0.25), and 0.21 in patients who did not respond (95% CI 0.15--0.28). Fig. 3**a** Proportion of patients with an arthropathy AE by response category on the psoriasis area and severity index (PASI) and physician global assessment (PGA) scales and **b** incidence of arthropathy AEs per patient-year by response category on the PASI and PGA scales When assessed using the PGA scale, arthropathy AEs also appeared to be less likely to occur in patients who had better clinical responses to treatment with efalizumab (Fig. [3](#Fig3){ref-type="fig"}). During the extended treatment phase, the incidence of arthropathy AEs per patient-year was 0.17 in patients with responses categorized as 'cleared', 'excellent' or 'good' on the PGA scale (95% CI 0.10--0.25) and 0.25 in patients with responses categorized as 'fair', 'slight', 'unchanged' or 'worse' on the PGA scale (95% CI 0.17--0.35). Discussion {#Sec13} ========== The placebo-controlled results of this large-scale pooled analysis of arthropathy data from seven clinical trials show that efalizumab does not appear to increase the risk of developing arthropathy AEs compared with placebo during the first 12 weeks of treatment. In addition, for patients treated with efalizumab, the incidence of arthropathy AEs did not appear to increase over time. The proportion of patients who had an arthropathy AE within any 12-week treatment period was low (\<4.1%) through all treatment phases (first treatment, first exposure, extended treatment, re-treatment, long-term treatment). Joint disease has also been reported as a side-effect of other approved biological treatments for psoriasis, namely infliximab (EMEA public statement on infliximab, <http://www.emea.eu.int/pdfs/human/press/pus/444500en.pdf>) \[[@CR2], [@CR18], [@CR19]\], alefacept (Biogen safety presentation on Alefacept to the FDA, <http://www.fda.gov/ohrms/dockets/ac/02/slides/3865S1_04_Biogen-Safety/sld007.htm>) \[[@CR20], [@CR24]\], and etanercept (EMEA Scientific discussion for the approval of Enbrel, <http://www.emea.eu.int/humandocs/PDFs/EPAR/Enbrel/014600en6.pdf>). Indeed, placebo-controlled studies of infliximab and alefacept indicate that in patients with moderate-to-severe psoriasis the incidence of arthralgia (joint pain) is 7 and 5%, respectively (Biogen safety presentation on Alefacept to the FDA, <http://www.fda.gov/ohrms/dockets/ac/02/slides/3865S1_04_Biogen-Safety/sld007.htm>) \[[@CR18], [@CR24]\]. Psoriatic arthritis has been reported as serious treatment-related AE in three placebo-controlled studies of etanercept in the treatment of chronic plaque psoriasis (incidence data have not been published) (EMEA Scientific discussion for the approval of Enbrel, <http://www.emea.eu.int/humandocs/PDFs/EPAR/Enbrel/014600en6.pdf>). The incidence of arthropathy AEs in the current analysis of efalizumab appears to be similar to that for arthralgia in studies of infliximab and alefacept. Moreover, the term 'arthropathy', used in the current study encompasses a variety of joint diseases, not just a single joint condition such as arthralgia or psoriatic arthritis, and therefore has greater potential to include more patients. However, this between-study comparison is indirect and thus should be treated with caution. Moreover, no arthropathy event (defined by any of the MedDRA or COSTART preferred terms) was excluded from the analysis. Also, data from first treatment phase of the study by Sterry et al. \[[@CR22]\] indicate that the proportion of patients with psoriatic arthropathy specifically was low (2.3%) in patients treated with efalizumab 1 mg/kg per week---in fact, lower than in the placebo group (3.8%). It should be noted, however, that psoriatic arthropathy events were not confirmed by a rheumatologist---this is a potential limitation of the study. However, the umbrella term 'arthropathy' was designed to capture all joint diseases, including 'psoriatic arthropathy'. Also, the incidence of psoriatic arthropathy in the study by Sterry et al. was in line with the incidence of 'arthropathy' in the overall pooled analysis. To put the results of this pooled analysis, which by its very nature included select patient populations (determined by the inclusion/exclusion criteria and study designs), in the context of routine clinical practice, post-marketing surveillance data were assessed. During post-marketing surveillance of efalizumab, which accounts for approximately 17,500 patient-years to date, serious arthropathies requiring hospitalization were reported with a frequency of about 4.8 per 1,000 patient-years in patients receiving efalizumab. It should be noted, however, that underreporting of AEs in routine clinical practice setting may lead to an underestimate of the true incidence of arthropathy. For both the 12-week first treatment and first exposure phases of the current analyses, the proportions of patients reporting an arthropathy AE appeared to be lower in the efalizumab groups than in the placebo group in the first treatment phase. Correspondingly, the incidences of AEs per patient-year in these treatment phases were also lower in the efalizumab groups than that observed in the placebo group in the first treatment phase. However, the proportion of AEs that were moderate or severe was greater in the efalizumab groups than in the placebo groups; too few patients had events to draw any meaningful conclusions. During the extension phase (weeks 13--24), the incidence of arthropathy AEs in efalizumab-treated patients remained similar to the placebo group in the first treatment phase. Previous history of arthropathy and poor clinical response may potentially indicate a risk for occurrence of new arthropathy AEs during treatment. Indeed, arthropathy AEs were most frequent in patients who did not respond to therapy with efalizumab or in patients with a history of arthropathy. Importantly, the data from the two long-term studies of efalizumab indicate that the incidence of arthropathy AEs remains stable and low for up to 3 years of continuous treatment. These results, coupled with efficacy data showing that the clinical improvements of the skin after 3 months of efalizumab therapy are maintained throughout 36 months of continuous dosing \[[@CR5]\], support the suitability of efalizumab for the chronic, continuous treatment of patients with psoriasis. Reasons for patients' discontinuations in the 36-month study by Gottlieb et al. \[[@CR6], [@CR7]\] were diverse and representative of the overall population included in this analysis and have been described previously. When considering the long-term analysis of the 36-month study (Fig. [2](#Fig2){ref-type="fig"}a), it should be noted that the number of patients who remained in the study decreased over time. This discontinuation rate is not unexpected for a study that is 3 years in duration but, by month 36, there is a small number of patients on which to base comparisons with the first treatment phase. Another factor that may confound between-phase analysis comparisons was the possible use of concomitant medications for psoriasis after the first treatment phase in the study by Gottlieb et al. \[[@CR6]\], which permitted the use of topical corticosteroids and ultraviolet B phototherapy. Accordingly, it should be noted that comparisons of the results between any of the treatment phases of this analysis are observational (i.e., not direct) but do confirm the results of the long-term treatment phase and the placebo-controlled 12-week first treatment phase studies, suggesting that the risk of joint disease is not increased with continued efalizumab treatment and that the incidence of arthropathy is low and similar to placebo. Further investigation is needed to confirm the results of this preliminary analysis of arthropathy events during long-term treatment with efalizumab. In patients who re-started treatment for a further 12 weeks following an intervention-free period, the proportion of patients who had an arthropathy AE was lower than during the first treatment phase; the same was true for the incidence of arthropathy AEs per patient-year in re-treated patients. Although this scenario is likely to occur infrequently in clinical practice, these data show that if a patient needs to stop (e.g., during pregnancy) and then restart treatment, there appears to be no increased risk of arthropathy AEs. Although arthritis in patients with psoriasis has a significant impact on QoL \[[@CR9], [@CR27]\], it can, in most cases, be managed effectively \[[@CR8]\]. In the small minority of patients who develop arthropathy during treatment, the symptoms can be managed successfully with non-steroidal anti-inflammatory drugs \[[@CR8]\]. In conclusion, the results of this pooled analysis show that efalizumab does not appear to increase the risk of developing arthropathy AEs compared with placebo. Long-term studies of efalizumab indicate that the incidence of arthropathy AEs remains stable and low for up to 3 years of continuous treatment. This study was sponsored by Serono International S.A. The authors thank Tom Potter, MSc, for his assistance with manuscript preparation.

Background {#Sec1} ========== Hypoxia is common after stroke, and associated with poor outcomes. In this article, we have reviewed the physiology of oxygen transport, the cerebrovascular response to hypoxia and pathophysiology, incidence and aetiology behind hypoxia in stroke and its subsequent clinical consequences. We have then reviewed all randomised clinical trials looking at the use of supplemental oxygen therapy in acute stroke and made conclusions regarding current evidence and recommendations for clinical practice. Oxygen physiology {#Sec2} ================= The normal adult range of arterial oxygen pressure (PaO2) is 11.0--14.4 kPa and the normal range for arterial oxygen saturation (SaO2) is 95--98% \[[@CR1]\]. The term hypoxia refers to oxygen levels below normal. It includes both tissue (e.g. brain, myocardium) hypoxia and hypoxia in the blood (hypoxaemia). Tissue hypoxia is defined by the concentration of oxygen in blood and also tissue perfusion, whilst hypoxaemia is defined by the concentration of oxygen in inspired air and its transfer into the blood \[[@CR2]\]. Following inhalation oxygen is taken up in the lung capillaries via diffusion down an oxygen concentration gradient across the alveoli \[[@CR3], [@CR4]\]. Oxygen binds to the haemoglobin molecule, which can carry four oxygen molecules; each binding and changing the shape of the haemoglobin molecule and increasing its affinity for oxygen \[[@CR5]\]. A small amount of oxygen is also dissolved in plasma. This proportion increases in hyperoxia, when all haemoglobin is saturated \[[@CR6]\]. Oxygen dissociates from the haemoglobin molecule in the tissues owing to the relatively hypercapnic and acidic environment (the Bohr effect) \[[@CR3]\]. Oxygen is a vital substrate that supports virtually all metabolic processes. 90% of oxygen intake is engaged in the cytochrome C oxidase system in the mitochondria \[[@CR7]\] generating adenosine triphosphate (ATP), which acts as the main energy substrate within cells. A continuous supply of oxygen is required to secure a continuous supply of ATP maintaining sufficient energy for cerebral neuronal and cellular activity. This facilitates an efficient energy producing process making 38 molecules of ATP during aerobic respiration, equivalent to 1270 joules (J) energy, in comparison to two molecules of ATP (67 J of energy) during anaerobic respiration \[[@CR7], [@CR8]\]. The anoxic brain {#Sec3} ================ 20% of all human oxygen consumption is utilised by the brain \[[@CR9]\]. The brain has no oxygen or glucose (the other important substrate in the ATP producing equation) stores. Thus complete disruption of cerebral blood flow very rapidly results in an anoxic, hypoglycaemic state, which via a variety of mechanisms ultimately leads to cell death. Excitatory neurotransmitters, such as glutamate, bind to a variety of receptors and allow for an influx of calcium ions that help formulate the chemical signal for depolarisation \[[@CR10], [@CR11]\]. Normally the re-uptake of glutamate is an active energy-driven process. In the absence of ATP this process fails, resulting in an extracellular accumulation of glutamate, which continually stimulates receptors leading to a persistent influx of calcium ions \[[@CR12]\]. Furthermore, the Na+/Ca2+ ATP driven pump normally used to eliminate calcium fails, also due to a lack of ATP \[[@CR13]\]. The resultant high intracellular calcium triggers multiple cascades that ultimately lead to mitochondrial dysfunction and cell death. Furthermore, instead of producing ATP, glial cells have been shown to release ATP extracellularly \[[@CR11]\]. Aside from rendering this unusable by mitochondria, ATP also stimulates the P2X7 receptor, which again leads to significant calcium influx and ultimately cell death \[[@CR13]\]. The other major mechanism of cellular demise is via the formation of free radicals facilitated by the reduction of iron from its ferric (Fe3+) to its ferrous (Fe2+) form and the initiation of inflammatory cascades \[[@CR12]\]. Cerebral blood flow in hypoxia {#Sec4} ============================== In normoxic states, cerebral blood flow is very tightly controlled by the partial pressure of carbon dioxide (PaCO~2~). Any hypocapnic state will result in vasoconstriction and reduction in regional cerebral blood flow and a hypercapnic state leads to the reverse with vasodilatation and an increase in cerebral blood flow. Cerebral blood flow is somewhat less responsive to changes in PaO~2~, which has the opposite effect to carbon dioxide; a hypoxic state causing cerebral vasodilatation with the aim of improving oxygen delivery and a hyperoxic state causing vasoconstriction \[[@CR14], [@CR15]\]. In a hypoxic state, whilst the vasodilatory response improves flow, the detection of hypoxia by peripheral chemoreceptors will in turn lead to an increase in respiratory drive, increasing arterial oxygen content. However, the consequence of this is also an increase in the clearance of carbon dioxide, which would theoretically cause vasoconstriction and reduced cerebral blood flow \[[@CR9]\]. It appears there is a threshold to which the hypoxic response predominates (and the carbon dioxide one attenuated) at a PaO~2~ of around 50--60 mmHg \[[@CR9], [@CR14]\]. Whilst the carbon dioxide mediated vascular response is mediated via a direct change in vessel wall pH \[[@CR15]\], the oxygen response appears to be mediated by the deoxygenated erythrocyte via a number of mechanisms; which include release of ATP and the subsequent actions of endothelial nitric oxide synthase on the vessel wall, reduction of nitrite to nitric oxide and the activity of S-nitrosohaemoglobin \[[@CR9]\]. The cerebral vascular response to hypoxia is not uniform. A study found that in an induced isocapnic hypoxic state increases in cerebral blood flow were most prominent in basal ganglia nuclei, the putamen, thalamus, nucleus accumbens and pallidum \[[@CR16]\]. Studies of blood flow in individual vessels have found that flow in the internal carotid artery is maintained during hypoxia and that vertebral artery flow is increased \[[@CR17]\]. This had led to the hypothesis that blood flow is increased in this region to preserve vital brainstem structures, or that possibly the posterior circulation vasculature is less susceptible to the effects of carbon dioxide for similar reasons. Neurological effects of hypoxia {#Sec5} =============================== The neurological consequences of hypoxia are dependent upon the speed of onset, the severity of hypoxia, and the level of tissue perfusion. Rapid decreases in PaO~2~, as in a cardiorespiratory arrest, can lead to permanent neurological damage within minutes. However, lower, less abrupt changes, can be tolerated if the decrease in oxygen occurs in a gradual manner, such as ascending at altitude, where individuals can acclimatise and develop tolerance to lower oxygen partial pressures or, (to a lesser degree), in chronic smokers. Initial clinical features include altered judgement, difficulty in completing complex tasks, and impairment in short term memory \[[@CR18], [@CR19]\], but in the longer term deficits can be more widespread and span physical and neuropsychological domains. Seizures occur in up to a third of individuals within a day of exposure to hypoxia, and are commonly partial complex or myoclonic in nature. Intractable forms of either of these types of seizure are associated with a poor prognosis \[[@CR20]\]. Cognitive impairment domains include amnesia, visuospatial deficits, frontal lobe symptoms, impairment of executive function, and impairments in language \[[@CR21]\]. These are covered in more detailed reviews on the subject \[[@CR22], [@CR23]\]. Involvement of the basal ganglia, a region particularly susceptible to hypoxic injury, can result in delayed Parkinsonism in older subjects, dystonia mainly in younger people, choreo-athetosis, and tremors \[[@CR20]\]. Varying degrees of unilateral or bilateral motor impairment may be observed depending on both the anatomical level and extent of corticospinal tract involvement. In very rare cases, the syndrome of delayed post-hypoxic leukoencephalopathy may occur weeks after a seemingly rapid recovery from the original insult. This condition is characterised by rapid deterioration in cognition, emergence of extra-pyramidal signs, and loss of executive function as a results of severe demyelination \[[@CR24]\]. The severity of leukoencephalopathy can be assessed by magnetic resonance imaging (MRI) \[[@CR20]\]. Electroencephalography, somatosensory evoked potentials and MRI can provide valuable information about the severity of hypoxic injury, but also aid in prognostication together with overall clinical state \[[@CR25]\]. Hypoxia in the context of a stroke {#Sec6} ================================== There is no specific definition as to what constitutes hypoxia in an acute stroke, and it is therefore reasonable to assume that normal values for the general population apply. Sulter and colleagues \[[@CR26]\] monitored 49 consecutive patients who presented with an acute stroke within 12 h duration using pulse oximetry for 48 h. Patients were considered hypoxic and treated with supplemental oxygen if saturations were below 96% for more than 5 min. This occurred in 63% \[[@CR31]\] of patients, with 28 of those returning to 'normal' oxygen saturations following administration of up to 5 L/min of oxygen. The remaining three required much higher concentrations. Factors associated with hypoxia in this group were stroke severity, presence of dysphagia, and older age. Roffe et al. \[[@CR27]\] recruited 118 patients (100 of whom had adequate measurements by pulse oximetry) and found that the mean daytime awake SO~2~ was 94.5 ± 1.7% in stroke patients and 95.8 ± 1.7% in healthy controls. Nocturnal saturations were reduced to 93.5 ± 1.9% in the stroke group and 94.3 ± 1.9% in controls. In the stroke group the average 4% oxygen desaturation index (ODI) (number of times per hour the saturation dipped more than 4% from baseline) was higher than in controls. At night almost a quarter of the stroke group had desaturations below 90%. The same group also looked further at the differences between day and night oxygen saturations \[[@CR28]\]. In stroke patients who were not hypoxic (defined as SaO~2~ less than 90%) during the day, baseline daytime saturations were measured between 9am and 9pm and nocturnal saturation between 10pm and 6am. In total 40 patients were recruited and in addition to SaO~2~, respiratory rate and sleep/awakeness was measured twice in each time period. The mean respiratory rate day vs. night was 20 and 18 breaths per minute respectively. The mean daytime SaO~2~ was 95.5% (87--98.6%) and 94.3% (80--98%) at night. There was a strong correlation between respiratory rate, SaO~2~ and the 4% ODI, making it clear that borderline daytime hypoxia could predict nocturnal hypoxic episodes. Comparisons in a later study were then made with matched controls overnight \[[@CR29]\]. In this study the mean nocturnal oxygen saturations were found to be 0.5% less than controls, with the lowest measured desaturation in this group of 79.4%, still almost 6% lower than the control group. The largest difference was in the percentage of patients with more than 10 desaturations per hour (42% stroke vs. 15% controls). Hand et al. \[[@CR30]\] performed a study looking at the feasibility of MRI as an imaging modality in hyperacute stroke assessment. One of the eligible 138 patients for the study could not be scanned owing to pulmonary oedema severe enough to cause considerable hypoxia. For a variety of reasons it was only possible to consistently measure oxygen saturations in 61 out of 85 patients. In those in whom saturations could reliably be measured, 11 out of 61 developed hypoxia (lowest 74%) and of those who received oxygen during the scan only two could be monitored successfully. This highlights not only the prevalence of hypoxia in acute stroke, but the logistical difficulties acute hypoxia may pose for assessment. Another study examined the effect of five different, but randomly ordered body positions, each for 10 min on the impact on oxygen saturation \[[@CR31]\]. Interestingly, lying on the left hand side reduced oxygen saturations, but only in those who hand a right hemiparesis. Those who were able to sit in a chair were able to achieve much higher mean SaO~2~, albeit suffering from more minor strokes. It was felt that a severe stroke, with a right hemiparesis and underlying chest disease were the greatest predictors of desaturation, but only when lying on the left side. A subsequent systematic review \[[@CR32]\] comprising of three randomised controlled trials (173 patients) and one case controlled trial (10 patients) found that body position only played a role in oxygen saturations if patients had underlying respiratory co-morbidities. The risk of aspiration is well documented in acute stroke (see below), but independent of this the question as to whether feeding (oral or nasogastric) contributes to hypoxia has also been examined. Dutta et al. \[[@CR33]\] reported that nasogastric feeding caused no decrease in SaO~2~. A later study \[[@CR34]\] found a small but statistically non-significant trend towards hypoxia when tube fed, in particular in patients that were fed overnight. Rowat and colleagues looked at the impact of oral feeding on oxygen saturations, using hospitalised elderly patients and young healthy controls as comparators \[[@CR35]\]. The baseline SaO~2~ was lower in the stroke cohort than the other two, with a very small decrease in SaO~2~ with oral feeding in the stroke (0.1%) and the elderly groups. Nearly a quarter of stroke patients dropped SaO~2~ to less than 90% (16% elderly, 0% young), but this did not occur in close relation to the time of swallowing, and thus no immediate risk could be attributed to oral feeding. There is relatively little research on the correlation between hypoxia and clinical outcome. Hypoxia has been shown to be an independent clinical risk factor for post stroke dementia \[[@CR36]\]. Rowat et al. \[[@CR37]\] found that hypoxic patients were more likely to have respiratory disease and this led to an increased mortality. A smaller study looked at the prevalence of hypoxia in patients undergoing rehabilitation and found no significant difference in mean SaO~2~ at baseline, in nocturnal SaO~2~, the lowest nocturnal SaO~2~ or in the 4% ODI \[[@CR38]\]. In conclusion, no association between SaO~2~ and functional outcome was found. Hypoxia has, however, been shown to correlate with the degree of white mater disease on MRI. White matter hyperintensity volumes were greatest in obstructive sleep apnoea (OSA) patients compared with non-OSA patients and more explicitly in hypoxic compared to non-hypoxic patients \[[@CR39]\]. Causes of hypoxia in acute stroke {#Sec7} ================================= Pneumonia is a frequent complication of acute stroke. A recent consensus defined the term stroke-associated pneumonia (SAP) as a representation of a spectrum of lower respiratory tract disorders occurring within 7 days after the onset of stroke \[[@CR40]\]. The criteria were based on a modified version of the centre for disease control (CDC) criteria, with a probable SAP fulfilling all CDC criteria but not meeting typical chest radiography changes and definite SAP fulfilling all CDC criteria including typical chest X-ray changes. In addition the consensus group concluded that there was a limited role for C-reactive protein, white blood cell count and other inflammatory biomarkers in the diagnosis \[[@CR40]\]. A meta-analysis of 64 studies showed that the definition of stroke---associated pneumonias varied widely \[[@CR41]\]. The incidence of pneumonia post stroke has been reported to range between 1 and 44% \[[@CR42], [@CR43]\] and has been shown to increase mortality (threefold) and overall hospital care costs \[[@CR42]\]. Two recent studies have looked at the utility of prophylactic antibiotics to reduce pneumonia. The STROKE-INF study \[[@CR44]\] randomised patients with acute stroke and dysphagia to 7 days of prophylactic antibiotics (to be commenced within 48 h of stroke onset) or standard care and found no reduction in the incidence of pneumonia (OR 1.21, 95% CI 0.71--2.08, p = 0.489). The PASS study \[[@CR45]\] investigated the effects of prophylactic ceftriaxone and found that this did not affect functional outcome at 3 months. While there was a significant reduction in infections overall, there was no effect on the incidence of pneumonia (OR 0.67 95% CI 0.39--1.15, p = 0.18). Therefore, current evidence does not support the use of antibiotic prophylaxis to prevent pneumonia. There are several validated risk scores which can help the clinician to identify patients at high risk of stroke-associated pneumonia \[[@CR46]--[@CR48]\]. Given the considerable morbidity and mortality, and the lack of benefit from prophylactic treatment, highlighting patients at high risk to allow early identification and treatment of established infection is important in the care of stroke patients. Aspiration is a frequent cause of pneumonia post stroke, especially in patients with dysphagia. Dysphagia is seen in up to 50% of ischaemic strokes \[[@CR49]\], although the reported incidence can vary between studies. Individuals suffering from dysphagia were three times more likely to develop pneumonia and this number increased to eleven times if they were shown to aspirate \[[@CR50]\]. Often aspiration occurs silently (reported in up to 40%), that is, with few or no clinical signs. The presence of either dysphagia or a subsequent pneumonia is predictors of a worse clinical outcome \[[@CR51]\]. An often neglected aspect of stroke treatment is oral care. Poor oral hygiene leads to proliferation of bacteria and debris in the oral cavity \[[@CR52]\], which are liable to be aspirated causing respiratory tract infection \[[@CR53]\]. This is particularly important in nasogastric tube fed patients in whom oral care can easily be missed. Sleep apnoea is a common cause of intermittent nocturnal hypoxia after stroke, affecting up to 60% of patients \[[@CR54]\]. This condition has also been shown to be a risk factor for future stroke and stroke mortality, if not appropriately treated. A few small studies have shown that nocturnal continuous positive airway pressure ventilation is feasible \[[@CR55], [@CR56]\], and can improve wellbeing in some stroke patients with sleep apnoea during the acute and rehabilitation phase, but compliance with the intervention is poor, especially in patients with delirium or cognitive impairment \[[@CR57]\]. Obstructive sleep apnoea can cause or accentuate many traditional vascular risk factors, in particular hypertension \[[@CR58]\] and atrial fibrillation \[[@CR59]\], and has been shown to be an independent risk factor for stroke \[[@CR54], [@CR57], [@CR60]\]. A review of the cohort of the Wisconsin Sleep Study found a significant association between sleep disordered breathing and stroke prevalence, the more severe the indices of sleep apnoea, the greater the risk \[[@CR61]\]. Respiratory muscle function is also a potential cause of hypoxia either directly by associated muscle paralysis or as a result of a secondary infection. Several studies have shown (some in comparison to matched controls) a significant reduction in forced vital capacity, forced expiratory volume in one second, peak expiratory flow rate and maximal inspiratory and expiratory pressures \[[@CR62]--[@CR64]\], suggesting impairment in function of accessory respiratory muscles as well as the diaphragm. This may pave the way for treatment strategies aiming to improve respiratory muscle function \[[@CR65]\]. Less common stroke complications resulting in hypoxia include pulmonary embolism, which despite its low incidence in most reported series (around 1%), is associated with increased in hospital mortality (31.5 vs 12.7% in a review of over 11,000 patients in the Registry of the Canadian Stroke Network), length of stay and severity of disability \[[@CR66]\]. The risk of pulmonary embolism can persist for up to 4 weeks post stroke \[[@CR67]\]. A review of a relatively small cohort of cryptogenic stroke patients found a significant incidence of silent pulmonary embolism (37%), but did not comment as to whether or not this led to a resultant hypoxic state \[[@CR68]\]. Improvements in mechanical, pharmacological and therapy based regimes are the likely reason pulmonary embolism is now a relative rarity. Cardiac failure and very rarely neurogenic pulmonary oedema \[[@CR69], [@CR70]\] are among the other causes. Oxygen therapy for acute stroke {#Sec8} =============================== Oxygen treatment can be used to maintain normal oxygen saturation or to increase the oxygen saturation above normal in patients with acute stroke. The rationale for the latter is that blood with higher oxygen content may improve oxygen action in ischaemic brain areas \[[@CR2]\]. When considering oxygen treatment it is important to weigh up potential adverse effects against benefits. Potential adverse effects of oxygen treatment after stroke {#Sec9} ---------------------------------------------------------- Oxygen treatment is not without side effects. Attachment to a wall delivery system as an inpatient restricts mobility in the acute phase and may represent an infection risk. In critical ill states or when bordering on the anaerobic threshold for exercise capacity, the body has several intrinsic systems to increase oxygen tension and deliver oxygen at the required rate in order to produce ATP and meet energy demands. One of the by-products of ATP formation is the formation of oxygen free-radical species, which, if not dealt with, can lead to cell apoptosis and developmental of tissue damage. In normal states the body has several intrinsic enzymes to neutralise free radicals by pairing them with so called donor electrons to form substances like oxygen or hydrogen peroxide which can then be efficiently removed. When high concentrations of oxygen are given this leads not only to increased oxygen delivery from red blood cells but also increased delivery via plasma. This then by-passes and overrides usual mechanism of clearance and is one the reasons tissue damage develops in inappropriately high concentrations of oxygen \[[@CR71]--[@CR73]\]. The cascade outlined above is only partially reversed during reperfusion, even though oxygen delivery has improved. Most of the clinical problems surrounding oxygen toxicity initially affect the lungs. High concentrations of oxygen may displace all nitrogen present in the alveoli and owing to the significant alveolar plasma gradient, the oxygen rapidly diffuses and dissolves into the plasma, effectively reducing the alveolar volume and leading to subsequent collapse. Hyperoxia may also impair mucilliary clearance and alter surfactant properties which may cause an 'adhesive collapse' \[[@CR73], [@CR74]\]. Neurological consequences outside of those described in the context of stroke include cerebral vasoconstriction, a by-product of excessive free radical formation, confusion, and seizures \[[@CR73]\]. Oxygen toxicity more often occurs during use of high concentrations of oxygen or in hyperbaric conditions. In the clinical setting a stroke patient is exposed to, these are highly unlikely scenarios to occur. Recommendations from national and international stroke guidelines {#Sec10} ----------------------------------------------------------------- A review of the most recent societal guidelines shows uniformity in the approach to oxygen therapy in acute ischaemic stroke. The Royal College of Physicians guidelines \[[@CR75]\] advise use of supplemental oxygen only if oxygen saturation drops below 95% and is not contraindicated, and recommends no supplemental oxygen for saturations of 95% or above. The European Stroke Organisation \[[@CR76]\] advises supplemental oxygen use for oxygen saturations of less than 95%. The American Heart Association/American Stroke Association Guidelines \[[@CR77]\] advise that in the pre-hospital setting, oxygen supplementation to maintain oxygen saturations above 94% is reasonable and recommended for suspected stroke patients and that on presentation to hospital saturations should be continually monitored to watch out for hypoxia. This guidance is based on the American Heart Association post cardiac arrest guidelines \[[@CR78]\] and thus the same advice applies to stroke patients. Again the guidelines do not support the use of hyperbaric oxygen therapy. Randomised controlled trials of supplemental oxygen in acute stroke {#Sec11} ------------------------------------------------------------------- A plausible solution to aid the correction of cerebral hypoxia in stroke would be to provide supplemental oxygen therapy in the acute phase, potentially helping to correct or prevent many of the catastrophic cerebral changes that may occur. To date 6 randomised controlled trials have tested this hypothesis. A quasi-randomised study of routine oxygen supplementation within the first 24 h of acute stroke by Ronning and Guldvog \[[@CR79]\] showed that routine oxygen treatment (3 L/min for 24 h) in unselected stroke patients did not reduce morbidity or mortality. Subgroup analyses suggested worse outcomes in patients with mild strokes treated with oxygen and a trend towards better outcomes in severe strokes (Scandinavian stroke scale score \<40), but the study was not large enough to identify with certainty those who are likely to derive benefit. Oxygen saturation before or during treatment was not reported and it is therefore impossible to determine whether or not oxygen was ineffective because it failed to improve oxygen saturation or because of a genuine lack of effect on the ischaemic brain. A small study (n = 16) delivered oxygen at a rate of 45 L/min for 8 h, commencing with 12 h of stroke onset. Perfusion-diffusion mismatch on MRI showed that cerebral blood volume and blood flow within ischaemic regions improved in the hyperoxia. Neurological deficit improved at 4 h (during treatment), 24 h and at 1 week. By 24 h MRI of the brain showed reperfusion and (asymptomatic) petechial haemorrhages in 50% of hyperoxia treated patients and 17% of controls (p = 0.06). No long-term clinical benefit was seen at 3 months \[[@CR80]\]. This study was too small to draw reliable conclusions, leading to a larger (unpublished) study by the same group (<http://www.clinicaltrials.gov/ct2/show/NCT00414726?term=singhal&rank=1>) which initially planned to enrol 240 patients, randomising to either room air or high flow oxygen (30--45 L/min for 8 h) within 9 h of acute stroke onset. After enrolment of 85 patients, the study was terminated early due to an imbalance of deaths favouring the control arm, though it is noted that the excess in mortality in the treatment group was not considered related to the treatment by an external blinded assessor. An Indian study \[[@CR81]\] enrolled 40 patients within 12 h of an acute anterior circulation ischaemic stroke and a National Institute Stroke Scale of more than 4 to receive either 10 L/min for 12 h via face mask in the treatment group versus room air or 2 L/min to keep oxygen saturation above 95%. There was no significant difference in NIHSS, modified Rankin or Barthel index scores between the two groups. There was also no statistically significant difference between DWI lesion volumes in either group, though there was a trend towards smaller lesions in the treatment group. In the Stroke Oxygen Pilot study \[[@CR82]\], oxygen was given for 72 h and the dose was dependent on baseline oxygen saturation (2 L/min if the saturation was \>93%, 3 L/min if the saturation was 93% or less). Initial results showed that the treatment regime increased oxygen saturation by about 2% in the treatment arm and this was associated with a small, but significant improvement in neurological recovery at one week. At 6 months \[[@CR83]\] there was no statistically significant difference between the two groups, although there remained a small trend towards overall benefit with supplemental oxygen. This data led to the Stroke Oxygen Study (SO~2~S) \[[@CR84]\], in which 8003 patients within 24 h of hospital admission with acute stroke were randomized 1:1:1 to receive either continuous supplemental oxygen, supplemental oxygen only at night (9pm--7am) oxygen, or no supplemental oxygen treatment for 72 h. This study has completed recruitment and is expected to report in 2016. Conclusion {#Sec12} ========== Oxygen is a vital substrate to the continual function and survival of cerebral tissue. Rapid reduction in partial pressures can very rapidly lead to catastrophic and permanent cerebral injury and physical disability. Whilst evidence does not currently support the additional supplementation of oxygen to stroke patients, it remains important to prevent hypoxia in stroke patients by identifying and treating reversible causes rapidly. Results of the Stroke Oxygen Study will provide new evidence of whether prophylactic oxygen treatment can prevent neurological deterioration and improve recovery. PF and CR both equally participated in the search of the literature and writing of the manuscript. Both authors read and approved the final manuscript. Acknowledgements {#FPar1} ================ None. Competing interests {#FPar2} =================== The authors declare that they have no competing interests. Consent for publication {#FPar3} ======================= No individual patient information used.

Raw deep sequencing cannot be provided because these data are derived from patient samples. Because of the Minimum Necessary Requirement of the HIPAA Privacy Rule these data may not be deposited into a public repository. A Supporting Information File accompanying the submission contains the de-identified results and interpretations of all clinical NGS tests included in our analysis next to the results of single-gene testing of the same specimen. Introduction {#sec001} ============ The advance of next-generation sequencing (NGS) is a cornerstone of a recent development in molecular pathology, variably referred to as "personalized," "precision," or "individualized" medicine. Much of the focus of clinical NGS has been on oncology, as there are clear diagnostic, prognostic, and therapeutic implications for a multitude of genomic mutations in both solid and liquid malignancies. For example, in non-small cell lung cancers, activating mutations of the *EGFR* gene predict therapeutic response to tyrosine kinase inhibitors (TKI) such as erlotinib, gefitinib, and afatinib \[[@pone.0152851.ref001]--[@pone.0152851.ref003]\]. The KRAS protein acts downstream of EGFR, and thus mutations of the *KRAS* gene predict resistance to TKI \[[@pone.0152851.ref002],[@pone.0152851.ref004]--[@pone.0152851.ref006]\]. In metastatic melanoma, *BRAF* V600 mutations predict response to dabrafenib, vemurafenib, and trametinib \[[@pone.0152851.ref007]\]. Mutations of the *FLT3* and *NPM1* genes affect the prognosis of karyotypically normal acute myeloid leukemia and aid in the decision whether or not to pursue hematopoietic stem cell transplantation \[[@pone.0152851.ref008]\]. In addition, activating mutations of *JAK2*, which encodes a tyrosine kinase essential for cytokine and growth factor signaling, are found in a large proportion of patients with myeloproliferative neoplasms \[[@pone.0152851.ref009]\]. Ruxolitinib is a JAK1/JAK2 inhibitor approved for the treatment of myelofibrosis \[[@pone.0152851.ref010]\], and additional JAK2 inhibitors are in clinical development \[[@pone.0152851.ref011]\]. Detection of mutations in *EGFR*, *KRAS*, *BRAF*, *FLT3*, *NPM1*, and *JAK2* is most commonly accomplished by targeted tests that are designed to detect one or at most a small number of mutations in a single gene. However, NGS is gaining momentum as a complementary test for a number of reasons. Firstly, clinical trials for targeted cancer therapies rely on detection of mutations that are frequently not covered by existing targeted tests. In contrast to the laborious and lengthy process of validating and implementing a new molecular assay testing for one or a few mutations, NGS greatly simplifies the task of providing coverage of one of more additional mutations of interest. Secondly, targeted tests can provide misleading results. For example, the widely used FDA-approved cobas® *EGFR* Mutation Test only detects exon 19 deletion and L858R mutations, which together only comprise the mutations found in 85% of *EGFR*-mutated lung cancers \[[@pone.0152851.ref012]\]. In a significant proportion of cases, this test fails to identify therapeutically targetable mutations. Thirdly, targeted tests may fail to detect the very mutation they are designed to detect. Our group has recently reported a striking failure of two separate single-gene tests for the *BRAF* gene to detect a V600E mutation in a melanoma specimen \[[@pone.0152851.ref013]\]. The mutation was clearly demonstrated by concurrent NGS analysis. Finally, as has recently become apparent, tumors frequently harbor mutations that are therapeutically targetable but are not typically seen in that tumor type. Due to its massively parallel nature, NGS is very well-suited for detecting mutations in unexpected genes. One study reported a three-fold increased yield of clinically actionable mutations with NGS as compared to traditional molecular approaches targeting mutation hotspots \[[@pone.0152851.ref014]\]. Multiple recent studies have investigated the potential utility of NGS for detection of clinically actionable cancer mutations with encouraging results \[[@pone.0152851.ref014]--[@pone.0152851.ref025]\]. With the exception of one study \[[@pone.0152851.ref021]\], all demonstrated excellent performance of NGS on various platforms as measured by detection of point mutations and small insertions and deletions (indels). In many cases, additional potentially important variants were uncovered by NGS. All of the aforementioned studies were designed to validate clinical NGS pipelines that were not yet in clinical practice. As a consequence, they enrolled selected samples from previously examined specimens. With the exception of one commercially-sponsored study \[[@pone.0152851.ref014]\], in all cases a very limited number of samples was re-examined (range 13--61), often from only a single tissue type. While these important contributions confirm the potential usefulness of clinical NGS, they do not address the important question whether a well-validated NGS pipeline performs at an acceptable level in day-to-day clinical practice. Here we present a summative analysis of mutation results and quality control metrics obtained during the first year (March 2013 through March 2014) of clinical solid and liquid malignancy NGS carried out at the Center for Personalized Diagnostics at the University of Pennsylvania Health System. More than 900 specimens were submitted and processed during this time frame. We report that using our validated molecular and bioinformatics pipeline \[[@pone.0152851.ref026]\] with pre-determined tumor percentage and DNA quality cutoffs, we achieved excellent NGS data quality as determined by virtually perfect concordance between NGS and targeted single-gene tests for various genes in a large number of solid and liquid malignancy specimens. Materials and Methods {#sec002} ===================== Specimen Characteristics and Processing {#sec003} --------------------------------------- Over the course of the study duration, 938 liquid and solid tumor specimens were submitted to the Center for Personalized Diagnostics ([Table 1](#pone.0152851.t001){ref-type="table"}). Specimens were eligible for NGS if they passed the tumor percentage, DNA quality, and DNA quantity thresholds that had been determined at the time of the validation of the NGS assay, which preceded the study period. Briefly, specimens with \<10% tumor were not eligible for NGS, because sequencing of samples with lower tumor percentages frequently yielded changes that were represented in fewer than five unique reads, making it difficult to distinguish true variants from sequencing artifacts. For similar reasons, DNA quality and quantity were judged to be insufficient, and the specimen was ineligible for NGS, if the DNA concentration was \<1 ng/μL; the DNA concentration was \<5 ng/μL with \>20% DNA degraded; the DNA concentration was \<50 ng/μL with \>45% DNA degraded; or DNA degradation was \>60%. Degraded DNA was defined as the proportion of DNA under 1000 bp in length. 10.1371/journal.pone.0152851.t001 ###### Characteristics of Specimens by Tumor Site. {#pone.0152851.t001g} Tumor site Number of specimens submitted for NGS Specimens with DNA quality or quantity inadequate for NGS analysis Number of "shared" specimens (i.e., results are available from both NGS and targeted tests) ------------------ --------------------------------------- -------------------------------------------------------------------- --------------------------------------------------------------------------------------------- Lung 196 13 (6.6%) 101 (51.5%) Brain 174 9 (5.2%) 6 (3.4%) Bone Marrow 161 0 95 (59.0%) Lymph Node 70 8 (11.4%) 32 (45.7%) Peripheral blood 56 0 29 (51.8%) Liver 36 4 (11.1%) 5 (13.9%) Skin 23 2 (8.7%) 5 (21.7%) Other 222 24 (10.8%) 31 (14.0%) Tumor site is not necessarily tissue of origin. Please note that in contrast to solid tumor specimens, all liquid specimens (bone marrow and peripheral blood) were adequate for NGS processing. To determine the tumor percentage and volume of solid tumors, hematoxylin- and eosin-stained tissue specimens were evaluated by an anatomic pathologist, and the region with the highest tumor burden was marked. Genomic DNA was extracted from fresh bone marrow or peripheral blood using the Gentra Puregene Cell Kit (Qiagen, Netherlands). For formalin-fixed, paraffin-embedded (FFPE) specimens, tissues were macro-dissected from 5 μM or 10μM slides. Scrapings were dewaxed with Qiagen Deparaffinization Solution and purified with Gentra Puregene Tissue reagents following the manufacturer's protocol (Qiagen, Netherlands). DNA quantification was performed using the Qubit Broad Range assay following manufacturer's protocols (Life Technologies, CA). Agilent Genomic TapeScreens were used following manufacturer's protocols to assess the degree of DNA degradation (Agilent, CA). Targeted Molecular Testing {#sec004} -------------------------- ### *EGFR*, *KRAS*, and *BRAF* Assays {#sec005} The mutational status of *EGFR* exons 19 and 21 was determined using a laboratory-developed test (LDT) as previously described \[[@pone.0152851.ref027]\]. Briefly, genomic DNA was extracted from FFPE tissue and amplified with primers covering two regions, one that is commonly deleted in exon 19, and a part of exon 21 that encompasses codon 858. The L858R missense mutation in exon 21 creates a new *Sau*96I cleavage site within exon 21. The amplification products were digested with *Sau*96I and then separated by capillary electrophoresis. *KRAS* mutations in codons 12 and 13 were assayed using a LDT as previously described \[[@pone.0152851.ref028]\]. Briefly, genomic DNA was extracted and amplified using primers designed to detect point mutations, hybridized to target-specific capture probes, and subjected to a bead assay (Lumina, TX). *BRAF* mutations were assayed by pyrosequencing of an amplified portion of the *BRAF* gene including codon 600, as previously described \[[@pone.0152851.ref029]\]. ### *FLT3* Assay {#sec006} DNA was extracted using the QIAamp DNA Blood Mini Kit (Qiagen, Netherlands). Mutation analysis of the *FLT3* gene was performed using multiplex PCR amplification with two sets of fluorescently labeled primers. For internal tandem duplication (ITD) detection, PCR was performed with the following primers: 5'-GCA ATT TAG GTA TGA AAG CCA GC-3' (forward) and 5'-CTT TCA GCA TTT TGA CGG CAA CC-3' (reverse); forward primers were labeled with 6-carboxyfluorescin (6-FAM), and reverse primers were labeled with VIC. An internal tandem duplication (ITD) was determined to be present if a product larger than the wild-type (329 bp) product was detected by capillary electrophoresis. Detection of the D835 mutation of *FLT3* (NM_004119.2: c.2503_2505) was based on the fact that this mutation abolishes an *Eco*RV cleavage site. PCR was performed with the following primers: 5'-GTA AAA CGA CGG CCA GCC GCC AGG AAC GTG CTT-3' (forward) and 5'-CAG GAA ACA GCT ATG ACG ATA TCA GCC TCA CAT TGC CCC-3' (reverse); forward primers were labeled with NED at the 5' end. After *Eco*RV digestion, PCR products were analyzed by capillary electrophoresis using a 3500xL Genetic Analyzer (Life Technologies, NY). A D835 point mutation was indicated by the presence of a 129 bp fragment. ### *NPM1* Assay {#sec007} Total RNA was extracted using the QIAamp RNA Blood Mini Kit (Qiagen, Netherlands), reverse transcribed, and amplified in a multiplex PCR reaction using primers designed to detect common mutations in *NPM1* (NM_002520.4) using the Signature *NPM1* Mutations Assay (Asuragen, TX). Labeled PCR products were hybridized to target-specific capture probes covalently bound to fluorescent microspheres in a liquid bead array followed by analysis with a Luminex 100 (Luminex, TX). Interpretation was based on the mean fluorescence intensity (MFI) obtained from a minimum of 50 microspheres. ### *JAK2* Assay {#sec008} Genomic DNA was isolated from leukocytes using the QIAamp DNA Blood Mini Kit (Qiagen, Netherlands) and amplified using real-time PCR with primers flanking *JAK2* codon 617. Allelic discrimination between the normal sequence and the *JAK2* V617F (NM_004972 c.1849G\>T) mutation was subsequently accomplished by simultaneous differential hybridization of two sequence-specific probes, each labeled with a different fluorescent marker (MutaScreen Assay, Qiagen, Netherlands). Next-Generation Sequencing and Bioinformatic Analysis {#sec009} ----------------------------------------------------- For solid tumors, target enrichment was performed with the TruSeq Amplicon Cancer Panel (Illumina, CA), a cancer gene panel consisting of 212 target amplicons covering mutation hotspots of 47 cancer genes. For hematologic malignancies, an in-house developed gene panel was utilized, which is composed of 382 amplicons covering 33 genes. For all successful sequencing runs, read depth was 250x at any given position, with 1000x mean coverage across the entire targeted sequence, and a Q30 at greater than 75% of reads. An in-house bioinformatics pipeline \[[@pone.0152851.ref026]\] was used to map reads, detect variants, and annotate them. Reads were de-multiplexed, mapped to the hg19 version of the human reference genome, filtered to remove off-target and poor-quality reads ([Fig 1](#pone.0152851.g001){ref-type="fig"}). Using custom scripts, four types of variants were extracted: single-nucleotide variants (SNVs), small indels, copy number variants, and large indels. Variants were then compared to an in-house developed knowledge base, which draws from publicly available sources such as PubMed, dbSNP database \[[@pone.0152851.ref030]\], COSMIC database \[[@pone.0152851.ref031]\], 1000 Genomes \[[@pone.0152851.ref032]\], and the Exome Variant Server ([http://evs.gs.washington.edu](http://evs.gs.washington.edu/)). Using this knowledge base, variants were classified into 1 of 5 categories: disease associated mutation (DAM), likely pathogenic mutation (LPM), variants of uncertain significance (VUS), likely benign (LB), or benign (B). DAMs include mutations previously reported and associated with disease, including gain-of-function mutations in oncogenes (e.g., the canonical *KRAS* G12D mutation) and truncating mutations in known tumor suppressor genes. LPMs were classified as variants that had some evidence of disease association, such as case reports, but are not well described otherwise. Variants were classified as VUS if they had not been previously reported either as a disease-associated mutation or as a normal variant on the Exome Variant Server, but whose pathogenicity could not be established with certainty. Variants were classified as LB if there was no report of association with disease and if they occurred in regions of the gene not predicted to have pathologic consequences. Variants noted on the clinical report, which are the ones included in this analysis, included the DAM, LPM and VUS (not LB or B). All reported variants were manually reviewed using the Integrative Genomics Viewer (IGV) \[[@pone.0152851.ref033]\] by at least two pathologists. {#pone.0152851.g001} Ethics Statement {#sec010} ---------------- Patient data were analyzed anonymously in accordance with institutional practice guidelines. The institutional review board of the University of Pennsylvania determined this study to be exempt. Results {#sec011} ======= Study Design and Specimen Characteristics {#sec012} ----------------------------------------- The University of Pennsylvania Health System began routine clinical NGS of solid tumors and hematologic malignancies in February of 2013. During validation of our NGS pipeline, we established thresholds for acceptable tumor percentage (10%) and DNA quantity and quality (see [Methods](#sec002){ref-type="sec"} for details). For a large number of specimens, clinicians ordered targeted single-gene tests and NGS analysis on the same specimen ([Table 1](#pone.0152851.t001){ref-type="table"}): this occurred frequently with pulmonary and hematological specimens, but relatively rarely with specimens derived from other sites such as brain. We reasoned that comparing results from targeted and NGS testing of all "shared" specimens (i.e. specimens that had been analyzed by both NGS and single-gene tests) would allow us to probe the robustness of our NGS pipeline. We therefore set out to compare NGS and targeted test results in specimens that underwent NGS analysis during the first year of operation of the NGS pipeline (March 1, 2013 through March 1, 2014). During this time, 938 specimens (717 solid and 221 hematologic) were submitted for NGS analysis. While the majority of solid tumor specimens were composed of \>50% tumor cells ([Fig 2](#pone.0152851.g002){ref-type="fig"}), a fraction of solid tumors was not tested due to low tumor percentage. The most common tumor sites were lung, brain, bone marrow, lymph nodes, and peripheral blood ([Table 1](#pone.0152851.t001){ref-type="table"}). A small fraction of solid tumor specimens could not be analyzed by NGS due to inadequate DNA quantity or quality (*n* = 60, 8.3%), however this was not an issue with peripheral blood and bone marrow specimens. {#pone.0152851.g002} Comparison of *EGFR* and *KRAS* Gene Mutations by NGS and Targeted Testing {#sec013} -------------------------------------------------------------------------- We were particularly interested in the performance of *EGFR* testing in our NGS assay, because mutations in this gene may be therapeutically targetable, and because it is often challenging to analyze, as DNA degradation and low tumor percentage are frequently encountered in lung cancer specimens. Since *KRAS* mutations were frequently evaluated alongside *EGFR* in cases of lung cancers at our institution, we also wanted to compare the performance of both methods on this gene. The tumor percentages for all specimens for which targeted *EGFR* mutation analysis was reported during the study period (*n* = 283) are shown in [Fig 3](#pone.0152851.g003){ref-type="fig"}. Approximately 10% of the cases evaluated by targeted testing had a tumor percentage below the cutoff for NGS. {#pone.0152851.g003} We compared NGS and targeted testing in 139 specimens that had been tested for *EGFR* mutations ([Fig 4A](#pone.0152851.g004){ref-type="fig"}) and 138 that had been tested for *KRAS* mutations by both methods ([Fig 4B](#pone.0152851.g004){ref-type="fig"}). Among the shared specimens, all generated a result with the targeted single-gene methods. However, 15/139 (11%) of shared *EGFR*-tested and 13/138 (9%) of shared *KRAS*-tested specimens were excluded from NGS analysis due to insufficient DNA quality or quantity. {#pone.0152851.g004} In the remaining specimens, all mutations detected by the targeted assays were also detected by NGS. Conversely, NGS identified a number of mutations that the targeted tests were not designed to detect. For example, our targeted *EGFR* mutation test only covers deletions in exon 19 and the L858R mutation in exon 21. Similarly, the targeted *KRAS* test only detects mutations in codons 12 and 13. In 10 *EGFR* shared cases (6%), additional pathogenic mutations were detected by NGS. In two of these cases, the T790M mutation was found, which predicts resistance to TKI therapy \[[@pone.0152851.ref004]\]. Additionally, NGS detected *EGFR* amplification in five cases; the predictive value of this copy number alteration in the context of TKI therapy is currently unclear. In four *KRAS* shared cases (3%), NGS detected mutations in codon 61, which predict resistance to TKI therapy \[[@pone.0152851.ref034]\]. Comparison of *BRAF* Gene Mutations by NGS and Targeted Testing {#sec014} --------------------------------------------------------------- The *BRAF* single-gene test was performed less frequently in parallel with solid tumor NGS than the *EGFR* or *KRAS* tests, because thyroid fine-needle aspiration samples, for which *BRAF* testing was frequently ordered, were not originally validated for the NGS assay. Of 224 specimens that were tested with targeted *BRAF* tests, 38 were also analyzed by NGS ([Fig 4C](#pone.0152851.g004){ref-type="fig"}). Among the shared specimens, all generated a result with the targeted *BRAF* test. In contrast, 8/38 shared cases (21%) could not be analyzed by NGS due to poor DNA quality or inadequate DNA quantity. All remaining shared specimens showed perfect concordance between NGS and targeted testing. Comparison *FLT3*, *NPM1*, and *JAK2* Mutations by NGS and Targeted Testing {#sec015} --------------------------------------------------------------------------- Of 221 hematologic specimens tested by NGS during the study period, 118 were also tested with the *FLT3*, and 98 with the *NPM1* targeted tests. In two cases, the targeted *FLT3* test detected an internal tandem duplication (ITD) that was not automatically called by the NGS analysis pipeline ([Fig 5A](#pone.0152851.g005){ref-type="fig"}). However, manual review of the sequencing data demonstrated ITD mutations at allele frequencies of 1.3% and 1.6% ([Fig 6A and 6C](#pone.0152851.g006){ref-type="fig"}, respectively). In one of these two cases, the hematologic malignancy NGS panel additionally detected a pathogenic *FLT3* D839G (c. 2516A\>G) mutation ([Fig 6B](#pone.0152851.g006){ref-type="fig"}); this mutation was not detected by the *FLT3* single-gene test that is designed to detect only ITDs and D835 mutations. {#pone.0152851.g005} {#pone.0152851.g006} In the remainder of *FLT3* shared cases, and also in all *NPM1* shared cases ([Fig 5B](#pone.0152851.g005){ref-type="fig"}), NGS and targeted tests were concordant. Additionally, a small number of specimens (*n* = 8) that had been tested for *JAK2* mutations by both modalities were analyzed and showed complete concordance ([Fig 5C](#pone.0152851.g005){ref-type="fig"}). Discussion {#sec016} ========== In this study, we report the properties of solid and liquid malignancy specimens processed during the first year of clinical oncologic NGS performed within the University of Pennsylvania Health System. We found that when we adhered to two predetermined quality control metrics, i.e., tumor percentage and DNA quantity and quality, we achieved excellent NGS data as determined by virtually perfect concordance between NGS and targeted, single-gene testing. While a number of recent studies have confirmed the potential utility of clinical NGS for oncology \[[@pone.0152851.ref014]--[@pone.0152851.ref025]\], there have been none, to our knowledge, that have evaluated the quality of data generated during day-to-day practice at a clinical oncologic sequencing facility. Additionally, in contrast to previous validation studies, we examined a larger number of specimens from a greater variety of tissues, and our data is therefore less subject to sample selection bias and may also more accurately reflect the expected annual case volume and distribution of tissue types encountered at a major academic medical center. Solid tumors and hematologic malignancy specimens differed considerably in their performance across the established quality control measures. We found that a sizable fraction of solid but not liquid specimens yielded DNA of insufficient quality or quantity for NGS testing. With respect to specimens tested for *EGFR* and *KRAS* (predominantly lung), this was likely largely due to formalin fixation, which degrades DNA through cross-linking as well as other less well understood mechanisms \[[@pone.0152851.ref035]\]. One potential approach to reduce the number of samples that are currently rejected from NGS analysis might be to determine the amplifiability of extracted genomic DNA, for example by using the human genomic DNA quantitation and quality control assay by Kapa Biosystems (Wilmington, MA). Our study highlights the complementarity of NGS and targeted tests for mutation detection. In a number of instances, NGS detected clinically important mutations that were not captured by targeted assays. For example, the *EGFR* test specifically interrogates potential exon 19 deletions and L858R mutations, which constitute about 85% of EGFR mutations in lung cancers \[[@pone.0152851.ref012]\]. In 9 shared cases, *EGFR* mutations were found in different regions of the gene by the NGS assay ([Fig 4A](#pone.0152851.g004){ref-type="fig"}). Similarly, clinically important *KRAS* mutations occur in codons 12, 13, and 61, but only codons 12 and 13 were evaluated by the in-house *KRAS* single-gene test. In 4 shared cases, *KRAS* mutations in codon 61 were detected by NGS only. Targeted assays generally outperformed NGS in specimens with low tumor percentage, DNA quantity, or quality. In fact, we identified actionable mutations by targeted analysis in multiple cases that were unable to be analyzed by NGS (data not shown). In two specimens, a *FLT3* ITD mutation, which was not automatically called by the NGS pipeline, was readily detected by the targeted assay. While *FLT3* ITDs can be challenging to detect by NGS due to the complex structure of the mutation \[[@pone.0152851.ref036]\], the problem in these two cases was that the allele frequency fell below the validated threshold for automatic detection of 4%. However, manual inspection of the *FLT3* exon 14 and the flanking intronic sequence using the Integrative Genomics Viewer (IGV) \[[@pone.0152851.ref033]\] clearly showed the presence of ITD mutations in both cases ([Fig 6A and 6C](#pone.0152851.g006){ref-type="fig"}). The detection of NGS for indel mutations, in contrast to single nucleotide variants, is not fundamentally limited by PCR artifacts and sequencing errors. Therefore, we validated our pipeline for the detection of *FLT3* ITDs with allele frequencies of 1%, and we altered the indel allele frequency calling threshold specifically for *FLT3* ITDs. Reanalysis of the two *FLT3* cases with updated parameters, which was part of the revalidation study, revealed the expected results. These findings highlight the importance of manual review of the sequencing data. Low tumor percentage was also limiting in solid tumor cases. Approximately 10% of specimens submitted for *EGFR* mutation testing were found to contain less than 10% tumor ([Fig 3](#pone.0152851.g003){ref-type="fig"}), thereby failing tumor percentage requirements for NGS testing. All of these specimens generated a result with the targeted assay. It should be noted, however, that as NGS methodologies continue to mature, the ability of NGS to detect single nucleotide variants in specimens with lower tumor percentage will improve. Rare variant detection by NGS is hampered by the high rate of amplification and sequencing errors, which can approach 1% for single nucleotide variants \[[@pone.0152851.ref037]\]. Various approaches to increase analytic sensitivity for rare variants by NGS have recently been described, including barcoding of each DNA fragment before amplification \[[@pone.0152851.ref038]--[@pone.0152851.ref040]\], barcoding of both strands of each fragment \[[@pone.0152851.ref041]\], and generation of multiple linked tandem copies of each DNA fragment by rolling circle amplification \[[@pone.0152851.ref042]\]. In particular, the latter method has been shown to improve sensitivity of rare variant detection by more than 100-fold without introducing excessive computational inefficiency. There are a number of limitations to this study. First, only genes for which both NGS analysis and single-gene tests are performed at our institution were included in the analysis. While the genes examined in this study currently represent the essential core of cancer gene testing, it is possible that NGS might function less reliably with certain other genes or specific mutations that were not assessed. Of note, our NGS pipeline produced excellent results for the challenging genes *EGFR* and *FLT3*. We therefore expect that most other genes covered by our cancer panels generate NGS data of similarly high quality. Second, not all specimens that were submitted for NGS analysis were also tested by single-gene methods. For example, only 3.4% of brain tumor specimens submitted for NGS were at the same time examined by targeted tests, and therefore our conclusions may not necessarily extend to all tumors at this time. Additionally, the specimens tested did not contain the entire spectrum of mutations that can be evaluated by our targeted tests. Finally, our findings may not be generalizable to other platforms, especially those that utilize more complex gene panels. Recently issued recommendations for validation and quality control of clinical NGS data \[[@pone.0152851.ref043]\] include monitoring of quality metrics (e.g., sequencing quality scores, depth of coverage, uniformity of coverage, mapping quality), proficiency testing, and confirmation of actionable results by independent methods. Multiple studies that appeared after these recommendations were published have established that particularly for solid tumors, tumor percentage and DNA quality are important additional metrics \[[@pone.0152851.ref014],[@pone.0152851.ref022],[@pone.0152851.ref023]\], and we found this as well in our validation studies (not shown). Proficiency testing for cancer NGS is not yet available but is currently being developed by the College of American Pathologists (CAP). A proposed requirement to confirm actionable mutations by independent molecular methods \[[@pone.0152851.ref043]\] has been called into question \[[@pone.0152851.ref023]\]. Since single-gene tests such as Sanger sequencing may not have inherently greater sensitivity than a well-scaled NGS pipeline \[[@pone.0152851.ref015],[@pone.0152851.ref044]\], verification by these methods might not improve the accuracy of test results. Accordingly, very recent guidelines by the College of American Pathologists leave the decision when and how to perform confirmatory testing of NGS results to the clinical laboratory \[[@pone.0152851.ref044]\]. Our *EGFR* single-gene test had greater analytical sensitivity than NGS, detecting mutations in samples with well under 10% tumor. However, our finding that within pre-defined tumor percentage and DNA quality cutoffs NGS showed perfect concordance confirms the notion that it is unnecessary to confirm each actionable mutation detected via NGS by a single-gene test. In addition, single-gene tests frequently do not cover important disease-associated mutations and in some instances fail to detect the very mutations they target \[[@pone.0152851.ref013]\]. Based on these considerations, we propose a workflow that integrates NGS as an adjunct diagnostic modality for solid and liquid neoplasms ([Fig 7](#pone.0152851.g007){ref-type="fig"}). The typical turn-around time of the NGS assay at the time of the study was around 7--10 days from receipt in the sequencing facility to the time the final report was signed out in the electronic medical record. Turnaround is generally faster for targeted tests, and thus it is advantageous to perform targeted tests, perhaps in addition to NGS, in clinically urgent situations. Specimens with low tumor percentage or DNA quantity or quality should be subjected to targeted tests that we found to be more analytically sensitive than NGS. On the other hand, when DNA quality and quantity is adequate and a turnaround time of 7--10 days is acceptable, NGS holds clear advantages, including increased clinical sensitivity within targeted genes of interest and additional information from other covered genes on the panel. In these cases, our data suggest that targeted tests may be safely omitted. {#pone.0152851.g007} Supporting Information {#sec017} ====================== ###### Raw Data for Comparison Analysis. (XLSX) ###### Click here for additional data file. We thank Daniel Desloover for comments on the manuscript. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: MCH DBR SK CDW RDD JJDM. Performed the experiments: MCH SK. Analyzed the data: MCH SK. Contributed reagents/materials/analysis tools: MCH SK DL JZ CDW RDD JJDM. Wrote the paper: MCH SK JZ CDW RDD JJDM. [^3]: Current address: BioReference Laboratories, Elmwood Park, New Jersey, United States of America [^4]: Current address: Department of Pathology, Massachusetts General Hospital, Boston, Massachusetts, United States of America [^5]: ‡ These authors also contributed equally to this work

Introduction {#s1} ============ All organisms live in environments that vary through time and such environmental heterogeneity can impose highly variable selection pressures on populations. In this situation, an allele may be beneficial during one environmental regime and subsequently deleterious during another. Such an allele would be subject to short bursts of directional selection, alternately being favored and disfavored. When this situation occurs in diploids, the heterozygote can have a higher geometric mean fitness than either homozygote and allelic variation at this locus could be maintained for long periods despite being subject to directional selection at any given time [@pgen.1004775-Gillespie1]--[@pgen.1004775-Hedrick1]. This situation is referred to as marginal overdominance and is a form of balancing selection. There is substantial evidence for the maintenance of phenotypic and genetic variation by temporally variable selection in a variety of organisms. For instance, evolutionary response to rapid changes in selection pressures has been demonstrated for morphological and life-history traits in mammals [@pgen.1004775-Gershenson1], [@pgen.1004775-Grant1], birds [@pgen.1004775-Tarwater1]--[@pgen.1004775-Wall1], plants [@pgen.1004775-Brakefield1], invertebrates [@pgen.1004775-Hairston1]--[@pgen.1004775-RodriguezTrelles1], and others (reviewed in [@pgen.1004775-Bell1], [@pgen.1004775-Siepielski1]). Chromosomal inversions and allozyme alleles in a variety of drosophilids vary among seasons [@pgen.1004775-Dobzhansky1]--[@pgen.1004775-Ananina1] suggesting that these polymorphisms confer differential fitness in alternating seasons. Further, in some species of drosophilids, life-history [@pgen.1004775-Bouletreaumerle1], [@pgen.1004775-Schmidt1], morphological [@pgen.1004775-Stalker1], [@pgen.1004775-Tantawy1] and stress tolerance traits [@pgen.1004775-Miyo1], [@pgen.1004775-Dev1] also fluctuate seasonally suggesting that these traits respond to seasonal shifts in selection pressures. Although theoretical models suggest that temporal variation in selection pressures can maintain fitness-related genetic variation in populations [@pgen.1004775-Gillespie1]--[@pgen.1004775-Hedrick1] and empirical evidence from a variety of species [@pgen.1004775-Gershenson1]--[@pgen.1004775-Dev1] demonstrates that variation in selection pressures over short time periods does alter phenotypes and allele frequencies, we still lack a basic understanding of many fundamental questions about the genetics and evolutionary history of alleles that undergo rapid adaptation in response to temporal variation in selection pressures. Specifically, we do not know how many loci respond to temporally variable selection within a population, the strength of selection at each locus, nor the effects of such strong selection on neutral genetic differentiation through time. We do not know whether adaptation at loci that respond to temporally variable selection is predictable nor do we know the relationship between loci that respond to temporally variable selection and spatially varying selection. Finally, it is unclear whether rapid adaptation to temporally variable selection pressures is primarily fueled by young alleles that constantly enter the population but cannot be maintained for long periods of time or, rather, by old alleles that have possibly been maintained by variable selection associated with environmental heterogeneity despite short bursts of strong directional selection. To address these questions, we estimated allele frequencies genome-wide from samples of *D. melanogaster* individuals collected along a broad latitudinal cline in North America and in the spring and fall over three consecutive years in a single temperate orchard. We demonstrate that samples of flies collected in a single Pennsylvania orchard over the course of several years are as differentiated as populations separated by 5--10° latitude. We identify hundreds of polymorphisms that are subject to strong, temporally varying selection and argue that genetic draft [@pgen.1004775-Gillespie3] in the wake of rapid, multilocus adaptation is sufficient to explain the high degree of genetic turnover that we observe in this population over several years. We examine the genome-wide relationship between spatial and temporal variation in allele frequencies and find that spatial genetic differentiation, but not clinality *per se*, in allele frequency is a good predictor of temporal variation in allele frequency. Moreover, at SNPs subject to seasonal fluctuations in selection pressures, northern populations are more similar to spring populations than southern ones are. Next, we show that allele frequencies at SNPs subject to seasonal fluctuations in selection pressures become more 'spring-like' (i.e., they move towards the average spring frequency) immediately following a hard frost event and that seasonally variably SNPs tend to be associated with two seasonally variable phenotypes, chill coma recovery time and starvation tolerance. Finally, we demonstrate that some of the loci that respond to temporal variation in selection pressures are likely ancient, balanced polymorphisms that predate the split of *D. melanogaster* from its sister species, *D. simulans*. Taken together, our results are consistent with a model in which temporally variable selection maintains fitness-related genetic variation at hundreds of loci throughout the genome for millions of generations if not millions of years. Results/Discussion {#s2} ================== Genomic differentiation through time and space {#s2a} ---------------------------------------------- To test for the genomic signatures of balancing selection caused by seasonal fluctuations in selection pressures, we performed whole genome, pooled resequencing of samples of male flies collected in the spring and fall over three consecutive years (2009--2011) in a temperate, Pennsylvanian orchard. We contrast changes in allele frequencies through time with estimates of allele frequencies we made from five additional populations spanning Florida to Maine along the east coast of North America over a number of years (2003--2010) largely during periods of peak abundance of *D. melanogaster* ([Fig. 1A](#pgen-1004775-g001){ref-type="fig"}, [Table S1](#pgen.1004775.s008){ref-type="supplementary-material"}). From each population and time point, we sampled approximately 50--100 flies and resequenced each sample to average read depth of 20--200× coverage ([Table S1](#pgen.1004775.s008){ref-type="supplementary-material"}, and see [Text S1](#pgen.1004775.s011){ref-type="supplementary-material"}). Estimates of allele frequency using this sampling design have been shown to be highly accurate [@pgen.1004775-Gillespie3]. ![Experimental design and genomic turnover through time and space.\ (A) Map of sampling locations in North America used in this study. Grey boxes represent individual samples from each locale. Genome-wide differentiation among spatially (B) and temporally (C) separated samples, measured as genome-wide average *F~ST~* (y-axis). Lines represent the predicted value of *F~ST~* based on the linear (A; y = a+bx) and non-linear (B; y = ab^X^) regression. Note: Pennsylvanian samples are not represented in (B) and the negative *F~ST~* in (B) results from the conservative correction of heterozygosity [@pgen.1004775-Rashkovetsky1], [@pgen.1004775-Turelli1]. In addition, please note that there are four estimates of pairwise *F~ST~* between the two replicate Maine and Florida samples (corresponding to a difference in latitude of 20°) and that there are two estimates of *F~ST~* between each of the remaining clinal populations and each Maine and Florida replicate sample. Error bars represent 95% confidence intervals based on 500 blocked bootstrap samples of ∼2000 SNPs.](pgen.1004775.g001){#pgen-1004775-g001} As a point of departure and to provide context for understanding the magnitude of genetic variation through the seasons, we first examined genetic differentiation along the cline ([Fig. 1B](#pgen-1004775-g001){ref-type="fig"}, [Fig. S1A](#pgen.1004775.s001){ref-type="supplementary-material"}). We calculated genome-wide average *F~ST~* among pairs of populations (excluding Pennsylvanian populations; hereafter 'spatial *F~ST~*') as well as the proportion of SNPs where average spatial *F~ST~* between a pair of populations is greater than expected by chance conditional on our sampling design and assuming panmixia using allele frequency estimates of 500,000 common polymorphisms ([Table S1](#pgen.1004775.s008){ref-type="supplementary-material"}). Genome-wide average spatial *F~ST~* ([Fig. 1B](#pgen-1004775-g001){ref-type="fig"}) as well as the proportion of SNPs where spatial *F~ST~* is greater than expected by chance ([Fig. S1A](#pgen.1004775.s001){ref-type="supplementary-material"}) is positively correlated with geographic distance (*r* = 0.75; *p* = 7e-5), a pattern consistent with isolation by distance [@pgen.1004775-Wright1]. Pooled resequencing did identify polymorphisms in or near genes previously shown to be clinal in North American populations (see [Text S1](#pgen.1004775.s011){ref-type="supplementary-material"}) demonstrating that clines are stable over multiple years. This suggests that populations sampled along the cline represent resident populations, and further confirms that our pooled resequencing design gives accurate estimates of allele frequencies [@pgen.1004775-Zhu1]. Next, we calculated genome-wide average *F~ST~* between samples collected through time in the Pennsylvanian population ('temporal *F~ST~*') as well as the proportion of SNPs where average temporal *F~ST~* is greater than expected by chance given our sampling design and assuming no allele frequency change through time ([Fig. 1C](#pgen-1004775-g001){ref-type="fig"}, [Fig. S1B](#pgen.1004775.s001){ref-type="supplementary-material"}). Genome-wide average temporal *F~ST~* ([Fig. 1C](#pgen-1004775-g001){ref-type="fig"}) as well as the proportion of SNPs where the observed temporal *F~ST~* is greater than expected by chance ([Fig. S1B](#pgen.1004775.s001){ref-type="supplementary-material"}) increases with the difference in time between samples. The temporal *F~ST~* increases non-linearly with duration of time between samples (*slope* ~log-log~ = 0.59, *p* ~log-log\ slope = 1~ = 0.0004, *df* = 19). Genome-wide average temporal *F~ST~* appears to asymptote by ∼7 months, corresponding to the duration of time between fall samples and the subsequent spring sample. Remarkably, samples of the Pennsylvanian population collected one to three years apart are as differentiated as populations separated by 5--10° latitude, demonstrating high genetic turnover through time. Identification and genomic features of seasonal SNPs {#s2b} ---------------------------------------------------- We sought to identify alleles whose frequency consistently and repeatedly oscillated between spring and fall over three years with the assumption that these polymorphisms would be the most likely to be adaptively responding to selection pressures that oscillate between the seasons. We identified seasonally variable polymorphisms that had a large and recurrent deviation from spring to fall around the average frequency using a generalized linear model (GLM) of allele frequency change as a function of season (spring or fall) that took into account read depth and the number of sampled chromosomes (see [Materials and Methods](#s3){ref-type="sec"} for details). Of the ∼500,000 common SNPs tested, we identified approximately 1750 sites that cycle approximately 20% in frequency between spring and fall at FDR less than 0.3 (hereafter 'seasonal SNPs'; [Fig. 2A](#pgen-1004775-g002){ref-type="fig"}, [Fig. S2A](#pgen.1004775.s002){ref-type="supplementary-material"}). Statistically significant changes in allele frequency of this magnitude at seasonal SNPs correspond to selection coefficients of 5--50% per locus per generation ([Fig. 2B](#pgen-1004775-g002){ref-type="fig"}, see [Materials and Methods](#s3){ref-type="sec"}), assuming 10 generations per summer or 1--2 generations per winter. Given the statistical power of our experiment ([Fig. 2B](#pgen-1004775-g002){ref-type="fig"}), we estimate there may be as many as 10 times as many sites that could cycle either directly in response to seasonally varying selection or could be linked to seasonal SNPs. {ref-type="sec"} for the definition of *S*. (C) Enrichment (log~2~ odds ratio) of seasonal SNPs that are annotated for each class of genetic element relative to control polymorphisms. (D) Seasonal *F~ST~* surrounding seasonal SNPs decays to background levels by ∼500 bp. (E) Allele frequency estimates at seasonal SNPs outside any large, cosmopolitan inversion (non-inv) or within the cosmopolitian inversions (diamonds) during the spring (blue) or fall (red). Allele frequency estimates at SNPs perfectly linked to the inversion during the spring and fall are denoted by circles. Error bars (C) and confidence bands (D) represent 95% confidence intervals based on blocked bootstrap resampling.](pgen.1004775.g002){#pgen-1004775-g002} Our rationale for focusing on the1750 seasonal SNPs at the FDR of 0.3 is that we are seeking to assess general molecular and evolutionary features of polymorphisms that may underlie rapid adaptive evolution in response to seasonal fluctuations in selection pressure. To assess these general features and enrichments, we require a sufficient number of true positive SNPs while maintaining as low a false positive rate as possible. Reducing FDR rates to lower values yielded an insufficient number of polymorphisms to assess enrichments with adequate precision (FDR of 10% yields 11 SNPs; FDR cutoff of 20% yields 200 SNPs). We note that our estimation of ∼1750 seasonal SNPs and their associated FDR should only be taken as a rough estimate of the number of seasonally varying SNPs: variance in linkage disequilibrium through the genome, heterscedasticity due to possible demographic events, limited statistical, unbalanced sampling of flies and variance in read-depth among samples, and modeling assumptions will affect our ability to infer the exact number of seasonally varying SNPs. One way to address some of these issues (e.g., heteroscedasticity) is to model allele frequency change through time with generalized linear mixed-effect (GLMM) or general estimation equation (GEE) models that account, to varying degrees, for the structured, time-series nature of our data. Seasonal SNPs inferred with these models are highly congruent with seasonal SNPs inferred using a simple GLM ([Fig. S2D,E](#pgen.1004775.s002){ref-type="supplementary-material"}) and *q-q* plots of the distribution of *p*-values from GLM, GLMM and GEE models suggest that GLM and GLMM modeling strategies fit the bulk of the genome well, with GEE models appearing to be anti-conservative ([Fig. S2B,C](#pgen.1004775.s002){ref-type="supplementary-material"}). However, the identification of a statistical excess of seasonally oscillating SNPs by any modeling strategy will be subject to a number of assumptions that will almost certainly be violated in some way or another and such violations could possibly lead to an increased false-positive rate. Because the false positive and false negative rates are inherently difficult to estimate, we adopt an empirical strategy to demonstrate that the seasonal SNPs identified though a simple GLM are not a random sample of SNPs but rather are enriched for true positive SNPs that directly underlie the adaptive response to seasonal fluctuations selection pressure. The identified seasonal SNPs are enriched for many signatures consistent with natural selection relative to control SNPs that are matched for several biologically and experimentally relevant parameters such as chromosome, recombination rate, allele frequency, and SNP quality coupled with a rigorous blocked-bootstrap procedure that accounts for the spatial distribution of seasonal SNPs along the chromosome (see [Materials and Methods](#s3){ref-type="sec"} and [Table S3](#pgen.1004775.s010){ref-type="supplementary-material"}). We now proceed to demonstrate these enrichments. Seasonal SNPs are enriched among functional genetic elements. These polymorphisms are likely to be in genic (i.e., 3′ and 5′ UTR, synonymous and non-synonymous, and long-intron SNPs; *p* = 0.054) and coding regions (synonymous and non-synonymous; *p*\<0.002) and are enriched among synonymous (*p*\<0.002), non-synonymous (*p* = 0.002) and 3′ UTR (*p* = 0.024, [Fig. 2C](#pgen-1004775-g002){ref-type="fig"}) relative to control, putatively neutral polymorphisms in short-introns [@pgen.1004775-Lawrie1]. The *p*-values of the enrichment tests were calculated after controlling for the spatial distribution of seasonal SNPs along the chromosome using a block bootstrap procedure coupled with the identification of paired control SNPs matched for several key genomic features ([Table S3](#pgen.1004775.s010){ref-type="supplementary-material"}), such as recombination rate, average allele frequency in the Pennsylvanian orchard, chromosome, and SNP quality (see 'Block Bootstrap' section in [Materials and Methods](#s3){ref-type="sec"}). Enrichment of adaptively oscillating polymorphisms among these genetic elements, including synonymous sites, suggests that these SNPs may affect organismal form and function through modification of protein function, translation rates, or mRNA expression and stability [@pgen.1004775-Lawrie1], [@pgen.1004775-Pechmann1]. Next, we show that rapid shifts in allele frequency at seasonal SNPs perturb allele frequencies at nearby SNPs. Adaptively oscillating polymorphisms are in regions of elevated temporal *F~ST~* ([Fig. 2D](#pgen-1004775-g002){ref-type="fig"}) and the elevation of temporal *F~ST~* decays, on average, by ∼500 bp, consistent with patterns of linkage disequilibrium in *D. melanogaster* [@pgen.1004775-Mackay1]. Elevation of temporal *F~ST~* within 500 bp of seasonal SNPs could contribute to high levels of genome-wide average *F~ST~* through time ([Fig. 1C](#pgen-1004775-g001){ref-type="fig"}). However, excluding SNPs within 500 bp of seasonal SNPs did not change patterns of genome-wide differentiation through time suggesting that genome-wide patterns of *F~ST~* through time are not driven by the seasonal SNPs themselves nor the SNPs in their immediate vicinity ([Fig. S3](#pgen.1004775.s003){ref-type="supplementary-material"}). Seasonal SNPs are spread throughout the genome ([Fig. 3A](#pgen-1004775-g003){ref-type="fig"}) and there is a 95% chance of finding at least one seasonal SNP per megabase of the euchromatic genome. This result suggests that seasonal SNPs are not exclusively concentrated in any single region (such as an inversion) nor distributed among a small number of regions (such as a limited number of genes). Although seasonal SNPs are distributed throughout the genome, their distribution is over-dispersed. To assess this, we calculated the number of seasonal SNPs per 1000 SNPs under investigation in non-overlapping windows of 1000 SNPs. If seasonal SNPs are homogeneously distributed throughout the genome, the rate of seasonal SNPs/1000 SNPs should follow a Poisson distribution with mean equal to the variance. After accounting for heterogeneity in recombination rate throughout the genome (see [Materials and Methods](#s3){ref-type="sec"}), we find that the variance in the rate of seasonal SNPs is ∼2.3 times greater than expected under a Poisson distribution (*p*\<10^−10^) implying that some regions have an excess of seasonal SNPs and some have a deficit of seasonal SNPs. The overdispersion of seasonal SNPs throughout the genome could be caused by several factors including variation in the density of functional elements, multiple functional and clustered seasonal SNPs, variance in the age of seasonal SNPs, or inversion status. {ref-type="supplementary-material"}). (B). Enrichment (log~2~ odds ratio) of seasonal SNPs with spatial *F~ST~* greater than or equal to value on *x*-axis relative to control SNPs. (C) Enrichment (log~2~ odds ratio) of seasonal SNPs with --log~10~(spatial q-value) greater than or equal to value on *x*-axis relative to control SNPs. (D) Absolute difference between average spring (blue) and fall (red) frequencies in the Pennsylvanian population and frequency estimates along the cline. Confidence bands represent 95% confidence intervals based on blocked bootstrap resampling.](pgen.1004775.g003){#pgen-1004775-g003} In general, we find no evidence that seasonal SNPs are enriched among large, cosmopolitan inversions segregating in North American populations (*p*\>0.05, [Fig. S4](#pgen.1004775.s004){ref-type="supplementary-material"}), with only one inversion, *In3R(Mo)*, marginally enriched for seasonal SNPs (*p* = 0.02, with *p* = 0.18 after Bonferroni correction for multiple testing). In addition, seasonal SNPs are significantly more common in the Pennsylvanian orchard population than polymorphisms perfectly linked [@pgen.1004775-Kapun1] to large cosmopolitan inversions ([Fig. 2E](#pgen-1004775-g002){ref-type="fig"}) and polymorphisms linked to inversions do not vary between seasons ([Fig. 2E](#pgen-1004775-g002){ref-type="fig"}, *p*\>0.05), including those linked to *In3R(Mo)*. Therefore, enrichment of seasonal SNPs within *In3R(Mo)*, if present, is most likely due to increased linkage disequilibrium caused by decreased recombination surrounding this inversion [@pgen.1004775-CorbettDetig1]. Taken together, these results indicate that the inversions themselves do not cycle seasonally in the Pennsylvanian population in any appreciable manner ([Fig. 2E](#pgen-1004775-g002){ref-type="fig"}) and suggests that adaptive evolution to seasonal variation in selection pressures may be highly polygenic. Relationship between spatial and temporal variation in allele frequencies {#s2c} ------------------------------------------------------------------------- To test the hypothesis that spatially varying selection pressures along the latitudinal cline reflect seasonally varying selection pressures in the Pennsylvanian population, we examined the relationship between temporal and spatial variation in allele frequencies. To quantify spatial variation in allele frequency, we calculated two statistics. First, we estimated average pairwise *F~ST~* among all populations for each SNP ('spatial *F~ST~*'). Second, we estimated clinality for each SNP by calculating the per-SNP false discovery rate (FDR) of the relationship between allele frequency and latitude using a generalized linear model that takes into account read depth and the number of sampled chromosomes (hereafter 'clinal *q*-value'). Spatial *F~ST~* and clinal *q*-value are highly correlated (*r* = 0.63, *p*\<1e-10; [Fig. S5](#pgen.1004775.s005){ref-type="supplementary-material"}) demonstrating that most, but not all, spatial variation along the latitudinal cline is represented by monotonic changes in allele frequency between northern and southern populations. We calculated the number of clinally varying polymorphisms (clinal *q*-value\<0.1) and the number of adaptively oscillating polymorphisms per common segregating SNP (average, North American MAF\>0.15) per megabase of the genome ([Fig. 3A](#pgen-1004775-g003){ref-type="fig"}). Approximately one out of every three common polymorphisms varies with latitude with FDR\<0.1 (i.e., clinal *q*-value\<0.1) whereas only one out of every three thousand polymorphisms varies predictably between seasons with seasonal FDR\<0.3 ([Fig. 3A](#pgen-1004775-g003){ref-type="fig"}). Although our ability to detect clinal SNPs at FDR\<0.1 is greater than our ability to detect seasonal SNPs at FDR\<0.3 (cf. [Fig. 2B](#pgen-1004775-g002){ref-type="fig"}, [Fig. S6](#pgen.1004775.s006){ref-type="supplementary-material"}), differences in power cannot explain the three order of magnitude difference in the number of detected clinal and seasonal SNPs (cf. [Fig. 2B](#pgen-1004775-g002){ref-type="fig"}, [Fig. S6](#pgen.1004775.s006){ref-type="supplementary-material"}). Next, we formally tested whether seasonal SNPs are enriched among spatially varying SNPs. Spatially varying SNPs, as defined by spatial *F~ST~*, are more likely to be seasonal SNPs than expected by chance ([Fig. 3B](#pgen-1004775-g003){ref-type="fig"}), and the odds of this enrichment increases with increasing spatial differentiation. In contrast, we cannot reject the null hypothesis of no enrichment of seasonal SNPs among clinal SNPs as defined by clinal *q*-value ([Fig. 3C](#pgen-1004775-g003){ref-type="fig"}). The observed differences in the enrichment of seasonal SNPs among SNPs with high spatial *F~ST~* and low clinal *q*-value may reflect aspects of our sampling design and differences in the evolutionary forces that shape allele frequencies through time and space. We sampled flies along the East Coast during different years and at different points of time relative to the progression of the growing season in each population ([Table S1](#pgen.1004775.s008){ref-type="supplementary-material"}). Thus, in each sampled clinal population, seasonal SNPs would be at different points in their adaptive trajectory. Consequently, seasonal SNPs would not likely have exceedingly low clinal *q*-values, a statistic which reflects the deviation of observed allele frequencies from the predicted value as estimated by a GLM. Rather, seasonal SNPs would likely be highly differentiated along the cline (i.e., have a large spatial *F~ST~*). SNPs with low clinal *q*-values, therefore, represent those SNPs that do not change in frequency between seasons and possibly reflect long-term demographic processes or adaptation to selection pressures that vary clinally, but not seasonally. Because of the relationship between spatial differentiation and seasonal variation in allele frequencies ([Fig. 3B](#pgen-1004775-g003){ref-type="fig"}) and because of parallels between spatial and seasonal variation in climate, we hypothesized that northern populations should be more 'spring-like' and southern populations should be more 'fall-like' in allele frequencies at the seasonal SNPs. To test this hypothesis, we calculated the absolute difference in allele frequencies for each population sampled along the cline with the average spring and fall allele frequency estimates for the Pennsylvanian population for all seasonal SNPs. Indeed, allele frequency estimates at seasonal SNPs from high latitude populations are more similar to spring Pennsylvanian populations and those from low latitude are more similar to fall populations ([Fig. 3D](#pgen-1004775-g003){ref-type="fig"}) demonstrating that latitudinally varying selection pressures at least partially reflect seasonally varying selection pressures. Immediate adaptive response to an acute frost event {#s2d} --------------------------------------------------- In the late fall of 2011, about two weeks after our 2011 fall sample was collected, a hard frost occurred in the Pennsylvanian orchard ([Fig. 4A](#pgen-1004775-g004){ref-type="fig"}). We were able to obtain a sample of *D. melanogaster* approximately one week after the frost and we estimated allele frequencies genome-wide from this sample. We hypothesized that allele frequencies at seasonal SNPs would predictably change following the frost event and would become more 'spring-like.' To test this hypothesis, we calculated the probability that post-frost allele frequencies at seasonal SNPs overshoot the long-term average allele frequency (i.e., become more 'spring-like'). We also estimated this probability for control polymorphisms, matched to adaptively oscillating polymorphisms by several characteristics ([Table S3](#pgen.1004775.s010){ref-type="supplementary-material"}) including, importantly, difference in allele frequency between the long-term average and the pre-frost allele frequency. This later control is essential given that some shift in the 'spring-like' direction is expected here simply by chance due to regression to the mean. The probability that seasonal SNPs overshoot the long-term average allele frequency is ∼43%, whereas only ∼35% of control polymorphisms overshoot the long-term average. This significant excess of adaptively oscillating polymorphisms that become more 'spring-like' following the frost event ([Fig. 4B](#pgen-1004775-g004){ref-type="fig"}; log~2~(OR) = 0.48, *p*\<0.002) suggests that these SNPs respond to acute changes in climate and that cold temperatures associated with winter is one selective force acting on this population shaping allele frequencies between seasons. {#pgen-1004775-g004} Association with seasonally variable phenotypes {#s2e} ----------------------------------------------- Chill-coma recovery time and starvation tolerance are two phenotypes that vary seasonally in drosophilid populations [@pgen.1004775-Gibert1]--[@pgen.1004775-Sisodia1]. Accordingly, we hypothesized that the winter-favored allele at seasonal SNPs would be associated with decreased chill-coma recovery time and increased starvation tolerance. To test this hypothesis, we used allele frequency data from previously published tail-based mapping of chill-coma recovery time and starvation tolerance [@pgen.1004775-Huang1]. We show that the winter favored allele at seasonal SNPs is more likely to be associated with fast chill coma recovery time than expected by chance across a range of GWAS *p*-values ([Fig. 5A](#pgen-1004775-g005){ref-type="fig"}). A similar analysis of starvation tolerance was equivocal but the general pattern is that the winter-adaptive allele is associated with increased starvation tolerance ([Fig. 5B](#pgen-1004775-g005){ref-type="fig"}). {#pgen-1004775-g005} Long-term balancing selection {#s2f} ----------------------------- Balancing selection caused by variation in selection pressures through time can in principle maintain allelic variation at adaptively oscillating loci and elevate levels of neutral diversity surrounding these balanced polymorphisms. Thus, if seasonal variation in selection pressures promotes balanced polymorphisms we hypothesized that seasonal SNPs would be old and present in regions of elevated polymorphism. We tested the hypothesis that seasonal SNPs are old by first examining their allele frequencies in a broad survey of African *D. melanogaster* populations [@pgen.1004775-Pool1]. Approximately 5% of seasonal SNPs are rare in Africa (MAF\<0.01), however these SNPs are not more likely to be rare in Africa than control polymorphisms (log~2~(odds ratio) = 0.96; *p* = 0.328). Interestingly, for seasonal SNPs where one allele is rare in Africa, the summer favored alleles are more likely to be rare in Africa than winter favored alleles (log~2~(odds ratio) = 0.475; *p* = 0.018). Because the vast majority of seasonal SNPs segregate in Africa, it appears that adaptation to temperate environments, and particularly winter conditions, relies primarily on old, standing genetic variation. Balancing selection acts to maintain alleles at intermediate frequencies for long periods of time and, in some instances, can maintain polymorphism across species boundaries [@pgen.1004775-Klein1], [@pgen.1004775-Leffler1]. We examined whether seasonal SNPs showed signatures of long-term balancing selection by examining patterns of polymorphism surrounding orthologous regions in *D. simulans*, the sister species to *D. melanogaster*. We note that the following analyses are conservative because we underestimate *D. simulans* diversity given the small number (\<6) of *D. simulans* haplotypes used. First, we demonstrate that seasonal SNPs are approximately 1.5 times more likely to be polymorphic and share the same two alleles identical by state in both species relative to control SNPs. This pattern is observed for all seasonal SNPs ([Fig. 6](#pgen-1004775-g006){ref-type="fig"}, *p*\<0.002) and for seasonal SNPs residing in genes ([Fig. 6](#pgen-1004775-g006){ref-type="fig"}, *p*\<0.002). The increased probability of shared polymorphism between *D. melanogaster* and *D. simulans* at seasonal SNPs could, in principle, be driven by an over-representation of synonymous, genic SNPs ([Fig. 2C](#pgen-1004775-g002){ref-type="fig"}). Unless synonymous SNPs are in four-fold degenerate positions, certain mutations may cause them to be non-synonymous thereby limiting the number of possible neutral allelic states and increasing the probability of shared polymorphism between species. However, adaptively oscillating SNPs that do not reside in synonymous sites are also more likely than expected by chance to be polymorphic and share the same two alleles by state in *D. melanogaster* and *D. simulans* ([Fig. 6](#pgen-1004775-g006){ref-type="fig"}, *p* = 0.014). {#pgen-1004775-g006} The co-occurrence of shared polymorphism between *D. melanogaster* and *D. simulans* could result from three evolutionary mechanisms. First, trans-specific polymorphisms could result from adaptive introgression. This scenario seems implausible given the high degree of pre- and post-zygotic isolating mechanisms between these two species [@pgen.1004775-Welbergen1], [@pgen.1004775-Sturtevant1]. Furthermore, if trans-specific polymorphisms resulted from recent adaptive introgression we would expect average pairwise divergence between *D. melanogaster* and *D. simulans* surrounding seasonal SNPs to be smaller than at control SNPs. However, there is no significant difference in estimates of divergence between seasonal and control SNPs (*p* = 0.7 for windows ±250 bp). Second, trans-specific polymorphisms could result from convergent adaptive evolution. Finally, trans-specific polymorphisms could be millions of years old [@pgen.1004775-Tamura1], predating the divergence of *D. melanogaster* from *D. simulans*. While we cannot differentiate these latter two mechanisms, we postulate that the most parsimonious explanation is that trans-specific seasonal SNPs predate the divergence of these two sister species. Seasonally variable selection is required to generate genome-wide patterns of allele frequency change through time {#s2g} ------------------------------------------------------------------------------------------------------------------ Despite empirical support for the conclusion that seasonal SNPs show many signatures consistent with adaptive response to seasonally variable selection, drift, caused by cyclic population booms and busts, or migration from neighboring demes are alternative mechanisms that could drastically perturb allele frequencies in the Pennsylvanian population and could generate some of the genome-wide patterns we observe. We address these possibilities here and conclude that neither cyclic changes in population size nor seasonal migration can plausibly explain the extent of genome-wide genetic differentiation through time, the observed number of seasonal SNPs, nor the enrichment of seasonal SNPs among many distinct genomic features (e.g., [Figs. 2](#pgen-1004775-g002){ref-type="fig"}--[6](#pgen-1004775-g006){ref-type="fig"}). At the same time, we also show through several simulation approaches that rapid adaptive evolution in response to seasonal fluctuations in selection pressure is sufficient to explain patterns of allele frequency change through time. Furthermore, we discuss how large-scale migration is internally inconsistent with certain aspects of our data. Taken together, we conclude that rapid adaptive evolution to seasonally variable selection is required to explain the patterns of allele frequency change through time at seasonal SNPs and at linked neutral loci that we observe in our dataset. First, we assessed the possibility that extensive drift caused by population contraction every winter [@pgen.1004775-Knibb1], [@pgen.1004775-Band1], [@pgen.1004775-Ives1] could generate genome-wide patterns of genetic differentiation through time observed in our data. To do so, we conducted forward genetic simulations that model biologically plausible variation in population size and included loci that cycle in frequency due to variable selection pressures [@pgen.1004775-Messer1]. For these simulations, we modeled a 20 Mb chromosome with constant recombination rate of 2 cM/Mb, representing the genome-wide average recombination rate in *D. melanogaster* [@pgen.1004775-Comeron1]. We simulated population contraction to one of various minimum, 'overwintering' population sizes followed by exponential growth over 10 generations in the 'summer' to a fixed maximum population size. In these models, we included various numbers of loci that respond to seasonally varying selection. Selection coefficients for each locus were set such that allele frequencies at selected sites oscillated by ∼20%, between 60 and 40%, representing the average change in allele frequency we actually see between spring and fall at seasonal SNPs. Finally, we placed 500 neutral loci randomly along the simulated chromosome and measured *F~ST~* at these neutral loci between three 'spring' (i.e., first generation of population expansion) and 'fall' (last generation of population expansion) samples. See [Materials and Methods](#s3){ref-type="sec"} for more details these models. In the absence of seasonal selection, these forward simulations suggest that overwintering *N~e~* would have to be exceedingly low (∼20; [Fig. 7A](#pgen-1004775-g007){ref-type="fig"}) to generate levels of *F~ST~* between spring and fall as high as we observe in our data (arrow in [Fig. 1C](#pgen-1004775-g001){ref-type="fig"}). However, with overwintering *N~e~* of 200 and 5--10 seasonally adaptive SNPs per chromosome arm, simulated *F~ST~* at neutral loci is on the order of 0.002 ([Fig. 7A](#pgen-1004775-g007){ref-type="fig"}), which we observe in our data (arrow in [Fig. 1C](#pgen-1004775-g001){ref-type="fig"}). While we do not know overwintering population size, we speculate it could be on the order of 200 flies or likely substantially larger [@pgen.1004775-Band1], [@pgen.1004775-Ives1] and conclude that at least 25--50 (5--10 per main chromosome arm) loci are sufficient to generate patterns of differentiation we observe through time. Note that increasing the overwintering population size requires concomitant increase in number of seasonally selected loci. {#pgen-1004775-g007} We regard overwintering population sizes of ∼20 flies to be inconsistent with certain aspects of our data and also implausible given what we know about the biology of the species. First, such a severe population contraction would result in reduction of genetic diversity, particularly for low frequency alleles. However, the observed allele frequency spectrum between fall and the following spring samples is similar and spring samples do not exhibit the expected loss of low frequency polymorphisms that would result from a population contraction to 20 individuals ([Fig. S7](#pgen.1004775.s007){ref-type="supplementary-material"}). Second, population contraction to 20 individuals would often lead to population extirpation in the Pennsylvanian orchard and would certainly lead to extirpation at localities further north that experience more severe winters. However, *D. melanogaster* are routinely collected in Northern orchards very early in the season [@pgen.1004775-Schmidt3] and are routinely found in populations at as far north as 45° (Schmidt pers. obs). Furthermore, certain rare alleles have persisted in northern *D. melanogaster* populations for upwards of 30 years [@pgen.1004775-Ives2] cf. [@pgen.1004775-Coyne1] and allele frequency clines are relatively stable over decadal scales [@pgen.1004775-Umina1] demonstrating that high latitude populations are not frequently extirpated and that overwintering bottlenecks cannot be so severe as our neutral simulations would require. In our forward simulations, seasonally variable selection is sufficient to generate high levels of genome-wide genetic differentiation through time. In addition, our forward simulations are consistent with the increase of genome-wide average *F~ST~* through time excluding polymorphisms that are within 500 bp of seasonal SNPs ([Fig. S3](#pgen.1004775.s003){ref-type="supplementary-material"}). In our simulations, 500 neutral loci were placed randomly along a 20 Mb chromosome and were initially completely unlinked to selected loci. Therefore, the high levels of simulated *F~ST~* are a consequence of genetic draft acting over long physical distances with low to moderate linkage disequilibrium between neutral and selected polymorphisms. Our observation that genome-wide average *F~ST~* (excluding polymorphisms near seasonal SNPs, [Fig. S3](#pgen.1004775.s003){ref-type="supplementary-material"}) increases with time resembles our simulations suggesting that draft can perturb allele frequencies over long genetic distances. We also note that long-range genetic draft, caused by rapid frequency shifts of ancient balanced alleles to seasonally variable selection would likely cause an asymptotic change in genome-wide temporal *F~ST~*, whereas a purely drift-based model would likely cause a linear increase in genome-wide *F~ST~* through time. Seasonal SNPs tend to be old and are therefore likely found on a diverse array of haplotypes. Therefore, the exact composition of haplotypes that rise and fall every seasonal cycle will be somewhat stochastic giving rise to a high genome-wide *F~ST~* over a duration of time less than ∼7 months (the duration of time between fall and the following spring). Among years, genome-wide average *F~ST~* would possibly plateau if local *N~e~* were large (as we suspect it is, see [Results and Discussion](#s2){ref-type="sec"}: The plausibility...), coupled with the effects of recombination, gene conversion, and low-level migration from neighboring demes or populations. Finally, we note that because seasonal SNPs likely exist on a diverse array of haplotypes we do not expect genome-wide average *F~ST~* to oscillate with a period corresponding to approximately 6--7 months. For such oscillations to occur, a large (i.e., much larger than we identify) number of loci would have to be repeatedly shifting between seasons. Next, we explore the possibility that migration could drastically alter allele frequencies in the Pennsylvanian population and generate the large number of loci that vary repeatedly among seasons. First, we examined a simple but general demographic model where the Pennsylvanian orchard population becomes extirpated every year and recolonized from a refugium such as a southern population or a large, local site such as a compost pile. Either situation is plausible given the purportedly high rates of migration in North American *D. melanogaster* populations [@pgen.1004775-Coyne1], [@pgen.1004775-Berry1] and what little is known about the overwintering biology of high latitude *D. melanogaster* [@pgen.1004775-Ives2]. In our model, we envisioned a resident, refugial population with stable allele frequencies across years that colonizes the orchard population. In this model, the orchard would be colonized early in the season with a random subsample of flies from the refugium and would therefore have aberrant allele frequencies. As more migrants arrived to the orchard from the refugium, allele frequencies at the orchard would stabilize to that of the source population. In such a scenario, allele frequencies in spring samples could vary considerably and a small fraction of SNPs might, by chance, have the same aberrant allele frequencies year after year and would appear to cycle seasonally. We calculated the expected number of SNPs that would cycle by chance alone as a function of the number of initial migrants ([Fig. 7B](#pgen-1004775-g007){ref-type="fig"}). For instance, if five migrants arrived at the orchard prior to our spring sample every year, approximately 1300 SNPs would cycle seasonally producing similar patterns to the observed change in allele frequency through time as at 'seasonal SNPs' ([Fig. 2A](#pgen-1004775-g002){ref-type="fig"}). However, if four migrants arrived at the orchard prior to our sampling, ∼2600 SNPs would vary repeatedly but if six migrants arrived, only ∼700 would. Although the expected number of sites that oscillate under this migration model with 5 migrants is approximately the number we observe, we note that the expected number is highly dependent on the exact number of migrants. It seems unlikely that exactly five flies would migrate from the refugium to the orchard before our first spring sample three times in a row. Therefore, the extreme sensitivity of the expected number of sites to the number of migrants makes this general demographic scenario implausible. We are therefore led to conclude that the simple migration model presented here is likely to be insufficient to explain changes in allele frequency through time in the Pennsylvanian orchard. In addition to our conclusion that a simple model of recolonization of the orchard is insufficient to explain the number of seasonally variable loci we observe, our data are also inconsistent with large-scale migrations from adjacent populations. For instance, if a large-scale migration from the South to resident northern populations were to occur, we would expect that clinally varying SNPs should also vary seasonally. Such a pattern would be expected both if a large-scale migration occurred randomly or were genotype dependent. However, seasonal SNPs are apparently not enriched among clinally varying polymorphisms ([Fig. 3C](#pgen-1004775-g003){ref-type="fig"}). A similar logic would apply for an early season migration from the North followed by a subsequent, late season migration from the South. We also note that this dual migration model is biologically implausible. The relationship between latitude and the onset of spring would suggest that far northern populations would be quite small in the early part of the growing season and the subsequent probability of emigration to southern locales would be low. Therefore, we conclude that large-scale migration does not play a major role shaping seasonal variation in allele frequencies in the Pennsylvanian orchard. Furthermore, even if seasonal SNPs were enriched among clinally varying polymorphisms (which they do not appear to be), adaptation to seasonally variable selection would need to be invoked in order to explain the yearly shift in allele frequencies every winter. Taken together, the models presented here demonstrate that seasonal boom-bust or migration-based scenarios are insufficient to explain allele frequency change through time in the Pennsylvanian population. While temperate populations of *D. melanogaster* clearly undergo cyclic population booms and busts due to changes in climate associated with the season, the extent of these population contractions necessary to generate the patterns of genetic variation through time that we observe would be too extreme to allow for stable population persistence. Similarly, the Pennsylvanian population certainly exists as a part of a complex metapopulation and experiences immigration and emigration. However, analysis of a simple demographic model of population recolonization during the spring is also insufficient to explain the patterns of allele frequency change through time that we observe and our data are internally inconsistent with a model of large-scale migration from neighboring populations. Finally, we point out that the boom-bust and recolonization models we presented here undoubtedly are oversimplifications and that there are other, more complex demographic models that we have not explored. Nonetheless, any stochastic demographic event would affect SNPs throughout the genome with equal probability. Many aspects of our data clearly show that seasonal SNPs are not a random set of common SNPs but rather show signatures consistent with both functional effect and long-term balancing selection such as enrichment in specific classes of genetic elements, association with seasonally variable phenotypes and predictable and virtually instantaneous shifts in allele frequency in response to frost. Therefore, while we cannot conclusively rule out the possibility that demographic events affect the temporal dynamics of allele frequencies at seasonal- and non-seasonal SNPs in the Pennsylvanian population, these demographic events are most likely coupled with adaptive evolution in response to temporally varying selection pressures. The plausibility of seasonally variable selection {#s2h} ------------------------------------------------- We have previously argued that adaptive response to seasonally fluctuating selection at no less than 25--50 loci is necessary to generate the high levels of genome-wide genetic differentiation through time observed in the Pennsylvanian population. Next, we considered the plausibility of such strong selection and estimated the upper bound of the number of loci that could independently respond to seasonally variable selection. To do so, we modeled independent selection at 1--10,000 simulated seasonal SNPs whose allele frequency change was drawn from the observed allele frequency change at seasonal SNPs. Using a simple Poisson model (see [Materials and Methods](#s3){ref-type="sec"}), we estimated the minimum fall census size required for that number of loci to shift in allele frequency during one or two rounds of truncation selection. Using these models, we sought to estimate the most likely number of seasonal SNPs that could independently respond to seasonally variable selection by contrasting model-based estimates of population size with our best estimates of population size in the field. Although fall census size of *D. melanogaster* in the focal Pennsylvanian population is unknown, some estimates of drosophilid population size have been made. Global population size of *D. melanogaster* is likely to be extremely large, greater than 10^8^ [@pgen.1004775-Karasov1]. However, estimates of local population size made from mark-release-recapture methods report census sizes on the order of 10^4^ to 10^5^ [@pgen.1004775-Mckenzie1]--[@pgen.1004775-Powell1]), with considerable variation among seasons, years and locales. *D. melanogaster* samples from orchards and vineyards often exceed 10^4^ flies [@pgen.1004775-Gravot1], [@pgen.1004775-Bastide1] and thousands of flies can easily be collected over large compost piles (Bergland pers. obs.). Therefore, we speculate that census size of temperate *D. melanogaster* populations at any locale is a function of the local ecology (e.g., amount of windfall fruit, number and size of compost piles, humidity) and given the favorable conditions in the focal Pennsylvanian orchard (Schmidt pers. obs.), large census sizes of more than 10^5^ are conceivable. If fall census size in the Pennsylvanian population is on the order of 10^5^, our truncation selection model suggests that no more than several hundred (200--700, [Fig. 7C](#pgen-1004775-g007){ref-type="fig"}) seasonal SNPs could respond to seasonally varying selection independently. We note that increasing the number of generations of winter-like selection pressures or the fall census size would lead to a concomitant increase in the number of seasonally selected loci that could independently respond to seasonally varying selection pressures. Our survey of temporal changes in allele frequency identified 1750 seasonal SNPs that cycle significantly by ∼20% between seasons at FDR of 0.3. Unless local census size in the Pennsylvanian population were unrealistically large -- on the order of 10^10^ or 10^20^ -- it is unlikely that all of these loci respond to selection independently. Our model suggests, however, that a large fraction, on the order of 200--700 could vary independently in every cycle. One explanation for cycling in the remaining SNPs is linkage with loci responding to seasonally variable selection. It is possible that this linkage is generated either stochastically and neutrally or, alternatively, by selective processes such as assortative mating [@pgen.1004775-Kirkpatrick1] or epistatic selection [@pgen.1004775-Lewontin1], [@pgen.1004775-Giesel1]. For instance, if winter adapted flies were more likely to mate with other winter adapted flies during the summer, winter adapted alleles may become coupled and linkage disequilibrium between these alleles could increase. Similarly, certain forms of epistatic interactions could also generate linkage disequilibrium between seasonal SNPs if, for instance, couplings of winter and summer favored alleles at multiple loci were particularly deleterious relative to winter-winter or summer-summer combinations. The net effect of selective mechanisms that promote positive linkage disequilibrium between seasonal SNPs is that the effective number of 'independently' seasonally selected loci decreases. If seasonal SNPs are in linkage disequilibrium due to selective processes, it would imply that more than 200--700 seasonal SNPs contribute to organismal form and function and modify fitness during the summer and winter. Conclusions -- Summary {#s2i} ---------------------- Herein, we present results from population based resequencing of samples of flies collected along a latitudinal cline in North America and over three years during the spring and fall in a Pennsylvanian orchard. We identify repeatable and dramatic changes in allele frequencies through time at hundreds of polymorphisms spread throughout the genome. Response to strong selection at these seasonal SNPs likely drives genetic differentiation through time at linked, neutral polymorphisms. This process leads to genome-wide differentiation between samples collected several years apart comparable to populations separated by 5--10° latitude. Seasonal SNPs are likely to be functional as they show enrichment at functional sites, vary predictably among populations sampled along the cline, respond immediately to a hard frost event, and are associated with phenotypes previously shown to vary seasonally in temperate *D. melanogaster* populations. Finally, our results suggest that some adaptively oscillating SNPs are possibly millions of years old, predating the split of *D. melanogaster* from its sister species *D. simulans*. Taken together, our results provide the first genomic picture of balancing selection caused by temporal fluctuations in selection pressures and provide novel insight into the biology of marginal overdominance. Conclusions -- Functional properties of adaptively oscillating polymorphisms {#s2j} ---------------------------------------------------------------------------- Temperate populations of *D. melanogaster* are exposed to high levels of environmental heterogeneity among seasons due to changes in various aspects of the environment including temperature, humidity, and nutritional quality and quantity. These shifts in the environment are primary determinants of cyclic population booms and busts [@pgen.1004775-Ives2], [@pgen.1004775-Mckenzie1], [@pgen.1004775-McInnis1] and impose strong temporally and spatially variable selection. Intuition, theoretical models [@pgen.1004775-Lewontin2], laboratory experimentation [@pgen.1004775-Schmidt1], and inference from patterns of clinal variation [@pgen.1004775-Schmidt4]--[@pgen.1004775-Arthur1] and seasonal variation in morphological, behavioral and life-history traits suggest that alternate seasons favor differing life-history strategies. In general, populations exposed to more harsh conditions such as those from Northern locales or those collected early in the season are larger [@pgen.1004775-James1], [@pgen.1004775-Robinson1], more stress tolerant [@pgen.1004775-Karan1]--[@pgen.1004775-Parkash1], [@pgen.1004775-Arthur1], longer lived [@pgen.1004775-Schmidt5], and are less fecund [@pgen.1004775-Schmidt5], [@pgen.1004775-Lazzaro1] than those collected in Southern locales or during the fall. The general picture that emerges, therefore, is that in temperate populations winter conditions select for hardier but less fecund individuals whereas summer selects for high reproductive output at the cost of somatic maintenance. Nonetheless, there is surprisingly little evidence directly linking adaptive differentiation between seasonally favored genetic polymorphisms, phenotypes and environmental perturbations (but see [@pgen.1004775-Schmidt1]). Herein we present several key results that link seasonal and spatial patterns of genotypic and phenotypic variation with environmental perturbations. First, our data suggest that that acute bouts of cold temperature elicit adaptive response at seasonally oscillating polymorphisms ([Fig. 4](#pgen-1004775-g004){ref-type="fig"}). Heretofore, the specific environmental factors altering allele frequencies through time and space among dipteran species has generally remained elusive largely stemming from the fact that many aspects of the environment co-vary over temporal and spatial scales. Here we show that acute exposure to sub-freezing temperatures in the field shifts allele frequencies in a spring like direction at seasonal SNPs but not at control polymorphisms, thereby suggesting that sharp modulation of temperature can act as a selective force in the field. While post-frost allele frequencies at seasonal SNPs move in a 'spring-like' direction, they do not reach average spring allele frequencies. This suggests that multiple frost events, long-term exposure to cold temperatures or other selective factors linked to winter conditions such as starvation also impose strong selection in temperate populations. Next, we demonstrate that environmental differences among populations predict, to a certain extent, changes in allele frequency at seasonal SNPs. Environmental factors that vary over seasonal time scales also vary with latitude. This fact has facilitated studies that substitute space for time and has led to a paradigm in many aspects of contemporary research in drosophilid evolutionary ecology of examining phenotypic and genetic differentiation along latitudinal (and altitudinal) clines as a proxy for studying adaptation to temperate environments [@pgen.1004775-Singh1]. Using allele frequency estimates that we made from populations sampled along the North American latitudinal cline, we demonstrate that southern populations are more 'fall-like' at seasonal SNPs whereas northern populations are more 'spring-like' ([Fig. 3D](#pgen-1004775-g003){ref-type="fig"}). Northern populations experience more severe winters and have shorter growing seasons; therefore, we speculate that the changes in allele frequency at adaptively oscillating polymorphisms along the cline is because (1) the summer favored allele would be at lower frequency due to stronger selection during the winter and (2) the summer favored allele would not rise in frequency as much during the summer because of the shorter growing season. The converse would be the case for Southern populations. Finally, we relate seasonally variable SNPs with ecologically relevant phenotypic variation. Previous studies have demonstrated that two important stress tolerance traits, chill coma recovery time and starvation resistance vary in predictable ways among temperate populations of *D. melanogaster*. Northern populations tend to have fast chill coma recovery time [@pgen.1004775-Ayrinhac1]--[@pgen.1004775-Overgaard1] recapitulating deeper phylogenetic patterns among drosophilids originating from temperate and tropical locales [@pgen.1004775-Gibert1]. Evidence for latitudinal variation in starvation tolerance is more equivocal with low latitude populations of *D. melanogaster* being more starvation tolerant in some studies but not significantly so in others [@pgen.1004775-Karan1], [@pgen.1004775-Gilchrist1] and closely related species showing equally ambiguous patterns [@pgen.1004775-Schmidt2], [@pgen.1004775-Hoffmann2], [@pgen.1004775-Gilchrist1]. However, diapause-competent genotypes that are at high frequency in Northern populations and in the spring show increased starvation tolerance [@pgen.1004775-Schmidt2] suggesting that spatial and temporal differentiation in starvation tolerance may be parallel in the context of specific polymorphisms. Nonetheless, because selection pressures along latitudinal clines are generally parallel with seasonal selection pressures (e.g., [Fig. 3D](#pgen-1004775-g003){ref-type="fig"}) we reasoned that winter adapted alleles at seasonal SNPs would be associated with fast chill coma recovery time and increased starvation tolerance. We show that winter adapted alleles at seasonal SNPs are likely to be associated with fast chill coma recovery time and, to a lesser extent, starvation tolerance ([Fig. 5](#pgen-1004775-g005){ref-type="fig"}). The strength of the relationship between seasonal SNPs with these two phenotypes likely differs for many reasons, including intrinsic differences in the statistical power and the complex genetic architecture of these traits. Nonetheless, the fact that seasonal SNPs are associated with chill coma recovery and starvation tolerance in the predicted direction given our prior knowledge of seasonal variation in these two traits strongly suggests that seasonal SNPs are functional and affect seasonally dependent fitness via stress tolerance traits. In addition, the concordance between seasonal SNPs and SNPs moderately associated with chill coma recovery time and starvation tolerance suggests that the intermediate frequency SNPs that we are investigating here have small effects on phenotype but nonetheless have large effects on average population fitness. Taken together, our analysis has linked adaptive oscillations at hundreds of polymorphisms in *D. melanogaster* to specific and persistent differences in climate and to phenotypes known to be under diversifying selection through time and space. Our results support the hypothesis that stress tolerance traits are favored during the winter and disfavored during the summer. Stress tolerance traits such as chill coma recovery time and starvation tolerance often have negative genetic correlations with reproductive output [@pgen.1004775-Schmidt2], [@pgen.1004775-Ayroles1] or development time [@pgen.1004775-Reynolds1], two phenotypes that would be favored during exponential growth during the summer. Therefore, it is likely that a subset of seasonal SNPs directly contribute to a tradeoff between stress tolerance and reproductive output. Because *D. melanogaster* originated in sub-Saharan Africa and colonized the world in the wake of human migration 200--10,000 years ago [@pgen.1004775-David1] it has been hypothesized [@pgen.1004775-Sezgin1] that phenotypes favored during the winter are derived whereas those favored during the summer are ancestral with respect to tropical, African populations. Although we show that the vast majority of seasonal SNPs are common in Africa, a small set (∼5%) are rare, segregating at less than 1%. Somewhat surprisingly, summer favored alleles are more likely to be rare in Africa than winter favored alleles (see [Results and Discussion](#s2){ref-type="sec"}: Long term...) suggesting that some environmental aspects of summer in temperate orchards are new for *D. melanogaster*. Consistent with the observation that flies sampled at low latitudes are likely subject to intense intra- and inter-specific competition [@pgen.1004775-James1], we speculate that the cornucopia of rotten fruit during the summer in mid- to high-latitude locales coupled with decreased inter-specific competition is a novel environment for *D. melanogaster* that has allowed formerly rare alleles associated with increased reproductive output to flourish. Conclusions -- Long-term, polygenic balancing selection, and ecological generality {#s2k} ---------------------------------------------------------------------------------- Herein, we present several lines of evidence demonstrating that hundreds of loci adaptively respond to seasonal fluctuations in the environment. Despite (or because of) the fact that these loci promote rapid adaptive evolution, many have remained polymorphic for millions of generations within *D. melanogaster* and some possibly predate the divergence of *D. melanogaster* and *D. simulans* ∼5 million years ago. Taken together, these observations suggest that alleles at these loci have may have been maintained by environmental heterogeneity for exceptionally long periods of time. Long-term balancing selection is typically regarded as an evolutionary oddity, found predominantly in the genetic systems regulating host-pathogen interactions, self-incompatibility, and sex-determination [@pgen.1004775-Klein1], [@pgen.1004775-Langley1]. Herein, we provide evidence that environmental heterogeneity might promote long-term balanced polymorphisms at hundreds of loci that affect quantitative, stress tolerance traits. Theory predicts that temporal variation in selection coefficients can maintain adaptive genetic variation for long periods of time when certain genetic and ecological conditions are met. Classic models suggest that the adaptive variation can be maintained in populations because of temporal shifts in selection pressure only when the heterozygote has a higher geometric mean fitness than either homozygote [@pgen.1004775-Gillespie1]. Such conditions are necessary for both finite and infinite populations and, moreover, in finite populations the persistence time of adaptive polymorphisms may be shorter than for neutral ones [@pgen.1004775-Hedrick1]. However, alternative models have demonstrated that overlapping generations [@pgen.1004775-Ellner2], the combination of spatial and temporal variation in selection pressures [@pgen.1004775-Ewing1], habitat fidelity [@pgen.1004775-GarciaDorado1], [@pgen.1004775-Hedrick2], and multiple liked loci subject to temporally variable selection [@pgen.1004775-Korol1] will increase the persistence time of balanced polymorphisms maintained by environmental heterogeneity. Each of these conditions are met in for *D. melanogaster*. First, flies are highly fecund [@pgen.1004775-Bergland1], iteroparous insects with generation time a fraction of lifespan [@pgen.1004775-Schmidt4], [@pgen.1004775-Schmidt5]. Therefore natural populations are likely to be highly age structured which will prevent the loss of balanced alleles during alternate seasons. Second, spatial selection pressures vary on the order of meters to kilometers [@pgen.1004775-McKechnie1], [@pgen.1004775-Rashkovetsky1], all well within the dispersal radius of flies [@pgen.1004775-McInnis1]. In addition, flies often return to the substrate they were collected on [@pgen.1004775-Turelli1], [@pgen.1004775-Hoffmann3] and flies collected within a locale show signatures of population structure on the order of tens of meters [@pgen.1004775-Wallace1], [@pgen.1004775-Oshima1]. Therefore, low to moderate levels of migration between demes separated by various distances [@pgen.1004775-Coyne1], [@pgen.1004775-Berry1], [@pgen.1004775-McInnis1] and environmental heterogeneity over small spatial scales may help mitigate the loss of balanced polymorphisms in any one orchard. Finally, our study identified hundreds of adaptively oscillating polymorphisms. Although the vast majority of these polymorphisms are unlinked due to the large physical distance between them, there is evidence of heterogeneity in the abundance of seasonal SNPs throughout the genome suggesting that some might be in partial linkage disequilibrium. Some models [@pgen.1004775-Korol1] have suggested that linkage between polymorphisms subject to temporally variable selection can allow for long-term persistence of both alleles at multiple sites. Taken together, we suspect *D. melanogaster* satisfies several key features required for the long-term maintenance of balanced polymorphisms due to temporal (and spatial) variation in selection pressures. Nonetheless, how do we account for the observation that these polymorphisms have been possibly maintained across different continents with clear differences in climate and between species with different ecologies [@pgen.1004775-Capy1]? The long-term persistence of these adaptively oscillating polymorphisms across populations, continents, and species suggests that these polymorphisms contribute to short-term and local adaptation in response to very generalized environmental conditions. This is in contrast to the hypothesis [@pgen.1004775-Ginzburg1] that adaptation to temperate environments in *D. melanogaster* was largely in response to novel environments, exclusively associated with life in northern, temperate locales. Rather, we speculate that the selective pressures associated with seasons in temperate environments are merely manifestations of general selective pressures resulting from cyclic population booms and busts. That is, during times of plenty, such as during the summer in temperate locales, populations rapidly expand and alleles that confer increased reproductive output or faster time to sexual maturity are strongly favored. Likewise, when population size contracts due to biotic and abiotic stressors such as those experienced during winter, alleles that confer increased stress resistance are favored. Cyclic population booms and busts are almost certainly a perennial feature of *D. melanogaster* populations, are a likely common occurrence in highly fecund species that exploit ephemeral resources, and may be an inherent property of most species in general [@pgen.1004775-Ginzburg1]. If true, we speculate that such species may harbor alleles that promote reproductive fitness during population growth (at the cost of somatic maintenance) and increase stress tolerance (at the cost of reproductive growth) during population contraction. Such balanced polymorphisms may be particularly common for species whose population cycles are decoupled from predictable environmental cues (e.g., photoperiod) but are rather linked to stochastic changes in resource abundance. For species such as these, including many microorganisms and invertebrates, balanced polymorphisms maintained by environmental heterogeneity through time and space may be the norm rather than the exception. Materials and Methods {#s3} ===================== Fly collections {#s3a} --------------- We resequenced samples of *D. melanogaster* from populations sampled over several years (2003--2010) largely during periods of peak abundance along a broad latitudinal cline in North America and during multiple time points over three consecutive years (2009 to 2011) at the Linvilla Orchard in Media, PA (39.9°N, 75.4°W). From each locality and sampling period, we collected ∼50--200 *D. melanogaster* largely by aspiration from individual fruits or baiting at strawberry fields and apple and peach orchards, established isofemale lines and collected male progeny at generation 1--5 for sequencing. One male progeny per isofemale line per population was pooled together to generate template DNA for high throughput sequencing ([Table S1](#pgen.1004775.s008){ref-type="supplementary-material"}). The only two exceptions are the second replicate sample from Maine which was derived from wild-caught males and the sample from North Carolina which was sampled from the Drosophila Genetic Reference Panel (DGRP) inbred lines. For the DGRP population, we resequenced a pooled sample consisting of one male from each of 92 DGRP strains and used allele frequency estimates from pooled samples when estimating clinality (see [@pgen.1004775-Wright1] for more information on this sample and [@pgen.1004775-Mackay1] for more information on this population). Note, there is evidence that two samples (Florida replicate 2 and post-frost Pennsylvania) show low levels of contamination with the sister species *D. simulans* (i.e., ∼ one wild caught *D. simulans* was accidentally included in our pooled sample). However, we have no evidence that the low level of contamination in two samples affects our results in any way (see [Text S1](#pgen.1004775.s011){ref-type="supplementary-material"}). Sample preparation, sequencing, and bioinformatics of pooled samples {#s3b} -------------------------------------------------------------------- DNA libraries were prepared for sequencing on the Illumina HiSeq2000 platform. To generate these libraries, we homogenized whole, male flies in 200 µL lysis buffer (100 mM Tris-Cl, 100 mM EDTA, 100 mM NaCL, 0.5% SDS) using a motorized pestle grinder. An additional 200 µL of lysis buffer was added to each sample and the homogenate was incubated at 65°C for 30 minutes. After lysis, we added 800 µL of 2 parts 5M potassium acetate, 5 parts 6M lithium chloride solution and incubated on ice for 15 minutes to precipitate proteins. The homogenate was centrifuged for at 12 K rotations per minute (RPM) for 15 minutes at room temperature, 1 mL supernatant was transferred to a new tube, and the sample was centrifuged again at 12K RPM for 15 minutes at room temperature. To precipitate DNA, we added 800 µL of isopropanol and centrifuged the sample at 12K RPM for 15 minutes. The supernatant was discarded and the DNA pellet was washed with 70% ethanol and centrifuged at 14K RPM for 10 minutes, washed with ethanol again and centrifuged once more. The ethanol was removed and the pellet was allowed to dry at room temperature. We resuspended the pellet in 100 µL TE buffer. DNA was prepared for Illumina sequencing by shearing, end-repair and ligation. To do so, 50 µL of DNA was mixed with an additional 50 µL of TE and this DNA was sheared to ∼500 bp using a Covaris machine. DNA was eluted to 30 µL using a QIAGEN PCR-purification kit (product number 28104). We performed end repair by incubating each sample of DNA with 5 µL T4 DNA ligation buffer (New England Biolabs \[NEB\] product number B0202S), 4 µL of 10 mM dNTPs, 2.5 µL T4 DNA polymerase (NEB product number M0203S), 0.5 µL Klenow large fragment (NEB product number M0210S), 2.5 µL T4 PNK (NEB product number M0201S), and 5.5 µL nuclease free water for 30 minutes at 20°C. Following incubation, DNA was purified using a QIAGEN PCR-clean up kit. Next, we performed dATP addition by incubating 32 µL of DNA with 5 µL 10× NEBuffer 2 (NEB product number B7002S), 1 µL 10 mM dATP, 3 µL Klenow Exo-minus (NEB product number M0212S), and 9 µL nuclease free water at 37°C for 30 minutes. Following incubation, DNA was purified using a QIAGEN MinElute kit (product number 28004) to a final volume of 11 µL. Sequencing adapters (custom synthesized by IDT) were ligated to DNA using T/A ligation by incubating 10 µL DNA with 2 µL T4 DNA ligation buffer, 1 µL T4 ligase (NEB product number M020S), 40 µL of 40 µM pre-annealed adapter mix and 6 µL nuclease free water for 15 minutes at 20°C followed by 65°C at 10 minutes to deactivate the DNA ligase. Finally, we performed size-selection and PCR amplification as a final step to prepare DNA sequencing libraries. Immediately following ligation, DNA was loaded into a 2%, pre-cast SizeSelect E-Gel (Life Technologies product number G661002) and run along side a 100 bp ladder. DNA at ∼500 bp was removed from the gel into a volume of ∼15 µL nuclease free water. To amplify ligated DNA, we performed two replicate PCR reactions for each sample where we used 7.5 µL template DNA, 0.25 µL of 100 µM forward and reverse primers (custom synthesized by IDT), 0.5 µL 10 mM dNTPs, 4 µL 5× High-Fidelity buffer (NEB product number B0518S), 0.5 µL Phusion High-Fidelity DNA polymerase (NEB product number M0530S), and 5 µL nuclease free water. Note, the use of two replicate PCR reactions and a high volume of template DNA was meant to prevent PCR-jackpotting. PCR was performed by 30 sections of initial denaturation at 98°C followed by 11 rounds of 10 seconds denaturation (98°C), 30 seconds annealing (65°C), 30 seconds elongation (72°C), followed by a final elongation at 72°C for 5 minutes. DNA was purified using a QIAGEN PCR-cleanup kit. Following PCR, DNA was quantified on a Life Technologies Qubit spectrophotometer as well as with a Agilent Bioanalyzer. Libraries were diluted to the appropriate concentration and sent to the Sequencing Service Center at the Stanford Center Genomics and Personalized Medicine for sequencing on the HiSeq 2000 platform. Raw, paired-end 100 bp sequence reads were mapped to the *D. melanogaster* reference genome version 5.39 using *bwa* version 0.5.9-r16 [@pgen.1004775-Li1] allowing for a maximum insert size of 800 bp and no more than 10 mismatches per 100 bp. PCR duplicates (∼5% per library) were removed using *samtools* version 0.1.18 [@pgen.1004775-Li2] and local realignment around indels was performed using GATK version 1.4--25 [@pgen.1004775-McKenna1]. We mapped SNPs and short indels (i.e., those occurring within the sequence reads) using CRISP [@pgen.1004775-Bansal1], excluding reads with base or mapping quality below 10. SNPs mapping to repetitive regions such as microsatellites and transposable elements, identified in the standard RepeatMasker library for *D. melanogaster* (obtained from <http://genome.ucsc.edu>) were excluded from analysis as were SNPs within 5 bp of polymorphic indels. SNPs with average minor allele frequency across all populations less than 15%, with minimum per-population coverage less than 10× or maximum per-population coverage greater than 400× were removed from analysis. Finally, to ensure that the examined SNPs were not artifacts of our pooled resequencing, we removed any SNP not present in the SNP tables provided by freeze 2 of the DGRP [@pgen.1004775-Mackay1] (<http://www.hgsc.bcm.tmc.edu/projects/dgrp/>). The inclusion of reads with read and mapping qualities greater than 10 (rather than greater than 20) is justified because we are restricting our analysis to common SNPs that have been previously identified in the DGRP. Of the 1,500,000 SNPs initially identified, ∼500,000 SNPs remained after applying these filters ([Table S2](#pgen.1004775.s009){ref-type="supplementary-material"}). SNPs were annotated using SNPeff version 2.0.5 [@pgen.1004775-Cingolani1]. Short intron annotations were taken from [@pgen.1004775-Lawrie1]. An annotated VCF file with allele frequency calls, genic annotations, and seaonsal/clinal p- and q-values is avaible on DataDryad (doi:10.5061/dryad.v883p). Raw sequence data are available from NCBI SRA (BioProject accession PRJNA256231, and see [Table S1](#pgen.1004775.s008){ref-type="supplementary-material"} for accession numbers of individual libraries). *F~ST~* estimates {#s3c} ----------------- To estimate average differentiation between populations or between samples collected trough time, we calculated genome-wide average (mean) *F~ST~* between pairs of populations. *F~ST~* was calculated as,where *H~total~* is the expected heterozygosity between two populations under panmixia and *H~with~* is the heterozygosity averaged between the two populations. Estimates of heterozygosity were corrected for read depth and number of sampled chromosomes by the factor,where,and where *N~chr~* is the number of sampled chromosomes and *N~rd~* is the number of reads at any site [@pgen.1004775-Nei1]--[@pgen.1004775-Feder1]. We performed a parametric permutation analysis to calculate the expected, genome-wide average *F~ST~* between pairs of populations under the null hypothesis of panmixia (spatial) or no allele frequency change through time (temporal) conditional on our experimental sampling design. To do so, we calculated the average allele frequency between any two pairs of populations or samples and randomly generated two estimates of allele frequency conditional on the average allele frequency, the number of reads at that site and the number of chromosomes sampled. To calculate the proportion of SNPs where observed *F~ST~* is greater than expected by chance, we generated 500 block bootstrap samples of ∼2300 SNPs, where one SNP was drawn per 50 kb interval. The proportion of SNPs where the observed *F~ST~* distribution is greater than expected by chance is thus,with standard deviation,where *i* refers to the *i^th^* SNP from *j^th^* block bootstrap sample. Identification of seasonally and clinally varying polymorphisms {#s3d} --------------------------------------------------------------- To identify clinally varying and seasonally oscillating polymorphisms, we used generalized linear models implemented in R 2.10 [@pgen.1004775-R1] with binomial error structure and weights proportional to the number of reads sampled at a site and the number of chromosomes sampled (see above, *N~eff~*). To identify clinal polymorphisms, we regressed allele frequency at each site (excluding all Pennsylvanian samples) on latitude ([Table S1](#pgen.1004775.s008){ref-type="supplementary-material"}) according to the form,where *y~i~* is the observed allele frequencies of the *i^th^* SNP and *ε~i~* is the binomial error given the number of effective reads (see above) at the *i^th^* SNP. To identify seasonally oscillating polymorphisms, we regressed allele frequency for the three sets of spring and fall samples on a binary variable corresponding to *spring* or *fall* according to the form,In addition, we modeled allele frequency change through time using generalized linear mixed models (GLMM) implemented in the *lme4* R package [@pgen.1004775-Bates1] and generalized estimation equations (GEE) implemented in the *geepack* R package [@pgen.1004775-Hjsgaard1]. We fit GLMMs with the model,where *(1\|population~i~)* corresponds to the random effect of population *j* and *ε~i~* corresponds to the binomial error. We fit GEEs with the model,where *population~j~* corresponds to the population level strata and *ε~i~* corresponds to the binomial error fit with an autoregressive order one correlation structure. *q-q* plots ([Fig. S2](#pgen.1004775.s002){ref-type="supplementary-material"}) demonstrate that these models (clinal and seasonal) fit the bulk of the data adequately, with the exception of the seasonal GEE model which appears to be exceedingly anti-conservative. The false discovery rate was estimated using the Benjamini & Hochberg procedure [@pgen.1004775-Benjamini1]. For seasonal SNPs, we estimated the cumulative selection coefficient as,where *f~Sp~* is the average allele frequency at seasonal SNPs in the spring and *f~Fall~* is the average allele frequency at seasonal SNPs in the fall. This estimation of S is derived from a basic model of logistic growth of a beneficial allele [@pgen.1004775-Ewens1], namely,Because we do not know the specific values of heterozygosity (*h*) nor the number of generations of selections during each season (*t*), we calculate *S* as the product of *s, h*, and *t*. Modeling the distribution of seasonal SNPs throughout the genome {#s3e} ---------------------------------------------------------------- We sought to test whether seasonal SNPs were homogeneously distributed throughout the genome. To do so, we grouped the genome into bins of 1000 non-overlapping SNPs (utilizing the ∼500,000 SNPs under investigation). For each window, we calculated the number of seasonal SNPs. The number of seasonal SNPs is Poisson distributed and we examined whether the observed distribution is over-dispersed after correcting for variation in rates of recombination within chromosomes and between the autosomes and X-chromsome. To do so, we fit the generalized linear model,where *n* is the count of seasonal SNPs per 1000 SNPs, *chrType* is the binary classification of autosome or X-chromosome, and *rec* is the average recombination rate estimated in [@pgen.1004775-Comeron1], and *ε* is the Poisson distributed error. To explicitly test if the number of seasonal SNPs is overdispersed, we used the *dispersiontest* function in the R package *AER* [@pgen.1004775-Kleiber1]. Control polymorphisms and the block bootstrap {#s3f} --------------------------------------------- Throughout our analysis, we contrasted seasonal SNPs with control polymorphisms ([Figs. 2](#pgen-1004775-g002){ref-type="fig"}--[6](#pgen-1004775-g006){ref-type="fig"}). For these analyses, we identified 500 sets of control polymorphisms matched to each seasonal SNP. For each test described in the results, control polymorphisms were identified based on different sets of characteristics that have been shown, or could plausibly, influence the parameter we sought to investigate. In general, we matched seasonal SNPs to control SNPs by chromosome, recombination rate, and allele frequency in either Pennsylvania, North Carolina, North America, and/or Africa. The choice of which population to match allele frequencies was determined by the specific test. These three parameters (chromosome, recombination rate, allele frequency) correspond with many important evolutionary processes as well as genetic patterns (e.g., [@pgen.1004775-Begun1]) and therefore control SNPs will be matched to seasonal SNPs with respect to long-term evolutionary history, gene-density, and background levels of genetic variation. In general, we used as many parameters as possible while still identifying a sufficient number of control SNPs for each test and a full list of the matched characters for each test are listed in [Table S3](#pgen.1004775.s010){ref-type="supplementary-material"}. For continuous characters, such as allele frequency, we typically rounded values so that a sufficient number of unique control sites could be identified. If no matched control SNPs were identified for a seasonal SNP, that seasonal SNP was removed from subsequent analyses. In addition, we implemented a block-bootstrap procedure to ameliorate positive dependence of our test-statistics due to linkage disequilbrium between seasonal SNPs. We generated 500 sets of seasonal SNPs where one seasonal SNP was sampled from each 50 kb consecutive interval of the genome. This block-bootstrap yielded ∼850 SNPs that were spaced approximately every 50 Kb. Estimates of expected values (E) of test statistics \[e.g. log~2~-odds-ratios ([Fig. 2C](#pgen-1004775-g002){ref-type="fig"}, [3B--C](#pgen-1004775-g003){ref-type="fig"}, [6A](#pgen-1004775-g006){ref-type="fig"}), *F~ST~* ([Fig. 2D](#pgen-1004775-g002){ref-type="fig"}), probability ([Fig. 4B](#pgen-1004775-g004){ref-type="fig"})\] and standard deviations (SD) about those expected values were calculated as, where *i* refers to control bootstrap set *i* and *j* refers to block bootstrap set *j* of any test-statistic, *TS*. Power calculations {#s3g} ------------------ To calculate statistical power of our experiment and to estimate the expected number of SNPs that are likely to vary repeatedly between seasons and along the cline we used Monte Carlo simulations based on the observed changes in allele frequency between spring and fall at seasonal SNPs or Maine and Florida at clinal SNPs. We calculated statistical power to detect seasonal SNPs as the probability of rejecting the null hypothesis of no repeatable change in allele frequency between spring and fall over three years given our sampling effort (e.g., number of chromosomes from nature and distribution of read depths in our Pennsylvanian samples) at α\<∼1e-5, corresponding to observed seasonal q-value of 0.3, conditional on *S*, the cumulative change in allele frequency between seasons calculated from the logistic function. Similarly, we calculated statistical power to detect clinal SNPs as the probability of rejecting the null hypothesis of no change in allele frequency with latitude given our sampling effort at α\<0.02, corresponding to the observed clinal q-value of 0.1, conditional on beta, the slope of the relationship between allele frequency and latitude. The expected number of seasonally (clinally) varying SNPs is then, the number of observed seasonal (clinal) SNPs at a particular value of S (beta) divided by the power to detect a seasonal (clinal) SNP at a selection coefficient S (beta). Comparison with D. simulans {#s3h} --------------------------- To estimate the extent of trans-specific polymorphism between *D. melanogaster* and *D. simulans*, we used *D. simulans* haplotype data available from the DPGP [@pgen.1004775-Begun2] (<http://www.dpgp.org/>). First, we remapped raw shot-gun sequences of each *D. simulans* strain (GenBank accessions AASS00000000 - AASW00000000) to the latest release of the *D. simulans* reference genome [@pgen.1004775-Hu1] with *bwa* version 0.5.9-r16 using the *bwa-sw* method. To convert the genomic coordinate system of the new *D. simulans* genome to the *D. melanogaster* genome, we generated a lift-over file using *lastz* [@pgen.1004775-Harris1] and components of the UCSC genome-browser toolkit [@pgen.1004775-Kent1]. Gap parameters corresponded to those used to generate the lift-over file between the first generation *D. simulans* genome and the *D. melanogaster* genome (<http://hgdownload.soe.ucsc.edu/goldenPath/dm3/vsDroSim1/>). The lift-over file to translate the coordinate system of the second generation *D. simulans* genome to the *D. melanogaster* version 5 genome is available on Data Dryad (doi:10.5061/dryad.v883p). We calculated average pairwise distance between *D. melanogaster* and *D. simulans* haplotypes at seasonal SNPs that were polymorphic in both species and shared the same two alleles by state. We calculated average pairwise distance at two windows surrounding seasonal SNPs, ±1--250 bp. Note, we excluded the focal, seasonal SNP. Pairwise distance calculations were performed using the *ape* [@pgen.1004775-Paradis1] package in R. Forward genetic simulations {#s3i} --------------------------- To simulate genome-wide allele frequency change due to cyclic changes in population size and selection at seasonally adaptive polymorphisms, we used a modified version of the forward genetic simulation software SLiM [@pgen.1004775-Messer1]. Source code for the modified version of SLiM is available upon request. In these simulations, we modeled a 20 Mb chromosome with constant recombination rate of 2 cM/Mb. For all simulations, we seeded the chromosome with 500 neutral mutations randomly placed along the chromosome all starting at 50% initial allele frequency and in complete linkage equilibrium. The number of loci under selection varied between 0 and 30 and loci under temporally heterogeneous selection were placed equidistantly along the chromosome. Selection coefficients for each selected locus were set to produce adaptive oscillations between 40 and 60% frequency every 2 (simulated 'winter') and 10 (simulated 'summer') generations. Genotypic state was assigned randomly to each simulated diploid genome at each selected locus. Population size varied over the course of each simulation. Populations grew exponentially each 'summer' to a maximum population size of 10^5^ over 10 generations. Population size instantaneously crashed at the start of winter to between 5 and 10^4^ individuals and was held constant for two generations. Simulations were run for 100 generations and *F~ST~* was estimated from the last three summer-winter cycles. Truncation selection model {#s3j} -------------------------- To estimate the upper bound of the number of loci that could plausibly respond to seasonally variable selection, we modeled a simple truncation selection scenario. For these models we calculated the expected number of winter adaptive alleles in the fall and the spring as the sum of average allele frequencies of the winter alleles in our fall and spring samples. If the oscillating alleles segregate independently, the variance in the number of winter alleles at any given time follows a Poisson distribution with mean and variance equal to the expected number of winter alleles. Therefore, the proportion of the population in the selected tail over winter is the probability of sampling the expected number of winter alleles in the spring from a Poisson distribution with mean equal to the number of winter alleles in the fall. To vary the number of independently oscillating polymorphisms in the spring and fall, we sub-sampled the number of oscillating polymorphisms 500 times for a range of values. Supporting Information {#s4} ====================== ###### Genomic turnover through space and time -- average *F~ST~*. Proportion of SNPs where average *F~ST~* among populations sampled along the cline (A) and through time (B) is greater than expected by chance conditional on our sampling design and panmixia among spatially separated populations or no allele frequency change through time, respectively. Lines represent the predicted values of Prop(Fst~Obs~\>Fst~Exp~) for the (A) linear relationship between Prop(Fst~Obs~\>Fst~Exp~) and difference latitude and (B) from non-linear relationship (y = ab^X^) between Prop(Fst~Obs~\>Fst~Exp~) and difference in months. Points represent mean *F~ST~*, error bars represent 95% confidence intervals based on blocked-bootstrap resampling. (TIF) ###### Click here for additional data file. ###### *q-q* plots and congruence of GLM, GLMM and GEE models. (A--C) Standard q-q plots of *p-values* of GLM, GLMM and GEE models, respectively. q-q plots show that GLM and GLMM models fit the bulk of the genome well whereas GEE models appear to be anti-conservative. (D) log~2~(odds-ratio) that the top 1750 seasonal SNPs identified with the GLM model are among the top 1750 seasonal SNPs identified with the GLMM model. (E) log~2~(odds-ratio) that the top 1750 seasonal SNPs identified with the GLM model are among the top 1750 seasonal SNPs identified with the GEE model. (TIF) ###### Click here for additional data file. ###### Genomic turnover through time excluding SNPs within 1 Kb of seasonal SNPs. (A) Genome-wide average *F~ST~* between samples of flies collected through time, excluding SNPs within 1 Kb of seasonal SNPs. (B) Proportion of SNPs where *F~ST~* between pairs of samples collected through time is greater than expected by chance given the null hypothesis of no allele frequency change through time and our sampling design. Solid line represents predicted relationship between genome-wide *F~ST~* and time excluding SNPs within 1 Kb; dashed line represents predicted relationship between genome-wide *F~ST~* for all common SNPs and time. The similarity between the solid and dashed line demonstrates that SNPs near seasonal SNPs are not driving genome-wide patterns of *F~ST~* through time. Lines represent the predicted values of Fst (A) and Prop(Fst~Obs~\>Fst~Exp~) (B) from non-linear regression (y = ab^X^). Points represent mean *F~ST~*, error bars represent 95% confidence intervals based on blocked-bootstrap resampling. (TIF) ###### Click here for additional data file. ###### Enrichment among cosmopolitan inversions. Log~2~ odds ratio that seasonal SNPs are enriched among the large cosmopolitan inversions relative to control polymorphisms. Inversion breakpoints are defined as ±2.5 Mb from the proximal or distal breakpoints. Error bars represent 95% confidence intervals based on blocked bootstrap resampling. (TIF) ###### Click here for additional data file. ###### Spatial *F~ST~* and clinal *q*-value. Scatter plot of the relationship between spatial *F~ST~* (x-axis) and --log~10~(clinal *q*-value). Colors of the hexagons represent the density of points in that interval. (TIF) ###### Click here for additional data file. ###### Power to detect clinal SNPs. Power to detect clinal SNPs (black line) is moderate and we estimate that we have identified ∼50% (red line) of all SNPs that change in frequency monotonically with latitude (black line). (TIF) ###### Click here for additional data file. ###### Site frequency spectrum of seasonal samples. Unfolded site frequency spectrum of spring (blue) and fall (red) samples from 2009--2010 (A) and 2010--2011 (B). Solid lines represent observed site frequency spectra, dashed lines represent simulated spring site frequency spectra conditional on one generation of bottleneck to 20 individuals and dotted lines represent simulated spring site frequency spectra conditional on two generations of bottleneck to 20 individuals. The increase in low frequency alleles in the spring 2010 sample (B, blue line) is due to the high coverage of this library. Site frequency spectra only included SNPs with allele frequencies greater than 2/(read depth) or less than 1--2/(read depth) to account for sequencing errors. (TIF) ###### Click here for additional data file. ###### Population sampling locales. (DOCX) ###### Click here for additional data file. ###### Basic SNP statistics. (DOCX) ###### Click here for additional data file. ###### Table of control characteristics. (DOCX) ###### Click here for additional data file. ###### Assessing the possibility of contamination with wild caught D. simulans. Discussion of previously identified clinal polymorphisms in relation to clinal resequencing described here. (DOCX) ###### Click here for additional data file. We thank the members of the Petrov and Schmidt labs for useful discussion and comments on previous versions of this manuscript. We also thank nine anonymous reviewers, Daniel Bolnick, and Hopi Hoekstra whose comments substantially improved the quality of this manuscript. [^1]: The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: AOB ELB KRO PSS DAP. Analyzed the data: AOB PSS DAP. Contributed reagents/materials/analysis tools: AOB ELB KRO PSS. Wrote the paper: AOB PSS DAP.

###### Strengths and limitations of this study - The study data were obtained through a questionnaire carried out by face-to-face interviews. The youth were trained to obtain dietary diet from their parent and grandparent. They were accompanied by the research assistant during data collection. - This study provides important insights of differences in food types consumed between the younger and older Pacific generation. - Exploratory factor analyses was used to examine dietary data between groups, providing robust findings. ```{=html} <!-- --> ``` - A major limitation is the small sample size of the study; however, this is a feasibility study, and the results are limited to the study participants. - The use of the dietary diversity questionnaire is based on a limited sampling frame of food group diversity over a 7-day period; therefore, it may not be completely representative of all food items and food groups that may have been consumed by the different age generations. Introduction {#s1} ============ Pacific peoples in New Zealand (NZ) are at greater risk of developing long-term conditions such as prediabetes, diabetes[@R1] and cardiovascular disease[@R4] due to obesity, defined as having a body mass index (BMI) \>30 kg/m^2^. The prevalence of obesity (67% in adults aged 15+ years) in Pacific peoples is twice that in the general population (33%).[@R5] Obesity has been viewed as a result of the changing westernised environmental pressures,[@R6] leading to an energy expenditure/energy storage mismatch that operates through dietary behaviours.[@R6] Often modernised dietary patterns, characterised by high energy dense and palatable food,[@R8] and eating habits, such as having more access to less healthy snacks and food,[@R9] explain the weight gain.[@R11] The findings from the NZ national nutritional survey on Pacific peoples reported similar patterns of energy, micronutrient intake, dietary supplement use and dietary habits (eg, consumption of breakfast and other food types, fruit and vegetable intake and salt intake), compared with the general population.[@R12] Obesity rates continue to rise for all young people aged 15--24 years, with the highest increase among young Pacific peoples (up by 50% between 2007 and 2012).[@R13] Previous NZ research has documented the impact of family meals on the dietary quality of young people[@R14] and found that eating family meals together had an overall positive effect on the home food environment[@R15] and that it encouraged the availability of more healthy food.[@R10] Results from the Youth'12 Survey highlighted similar findings that shared family meals were associated with positive outcomes for young people, which were accentuated for those living in the lower deprivation neighbourhoods (66%), compared with those in the higher deprivation areas (58%).[@R16] However, the survey results do not provide any clues as to the obesity-related mechanisms to further understanding about why young people continue to eat unhealthily. It has also been shown that young people experiencing poverty, irrespective of living in either low deprivation or more affluent neighbourhoods, does not explain overweight and obesity issues for young people.[@R17] Little research has focused on understanding the Pacific concept of socialisation and food as inter-related activities, and there is increasing recognition of the importance of this if obesity prevention strategies are to effective.[@R18] In 2014, we investigated obesity-related health issues among 30 Pacific youth from the Wellington and Auckland regions of NZ in the pilot study, '*Chewing the facts on fat! What does that say about me?*'. The methodology and scope of the study has been published elsewhere.[@R19] As part of that study, we also examined Pacific youths' diet and eating habits as it relates to obesity development. In particular, we explored dietary diversity as a form of investigating diet quality, which could be useful in identifying dietary component needs and provide insights to guide development of intervention strategies to improve diet quality and health outcomes for young Pacific peoples. The aim of the current paper is to examine the intergenerational dietary patterns among 30 young Pacific adults aged 15--24 years, and 34 parents and grandparents, from Wellington and Auckland regions. Method {#s2} ====== From a pilot study of 30 young Pacific adults aged 15--24 years in Wellington and Auckland, NZ, we investigated the social cultural determinants of the obesogenic environment. The study was conducted in two phases, and the original study methodology has been previously published,[@R19] which described the recruitment, questionnaire data and the social demography of the participants, from phase 1. This paper presents data mainly from phase 2, obtained from young Pacific peoples who were trained to interview and obtain information from one Pacific parent and one Pacific grandparent to examine and compare older generation's dietary habits with those of the Pacific youth group (phase 1) using the Pasifika dietary diversity questionnaire (described below). As Pacific peoples are not a homogenous group, we will refer to them as Pasifika, defined as a collective group of people representing the different Pacific Island nations and their respective languages, social cultural realities and protocols. Patient involvement {#s2a} ------------------- Patients were not involved in the planning or design of the study. Demography {#s2b} ---------- Basic descriptive data were obtained from the parent and grandparent as per the protocol (face to face) described in phase 1 of the study,[@R19] including measured weight and height, from which BMI was determined. We used the international standard cut-offs in defining obesity.[@R20] BMI was analysed as a continuous variable, with a BMI of ≥30 kg/m^2^ and 25--29.9 kg/m^2^ defined as being obese and overweight, respectively.[@R21] Waist-to-hip circumference was also measured, from which the waist--hip ratio (WHR) was determined to provide a measure of central adiposity to indicate associated risk of incident cardiovascular events.[@R22] The waist-to-height ratio (WtHR) was also calculated, as an adjunct measure of central obesity, which is less prone to measurement error than WHR.[@R23] Deprivation was assessed using the NZDep2013 measure,[@R25] a small area-based measure of deprivation derived from the 2013 Census, which uses nine variables (benefit income, employment, household income, communication, transport, support, qualifications, living space and home ownership) from the Census to place small area blocks on a deprivation scale from 1 to 10 with 10 representing the most deprived 10% of NZ areas, while 1 represents the 10% least deprived areas. For analyses, deprivation was categorised into quintiles combining deciles 1--2, 3--4, 5--6, 7--8 and 9--10. Ethnicity was defined according to standard NZ Census data[@R26] and self-identified by each participant; however, for participation in the project, parents and grandparents needed to be self-identifiable as being of Pacific ethnicity. Where possible, the participants were also invited to record any physician-diagnosed conditions to highlight the presence of morbidity. Dietary diversity {#s2c} ----------------- The Pacific dietary diversity questionnaire was compiled by the research team to assess the scope of individual foods and food groups in Pacific peoples' diets. The dietary diversity questionnaire aims to capture the 'range of food' that people consume over a 7-day period and not to measure quantity of food items like the 'food frequency questionnaire'. This type of assessment has been used among other indigenous groups successfully, and it has been proven to be effective than a quantitative assessment of describing the quality of the diet.[@R27] This questionnaire was pretested among an independent community group of Pasifika youth (n=30) from Wellington, and it was adapted to include common food that would be consumed by Pasifika people in NZ (eg, povi masima \[salted meat\]). Data on different individual foods and food groups that were consumed over a 7-day period (reference period) were collected and recorded (dichotomised: yes/no) by the Pacific youth who were participants in phase 1 and were trained by the research team at a single day workshop. The training involved familiarising and understanding the questionnaire and prompts. Following the training day, each youth arranged and organised a face-to face interview (accompanied by a research assistant), with a parent and grandparent. The questionnaire included both nutritious and discretionary food and food groups to encapsulate the diversity of food groups and of food items. The questionnaire data produced a total of 26 food groups (15 nutritious and 11 discretionary) specifically consumed by the study participants. However, using exploratory factor analyses (EFA) (see below) in order to create meaningful summary patterns that describe types of diet, we had refined the food groups down to 13, as determined by the total percentage of items consumed within a food grouping, per person. The groupings are: group 1: meats, poultry and fish diversity; group 2: dairy products diversity; group 3: bread, cereals and starchy vegetable diversity; group 4: legumes and nut diversity; group 5: fruit diversity; group 6: vegetable diversity; group 7: oil and fat diversity; group 8: drinks diversity; group 9: alcohol diversity; group 10: sauces, spreads and flavouring diversity; group 11: sweets and sweet snacks diversity; group 12: savoury snacks diversity; and group 13: take way food diversity. Eating habits and meal patterns, food choices and related cultural and social influences were also investigated, but the results are not presented here. We also included a measure of acculturation using a tool developed by coauthor JK and his colleagues[@R33] to examine affiliation to traditional or mainstream culture. The tool was used in this project to examine metabolic health outcomes in relation to the participants' affiliation with their Pacific heritage and or mainstream culture. This has been described in detail previously,[@R19] and the responses were grouped based on a summation of the following categories: *integrated* (high affiliation with Pacific heritage and mainstream culture); *tradition* (high affiliation with Pacific heritage only); *assimilated* (high affiliation with mainstream culture only); and *marginalised* (low affiliation with both Pacific heritage and mainstream culture). Data analysis {#s2d} ------------- Descriptive data of the parent/grandparent characteristics (ie, sex, ethnicity, education, diagnosed comorbidities, deprivation, body weight anthropometrics and acculturation mode affiliation) were analysed by distribution proportions. EFA was used for analysis of the 13 food data groups (described above). With this approach, we originally intended to identify dietary patterns between three generational groups. Due to small numbers in the older generational group, the parents and grandparents were combined to form a single group ('old'), which was compared with the youth responses ('young') obtained from phase 1. Each dietary pattern was allocated weights for each food group, which were used to calculate a standardised mean score for each dietary pattern. EFA analysis was conducted in SAS for Windows V.9.1 (SAS Institute, Cary, NC, USA), and each factor was rotated and compared (promax and varimax) to identify and improve interpretability of each factor loading. Parallel analyses and scree plots were also used to check for data interpretability. Each of the dietary pattern scores was standardised to have a mean of zero and a variance of one. Each participant was assigned a score for each dietary pattern, since a typical person's diet may include characteristics of more than one pattern. Thus, the dietary pattern scores are a constant measure of how closely the participant's diet matches each type of diet. Based on the EFA of all the participants' dietary intake, standardised scores above 0.3 (ie, threshold) for any given food grouping indicated a strong propensity matched to a particular dietary pattern. Thus, we have identified three distinctive dietary pattern groups (from 65 potential dietary groups). Our selection of the three-factor loadings was confirmed by parallel analyses. Negative scores are indicative of participants 'less likely' to consume dietary patterns. From our univariate logistic regression analyses (data not presented here), we identified potential independent factors that may explain differences in dietary scores between the young and old participants. Therefore, multivariate logistic regression was applied to further examine the association between participant's dietary patterns and those variables. (acculturation status, sex, treatment for comorbidities such as asthma, high blood pressure, heart troubles, diabetes, stroke, thyroid problems, psychological problems and sleep problems). Results {#s3} ======= Demographics of the old and young participant groups {#s3a} ---------------------------------------------------- [Table 1](#T1){ref-type="table"} highlights the main characteristics of the young and old generation groups that participated in the study. In addition to the 30 young people, a total of 34 parents and grandparents (ie, old group) took part in phase 2 of the study. There were no significant differences between the young and old gender groups. The average age of the parent's group was 50.2 years old and for the grandparent's group 72.2 years old (for both sexes). Education background showed that the more than half (52.9%) of the old group had postschool qualifications; the majority (80%) of the young group had any school qualification. Thus, education status was non-remarkable with no significant differences between the two groups. The participants self-identified across a range of ethnic groupings including different Pacific Island nations, such as Samoa, Tonga and Tokelau representing the dominant Pacific ethnic affiliations. Other ethnic groups included: Māori, Chinese and NZ European. This is an indicator of the growing 'diversity' of intermixing between Pacific and other ethnic groups. However, the important thing to note, is that the participants continue to self-identify and affiliate strongly with their Pacific culture and values. The majority of the young (80%) and old (73%) groups assessed their acculturation mode as being 'integrated', illustrating that both groups have an *equally* high affiliation with both Pacific and mainstream cultures, although this was not statistically significant. The old group rated their affiliation with living a more 'traditional' (17.7%) lifestyle, compared with the young group (6.7%), suggesting young people have less involvement within the Pacific community and other wider activities, at the time of the study. When examining socioeconomic position using the NZ Deprivation (2013) scale, comparing the old and young groups, those living in the 'least deprived' (NZDep 1--2) areas, with those living the 'most deprived (NZDep 9--10) areas, there were no differences between each group. For this study, the majority of both young (46.7%) and old (51.5%) participants lived the highest deprived areas of the region. ###### Distribution of participant characteristics Proportion (%) or mean (SD) by young and old -------------------------------- ---------------------------------------------- -------------- ---- -------------- ---- -------------- -------------- Gender  Male 19 29.7 11 36.7 8 23.5  Female 45 70.3 19 63.3 26 76.5 0.251 Ethnicity  NZ European 6 9.4 5 16.7 1 2.9 0.091  Maori 2 3.1 1 3.3 1 2.9 1.000  Samoan 30 46.9 14 46.7 16 47.1 1.000  Cook Island Maori 3 4.7 3 10.0 0 0.0 0.097  Tongan 15 23.4 7 23.3 8 23.5 0.985  Niuean 6 9.4 4 13.3 2 5.9 0.407  Tokelauan 10 15.6 5 16.7 5 14.7 1.000  Tahitian 1 1.6 1 3.3 0 0.0 0.469  Kiribatian 1 1.6 0 0.0 1 2.9 1.000  Fijian 5 7.8 3 10.0 2 5.9 0.659  Chinese 6 9.4 5 16.7 1 2.9 0.091 Education  Any school qualification 45 70.3 24 80.0 21 61.8 0.111  Postschool qualification 30 46.9 12 40.0 18 52.9 0.301 NZDep2013  1--2 4 6.4 2 6.7 2 6.1  3--4 7 11.1 4 13.3 3 9.1  5--6 5 7.9 1 3.3 4 12.1  7--8 16 25.4 9 30.0 7 21.2  9--10 31 49.2 14 46.7 17 51.5 0.673  Missing 1 Comorbidities  Asthma 16 25.0 6 20.0 10 29.4 0.386  High blood pressure 19 29.7 0 0.0 19 55.9 \<0.0001\*\*  Diabetes 11 17.2 0 0.0 11 32.4 0.0006\*\*  Other conditions 17 26.6 4 13.3 13 38.2 0.024\*  Metabolic syndrome conditions 22 34.4 0 0.0 22 64.7 \<0.0001\*\* Acculturation  Integrated 49 76.6 24 80.0 25 73.5  Traditional 8 12.5 2 6.7 6 17.7  Assimilated 4 6.3 3 10.0 1 2.9  Marginalised 3 4.7 1 3.3 2 5.9 0.462 Anthropometrics P value‡  Age 64 40.6 (22.1) 30 19.5 (2.4) 34 59.3 (12.7) \<0.0001\*\*  Weight (kg) 64 92.9 (20.6) 30 90.5 (18.3) 34 94.9 (22.5) 0.399  Height (cm) 62 166.1 (10.9) 29 170.4 (10.2) 33 162.2 (10.2) 0.002\*  BMI (kg/m^2^) 62 33.7 (7.5) 29 31.0 (6.9) 33 36.1 (7.2) 0.006\*  Waist-to-hip ratio 59 0.9 (0.11) 26 0.84 (0.05) 33 0.94 (0.13) 0.0002\*\*  Waist-to-height ratio 58 0.62 (0.10) 25 0.56 (0.09) 33 0.66 (0.08) \<0.0001\*\* NZDep2013=quintiles: 1 (lowest deprivation) to 5 (highest deprivation). \*p values \<0.05; \*\*p values \<0.001. †χ^2^. ‡Test. BMI, body mass index. Regarding the known comorbidities, these were evident primarily in the old group, and in particular, high blood pressure (59.9%) and diabetes (32.4%), with both conditions showing statistically significant differences between the young and old groups. Of note, other conditions (gout, low blood pressure, leg ulcer, arthritis, eczema, allergies) demonstrated significant differences between the old and young. In addition, we combined several conditions (high blood pressure, heart trouble and diabetes) that form part of the metabolic syndrome.[@R34] This condition indicated a significant statistical difference. Means and SD for the anthropometric measures comparing both groups illustrated that, on average, the old group were heavier (94.9 kg), compared with the young (90.5 kg) group. Across both groups, the average BMI was 33.7 kg/m^2^ (7.5 kg), and the old group showed a higher BMI (36.6 kg/m^2^) compared with the young (31.0 kg/m^2^). Nonetheless, the average BMI data indicated that the study participants were significantly obese. Moreover, central fatness as measured by the WHR and WtHR ratios, the old group showed a very high WHR ratio (mean: 0.94; 0.13), and for the young group, the average WHR ratio was 0.84 (0.05). According to the WHO, this is indicative of a higher than normal risk of developing heart disease and other serious conditions.[@R37] Both young and old groups attained an overall average (WtHR) of 0.62 (0.10), as a measure of body fat distribution,[@R38] and the scores are indicative of being at high risk of developing disease morbidities such as heart attack and stroke[@R23] but not of any new disease diagnoses, like diabetes.[@R39] Using EFA, standardised scores above 0.3 for any given food grouping were matched to a particular dietary pattern, and we identified three-factor loaded patterns ([table 2](#T2){ref-type="table"}). The patterns provided a reasonable summary of dietary patterns with the scree plot elbow being at approximately 3 and the three factors cumulatively explaining 28% of the variance in dietary intake data, resulting in interpretable dietary patterns. The following are suitable descriptions of the three dietary patterns: (1) dietary pattern 1: 'healthy food': characterised by moderate-to-high intake of proteins/meats, fruits and vegetables; (2) dietary pattern 2: 'processed food': characterised by high intake of sweets, snacks, takeaways and sugary beverages; and (3) dietary pattern 3: 'mixed food': characterised by limited healthy and more processed food group intake. Of particular note, the high intake (ie, ≥70% per food item checked) of sugary beverages (eg, soft drinks, juice and tea), alcohol (eg, beer, wine and kava) and dairy food (eg, ice cream, cheese and dairy food) consumption marks this dietary pattern as being remarkably different from the other two patterns. ###### Food grouping factor loadings for all participants by dietary pattern Food items Healthy diet Processed diet Mixed diet --------------------------------------------------------- -------------- ---------------- ------------ Group 1: meat, poultry, fish diversity 0.63 0.17 0.39 Group 2: dairy products diversity 0.38 0.19 0.68 Group 3: bread, cereals and starchy vegetable diversity 0.74 0.36 0.21 Group 4: legume and nut diversity 0.59 0.09 0.27 Group 5: fruit diversity 0.92 0.16 0.16 Group 6: vegetable diversity 0.76 0.14 0.32 Group 7: oil and fat diversity 0.38 0.28 0.47 Group 8: drinks diversity 0.30 0.44 0.57 Group 9: alcohol diversity 0.19 0.07 0.46 Group 10: sauces, spreads and flavouring diversity 0.61 0.41 0.48 Group 11: sweets and sweet snacks diversity 0.16 0.63 0.56 Group 12: savoury snacks diversity 0.22 0.97 0.05 Group 13: take away food diversity 0.18 0.64 0.42 [Table 3](#T3){ref-type="table"} compares the dietary patterns of the young and old generation using the mean scores and two-sample t tests (p value). It showed there were statistically significant dietary differences between these two groups, predominantly in food groupings 1, 3, 5 and 6. The types of food items young people consumed in very high proportions (ie, \>80% consumption) over a 7-day period were: group 1 (meat, poultry and fish diversity): beef (87%), chicken (87%) and eggs (83%); group 3 (bread, cereals and starchy vegetable diversity): white/brown bread (100%), rice (90%) and potatoes (80%); group 5 (fruit diversity): banana (100%); and group 6 (vegetable diversity): onions (93%), carrots (96.7%), lettuce (83%), garlic (80%), broccoli (80%) and tomatoes (80%). ###### Comparing young versus old dietary patterns (7-day measure) ------------------------------------------------------------------------------------------------------------- Food groups Young\ Old\ P value\* Phase 1, n=30 Phase 2,\ n=34 ----------------------------------------------------- --------------- ----------- ----------- ------- ------- Group 1: Meat, poultry, fish diversity 38.27 15.63 52.72 21.36 0.003 Group 2: Dairy products diversity 33.70 14.29 31.04 20.15 0.550 Group 3: Bread, cereals starchy vegetable diversity 40.42 17.31 49.17 17.00 0.046 Group 4: Legume and nut diversity 28.14 19.29 34.96 23.39 0.212 Group 5: Fruit diversity 31.11 21.75 48.04 25.40 0.006 Group 6: Vegetable diversity 38.65 17.72 53.26 23.45 0.007 Group 7: Oil and fat diversity 43.70 17.00 49.01 20.85 0.272 Group 8: Drinks diversity 46.22 19.57 41.17 19.86 0.311 Group 9: Alcohol diversity 7.92 16.57 8.09 14.06 0.964 Group 10: Sauces, spreads and flavouring diversity 39.84 15.18 44.65 20.22 0.292 Group 11: Sweets and sweet snacks diversity 50.83 22.88 48.04 28.21 0.668 Group 12: Savoury snacks diversity 41.67 20.86 40.59 29.02 0.864 Group13: Take away food diversity 44.05 20.47 38.44 25.66 0.342 Dietary patterns, by mean scores (%)  Healthy dietary pattern 36.07 14.93 47.13 17.90 0.010  Processed dietary pattern 45.51 19.16 42.36 24.38 0.570  Mixed dietary pattern 32.88 13.28 32.33 14.72 0.875 Dietary patterns, by standardised scores  Healthy dietary pattern −0.26 0.90 0.23 1.04 0.052  Processed dietary pattern −0.04 0.93 0.03 1.07 0.789  Mixed dietary pattern −0.09 0.94 0.08 1.06 0.496 ------------------------------------------------------------------------------------------------------------- \*Based on means score and two-sample t-test. The old generation demonstrated somewhat more variety in the items they consumed (on average) per food grouping. These were: group 1 (meat, poultry and fish diversity): chicken whole/thighs/drumstick (94%), eggs (94%), fresh white fish (eg, hoki and snapper) (91%) and chicken wings (82%); group 3 (bread, cereals and starchy vegetable diversity): white/brown bread (100%), rice (88%) and taro (88%); group 5 (fruit diversity): banana (100%), apple (88%) and orange (85%); and group 6 (vegetable diversity): onions (94%), lettuce (85%), garlic (85%), cabbage (82%) and cauliflower (80%), as indicated by listing the individual food items. There were also differences between the two old generations (ie, parent vs grandparent) (data not shown) by: group 9 (alcohol diversity): examples of items consumed included wine and kava. The median score for parents was 0.65 and for grandparents was 0.28 (p value=0.004); and group 11 (sweets and sweet snacks diversity): examples of items consumed included plain biscuits, cakes and chocolates. The median score was 0.65 for parents and 0.28 for grandparents (p value=0.033). This indicates that the parent generation consumes statistically significantly more alcohol and high sugar snacks than the grandparents. Additionally, based on '*percentage dietary patterns scores*', there was a significant difference (p value=0.010) between the young and old groups for the 'healthy dietary pattern', with the older group showing a higher percentage intake of this dietary pattern. Based on the '*standardised dietary pattern scores*', comparing the young and old groups, there was a trend for significance (p value=0.052). The scores indicated that the young group were less likely to consume a healthy dietary pattern than the old group. We carried out univariate analyses (data not presented here) and identified certain characteristics that contributed to the difference in dietary pattern scores. Therefore, [table 4](#T4){ref-type="table"} illustrates a multivariate regression model using standardised regression coefficients to measure the strength of independent relationships of (known relationships and statistically significant) variables identified from earlier analyses. The key independent variables shown to have a positive influence on the dietary patterns were predominantly *acculturation*, particularly 'assimilation' and 'marginalised' modes, and high *socioeconomic deprivation*, particularly quintiles 7--8 and 9--10. After controlling for other variables, trending for significance for BMI had dissolved, and age indicated no relationship for processed and mixed dietary patterns. ###### Multivariate analyses examining dietary factor patterns and demographic characteristics Variable (number) Factor 1: healthy diet Factor 2: processed diet Factor 3: mixed diet --------------------- ------------------------ -------------------------- ---------------------- ------- --------------- -------------- ------- --------------- -------------- Phase  Old (30) Ref Ref Ref  Young (34) 0.05 −0.81 to 0.92 0.904 0.09 −0.86 to 1.04 0.850 0.12 −0.70 to 0.95 0.762 Gender  Male (19) Ref Ref Ref  Female (45) −0.02 −0.58 to 0.54 0.941 −0.14 −0.64 to 0.36 0.575 −0.19 −0.74 to 0.37 0.504 Acculturation  Integrated (49) Ref Ref Ref  Traditional (8) −0.65 −1.15 to 0.15 0.013\* −0.36 −0.86 to 0.14 0.155 −0.62 −1.14 to 0.09 0.021\*  Assimilated (4) −0.88 −1.47 to 0.29 0.004 −1.06 −1.57 to 0.55 0.0001\*\* −1.03 −1.59 to 0.47 0.0006\*\*  Marginalised (3) −1.30 −2.27 to 0.33 0.010\* −1.23 −2.17 to 0.29 0.011\* −1.31 −2.13 to 0.49 0.002\* NZDep13 (quintiles)  1--2 (4) Ref Ref Ref  3--4 (7) 0.39 0.08 to 0.69 0.013\* 0.45 0.25 to 0.65 \<0.0001\*\* 0.48 0.17 to 0.79 0.003\*  5--6 (5) 0.96 0.14 to 1.78 0.022\* 0.94 0.28 to 1.60 0.006\*\* 0.82 −0.06 to 1.69 0.067  7--8 (16) 1.26 0.81 to 1.71 \<0.0001\*\* 1.10 0.66 to 1.54 \<0.0001\*\* 1.08 0.67 to 1.47 \<0.0001\*\*  9--10 (31) 1.00 0.65 to 1.32 \<0.0001\*\* 0.97 0.66 to 1.28 \<0.0001\*\* 0.90 0.57 to 1.23 \<0.0001\*\* BMI (n) (SD) (SD) (SD)  62 0.015 0.01 0.226 0.013 0.01 0.302 0.01 0.01 0.140 Age  64 0.012 0.01 0.240 −0.00 0.01 0.952 0.00 0.00 0.797 P value=χ^2^ test. \*p\<0.05; \*\*p\<0.001. BMI, body mass index; RCE, regression coefficient estimate; P value=chi square test. Discussion {#s4} ========== In a sample of 30 Pasifika youths and 34 older Pasifika adults (parents and grandparents), we identified three distinctive dietary pattern groups: healthy diet, processed diet and mixed diet. In relation to these dietary patterns, there were three major findings. First, although the mixed dietary pattern contained 8 of the total 13 food groups, it had the highest average consumption of dairy products, sweetened beverages and alcohol intake, potentially making it the unhealthiest of the three dietary patterns. Although the questionnaire did not measure the 'actual' amounts, frequency or average volume of drinks per week of the sweetened and alcohol beverages, it has been well established that alcohol is the second most energy dense macronutrient[@R40] and favours fat storage resulting in weight gain.[@R37] Similarly, sugar-sweetened beverages and fast food were more available within the home environment. A recent study reported 58% of Pacific children consume high amounts of fizzy drinks compared with non-Pacific children (15%). Moreover, Pacific youth consumed high (29%) and moderate (45%) amounts (ie, 1--3 times a week) of soft drinks in the past week compared with 25% of children who were considered to be low consumers.[@R41] Data from the NZ Youth Survey 2012 revealed that in the previous 7 days[@R42] youth residing in the highest deprivation decile were almost four times more likely to consume four or more energy drinks, three times more likely to eat fast food at least four times a week and five times more likely to eat other takeaways at least four times. This is compounded by social marketing targeting the younger generation as their main customers due to their spending power, purchasing influence, as emergent behaviours of adult customers.[@R43] With the advent of online advertising and shopping and mobile media, public health policy and researchers are lagging well behind in being aware of and understanding the potential impact of the latest and continuously developing advertising techniques used on children and youth. More work and comprehensive monitoring are needed in this space.[@R44] Second, there were major significant differences reported from [table 3](#T3){ref-type="table"} between the young and old in relation to the four food groups: (1: meat, poultry and fish; 2: breads, cereal and starchy vegetables; 3: fruit diversity; and 4: vegetable diversity). More specifically, the old generation ate a greater range of items from each food group than the young generation. Furthermore, dietary patterns between the two groups also differed. The mean food scores for the old group being primarily characterised by the 'Healthy diet', in contrast with the younger generation who were more likely to consume a 'Processed diet'. Multiple factors may influence eating behaviours and food choices between young and old. For instance, the idea that eating family meals promotes togetherness and is positively associated with overall well-being within the home food environment is an important factor. However, a NZ-based research study reported no significant relationship between the frequency of family meals and the availability of high sugar/high fat snack foods and fast food. In fact, adolescents who regularly ate with their family were just as likely to consume less healthy snack food. Parental education and family work schedules were regarded as potential explanatory factors for the lack of association.[@R10] For our study, all of our youth lived at home with their parents, and although we did not specifically collect data on the home food environment, we have previously reported data on this study groups' purchasing behaviour. That is, the youth purchased savoury and sweet snack food from specific shops in their local neighbourhoods spending up to \$24 over a 7-day period.[@R19] It is likely that this behaviour is contributing to the difference in dietary diversity between the two generational groups. Food purchasing and eating behaviours outside of the home environment is a tremendously influential force, particularly in regards to the obesogenic environment.[@R45] More understanding around the impact of health education and health promotion messages targeting healthy eating is needed to make mindful decisions and subsequent behavioural changes to support vulnerable communities for the sake of healthier living. However, by comparing standardised food scores, the difference between young and old (particularly for the healthy dietary pattern) trended towards significance (p value=0.052), and it is likely that other factors may explain the difference rather than just the food types consumed between the two generation groups. Finally, from our multivariate regression analyses, we found that *acculturation* (categorised affiliation with Pasifika and mainstream cultures) and *socioeconomic position* (NZ Deprivation) played a significant role in influencing dietary patterns. For acculturation, those who described themselves as being 'assimilated' (ie, a high degree of affiliation with NZ's mainstream culture) and 'marginalised' (ie, a low affiliation with both Pacific heritage and mainstream culture) were significantly less likely to exhibit characteristics of these dietary pattern groups, based on their high negative coefficients. Although they were significant negative relationships for 'traditional', 'assimilation' and 'marginalised' modes across all dietary patterns, it was more marked (high negative coefficients) for those who affiliated with the 'marginalised' mode for healthy diet (regression coefficient estimate \[RCE\]: −1.30), processed diet (RCE: −1.23) and mixed diet (RCE:−1.31), compared with those identified as being 'integrated' with both Pacific heritage and mainstream culture. Similarly, those who aligned with the 'assimilation' mode also indicated significantly high negative relationships with processed diet (RCE: −1.06) and mixed diet (RCE: −1.03). However, this finding needs to be interpreted with caution (ie, the wide CIs) because of the small number of participants for this specific analyses. Nonetheless, this finding suggests that the 'assimilated' and 'marginalised' groups have different behaviours that may be protective for their overall health (eg, they may do more physical activity and eat more healthily), compared with the majority of this sample (who were 'integrated'). There are studies that suggest greater acculturation increases the risk of obesity among indigenous groups.[@R46] Some studies suggest that the obesogenic environments of host countries promote weight gain among migrants that significantly increase over 10--15 years postmigration, by which time migrants' obesity rates match or become greater than that of the host population.[@R49] The impact of the relationship between acculturation and obesity continues to remain unclear, however, there is a need for validated and comprehensive acculturation scales to be used between studies to enable comparability of results.[@R33] Socioeconomic deprivation as measured using the NZ Deprivation scale also highlighted strong positive relationship with the three dietary patterns. In particular, those people residing in the high-to-highest deprived areas (quintiles 7--8 and 9--10) had strong positive relationships, compared with those living in the least deprived areas (quintile 1--2), where the coefficients were closer to 0. This finding is not new but provides added support to the knowledge-base that high deprivation plays an important role in financial and economic barriers to have access to good quality food (particularly for processed and mixed dietary patterns).[@R51] However, for those living in the high-to-highest deprived areas, showing a positively strong relationship (quintile 7--8 RCE: 1.26 and quintile 9--10 RCE: 1.00) to a 'healthy diet' maybe reflective of some people instigating the efforts to improve behaviours and preparing and eating food at home, rather than purchasing meals and snacks prepared away from the home, that are higher in fat and saturated fat and contain less nutritional value[@R51] Study limitations {#s4a} ----------------- A major limitation of the study is the sample size that restricts our ability to make any definitive inferences of the observed dietary patterns described above. As this is a feasibility study, these patterns reflect the behaviours of the study participants and cannot be generalised to Pasifika youth and adults in general. In addition, the Pacific-focused dietary diversity questionnaire was developed, including both nutritious and discretionary foods and food groups to capture diversity (food groups) and variety (food items) and does not measure quantity and so we cannot be definitive about the defined dietary patterns, which is also reflective of the study participants. The EFA analyses allowed us to examine the highest and lowest variety of food groups that were consumed, and this provided important insight in identifying which dietary components needs to be addressed, and it could guide intervention strategies to improve diet quality to improve health outcomes. Furthermore, we are cognizant that the data collected from the limited sampling frame of food group diversity over a 7-day period may not be completely representative of all food items and food groups that may have been consumed by the different age generations. Finally, although the study highlighted behavioural patterns of diet, the study also showed 'no relationship' between the dietary patterns and obesity-related parameters (BMI, WHR and WtHR) and poor health, and this may be due to limited sample size, as other studies have shown clear associative links. Conclusion {#s5} ========== Our investigation of the food consumed by young and old Pacific people allowed us to identify three distinctive dietary patterns based on the high food scores. There was an intergenerational difference in dietary patterns, with the younger generation tending to consume a more limited diet, compared with the old group. The older generation consumed more diversity in their food groups and a healthier dietary pattern. These patterns may be related to the social cultural aspects of obesity in relation to dietary habits, such as the young generation who are opting to eat more high fat, high sugar snacks outside of the home food environment. There was a strong negative relationship between the level of acculturation and dietary patterns, and a strong positive relationship of dietary patterns among those living in the high-to-highest deprivation areas. Our findings provide added social cultural insights that could guide improved health promotion strategies to increase health outcomes of young and older Pacific family members. In particular, improvements in highlighting more healthy food group options, encourage food diversity within the home and social environments and enhance access to health education about food items for the whole family. Supplementary Material ====================== ###### Reviewer comments ###### Author\'s manuscript The authors would like to acknowledge the partnership funding from the Health Research Council of NZ and the Ministry of Health who supported this study. We also thank our collaborating partners: The Fono and Evolve for participating in this project. We acknowledge the efforts of collaborators: Bernhard Breier, Rozanne Kruger and Te Kani Kingi, and our research assistants: Ms Hana Tuisano and Mrs Moana Manukia for conducting the interviews with the young participants. Finally, the authors are indebted to the efforts and participation of the Pasifika youth and their families. **Contributors:** RTT-F designed the study and drafted the manuscript. JK, BB and LE-L help informed the study design and JK assisted in interpretation of the acculturation tool. SC analysed the data. All authors read and approved the final manuscript. **Funding:** This work was supported by the Health Research Council of NZ and the Ministry of Health grant (12/953). **Competing interests:** None declared. **Ethics approval:** Ethical approval for the overall study was received from the Central Health and Disability Ethics Committee, NZ (13/CEN/22). **Provenance and peer review:** Not commissioned; externally peer reviewed. **Data sharing statement:** No additional data available. **Patient consent for publication:** Not required.

1. Introduction {#sec1-ijerph-17-00783} =============== The decline in youth's healthy behaviors and related consequences \[[@B1-ijerph-17-00783]\] is of concern for public health in general and for national defense in particular, which requires a sufficient number of physically fit and mentally healthy military personnel. In Lithuania, there is evidence that young people do not partake in adequate physical activity; that is, only around 30% of students aged 18 years comply with the recommendation to be active ≥1 h on at least 5 days a week \[[@B2-ijerph-17-00783]\]. The trends for health-related physical fitness among adolescents have deteriorated in the past 20 years, and the indicators of cardiorespiratory fitness have decreased by nearly 50% during this period \[[@B3-ijerph-17-00783]\]. Only 13--14% of high school students comply with recommendations for healthy nutrition \[[@B4-ijerph-17-00783],[@B5-ijerph-17-00783]\], 11% of young people consume alcohol \[[@B6-ijerph-17-00783]\]. Moreover, 22% experience psychological distress \[[@B7-ijerph-17-00783]\]. Multiple health behaviors, psychological distress, and their associated risk factors, such as obesity and poor physical fitness, are associated with poor mental health outcomes, the risk of cardiovascular disease, type 2 diabetes, certain types of cancer \[[@B8-ijerph-17-00783],[@B9-ijerph-17-00783]\], depression \[[@B10-ijerph-17-00783],[@B11-ijerph-17-00783]\], and anxiety \[[@B12-ijerph-17-00783]\]. Unhealthy behaviors also impose an economic burden; for example, one study estimated that physical inactivity alone costs USD 53.8 billion in 2013 for healthcare worldwide \[[@B13-ijerph-17-00783]\]. Health is the main criterion for accepting or rejecting young men into military service (MS) and is strongly related to the ability to perform military duties. Unsatisfactory results of youth recruitment to Lithuanian MS have been presented: on average, 58.6% of Lithuanian young men proceed through the full procedure of military enlistment. Among those who do not pass these procedures, 33%--37% experience psychological problems, 29%--33% have cardiovascular diseases, and 13% have musculoskeletal problems \[[@B14-ijerph-17-00783]\]. Other countries face similar problems of rejection from MS. For instance, analysis of the reasons for rejection from MS in the USA found that about 22% of rejections were because of problems with bones or joints, flat feet, or hernias, 15% because of organ defects, 13% because of defects of the cardiovascular system, 12% because of nervous system or mental problems, and 10% because of communicable diseases \[[@B15-ijerph-17-00783]\]. In the USA, Hispanic men appear to have a better health profile than their white and black peers, except for the prevalence of overweight, which is higher in Hispanic men \[[@B16-ijerph-17-00783]\]. Although health behaviors alone are not criteria for enlistment into MS, they might explain the reasons for some instances of rejection. Given that some health behaviors are risk factors for the occurrence of many lifestyle-related diseases \[[@B17-ijerph-17-00783]\], it is critical to identify whether and how health behaviors differ between young men who are deemed eligible and those who are ineligible for MS and to take appropriate actions to prevent adverse health behaviors from their onset. Soldiers must be both physically and mentally healthy. However, mental health issues are among the main factors for rejection from MS \[[@B14-ijerph-17-00783],[@B15-ijerph-17-00783]\]. Identifying psychological distress along with health behaviors might help to provide a more complete understanding of health indicators in conscripts as psychological distress is an indicator of mental health \[[@B18-ijerph-17-00783]\]. Given the associations between many health-related behaviors, the complex analysis of a set of risk behaviors instead of evaluation of individual associations may help to reduce the risk of missing potential confounders for enlistment into MS \[[@B19-ijerph-17-00783]\]. Several studies have examined psychological distress, health behaviors \[[@B20-ijerph-17-00783],[@B21-ijerph-17-00783]\], and changes in health behavior \[[@B22-ijerph-17-00783]\] during MS. Enlistment is based on a medical examination of draftees, which may reject unhealthy individuals. As a result, these studies have evaluated health behaviors in relatively healthy youth but have not examined whether health behavior is related to the rejection of military recruits. The aim of this study is to identify and compare health behaviors and psychological distress between male conscripts rejected for and enlisted into MS. We expect to find that healthier behaviors and low distress level would be related to a higher rate of enlistment into MS. 2. Materials and Methods {#sec2-ijerph-17-00783} ======================== 2.1. Study Design and Procedure {#sec2dot1-ijerph-17-00783} ------------------------------- This nationally representative cross-sectional study was performed among Lithuanian conscripts. In Lithuania, 9-month-long MS is compulsory. In accordance with the conscription procedure of the Lithuanian Armed Forces, a list of potential draftees is created by an automatic electronic selection system each year and includes male Lithuanian citizens of compulsory MS-eligible age (19--26 years). Other male and female young adults can express their wish to be conscripted on a priority basis. A stratified random sampling was used. There are four centers for military recruitment in Lithuania that recruit conscripts during the whole year. The data were gathered in all four centers for 4 months from June to October 2018. Each conscript in these centers was approached during this period and asked to sign a consent form to participate in the study and to complete the study questionnaire. There was an equal chance (probability) that participants included in the study would be enlisted or rejected for MS. The decision to accept a recruit for enlistment into MS is made by medical experts and is based on an individual medical examination and previous medical reports. The criteria for rejection from MS are defined in Order V-1142/V-1139, which provides a list of disorders along with their severity, and is signed by the Ministers of State Defense and Health Care. These criteria also indicate the minimum height requirement: 160 and 155 cm for men and women, respectively. Obesity without comorbid illness is not a reason for rejection from MS \[[@B23-ijerph-17-00783]\]. Enlistment in and rejection from MS were identified from the medical records for each study participant. 2.2. Participants {#sec2dot2-ijerph-17-00783} ----------------- In 2018, there was a list of 10,340 conscripts, 209 of them female. This study included 1427 conscripts (which represented 13.80% of the total population), of whom 1296 returned their completed questionnaire along with their consent to participate in the study. The response rate was 90.82%. Female recruits were invited to participate in the study, but because only 53 agreed to participate, they were later excluded because of the small sample size. Finally, 1243 young male potential conscripts were included in the analysis. The participants were aged 19--26 years and their mean age was 22.50 ± 2.43 years. The Ministry of National Defense of the Republic of Lithuania approved the research. Ethics approval (No SMTEK-28, 2018) was obtained from the Ethics Committee of Lithuanian Sports University. The investigations were carried out following the rules of the Declaration of Helsinki of 1975, revised in 2013. Participants were informed of the tasks in the study before data collection, and all participants gave their informed consent for inclusion before they participated in the study. 2.3. Measurements {#sec2dot3-ijerph-17-00783} ----------------- ### 2.3.1. Physical Activity {#sec2dot3dot1-ijerph-17-00783} The World Health Organization (WHO) defines moderate physical activity as activity that noticeably accelerates the heart rate and includes activities equivalent in intensity to brisk walking or bicycling. Vigorous physical activity causes rapid breathing and substantially increases heart rate, and includes activities such as jogging, aerobic dance, and bicycling uphill \[[@B24-ijerph-17-00783]\]. To assess physical activity in the participants in this study, we used the 2005 US Department of Defense Survey of Health Related Behaviors Among Active Duty Military Personnel \[[@B21-ijerph-17-00783]\]. The study participants were asked, "During the past 7 days, for leisure-time physical activity, how often did you usually do each of the following?". Participants were also asked, "During the past 7 days, when you did leisure-time physical activity, how long did you usually do each of the following?" For both questions, detailed descriptions and examples of what constitutes moderate and vigorous physical activity were presented. In this study, we assessed physical activity during the preceding 7 days instead of the 30 days in the original questionnaire. Also, instead of using categorical answers (such as "5 or 6 days" and "at least 20 minutes") for each question, we provided the opportunity to write the exact numbers of days, hours, and/or minutes per day. The number of minutes spent in was totaled. Participants whose MVPA was \<2.5 h/week were coded as not meeting health-related physical activity requirements and those whose MVPA was ≥2.5 h/week were coded as meeting health-related physical activity requirements \[[@B25-ijerph-17-00783]\]. ### 2.3.2. Eating Patterns {#sec2dot3dot2-ijerph-17-00783} A healthy eating pattern was defined as a diet based on whole or minimally processed foods and that included health-protective ingredients and lacked unhealthy items such as fast food and sugar-sweetened beverages \[[@B26-ijerph-17-00783]\]. Adherence to a healthy eating pattern was evaluated using the Mediterranean Diet Adherence Screener (MEDAS) \[[@B25-ijerph-17-00783]\], which was previously validated in Spanish and German adult populations \[[@B26-ijerph-17-00783],[@B27-ijerph-17-00783]\]. The MEDAS includes 14 items: two indicate food-intake habits such as the use of olive oil and preference for white versus red meat, and the other 12 items capture the frequency of consumption of olive oil, animal fat, vegetables and fruits, fish, nuts, commercial pastries, sugar-sweetened beverages, and dishes with homemade sauce. Each item was scored as 0 (does not meet the healthy eating criteria) or 1 (meets the healthy eating criteria). The total score was calculated by summing all item scores. The MEDAS score was classified into three categories: ≤7 indicated low adherence, 8--9 indicated medium adherence, and ≥10 indicated high adherence to the Mediterranean diet \[[@B26-ijerph-17-00783]\]. ### 2.3.3. Alcohol Consumption {#sec2dot3dot3-ijerph-17-00783} Alcohol consumption was evaluated according to the 2008 US Department of Defense Survey of Health Related Behaviors Among Active Duty Military Personnel \[[@B20-ijerph-17-00783]\]. Participants were asked to indicate the number of drinking occasions during the past month with the question, "During the past 30 days, on how many days did you drink alcohol?". Participants had to identify the quantity of drinks per typical drinking occasion in the table, where columns presented the type of beverage classified as the number of beers (bottle or can = 330 mL), the number of wines (glass = 125 mL), or the quantity of whiskey/vodka (mL), liquor (mL), and other drinks (mL) by indicating the quantity or amount of each of the drinks in the columns. The number of occasions per month when ≥5 drinks were consumed was determined with the question, "During the past 30 days, on how many days did you have 5 or more drinks of beer, wine, or liquor on the same occasion?" (By "drink," we mean a bottle or can of beer, a wine cooler or a glass of wine, a shot of liquor, or a mixed drink or cocktail. By "occasion," we mean at the same time or within a couple of hours of each other.) Following the 2008 Department of Defense Survey of Health-Related Behaviors Among Active Duty Military Personnel \[[@B20-ijerph-17-00783]\] paper, participants were allocated into three groups according to alcohol consumption: abstinent (0 alcoholic drinks in the past month), light--moderate drinker (1--4 drinks per typical drinking occasion or ≥5 drinks per typical drinking occasion 1--3 times/month), heavy drinker (≥5 drinks per typical drinking occasion ≥1/week (on average) in the 30 days before the survey). The classification was adapted from Mulford and Miller (1960) \[[@B28-ijerph-17-00783]\]. To provide a continuous variable indicating the total amount of alcohol consumed, the total number of alcohol units per month was calculated as one standard unit equal to 10 g ethanol. In accordance with the WHO guidelines, the following formula was used: volume (mL) multiplied by the percentage of alcohol and divided by 1000 \[[@B29-ijerph-17-00783]\]. ### 2.3.4. Cigarette Smoking {#sec2dot3dot4-ijerph-17-00783} Cigarette smoking was evaluated as indicated in the 2008 US Department of Defense Survey of Health Related Behaviors Among Active Duty Military Personnel \[[@B20-ijerph-17-00783]\]. Participants were asked to indicate the number of cigarettes smoked in the past month. Given that there is no safe amount of tobacco use, as reported by the 2014 Surgeon General's Report \[[@B30-ijerph-17-00783]\], two categories of smoking were created: current nonsmoker (smoked 0 cigarettes in the past 30 days and never smokers) and current smoker (smoked ≥1 cigarette in the past 30 days). ### 2.3.5. Psychological Distress {#sec2dot3dot5-ijerph-17-00783} Psychological distress was assessed using the six-item Kessler scale \[[@B31-ijerph-17-00783]\]. Participants were asked to evaluate their nervousness, hopelessness, anxiety, restlessness or fidgety feelings, worthlessness, and depression. Each item was scored from 0 (none of the time) to 4 (all the time). The total score was calculated by summing the scores for each item and ranged from 0 to 24 points, with a lower score indicating a lower level of psychological distress. The internal consistency of the scale was good (Cronbach α = 0.876). The summed score was dichotomized as low psychological distress (0--12 points) and high psychological distress (≥13 points) \[[@B31-ijerph-17-00783]\]. 2.4. Covariates {#sec2dot4-ijerph-17-00783} --------------- Body mass index (BMI) was evaluated using the self-reported height and weight and was calculated as weight (kg)/height (m^2^). Family income was evaluated by asking participants to rate their current family financial situation as having insufficient income, having average income, having higher than average income, wealthy, and rich. Age (in years) was also used as a covariate. 2.5. Statistical Analysis {#sec2dot5-ijerph-17-00783} ------------------------- The data were analyzed using SPSS (ver. 24, IBM SPSS Statistics; IBM Corp., Armonk, NY, USA). Categorical variables are presented as frequencies and percentages. Differences in categorical variables were tested using the chi-squared test. The relationships between independent continuous variables were examined using Pearson correlational analysis because all scaled variables had a normal distribution. Logistic regression analysis was used to identify significant associations between dependent (enlistment into MS) and independent variables (covariates, health behaviors, and psychological distress level) by adding independent variables individually into the next model and producing odds ratios (ORs) and 95% confidence intervals (95% CIs). The final model contained adjusted OR (95% CI). Statistical significance was set at *p* \< 0.05. 3. Results {#sec3-ijerph-17-00783} ========== Among 1243 conscripts who went through the enlistment procedures, 624 (50.2%) were rejected for and 619 (49.8%) were enlisted into MS ([Table 1](#ijerph-17-00783-t001){ref-type="table"}). Around half were adequately physically active. Among youth enlisted to MS, there were 63.1% of physically active versus 47.3% of physically active among youth rejected for MS (p \< 0.01). The level of adherence to the Mediterranean diet was low among half of Lithuanian male conscripts, a little bit less than half of them had medium adherence, and only a small percentage (3.5%) of male conscripts complied with healthy eating patterns. However, there were no significant differences between rejected and enlisted youth (*p* \> 0.05). Almost a quarter of all conscripts reported their alcohol abstinence for the previous 30 days, two-thirds reported a light--moderate drinking pattern, and almost one out of ten (9.6%) reported heavy drinking. The percentage distribution after Bonferroni correction indicated that among youth enlisted to MS, there were 6.8% of heavy drinkers versus 12.3% of heavy drinkers among rejected youth (*p* \< 0.01). Percentages of light--moderate drinking and abstinence for the 30 previous days were not significantly different (*p* \> 0.05) between rejected and enlisted youth. Almost two-thirds of male conscripts were current smokers; however, smoking was more frequent among those who were rejected for MS (67.7%) than among those who were enlisted (57.0%) (*p* \< 0.0001). A high level of distress was reported in 9.1% of the total population of male conscripts. However, among rejected youth, high distress had 14.7% of prevalence versus 3.4% of high distress prevalence among enlisted youth (*p* \< 0.0001). Continuous variables for health behaviors and distress were included in the correlational analysis to examine the relationships between the main predictors of enlistment, and separate analyses were performed for the rejected and enlisted conscripts ([Table 2](#ijerph-17-00783-t002){ref-type="table"}). In both groups, there were significant relationships between better adherence to healthy eating patterns and lower smoking levels (r = −0.011 and r = −018 for rejected and enlisted, respectively) and lower psychological distress levels (r = −0.12 and r = −0.22 for rejected and enlisted, respectively). Smoking and distress were also significantly correlated (r = 0.12 in both groups), which showed that a higher psychological distress level was associated with a higher average number of cigarettes smoked per day. In both groups, higher alcohol consumption was related to higher smoking levels (r = 0.22 and r = 0.24 for rejected and enlisted, respectively) and higher distress levels (r = 0.21 and r = 0.27 for rejected and enlisted, respectively). Among the men enlisted into MS, there were also significant associations between greater alcohol consumption and lower adherence to healthy eating (r = −0.12), and between higher compliance to healthy eating patterns and a higher physical activity level (r = 0.13) (*p*~s~ \< 0.05). Enlistment into MS was predicted using multiple logistic regression analysis ([Table 3](#ijerph-17-00783-t003){ref-type="table"}). When entered individually into the model, enlistment into military service was associated with MVPA (OR = 1.97; CI 1.57--2.47), adversely associated with heavy drinking (OR = 0.60; CI 0.45--0.79), current smoking (OR = 0.61; CI 0.49--0.78), and psychological distress (OR = 0.20; CI 0.13-0.33). After controlling for all variables included in the regression analysis in Model 7, the significant predictors of enlistment remained higher MVPA, current non-smoking, and low level of distress. Adherence to healthy eating was not significantly related to enlistment neither entered individually nor after controlling for covariates and other study variables (*p* \> 0.05). Among the covariates, only older age predicted a lower chance of being enlisted (OR = 0.77; 95% CI 0.71--0.83). Neither BMI nor perceived family income was associated with enlistment (*p* \> 0.05). 4. Discussion {#sec4-ijerph-17-00783} ============= This study was aimed at determining the prevalence of healthy behaviors and psychological distress in recruits for MS and at identifying differences in these indicators between young men rejected for and enlisted into MS. The results partly support the premise that young men who are enlisted into MS are more likely to adhere to healthy behaviors and have lower psychological distress levels compared with those rejected for MS. 4.1. Physical Activity {#sec4dot1-ijerph-17-00783} ---------------------- In the regression analysis, after controlling for covariates and other study variables, meeting the recommendations for MVPA was significantly associated with being enlisted into MS. Previous studies have shown that physical activity can help prevent 26 different chronic diseases \[[@B17-ijerph-17-00783]\] and is related to a lower incidence of metabolic syndrome, lower rates of stress, and better mental health and cognitive functions \[[@B17-ijerph-17-00783],[@B32-ijerph-17-00783],[@B33-ijerph-17-00783]\], all of which are important to being able to serve in the military. Physical activity is also a strong predictor of physical fitness \[[@B34-ijerph-17-00783]\], which is a key indicator of physical health during MS. Other research has suggested that physical activity dampens the inflammatory processes in the body \[[@B34-ijerph-17-00783]\]. Higher-intensity physical activity confers greater health benefits, such as reduced body fat and central adiposity, and higher cardiovascular fitness in youth, compared with lower-intensity activity \[[@B35-ijerph-17-00783]\]. In the current study, 55.3% of all conscripts had an adequate level of MVPA, but this percentage was lower in those rejected for MS (47.4%) than in those enlisted (63.1%) into MS. These results are similar to data reported in the general population of similar age in the Eurobarometer (2017) data, which showed that 55% of male youth in Lithuania and 57% of male youth from other European countries aged 15--24 years engage in physical activity regularly or with some regularity \[[@B36-ijerph-17-00783]\]. The physical activity level generally increases in those enlisted into MS after they start their duty because MS enforces its own standards for health behaviors such as physical activity by, for example, requiring daily physical activities. Surprisingly, however, it has been reported that 13% of those serving in the US active duty military perform insufficient exercise \[[@B37-ijerph-17-00783]\]. 4.2. Cigarette Smoking {#sec4dot2-ijerph-17-00783} ---------------------- Along with physical activity, current smoking also predicted enlistment into MS. The logistic regression analysis after controlling for covariates and other study variables indicated that current smokers had a 72% lower chance of being enlisted than nonsmokers. In total, around two-thirds of Lithuanian male youth called to serve in the military are current smokers. However, among those enlisted to the military, there are 10.7% less of them than among those rejected for MS. A study in Taiwan reported that 50.8% of young men in their MS smoked before service \[[@B38-ijerph-17-00783]\]; this percentage is 10 percentage points lower than in Lithuania. In the US, 23% of active duty personnel smoke \[[@B36-ijerph-17-00783]\], which is much lower than the 57% smoking rate in Lithuanian conscripts enlisted into MS in our study. The statistics for the general Lithuanian population confirm the high prevalence of smoking: one in three men smoke every day in Lithuania. This is the fourth highest rate in the European Union \[[@B39-ijerph-17-00783]\]. Smoking is related to obesity \[[@B40-ijerph-17-00783]\] and hypertension \[[@B41-ijerph-17-00783]\], and is the most relevant risk factor for the burden of disease and mortality \[[@B39-ijerph-17-00783],[@B42-ijerph-17-00783]\]. A low physical activity level and smoking are the most prevalent health risk factors \[[@B43-ijerph-17-00783]\] and were also negative predictors of enlistment for MS in our study. 4.3. Psychological Distress {#sec4dot3-ijerph-17-00783} --------------------------- Psychological distress was another significant predictor of enlistment into MS in this study. We found by 11% higher prevalence of high psychological distress among those rejected for than among those enlisted into MS. This effect was independent of health-related behaviors such as physical activity, nutrition, cigarette smoking, and alcohol consumption, which might be interrelated to psychological distress given that these behaviors can act as coping strategies \[[@B7-ijerph-17-00783],[@B44-ijerph-17-00783],[@B45-ijerph-17-00783]\]. However, in total, only 9.1% of these male conscripts reported a high psychological distress level. Another study of the Lithuanian high school male student population, from which the military draws its enlistees, reported a 12.5% prevalence of psychological distress measured using the same Kessler scale \[[@B7-ijerph-17-00783]\]. A study in Massachusetts in the USA found that 20% of high school students of both sexes reported depressive symptoms \[[@B46-ijerph-17-00783]\]. An Australian study of university students reported a prevalence of high psychological distress of 11.1%--22.5% \[[@B44-ijerph-17-00783]\]. A prevalence of high psychological distress of 11%--18% was reported for active duty personnel in the US military, and the prevalence was higher among heavy drinkers than among those who consumed less alcohol \[[@B20-ijerph-17-00783]\]. This finding suggests an interdependence of heavy alcohol consumption and psychological distress. This association was identified in the correlational analysis in our study; in particular, of the associations of psychological distress with other health behaviors, the association with alcohol consumption was the strongest. 4.4. Alcohol Consumption and Nutrition {#sec4dot4-ijerph-17-00783} -------------------------------------- Alcohol was consumed by 75.8% of the included youth, and 9.6% were heavy drinkers. These figures are similar to the heavy or risky drinking behavior noted in about 12% of Swiss conscripts \[[@B19-ijerph-17-00783]\] and 7.8% of US active duty personnel \[[@B36-ijerph-17-00783]\]. In a report on healthy lifestyle factors among adult Lithuanians, the Lithuanian Department of Statistics reported that 34% of the population consumes alcoholic beverages at least once a month \[[@B47-ijerph-17-00783]\]. The adherence rates for healthy nutrition among potential conscripts in our study are problematic because only 3.5% fully complied and 46.3% partly complied with the healthy eating recommendations. The Country Health Profiles (2019) reported that 32% of deaths in Lithuania could be attributed to dietary risk factors, a rate that is higher than the 18% in other European countries. Dietary risk has the highest risk for mortality among other health behaviors: smoking is related to 15% and 17%, alcohol consumption to 10% and 6%, and physical inactivity to 5% and 3% of deaths in Lithuania and Europe, respectively \[[@B39-ijerph-17-00783]\]. The regression analysis did not find any associations between enlistment and alcohol consumption or dietary habits in the final model after controlling for all study variables included, although the differences in the percentages were shown significant for alcohol consumption. Significantly more of the recruits rejected for MS (12.3%) were heavy drinkers than those enlisted (6.8%). When entered individually into logistic regression, heavy drinking lowered the chances of being enlisted into MS. However, after the addition of covariates and other study variables, this relationship lost its significance. Moreover, study participants are young men and their history of alcohol consumption may be too short to have a noticeable direct effect on their health. This could also be said about their dietary habits. However, other research has shown that both alcohol use and nutrition over the long term are associated with health outcomes such as the risk of diabetes, cardiovascular disease, Parkinson's disease, certain types of cancer, myocardial infarction, and high blood cholesterol level \[[@B48-ijerph-17-00783],[@B49-ijerph-17-00783],[@B50-ijerph-17-00783],[@B51-ijerph-17-00783]\]. Overall, our results correspond with those of other studies where health-related behaviors and psychological distress are linked to physical comorbidities \[[@B52-ijerph-17-00783]\], which were the main reasons for rejection from MS. Public health policies are targeted at strengthening health in the total population and its subpopulations, such as young people. Similarly, state authorities are also interested in having physically fit and healthy conscripts, who are mostly young men. Improving health among youth is most effective when started early in life during behavior development and when behavioral changes are easier to implement. Health education in an organized way that targets the young population may be best implemented at school, which provides the best setting for delivering public health interventions at a young age. In particular, physical development of school children could be based on physical literacy that addresses attributes, characteristics, skills, and behaviors that are related to the ability for, and commitment to, a healthy, active lifestyle \[[@B53-ijerph-17-00783]\]. From the military perspective, these interventions could also be modeled to capture physical capacity (e.g., physical endurance) and psychological resilience (e.g., resistance to stress) and to emphasize the importance of avoiding substance use or abuse. The development of a healthy future military could start at a young age and involve the objectives of both public health and state defense. To implement this, governmental policies and strategies are required to enable intersectional collaboration and shared responsibility among the education, military and health sectors as well as to set up coordinating mechanisms among sectors. Funding and financial incentives should be provided for intervention programs and implementers. National regulations are needed to enable school administrative and staff to expand their competencies and skills in physical and mental health promotion on a basis of compulsory professional training. 4.5. Study Limitations {#sec4dot5-ijerph-17-00783} ---------------------- The study has some limitations. We used a cross-sectional design, and no causal inferences can be drawn. It remains unclear whether the poor health behaviors were the causes or the consequences of the health outcomes, which were the reasons for rejection from MS. Longitudinal studies are needed to evaluate the direction of, and interactions between, the associations demonstrated here. Another limitation is that we did not include objective measures of physical and mental health. Physical and/or mental health may be confounding variables that both prevent conscripts from serving in the military and increase the likelihood of exhibiting unhealthy behaviors and having a high psychological distress level. Another possible confounding factor may be the interaction between psychological distress and health behaviors because the latter can result from the former. However, the logistic regression analysis allowed us to control for this possible confounding effect. That is, the health behaviors that were significant for enlistment remained significant even after psychological distress levels were included in Model 7. Another limitation is that we did not measure the onset of health-damaging behavior, which may have served as a control variable when examining the predictors of enlistment, especially for addictive behaviors such as smoking and alcohol consumption. In addition, we did not assess the use of other substances such as "party drugs," and we did not obtain a full picture of these conscripts' use patterns. Our study also has some strengths. This was a nationally representative sample that provides a picture of the health behaviors and prevalence of psychological distress among young men called to serve in the military and the differences between those who were enlisted and rejected for MS. Further research should investigate the relationships between health behaviors in conscripts and the reasons they are rejected. 5. Conclusions {#sec5-ijerph-17-00783} ============== Health behaviors in male conscripts are unsatisfactory because about half are physically inactive, have a poor diet, or smoke, and almost one out of 10 is a heavy drinker and has a high psychological distress level. The latter is a concern because of its association with adverse health behaviors. The enlisted conscripts were more likely to be sufficiently physically active and less likely to be current smokers or have a high psychological distress level. Early intervention programs to provide a heathier population of young men for conscription should focus on mental well-being and target health-related behaviors such as physical activity and smoking. Preferably, these should be implemented as health education programs in schools to help prevent the development of adverse health behaviors among young men. Governmental policies and strategies are required to enable intersectional collaboration and shared responsibility among the education, military and health sectors. The authors thank the Lithuanian Ministry of Defense and all centers for military recruitment for administrative support. Conceptualization, B.M. and A.E.; Methodology, B.M., V.J.C., and D.V.; Formal Analysis, B.M. and A.E.; Investigation, D.V., V.J.C., and R.Z.-S.; Resources, A.E.; Data Curation, D.V., V.J.C., and R.Z.-S.; Writing---Original Draft Preparation, B.M.; Writing---Review & Editing, A.E.; Supervision, B.M. All authors have read and agreed to the published version of the manuscript. The study was supported by the Lithuanian Research Council. The authors declare no conflict of interest. The sponsors had no role in the study design, execution, interpretation, or writing of the study. ijerph-17-00783-t001_Table 1 ###### Descriptive statistics and comparison (chi-squared test) between recruits rejected for and enlisted into military service. ----------------------------------------------------------------------------------------------------------------------------- Total\ Rejected for Military Service\ Enlisted Into Military Service\ χ^2^; df; *p* (%) (%) (%) ------------------------- ------------ -------------------------------- --------------------------------- ------------------- MVPA *N = 1219* 31.0; 1; \<0.0001 Low (\<2.5 h/week) 44.7 52.6 **36.9** Adequate (≥2.5 h/week) 55.3 47.4 **63.1** Mediterranean diet *N = 630* 2.3; 2; 0.315 Low adherence 50.2 51.5 49.4 Medium adherence 46.3 44.9 48.0 High adherence 3.5 3.6 2.6 Alcohol consumption *N* = 1082 9.2; 2; 0.010 Abstinent (≥30 days) 24.2 23.4 25.0 Light--moderate drinker 66.2 64.3 68.1 Heavy drinker 9.6 12.3 **6.8 \*** Cigarette smoking *N = 1179* 16.1; 1; \<0.0001 Current nonsmoker 37.7 32.3 **43.0** Current smoker 62.3 67.7 **57.0** Psychological distress *N = 1159* 44.4; 1; \<0.0001 Low 90.9 85.3 **96.6** High 9.1 14.7 **3.4** ----------------------------------------------------------------------------------------------------------------------------- Note: Boldface indicates significant values (*p* \< 0.05); \* adjusted *p* values for 2 × 3 comparisons using the Bonferroni method. df, degrees of freedom; MVPA, moderate to vigorous physical activity. ijerph-17-00783-t002_Table 2 ###### Correlations between the predictors of enlistment for recruits rejected for (above the diagonal) and enlisted into (below the diagonal) military service. MVPA (h/Week) Mediterranean Diet Alcohol (Units/Month) Cigarettes (Number/Day) Psychological Distress ------------------------- --------------- -------------------- ----------------------- ------------------------- ------------------------ MVPA (h/week) -- --0.02 0.01 0.05 --0.07 Mediterranean diet **0.13** -- --0.06 **--0.11** **--0.12** Alcohol (units/month) --0.03 **--0.12** -- **0.22** **0.21** Cigarettes (number/day) --0.01 **--0.18** **0.24** -- **0.12** Psychological distress --0.07 **--0.22** **0.27** **0.12** -- Note: Boldface indicates significant values (*p* \< 0.05). MVPA, moderate to vigorous physical activity. ijerph-17-00783-t003_Table 3 ###### Multiple logistic regression predicting enlistment to military service from health behaviors and distress in Lithuanian male youth, controlling for family income, body mass index, and age (crude and adjusted odds ratios (ORs)). Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 ------------------------------- ----------------------- ------------------------------ ------------------- ----------------------- ----------------------- ---------------------- ----------------------- Predictors Odds ratio (95% CI) Adjusted odds ratio (95% CI) Family income 1.15 (0.93--1.43) 1.09 (0.87--1.36) BMI 0.99 (0.95--1.03) 0.99 (0.95--1.03) Age **0.76 (0.71--0.82)** **0.77 (0.71--0.83)** MVPA (≥2.5 h/week) **1.97 (1.57--2.47)** **1.42 (1.11--2.03)** Mediterranean diet Low adherence Ref. Ref. Medium adherence 1.27 (0.96--1.74) 0.90 (0.63--1.29) High adherence 0.81 (0.34--1.93) 0.83 (0.29--2.39) Alcohol consumption Abstinent (≥30 days) Ref. Ref. Light--moderate drinker 0.52 (0.82--1.49) 1.15 (0.73--1.83) Heavy drinker **0.60 (0.45--0.79)** 0.64 (0.31--1.34) Cigarette smoking Current nonsmoker Ref. Ref. Current smoker **0.61 (0.49--0.78)** **0.58 (0.39--0.86)** Psychological distress (high) **0.20 (0.13-0.33)** **0.26 (0.12--0.55)** Nagelkerke R^2^ 0.06 0.04 0.01 0.02 0.02 0.06 0.16 Note: Boldface indicates significant values (*p* \< 0.05). CI, confidence interval; BMI, body mass index; MVPA, moderate to vigorous physical activity; Ref., reference group.

Background {#Sec1} ========== The prevalence of obesity in the general population has increased dramatically over the last 30 years and it seems likely that the environmental changes that have provoked these increases have also affected people with severe mental illness (SMI); in fact, the rates of overweight and obesity have increased even more rapidly in this cohort \[[@CR1]\]. Obesity adversely affects the physical health and psychological well-being of people with SMI and if weight gain is attributed to treatment, this can lead to non-adherence and risk of relapse. Schizophrenia is a major psychiatric disorder that alters the individual's perception, thoughts, affect and behaviour and may involve a loss of insight and has a lifetime prevalence of approximately 1% \[[@CR2]\]. Schizoaffective disorder is recognised as a separate condition to schizophrenia and is more likely to occur in women at a later age. This disorder affects an individual's thoughts and emotions \[[@CR3]\]. Although individuals with first-episode psychosis do not fulfil the diagnostic criteria for schizophrenia or schizoaffective disorder, 90--95% of people presenting with a non-affective psychotic episode (i.e. not mania and not depressive psychosis) will still meet the criteria for a schizophrenia spectrum disorder 2 years later. Mortality rates are increased two to three fold in people with SMI and life expectancy is reduced by 10--20 years. Approximately 75% of all deaths in people with schizophrenia are caused by physical illness with cardiovascular disease being the commonest cause \[[@CR4]\]. Overweight and obesity contribute to this excess morbidity and mortality. Recent studies indicate that obesity is two to three times more common among people with SMI \[[@CR5]\]. Obesity occurs early in the natural history of schizophrenia with a significant proportion of people with first-episode psychosis being overweight prior to any treatment. Substantial weight gain (\> 7%) often occurs rapidly within 6--8 weeks after antipsychotic-treatment initiation \[[@CR6]\]. While most weight gain occurs early in treatment, longer-term observational studies suggest that weight gain continues for at least 4 years albeit at a slower rate \[[@CR7]\]. Individuals with schizophrenia are more likely to consume a diet that is rich in fat and refined carbohydrates while containing less fibre, fruit and vegetables than the general population \[[@CR8]\]. Although there are fewer studies, people with first-episode psychosis also have poor diets \[[@CR9]\]. Physical inactivity and the social and urban deprivation experienced by those with schizophrenia may contribute further to the increased obesity rates \[[@CR8], [@CR10]\]. There may be disease-specific effects of schizophrenia, such as genetic susceptibility, that have additive or synergistic actions to increase body weight further \[[@CR5]\]. However, the most important factor related to weight gain in people with SMI is the use of antipsychotic medications, which are among the most obesogenic drugs. Weight gain is the commonest side effect of second-generation antipsychotic medication, affecting between 15 and 72% of patients \[[@CR11]\]. Other psychotropic drugs are often prescribed to people with schizophrenia and include some antidepressants and mood-stabilising drugs, such as lithium and sodium valproate; these may also induce significant weight gain \[[@CR12]\]. Both lifestyle and pharmacological interventions lead to significant reductions in body weight in the general population. Not only are the interventions clinically effective, they are also cost-effective because of the benefits of long-term improved health outcomes, including decreased mortality \[[@CR13]\]. It is likely that similarly effective interventions for people with schizophrenia will also lead to improvements in health and would be a major step towards reducing the health inequalities experienced by people with schizophrenia. As the weight gain associated with antipsychotic medication result in some people discontinuing their medication, we hypothesis that effective weight-management strategies may also lead to improved adherence to antipsychotic medication and reduced relapse and hospitalisation rates. Some studies have suggested that short-term lifestyle interventions could support weight reduction in people with SMI. A meta-analysis of non-pharmacological interventions in people with SMI \[[@CR14]\] reported a mean reduction in weight of 3.12 kg over a period of 8--24 weeks. However, the results of longer-term studies are more mixed. A recent meta-analysis found significant weight loss in only two of six studies with interventions lasting longer than a year \[[@CR15]\]. Most studies have included a mixed population of people with SMI and two large studies, which included only people with schizophrenia, found no effect of a lifestyle intervention on body weight \[[@CR16], [@CR17]\]. These latter studies suggest the weight management in people with schizophrenia may require a different approach from other SMIs such as bipolar disorder. Given the challenges of implementing lifestyle change in people with schizophrenia and the lack of long-term effectiveness, alternative approaches are needed to manage overweight and obesity. A wide variety of treatments have been subject to clinical studies but currently no drug treatments are licensed for the treatment of antipsychotic-medication-associated weight gain or obesity in people with SMI with the exception of orlistat \[[@CR18]\]. The long-term use of the latter, however, is extremely limited by high discontinuation rates, making it of little value in routine clinical practice \[[@CR19]\]. To date, there have also been three completed trials of glucagon-like peptide 1 (GLP-1)-receptor agonists in people with SMI, two of which used exenatide and one used liraglutide (maximum dosage 1.8 mg) \[[@CR20]--[@CR22]\]. Liraglutide is a GLP-1-receptor agonist, with 97% homology to human GLP-1, which induces weight loss in humans mainly by reducing appetite and caloric intake, rather than increasing energy expenditure. There were contrasting results in the exenatide studies with one showing no difference between groups after 12 weeks of treatment \[[@CR20]\] but in the other the exenatide arm had greater mean weight loss (− 5.29 vs − 1.12 kg; *P* = 0.015), and reduced glycosylated haemoglobin (HbA~1c~) levels (− 0.21% vs 0.03%; *P* = 0.004) \[[@CR21]\]. In the liraglutide (maximum dosage 1.8 mg) study, glucose tolerance improved in the liraglutide group and body weight decreased compared with placebo (− 5.3 kg; 95% confidence interval (CI) − 7.0 to − 3.7 kg) \[[@CR22]\]. Liraglutide is approved for the management of obesity at a dosage of 3.0 mg daily, which is higher than the dosage used to treat diabetes \[[@CR23]\]. In a 56-week, double-blind trial involving 3731 participants without type-2 diabetes, 63.2% of the intervention arm compared with 27.1% in the placebo arm group lost at least 5% of their body weight, and 33.1% and 10.6%, respectively, lost more than 10% of their body weight \[[@CR23]\]. We have, therefore, chosen to use liraglutide (maximum dosage 3.0 mg) as we postulate that a higher dosage of liraglutide may offer even greater weight loss in people with SMI than the 1.8-mg dosage employed in the previous study or other currently available GLP-1-receptor agonists. Aims and objectives {#Sec2} ------------------- The aim of this pilot study is to undertake a double-blind, randomised controlled trial (RCT) of the use of liraglutide (maximum dosage 3.0 mg daily) in comparison to placebo in 60 obese or overweight people with schizophrenia, schizoaffective disorder or first-episode psychosis to assess the feasibility and acceptability of delivering a full-scale trial evaluating treatment with liraglutide in people with schizophrenia, schizoaffective disorder and first-episode psychosis. Methods {#Sec3} ======= Design {#Sec4} ------ This study is a double-blind, randomised pilot study of the use of liraglutide (maximum dosage 3.0 mg daily) in comparison to placebo (Fig. [1](#Fig1){ref-type="fig"}). It is important to include a double-blind placebo for two main reasons; there is evidence that people would be less likely to consent to a trial that includes a placebo arm because of the risk of not receiving an active treatment. As our ability to recruit to the trial was one of our key aims, it is important to assess whether the inclusion of a placebo prevented us from recruiting to the trial. Previous experience from weight-management trials that include pharmaceuticals have been plagued by high dropout rates in the placebo arm as the participants are able to assess the effectiveness of treatment. To adequately power a full RCT of liraglutide, we will need to know the likely dropout rate in the placebo group in this patient population. Fig. 1Consolidated Standards of Reporting Trials (CONSORT) study diagram Setting {#Sec5} ------- The study will be take place in a variety of community and inpatient mental health locations in the Southern Health NHS Foundation Trust. Public and Patient Involvement was actively included in the development of the trial and will continue throughout the trial. Ethics approval and consent to participate {#Sec6} ------------------------------------------ South Central -- Hampshire B Research Ethics Committee (REC) approved the study on 17 April 2018 with REC reference: 18/SC/0085. Only those who agree to provide written informed consent will be included in the study. The study will be conducted in keeping with Good Clinical Practice (GCP) and the International Conference of Harmonisation (ICH) standards. The Trial Steering Committee (TSC) includes an independent chair and two other independent members, including a service user. The main trial investigators will also attend the TSC. Participants {#Sec7} ------------ Adults are eligible to participate in the study if they: Are aged 18--75 yearsHave a clinical diagnosis of schizophrenia or schizoaffective disorder (defined by *International Classification of Diseases, version 10* (*ICD-10*) codes F20 and F25) or first-episode psychosis using case note review. There is no limit on the duration of illness for those with schizophrenia or schizoaffective disorder but first-episode psychosis is defined as less than 3 years since presentation to the mental health team or the use of the first antipsychotic medication prescriptionAre being treated with an antipsychotic medication, with a minimum duration of 1 month prior to entry in to the trial. No restriction is placed on the class or generation of the antipsychotic medicationAre able to give written informed consentAbility and willingness to take liraglutide or placeboAre able to speak and read EnglishHave a Body Mass Index (BMI, calculated as weight in kilograms divided by height in meters squared) *≥* 30 kg/m^2^ (obese), or *≥* 27 kg/m^2^ to \< 30 kg/m^2^ (overweight) in the presence of at least one weight-related consequence such as dysglycaemia (pre-diabetes or type-2 diabetes), hypertension, dyslipidaemia or obstructive sleep apnoea People are excluded from the study if they have a: Physical illnesses, e.g. cancer, that could seriously reduce their life expectancy or ability to participate in the trialA co-existing physical health problem that would, in the opinion of the principal investigator, independently impact on metabolic measures or weight, e.g. Cushing's syndrome, poorly controlled type-2 diabetes defined by HbA~1c~ level \> 8% (64 mmol/mol)Inflammatory bowel disease and diabetic gastroparesisContraindications to liraglutide: hypersensitivity to liraglutide or to any of the excipients Any condition which, in the investigator's opinion, might jeopardise a participant's safety or their compliance with the protocolFamily or personal history of multiple endocrine neoplasia type-2 or medullary thyroid carcinoma. Family is defined as a first-degreerelativeHistory or presence of pancreatitis (acute or chronic)History of diabetic ketoacidosisAny of the following: myocardial infarction, stroke, hospitalisation for unstable angina or transient ischaemic attack within the past 180 days prior to the day of screeningParticipants presently classified as being in New York Heart Association Class IVPlanned coronary, carotid or peripheral artery revascularisation known on the day of screeningRenal impairment measured as an estimated glomerular filtration rate (eGFR) value of \< 30 ml/min/1.73 m^2^ as defined by the Kidney Disease Improving Global Outcomes (KDIGO) 2012 classificationImpaired liver function, defined as alanine aminotransferase (ALT) ≥ 2.5 times the upper normal limit at screeningProliferative retinopathy or maculopathy requiring acute treatmentPresence or history of malignant neoplasms within the past 5 years prior to the day of screening. Basal and squamous-cell skin cancer and any carcinoma in situ is allowedUse of other pharmacological products for weight managementMental illnesses that could seriously reduce their ability to participate in the trial, including significant suicidalityCurrent pregnancy or a desire to become pregnant. Mothers who are less than 6 months post-partum or breastfeeding will also be excluded. In line with the current EU licence and advice from the Medicines and Healthcare products Regulatory Agency (MHRA), sponsor and manufacturer, any women who may become pregnant during the trial but are unwilling to use a highly effective method of birth control (e.g. such as implants, injectables, combined oral contraceptives, intrauterine devices, sexual abstinence or having a vasectomised partner) will not be eligible for the trialSignificant alcohol or substance misuse which, in the opinion of the principal investigator, would limit a participant's ability to participate in the trialA diagnosis or tentative diagnosis of psychotic depression or maniaA primary diagnosis of learning disability or cognitive impairment which would impair a participant's ability to self-administer trial medicationLack of capacity. Those who lose capacity any time during the study will not be eligible to continue and will be withdrawn from the study immediately with no further study procedures carried outHistory of type-1 diabetes.Current or previous use of incretin-based therapies (GLP-1-receptor agonist or dipeptidyl peptidase 4 (DPP-4) inhibitors) or insulin Sample size {#Sec8} ----------- This pilot trial will explore the feasibility and practical issues of conducting a future definitive trial and estimate the important parameters to help its design. In this regard, sample size is based on the need to estimate study parameters within a reasonable degree of precision rather than on hypothesis testing. Simulation work by Sim et al. (2012) recommended a minimum of 50 participants (25 per group) in order to achieve the pilot/feasibility objectives \[[@CR24]\]. A further paper by Whitehead et al. recommended pilot trial sample sizes per treatment arm of 75, 25, 15 and 10 for standardised effect sizes that are extra small (≤ 0.1), small (0.2), medium (0.5) or large (0.8), respectively \[[@CR25]\]. In the previous liraglutide study in people with SMI, 10% of the liraglutide arm and 2% in the placebo arm had dropped out of the trial after 16 weeks \[[@CR22]\]. Although dropout from obesity trials is non-linear as those on active treatment drop out earlier because of side effects while those on placebo drop out later because of lack of efficacy, we assume a conservative dropout rate at 6 months of between 15 and 20%. As such, we will need to recruit at least 60 participants (30 per group) to provide robust estimates that will inform the design of the definitive trial. In a pilot trial looking at the use of liraglutide (maximum dosage 1.8 mg) of 214 potential participants assessed for eligibility, 103 were randomised. Of the 111 excluded, 86 did not meet the final inclusion/exclusion criteria, 23 declined to participate and 2 had too severe a degree of mental illness to participate \[[@CR22]\]. However, in a similar study examining the use of once-daily exenatide in people with schizophrenia, out of 123 potentially eligible participants, only 28 were randomised with 63 declining to participate \[[@CR21]\]. We used these data to estimate our screen-to-randomisation rate. Selection {#Sec9} --------- The study will be promoted within clinical teams and in areas where community mental health services are delivered in the Southern Health NHS Foundation Trust. It is estimated that \~ 70% of those with a recorded diagnosis of schizophrenia or one of its subtypes attending community and inpatient settings of the Southern Health NHS Foundation Trust fulfil the inclusion criteria. This equates to approximately 500 individuals across the trust. Assuming that 30% of patients were willing to take part, we would be able to recruit up to 60 patients for the study within 12 months. Confirming these screening and enrolment rates is part of the reason for performing the study. In addition, the Early Intervention in Psychosis team in Southampton receives approximately 20 new referrals a month. Although these individuals do not fulfil the diagnostic criteria for schizophrenia, we intend to include them in the study for the following reasons: Up to 80% of individuals treated with antipsychotic medication during a first episode of psychosis gain more than 7% of their body weight within 12 weeks of treatment.People with first-episode psychosis are more likely to develop weight-induced metabolic abnormalities and consequently the benefits may be greater Participants will be advised verbally and in writing that they will be able to end their participation in the study at any point without affecting their clinical care. The investigators will also have the right to withdraw participants from the study if they lose capacity, develop any of the exclusion criteria or miss their 3-month study visit. The rationale for the withdrawal will be recorded and dated in the Case Report Form (CRF) accordingly. Participants withdrawing from the trial treatment will be encouraged to undergo the same final clinical evaluations. Randomisation {#Sec10} ------------- After baseline assessments, participants will be randomised to either daily subcutaneously administered liraglutide (maximum dosage 3.0 mg) or matching placebo. Equal numbers of participants will be randomised to each arm of the trial using simple randomisation with permuted, blinded, block size. Novo Nordisk prepares and provides the subject randomisation list (SRL) using a computer-based programme. All participants, carers and study personnel except the pharmacy team will be blinded to treatment assignment. This includes the statistician undertaking the data analysis. Emergency un-blinding will be undertaken if a participant develops an adverse event that requires knowledge of the treatment, an overdose of trial medication or there is a clinical need to start a participant on medication which has a risk of interaction with the trial drug. Intervention {#Sec11} ------------ Before screening, all potential participants will be provided with written information about the trial including the most common adverse event and the procedures involved in the study. All participants will receive standardised written information about healthy eating, physical activity, alcohol and smoking. Liraglutide will be used according to the current EU licence for Saxenda®; the starting dose will be 0.6 mg per day. Participants will be taught face-to-face about using the injection pen and will be witnessed giving the first injection. Participants will be given an instruction leaflet to take away with them. The dose will be increased each week by 0.6 mg to a maximum dosage of 3.0 mg per day. Participants who do not tolerate up-titration will remain on the highest tolerable dosage. Each participant will attend 4-weekly visits where concomitant medications and adverse events will be documented. In addition at the baseline, 3- and 6-month visits they will also have clinical data collected (secondary endpoints) including drawing blood samples. The blood samples will be analysed for fasting plasma glucose (FPG), lipid profile and HbA~1c~ level~.~ Participants will be invited at baseline and study completion to take part in one-to-one interviews with a psychologist, trained and experienced in qualitative research methodology, to explore expectations and experience of their participation in the trial. Outcome measures {#Sec12} ---------------- ### Primary objective {#Sec13} The primary objective of the pilot trial is to investigate the feasibility and acceptability of delivering a full-scale trial evaluating whether liraglutide 3.0 mg, once-daily injectable therapy, may be an effective treatment of overweight and obesity in people with schizophrenia, schizoaffective disorder and first-episode psychosis. In order to achieve our primary objective, the study will gather data on: Time to reach the recruitment targetThe number of eligible participants required to be screened in order to reach the recruitment target. Key characteristics and reasons for not joining the trial will be recorded, in line with the Consolidated Standards of Reporting Trials (CONSORT) criteria for clinical trialsTo estimate participant attrition rateTo estimate adherence to the investigational medicinal product (IMP) ### Secondary exploratory outcomes {#Sec14} To estimate effect size and standard deviation (SD) of the change in weight at 6 months in order to inform a power calculations for a fully powered RCT based on this feasibility pilot study. Changes in waist circumference, BMI, FPG, HbA~1c~ level, blood pressure, lipid profile, adverse events and Brief Psychiatric Rating Scale (BPRS) score at 3 and 6 months will also be assessed.. The BPRS is an instrument used for assessing the positive, negative and affective symptoms of psychotic disorders, especially schizophrenia. The BPRS will, therefore, be used to assess any changes in mental health during the trial. Follow-up windows for 3- and 6-month follow-ups will be defined as minus and plus 2 weeks to allow for missed appointments. A schedule of follow-up activities is shown in Fig. [2](#Fig2){ref-type="fig"}. Fig. 2Schedule of outcome measures and trial-related activities Safety assessments {#Sec15} ------------------ Heart rate and glycaemia assessments will be taken at baseline and at 3 and 6 months post randomisation. Adverse events (AEs) are defined as the side effects listed in the Summary of Product Characteristics (SmPC) and include nausea, vomiting and diarrhoea. Medication error and laboratory outliers will also be considered as AEs. AEs will be monitored every 4 weeks. Serious Adverse Events (SAEs) are defined as per GCP. Any SAE which a member of the study team deems to be associated with the trial intervention will be assessed by the principal investigator. There are SAEs that are expected for the patient population: Psychiatric hospitalisationWorsening of psychiatric symptomsSelf-harmSuicide attemptDeath from suicide If the investigator deems that any of these expected events are at least possibly related to the study drug then these will be reported as serious adverse reactions (and not events). All SAEs that are both 'unexpected' (that is, the type of event is not listed in the protocol as an expected occurrence); and 'related' (that is, it resulted from administration of any of the research procedures) will be reported to the sponsor for expedited reporting to the TSC and the REC. No serious adverse outcomes are anticipated associated with use (or not) of the trial medication; therefore, no interim statistical analysis is planned regarding safety; however, the TSC will review all SAEs periodically. Data analysis {#Sec16} ------------- Data will be analysed in accordance with the trial's detailed Statistical Analysis Plan (SAP) (Additional files [1](#MOESM1){ref-type="media"} and [2](#MOESM2){ref-type="media"}), a summary of the main methods are as follows: All data will be analysed based on the intention-to-treat population. Continuous variables will be analysed by either mean or median, with groups compared statistically using either a paired *t* test or a Mann-Witney *U* test as appropriate. Categorical variables will be presented as *n* (%) with groups compared using chi-squared or Fisher's exact tests. The difference in weight change between groups will be analysed using generalised linear models (GLMs), both unadjusted and adjusted for covariates that are identified as potential confounders in univariate testing. All statistical tests will be two-sided with statistical significance assumed at 0.05. Key characteristics and reasons for not joining the trial will be recorded for all participants screened. A CONSORT diagram will be used to summarise this information and reasons will also be provided if applicable. Demographics and person-reported outcome data at baseline will be summarised using number and proportion, mean and SD or median and interquartile range (IQR) as appropriate. ### Analysis of primary objectives {#Sec17} This study focusses on the feasibility of recruiting from this patient population for a fully powered RCT. The following will be reported: Time to reach recruitment target (in weeks) ◦ Mean number of participants recruited per week◦ Rate of successful screens◦ Participant attrition rateAdherence to the IMP (defined as the proportion of medication used by each person ranging from 0 to 100%) ◦ Either mean (SD) or median (IQR) adherence will be presented as appropriate◦ Number of participants using at least 70% of the prescribed trial medication. Adherence to the IMP is defined as the number of empty cartridges returned at each visit divided by the total number of cartridges prescribed ### Analysis of secondary exploratory outcomes {#Sec18} Changes in weight (defined as weight in kilograms at 3 or 6 months minus weight in kilograms at randomisation), BMI, waist circumference, BPRS score, HbA1c level, FPG, lipid level, systolic and diastolic blood pressure, and adherence to randomised treatment (including the effect of the using the optional text-messaging reminder service or not), type of diabetes medication, change in type or dose of diabetes medication, type of antipsychotic medication, change in type or dose of antipsychotic medication between the two treatment groups will be summarised and tested for significance, as will the number of participants experiencing a weight loss of at least 5% from baseline to 3 to 6 months. Change in body weight between the two groups at 26 weeks will be further analysed using a GLM adjusted for baseline weight and other covariates. ### Missing data {#Sec19} Analysis will be completed using list-wise deletion of missing data. Participants with and without missing data will be compared for differences in demographic and physiological data where possible. ### Harms {#Sec20} The number (and percentage) of participants experiencing each AE/SAE will be presented for each treatment arm categorised by severity. Qualitative component of study {#Sec21} ------------------------------ One-to-one semi-structured interviews will be held with a sub-sample of liraglutide-treated participants and healthcare professionals delivering the intervention to provide data on the drug treatment. Purposive sampling will ensure diversity in terms of demographic and disease characteristics. Participants (10--12 from each trial arm) will be interviewed until data saturation is met. Semi-structured interview-topic guides will contain questions intended to elicit themes outlined in the existing published literature, after exploring more general open questions on the experience and acceptability of the treatments. Interviews will be held before, during and at the end of the trial. For mental health workers, key themes relate to the perception of pharmacological interventions as part of the care-coordinating role, workload, the need for specialist knowledge and views on client adherence. We will use May's normalisation process model as a theoretical framework to understand the conditions necessary to support the introduction, embedding and integration of a weight intervention as a routine element of care. All semi-structured interviews will be audio-taped and fully transcribed. Content and thematic analysis will use the National Centre for Social Research 'Framework' approach. Trial status {#Sec22} ============ Recruitment opened on 2 July 2018 and we aim to complete recruitment by 31 July 2019. Protocol version 1.6, dated 22 May 2018, is being used. Discussion {#Sec23} ========== Obesity adversely affects the physical health and psychological well-being of people with SMI. If weight gain is attributed to treatment, this can lead to non-adherence and risk of relapse. The provision of an effective intervention to reduce the burden of overweight and obesity in people with schizophrenia, schizoaffective disorder and first-episode psychosis would improve physical health and reduce the risk of developing obesity-related illnesses as well as improving psychological well-being. This pilot study is a double-blind, randomised, placebo-controlled trial investigating the use of once-daily liraglutide subcutaneous injection in obese or overweight people with schizophrenia, schizoaffective disorder or first-episode psychosis. It aims to explore the feasibility and practical issues of conducting a future definitive RCT evaluating weight change with liraglutide in overweight or obese people with SMI. This feasibility trial should estimate important parameters to help its design. One potential limitation of this feasibility trial is the funding by the investigational drug manufacturer. In order to mitigate against this bias, the trial was sponsored by the Southern Health NHS Foundation Trust, which has the responsibility for the initiation, management, conduct, analysis, reporting and publication of the trial. Although Novo Nordisk is providing support financially and the product for the trial, Novo Nordisk is not involved in the conduct, management and delivery of the trial. Additionally, the initial idea, rationale and design for the trial came from the chief investigator. Nevertheless, the results of this study will need to be confirmed in a fully powered, investigator-led trial. This research should contribute to the development of effective weight-management intervention programmes for people with SMI. It should provide information on whether injectable GLP-1-receptor agonists are an acceptable weight-loss medication in this group of people. Further, potentially more effective, GLP-1-based medications are in development including once weekly and orally administered versions which may prove to be a better option for people with SMI if once daily injections are shown to be feasible. . Additional files ================ {#Sec24} Additional file 1:Statistical Analysis Plan (SAP). L. O. S. E. Weight Pilot Study. (DOCX 39 kb) Additional file 2:Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) 2013 Checklist: recommended items to address in a clinical trial protocol and related documents. (DOCX 121 kb) AE : Adverse event BMI : Body Mass Index BPRS : Brief Psychiatric Rating Scale CI : Confidence interval CRF : Case Report Form eGFR : Estimated glomerular filtration rate FPG : Fasting plasma glucose GCP : Good Clinical Practice GLM : Generalised linear model GLP-1 : Glucagon-like peptide 1 ICH : International Conference on Harmonisation RCT : Randomised controlled trial REC : Research Ethics Council SAE : Serious adverse event SD : Standard deviation SmPC : Summary of Product Characteristics TSC : Trail Steering Committee **Publisher's Note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Not applicable. RIGH, HCP, SR, PP, KT, KBK and CAW wrote the original protocol. RIGH, HCP, SR, PP, KBK and CAW were co-applicants on the investigator-led Novo Nordisk grant application. CAW, HCP, SR, PP, KBK, KT, CR, RP, JM and RIGH refined the protocol. CA reviewed the protocol from a Patient and Public Involvement (PPI) point of view. All authors critically reviewed the protocol manuscript. All authors read and approved the final manuscript. The study team was awarded an investigator-led grant by Novo Nordisk Ltd. The protocol was designed by the investigators and the funder will have no role in the analysis of the results. Sponsor: Southern Health NHS Foundation Trust. Not applicable. South Central -- Hampshire B Research Ethics Committee (REC) approved the study on 17 April 2018 with REC reference: [@CR18]/SC/0085. Only those who agree to provide written informed consent will be included in the study. Not applicable. CAW, HCP, PP, SR and KT are employees of the Southern Health NHS Foundation Trust. HCP received fees for lecturing, consultancy work and attendance at conferences from Novo Nordisk. SR received fees for speaking at conferences from Janssen, Lundbeck and Otsuka. KBK received fees for advisory board participation, consultancy work and attendance at conferences from the following: Sanofi, Roche Diabetes Care, Lifescan, Novo Nordisk, Silvercloud, and Senseonics. RIGH received fees for lecturing, consultancy work and attendance at conferences from the following: Boehringer Ingelheim, Eli Lilly, Janssen, Lundbeck, Novo Nordisk, Novartis, Otsuka, Sanofi, Sunovion, Takeda, and MSD.

Introduction {#s1} ============ Traditionally, ecological and eco-evolutionary epidemiological models describe the dynamics of infectious diseases by considering susceptible, infected and recovered hosts with host-to-host, or host-environment-host transmission [@pone.0071621-Kermack1]--[@pone.0071621-Hudson1]. A number of modifications--such as seasonality [@pone.0071621-Altizer1], or within-host dynamics [@pone.0071621-Mideo1]--have been introduced to the SI- and SIR-models in various attempts to explain recurrent outbreak disease dynamics. However, natural epidemics often show a variety of dynamics that do not correspond to the predictions made by the classical models. One reason for this is that the underlying assumptions on disease transmission are unrealistic for pathogens that spend a considerable amount, or even the most part of their life cycle, in the outside-host environment. A large proportion of opportunist pathogen species also grow actively in the outside-host environment. These environmental pathogens form an increasing problem for human health [@pone.0071621-Cangelosi1], and thus a better theoretical understanding of their epidemiology is required. Currently, most models for environmental transmission allow only decay of pathogens in the outside-host environment [@pone.0071621-Codeco1], [@pone.0071621-Day1], but see [@pone.0071621-Merikanto1]. In addition, the outside-host environment is riddled with other microbes which frequently interact with the pathogen. This means that while active growth as well as ecological interactions in the environment are likely to be profoundly important, they are yet poorly understood factors in disease dynamics. The role of environmental transmission in disease dynamics and the evolution of virulence has attracted increasing interest [@pone.0071621-Day1], [@pone.0071621-Roche1], both due to human pathogen outbreaks such as cholera [@pone.0071621-Colwell1] and emergent animal diseases, e.g. columnaris disease [@pone.0071621-Pulkkinen1]. Worldwide, there is an ongoing battle against opportunistic infections, which are often persistent due to the pathogens' ability to grow outside hosts. Broad-spectrum antibiotics and disinfectants are used *en masse* to prevent environmental infections to humans and cultivated animals. This is likely to cause changes in the composition of environmental communities and have an impact on ecosystem functioning, health and disease [@pone.0071621-Ding1]. Theoretically, an environmentally transmitted pathogen can be highly lethal as the trade-offs between transmission and virulence associated with obligate pathogens are reduced [@pone.0071621-Walther1]. This is because by killing a host--which is not required to be alive for pathogen transmission--the pathogen gains access to an enormously rich resource for saprotrophic growth, which can lead to a positive transmission-virulence relationship [@pone.0071621-Kunttu1]. In most studies, however, the environment represents simply a reservoir into which the pathogen particles are shed from infected hosts and from which the surviving pathogen individuals may re-enter susceptible hosts, without explicit description of the outside-host dynamics other than the decay rate of the pathogen. Interspecific interactions, such as competition, mutualism, predation, and parasitism constitute the core of ecological research. An important implication of these interactions is that the dynamics and stability of individual populations within ecological networks (e.g., communities or food webs) can strongly depend on the composition of these networks and the details of between-species interactions [@pone.0071621-DeRuiter1]--[@pone.0071621-Pimm1]. In general, similar co-occurring species compete for limiting resources [@pone.0071621-Hibbing1] and are attacked by parasites and predators [@pone.0071621-Abedon1]. All of these different ecological interactions can affect the density of pathogens and other interacting species in the community, thereby affecting the probabilities of infection outbreaks. Therefore, understanding the role of ecological interactions in the outside-host environment is likely to be of great importance for uncovering mechanisms behind the dynamics of many environmentally transmitted diseases such as *Vibrio cholera* [@pone.0071621-deMagny1], group A *streptococci* [@pone.0071621-Beres1], *Staphylococcus aureus* [@pone.0071621-Eveillard1], and *Flavobacterium columnare* [@pone.0071621-Pulkkinen1]. We explore the dynamics of a model that combines environmental opportunist pathogen--host dynamics to community dynamics outside the host. By the term environmental opportunist pathogen we mean an organism that is both (i) able to grow in the environment in the absence of hosts and (ii) infect susceptible hosts. Whether the pathogen can infect one or several host species is not important in this simple model in which we consider all (susceptible) hosts similar from the pathogens perspective. The model contains susceptible and infected hosts as in a classical SI-model and a competitive community in the outside-host environment. One of the competitors is a pathogen that can return from a dead host to the environment and thus the host does not represent an ecological or evolutionary dead end for the pathogen. This is opposite to the common assumption of the theory of "co-incidental virulence" [@pone.0071621-Brown1]. Further, we assume that there is a trade-off between virulence and environmental competitive ability. This assumption also differs from the expectation under co-incidental virulence theory [@pone.0071621-Brown1], where resource acquisition or fighting against natural enemies outside the host are positively linked to a pathogens ability to cause infections. Life-history trade-offs can reduce virulence because the machineries for resource acquisition and defence in the outside- vs. inside-host environments require specialisation [@pone.0071621-Gower1], [@pone.0071621-Mikonranta1]. The choice of the functional form of the infectivity response can crucially affect the model dynamics [@pone.0071621-Boldin1], [@pone.0071621-McCallum1]. We explore both linear and sigmoidal infection rate in response to pathogen density, and assume that the infected hosts can either recover back to the susceptible class or die from the disease. The sigmoidal infectivity response incorporates dose-dependence, i.e., exposure to pathogen densities below a certain level is unlikely to cause an infection, as observed in many laboratory experiments [@pone.0071621-Aaby1]--[@pone.0071621-McLean1]. The model behavior is explored in a parameter range that is likely to cover typical environmentally growing opportunist micro-parasites (e.g., bacteria, protozoa, or fungi) that infect multicellular hosts ranging from taxa with fast growth rates (e.g., nematodes and insects) to taxa with slow growth rates (e.g., vertebrates). A striking feature of the model is that reducing competitor species richness in the outside-host environment can lead to an abrupt emergence of disease outbreaks. If the infectivity response of the pathogen is a sigmoidal function of pathogen density in the environment, epidemiological dynamics are sensitive to the intensity of competitive suppression by the outside-host community. With linear infectivity response and pathogen growth in the environment no disease outbreaks are observed, i.e. population dynamics remain stable, and pathogen and host densities are relatively insensitive to manipulation of diversity. Under sigmoidal infectivity, reduction of competitive pressure on the pathogen, either due to loss of diversity from the outside-host community (e.g., due to use of disinfectants), or increased loss rate of all species in the outside-host community (e.g., due to application of non-specific antibiotics) can lead to catastrophic disease outbreaks. Methods {#s2} ======= The dynamical model is a combination of the classical SI-disease dynamics [@pone.0071621-Kermack1] for hosts and the Lotka--Volterra competition model for an outside-host community. The susceptible and infected hosts at time *t* are denoted by *S*(*t*) and *I*(*t*), respectively. The pathogen, *P*(*t*), has *n* competitors, with densities denoted by *B~i~*(*t*), i.e., community size is *N* = *n* +1. The population densities vary according to the following differential equations: In the absence of an infection, susceptible hosts follow a logistic growth model with a growth rate *r~h~* and a carrying capacity *K~h~* (eqn. 1.1). In the presence of an infective pathogen infected host individuals are formed. The infected individuals compete for resources with the susceptible individuals, but do not contribute to host reproduction (this assumption has no qualitative effect on the dynamics). Susceptible hosts are infected with a rate *βSf*(*P*) depending on the infectivity response *f*(*P*). Two alternative infectivity response functions were explored. Most previous theoretical work has assumed a linear infectivity response. However, we argue that a more realistic assumption for many circumstances is a sigmoidal dose-dependent response that is supported by empirical data [@pone.0071621-Aaby1]--[@pone.0071621-McLean1] as well as theoretical analysis [@pone.0071621-Pujol1]. The sigmoidal infectivity function has the following mechanistic interpretation. With low pathogen densities the immune system can effectively overcome most pathogen invasions and therefore the probability of an infection per unit time must increase nonlinearly at low densities. With high pathogen densities the effect of increasing pathogen density on the probability of infection per time unit must saturate since the development of an infection is not an instantaneous process. This type of functional response has been studied by Regoes et al. [@pone.0071621-Regoes1] for direct transmission in the context of the classical SIR-model. Mechanisms behind sigmoidal infectivity response can also include saturation of the immune system with a large number of invaders, density-dependent accumulation of enzymes that allow breaching the immune system or expression of virulence factors due to density-dependent bacterial communication, i.e., quorum sensing [@pone.0071621-Kumar1]. A Hill function ([eqn. 2](#pone.0071621.e005){ref-type="disp-formula"}) was chosen as the functional form for the sigmoid infectivity response *f*(*P*), because it is a simple way to conceive this type of dose-dependence for infections [@pone.0071621-Regoes1]:where parameters *ID* ~50~ and *κ* affects the shape of the response. [Eqn. 2](#pone.0071621.e005){ref-type="disp-formula"} is used as a convenience function with no particular mechanistic underpinning. A classical linear term with *f*(*P*)* = P* was used for comparison. In either case, the infectivity rate is scaled by parameter *β* in eqn (1.1). In eqn. (1.2) infected hosts die from the disease with rate *ν* and recover from the infection with rate *δ*. With the parameter set in [Table 1](#pone-0071621-t001){ref-type="table"}, every 3/7:th infection leads to host death. In case of recovery the pathogens inside a host are killed by the host immune system. 10.1371/journal.pone.0071621.t001 ###### Model parameters and values used in the model. {#pone-0071621-t001-1} Parameter Interpretation Values used in simulations ----------- ----------------------------------------------------- ---------------------------- *n* Number of competitors in the environment 0--17 *r~h~* Host growth rate 0.1 *r~p~* Pathogen growth rate 2.0 *r~b~* Competitor growth rate 2.0 *K~h~* Host carrying capacity 100 *K~p~* Outside host pathogen carrying capacity 10000 *K~b~* Outside host community carrying capacity 10000 *ID* ~50~ Infectious dose at which 50% of hosts are infected. 3150 *κ* Slope parameter of the sigmoid infectivity function 4 *β* Maximum infectivity 4 *δ* Host recovery rate from infection 0.6 *ν* Infection kill rate 0.1 Between species competition strength 0.5 *α* Competitive ability reduction 0.05, 0--0.15 *k* Number of pathogens released at host death 500 *η* Pathogen mortality 1.0, 0--2.5 The parameters are chosen to represent a realistic scenario with a bacterial pathogen and a small multi-cellular host (see text). The pathogen (*P*) and the competing non-pathogenic strains (*B~i~*) grow logistically with rates *r~p~* and *r~b~*. Carrying capacities are *K~p~* and *K~b~* in a Lotka--Volterra competition setting, respectively (for simplicity, *r~p~* = *r~b~*, and *K~p~* = *K~b~*). The competition in the outside-host community was modeled with a very general diffuse competition model where the competitive ability of non-pathogens against the pathogen was varied. The intraspecific competition coefficients were set to 1, whereas between species interaction strength is given by (eqns. 1.3, 1.4; here set  = 0.5). It is assumed that the pathogen pays a cost of its ability to cause infections (e.g., due to extra biochemical machinery) in form of reduced competitive ability. This is realised as competition against the pathogen equal to +α, and for the pathogen in competition with other species -- α, with α being the reduction in pathogen competitive ability. A diffuse competition model is used as a simplification in this model. This is justified since generating between species coefficients randomly with a mean produces, on average, the same dynamics. In the case of positive correlation between virulence and competitive ability, i.e. negative α, a trivial result in our model is that the host is driven to extinction by the pathogen. A mortality term with rate *η* was included for scenarios where growth is not only self-limited but there is another density-independent mortality factor that removes bacteria from the system (e.g., antibacterial substances, or physical outflow from the system). If hosts are sparse, intense outside-host competition in addition to out-flow mortality *η* can drive the pathogen extinct while waiting for the next infection. The effect competition has on the pathogen species could be realised either through reduction in pathogen competitive ability (α) or increasing the number of competitors present in the outside host community (*n*). The equilibrium pathogen density reaches zero with *α = *1/(2*n*) in the absence of the host, due to competition. The survival of the pathogen beyond this level of competition is still possible via coupling to host dynamics due to fitness benefits gained by inside-hosts growth. On the other hand if the pathogen is very infective and lethal, it may kill all the suitable hosts and thereafter become extinct by competitive exclusion. This outcome resembles that of the classical trade-off between virulence and transmission, albeit via an entirely different mechanism. Equilibrium densities for the pathogen and non-pathogenic species in the absence of hosts are: where *r* is the common growth rate for all species. Note that *P\** is only positive when and *r*\>*η.* The model assumes that a burst of pathogens is released to the environment with the death of an infected host. This identifies to saprotrophy where the dead host body is consumed to some degree. As the dead host typically represents an extremely rich resource in comparison to the typical outside-host environment [@pone.0071621-Kunttu1], the number of pathogens released from the host can be extremely high. Thus the large flow of pathogens to the environment from the host can in turn lead to a rapid cascade of infections and host deaths that can ultimately result in host extinction [@pone.0071621-Godfray1]. The parameter ranges used in the simulations are given in [Table 1](#pone-0071621-t001){ref-type="table"}. The pathogen growth rate of two divisions per day represents the lower end of bacterial growth rates. The infected hosts remain infected for a relatively long period since both infection kill rate and recovery rate are low. This is plausible e.g. for an untreated bacterial disease many of which are very persistent. The host growth rate is suitable for a fish host [@pone.0071621-Merikanto1]. Outside host community densities are scaled down to make the numbers comparable in order of magnitude to that of hosts. Multiplying outside host carrying capacity and the number of pathogen units released from a dead host by 10^6^ leads to a scenario where the unit of area could be a cubic metre of water with 100 small fish hosts and 10^10^ bacterial cells. The infectivity parameter values used here allow for effective infecting without driving the host extinct too easily. Extensive simulations with randomly selected parameter values indicate that the results presented here are qualitatively robust ([Materials S1](#pone.0071621.s004){ref-type="supplementary-material"}, [Fig. S2](#pone.0071621.s002){ref-type="supplementary-material"}, [S3](#pone.0071621.s003){ref-type="supplementary-material"}). The model (eq. 1) behaviour was analysed numerically using a Runge-Kutta fourth order routine. Means, maxima, and standard deviations of population densities were recorded from the final 100 time units from each simulation in order to assess the type of the dynamics (e.g. outbreaks). The analysis focused on the asymptotic behavior of the model and excluded the initial transients. Trajectories were simulated for 200 time units that was a sufficiently long time interval for analysing the asymptotic behaviour of the model in all cases. The initial state in all simulations was *S = *100, *I = *0, *P* and *B~i~* chosen from uniform random distribution between 100 and 1600. Results {#s3} ======= The choice of the infectivity response function is crucial to the behaviour of the model. Therefore the following results are presented according to this dichotomy between linear and sigmoidal infectivity responses. A common assumption in epidemiological models is that there is a linear relationship between pathogen density and the number of infections (i.e., *f*(*P*) = *P*, [Fig. 1](#pone-0071621-g001){ref-type="fig"}). The linear response is a benefit to the pathogen since it remains infective even in small doses ([Fig. 2a](#pone-0071621-g002){ref-type="fig"}). However, the host is easily driven to extinction unless its growth is sufficiently fast. Under this assumption, increasing outside-host community size is associated with an initial reduction in the density of free-living pathogen, as expected from classical competition theory ([Fig. 3a](#pone-0071621-g003){ref-type="fig"}, see e.g., May 1972). Increasing competition in the outside-host community quickly drives the pathogen extinct in the absence of hosts. However, the pathogens ability to infect hosts and use them as resources for reproduction compensates for the reduced competitive ability in the outside-host environment ([Fig. 3a](#pone-0071621-g003){ref-type="fig"}). Pathogens persist at a stable density, which is relatively independent of the number of competitors. With the linear response cyclic pathogen dynamics were only observed when pathogens did not grow in the environment (*r~p~*\< *η*) and that there was no significant recovery of infected hosts. Linear response stabilises the dynamics because it prevents host supply re-growth by making the pathogen efficient in infecting also when the pathogen density is low. While pathogen infections reduce the number of susceptible hosts well below their carrying capacity ([Fig. 3b](#pone-0071621-g003){ref-type="fig"}), competition in the outside-host environment can prevent the pathogen from driving the host extinct. {#pone-0071621-g001} {#pone-0071621-g002} {#pone-0071621-g003} A pathogen with a sigmoidal transmission differs from the linear transmission in its response to increasing competition pressure in three important ways ([Fig. 2](#pone-0071621-g002){ref-type="fig"}, [3c, 3d](#pone-0071621-g003){ref-type="fig"}): (1) Near the point where pathogen growth rate in the absence of hosts approaches zero due to competitive exclusion, increasing outside-host community size gives rise to cyclic pathogen dynamics. (2) After this bifurcation an alternative attractor appears where the pathogen is excluded from the system. This happens because of the Allee effect associated with the sigmoidal dose-response function [@pone.0071621-Regoes1]; if the pathogen is initiated at a sufficiently low density, it is unable to infect susceptible hosts, preventing fitness gains through within-host growth. Also when the initial pathogen density is too high there is a rapid increase in density followed by a drop to very low density after which the pathogen density stays close to zero. (3) Increasing outside-host community size further amplifies the cyclic dynamics on the attractor where the pathogen is present. This continues up to a point where competitive pressure is sufficiently high to prevent pathogen infections independently of initial conditions, leading to an abrupt disappearance of the pathogen from the system. The range of cyclic dynamics depends on host growth rate ([Fig. 2B](#pone-0071621-g002){ref-type="fig"}). Outbreaks arise when the time scale of pathogen release from hosts is comparable to that of host growth rate. If pathogen release happens much faster than hosts grow or recover then no hosts are available for the released pathogen making further cycling impossible. The host growth rate of 0.1 per day used in our simulations is quite high for most large multi-cellular organisms. To represent diseases of slowly growing hosts the infectivity and host-pathogen interaction parameters need to be scaled accordingly to retain the same range of qualitative dynamics. For example cyclic dynamics can be retained at lower host growth rates by lowering infected host recovery and death rates at the same time. Similar patterns to those shown in [Fig. 3](#pone-0071621-g003){ref-type="fig"}, due to increasing size of the outside-host community *N*, can be generated by varying the competitive disadvantage of the pathogen, α, for this parameter set ([Fig. S1](#pone.0071621.s001){ref-type="supplementary-material"}). This is easy to understand, as increasing either the number of competitors (*n*) or competitive disadvantage (α) decreases the equilibrium density of the pathogen (eqn. 3). Under sigmoidal transmission this means that when the pathogen is unable to survive in the outside-host environment without fitness gains from host infection, the presence of susceptible hosts can be associated with two alternative attractors, with pathogens either present or absent. When competition is intensive enough, pathogens are unable to reach sufficiently high densities for infections to arise, inevitably leading to pathogen extinction. Increasing mortality (*η*) in outside-host environment reduces the competitive pressure from the rest of the outside-host community on the pathogen. When this density-independent mortality is intermediate, the pathogen can thrive in environments where competition would otherwise drive it to extinction ([Fig. 4c](#pone-0071621-g004){ref-type="fig"}). The reason for this is that increasing *η* reduces the density-dependent negative feedback from competitors to the pathogen, leading to better utilisation of the fitness benefits from within-host reproduction. This effect is not seen when the infectivity response is linear ([Fig. 4a](#pone-0071621-g004){ref-type="fig"}). {#pone-0071621-g004} If the mortality term is taken as non-specific antibiotic treatment, the effect of increasing mortality can depend on initial conditions with the sigmoidal infectivity: If the pathogen is present in the system, sufficient antibiotic treatment leads to pathogen extinction. In contrast, if the pathogen is initially at a very low density and antibiotics are applied as a precautionary measure, the treatment can paradoxically result in pathogen outbreaks. The scenario, where *η* is equal for all environmental species, represents a limiting case along a continuum where the pathogen is affected by the antimicrobial substance either more or less than on the competing species on average. If the treatment targets the pathogen more than other species, the probability of an outbreak is reduced. In the worst case the treatment targets the pathogen less than the competitors leading to an increase in pathogen density by reducing competition. Discussion {#s4} ========== Environmentally growing opportunists are common class of pathogens but there are few attempts to understand how environmental growth and ecological interactions outside the host affect epidemiological dynamics. We present here an analysis of the epidemiology of an environmental pathogen that has sustained growth in the absence of hosts and interacts with non-pathogenic organisms through competitive interactions. The results stress three important factors affecting the host--pathogen interaction: (1) The shape of the infectivity response has a strong impact on the dynamical behavior of the system; (2) Under a sigmoidal dose-dependent pathogen infectivity response reducing species richness of non-pathogenic competitors in the outside-host environment provides a novel mechanism for disease outbreaks; (3) With the sigmoidal infectivity too high pathogen virulence can lead to host extinction, which leads to virulence becoming useless for the pathogen. Microbial Diversity and Disease Dynamics {#s4a} ---------------------------------------- The importance of biodiversity on the stability and functioning of ecological communities continues to motivate ecological research [@pone.0071621-McCann1]--[@pone.0071621-Loreau2]. A common observation is that increasing diversity tends to promote stability of community biomass [@pone.0071621-Tilman1], [@pone.0071621-Hector1]. The importance of diversity in ecological systems and the ubiquitousness of environmentally growing pathogens are well known [@pone.0071621-Cangelosi1], but the theory connecting diversity to disease outbreaks is centred mainly on host diversity [@pone.0071621-Keesing1] with some work on pathogen diversity [@pone.0071621-Gupta1]. Our results show that the intensity of competitive interactions--modified either through community size or the strength of interspecific interactions--in outside-host environments can be very important for the dynamics of environmental opportunistic pathogens and the occurrence of disease outbreaks. These results are in agreement with numerous empirical observations promoting the importance of biodiversity for disease dynamics. For example, biodiversity loss has been associated with both increases and decreases in disease transmission [@pone.0071621-Keesing2]. Loss of fungal diversity in agricultural soil has been shown to result in higher incidence of fungal plant diseases, and it is noteworthy that even generally non-pathogenic fungi can cause diseases if they are the predominant species [@pone.0071621-Nitta1]. Loss of bacterial diversity has been linked to problems in the human intestine such as inflammatory bowel diseases [@pone.0071621-Round1], and low microbial diversity is also suspected to be the cause of several allergies [@pone.0071621-Hanski1]. The Role of the Pathogen's Infectivity Response {#s4b} ----------------------------------------------- Our results indicate that the way increasing competitive pressure on the pathogen in the outside-host community affects pathogen dynamics depends crucially on the shape of the pathogen's infectivity response (i.e., how pathogen infectivity depends on its own density) ([Figs. 3](#pone-0071621-g003){ref-type="fig"}, [S1](#pone.0071621.s001){ref-type="supplementary-material"}). A common assumption in epidemiological models is that the relationship between pathogen number and infections is linear [@pone.0071621-McCallum1]. In this case the pathogen (suffering a cost in competitive ability) is able to compensate for reduced density due to competition via fitness benefits from host infection, and is relatively unaffected by varying, e.g., community size in the outside-host environment ([Fig. 2a](#pone-0071621-g002){ref-type="fig"}). This is because the pathogen remains infective even at low densities. In contrast, under a sigmoidal infectivity response that incorporates an infective dose [@pone.0071621-Regoes1], [@pone.0071621-Joh1] increasing community size can either generate cyclic pathogen outbreaks, or drive the pathogen extinct ([Fig. 2c](#pone-0071621-g002){ref-type="fig"}). If pathogen densities are reduced sufficiently, the pathogen is unable to (re)enter the infection cycle. Similarly to the direct host-to-host transmission of obligatory pathogens studied by Regoes et al. [@pone.0071621-Regoes1], the sigmoidal infectivity response is a disadvantage to the pathogen also in our model with environmental transmission and growth. This is due to the Allee effect associated with sigmoid response [@pone.0071621-Regoes1]. Increasing competitive pressure on the pathogen eventually leads to a situation where the pathogen is unable to recover from low densities and cause infections, resulting in pathogen extinction. On the other hand, when competition is weak the pathogen can drive the host extinct, where after infectiveness becomes useless to the pathogen. This phenomenon resembles the consequences of the classical trade-off between virulence and transmission in obligate pathogens [@pone.0071621-Levin1], [@pone.0071621-Paul1], but arises from a completely different mechanism. The host represents simply a resource for the pathogen that may in cases be 'over-exploited' and lost, reducing the system to the outside-host community. Over-exploitation of the host is the mechanism behind cyclic host-pathogen dynamics, similarly to the mechanism underlying cyclic consumer-resource dynamics [@pone.0071621-Rip1]. As indicated above, the shape of the infectivity response at low pathogen densities has a profound impact on the dynamics. If infectivity is low (and increases slowly) at low pathogen densities, the pathogen needs to reach relatively high densities in the environment to become infective. This can in turn be prevented by, e.g., ecological interactions in the outside-host environment that reduce pathogen densities (such as competition, predation, and parasitism). The importance of the shape of infectivity response in environmental transmission has been recognized by Boldin & Kisdi [@pone.0071621-Boldin1] in evolutionary context. They argued that a concave infectivity response enables evolutionary branching. Here we have shown that convexity of the infectivity response at low pathogen densities can be very important for epidemiological dynamics of opportunistic, environmentally growing pathogens. Environmental Growth {#s4c} -------------------- Outside-host growth of pathogens is widespread in nature. Models of environmentally transmitted diseases typically allow only exponential decay of pathogens [@pone.0071621-Day1], [@pone.0071621-Roche1] or in some cases density-independent growth [@pone.0071621-Codeco1], [@pone.0071621-Joh1], [@pone.0071621-Capasso1], [@pone.0071621-Fisher1]. To our knowledge, density-dependent growth has been analysed only by Merikanto et al. [@pone.0071621-Merikanto1]. Outside-host growth and the recovery of infected hosts have a strong stabilising influence on pathogen dynamics. Density dependent growth dampens the influence of pathogens shed from infected hosts, when the outside-host pathogen population is near its carrying capacity [@pone.0071621-Merikanto1]. Recovery is stabilising because it moderates the decline of susceptible hosts after a disease outbreak. Under a linear infectivity response density-dependent pathogen growth in the outside-host environment effectively filters the inflow pathogens from infected hosts, resulting in stable population dynamics [@pone.0071621-Merikanto1]. Cyclic outbreaks can arise only when pathogens do not exhibit active growth in the outside-host environment (i.e., there is only an exponential decay of pathogen densities), and when the pathogen is extremely lethal. A sigmoidal infectivity response is more prone to generate cyclic population dynamics because after an outbreak host growth is more rapid than the increase in pathogens infectivity at low densities, allowing the host supply to recover before the next outbreak. Implications {#s4d} ------------ Multi-cellular hosts are high resource environments with a potentially deadly immune system for parasites. Parasites often have to allocate considerable amount of resources for infecting and overcoming the immune system, and sacrifice part of their competitive ability in the outside host environment [@pone.0071621-Godfray1]. However, the payoff is that once the immune system is defeated the nutrient rich body may be consumed to gain massive fitness benefits. An example of an effective and highly virulent environmental pathogen that grows slowly in the environment is *Flavobacterium columnare*, a saprotrophic fish pathogen that causes considerable economical losses in fisheries [@pone.0071621-Kunttu1]. Other examples of well-studied environmental saprotrophic pathogens are *Serratia marcescens* and *Pseudomonas sp*. both capable of infecting a wide range of hosts. Opportunism coupled with the ability to grow in the free-living environment, may be an important step in the evolution of virulence for bacteria. This naturally requires that pathogens are released from the infected host to the environment, i.e., the host is not an ecological or evolutionary dead end. Free-living bacteria may develop infectivity but lack the means for effective host-to-host transmission. Thus they would benefit from infectivity only if they are virulent enough to gain reproductive output from the host and survive in the environment until encountering a new host. The waiting time before a new infection may be long and the costs of maintaining infectivity traits are likely to restrict the growth rate of the pathogen. If the resources are sparse and there is competition for them, the non-pathogenic competitors are likely to out-compete the pathogen, and this may lead to local pathogen extinction. Applying a non-specific mortality factor upon the outside-host community can reduce competitive pressure on the pathogen. While the density of all species, including the pathogen, is reduced equally in the outside-host environment, the pathogen gets an indirect advantage through within-host reproduction ([Fig. 3](#pone-0071621-g003){ref-type="fig"}). This suggests that the use of antimicrobial substances as a means of controlling pathogen growth does not always have the desired effect. If all species in the outside-host community are similarly susceptible to the substance, its action is reasonably modeled as a linear mortality term imposed on all members of the outside-host community. This can in some cases help the pathogen by removing competition. It would then be crucial to use enough antimicrobial substances to reach the range where the pathogen cannot survive. It would also be possible that the introduction of effective non-pathogenic competitors could work as a defense measure in such conditions because treatment of hosts in environmentally growing pathogens is ineffective [@pone.0071621-Merikanto1], [@pone.0071621-Joh1]. Supporting Information {#s5} ====================== ###### **Equilibrium pathogen and susceptible host densities as a function of competitive disadvantage of the pathogen (α).** Black dots represent mean densities. At cyclic ranges open circles represent an alternative attractor and filled symbols indicate minima and maxima. Here the number of competitors is fixed to *n = *7 and linear mortality parameter *η* to 1.0. Linear infectivity response with rate constant *β*/2ID~50~ was used in panels (A) and (B), and sigmoidal response ([eqn. 2](#pone.0071621.e005){ref-type="disp-formula"}) in panels (C) and (D). The solid line is the equilibrium pathogen density without hosts. (TIF) ###### Click here for additional data file. ###### **Mean pathogen densities versus competitive disadvantages of the pathogen (α) from sensitivity analysis replicates.** In panel (A) with linear infectivity response red dots represent individual simulation outcomes. In panel (B) with sigmoidal infectivity response cyan dots represent outcomes from simulations resulting in stable dynamics (s.d.(*P*) \<10). Red and blue dots represent outcomes from simulations resulting in cyclic dynamics (s.d.(*P*) \>10). To distinguish between alternative outcomes, values above 1000 are coloured red and those below 1000 are blue. Grey dots represent standard deviations (s.d.(*P*)) above and below the mean value. The replicates resulting in host extinction (mean(*S*) \<5) have been excluded from both panels. Panel (A) has 50720 points and panel (B) has 86615 coloured points of which 14159 are cyclic and 72456 are stable. Total number of simulations is 100000 in both cases. (TIF) ###### Click here for additional data file. ###### **Posterior parameter distributions from sensitivity analysis replicates not resulting in host extinction (mean(*S*) \>5)**. The parameter values for unspecific mortality rate (*η*), host growth rate (*r~h~*), host recovery rate (*δ*), infected death rate (*ν*), and pathogen release (*k*) were picked from uniform random distributions. Selecting the cases with host persistence resulted in 50720 replicates in the upper panels (linear infectivity) and 86615 replicates in the lower panels (sigmoidal infectivity) from a total of 100000 replicates for each infectivity response. (TIF) ###### Click here for additional data file. ###### **Sensitivity analysis.** (DOCX) ###### Click here for additional data file. We thank Eva Kisdi, Lotta-Riina Sundström, Anna-Liisa Laine, Tarmo Ketola, and two anonymous reviewers for insightful comments and suggested corrections. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: JA LR VK JL. Performed the experiments: JA. Analyzed the data: JA LR. Wrote the paper: JA LR VK JL. Provided the original idea: JL. Constructed the model: JA.

All coordinates and structure factor files are available from the PDB database (accession numbers: 6KR2 and 6KR3). Introduction {#sec001} ============ The flaviviruses are a large group of positive-strand RNA viruses, including dengue virus (DENV), West Nile virus (WNV), Japanese encephalitis virus (JEV), Zika virus (ZIKV), tick-borne encephalitis virus (TBEV), and Omsk hemorrhagic fever virus (OHFV). The majority of flaviviruses are mosquito-borne or tick-borne, sometimes causing human encephalitis or hemorrhagic diseases. The recent ZIKV outbreaks in South and North America, and more recently in Southeast Asia, have intensified the global threats of flaviviruses, in part due to the capabilities of the virus to cause birth defects through maternal-fetal transmission \[[@ppat.1008484.ref001]\]. The flavivirus RNA genome is 10--11 kilo-bases in length, bearing a type 1 cap and lacking a poly-adenine tail. It encodes a large polyprotein that is further processed by viral and host proteases, yielding three structural proteins C/prM/E, and seven nonstructural proteins NS1/NS2A/NS2B/NS3/NS4A/NS4B/NS5 \[[@ppat.1008484.ref002]\]. Being a unique natural fusion of an N-terminal methyltransferase (MTase) and a C-terminal RNA-dependent RNA polymerase (RdRP), the NS5 is the largest and most conserved protein encoded by flaviviruses. The NS5 MTase catalyzes the guanylyltransfer and both the guanine N7 and nucleoside 2′-O methylation steps in the capping process, and is a single-domain module adopting a common S-adenosyl-L-methionine (SAM)-dependent MTase fold \[[@ppat.1008484.ref003],[@ppat.1008484.ref004]\]. The K-D-K-E catalytic tetrad sits in the center of the MTase catalytic cleft, with the methyl donor SAM binding site and the cap binding site residing on the opposite sides. The RdRP module is the central molecular machine governing the viral genome replication, and has an encircled human right hand architecture with palm, fingers, and thumb domains surrounding the active site \[[@ppat.1008484.ref005],[@ppat.1008484.ref006]\]. The fingers domain can be further divided into index, middle, ring, and pinky subdomains according to nomenclatures first used in describing the poliovirus (PV) RdRP ([Fig 1A](#ppat.1008484.g001){ref-type="fig"}) \[[@ppat.1008484.ref007],[@ppat.1008484.ref008]\]. Among the seven viral RdRP catalytic motifs, A/B/C/D/E are within the most conserved palm, and F/G are part of the ring and pinky fingers, respectively. Motifs A/B/C/F contain amino acids highly conserved in all viral RdRPs, and these conserved residues have highly analogous spatial arrangements around the polymerase active site \[[@ppat.1008484.ref009],[@ppat.1008484.ref010]\]. Being an entity bearing two active sites and multiple essential viral enzymatic activities, NS5 has become a very attractive system in flavivirus research, and understanding the interplay between its MTase and RdRP modules is undoubtedly critical. {#ppat.1008484.g001} Among the evidence related to MTase-RdRP crosstalk, high-resolution structures of full-length NS5 are essential in providing direct and informative readout of the interactions between the two modules. To date, two types of global conformations have been observed in full-length NS5 structures \[[@ppat.1008484.ref011]\]. The conformation revealed by the JEV NS5 structure (named JEV-mode hereinafter) features a medium size interface (\~1540 Å^2^, for all interface area values presented in this study, the total buried solvent accessible surface from both side of the interface was accounted) with a conserved hydrophobic core \[[@ppat.1008484.ref008],[@ppat.1008484.ref012]\], and is also observed in recently reported ZIKV, yellow fever virus (YFV), and DENV serotype 2 (DENV2) full-length NS5 crystal structures \[[@ppat.1008484.ref013]--[@ppat.1008484.ref018]\]. In such a conformation, the MTase approaches the RdRP from its backside and interacts with the RdRP middle finger, ring finger, and an index finger helix bearing part of a nuclear localization signal (NLS-helix) \[[@ppat.1008484.ref019]\] ([Fig 1A](#ppat.1008484.g001){ref-type="fig"}). The second conformation was observed in two different crystal forms of DENV serotype 3 (DENV3) NS5 (named DENV3-mode hereinafter) \[[@ppat.1008484.ref020],[@ppat.1008484.ref021]\] ([Fig 1B](#ppat.1008484.g001){ref-type="fig"}). In this case, the MTase also approaches the RdRP from the backside, but it is related to the JEV conformation by an approximately 110° rotation around an axis passing near its center of mass with less than 6 Å translation. The RdRP index and middle fingers are still involved in the interactions in a slightly larger interface (\~1650--1780 Å^2^), and the nature of the interactions is instead primarily polar. Notably, the NTP binding ring finger (motif F) contacts with the MTase are absent in the DENV3 structures and the ring finger itself and the adjacent motif G residues in the pinky finger are largely disordered ([Fig 1B](#ppat.1008484.g001){ref-type="fig"}). Motif G participates in RNA template binding and has been proposed to participate in the translocation step after every phosphoryl transfer reaction \[[@ppat.1008484.ref022],[@ppat.1008484.ref023]\]. Hence, the JEV-mode conformation likely represents a state more suitable for polymerase synthesis from the structural perspective. With two monomer conformation modes and eight crystal forms identified, more than 10 NS5 dimer interfaces can be recognized in the aforementioned NS5 crystal structures with no obvious conservative features. A couple of studies did focus on some of these dimer interface interactions, even though the primary NS5 solution state is monomer \[[@ppat.1008484.ref016],[@ppat.1008484.ref021]\]. Either by probing the inter-molecular interactions, deleting the MTase domain, or mutating the MTase-RdRP linker, multiple *in vitro* polymerase assay-based studies together suggest that the MTase regulates the RdRP catalytic activities, albeit to overall moderate extents \[[@ppat.1008484.ref012],[@ppat.1008484.ref014],[@ppat.1008484.ref016],[@ppat.1008484.ref018],[@ppat.1008484.ref024],[@ppat.1008484.ref025]\]. However, the RdRP assays established in all these studies, no matter in primer-dependent or *de novo* (including dinucleotide driven) format, did not demonstrate the formation of a processive RdRP elongation complex (EC), at least for the majority of the polymerase molecules, and all these assays require the manganese ion (Mn^2+^) for catalysis, albeit in combination with the magnesium ion (Mg^2+^) in some cases. In other words, the mutation-derived effect on RdRP synthesis observed in these studies may only reflect overall changes in non-processive RdRP synthesis activity, while specific alteration of either RdRP initiation or elongation cannot be clearly judged. Furthermore, none of these studies characterized both the JEV- and DENV3-mode monomer conformations to distinguish their differences in RdRP synthesis, except for our previous JEV NS5 study that only probed the JEV-mode conformation \[[@ppat.1008484.ref012]\]. Therefore, the precise mechanism of the regulation and the explicit contribution of either NS5 conformation remain to be clarified. In this work, we report two crystal forms of DENV2 NS5 that reveal two conformational states bearing clear analogies to those observed in the JEV-mode and DENV3-mode NS5 structures, respectively. Virological data further support the conservation and the functional importance of both conformation modes. NS5 constructs bearing mutations specifically probing two modes of MTase-RdRP intra-molecular interfaces were tested in *in vitro* polymerase assays, and only the JEV-mode interface related mutants inhibited polymerase initiation primarily through a three-fold reduction in the Michaelis constant of the initiating NTP (*K*~*M*,*NTP*~), while polymerase EC properties were not much affected by mutations probing both modes of interactions. Collectively, our work demonstrates the conformational conservation and diversity of the flavivirus NS5 and highlights the specific contribution of the JEV-mode conformation to polymerase initiation. Results {#sec002} ======= The first form of DENV2 NS5 structure contains a partially open MTase-RdRP interface and is clearly related to the JEV-mode NS5 structures {#sec003} ------------------------------------------------------------------------------------------------------------------------------------------ With an aim to further understand the conformational diversity of NS5 and related functional relevance, we crystallized and solved the structures of DENV2 NS5 in two different crystal forms at 3.1 Å (form 1) and 2.9 Å (form 2) resolution ([Table 1](#ppat.1008484.t001){ref-type="table"}). Each structure has two NS5 molecules in the crystallographic asymmetric unit, and has the two molecules arranged in a dimer through pseudo two-fold symmetries with highly consistent global conformation (root mean square (RMS) deviation values for superimposable C-α atoms are 0.74 Å and 0.57 Å, respectively; chain A as the reference). Strikingly, the NS5 conformations between the two crystal forms are quite different ([Fig 1C and 1D](#ppat.1008484.g001){ref-type="fig"}). Using a maximum likelihood superpositioning method \[[@ppat.1008484.ref026]\], the RdRP palm and the majority of the fingers domain were identified as the structurally most conserved regions in the superpositioning including these and the representative full-length flavivirus NS5 structures ([Fig 1](#ppat.1008484.g001){ref-type="fig"}). The DENV2 form 1 conformation is clearly related to the JEV-mode conformation with the JEV interface partially opened through a pure 20° rotation along an axis passing the vicinity of the highly conserved GTR residues that were proposed to pivot the MTase movement relative to RdRP \[[@ppat.1008484.ref008],[@ppat.1008484.ref027]\] ([Fig 1A and 1C](#ppat.1008484.g001){ref-type="fig"}; [Fig 2A](#ppat.1008484.g002){ref-type="fig"}). The partial opening of the interface results in the reduction of the interface area to only about 900 Å^2^ or 59% of the JEV interface. Among the six conserved hydrophobic residues forming the interface core in the JEV structure, only three of them (DENV2 NS5 residues W121, F349, and P583) remained as part of the interface. The tip of the ring finger no longer contacts the MTase in the DENV2 form 1 structure, and its electron density becomes weak but still readily traceable and its folding is largely consistent with the JEV conformation ([Fig 1A and 1C](#ppat.1008484.g001){ref-type="fig"}; [Fig 2A](#ppat.1008484.g002){ref-type="fig"}). However, the motif G region in the pinky finger is largely disordered as observed in the DENV3 structures \[[@ppat.1008484.ref020],[@ppat.1008484.ref021]\]. Based on these observations, we propose that the interactions between the MTase residues 113 and 115 and the phenylalanine (residue 465 in DENV2 NS5) in the tip of the RdRP ring finger are essential for maintaining the canonical folding of the ring and pinky fingers, and the folding of the motif G region in the pinky finger is likely dependent on the dynamics of the adjacent ring finger. It is also worth noting that, in addition to ring and pinky fingers, the index finger is also partially disordered in most of the RdRP-only flavivirus NS5 structures \[[@ppat.1008484.ref005],[@ppat.1008484.ref006]\]. These observations together suggest that MTase interaction likely contributes to the folding of RdRP fingers domain, which in turn could affect polymerase properties including RNA binding and subsequent catalytic events. {ref-type="fig"}. For panels A and B, the rotational movements correlate the both structure pairs are indicated.](ppat.1008484.g002){#ppat.1008484.g002} 10.1371/journal.ppat.1008484.t001 ###### X-ray diffraction data collection and structure refinement statistics. {#ppat.1008484.t001g} Crystal form--PDB 1 -- 6KR2 2 --6KR3 --------------------------------------------------------------- ------------------------- ------------------------- **Data collection**[^**1**^](#t001fn001){ref-type="table-fn"} Space group P2~1~ C222~1~ Cell dimensions *a*, *b*, *c* (Å) 87.2, 146.4, 98.4 178.8, 210.0, 157.9 α, β, γ (°) 90, 105.8, 90 90, 90, 90 Resolution (Å)[^2^](#t001fn002){ref-type="table-fn"} 60.0--3.06 (3.17--3.06) 60.0--2.93 (3.03--2.93) R~merge~ 0.160 (0.50) 0.138 (0.55) R~meas~ 0.191 (0.60) 0.153 (0.61) CC~1/2~ 0.933 (0.781) 0.952 (0.832) I / σI 6.8 (2.1) 11.6 (3.0) Completeness (%) 98.3 (99.1) 99.9 (100.0) Redundancy 3.2 (3.2) 5.6 (5.5) **Refinement** Resolution (Å) 3.06 2.93 No. reflections 44,340 63,847 R~work~ / R~free~ [^3^](#t001fn003){ref-type="table-fn"} (%) 22.4 / 27.5 23.3 / 27.0 No. atoms     Protein 12638 12758     Ligand / Ion / Water 52 / 4 / 33 134 / 19 / 47 B-factors (Å^2^)     Protein 27.9 60.3     Ligand / Ion / Water 54.5 / 76.4 / 22.5 59.9 / 93.2 / 52.5 RMS deviations     Bond lengths (Å) 0.011 0.012     Bond angles (°) 1.42 1.25 Ramachandran stat.[^4^](#t001fn004){ref-type="table-fn"} 79.9 / 17.2 / 1.7 / 1.2 83.1 / 15.1 / 1.0 / 0.8 ^1^ One crystal was used for data collection for each structure. ^2^ Values in parentheses are for highest-resolution shell. ^3^ 5% of data are taken for the R~free~ set. ^4^ Values are in percentage and are for most favored, additionally allowed, generously allowed, and disallowed regions in Ramachandran plots, respectively. The second form of DENV2 NS5 structure resembles the DENV3 NS5 structures {#sec004} ------------------------------------------------------------------------- The DENV2 form 2 conformation is instead analogous to the DENV3-mode conformation but contains previously unidentified features. It is related to the DENV3-mode conformation by a 14° rotation along an axis near the interface and the MTase-RdRP linker region (residues 264--273 in DENV2 NS5) and a translation less than 2 Å ([Fig 1B and 1D](#ppat.1008484.g001){ref-type="fig"}; [Fig 2B](#ppat.1008484.g002){ref-type="fig"}). The primarily polar interactions between the NLS-helix and the MTase are largely retained ([Fig 2B](#ppat.1008484.g002){ref-type="fig"}), and interface area is about 1460 Å^2^ and is only reduced for about 11% as compared to the first reported DENV3-mode interface \[[@ppat.1008484.ref020]\]. The rotational movement widens the cleft between the relatively conserved residue pair E67-R68 in the MTase and the RdRP middle finger, creating a pocket that allows the binding of a putative SAH molecule with high occupancy (0.90 and 0.97) in addition to the SAH molecule usually bound in the SAM binding pocket of the MTase ([Fig 1D](#ppat.1008484.g001){ref-type="fig"}; [Fig 2B and 2C](#ppat.1008484.g002){ref-type="fig"}). Usually, SAH can be co-crystallized with the flavivirus MTase at a 1:1 molar ratio with the SAH co-purified with the MTase after its overexpression in bacterial culture \[[@ppat.1008484.ref004],[@ppat.1008484.ref008],[@ppat.1008484.ref020],[@ppat.1008484.ref028]\]. It is possible that the observed NS5 conformation allowed SAH co-purification with NS5 at a higher molar ratio. Such a secondary SAH binding site has not been observed in numerous MTase-containing flavivirus NS5 structures. This binding pocket appears to be not tight enough, as the SAH molecules bind at moderately different positions in the two NS5 proteins in the crystallographic asymmetric unit, and specific interactions between the non-carbon atoms of the SAH and the side chains of the NS5 are largely lacking ([Fig 2B and 2C](#ppat.1008484.g002){ref-type="fig"}). Nevertheless, based on the fact that this secondary SAH binding pocket is created at the MTase-RdRP interface and is reasonably conservative ([Fig 2C](#ppat.1008484.g002){ref-type="fig"}), it might have potential as a target to develop small molecule inhibitors against flaviviruses. The correlation among all four NS5 conformational states and the structural elements that may mediate NS5 conformational dynamics {#sec005} --------------------------------------------------------------------------------------------------------------------------------- The two DENV2 NS5 structures nicely fill the gap of major conformational differences between the JEV-mode and DENV3-mode structures, suggesting a plausible order from JEV-mode to DENV2 form 1, then to form 2, and finally to DENV3-mode, primarily through rotational movements in more or less consistent directions ([Fig 1](#ppat.1008484.g001){ref-type="fig"}; [Fig 3A and 3B](#ppat.1008484.g003){ref-type="fig"}; [S1 Movie](#ppat.1008484.s001){ref-type="supplementary-material"}). Three structural elements may play critical roles in the transitioning among these states. The first is the universally conserved GTR sequence at the C-terminal end of the MTase ([Fig 3C](#ppat.1008484.g003){ref-type="fig"}). This tripeptide sequence was proposed as a pivoting element in the work of the full-length JEV NS5 structure and was proved to be functionally important in both JEV and DENV2 replications \[[@ppat.1008484.ref008],[@ppat.1008484.ref027]\]. The second element is the 10-residue MTase-RdRP linker that overall exhibits low sequence conservation in flaviviruses ([Fig 3C](#ppat.1008484.g003){ref-type="fig"}). A comparison of all full-length NS5 structures demonstrated that the N-terminal half of the linker undergoes a swinging motion with a partial refolding to become helical when transitioning from the JEV-mode conformations to the DENV3-mode ones, while the C-terminal half remains unaffected ([Fig 3A and 3B](#ppat.1008484.g003){ref-type="fig"}). Not surprisingly, mutations in the linker region or linker substitutions using sequences of other flavivirus NS5 have been found to affect NS5 conformation distribution, NS5 enzymatic activities, and virus proliferation \[[@ppat.1008484.ref018],[@ppat.1008484.ref024],[@ppat.1008484.ref029]\], possibly by altering the flexibility of the linker. The third element is the NLS-helix (residues 348--358 in DENV2 NS5) in the RdRP index finger. On one hand, this helix is critical to both the JEV-mode and DENV3-mode of interface interactions by contributing the F349 and R353 residues ([Fig 2A and 2B](#ppat.1008484.g002){ref-type="fig"}; [Fig 3B](#ppat.1008484.g003){ref-type="fig"}). On the other hand, it is at the central region of a long stretch of conserved sequences (residues 341--366 in DENV2 NS5) that may also be related to NS5 nuclear localization, nuclear export, and interaction with another important viral protein NS3 \[[@ppat.1008484.ref019],[@ppat.1008484.ref030],[@ppat.1008484.ref031]\], emphasizing its possible importance when not participating in the intra-molecular MTase-RdRP interactions. We propose that the largely rotational movement of the MTase from the JEV-mode conformations to the DENV3-mode conformations may utilize this highly conserved helix as a guiding track ([S1 Movie](#ppat.1008484.s001){ref-type="supplementary-material"}). At the starting and end points of the movement, the conserved hydrophobic residue patch P113/X115/W121 (the majority of X are L and M) and the conserved polar residue pair E67-R68 provide the anchoring points on the MTase side ([Fig 3B--3D](#ppat.1008484.g003){ref-type="fig"}). On the other hand, both of these stable conformations make the NLS-helix inaccessible to other factors, and the helix may only become solvent exposed at certain stages of virus life cycle. {ref-type="fig"}. The α-carbon atoms of two important patches of residues are shown as blue spheres. C) The sequence logo plot showing the conservation of the two RdRP interacting regions, the GTR-linker region (top panel), and the NLS-helix that is important for both JEV- and DENV3-mode conformational states, the middle and ring finger regions only critical in JEV-mode states (bottom panel). The triangles indicated key residues involved in the intra-molecular interactions. D) A schematic free energy diagram for all four conformational states of flavivirus NS5. The relative free energy was crudely estimated by the solvent accessible surface area occluded by the NS5 intra-molecular interface interactions. E) A list of DENV2 NS5 mutants that were designed based on both the JEV- and DENV3-mode conformations, with abbreviations and full descriptions, including mutation site and mutation type.](ppat.1008484.g003){#ppat.1008484.g003} Virological data support the functional relevance of both the JEV-mode and the DENV3-mode conformations {#sec006} ------------------------------------------------------------------------------------------------------- Previously, we tested the functional significance of the JEV-mode conformation using both JEV and DENV2 systems \[[@ppat.1008484.ref027]\]. When arginine/aspartic acid/serine (R/D/S) mutations were introduced at the six hallmark hydrophobic residue sites, virus proliferation was significantly inhibited (the corresponding mutation sites in DENV NS5 were listed in [Fig 3E](#ppat.1008484.g003){ref-type="fig"}). In order to understand functional relevance of the DENV3-mode conformation, we designed five mutations at the NS5 residues 67 and 68 for each virus system (E67A, E67D, K68A, K68R and E67A/K68A in JEV; E67A, E67D, R68A, R68K and E67A/R68A in DENV2, and correspond to the same mutations with an "M\_" prefix in [Fig 3E](#ppat.1008484.g003){ref-type="fig"}) and compared the mutant constructs with the wild type (WT) viruses ([Fig 4](#ppat.1008484.g004){ref-type="fig"}). The residues 67--68 were chosen as the mutation sites because these two residues are highly conserved among the MTase residues that participate in the DENV3-mode MTase-RdRP interface but are not involved in the JEV-mode interface interactions. We first introduced each mutation into a JEV infectious clone \[[@ppat.1008484.ref032]\]. After viral RNA was transfected into baby hamster kidney cells BHK-21, viral protein expression and virus production were monitored. The expression level of the viral envelope (E) protein in transfected cells was detected by an immunofluorescence assay (IFA) ([Fig 4A](#ppat.1008484.g004){ref-type="fig"}). Both the E67A and K68R mutant viruses produced similar IFA positive cells in comparison with the WT virus (100% IFA positive cells observed at 72 h post transfection (hpt)); The K68A and E67A/K68A mutants showed only around 10% positive cells; the E67D mutant produced very few IFA-positive cells. Virus productions were then quantified by a plaque assay at three time points (48, 72, and 96 h) post transfection. Consistent with the IFA data, the E67A and K68R mutant RNAs yielded similar amounts of viruses as the WT at each time point, the K68A and E67A/K68A mutants moderately impaired virus production, and viruses derived from the E67D mutant RNA-transfected cells were only detected at 72 and 96 hpt ([Fig 4A](#ppat.1008484.g004){ref-type="fig"}). Overall, the results indicated residues 67 and 68 are important for JEV proliferation. We also performed similar analyses using a DENV2 infectious clone ([Fig 4B](#ppat.1008484.g004){ref-type="fig"}) \[[@ppat.1008484.ref033]\]. No IFA-positive cells were observed in the R68A, R68K and E67A/R68A transfected cells; the E67A and E67D produced around 70%-80% IFA-positive cells relative to the WT. Data from the plaque assay indicated that virus production was blocked by the R68A, R68K, and E67A/R68A mutations, while the E67A and E67D mutations had slightly less effect on virus production at each time point post transfection comparing with the WT ([Fig 4B](#ppat.1008484.g004){ref-type="fig"}). Taken together, these virological data suggest that the DENV3-mode conformation is also functionally important and is likely conserved in flaviviruses, consistent with our structural observation of both conformations in DENV2 NS5. {#ppat.1008484.g004} The establishment of effective *in vitro* RdRP assays to characterize both initiation and elongation in DENV2 NS5 {#sec007} ----------------------------------------------------------------------------------------------------------------- We previously tested the JEV-mode interface mutants in JEV NS5 using *in vitro* assays derived from an HCV study \[[@ppat.1008484.ref012],[@ppat.1008484.ref034]\]. While the HCV assays have Mg^2+^ as the only divalent metal ion and allow the formation of processive ECs \[[@ppat.1008484.ref034]\], the JEV assays require Mn^2+^ for RdRP activity and the stability and reactivity of the assembled complexes are far from optimal \[[@ppat.1008484.ref008]\]. When the HCV assay format was used in the DENV2 NS5, the RdRP enzyme behavior is consistent with the HCV enzyme, and therefore the DENV2 assays established in this study can serve as effective systems to assess RdRP catalytic properties and to analyze whether the interface interactions observed in both the JEV-mode and the DENV3-mode conformations modulate RdRP catalysis. Note that dinucleotide-driven assays, such as those in the HCV study \[[@ppat.1008484.ref034]\], have been used in multiple polymerase systems to reasonably mimic the *de novo* initiation process \[[@ppat.1008484.ref035],[@ppat.1008484.ref036]\]. When ATP and UTP were provided as the only NTP substrates, a GG dinucleotide primer (P2) was extended to yield a 9-mer product (P9) as directed by a 30-mer template (T30) after a 45-min incubation ([Fig 5A](#ppat.1008484.g005){ref-type="fig"}). We assessed the reactivity of the P9-containing complex by a single-nucleotide extension assay. Because the complex was in the precipitate form under the low-salt reaction condition (20 mM NaCl) but was soluble under high-salt condition (e.g. 190 mM NaCl), we removed the excess ATP and UTP by centrifugation, pellet wash, and pellet resuspension, and then added CTP to allow the single-nucleotide addition to make a 10-mer product (P10). It turned out that the conversion from P9 to P10 was very rapid and was completed immediately after manual mixing on ice without further incubation ("0 min"; [Fig 5A](#ppat.1008484.g005){ref-type="fig"}, lane 3). Although not accurately determined, the expected catalytic rate constant of the P9-containing polymerase complex is at least magnitudes larger than that determined in the JEV study (0.14 min^-1^ for WT NS5) \[[@ppat.1008484.ref012]\]. This observation strongly suggests that the P9-containing complex of DENV2 NS5 has completed the transition from initiation to elongation and is a *bona fide* EC. Therefore, the production of this P9-containing EC (EC9) can be used to assess the overall process of initiation followed by the transition to elongation. ![Mutations perturbing the JEV-mode interface interactions impaired DENV2 RdRP initiation.\ A) Left: A diagram of construct T30/P2 used in all polymerase assays and related NTP-driven reactions to generate products with different lengths. Middle: reaction flow chart of the P2-driven EC formation (IC2 to EC9) and the subsequent single-nucleotide extension (EC9 to EC10). Right: the EC9 was in a form of precipitate and was able to extend to EC10 upon CTP addition under high-salt condition. B-C) The EC9 formation comparison with the WT NS5 for the JEV-mode (B) and DENV3-mode (C) mutants. The relative intensity of the 9-mer was used to estimate the polymerase activities (the WT value for each time point series was set to 100). D) Comparison of the WT and two representative mutants (R3 for the JEV-mode; M_67A/68A for the DENV3-mode) in the multiple-turnover P3 formation and in the single-turnover P9 formation assays. The pppGGA was synthesized when GTP and ATP were provided as the only NTP substrates using the P2-free T30 template and was used as a migration marker. Note that pppGGA migrated at similar position as pGG, but faster than pGGA since it contains two extra phosphate groups at the 5′ end. A chemically synthesized 9-mer loaded with an equal molar amount to T30 was used as a quantitation standard (STD, lanes 62 and 81). The average intensity of the STD bands was set to 1. Based on previously reported evaluation, the intensity-molar amount starts to deviate from a linear relationship when the relative intensity approaches 4--5 under similar experimental settings \[[@ppat.1008484.ref012]\]. Therefore, the intensity reported in lanes 64--65 and 76--77 are underestimated. The P9 product migrated faster than the chemically synthesized STD RNA due to its 5′-phosphate inherited from the pGG dinucleotide.](ppat.1008484.g005){#ppat.1008484.g005} Perturbing the JEV-mode but not the DENV3-mode intra-molecular interactions impairs the DENV2 NS5 RdRP initiation but not elongation {#sec008} ------------------------------------------------------------------------------------------------------------------------------------ To test the impact on polymerase catalysis brought by the intra-molecular MTase-RdRP interactions, we made two sets of DENV2 NS5 mutants. The first set contains equivalent mutations utilized in the JEV study to perturb the hydrophobic JEV-mode interface \[[@ppat.1008484.ref012]\], and the second set that contains mutations at residues 67--68 was used to probe the polar DENV3-mode interface ([Fig 3E](#ppat.1008484.g003){ref-type="fig"}). We compared the EC9 formation of these mutants with the WT NS5 at three incubation time points. For the WT NS5, only limited 9-mer accumulation was observed beyond the first time point (15 min), and small amount of misincorporation-related 10-mer products became obvious at the last time point (90 min) ([Fig 5B](#ppat.1008484.g005){ref-type="fig"}, lanes 11, 18, and 25; [Fig 5C](#ppat.1008484.g005){ref-type="fig"}, lanes 41, 47, and 53). These observations again suggest that: although the EC9 formation is a slow process due to slow pre-initiation and initiation steps, it likely produces a stable EC that is not turning over to carry out multiple rounds of P9 synthesis. Among the six JEV-mode NS5 mutants, three of them exhibited obvious slower accumulation of P9 products, in particular at shorter incubation time points ([Fig 5B](#ppat.1008484.g005){ref-type="fig"}, compare lanes 15--17, 22--24, 29--31 to the WT lanes). In contrast, all five DENV3-mode NS5 mutants showed very similar trend of P9 accumulation as the WT enzyme ([Fig 5C](#ppat.1008484.g005){ref-type="fig"}). These data together suggest that perturbing the JEV-mode but not the DENV3-mode interactions inhibits the overall process to produce a processive EC. To further dissect the mechanism of inhibition brought by the JEV-mode interface mutations, we compared representative NS5 mutants (R3 for JEV-mode interface and M_67A/68A for DENV3-mode interface) with the WT enzyme in a P2-driven initiation assay to explicitly assess the continuous production of the 3-mer (P3) when ATP was provided as the only NTP substrate, while the P9 synthesis was monitored in parallel for comparison ([Fig 5D](#ppat.1008484.g005){ref-type="fig"}). Here we use a chemically synthesized 9-mer as a quantitation standard (STD; [Fig 5D](#ppat.1008484.g005){ref-type="fig"}, lanes 62 and 81) loaded with an equal molar amount of the T30 template. For the WT, the R3 mutant, and the M_67A/68A mutant, the P9 amount was relatively consistent at the 60 and 90 min time points ([Fig 5D](#ppat.1008484.g005){ref-type="fig"}, compare lanes 67, 73, 79 to lanes 68, 74, 80), suggesting that all three constructs had formed stable ECs and did not turn over to accumulate the P9 products over time. In contrast, the P3 accumulation proceeded continuously during the same period for all three constructs, indicating that the P3-containing complex is an initiation complex (IC) that carried out multiple rounds of synthesis in an abortive fashion ([Fig 5D](#ppat.1008484.g005){ref-type="fig"}, lanes 63--65, 69--71, and 75--77). The R3 mutant had a slower P3 accumulation than the WT and the M_67A/68A mutant had, clearly suggesting that the perturbation of the JEV-mode interface impaired the RdRP initiation process. We next performed two tests regarding the EC9 properties using the WT, R3, and M_67A/68A constructs ([Fig 6](#ppat.1008484.g006){ref-type="fig"}). To test the EC9 reactivity, we compared the P9-to-P10 conversion under high and low CTP substrate concentrations for these constructs ([Fig 6B](#ppat.1008484.g006){ref-type="fig"}). At 300 μM CTP concentration, all three constructs converted the majority of P9 to P10 at "0-min" time point (89--94% converted suggested by intensity-based quantitation) ([Fig 6B](#ppat.1008484.g006){ref-type="fig"}, lanes 12, 14, and 16). When CTP was supplied at 5 μM, the conversion became slower, but all three constructs showed consistent progress of conversion (67--70% and 84--88% converted at "0-min" and at 1 min, respectively) ([Fig 6B](#ppat.1008484.g006){ref-type="fig"}). To test the stability of the EC9, we used NaCl as the challenging agent in a high-salt challenge stability assay similar to those described in previous work characterizing the PV, HCV and the classical swine fever virus (CSFV) RdRPs \[[@ppat.1008484.ref034],[@ppat.1008484.ref037],[@ppat.1008484.ref038]\]. We found that EC9 formed by all three constructs were quite stable and exhibited comparable inactivation rate constants (0.02--0.04 h^-1^, corresponding to half life values of 18--35 h) upon a NaCl challenge at 500 mM concentration ([Fig 6C and 6D](#ppat.1008484.g006){ref-type="fig"}). These data together suggest that the EC9 is highly stable and reactive, and these properties were not much affected by both types of mutations. {#ppat.1008484.g006} The representative R3 mutant perturbing the JEV-mode interface leads to about 4-fold reduction in initiation efficiency and primarily by affecting the initiating NTP binding {#sec009} ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- To specifically investigate the enzymatic properties of the WT and the two representative mutants during the conversion of the P2 to P3, we measured the relative catalytic rates under different ATP concentrations for each construct ([Fig 7A--7C](#ppat.1008484.g007){ref-type="fig"}), and these data were used to determine the Michaelis constants (*K*~*M*~) of these constructs ([Fig 7D--7F](#ppat.1008484.g007){ref-type="fig"}). By optimizing the reaction time points selection for each NTP concentration and each construct, the 3-mer band intensities were controlled to be within the linear range of the Stains-All based staining method, to facilitate quantitation accuracy (refer to our previous analysis in the JEV study \[[@ppat.1008484.ref012]\]) ([Fig 7A--7C](#ppat.1008484.g007){ref-type="fig"}). It turned out that the *K*~*M*~ value of the R3 mutant is about three times the values of the WT and M_67A/68A mutant (1046 μM vs. 344 μM and 312 μM), indicating that the initiating NTP binding is clearly impaired by the JEV-mode interface mutations ([Fig 7D--7F](#ppat.1008484.g007){ref-type="fig"}). Higher NTP concentrations were not tested in these trials due to a substrate inhibition effect observed in preparatory experiments. We then use the relative catalytic rates determined at 1000 μM for the WT and M_67A/68A mutant, and 1500 μM for the R3 mutant, respectively, to correlate the curve fittings of all three constructs ([Fig 7G](#ppat.1008484.g007){ref-type="fig"}, see [Materials and Methods](#sec014){ref-type="sec"}). The estimated relative specificity constant (rel. *k*~*cat*~*/K*~*M*~) were approximately 0.27 and 1.45 for R3 and M_67A/68A mutants, respectively ([Fig 7G](#ppat.1008484.g007){ref-type="fig"}, using the WT as reference). Taken together, these data suggest that perturbation of the JEV-mode interface clearly impaired DENV2 RdRP initiation mainly by affecting the initiation NTP binding, while perturbation of the DENV3-mode interface only had a moderate effect. {ref-type="sec"}). The STD samples were used as a reference and NTP concentration range was adjusted for each NS5 construct, both for ensuring that all P3 band intensities were within the linear range of the staining method used. D-F) The relative reaction rates (left) and *K*~*M*~ fitting (right) analyses for the WT (D), R3 (E), and M_67A/68A (F) constructs. Left: The adjusted intensity of the 3-mer products as a function of time for WT, R3, and M_67A/68A with five ATP concentrations. (G) An analysis of the relative specificity constants of the R3 and M_67A/68A mutants to the WT. Single-gel based reaction rate correlation analysis (left) for the R3-WT (purple points) and M_67A/68A-WT pairs to correlate three Michaelis-Menten curves for determination of the relative specificity constants (right).](ppat.1008484.g007){#ppat.1008484.g007} In the multiple-turnover P3 accumulation process, either the catalysis of the P2-to-P3 conversion and the dissociation of the P3 product could be rate limiting. Hence, the observed *k*~*cat*~ related differences in the P3 conversion initiation assay could be affected by the variation in the P3 dissociation rates among different constructs. To further validate our judgment, we monitored the P2-driven P9 formation for the WT, R3, and M_67A/68A constructs ([Fig 8](#ppat.1008484.g008){ref-type="fig"}). In this case, the observed difference in the P9 formation is highly dependent on the catalytic rate (presumably the P2-to-P3 conversion step), while not much related by dissociation rates at subsequent steps due to the single-turnover feature of the reaction and the absence of the intermediate products (3-8-mer) in the denaturing gel analyses. When the initiating ATP was supplied at the *K*~*M*~-level concentration of the WT and M_67A/68A mutant (300 μM), the overall conversion rate (P2-to-P9) of the R3 mutant was much slower than those of the other two constructs (0.05 min^-1^ vs. 0.23--0.24 min^-1^). When the ATP concentration was lifted to the *K*~*M*~-level of the R3 mutant (1000 μM), the conversion rate of the R3 was still lower than those of the other two constructs, while the difference between the R3 and WT/ M_67A/68A constructs became smaller (0.11 min^-1^ vs. 0.20--0.27 min^-1^). These data validate the judgment derived from the P3 formation assay and strongly suggest that the JEV-mode interface interactions are critical in NS5 initiation, and in particular, in the initiation NTP binding. {#ppat.1008484.g008} Discussion {#sec010} ========== On the conformation diversity and conformation-function relationship of the flavivirus NS5 {#sec011} ------------------------------------------------------------------------------------------ By solving the DENV2 NS5 crystal structures in two different global conformations and characterizing the functional relevance of both conformations, the work presented here helps the understandings of the conformational diversity and conservation in flavivirus NS5, highlighting the important role of the MTase module, a natural fusion partner of the NS5 RdRP module, in the initiation phase of RNA synthesis. Interestingly, the MTase only facilitates RdRP initiation through the JEV-mode conformation, and this specific mechanism is in agreement with the structural observation that the MTase stabilizes the folding of the NTP-binding ring finger only through this conformation \[[@ppat.1008484.ref008],[@ppat.1008484.ref039]\] ([Fig 1A](#ppat.1008484.g001){ref-type="fig"}). Both conformation modes seem not to apparently contribute to the RdRP catalysis in the elongation phase, supporting our previous proposal that the RdRP may only need the assistance from the MTase in the unstable initiation phase \[[@ppat.1008484.ref012],[@ppat.1008484.ref040]\]. Based on the virological data, both conformation modes are important for virus proliferation. Therefore, the DENV3-mode conformation could contribute to other NS5-involved processes. However, whether and how it is related to the catalysis of the MTase, or to the interactions with other viral proteins or host factors remain to be clarified. Note that, the index finger NLS-helix (residues 348--359 in DENV2 NS5), possibly participating in NS3-binding or NS5 shuttling between cytoplasm and nucleus, was occluded by the MTase in both JEV- and DENV3-mode conformations \[[@ppat.1008484.ref019],[@ppat.1008484.ref030],[@ppat.1008484.ref041]\]. Interestingly, residues in the same helix were found to mediate RANTES ([r]{.ul}egulated on [a]{.ul}ctivation, [n]{.ul}ormal [T]{.ul} cell [e]{.ul}xpressed and [s]{.ul}ecreted, also known as CCL5) expression in TBEV \[[@ppat.1008484.ref031]\]. Therefore, additional functional relevant conformation modes likely exist, as also suggested by the small-angle X-ray scattering and reverse genetics data \[[@ppat.1008484.ref005],[@ppat.1008484.ref042],[@ppat.1008484.ref043]\]. Based on the conformational variation of the N-terminal half of the linker region observed in crystallography, the MTase is able to reach the fingers side without much difficulty. However, a conformational change of the entire linker, or additional rearrangements of the N-terminal extension (NE, residues 274--301) of the RdRP is probably necessary to allow the MTase to reach the front of RdRP as we previously suggested \[[@ppat.1008484.ref008]\]. Naturally fused regions as common regulators of viral RdRPs {#sec012} ----------------------------------------------------------- Although from the same virus family and utilizing the same *de novo* initiation mechanism, the overall structure is quite different for *Flaviviridae* RdRP molecules from representative virus genera. The 66-kD HCV (the type species of the hepacivirus genus) NS5B comprises the RdRP catalytic core and a 21-residue C-terminal membrane anchor. The 82-kD pestivirus NS5B has a 24-residue C-terminal membrane anchor and a \~90-residue unique N-terminal domain (NTD) specifically modulate the RdRP fidelity through intra-molecular interactions with the RdRP palm \[[@ppat.1008484.ref037]\]. The flavivirus NS5 does not have the C-terminal membrane anchor, but is naturally fused to the capping related MTase at its N-terminus. Compared to the NTD-RdRP regulation in pesitivirus NS5B, the crosstalk between the flavivirus MTase and RdRP seems to be more versatile with respect to conformational diversity and functional relevance, as discussed above. While the *Flaviviridae* RdRPs exhibit diversities in global structure and regulatory mechanisms involving the naturally fused regions of the RdRP module, this phenomenon appears to be commonly occurring in viral RdRPs \[[@ppat.1008484.ref044]\]. The *Rhabdoviridae* L proteins contain four additional functional regions including an MTase and a polyribonucleotidyl transferase (PRNTase) \[[@ppat.1008484.ref045]\]; the *Bunyavirales* L proteins include two additional regions with analogy to the PA and PB2 subunits of the *Paramyxorviridae* RdRP complex that participate in the cap-snatching process to generate a capped primer for RdRP synthesis \[[@ppat.1008484.ref046],[@ppat.1008484.ref047]\]; the *Alphatetraviridae* RdRPs contain an MTase and a helicase at the N-terminal region \[[@ppat.1008484.ref048]\]; the *Nidovirales* RdRPs, including the *Coronaviridae* nsp12 and *Arteriviridae* nsp9, contain a \~200-400-residue N-terminal region with the nucleotidyltransferase (NiRAN) function \[[@ppat.1008484.ref049],[@ppat.1008484.ref050]\]. Likewise, viral RdRPs may evolve from common ancestors comprising only the catalytic module with relative independency in carrying out RNA synthesis, similar to the PV 3D^pol^ and HCV NS5B. Co-evolution with versatile host species and the low coding capacity-driven protein function combination may contribute to the structural and functional diversity of current viral RdRPs with respect to regions beyond the RdRP catalytic module. We propose that the functional roles currently offered by these regions, such as the fidelity modulation in pestivirus NS5B and initiation enhancement flavivirus NS5, are likely a result of their co-evolution with the RdRP module through the establishment of specific interactions. Moreover, the RdRP fidelity and initiation enhancement offered by the pesitvirus NS5B NTD and flavivirus NS5 MTase, respectively, is likely not an improvement of a specific function relative to the corresponding RdRP ancestors, but a balance between gaining an extra module and maintaining the levels of key enzymatic properties for the RdRPs. Implications for conformation-based protein function {#sec013} ---------------------------------------------------- Large-scale protein conformational changes are amazing events in biological systems, albeit associated with very different characteristics. Driven by multiple cycles of phosphoryl transfer reactions and accompanied by dramatic changes in protein-nucleic acid interactions, the N-terminal one third of bacteriophage T7 RNA polymerase undergoes a dramatic rearrangement as it makes an irreversible transition from the promoter-bound initiation state to the promoter-free elongation state in the transcription process \[[@ppat.1008484.ref051]--[@ppat.1008484.ref053]\]. By contrast, the observed conformational diversity in flavivirus NS5 has not yet involved continuous events of nucleotide addition, and the energy barrier between different observed states may be low enough to allow a distribution of several states in solution. As both captured by crystallography in multiple virus species, the JEV-mode and DENV3-mode conformations probably represent relatively stable states of apo NS5, thus forming the foundation for further understanding of the NS5 function when it is participating in enzymatic reactions or interactions with essential viral or host factors. The flavivirus NS5 is also a great example that the different interaction modes can be established between regions of a single protein, and are related to different functions of the protein and different processes in the life cycle of the corresponding species. In a broader context, the conformation-based function diversity extends the function capacity beyond the traditional consideration of protein folding, and is therefore an important factor when understanding protein function. Materials and methods {#sec014} ===================== Cloning and protein production {#sec015} ------------------------------ The full-length WT DENV2 NS5 gene within the DNA clone of TSV01 strain (GenBank: AY037116) was cloned into a pET26b vector to yield the pET26b-DENV2-NS5 plasmid. Eleven full-length NS5 constructs with point mutations ([Fig 3E](#ppat.1008484.g003){ref-type="fig"}) were made by using the QuickChange site-directed mutagenesis method and the WT plasmid as the template. NS5 expressing plasmids were transformed into *Escherichia coli* strain BL21-CodonPlus(DE3)-RIL for expression of NS5 constructs with a hexa-histidine tag at the C-terminus. Cells were grown at 30°C overnight in the NZCYM medium containing 25 μg/mL kanamycin (KAN25) and 20 μg/mL chloramphenicol (CHL20) until the optical density at 600 nm (OD~600~) was 1.0. The overnight culture was used to inoculate 1 L of NZCYM medium with KAN25 and CHL20 to reach an initial OD~600~ around 0.025. The cells were grown at 37°C at 220 rpm to an OD~600~ of 1.0 and then cooled to room temperature (r.t.). Isopropyl-β-D-thiogalactopyranoside (IPTG) was added to a final concentration of 0.5 mM, and the cells were grown for an additional 6 h at r. t. or 20 h at 16°C before harvesting. Purification of DENV NS5 and its variants {#sec016} ----------------------------------------- Cell lysis, subsequent purification and storage procedures were as previously described in the JEV NS5 study \[[@ppat.1008484.ref008]\], except that the centrifugation duration to remove cell debris was 1 h, a 50 mM imidazole wash was applied prior to the elution step of the nickel-affinity chromatography, and 5 mM Tris (pH 7.5) was used as the buffering agent in the gel filtration chromatography. Tris-(2-carboxyethyl)phosphine (TCEP) was added to the pooled fractions to a final concentration of 5 mM. The molar extinction coefficient for the DENV NS5 was calculated based on protein sequence using the ExPASy ProtParam program (<http://www.expasy.ch/tools/protparam.html>). The yield is typically 2 mg of pure protein per L of bacterial culture. Protein crystallization and crystal harvesting {#sec017} ---------------------------------------------- The DENV NS5 crystals were grown by sitting drop vapor diffusion at 10 or 16°C using 10--12 mg/mL protein sample. Crystals grew to its final dimension in about 3 weeks with a precipitant/well solution containing 1.7% (vol./vol.) dioxane, 0.085 M bicine (pH 8.8), 4.9% (wt./vol.) PEG2000, and 15% (vol./vol.) glycerol for crystal form 1, and 0.2 M NH~4~I, 8% (vol./vol.) Tacsimate (pH 6.1), and 20% (wt./vol.) PEG3350 for crystal form 2. Crystals were transferred into cryo-stablizer solutions and stored in liquid nitrogen prior to data collection. Crystallographic data processing and structure determination {#sec018} ------------------------------------------------------------ Diffraction datasets were collected at the Shanghai Synchrotron Radiation Facility (SSRF) beamlines BL17U1 (crystal form 1, wavelength 0.9792 Å) and BL18U1 (crystal form 2, wavelengths: 0.9788 Å) at 100 K. Typically, at least 180° of data were collected in 0.4--0.5° oscillation steps. Reflections were integrated, merged, and scaled using HKL2000 or D\*Trek \[[@ppat.1008484.ref054],[@ppat.1008484.ref055]\]. The initial structure solution for crystal form 1 was obtained using the molecular replacement program PHASER \[[@ppat.1008484.ref056]\] and separated MTase and RdRP ensembles derived from JEV and DENV3 NS5 structures (PDB entries 4K6M and 4V0Q) \[[@ppat.1008484.ref008],[@ppat.1008484.ref020]\]. The final model of crystal form 1 was split into three ensembles (MTase, RdRP thumb with index tip, and the rest of RdRP) in the molecular replacement trial to obtain the initial structure solution for crystal form 2. Manual model building and structure refinement were done using Coot and PHENIX, respectively \[[@ppat.1008484.ref057],[@ppat.1008484.ref058]\]. The 3,500 K composite simulated-annealing omit 2F~o~-F~c~ electron density maps were generated using CNS \[[@ppat.1008484.ref059]\]. All NS5 superimpositions were done using the maximum likelihood based structure superpositioning program THESEUS \[[@ppat.1008484.ref026]\]. Relative domain motions were analyzed by DynDom \[[@ppat.1008484.ref060]\]. The occlusion of the solvent accessible area by the MTase-RdRP interactions was analyzed by program SurfRace with a probe radius of 1.4 Å and the 10-residue linker excluded in the calculation \[[@ppat.1008484.ref061]\]. The multiple sequence alignment of NS5 was carried out using 47 available complete NS5 sequences among the flavivirus species documented by the International Committee on Taxonomy of Viruses (ICTV) ([http://www.ictvonline.org](http://www.ictvonline.org/)), and the alignment was then used to generate the sequence logos ([http://weblogo.berkeley.edu](http://weblogo.berkeley.edu/)) and the conservation score projected structural representations. The projection of the conservation score onto structural model was done by the ConSurf server \[[@ppat.1008484.ref062]\]. Cells and antibodies {#sec019} -------------------- BHK-21 cells (American Type Culture Collection (ATCC), CCL-10) was propagated in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 units/mL of penicillin and 100 μg/mL of streptomycin in 5% CO~2~ at 37°C. Monoclonal antibody (4G2) against envelope protein of flavivirus was used to detect viral protein expression of both JEV and DENV2. FITC-conjugated goat anti-mouse IgG was used as secondary antibody. Construction of mutant genome-length cDNA of JEV and DENV2 {#sec020} ---------------------------------------------------------- The infectious clones of pACYC-JEV-SA14 \[[@ppat.1008484.ref063]\] and pACYC-DENV2-TSV \[[@ppat.1008484.ref033]\] were used as the backbone to construct recombinant JEV and DENV2 with different NS5 mutations, respectively. All mutations were engineered by fusion PCR. The JEV NS5 mutations were engineered at *BamH*I and *Xba*I restriction sites of pACYC-JEV-SA14. All DENV2 NS5 mutations were inserted into to pACYC-DENV2-TSV by restriction digestion with *Nru*I and *Cla*I. All constructs were verified by DNA sequencing before they were used in the subsequent experiments. Recombinant JEV and DENV2 study with different NS5 mutations {#sec021} ------------------------------------------------------------ The infectious clone of JEV and DENV2 with corresponding NS5 mutations were linearized with *Xho*I and *Cla*I, respectively, and then subjected to *in vitro* transcription using a T7 *in vitro* transcription kit (Thermo Fisher Scientific). Approximately 1 μg of transcribed recombinant genomic RNAs were transfected into BHK-21 cells with reagent DMRIE-C (Invitrogen). Then the cell slides were fixed in cold (-20°C) 5% (vol. to vol.) acetone in methanol at r. t. for 10 min. After washing three times with phosphate buffer saline (PBS) (pH 7.4), the fixed cells were subjected to IFA with 4G2 monoclonal antibody to examine viral envelope expression of both JEV and DENV2. At the same time, the supernatants of RNA-transfected BHK-21 cells were harvested as viral stocks for plaque assay to quantify viral titers and examine plaque morphologies. Briefly, confluent BHK-21 cells (1×10^5^ cells per well, plated 1 day in advance) in 24-well plates were infected with serially 10-fold diluted viral supernatants and incubated at 37°C with 5% CO~2~ for 1 h before the layer of medium containing 1% methylcellulose was added. After 4 days of incubation at 37°C with 5% CO~2~, the cells were fixed in 3.7% formaldehyde and then stained with 1% crystal violet. The viral titer was calculated as plaque formatting unit (PFU) per mL. RNA preparation {#sec022} --------------- The 30-mer template RNA (T30, [Fig 5A](#ppat.1008484.g005){ref-type="fig"}) used for *de novo* polymerase assays was chemically synthesized (Integrated DNA Technologies or Dharmacon) and purified by 12% (wt./vol.) polyacrylamide/7 M urea gel electrophoresis. The target RNA was excised from the gel, electro-eluted by using an Elu-Trap device (GE Healthcare), ethanol precipitated, dissolved in an RNA annealing buffer (RAB: 50 mM NaCl, 5 mM Tris (pH 7.5), 5 mM MgCl~2~), and stored at -80°C after a self-annealing process (a 3-min incubation at 95°C followed by snap-cooling to minimize inter-molecular annealing). A 5′-phosphorylated dinucleotide primer pGG (P2) (Jena Biosciecnes) was mixed with T30 at 5:1 or 20:1 molar ratio to make the T30/P2 construct. *In vitro* polymerase assays {#sec023} ---------------------------- All *in vitro* polymerase assays were based on the dinucleotide (P2)-driven reactions. The standard reaction condition was derived from the JEV NS5 work with the 4 μM of the T30 RNA and 6 μM of NS5 and with a couple of adjustments \[[@ppat.1008484.ref012]\]. Firstly, a manganese-free reaction condition (50 mM Tris (pH 7.5), 20 mM NaCl, 5 mM MgCl~2~, 5 mM DTT) (hereinafter referred as reaction buffer) was established. Secondly, a 5′-phosphorylated dinucleotide (pGG) and the T30 RNA was mixed at a difference ratio. In the assay to characterize the conversion of P2 to the 3-mer product (P3), ATP was supplied as the only NTP substrate and a high P2:T30 ratio (20:1) was used to achieve multiple turnovers within a reasonable duration. For the P9-containing EC (EC9) formation assay to achieve the P2-to-P9 conversion, ATP and UTP were supplied and the P2:T30 ratio was 5:1. For the P9-to-P10 single nucleotide elongation assay, reactions were first carried out as described in the EC9 formation assay. The reaction mixtures were centrifuged at 16,000 g for 5 min, and the pellet was washed twice by the reaction buffer, and was then resuspended in a modified reaction buffer with NaCl concentration lifted to 200 mM. CTP was supplemented to the resuspended mixture to allow the single-nucleotide elongation at 30°C, and the reaction was quenched immediately ("0" min) or at a certain time point after the addition of CTP. The concentrations of NaCl and CTP in the final reaction mixture are 200 mM and 300 μM, respectively, unless otherwise indicated. For all assays, the procedures for denaturing polyacrylamide gel electrophoresis (PAGE), gel staining, and quantitative analyses were performed as previously described \[[@ppat.1008484.ref012]\]. The Stains-All (Sigma-Aldrich)-based staining method is reasonably accurate when quantitating RNA bands with the same length, and in the majority of our experiments we tried to keep the band intensity within the linear range estimated in the previous study by adjusting the range of reaction time points \[[@ppat.1008484.ref012]\]. For the P2-to-P3 conversion assay, 50, 100, 200, 500, and 1000 μM ATP concentrations were used for WT and M-67A/68A, and 200, 400, 600, 1000, and 1500 μM were used for R3. To account for gel-to-gel intensity variations, samples from the same reaction mixture (indicated by the same icon above corresponding lanes in [Fig 7A--7C](#ppat.1008484.g007){ref-type="fig"}) were loaded on different gels to normalize the intensities (e.g. lanes 4/17, 14/27, 24/37 in [Fig 7A](#ppat.1008484.g007){ref-type="fig"}). The normalized intensity was then used to calculate the relative reaction rates ([Fig 7D--7F](#ppat.1008484.g007){ref-type="fig"}, left), which in turn were fitted to the Michaelis-Menten equation ([Fig 7D--7F](#ppat.1008484.g007){ref-type="fig"}, right). To estimate the relative *k*~*cat*~ and specificity constant values of all three constructs, The Michaelis-Menten curves of all three constructs were normalized based on the measurement of the relative reaction rates of each WT-mutant pair in the same gel ([Fig 7G](#ppat.1008484.g007){ref-type="fig"}). For the stability assessment of EC9, the P2-to-P9 conversion was first carried out as described above. The reaction mixture was subsequently centrifuged at 16,000 g for 5 min, and the pellet was washed twice by the reaction buffer and was then resuspended in a modified reaction buffer with NaCl concentration lifted to 500 mM. The resuspension was incubated in 37°C for various duration (0--78 h), and then CTP was added to reach a final concentration of 300 μM to trigger the P9-to-P10 conversion by EC9 that survived the incubation. The fraction of 10-mer intensity values were fitted to a single-exponential decay model to estimate the inactivation rates of all three constructs. Supporting information {#sec024} ====================== ###### A modelled NS5 conformational transition from the JEV-mode to the DENV3-mode. The movie starts with the JEV NS5 structure (PDB entry 4K6M), switches to the first form of DENV2 structure (PDB entry 6KR2), then the second form of DENV2 structure (PDB 6KR3), and finally the DENV3 structure (PDB entry 4V0Q). Three rotation axes related to the MTase movement are indicated by black lines, and one axis related to the RdRP thumb movement is indicated by a grey line. Note that the non-superimposable regions are not included in the structural models. (MP4) ###### Click here for additional data file. We thank Dr. Pei-Yong Shi for providing the cloning material for the DENV2 NS5 gene, Dr. Xiao-Dan Li for construction of the DENV2 NS5 M3 mutant plasmid, Dr. Bo Shu for X-ray diffraction data collection, Liu Deng and Yancheng Zhan for laboratory assistance, Dr. Yunhuang Yang for helpful discussions, synchrotron SSRF (beamlines BL17U and BL18U1 Shanghai, China) for access to beamlines, and The Core Facility and Technical Support, Wuhan Institute of Virology, for access to instruments. 10.1371/journal.ppat.1008484.r001 Decision Letter 0 Diamond Michael S. Section Editor Rey Félix A. Associate Editor © 2020 Diamond, Rey 2020 Diamond, Rey This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. 25 Feb 2020 Dear Dr. Gong, Thank you very much for submitting your manuscript \"A conformation-based intra-molecular initiation factor identified in the flavivirus RNA-dependent RNA polymerase\" for consideration at PLOS Pathogens. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by two independent reviewers. As you will see, the reviewers provided contradictory reports. After careful consideration, we have editorially sided with reviewer 2.   Flavivirus NS5 is a multifunctional protein, composed of an N-terminal methyl-transferase (MTase) domain and a larger RNA-dependent RNA polymerase (RdRp) C-terminal domains. Multiple structures have been reported previously, some identifying a functional dimeric complex. As a polymerase, the NS5 protein has to adopt several conformation, which are different for initiation of RNA synthesis, for capping the newly synthesized RNA, or for elongation of the latter. In your study, you obtained crystals of DENV2 NS5 adopting two different conformations : one that resembles a structure reported for NS5 from the Japanese encephalitis flavivirus (JEV), in which the MTase domain adopts a "closed conformation" with respect to the RdRp domain, and another in which it adopts an \"open conformation\", similar to the one seen before in a structure of NS5 from the dengue virus serotype 3 (DENV3). As the two MTase/RdRp interfaces are different,  site directed mutagenesis allowed the selective targeting of the two interfaces. The use of these mutants for functional studies in an assay that allows to distinguish RNA initiation from RNA elongation showed that mutations at the interface seen in the open form interfere with the initiation process but do not affect the elongation process. Although there was no effect of the mutants at the interface observed in the closed form, it was considered editorially that it  is the first time that a given conformation of NS5 can be clearly attributed to a specific step in the replication process, and even though no MTase studies were performed, as pointed out by reviewer 1, that this is still a meaningful study. Please prepare and submit your revised manuscript within 30 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email.  When you are ready to resubmit, please upload the following: \[1\] A letter containing a detailed list of your responses to all review comments, in particular to those raised by reviewer 2, but also, whenever possible, to those raised by reviewer 1, together with a description of the changes you have made in the manuscript.  Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out \[2\] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file). Important additional instructions are given below your reviewer comments. Thank you again for your submission to our journal. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don\'t hesitate to contact us if you have any questions or comments. Sincerely, Félix A. Rey Associate Editor PLOS Pathogens Michael Diamond Section Editor PLOS Pathogens Kasturi Haldar Editor-in-Chief PLOS Pathogens ​[orcid.org/0000-0001-5065-158X](http://orcid.org/0000-0001-5065-158X) Michael Malim Editor-in-Chief PLOS Pathogens [orcid.org/0000-0002-7699-2064](http://orcid.org/0000-0002-7699-2064) \*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\* The paper describes the X-ray structure of the NS5 protein of Dengue virus serotype 2. Flavivirus NS5 is a multifunctional protein, composed of an N-terminal methyl-transferase (MTase) domain and a larger RNA-dependent RNA polymerase (RdRp) C-terminal domains. Multiple structures have been reported previously, some identifying a functional dimeric complex. As a polymerase, this protein has to adopt several conformation, which are different for initiation of RNA synthesis, for capping the newly synthesized RNA, and for elongation of the later. Here, the authors have identified two different conformations of NS5 in their crystals: one that resembles a structure reported for NS5 from the Japanese encephalitis flavivirus (JEV), in which the MTase domain adopts a "closed conformation" with respect to the RdRp domain, and another in which it adopts an open conformation, similar to the one seen before in a structure of NS5 from the dengue virus serotype 3 (DENV3). What I found novel is that, as the two MTase/RdRp are different, they used directed mutagenesis to alter the interfaces in the closed or in the open conformation. They used these mutants for functional studies in an assay that allowed them to distinguish RNA initiation from RNA elongation. They found that mutations in the interface seen in the open form interfere with the initiation process but do not alter the elongation process. Although they didn't found a phenotype for the mutants at the interface observed in the closed form, this is the first time that a given conformation of NS5 can be assigned to a particular step in the replication process, and even though no MTase studies were performed, as pointed out by reviewer 1, I found that the authors have still presented a meaningful study. Reviewer Comments (if any, and for reference): Reviewer\'s Responses to Questions **Part I - Summary** Please use this section to discuss strengths/weaknesses of study, novelty/significance, general execution and scholarship. Reviewer \#1: In this manuscript, Wu et al. report the crystal structures Dengue virus type 2 (DENV2) NS5 and show that it exists in two distinct conformations representing both of the previously reported structures of Japanese encephalitis virus (JEV) and DENV3 NS5. These are defined as JEV-mode and DENV3-mode due to closeness of these DENV2 structures to the JEV and DENV3 NS5 structures. Moreover, the authors present additional data by mutational analysis to distinguish the functional differences between the two conformations of DENV2 NS5. They performed in vitro polymerase assays with wild-type (WT) and site-specific NS5 mutants to distinguish the initiation and elongation steps of RNA synthesis as well as cell-based replication using immunofluorescence assays. The authors conclude that the methyltransferase (MTase) of NS5 serves as a unique initiation factor only through NS5 conformation in the JEV-mode. But no MTase assays were performed. Overall, the crystallographic work is well-done. However, their biochemical and virological assays raise several questions. The functional roles of these two conformational states of DENV2 NS5 in viral replication remains unclear. Moreover, another group reported recently that DENV2 NS5 exists in two distinct conformational states (El Sahili, Soh, Schiltz, Gharbi-Ayachi, She, Shi, Lim, Lescar, 2020. J. Virol. 94: e01294-19). These authors in El Sahili et al. provide convincing evidence that it is the inter-domain linker between the MTase and POL domains the functional determinant of the two conformational states of DENV2 NS5. Reviewer \#2: The flaviviral NS5 proteins contain a methyltransferase (MTase) domain followed by an RNA-dependent RNA polymerase (RdRP) domain, both of which have essential functions during viral replication. Structures of the individual domains have been known for over a decade and biochemical approaches have been applied to studying the seemingly independent activities of these domains; this has worked moderately well for the MTase, but not so well for the RdRP due to the presence of a priming loop on the thumb domain that limits RNA binding and has precluded detailed biochemical studies of elongation kinetics the solving of an active elongation complex structure. Over the past few years that have several structure of full length NS5 solved from multiple flaviviruses (JEV, ZIKV, DENV3), and interestingly these have shown different interactions between the MTase and RdRP domains. However, it has not yet been clear if the observed inter-domain interactions are biologically relevant or due to crystal packing effects. The work presented in this manuscript provides a thorough study of MTase-RdRP interactions using structural biology, biochemistry, and virus replication data to address functional roles of the flaviviral methyltransferase (MTase) during the initiation of replication by the RdRP domain. Two new structures of DENV2 NS5 are solved that capture domain orientations which are intermediate between those solved previously for DENV3 and JEV. Structure-based mutagenesis is then used to show that one specific mode of MTase-RdRP interaction (the JEV mode) is important for both the growth of infectious virus and for in vitro initiation by the RdRP domain. These new results provide key insights into understanding the molecular mechanisms whereby flaviviral NS5 initiates replication. Using initiation of a pGG dinucleotide to form pGGA or to extent to a longer 9mer when UTP is also added, the authors use mutations of the show that formation of both products is diminished upon mutation of the JEV-mode interface but not the DENV3 interface, indicating that the JEV-mode is the active form. Furthermore, the researchers demonstrate that mutating the MTase interaction only affects RdRP initiation to form short products (pGG to pGGA), and does not affect on the elongation of a longer 9mer product to a 10mer. This latter elongation step is demonstrated to be very rapidly done by a complex that is stable in high salt, showing that the authors have assembled authentic stalled elongation complexes form a flaviviral polymerase. This is a major breakthrough that sets the stage for future studies of both RdRP biochemistry and elongation complex structure. The core experiments and conclusions are solid and well presented in the manuscript and there are no major concerns with the interpretation of the results. As usual from Gong lab, the illustrations are excellent and the polymerase biochemistry is well done with rigorous interpretations, all the more notable in this work because of the extensive steps taken to normalize Stains-All data across multiple gels. However, there are a few points where additional clarification or grammar corrections would be appreciated, as outlined below. Line 146-149: It may be worth discussing that the ring finger and part of index finger are not resolved in many of the early flaviviral RdRP domain structures (2hcn, 2hfz, 2j7u, 2j7w... see the flaviviral structures at DOI: 10.5281/zenodo.3361874). Might this be because the MTase domain is needed to stabilize the fingers folding, which could have big effects on RNA binding and initiation? The section from lines 162-167 describing SAH binding site is weak and the evidence for an "allosteric" site sounds somewhat circumstantial. Is there any experimental evidence for SAH binding being important, or is it perhaps a crystallization artifact? The section should either be strengthened with experimental data or references, or the idea of "allosteric" should be removed and the observed SAH density instead be described as an interesting structural observation that merits further biochemical validation. Also, was SAH added to the crystallization conditions, or did this SAH co-purify with the protein much as the SAH bound to the MTase domain does? Is there full occupancy of this "allosteric" site and of the native MTase site? Is there any (weak) anomalous density from the sulfur that would definitively identify the SAH? Please explicitly state the actual concentration of polymerase, RNA template and pGG primer used in the biochemical experiments. These are not stated anywhere in the manuscript, where only mole ratios are listed. \*\*\*\*\*\*\*\*\*\* **Part II -- Major Issues: Key Experiments Required for Acceptance** Please use this section to detail the key new experiments or modifications of existing experiments that should be [absolutely]{.ul} required to validate study conclusions. Generally, there should be no more than 3 such required experiments or major modifications for a \"Major Revision\" recommendation. If more than 3 experiments are necessary to validate the study conclusions, then you are encouraged to recommend \"Reject\". Reviewer \#1: In this manuscript, Wu et al. report the crystal structures Dengue virus type 2 (DENV2) NS5 and show that it exists in two distinct conformations representing both of the previously reported structures of Japanese encephalitis virus (JEV) and DENV3 NS5. These are defined as JEV-mode and DENV3-mode due to closeness of these DENV2 structures to the JEV and DENV3 NS5 structures. Moreover, the authors present additional data by mutational analysis to distinguish the functional differences between the two conformations of DENV2 NS5. They performed in vitro polymerase assays with wild-type (WT) and site-specific NS5 mutants to distinguish the initiation and elongation steps of RNA synthesis as well as cell-based replication using immunofluorescence assays. The authors conclude that the methyltransferase (MTase) of NS5 serves as a unique initiation factor only through NS5 conformation in the JEV-mode. But no MTase assays were performed. Overall, the crystallographic work is well-done. However, their biochemical and virological assays raise several questions. The functional roles of these two conformational states of DENV2 NS5 in viral replication remains unclear. Moreover, another group reported recently that DENV2 NS5 exists in two distinct conformational states (El Sahili, Soh, Schiltz, Gharbi-Ayachi, She, Shi, Lim, Lescar, 2020. J. Virol. 94: e01294-19). These authors in El Sahili et al. provide convincing evidence that it is the inter-domain linker between the MTase and POL domains the functional determinant of the two conformational states of DENV2 NS5. Specific comments 1\. Line 71, p. 4: This study focuses only on the polymerase activity of NS5 although the interdomain interaction between MTase and POL domains could impact on both enzyme activities. Why the authors did not perform any MTase assays to study the impact of the two conformations of DENV2 NS5? 2\. Lines 99-101, p. 5: "This statement is not correct. Once initiated de novo, RdRp can continue RNA synthesis until the template is copied. This has been demonstrated for dengue virus type 2 RdRp in many studies using subgenomic RNA templates containing 5'- and 3'-terminal regions and conserved cis-active elements that have been shown to be important for both initiation and elongation steps (see refs. Ackermann and Padmanabhan, 2001; Niyomrattanakit et al. 2010). The statement on line 101 is not accurate because in one study, processivity of RdRp was demonstrated by including heparin after initiation complex is formed to block re-initiation events (Nomaguchi et al. 2003). The statement on line 102 needs to modified to indicate that both Mg++ and Mn++ were used (in refs. 18, 24 and 25 cited in this study). 3\. The effects of mutations on MTase activity were not studied whereas these were examined in ref. 18 cited in this study. 4\. Lines 204, 208, 264: There is no Fig. 3E included in the manuscript. 5\. Lines 235-239 and lines 239-242: Please cite the ref(s) here; 12 or 33?. 6\. One concern with this study is the use of artificial GG dinucleotide primer and template for RdRp initiation and elongation assays. The authors' definition of de novo initiation in mosquito-borne RdRp assays is unconventional and may lead to wrong conclusions regarding the mutational effects in RdRp activities. For example, the catalytic rate constant could not be accurately determined under their conditions (lines 252-257). Reviewer \#2: None \*\*\*\*\*\*\*\*\*\* **Part III -- Minor Issues: Editorial and Data Presentation Modifications** Please use this section for editorial suggestions as well as relatively minor modifications of existing data that would enhance clarity. Reviewer \#1: Minor Comments: 1\. Line 144: Which DENV3 structures the authors are referring to here? It would be helpful if the authors label the motif G in Fig. 1. 2\. Lines 160-162: The SAH binding site has been observed and labeled in Klema et al. (ref. 21). 3\. Define "pt" ( line 229) andNS5B NTD (line 403). Reviewer \#2: Line 40: calling differences in prior structures "drastically different global folds" is perhaps an overstatement since the effects are really at the level of relative domain orientations and not protein folding. Line 51: remove "about" as it redundant with giving a range of 10-11 kb for genome length. Line 71: change are to is Line 72: make evidences singular Line 76: typo in hereafter Line 86: in discussing interface areas, please explicitly state if the values presented reflect total buried solvent accessible surface from both sides of the interface, or from only one side of the interface. Also on lines 138, 157 and others. Line 115: Consider deleting "apparently" as the data for this are convincing. Line 120: change in to to. Line 127: arrange should be in past tense (arranged). Line 301: The same high-salt challenge studies are also used in studies of picornaviral polymerases, where they led to the crystallization of elongation complexes. Line 312-316: What really needs to the optimized here is the amount of material loaded on the gel, not necessarily what was used in the reactions, which could be done at higher concentrations and then diluted prior to gel loading (a minor conceptual point, no need to edit the text). Line 316: Change 3-fold to "3-fold higher" Line 375: I do not agree with the statement that the global structural organization is quite different from flaviviral polymerases versus others. Structure comparisons have shown the core polymerase fold and structure to be very similar across all viral polymerases, and certainly among positive strand RNA viruses, with few insertions or topology changes within the core fold. There are added domains, such as the N-terminal MTase and NiRAN domains or C-terminal membrane anchors, but these are additions to the protein and not a reorganization of the protein fold as implied by the manuscript as written. Line 394: I do believe there is convincing evidence that the N-terminal region of nsp9 is a nucleotidylating NiRAN domain. Line 518: Include actual concentrations of macromolecules used in the experiments. Figure 3: Are the stereo panels in A and B flipped right to left? Stereo viewing of these images does not seem to work correctly. Figure 5: In panel D, why is the 9mer STD band higher than the 9mer product band from the polymerase? Effect of the 5' phosphate from pGG primer, perhaps? Line 816: Consider changing to "...in the multiple turnover P3 formation and single turnover P9 formation assays" so as to explain the vast difference in band intensities in the figure legend itself (in addition to the comments already in the text of the manuscript). Line 833: Technically speaking, the experiment in really testing inactivation of the elongation complex and not the dissociation rate of the RNA, although that is probably a valid interpretation in this case. Figure 7: In some cases, linear curve fits to determine initial rates are being applied to data that have clear curvature (panel E-200uM, panel F-50 & 100 uM, panel G-R3). This is a minor point. \*\*\*\*\*\*\*\*\*\* PLOS authors have the option to publish the peer review history of their article ([what does this mean?](https://journals.plos.org/plospathogens/s/editorial-and-peer-review-process#loc-peer-review-history)). If published, this will include your full peer review and any attached files. If you choose "no", your identity will remain anonymous but your review may still be made public. **Do you want your identity to be public for this peer review?** For information about this choice, including consent withdrawal, please see our [Privacy Policy](https://www.plos.org/privacy-policy). Reviewer \#1: No Reviewer \#2: No [Figure Files:]{.ul} While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, [[https://pacev2.apexcovantage.com](https://pacev2.apexcovantage.com/)]{.ul}. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at [<figures@plos.org>]{.ul}. [Data Requirements:]{.ul} Please note that, as a condition of publication, PLOS\' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example see here: <http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5>. [Reproducibility:]{.ul} To enhance the reproducibility of your results, PLOS recommends that you deposit laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see [[http://journals.plos.org/plospathogens/s/submission-guidelines\#loc-materials-and-methods](http://journals.plos.org/plospathogens/s/submission-guidelines)]{.ul} 10.1371/journal.ppat.1008484.r002 Author response to Decision Letter 0 27 Feb 2020 ###### Submitted filename: point_by_point_responses.docx ###### Click here for additional data file. 10.1371/journal.ppat.1008484.r003 Decision Letter 1 Diamond Michael S. Section Editor Rey Félix A. Associate Editor © 2020 Diamond, Rey 2020 Diamond, Rey This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. 18 Mar 2020 Dear Dr. Gong, We are pleased to inform you that your manuscript \'A conformation-based intra-molecular initiation factor identified in the flavivirus RNA-dependent RNA polymerase\' has been provisionally accepted for publication in PLOS Pathogens. Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests. Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated. IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript. Should you, your institution\'s press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS. Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Pathogens. Best regards, Félix A. Rey Associate Editor PLOS Pathogens Michael Diamond Section Editor PLOS Pathogens Kasturi Haldar Editor-in-Chief PLOS Pathogens ​[orcid.org/0000-0001-5065-158X](http://orcid.org/0000-0001-5065-158X) Michael Malim Editor-in-Chief PLOS Pathogens [orcid.org/0000-0002-7699-2064](http://orcid.org/0000-0002-7699-2064) \*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\* Reviewer Comments (if any, and for reference): 10.1371/journal.ppat.1008484.r004 Acceptance letter Diamond Michael S. Section Editor Rey Félix A. Associate Editor © 2020 Diamond, Rey 2020 Diamond, Rey This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. 24 Apr 2020 Dear Dr. Gong, We are delighted to inform you that your manuscript, \"A conformation-based intra-molecular initiation factor identified in the flavivirus RNA-dependent RNA polymerase,\" has been formally accepted for publication in PLOS Pathogens. We have now passed your article onto the PLOS Production Department who will complete the rest of the pre-publication process. All authors will receive a confirmation email upon publication. The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any scientific or type-setting errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript. Note: Proofs for Front Matter articles (Pearls, Reviews, Opinions, etc\...) are generated on a different schedule and may not be made available as quickly. Soon after your final files are uploaded, the early version of your manuscript, if you opted to have an early version of your article, will be published online. The date of the early version will be your article\'s publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers. Thank you again for supporting open-access publishing; we are looking forward to publishing your work in PLOS Pathogens. Best regards, Kasturi Haldar Editor-in-Chief PLOS Pathogens ​[orcid.org/0000-0001-5065-158X](http://orcid.org/0000-0001-5065-158X) Michael Malim Editor-in-Chief PLOS Pathogens [orcid.org/0000-0002-7699-2064](http://orcid.org/0000-0002-7699-2064) [^1]: The authors have declared that no competing interests exist.

1. Introduction =============== The synthesis of metal sulfide nanocrystals has attracted much interest for both fundamental research and technological applications in the past decades \[[@B1-nanomaterials-02-00113],[@B2-nanomaterials-02-00113],[@B3-nanomaterials-02-00113],[@B4-nanomaterials-02-00113],[@B5-nanomaterials-02-00113],[@B6-nanomaterials-02-00113]\]. Metal sulfide nanocrystals have been prepared by a wide range of synthetic methods, one of which involves the direct decomposition of molecular precursors \[[@B7-nanomaterials-02-00113],[@B8-nanomaterials-02-00113],[@B9-nanomaterials-02-00113],[@B10-nanomaterials-02-00113],[@B11-nanomaterials-02-00113],[@B12-nanomaterials-02-00113],[@B13-nanomaterials-02-00113],[@B14-nanomaterials-02-00113]\]. Molecular precursor approach has recently been developed as an efficient route to prepare monodispersed semiconductor nanocrystals \[[@B7-nanomaterials-02-00113],[@B10-nanomaterials-02-00113],[@B15-nanomaterials-02-00113]\] and, in some cases, unique shape-control has been achieved \[[@B9-nanomaterials-02-00113],[@B11-nanomaterials-02-00113],[@B16-nanomaterials-02-00113]\]. One of the earliest precursors used for preparing metal sulfides in the literature is *N*,*N*'-dialkyl dithiocarbamate. In this preparation, the precursor was injected into hot coordination solvents such as trioctylphosphine oxide (TOPO) under nitrogen atmosphere at high temperatures \[[@B7-nanomaterials-02-00113],[@B8-nanomaterials-02-00113],[@B9-nanomaterials-02-00113],[@B10-nanomaterials-02-00113],[@B11-nanomaterials-02-00113]\]. Metal bis(benzylthiolates) \[[@B12-nanomaterials-02-00113]\] and metal salts of alkylxanthate \[[@B17-nanomaterials-02-00113],[@B18-nanomaterials-02-00113],[@B19-nanomaterials-02-00113]\] have also been used as precursors to prepare nanocrystalline sulfides of zinc, cadmium or lead via pyrolysis at 150--400 °C. It was found that Lewis base such as hexydecylamine (HDA), trioctylphosphine (TOP) or tributylphosphine (TBP) could lower the reaction temperature for the alkylxanthate precursors. Recently, there are also several reports on the preparation of ternary and metal sulfide nanocrystals via the decomposition of thiocarboxylate precursors \[[@B20-nanomaterials-02-00113],[@B21-nanomaterials-02-00113],[@B22-nanomaterials-02-00113],[@B23-nanomaterials-02-00113],[@B24-nanomaterials-02-00113]\]. Most of these syntheses, nevertheless, require elevated or refluxing temperatures. In this paper, we report a generalized precursor method operating at room temperature for the synthesis of various metal sulfide nanocrystals. The precursors we use are metal thiobenzoates \[M~x~(SCOC~6~H~5~)~y~ or simply MTB\]. These MTB precursors are air-stable and could be readily prepared from thiobenzoic acid and the corresponding metal salts following the known literature method \[[@B25-nanomaterials-02-00113]\]. We illustrate the generality of this MTB method by preparing four types of semiconductors from both transition and main group metals: Ag~2~S, Cu~2−x~S, In~2~S~3~ and CdS. We discuss in this report the basis of our approach and also attempt to understand the reaction mechanism through theoretical Density Functional Theory (DFT) calculations. With this insightful knowledge, we demonstrate how to manipulate the stability of the precursor, and thus the reaction kinetics, to generate different nanostructures. 2. Experimental Section ======================= 2.1. Synthesis of Precursors and Metal Sulfide Nanocrystals ----------------------------------------------------------- Commercially available compounds such as thiobenzoic acid (Fluka), ether, ethanol, chloroform (all from J. T. Baker), sodium bicarbonate (Dumont), silver nitrate (Merck), indium chloride (Fluka), acetonitrile, cadmium acetate, copper chloride, 2,2'-bipyridine, octylamine, dodecylamine and oleylamine (all from Aldrich) were used as received. All of the MTB precursors used in this paper (M = Ag, Cu, In, Cd; TB = thiobenzoate) were prepared according to literature methods \[[@B25-nanomaterials-02-00113]\]. The products obtained were washed with ethanol, dried, and recrystallized from chloroform or ether. Purity of the crystals was checked with microanalysis and their decomposition profiles were investigated by thermogravimetric analysis (TGA). For the synthesis of silver sulfide nanoparticles, AgTB (3 mmol) was stirred in toluene (5 mL) at room temperature, and then added with 1.8 mmol of an amine. A homogeneous clear brown solution formed quickly and the solution was stirred for a further 3 hours. After that, 10 mL of ethanol was added to induce turbidity in the mixture. The brown or dark brown nanoparticles are isolated by centrifugation, washed several times with ethanol and acetone, and then dried under vacuum. The resulting powder can be easily re-dispersed in toluene, hexane, chloroform and other non-polar solvents. The experimental procedure is similar for the synthesis of copper sulfide nanoparticles, except that the reaction mixture turned blue and was stirred overnight. Dark brown or black copper sulfide nanoparticles were isolated, which can be re-dispersed in non-polar solvents such as toluene, hexane and chloroform. The synthesis of cadmium sulfide nanoparticles is similar to silver sulfide and copper sulfide, except that the reaction mixture turned yellow and was stirred for 4 hours at room temperature. Pale yellow product was separated by centrifugation after adding 10 mL of ethanol. The product isolated, after washing with ethanol and acetone, can be re-dispersed in chloroform, hexane and toluene. For the synthesis of indium sulfide nanoparticles, InTB (0.3 mmol) was stirred in toluene (5 mL) at room temperature, and then 1.2 mmol of octylamine (OA) was injected to give a yellow solution. After stirring for 6 hours, 10 mL of ethanol was added to induce the formation of turbidity. The particles were purified similarly to the previous procedure. For the preparation of oleylamine-capped In~2~S~3~ nanoparticles, it was found necessary to further add 40 µL of propylamine to speed up the reaction. 2.2. Characterizations ---------------------- TGA was recorded on a SDT 2960 Simultaneous DTA-TGA by heating approximately 10 mg of the precursor under inert N~2~ flow (flow rate 90 mL/min) at a heating rate of 10 °C/min. X-ray diffraction (XRD) analysis was carried out on a Siemens D5005 X-ray powder diffractometer with Cu Kα radiation (40 kV, 40 mA). The powdered sample was mounted on a sample holder and scanned with a step size of 2θ = 0.05° in the range of 20° to 90°. UV-Visible spectra were recorded using a Shimadzu UV-2550 UV/Vis spectrophotometer. FT-IR spectra were recorded using a FTS 165 Bio-Rad FTIR spectrophotometer in the range of 4000--400 cm^−1^ on KBr or nujol mulls. Transmission electron microscopy (TEM) images were obtained on a 100 kV JEM-100CXII TEM and 200 kV JEOL 2010F microscope. Samples were prepared by placing a drop of the dispersed nanoparticles onto a copper grid with carbon film, and were allowed to dry in a desiccator. 2.3. Computational Methodology ------------------------------ Calculations were performed using the hybrid density functional B3LYP \[[@B26-nanomaterials-02-00113],[@B27-nanomaterials-02-00113]\] method with the effective core potential LanL2DZ \[[@B28-nanomaterials-02-00113],[@B29-nanomaterials-02-00113],[@B30-nanomaterials-02-00113],[@B31-nanomaterials-02-00113]\] basis set. All reported energies include zero point energy corrections. All calculations were performed using the Gaussian 98 \[[@B32-nanomaterials-02-00113]\] suite of programs. Optimization was performed without any constraints and the optimized structures were verified to be equilibrium structures or transition states from frequency calculations. An equilibrium structure is characterized by all real frequencies while a transition state has one and only one imaginary frequency. 3. Results and Discussion ========================= 3.1. Thermal Behavior of the MTB Precursors ------------------------------------------- The decomposition profiles of the four MTB precursors were investigated and their TGA curves are presented in [Figure 1](#nanomaterials-02-00113-f001){ref-type="fig"} and [Table 1](#nanomaterials-02-00113-t001){ref-type="table"}. The analysis indicated a clear onset of decomposition at \~ 200--300 °C in each case. The residual weight after each complete decomposition was found to be close to the expected remaining weight of the corresponding metal sulfide. The slightly higher measured values are possibly due to the presence of some nonvolatile carbon-containing side-products \[[@B33-nanomaterials-02-00113]\]. {ref-type="table"}.](nanomaterials-02-00113-g001){#nanomaterials-02-00113-f001} nanomaterials-02-00113-t001_Table 1 ###### Decomposition onset and weight of residue obtained from TGA plots in [Figure 1](#nanomaterials-02-00113-f001){ref-type="fig"}. Precursors Decomposition onset (°C) Weight of residue (%) ------------------------ -------------------------- ----------------------- ------ **AgSCOC~6~H~5~** 172 51.4 50.6 **CuSCOC~6~H~5~** 155 36.5 35.4 **In(SCOC~6~H~5~)~3~** 196 33.6 31.0 **Cd(SCOC~6~H~5~)~2~** 163 38.5 35.9 The decomposition pathway is expected to be similar to typical thiocarboxylate compounds as proposed by Hampden-Smith \[[@B33-nanomaterials-02-00113]\]: M (SCOC~6~H~5~)~x~ → MS~x/2~ + x/2 (C~6~H~5~CO)~2~S   (1) Hence, in solvents such as carboxylic acid and TOPO, these complexes were found to decompose to the corresponding metal sulfides at elevated temperatures. In our studies, nevertheless, we have found that the MTB precursors decompose readily at room temperature by reacting with various alkylamines. In [Section 3.2](#sec3dot2-nanomaterials-02-00113){ref-type="sec"} below, we first characterize the sizes and morphologies of the nanoparticles prepared. We then present the DFT results on our proposed mechanism in [Section 3.3](#sec3dot3-nanomaterials-02-00113){ref-type="sec"}. 3.2. Preparation and Characterization of Metal Sulfide Nanoparticles {#sec3dot2-nanomaterials-02-00113} -------------------------------------------------------------------- ### 3.2.1. Silver Sulfide Nanoparticles Ag~2~S is a semiconductor with a narrow band gap of 1.08 eV \[[@B34-nanomaterials-02-00113]\]. When AgTB precursor was stirred with octylamine (OA) in toluene at room temperature, monodispersed Ag~2~S nanocrystals can be isolated after 3 hours ([Figure 2](#nanomaterials-02-00113-f002){ref-type="fig"}a). Typical XRD pattern of the nanoparticles prepared from AgTB is shown in [Figure 2](#nanomaterials-02-00113-f002){ref-type="fig"}d. The diffraction pattern revealed good monoclinic crystallinity and fitted well with the α-phase of bulk Ag~2~S (JCPDS 14-72). This is known to be the stable silver sulfide phase which commonly exists at room temperature. The same α-phase silver sulfide nanocrystals were obtained when the reaction was carried out using other amines such as dodecylamine (DDA) and oleylamine (OLA). TEM analysis showed that spherical nanoparticles with reasonable size distribution were produced at room temperature in all these cases. The Ag~2~S nanocrystals obtained from different amines, however, are slightly different in sizes as shown by the size histograms in [Figure 2](#nanomaterials-02-00113-f002){ref-type="fig"}a--c. Thus, average particle diameters of 9.2 ± 1.9, 8.3 ± 1.5 and 7.5 ± 0.9 nm are obtained for reaction with OA, DDA and OLA respectively. In conclusion, the Ag~2~S particle sizes can be controlled by varying the chain length of the alkylamine used, and longer-chain amines tend to produce smaller-sized particles. {#nanomaterials-02-00113-f002} Typical XPS spectra of the Ag~2~S nanocrystals are shown in [Figure 3](#nanomaterials-02-00113-f003){ref-type="fig"}, with binding energies corrected with reference to the C 1s peak at 284.7 eV. The doublet arising from Ag 3d~5/2~ and 3d~3/2~ was detected at 368.0 and 374.1 eV respectively, while the S 2p photoelectron peak appears at 161.7 eV. These values are close to those of bulk Ag~2~S \[[@B35-nanomaterials-02-00113]\]. There is no O 1s peak (531.0 eV) detected on the spectrum, indicating that there is no by-product such as Ag~2~SO~4~ (368.3 eV) or Ag~2~O (368.4 eV) produced. Peak area analysis of the Ag 3d~5/2~ and S 2p peaks, after accounting for elemental sensitivity factors, gives an elemental ratio of 1.97:1 for Ag to S. {#nanomaterials-02-00113-f003} ### 3.2.2. Copper Sulfide Nanoparticles Similarly, we found that CuTB precursors readily decompose by reacting with alkylamines at room temperature to give uniform Cu~2−x~S nanoparticles. TEM analysis ([Figure 4](#nanomaterials-02-00113-f004){ref-type="fig"}a--c) indicated average diameters of 8.1 ± 1.1, 6.1 ± 0.5 and 5.8 ± 0.4 nm for the nanoparticles produced from OA, dioctylamine (DOA) and OLA, respectively. It is well known that copper sulfides exist in many different phases and compositions (Cu~x~S, x: 1 → 2). Non-stoichiometric copper sulfide is readily formed and has been utilized as superionic conductor or p-type semiconductor. XPS peak area analysis gave an elemental Cu to S ratio of 1.72:1 for our samples, thus corresponding to x \~0.28 in the Cu~2−x~S general formula. The XRD pattern in [Figure 4](#nanomaterials-02-00113-f004){ref-type="fig"}d clearly revealed the rhombohedral structure, which fits well with the standard bulk Cu~2−x~S phase (JCPDS 85-1693). {#nanomaterials-02-00113-f004} While DOA also reacts with CuTB to produce copper sulfide nanoparticles at room temperature, we notice that the growth rate is slightly slower and the particles produced are smaller than those prepared with primary amine (*i.e*., OA). We believe that, since OA is less bulky compared to DOA, it attacks the precursor with less hindrance and thus causes the reaction to occur at a faster rate. In addition, DOA is expected to play a better role to efficiently prevent the nanoparticles from aggregation and thus will tend to produce smaller particles. On the other hand, our experiments confirmed that tri-substituted amine does not result in the formation of Cu~2−x~S, probably due to its bulkiness. ### 3.2.3. Indium Sulfide Nanoparticles β-In~2~S~3~ is an n-type semiconductor with a band gap of 2.0--2.2 eV \[[@B36-nanomaterials-02-00113]\]. It has promising applications in the preparation of green and red phosphors for photoconductors and photovoltaics \[[@B37-nanomaterials-02-00113],[@B38-nanomaterials-02-00113],[@B39-nanomaterials-02-00113]\]. In addition, it can serve as a host for a number of metal ions to form semiconducting and/or magnetic materials. In~2~S~3~ nanocrystals have not received as much attention so far, due to the lack of a straightforward preparation methodology. In this case, we have successfully prepared these nanoparticles by decomposing InTB precursor in alkylamines. Reacting InTB with OA could produce spherical In~2~S~3~ with an average diameter of 3.7 ± 0.6 nm ([Figure 5](#nanomaterials-02-00113-f005){ref-type="fig"}a). However, reaction of InTB with OLA or DOA could not proceed at room temperature until the addition of a trace amount of a short chain amine. Thus, upon adding propylamine into OLA, In~2~S~3~ nanoparticles with average diameter of 2.6 ± 0.4 nm could be isolated ([Figure 5](#nanomaterials-02-00113-f005){ref-type="fig"}b). The In~2~S~3~ nanoparticles prepared from OLA and OA show a UV-Vis absorption onset at 355 nm (3.50 eV) and 437 nm (2.84 eV), respectively ([Figure 5](#nanomaterials-02-00113-f005){ref-type="fig"}c). Compared to the band-gap of bulk In~2~S~3~ (2.2 eV or 564 nm) \[[@B36-nanomaterials-02-00113]\], it is clear that the excitonic transition is blue-shifted due to strong quantum confinement in these In~2~S~3~ nanocrystals. [Figure 5](#nanomaterials-02-00113-f005){ref-type="fig"}d shows the XRD pattern of the as-prepared In~2~S~3~ nanoparticles, confirming the tetragonal β-phase of In~2~S~3~ (JCPDS 32-0456). {#nanomaterials-02-00113-f005} ### 3.2.4. Cadmium Sulfide Nanoparticles CdS is one of the most studied metal sulfides, due to its various applications. When CdTB precursor was mixed with OA, spherical CdS nanoparticles with an average diameter of 5.3 ± 0.7 nm were produced ([Figure 6](#nanomaterials-02-00113-f006){ref-type="fig"}a). No reaction had happened, however, in the sole presence of OLA at room temperature. Again, CdS nanoparticles with an average diameter of 4.4 ± 0.3 nm were obtained upon adding a small amount of propylamine into the OLA ([Figure 6](#nanomaterials-02-00113-f006){ref-type="fig"}b). {#nanomaterials-02-00113-f006} The monodispersity of the prepared CdS nanoparticles is manifested in their absorption spectra, which exhibit a clear distinct band rather than a shoulder or threshold ([Figure 6](#nanomaterials-02-00113-f006){ref-type="fig"}c). The band-edge absorption is blue-shifted, occurring at 432 nm for OA-capped and 424 nm for OLA-capped nanocrystals. These gave an estimated particle diameter of 4.9 and 4.0 nm respectively from the Brus equation, in good agreement with those obtained from TEM analysis. CdS is known to exist in two structures: Cubic zinc blende phase and hexagonal wurtzite phase. In our sample, XRD analysis ([Figure 6](#nanomaterials-02-00113-f006){ref-type="fig"}d) suggested a wurtzite crystal structure (JCPDS 41-1049). This observation is interesting, since CdS nanoparticles with zinc blende structure are often produced in ambient conditions, whereas wurtzite structure is often obtained at high temperatures. 3.3. Proposed Mechanism for the Formation of Nanoparticles from the MTB Precursors {#sec3dot3-nanomaterials-02-00113} ---------------------------------------------------------------------------------- ### 3.3.1. The Role of Amine in the Reaction From the above experimental results, we have confirmed that long chain alkylamines such as OA and OLA can react with the thiobenzoate precursors at room temperature to produce monodispersed metal sulfide nanoparticles. Comparatively short chain amines such as propylamine and ethylamine yield larger particles that precipitate quickly from the reaction mixtures. By mixing a small amount of these shorter chain amines into the long chain amines, on the other hand, enables monodispersed nanoparticles to be isolated again. Thus, the long chain amines are needed in this reaction as capping groups to efficiently prevent aggregation of the nanoparticles. After repeating the reactions using different types of amine, we could also conclude that the types of amine used will affect the rate of formation of the specific nanocrystals. Thus, for OLA and aliphatic amines with chains longer than 14 carbon atoms, In~2~S~3~ nanoparticles cannot be produced at room temperature. Hence, heating was needed to initiate this particular reaction. On the other hand, the reaction can be induced at room temperature when a small amount of shorter chain amines are added into the long chain amine. An important aspect of our findings is that MTB precursors are broken down by amines at room temperature. We have found that in other common capping reagents, e.g., TOPO, thiols or carboxylic acids, the decomposition of MTB precursors can only occur at elevated temperatures. We determined the onsets of these reactions by slowly heating up the MTB precursors in these media, and the results are summarized in [Table 2](#nanomaterials-02-00113-t002){ref-type="table"}. We could hence conclude that the decomposition of thiobenzoates in amines is not a simple pyrolysis process. nanomaterials-02-00113-t002_Table 2 ###### Comparison of the decomposition onsets from TGA and reaction in different media for the various MTB precursors. AgTB CuTB InTB CdTB --------------------------- ---------- ---------- ---------- ---------- Decomposition onset (TGA) 172 °C 155 °C 196 °C 163 °C OA Room T Room T Room T Room T Octanethiol \~85 °C \~90 °C \~130 °C \~115 °C TOPO \~130 °C \~130 °C \~160 °C \~150 °C Octanoic acid \~145 °C \~140 °C \~170 °C \~155 °C In addition to the above, we have also confirmed that tertiary amines such as trioctylamine do not react with the MTB precursors even at reflux conditions. Secondary amines, such as DOA, can produce monodispersed nanoparticles from CuTB and AgTB precursors at room temperature, but not with InTB and CdTB. Thus, it seems that amines with an active hydrogen atom are needed for the reaction to proceed at room temperature. On the basis of all the above observations, we propose a general reaction mechanism as follows. The reaction probably arises from an initial attack of the amine group onto the electron-deficient (*i.e*., electrophilic) carbonyl carbon of the MTB precursor \[[@B40-nanomaterials-02-00113]\], and this is followed by an elimination of an amide to produce the metal sulfides ([Scheme 1](#nanomaterials-02-00113-scheme1){ref-type="scheme"}): {#nanomaterials-02-00113-scheme1} For the InTB and CdTB precursor, the reaction with secondary DOA occurs only at elevated temperatures. We believe this is because the bulkier DOA could not access the carbonyl carbon of these di- and tri-substituted MTB precursors readily. As a comparison, we have also performed reactions with shorter chain secondary amines, such as diethylamine and dipropylamine. It was found that reactions of these secondary amines with MTB can occur readily at room temperature. The particles produced, however, are rather large in size, as expected from our proposition that amines are playing dual roles as the nucleophilic reagent and the capping agent. From these results, it can be seen that properties such as coordination ability, steric hindrance, solubility, as well as stability constants are all important factors in the successful preparation of nano-sized crystals from the MTB precursors. ### 3.3.2. Results of DFT Calculations {#sec3dot3dot2-nanomaterials-02-00113} In order to support the proposed mechanism in [Scheme 1](#nanomaterials-02-00113-scheme1){ref-type="scheme"}, we performed DFT calculations by modeling the reaction between MTB precursors with a simple amine-ethylamine. For all the four types of precursors, B3LYP/LanL2DZ optimized transition state structures were located as shown in [Figure 7](#nanomaterials-02-00113-f007){ref-type="fig"}. The zero-point corrected enthalpy change values and energy barriers are tabulated in [Table 3](#nanomaterials-02-00113-t003){ref-type="table"}. For comparison, important geometric parameters of the optimized ground and transition state structures are presented in [Table 4](#nanomaterials-02-00113-t004){ref-type="table"}. Thus, as can be seen in [Figure 7](#nanomaterials-02-00113-f007){ref-type="fig"}, a transition state structure resembling the nucleophilic addition-elimination mechanism proposed is obtained for all the MTB precursors studied. Clearly, a C-N bond (\~1.5--1.6 Å in [Table 3](#nanomaterials-02-00113-t003){ref-type="table"}) is formed between the nucleophilic amine and the electrophilic carbonyl carbon of MTB in the transition state. A simultaneous elimination of amide is also evident from the lengthening of one C-S bond from \~1.8 Å to \~2.7 Å. We notice also that one of the metal-sulfur bonds is shortened during this addition-elimination process towards the formation of metal-sulfur monomer. In [Table 3](#nanomaterials-02-00113-t003){ref-type="table"}, we noted that the DFT predicted activation energy values are similar, (\~110 kJ mol^−1^) for all the four precursors studied. This low energy barrier is in agreement with our observation that these reactions can occur readily at room temperature. Moreover, all the reactions were predicted to give an exothermic enthalpy ranging from 4 to 33 kJ mol^−1^. The calculations further suggested that reactions of AgTB and CuTB will be fast, partly driven by the large exothermic enthalpy change. {#nanomaterials-02-00113-f007} nanomaterials-02-00113-t003_Table 3 ###### B3LYP/LanL2DZ predicted values of energy barrier and enthalpy change for the ethylamine-mediated decomposition of the respective MTB precursors. Precursor Energy barrier (kJ mol^−1^) Enthalpy change (kJ mol^−1^) -------------- ----------------------------- ------------------------------ Ag(SCOPh) 109.7 −33.3 Cu(SCOPh) 109.2 −27.3 Cd(SCOPh)~2~ 108.0 −6.7 In(SCOPh)~3~ 115.3 −4.0 nanomaterials-02-00113-t004_Table 4 ###### Comparison between the B3LYP/LanL2DZ optimized initial state and transition state structures for the ethylamine-mediated decomposition of the respective MTB precursors. Precursor Bond length in precursor (Å) Bond length in transition state structures (Å) -------------- ------------------------------ ------------------------------------------------ ------- ------- ------- ------- Ag(SCOPh) 2.540 1.820 2.398 2.660 1.603 1.061 Cu(SCOPh) 2.348 1.812 2.173 2.669 1.601 1.060 Cd(SCOPh)~2~ 2.632 1.803 2.454 2.801 1.564 1.102 In(SCOPh)~3~ 2.725 1.777 2.373 2.608 1.628 1.035 ^a^Length of the shortened M-S bond; ^b^N-H bond length of ethylamine = 1.017 Å. In order to further elucidate the role of different amines, we also performed B3LYP/LanL2DZ calculations for reactions mediated by different amines. Using a AgTB precursor as the example, transition state structures similar to that shown in [Figure 7](#nanomaterials-02-00113-f007){ref-type="fig"}a were obtained and the calculation results are summarized in [Table 5](#nanomaterials-02-00113-t005){ref-type="table"}. Similarly, the formation of a new C-N bond, lengthening of the C-S bond and shortening of the Ag-S bond are predicted. It is important to highlight that DFT calculations predicted the bond dissociation energy of AgS-COPh as 266.0 kJ mol^−1^, while that of Ag-SCOPh was predicted as 206.4 kJ mol^−1^. Thus, our studies suggest that the stronger S-C bond is broken instead of the weaker Ag-S bond in the presence of amine. In [Table 5](#nanomaterials-02-00113-t005){ref-type="table"}, variations between different amines was predicted to be rather similar, except that the energy barrier is slightly smaller and the exothermic enthalpy is larger in the case of OA and had, as compared to DOA. All these DFT results provided good support to our experimental observations and the proposed mechanism. nanomaterials-02-00113-t005_Table 5 ###### B3LYP/LanL2DZ calculated results for the various amine-mediated decomposition of AgTB. Amine ^a^ N-H bond length (Å) Energy barrier (kJ mol^−1^) Enthalpy (kJ mol^−1^) Bond length in transition state structures (Å) ----------- --------------------- ----------------------------- ----------------------- ------------------------------------------------ ------- ------- ------- EA 1.017 109.7 −33.3 2.398 2.660 1.603 1.061 OA 1.017 107.6 −35.1 2.399 2.670 1.600 1.061 HDA 1.017 107.0 −36.2 2.398 2.680 1.599 1.061 DOA 1.019 108.8 −22.1 2.393 2.945 1.573 1.060 ^a^EA = Ethylamine, OA = Octylamine, HDA = Hexadecylamine, DOA = Dioctylamine;^b^Bond lengths in AgTB precursor: Ag-S = 2.540 Å, C-S = 1.820 Å. ### 3.3.3. FT-IR Analysis of the Reaction Mixtures In order to further support the proposed mechanism, we monitored the reaction at room temperature using FTIR analysis. Reaction mixtures of the respective MTB precursor with OA were concentrated through rotary evaporation and their FTIR spectra were measured. [Figure 8](#nanomaterials-02-00113-f008){ref-type="fig"} shows a typical FTIR spectrum of such reaction mixture, together with the spectra of pure OA and the precursor for comparison. The FTIR spectrum of pure AgTB shows two strong bands at 1602 and 1573 cm^−1^, arising from C=O stretching. These are slightly lower in frequency as compared to υ(C=O) in thiobenzoic acid, due to its strong coordination to the metal ions \[[@B41-nanomaterials-02-00113]\]. Lower frequency peaks at 915, 901 and 647 cm^−1^ resulting from the C-S bending vibration and O-C-S deformation were also observed in the pure AgTB spectrum. On the other hand, the FTIR spectrum of pure OA shows a pair of fairly strong asymmetric and symmetric stretching of primary N-H bands at 3376 and 3292 cm^−1^, and also another N-H bending mode at 1610 cm^−1^. In comparison, the reaction mixture AgTB + OA gives a IR peak of the C=O group at 1641 cm^−1^ ([Table 6](#nanomaterials-02-00113-t006){ref-type="table"}). This position is characteristic of the −C=O bond stretching in amides \[[@B41-nanomaterials-02-00113]\]. Besides, secondary amides often exhibit a relatively strong bending band at about 1543 cm^−1^, attributed to a combination of a C-N stretching band and an N-H bending band. Meanwhile, the doublet of N-H is replaced by a single sharp band at 3310 cm^−1^, typical of the N-H stretching frequency of a secondary amine or an amide. Clearly, the N-H bending mode, existing at 1610 cm^−1^ in pure OA, has disappeared in the reaction mixture. {#nanomaterials-02-00113-f008} nanomaterials-02-00113-t006_Table 6 ###### A comparison of IR spectral bands of AgTB, OA and the mixture of AgTB + OA. Sample Peak Assignments \* (cm^−1^) ----------- ------------------------------ ------ ------ ------ ----- ----- OA 3376 2923 1610 \- \- \- 3292 2861 AgTB \- 3049 \- 1602 915 647 1573 901 AgTB + OA 3310 2928 \- 1641 \- \- 2853 1543 \* υ~s~: stretching; δ~s~: N-H bending; δ: deformation. Thus, the above FTIR analysis confirms the attack of the OA molecule onto the carbonyl carbon of the AgTB precursors to form an intermediate secondary amide adduct. In addition, it was also found that the vibrational peaks of C-S bending and O-C-S deformation detected in AgTB had disappeared from the mixture. This further confirms the breaking down of these bonds as suggested by DFT calculations in [Section 3.3.2](#sec3dot3dot2-nanomaterials-02-00113){ref-type="sec"}. 3.4. Stability of Precursor and Its Influence on the Formation of Nanoparticles ------------------------------------------------------------------------------- In summary, we described above a general precursor method that can be used to prepare metal sulfides at room temperature. Since the injection of reagent is carried out at ambient conditions, special handling techniques are thus not necessary during preparation. In addition, there is just one parameter to monitor once the type of amine is chosen, *i.e.*, the relative amine-to-precursor ratio. This particular simplicity has allowed us to advantageously adopt the method in the preparation of homogeneous nanoparticles/polymer composites, for example, for electrode applications \[[@B42-nanomaterials-02-00113]\]. With a good understanding of the reaction mechanism, we propose to modify the stability of the precursors for controllable growth and crystal engineering. It is known that the growth of colloidal nanocrystals is controlled by a balance between nucleation and growth processes \[[@B43-nanomaterials-02-00113],[@B44-nanomaterials-02-00113]\]. A successful synthetic scheme should start with a burst of nucleation events followed by a controllable growth stage, without either prolonged nucleation or ripening. In addition, anisotropic growth is favored by a sustained high monomer concentration. Hence, as a rule of thumb, a good precursor should be stable enough to prevent individual nucleation to occur, while sufficiently easy to break down to ensure a steady supply of monomers. We illustrate below our attempt to modify the stability of a AgTB precursor to achieve controllable formation of nanoparticles. By dissolving AgTB in trioctylphosphine (TOP) prior to the addition of amine, we found that heating to a temperature greater than 80 °C was required to produce Ag~2~S nanoparticles. In addition, we could obtain β-phase Ag~2~S as the kinetically-driven product by optimizing the reaction temperature and amine-to-precursor ratio \[[@B22-nanomaterials-02-00113]\]. Thus, TOP acts as the stabilizing reagent to AgTB in this case. We performed DFT calculations on complexes of AgTB coordinated to PH~3~ molecules (as a simpler analog of TOP) and the results are shown in [Figure 9](#nanomaterials-02-00113-f009){ref-type="fig"}. In [Table 7](#nanomaterials-02-00113-t007){ref-type="table"}, DFT calculations predicted that three PH~3~ groups could be coordinated to AgTB, with an overall stabilization of \~109 kJ mol^−1^. nanomaterials-02-00113-t007_Table 7 ###### B3LYP/LanL2DZ predicted effect of PH~3~ coordination to AgTB. Structure ^a^ Complex Energy + ZPE (Hartree ^b^) Enthalpy stabilization (kJ mol^−1^) --------------- ---------------- ---------------------------- ------------------------------------- \- AgTB −500.73344 \- \- PH~3~ −8.24658 \- 1 AgTB(PH~3~) −509.01069 −80.5 2a AgTB(PH~3~)~2~ −517.26284 −95.1 2b AgTB(PH~3~)~2~ −517.26420 −98.7 3 AgTB(PH~3~)~3~ −525.51484 −109.4 ^a^ The fully optimized structures are shown in [Figure 9](#nanomaterials-02-00113-f009){ref-type="fig"}; ^b^ 1 Hartree = 2625.502 kJ mol^−1^. {#nanomaterials-02-00113-f009} In addition, we have also located the transition states for the proposed amine-mediated nucleophilic attack onto the complexes by DFT calculations. We modeled the reaction using ethylamine attacking onto three different precursors: AgTB, AgTB(PH~3~) and AgTB(PMe~3~). The optimized transition state structures and their important geometrical parameters are compared in [Figure 10](#nanomaterials-02-00113-f010){ref-type="fig"} and [Table 8](#nanomaterials-02-00113-t008){ref-type="table"}. The DFT predicted energy barriers are 109.7 kJ mol^−1^, 124.1 kJ mol^−1^ and 128.6 kJ mol^−1^ for these three precursors respectively. The slightly higher barrier (\~14--19 kJ mol^−1^) predicted for reactions involving AgTB(PH~3~) and AgTB(PMe~3~) is in agreement with our experimental observations. Thus, we demonstrated that the desired control over nucleation and growth can be achieved through optimizing the delicate balance between the amount of activator (amine) and stabilizer (TOP). We have indeed found this to be general and applicable to most of the MTB precursors studied. For instance, we have varied the type of activator (from OA to dodecanethiol) and stabilizer (from TOP to tributylphosphite) to produce Cu~2−x~S of different crystalline phases \[[@B23-nanomaterials-02-00113]\]. By further changing the activator to difunctional amine such as ethylenediamine, dendritic Cu~2−x~S crystals were produced, giving well-defined snowflakes and blossom-like shapes \[[@B24-nanomaterials-02-00113]\]. {#nanomaterials-02-00113-f010} nanomaterials-02-00113-t008_Table 8 ###### Important B3LYP/LanL2DZ geometrical parameters for the various transition state (T.S) structures, compared with those in the precursor complex (P) and ethylamine (EA). Precursor complex (P) Bond length (Å) ----------------------- ----------------- ------- ------- ------- ------- ------- ------- AgTB 2.540 2.398 1.820 2.660 1.603 1.017 1.061 AgTB(PH~3~) 2.495 2.382 1.822 2.716 1.585 1.079 AgTB(PMe~3~) 2.496 2.390 1.822 2.731 1.579 1.087 4. Conclusions ============== In summary, we have reported here a simple route to prepare various types of metal sulfide nanoparticles readily from metal thiobenzoate precursors under ambient conditions. Such a simple and specific process may be utilized for the production of nanostructures and nanocomposites, and for applications such as nanofabrication and nanopatterning at ambient temperature. Principles we have demonstrated in this report include: (1) by choosing a suitable capping agent that can initiate an attack onto the functionality of the precursor, a relatively less stable intermediate may be generated and this can lead to breaking down of the precursor at room temperature. We believe this also explains why precursors such as metal dithiocarbamate or metal alkylxanthate could react with amine to generate the corresponding semiconductors at temperatures well-below their decomposition onsets \[[@B12-nanomaterials-02-00113],[@B15-nanomaterials-02-00113]\]; (2) by choosing a suitable coordination group, stability of the precursor may be optimized such that the decomposition can be moved to a kinetically-controlled regime. We believe this general principle can be used to guide the preparation of other types of metal chalcogenides from suitable precursors. This research work is supported by the National University of Singapore research grant R-143-000-167-112.

Introduction ============ Porcine reproductive and respiratory syndrome (PRRS), first reported in 1987 in the USA, is a global infectious disease that has resulted in widespread economic loss in the swine industry. PRRS is characterized by reproductive failure in pregnant sows (*e.g.* abortions and stillbirths), respiratory distress in pigs of different ages, and severe immune suppression \[[@B13],[@B16]\]. The PRRS virus (PRRSV) is the recognized causative agent of this syndrome. Two different genotypes of PRRSV have been described: European or type 1, and North American or type 2 \[[@B6],[@B21]\]. In China, most isolated strains are of type 2 \[[@B3]\]. PRRSV is an enveloped, positive, single-stranded RNA virus in the genus *Arterivirus*, which belongs to the family Arteriviridae within the order Nidovirales \[14\].The PRRSV genome is 15 kb in length and contains at least 10 open reading frames (ORFs). Of these, ORF1a and ORF1b represent nearly 75% of the viral genome and encode two large polyproteins (pp), pp1a and pp1ab, respectively. These pps can be hydrolyzed into at least 16 small nonstructural proteins (Nsps). Of these 16 Nsps, at least 14 are involved in viral genome replication and transcription \[[@B10]\]. The Nsp1, Nsp2, and Nsp7 proteins elicit a strong immune response in pigs. Fourteen days after pigs are infected with PRRSV, anti-PRRSV Nsp7 antibodies can be detected, and high levels of antibodies in pigs can last for up to 202 days \[[@B2]\]. A diagnosis of PRRSV is important for its prevention and control. The laboratory diagnostic tests commonly used at present include reverse transcriptase polymerase chain reaction (PCR), quantitative real-time PCR \[[@B19],[@B22]\], and four serological detection methods \[[@B1],[@B4],[@B5],[@B20]\]: the indirect fluorescence assay (IFA), the immunoperoxidase monolayer assay (IPMA), the serum neutralization test (SN), and the enzyme-linked immunosorbent assay (ELISA). In recent years, by using the PRRSV N protein as an antigen, the commercial IDEXX ELISA kit (IDEXX Laboratories, USA) has been widely used for the detection of PRRSV antibodies. Recently, an ELISA kit that uses recombinant Nsp7 protein as the antigen has been introduced, and the concurrence rate between that kit and the IDEXX ELISA assay has been reported to be up to 97.6% \[[@B2]\]. Although traditional laboratory tests offer good sensitivity and specificity, these detection methods require professional and technical personnel and specialized equipment. In addition, there are many methodological aspects that must be constantly improved and complemented. In developing countries, vaccination is an important strategy for preventing and controlling PRRSV. To monitor the titer of anti-PRRSV antibodies after vaccination in order to confirm the effectiveness of the vaccination, quick and easy techniques suitable for field testing are needed. Compared with the traditional detection methods, an immunochromatographic strip method has the advantages of being easy to use, providing an answer rapidly and at a low cost, and not requiring specialized equipment or technical personnel, thus such a method is suitable for field testing for antigens or antibodies. In this study, we developed an immunochromatographic test strip for detecting anti-PRRSV antibodies and conducted a preliminary field study of the strip. Materials and Methods ===================== Serum samples ------------- Pigs were experimentally inoculated with either of two strains of PRRSV having differing levels of virulence, HN07-1 and BJ-4, in order to provide positive test serum samples. HN07-1 strain was isolated during an atypical PRRSV outbreak in Henan Province, China in 2007 \[[@B17]\]. The BJ-4 strain was a typical North American (VR2332)-like PRRSV isolated in 1996 in China \[[@B17]\]. The field sera for antibody testing were collected from several swine herds (all sampled pigs were more than 2-weeks-old) within Henan Province, China. Positive serum samples for porcine circovirus-2 (PCV-2), pseudorabies virus (PRV), classical swine fever virus (CSFV), porcine parvovirus (PPV), and foot-and-mouth disease virus (FMDV) were collected and stored in the Henan Provincial Key Laboratory of Animal Immunology, Henan Academy of Agricultural Sciences. Preparation of recombinant Nsp7 protein of PRRSV ------------------------------------------------ For the expression of a recombinant Nsp7 proteins in *Escherichia* (*E.*) *coli*, reverse transcriptase PCR was performed on viral genomic RNA from PRRSV isolate BJ-4, obtained from Dr. Hanchun Yang of China Agricultural University. To amplify the Nsp7 gene the following oligonucleotide primers were used: forward primer, 5′-CGC**GGATCC**TCTCTGACTGGTGCCCTCGCTATG-3′ and reverse primer, 5′-CCG**CTCGAG**TTCCCATTGAACTCTTCCAT-3′ (restriction sites in bold font). The amplified gene was then ligated into the pET-28a vector. The recombinant plasmid was transformed to *E. coli* BL21-competent cells. Single colonies were obtained and tested by PCR and sequencing, and a positive clone was grown at 37℃ in LB broth supplemented with 100 µg/mL ampicillin to an optical density of 0.8 at 600 nm. Expression of the recombinant protein was induced by 100 mM isopropyl-β-D-thiogalactopyranoside (IPTG, TAKARA Bio, China) for 8 h at 37℃. Cells were then harvested by centrifugation (7000 × g for 30 min). Purification of recombinant Nsp7 protein ---------------------------------------- The recombinant Nsp7 protein was purified by immobilized-metal affinity chromatography (IMAC) using a polyhistidine tag and further purified by a gel filtration column Superdex200 (GE Healthcare, Sweden). The cell pellet was suspended and lysed by sonication on ice. The lysate was centrifuged at 16,000 × g for 30 min, and the supernatant was collected and transferred to a Ni-NTA His Band Resin column pre-equilibrated with binding buffer (500 mM NaCl, 20 mM Tris, 5 mM imidazole). More than five column-volumes of washing buffer (500 mM NaCl, 20 mM Tris, 20 mM imidazole) was added to remove the nonspecific binding proteins. The target protein was eluted with elution buffer (500 mM NaCl, 20 mM Tris, 400 mM imidazole). The purity and relative concentration of the recombinant Nsp7 was determined by SDS-PAGE. The protein was further fractionated by gel filtration on a column of Superdex200 in a buffer of 50 mM Tris, 150 mM NaCl by using the Bio-Rad BioLogic system (Bio-Rad Laboratories, USA). The protein of interest was collected in different fractions according to its different states of aggregation. The final protein products were examined by SDS-PAGE before storing at −80℃. Western blot ------------ For western blot analysis, 4 µg purified recombinant Nsp7 protein were subjected to 15% SDS-PAGE gel and transferred to polyvinylidene difluoride membranes. The membrane was washed with phosphate-buffered saline-Tween20 (PBST) and blocked with 5% skimmed milk. After washing three times with PBST, the membranes were reacted with PRRSV-positive sera; a PRRSV-negative serum were used as a negative control. After incubating at 37℃ for 1 h, the resulting blot was treated with secondary antibody horseradish peroxidase-conjugated rabbit-anti-pig IgG (Abbkine; WuHan AmyJet Scientific, China) for 1 h. As the substrate for color development, 3-amino-9-ethylcarbazole (AEC) was used. The antigenicity of the separated protein fractions compared by ELISA --------------------------------------------------------------------- The antibody binding capability of the monomer, dimer, and larger aggregate of the recombinant Nsp7, which were separated by Superdex200 gel filtration column, were compared by indirect ELISA assay. The separated proteins were diluted to the appropriate concentration in 50 mM sodium carbonate bicarbonate buffer (pH 9.6). After incubation for 14 h at 4℃, antigen-coated plates were washed five times with phosphate-buffered saline (PBS) containing 0.05% Tween 80 then blocked with 5% skimmed milk powder dissolved by PBST for 1 h at 37℃. Then, appropriate dilutions in PBST of PRRSV-positive HN07-1, PRRSV-positive BJ-4, and PRRSV-negative pig sera were incubated in the antigen-coated wells at 37℃ for 30 min. Secondary antibody horseradish peroxidase-conjugated rabbit-anti-pig IgG was added at a final dilution of 1:2,000, and the mixture incubated for a further 30 min at 37℃. Finally, 3′,3′,5′,5′-tetramethylbenzidine was added as a substrate. Color development was stopped with 2 M H~2~SO~4~, and the OD value at 450 nm was read on a spectrophotometer. Conjugation of antigen with colloidal gold ------------------------------------------ Colloidal gold with an average particle diameter of approximately 20 to 25 nm was obtained by reduction of a HAuCl4 solution with 1% trisodium citrate. Three milliliters of 1% trisodium citrate (w/v) was added to 100 mL of 0.01% HAuCl~4~·3H~2~O solution (w/v) with stirring. Then the mixture was heated to 100℃ for 20 min. The colloidal gold solution was then cooled to room temperature and stored at 4℃. The colloidal gold-labeled antigen was prepared according to a previously reported method \[[@B23]\]. Briefly, 4 mL of purified protein (0.2 mg/mL) was incubated with 80 mL of colloidal gold solution for 30 min at room temperature with the pH adjusted to 9.4. After the addition of 1 mL of 3% casein solution, the mixture was incubated at room temperature for a further 10 min, and the resulting suspension was centrifuged 25,000 × g at 4℃ for 30 min. Finally, the colloidal gold-labeled antigen was suspended in 4 mL of 0.02 M sodium borate buffer (containing 2% bovine serum albumin \[BSA\], 1% sucrose, 0.1% NaN~3~) and stored at 4℃. Preparation of the conjugate pad and sample pad ----------------------------------------------- Ten percent BSA was sprayed onto a fiberglass conjugate pad by using a XYZ-3000 dispensing platform (BioDot, USA). The pad was then dried at 42℃ for 45 min. Conjugate solution was dispensed onto the modified fiberglass at a speed of 25 µL/cm (about 2 µg per pad) by using the dispensing platform. After drying for 1 h at 42℃, the pad was then stored in a desiccator at room temperature. The sample pad was saturated with a buffer containing 20 mM sodium borate, 2.0% (w/v) sucrose, 2.0% (w/v) BSA, and 0.1% (w/v) NaN~3~ then dried and stored as described above. Pig anti-Nsp7 antibody preparation ---------------------------------- Two-month-old pigs were injected intramuscularly with 100 µg recombinant Nsp7 protein of PRRSV in 2.5 ml of an oil-in-water emulsion per pig. The injection was repeated two times at two-week intervals. Fifteen days after the third immunization, blood containing a high titer of anti-Nsp7 antibody was collected from the immunized pigs. Porcine anti-Nsp7 IgG was salted out from the hyperimmune sera through successive treatment with 50%, 40%, and 33% saturated ammonium sulfate and dialyzed with 0.01 M PBS. ELISA was carried out to determine the titer of anti-Nsp7 protein IgG. The ELISA plate was coated with 0.2 µg/mL Nsp7 protein at 37℃ for 2 h. After washing with PBST five times and blocking with 5% skimmed milk, the anti-Nsp7 protein IgG was two-fold serially diluted with physiological saline and added to the ELISA plate. The anti-Nsp7 IgG titer level was up to 1:20,000. The purified Nsp7 IgG was aliquoted into tubes (1 mL/tube) and stored at −80℃. Immobilization of capture reagents ---------------------------------- Staphylococcal protein A (SPA; Sigma Chemical, USA) at 1.5 mg/mL and porcine anti-Nsp7 IgG at 1 mg/mL were sprayed by using Quanti 3000 Biojets (BioDot) attached to a XYZ Biostrip Dispenser (BioDot) onto a nitrocellulose filter membrane (HiFlow Cellulose Membrane; Millipore, USA) to form zones of 300 mm × 25 mm at the test line (T line) and quality control line (C line), respectively. The T and C lines were in the middle of the membrane, and 5 mm apart from each other. The flow rate of the liquid was 180 sec/4 cm on the membrane, the use level was 1 µL/cm, and forced air drying was carried out at 40℃ for 1 h. The membrane was then sealed in plastic with desiccant and kept at room temperature. Assembly principle and use of test strips ----------------------------------------- Assembly of the test strips used the same method as is previous descriptions \[[@B24],[@B25],[@B26]\]. The strip assembly consists of a sample pad, a conjugate pad, a nitrocellulose membrane (NC membrane), and an absorbent pad. The NC membrane was adhered to the center of the plastic solid support, and the absorbent pad and conjugate pad were adhered to upper and lower sides, respectively of the cellulose nitrate membrane with a 1 mm overlap. The main card was cut to the required width with a paper cutting machine. Each strip was packed in a plastic block and then sealed into a plastic bag with desiccant. When stored at room temperature, the strip\'s period of validity is one year. For testing, serum samples were diluted 1/200 in physiological saline. The 100 µL diluted sample was added to the sample pad and allowed to migrate to the absorbent pad. If the serum sample to be tested contained PRRSV specific antibodies, these antibodies will form a "gold-labeled antigen-antibody" complex with the gold-labeled antigen on the conjugate pad. The complex, while migrating to the NC membrane, binds to SPA, thereby showing a red line in the T line area. Some of the gold-labeled antigen and some of the gold-labeled antigen-antibody complex migrates further to react with the anti-Nsp7 antibody, showing another red line in the C line area. Negative samples do not produce a red line in the T area, whereas the C line will always show a red line whether or not the serum contains anti-PRRSV antibodies. Cross-reactivity and detection limit of the immunochromatographic strip test ---------------------------------------------------------------------------- Cross-reactivity of the strip was evaluated with 18 PRRSV-positive sera (individual S/P ratios were 0.98, 1.37, 1.57, 1.66, 1.66, 1.70, 1.71, 1.72, 1.75, 1.77, 1.78, 1.80, 2.00, 2.28, 2.46, 2.611, 2.75, and 2.79), 10 standard PRRSV-free/negative sera (S/P \< 0.4) from PRRSV-free and non-vaccinated pigs and 32 CSFV-, FMDV-, PCV-2-, PRV-, and PPV-infected serum samples. The detection limit of the strip was evaluated by using PRRSV-positive sera. The titers of the sera were determined by using a commercial ELISA Kit (IDEXX PRRS X3 ELISA kit; IDEXX Laboratories, USA) as the reference standard. Physiological saline solution was used to dilute sera from 1:100 to 1:3,200. The diluted sera were tested with the test strip and the intensity of the T line was scanned with a BioDot TSR3000 film scanner to provide a numerical value. The whole process was repeated three times by three different operators. Comparison of the strip with other methods ------------------------------------------ The strip and the commercial IDEXX PRRS X3 ELISA kit were simultaneously used to test and evaluate 1034 sera obtained from several swine herds in Henan Province, China. A comparison of the results obtained by the two methods was used to evaluate the effectiveness of the test strip. To further evaluate the immunochromatographic test strip for detection, anti-PRRSV antibody sera were collected at intervals after infection. Piglets from PRRSV-free, specific pathogen-free herds were experimentally inoculated with two PRRSV strains of differing virulence, HN07-1 and BJ-4. Groups of three animals were inoculated with HN07-1 or BJ-4 and a further three uninfected piglets were sham-inoculated with a control cell culture supernatant. Blood samples were collected at 0, 7, 14, 21, 28, and 35 days after infection. Sera from these blood samples were simultaneously tested by the immunochromatographic test strip and two reference standard methods: the IDEXX PRRS X3 ELISA kit and the IPMA. A similar experiment was undertaken to evaluate sera from PRRSV-vaccinated pigs. Five piglets were intramuscularly vaccinated with highly pathogenic porcine reproductive and respiratory syndrome virus vaccine (HuN4-F112) following the manufacturer\'s instruction, which were produced by Harbin WEIKE Biological Technology Development Company (batch No. 2014001). Sera were collected at 3, 7, 14, 21, and 28 days post vaccination, and tested with the strip and the IDEXX PRRS X3 ELISA kit. The IPMA was performed according to standard methods. Marc-145 cells were inoculated into 96-well cell plates and cultured in Dulbecco\'s modified Eagle medium (DMEM; Shanghai Bioleaf Biotech, China) containing 10% fetal bovine sera (Shanghai Bioleaf Biotech) at 37℃ in 5% CO~2~ atmosphere for 24 h. When the cells had grown to cover over 90% of the available area, serum samples to be assayed were diluted 1:100 with PBS, and 100 µL of each dilution were transferred to paired wells of HN07-1-infected and HN07-1-uninfected Marc-145 cells. The plates were incubated at 37℃ for 1 h and washed three times with PBST, the secondary antibody, horseradish peroxidase-conjugated rabbit-anti-pig IgG, was added at a final dilution of 1:2,000 in PBST. After washing 3 times, the monolayers were stained with AEC. The plates were observed with a light microscope, and wells containing distinctly stained individual cells were scored as positive. Results ======= Expression and purification of the Nsp7 protein in *E. coli* ------------------------------------------------------------ After IPTG induction, the *E. coli* BL21 (DE3) carrying pET28a-Nsp7 plasmid was shown by using SDS-PAGE to express a recombinant protein with a molecular weight *circa* 34 kDa. The expressed protein was present in the soluble, as well as in the insoluble, fraction of the bacterial cell pellet. After sonication treatment, the completely soluble target protein was dissolved in phosphate buffer solution and purified by Ni-chelation chromatography. The purified protein was up to 90% pure as determined by SDS-PAGE and western blot analysis revealed that the purified protein reacted with anti-PRRSV sera, confirming its antigenicity ([Fig. 1](#F1){ref-type="fig"}). The target protein self-aggregated and was further fractionated by using molecular sieve chromatography. The different assemblages, referred to as the "larger aggregate", the "dimer", and the "monomer" were then identified by SDS-PAGE ([Fig. 2](#F2){ref-type="fig"}). Specific immunogenicity of larger aggregate, dimer, and monomer Nsp7 -------------------------------------------------------------------- The immunogenicity of the Nsp7 larger aggregate, the dimer, and the monomer assemblages were examined by ELISA. The Nsp7 larger aggregate and dimer proteins showed some reactivity with PRRSV-negative sera. The Nsp7 monomer reacted strongly with pig sera against both classical and atypical PRRSV but reacted weakly with negative sera ([Fig. 3](#F3){ref-type="fig"}). Thus, the monomer was selected for use as the colloidal gold-labeled antigen probe for further study. Cross-reactivity of the immunochromatographic strip --------------------------------------------------- Serum samples after 200-fold dilution with physiological saline solution were applied to the immunochromatographic test strip. The combination of antigen and antibody forms an "antigen-antibody complex" which binds to SPA and gives a red line (positive serum) in the test region; no red indicates a negative serum. As a product quality control check, the C line will always show a red line, regardless of whether anti-PRRSV antibodies are present. Cross-reactivity of the immunochromatographic test strip was evaluated with 60 reference serum samples, including 18 PRRSV standard positive sera, 10 negative sera samples, and 32 serum samples positive for CSFV, FMDV, PRV, PCV-2, and PPV. As shown in [Fig. 4](#F4){ref-type="fig"}, the positive sera appeared as two visible red bands at the T and C line positions on the strip, while the negative sera and other sera appeared as only one red band at the C line position. As shown in [Table 1](#T1){ref-type="table"}, the strip gave 18 positive results from the 18 PRRSV-positive samples, and all of the 10 PRRSV-negative samples and the serum samples positive for other non-PRRSV pathogens showed negative results on the immunochromatographic strip. The results indicated that the specificity of the immunochromatographic strip was 100% (42/42). The detection limit of the immunochromatographic strip ------------------------------------------------------ To evaluate the detection limit of the immunochromatographic strip, the PRRSV standard positive sera (S/P ratios were 2.428 and 2.715) were two-fold serially diluted with physiological saline from 1:100 to 1:3,200. A red band could be seen clearly at the T line position when the dilution of the serum sample was 1:3,200 or less, indicating that the strip has a high sensitivity for detecting small amounts of anti-PRRSV antibodies. The strip test lines were scanned with a BioDot TSR3000 membrane strip reader and showed that the G/D×A-ROD×pixel of PRRSV-positive sera for HN07-1 and BJ-4 diluted 1:100, 1:200, 1:400, 1:800, 1:1600, 1:3200 were 95.3926, 118.0526, 187.4653, 114.9801, 78.9306, 52.1133 and 44.6651, 49.7877, 103.5346, 105.3718, 65.5187, 36.0817, respectively. The highest G/D×A-ROD×pixel of the negative/control sera was 13.7273, which was very weak and could not be seen by the naked eye ([Table 2](#T2){ref-type="table"}). The results indicated that the strip had a high sensitivity for detecting PRRSV antibodies and could detect antibodies in serum samples at low concentrations. Comparison of the strip with other methods for detection of experimental infection in pigs ------------------------------------------------------------------------------------------ A comparison of the serological reactivity of the strip with the commercial IDEXX PRRS X3 ELISA test and the IPMA was performed by using serum samples that were collected from pigs experimentally infected with isolates HN07-1 and BJ-4 ([Table 3](#T3){ref-type="table"}). All sera collected prior to 7 days postinoculation were negative in the strip test, ELISA, and IPMA results. At 14 days postinoculation, all samples were detected as seropositive (S/P \> 0.4) by the IDEXX ELISA. The strip test detected seroconversion at 14 days post inoculation in pig numbers 10 and 26. At 21 days postinoculation, all samples were identified as positive by all three methods. In summary, 33 of 36 (91.7%) IDEXX results agreed with the results obtained by the strip and 32 of 36 (88.9%) IPMA results were in agreement with the strip. Comparison of postvaccination PRRSV antibody detection by the strip and IDEXX PRRS X3 ELISA ------------------------------------------------------------------------------------------- All sera collected prior to 7 days post vaccination were negative in both the strip and the IDEXX PRRS X3 ELISA results. At 14 days post vaccination, both the strip and ELISA detected seroconversion in the vaccinated pigs, but both tests also recorded animals which were negative. However, at 21 days postvaccination all serum samples were positive in both tests ([Table 4](#T4){ref-type="table"}). Comparison of the strip with the IDEXX PRRS X3 ELISA for the detection pigs infected in the field ------------------------------------------------------------------------------------------------- To further demonstrate the usefulness and reliability of the immunochromatographic strip for detecting the serum antibody responses of pigs infected with field strains of the virus a 1034 field serum samples were tested in parallel with the strip and IDEXX PRRS X3 ELISA ([Table 5](#T5){ref-type="table"}). According to the formula TP/(TP + FN) and TN/(TN + FP): TP, true positive, TN, true negative, FP, false positive, and FN, false negative, which was used to calculate the sensitivity and specificity of the test, the corresponding values were 93.04% (682/733) and 88.70% (267/301), respectively. Discussion ========== Serological testing for antiviral antibodies is a standard diagnostic and surveillance method for determining whether pigs have been exposed to PRRSV. Currently, the IDEXX PRRSV ELISA kit, which is based on the antigen PRRSV N protein, is widely used for determining the viral infection status of swine herds. The recombinant Nsp7 protein of PRRSV has also been used as an antigen in indirect ELISA tests to detect the antibodies against PRRSV \[[@B2]\]. These ELISA methods require professional/technical personnel and are time-consuming. The proposed immunochromatographic test strip described here is simple, easy to operate, cost efficient, and provides rapid results, which is ideal for wide application, including developing countries. As mentioned, PRRSV has European and North American types. Nucleotide sequences within the same genotype have some variability, whereas the nucleotide sequences between the two different genotypes differ significantly \[[@B11],[@B12]\]. The amino acid sequence identity of the Nsp7 protein in the European-type PRRSV is within 96.7% to 97.4%, whereas that in the North American type is within 84.9% to 100%. However, the amino acid sequence identity of Nsp7 protein between the European and North American types overlapped by only 45% \[[@B8],[@B9],[@B15],[@B18]\]. PRRSV isolates in China are mainly of the North American type, and the BJ-4 strain is a North American type. Thus, we selected the Nsp gene from the BJ-4 strain, a typical North American (VR2332)-like PRRSV, for the preparation of the recombinant antigen for the detection of serum antibodies for the diagnosis of PRRSV infection in China. The developed immunochromatographic strip that uses the recombinant Nsp7 antigen of the BJ-4 strain can detect antibodies against the atypical PRRSV isolate HN07-1, which has been the dominant virus in China after 2006. Recombinant Nsp7 protein has a number of practical advantages. First, it is expressed in a soluble form, which is convenient for protein purification. The Nsp7 protein is sufficiently immunogenic for antibodies against the Nsp7 protein to be detected 14 days after PRRSV infection, and these antibodies remain at elevated levels for 202 days \[[@B2]\]. Therefore, the recombinant Nsp7 protein can be used as the target antigen for detecting anti-Nsp7 antibodies in serological testing. The immunochromatographic strip for detecting PRRSV antibodies on the basis of N protein was developed in 2008 \[[@B7]\]. The movement to eliminate PRRSV may require a variety of serological diagnostic technologies that can accurately detect persistent infection in pigs. In this study, we used a gold-labeled recombinant Nsp7 protein to bind the anti-PRRSV antibodies. This gold-labeled antigen-antibody complex can be captured by SPA and the result is then seen as a red line indicating the presence of PRRSV antibodies. The Nsp7 recombinant protein was further purified by gel filtration to yield the monomer form because aggregates of the protein were found to nonspecifically bind IgG. We observed cross reaction of the Nsp7 large aggregate and dimer protein with PRRSV-negative sera, and the Nsp7 monomer reacted strongly and specifically with porcine antibodies against both classical and atypical PRRSV. On this basis, the monomer was selected for a colloidal gold-labeled antigen probe to increase the specificity of the test strip. The developed test strips have high sensitivity in the detection of antibodies against PRRSV and exhibit no cross reaction with antibodies of other pathogens such as FMDV, PRV, CSFV, PPV, PCV-2, and negative PRRSV sera. We also used the test strips to screen 1034 clinical sera from several swine herds in Henan Province and then compared our results with those from the commercial IDEXX PRRS X3 ELISA kit. The results showed that the sensitivity and specificity of the strip were 93.04% and 88.70%. We used the test strips to detect antibodies in pigs experimentally infected with both classical and atypical PRRSV and compared the results with those of the IDEXX PRRS X3 ELISA kit and the IPMA. The overall coincidence rate was 91.7% and 88.9%, respectively. Post vaccination test results from piglets demonstrated that the strip and the IDEXX PRRS X3 ELISA assay had a high coincidence rate (92%). Considering all of the results, the test strip shows comparable performance to the reference ELISA and should be considered of similar value to the ELISA kit in clinical applications. In summary, the chromatographic strip developed in this study is a simple, sensitive, and serologically-specific detection method. The study results show that the strip accurately identifies antibodies against PRRSV in experimentally infected swine sera, field-infected porcine sera, and sera from piglets after vaccination. Results using the new method are well within the ranges achieved by the commercial IDEXX ELISA assay and the IPMA test. However, the strip cannot distinguish antibodies induced by vaccination from those resulting from viral infection. Further investigation to develop an antibody differentiation test is required. This work was supported by grants from China Agriculture Research System (CARS-36), the Pig Industry Technology System Innovation Team Project of Henan Province (S2012-06), and the Special Fund for Agro-scientific Research in the Public Interest (201203039). We would like to thank Professor Gregson from IoN University for revising the paper. **Conflict of Interest:** The authors declare no conflict of interests. {#F1} {#F2} {#F3} {#F4} ###### Specificity of the immunochromatographic strip with 200-fold diluted porcine sera  PRRSV, porcine reproductive and respiratory syndrome virus; PCV-2, porcine circovirus-2; PCV-1, porcine circovirus-1; CSFV, classical swine fever virus; PRV, pseudorabies virus. ###### G/D×area, and G/peak of the relative optical density (ROD) of test lines of the standard porcine reproductive and respiratory syndrome virus samples^\*^  ^\*^Standard sera samples were tested by using the immunochromatographic test strips and the test lines were scanned with a TSR3000 membrane strip reader. G/D×A indicates graph/density×area, which means density value of the sampled outline, multiplied by its area. G/peak indicates graph peak density, maximum density value of sampled line points. ROD means relative optical density, inverse of gray level value with a logarithmic transformation. ###### Detection porcine reproductive and respiratory syndrome virus (PRRSV) antibodies in swine sera at different days after infected with PRRSV HN07-1 or BJ-4: test strip vs. IDEXX PRRS 3X ELISA kit and immunoperoxidase monolayer assay (IPMA)  ELISA, enzyme-linked immunosorbent assay. ###### Detection of porcine reproductive and respiratory syndrome virus antibodies in swine sera at different days after vaccination using the strip and the IDEXX PRRS 3X ELISA kit  ELISA, enzyme-linked immunosorbent assay; −, Negative; + positive. Samples with S/P ratio \> 0.4 were considered positive. ###### Comparison of results from the strip and the IDEXX PRRS X3 ELISA from pigs infected with field virus  ELISA, enzyme-linked immunosorbent assay. [^1]: ^†^The first two authors contributed equally to this work.

Background ========== Human immunodeficiency virus type-1 (HIV-1) is characterized by extensive genetic heterogeneity. Molecular epidemiologic studies have demonstrated that globally, the most prevalent forms of HIV-1 are subtypes (clades) C, B and A \[[@B1]-[@B3]\]. Subtype C, which accounts for almost 50% of all HIV-1 infections globally, predominates in sub-Saharan Africa and India \[[@B1]-[@B3]\]. Subtype B is the main genetic form in the Americas, Australia and western Europe; subtype A predominates in areas of central and eastern Africa (Kenya, Uganda, Tanzania and Rwanda) and in eastern Europe \[[@B1]-[@B3]\]; and subtype D is distributed mainly in east Africa, including Uganda \[[@B1]\]. HIV-1 subtypes differ by as much as 20-25% at the genetic level \[[@B2]\], and have varying biological characteristics, including differences in disease progression, pathogenicity, transmissibility and co-receptor usage \[[@B1],[@B2],[@B4]-[@B7]\]. Studies of HIV-1 co-receptor tropism, which have been conducted primarily in populations where subtype B infections predominate, have demonstrated a relationship between HIV-1 co-receptor use and disease stage. In general, early stages of infection and disease are characterized by greater prevalence of only C-C chemokine type 5 (CCR5)-tropic (R5) HIV-1, which has been associated with slower progression to AIDS \[[@B8]-[@B12]\]. The emergence of C-X-C chemokine receptor type 4 (CXCR4)-using virus (X4) has been associated with greater treatment experience and higher risk of death, and coincides with more rapid CD4^+^T-cell depletion and disease progression \[[@B6],[@B8],[@B9],[@B12],[@B13]\]. Some variants of HIV-1 can use either co-receptor (dual/mixed-tropic \[DM\] HIV-1); these can be found in all stages of infection, but are more common in infections of longer duration, with lower CD4^+^cell counts and higher viral loads \[[@B12]-[@B14]\]. Despite the emergence of X4-using variants in some patients, only R5 infection typically persists in the majority of patients. Nearly 50% of patients who die of HIV-1 disease have only R5 HIV-1 detectable at the time of their death, indicating that CCR5 remains a critical co-receptor throughout the course of HIV infection \[[@B12],[@B15]\]. Although HIV-1 co-receptor usage and its relationship to disease stage have been studied in the developed world, where subtype B predominates, such relationships are less well understood for subtypes A, C and D. The R5 phenotype is predominant in subtype C HIV-1 infections, whereas X4-using virus has been reported infrequently, even in advanced disease. R5-using virus is more common in subtype A than subtype D HIV-1 infections, and a high proportion of subtype D infections shows D/M tropism throughout the course of disease \[[@B16]-[@B24]\]. However, some of these previous studies have been limited by small sample sizes. The introduction of the CCR5 antagonist, maraviroc, for HIV-1 therapy \[[@B25]\] has increased interest in the epidemiology of tropism and relationships with HIV-1 subtype. A greater understanding of the tropism of non-B subtype HIV-1 is key for the optimal use of CCR5 antagonists in the treatment of these infections in the developing world, and HIV-1 prevention strategies, such as topical microbicides and systemic pre- or post-exposure prophylaxis. In addition, this information will be important for management of clinic populations in the developed world that include individuals with non-B subtype infections who have migrated from endemic countries \[[@B26]\]. HIV-1 tropism can be determined by genotypic and phenotypic methods. While genotypic assays may have lower specificity and sensitivity, retrospective analyses have found that they are comparable to phenotypic tropism assays for prediction of response to treatment with CCR5 antagonists, in populations pre-screened with a phenotypic assay \[[@B27],[@B28]\]. The clinical development programme for maraviroc, the first-in-class CCR5 antagonist, used the Trofile^®^phenotypic assay (Monogram Biosciences, South San Francisco, California) \[[@B29]-[@B31]\], which determines tropism via the expression of full-length *env*genes of multiple viruses isolated from patient plasma and can detect 10% of X4 variants with 100% sensitivity. More recently, a Trofile^®^assay with enhanced sensitivity to improve detection of low-level X4-using variants has been developed that can detect 0.3% of these variants with 100% sensitivity. Otherwise, assay validation performance characteristics are equivalent between the original and enhanced Trofile^®^assays \[[@B31]\]. The enhanced Trofile^®^assay has been validated in a number of studies by re-testing the co-receptor tropism of clinical samples that were initially determined using the original assay \[[@B31]-[@B34]\]: in a re-analysis of samples from the Phase 3 MERIT study, which evaluated the efficacy of maraviroc in treatment-naïve patients with CCR5-tropic virus, 15% of enrolled patients (n = 106/721) were reclassified as having D/M virus with the enhanced assay \[[@B34]\]. The aim of this study was to estimate the prevalence of R5-, D/M-, and X4-tropic HIV-1 among isolates obtained from patients with HIV-1 subtype C infection from India and South Africa, and with subtype A/A1 and D infection from Uganda, and to explore the demographic and clinical characteristics associated with R5 infection. In addition, the study examined the ability of the Trofile^®^assay to determine tropism of non-B subtypes of HIV-1, which previously had not been explored in a large study. Methods ======= Study design ------------ HIV-1-infected, antiretroviral therapy (ART) treatment-naïve (TN) and treatment-experienced (TE) viremic patients were recruited into this prospective, cross-sectional, epidemiologic study from HIV clinics in South Africa (four sites), Uganda (one site), and India (seven sites) in 2007 and 2008. Sites were selected if they had considerable experience with both HIV management and HIV research. The study protocol was approved by the institutional review board at each site. Both TN and TE adults (aged 18 years or older) were eligible for enrolment in India and Uganda. In South Africa, where the MERIT study had previously been conducted in TN patients \[[@B35]\], only TE adults were eligible for inclusion in the present study. No patients were recruited from any other study. Patients who had received less than 10 days of ART were considered to be TN; those who had experienced failure of at least one three-drug ART regimen were considered to be TE. To maximize the external validity and generalizability of the study, only one member of a known infection cluster (i.e., only one member of a family affected by HIV) was eligible for enrolment. Study procedure --------------- At a single visit, the study was explained to patients, and verbal and written informed consent were obtained prior to conduct of any study procedures. Demographic, and HIV clinical history and treatment data were collected; blood was drawn and analyzed for CD4^+^and CD8^+^T cells using BD FACS™ CAP (Becton Dickinson and Company, Franklin Lakes, New Jersey), and for HIV-1 RNA levels using Amplicor HIV-1 Monitor™ UltraSensitive Assay (Roche Diagnostics, Indianapolis, Indiana). For individuals with viral loads exceeding 500 copies/mL, HIV-1 subtype was determined based on reverse transcriptase and protease gene sequence (Monogram Biosciences). *Pol*subtyping was performed by generating three distinct sets of partial-length nucleotide sequences from patient-derived reverse transcriptase and protease genes. Genetic sequence comparison was determined by the Basic Local Alignment Search Tool algorithm, and subtype was resolved by a customized software package that was validated against the publically available tool on the National Center for Biotechnology Information web site at the National Institutes of Health. Samples determined to be HIV-1 subtypes of interest (A/A1 and D in Uganda; C in South Africa and India) were further tested for viral tropism. HIV-1 co-receptor tropism was determined using the original Trofile^®^assay \[[@B29],[@B30]\] for samples collected from patients in South Africa and Uganda. The Indian cohort was enrolled after the South African and Ugandan cohorts, thus Indian patients were tested using the newly available enhanced Trofile^®^assay, which had replaced the original assay. Samples were prepared for both the original and enhanced assays in the same way: 1 mL plasma samples underwent centrifugation and viral RNA was isolated, purified and subjected to polymerase chain reaction (PCR) amplification of the entire HIV envelope gene (*env*). Co-transfection of HIV *env*expression vectors and HIV-1 genomic vectors produced pseudoviruses containing full length *env*genes derived from patient virus populations. Tropism was determined by measuring the ability of the pseudovirus population to efficiently infect target cells co-expressing CD4 with either the CXCR4 or CCR5 co-receptor \[[@B29],[@B31]\]. For commercial Trofile^®^testing, up to 3 mL of plasma and up to three attempts at RNA PCR amplification are performed for each patient specimen. For this study of non-subtype B specimens, however, only a single attempt at amplification was made with 1 mL samples and primers that were optimized on subtype B specimens (hereafter referred to as \"standard primers\"). In the event that samples yielded a non-reportable tropism result, re-testing was performed using modified primers that were optimized for subtypes A, C and D but retained high performance for subtype B (hereafter referred to as \"optimized primers\"). All re-testing with optimized primers was performed using the same version of Trofile^®^that was used for the initial 1 mL test. Thus for South African and Ugandan specimens, all re-testing was performed with the original assay, and for Indian specimens, all re-testing was performed with the enhanced assay. Sample size and statistical analysis ------------------------------------ The study goal was to include at least 171 patients with reportable tropism results in each country, subtype and treatment experience stratum to provide approximately 5% precision for estimating prevalence of R5 in TN patients (anticipated to be 85%) and 7.5% precision for estimating prevalence of R5 in TE patients (anticipated to be 50%). Demographic and HIV disease characteristics were summarized by country, subtype and treatment experience strata. The prevalence of R5, X4, and D/M HIV-1 among patients with reportable tropism results was estimated in each stratum. The relationship between R5 infection and the following variables was examined in univariate logistic regression analyses: age (per year); gender; CD4^+^cell count (per 50 cells/mm^3^); HIV-1 RNA level (per 10,000 copies/mL); ART experience status; time since HIV diagnosis (months); mode of transmission; and Centers for Disease Control and Prevention (CDC) HIV disease category. CD4^+^count and HIV-1 RNA level were also analyzed using quartiles. Variables significantly associated with tropism in univariate analyses (p \< 0.05) were entered into multivariate logistic regression analyses, in which two-way interactions between all covariates were examined categorically. Backward elimination from a saturated model was used to determine the final predictive model. Results ======= Patients -------- The demographic and clinical characteristics of patients with reportable HIV-1 tropism results are summarized in Table [1](#T1){ref-type="table"}. All but four patients in South Africa and Uganda identified themselves as African: three self-identified as mixed race, and one as Caucasian. All but one patient in India (who self-identified as Asian) identified themselves as Indian. In all countries, nearly all patients reported heterosexual contact as the mode of HIV transmission. ###### Demographics and clinical characteristics of patients with reportable tropism results India South Africa Uganda ---------------------------------------------------- --------------- --------------- ----------------- --------------- ------------- --------------- --------------- **Age:**mean (SD) 36 (8.2) 39 (7.5) 37 (7.9) 38 (8.7) 38 (8.7) 35 (8.4) 33 (8.3) **Gender:**% male 65% 82% 41% 32% 31% 35% 22% **Time since HIV diagnosis**(months): mean (SD) 23 (36.2) 68.4 (43.7) 48.9 (31.5) 37.7 (35.4) 62.8 (48.1) 32.6 (26.6) 46.6 (19.0) **Number of treatment failures:**mean (SD) NA 1.5 (1.6) 1.3 (0.58)\* NA 1.2 (0.4) NA 1.0 (0.2) **Absolute CD4^+^cell count (cells/μL):**mean (SD) 155.7 (128.5) 114.0 (110.1) 204.0 (146.1)\* 334.3 (178.9) 83.4 (64.8) 361.0 (198.0) 109.3 (124.2) **CD4^+^count:**range 3-775 6-504 1-681 5-1171 8-223 10-1255 1-589 **Viral load (log~10~copies/mL):**mean (SD) 5.52 (5.46) 5.40 (5.42) 5.09 (5.55) 5.27 (5.33) 5.60 (5.20) 5.11 (5.19) 5.15 (5.22) **Log viral load: r**ange 3.16-5.88 3.16-5.88 3.04-6.63 3.40-5.88 3.13-5.86 3.33-5.85 3.07-5.88 NA, not available; SD, standard deviation; TE, treatment-experienced; TN, treatment-naïve.\* Missing 1; n = 204. Haemophilia/coagulation disorder and/or blood transmission were reported as transmission factors by three patients (1.5%) in South Africa and two patients (1.1%) in India. Four patients (one in South Africa, one in Uganda, and two in India) reported other unspecified risk behaviours. Risk was not reported or identified by 21 patients (8.8%) in India, two patients (1.0%) in South Africa, and one patient (0.9%) in Uganda. Most patients in South Africa and Uganda were female; most Indian patients were male. Indian patients had lower mean CD4^+^counts than South African and Ugandan patients; however, mean viral loads were consistent across the three countries (Table [1](#T1){ref-type="table"}). TE patients were more likely (53-79%) to have CDC category C AIDS-defining events than TN patients (12-42%), regardless of country (Figure [1](#F1){ref-type="fig"}). Across treatment experience, the proportions of patients with CDC category C HIV disease were higher in India (79% of TE and 42% of TN patients) than in South Africa or Uganda (Figure [1](#F1){ref-type="fig"}). {#F1} Subtype analysis ---------------- Samples for subtype analysis were collected from 307 patients (95 TE, 212 TN) in India, 678 (96 TE, 582 TN) in Uganda, and 297 (all TE) in South Africa. All samples with HIV-1 RNA exceeding 500 copies/mL were submitted for subtype testing. HIV-1 subtype could not be determined in 2.9%, 2.4%, and 5.9% of samples from India, Uganda, and South Africa, respectively, due to samples compromised by collection or handling problems and/or atypical genetic sequences (inability to generate HIV-1 pseudoviruses due to a rare genetic sequence within the patient\'s virus that is cleaved by the restriction enzyme used in the assay). Recruitment of TE patients with reportable tropism in Uganda and India, where HIV treatment is limited, was a challenge and required reducing the numbers of patients in these strata. TN samples were submitted for subtyping up until the enrolment goal of at least 171 patients had been reached. Subtype C was predominant in India (overall, 280 \[91%\]: 202 \[95%\] TN and 78 \[82%\] TE patients) and South Africa (275 \[93%\] TE patients). The predominant subtypes in Uganda were A/A1 (overall, 324 \[48%\]: 285 \[49%\] TN and 39 \[41%\] TE patients) and D (overall, 203 \[30%\]: 167 \[29%\] TN and 36 \[38%\] TE patients). Other subtypes representing ≥ 1% of the Ugandan cohort were A/D (8%), AE (2.4%), C (2.2%), and Complex (1.5%). Tropism ------- Samples from patients with subtype C virus from India and South Africa and subtypes A/A1 and D from Uganda that were initially tested for co-receptor tropism using a single 1 mL aliquot of patient plasma yielded the following reportable results, respectively, when RNA amplification was performed with the standard primers: 240 (86%), 205 (75%), 236 (81%) and 136 (67%) (Figure [2](#F2){ref-type="fig"}). After re-testing all South African, Ugandan and Indian non-reportable samples using optimized primers for RNA amplification, the reportable rate improved to 90-97% in the different strata. {#F2} Overall, most patients with reportable tropism results had only R5 HIV-1, but the proportion with R5 virus varied substantially by country, subtype and treatment experience (Figure [3](#F3){ref-type="fig"}). The proportion with only R5 virus ranged from 30% among TE Ugandan patients infected with subtype D to 97% among TN Indian patients infected with subtype C. D/M-tropic HIV-1 was less frequent, ranging from 3% among TN subtype C-infected patients in India to 65% among TE patients from Uganda with subtype D infection. X4 HIV-1 was rare and found in only eight TE patients from South Africa, and in two TN subtype A-infected patients and one TE subtype D-infected patient from Uganda. {#F3} Although co-receptor tropism varied according to country in TE patients with subtype C infection, R5 tropism was consistently high in TN patients with subtype C. R5 tropism was identified in 97% of TN patients with subtype C from India (Figure [3](#F3){ref-type="fig"}) and, in the MERIT study of maraviroc \[[@B35]\], a similar analysis identified R5 tropism in 94% of 293 TN patients with subtype C, of whom a large majority (92%) were South African. Variables significantly associated with tropism appear in Figure [4](#F4){ref-type="fig"}. Higher CD4^+^count was associated with an increased likelihood of R5 tropism in patients from both India and South Africa (Figures [4A](#F4){ref-type="fig"} and [4B](#F4){ref-type="fig"}). Additionally, in South African patients, older age and higher HIV-1 viral load predicted R5 tropism (Figure [4B](#F4){ref-type="fig"}). Among Ugandan patients infected with subtype A/A1 HIV-1, higher CD4^+^counts were associated with R5 tropism and CDC category B disease was negatively associated with R5 tropism (Figure [4C](#F4){ref-type="fig"}), whereas among patients infected with subtype D, only higher CD4^+^counts predicted R5 tropism (Figure [4D](#F4){ref-type="fig"}). {#F4} Discussion ========== In this study, CCR5 was the predominant co-receptor used in non-B subtype HIV-1 in ART-naïve individuals from India and Uganda, and in TE individuals from South Africa. In contrast to prior reports, this study evaluated a large number (n = 817) of both TN and TE patients infected with non-B subtypes (C, A, D). There were difficulties recruiting TE patients in India and Uganda, which may indirectly reflect the availability of ART, estimated by the World Health Organization to reach only 31% of HIV-infected persons in these countries \[[@B36]\]. Most (83%) study subjects with reportable tropism results had only R5 HIV-1 detectable, including 90% of TN and 72% of TE patients. However, tropism distribution varied by subtype, CD4^+^cell count and treatment experience, which can be considered a marker of advanced disease. Indian subjects with subtype C virus were almost exclusively infected with R5 regardless of their CD4^+^count, HIV-1 disease status, or treatment experience. The prevalence of R5 virus was similarly high in TN South African patients with subtype C infection but, while most common and relatively well-preserved, was somewhat lower in the South African TE cohort of this study (71%). The difference observed between R5 prevalence in India and South Africa may be a result of differences in CD4+ cell count. Compared with the Indian cohort, there was a higher proportion of individuals in the South African cohort with a D/M phenotype (25%) despite similar time since HIV diagnosis and similar exposure to prior therapies. Exclusive use of X4 was rare (4%). Our findings are consistent with previous reports of subtype C strains from India and elsewhere, where R5 HIV-1 is maintained even in advanced stages of disease. For example, some similarities can be observed between the findings of the present study and a previous study that assessed co-receptor tropism of HIV-1 strains in China: mainly that R5 tropism was seen exclusively in subtype C infections, and remained stable over time \[[@B24]\]. These data highlight the persistence of CCR5 tropism even in long-term infections, possibly related to differences within the *env*variable loops \[[@B37]\], and contrasts with other subtypes, especially subtype B, where X4 HIV-1 is more frequent during later stages of infection \[[@B6],[@B8]-[@B10],[@B16],[@B18],[@B21],[@B37]-[@B40]\]. The distribution of HIV-1 subtypes in India and South Africa was consistent with published epidemiologic data \[[@B6]\]. However, this study found more subtype diversity within the Ugandan cohort than previously reported \[[@B41],[@B42]\]. Despite the existence of multiple subtype variations in Uganda, the predominant subtypes were A/A1 (49.0%) and D (28.7%), and there were no marked differences in demographics or HIV disease characteristics. Among Ugandan subjects with subtype A/A1 virus, R5 infection predominated in TN individuals but was reduced in TE individuals, similar to the pattern seen with subtype B. While subtype D TN patients were predominantly R5, X4-using virus (D/M and X4) was detected in nearly 70% of the small number of TE patients. This agrees with earlier reports, where Ugandan individuals infected with subtype D virus showed predominantly D/M tropism \[[@B6],[@B16]-[@B19],[@B21],[@B22]\]. Multivariate analyses of other cohorts have identified higher CD4^+^cell count as a predictor of R5 infection in both TN and TE patients with subtype B and C HIV-1 infections \[[@B43],[@B44]\]. In the present study, multivariate analysis identified CD4^+^count as a significant predictor of CCR5 tropism for subtypes C, A/A1, and D. In Ugandan patients with subtype A/A1, R5 use was negatively associated with the presence of CDC category B disease, with a negative trend for category C. This study provided an opportunity to assess the ability of the Trofile^®^assay to provide results in different non-B subtype HIV-1 viral subtypes collected in developing countries. While initial testing using standard amplification primers with just a single 1 mL aliquot of patient plasma yielded a lower reportable rate (67-86%) than that expected from experience with subtype B virus, non-reportable samples that were re-tested with the original Trofile^®^assay (South Africa and Uganda) or enhanced assay (India) using modified primers optimized for non-B subtype virus, yielded a much higher percentage of reportable results (90-97%), and the distribution of R5 and non-R5 HIV-1 did not change. These improvements in amplification primers - which also retain high performance for subtype B specimens and are currently the primers used regularly for RNA amplification in the Trofile assay - may be advantageous when Trofile testing is used in clinical populations where non-subtype B virus is common \[[@B45]\]. Globally, these data offer guidance on the potential use of CCR5 antagonists and tropism tests in resource-poor clinical settings. Since HIV transmission appears to be mostly associated with R5 HIV-1 regardless of subtype \[[@B46]-[@B48]\] and a large majority of chronic HIV-1 infections in India and South Africa are R5, CCR5 antagonists may have a role in prevention efforts: for example, use as microbicides or in pre- and post-exposure prophylaxis. Combining a CCR5 antagonist with another ARV agent active against all or X4-using viruses may improve the potential utility of CCR5 antagonists in ARV-based HIV-1 prevention strategies. In the context of evaluating patients for CCR5 antagonist therapy, the high proportions of R5 HIV-1 in both TN (97%) and TE (94%) patients in India suggests that alternative methods for determining co-receptor status could be explored in this setting. Tropism could potentially be tested using genotypic or phenotypic co-receptor tropism tests and/or virologic response to short-term CCR5 antagonist monotherapy. Conclusions =========== This study examined the correlates of HIV-1 co-receptor tropism in virus from individuals with non-B subtype HIV-1 infection from India, Uganda and South Africa. While only R5 virus was detectable in the majority of TN individuals infected with subtypes C, A and D, infection with only R5 virus was maintained, albeit at lower levels, in TE individuals. However, R5 prevalence varied by subtype: patients with subtype C, especially in India, were almost exclusively infected with R5, while subtype D patients were more frequently infected with D/M viruses. The high prevalence of R5 HIV-1 in subtype C infections makes a CCR5 antagonist an attractive option for ART-based prevention strategies in areas where subtype C predominates. While a higher CD4^+^cell count was highly correlated with CCR5 use regardless of subtype, a further understanding of HIV-1 co-receptor use in individuals and populations infected with non-subtype B HIV-1 is still needed. Given the heterogeneity of HIV-1 co-receptor use in non-subtype B HIV-1, there remains a need for prospective determination of HIV-1 tropism prior to the initiation of CCR5 antagonist therapy in geographic locations where the burden of infection is greatest. Competing interests =================== Quazi Ataher and Simon Portsmouth are employees of Pfizer Inc. and own Pfizer stock. Sybil Eng, Anna Greenacre and Randy Tressler were employees at Pfizer Inc. at the time that this study was conceived and conducted. Randy Tressler currently owns Pfizer stock and is eligible for a Pfizer pension. Laura Napolitano is an employee of Monogram Biosciences and its parent company, LabCorp, and owns LabCorp stock. Andrew Kambugu has received honoraria from Abbott. Robin Wood has received advisory fees from Merck, GlaxoSmithKline, Abbot and Tibotec. Sharlaa Badal-Faesen declares that she has no competing interests. Authors\' contributions ======================= QA conducted analyses of clinical samples and led manuscript preparation. LAN led and performed the standard and optimized primer data analysis. SP and SE participated in data interpretation. AG conducted statistical analyses of study data. AK, RW and SBF participated in study design, enrolled study subjects and collected clinical samples for analysis. RT conceived of the study, and participated in its design and in data interpretation. All authors reviewed and edited the manuscript. Acknowledgements and funding ============================ The study was supported by funding from Pfizer Inc. Editorial assistance was provided by William Hatch of Health Interactions and Lynsey Stevenson of Complete Medical Communications, and funded by Pfizer Inc. The authors thank the study participants, as well as the investigators: Dr JK Maniar, Jaslok Hospital and Research Centre, Mumbai; Dr GD Ravindran, St. John\'s Medical College & Hospital, Bangalore; Dr Tripathi, Chattrapati Shahuji Maharaj Medical University (Upgraded King George\'s Medical College), Lucknow; Dr S Pujari, Ruby Hall Clinic, Pune; Dr N Kumarasamy, YRG Center for AIDS Research & Education, Chennai; Dr AR Pazare, Seth GS Medical College & King Edward Memorial Hospital, Mumbai; Dr N Khippal, S.M.S. Medical College and Hospital, Jaipur; and staff at the participating study sites. We also thank Monogram Process Development for its role in the construction and implementation of optimized primers at its own expense. Jeannette M Whitcomb, of Monogram Biosciences, initiated the plan to re-test study specimens with optimized primers and led the optimization of the Trofile^®^assay primers. Carmeliza Santos, also of Monogram Biosciences, managed all initial testing and re-testing procedures and data reporting for this study. The Monogram Biosciences Clinical Reference Laboratory performed all assays.

Introduction ============ Myopia is the main cause of preventive blindness worldwide, especially in adolescents ([@B1], [@B2]). Thus, it is one of the main priorities among the five projects under the 'Vision 2020 Action' launched by WHO ([@B3]). In recent years, the incidence of myopia has increased rapidly worldwide ([@B4]), especially among adolescents in East and Southeast Asia ([@B5], [@B6]). The prevalence of myopia among adolescents is at 96.5% ([@B7]) in South Korea, 81.6% ([@B8]) in Singapore and 95.5% ([@B9]) in Shanghai. Myopia not only affects adolescents' school performance and future career choice ([@B10]) but also causes glaucoma, cataract and other serious complications ([@B11]). Thus, many researchers have devoted themselves to gaining a more in-depth understanding of the prevention and control of adolescent myopia ([@B12], [@B13]). Collaborative research networks can help other researchers expand their field of research or join groups conducting related studies. Bibliometric studies of scientific collaboration have been conducted in various fields ([@B14], [@B15]), providing different levels of cooperation frequency in research practice. One of the methods used to study such collaboration is the co-authorship network analysis, which focuses on finding patterns of contacts or interactions between social actors. Author, country, and institution are the subjects of cooccurrence relationship; thus, analyzing their cooccurrence relationship can better reflect the truth of scientific research and academic communication, because the cooperation of authors, institutions and countries can measure the cooperation at different levels ([@B15]). However, to date, no bibliometric analysis of scientific literature in myopia prevention and control had been carried out and published. As such, this study aimed to describe the diversity of cooperation among authors, institutions, and countries in the study of adolescent myopia prevention and control. Specifically, for adolescent myopia prevention and control research, our main goal is to explore the following content: firstly, analysing the overall status of collaborative research among authors, institutions and countries; secondly, determining the institutions and authors at the core of the cooperative research network; and thirdly, identifying countries that have a strong cooperative relationship. Materials and Methods ===================== The search for papers to be included in the analysis was conducted in one day (Sep 25, 2017) to avoid bias resulting from daily updating in the database. The Web of Science Core Collection annually collects a large number of journals and records each publication, including bibliographic information (i.e., author, institution and country or region), which we used to locate publications. All papers published within the period of 1997--2016 were evaluated. Search terms included combinations of terms, such as 'adolescent', 'children', 'student', 'myopia', 'myopic', 'prevention', 'control' and 'management'. Literature types, such as meeting abstracts, letters, correction, news item, book chapter, retracted publication, editorial material, non-English literature and repeated articles, were excluded. To ensure reliability, profile information of each included article was extracted by two independent reviewers, resulting in a reliability check of 100% of the selected abstracts. A search query that was used for data extraction from Web of Science database looked like this: TS= ((adolescent myopia OR children myopia OR student myopia OR adolescent myopic OR children myopic OR student myopic) AND (prevention OR control OR management)). Social network analysis (SNA) is a method of structural analysis applied in many research fields. It focuses on relationship research and is mainly used to describe and measure relationships and information between individuals ([@B16]). SNA has been proven to be effective in studies on scientific collaboration network ([@B17], [@B18]). The same method is used in the current study. To analyze and identify critical issues, we used SATI (Statistical Analysis Toolkit for Informetrics) (ver. 3.2) to build the co-occurrence matrix ([@B19]) and transformed the data format with Ucinet 6.0 ([@B20]) to finally obtain co-occurrence mapping. VOS viewer (Visualisation of Similarities viewer) software (ver. 1.6.6) was employed to draw the co-country (region) maps by using literature title packets ([@B21]). Excel 2016 (Microsoft, Redmond, DC, USA) and Netdraw (ver. 2.118) were also used in the research. In addition, some measures of our network, including degree centrality, betweenness centrality, closeness centrality, density, and diameter, were evaluated ([@B22]). Degree centrality refers to the number of neighbors to a node in the network ([@B15]). In this case, the greater its connection to other nodes in the network, the more important is the node. Betweenness centrality refers to the number of the shortest paths passing through a given node ([@B23]). The higher the betweenness centrality of the node, the greater the ability to control the information passed between the other nodes. The closeness centrality is used to measure the distance of one node to other nodes in a network. Nodes with high closeness centrality obtain information better than other nodes or tend to have a more direct influence on other nodes ([@B16]). Density is calculated through the actually observed ties divided by all possible ties whose value is between 0 and 1 ([@B24]). Density values tend to reach 0 in sparse networks, and close to one in tightly connected networks ([@B24]). The diameter represents the longest measuring distance in a connected network; it shows the number of steps required from one side of the network to the other ([@B16]). Ethical considerations ---------------------- This study did not require any ethical consideration as it does not include any human or animal to be the object of study. Results ======= A systematic search for publications on adolescent myopia prevention and control retrieved 624 articles in Web of Science Core Collection, excluding one duplicate. After further screening of titles and abstracts, 9 editorial materials, 4 letters and a meeting abstract were removed, leaving 610 eligible papers. The scale and overall trend of collaborative research ----------------------------------------------------- [Figure 1](#F1){ref-type="fig"} shows the number of publications issued annually and the number of papers published through collaboration with authors, institutional cooperation and country (region) cooperation. The number of papers, co-authors, co-institutions and country (region) cooperative papers has increased significantly from 1997 to 2016, particularly after 2011. In general, the total number of published articles since 1997 has increased more than six-fold, from 11 in 1997 to 79 in 2016; the institutional cooperation increased more than five-fold, the author cooperation increased by twelve-fold and the country (region) cooperation increased by fifteen-fold. {#F1} [Fig. 2](#F2){ref-type="fig"} reveals the average number of authors, institutions and countries per article from 1997 to 2016. The average number shows a gradually increasing trend. {#F2} The increase in number of authors was from 3.91 to 4.34, from 1.36 to 2.51 for institutions and from 0.82 to 1.25 for countries per paper. Overall, the rates of cooperation among authors, institutions and countries were 93%, 57.9% and 21.5%, respectively. In general, the number of SCI journal papers produced by institutional cooperation is the largest (accounting for 56.6%), followed by papers generated through intra-institutional collaboration (accounting for 36.1%) and papers produced without collaboration (accounting for only 7.4%). [Figure 3](#F3){ref-type="fig"} shows the percentage of papers studied in each of the different institution collaboration types and their changes over time. The percentages of single-author papers have decreased by 26.7% from 1997 to 2016, whereas that of institution-collaborated papers increased by 24.4%. The percentage of papers produced through single authorship has always been higher than that of institutional collaboration from 1997 to 2000 but decreased after 2006. {#F3} Authors' collaborative research ------------------------------- Results of scientific research are published in the form of papers, and the status of co-authorship in papers reflects the collaboration among authors. Researchers who study the growth of co-authorship articles produced by multiple authors regard co-authorship of papers as a significant scientometric indicator of researching on cooperation among authors ([@B25]). More important researchers were expected to have published more articles, thus scholars who published more than four articles were included in the co-authorship networks. Overall, 75 researchers with 371 co-authored experiments meet this condition. Five authors not cooperated with other authors were excluded. The research collaboration network between authors is shown in [Fig. 4](#F4){ref-type="fig"}. {#F4} Each node of the figure represents an author, and the connections among the nodes represent the collaboration relationships among authors. The weight of a link indicates the number of publications co-authored by two scholars. In this author's collaboration network, the highest degree centrality of Allen, Peter M. and O\'Leary, Daniel J. was 5.83, indicating that they had 5.83 collaborators and that they played a pivotal role in the co-authorship network. Saw, Seang Mei obtained the highest betweenness centrality manifesting that he had the ability to control collaborative relationship and that he possessed and controlled a large number of research resources. In collaborative network, the closer the distance between one author and the other, the easier it is to exchange information and build cooperative research relationship. Saw, Seang Mei and Mitchell, Paul had the highest closeness centrality, which manifested that they had the most opportunity to exchange information and establish a cooperative relationship with other authors ([Table 1](#T1){ref-type="table"}). Furthermore, we assessed the density and diameter of co-authorship network, which were 0.043 and 5, respectively. ###### Top 10 authors on centrality measures in collaborative network ***Degree*** ***Score*** ***Betweenness*** ***Score*** ***Closeness*** ***Score*** -------------------- ------------- ------------------- ------------- ----------------- ------------- Allen, Peter M 5.83 Saw, Seang Mei 72.00 Saw, Seang Mei 1.85 O\'Leary, Daniel J 5.83 Mitchell, Paul 56.00 Mitchell, Paul 1.85 Guo, Xiangming 5.56 Gao, Yang 48.00 Liu, Luo-Ru 1.84 Wang, Panfeng 5.56 Congdon, Nathan 48.00 Li, Shi-Ming 1.84 Xiao, Xueshan 5.56 Manny, Ruth E 42.00 Li, He 1.84 Li, Shiqiang 5.56 Davitt, Bradley V 40.00 Li, Si-Yuan 1.84 Jia, Xiaoyun 5.56 Qu, Jia 26.00 Kang, Meng-Tian 1.84 Zhang, Qingjiong 5.56 Tan, Donald 14.00 Qu, Jia 1.84 Price, Holly 5.39 He, Mingguang 13.00 Tan, Donald 1.84 Rae, Sheila 5.39 Liu, Luo-Ru 3.60 Pan, Chen-Wei 1.84 Institutions' collaborative research ------------------------------------ It would be helpful for us to study the academic information exchange mode in scientific collaboration by analyzing the institutional relationship network of research collaboration ([@B26]). Among 610 papers, 353 documents were produced by inter-institution cooperation and 45 papers produced by intra-institutional cooperation. These papers covered 480 actual institutions and the total appearing frequency of institutions is 1674. The largest collaboration in the sample of this study involved 22 institutions. We deleted five institutions not cooperated with other institutions and eventually selected the top 69 institutions with appearance frequencies excessing five to form a map visualizing the structure of institution's collaboration network in the field of adolescent myopia prevention and control during 1997 to 2016 ([Fig. 5](#F5){ref-type="fig"}). The size of the node indicated centrality in collaboration network. The network's density is 0.011 and diameter is 6. [Table 2](#T2){ref-type="table"} lists the top ten institutions in the adolescent myopia prevention and control research dataset based on three measures of centrality: degree, betweenness, and closeness centrality. The Smith-Kettlewell Eye Research Institute had the highest degree centrality and that the University of Melbourne had the highest betweenness centrality and the lowest closeness centrality. Furthermore, the Smith-Kettlewell Eye Research Institute is found to be the most critical institution in the cooperation network and enjoys a high level of cooperation with the University of Melbourne. {#F5} ###### Top 10 institutions on centrality measures in collaborative network ***Degree*** ***Score*** ***Betweenness*** ***Score*** ***Closeness*** ***Score*** ------------------------------------------------------ ------------- ----------------------------------------- ------------- -------------------------------------- ------------- The Smith-Kettlewell Eye Research Institute 3.19 University of Melbourne 397.14 the University of Melbourne 12.67 Saint Louis University 3.19 Sun Yat-sen University 262.38 National University of Singapore 12.50 University of Pennsylvania Health System 3.08 National University of Singapore 255.13 Sun Yat-sen University 12.48 University of Texas Health Science Center at Houston 2.59 Duke University 238.41 Singapore National Eye Centre 12.41 University of Arizona 2.35 University of California, San Francisco 228.00 Duke University 12.27 Children\'s Hospital of Philadelphia 2.09 University of New South Wales 150.88 The Hong Kong Polytechnic University 12.23 Orbis International 1.82 The Hong Kong Polytechnic University 140.97 The Chinese University of Hong Kong 12.21 Stanford University 1.71 The University of Arizona 119.22 Peking University 12.18 Chinese Academy of Sciences 1.58 Aston University 118.00 Capital Medical University 12.18 European University of Madrid 1.56 The Chinese University of Hong Kong 117.56 Anyang Eye Hospital 12.16 Countries' (regions') collaborative research -------------------------------------------- The research articles produced by international cooperation have greater influence ([@B26]). Our dataset involved 53 countries, excluding 15 countries whose paper production had been less than two in the past 20 years. Using the rest of the 38 countries, we constructed a collaboration network through the VOS viewer, which helped depict the relationships between these countries. [Figures 6](#F6){ref-type="fig"} and [7](#F7){ref-type="fig"} are visual presentations of the collaborative networks among different countries (regions). In [Fig. 6](#F6){ref-type="fig"}, the size of the node represents the number of papers produced in that country or region, wherein the thickness of the links is positively correlated with the strength of the collaboration ([@B27]). The colors represent the collaboration clusters and each color represents a separate cluster ([@B27]). Hot spots are colored in red and appear lighter as they go farther away from the center of gravity. The thickness of the links represents the strength of collaborations between the countries (regions) it connects. In [Fig. 7](#F7){ref-type="fig"}, seven major clusters can be distinguished: the largest one gathering around China, the next one around the USA and the other clusters gathering around Australia, England, Singapore, Germany and Ireland. The highest density in the network belonged to China, USA and Australia. {#F6} {#F7} Discussion ========== The increase of myopia incidence is related to many factors, such as genetic ([@B28]), environmental risk factors ([@B1]) and unhealthy lifestyles ([@B29]). The early onset of myopia is reported to cause faster development and higher possibility of occurrence in adulthood ([@B30]). Therefore, prevention and control of myopia in adolescence are of great significance. We can try to do more intervention research on the risk factors that are controllable or changeable. Research collaboration is an important way to improve the overall strength of research and enables researchers to supplement each other\'s strengths and share information ([@B31]). We found that, in the past 20 years, the total number of papers in the field of prevention and control of myopia in adolescents, as well as the total number of co-authored articles, increased especially since 2011. Previous studies in other research fields showed a similar trend ([@B15], [@B32]). The average number of authors, institutions and countries (regions) per paper has also increased over time. In general, 93.0% of the publications are co-authored by more than one author, whereas 57.9% of the publications have been co-authored by a number of institutions over the past 20 years. On the one hand, researchers not only benefit from knowledge exchange but also save on research costs due to the sharing of information, technology and resources ([@B33]). On the other hand, as research goes deeper, it becomes increasingly difficult to achieve a great breakthrough through a single person or institution, which forces researchers in adolescent myopia prevention and control field to cooperate with others. The above analysis indirectly indicates that scientists and research institutions attach importance to cooperation; thus, researchers will have great propensity to cooperate in the adolescent myopia prevention and control field. As social networks are developed by adding new nodes and links, and the new knots are connected to old high central knots based on the principle of preferential connection ([@B34]), scholars or institutions with high degree centrality can play a vital role in the development of co-authorship networks. From the results of degree centrality analysis, Allen, Peter M. (5.83) and O\'Leary, Daniel J. (5.83) emerged as the top researchers with the most frequent collaborative activities, followed by Guo, Xiangming, Xiao, Xueshan, Li, Shiqiang, Jia, Xiaoyun and Zhang, Qingjiong. The Smith-Kettlewell Eye Research Institute and Saint Louis University are institutions with the most frequent cooperative activities. Hence, Allen, Peter M. and O\'Leary, Daniel J. had the highest number of opportunities to communicate with other members of the network and had the greatest ability to build collaborative teams in the growth and dynamics of network, as well as the Smith-Kettlewell Eye Research Institute and Saint Louis University. Our betweenness results showed that Saw, Seang Mei and the University of Melbourne had the highest scores, indicating that they play a good mediating role in the cooperative network and control in the flow of information. The closeness centrality measures the distance of one member from the other members in a network ([@B16]). In this study, two authors (Saw, Seang Mei and Mitchell, Paul) and one institution (University of Melbourne) had the highest closeness centrality, indicating that they receive information faster than others because of fewer intermediaries among them. Whether author collaboration or institutional collaboration network, network density is very low, which means that the cooperation among the authors and the relationship between the institutions are not tight enough. The low network diameter also implies that the communication between the authors and the information exchange between the institutions is slow. Visualization analysis of cooperation among countries suggests that China, the USA and Australia are the most productive countries and are also the hotspots for the study of adolescent myopia prevention and control. Compared with other studies indicating that collaborative countries are often geographically interrelated ([@B32]), our results present conflicting outcomes considering the distribution of seven major clusters. In the co-authorship network of countries, the developed countries occupy the vast majority. Although our study is one of the first attempts to systematically describe the research collaboration in this field, it also has some limitations. Firstly, in order to guarantee the homogeneity of the research samples, the books, meeting records, reports, letters, editing materials and non-English articles were excluded in the analysis, which may bring about the issue of incomplete information. Secondly, our search terms may not be used in the title of some relevant research and cannot be retrieved by our method. Thirdly, since no database is perfect and some might have bias by over-representing journals using the English language, bibliometric results should always be considered with caution ([@B35]). Conclusion ========== This study provides a systematic description of collaboration at the levels of author, institution and country (region) in the research on adolescent myopia prevention and control. Although the number of publications has been increasing, the information flow is slow and no close collaboration occurs among the authors as well as among institutions in the network. Multiple collaboration types should thus be encouraged in this field, especially among countries. Moreover, middle and low-income countries need to strengthen cooperation with developed countries. Ethical considerations ====================== Ethical issues (Including plagiarism, informed consent, misconduct, data fabrication and/or falsification, double publication and/or submission, redundancy, etc.) have been completely observed by the authors. None declared. **Funding** This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. **Conflict of interest** The authors report no conflicts of interest regarding this manuscript.

Introduction ============ Resistin is a cysteine-rich protein, which is mainly secreted from monocytes and macrophages in humans ([@b1-or-40-06-3392]--[@b3-or-40-06-3392]). It is associated with inflammation and malignant neoplasms ([@b4-or-40-06-3392]--[@b5-or-40-06-3392]). Blood resistin levels are demonstrated to be increased in certain cancer patients compared with healthy controls, including esophageal squamous cancer, gastric, colorectal, breast and endometrial cancer, and malignant lymphoma ([@b6-or-40-06-3392]--[@b12-or-40-06-3392]). Resistin is considered to be a risk factor for breast cancer ([@b9-or-40-06-3392],[@b13-or-40-06-3392]) and biomarker of disease progression of esophageal squamous cancer, gastric and colorectal cancer ([@b6-or-40-06-3392]--[@b8-or-40-06-3392]). It is an independent prognostic factor of pancreatic ductal adenocarcinoma ([@b5-or-40-06-3392]). Resistin can promote prostate cancer cell proliferation through the phosphatidylinositol 3 kinase (PI3K)/protein kinase B signaling pathway in human prostate cancer cell lines PC-3 and DU-145 ([@b14-or-40-06-3392]). However, most of the reported studies only demonstrated the association between serum or plasma resistin and malignancy. Few reports measured the level of resistin expression in cancer tissues, even though it is less well studied and controversial in lung cancer. Certain reports demonstrated a higher concentration of resistin in the blood was demonstrated in non-small cell lung cancer (NSCLC) patients compared with the controls ([@b15-or-40-06-3392]--[@b17-or-40-06-3392]). One of the reports assessed resistin expression in the marginal area of lung cancer tissue and non-cancer region by immunofluorescence staining in 10 cases ([@b17-or-40-06-3392]). Another revealed that blood resistin levels were similar between cancer group and non-cancer group ([@b18-or-40-06-3392]). Furthermore, the clinical significance and biological function remain largely unknown. However, lung cancer is one of the most common malignancies worldwide, with higher morbidity and poorer prognosis ([@b19-or-40-06-3392]--[@b20-or-40-06-3392]). The most common form of lung cancer is NSCLC, which includes lung adenocarcinoma and squamous carcinoma. At present, lung adenocarcinoma replaces squamous carcinoma as the dominating type of NSCLC. The aim of the present study is to determine the resistin expression in lung adenocarcinoma tissues, clinical significance and biological function *in vitro* and *in vivo*. Materials and methods ===================== ### Patients and tissue samples A total of 70 consecutive cases of newly diagnosed lung adenocarcinoma patients at Tianjin Medical University Cancer Institute and Hospital (Tianjin, China) from January to December 2008, with complete clinical and pathological data, were selected retrospectively in the present study and followed up for at least five years. Paired cancer and adjacent non-cancerous tissue samples, which were located more than 1 cm away from the tumor, were obtained through open surgeries. The paraffin-embedded tissue samples were stained with hematoxylin-eosin and confirmed lung as adenocarcinoma again. The clinicopathological characteristics of the patients were recorded. The tumor staging of NSCLC was defined according to the tumor, node and metastasis system. The study was comprised of 70 cases of lung adenocarcinoma (38 male cases, 32 female cases), with an average age of 61 years old (36--77 years old). All patients received treatments (including operation, chemotherapy or radiotherapy), which conformed to the guidelines of NSCLC. ### Immunohistochemistry For immunohistochemical staining, 5-µm paraffin-embedded tissue sections were heated for 1 h at 70°C, deparaffinized with a xylene soak, followed by rehydration via the addition of alcohol at decreasing concentrations (100, 95, 85 and 75%) for 5 min/step. A 96°C water-bath was used for antigen retrieval in 0.01 mol/l sodium citrate buffer (10 mM, pH 6.0) for 20 min. Next, endogenous peroxidase activity was quenched by incubation in 3% hydrogen peroxide for 15 min at room temperature. Subsequently, slides were blocked with goat serum (cat. no. ZLI-9022; OriGene Technologies, Inc., Beijing, China; 1:1) at 37°C in a wet box for 30 min and then incubated by the primary antibody (rabbit polyclonal antibody against human resistin; cat. no. BS7730; Bioworld Technology, Inc., St. Louis Park, MN, USA; 1:100) overnight at 4°C in moist chambers. Following washing with 0.01 mol/L PBS (pH 7.2) three times, slides were incubated with a biotinylated secondary rabbit anti-mouse antibody (cat. no. PV-6000; OriGene Technologies, Inc.; 1:1) for 25 min at 37°C. Following incubating in horseradish peroxidase marked streptomycin avidin working fluid at 37°C for 30 min, slides were treated with avidin biotin-peroxidase complex using 3,3′-diaminobenzidine as a chromogen, and then counterstained with hematoxylin for 30 sec at room temperature and examined by light microscopy (Olympus Corporation, Tokyo, Japan). PBS was used as a negative control instead of a primary antibody. ### Evaluation of immunohistochemical staining Existing tan or brown particles in the nucleus or cytoplasm indicated positive cells, which must conform to the following conditions: i) The cellular structure was clear; ii) the location of positive granules was accurate; iii) staining was significantly increased compared with the background. In a 400× high power filed, randomly selected from 10 different cancer cell fields of view, the percentage of (a) positively stained cells was calculated as follows: 0--5% positive cells, score 0; 6--25% positive cells, score 1; 26--50% positive cells, score 2; 51--75% positive cells, score 3; 76--100% positive cells, score 4. Then, the (b) staining intensity was evaluated: Colorless, score 0; light yellow, score 1; deep yellow and tan, score 2; brown, score 3. The expression of resistin was based on the product of (a) × (b): Score 0, negative (−); score 1--4, weakly positive (+); score 5--8, positive (++); score 9--12, strongly positive (+++). (−) and (+) were regarded as low expression group, (++) and (+++) were categorized as the high expression group ([Fig. 1A](#f1-or-40-06-3392){ref-type="fig"}). The result of each specimen was independently evaluated by two qualified and expert pathologists, blinded to the patients\' clinical data. The few cases with discordant results were reevaluated and final scores were consensual. ### Cell culture Human lung adenocarcinoma cell lines A549 and H1975 were obtained from the Committee of Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). The cells were maintained in RPMI-1640 (Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Gibco; Thermo Fisher Scientific, Inc.) at 37°C in a humidified atmosphere of 95% air and 5% CO~2~. Cells were divided according to transfection of overexpression resistin plasmid for the resistin group and empty vector for the control group. ### Transfection and isolation of stable transfectants Lipofectamine™ 2000 Reagent (Invitrogen; Thermo Fisher Scientific, Inc.), endo-free maxiplasmid kit (Tiangen Biotech Co., Ltd., Beijing, China); pcDNA3.1-(+)/resistin plasmids were established by OriGene Technologies, Inc. (cat. no. RC210942) and the primers were as follows: Forward primer, 5′-CCCACCGAGAGGGATGAAAG-3′ and reverse primer, 5′-CAGTGACATGTGGTCTCGGC-3′; forward primer, 5′-CAGCTCACCATGGATGATGATATC-3′ and reverse primer, 5′-AAGCCGGCCTTGCACAT-3′ (β-actin). A fragment of the rat resistin cDNA fragment (285 bp) was inserted at the unique *Eco*RI site in the anti-sense orientation as determined by sequencing. The final concentration of resistin was 100 nM. A total of 2×10^6^ A549 and H1975 cells grown in 60 mm Petri dishes were transfected with 10 µg of the recombinant plasmid using lipofectamine, as described by the supplier (Gibco; Thermo Fisher Scientific, Inc.). A total of 24 h later, fresh RPMI-1640 media containing 10% FBS was added and replaced 48 h later. Monoclonal cells were selected with NeoR. Then the cells were cultured for 24 h following transfection. ### Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) Total RNA was extracted from the cells by TRIzol (cat. no. 15596026; Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer\'s protocol. Prime Script RT reagent kit (DRR037A; Takara Bio, Inc., Otsu, Japan) were used for cDNA generation (42°C for 30--60 min, 70°C for 15 min). RT-qPCR was performed with Super Real PreMix (cat. no. FP204-01; Tiangen Biotech, Co., Ltd., Beijing, China) by the following program: 95°C for 3 min in 1 cycle; 95°C for 5 sec, 58°C for 30 sec, and 72°C for 30 sec in 35 cycles. To ensure the DNA production, a melting curve analysis was performed according to ABI Step One system. The relative gene expression was normalized to the internal standard β-actin using 2^−ΔΔCq^ method ([@b4-or-40-06-3392],[@b17-or-40-06-3392],[@b21-or-40-06-3392]). The primers are 5′-TGGAGTGCCAGAGCGTCACCT-3′ (forward) and 5′-ACTGGCAGTGACATGTGGTCTC-3′ (reverse). ### Western blotting Western blotting was used to verify successful transfection. Whole cell extracts of A549 and H1975 were prepared with a CellLytic™ M reagent (cat. no. C2978; Sigma-Aldrich; Merck KGaA, Darmstadt, Germany), the protein was quantified by a Bicinchoninic acid (Pierce; Thermo Fisher Scientific, Inc.) assay. Then, the protein samples (50 µg) were separated by 10% SDS-PAGE. The samples were blocked with goat serum (cat. no. ZLI-9022; OriGene Technologies, Inc.; 1:1) for 60 min at room temperature and detected by western blotting using rabbit polyclonal resistin antibody (cat. no. BS7730; Bioworld Technology, Inc.; 1:500), mouse anti-β-actin (cat. no. ab8226; Abcam, Cambridge, UK; 1:1,000), mouse anti-proliferating cell nuclear antigen (PCNA; cat. no. ab29; Abcam; 1:500), rabbit anti-Ki67 (cat. no. ab16667; Abcam; 1:1,000), rabbit polyclonal to caspase-3 (cat. no. ab13847; Abcam; 1:500), rabbit monoclonal to caspase-7 (cat. no. ab32522; Abcam; 1:1,000), mouse anti-matrix metalloproteinase (MMP)2 (cat. no. ab37150; Abcam; 1:1,000) and mouse anti-MMP9 (cat. no. ab38898; Abcam; 1:1,000) monoclonal antibodies at 4°C overnight, and the secondary antibody was goat anti-rabbit horseradish peroxidase (cat. no. ZDR-5306; ZSGB-BIO; OriGene Technologies, Inc.; 1:10,000) for rabbit polyclonal resistin antibody, rabbit anti-Ki67, rabbit polyclonal to caspase-3, rabbit monoclonal to caspase-7 or goat anti-mouse IgG(H+L)-HRP (cat. no. LK2003; Tianjin Sungene Biotech Co., Ltd.; 1:5,000) for mouse anti-β-actin, mouse anti-proliferating cell nuclear antigen, mouse anti-MMP2 and mouse anti-MMP9 monoclonal antibodies. The gray values (cat. no. C8420; Coomassie brilliant blue G-250, Beijing Solarbio Science & Technology Co., Ltd., Beijing, China) were analyzed by using the Odyssey V3.0 software (Thermo Fisher Scientific, Inc.). ### Cell proliferation assay by colony formation A549 and H1975 Cells were cultured in 6-well plates, 300 cells/well prior to being fixed in methyl hydrate room temperature for 10 min. Then the colonies were stained by 1% Crystal Violet Staining Solution at room temperature for 5 min and counted with a light microscope (Inverted microscope; Leica Microsystems GmbH, Wetzlar, Germany). ### MTT assay MTT assay was also used to observe and compare cell proliferation ability of A549 and H1975. A total of 2×10^3^ cells were plated into a well of 96-well plates and 10 ml, 5 mg/ml MTT was added to each well and continued to culture for 4 h. Then following dimethyl sulfoxide addition, the plates were placed on a microplate autoreader (Bio-Rad, Laboratories, Inc., Hercules, CA, USA). Optical density was read at 570 nm wavelength and cell growth curves were determined according to the optical density value. ### Apoptosis analysis by flow cytometry The A549 and H1975 cells were dosed 24 h following plating and then tested according to the protocol of Biolegend kit (cat. no. 640906, Biolegend, Inc., San Diego, CA, USA). Cells were resuspended in Annexin V binding buffer at a concentration of 10^6^ cells/ml. Following transferring 100 µl cell suspension to 5 ml test tube, 5 ml of Annexin V-fluorescein isothiocyanate and 10 µl of propidium iodide solution were added to the cell suspension. A total of 400 µl binding buffer was added to each tube 15 min later, the apoptosis was analyzed using a flow cytometer (CytExpert analysis software 2.0; Beckman Coulter, Inc., Brea, CA, USA). ### Cell scratch assays The A549 and H1975 cells were seeded to full confluence in 6-well plates overnight. A scratch was introduced in the middle of the well with a sterile pipette tip the following day. The medium was discarded and replaced. The rate of migration towards the center of the wound was determined using calipers in the image under a light microscope 48 h later. ### Cell invasion assays The invasion assays were performed with an 8.0-µm pore inserts in a 24-well Transwell chambers (Corning Incorporated, Corning, NY, USA). The A549 and H1975 cells (2×10^5^/well) and Dulbecco\'s modified Eagle\'s medium were added to the upper chamber of Transwell coated with Matrigel (BD Biosciences, Franklin Lakes, CA, USA). The RPMI-1640 with 10% FBS was added to the lower chamber and incubated at 37°C for 24 h. Cells that had migrated to the bottom of the filter were stained with a three-step stain kit (Thermo Fisher Scientific, Inc.) at room temperature for 5 min. The cells were counted by light microscope from each chamber. A total of 5 fields of view were counted. ### Xenografts assays in vivo All surgery was performed under sodium pentobarbital anesthesia. A total of 3% sodium pentobarbital (50 mg/kg) by intraperitoneal injection was used. Athymic BALB/c nude mice (16 female, aged 5 weeks, 25 g) were provided by Slac Laboratory Animal Co., Ltd. (Shanghai, China). Mice were housed in a pathogen-free animal facility at 18--29°C. The humidity is 40--70%. The mice had access to food and water with 12-h light/dark cycle. They were randomly divided into 2 groups according to the A549 cell groups described above. There were 8 mice/group. A total of 0.1 ml serum-free RPMI-1640 with 2×10^6^ cells were subcutaneously injected into the right flank of each mouse. The control mice were injected with stable A549 cell lines transfected with empty vector. The tumors were measured by vernier caliper on day 14, 17, 21, 23, 26 and 29. Mice were sacrificed at 29 days post inoculation. The final volume of tumor tissues was calculated with the following equation: Tumor volume (mm^3^) = tumor length (mm) × tumor width (mm) × tumor height (mm)/2. ### Statistical analysis Each experiment was repeated three times. All statistical analyses were performed using SPSS 20.0 software (IBM, Corps., Armonk, NY, USA). The Spearman method was used to analyze the correlation of resistin expression with clinicopathological variables. Kaplan-Meier method was used to perform survival analysis and evaluate the differences between survival curves by log-rank test. The hazard ratio and confidence interval was calculated by univariate and multivariate Cox regression model. The experiments\' results *in vitro* and *in vivo* were recorded as the mean ± standard deviation. A student\'s two-sided t-test was used to compare values of test and control samples. P*\<*0.05 was considered to indicate a statistically significant difference. Results ======= ### The expression of resistin in lung adenocarcinoma tissues Lung adenocarcinoma tissues exhibited different levels of resistin expression and adjacent normal lung tissues/non-cancer tissues stained negative. A total of 41.4% of cancer tissues demonstrated high resistin expression (58.6% demonstrated low resistin expression; [Table I](#tI-or-40-06-3392){ref-type="table"}). The expression of resistin was different between ≤65 and \>65 years, tumor size ≤3 and \>3 cm, non-lymph node metastasis and lymph node metastasis as well as early stage and advanced stage ([Table I](#tI-or-40-06-3392){ref-type="table"}). ### The association between resistin expression and clinicopathological characteristics The expression of resistin in lung adenocarcinoma tissues was significantly, positively correlated with tumor size, lymph node status and clinical stage (P\<0.05) and significantly, negatively correlated with progression-free survival (PFS) and overall survival (OS; P\<0.01). There is no correlation with age at diagnosis, smoking, drinking, body mass index (BMI) and blood type ([Table II](#tII-or-40-06-3392){ref-type="table"}). ### Survival analysis Increased PFS and OS were observed in the patients with low resistin expression and low expression groups as determined by the log-rank test. Comparing the low resistin expression group with the high one, the prognosis of the former was improved (P\<0.01; [Fig. 1B](#f1-or-40-06-3392){ref-type="fig"} and [1C](#f1-or-40-06-3392){ref-type="fig"}). Survival analysis demonstrated that factors were significantly associated with PFS and OS, including smoking, age, lymph node status, resistin expression, clinical stage and chemoradiotherapy (P\<0.05). The last three one were independent risk factors of PFS and OS in patients with lung adenocarcinoma ([Table III](#tIII-or-40-06-3392){ref-type="table"}). ### The influence of resistin on biological behavior of A549 and H1975 cell lines To test the influence of resistin on biological behavior of A549 and H1975 cell lines, the resistin overexpression cell lines was established through resistin plasmids. Although the expression of resistin in the lung adenocarcinoma cell lines was demonstrated, in the reported studies, the level of resistin in lung adenocarcinoma cells A549 and chondrosarcoma cells was low ([@b17-or-40-06-3392],[@b21-or-40-06-3392]). The results of western blotting also indicated that the bands of resistin in cell lines A549 and H1975 were weak. In addition, one study reported that the resistin gene was transfected into the PC-3 cells to assess the effect of overexpression of resistin in prostate cancer cell line PC-3 ([@b14-or-40-06-3392]). Same as above, the lung adenocarcinoma cells are not resisitin-dominant expression cells and secretory cells, the resistin level using overexpression was manipulated rather than being knocked-down in A549 cells and H1975 cells. The level of resistin was detected by RT-qPCR and western blotting in A549 and H1975 cell lines, which demonstrated a significantly increased level compared with the control (P\<0.05; [Fig. 2](#f2-or-40-06-3392){ref-type="fig"}). Then, colony formation and MTT assays were performed to test the proliferation of A549 and H1975 cells, and the results demonstrated that compared with the control cells, stable overexpressed resistin cells obviously increased the ability of proliferation *in vitro* ([Fig. 3A and B](#f3-or-40-06-3392){ref-type="fig"}). Furthermore, to investigate the mechanism, the expression level of proliferation-associated proteins including PCNA and Ki67 were tested, and the results demonstrated that the protein expression of PCNA and Ki67 significantly increased in the group resistin compared with the control (P\<0.05; [Fig. 3C and D](#f3-or-40-06-3392){ref-type="fig"}). Taken together, the results of the present study indicated that resistin could regulate the proliferation by increasing the associated proteins including Ki67 and PCNA *in vitro*. Flow cytometry was used to detect alterations in cell apoptosis and the results indicated that the resistin overexpressing A549 and H1975 cells demonstrated significantly decreased apoptotic ability (P\<0.05; [Fig. 4A](#f4-or-40-06-3392){ref-type="fig"}). Furthermore, the results were confirmed by detecting the level of the apoptosis associated proteins, including caspase-3 and caspase-7. A significantly decreased level in the resistin overexpression group was demonstrated compared with the control group (P\<0.05; [Fig. 4B and C](#f4-or-40-06-3392){ref-type="fig"}). In addition, a scratch test and Transwell invasion assay was used to detect alterations in cell migration and invasion. The results were as follows: Compared with the control cells, the resistin overexpression cells demonstrated significantly reduced scratches and more invading cells through the membrane (P\<0.05; [Fig. 5A and B](#f5-or-40-06-3392){ref-type="fig"}). Resistin promoted A549 and H1975 cell migration and invasion *in vitro*. Furthermore, to investigate the mechanism of the above, the expression level of the proteins that reflect invasion and metastasis including MMP2 and MMP9 was tested. The expression of MMP2 and MMP9 was demonstrated to be significantly increased in the resistin group compared with the control (P\<0.05; [Fig. 5C and D](#f5-or-40-06-3392){ref-type="fig"}). In summary, the results indicated that resistin could strengthen the invasive capacity of cancer cells by regulating the associated proteins including MMP2 and MMP9. The results above indicated that resistin could promote A549 and H1975 cell proliferation, migration and invasion as well as suppress apoptosis *in vitro*. ### The promotional effect of resistin for lung cancer in vivo The biological roles of resistin in lung cancer tumorigenesis were further examined by xenograft studies in nude mice. A549 cells transfected with resistin plasmids or an empty vector were inoculated subcutaneously into the upper back of nude mice. Following four weeks the tumor volume of resistin group was significantly larger compared with the control group (P\<0.05), as presented in [Fig. 6A and B](#f6-or-40-06-3392){ref-type="fig"}. To validate the expression of resistin in mice tumors, immunoblotting analysis was performed. The results revealed that the resistin protein level was significantly increased in the resistin plasmids group (P\<0.05; [Fig. 6C](#f6-or-40-06-3392){ref-type="fig"}). Taken together, these indicated the tumorigenic effect of resistin *in vivo*. Discussion ========== In the present study, the resistin expression of 70 patients with lung adenocarcinoma was analyzed by immunohistochemistry. Resistin expression was demonstrated to be increased in lung adenocarcinoma tissues compared with the paired adjacent normal lung tissues. It was in accordance with previous reports that NSCLC patients exhibited a higher blood level of resistin in contrast to controls ([@b15-or-40-06-3392]--[@b17-or-40-06-3392]). Compared with the non-cancerous regions, Kuo *et al* ([@b17-or-40-06-3392]) also demonstrated a higher level of resistin in the marginal areas of human lung cancer tissue by immunofluorescence staining. It was also similar to that presented in cancer tissues of breast cancer, colorectal cancer, pancreatic ductal adenocarcinoma and prostate cancer ([@b5-or-40-06-3392],[@b13-or-40-06-3392]--[@b14-or-40-06-3392],[@b22-or-40-06-3392]). It was also demonstrated that the expression of resistin in lung adenocarcinoma tissues increased as the size of tumor and clinical stage of the cancer progressed. The resistin level was increased in patients with lymph node metastasis compared with the ones without lymph node metastasis. As well as in breast cancer tissues, the resistin expression was positively associated with tumor stage, tumor size and lymph node status ([@b13-or-40-06-3392],[@b23-or-40-06-3392]). In pancreatic ductal adenocarcinoma, resistin expression was strongly and positively associated with tumor stage ([@b5-or-40-06-3392]). The blood level of resistin was also elevated with progression in tumor stage in patients with gastric cancer and colorectal cancer ([@b7-or-40-06-3392],[@b24-or-40-06-3392]). In general, resistin is correlated with poor clinicopathological status. It was also demonstrated that there was no correlation of resistin expression with sex, age at the point of diagnosis, smoking, drinking and blood type in patients with lung adenocarcinoma in the present study. To summarize the existing reports, resistin was associated with sex in certain types of cancer, which mainly occurred in females, including breast cancer and endometrial cancer ([@b10-or-40-06-3392],[@b11-or-40-06-3392]); but in certain cancer that mainly occurs in males, resistin level demonstrated no significant sex differences, including esophageal ([@b6-or-40-06-3392]), gastric ([@b7-or-40-06-3392],[@b25-or-40-06-3392]--[@b26-or-40-06-3392]), colorectal ([@b1-or-40-06-3392]--[@b2-or-40-06-3392],[@b8-or-40-06-3392],[@b27-or-40-06-3392]--[@b28-or-40-06-3392]) and pancreatic cancer ([@b29-or-40-06-3392]). Karapanagiotou *et al* ([@b16-or-40-06-3392]) also demonstrated that serum resistin level was unassociated with sex, age and BMI in NSCLC patients. In the present study, the expression of resistin in lung adenocarcinoma tissues also demonstrated no sex differences prior to and following being divided into low and high expression groups. This may be due to there being no significant sex differences for resistin in lung adenocarcinoma patients; or that the morbidity of lung cancer is increasing annually in female in recent years, the difference between male and female is decreasing; or that the sample size is small in the present study. Further research with an expanded sample size is required. Nevertheless, resistin expression level exhibited a negative correlation with PFS and OS by bulk analysis. Increased PFS and OS were observed in the patients with low resistin expression (−/+) and in the low resistin expression groups (− to +). The prognosis of the low expression group was improved compared with the high one (++ to +++). Multivariate Cox regression analysis demonstrated that resistin expression, clinical stage and chemoradiotherapy were independent prognostic factors of PFS and OS in the patients with lung adenocarcinoma. To a certain extent, this was in line with a few reports. Higher resistin expression in cancer tissues was associated to poor prognosis in breast cancer patients ([@b13-or-40-06-3392],[@b23-or-40-06-3392]). It may be an independent prognostic factor of breast cancer and pancreatic ductal adenocarcinoma ([@b5-or-40-06-3392],[@b13-or-40-06-3392]). With respect to lung cancer, one report demonstrated that the serum resistin levels exhibited a trend of association with time to relapse, but it was not a predictive factor for OS ([@b15-or-40-06-3392]). Another study reported that it was unconnected with diagnosis and prognosis ([@b16-or-40-06-3392]). However, the present study\'s results were mainly discussing resistin level within the tumor tissues. Most researchers believed that the levels of adipokine (including leptin, resistin and adiponectin) in blood circulation cannot accurately reflect their true levels in human body ([@b30-or-40-06-3392]--[@b31-or-40-06-3392]). Resistin level in cancer tissues or the tumor microenvironments was higher than in blood circulation and more closely correlated with tumorigenesis and tumor progression ([@b30-or-40-06-3392]--[@b31-or-40-06-3392]). It demonstrated that NSCLC cancer tissue-specificity and is more reliable than circulating level. Furthermore, a series of *in vitro* and *in vivo* assays were performed. Resistin overexpression was verified in A549 and H1975 cell lines, and could promote cell proliferation, migration and invasion as well as inhibit apoptosis *in vitro*, and serve a tumorigenic function of lung adenocarcinoma *in vivo*. The present study demonstrated suggested that resistin may work by increasing proliferation associated proteins Ki67 and PCNA, while decreasing the apoptosis associated proteins caspase-3 and caspase-7. In 2015, it was demonstrated that resistin could promote chondrosarcoma metastasis and MMP2 expression through activation of the AMP-activated protein kinase/p38 signaling pathway and downregulation of microRNA-519d expression ([@b21-or-40-06-3392]). Consistent with the above study, it was demonstrated that resistin may promoted lung adenocarcinoma migration and invasion by increasing MMP2 and MMP9. Resisitin also promoted breast cancer progression via Toll-like receptor 4 (TLR4)/nuclear factor (NF)-κB/signal transducer and activator of transcription 3 signaling ([@b23-or-40-06-3392]). Recently resistin was reported to be strongly expressed in lung adenocarcinoma tissues. Resistin promoted lung adenocarcinoma metastasis via TLR4/Src/epidermal growth factor receptor/PI3K/NF-κB pathway ([@b32-or-40-06-3392]). These studies have important referential significance and value for the potential molecular mechanisms of resistin in lung adenocarcinoma. There were certain limitations in the present study. First, the study only investigated adenocarcinoma and not other pathological subtypes of NSCLC. Nevertheless, lung adenocarcinoma is considered as the main type of NSCLC at present. One study reported that the serum level of resistin was not correlated to the histological type of NSCLC ([@b15-or-40-06-3392]). Furthermore, in the present study, the potential molecular mechanism was only investigated *in vitro*. In conclusion, the expression of resistin in pathological tissues and its association with corresponding clinicopathological data in 70 consecutive patients with lung adenocarcinoma was studied for the first time to the best of our knowledge. It was demonstrated that high resistin expression was predominantly observed in lung adenocarcinoma tissues. It is associated with a more malignant clinicopathological status and poorer survival. Analysis demonstrates resistin expression is an independent prognostic factor for PFS and OS. Resistin could promote A549 and H1975 cell proliferation, migration and invasion while inhibit their apoptosis *in vitro*. Resistin also serves a tumorigenic function *in vivo*. The present study will be helpful to make clear the exact role of resistin in lung adenocarcinoma. Not applicable. Funding ======= The present study was supported by the National Natural Science Foundation of China (grant no. 81401957) and Tumor translational medicine seed fund of Tianjin Medical University Cancer Institute and Hospital (grant no. 1317). Availability of data and materials ================================== The data used during the present study are available from the corresponding author upon reasonable request. Authors\' contributions ======================= CCZ, JC, RFN and CGZ conceived and designed the study. CCZ, JC and YL collected the data. CCZ, JC and RFN performed the data analysis and interpretation. CCZ and JC wrote the manuscript and revised the important intellectual content. All authors edited and approved the final manuscript. Ethics approval and consent to participate ========================================== The research involving human samples and animal experiments had been approved by the Ethics Committee of Tianjin Cancer Hospital (Tianjin, China). All experiments were conducted according to relevant national and international guidelines. Informed consent was obtained from all participants included in the study. Patient consent for publication =============================== Informed consent was obtained from all participants included in the study. Competing interests =================== The authors declare that they have no competing interests. ###### The expression of resistin in lung adenocarcinoma tissues and its association with survival. (A) Immunohistochemical staining of lung adenocarcinoma tissues with a resistin antibody. The low and high expression of resistin in lung adenocarcinoma tissues are presented (magnification, ×100). High resistin expression indicated poor prognosis in lung adenocarcinoma patients. (B) PFS a curves of 70 patients with different level of resistin expression are presented. The PFS curves of age, smoking, tumor size, lymph node metastasis and clinical stage are presented. (C) OS curves of 70 patients with different level of resistin expression are presented. The OS curves of age, smoking, tumor size, lymph node metastasis and clinical stage are presented. PFS, progression-free survival; OS, overall survival; Cum, cumulative; (−) and (+), low expression group, (++) and (+++), high expression group.    {#f2-or-40-06-3392} {#f3-or-40-06-3392} {#f4-or-40-06-3392} {#f5-or-40-06-3392} {#f6-or-40-06-3392} ###### Difference of resistin expression in patients with lung adenocarcinoma. Clinicopathological characteristics All (n=70) Low expression (n=41) High expression (n=29) χ^2^ P-value ------------------------------------- ------------ ----------------------- ------------------------ ------- --------- Sex 0.721 0.396   Male 32 17 15   Female 38 24 14 Age at diagnosis (years) 9.587 0.002   ≤65 46 33 13   \>65 24 8 16 Smoking 0.170 0.681   Yes 39 22 17   No 31 19 12 Drinking 0.059 0.808   Yes 23 13 10   No 47 28 19 Tumor size (cm) 3.934 0.047   ≤3 34 24 10   \>3 36 17 19 Lymph node metastasis 4.749 0.029   Yes 28 12 16   No 42 29 13 Clinical stage 6.346 0.012   Early stage (I--II) 39 28 11   Advanced stage (III--IV) 31 13 18 ###### Correlation of resistin expression and clinicopathological characteristics. The expression of resistin in lung adenocarcinoma tissues ------------------- ----------------------------------------------------------- --------- -------- --------- Sex −0.036 0.768 −0.101 0.403 Age 0.199 0.099 −0.216 0.072 Smoking 0.191 0.425 0.049 0.686 Drinking 0.021 0.861 0.029 0.811 Blood type −0.158 0.192 −0.097 0.425 Body mass index −0.106 0.384 −0.077 0.527 Tumor size 0.307 0.010 0.237 0.048 Lymph node status 0.261 0.029 0.260 0.029 Clinical stage 0.408 \<0.001 0.394 0.001 PFS −0.419 \<0.001 −0.379 0.001 OS −0.429 \<0.001 −0.416 \<0.001 PFS, progression-free survival; OS, overall survival. ###### Univariate and multivariate analysis of PFS and OS. PFS OS ----------------------- --------- ------- -------------- --------- --------- ------- -------------- ------- Sex 0.385 0.584 Drinking 0.536 0.617 Smoking 0.029 1.022 0.496--2.104 0.954 0.040 0.826 0.408--1.671 0.594 Age at diagnosis 0.001 1.595 0.835--3.045 0.157 0.002 1.190 0.597--2.370 0.621 Tumor size 0.019 1.725 0.936--3.176 0.080 0.016 1.659 0.860--3.198 0.131 Lymph node metastasis \<0.001 1.603 0.868--2.960 0.132 0.001 1.612 0.831--3.129 0.158 Clinical stage \<0.001 2.349 1.268--4.352 0.007 \<0.001 2.028 1.075--3.826 0.029 Resistin expression 0.002 1.856 1.003--3.436 0.049 0.007 1.895 1.005--3.574 0.048 Chemoradiotherapy \<0.001 0.059 0.013--0.256 \<0.001 \<0.001 0.027 0.004--0.210 0.001 PFS, progression-free survival; OS, overall survival; HR, hazard ratio; CI, confidence interval. [^1]: Contributed equally

INTRODUCTION {#sec1-1} ============ The main goal of periodontal treatment is to control the inflammation in periodontal tissues and to regenerate the lost tissues predictably. To meet this goal it is critical to guide the tissues capable of regeneration.\[[@ref1][@ref2][@ref3]\] Guided tissue regeneration is an accepted method for enhancement of lost periodontal tissue. In this technique a barrier membrane is used to prevent epithelial cell migration and stabilization of the clot into the defect. This prevention results in the migration of periodontal ligament cells and osteoblasts into defect and these cells are known to be responsible for tissue regeneration.\[[@ref4]\] Different types of barrier membranes are introduced that had shown favorable results due to different studies.\[[@ref5]\] These membranes are different in composition and structure, but all of them prevent the migration of epithelial and gingival connective tissue cells into the defect and ideally, a barrier membrane should enhance the cell attachment and migration of the progenitor cells.\[[@ref5][@ref6][@ref7][@ref8][@ref9][@ref10]\] Wound healing is a complex process which includes cell migration, cell attachment to various extracellular matrix components, and cell proliferation.\[[@ref11][@ref12]\] Cell attachment process is a four-step sequence which includes adsorption of glycoproteins to the substrate surface, cell contact, attachment, and spreading.\[[@ref9][@ref10]\] Cell proliferation begins after these events.\[[@ref5]\] Tissue integration property ensures the stabilization of the wound and inhibits the migration of epithelial cells, which results in better gain of clinical attachment levels.\[[@ref13][@ref14][@ref15]\] According to their degradation characteristics, barrier membranes are divided into two groups of resorbable and non-resorbable membranes. Collagen is the most common material used as resorbable membranes.\[[@ref5]\] It facilitates hemostasis and wound stability by promotion of platelet aggregation along with fibroblast migration which accelerates wound closure,\[[@ref16][@ref17]\] but collagenous membranes are not stiff enough to resist soft tissue pressure during healing.\[[@ref16][@ref18]\] Polytetrafluroethylene (PTFE) is the main composition of non-resorbable membranes.\[[@ref19]\] Although their biocompatibility and positive effect on bone regeneration was shown, but a second surgery is required for their removal which may traumatize the newly formed immature periodontal tissue and causes patient discomfort and increases the treatment time and cost.\[[@ref20]\] Also, the membrane stiffness may result in tissue dehiscence which is the main reason of treatment failure 3 weeks after membrane placement and exposes the membrane which leads to bacterial infection and decrease in the levels of gained clinical attachment.\[[@ref21][@ref22][@ref23][@ref24]\] An alternative to an expanded PTFE membrane is a high-density polytetrafluroethylene (d-PTFE) membrane which is commercially available as TXT-200 and GBR-200. High-density polytetrafluroethylene membranes have small porosities, so bacterial contamination is eliminated and therefore there is no need of primary closure when they are being used and they can be left exposed to the oral cavity.\[[@ref25][@ref26][@ref27][@ref28]\] The acellular dermal matrix (Alloderm) was originally introduced in medicine for reconstructive plastic surgeries but is also used in dentistry in various periodontal procedures like root coverage and keratinized tissue augmentation around teeth and implants.\[[@ref29][@ref30][@ref31]\] It has many advantages, but the absence of cells and vessels makes tissue incorporation slower, therefore, attempts of culturing fibroblasts on Alloderm were performed to achieve early wound healing and decrease wound contraction in periodontium.\[[@ref32][@ref33][@ref34][@ref35]\] Fibroblasts play an important role in the healing process. It has been shown that the key factor in the success of regenerative treatment is the recruitment or delivery of cells to the defect site and the production of suitable extracellular matrix along with the periodontal tissues.\[[@ref36][@ref37]\] Introduction of specific cell adhesion molecules to the membrane surfaces may lead to specific tissue responses. Different growth factors and proteins have been introduced and one of them is enamel matrix derivatives. A commercially available product of enamel matrix derivatives is called Emdogain^®^ (EMD). It is an acidic extract of low molecular weight procine enamel proteins mainly amelogenin and a propylene glycol alginate vehicle.\[[@ref38][@ref39]\] Different studies showed that EMD enhances the adhesion, proliferation, and matrix production of periodontal ligament fibroblasts, stimulates cell growth, and production of insulin growth factor-1 and transforming growth factor-β1 in periodontal ligament cells although it has no appreciable effect on osteoblastic differentiation and has no effect on epithelial cells.\[[@ref37][@ref38]\] All of the described characteristics of EMD make it a suitable functional material for regenerative treatments. Therefore, its effects on cell adhesion to different materials were investigated in the present study. There was also no available study that had compared the fibroblast adhesion among TXT-200, GBR-200, Alloderm, and collagenous membrane (RTM Collagen, Cytoplast^®^) or the effect of EMD on fibroblast attachment to these common barrier membranes. The present study was performed to compare cell adhesion among the prementioned membranes and also to investigate the effect of EMD on gingival fibroblast attachment. MATERIALS AND METHODS {#sec1-2} ===================== For this experimental *in vitro* study, gingival fibroblast cells (NCBI Codece C165) were provided by Pasteur Institute of Iran. Cells were cultured in a culture flask and cultured in the presence of Dulbecco\'s modified Eagle medium (DMEM, Sigma-Aldrich, St. Louis, MO, USA) containing 10% Fetal Calf Serum and 100 μg/ml of penicillin, streptomycin, and amphotericin B. The flask was kept in 37°C in a 5% CO~2~ atmosphere in an incubator with humidity. The medium was changed twice a week. Cells were cultured for 3 weeks and passaged for five times. Four different barrier membranes were used in this study. Two non-resorbable dense polytetrafluoroethylene membranes GBR-200 (GBR1224, LOT: 2541) (Cytoplast®, Osteogenic Biomedical, Lubbock, TX, USA), TXT-200 (TXT1224, LOT: 3688) (Cytoplast®), RTM Collagen (RTM2030, LOT:C2030263) (Cytoplast®) and acellular dermal matrix (ADM, 302111, LOT: B42234) (Alloderm, Biohorizons, Birmingham, AL, USA). Each membrane was cut into two 6×6-mm pieces and washed with sterile saline solution according to the supplier\'s instructions. In RTM Collagen and ADM groups, membranes were washed with sterile saline solution until the protect paper was floating. A 48 wells culture plate was used in this experiment. Five groups of four close wells were selected. Four groups were used for membranes (each group containing four wells for each membrane). All of the membranes were adapted at the bottom of the selected group of wells. No membrane was added to the fifth group and it served as a control group to check the growth of seeded cells. 10 μg/mL of EMD (LOT: C2822, Emdogain®, Straumann, Malmö, Sweden) was added to two wells of each group (EMD+) and two wells were left without any EMD (EMD-). Cells were seeded at a density of 100,000 cell/well on the membranes. Plate was placed in a 37°C incubator with humidity and 5% CO~2~ atmosphere for 24 hours. The growth of seeded cells in the fifth group was evaluated by means of a light microscope. Then cells were washed four times with phosphate buffer saline (PBS) to remove non-adherent cells. The membranes were fixed in 2.5% glutaraldehyde for 2 hours, washed five times with distilled water for 20 minutes, treated with 1% osmium tetroxide for 1 hour, washed again five times with distilled water for 20 minutes and finally dehydrated through a series of graded ethanol solutions and left for 24 hours in room temperature to dry. To finish the process, they were coated with gold and analyzed with Field Emission Scanning Electron Microscope (Hittachi s4160, Stanford, CA, USA). An operator not aware of the experimental set up analyzed the membranes with SEM. Each membrane was divided into four intellectual parts under SEM with ×300 magnifications and one image was taken from each part. Another two observers totally unaware of the experiment counted the cells on each image and if there was a difference, the least cell count was recorded. Data was analyzed by independent t-test, one-way ANOVA, two-way ANOVA, and *post hoc* LSD test with SPSS18 (version 18;SPSS Inc, Chicago, IL, USA). *P* \< 0.05 in independent t-test analysis and *P* \< 0.001 in one-way ANOVA, two-way ANOVA, and *post hoc* LSD analysis was considered statistically significant. RESULTS {#sec1-3} ======= Figures [1](#F1){ref-type="fig"}--[4](#F4){ref-type="fig"} illustrates the membranes in EMD- and EMD+ groups under SEM with ×300 magnifications and [Table 1](#T1){ref-type="table"} shows the gained data after cell counting process by two observes. {#F1} {#F2} {#F3} {#F4} ###### The mean of attached cells to membranes in EMD+ and EMD- groups  [Figure 5](#F5){ref-type="fig"} shows the mean of attached gingival fibroblasts to the barrier membranes used in this study in EMD+ and EMD- groups. {#F5} Two-way ANOVA test showed the membrane type (*P* \< 0.001) and the presence of Emdogain (*P* = 0.04) affect the gingival fibroblast adhesion efficacy. The quality of cell adhesion to each membrane in EMD+ and EMD- groups was evaluated by independent *t*-test and it was shown that cell adhesion in GBR-200 was slightly higher in EMD- group, but this difference was not statistically significant (*P* = 0.060). On the other hand, cell adhesion to TXT-200 membrane was higher in EMD- group and the difference was statistically significant (*P* = 0.020). Cell adhesion to RTM Collagen showed no significant difference between EMD+ and EMD- groups (*P* = 0.310). Unlike other membranes, ADM showed higher cell adhesion efficacy in EMD+ group and the difference was statistically significant (*P* = 0.004). All of the above results are illustrated in [Figure 6](#F6){ref-type="fig"}. {#F6} One-way ANOVA also showed that ADM has the highest cell adhesion capacity in EMD+ group and the difference was statistically significant (*P* \< 0.001). It was also shown that in EMD- group gingival fibroblasts adhesion to TXT-200 and ADM is statistically significantly higher in comparison to GBR-200 and RTM Collagen (*P* \< 0.001). *Post hoc* LSD test was used to compare membranes two by two. As it is shown in [Figure 3](#F3){ref-type="fig"}, this test revealed when EMD is present, cell adhesion to ADM is higher than GBR-200 (*P* \< 0.001), TXT-200 (*P* \< 0.001), and RTM Collagen (*P* \< 0.001). This test also showed when EMD is not present, cells significantly adhere to TXT-200 more than RTM Collagen (*P* \< 0.001) and GBR-200 (*P* \< 0.001). Also when EMD was not present, cell adhesion to TXT-200 was slightly higher than ADM, but it was not statistically significant (*P* = 0.156). DISCUSSION {#sec1-4} ========== Tissue engineering represents very exciting advances in regenerative medicine; however, periodontal literature only contains few reports.\[[@ref40][@ref41][@ref42][@ref43][@ref44]\] ADM has been shown as an useful material in gingival augmentation.\[[@ref37]\] It has many advantages, but the absence of cells and vessels makes tissue incorporation slower.\[[@ref36]\] In an attempt to solve this problem, fibroblasts were cultured on Alloderm as an alternative to achieve early wound healing and decrease wound contraction in periodontium.\[[@ref32][@ref33][@ref34][@ref35]\] In this study, EMD was used to enhance the gingival fibroblast adhesion to different membranes including Alloderm. The highest cell efficacy in all of the studied groups belonged to TXT-200 in absence of EMD followed by ADM in the presence of EMD and then ADM in the absence of EMD. When EMD was not present, GBR-200 had slightly higher cell adhesion in comparison to the presence of EMD, but this difference was not significant (*P* = 0.060). Same happened to TXT-200, but the difference was significant (*P* = 0.02). Cell adhesion to RTM Collagen was slightly higher when EMD was present but the difference was not significant in comparison to the absence of EMD (*P* = 0.310). The difference in the cell adhesion efficacy when EMD is present can be related to its mitogenic properties. Bertl *et al*.\[[@ref45]\] observed that 0.1-50 μg/mL of EMD promotes cell migration in the wound healing process and it is inhibited at 100 μg/mL. Also, in other studies it was reported that the EMD with the concentration of 25 μg/mL and lower leads to better results,\[[@ref46][@ref47][@ref48]\] so in the present study the concentration of EMD was considered 10 μg/mL for the EMD+ groups. Hoang *et al*.\[[@ref49]\] had shown that under physiologically relevant conditions, amelogenin (the main composition of EMD) does not bind to collagen. Van der Pauw *et al*.\[[@ref48]\] declared that with collagen as a substratum, EMD has an inhibitory influence on periodontal ligament cells attachment and spreading. Lyngstadaas *et al*.\[[@ref50]\] found a five-fold increase in cell adhesion on plates coated with EMD. These conflicting results may be due to the higher concentration of EMD (500 μg/mL) which was used by these authors. In the present study, cell adhesion to RTM Collagenmembrane showed no significant difference in EMD+ and EMD- groups which was similar to some of the mentioned studies.\[[@ref49][@ref50]\] ADM which has a collagenous composition showed higher cell adhesion efficacy in the presence of EMD. This result was similar to Lyngstadaas *et al*.\[[@ref50]\] study but the concentration of EMD which was used in the present study (10 μg/mL) was different form theirs (500 μg/mL). It can be concluded that ADM, *per se* has a good cell adhesion efficacy. It is derived from human skin and is prepared by a controlled process that removes epidermis and the cells from the dermis but leaves the basement membrane and extracellular matrix organization and collagen and elastin fibers undamaged.\[[@ref29][@ref32]\] Although RTM Collagen is a collagenous membrane, but similarity of ADM structure to human skin may be the reason of its better cell adhesion efficacy in comparison with RTM Collagen. In EMD- groups, TXT-200 showed statistically higher cell adhesion in comparison to GBR-200 (*P* \< 0.001) but in the presence of EMD this difference was not significant (*P* = 0.118). Although their composition is the same and they are both made of dense polytetrafluoroethylene, but their surface texture is different. TXT-200 has a roughened surface that is caused by the presence of macro-porosities on its surface but GBR-200 lacks these porosities \[Figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}\]. It seems that EMD may cover the porosities of TXT-200and decrease the cell adhesion efficacy of this material. These results show that surface texture and material structure play an important role on the cell adhesion efficacy. Cell adhesion affects the tissue integrity efficacy of biomaterials and higher tissue integrity efficacy results in better gain of clinical attachment levels.\[[@ref13][@ref14][@ref15]\] CONCLUSION {#sec1-5} ========== Within the limits of the present study, it is shown that the membranes used in this study affect cell adhesion, proliferation and differentiation of gingival fibroblasts. Also, EMD may lower the cell adhesion efficacy of GBR-200, TXT-200, and RTM Collagen but it can promote this efficacy in ADM. When membranes are used without EMD, TXT-200 shows the highest cell adhesion efficacy followed by ADM without a statistically significant difference. This study also showed not only composition of biomaterials, but also their surface texture and internal structures may play an important role in their cell adhesion efficacy. The authors are grateful to Dr. Omid Moghaddas (Islamic Azad Dental School of Tehran, Iran) for his precious concepts and Dr. Fatemeh Rahbarizadeh, Dr. Farnaz Jafari and Eng. Mohammad Mohaghegh (Tarbiat Modares University, Tehran, Iran) for their cooperation in this study. **Source of Support:** Nil **Conflict of Interest:** None declared

Background {#Sec1} ========== Bronchopleural fistula (BPF) is a relatively infrequent but potentially fatal complication of pulmonary resection. BPF can be divided into peripheral or central, based on the location of the leakage, and BPF occurs in about 1.5 to 28 % of pneumonectomy cases, and is associated with high death rate \[[@CR9], [@CR30]\]. It is estimated that incidence of BPF after pneumonectomy and lobectomy for lung cancer is 4.5--20 % and 0.5 %, respectively, and the incidence of BPF is highest after right pulmonary resection and right lower lobectomy \[[@CR31]\]. The etiology of BPF includes incomplete tumor resection, use of steroids, intraoperative infection and prolonged postoperative mechanical ventilation as major risk factors of BPF \[[@CR31]\]. The clinical manifestations of BPF can be frequently classified as acute, subacute, and chronic. An acute BPF presents as tension pneumothorax, with pleural cavity communicating abnormally with the airways, and is associated with purulent sputum expectoration, dyspnea, and reduction in established pleural effusion \[[@CR22]\]. The presentations of subacute and chronic BPF are commonly related to a pleural space with infection, manifesting as a more invisible form with fever, dry cough, and malaise with different levels of respiratory disorder \[[@CR33]\]. Traditional treatments of BPF include thoracotomy after drainage and primary repair, which is based on vascularized muscular flaps and omental grafts tissues \[[@CR20]\]. Amplatzer vascular plug, which was originally designed for the transcatheter closure of vascular structures, has also been reported as a safe and effective method to treat small postoperative BPF \[[@CR9]\]. Fruchter et al. also found that the technique of Amplatzer double-disk occluder implantation may be suitable for both large and small BPFs which originate from the main bronchi and lobar bronchi, respectively \[[@CR8]\]. Additionally, endoscopic approaches and bronchoscopy are common methods of treating BPF to avoid thoracotomy \[[@CR27], [@CR36]\]. Bronchofiberscope (BFS) is a precision instrument employed to diagnose bronchial diseases using of the light guide composed by the fine fibers formed by tens of thousands of high transmittance glass or acrylic resin \[[@CR12], [@CR16]\]. BFS is designed to offer advantageous features such as easy operation method, clear vision, mild trauma, tolerance of surgery by patients, and high safety profile, which reduces or avoids complications associated with tracheotomy and prevents local infection \[[@CR25], [@CR32]\]. Clinically, BFS has multiple uses, including removing foreign bodies, eliminating secretions, treating nasopharyngeal carcinoma, central lung cancer, alveolar cell carcinoma, esophageal fistula, hemoptysis, obstruction, assisting endotracheal intubation treatment and placing gastric tube \[[@CR1], [@CR11], [@CR21]\]. Previous studies have revealed that BFS is also an excellent diagnostic tool for early detection of various intrabronchial injuries, and the attached biopsy sampling feature is helpful in the identification of early lesions, and to carry out poly excision surgery for the studies on bronchus and lung diseases \[[@CR15], [@CR18], [@CR19]\]. Previous studies reported various treatment methods for BPF using BFS, and the methods include gelfoam, shot put plugs, and tissue adhesives. However, these methods have significant deficiencies, evident from the fact that treatment fistula under 3 mm was efficient using these methods, but they show poor efficacy in treatment of BPF beyond 3 mm, particularly those beyond 10 mm \[[@CR6], [@CR28], [@CR35], [@CR37]\]. Phenol, also named carbolic acid, is a sweet-smelling colorless liquid used to prepare resins, preservatives, fungicides, drugs (e.g., aspirin), and also is used to disinfect surgical instruments \[[@CR4], [@CR24], [@CR38]\]. 88 % carbolic acid was found to be efficacious with all alopecia areata patients and can be considered as a treatment of choice for stable alopecia areata \[[@CR3]\]. Moreover, spot peel with 88 % phenol can be a cost-effective procedure for idiopathic guttate hypomelanosis, which can be combined with other medical therapies \[[@CR26]\]. There are no studies using carbolic acid to treat BPF with the help of BFS at present. Therefore, we investigated the efficiency of carbolic acid treatment of BPF in post-pulmonectomy patients, by instilled 100 % carbolic acid with the aid of BFS. Methods {#Sec2} ======= Ethics statement {#Sec3} ---------------- This study was conducted with the approval of the Institutional Review Board of Liaoning Tumor Hospital, Shenyang. The informed written consent was collected from each eligible patient and the whole study was performed based on the Declaration of Helsinki \[[@CR14]\]. Study population {#Sec4} ---------------- A total of 12 patients with post-pulmonectomy BPF were enrolled at the Department of Thoracic Surgery, Liaoning Tumor Hospital, Shenyang between February 2009 and March 2012. Orificium fistulae were confirmed by bronchoscope and the average diameter was 4.5 mm. The eligible patients included eight males and three females, with an average age of 56 years (range, 45 \~ 71 years). Three patients had BPF after the right pneumonectomy, six after the left pneumonectomy, one after the right middle and low lobectomy and two after left upper lobectomy. Preoperotive preparation {#Sec5} ------------------------ Electrocardiogram, routine blood tests and biochemical examination were performed in all the patients. Patients were fasted for 4 \~ 6 h in preparation for surgery and received 10 mg diazepam and 1 mg atropine via intramuscular injection about 30 min before operation. In addition, 1 % lidocaine was used for nasopharyngeal anesthesia by nebulizer. Intraoperative methods {#Sec6} ---------------------- All patients were instructed to take supine position except 2 patients with short breath in sitting position. The BFS (Olympus BF1T40) was inserted into the trachea through nasal cavity. Heart rate, blood pressure and SpO2 was monitored. Patients received local nasopharyngeal anesthesia with 2 % lidocaine to alleviate irritant reaction. The bronchus around the suture was bubbling when the patient breathed deeply. The fistula was observed via BFS. After the drainage of secretion, hematocele or pus around the BPF, a bronchoscopy biopsy forceps was used for removing necrotic tissues and a 1.8 mm flexible tube was guided through the biopsy hole. The distal end of BFS was brought out and fixed 0.3 cm above the fistula. With breath holding, 100 % carbolic acid solution (0.5--1.0 ml) was instilled to bronchial mucosa through the BFS. The bronchial mucosa became pale after treatment and finally the flexible tube and bronchoscope were removed. Postoperative and histological observation {#Sec7} ------------------------------------------ After the surgery, the patients were treated with closed drainage of thoracic cavity, anti-inflammatory, symptomatic and supportive treatments. Gas discharge in thoracic drainage tube was observed, and fistula healing were measured via BFS. The treatments were repeated if there was gas discharge from thoracic drainage tube, or further observations were made. Patients could leave hospital after blood routine test showing no evidence of dyspnea, fever, positive culture of fluid drainage (3 times). Paraffin sections (4 \~ 6 μm) of bronchial stump were stained by hematoxylin and eosin (HE) to observe the irritation of bronchial stump after instilled with carbolic acid solution. Results {#Sec8} ======= Outcome characteristics of BPF {#Sec9} ------------------------------ In the 12 patients with BPF, the median diameter of the BPF orifice was 4.5 mm, according to the intraoperative observation. Specifically, 3 patients showed a fistula diameter of 3 mm or smaller, 6 patients showed a fistula diameter of 3 \~ 5 mm, and 3 patients exhibited a fistula diameter of 5 mm or larger, 1 of whom had a fistula diameter of 7 mm (Table [1](#Tab1){ref-type="table"}). Serious complications, such as haemorrhage, severe dyspnea and SpO2 declines, did not occur in all the 12 patients during bronchoscopic therapy. Of note, BPF orifices in 5 patients closed after 5 treatments with carbolic acid, 1 patient through 2 treatments, 1 patient through 3 treatments, 2 patients through 4 treatments and 3 patients through 7 treatments (Fig. [1](#Fig1){ref-type="fig"}). Follow-up was conducted for six months after bronchoscopy. Based on the data collected, the average treatment time of the 12 patients was calculated as 20 min and the average time of fistula closure was 30 days. Importantly, the cure rate was 100 %.Table 1Characteristics and outcomes of BPF patientsAge\ (y)GenderInitial symptomSurgical methodBronchopleural fistulaeSize (mm)Treatment timesCure time (d)Follow up148maleLow feverRight PNY3535alive256maleHigh feverRight PNY3.5535alive350maleBlood sputumLeft PNY1321alive471maleIrritating coughLeft upper LBY1.5428alive557femaleLow feverRight upper LBY4535alive665maleFever/air bubbleLeft PNY5749unknown764maleCough/feverLeft PNY7749alive859maleCough/feverRight PNY1214alive962maleLow feverLeft upper LBY4749alive1058maleFever/sputumLeft PNY5535alive1145maleLow feverLeft PNY3.5428alive1256malesputumLeft PNY4535alive*BPF* bronchopleural fistula, *y* years, *PNY* pneumonectomy, *LBY* lobectomy, *d* daysFig. 1The treatment process of carbolic acid: **a** the BPF (about 1 mm); **b** infused by the carbolic acid; **c** BPF narrowed after 2 times instillation, and bubbles appeared after carbolic acid infused with saline; **d** repositioned the fistula, and three times later the BPF healed. Note: BPF, bronchopleural fistula HE staining {#Sec10} ----------- HE staining was performed to observe bronchial stumps stimulated by carbolic acid infusion. The biopsy showed that the white flat hyperplasia tissue (bronchial stumps tissue) after carbolic acid treatment was inflammatory granulation tissue. Furthermore, it was found that the tissue was loose and dropsical, with a small amount of proliferation of irregularly distributed fibroblast, small vascular proliferation, a large number of plasma cells and lymphocyte infiltration, and a small amount of irregularly arranged squamous epithelial hyperplasia, as shown in Fig. [2](#Fig2){ref-type="fig"}.Fig. 2Results of HE staining observing bronchial stumps tissues stimulated by carbolic acid infusion. Note: HE, hematoxylin and eosin Discussion {#Sec11} ========== BPF is defined as an abnormal communication between a lobar or the main bronchus and the pleural space, and continues to be a severe surgery complication, which is related to high morbidity and mortality \[[@CR13]\]. Risk factors associated with BPF incidence are fever, steroid use, anemia, leukocytosis and tracheostomy, elevated erythrocyte sedimentation rate, Haemophilus influenzae in sputum and bronchoscopy for sputum suction or mucus plugging \[[@CR2]\]. Recently, a number of flexible bronchoscopic techniques have been used to seal BPFs. These materials include cyanoacrylate-based glues, absorbable gelatine sponge, vascular embolisation devices, and fibrin compounds \[[@CR7], [@CR10]\]. In this study, we describe a novel approach of using carbolic acid for the closure of fistulas. Carbolic acid has a strong reaction with mucosal tissues, and pure carbolic acid corrodes mucosa completely in 60 s. When carbolic acid contacts the mucosal surface, the mucosa tissues is rapidly degenerated (pale), and mucosal inflammation stimulate exudation and proliferation, finally resulting in the closure of fistula \[[@CR34]\]. Carbolic acid is widely used for disinfecting appendiceal stump in appendicitis operation and in the treatment of suspected TB contaminants in tuberculosis surgery \[[@CR5], [@CR17]\]. We describe a simple, safe and effective way to instill 100 % carbolic acid through BFS in the treatment of BPF. The 12 patients treated with carbolic acid successfully reached fistula closure, with the total effective rate at 100 % without any adverse reactions of hemoptysis and dyspnea. A reasonable explanation might be that carbolic acid is relatively safe, a small amount of acid liquid overflow will not cause serious injuries to normal mucosal \[[@CR29]\]. A number of advantages are embodied in the instillation of carbolic acid under BFS. First, this method is easy to perform and, based on it success rate in this study, likely to be readily accepted by the patients, and patient hospitalization is unnecessary if they are in good condition. Second, for patients who have larger fistula with significant pleural effusion and sputum, this therapeutic tool can rapidly relieve the symptoms and avoid aspiration pneumonia. Third, the fistula location, size, shape can be clearly orientated. Finally, it can help reduce operative risks, decrease mortality rate as well as the cost of treatment for BPF \[[@CR13]\]. Our results also revealed that the healing time of fistula is positively correlated to its size. Based on the clinical observations of series of cases, we summarize that fistula \< 5 mm healed in much shorter time compared to fistula ≥ 5 mm. However, if the size of fistula exceeded certain limit, it might be difficult to heal due to the potential lack of healthy mucosa to stimulate proliferation and regenerate tissue \[[@CR23]\]. Our study presents clear evidence that use of carbolic acid for BPF treatment under the inspection of BFS is safe and 100 % effective. However, our findings need to be interpreted with caution due to limitations in the study. A limitation is the small number of patients with BPF who underwent BFS. Therefore, our study contained a relatively smaller sample size, which might restrict the application of our results to a wider population. Further studies using large sample size and better study designs will be necessary to confirm our findings. Conclusion {#Sec12} ========== In conclusion, we achieved 100 % efficacy in treatment of BPF with carbonic acid through BFS, with BPF size ranging from 3--7 mm in diameter. The described procedure is simple, safe and an effective choice for BPF patients, with little pain and at relatively low cost. **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** Z Wang and YY Liu carried out the molecular genetic studies, HB Yu participated in the sequence alignment and Q Luo drafted the manuscript. All authors read and approved the final manuscript. We would like to acknowledge the reviewers for their helpful comments on this paper.

The content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. Introduction ============ Ganglion cysts are soft tissue masses that arise from the joint capsule and tendons which are commonly seen on the dorsal side of the hand and wrist \[[@REF1]\]. One of the treatment options is surgical excision \[[@REF2]\]. Surgery is usually performed with local anesthesia infiltration in sedated-unsedated patients or under general anesthesia. In recent years, ultrasound (US) guidance has made distal peripheral nerve blocks of the upper extremity technically safe and feasible options to achieve anesthesia for hand and wrist surgery \[[@REF3]\]. Nowadays, US-guided regional blocks should be the first choice for such surgical procedures where the possible risks of general anesthesia and the necessity of clear airways are considered \[[@REF4]\]. US-guided nerve blocks have many advantages, such as the avoidance of nerve damage, with a clear definition of nerves from the surrounding structures, control of the distribution of local anesthetics, visualization of needle position, faster block onset time, improved block qualities, and a reduction in the volumes of local anesthetics \[[@REF5]-[@REF6]\]. The success rate of regional anesthesia has increased when performed under ultrasound guidance. Ultrasound identifies nerves easily, which cannot be defined by surface anatomic landmarks and alternative sites can be determined by scanning the nerve along its route \[[@REF6]-[@REF7]\]. Forearm blocks may especially be useful for minor surgeries of the hand. However, background data are limited regarding the performance of a peripheral nerve blockade at the level of the upper arm for hand surgery. There is also a lack of reports about the optimal sites for needle insertion regarding US-guided radial nerve blocks. Most studies about median, ulnar, and radial nerves present these blocks as rescue techniques for a failed or incomplete proximal (infraclavicular, axillary, interscalene, or supraclavicular) upper extremity block \[[@REF4],[@REF8]-[@REF11]\]. The mid-humeral region enables anesthesiologists to selectively administer local anesthetics to different nerves and block the four main nerves of the upper extremity separately \[[@REF12]\]. This technique may have advantages over proximally performed approaches, such as the avoidance of needle trauma to central structures, decreased motor blockade, and smaller amounts of local anesthetic drugs used to achieve anesthesia for a narrower area \[[@REF4]\]. In this case report, we present three cases of US-guided radial nerve blocks performed at the mid-humeral region. Efficient surgical anesthesia was achieved for ganglion cyst excision at the hand dorsum using the minimal dose of local anesthesia. We discuss the potential indications and advantages of a US-guided radial nerve block at the mid-humeral region for patients undergoing hand surgery. Case presentation ================= Written informed consent has been obtained from all patients for the anesthetic/surgical procedures and for the publication of this case report. All patients underwent the excision of ganglion cysts at the dorsum of the hand (Figure [1](#FIG1){ref-type="fig"}). {#FIG1} Standard monitorization (electrocardiogram, blood pressure, pulse oximetry) was applied at the block room. An intravenous cannula was placed and infusion of 0.9% NaCl solution was started. The patients were sedated by administering 0.03 mg/kg of midazolam and 0.5 µg/kg of fentanyl intravenously. After sedation, the position of the arm was set at the 90 degrees abduction of the arm and 90 degrees of the forearm on the surgery table. The ultrasound probe was placed transversally at the mid-humeral region, on the posterolateral aspect of the arm. The needle was entered from the lateral side of the arm in the medial direction within the plane of the ultrasound beam (Figure [2](#FIG2){ref-type="fig"}). {#FIG2} The block area was sterilized with iodine solution and the US transducer was covered with a sterile cap. We performed needle insertion at the mid-humeral region, which is located midway between the anterior process of the acromion and the lateral epicondyle of the humerus, as described by Foxall et al. \[[@REF6]\]. The radial nerve could be visualized easily by locating the ultrasound probe on the posterolateral aspect of the arm at this level. A 13 MHz linear transducer was used (LOGIQ P5®, General Electric, USA). The radial nerve was visualized as an oval, heterogeneous structure that consisted of hypoechoic and hyperechoic structures, which represent the nerve fascicles and connective tissue. Using the in-plane technique, a 22 Ga 8 cm echogenic needle (Stimuplex® Ultra 360® Braun, Melsungen, Germany) was introduced into the mid-humeral region, aiming to enter the fascial plane next to the radial nerve. After the localization of the needle tip on the radial nerve, 0.15 ml/kg of 0.5% bupivacaine was injected. The distribution of the local anesthetic drug and the tip of the needle was visualized in real time during the procedure, aiming to spread around the nerve (Figure [3](#FIG3){ref-type="fig"}). {#FIG3} We did not use a nerve stimulator, as the nerve was clearly identified under ultrasound guidance. The patients were followed up with pinprick tests after the blocks for loss of sensation. The incisions were within the sensory dermatomal innervation area of the radial nerve (Figure [4](#FIG4){ref-type="fig"}). {#FIG4} Case 1 A 33-year-old female patient presented with complaints of swelling and pain in the dorsum of the right hand. She had a 2x2 cm mass and a soft cystic lesion, which was diagnosed as a ganglion cyst at the dorsum of the hand (Figure [1](#FIG1){ref-type="fig"}). Past medical history was unremarkable with American Society Anesthesiology (ASA) classification I. Surgical history examination revealed two cesarian-section operations. Ten milliliters of local anesthetic (bupivacaine 0.5%) was administered to encircle the radial nerve without entering the humeral side. Surgical anesthesia was achieved at the 25th minute after local anesthetic administration. A mild motor block was observed, as the patient could move her hand parallel to gravity. The patient was cooperative and reported minor discomfort during the excision of the cyst from its base where it originated, but there was no need for additional analgesic drugs during the surgery. There were no symptoms of cardiovascular, respiratory, or central nervous system side-effects. The surgery proceeded uneventfully and lasted about 20 minutes. The block was considered successful without the necessitation of conversion to general anesthesia. Case 2 A 28-year-old female patient with complaints of swelling in the wrist dorsum of the right hand. The patient\'s medical and surgical history was unremarkable and evaluated as ASA I class. Fifteen milliliters of local anesthesia (10 ml bupivacaine 0.5 % and 5 ml lidocaine 2%) was administered around the radial nerve under ultrasound guidance. The block procedure was uneventful. The patient was cooperative during the operation and did not report pain at the beginning of the surgery. During the excision of the cyst from its base, the patient complained of discomfort. Fentanyl 50 µg intravenous was administered and 3 milliliters of 2% prilocaine was infiltrated to the surgical area. The surgery lasted 30 minutes, uneventfully. The block was considered successful without the need for conversion to general anesthesia. Case 3 A 24-year-old male patient with a complaint of swelling at the wrist dorsum of the right hand was diagnosed with a ganglion cyst. The patient was evaluated as ASA I class with no remarkable medical and surgical history. After identifying the radial nerve under ultrasound guidance, 10 milliliters of 0.5% bupivacaine was administered. There were no symptoms of side effects during the block procedure. The patient reported minor discomfort, which was resolved with the administration of 50 µg intravenous fentanyl and infiltration of 3 milliliters 2% prilocaine into the surgical area. The surgical procedure was completed in 30 minutes without any complications. The block was considered successful, with no need of conversion to general anesthesia. Discussion ========== In this case report, a US-guided radial nerve block was applied from the mid-humeral level, which allowed sufficient surgical anesthesia for the excision of the ganglion cyst from the dorsal side of the hand. The course of the radial nerve in the distal part of the upper arm has a great variety \[[@REF13]\]. The radial nerve passes obliquely across the back of the humerus between the lateral and medial heads of the triceps. It enters the anterior compartment from the deltoid tuberosity to the lateral epicondyle and passes between the brachioradialis and brachialis muscles \[[@REF6]\]. It usually divides approximately 2-3 cm proximal to the elbow. The radial nerve can be easily visualized as a single structure 5 cm proximal to the elbow lateral to the humerus and curves posterior at about 15 cm proximal from the medial epicondyle \[[@REF7],[@REF13]\]. The nerve could be scanned at this level and can be found slightly posterolateral. The deep branch is sometimes difficult to visualize by ultrasound \[[@REF14]\]. The superficial radial nerve is cutaneous, whereas the deep branch forms the posterior interosseous nerve and mainly supplies motor function. The deep branch also supplies sensation to the interosseous membrane, the periosteum of the radius and ulna, and the extensor surfaces of the carpal joints \[[@REF6]\]. At the mid-humeral level, the four main nerves are anatomically well-separated. This allows for the selective administration of local anesthetics on both the radial and the musculocutaneous nerves with the ulnar and median nerves \[[@REF12]\]. Before its division into two terminal branches, a mid-humeral radial nerve block would produce more widespread anesthesia of the wrist and can be used for anesthesia in a wider variety of surgery \[[@REF6]\]. Foxall et al. reported the ease of identification of the radial nerve at the distal third level of the arm \[[@REF6]\]. We could easily visualize the radial nerve at the mid-humeral level just posterior to the humerus as a single, separate, and bright hyperechoic, oval-shaped structure. When it is slightly scanned to distally closer to the elbow, it becomes more distant to the humerus and can be blocked at this level, before it travels into the intermuscular septum. Reports about ultrasound-guided radial nerve blocks in the literature are rare and limited to very few anecdotal reports. These reports are limited to performance in emergency departments to provide analgesia for the reduction of distal radius fractures \[[@REF15]-[@REF18]\]. Blocking the radial nerve at the mid-humeral region has the advantage of blocking the motor function that the radial nerve innervates as well, which may be preferable in many operations. However, in certain operations, such as tendon repair, preservation of motor function may be desired intraoperatively. The maintenance of motor function is an important advantage of distal nerve blocks over proximal blocks. A more distal nerve block may be preferred to block the sensory function of the nerve. Also, a differential infraclavicular block (by using a low-dose local anesthetic) may be performed in such cases. The general advantages of distal peripheral nerve blocks could be expressed as: Distal approaches to upper extremity blocks need a lower amount of local anesthetic drugs with a blockade of a specific area. These approaches also avoid the risk of inadvertent puncture to critical structures, such as pleura or big vessels, which may occur when a proximal block is applied \[[@REF3],[@REF5]\]. Distal peripheral nerve blocks allow the preservation of proximal muscle function of the upper limb and patients may be discharged earlier from the hospital \[[@REF19]\]. These blocks may be suitable in patients where local anesthetic infiltration is contraindicated, such as in cases of infection or suspected malignancy. The surgical site anatomy is preserved and may provide for better operating conditions by not causing tissue distortion with infiltrated local anesthetic \[[@REF4]\]. Selective blockade of the specific nerve may shorten the time for anesthetic administration \[[@REF4]\]. The disadvantages are that distal nerve blocks would not prevent tourniquet pain. Peripheral nerve blocks may not always be sufficient for all surgical procedures on the forearm that involve wider surgical areas \[[@REF3]\]. Among our cases, two cases needed additional local anesthetic infiltration during the surgery. This may be due to a wider surgical dissection that exceeded beyond the analgesic area of radial nerve block. A proximal, brachial plexus block is an alternative for operations of the hand in which larger surgical areas involve multiple peripheral nerve sensory dermatomes. Conclusions =========== Our experience with an ultrasound-guided mid-humeral radial nerve blockade suggests that this block provides successful surgical anesthesia and postoperative analgesia for the excision of dorsal-side ganglion cysts of the hand when applied from the mid-humeral region. This regional anesthesia technique may be a feasible and more comfortable alternative to general anesthesia for patients undergoing hand surgery. A mid-humeral radial nerve block has the advantage of requiring a lower amount of local anesthetic drug and less motor block of the arm, which is favorable for outpatient settings. The authors have declared that no competing interests exist. Consent was obtained by all participants in this study

All relevant data are within the the paper and its Supporting Information file. Introduction {#sec001} ============ There is strong evidence that insufficient physical activity (PA) is associated with increased risk of developing lifestyle-related diseases and premature death \[[@pone.0175190.ref001]\]. Metabolic syndrome (MetS) is not consistently defined, but includes: overweight, abdominal obesity, insulin resistance, dyslipidemia, and hypertension in various combinations \[[@pone.0175190.ref002]\]. The presence of MetS carries a high risk for developing cardiovascular disease and type 2 diabetes \[[@pone.0175190.ref003]\]. Importantly, MetS is also associated with physical inactivity, further aggravating the risk of cardiovascular events \[[@pone.0175190.ref004]\]. The definition of PA is "any bodily movement produced by skeletal muscles that results in energy expenditure" and can be categorized as e.g. a household, occupational, leisure time, and sporting activity \[[@pone.0175190.ref005]\]. Exercise is PA with the objective to improve or maintain physical fitness components and is categorized in terms of the type, frequency, duration, intensity, and purpose \[[@pone.0175190.ref006]\]. The internationally recommended minimum level of PA \[[@pone.0175190.ref007]\] is moderate-intensity aerobic PA 150 min per week or, alternatively, vigorous-intensity aerobic PA 75 min per week, which has been associated with a clinically relevant risk reduction. Additional health benefits can be achieved by increased PA, above the national recommendation levels \[[@pone.0175190.ref008]\]. Despite the evidence-based positive effects of regular PA on health, implementing PA as an integrated method of treatment in health care remains a major challenge \[[@pone.0175190.ref009]\]. The Swedish National Board of Health's guidelines for disease prevention methods recommend the use of individual-based dialogue, written information, training diaries, a pedometer, and structured follow-up when the patient's PA level is insufficient \[[@pone.0175190.ref010]\]. An example of such a treatment strategy is physical activity on prescription (PAP), which is individually tailored for each patient and prescribed for preventive and therapeutic purposes as a first-line treatment. Meta-analyses of international PAP studies show varying results, with small to medium positive intervention effects when comparing increased PA levels with usual care. However, there is uncertainty due to the lack of high quality studies and further research is needed with more homogenized, comparable PAP interventions, longer follow-up, and objective measures of outcome \[[@pone.0175190.ref011], [@pone.0175190.ref012]\]. Swedish lifestyle interventions, including PAP, has shown to be cost-effective \[[@pone.0175190.ref013], [@pone.0175190.ref014]\] and the Swedish PAP intervention method had positive effects on PA levels, body composition, cardio metabolic risk factors, and health related quality of life (HRQOL) \[[@pone.0175190.ref015], [@pone.0175190.ref016]\]. Although scientific evidence has resulted in clinical treatment guidelines \[[@pone.0175190.ref010]\] and there are some evaluated Swedish PAP studies \[[@pone.0175190.ref015]--[@pone.0175190.ref018]\], PA is still underutilized as a treatment strategy in Swedish health care \[[@pone.0175190.ref019], [@pone.0175190.ref020]\]. There is still a lack of knowledge about PAP interventions suitable for different patient groups to improve their PA level and health outcomes. Further studies are needed evaluating clinical feasible PAP strategies on a large sample \[[@pone.0175190.ref021]--[@pone.0175190.ref023]\]. In primary care in the city of Gothenburg, health care centers have implemented PAP-treatment, individualized for patients with metabolic risk factors, with the purpose of increasing the PA level and health benefits. This specific model of PAP-treatment in daily clinical work has not been evaluated and may add new insights on how the extent of the intervention affects the PA level and health status. The aim of this observational study was to explore the association between PAP-treatment and the PA level of patients with metabolic risk factors and the relationship between changes in the PA level and health outcomes, including metabolic risk factors and HRQOL at the 6-month follow-up. Methods {#sec002} ======= Study design {#sec003} ------------ This is a prospective, longitudinal observational study with a 6-month follow-up of PAP-treatment in a daily clinical primary care practice. The present study is part of an ongoing study with a 5 year follow-up. Study population {#sec004} ---------------- The study population included 444 patients, aged 27--85 years. The patients were selected as a convenience sample from 15 primary health care centers in Gothenburg center/west. The patients agreed with their health care provider to participate in the study before they were prescribed PA and were included prospectively from 2010 to 2014. The population of central/western Gothenburg is 220 000 and has a higher socio-economic status compared with Gothenburg overall \[[@pone.0175190.ref024]\]. The inclusion criteria were: physically inactive, having at least one component of MetS present, and receiving PAP-treatment. The patients also had to understand the Swedish language to fill in the questionnaires. Intervention {#sec005} ------------ The patient was informed of the possibility to receive treatment with PAP by written information in the waiting room and orally by their caregiver. All authorized personnel were educated on the effects of PA according to the *Physical activity in the prevention and treatment of disease* (FYSS) \[[@pone.0175190.ref025]\] and the concept of the Swedish PAP model. Authorized personnel, mainly nurses, at the health care centers prescribed PA to the patients. The PAP included a dialogue with the patient, based on the principles of motivational interviewing (MI) \[[@pone.0175190.ref026]\]. Each patient's previous and current level of PA and their preferences for different kinds of PAs were elucidated. Furthermore, the patient's motivation, self-efficacy, and readiness to change PA behaviour were evaluated. This information served as the basis for the selection of the type and volume of the PA. The volume of the chosen PA was determined using the FYSS reference book, and the most suitable activity was prescribed at the appropriate relative intensity using the Borg's rate of perceived exertion scale \[[@pone.0175190.ref027]\] as well as duration and frequency. To help the patient to choose a suitable PA, a registry of the local supply of PA's was presented. It was possible to recommend two different types of PA in the PAP. This resulted, for each patient, in an individually tailored PA recommendation planned in dialogue with the patient and followed by a structured follow-up. The patients were offered individually adjusted support during the 6 month intervention period, either by revisits or telephone contacts. Measurements {#sec006} ------------ The measurements described below were conducted when PA was first prescribed as well as at the 6-month follow-up. ### PA level {#sec007} The PA level was the primary outcome and four questionnaires were used due to the known complexity of PA assessments. 1. Self-assessment was according to the American College of Sports Medicine (ACSM) and American Heart Association (AHA) public health recommendations. The patient responded to two PA questions (ACSM/AHA questionnaire), where 30 min of moderate-intensity PA per day resulted in 1 point and 20 min of more vigorous-intensity PA per day resulted in 1.7 point during each specific day of the week. A value of \<5 points indicated an inadequate PA level \[[@pone.0175190.ref028]\]. 2. The International Physical Activity Questionnaire (IPAQ) assessed the level of PA during the last 7 days. This instrument is extensively tested and translated into Swedish and can assess vigorous- and moderate-intensity PA, walking, and sitting time \[[@pone.0175190.ref029], [@pone.0175190.ref030]\]. 3. The Saltin-Grimby Physical Activity Level Scale (SGPALS) assessed leisure time PA during the past year at four different levels, from sedentary/physically inactive to vigorous physically active \[[@pone.0175190.ref031]\]. The levels have been validated against e.g. metabolic risk factors \[[@pone.0175190.ref032], [@pone.0175190.ref033]\] and the SGPALS has been published in an updated Swedish form \[[@pone.0175190.ref034]\]. 4. A six-grade PA scale, which is a further development of the SGPALS (Frändin/Grimby), was used and includes household activities \[[@pone.0175190.ref035]\]. This scale correlates with physical performance and self-assessed fitness and is used to classify PA among the elderly \[[@pone.0175190.ref036]\]. ### Anthropometrics {#sec008} Body weight was measured with light clothing and without shoes to the nearest 0.1 kg using an electric scale (Carl Lidén AFW D300, Jönköping, Sweden). Body height was measured in an upright position without shoes to the nearest 0.5 cm using a scale fixed to the wall (Personmått PEM 136, Hultafors, Sweden), and the body mass index (BMI) was calculated. Waist circumference (WC), to the nearest 0.5 cm, was measured in a standing exhaled position, with a measuring-tape (Kirchner Wilhelm, Aspberg, Germany) placed on the patient's skin between the lower rib and the iliac crest. ### Systolic and Diastolic Blood Pressure (SBP, DBP) {#sec009} The SBP and DBP were measured in mmHg according to guidelines \[[@pone.0175190.ref037]\] after 5 min rest with the patient seated with a blood pressure sphygmomanometer (Omron HEM-907, Kyoto, Japan) attached to the right upper arm at the level of the heart. ### Blood samples {#sec010} Blood samples were used to measure (in mmol/l) fasting plasma glucose (FPG) after an overnight fast, triglycerides (TG), cholesterol (Chol), High Density Lipoprotein (HDL), and Low Density Lipoprotein (LDL). Values were analyzed according to the European Accreditation system \[[@pone.0175190.ref038]\]. ### The cut-off values of MetS components {#sec011} Cut-off values were according to the National Cholesterol Education Program (NCEP) classification and were: WC \>88 cm for women, \>102 cm for men; BP ≥130/85 mm Hg; FPG ≥6.1 mmol/l; TG ≥1.7 mmol/l; or HDL \<1.3 mmol/l for women, \<1.0 mmol/l for men \[[@pone.0175190.ref039]\]. ### Health related quality of life {#sec012} The HRQOL was assessed with the Swedish version of the Short Form 36 (SF-36 Standard Swedish Version 1.0), which includes 36 questions \[[@pone.0175190.ref040]\]. It generates eight health concepts: physical functioning (PF), role physical functioning (RP), bodily pain (BP), general health (GH), vitality (VT), social function (SF), role emotional functioning (RE), and mental health (MH). The health concepts were converted to 0--100 points, where higher values represented a better HRQOL. The different health concepts of the SF-36 were also grouped into a physical component summary (PCS) and mental component summary (MCS). The SF-36 has shown good to excellent internal consistency and reliability and was validated in a representative sample of the Swedish population \[[@pone.0175190.ref040]\]. ### Readiness to change the PA level {#sec013} The readiness to change the PA level was measured at baseline using three questions estimated on a 100 mm visual analogue scale (VAS): How prepared are you? How important is it to you? How confident are you to succeed (self-efficacy)? The questions were derived from MI and behaviour change counseling, according to Rollninck et al. \[[@pone.0175190.ref041], [@pone.0175190.ref042]\], where higher values on the VAS indicated increased readiness to change. ### Support from the PAP-responsible nurse {#sec014} The support from the nurse responsible for the PAP was assessed at the 6-month follow-up by questioning the patient about the frequency of visits at the health care center. Statistical analysis {#sec015} ==================== Interval and ratio data are presented as the mean (m) and the dispersion as a standard deviation (SD) or 95% confidence interval (CI). Nominal and ordinal data were presented as the median (md) and minimum---maximum (min---max). A per-protocol analysis was used and differences between baseline and the 6-month follow-up, within the group, were analyzed using the paired sample t-test or Wilcoxon sign-rank test, based on the data level. Subgroup analyses between women *vs*. men and the completed group *vs*. dropout group were performed using the independent sample t-test or Mann Whitney U-test. A standardized measure of effect size (*d*) in the within-subject comparisons ($Cohen^{\prime}s\,~d_{z} \times \sqrt{2}$) was reported to quantify the degree of differentiation in values between baseline and the 6-month follow-up. The effect size was considered small when *d* = 0.2--0.3, as medium when *d* = 0.5, and as large when *d* = 0.8 \[[@pone.0175190.ref043]\]. In regression analysis, both multivariate and univariate methods were used to evaluate associations between changes in the PA level and changes in health outcomes, when adjusting for potential confounders. The predictor of interest was change in PA level, calculated as a Delta (Δ)-value (6-month value minus baseline value), and the PA level at baseline, age, sex, social situation, economy, education, and smoking were examined as potential confounders. The outcomes could be classified into two clusters. The first cluster contained Δ-values of the metabolic risk factors (BMI, WC, SBP, DBP, FPG, TG, Chol, HDL, LDL) and the second cluster contained Δ-values of self-reported health, using the SF-36 HRQOL (PF, RP, BP, GH, VT, SF, RE, MH, PCS, MCS). Univariate multiple linear regression was used to check whether or not a change in PA was significantly correlated with the nineteen independent variables, one at a time, when all the potential confounders were considered. Multivariate linear regression was then used to test if changes in PA were significantly associated with the two clusters (change in metabolic risk factors and change in self-rated health) and not just the specific variables in the clusters. The significance was tested using a regression-based MANOVA and test-statistic for Pillai's trace. Assumptions of normality, linearity, and outliers were checked using residual plots. All data in regression analysis were analyzed using SPSS for Windows, version 23 (IBM Corp. Armonk, NY, USA). A *p*-value \<0.05 was considered statistically significant. The patient's contact frequency with the PAP-responsible nurse was categorized in 1--2, 3--5, 6--10, 11--20, and ≥21 contacts. All statistical analyses, except the regression analysis, were calculated in SPSS version 22.0 (IBM Corp. in Armonk, NY, USA). Statistical significance was set at *p* ≤0.05. Ethical aspects {#sec016} =============== The study was approved by the Regional Ethical Review Board in Gothenburg, Sweden (Dnr 678--14). Results {#sec017} ======= Study population {#sec018} ---------------- Of the 444 included individuals, 368 completed the 6 months of follow-up. The dropout rate was 17% ([Fig 1](#pone.0175190.g001){ref-type="fig"}). {#pone.0175190.g001} Baseline characteristics {#sec019} ------------------------ The mean age of the study population was 57 years, with 56% female. Overweight/obesity was present in 91%, hypertension in 78%, and 58% had hyperlipidemia ([Table 1](#pone.0175190.t001){ref-type="table"}). Two components of MetS, WC (\>88 cm for women, \>102 cm for men) and BP (≥130/85 mm Hg) were present in 72% and WC (\>88 cm for women, \>102 cm for men) and TG (≥1.7 mmol/l) in 53%. At baseline, 61% were taking medications for components of MetS, including 54% for hypertension and 22% for hyperlipidemia. The PA level was estimated to be low, using all four PA instruments. A total of 36% of subjects were sedentary according to SGPALS, 80% reported PA equivalent to a 30 min brisk walk three times per week or less and 47% reported ≤1 time per week, according to the ACSM/AHA questionnaire ([Table 2](#pone.0175190.t002){ref-type="table"}). 10.1371/journal.pone.0175190.t001 ###### Baseline characteristics of the follow-up and dropout group. {#pone.0175190.t001g} ------------------------------------------------------------------------------------------------------------------------------ Variable[^a^](#t001fn001){ref-type="table-fn"} Follow-up\ Dropout\ *p* value[^b^](#t001fn002){ref-type="table-fn"} (n = 368) (n = 76) ------------------------------------------------ ------------- ------------- ------------------------------------------------- **Age**---years 57.4 (10.9) 57.6 (13.1) 0.955 **Sex** **0.011** Female 198 (53.8) 53 (69.7) Male 170 (46.2) 23 (30.3) **Nationality** 0.915 Sweden 312 (86.0) 62 (84.9) Other 51 (14.0) 11 (15.1) **Social situation** 0.144 Single 135 (37.9) 35 (48.6) Married/ cohabit 205 (57.6) 33 (45.8) Other 16 (4.5) 4 (5.6) **Economy**---perceived 0.467 Good 213 (59.3) 36 (50.7) Neither nor 107 (29.8) 19 (26.8) Bad 39 (10.9) 16 (22.5) **Education** 0.117 Elementary grade 69 (19.2) 14 (19.4) Upper secondary school 131 (36.4) 36 (50) University college 160 (44.4) 22 (30.6) **Tobacco** 0.871 Smokers 34 (9.5) 10 (13.9) Non-smokers 229 (63.8) 41 (56.9) Ex-smokers 96 (26.7) 21 (29.2) **Part of metabolic syndrome** Overweight/Obesity 333 (90.5) 71 (93.4) 0.245 Hyperglycemia 144 (39.1) 30 (39.5) 0.672 Hypertension 293 (79.6) 53 (69.7) 0.117 Hyperlipidemia 212 (57.6) 41 (53.9) 0.801 Other diagnosis Mental health, depression 52 (14.1) 13 (17.1) 0.446 Musculoskeletal disorders 58 (15.8) 19 (25) **0.040** Other 155 (42.1) 38 (50) 0.172 **Drug treatment** Overweight/Obesity 1 (0.3) 1 (1.3) 0.207 Hyperglycemia 46 (12.5) 13 (17.1) 0.246 Hypertension 196 (53.3) 40 (52.6) 0.901 Hyperlipidemia 77 (20.9) 17 (22.4) 0.694 Other drug treatment Mental health, depression 52 (14.1) 12 (15.8) 0.642 Musculoskeletal disorders 49 (13.3) 11 (14.5) 0.723 Other 132 (35.9) 35 (46.1) 0.064 ------------------------------------------------------------------------------------------------------------------------------ ^a^ Age data are given as mean (standard deviation) and data for other variables are given as number (percentage). ^b^ Difference between follow-up and dropout group. *P*-value for age was determined by an independent samples *t*-test and all the other characteristics were determined by Mann-Whitney U-test. Statistical significance was set at *p*≤ 0.05. 10.1371/journal.pone.0175190.t002 ###### Baseline characteristics in anthropometrics, metabolic risk factors, physical activity level and health related quality of life---Follow-up and dropout group. {#pone.0175190.t002g} -------------------------------------------------------------------------------------------------------------------------------- Variable[^a^](#t002fn002){ref-type="table-fn"} Follow-up\ Dropout\ *p* value[^b^](#t002fn003){ref-type="table-fn"} (n = 368) (n = 76) ------------------------------------------------ -------------- -------------- ------------------------------------------------- BMI, kg/m^2^ 32.0 (5.2) 33.0 (5.8) 0.104 Waist circumference, cm 107.9 (13.1) 109.2 (13.5) 0.423 Blood pressure, mm/Hg: Systolic 137.3 (17.4) 135.8 (19.3) 0.515 Diastolic 82.7 (10.2) 79.6 (9.9) **0.017** Metabolic components, mmol/l: Fasting plasma glucose 6.3 (1.9) 6.3 (1.8) 0.894 Triglycerides 1.7 (1.0) 1.8 (0.9) 0.597 Cholesterol 5.6 (1.2) 5.6 (1.2) 0.949 HDL 1.4 (0.4) 1.4 (0.6) 0.549 LDL 3.6 (1.1) 3.6 (1.0) 0.734 Physical activity level, score: ACSM/AHA questionnaire 1.7 (1.5) 1.7 (1.4) 0.975 HRQOL SF-36, score: Physical functioning 81.1 (18.3) 70.4 (24.1) **0.001** Role limitation, physical 69.9 (37.3) 58.1 (42.1) **0.030** Bodily pain 67.0 (26.7) 55.1 (28.1) **0.001** General health 60.5 (20.4) 53.5 (20.1) **0.009** Vitality 52.4 (23.2) 44.7 (23.6) **0.010** Social function 78.9 (25.2) 68.8 (29.2) **0.007** Role limitation, emotional 72.9 (39.0) 58.6 (45.2) **0.015** Mental health 72.1 (19.9) 65.8 (21.9) **0.016** Physical component summary 45.7 (9.9) 41.4 (10.8) **0.001** Mental component summary 44.4 (13.1) 40.0 (14.6) **0.012** -------------------------------------------------------------------------------------------------------------------------------- BMI, body mass index; LDL, low density lipoprotein; HDL, high density lipoprotein; ACSM, American College of Sports Medicine; AHA, American Heart Association; HRQOL SF-36, Health Related Quality of Life 36-Item Short Form Health Survey. ^a^ Data are given as mean (standard deviation). ^b^ Difference between follow-up and dropout group. *P*-value for the variables was determined by an independent samples *t*-test. Statistical significance was set at *p*≤ 0.05. A higher proportion of women and musculoskeletal disorders and a lower level of SF-36 HRQOL were seen in the dropout group (Tables [1](#pone.0175190.t001){ref-type="table"} and [2](#pone.0175190.t002){ref-type="table"}). There was a significantly lower SF-36 value, not presented in table, for BP (*p* = 0.048) among women compared with men in the dropout group. A lower DBP was also present in the dropout group ([Table 2](#pone.0175190.t002){ref-type="table"}). In a subgroup analysis between men and women, inferior values were seen in metabolic health (DBP, FPG, TG, Chol, HDL) for men and in HRQOL (PF, BP, SF, RE, MH, PCS) for women at baseline ([Table 3](#pone.0175190.t003){ref-type="table"}). 10.1371/journal.pone.0175190.t003 ###### Baseline characteristics in anthropometrics, metabolic risk factors, physical activity level and health related quality of life---Men and women. {#pone.0175190.t003g} -------------------------------------------------------------------------------------------------------------------------------- Variable[^a^](#t003fn002){ref-type="table-fn"} Men\ Women\ *p* value[^b^](#t003fn003){ref-type="table-fn"} (n = 193) (n = 251) ------------------------------------------------ -------------- -------------- ------------------------------------------------- BMI, kg/m^2^ 32.2 (5.0) 32.1 (5.6) 0.788 Waist circumference, cm 113 (12.4) 104 (12.5) **\<0.001** Blood pressure, mm/Hg: Systolic 137.4 (17.1) 136.7 (18.2) 0.668 Diastolic 83.5 (10.9) 81.1 (9.4) **0.016** Metabolic components, mmol/l: Fasting plasma glucose 6.6 (2.2) 6.0 (1.5) **0.003** Triglycerides 2.0 (1.2) 1.5 (0.7) **\<0.001** Cholesterol 5.4 (1.2) 5.7 (1.2) **0.004** HDL 1.2 (0.4) 1.6 (0.5) **\<0.001** LDL 3.5 (1.0) 3.7 (1.1) 0.168 Physical activity level, score: ACSM/AHA questionnaire 1.8 (1.5) 1.7 (1.5) 0.403 HRQOL SF-36, score: Physical functioning 83.2 (17.5) 76.3 (20.9) **\<0.001** Role limitation, physical 71.5 (36.2) 65.1 (39.8) 0.087 Bodily pain 69.2 (26.5) 61.7 (27.4) **0.005** General health 59.4 (19.5) 59.2 (21.4) 0.911 Vitality 52.5 (23.2) 50.0 (23.6) 0.275 Social function 80.0 (24.4) 75.0 (27.3) **0.049** Role limitation, emotional 76.0 (36.8) 66.2 (42.6) **0.011** Mental health 73.2 (19.2) 69.3 (21.0) **0.049** Physical component summary 46.2 (9.6) 44.0 (10.5) **0.035** Mental component summary 45.0 (12.4) 42.7 (14.2) 0.071 -------------------------------------------------------------------------------------------------------------------------------- BMI, body mass index; LDL, low density lipoprotein; HDL, high density lipoprotein; ACSM, American College of Sports Medicine; AHA, American Heart Association; HRQOL SF-36, Health Related Quality of Life 36-Item Short Form Health Survey. ^a^ Data are given as mean (standard deviation). ^b^ Difference between men and women. *P*-value for the variables was determined by an independent samples *t*-test. Statistical significance was set at *p*≤ 0.05. The three questions regarding the readiness to change PA levels were: How prepared? (mean 81.4 mm, SD 17.7 mm), How important? (mean 83.9 mm, SD 17.9 mm), and How confident? (self-efficacy) (mean 64.3 mm, SD 23.9 mm). A subgroup analyses between the completed group *vs*. dropout group showed significantly lower values for the dropout group regarding confidence to succeed (65.9 mm *v* 56.3 mm, 95% CI: 2.7--16.6, *p* = 0.007). Six-month follow-up {#sec020} ------------------- Statistically significant positive changes were reached using all four PA instruments ([Table 4](#pone.0175190.t004){ref-type="table"}). The most commonly prescribed PA modality overall was moderate intensity walking, 30--44 min, 2--5 times/week. At the 6-month follow-up, 270 patients (73%, *d-value* = 1.17) in the study group had increased their PA level, measured with the ACSM/AHA questionnaire and a total of 153 patients (42%) had improved their PA from inadequate to sufficient, according to the public health recommendations of the ACSM/AHA. There were also significant improvements in BMI, WC, SBP, FPG, TG, Chol, and LDL with small *d*-values except for women's WC reaching a medium *d*-value at the 6-month follow-up. The SF-36 showed a significant increase in 6 of 8 health concepts: RP, GH, VT, SF, RE, and MH and in the physical and mental component summary with small *d*-values ([Table 5](#pone.0175190.t005){ref-type="table"}). 10.1371/journal.pone.0175190.t004 ###### Descriptive statistics and differences for physical activity level at baseline and 6-months follow-up. {#pone.0175190.t004g} ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Variable\ Baseline 6-months follow up Mean difference\ 95% CI *p* value (n) (6-months---baseline) --------------------------------------------------------------------------- ------------- -------------------- ----------------------- --------- --------------------------------------------------- ACSM/AHA questionnaire, score (361)[^a^](#t004fn002){ref-type="table-fn"} 1.75 (1.55) 4.57 (3.29) 2.8 (3.4) 2.5;3.2 **\<0.001**[^c^](#t004fn004){ref-type="table-fn"} IPAQ 1--3, score (236)[^b^](#t004fn003){ref-type="table-fn"} 1 (1--2) 2 (1--2) \- \- **\<0.001**[^d^](#t004fn005){ref-type="table-fn"} IPAQ 1--3, category, No (%)  • Low 222 (62.4) 130 (47.3) \- \-  • Moderate 134 (37.6) 145 (52.7) \- \-  • High 0 0 \- \- SGPALS 1--4, score (337)[^b^](#t004fn003){ref-type="table-fn"} 2 (1--3) 2 (1--3) \- \- **\<0.001**[^d^](#t004fn005){ref-type="table-fn"} SGPALS 1--4, category, No (%)  • 1 158 (36.5) 66 (19.2) \- \-  • 2 268 (61.9) 223 (65.1) \- \-  • 3 7 (1.6) 54 (15.7) \- \-  • 4 0 0 \- \- Frändin/Grimby 1--6, score (338)[^b^](#t004fn003){ref-type="table-fn"} 3 (1--5) 3 (1--6) \- \- **\<0.001**[^d^](#t004fn005){ref-type="table-fn"} Frändin/Grimby 1--6, category, No (%)  • 1 31 (7.1) 7 (2.0) \- \-  • 2 75 (17.3) 41 (12.0) \- \-  • 3 243 (56.0) 169 (49.1) \- \-  • 4 81 (18.7) 107 (31.1) \- \-  • 5 4 (0.9) 17 (4.9) \- \-  • 6 0 3 (0.9) \- \- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- CI, confidence intervals; ACSM, American College of Sports Medicine; AHA, American Heart Association; IPAQ; International Physical Activity Questionnaire ^a^Values are given as mean (standard deviation). ^b^Values are given as median (minimum-maximum). ^c^*P* values were determined by a paired samples *t*-test for the difference between baseline and 6-months follow up. ^d^*P* values were determined by Wilcoxon Signed Ranks Test for the difference between baseline and 6-months follow up. Statistical significance was set at *p*≤ 0.05. 10.1371/journal.pone.0175190.t005 ###### Descriptive statistics and differences for anthropometric-, metabolic characteristics and health related quality of life at baseline and 6-month follow-up. {#pone.0175190.t005g} ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Variable[^a^](#t005fn002){ref-type="table-fn"}\ Baseline 6-months follow up Mean difference\ 95% CI *p* value[^b^](#t005fn003){ref-type="table-fn"} Cohen's *d*[^c^](#t005fn004){ref-type="table-fn"} (n) (6-month---baseline) ------------------------------------------------- -------------- -------------------- ---------------------- ----------- ------------------------------------------------- --------------------------------------------------- BMI, kg/m^2^ (353) 32.0 (5.2) 31.7 (5.4) -0.3 (1.7) -0.5;-0.1 **0.001** 0.25 Waist circumference, cm (352) 107.8 (13.2) 106.2 (13.9) -1.7 (5.8) -2.3;-1.1 **\<0.001** 0.41  • female (187) 103.4 (12.2) 101.4 (13.2) -2.1 (5.9) -2.9;-1.2 **\<0.001** 0.50  • male (165) 112.8 (12.6) 111.6 (12.6) -1.3 (5.6) -2.1;-0.4 **0.005** 0.32 Blood pressure, mm/Hg: Systolic (358) 137.5 (17.3) 133.9 (16.2) -3.6 (16.4) -5.3;-1.9 **\<0.001** 0.31 Diastolic (358) 82.8 (10.1) 82.5 (9.3) -0.4 (9.9) -1.4;0.6 0.466 0.05 Metabolic components, mmol/l: Fasting plasma glucose (352) 6.26 (1.92) 6.01 (1.44) -0.3 (1.2) -0.4;-0.1 **\<0.001** 0.29 Triglycerides (355) 1.69 (0.99) 1.59 (0.88) -0.1 (0.8) -0.2;0.0 **0.016** 0.18 Cholesterol (358) 5.57 (1.21) 5.39 (1.16) -0.2 (0.9) -0.3;-0.1 **\<0.001** 0.27 HDL (357) 1.41 (0.45) 1.43 (0.45) 0.0 (0.3) 0.0;0.1 0.196 0.10 LDL (353) 3.63 (1.06) 3.52 (1.03) -0.1 (0.8) -0.2;0.0 **0.009** 0.20 HRQOL SF-36, score: Physical functioning (335) 81.3 (18.1) 81.8 (19.1) 0.4 (14.1) -1.1;2.0 0.558 0.04 Role limitation, physical (323) 70.0 (37.4) 77.4 (33.3) 7.4 (39.7) 3.1;11,8 **0.001** 0.26 Bodily pain (334) 67.3 (26.6) 69.6 (27.2) 2.3 (22.9) -0.2;4.7 0.069 0.14 General health (335) 60.7 (20.2) 64.2 (20.8) 3.6 (14.4) 2.0;5.1 **\<0.001** 0.35 Vitality (333) 52.6 (23.1) 58.3 (21.6) 5.7 (19.4) 3.6;7.8 **\<0.001** 0.42 Social function (334) 79.2 (25.2) 83.6 (21.8) 4.4 (24.6) 1.8;7.1 **0.001** 0.26 Role limitation, emotional (324) 73.4 (38.8) 77.8 (36.2) 4.4 (39.0) 0.2;8.7 **0.042** 0.16 Mental health (333) 72.2 (19.8) 74.3 (18.8) 2.2 (16.3) 0.4;3.9 **0.017**[^b^](#t005fn003){ref-type="table-fn"} 0.19 Physical component summary (318) 45.8 (9.9) 46.8 (9.9) 1.0 (8.0) 0.1;1.9 **0.029** 0.17 Mental component summary (318) 44.6 (13.2) 46.6 (11.8) 2.0 (10.9) 0.8;3.2 **0.001** 0.19 ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ CI, confidence intervals; BMI, body mass index; HDL, high density lipoprotein; LDL, low density lipoprotein; HRQOL SF-36, Health Related Quality of Life 36-Item Short Form Health Survey. ^a^Values are given as mean (standard deviation). ^b^*P* values were determined by a paired samples *t*-test for the difference between baseline and 6-months follow up. ^c^Effect size in within-subjects comparisons ($\text{Cohen's }d_{\text{z}}\, \times \sqrt{2}\, = \text{Cohen's }d$) was measured quantifying the degree of differentiation in values between baseline and 6-months follow-up. Statistical significance was set at *p*≤ 0.05. A multivariate regression analysis showed positive significant associations between changes in the PA level and the health outcomes of metabolic risk factors (Pillai's Trace = 0.063, *p* = 0.032) and SF-36 HRQOL (Pillai's Trace = 0.095, *p*\<0.001) at the 6-month follow-up. Univariate linear regression analysis showed positive significant associations between changes in the PA level and the BMI ([Table 6](#pone.0175190.t006){ref-type="table"}), and the SF-36: RP, BP, GH, VT, SF, MH, PCS, and MCS ([Table 7](#pone.0175190.t007){ref-type="table"}). The univariate regression models could only partially explain how the changes in the PA level were related to metabolic risk factor outcomes (R^2^-values 0.03--0.05) and health concepts in the SF-36 (R^2^-values 0.03--0.12). 10.1371/journal.pone.0175190.t006 ###### Results for univariate linear regression analysis investigating the association between change in PA and metabolic risk factors at 6-month follow-up[^a^](#t006fn002){ref-type="table-fn"}. {#pone.0175190.t006g} Dependent variable Independent variable β 95% CI *p* value ----------------------------------- ---------------------- -------- --------------- ----------- Δ BMI, kg/m^2^ Δ Change PA -0.069 -0.128;-0.010 **0.022** Δ Waist circumference, cm Δ Change PA -0.124 -0.340;0.091 0.256 Δ Blood pressure systolic, mm/Hg Δ Change PA 0.489 -0.108;1.086 0.108 Δ Blood pressure diastolic, mm/Hg Δ Change PA 0.277 -0.083;0.637 0.131 Δ Fasting plasma glucose, mmol/l Δ Change PA -0.035 -0.076;0.006 0.097 Δ Triglycerides, mmol/l Δ Change PA -0.013 -0.039;0.014 0.341 Δ Cholesterol, mmol/l Δ Change PA -0.025 -0.059;0.008 0.135 Δ HDL, mmol/l Δ Change PA 0.001 -0.007;0.009 0.739 Δ LDL, mmol/l Δ Change PA -0.012 -0.042;0.017 0.410 PA, physical activity according to ACSM/AHA questionnaire; β, change in value; CI, confidence intervals; Δ, the difference between 6-month value and start value; BMI, body mass index; HDL, high density lipoprotein; LDL, low density lipoprotein. ^a^Adjusted for PA level at baseline, age, sex, social situation, economy, education and smoking. Statistical significance was set at *p*≤ 0.05. 10.1371/journal.pone.0175190.t007 ###### Results for univariate linear regression analysis investigating the association between change in PA and health related quality of life (SF-36) at 6-month follow-up[^a^](#t007fn002){ref-type="table-fn"}. {#pone.0175190.t007g} Dependent variable Independent variable β 95% CI *p* value ------------------------------ ---------------------- ------- -------------- ------------- HRQOL SF-36, score: Δ Physical functioning Δ Change PA 0.432 -0.077;0.941 0.096 Δ Role limitation, physical Δ Change PA 1.713 0.289;3.138 **0.019** Δ Bodily pain Δ Change PA 1.184 0.380;1.988 **0.004** Δ General health Δ Change PA 0.785 0.265;1.306 **0.003** Δ Vitality Δ Change PA 1.702 1.017;2.388 **\<0.001** Δ Social function Δ Change PA 1.435 -0.561;2.309 **0.001** Δ Role limitation, emotional Δ Change PA 1.004 -0.449;2.456 0.175 Δ Mental health Δ Change PA 1.171 0.576;1.765 **\<0.001** Δ Physical component summary Δ Change PA 0.338 0.048;0.628 **0.022** Δ Mental component summary Δ Change PA 0.659 0.265;1.053 **0.001** PA, physical activity according to ACSM/AHA questionnaire; β, change in value; CI, confidence intervals; HRQOL SF-36, Health Related Quality of Life 36-Item Short Form Health Survey; Δ, the difference between 6-month value and start value. ^a^Adjusted for PA level at baseline, age, sex, social situation, economy, education and smoking. Statistical significance was set at *p*≤ 0.05. In a within subgroup analysis between men and women, there were improved metabolic risk factor values for both men (6 of 9 measured parameters) and women (5 of 9) at the 6-month follow-up. Increased HRQOL values were seen in 2 of 10 health concepts among men and 7 of 10 among women ([Table 8](#pone.0175190.t008){ref-type="table"}). In a between group analysis, not presented in table, women increased their PA level more than men. 10.1371/journal.pone.0175190.t008 ###### Descriptive statistics and differences for anthropometric-, metabolic characteristics, physical activity level and health related quality of life at baseline and 6-month follow-up---Men and women. {#pone.0175190.t008g} ------------------------------------------------------------------------------------------------------------------------- Variable[^a^](#t008fn002){ref-type="table-fn"}\ Men Women (n; men/women) ------------------------------------------------- ------- ------------- ------------- ------- ------------- ------------- BMI, kg/m^2^ (164/189) -0.18 -0.43;0.06 0.139 -0.39 -0.64;-0.14 **0.002** Waist circumference, cm (165/187) -1.26 -2.12;-0.39 **0.005** -2.09 -2.94;-1.23 **\<0.001** Blood pressure, mm/Hg: Systolic (166/192) -3.38 -5.74;-1.01 **0.005** -3.76 -6.21;-1.29 **0.003** Diastolic (166/192) -0.04 -1.59;1.51 0.957 -0.68 -2.08;0.72 0.342 Metabolic components, mmol/l: Fasting plasma glucose (161/191) -0.36 -0.56;-0.14 **0.001** -0.17 -0.33;-0.01 **0.032** Triglycerides (163/192) -0.18 -0.33;0.02 **0.020** -0.03 -0.10;0.04 0.417 Cholesterol (163/195) -0.23 -0.37;-0.07 **0.003** -0.14 -0.27;-0.01 **0.039** HDL (164/193) 0.02 -0.03;0.07 0.464 0.02 -0.01;0.05 0.213 LDL (161/192) -0.17 -0.30;-0.03 **0.017** -0.08 -0.20;0.04 0.184 Physical activity level, score: ACSM/AHA questionnaire (168/193) 2.44 1.97;2.91 **\<0.001** 3.16 2.63;3.68 **\<0.001** HRQOL SF-36, score: Physical functioning (153/182) -1.2 -3.53;1.05 0.286 1.9 -0.14;3.89 0.068 Role limitation, physical (148/175) 6.2 -0.02;12.41 0.051 8.5 2.4;14.64 **0.007** Bodily pain (152/182) 2.8 -0.71;6.31 0.117 1.8 -1.62;5.32 0.294 General health (153/182) 3.2 0.98;5.51 **0.005** 3.8 1.72;5.99 **\<0.001** Vitality (152/181) 6.2 3.70;8.71 **\<0.001** 5.3 2.10;8.58 **0.001** Social function (153/181) 2.5 -1.11;6.18 0.172 6.1 2.27;9.88 **0.002** Role limitation, emotional (146/178) 0.2 -5.62;6.08 0.939 7.9 1.76;13.97 **0.012** Mental health (152/181) 0.2 -2.32;2.67 0.888 3.8 1.35;6.28 **0.003** Physical component summary (144/174) 1.0 -0.29;2.27 0.128 1.0 -0.25;2.22 0.117 Mental component summary (144/174) 0.9 -0.81;2.53 0.311 3.0 1.28;4.72 **0.001** ------------------------------------------------------------------------------------------------------------------------- CI, confidence intervals; BMI, body mass index; HDL, high density lipoprotein; LDL, low density lipoprotein; ACSM, American College of Sports Medicine; AHA, American Heart Association; IPAQ; International Physical Activity Questionnaire; HRQOL SF-36, Health Related Quality of Life 36-Item Short Form Health Survey. ^a^Values are given as mean. ^b^*P* values were determined by a paired samples *t*-test for the difference between baseline and 6-months follow up. Statistical significance was set at *p*≤ 0.05. During the first 6 month period, excluding the start visit, 80% of the patients received PAP support from caregivers 1--2 times, either through visits at the health care center or by telephone contact ([Fig 2](#pone.0175190.g002){ref-type="fig"}). {#pone.0175190.g002} Discussion {#sec021} ========== The main results of this study were the improvements in PA level, metabolic risk factors, and self-reported quality of life at the 6 month follow-up of PAP-treatment in primary health care for 27--85 year olds, having at least one component of MetS. Multivariate regression analysis also showed positive significant associations between changes in the PA level and health outcomes, similar to previous studies \[[@pone.0175190.ref044], [@pone.0175190.ref045]\]. To explore the association between PAP-treatment and the PA level, four self-reported questionnaires were used and all showed significant increments in the PA level. The difference measured with the ACSM/AHA questionnaire had a *d* value (*d* = 1.17) that indicated a large effect size. The finding that several PA instruments showed positive changes increases the dependability of the overall result and may be important since the concept of PA is contextualized, complex and difficult to measure \[[@pone.0175190.ref046]\]. Two previous meta-analyses \[[@pone.0175190.ref011], [@pone.0175190.ref012]\] regarding the clinical effects of PAP-interventions also demonstrated difficulties with measuring PA due to heterogeneity in the quality and nature of the studies. In addition, several of the included studies had a fixed short-time intervention (10--12 weeks) \[[@pone.0175190.ref047]--[@pone.0175190.ref050]\] and were linked to predetermined activities, e.g. leisure center-based or community walk programs \[[@pone.0175190.ref048]--[@pone.0175190.ref051]\], and not individualized to the needs and opinions of the patients. There was also uncertainty about the customized structured support and follow-up during the intervention. The lack of patient-related individualization of the PAP-intervention may have influenced the outcome. This present study was a totally individualized intervention and the PAP-treatment consisted of individual-based dialogue with the patient, an individually tailored recommendation of PA, and customized, structured support over 6 months. The majority of the patients chose a PA to be carried out, on their own, in everyday life near their residential area or workplace. The majority of the patients received PAP support 1--2 times during the 6 month period; thus, the effort made by the primary health care workers was small and the cost was relatively low \[[@pone.0175190.ref052]\]. The present study showed significant improvements in the majority of metabolic risk factors, measured at the 6-month follow-up and positive associations between changes in the PA level and metabolic health outcomes. Previous PAP studies have shown some, but varied outcomes regarding metabolic risk factors \[[@pone.0175190.ref047], [@pone.0175190.ref049], [@pone.0175190.ref050]\] in within group analyses, but often non-significant effects in the between group analysis. Though, Kallings et al. \[[@pone.0175190.ref015]\], in a Swedish PAP study among elderly subjects, showed an increase in metabolic health in both within- and between group analyses with patient-individualized PAP-intervention versus usual care. In this study, the results were even more obvious with regard to the effects on metabolic risk factors and may strengthen the individualization perspective of the PAP-treatment. The relatively small effect sizes regarding changes in metabolic parameters can be explained by the existing dose-response relationship between PA level and health outcomes \[[@pone.0175190.ref044], [@pone.0175190.ref045]\]. In this study, 42% of the patients reached the public health recommendations and the most commonly prescribed exertion was moderate intensity. A more extensive increase in PA level among the patients not reaching the public health recommendations or PA on a vigorous-intensity level possibly would have increased the effect size. Even HRQOL, measured with the SF-36, was significantly improved for 8 of 10 health concepts with a small effect size (*d*) and five of the concepts considered clinically relevant \[[@pone.0175190.ref053]\]. Small differences on the SF-36 health survey (i.e. 3- to 5-points) are considered clinically important even if the effect size is estimated as small \[[@pone.0175190.ref054], [@pone.0175190.ref055]\]. Importantly, the improvements in our study were associated with changes in the PA level. Several previous PAP studies found improvements in quality of life \[[@pone.0175190.ref017], [@pone.0175190.ref047], [@pone.0175190.ref049], [@pone.0175190.ref051], [@pone.0175190.ref056]--[@pone.0175190.ref058]\]. However, the studies that used the SF-36 reported smaller numbers of improved health concepts, with the exception of two Swedish studies showing significant positive changes in the majority of health concepts at 6 months \[[@pone.0175190.ref059]\] and 2 years follow-up \[[@pone.0175190.ref060]\]. This study found similar changes in HRQOL, although the inclusion criteria were different. However, the amount of PAP support is only reported in the present study, where 80% of the patients received support from caregivers 1--2 times during the 6 month period. Although univariate regression models showed positive significant associations, they could not explain all the changes in health outcomes. An increased PA level explained 3--5% of the metabolic health effects and 3--12% of the health effects in the HRQOL SF-36. However, the role of PA is difficult and complex to measure and not fully clarified due to its multifactorial effects. The relatively low correlation between changes in physical fitness and metabolic risk factors has been discussed due to exercise-induced improvements in body composition, e.g. reduced abdominal visceral fat, per se, has positive effects on the metabolic risk profile \[[@pone.0175190.ref061]\]. There were some differences in baseline characteristics between those who could be followed up at 6 months and the dropout group. The most interesting findings were that women with an additional musculoskeletal diagnosis and lower self-reported quality of life (BP *p* = 0.048), were more frequent in the dropout group. This might give an indication that some patients are in need of more enhanced support to increase their PA level or would benefit from a period of rehabilitation before or in combination with a PAP intervention. There were some differences between men and women, within the study group, with inferior values in metabolic variables for men and HRQOL for women at baseline. At the 6 month follow-up, the women had increased their PA and HRQOL values more than men; whereas, the improvements in metabolic parameters were similar between the groups. In contrast, previous studies of lifestyle interventions have shown that men were more likely to increase their PA level \[[@pone.0175190.ref062], [@pone.0175190.ref063]\]. The most common recommendation for PAP was for PA to be carried out in everyday life (e.g. taking walks) near the subject's residential area or workplace, a recommendation which may be suitable for many women. These gender-related differences should be considered when individualizing the intervention and support for the patient. Various subgroup analyses are needed to explore the possible needs for the patient to succeed when making this behavioral change. Limitations {#sec022} ----------- The dropout rate was 17% between baseline and the 6-month follow-up, and this may be normal in this type of study \[[@pone.0175190.ref059], [@pone.0175190.ref064]\]. Importantly, the present study was a "daily clinical work" survey with ordinary primary health care patients visiting their local health care center where the personnel had no extra time to manage this part of their duties. A per-protocol analysis was used where the drop-out group was excluded from the analysis between the baseline and 6-month follow-up, a method that increases the risk of bias. Although, when using an intention to treat analysis (ITT) that includes the dropout group, there would be a risk of attributing characteristics to the patients in the dropout group that they did not have. The ITT analysis has been criticized for potentially leading to a biased treatment effect \[[@pone.0175190.ref065], [@pone.0175190.ref066]\]. This survey was an observational, follow-up study in daily clinical practice without any control group. The lack of a control group complicates the interpretation of estimating the effects of the PAP-intervention on the increased PA level and positive health effects. However, the results in this study are comparable to other studies evaluating intervention groups in randomized controlled trials (RCTs) \[[@pone.0175190.ref015], [@pone.0175190.ref064]\] and two Swedish PAP studies without a control group that reported increased PA levels \[[@pone.0175190.ref016], [@pone.0175190.ref059]\]. Notably, a review assessing the impact of RCTs versus an observational study design revealed no differences in effect estimates (pooled ratio of odds ratios of 1.08) \[[@pone.0175190.ref067]\]. In addition, there are limitations of RCTs due to the possible lack of external validity resulting in non-appropriate evidence on which clinical decisions are based \[[@pone.0175190.ref068], [@pone.0175190.ref069]\]. An observational study design requires caution concerning bias that can undermine the internal validity \[[@pone.0175190.ref070]\]. The study population was non-consecutively included. No data were collected for patients with the same inclusion criteria who visited the health care centers and were not included in the study. It is not possible to estimate how many patients, in total, were candidates for treatment with PAP. The risk of selection bias increases when including patients more willing to change their PA level. The patients estimated high values for how prepared they were and how important it was to increase their PA level, but lower values for the confidence (self-efficacy) to succeed. The dropout group had even lower values for self-efficacy (*p* = 0.007), which is predictive for exercise adherence and compliance in both the adoption and maintenance of PA among adults \[[@pone.0175190.ref071], [@pone.0175190.ref072]\]. The PAP-treatment method is patient-centered and dependent on the patient's attitude to change living habits and the health care professional's ability to evaluate what stage of change the patient is in. The PAP method probably has the most potential for patients in contemplation or preparation stages \[[@pone.0175190.ref073]\]. Using multiple measurements to assess the outcomes may incur a risk for type I errors. However, multiple predefined outcomes in this study were needed to indicate several effects and they seem to be justified without the need for adjustment \[[@pone.0175190.ref074]\]. Measuring PA seems to be a complex issue with methodological problems due to the diversified designs of several studies \[[@pone.0175190.ref075]\]. Using self-reported questionnaires may increase the risk of recall or social desirability bias, but is both practical and valid \[[@pone.0175190.ref064]\]. In this study, we used four self-reported instruments, and all showed similar results. Conclusions {#sec023} =========== This observational study indicated that an individual-based PAP-treatment has the potential to change people's PA behavior with improved metabolic risk factors and self-reported quality of life at the 6-month follow-up. The PAP-treatment proved to work in daily clinical primary care with educated, authorized personnel, structured routines, and readily available information for both patients and workers. If implemented widely, PAP-treatment has the potential to become an important method and may result in major health benefits for physically inactive, sedentary patients with a minimum effort from healthcare professionals. However, further research about the outcome of PAP-treatment strategies is needed regarding individualized, well-defined interventions and long term follow-up in RCTs. Supporting information {#sec024} ====================== ###### Source data file. (XLSX) ###### Click here for additional data file. The authors gratefully acknowledge the PAP-responsible co-workers at Närhälsan Askim HCC, Capio HCC Axess, Carlanderska HCC, Närhälsan Frölunda HCC, Närhälsan Gibraltargatan HCC, Närhälsan Högsbo HCC, Johannesvården HCC, Närhälsan Kungshöjd HCC, Närhälsan Kungssten HCC, Närhälsan Majorna HCC, Närhälsan Masthugget HCC, Närhälsan Opaltorget HCC, Närhälsan Slottsskogen HCC, Närhälsan Styrsö HCC and Västerleden HCC for skilful assistance in this study. Helpful advice on study modeling and clinical planning were provided by Gunilla Sigurdsdotter, Carin Häggström and Britt-Marie Finbom Forsgren. The authors would like to thank Irja Burhöi, Susy Caraza Rivera and Sara Gustavsson for technical assistance with data processing and statistical analyzes. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: **Conceptualization:** SL LH ÅC.**Data curation:** SL ÅC.**Formal analysis:** SL ÅC MEHL MB LH.**Investigation:** SL ÅC.**Methodology:** SL LH ÅC MB.**Project administration:** SL.**Resources:** SL ÅC.**Software:** SL ÅC.**Validation:** SL ÅC MB MEHL LH.**Writing -- original draft:** SL ÅC MB MEHL LH.**Writing -- review & editing:** SL ÅC MB MEHL LH.

1. Introduction {#s0005} =============== Human saliva is the complex fluid secreted by major and minor salivary glands. Saliva secretion is under the control of the autonomic nervous system. The three major salivary glands are parotid, sublingual and submandibular. Human saliva consists of water, glycoproteins, enzymes, antimicrobial substances and electrolytes. Glycoproteins in saliva are responsible for the viscoelastic characteristics giving a lubricative film, which enables the free movement of oral tissues. The mucin and electrolytes in saliva maintain the oral mucosa in its hydrated state and thus providing mucosal integrity ([@b0140]). The major functions of human saliva are lubrication, antimicrobial and cleansing activity, remineralisation of enamel with calcium and phosphate and facilitating eating and speech. Salivary gland dysfunction such as xerostomia (subjective sensation of dry mouth) and hyposalivation (diminished salivary flow) ([@b0055]) are relatively common problems that can give difficulties in speech, problems with eating, mucosal infections, denture intolerance, increased dental caries, periodontal disease and loss of life quality. The usual therapies for xerostomia and hyposalivation are drinking large quantities of water, using chewing gum, candies and artificial saliva. The aims of using artificial saliva are to ensure lubrication of oral tissues, relieve the sensation of dry mouth, and protect the tooth tissues from decay. There are several approaches to produce artificial saliva, including the imitation of natural saliva which is quite complex. Usually, the commercially available artificial saliva formulation composes of carboxymethylcellulose (CMC), sodium carboxymethylcellulose (SCMC) and hydroxyethylcellulose as thickening and lubricating agents. In fact, mucilage can be found in various parts of many plants. These mucilages can be used as thickening, moisturizing and lubricating agents in artificial saliva formulations. Beside mucilages in these plants, there are also several bioactive compounds which are advantageous for the development as artificial saliva, e.g. antioxidant and anti-adherent activity. Antioxidants can enhance the immune system and prevent cancer in oral cavity, while the anti-adherent activity is one of the important properties to prevent the adhesion of microorganisms on the teeth. Mucilage is water soluble polysaccharide found in a widespread number of plants and also in some microorganisms. There are many plants which contain mucilage such as *Basella alba* Linn., *Hibiscus esculentus* Linn., *Litsea glutinosa* (Lour.) C.B. Robinson, *Ocimum canum* Sims., *Plantago ovate* Forssk., *Scaphium scaphigerum* G. Don. and *Trigonella foenum-graecum* Linn. ([@b0125]). *Basella alba* Linn. (called in common name as Ceylon Spinach or Phak Plung in Thai), family Basellaceae is a wildly cultivated, cool season vegetable (plants that have adapted to cool climates. They prefer temperatures between 55° and 75° F, which are late winter, early spring, late summer, autumn and early winter) with climbing growth habit ([Fig. 1](#f0005){ref-type="fig"}). Ceylon Spinach is an edible vegetable and has long been used as thickening agents in soups and stews. It is rich in vitamins A and C, as well as iron, calcium and soluble fiber. In addition, mucilage from this plant has also been used as topical Thai traditional medicines for the treatment of irritant, bruise, ringworm and laboring. Its stem and leaves are used as mild laxative, diuretic and antipyretic. The Ayurvedic treatment in India has used its leaves and stems for anticancer such as melanoma, leukemia and oral cancer ([@b0005]). Its mucilage is composed of mainly polysaccharides with the pH ranging between 5.3 and 5.4, containing D-galactose as a major monosaccharide and exhibiting slow swelling capacity ([@b0020]). Our previous study has demonstrated that the mucilages of *B. alba* Linn. extracted by distilled water at pH 11 using microwave for 3 min gave the highest DPPH radical scavenging and metal chelating activity of 1.01 and 11.14 folds of vitamin C and EDTA, respectively ([@b0105]). Therefore, the objective of this present study was to develop a biological active artificial saliva formulation containing mucilage from Ceylon Spinach by evaluating the physicochemical properties (viscosity, rheology wetting time) and biological activities including anti-oxidant activity and anti-adherent activity.Fig. 1The whole plant of Ceylon Spinach (*Basella alba* Linn.) showing its flowers, Family Basellaceae. 2. Materials and methods {#s0010} ======================== 2.1. Materials {#s0015} -------------- Ceylon Spinach was purchased from the local fresh market in Chiang Mai province, Thailand during April-May 2012 and identified by a botanist of the botanical garden at Faculty of Pharmacy, Chiang Mai University in Thailand. A voucher specimen (BAL-258) was deposited in the herbarium of Chiang Mai University (CMU) herbarium and flora database, Department of Biology, Faculty of Science, Chiang Mai University, Thailand. Vitamin C (L-(+)-ascorbic acid), 2, 2-diphenyl-1-picryhydrazyl (DPPH), sulforhodamine B (SRB) and resazurin sodium were from Sigma Chemical Co. (St. Louis, MO, USA). Calcium chloride was purchased from Merck, Germany. Potassium chloride and sodium fluoride from Ajax Finechem Pty Ltd., Australia were used. Tris (hydroxymethyl) methylamine was purchased from Fisher Scientific UK Limited, UK. All other chemicals and reagents were of analytical grade. 2.2. Mucilage extraction {#s0020} ------------------------ The fresh flowers of Ceylon Spinach (100 g) were cut into small pieces, macerated with 700 ml distilled water for 24 h and microwaved at 600 W intensity for 5 min. The mixture was then pressed through a muslin cloth. The filtrate containing the mucilage was centrifuged at 4,660g (centrifuge machine, Fisher Scientific Inc., New York, U.S.A) 25 °C for 30 min. The supernatant was collected, mixed with 95% ethanol (3 folds in volume) to precipitate the mucilage and re-centrifuged at 4,660g for 15 min. The precipitate was collected and the remaining ethanol in the precipitate was removed by a rotary evaporator (Buchi, Flawil, Switzerland) (50 ± 2 °C) until all ethanol was evaporated and lyophilized by a lyophilizer (Christ FOC-1 Model K-40 equipment, Balzers-Pfeiffer GmbH, Asslar, Germany) at −50 ± 2 °C. The dried lyophilized powder of the mucilage was kept at room temperature (25 ± 2 °C) until use. 2.3. Phytochemical assays {#s0025} ------------------------- The mucilage was analyzed for phytochemical constituents (anthraquinones, glycosides, tannin, carotenoids, flavonoids and alkaloids) using the standard methods ([@b0100]). For anthraquinone, 0.05 g of the mucilage was put into a dry test tube, added with 2 ml of chloroform and shaken for 5 min. The mucilage was filtered. The filtrate was mixed with an equal volume of 10% ammonia solution and shaken. A pink violet or red color in the ammoniacal layer (lower layer) indicated the presence of anthraquinone. The qualitative assay of reducing sugars was performed by TLC method. The mucilage dissolved in water was spotted on the silica gel plate in comparing to the standard reducing sugars (glucose, fructose and sucrose). The filtrate was resolved on the TLC plate coated with silica gel 60. The mobile phase was butanol/acetic acid/diethyl ether/water (9:6:1:3). The spot on the plate was sprayed with 10% H~2~SO~4~ and heated. Sucrose, glucose and fructose were used as the standards. For tannins, 0.05 g of the mucilage was mixed with 2 ml of 15% FeCl~3~ solution. The blue-black precipitate indicated the presence of tannins. For carotenoid, each mucilage sample was extracted with chloroform in a test tube with vigorous shaking. The resulting mixture was filtered and 0.1 ml of H~2~SO~4~ was added. The blue color at the interface showed the presence of carotenoids. For the presence of flavonoid, 2 ml of the mucilage solution mixed with 1 ml of the concentrated HCl and magnesium ribbon gave the pink tomato-red color. For alkaloids, an amount of 0.05 g of the mucilage in 2 ml of 1.5%v/v HCl was boiled on a water bath and 6 drops of the Dragendorff's reagent were added. The orange precipitate indicated the presence of alkaloids. 2.4. The development of artificial saliva formulations containing mucilage from Ceylon Spinach {#s0030} ---------------------------------------------------------------------------------------------- ### 2.4.1. Preparation of the artificial saliva formulations {#s0035} Five developed artificial saliva formulations were prepared ([Table 1](#t0005){ref-type="table"}). The compositions of the formulations were mucilage from Ceylon Spinach, calcium chloride (CaCl~2~), potassium chloride (KCl) and sodium fluoride (NF) as electrolytes and minerals, and 20% concentrated paraben (18% methyl paraben and 2% propyl paraben) as preservative. Briefly, the mucilage from Ceylon Spinach together with the electrolytes and minerals were dispersed in 100 ml of 20 mM phosphate buffer (pH 7.4). The mixture was then homogenized by a homogenizer (Brinkmann, Kinematica GmbH, Switzerland) at room temperature (25 ± 2 °C) for 15 min and left for 30 min. Then, 1% of the concentrated paraben was added. The formulations were filtered through a Whatman® No. 1 filter paper and kept in a tight bottle for the further experiment.Table 1Compositions of the 5 developed artificial saliva formulations.Compositions (g)No.1No.2No.3No.4No.5Ceylon Spinach mucilage0.55000.73000.83000.66000.6100CaCl~2~·2H~2~O0.02420.02260.02160.02320.0237KCl0.14280.13340.12820.13710.1397Na~2~HPO~4~0.12650.12650.12650.12650.1265NaH~2~PO~4~·H~2~O0.15300.15300.15300.15300.1530NaF0.00100.00100.00100.00100.001020% concentrated paraben11111[^1] ### 2.4.2. Physico-chemical properties of the developed artificial saliva formulation containing mucilage from Ceylon Spinach {#s0040} #### 2.4.2.1. Natural human saliva collection {#s0045} The natural human saliva was collected from healthy volunteers (N = 5, 25--35 years old). The volunteers gargled their mouths with drinking water prior to saliva collection. #### 2.4.2.2. Viscosity and rheology determination {#s0050} Viscosity and rheology of the developed artificial saliva formulations were determined by a viscometer (Myr VR 3000 model, Tarragona, Spain) with the range of the shear rate between 5 and 200 rpm. All measurements were carried out at 25 ± 2 °C and 50 ml volume of the formulation were used in each test. The shear stress of the artificial saliva formulations was calculated by the following equation:$$\text{Shear}\,\text{stress} = \frac{\text{shear}\,\text{rate}}{\text{Viscosity}}$$ #### 2.4.2.3. Wetting time determination {#s0055} Wetting time of the developed artificial saliva formulations was measured by the modified method previously described ([@b0080]). Briefly, a filter paper (Whatman® No. 1) was placed in a Petri dish containing 6 ml of the sample (distilled water, natural saliva from human volunteer and the developed artificial saliva formulation). A tablet with methylene blue dye powder on the upper surface was carefully placed on the filter paper in the Petri dish. The time (mins) required for the sample to reach the upper surface of the tablet by observing from the completely wetting with methylene blue was noted as the wetting time. #### 2.4.2.4. Selection of the best artificial saliva formulation {#s0060} The formulation with the best physical properties (clarity, precipitation, viscosity/rheology and wetting time) was selected for the further study. 2.5. Biological activities of the selected artificial saliva formulation containing mucilage from Ceylon Spinach {#s0065} ---------------------------------------------------------------------------------------------------------------- ### 2.5.1. DPPH radical scavenging activity {#s0070} Free radical scavenging activity of the selected artificial saliva formulations was determined by the modified DPPH assay ([@b0095]). Briefly, 50 µl of five serial concentrations (0.001--10 mg/ml) of the sample and 50 µl of ethanolic solution of DPPH were added into each well of a 96-well microplate (Nalge Nunc International, NY, USA). The reaction mixtures were allowed to stand for 30 min at 25 ± 2 °C and the absorbance was then measured at 515 nm by a well reader (Bio-Rad, Model 680 Microplate Reader, USA) against the negative control (distilled water). Ascorbic acid (0.001--10 mg/ml) was used as a positive control. The experiment was done in triplicate. The percentages of free radical scavenging activity were calculated as the following: Scavenging (%) = \[(A − B)/A\] × 100, where A was the absorbance of the negative control and B was the absorbance of the sample. The sample concentration providing 50% of scavenging (SC~50~) activity was calculated from the graph plotted between the percentages of the scavenging activity and the sample concentrations. ### 2.5.2. Cytotoxicity on normal human gingival fibroblasts {#s0075} #### 2.5.2.1. Cell culture {#s0080} Normal human gingival fibroblasts were obtained from the gingival tissue by the explant technique at Faculty of Dentistry, Chiang Mai University in Chiang Mai, Thailand. They were cultured in the 30-mm diameter tissue culture dishes in the complete culture medium containing α-Modified Eagles culture medium (MEM-Alpha, Hyclone, Utah, USA) supplemented with 10% (v/v) fetal bovine serum (FBS, Hyclone, Utah, USA), penicillin (100 U/ml; Hyclone, Utah, USA) and streptomycin (100 mg/ml; Hyclone, Utah, USA). Cells were incubated in a temperature-controlled, humidified incubator (Shel Lab, model 2123TC, USA) with 5% CO~2~ at 37 °C and subcultured every 5--7 days. The cells at the 5th to 8th passage were used in this study. #### 2.5.2.2. Sample preparation {#s0085} The extract at 6.1 mg/ml and the developed artificial saliva formulations containing 6.1 mg/ml of mucilage from Ceylon Spinach were used for cytotoxicity test compared to natural human saliva at 1 mg/ml. All samples were filtered through a membrane filter (0.2 µm) before use. #### 2.5.2.3. Cytotoxicity by SRB assay {#s0090} The cells were seeded in 96-well plates at an amount of 10,000 cells/well and allowed to attach overnight at 37 °C in 5% CO~2~ incubator. Then, the cells were exposed to the sample for 24 h. After incubation, the adherent cells were fixed by adding cold 50%w/v trichloroacetic acid and further incubated for 1 h at 4 °C. Then, the cells were rinsed with distilled water, air-dried and stained with 0.4% SRB in 1%glacial acetic acid for 30 min at room temperature (27 ± 2 °C). The unbound SRB was removed by washing with 1% glacial acetic acid solution for four times. After air-drying, 100 µl per well of 10 mM Tris base were added to dissolve the bound stain. After mixing, the absorbance was measured at 540 nm with a microplate reader (Biorad, Milan, Italy). The untreated cells were used as a negative control. Cell viability (%) was calculated by the following equation:$$\text{Cell}\,\text{viability}\,\left( \% \right) = \left( {\text{Absorbanc}\text{e}_{\text{treated}\;\text{cells}}/\text{Absorbanc}\text{e}_{\text{untreated}\;\text{cells}}} \right) \times 100$$ ### 2.5.3. Anti-adherent activity determination {#s0095} #### 2.5.3.1. Bacterial culture {#s0100} *Streptococcus mutans* (obtained from Faculty of Dentistry, Chiang Mai University, Thailand) was inoculated into tryptic soy broth and incubated at 37 °C for 24 h. The absorbance at 550 nm (OD~550~) was measured and the cell concentration was adjusted to obtain the OD~550~ of 0.5. The bacterial concentration was diluted one fold to obtain the bacterial concentration of 1 × 10^6^ CFU/ml. #### 2.5.3.2. Sample preparation {#s0105} The selected artificial saliva formulation containing the Ceylon Spinach mucilage and the natural human saliva from the human volunteers were used. Briefly, an amount of 20 ml of the natural saliva was collected from the human volunteers, centrifuged at 10,000g, 4 °C for 15 min and the clear solution was collected. Distilled water was used as a negative control. The developed artificial saliva formulation, the natural saliva and the distilled water were filtered through a membrane filter (0.2 µm) before use. #### 2.5.3.3. Anti-adherent activity {#s0110} The anti-adherent activity was performed as previously described ([@b0105]). Briefly, hydroxyapatite (HA) was dispersed in phosphate buffer (pH 6.8) at the concentration of 5 mg/ml. An amount of 200 µl of the HA suspension was added into each well of the 96-well plate, centrifuged at 1,000g for 15 min and the supernatant was discarded. The natural saliva (200 µl) was added into each well, incubated at 37 °C with shaking at 80 rpm for 120 min and centrifuged at 1,000g for 15 min. The HA was washed by phosphate buffer. The 100 µl of *S. mutans* with 100 µl each of either the selected artificial saliva formulation, natural saliva or distilled water were added, incubated at 37 °C with shaking at 80 rpm for 120 min and then incubated without shaking at 37 °C for 60 min. The plate was centrifuged at 1,000g for 15 min and washed with phosphate buffer. The resazurin solution (100 µl) was added into each well and incubated at 37 °C with shaking at 80 rpm for 60 min. The fluorescence intensity was measured by a spectrofluorometer (Jusco, Hachioji, Japan) with the excitation/emission at the λ of 562/595. The % adhesion was calculated in comparing to the control (distilled water) as the followings:$$\%\;{\text{Decrease\ of\ adherence\ =}\mspace{6mu}}\frac{\text{F}_{\text{control}} - \text{F}_{\text{sample}}}{\text{F}_{\text{control}}} \times 100$$$$\%\;{\text{Increase\ of\ adherence\ =}\mspace{6mu}}\frac{\text{F}_{\text{sample}} - \text{F}_{\text{control}}}{\text{F}_{\text{control}}} \times 100$$where F~control~ was the fluorescence intensity of the control and F~sample~ was the fluorescence intensity of the sample. 2.6. Statistical analysis {#s0115} ------------------------- The results were presented as mean ± SD of three independent experiments. ANOVA was used for the analysis of the test results at the significance level of *p*-value \< 0.05. 3. Results and discussion {#s0120} ========================= 3.1. Physical properties of the Ceylon Spinach mucilage and the developed artificial saliva formulations containing mucilage from Ceylon Spinach {#s0125} ------------------------------------------------------------------------------------------------------------------------------------------------ ### 3.1.1. Physical appearances {#s0130} The lyophilized powder of the mucilage from Ceylon Spinach was in green-brown appearance with the characteristic greenish odor. The percentage yield of the mucilage was 0.44% w/w of the fresh plant. The Ceylon Spinach mucilage gave the positive results of phytochemical test of flavonoids and fructose. The 5 developed artificial saliva formulations gave different physical appearance ([Fig. 2](#f0010){ref-type="fig"}). Formula No.5 containing 0.61 g of the mucilage from Ceylon Spinach, 0.0237 g of CaCl~2~, 0.1397 g of KCl and 0.1530 g of NaF gave the best physical appearance which was translucent light green solution without precipitation.Fig. 2Physical appearances of the 5 developed artificial saliva formulations conatining mucilage from flowers of Ceylon Spinach. ### 3.1.2. Viscosity and rheology of the developed artificial saliva formulations {#s0135} The formulations which showed good physical appearance with no precipitation having the viscosity value and rheology behavior similar to natural human saliva which is a non-Newtonian pseudoplastic flow that is the most important physical characteristic of human saliva and gave no microbial contamination were selected. All of the 5 developed artificial saliva formulations gave good physical appearance with no precipitation with formulation nos. 1, 2, 3, 4 and 5 showing the non-Newtonian pseudoplastic flow with difference viscosity values of 6.8 ± 0.1, 14.4 ± 0.5, 17.3 ± 0.3, 10.1 ± 0.5 and 8.9 ± 0.2 cP, respectively. Thus, only the developed artificial saliva formulation No. 5 gave the viscosity of 8.9 ± 0.2 cP which was very close to the natural human saliva (9.0 ± 0.1 cP) was selected for the further experiment. The viscosity of the artificial saliva formulation No.5 decreased with increased shear rate, which was an essential property of saliva as shown in [Fig. 3](#f0015){ref-type="fig"}. This formulation also indicated the non-Newtonian pseudoplastic flow the same as the natural saliva ([@b0135]). Saliva has important rheological properties that may affect mouth feel and other sensory perceptions. Mucin glycoproteins containing in natural saliva are known to be important factors for the extensional rheological properties (viscosity, elasticity and stickiness). Also, different rheological behaviours of saliva from each gland secretion could be due to the influence of mucin concentration, mucin conformation and/or the mucin type within the glandular saliva ([@b0035]). As known, the main constituents of plant mucilages are galactose, arabinose, rhamnose, uronic acids, galacturonic acid, protein, Ca and Mg ([@b0180]). The major composition in the Ceylon Spinach mucilage which is responsible for viscosity is pectin, a polysaccharide with rhamnose and neutral sugar with the α-(1--4) glycoside linkage. Pectins containing a significant amount of galacturonic residues in which the presence of the counterions in their solution can affect the inter- and intra-molecular interactions. The reduction of these intramolecular forces can let the coils to contract to the more compact conformation, with the consequent of the reduction in intrinsic viscosity ([@b0075]). In fact, the viscosity of polysaccharide solution has been reported to increase with increased concentrations, but decrease with increased pH and temperatures ([@b0010]). It has been previously indicated that the higher concentration of the Ceylon Spinach mucilage, the lower the pH and the higher viscosity were obtained ([@b0020]). At higher concentration, the mucilage can have a very rigid structure which can attribute to the formation of hydrogen bonds between the polysaccharide and water ([@b0145]). The selected artificial saliva formulation No.5 containing mucilage from Ceylon Spinach exhibited similar rheological properties of non-Newtonian pseudoplastic flow to the natural saliva with the viscosity of 8.9 cP.Fig. 3Viscosity (A) and rheology behavior (B) of the selected artificial saliva formulation No.5 containing mucilage from Ceylon Spinach showing the non-Newtonian pseudoplastic flow similar to the natural human saliva. ### 3.1.3. Wetting times of the developed artificial saliva formulations {#s0140} The wetting time is an important property of saliva, since the shorter wetting time can improve food swallowing in xerostomia and hyposalivation patients. Thus, food can be wet in few minutes to make it easier to chew and be swollen. Out of the 5 developed formulations, the wetting time of the artificial saliva formulation No.5 containing mucilage from Ceylon Spinach was 12.50 ± 2.24 min which was close to the normal human saliva (10.87 ± 1.79 min) ([Table 2](#t0010){ref-type="table"}). The mucin glycoprotein in natural saliva has a lubricating, wetting and softening effect from its hydrophilic and hydrophobic segments that acts as a surfactant resulting the surface tension reduction between water and air ([@b0110]). Negative charge moieties of the mucin (glycosylated region) can interact with water molecules, create a hydration shell and improve hydration and lubrication ([@b0030]). Thus, loss of negatively charged glycan residues is a proposed mechanism for oral dryness through the reduced water retention capacity of mucin, leading to reduced mucosal hydration. Moreover, the contact angle of saliva can reflect the degree of wetting of saliva on surfaces and hence gives an insight into the interaction of saliva with surfaces ([@b0175]). Thus, the wetting property of formulation No.5 may be from the bound moisture of the mucilage that improves wettability by the mechanism of surface tension reduction. In addition, polymers in the Ceylon Spinach mucilage have been reported to have hydrophilic constituents, with hydroxyl and carboxyl groups ([@b0130]), that can be swollen in water thereby exposing to the maximum number of the adhesive sites ([@b0150]). Hence, the formula No.5 which gave superior physical properties of viscosity and wetting time similar to natural saliva was selected for the further study.Table 2Physical properties of the selected artificial saliva formulation No.5 containing mucilage from Ceylon Spinach.SamplePhysical appearancepH valueRheologyWetting time (min)Natural human salivaturbid solution7.32 ± 0.51non-Newtonian\ pseudoplastic10.87 ± 1.79Ceylon Spinach mucilagegreen-brown solid5.58 ± 1.09non-Newtonian\ pseudoplastic10.65 ± 0.69The artificial saliva formulation No.5 containing mucilage from Ceylon Spinachtranslucent light green solution7.41 ± 0.35non-Newtonian\ pseudoplastic12.50 ± 2.24 3.2. Antioxidative activities of the selected artificial saliva formulation No.5 containing mucilage from Ceylon Spinach {#s0145} ------------------------------------------------------------------------------------------------------------------------ Antioxidative activities of the selected artificial saliva formulation containing mucilage were shown in [Table 2](#t0010){ref-type="table"}. The formulation gave the DPPH scavenging activity with the SC~50~ value of 14.26 ± 2.00 mg/ml which was 0.05 folds of ascorbic acid, and was similar to mucilage from the Ceylon Spinach (the SC~50~ value of 13.84 ± 0.14 mg/ml). However, DPPH scavenging activity was not observed in natural human saliva. The SC~50~ value of the mucilage from the Ceylon Spinach observed in this study was different from the previous study which gave the SC~50~ value of 514.41 µg/ml ([@b0020]). This difference may be from the different extraction method. The observed antioxidant activity of the developed artificial saliva formulation No.5 may be from the presence of polysaccharide containing in the mucilage of Ceylon Spinach such as galactose, arabinose, glucose and mannose ([@b0040], [@b0045]). Several previous works have reported the antioxidant activity of polysaccharides from plants. Polysaccharides including rhamnose, glucuronic acid, galacturonic acid, glucose, galactose and arabinose from Wampee \[Clausena lansium (Lour.) Skeels\] gave high antioxidant activities of 78% hydroxyl radical-scavenging ability and 48.35% antioxidative activity ([@b0185]). Water-soluble polysaccharides from Wolfberry (*Lycium barbarum* L.), Sweet Cherry (*Prunus avium* L.), Kiwi (*Actinidia chinensis* L.) and Cranberry fruits (*Vaccinium macrocarpon* Aiton) indicated the antioxidant activities assayed by the oxygen radical absorbance capacity (ORAC) and Trolox equivalent antioxidant capacity (TEAC) ([@b0050]). The mucilage from Ceylon Spinach of this present study contained not only polysaccharides, but also flavonoids the same as the previous study ([@b0115]). Functional hydroxyl groups in flavonoids mediate their antioxidant effects by scavenging free radicals and chelating metal ions. They also found ascorbic acid and especially phenolic compounds, which have been reported to have antioxidant activities from their redox properties of acting as a reducing agents hydrogen donors and singlet oxygen quenchers. The radical scavenging activities of Ceylon Spinach may be related not only to the phenolic contents, but also the flavonoid, anthocyanin and proanthocyanin contents ([@b0120]). Kaempherol which was a flavonoid found in Ceylon Spinach has been reported to exhibit strong DPPH radical scavenging activity with the SC~50~ value of 4.349 µg/ml ([@b0190]). Hence, the synergistic effect of all phytochemicals in Ceylon Spinach mucilage may be responsible for this free radical scavenging biological activity. 3.3. Cytotoxicity of the selected artificial saliva formulation No.5 containing mucilage from Ceylon Spinach {#s0150} ------------------------------------------------------------------------------------------------------------ Several studies have investigated cytotoxicity of mucilaginous polysaccharide content in plants. For examples, Okra (*Abelmoschus esculentus* Linn. moench) polysaccharides showed no cytotoxicity on human cancer cell line with nearly 100% cell viability and no morphological changes ([@b0065]). Mucilage from *Grewia optiva* at 10, 20 and 30 μg/ml gave non-toxic on normal (NIH3T3) cell lines ([@b0085]). Sushila et al. have also demonstrated that the extract from *Basella alba* whole plant indicated significant dose dependant cytotoxicity on the Jurkat cell lines, but gave no toxicity on lung cancer cell line (A549) with 88.33 ± 1.20% cell viability at 25 µg/ml ([@b0170]). There was no report of Ceylon Spinach mucilage and its artificial saliva formulation on human gingival fibroblast cytotoxicity. This present study has been the first report to determine the safety for oral administration of the artificial saliva formulation containing mucilage from Ceylon Spinach on human gingival fibroblasts. Cytotoxicity of the selected artificial saliva formulation containing mucilage from Ceylon Spinach determined by SRB assay was shown in [Table 3](#t0015){ref-type="table"}. The formulation with the pH value of 7.41 ± 0.35 exhibited no cytotoxic effect on the treated cells with 99.20 ± 12.09% cell viability. However, the mucilage from Ceylon Spinach showed high cytotoxic effect of only 23.90 ± 10.43% cell viability, owing to the acidity (pH 5.58 ± 1.09) of the mucilage. pH is one of the most important factors of growth promoting properties including contact inhibition, growth rate and cell mobility. This result was different from the previous study showing that the aqueous extract of *Basella alba* (Ceylon Spinach) indicated the LC~50~ value at \>1000 ppm assayed by the brine shrimp (*Artemia salina*) lethality bioassay, indicating of no cytotoxicity ([@b0015]). As previously reported, the pH range of 6.8--8.2 enhanced human-human hybridoma (HB4C5) cell viability, whereas the outside ranges demonstrated a decrease in cell viability ([@b0195]). Also, it has been previously demonstrated that Ceylon Spinach mucilage at 2 mg/ml showed relatively mild toxicity on Chang liver cell line with 84.4% cell viability ([@b0020]).Table 3Antioxidative activity and cytotoxicity on human gingival fibroblasts of Ceylon Spinach mucilage and the artificial saliva formulation No.5 containing mucilage from Ceylon Spinach.SampleAntioxidative (SC~50~; mg/ml)Cell viability (%)Ascorbic acid0.72 ± 0.0087.44 ± 10.38Natural human salivaNA98.28 ± 8.92Ceylon Spinach mucilage13.84 ± 0.14\*23.90 ± 10.43\*The artificial saliva formulation No.5 containing mucilage from Ceylon Spinach14.26 ± 2.00\*99.20 ± 12.09[^2] Ascorbic acid (vitamin C) is a common positive control of antioxidant activity test by DPPH free radical scavenging. For cytotoxicity test, several studies have shown that vitamin C increases numbers of collagen bundles in the regenerating periodontal tissue, detoxifies histamine in gingival inflammation and reduces gingival oxidative stress. Vitamin C reduces the cytotoxic and apoptotic effects of *P. gingivalis* on human gingival fibroblasts (HGF) ([@b0165]). This present study has expected that the effect of the developed artificial saliva formulation on HGF was similar to vitamin C and showed the safety of the developed artificial saliva formulation for oral use. 3.4. Anti-adherent activity of *S. mutans* of the selected artificial saliva formulation No.5 containing mucilage from Ceylon Spinach {#s0155} ------------------------------------------------------------------------------------------------------------------------------------- *S. mutans* has been implicated as a primary causative agent of dental caries. Thus, inhibition of *S. mutans* adhesion to the tooth surface is a major goal for the prevention of dental caries. Anti-adherent activity of the selected artificial saliva formulation containing mucilage from Ceylon Spinach were shown in [Table 4](#t0020){ref-type="table"}. In this study, the natural saliva demonstrated an increase adherence of *S. mutans* on the HA (hydroxyapatite) beads of 33.10 ± 9.70% over the control (distilled water). This may be from the compositions in the natural saliva, such as proteins and polysaccharides. Natural saliva contains a multitude of proteins which are important for oral microbial ecology and biofilm formation. Adsorption of specific salivary proteins, such as acidic proline-rich proteins and agglutinin, promotes the adhesion of *S. mutans* onto the HA surfaces by providing ligands for bacterial attachment ([@b0155]). Moreover, high molecular weight salivary glycoprotein promotes adhesion of *S. mutans* ([@b0160]). Hydrophobic interactions have been demonstrated to be important in bacterial adhesion. Also, the negative charges containing in polysaccharides might be necessary for adsorption onto the HA surfaces. Ceylon Spinach mucilage increased the adherence of only 4.83 ± 3.47% which was less than the natural saliva of 6.85 times. Thus, Ceylon Spinach mucilage appeared to reduce *S. mutans* adherence on the HA beads in comparing to the natural saliva. The mechanism of the mucilage from Ceylon Spinach on the reduction of *S. mutans* adhesion might be due to the amino sugars, such as glucosamine containing in the mucilage which can inhibit the binding of *S. mutans* to the saliva-coated HA ([@b0060]). The low molecular weight proteins reduced the saliva-promoted *S. mutans* adherence by competitively inhibiting *S. mutans*-binding glycoprotein adsorption onto the HA surface has also previously reported ([@b0155]). Thus, the amino acid such as arginine, leucine, isoleucine, lysine, threonine and tryptophan containing in the Ceylon Spinach mucilage may inhibit *S. mutans* adhesion to the HA surface ([@b0070]).Table 4Anti-adherent activity of *S. mutans* of the selected artificial saliva formulation No.5 containing mucilage from Ceylon Spinach.Sample%Decrease of adherence%Increase of adherenceNatural human saliva--33.10 ± 9.70Ceylon Spinach mucilage--4.83 ± 3.47Vehicle of the artificial saliva formulation3.66 ± 0.88--The artificial saliva formulation No.5 containing mucilage from Ceylon Spinach17.01 ± 7.75--[^3][^4][^5] The selected artificial saliva formulation containing mucilage from Ceylon Spinach exhibited a decrease adhesion of *S. mutans* on HA beads of 17.01 ± 7.75%, whereas the vehicle of the formulation decreased the adherence of 3.66 ± 0.88%. Hence, it appeared that the Ceylon Spinach mucilage containing in the developed artificial saliva formulation may enhance the anti-adherent activity of *S. mutans* of the vehicle of about 13%. The reduction of bacterial adhesion of the vehicle may be due to the bacterial growth reduction by its compositions of parabens and fluoride. Parabens are antimicrobial agents which can inhibit irreversibly glycolysis by the cariogenic dental plaque bacterium, *S. mutans* ([@b0090]). Sodium fluoride can reduce the growth of S. mutans by the sub-minimum inhibitory concentration (MIC) levels of fluoride (MIC: 282 ppm F^−^) ([@b0040], [@b0045]). Also, the commercial fluoride varnishes showed a strong inhibitory effect on the bacterial adhesion of approximately 67--98% reduction ([@b0025]). The anti-adherent activity of *S. mutans* of the selected artificial saliva formulation No.5 was thus the synergistic effects from the Ceylon Spinach mucilage and other compositions in the vehicle. Hence, the selected artificial saliva formulation No.5 containing mucilage from Ceylon Spinach indicated superior anti-adherent activity of *S. mutans* to the normal human saliva which is one of the important advantages of this developed artificial saliva formulation. 4. Conclusions {#s0160} ============== The results from this study have suggested the potential application of mucilage from Ceylon Spinach for artificial saliva because of not only its superior physical properties including rheological property and wetting time similar to the natural human saliva, but also having higher antioxidant and anti-adherence activities than the natural human saliva, and with no cytotoxic effect on normal human gingival fibroblasts. Moreover, Ceylon Spinach is an edible vegetable and has been used for foods. Thus, the developed artificial saliva formulation containing mucilage from this plant is safer for consumption than other chemical artificial saliva formulations. This present study has also demonstrated the possibility of using mucilage from edible plants which are safe and cost effective to prepare a biological active artificial formulation for the substitution of natural human saliva and other chemical artificial saliva formulations. Declaration of Competing Interest ================================= The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. This work was supported by Thailand Research Fund (TRF) for small and medium size enterprises (SMEs), RDG5450019, Manose Health and Beauty Research Center in Thailand and Thai-China Flavours and Fragrances Industry Co., Ltd., Phra Nakhon Si Ayutthaya, Thailand. Peer review under responsibility of King Saud University. [^1]: Note : 20% concentrated paraben solution contained 18% methyl paraben and 2% propyl paraben. [^2]: Note: NA = no activity; Cell viability \> 90% = non-toxic, 60--90% = slightly toxic, 30--59% = moderately toxic and \<30% = highly toxic, \* significant difference (p \< 0.05) in comparing with ascorbic acid. [^3]: Note: [^4]: $\;\%\;{\text{Decrease\ of\ adherence\ =}\mspace{6mu}}\frac{\text{F}_{\text{control}} - \text{F}_{\text{sample}}}{\text{F}_{\text{control}}} \times 100.$ [^5]: $\;\%\;{\text{Increase\ of\ adherence\ =}\mspace{6mu}}\frac{\text{F}_{\text{sample}} - \text{F}_{\text{control}}}{\text{F}_{\text{control}}} \times 100.$

1.. INTRODUCTION ================ The pattern of stresses transferred to the bone have great influence on the success or failure of an orthopedic implant. The evaluation of bone stresses is so complex that it cannot be accomplished analytically, necessitating the application of FEA (Finite Element Analysis) techniques; three types of FE model, axisymmetric, bi-dimensional and three-dimensional, are considered in the literature \[[@R1]-[@R5]\]. Two-dimensional models are flawed in that stresses outside the plane of analysis are disregarded, while 3D models require a large number of elements and, consequently, long calculation times. The axisymmetric model, where the only simplification is that the screw thread is modeled as a disk, can be considered a good compromise between 2D and 3D models. With regard to the constraints, the condition of osseointegration is usually simulated, and the post-operative condition it is rarely considered in the literature \[[@R3],[@R4],[@R6]\]. However, these two conditions produce significantly different results and can be considered the limit conditions under which orthopedic screws operate, so that the analysis of both cases can provide useful information. Two boundary conditions (pull out test; alternative condition) are simulated in the literature \[[@R4],[@R7]\], but the relation between their respective results has not been analyzed. This relation is relevant because the pull out test is the standard test for orthopedic screws, while the screws, once implanted, are subjected to different loading conditions. Regarding bone material, most models in the literature consider bone an isotropic, elastic and homogenous material \[[@R1]-[@R4],[@R7]-[@R9]\], while in reality bone is anisotropic because of its trabecular structure. In this study, bone isotropy was assumed in order to obtain more general results, while bone structure was defined by a single parameter, its volumetric density. In the literature, the geometric and mechanical parameters of the screw generally considered are: pitch \[[@R10],[@R11]\], length, flank angle, and material \[[@R2]-[@R5]\]. Few authors have considered the fillet radius \[[@R4],[@R7]\], while many models have sharp edges \[[@R1],[@R3],[@R12]\]. Little emphasis has been given to screw performance in relation to bone density \[[@R5],[@R13]\] even though it is well known that the density of the bone determines its mechanical properties \[[@R14]-[@R18]\]. The present study evidences that it is not possible to select the appropriate screw without considering bone density. Different models were developed, considering the geometry and mechanical properties of commercial screws, and a parametric analysis was undertaken to assess how the strength of the bone-screw system varies for different values of thread pitch and bone density. Actually, this paper introduces a methodology where a multi-parametric structural numerical analysis is integrated with a multi-factorial analysis in order to be able to summarize a great amount of results and to build predictive analytic models. Overall, it was possible to determine some criteria for system optimization. 2.. MATERIALS AND METHODOLOGY ============================= Stress analysis required the construction of apposite FE models, which were then validated by means of experimental tests. 2.1.. Finite Element Model -------------------------- The model shown in Fig. (**[1](#F1){ref-type="fig"}**) was developed using MSC MARC^®^ 2003 software. The bone consists of cortical bone (E =11 GPa, ν=0.33, 5 mm in thickness) and trabecular bone. Two different trabecular bone densities were simulated (Table **[1](#T1){ref-type="table"}**), and their respective mechanical properties were obtained from the relations \[[@R14], [@R16]\]: $$E = 2015 \cdot \mathit{\rho}^{25}$$ $$\mathit{\sigma}_{U} = 0.0042 \cdot E - 0.039$$ where: ρ is bone density (g/cm^3^); E is Young's modulus (MPa); σ~U~ is the ultimate tensile stress (MPa). The bone density values were chosen to simulate recurrent clinical situations (adult patient with mild bone resorbing) and significantly different Young's modulus (1:2 ratio). With regard to the screws, four different constitutive materials were simulated in order to study both currently used materials such as stainless steel (screw models n. 9-10) and titanium Ti6Al4V (screw models n. 3-8), and more innovative solutions such as low stiffness titanium (Ti12Mo5Ta, screw model n.2) and PMMA reinforced by an inner Ti12Mo5T cylinder (screw model n.1); the last two materials have been chosen in order to better approach trabecular bone's Young modulus. The general geometry of the simulated screws and the mesh details are shown in Fig. (**[2](#F2){ref-type="fig"}**) (symbols as in Table **[2](#T2){ref-type="table"}**); they were generated from an effectively produced geometry (screw model n. 9), varying all parameters, one by one. All the screw threads were simulated with symmetrical flanks, fillet radii were all equal, and the pitch was constant along the screw axis. The axisymmetric model entails shorter calculation times. The constraints and boundary conditions are shown in Fig. (**[1](#F1){ref-type="fig"}**): they correspond to the pull out test (Fig. **[1a](#F1){ref-type="fig"}**) or to an alternative condition, with a different position of the constraint (Fig. **[1b](#F1){ref-type="fig"}**); this alternative condition should represent a more realistic working condition. According to Saint Venant's, the second condition is equivalent to the first one, however the small displacement hypothesis cannot be assumed a priori in this case, due to the low stiffness of trabecular bone. For both configurations, the immediate post-operative condition and that of complete osseointegration were simulated: the first condition was created through a 'touch' contact between the screw and trabecular bone (whatever displacement is allowed with the exception of penetration), while the second was obtained through a 'glue' contact (the two components are completely bounded to each other). The mean number of elements per model was 7500. 2.2.. Maximum Allowed Force Evaluation -------------------------------------- The maximum allowed force was calculated by iteration (non-linear case), specifying that the maximum stress on the bone must not exceed its ultimate tensile stress. This hypothesis is quite restrictive: actually bone is not a fragile material, tension peaks produce localized plastic deformations and stresses are redistributed over a wider area. However bone is likely to be stressed by dynamic loads and fatigue damage may occur when stress peaks produce a crack which progressively propagates during loading, until structural failure occurs. 2.3.. Experimental Tests ------------------------ The finite element model with 'touch' boundary condition was validated by means of experimental tests. All tests were performed on an hydraulic INSTRON 8872 test machine, equipped with a ±5 kN load cell. Pull out tests \[[@R19]\] were conducted replacing bone by polyurethane foam, as reported in the literature \[[@R20]\]. The mechanical properties of polyurethane foam were determined by ten compression and ten tensile static tests, where a linearly increasing displacement was applied. The nominal stress was calculated from the force, divided by the initial specimen section; the nominal strain was calculated from the current displacement, divided by the initial specimen height. Young's modulus was determined through five dynamic tests, applying a pulsating sinusoidal load (Fig. **[3](#F3){ref-type="fig"}**, left). Young's modulus was calculated as 6 MPa (s.d. = ± 1.21 MPa), while ultimate tensile stress was 0.42 MPa (s.d. = ± 0.084 MPa); these data were input into the numeric model to be validated. Compression tests produced a progressive packing of the foam (reaching up to 60% height reduction) which does not break but shows buckling. The pull out tests required specific equipment: two holes drilled in the opposite sides of a rectangular metal profile (obtained from an extruded bar) and a polyurethane block was placed within the box profile (Fig. **[4](#F4){ref-type="fig"}**). A bolt screwed into the lower hole was then held by the lower jaw of the machine (the bolt was left free to move in the cross-sectional plane, in order to avoid bending moments acting on the screw). A screw was inserted through the upper hole and inserted into the polyurethane foam. The metal profile simulated the cortical layer of the FE model, and prevented the polyurethane foam deforming when the screw was pulled. A preparatory hole (6.5 mm in diameter) was drilled in the foam before inserting the screw, which was always set at the same height. The tests were conducted on a standard metrical screw (M10, UNI 4536) whose geometry was known in detail, allowing an accurate FE model to be constructed. The lower end of the shaft was threaded with 17 threads. The screw was pulled at a speed of 2 mm/min. The pull out test was repeated 23 times and the mean force determined was 120 N (s.d. = ±14 N). A typical force versus displacement curve is shown in Fig. (**[4](#F4){ref-type="fig"}**). 2.4.. Analysis of FE Model Results ---------------------------------- Ten different screw models were realized (Table **[2](#T2){ref-type="table"}**). Each model was implemented with four different pitches (1.4 mm, 2.4 mm, 3.4 mm, 4.4 mm) and two different values of bone density (0.35 g/cm^3^; 0.47 g/cm^3^). Overall, each screw model was analyzed in eight FE models. The numerical results were summarized through a multifactor model \[[@R22]\], examining both the effects of each single factor and the interaction between the pitch and bone density. $$Y = \mathit{\mu} + \mathit{\beta} \cdot d + \mathit{\gamma}_{L} \cdot p + \mathit{\gamma}_{Q} \cdot p^{2} + \mathit{\gamma}_{C} \cdot p^{3} + \mathit{\beta}\mathit{\gamma}_{L} \cdot d \cdot p + \mathit{\beta}\mathit{\gamma}_{Q} \cdot d \cdot p^{2} + \mathit{\beta}\mathit{\gamma}_{C} \cdot d \cdot p^{3}$$ where: Y = pull-out limit load; d = bone density; p = pitch μ = mean β = linear effect of bone density γ~L~ = linear effect of the pitch; γ~Q~ = square effect of the pitch γ~C~ = cube effect of the pitch βγ~L~= effect of interaction between the pitch and bone density βγ~Q~= effect of interaction between the bone density and the square effect of the pitch βγ~C~=effect of interaction between the bone density and the cube effect of the pitch Up to the third power of p could be considered because four different pitches were simulated; the only linear effect of bone density could be considered because only two density values have been simulated. Eq. (3) can be reformulated in order to isolate the contribution of the screw diameter: $$Y = \mathit{\mu} + d \cdot \left\lbrack {\mathit{\beta} + \mathit{\beta}\mathit{\gamma}_{L} \cdot p + \mathit{\beta}\mathit{\gamma}_{Q} \cdot p^{2} + \mathit{\beta}\mathit{\gamma}_{C} \cdot p^{3}} \right\rbrack + \mathit{\gamma}_{L} \cdot p + \mathit{\gamma}_{Q} \cdot p^{2} + \mathit{\gamma}_{C} \cdot p^{3} = \mathit{\mu} + d \cdot D + \left( {\mathit{\gamma}_{L} \cdot p + \mathit{\gamma}_{Q} \cdot p^{2} + \mathit{\gamma}_{C} \cdot p^{3}} \right)$$ or in order to isolate the contribution of the screw pitch: $$Y = \mathit{\mu} + \mathit{\beta} \cdot d + p \cdot \left\lbrack {\mathit{\gamma}_{L} + \mathit{\beta}\mathit{\gamma}_{L} \cdot d} \right\rbrack + p^{2} \cdot \left\lbrack {\mathit{\gamma}_{Q} + \mathit{\beta}\mathit{\gamma}_{Q} \cdot d \cdot p^{2}} \right\rbrack + p^{3} \cdot \left\lbrack {\mathit{\gamma}_{C} + \mathit{\beta}\mathit{\gamma}_{C} \cdot d \cdot p^{3}} \right\rbrack = \mathit{\mu} + \left( {\mathit{\beta} \cdot d} \right) + f\left( p \right)$$ The coefficients of this numerical model were calculated for every screw, solving a mathematical system with eight parameters (the effects) and eight equations (pull out forces, calculated from the FE models). 3.. RESULTS AND DISCUSSION ========================== 3.1.. Validation of FE Model ---------------------------- In order to validate FE models, it is necessary to select a failure criterion. The experimental tests showed that the polyurethane foam exhibited fragile behavior in the tensile tests, while compressive tests were much less critical. The maximum (positive) main stress failure criterion was chosen with an ultimate tensile stress of 0.42 MPa. As shown in Fig. (**[5](#F5){ref-type="fig"}**), the experimentally evaluated pull out force (120 N) generates ultimate tensile stresses at the last thread, close to the fillet. The FE model is therefore validated because it correctly predicts a localized failure when the pull out force is applied. Actually this validation is quite specific: it has been demostrated that FE model correctly predicts the failure load which is the object of this paper; a more detailed validation would be required if other information were sought from the finite element model (stiffness behavior, stress distribution, etc.). The location of ultimate tensile stress at the last thread may be unexpected, given that in the literature, the first thread is considered the most critical \[[@R21]\]. However, the pull out test is a static test, while the above considerations refer to fatigue life. Further, the mechanical behavior of threaded joints using polyurethane foam (or bone) differs from the behavior of metal joints: both foam and bone are porous, leading to completely different modalities of failure for tension and compression. The first "female" thread is subjected to compression stresses (contact between the screw and the bone), while the last is subjected to tensile stresses (as evidenced in Fig. **[5](#F5){ref-type="fig"}**). 3.2.. Multifactor Analysis -------------------------- The results of the multifactor analysis are illustrated in Fig. (**[6](#F6){ref-type="fig"}**), where the trends of the 'f(p)' function (eq. 5) *vs* pitch are plotted for different screw models, bone densities and boundary conditions (post-operative or complete osseointegration). Table **[3](#T3){ref-type="table"}** shows the values of the D coefficient (eq. 4) reported for different screw models. Some considerations can be made: In general, the differences in the stress distribution between pull out and the alternative constraint condition (with all other conditions constant) are minimal: the position of the full constraint does not influence the stress pattern on the "female" thread. In this sense, the pull out test is representative of a wide range of operating conditions.The post-operative and full osseointegration conditions are very different: the latter is more sensitive to bone density (in Table **[3](#T3){ref-type="table"}** the D coefficient values calculated for the post-operative condition are lower than that calculated for complete osseointegration) and to thread pitch (in Fig. (**[6](#F6){ref-type="fig"}**), curves B, D, F are steeper than curves A, C, E). This can be explained considering that, in full osseointegration, the screw and the bone are forced to undergo the same deformation. The post-operative condition is more critical because a lower pull out force is required. It is therefore advisable to construct non-linear numerical models where this situation is analyzed, implementing a 'touch' contact between the screw and bone.The coefficients of eq. 5 vary from model to model.The influence of pitch varies more for high values of bone density (Fig. **[6](#F6){ref-type="fig"}**). Effectively, greater bone stiffness implies a greater sensitivity to thread geometry because stresses are more localized.For a given screw, the optimal pitch (that providing maximum pull out force) is not always the maximum simulated pitch. For example, in the pull out test regarding the post-operative condition, given the screw model N. 8, the optimal pitch is between 3.4 and 4.4 mm for a bone density of 0.35 g/cm^3^, while the optimal pitch is between 2.4 mm and 3.4 mm for a bone density of 0.47 g/cm^3^.With regard to the interaction between the pitch and bone density, coefficients βγ~L~, βγ~Q~ and βγ~C~ can assume either positive and negative values (for example, βγ~L~\>0 in the screw model 1, while βγ~L~has a negative value in model 2). This means that in some models, the effect of bone density on pull out force is smaller for greater pitches. This result further confirms the conclusion that the optimal screw geometry will be different for different bone densities.Many screw models show a similar pattern of function p *vs* pitch (Fig. **[6](#F6){ref-type="fig"}**).There is a difference between load distribution in mechanical threaded joint and biomechanical threaded joint, as demonstrated by further numerical models, which implemented different values of "r", the parameter which defines the ratio between the Young's modulus of bone and that of the screw (Fig. **[7](#F7){ref-type="fig"}**). In the biomechanical case, r is very low because the Young's modulus of bone is approx. 300 MPa, while the Young's modulus of the screw is about one thousand times greater. Conversely, they have similar values in the mechanical and consequently a much more homogeneous load distribution is obtained on the second, third and fourth threads the biomechanical case.According to our results, screw model n. 10 is generally to be preferred; the performance of this model is moderately influenced by the thread pitch; considering the post-operative condition which is the most critical, a pitch larger than 1.4 mm should be recommended for the higher bone density. 4.. CONCLUSIONS =============== An FE model of the bone-screw system was constructed and experimentally validated. The model evidenced that the pull out test, the standard test for orthopedic screws, is representative of the operating conditions. It was demonstrated that the hypothesized level of osseointegration has a significant effect on the results obtained. In the light of this study, it is clear that the optimal screw geometry depends on the bone density: the pitch producing a higher load changes as a function of different bone densities (for example, in screw model 1 the optimal pitch is 3.75 mm for a bone density of 0.35 g/cm^3^, against 3.13 mm for 0.47 g/cm^3^, while for model 2 the optimal pitch is 3.87 mm for a bone density of 0.35 g/cm^3^ against 4.93 mm for a density of 0.47 g/cm^3^). It was found that threaded joints cut for biomechanical applications have some particularities compared to classic mechanical threaded joints: bone is much more compliant compared with the material of the screw and consequently a more homogenous load distribution over the threads can be obtained. Further, bone is porous and shows very different compression/tensile behaviors. {#F1} {#F2} {#F3} {#F4} {#F5} {#F6} {#F7} ###### Mechanical Properties of Trabecular Bone as a Function of Bone Density Density \[g/cm^3^\] Young's Modulus \[MPa\] Ultimate Stress \[MPa\] --------------------- ------------------------- ------------------------- 0.35 147 0.57 0.47 300 1.24 ###### Geometric Parameters of Screws Screw Model N d~e~ (mm) d~n~ (mm) R~T~ (mm) R~F~ (mm) α β E~screw~ (GPa) E~ext~ (GPa) L (mm) ------------- --- ----------- ----------- ----------- ----------- ---- --- ---------------- -------------- -------- 1 5 6 2 0.5 0.10 10 0 74 3.2 27 2 5 6 2 0.5 0.10 10 0 74 \- 27 3 5 4 2 0.5 0.10 10 0 105 \- 27 4 5 6 2 0.5 0.10 10 1 105 \- 27 5 5 6 2 0.5 0.10 10 2 105 \- 27 6 5 6 2 0.6 0.10 10 0 105 \- 27 7 5 6 2 0.5 0.10 10 0 105 \- 27 8 5 6 2 0.5 0.15 10 0 105 \- 27 9 5 6 2 0.5 0.10 10 0 186 \- 27 10 5 6 2 0.5 0.10 10 0 186 \- 33 ###### D\[10^-3^Nm^3^/kg\] from Eq. (5) Pull Out Test (Touch) ----------------------------------- --- ------- ------- ------- -------- ------- ------- ------- ------- ------- ------- 1.4 D 162.5 125.0 75.0 100.0 104.2 91.7 91.7 133.3 100.0 87.5 2.4 D 191.7 116.7 83.3 112.5 125.0 108.3 108.3 158.3 104.2 120.8 3.4 D 179.2 129.2 100.0 137.5 175.0 125.0 145.8 191.7 145.8 116.7 4.4 D 104.2 175.0 83.3 150.0 191.7 137.5 145.8 191.7 145.8 141.7 **PullOut Test (Glue)** 1.4 D 341.7 483.3 508.3 454.2 391.7 416.7 441.7 441.7 525.0 412.5 2.4 D 300.0 475.0 483.3 420.,8 458.3 470.8 508.3 537.5 533.3 504.2 3.4 D 300.0 454.2 500.0 495.8 458.3 470.8 512.5 529.2 554.2 462.5 4.4 D 408.3 583.3 566.7 625.0 575.0 633.3 616.7 575.0 716.7 579.2 **Alternative Condition (Touch)** 1.4 D 133.3 125.0 62.5 91.7 87.5 83.3 87.5 125.0 91.7 87.5 2.4 D 170.8 116.7 75.0 116.7 125.0 108.3 100.0 158.3 112.5 116.7 3.4 D 170.8 137.5 83.3 133.3 141.7 108.3 150.0 187.5 137.5 125.0 4.4 D 108.3 145.8 87.5 154.2 187.5 141.7 145.8 191.7 141.7 133.3 **Alternative Condition (Glue)** 1.4 D 362.5 537.5 525.0 470.8 354.2 450.0 470.8 395.8 520.8 416.7 2.4 D 312.5 408.3 508.3 533.3 479.2 404.2 537.5 558.3 541.7 445.8 3.4 D 312.5 466.7 483.3 529.2 466.7 466.7 566.7 500.0 562.5 487.5 4.4 D 420.8 620.8 570.8 720.8 575.0 641.7 683.3 512.5 691.7 591.7

Introduction {#s1} ============ Iron (Fe) deficiency is among the most prevalent micronutrient deficiencies in humans. Since plants constitute the primary source of nutrients for a large part of the world's population, the improvement of plants in terms of nutrient bioavailability is considered a priority [@pone.0099234-deBenoist1]. Micronutrients like Fe are often present in an un-soluble form in the soil. Plants are able to mobilize such nutrients for uptake into the roots. Plants can also mobilize Fe from internal stores. Understanding the regulation of Fe acquisition and internal Fe utilization is of high importance for precision breeding of crops that are improved to either tolerate growth on alkaline and calcareous soils with poor Fe bio-availability or to accumulate a higher content of this micronutrient in bio-available form in the edible plant parts. Genetic traits have been associated with micronutrient content and usage in plants, for example [@pone.0099234-Uauy1], [@pone.0099234-Baxter1]. Another trait was found in soybean as being linked to transcription factor genes encoding the soybean homologs of *BHLH38* and *BHLH39* [@pone.0099234-Peiffer1]. The potential importance of these two transcription factor genes for Fe mobilization had previously been uncovered in studies on the plant model *Arabidopsis thaliana*. *BHLH38* and *BHLH39* belong to the so-called subgroup Ib(2) *BHLH* genes [@pone.0099234-Pires1] and they are functionally redundant [@pone.0099234-Wang1]--[@pone.0099234-Wang2]. In fact, *BHLH38* and *BHLH39* are tandem duplicates on the chromosome, and they share similarity with two other *BHLH* genes, namely *BHLH100* and *BHLH101* [@pone.0099234-Wang1], [@pone.0099234-Heim1]--[@pone.0099234-ToledoOrtiz1]. All these four subgroup Ib(2) *BHLH* genes are highly induced by low Fe supply in roots and leaves while they are not usually found expressed under sufficient Fe supply [@pone.0099234-Wang1]. Expression of *BHLH39* and *BHLH101* in response to iron can be followed using the public microarray data in Arabidopsis [@pone.0099234-Bauer1]--[@pone.0099234-Yang1] and it was found that they occur in a co-expression network along with several Fe homeostasis genes like *FERRIC REDUCTASE OXIDASE3* (*FRO3*), *NATURAL RESISTANCE-ASSOCIATED MACROPHAGE PROTEIN4* (*NRAMP4*) and *NICOTIANAMINE SYNTHASE4* (*NAS4*) [@pone.0099234-Ivanov1]. From the co-expression with Fe homeostasis genes it can be concluded that the subgroup Ib(2) *BHLH* transcription factor genes likely perform regulatory functions in the context of Fe homeostasis and internal Fe mobilization. The bHLH protein POPEYE (PYE, belonging to another bHLH subgroup) is also induced by Fe deficiency within this co-expression network and it acts as a negative regulator of *FRO3*, *NRAMP4* and *NAS4,* presumably to avoid over-activation of Fe mobilization [@pone.0099234-Long1]. PYE is regulated by BRUTUS (BTS) that is also found in this co-expression network [@pone.0099234-Ivanov1], [@pone.0099234-Long1]. bHLH subgroup Ib(2) can physically interact with the bHLH FER-LIKE IRON DEFICIENCY-INDUCED TRANSCRIPTION FACTOR (FIT) [@pone.0099234-Wang2], [@pone.0099234-Yuan1]. FIT is expressed specifically in roots and has been shown to be essential for Fe uptake [@pone.0099234-Bauer2]--[@pone.0099234-Yuan2] by regulating the expression of the genes encoding ARABIDOPSIS H^+^-ATPASE2 (AHA2) [@pone.0099234-Ivanov1], Fe reductase FERRIC OXIDASE2 (FRO2) [@pone.0099234-Jakoby1], [@pone.0099234-Robinson1] and the IRON-REGULATED TRANSPORTER1 (IRT1) [@pone.0099234-Jakoby1], [@pone.0099234-Eide1]. From ectopic FIT expression experiments along with yeast promoter activation assays and inducible FIT activation in plants, it can be concluded that FIT targets *FRO2* and *IRT1* gene promoters [@pone.0099234-Jakoby1], [@pone.0099234-Yuan2], [@pone.0099234-Meiser1], [@pone.0099234-Sivitz2]. However, FIT induces *IRT1* and *FRO2* only upon Fe deficiency even when overexpressed [@pone.0099234-Jakoby1], [@pone.0099234-Meiser1]. The activation of FIT at low Fe can be explained with the presence of bHLH subgroup Ib(2) factors. Indeed, the double overexpression of FIT together with either bHLH subgroup Ib(2) protein leads to an increase of Fe acquisition responses under sufficient Fe supply conditions, and it was therefore proposed that the function of bHLH subgroup Ib(2) might be to induce Fe deficiency responses in conjunction with FIT [@pone.0099234-Wang2], [@pone.0099234-Yuan1]. However, the occurrence of *BHLH* subgroup Ib(2) genes in the *PYE* coexpression network, their non-expression upon sufficient Fe (where *FIT* and *IRT1* are active although at low level) and their high induction upon Fe deficiency not only in roots but also in leaves (in contrast to Fe acquisition genes) renders this hypothesis questionable. Moreover, contradictory results have been published with regard to the function of bHLH subgroup Ib(2) proteins. In one report, double *bhlh100 bhlh101* knockout mutants were demonstrated to develop a more severe leaf chlorosis than the wild type upon Fe deficiency, while no phenotype was apparent upon Fe sufficiency. Although some Fe homeostasis genes appeared mis-expressed, the gene knockouts did not affect the plants' abilities for Fe uptake and the regulation of *FRO2* and *IRT1* upon sufficient or deficient Fe supply [@pone.0099234-Sivitz1]. In contrast to that, in another report, bHLH subgroup Ib(2) knockouts including *bhlh100 bhlh101* and a triple knockout *bhlh39 bhlh100 bhlh101* were demonstrated to affect Fe acquisition responses and to have low *FRO2* and *IRT1* expression upon sufficient or deficient Fe supply [@pone.0099234-Wang3]. This latter finding was rather puzzling, and it was not further explained how this finding fits to the observation that the *BHLH* genes are not normally expressed upon sufficient Fe supply, when Fe also needs to be acquired via FRO2 and IRT1 [@pone.0099234-Jakoby1], [@pone.0099234-Vert1]. Thus, the function of the bHLH subgroup Ib(2) transcription factors in Fe uptake is still open for debate. Very interestingly, it has been shown that *BHLH38* and *BHLH39* were induced after application of salicylic acid ( = SA) by the SA-inducible Dof ( = DNA binding with one finger) transcription factor OBF BINDING PROTEIN3 (OBP3) [@pone.0099234-Kang1]. Binding of OBP3 to promoter elements in *BHLH38* and *BHLH39* genes and their subsequent activation was demonstrated (in these studies *BHLH38* and *BHLH39* were named *OBP3 RESPONSIVE GENE2*, *ORG2*, and *OBP3 RESPONSIVE GENE3*, *ORG3*) [@pone.0099234-Kang1]. Jasmonic acid negatively affects the onset of Fe mobilization and the induction of *FRO2* and *IRT1* [@pone.0099234-Maurer1], while ethylene enhances the responses [@pone.0099234-Garca1]--[@pone.0099234-Lingam1]. Since SA, jasmonic acid and ethylene act in stress response networks, the possibility exists that perhaps, there is a link between SA and the up-regulation of Fe deficiency responses. Here, we made use of the triple knockout mutant *bhlh39 bhlh100 bhlh101* (*3xbhlh*) that we constructed to investigate the functions of these *BHLH* genes in the Fe deficiency response and to further shed light on the question whether SA is involved in mediating the onset of Fe uptake via the induction of *BHLH* subgroup Ib(2) genes. We discuss that *BHLH39*, *BHLH100* and *BHLH101* are essential for a subset of Fe deficiency responses but not including up-regulation of *IRT1* and *FRO2*. We suggest that these transcription factors are involved in adapting stress responses and internal metabolic responses to Fe deficiency. Materials and Methods {#s2} ===================== Plant Materials {#s2a} --------------- Wild type was Col-0. The *3xbhlh* mutant line was generated from the single T-DNA insertion mutants *bhlh39-1* (SALK_025676), *bhlh100-1* (SALK_074568) and *bhlh101-1* (SALK_011245) [@pone.0099234-Wang1]. A homozygous *bhlh39-1* plant was crossed with a *bhlh100-1 bhlh10-1* double mutant plant. In the F2 progeny a triple homozygous *bhlh39-1 bhlh100-1 bhlh101-1* plant was identified by genotyping and multiplied to obtain a triple homozygous line, hereafter named *3xbhlh*. The SA mutant line *npr1-1* (hereafter named *npr1*) with a defect in a central regulator component of SA signaling resulting in the failure of the expression of the *PR1* gene was obtained from the NASC stock center (N3726) [@pone.0099234-Cao1]. The lines *NaHG* and *sid2-2* were provided by Fred Ausubel, Massachusetts General Hospital [@pone.0099234-Ferrari1]. Plant Growth {#s2b} ------------ Arabidopsis seeds were surface-sterilized with 6% NaOCl, 0.1% Triton-X for 10 minutes, and washed 5 times with distilled water. Seeds were stratified for 2 days in 0.1% plant agar in the dark at 4°C. For the 6-day growth assay seeds were placed on Hoagland agar medium containing 50 µM FeNaEDTA (sufficient Fe supply, hereafter termed +Fe) or 0 µM FeNaEDTA (deficient Fe supply, hereafter termed −Fe), germinated and grown for 6 d under long-day conditions with 8 h dark and 16 h light [@pone.0099234-Lingam1]. On day 6, seedlings were harvested for analysis. For the 2-week growth assay, seeds were germinated and seedlings grown for 14 d on Hoagland agar medium as described above containing 50 µM FeNaEDTA, then transferred for 3 days to fresh medium containing either 0 µM FeNaEDTA (−Fe) or 50 µM FeNaEDTA (+Fe). Then, leaves and roots were harvested separately for RNA or protein analysis. If indicated in the text 100 µM methyl-salicylic acid (hereafter named SA, Sigma-Aldrich, USA) was added to the growth medium and plants exposed for the indicated time. Physiological Analysis {#s2c} ---------------------- The degree of leaf chlorosis was assessed according to a previously published procedure [@pone.0099234-Schuler2]. The leaf chlorosis scale is mentionned in the figure legend of [Fig. 1](#pone-0099234-g001){ref-type="fig"}. {#pone-0099234-g001} For metal determination, plant parts exposed to plant medium were washed with 100 mM Ca(NO~3~)~2~ prior to harvest to eliminate metal residues from the growth medium. Plant material was dried overnight at room temperature, then for 1 d at 120°C and powdered with an achat mortar. Quantification of metal contents of the plant samples was performed using atomic absorption spectroscopy coupled with a graphite tube atomizer as described [@pone.0099234-Lingam1]. Four technical replicate measurements were carried out with weighted samples of 50--120 µg for every atomisation (2300°C) and mean mass per dry weight values were calculated for each biological sample. Four biological replicates were produced and mean values calculated. Iron reductase activity assays were performed using a liquid ferrozine assay [@pone.0099234-Klatte1] and biological replicates were performed as described in the figure legends. Statistical analysis was performed using the t-test. Gene Expression using Reverse Transcription-qPCR {#s2d} ------------------------------------------------ Reverse transcription-quantitative real-time PCR was performed as previously described [@pone.0099234-Klatte2]. Briefly, DNase-treated RNA was used for cDNA synthesis. SYBR green I-based real-time PCR analysis was performed by using the TaKaRa Premix (TaKaRa, Japan) in the real-time ICycler (Bio-Rad, USA). For each gene, the absolute quantity of initial transcript was determined by standard curve analysis using mass standards. Absolute expression data was normalized against the averaged expression values of the internal control gene *EF1BALPHA2* [@pone.0099234-Klatte2]. Each biological cDNA sample was tested in two technical replicates and the values averaged. Statistical analysis was performed by ANOVA using the values of biological replicates. Information on oligonucleotide primer sequences is available in [Table S1](#pone.0099234.s005){ref-type="supplementary-material"}. Gene Expression using Microarray Analysis {#s2e} ----------------------------------------- Wild type and *3xbhlh* seedlings were grown in the 6-day growth system at --Fe, treated for six hours with 100 µM SA or were mock-treated and harvested for RNA preparation. Three biological replicates were generated. RNA was purified using the Qiagen kit and checked for integrity. Microarray hybridization was performed using the Agilent one-color gene expression V4 chip (4×44 k) for *Arabidopsis thaliana*. Microarray chip hybridization and processing were done by ATLAS Biolabs GmbH, Berlin, Germany. The obtained data were further processed, checked for quality and filtered using the GeneSpring Software, Agilent Technologies, USA, according to the GeneSpring protocol. Full microarray data are available from the NCBI site of Gene Expression Omnibus under the series GSE41774. Interesting probes were identified based on fold change analysis with a fold change cut-off of 1.5 in four pairs of conditions which were *3xbhlh* versus wild type, *3xbhlh* + SA versus wild type + SA, wild type + SA versus wild type and *3xbhlh* + SA versus *3xbhlh*. Probes were retained if they passed the moderated t-test with p\<0.05. The differentially regulated probes of the four pairs of conditions were then used to construct Venn diagrams to identify groups with unique and commonly regulated probes. Probe names of these groups were converted into Arabidopsis gene ID numbers. The groups of differentially expressed genes were then further analyzed using Venn diagrams, the ATTED co-expression tool [@pone.0099234-Obayashi1], the GOrilla GO annotation tool [@pone.0099234-Eden1] and the Genevestigator tool [@pone.0099234-Zimmermann1]. Results {#s3} ======= The *3xbhlh* Mutant was Sensitive to Fe Deficiency but not Affected in Fe Acquisition and Fe Transport to Shoots {#s3a} ---------------------------------------------------------------------------------------------------------------- To analyze the functions of *BHLH* subgroup Ib(2) genes, we generated a multiple loss of function mutant. The triple homozygous *bhlh39-1 bhlh100-1 bhlh101-1* knockout mutant, hereafter named *3xbhlh*, was fully fertile and did not express any full-length *BHLH39*, *BHLH100* and *BHLH101* transcripts while it expressed *BHLH38* at a higher level at Fe deficiency (hereafter termed −Fe) in roots and leaves. *BHLH38* was also found induced in the *3xbhlh* mutant roots at sufficient Fe (hereafter termed +Fe) compared to wild-type roots ([Fig. S1A, B](#pone.0099234.s001){ref-type="supplementary-material"}). In the same experiment, all four *BHLH* genes were highly expressed at --Fe in wild type plants, while they were not expressed or expressed at low level at +Fe ([Fig. S1A, B](#pone.0099234.s001){ref-type="supplementary-material"}), as expected [@pone.0099234-Wang1]. The increased *BHLH38* expression in the *3xbhlh* mutant especially at --Fe might be due to a feedback control conferred by the triple loss of function phenotype. To determine whether the triple *3xbhlh* mutant had any Fe deficiency phenotype at + or --Fe, we grew *3xbhlh* and wild type plants at +Fe and --Fe. Morphological alterations in root and shoot growth were not apparent at +Fe. However, at --Fe, the *3xbhlh* mutant plants had a stronger leaf chlorosis than wild type, while root growth was normal ([Fig. 1A, B](#pone-0099234-g001){ref-type="fig"}). This observation suggested that *3xbhlh* mutants might be perhaps Fe deficient. This assumption was tested and could be rejected after the determination of shoot Fe contents. Neither under + nor --Fe, we could detect any differences in Fe content in the mutant versus the wild type ([Fig. 1C](#pone-0099234-g001){ref-type="fig"}), suggesting that the mutant was not Fe deficient. After retransfer of *3xbhlh* mutants from --Fe to +Fe the leaf chlorosis phenotype disappeared within 1--2 days (data not shown). This observation confirmed that indeed the triple *3xbhlh* mutant was able to acquire Fe. We verified this point by analyzing gene expression of Fe acquisition genes. We observed that *IRT1*, *FRO2* and *FIT* were significantly up-regulated at --Fe in wild type and *3xbhlh* plants, and that no significant difference in the expression levels between wild type and mutant was detected ([Fig. 2A](#pone-0099234-g002){ref-type="fig"}). A reduced expression of these Fe acquisition genes is typical for mutants affected in the regulation of Fe deficiency responses, such as the chlorotic *fit* mutant [@pone.0099234-Jakoby1]. Fe reductase activity was also detected at comparable levels in mutant and wild type ([Fig. 2B](#pone-0099234-g002){ref-type="fig"}). An increase of Fe reductase activity was noted at --Fe but due to high standard deviations it was not found significant, but clearly was not lower than in the wild type ([Fig. 2B](#pone-0099234-g002){ref-type="fig"}; [@pone.0099234-Jakoby1]). {#pone-0099234-g002} In addition, we tested whether two of the PYE-regulated Fe homeostasis genes of the *PYE*/*BHLH39*/*BHLH101* co-expression network that function in internal Fe mobilization were affected in the mutant. No differences in gene expression were noted for *NRAMP4* and *FRO3* in roots and leaves ([Fig. 2C, D](#pone-0099234-g002){ref-type="fig"}). Taken together, *3xbhlh* mutant plants were fully capable of regulating internal and external Fe mobilization, Fe acquisition and Fe transport genes despite of the lack of the three transcription factors. The *3xbhlh* mutant was also able to mobilize Fe from roots into the shoots. Thus, the strong leaf chlorosis of the triple mutant at --Fe cannot be the consequence of a defect in Fe acquisition and mobilization responses. SA and SA Signaling do not Affect Fe Deficiency Responses {#s3b} --------------------------------------------------------- Since the *BHLH39*, *BHLH100* and *BHLH101* genes were dispensable for Fe uptake, but yet the plants showed a chlorosis phenotype at --Fe, we reasoned that the chlorosis phenotype could be the result of an altered adaptation to Fe deficiency stress. Since previous reports established a connection between the *BHLH* subgroup Ib(2) genes and SA signaling [@pone.0099234-Kang1] we hypothesized that perhaps SA responses interfered with Fe deficiency regulation via the bHLH subgroup Ib(2) proteins in the process of adaptation to --Fe. To test this hypothesis, we analyzed available gene expression data of *OBP3* (At3g55370), the regulator of *BHLH38* and *BHLH39* [@pone.0099234-Kang1]. *OBP3* is not in a co-expression network with any known Fe-regulated metal homeostasis genes. However, its expression was reported to occur in the root stele where the subgroup Ib(2) gene promoters are also active [@pone.0099234-Wang1], [@pone.0099234-Brady1]. SA plays a role throughout plant development and hence this could require an adaptation to Fe homeostasis [@pone.0099234-RivasSanVicente1]. To test a possible interference of SA and Fe signaling, we first tested using the seedling growth assay whether the SA response gene *PATHOGENESIS RELATED1* (*PR1*) as a marker for SA responses [@pone.0099234-Cao1], *OBP3* and *BHLH38* were regulated by SA and Fe deficiency treatments in wild type and *3xbhlh* mutant plants. As expected, *PR1* was induced by 100 µM SA in the wild type, while *BHLH38* was induced by --Fe ([Fig. 3A](#pone-0099234-g003){ref-type="fig"}). These two marker genes were not per se induced by the respective other treatment, and in two out of three experiments, the expression levels were not affected in the *3xbhlh* mutant ([Fig. 3A](#pone-0099234-g003){ref-type="fig"}). *OBP3* was hardly induced by SA treatment and did not show any regulation by Fe or in response to the *3xbhlh* mutant. We did also not detect any differences in gene expression levels of *IRT1*, *FRO2* and *FIT* at + versus --SA treatment ([Fig. 3B](#pone-0099234-g003){ref-type="fig"}). Therefore, we conclude that a clear SA response did not take place upon --Fe and that the *3xbhlh* mutant did not show an altered SA response upon SA application. {#pone-0099234-g003} Since it is possible, that SA might have a more subtle effect in the regulation of Fe deficiency responses, we also tested Fe deficiency gene regulation in various SA mutants grown under + and −Fe supply. The *non-repressor of pr1* (*npr1*) mutant is defective in SA signaling [@pone.0099234-Cao1]. *sid2* mutant plants are defective in SA production due to lack of isochorismate synthesis, and *NahG* plants overexpress a bacterial SA hydroxylase so that SA is rapidly transformed into catechol and consequently does not accumulate [@pone.0099234-Ferrari1]. We found that in these mutants Fe acquisition genes *IRT1*, *FRO2* and *FIT* were up-regulated by --Fe as in the wild type ([Fig. 4A](#pone-0099234-g004){ref-type="fig"}). *IRT1* was slightly up-regulated at +Fe in *npr1* compared to the wild type in one experiment. *BHLH38* and *BHLH39* expression was not affected in the mutants ([Fig. 4B](#pone-0099234-g004){ref-type="fig"}). {#pone-0099234-g004} Hence, we exclude an apparent effect of SA treatment or SA signaling on Fe deficiency regulation. The up-regulation of *BHLH* subgroup Ib(2) genes under --Fe was not likely the consequence of a SA signal. Wild Type and *3xbhlh* Mutants Differ in Gene Expression Patterns at --Fe and in the Presence of SA {#s3c} --------------------------------------------------------------------------------------------------- To get further hints on the Fe deficiency phenotype we performed a transcriptome comparison between *3xbhlh* and wild type seedlings ([Fig. S2](#pone.0099234.s002){ref-type="supplementary-material"}). This analysis was conducted under --Fe conditions since we expected most of the differential gene expression to occur at --Fe as deduced from the *3xbhlh* leaf chlorosis phenotype at --Fe. SA was included as a treatment to search for changes in gene expression in response to SA. 6-day old seedlings were grown on --Fe, transferred for 6 h to + or −100 µM SA containing --Fe medium and harvested. Microarray analysis was performed using the Agilent V4 gene chip (26.283 Arabidopsis genes). In parallel, we used the same RNA samples to perform real time quantitative reverse transcription-PCR to assess the technical quality of the method and the biological quality of the samples in this experiment ([Fig. S2](#pone.0099234.s002){ref-type="supplementary-material"}). At first, we identified from the microarray data those genes that showed differential regulation at least 1.5-fold between wild type and the *3xbhlh* mutant in the absence and presence of SA in a statistically significant manner. The differentially regulated genes were grouped according to expression patterns using Venn diagrams ([Fig. 5](#pone-0099234-g005){ref-type="fig"}). 198 genes were identified as being regulated between mutant and wild type ([Table S2](#pone.0099234.s006){ref-type="supplementary-material"}; [Fig. 5](#pone-0099234-g005){ref-type="fig"}). 25 of these genes were only found deregulated in the mutant in the absence of SA (group I) and 61 genes only in the presence of SA (group II). 112 genes were regulated between mutant and wild type in the presence and absence of SA (group III). An important control for the quality of our microarray hybridization experiment was represented by the *BHLH39*, *BHLH100* and *BHLH101* genes which we found down-regulated in the *3xbhlh* mutant compared to the wild type in the group III, as was expected ([Table S2](#pone.0099234.s006){ref-type="supplementary-material"}). On the other hand, 9892 genes were differentially regulated between + and − SA in the wild type or in the *3xbhlh* mutant ([Fig. 5](#pone-0099234-g005){ref-type="fig"}). Among these latter genes 1718 were only differentially expressed in response to SA in the wild type (group IV), while 1472 other genes were specifically affected by SA in the mutant (group V) ([Table S3](#pone.0099234.s007){ref-type="supplementary-material"}; [Fig. 5](#pone-0099234-g005){ref-type="fig"}). 6702 genes were regulated by SA in wild type and in the *3xbhlh* mutant (group VI) and were not further investigated as they reflected purely SA-dependent genes. ([Fig. 5](#pone-0099234-g005){ref-type="fig"}). The high numbers of SA-regulated genes might indicate that the SA treatment affected the transcriptomes in a strong manner than did the *3xbhlh* mutations. However, we found that genes of groups IV and V were only mis-regulated by a maximum level of 5-fold ([Table S3](#pone.0099234.s007){ref-type="supplementary-material"}). On the other hand, among the small number of genes mis-regulated between mutant and wild type (groups I--III) some genes reached differential expression up to 60-fold ([Table S2](#pone.0099234.s006){ref-type="supplementary-material"}). {ref-type="supplementary-material"}, [S4](#pone.0099234.s008){ref-type="supplementary-material"}. Groups I to III contain genes differentially expressed between *3xbhlh* and wild type. Groups IV and V contain genes that show differential regulation between + and --SA treatment but not between wild type and mutant.](pone.0099234.g005){#pone-0099234-g005} Analysis of Functional Categories Differentially Regulated between *3xbhlh* and Wild Type Plants (Groups I to III) {#s3d} ------------------------------------------------------------------------------------------------------------------ Next, we assessed whether any specific functional pathways were affected in the *3xbhlh* mutant. At first, we analyzed the functions of the genes of groups I, II and III by analyzing whether specific gene ontology categories were hit among them using the GOrilla tool [@pone.0099234-Eden1]. Among the DOWN-regulated genes of group I we found weak enrichment of categories related to lactate metabolism and cell wall, and among the UP-regulated genes of group I of categories RNA metabolism and transcription ([Table S4](#pone.0099234.s008){ref-type="supplementary-material"}-1, S4-2). In group II we found an enrichment for the categories flavonoid biosynthesis, UV responses, inositol metabolism and transpiration among the UP-regulated genes ([Table S4-1](#pone.0099234.s008){ref-type="supplementary-material"}, [S4-2](#pone.0099234.s008){ref-type="supplementary-material"}). In group III the categories metal response and copper binding were among the DOWN-regulated genes and again inositol metabolism among the UP-regulated genes ([Table S4-1, 4-2](#pone.0099234.s008){ref-type="supplementary-material"}). This analysis indicates that loss of the BHLH subgroup Ib(2) functions in the *3xbhlh* mutant at --Fe resulted in altered stress regulation and adaptation to stress. Then, we checked whether the three groups I--III contained known Fe-regulated genes. In a previous work, we have grown six day-old Arabidopsis seedlings in + and −Fe conditions in the same system as utilized here. From these previous experiments we have obtained a list of iron-regulated genes in the seedlings [@pone.0099234-Bauer1], [@pone.0099234-Schuler1], [@pone.0099234-Ivanov1]. The list of Fe-regulated genes in wild type seedlings published in [@pone.0099234-Bauer1] was used to compare with the list of genes in groups I--III. To our surprise we could only find 29 Fe-regulated genes among the 197 genes of the groups I--III, which corresponded to only 15% of these genes ([Fig. 6A](#pone-0099234-g006){ref-type="fig"}). Interestingly, these 15% of the *3xbhlh*-regulated genes could be further subdivided into 9 genes that were up-regulated by --Fe in the wild type and up-regulated in the mutant versus the wild type at --Fe as well as 15 genes that were down-regulated by --Fe in the wild type and down-regulated in the mutant versus the wild type at --Fe. Hence, the expression patterns of these 24 genes reflected the situation that the *3xbhlh* mutant was more sensitive to Fe deficiency than the wild type and that the Fe deficiency leaf chlorosis response was enhanced. Very interestingly, five genes showed an opposite expression pattern of being highly expressed at --Fe in the wild type, but expressed at low level in the *3xbhlh* mutant versus the wild type. These five genes were At1g53310 encoding PPC1, At3g07720, encoding a putative Kelch-repeat protein, At3g12900, an oxidoreductase gene, At4g31940 encoding CYP82C4 and At3g58810 encoding the metal transporter MTPA2. Therefore, it can be assumed that these five genes could represent specific target genes of the transcription factors bHLH39, bHLH100 and bHLH101. We also noted that the genes encoding the Fe acquisition machinery in Arabidopsis like *IRT1* and *FRO2* were not found among the differentially expressed genes which confirms above expression studies. We used the 29 genes to build co-expression networks using the ATTED tool [@pone.0099234-Obayashi1], and in total we could identify four co-expression networks ([Fig. 6B](#pone-0099234-g006){ref-type="fig"}, [Fig. S3](#pone.0099234.s003){ref-type="supplementary-material"}). One network contained four out of the five putative bHLH targets. As mentioned above, this co-expression network was previously identified as the FIT target network [@pone.0099234-Ivanov1]. Another network contained six genes up-regulated at --Fe and in the chlorotic triple mutant and one gene down-regulated in these respective conditions. This network was enriched with gene functions in flavonoid synthesis, secondary metabolism and circadian rhythm. A third network contained seven genes being down-regulated by --Fe and in the triple mutant, and this network was enriched for circadian rhythm functions. The fourth co-expression network comprised another five genes that were down-regulated at --Fe and in the *3xbhlh* mutant, and enrichment was found for photosynthetic functions and secondary metabolism. Similar functional categories were also evident for the 29 genes when using GOrilla as a GO annotation tool, namely metal response, inositol metabolism, protein folding, photosystem II and other chloroplast functions and circadian clock ([Table S4-3, S4-4](#pone.0099234.s008){ref-type="supplementary-material"}). In summary, co-expression analysis, functional annotation and enrichment analysis indicated that the expression of the 24 Fe-regulated genes was different in *3xbhlh* as a consequence of the increased leaf chlorosis of the triple mutant at --Fe, while the expression of five co-expressed genes indicates that they might be targets of the transcription factors in roots. ![Regulation of the subset of 29 Fe-regulated genes out of groups I, II and III identified in microarray analysis.\ The list of Fe-regulated genes in wild type seedlings that we had published earlier [@pone.0099234-Bauer1] was used to compare with the list obtained in this work for the groups I--III. 29 genes of the groups I--III were found Fe-regulated in [@pone.0099234-Bauer1]. A, Regulation patterns, annotation and co-expression of the subset of 29 genes; the regulation at +Fe versus --Fe in the wild type is represented in the left-most column (in dark green up = up-regulated, in red down = down-regulated); the regulation in the *3xbhlh* mutant versus WT is shown in the middle column (in light green up = up-regulated and in yellow or violet down = down-regulated; note that the yellow color indicates that these genes do not follow in the *3xbhlh* mutant the regulation expected from --Fe versus +Fe in the left column and hence could be direct targets of bHLH39, bHLH100 and bHLH101). The Arabidopsis gene identification (AGI) numbers and annotations are shown on the right side, whereby the color code indicates the belonging to different co-expression networks as determined using the ATTED tool (ref), represented in B; B, Co-expression network analysis of the 29 genes; the ATTED tool was utilized for construction; the different networks are highlighted in color and the AGI numbers belonging to those networks are highlighted by the same color in A. The grey color indicates genes that are part of isolated co-expression networks. A high-resolution image of the co-expression networks is shown in [Fig. S3](#pone.0099234.s003){ref-type="supplementary-material"}.](pone.0099234.g006){#pone-0099234-g006} On the other hand, 85% of the genes of group I--III were not regulated by Fe deficiency. It was then interesting to determine whether these genes represented specific pathways. The functional categories were again related to inositol metabolism, circadian rhythm and UV response ([Table S4-3, S4-4](#pone.0099234.s008){ref-type="supplementary-material"}). Hence, the Fe-regulated and non-Fe-regulated genes of groups I to III play functions in regulating adaptive stress processes suggesting that these functions are central for the bHLH proteins. Validation of Differential Gene Expression in Groups I to III by Reverse Transcription-qPCR {#s3e} ------------------------------------------------------------------------------------------- To verify the regulatory expression patterns identified in the microarray experiments, we verified some of the gene expression results by reverse transcription-qPCR studies, especially by studying the Fe-regulated genes. Reverse transcription-qPCR was performed on the same biological samples as used for the microarray ([Fig. S2](#pone.0099234.s002){ref-type="supplementary-material"}). We also used the samples derived from plants grown at +Fe that we had raised in parallel. Three biological replicates have been analyzed. Due to the low number of three samples the differences were not found to be significant with p\<0.05 in all expected cases, which we designated then as "tendency". We found a tendency for up-regulation of *IRT1*, *FRO2* and *BHLH038* at --Fe versus +Fe irrespective of genotype and SA treatment which was significant for *IRT1* at −Fe versus +Fe in the *3xbhlh* mutant ([Fig. S4A](#pone.0099234.s004){ref-type="supplementary-material"}, B, C). *PR1* expression was significantly increased in all +SA versus --SA samples, as expected ([Fig. S4D](#pone.0099234.s004){ref-type="supplementary-material"}). Hence, these control gene expression results confirm that the plants had reacted as expected to Fe and salicylic acid supply in the experiment. Then, we studied gene expression of the Fe-regulated targets of the bHLH factors from groups I--III, namely of the four coexpressed genes At3g07720, *CYP82C4*, At3g12900 and *MTPA2* as well as of *PPC1*. All five genes showed a tendency to be up-regulated by --Fe in the wild type and in the *3xbhlh* mutant compared to +Fe, which was significant with p\<0.05 for *CYP82C4* in *3xbhlh* and for At3g12900 ([Fig. 7A--E](#pone-0099234-g007){ref-type="fig"}; compare to [Fig. 6](#pone-0099234-g006){ref-type="fig"}, yellow co-expression network and *PPC1*). The expression of these five genes had a tendency to be lower in the mutant than in the wild type at --Fe. In the presence of SA the five genes were similarly regulated than in the absence of SA, whereby the basal expression levels at +Fe in the wild type were found higher in tendency at +SA than at −SA for At3g07720 and *MTPA2*. {ref-type="fig"}). *3xbhlh* and wild type seedlings were grown for 6 d at + and −Fe and exposed for 6 h to 100 µM SA (+SA) or were mock-treated (−SA). Whole seedlings were harvested for analysis. n = 3; the --Fe cDNA samples were derived from the RNAs used in the microarray ([Fig. S2](#pone.0099234.s002){ref-type="supplementary-material"}); \* indicates a significant change (p\<0.05) of −Fe versus +Fe; + indicates a significant change (p\<0.05) of *3xbhlh* versus WT; § indicates a significant change (p\<0.05) of +SA versus --SA. Gene expression was studied using reverse transcription-qPCR.](pone.0099234.g007){#pone-0099234-g007} Next, we verified gene expression of the Fe-regulated group III genes, namely *LHY1*, *PSAF* and At1g07050 ([Fig. 7F--H](#pone-0099234-g007){ref-type="fig"}; compare to [Fig. 6](#pone-0099234-g006){ref-type="fig"}, green, violet and pink coexpression networks, respectively). We found that *LHY1* followed a tendency to be more expressed at --Fe versus +Fe in the mutant and in the presence of SA. At1g07050 and *PSAF* followed a tendency to be expressed at lower level at --Fe versus +Fe, especially in the *3xbhlh* mutant. Taken together, the reverse transcription-qPCR data confirmed in their tendency the gene regulation changes detected in the comparative transcriptome analysis. This result underlined the technical and biological reproducibility, however the number of only three biological replicates did not allow obtention of p\<0.05 statistical values in each comparison. Analysis of Functional Categories Differentially Regulated between + and −SA (Groups IV and V) {#s3f} ---------------------------------------------------------------------------------------------- Different genes were regulated by SA in the wild type (group IV) and *3xbhlh* plants (group V). We used again the GOrilla tool [@pone.0099234-Eden1] to analyze whether specific categories were hit in the two cases. In group IV we identified the categories RNA and post-transcriptional genes silencing, organelle organization, GTP activity and chloroplast functions among the DOWN-regulated genes in the wild type, while the categories lipid metabolism, cellular starvation response and transition metal transport were hit among the UP-regulated genes ([Table S4-1, 4-2](#pone.0099234.s008){ref-type="supplementary-material"}). In group V the categories DNA replication, transmembrane receptor signaling, cell division, histone modification, protein binding and chloroplast thylakoid were hit among the DOWN-regulated genes and defense and heat response, signaling, kinase activity and ADP binding were identified as pathways among the UP-regulated genes ([Table S4-1, 4-2](#pone.0099234.s008){ref-type="supplementary-material"}). Hence from the high number of regulated genes and the categories we deduce that SA affects stress responses and cell division in a different manner in the wild type and the *3xbhlh* mutant upon SA treatment. Thus, the *3xbhlh* mutant is in a different manner sensitive to SA than the wild type. Discussion {#s4} ========== In this report we show that *BHLH* subgroup Ib(2) genes *BHLH39*, *BHLH100* and *BHLH101* acted in the adaptation to the stress caused by the Fe deficiency but not for Fe acquisition itself. Although *BHLH38* and *BHLH39* may act in the SA pathway we did not find any evidence that SA responses interfered with Fe deficiency regulation via the *BHLH* genes. Contradictory reports in the literature rendered it difficult to establish a clear function for *BHLH* subgroup Ib(2) genes in the regulation of Fe deficiency responses in Arabidopsis. On one side, it was shown that a triple mutant *bhlh39 bhlh100 bhlh101* was affected in the ability to up-regulate Fe reductase activity upon --Fe [@pone.0099234-Wang3], while a double mutant *bhlh100 bhlh101* was not [@pone.0099234-Sivitz1]. However, in both these studies the mutants were described to display a leaf chlorosis at --Fe but not at +Fe. Here we demonstrate that the *3xbhlh* triple mutant *bhlh39 bhlh100 bhlh101* developed a leaf chlorosis only at --Fe but not at +Fe. We found and confirmed that this phenotype was clearly not associated with a reduced uptake of Fe. When we analyzed the triple *3xbhlh* mutant we observed that this mutant was able to mobilize and acquire Fe, in contrast to the report mentioned above [@pone.0099234-Wang3]. The first evidence came from the fact that the leaf chlorosis was restrained to --Fe, while no such phenotype was observed at +Fe. The *3xbhlh* mutants had similar Fe levels as the wild type, and hence, these plants were able to utilize Fe at +Fe like the wild type. In addition, the *3xbhlh* mutants were able to induce *FIT*, *IRT1* and *FRO2* at --Fe versus +Fe. The three Fe uptake genes did also not occur among the groups I--III that comprised the genes differentially expressed between mutant and wild type. Fe reductase activity was clearly inducible in the triple mutant at least to the same extent as in the wild type. An alternative possibility to explain the increased leaf chlorosis of *3xbhlh* mutant compared to wild type plants at --Fe could be linked with a reduced ability of the mutant to utilise internal iron. However, we did not find evidence that typical genes for internal Fe utilization were changed in their expression, like *NRAMP4* [@pone.0099234-Lanquar1], *FRO3* [@pone.0099234-Mukherjee1], *OPT3* [@pone.0099234-Stacey1] and *NAS4* [@pone.0099234-Koen1], [@pone.0099234-Palmer1]. Taken together, we can say that the leaf chlorosis of *3xbhlh* plants was not the consequence of impaired Fe uptake and utilisation. Since *BHLH38* was still expressed in the *3xbhlh* mutant, even to a higher level than in the wild type, we cannot conclude that none of the subgroup Ib(2) *BHLH* genes are required for Fe uptake or Fe mislocalization. Since *BHLH38* is hardly expressed at +Fe (when Fe is available for uptake), it is not likely required for Fe uptake, either. Due to the tandem location of *BHLH38* and *BHLH39* in the genome, a quadruple insertion mutant with a full knockout of all *BHLH* functions cannot be readily generated to confirm this. We predict that a quadruple mutant would have more severe leaf chlorosis symptoms than the wild type at --Fe. The question remains what is the cause of the increased leaf chlorosis of *3xbhlh* mutants at --Fe. One clue to that question could come from the functions of the genes that we found differentially expressed in the mutant versus the wild type. Five genes were regulated in an opposite manner in the *3xbhlh* mutant than in the wild type in response to iron. These genes might be targets for bHLH39, bHLH100 and bHLH101, namely *PPC1*, At3g07720, the oxidoreductase At3g12900, *CYP82C4* and *MTPA2*. The latter four are coexpressed with target genes of the transcription factor FIT, such as *IRT1*. However, some of the FIT targets of the co-expression network were found here not be mis-regulated in the *3xbhlh* mutant. This result can be explained. We suggest that the four bHLH transcription factors of the subgroup Ib(2) show functional divergence. bHLH38 in conjunction with FIT namely acts on the induction of the targets *IRT1* and *FRO2*, while bHLH39, bHLH100 and bHLH101 presumably together with FIT more specifically target the other genes in this co-expression network that we have found here. Another set of Fe-regulated genes of groups I--III, which are differentially expressed in the mutant versus the wild type, indicated that the Fe deficiency responses were augmented in the mutant. For example, genes which are up-regulated at --Fe versus +Fe in the wild type showed an exaggerated up-regulation response in the comparison of mutant versus wild type at --Fe. Or, genes which are down-regulated at --Fe versus +Fe in the wild type showed a stronger down-regulation response in the mutant versus wild type at --Fe. From these expression patterns it is unlikely that these genes are regulated directly by the bHLH subgroup Ib proteins since otherwise we would have expected opposite trends of regulation between +/−Fe and mutant/wild type. The differential regulation of these genes is rather a pleiotropic effect due to the *3xbHLH* mutations. Since the most obvious phenotype was the increased leaf chlorosis, the differential regulation of these genes was likely an effect of the severe leaf chlorosis. Indeed, several of these genes have functions in the leaves related to the clock and in chloroplasts. Previous reports have established an effect of iron deficiency on the biological clock [@pone.0099234-Chen1]--[@pone.0099234-Salome1]. On the other hand, 170 genes were differentially expressed between mutant and wild type without that any differential expression of these genes had been noted in response to Fe supply in previous experiments [@pone.0099234-Bauer1]. Therefore, we deduce that the three *BHLH* genes regulate an adaptation response to the stress caused by the Fe deficiency. This would also explain why these *BHLH* genes are only activated at --Fe, but not at +Fe, and why they are induced in leaves and in roots. And in addition, it explains that in the absence of the genes in the *3xbhlh* mutant such a high number of genes are differentially expressed whereby in the wild type the expression appears unchanged. A further question is what could be the function of the 170 genes not regulated by Fe supply. One possibility could have been that these 170 genes are regulated by salicylic acid, a stress hormone. *BHLH38* and *BHLH39* have been brought into the context of SA regulation [@pone.0099234-Kang1], and as a stress hormone SA signaling could be effective upon Fe deficiency stress. However, we exclude that the 170 genes are SA-regulated since none of them was in the intersections with the 50 times higher number of SA-regulated genes in the Venn diagrams, as revealed in our microarray experiments. The main functional categories that these 170 genes belong to are inositol metabolism, circadian rhythm again and UV responses. Hence, it can be concluded that the bhlh functions are important for adapting leaf responses to Fe deficiency stress. The absence of proper regulation would thus result in a more increased leaf chlorosis due to improper adaptation. Since the bHLH genes of the subgroup Ib(2) are expressed in leaves and there induced by Fe deficiency, it is plausible to predict that the leaf regulation is the main function of the bHLH genes which takes place independent of FIT. In conclusion, it will be very interesting in the future to study the connection between leaf responses as an adaptation to Fe deficiency conferred by bHLH subgroup Ib(2) proteins. Supporting Information {#s5} ====================== ###### Gene expression of *BHLH* subgroup Ib genes in *3xbhlh* and wild type plants in response to Fe. A, in roots; B, in leaves; *3xbhlh* and wild type seedlings were grown for 14 d at +Fe and exposed for 3 d to + or −Fe. Roots and leaves were harvested for analysis. n = 4; \* indicates a significant change (p\<0.05) of −Fe versus +Fe; + indicates a significant change (p\<0.05) of *3xbhlh* versus WT. Gene expression was studied by reverse transcription-qPCR. (TIFF) ###### Click here for additional data file. ###### Overview of one biological replicate set of microarray and reverse transcription-qPCR experiments; *3xbhlh* and wild type seedlings were grown for 6 d at + and −Fe and exposed for 6 h to 100 µM SA (+SA) or were mock-treated (−SA). RNA and cDNA was prepared from all samples for use in reverse transcription-qPCR analysis (see [Fig. 7](#pone-0099234-g007){ref-type="fig"}). Microarray analysis was only conducted with --Fe samples (see [Fig. 5](#pone-0099234-g005){ref-type="fig"}, [6](#pone-0099234-g006){ref-type="fig"}). For reverse transcription-PCR and microarray analysis a total of three biological replicates was performed. (TIFF) ###### Click here for additional data file. ###### High-resolution image of the co-expression network analysis of the 29 Fe-regulated genes out of groups I, II and III. The ATTED tool was utilized for construction. Further analysis and additional information are provided in [Fig. 6](#pone-0099234-g006){ref-type="fig"}. (TIF) ###### Click here for additional data file. ###### Gene expression of Fe deficiency and SA marker genes in the samples used for microarray analysis. A, *FIT*; B, *IRT1*; C, *FRO2*; D, *PR1*; *3xbhlh* and wild type seedlings were grown for 6 d at + and −Fe and exposed for 6 h to 100 µM SA (+SA) or were mock-treated (−SA). Whole seedlings were harvested for analysis. n = 3; the --Fe cDNA samples were derived from the RNAs used in the microarray ([Fig. S2](#pone.0099234.s002){ref-type="supplementary-material"}); \* indicates a significant change (p\<0.05) of −Fe versus +Fe; + indicates a significant change (p\<0.05) of *3xbhlh* versus WT; § indicates a significant change (p\<0.05) of +SA versus --SA. Gene expression was studied using reverse transcription-qPCR. (TIF) ###### Click here for additional data file. ###### Primer sequences. (DOT) ###### Click here for additional data file. ###### List of genes of the groups I, II and III with differential expression between *3xbhlh* and wild type at --Fe in the presence and absence of 100 µM SA. (XLS) ###### Click here for additional data file. ###### List of genes of the groups IV and V with differential expression of either *3xbhlh* or wild type at --Fe in the presence and absence of 100 µM SA. (XLS) ###### Click here for additional data file. ###### GO annotation of the differentially regulated genes of groups I to V. (XLS) ###### Click here for additional data file. This work has been funded by the Saarland University and the Heinrich Heine University Düsseldorf. Funding by the Deutsche Forschungsgemeinschaft grant No. DFG-Ba1610/5-1 to PB is kindly acknowledged. The authors thank Dr. Claudia Fink-Straube, Saarbrücken, for iron content determination. We thank Angelika Anna for excellent technical assistance regarding plant growth. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: PB. Performed the experiments: FM MN. Analyzed the data: FM MN PB. Wrote the paper: PB. [^3]: Current address: Department of Anaesthesiology, Intensive Care and Pain Therapy, Saarland University Medical Center and Saarland University Faculty of Medicine, Homburg, Germany

Introduction ============ RNA interference (RNAi) is a highly efficient gene-silencing mechanism in which a small interfering RNA (siRNA) binds a target mRNA, guiding mRNA cleavage via an RNA-induced silencing complex (RISC).^[@bib1],[@bib2]^ This biological phenomenon is widely used as a genetic tool in biomedical research. Advances in RNA chemistry have expanded siRNA applications toward therapeutic development, with robust efficacy seen in phase 2 clinical trials for liver diseases (*e.g.*, transthyretin amyloidosis).^[@bib3],[@bib4],[@bib5]^ Despite its prevalence in biomedical research, the use of RNAi in neurodegenerative research has been limited.^[@bib6]^ There is a significant unmet need for simple, effective, and nontoxic siRNA delivery methods to modulate gene expression in primary neurons and brain. A range of approaches has been evaluated,^[@bib7]^ including AAV viruses,^[@bib8],[@bib9]^ peptide conjugates,^[@bib10]^ oligonucleotide formulations,^[@bib11]^ infusion of naked or slightly modified siRNAs,^[@bib12],[@bib13]^ ultrasound,^[@bib14]^ and convection-enhanced based delivery.^[@bib15]^ None of these approaches has received wide acceptance due to toxicity, a requirement for extensive repetitive dosing, and/or limited spatial distribution. Lipofection and electroporation of siRNAs are challenging in primary neurons due to low transfection efficiencies and their extreme sensitivity to external manipulation.^[@bib16]^ Delivery of siRNA precursors (Lentiviruses and AAV) has been used successfully, but viral transduction cannot readily be turned off and requires extensive formulation and experimental optimization to achieve reproducible, nontoxic silencing in neuronal cells.^[@bib17],[@bib18],[@bib19],[@bib20],[@bib21],[@bib22]^ In this study, we describe the delivery, distribution, and silencing capacity of hydrophobically modified siRNAs (hsiRNAs) in primary neurons and in mouse brain. hsiRNAs are siRNA-antisense hybrids containing numerous chemical modifications (see **[Figure 1](#fig1){ref-type="fig"}** and **Supplementary Table S1** for exact chemical composition of compounds used) designed to promote biodistribution and stability while minimizing immunogenicity. As a model for our studies, we silenced the huntingtin (*Htt*) gene, the causative gene in Huntington\'s disease (HD). HD is an autosomal-dominant neurodegenerative disorder caused by a toxic expansion in the CAG repeat region of the huntingtin gene leading to a variety of molecular and cellular consequences. Tetrabenazine, the only FDA-approved therapy for HD, seeks to alleviate disease symptoms but does not treat the actual problem: the gain of toxic function caused by mutant *Htt*. Recent studies suggest that transient neuronal knockdown of *Htt* mRNA can reverse disease progression without compromising normal cellular function *in vivo*.^[@bib23]^ At present, RNA interference via siRNA or antisense oligonucleotide is one of the most promising therapeutic approaches for transient *Htt* mRNA silencing. We performed a screen of hsiRNAs targeting *Htt* mRNA and identified multiple functional compounds. We showed that primary neurons internalize hsiRNA added directly to the culture medium, with membrane saturation occurring by 1 hour. Direct uptake in neurons induces potent and long-lasting silencing of *Htt* mRNA for up to 3 weeks *in vitro* without major detectable effects on neuronal viability. Additionally, a single injection of unformulated (without cationic lipid or AAV formulation) *Htt* hsiRNA into mouse brain silences *Htt* mRNA with minimal neuronal toxicity. Efficient gene silencing in primary neurons and *in vivo* upon direct administration of unformulated hsiRNA represents a significant technical advance in the application of RNAi to neuroscience research, enabling technically achievable genetic manipulation in a native, biological context. Results ======= hsiRNAs are efficiently internalized by primary neurons ------------------------------------------------------- hsiRNA is an asymmetric compound composed of a 15-nucleotide modified RNA duplex with a single-stranded 3′ extension on the guide strand (**[Figure 1a](#fig1){ref-type="fig"}** and **Supplementary Table S1**).^[@bib24],[@bib25]^ Pyrimidines in the hsiRNA are modified with 2′-O-methyl (passenger strand) or 2′-fluoro (guide strand) to promote stability, and the 3′ end of the passenger strand is conjugated to a hydrophobic teg-Chol (tetraethylene glycol cholesterol) to promote membrane binding and association.^[@bib26]^ The single-stranded tail contains hydrophobic phosphorothioate linkages and promotes cellular uptake by a mechanism similar to that of antisense oligonucleotides.^[@bib27]^ The presence of phosphorothioates, ribose modifications, and a cholesterol conjugate contribute to overall hydrophobicity and are essential for compound stabilization and efficient cellular internalization. Previous studies have shown that hydrophobically modified siRNAs bind to a wide range of cells and is readily internalized without the requirement for a transfection reagent.^[@bib26],[@bib28],[@bib29]^ Here, we evaluated whether asymmetric hydrophobically modified siRNAs are efficiently internalized by primary neurons. We found that, when added to the culture medium, Cy3-labeled hsiRNAs rapidly associated with primary cortical neurons (**[Figure 1b](#fig1){ref-type="fig"}**). These Cy3-labeled hsiRNAs were observed in every cell in the culture, demonstrating efficient and uniform uptake. Initially, hsiRNAs mainly associate with neurites and, over time, accumulate in the cell bodies. Treatment of primary neurons with a previously identified hsiRNA targeting *Ppib*^[@bib26],[@bib28]^ (encodes cyclophilin B) reduced target mRNA levels by 90%, further supporting that the observed compound internalization results in potent gene silencing (**[Figure 1c](#fig1){ref-type="fig"}**). Identification of hsiRNAs that silence huntingtin mRNA ------------------------------------------------------ Robust uptake and efficacy observed with hsiRNAs in primary cortical neurons encouraged us to identify functional compounds that target *Htt* mRNA, the single gene responsible for the development of Huntington\'s disease. The hsiRNA\'s extensive chemical scaffold^[@bib26],[@bib28]^ is essential for stability, minimization of innate immune response,^[@bib30],[@bib31]^ and cellular internalization but imposes significant restrictions on sequence space by potentially interfering with the compound\'s RISC-entering ability. To maximize the likelihood of identifying functional *Htt* hsiRNAs and to evaluate the hit rate for this type of chemistry, we designed (using conventional criteria described in Materials and Methods) and synthesized hsiRNAs targeting 94 sites across the human *Htt* mRNA (**Supplementary Table S1**). The panel of hsiRNAs was initially screened for efficacy in HeLa cells by adding hsiRNA directly to the culture medium (without lipofection or electroporation) to a final concentration of 1.5 µmol/l and evaluating impact on levels of *Htt* and housekeeping (*Ppib*) gene mRNA expression using the QuantiGene (Affymetrix, Santa Clara, CA) assay. At this concentration, 24 hsiRNAs reduced *Htt* mRNA levels to less than 50% of control levels, including 7 hsiRNAs that reduced *Htt* mRNA levels below 30% of control (**[Figure 2a](#fig2){ref-type="fig"}**). Unlike unmodified siRNA libraries, creating a library with extensive 2\'-O-methyl and 2\'-fluoro modifications introduces additional constraints on sequence selection. As a result, hit rates for modified siRNA screens are lower than that seen for conventional unmodified siRNA.^[@bib32],[@bib33],[@bib34],[@bib35]^ Functional hsiRNAs targeted sites distributed throughout the mRNA, except the distal end of the 3′ UTR, which later was shown to be part of the alternative *Htt* gene isoform^[@bib36]^ not expressed in HeLa cells (data not shown). Discounting the \~32 hsiRNAs targeting long 3′ UTR sites absent from the *Htt* isoform in HeLa cells, almost 40% of hsiRNAs showed some level of activity at 1.5 µmol/l, demonstrating that the evaluated chemical scaffold is well tolerated by the RNAi machinery and a functional compound can be easily identified against a wide range of targets. Half-maximal inhibitory concentrations (IC~50~) for passive uptake of hsiRNAs ranged from 82 to 766 nmol/l (**Supplementary Table S1 and Figure S1**). In lipid-mediated delivery, eight of the most active hsiRNAs had IC~50~ values ranging from 4 to 91 pmol/l (**Supplementary Table S1**). The best clinically active siRNAs are usually characterized by IC~50~ values in the low pmol/l range.^[@bib37]^ An ability to identify highly potent compounds with low picomolar IC~50~ values suggests that the hsiRNA chemical scaffold does not interfere with siRNA biological activity in selected compounds. The most potent hsiRNA targeting position, 10150 (HTT10150), and an unmodified conventional siRNA version of HTT10150 showed similar IC~50~ values in lipid-mediated delivery (4 and 13 pmol/l respectively, **[Figure 2c](#fig2){ref-type="fig"}**), further confirming that the hsiRNA chemical scaffold does not interfere with RISC loading or function. Only the fully modified hsiRNA, and not the unmodified version, silenced *Htt* mRNA by passive uptake (**[Figure 2b](#fig2){ref-type="fig"}**). Thus, the chemical scaffold described here does not interfere with RISC assembly and is sufficient to support unformulated compound uptake and efficacy. HTT10150 was used for subsequent studies. Potent and specific silencing with unformulated hsiRNAs in primary neurons -------------------------------------------------------------------------- HTT10150 induced a concentration-dependent silencing at 72 hours and 1 week after unformulated addition to either primary cortical or primary striatal neurons isolated from FVB/NJ mice (**[Figure 3a](#fig3){ref-type="fig"}**). At 1.25 µmol/l, HTT10150 induced maximal silencing, reducing both *Htt* mRNA levels and HTT protein levels by as much as 70 and 85%, respectively (**[Figure 3a](#fig3){ref-type="fig"}**--**[c](#fig3){ref-type="fig"}** for original westerns). HTT10150 hsiRNA did not affect the expression levels of housekeeping controls (*Ppib* and *Tubb1*) or the overall viability of primary neuronal cells, as measured by the alamarBlue assay, up to a 2 µmol/l concentration (**Supplementary Figure S2**). Similar results were obtained with another hsiRNA targeting *Htt* mRNA (**[Figure 3c](#fig3){ref-type="fig"}**), supporting that the observed phenomena is not unique to HTT10150. These experiments, in conjugation with the results seen from targeting *Ppib* (**[Figure 1b](#fig1){ref-type="fig"}**), indicate that a diversity of genes and target sequences can be silenced by hsiRNAs in primary neurons simply upon direct addition of compounds into cellular media. Since loaded RISC has a typical half-life of weeks,^[@bib38]^ silencing is expected to be long lasting in nondividing cells. To evaluate duration of silencing after a single HTT10150 treatment of primary cortical neurons, *Htt* mRNA levels were measured at 1-, 2-, and 3-week intervals (**[Figure 3d](#fig3){ref-type="fig"}**). A single treatment with hsiRNA induced *Htt* silencing that persisted for at least 3 weeks, the longest time that primary cortical neurons can be maintained in culture. Together, these data demonstrate that hsiRNAs are a simple and straightforward approach for potent, specific, nontoxic, and long-term modulation of gene expression in primary neurons *in vitro*. hsiRNA distribution *in vivo* in mouse brain after intrastriatal injection -------------------------------------------------------------------------- Having shown that hsiRNAs effectively silence their targets in primary neurons *in vitro*, we sought to evaluate the ability of HTT10150 to silence *Htt* mRNA in the mouse brain *in vivo*. The distribution of HTT10150 was evaluated in perfused brain sections prepared 24 hours after intrastriatal injection with 12.5 µg Cy3-labeled hsiRNA in artificial cerebral spinal fluid (ACSF). We observed a steep gradient of fluorescence emanating from the injection site and covering most of the ipsilateral striatum (**[Figure 4a](#fig4){ref-type="fig"}**,**[b](#fig4){ref-type="fig"}**), while no fluorescence was visually detectable in the contralateral side of the brain. In high magnification images of the ipsilateral side, hsiRNAs appeared preferentially associated with the tissue matrix and fiber tracts. In addition, efficient internalization was observed in a majority of cell bodies (**[Figure 4c](#fig4){ref-type="fig"}**,**[d](#fig4){ref-type="fig"}**). Consistent with *in vitro* studies, we observed Cy3-labeled hsiRNA in neuronal processes and as punctae in the perinuclear space of multiple cell types, including NeuN-positive neurons^[@bib39],[@bib40]^ (**[Figure 4d](#fig4){ref-type="fig"}**,**[e](#fig4){ref-type="fig"}**). In summary, a single intrastriatal injection delivers hsiRNA to neurons in the striatum of the injected side. hsiRNA effectively silences Htt in vivo with minimal cytotoxicity or immune activation -------------------------------------------------------------------------------------- To measure HTT10150 efficacy *in vivo*, we performed dose-response studies in wild type FVB/NJ mice injected intrastriatally with 3.1, 6.3, 12.5, or 25 µg of HTT10150. As controls, we injected mice with a non-targeting control hsiRNA (NTC), ACSF, or PBS. In punch biopsies taken from the ipsilateral and contralateral striatum, HTT10150 reduced *Htt* mRNA levels in a dose-dependent manner (**[Figure 5a](#fig5){ref-type="fig"}**). This experiment was repeated several times with similar results. The *Htt* mRNA is significantly reduced in the ipsilateral side of striatum in all experiments. We observed robust dose-dependent silencing with up to 77% (one-way analysis of variance, *P* \< 0.0001) reduction in *Htt* mRNA expression levels at the highest dose. Interestingly we observe statistically significant, but less pronounced silencing in the contralateral striatum and the cortex. The silencing reaches statistical significance with both one-way and two-way analysis of variance (values for two-way analysis of variance are presented in **[Figure 5](#fig5){ref-type="fig"}**). While some level of fluorescence is detectable in these brain regions with high laser intensity, it is very close to the tissue auto-fluorescence and thus is not reported here. We will be investigating this phenomenon further, but it is clear that the level of silencing is at least correlative to the sharp gradient of diffusion from the injection site. Finally, *Htt* mRNA silencing is observed with HTT10150 but not with NTC or ACSF (**[Figure 5](#fig5){ref-type="fig"}**). In addition, the HTT10150 does not affect expression of several housekeeping genes (PPIB, HPRT). In combination, this is indicative of *Htt* mRNA silencing being caused by HTT10150 hsiRNA and not by off-target effects. Nucleic acids, including siRNAs, are potent stimulators of the innate immune response,^[@bib41]^ but extensive chemical modifications, like 2′-O-methyl, are expected to suppress the immunostimulatory effects of siRNAs *in vitro* and *in vivo*.^[@bib42]^ To assess innate immune response activation by hsiRNAs *in vivo*, we quantified IBA-1-positive microglial cells in brain sections from mice injected with 12.5 µg HT10150 or artificial CSF. IBA-1 is specific to microglial cells and is upregulated following injury to the brain, allowing us to distinguish between resting and activated microglia.^[@bib43],[@bib44],[@bib45]^ In the case of a major innate immune response, an increase of 200--300% in total microglia is customary.^[@bib46]^ Total microglia counts showed only a 25% increase in the ipsilateral striatum at 5 days post-injection indicating a lack of any major inflammatory response (**Supplementary Figure S3**). Thus, the observed activation is relatively minor but reaches statistical significance, indicating some level of response. Levels of innate immune response might be more pronounced immediately following compound administration. To assess the level of stimulation in more detail, we separately evaluated the number of activated and resting microglia at both 6 hours and 5 days post-injection. At 6 hours post-injection, we observed a significant increase in the number of activated microglia in the injected side of the brain with both ACSF and HTT10150. The injection event itself causes trauma and induces a major increase in activated microglia (ninefold) compared to the contralateral side of the brain (**[Figure 6b](#fig6){ref-type="fig"}**).^[@bib12],[@bib47]^ In the presence of HTT10150, the number of activated microglia was additionally increased twofold compared to ACSF, indicating enhancement of trauma-related microglia activation in the presence of oligonucleotide, although the relative contribution of the oligonucleotide to the trauma-related induction is minor. HTT10150-treated mice also showed some elevation of activated microglia in the contralateral striatum 6 hours post-injection (**[Figure 6b](#fig6){ref-type="fig"}**); however, after 5 days, all changes in number of microglia in the contralateral side of the brain disappeared (**[Figure 6a](#fig6){ref-type="fig"}**,**[c](#fig6){ref-type="fig"}** for representative images), suggesting that HTT10150-dependent activation of microglia in the contralateral striatum is transient. Despite the mild immune stimulation in the brains of animals injected with HTT10150, we did not observe any overall significant reduction of DARPP-32, an established marker for striatal neuron viability^[@bib48]^ (**[Figure 7](#fig7){ref-type="fig"}**). The only observed effect was at a small area directly around the injection site in animals treated with 25 µg HTT10150 (**Supplementary Figure S4**). Taken together, our data show that a single intrastriatal injection of hsiRNA induces potent gene silencing with a mild immune response and minimal neuronal toxicity *in vivo*. Discussion ========== Simple, effective, and nontoxic delivery of synthetic oligonucleotides to primary neurons and brain tissue represents a challenge to the use of RNAi as a research tool and therapeutic for neurodegenerative diseases like HD.^[@bib7]^ We have shown that hsiRNAs elicit potent silencing in primary neurons in culture, without effect on housekeeping gene expression, and with minimal toxicity at effective doses Additionally, a non-targeting control hsiRNA did not silence any of the mRNAs tested (*Htt*, *Ppib*, *Hprt*), suggesting that these compounds are both sequence specific and on-target. Interestingly, the level of silencing is more pronounced on the protein level (\>90%) compared to the mRNA level (\>70%). The mRNA plateau effect is reproducible and is specific to *Htt* mRNA, as housekeeping genes like *Ppib* can be silenced by 90%. One potential explanation is that some fraction of huntingtin mRNA is translationally inactive and poorly accessible by RNAi machinery. We are continuing to investigate this phenomenon. Silencing in primary neurons persists for multiple weeks after a single administration, consistent with the expected half-life of active RISC.^[@bib49]^ Moreover, efficient intracellular delivery does not require the use of lipids or viral packaging. Currently, the most impressive *in vivo* modulation of *Htt* mRNA expression is demonstrated with 2′-O-methoxyethyl GapmeR antisense oligonucleotides. A single injection of 50 µg of antisense oligonucleotides or infusion of around 500 µg results in potent and specific *Htt* mRNA silencing and marked improvement in multiple phenotypic endpoints.^[@bib23],[@bib50],[@bib51],[@bib52]^ However, 2′-O-methoxyethyl GapmeR antisense oligonucleotides are not readily commercially available making them inaccessible for the majority of academic labs. Here, we show *Htt* mRNA silencing in the ipsilateral striatum and cortex, two brain areas significantly affected in HD disease progression, with a single intrastriatal injection. As a considerably reduced level of silencing was observed on the contralateral side of the brain, bilateral injections might be necessary to promote equal gene silencing in both hemispheres. The limited distribution profile observed *in vivo* restricts immediate adoption of this technology for use in larger brains and eventually as a therapeutic for neurodegenerative disease. Tissue distribution can be improved by tailoring the chemical scaffold (*e.g.*, number and type of sugar modifications, position of phosphorothioate linkages) or by changing the conjugation moiety to promote receptor-mediated cellular internalization. Formulation of hsiRNA in exosomes, exosome-like liposomes, or shielding the compounds with polyethylene glycol may also provide an alternative strategy to improve tissue distribution.^[@bib53],[@bib54]^ Here, we describe a class of self-delivering therapeutic oligonucleotides capable of targeted, nontoxic, and efficient *Htt* gene silencing in primary neurons and *in vivo*. This chemical scaffold can be specifically adapted to many different targets to facilitate the study of neuronal gene function *in vitro* and *in vivo*. The development of an accessible strategy for genetic manipulation in the context of a native, biological environment represents a technical advance for the study of neuronal biology and neurodegenerative disease. Materials and methods ===================== *hsiRNA design.* We designed and synthesized a panel of 94 hsiRNA compounds (**Supplementary Table S1**) targeting the human huntingtin gene. These sequences span the gene and were selected to comply with standard siRNA design parameters^[@bib24]^ including assessment of GC content, specificity and low seed compliment frequency,^[@bib55]^ elimination of sequences containing miRNA seeds, and examination of thermodynamic bias.^[@bib56],[@bib57]^ *Oligonucleotide synthesis, deprotection, and purification.* Oligonucleotides were synthesized using standard phosphoramidite, solid-phase synthesis conditions on a 0.2--1 µmole scale using a MerMade 12 (BioAutomation, Irving, TX) and Expedite DNA/RNA synthesizer. Oligonucleotides with unmodified 3′ ends were synthesized on controlled pore glass (CPG) functionalized with long-chain alkyl amine and a Unylinker terminus (Chemgenes, Wilmington, MA). Oligonucleotides with 3′-cholesterol modifications were synthesized on modified solid support (Chemgenes). Phosphoramidite solutions were prepared at 0.15 mol/l in acetonitrile for 2′-TBDMS, 2′-O-methyl (Chemgenes), and Cy3 modifications or 0.13 mol/l for 2′-fluoro (BioAutomation) modifications. Phosphoramidites were activated in 0.25 mol/l 4,5-dicyanoimidazole in acetonitrile. Detritylation was performed in 3% dichloroacetic acid in dichloromethane for 80 seconds. Capping was performed in 16% *N*-methylimidazole in tetrahydrofuran and acetic anhydride:pyridine:tetrahydrofuran, (1:2:2, v/v/v) for 15 seconds. Oxidation was performed using 0.1 mol/l iodine in pyridine:water:tetrahydrofuran (1:2:10, v/v/v). The CPG was removed from the solid-phase column and placed in a polypropylene screw cap vial. Dimethylsulfoxide (100 µl) and 40% methylamine (250 µl) are added directly to the CPG and shaken gently at 65 °C for exactly 16 minutes. The vial was cooled on dry ice before the cap was removed. The supernatant was transferred to another polypropylene screw cap vial, and the CPG was rinsed with two 150 µl portions of dimethylsulfoxide, which were combined with original supernatant. Oligonucleotides without 2′-TBDMS-protecting groups were lyophilized. Oligonucleotides with 2′-TBDMS-protecting groups were desilylated by adding 375 µl triethylamine trihydrofluoride (\~1.5 volumes relative to 40% methylamine) and incubated for exactly 16 minutes at 65 °C with gentle shaking. Samples were quenched by transferring to a 15 ml conical tube containing 2 ml of 2 mol/l triethylammonium acetate buffer (pH 7.0). The sample was stored at −80 °C until high-performance liquid chromatography purification. Oligonucleotides were purified by reverse-phase high-performance liquid chromatography on a Hamilton PRP-C18 column (21 × 150 mm) using an Agilent Prostar 325 high-performance liquid chromatography (Agilent, Santa Clara, CA). Buffer A 0.05 mol/l tetraethylammonium acetate with 5% acetonitrile, Buffer B 100% acetonitrile, with a gradient of 0% B to 35% B over 15 minutes at 30 ml/minutes. Purified oligonucleotides were lyophilized to dryness, reconstituted in water, and passed over a Hi-Trap cation exchange column to exchange the tetraethylammonium counter-ion with sodium. *Cell culture.* HeLa cells (ATCC, Manassas, VA; \#CCL-2) were maintained in Dulbecco\'s Modified Eagle\'s Medium (Cellgro, Corning, NY; \#10-013CV) supplemented with 10% fetal bovine serum (FBS; Gibco, Carlsbad, CA; \#26140) and 100 U/ml penicillin/streptomycin (Invitrogen, Carlsbad, CA; \#15140) and grown at 37 °C and 5% CO~2~. Cells were split every 2 to 5 days and discarded after 15 passages. *Preparation of primary neurons.* Primary cortical neurons were obtained from FVB/NJ mouse embryos at embryonic day 15.5. Pregnant FVB/NJ females were anesthetized by intraperitoneal injection of 250 mg Avertin (Sigma, St Louis, MO; \#T48402) per kg weight, followed by cervical dislocation. Embryos were removed and transferred into a Petri dish with ice-cold Dulbecco\'s Modified Eagle\'s Medium/F12 medium (Invitrogen; \#11320). Brains were removed, and meninges carefully detached. Cortices were isolated and transferred into a 1.5-ml tube with prewarmed papain solution for 25 minutes at 37 °C, 5% CO~2~, to dissolve tissue. Papain solution was prepared by suspending DNase I (Worthington, Lakewood, NJ; \#54M15168) in 0.5 ml Hibernate E medium (Brainbits, Springfield, IL; \#HE), and transferring 0.25 ml DNase I solution to papain (Worthington, Lakewood, NJ; \#54N15251) dissolved in 2 ml Hibernate E medium and 1 ml Earle\'s balanced salt solution (Worthington; \#LK003188). After the 25-minute incubation, papain solution was replaced with 1 ml NbActiv4 medium (Brainbits; \#Nb4-500) supplemented with 2.5% FBS. Cortices were dissociated by repeated pipetting with a fire-polished, glass, Pasteur pipette. Cortical neurons were counted and plated at 1 × 10^6^ cells per ml. For live-cell imaging, culture plates were precoated with poly-[l]{.smallcaps}-lysine (Sigma; \#P4707), and 2 × 10^5^ cells were added to the glass center of each dish. For silencing assays, neurons were plated on 96-well plates precoated with poly-[l]{.smallcaps}-lysine (BD BIOCOAT, Corning, NY; \#356515) at 1 × 10^5^ cells per well. After overnight incubation at 37 °C, 5% CO~2~, an equal volume of NbActiv4 supplemented with anti-mitotics, 0.484 µl/ml of UTP Na~3~ (Sigma; \#U6625), and 0.2402 µl/ml of FdUMP (Sigma; \#F3503), was added to neuronal cultures to prevent growth of nonneuronal cells. Half of the media volume was replaced every 48 hours until the neurons were treated with siRNA. Once the cells were treated, media was not removed, only added. All subsequent media additions contained antimitotics. *Direct delivery (passive uptake) of oligonucleotides.* Cells were plated in Dulbecco\'s Modified Eagle\'s Medium containing 6% FBS at 10,000 cells per well in 96-well tissue culture plates. hsiRNA was diluted to twice the final concentration in OptiMEM (Gibco; \#31985-088), and 50 μl diluted hsiRNA was added to 50 μl of cells, resulting in 3% FBS final. Cells were incubated for 72 hours at 37 °C and 5% CO~2~. Based on previous experience, we know that 1.5 µmol/l active hsiRNA supports efficient silencing without toxicity. The primary screen for active *Htt* siRNAs, therefore, was performed at 1.5 µmol/l compound, which also served as the maximal dose for *in vitro* dose--response assays. *hsiRNA lipid-mediated delivery.* Cells were plated in Dulbecco\'s Modified Eagle\'s Medium with 6% FBS at 10,000 cells per well in 96-well tissue culture--treated plates. hsiRNA was diluted to four times the final concentration in OptiMEM, and Lipofectamine RNAiMAX Transfection Reagent (Invitrogen; \#13778150) was diluted to four times the final concentration (final = 0.3 µl/25 µl/well). RNAiMAX and hsiRNA solutions were mixed 1:1, and 50 µl of the transfection mixture was added to 50 µl of cells resulting in 3% FBS final. Cells were incubated for 72 hours at 37 °C and 5% CO~2~. *mRNA quantification in cells and tissue punches.* mRNA was quantified using the QuantiGene 2.0 Assay (Affymetrix; \#QS0011). Cells were lysed in 250 μl diluted lysis mixture composed of 1 part lysis mixture (Affymetrix; \#13228), 2 parts H~2~O, and 0.167 μg/μl proteinase K (Affymetrix; \#QS0103) for 30 minutes at 55 °C. Cell lysates were mixed thoroughly, and 40 μl (\~8,000 cells) of each lysate was added per well to a capture plate with 40 μl diluted lysis mixture without proteinase K. Probe sets were diluted as specified in the Affymetrix protocol. For HeLa cells, 20 μl human *HTT* or *PPIB* probe set (Affymetrix; *\#*SA-50339, \#SA-10003) was added to appropriate wells for a final volume of 100 μl. For primary neurons, 20 μl of mouse *Htt* or *Ppib* probe set (Affymetrix; *\#*SB-14150, \#SB-10002) was used. Tissue punches (5 mg) were homogenized in 300 μl of Homogenizing Buffer (Affymetrix; \#10642) containing 2 μg/μl proteinase K in 96-well plate format on a QIAGEN TissueLyser II (Qiagen, Valencia, CA; \#85300), and 40 μl of each lysate was added to the capture plate. Probe sets were diluted as specified in the Affymetrix protocol, and 60 μl of *Htt* or *Ppib* probe set was added to each well of the capture plate for a final volume of 100 μl. Signal was amplified according to the Affymetrix protocol. Luminescence was detected on either a Veritas Luminometer (Promega, Madison, WI; \#998--9100) or a Tecan M1000 (Tecan, Morrisville, NC). *Western blot.* Cell lysates (25 µg) were separated by SDS--PAGE using 3--8% Tris-acetate gels (Life Technologies, Grand Island, NY; \#EA03785BOX) and transferred to nitrocellulose using a TransBlot Turbo apparatus (BioRad, Hercules, CA; \#1704155). Blots were blocked in 5% nonfat dry milk (BioRad; \#1706404) diluted in Tris-buffered saline with 0.1% Tween-20 (TBST) for 1 hour at room temperature then incubated in N-terminal antihuntingtin antibody Ab1^[@bib58]^ diluted 1:2,000 in blocking solution overnight at 4 °C with agitation. After washing in TBST, blots were incubated in peroxidase-labeled antirabbit IgG (Jackson ImmunoResearch, West Grove, PA; \#711035152) diluted in blocking buffer for 1 hour at room temperature, washed in TBST, and proteins were detected using SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific, Rockford, IL; \#34080) and Hyperfilm ECL (GE Healthcare, Buckinghamshire, UK; \#28906839). Blots were reprobed with anti-β tubulin antibody (Sigma; \#T8328) as a loading control. Films were scanned with a flatbed scanner (Epson Perfection V750 Pro; Epson, Long Beach, CA), and densitometry was performed using NIH ImageJ software to determine total intensity of each band. The huntingtin signal was divided by the tubulin signal to normalize to protein content, and percent of untreated control was determined for each set of samples (*N* = 5). *Live cell imaging.* To monitor live cell hsiRNA uptake, cells were plated at a density of 2 × 10^5^ cells per 35-mm glass-bottom dish. Cell nuclei were stained with NucBlue (Life Technologies; \#R37605) as indicated by the manufacturer. Imaging was performed in phenol red-free NbActiv4 (Brainbits; \#Nb4-500). Cells were treated with 0.5 μmol/l Cy3-labeled hsiRNA, and live cell imaging was performed over time. All live cell confocal images were acquired with a Leica DM IRE2 confocal microscope using 63x oil immersion objective (Buffalo Grove, IL), and images were processed using ImageJ (1.47v) software. *Stereotaxic injections.* FVB/NJ mice (50% male and 50% female for each dose group, 6--8 weeks old) were deeply anesthetized with 1.2% Avertin (Sigma; \#T48402) and microinjected by stereotactic placement into the right striatum (coordinates relative to bregma: 1.0 mm anterior, 2.0 mm lateral, and 3.0 mm ventral). For both toxicity (DARPP-32 staining) and efficacy studies, mice were injected with either PBS or artificial CSF (2 μl per striata), 12.5 μg of nontargeting hsiRNA (2 μl of 500 µmol/l stock per striata), 25 μg of HTT10150 hsiRNA (2 μl of 1 mmol/l stock per striata), 12.5 μg of HTT10150 hsiRNA (2 μl of 500 μmol/l stock per striata), 6.3 μg of HTT10150 hsiRNA (2 μl of 250 μmol/l stock per striata), or 3.1 μg of HTT10150 hsiRNA (2 μl of 125 μmol/l stock per striata). For toxicity studies, *n* = 3 mice were injected per group, and for efficacy studies, *n* = 8 mice were injected per group. Mice were euthanized 5 days post-injection, brains were harvested, and three 300-μm coronal sections were prepared. From each section, a 2-mm punch was taken from each side (injected and noninjected) and placed in RNAlater (Ambion, Carlsbad, CA; \#AM7020) for 24 hours at 4 °C. Each punch was processed as an individual sample for Quantigene 2.0 assay analysis (Affymetrix) and averaged for a single animal point. All animal procedures were approved by the University of Massachusetts Medical School Institutional Animal Care and Use Committee (protocol number A-2411). *Immunohistochemistry/immunofluorescence.* Mice were injected intrastriatally with 12.5 µg of Cy3-labeled hsiRNA. After 24 hours, mice were sacrificed and brains were removed, embedded in paraffin, and sliced into 4-μm sections that were mounted on glass slides. Sections were deparaffinized by incubating in Xylene twice for 8 minutes. Sections were rehydrated in serial ethanol dilutions (100%, 95%, and 80%) for 4 minutes each, and then washed twice for 2 minutes with PBS. For NeuN staining,^[@bib39],[@bib40]^ slides were boiled for 5 minutes in antigen retrieval buffer (10 mmol/l Tris/ 1mmol/l EDTA (pH 9.0)), incubated at room temperature for 20 minutes, and then washed for 5 minutes in PBS. Slides were blocked in 5% normal goat serum in PBS containing 0.05% Tween 20 (PBST) for 1 hour and washed once with PBST for 5 minutes. Slides were incubated with primary antibody (Millipore, Taunton, MA; MAB377, 1:1,000 dilution in PBST) for 1 hour and washed three times with PBST for 5 minutes. Slides were then incubated with secondary antibody (Life Technologies; \#A11011, 1:1000 dilution in PBST) for 30 minutes in the dark and washed three times with PBST for 5 minutes each. Slides were then counterstained with 250 ng/ml 4\',6-diamidino-2-phenylindole (Molecular Probes, Life Technologies; \#D3571) in PBS for 1 minute and washed three times with PBS for 1 minute. Slides were mounted with mounting medium and coverslips and dried overnight before imaging on a Leica DM5500 microscope fitted with a DFC365 FX fluorescence camera. For toxicity studies, injected brains were harvested after 5 days. For microglial activation studies, brains were harvested after 6 hours or 5 days. Extracted, perfused brains were sliced into 40-µm sections on the Leica 2000T Vibratome (Leica Biosystems, Wetzlar, Germany) in ice-cold PBS. Every sixth section was incubated with DARPP-32 (Abcam, Cambridge, UK; \#40801; 1:10,000 in PBS) or IBA-1 (Wako; \#019-19741; 1:1,000 in PBS) antibody, for a total of nine sections per brain and eight images per section (four per hemisphere). IBA-1 sections were incubated in blocking solution (5% normal goat serum, 1% bovine serum albumin, 0.2% Triton-X-100, and 0.03% hydrogen peroxide in PBS) for 1 hour, and then washed with PBS. Sections were incubated overnight at 4 °C in primary antibody, anti-Iba1 (polyclonal rabbit anti-mouse/human/rat; dilution: 1:1,000 in blocking solution) (Wako; \#019-19741). Sections were then stained with goat antirabbit secondary antibody (1:200 dilution) (Vector Laboratories, Burlingame, CA), followed by a PBS wash, the Vectastain ABC Kit (Vector Laboratories), and another PBS wash. IBA-1 was detected with the Metal Enhanced DAB Substrate Kit (Pierce, Rockford, IL). For DARPP32 staining, sections were washed for 3 minutes in 3% hydrogen peroxide, followed by 20 minutes in 0.2% TritonX-100 and 4 hours in 1.5% normal goat serum in PBS. Sections were incubated overnight at 4 °C in DARPP32 primary antibody (1:10,000 dilution) (Abcam; \#40801) made up in 1.5% normal goat serum. Secondary antibody and detection steps were conducted as described for IBA-1 staining. DARPP-32 sections were mounted and visualized by light microscopy with 20× objective on a Nikon Eclipse E600 with a Nikon Digital Sight DSRi1 camera (Nikon, Tokyo, Japan). The number of DARPP-32-positive neurons was quantified manually using the cell counter plug-in on ImageJ for tracking. Activated microglia were quantified by morphology of IBA-1-positive cells^[@bib42],[@bib43],[@bib44],[@bib45]^ from the same number of sections captured with 40× objective. Counting of both IBA-1- and DARPP-32-positive cells was blinded. Coronal section images were taken with a Coolscan V-ED LS50 35-mm Film Scanner (Nikon, Tokyo, Japan). *Statistical analysis.* Data were analyzed using GraphPad Prism 6 software (GraphPad Software, Inc., San Diego, CA). Concentration-dependent IC~50~ curves were fitted using a log(inhibitor) versus response--variable slope (four parameters). The lower limit of the curve was set at zero, and the upper limit of the curve was set at 100. For each independent mouse experiment, the level of knockdown at each dose was normalized to the mean of the control group (the noninjected side of the PBS or artificial CSF groups). *In vivo* data were analyzed using a two-way repeated-measures analysis of variance with Tukey\'s multiple comparisons test for dose and side of brain. Differences in all comparisons were considered significant at *P* values less than 0.05 compared with the NTC- injected group. *P* values reported represent significance of the entire dose group relative to NTC and are not specific to the ipsilateral or contralateral side. For microglial activation, significance was calculated using a parametric, unpaired, two-tailed *t*-test for comparison between dose groups, and paired *t*-test for comparison between ipsilateral and contralateral hemispheres within the same dose group. [**SUPPLEMENTARY MATERIAL**](#sup1){ref-type="supplementary-material"} **Figure S1.** Active hsiRNAs silence huntingtin mRNA in a concentration dependent manner in HeLa cells. **Figure S2.** HTT10150 does not affect primary cortical neuron viability. **Figure S3.** HTT10150 causes a slight increase in total resting microglia 5 days post injection. **Figure S4.** HTT10150 shows limited toxicity at the site of injection at the 25 µg dose. **Table S1.** Detailed sequence, chemical modification patterns, and efficacy of hsiRNAs. This project was funded by the CHDI Foundation (Research Agreement A-6119 and JSC A6367), NIH Strategic Fund (TR000888-02), and NIGMS (GM108803-01A1). We especially thank Darryl Conte for help with manuscript writing and editing, and all the members of the Khvorova, Aronin, and DiFiglia Labs and CHDI Foundation Inc. for stimulating discussions and thoughtful guidance. We also thank M. Sena Esteves and C. Mello Lab for guidance on microscopy. A.K. (University of Massachusetts Medical School (UMMS)), N.A. (UMMS), and M.D. (Mass General Institute for Neurodegenerative Disease) conceived of the study. M.R.H. (UMMS) synthesized and purified siRNAs. J.F.A. (UMMS), L.M.H. (UMMS), A.H.C. (UMMS), M.-C.D. (UMMS), K.C. (UMMS), J.A. (UMMS), E.S. (UMMS), E.J. (UMMS), and E.S. (Mass General Institute for Neurodegenerative Disease) contributed experimentally. M.F.O. (UMMS) performed PyMol modeling. J.F.A., A.K., N.A., and M.D. wrote and edited the manuscript. A.K. owns stock at RXi Pharmaceuticals and Advirna LLC, which holds a patent and license on asymmetric, hydrophobically modified siRNAs. Other authors do not have any competing financial interest to disclose. Supplementary Material {#sup1} ====================== ###### Active hsiRNAs silence huntingtin mRNA in a concentration dependent manner in HeLa cells. ###### Click here for additional data file. ###### HTT10150 does not affect primary cortical neuron viability. ###### Click here for additional data file. ###### HTT10150 causes a slight increase in total resting microglia 5 days post injection. ###### Click here for additional data file. ###### HTT10150 shows limited toxicity at the site of injection at the 25 µg dose. ###### Click here for additional data file. ###### Detailed sequence, chemical modification patterns, and efficacy of hsiRNAs. ###### Click here for additional data file. {#fig1} {#fig2} {#fig3} {#fig4} {#fig5} {#fig6} {#fig7}

Introduction {#s1} ============ Immunity is essential for survival, yet energetically expensive and potentially self-damaging. The immune system needs to be tightly regulated and highly responsive to changes in external and internal environments, adapting to developmental stage [@ppat.1003720-Kollmann1], nutritional state [@ppat.1003720-Becker1], and stress [@ppat.1003720-Cohen1]. Hormonal control of the immune system is not well-understood, and lies at the heart of a diverse range of clinically relevant phenomena including immune circadian rhythm [@ppat.1003720-Silver1], [@ppat.1003720-Stone1], obesity induced inflammation [@ppat.1003720-Fantuzzi1] and age- and gender-differences in immune function [@ppat.1003720-Kollmann1], [@ppat.1003720-Gilliver1]. *Drosophila* has proven to be a fruitful model of innate immunity [@ppat.1003720-Lemaitre1]. The innate immune system in the fly comprises humoral and cellular responses [@ppat.1003720-Lemaitre1]. The humoral response is best characterized by the secretion of antimicrobial peptides (AMPs) locally by epithelia such as the gut, or systemically by the fat-body, a functional analogue of the mammalian liver. This response to bacterial or fungal infection is mainly regulated by the Toll and Imd signaling pathways. Cellular immunity is performed by hemocytes, phagocytic circulating immune cells. *Drosophila* has become a powerful system to study phagocytosis due to high conservation of the processes involved [@ppat.1003720-Stuart1], [@ppat.1003720-Stuart2]. Cellular immunity is required to survive various types of bacterial infections, complementing the humoral response [@ppat.1003720-AvetRochex1]--[@ppat.1003720-Ulvila1] and is involved in the response to wasp parasitoid infestation [@ppat.1003720-Krzemien1]. In addition, in the larvae, hemocytes play a role in inter-organ communication, as they are required for the fat-body to mount a full humoral antimicrobial response after an intestinal infection [@ppat.1003720-Charroux1], [@ppat.1003720-Basset1], [@ppat.1003720-Wu1]. Hemocytes are also recruited to wounds in embryo and larvae [@ppat.1003720-Babcock1]--[@ppat.1003720-Wood1], potentially clearing damaged tissue and preventing further dispersal of microorganisms from unsterile wound sites. In parallel to their role in immunity, hemocytes perform developmental and homeostatic functions such as the phagocytosis of apoptotic cells and extracellular matrix secretion. These functions are essential at embryonic stages, for example, for central nervous system and renal tubule development [@ppat.1003720-Bunt1], [@ppat.1003720-Olofsson1]. The crosstalk between immunity and nutritional state or stress [@ppat.1003720-Becker1], [@ppat.1003720-DiAngelo1]--[@ppat.1003720-Storelli1] in *Drosophila* is beginning to be unravelled, however, developmental regulation of the immune system remains enigmatic. Ecdysone is a steroid hormone, similar to mammalian estrogens and androgens; peaks in ecdysone titer regulate the major developmental transitions in the fly, including metamorphosis. The pupal stage lasts 4 days from the end of the 3^rd^ larval instar, after which adult flies eclose [@ppat.1003720-Thummel1]. The biologically active form of ecdysone, 20-hydroxyecdysterone (20-E, hereafter referred to as 'ecdysone') coordinates tissue remodelling at metamorphosis [@ppat.1003720-Thummel1]. This hormone activates a nuclear receptor, the Ecdysone Receptor (EcR), which acts as a heterodimer with its partner Ultraspiracle (USP), a homologue of the mammalian Retinoid X Receptor. Together, they activate the transcription of primary response genes, which in turn activate the transcription of a battery of late response genes [@ppat.1003720-Thummel1]. This transcriptional cascade ultimately leads to the induction of both cell death in larval tissues and differentiation and proliferation of the imaginal discs into adult tissues [@ppat.1003720-Thummel1]. In addition, several lines of evidence indicate that ecdysone regulates some aspects of hemocyte behaviour. In the larva, while the majority of hemocytes are in circulation, approximately one third of the total population interacts with tissues, attaching in repeated patches to the dorsal epithelium along the longitudinal axis [@ppat.1003720-Lanot1]. Despite the fact that hemocytes from these patches are mostly immotile at larval stages, it has been noticed that they disperse at metamorphosis [@ppat.1003720-Lanot1]. This observation correlates with the recent finding that *ex vivo*, hemocytes activate motility and morphological changes after metamorphosis [@ppat.1003720-Sampson1]. Cell shape changes can be prematurely triggered in larvae by ecdysone injection [@ppat.1003720-Lanot1]. Last, it is long known that ecdysone treatment is important to potentiate AMP gene expression and phagocytosis after an immune challenge in hemocyte-derived cell culture lines [@ppat.1003720-Dimarcq1]. Although these results suggest an ecdysone-dependent regulation of hemocyte function, *in vivo* evidence of a direct effect of ecdysone signaling on hemocyte behaviour, and its functional relevance, is lacking. Here, we explore the hormonal regulation of *Drosophila* hemocytes at metamorphosis and its impact on *Drosophila* immunity using an *in vivo* approach. We demonstrate that direct activation of ecdysone signaling in hemocytes is necessary to increase their developmental and immune activities at metamorphosis, including phagocytosis. We show that this activation is essential to respond efficiently to and survive pathogenic challenge. Results {#s2} ======= 1. The Ecdysone Receptor is required in the hemocytes for their activation at metamorphosis {#s2a} ------------------------------------------------------------------------------------------- To test whether ecdysone signaling cell-autonomously regulates hemocyte shape changes at metamorphosis, we used the *Hml-(Hemolectin-)ΔGal4* driver [@ppat.1003720-Sinenko1] to specifically express green fluorescent protein (GFP) and dominant-negative constructs of the three known EcR isoforms under the control of the UAS sequence. We imaged hemocytes *ex vivo* by bleeding larvae or pupae at precise time points after puparium formation (APF) ([Fig. 1A--B](#ppat-1003720-g001){ref-type="fig"}) and *in vivo* through the dorsal epidermis ([Fig. 1C--D](#ppat-1003720-g001){ref-type="fig"}). The first hour of the 12h 'prepupal' period is characterized by a translucent pupal case (which then darkens). Control hemocytes (from *HmlΔGal4, UAS-GFP/+* pupae, here after referred to as *HmlΔ*\>GFP) displayed a clear change of morphology over metamorphosis: they became gradually more polarized, with many cytoplasmic protrusions and a higher number of vacuoles, increasing in size dramatically, likely due to their greater spread and many phagocytic vesicles ([Fig. 1A and C](#ppat-1003720-g001){ref-type="fig"}). In striking contrast, hemocytes expressing a dominant negative (DN) form of the EcRB1 isoform (from *HmlΔGal4, UAS-GFP/UAS-EcRB1DN* pupae, here after referred to as *HmlΔ\>EcRB1DN*) did not show any obvious change of size or morphology ([Fig. 1B and D](#ppat-1003720-g001){ref-type="fig"}). Similar results were obtained when we analysed by flow cytometry the properties of the hemocyte population at the onset of pupariation. Control hemocytes displayed a clear shift both in Forward Scatter (FSC) and Side Scatter (SSC), indicating an increase in cell size and granularity, respectively ([Fig. 1E](#ppat-1003720-g001){ref-type="fig"} and [S1](#ppat.1003720.s001){ref-type="supplementary-material"}). In contrast, the FSC and SSC of hemocytes expressing *EcRB1DN* remained stable over metamorphosis and similar to the parameters observed for control hemocytes in late 3^rd^ instar larvae (L3 wandering; L3W; [Fig. 1F](#ppat-1003720-g001){ref-type="fig"} and [S1](#ppat.1003720.s001){ref-type="supplementary-material"}). {#ppat-1003720-g001} In control pupae, the cell shape changes were concomitant with a dispersal of the dorsal patches of hemocytes visualised by *in vivo* cinemicroscopy and cell tracking analysis ([Fig. 1G--J](#ppat-1003720-g001){ref-type="fig"}). These so-called 'sessile hemocytes' changed their behaviour at pupariation and began to actively migrate under the epithelium (see video 1 and [Fig. 1G, I](#ppat-1003720-g001){ref-type="fig"}). This dispersal did not occur when hemocytes expressed EcRB1DN ([Fig. 1H, J](#ppat-1003720-g001){ref-type="fig"} and video 2). In these individuals, hemocytes were more numerous, which correlated with higher proliferation at larval stages ([Fig. S2](#ppat.1003720.s002){ref-type="supplementary-material"}). Moreover, they were unable to migrate away from their original location ([Fig. 1J](#ppat-1003720-g001){ref-type="fig"}). Measurements of cell velocity clearly indicated an impaired motility in EcRB1DN-expressing hemocytes, as compared to control hemocytes ([Fig. 1K](#ppat-1003720-g001){ref-type="fig"}). Importantly, this defect in motility was not due to a decrease in viability, and appeared specific to the pupal stage as neither localisation of hemocytes to the gut in larvae, nor embryonic hemocyte migration were affected when EcRB1DN was expressed specifically in hemocytes ([@ppat.1003720-Wood1], [@ppat.1003720-ZaidmanRemy1]; [Fig. S3](#ppat.1003720.s003){ref-type="supplementary-material"}). In addition, when we expressed EcRB1DN just before the initiation of puparium formation, as opposed to throughout larval stages, we also saw a significant reduction in motility ([Fig. S4](#ppat.1003720.s004){ref-type="supplementary-material"}). The expression of dominant negative constructs for each of the three isoforms of EcR showed a similar phenotype to the EcRB1 isoform ([Fig. 1K](#ppat-1003720-g001){ref-type="fig"}). EcR function in the regulation of hemocyte motility was further confirmed by RNAi ([Fig. 1K](#ppat-1003720-g001){ref-type="fig"}). Last, the expression of *USP* RNAi in the hemocytes led to a similar lack of motility ([Fig. 1K](#ppat-1003720-g001){ref-type="fig"}), indicating that the functional ecdysone receptor complex (EcR/USP) is required for this process. 2. Hemocytes insensitive to ecdysone are unable to perform dead cell phagocytosis during metamorphosis {#s2b} ------------------------------------------------------------------------------------------------------ We next addressed the question of whether ecdysone signaling regulates 'house keeping' functions of hemocytes, required for homeostasis in the organism. During development, one major task for hemocytes is to scavenge and engulf dying cells [@ppat.1003720-Franc1]. We recovered hemocytes at different time points after the initiation of metamorphosis and stained them for DNA and F-actin. Hemocytes from control animals barely contained any stained vesicles at 1 h APF ([Fig. 2A](#ppat-1003720-g002){ref-type="fig"}), but progressively presented a growing number of DNA and/or F-actin-containing vesicles (4 h, 8 h and 18 h APF; [Fig. 2A](#ppat-1003720-g002){ref-type="fig"}). By 8 h APF, most of the control hemocytes had engulfed up to ten apoptotic cells, identifiable by the fragmented aspect of their nuclei, and/or huge pieces of muscles characterized by a strong, striated phalloidin staining. Such vesicles were not seen in the hemocytes expressing *EcRB1DN*, which were stained only for their own nucleus ([Fig. 2A](#ppat-1003720-g002){ref-type="fig"}). We quantified the proportion of cells with phagosomes containing phalloidin-positive muscle fibres and found that the difference between control and *EcRB1DN* hemocytes was striking. In control pupae, we observed a significant increase in dead cell engulfment over time (p\<0.001): just under 20% of hemocytes contained muscle tissue at 1 h APF, rising to approximately 80% of cells by 18 h APF ([Fig. 2B](#ppat-1003720-g002){ref-type="fig"}). In *EcRB1DN* hemocytes the proportion was \<10% at 1 h APF and remained at this low level throughout early pupal stages ([Fig. 2B](#ppat-1003720-g002){ref-type="fig"}; no statistical differences between the five stages). The difference between *EcRB1DN* and control hemocytes was significant at every stage ([Fig. 2B](#ppat-1003720-g002){ref-type="fig"}). Despite this effect on phagocytosis of larval tissue, viability of *HmlΔ\>EcRB1DN* individuals from prepupal stage to adulthood was not affected as compared to control individuals, nor was their duration of metamorphosis ([Fig. S5](#ppat.1003720.s005){ref-type="supplementary-material"}). {#ppat-1003720-g002} 3. Activation of hemocytes at metamorphosis is important for survival after infection {#s2c} ------------------------------------------------------------------------------------- To find out if hemocyte activation by ecdysone is important in critical situations, such as wounding or infection, we tested the ability of *HmlΔ\>EcRB1DN* pupae to resist clean or septic injuries with the Gram-negative bacteria *Erwinia carotovora* (*E.carotovora*) or the Gram-positive bacteria *Enteroccocus faecalis* (*E.faecalis*) when wounded at end of prepupal/early pupal stage (10 h--30 h APF). Interestingly, *HmlΔ\>EcRB1DN* pupae survived as well as control individuals to clean injury, but were significantly less resistant to both types of bacterial infections ([Fig. 3A](#ppat-1003720-g003){ref-type="fig"}). In order to understand better if the difference in resistance between control and *HmlΔ\>EcRB1DN* animals is more pronounced at later stages when hemocytes are fully activated in control individuals, we imposed Gram-negative septic infection on closely-staged pupae at either 3 h APF or 24 h APF. We found that the difference between overall survival of control and *HmlΔ\>EcRB1DN* pupae wounded at 3 h APF was not significant, whereas *HmlΔ\>EcRB1DN* pupae wounded at 24 h APF survived significantly less than controls (p\<0.001; [Fig. 3B](#ppat-1003720-g003){ref-type="fig"}). When we analysed the time of death of infected pupae more precisely, we saw that although the overall survival of *HmlΔ\>EcRB1DN* and control pupae wounded at 3 h APF were not significantly different, *HmlΔ\>EcRB1DN* pupae tended to succumb to infection much sooner than controls ([Fig. 3B](#ppat-1003720-g003){ref-type="fig"}). Altogether, the data suggest that the difference in resistance between *HmlΔ\>EcRB1DN* and control is more pronounced at later stages when control hemocytes are fully activated, although we cannot exclude the possibility that other differences between the two stages could have an effect on the severity of the infection. {#ppat-1003720-g003} We next tested whether feeding larvae with bacteria ('natural infection') would affect their survival over subsequent metamorphosis. Control individuals were not affected by the oral infection with *E. carotovora*, and eclosed normally ([Fig. 3C](#ppat-1003720-g003){ref-type="fig"}). In contrast, we observed a significant decrease in the survival of *HmlΔ\>EcRB1DN* pupae ([Fig. 3C](#ppat-1003720-g003){ref-type="fig"}). For both septic and natural infections, the observed decreased in survival was similar to that caused by genetic ablation of hemocytes (by expression of the pro-apoptotic gene Bax in hemocytes [@ppat.1003720-Defaye1]; see "Phago^less^" survival in [Fig. 3A and C](#ppat-1003720-g003){ref-type="fig"}). Thus hemocyte activation at metamorphosis is essential for the efficient participation of the cellular immune response in resistance to bacterial infections. 4. Hemocytes require ecdysone signaling to become wound-responsive at metamorphosis {#s2d} ----------------------------------------------------------------------------------- Because pupae in which hemocytes are not activated succumb to infections more often than controls, we surmise that ecdysone signaling has a role in the regulation of the hemocyte response to infections. Therefore, we analysed the three main known responses of hemocytes to infections to determine whether they were affected by ecdysone signaling: i) wound response, ii) participation in AMP systemic expression (inter-organ signaling) and iii) phagocytosis. Upon injury, hemocytes are recruited to damaged tissues and wounded epithelia. This may be important to limit infection in the case of septic injury: once recruited to the wound, hemocytes phagocytose bacteria and prevent their spreading inside the organism. In the embryo, hemocyte recruitment involves active migration toward the wound by chemotaxis [@ppat.1003720-Stramer1], [@ppat.1003720-Wood1]. In the larva by contrast, hemocytes are described as being recruited passively from the circulation by 'capture' at the wound site [@ppat.1003720-Babcock1]. Hemocytes attached under the epithelium do not respond to wounds at larval stages, even to those in their close vicinity [@ppat.1003720-Babcock1]. Interestingly, when we wounded the dorsal epithelium of prepupae using a laser-ablation system (which allows immediate high resolution imaging after wounding and tracking of hemocyte movement), we observed that most recruited cells had not come from circulation but by active migration under the epithelium, often from the so-called 'sessile patches' ([Fig. 4A--C](#ppat-1003720-g004){ref-type="fig"} and video 3). Analysis of hemocyte tracks after wounding confirmed their directed migration towards the wound ([Fig. 4D](#ppat-1003720-g004){ref-type="fig"}; compare with the center of mass close to (0;0) in the absence of a wound, indicative of a random migration, [Fig. 1I](#ppat-1003720-g001){ref-type="fig"}). Importantly, the capacity of hemocytes to respond to wounds clearly increased over early prepupal stages. Indeed, hemocyte recruitment rate increased with pupal age: after wounding prepupae at 120 min APF, we observed a higher rate of recruitment of hemocytes to wounds as compared with prepupae wounded at 55 min APF ([Fig. 4E](#ppat-1003720-g004){ref-type="fig"}). Strikingly, *EcRB1DN*-hemocytes remained unresponsive to wounds and very few were recruited to the sites of damage ([Fig. 4F--H](#ppat-1003720-g004){ref-type="fig"} and video 4). {ref-type="supplementary-material"}). Tracks of individual cells measured over 80 min, 70 min after wounding (wound performed at 1h30 APF) are superimposed at (0;0) in the tracking plot; all trajectories were rotated to maintain their relative positions towards the wound center (D). The center of mass of all tracks, indicated with red cross, is shifted towards the wound (positive *x*-axis) and significantly different from (0;0) (p = 0.01). Recruitment rate (cell/min) is significantly higher in pupae wounded at 120 min APF than in pupae wounded 55 min APF (E). Profile of hemocyte recruitment to epithelial wounds in control (brown) or *HmlΔ\>EcRB1DN* pupae (blue) (F). Tracks corresponding to trajectories of *EcRB1DN*-hemocytes after epithelial wounding were superimposed and rotated as above in tracking plot (G); note that the center of mass (red cross) remains close to (0;0). Rate of recruitment (cell/min) in *HmlΔ\>EcRB1DN* pupae and controls wounded 1h30 APF (H). Mean and SEM are displayed on graphs.](ppat.1003720.g004){#ppat-1003720-g004} Thus, our data reveal that the mode of recruitment of hemocytes to wounds is modified at pupal stage, switching from a passive 'capture' of circulating cells in larvae [@ppat.1003720-Babcock1], to active migration of hemocytes toward the wounds in pupae. This activation is lost in hemocytes insensitive to ecdysone. 5. Ecdysone signaling is required for bacterial phagocytosis and to fight infection at metamorphosis {#s2e} ---------------------------------------------------------------------------------------------------- Hemocytes are involved in the systemic induction of the Imd pathway after oral infection in the larva [@ppat.1003720-Charroux1], [@ppat.1003720-Basset1], [@ppat.1003720-Wu1], [@ppat.1003720-Foley1], and therefore we next tested the hypothesis of an ecdysone-dependent regulation of "gut-to-fat-body" communication by hemocytes. However, when we analysed the expression of four AMPs (*Diptericin*, *Attacin*, *Metchnikowin* and *Drosomycin*) in prepupae after oral infection with *E. carotovora* at larval stage ([Fig. S6](#ppat.1003720.s006){ref-type="supplementary-material"}), we found that *HmlΔ\>EcRB1DN* prepupae induce a systemic humoral response similar to that of control. Thus, our data indicate that ecdysone signaling in hemocytes is not required for inter-organ communication after oral infection. In addition, we analyzed the local epithelial response to oral infection by examining expression of *Diptericin* specifically in the gut, and found that this response is also unaffected in *HmlΔ\>EcRB1DN* prepupae ([Fig. S7](#ppat.1003720.s007){ref-type="supplementary-material"}). We wondered if phagocytic activity, on the other hand, would be ecdysone-regulated at metamorphosis and could account for the observed decrease in survival to infection. To first address this question, we bled late L2/young L3 larvae or young prepupae, incubated the retrieved hemocytes with a high concentration of pHRodo bacterial particles (that emit a strong fluorescent signal only at low pH, i.e. in the acidic environment of the phagosome) and assessed the ability of hemocytes to perform phagocytosis using flow cytometry ([Fig. 5A--B](#ppat-1003720-g005){ref-type="fig"}). Strikingly, hemocytes presented a clear shift in their ability to perform phagocytosis from L2/L3 to prepupal stages ([Fig. 5A](#ppat-1003720-g005){ref-type="fig"}); the prepupal hemocyte phagocytic index (number of phagocytosing hemocytes multiplied by mean pHRodo intensity per hemocyte)/total number of hemocytes) was almost four times the index of L2/L3 larval hemocytes ([Fig. 5B](#ppat-1003720-g005){ref-type="fig"}). Moreover, *EcRB1DN*-hemocytes did not present such a shift and maintained the same phagocytic ability at prepupal stage as at L2/L3 larval stages ([Fig. 5A--B](#ppat-1003720-g005){ref-type="fig"}). {#ppat-1003720-g005} To confirm this phenotype *in vivo*, we injected pHRodo bacterial particles in control and *HmlΔGal4\>EcRB1DN* prepupae and imaged hemocytes live through the dorsal epithelium ([Fig. 5C](#ppat-1003720-g005){ref-type="fig"}). While under these conditions, all hemocytes in control prepupae had performed phagocytosis of numerous particles, only very few *EcRB1DN*-hemocytes contained particles and then in low numbers ([Fig. 5D--E](#ppat-1003720-g005){ref-type="fig"}). We complemented this analysis with a flow cytometry method that combines *in vivo* assay with high-throughput quantitative analysis. One hour after injection with pHRodo, hemocytes were retrieved from prepupae and analysed for their fluorescence by flow cytometry. In accordance with the results obtained with confocal analysis, fewer *EcRB1DN*-hemocytes were pHRodo-positive as compared to control (40% versus 88%; [Fig. 5F](#ppat-1003720-g005){ref-type="fig"}). Those few hemocytes positive for pHRodo had a lower fluorescence intensity than control hemocytes ([Fig. 5F](#ppat-1003720-g005){ref-type="fig"}), contributing to a global decrease in phagocytic index ([Fig. 5G](#ppat-1003720-g005){ref-type="fig"}). Interestingly, hemocytes expressing dominant negative forms of any of the three EcR isoforms were affected in their ability to perform phagocytosis, as well as hemocytes expressing *EcR RNAi* ([Fig. 5G](#ppat-1003720-g005){ref-type="fig"}). As pHRodo particles only fully emit fluorescence once in the mature phagosome, the above results could also be indicative of a defect of *EcRDN*-hemocytes in phagosome maturation. To test this hypothesis we analysed the ability of control and *EcRB1DN*-hemocytes to bind and engulf live *Escherichia coli* (*E.coli*). *E.coli* expressing the red fluorescent protein (*E.coli-RFP*) were injected in prepupae and phagocytosis was monitored *in vivo* (video 5 and 6). *EcRB1DN*-hemocytes were significantly impaired in their ability both to bind (attachment index: the average number of bacteria bound per hemocyte; [Fig. 5H](#ppat-1003720-g005){ref-type="fig"}) and engulf bacteria (internalisation index: the average number of bacteria ingested per hemocyte; [Fig. 5I](#ppat-1003720-g005){ref-type="fig"} and [S7](#ppat.1003720.s007){ref-type="supplementary-material"}). Altogether, our results demonstrate that EcR-dependent activation of hemocytes is required for optimal phagocytic activity of hemocytes at metamorphosis, and modulate their ability to both bind and internalise bacteria. 6. EcR transcriptionally regulates genes involved in phagocytosis {#s2f} ----------------------------------------------------------------- Our phenotypic analysis has revealed the importance of ecdysone signaling in the regulation of several hemocyte functions, including bacteria phagocytosis. To better understand the molecular basis of the ecdysone-dependent regulation of hemocyte biology, we performed a hemocyte-specific transcriptomic analysis, combining FACS and Affymetrix microarrays. We first compared the transcriptome of "control" hemocytes retrieved from *w; Hml\>GFP/+* 3^rd^ instar larvae and prepupae. We selected late feeding 3^rd^ instar larvae prior to the peak of ecdysone that induces metamorphosis, and compared them with hemocytes retrieved from 1 h--2 h APF prepupae, the stage we have used for the majority of our phenotypic analyses. This allowed us to determine which genes are induced or repressed at metamorphosis in hemocytes. Comparison with *EcRDN* hemocytes allowed us to determine which of these genes are dependent on EcR signaling. A high number of genes are transcriptionally regulated in hemocytes at the onset of metamorphosis (3331 genes, among which 659 are up-regulated more than two fold; p\<0.001; see repartition by fold change in [Fig. 6A](#ppat-1003720-g006){ref-type="fig"}. The full microarray data can be found in [Table S2](#ppat.1003720.s010){ref-type="supplementary-material"} and on GEO database accession GSE49326, <http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE49326>). Strikingly, our data indicate that most of the changes in gene expression are EcR-dependent at this stage (1955 genes out of 3331 are EcR-regulated; p\<0.001; see repartition by fold change in [Fig. 6A](#ppat-1003720-g006){ref-type="fig"} and the list of genes in [Table S1](#ppat.1003720.s009){ref-type="supplementary-material"}). In fact, 83% of genes that present a fold change in expression greater than 2 at metamorphosis are regulated by EcR. The microarray analysis was further confirmed by hemocyte-specific quantitative PCR of 12 genes, whose expression was clearly up-regulated in control hemocytes at metamorphosis, and significantly lower in *EcRDN*-expressing hemocytes at the same stage ([Fig. 6B](#ppat-1003720-g006){ref-type="fig"}). These data confirm and explain the huge impact of ecdysone signaling on hemocyte biology at the onset of metamorphosis. Interestingly, we discover that in prepupal hemocytes, ecdysone signaling up-regulates the expression of numerous genes involved in bacteria phagocytosis. Indeed, we found that 35 genes known to be involved in phagocytosis are significantly up-regulated in hemocytes from control pupae and significantly less expressed in hemocytes expressing *EcRDN* ([Table 1](#ppat-1003720-t001){ref-type="table"}). Among them, several genes encode phagocytic bacterial receptors (*Nimrod*, *Dscam*, *PGRP*-*LC*), a receptor for dead cell phagocytosis (*crq*), or are involved in the actin remodelling required for this process (such as *Rac2* and *SCAR*). As loss of expression of several of these genes has been shown to have a dramatic effect on phagocytosis, these molecular data are sufficient to explain how insensitivity to ecdysone in hemocytes results in the strong phagocytosis phenotype we describe. Similarly, we found several genes involved in cell motility or cell shape regulation to be up-regulated in the control hemocytes at metamorphosis, and not in *EcRDN*-expressing hemocytes ([Table S3](#ppat.1003720.s011){ref-type="supplementary-material"} and [S4](#ppat.1003720.s004){ref-type="supplementary-material"}, respectively). These genes represent good candidates responsible for the phenotypes observed in our analyses; additionally, it is likely that this data contains further genes involved in phagocytosis, cell motility or cell shape as yet unrevealed. {ref-type="supplementary-material"} for gene list). (B) qPCR confirmation of the expression of 12 genes from the microarrays. qPCR was performed on RNA extracted from FACS-sorted hemocytes. The results displayed represent the mean and SEM of three biological repeats; samples were independent from samples used for microarrays. Statistical analysis was performed using one-way ANOVA and Tukey-Kramer post-test. For comparison between control larval hemocytes and control pupal hemocytes, as well as between control pupal hemocytes and EcRDN pupal hemocytes, significance are indicated on the graph. There was no significant difference between control larval hemocytes and EcRDN pupal hemocytes except for GstS1 (\*), NimC3 (\*), Mmp2 (\*) and Dscam (\*\*\*).](ppat.1003720.g006){#ppat-1003720-g006} 10.1371/journal.ppat.1003720.t001 ###### EcR transcriptionally regulates genes involved in phagocytosis. {#ppat-1003720-t001-1} Gene Symbol Gene Accession Control pupae vs control larvae Control pupae vs EcRDN pupae References on phagocytosis phenotype ------------- ---------------- --------------------------------- ------------------------------ -------------------------------------- ---------- ------------------------------------------------------------------------------------------------------------------------------------------------ NimC2 FBgn0028939 57.08 0 −4.40 0 [@ppat.1003720-Kurucz1] (Prediction) **Eip93F** FBgn0264490 26.40 0 −2.85 8.00E-06 [@ppat.1003720-Thummel1] (Prediction) scb FBgn0003328 25.72 0 −1.98 1.30E-05 [@ppat.1003720-StroscheinStevenson1], [@ppat.1003720-Nonaka1] Eip78C FBgn0004865 11.88 0 −3.80 0 [@ppat.1003720-Stuart1] Tep2 FBgn0041182 9.19 0 −2.14 8.00E-06 [@ppat.1003720-StroscheinStevenson1] **NimC4** FBgn0260011 6.26 0 −5.34 0 [@ppat.1003720-Kurucz1] (Prediction) **NimC3** FBgn0001967 5.83 6.90E-05 −2.54 2.20E-05 [@ppat.1003720-Kurucz1] (Prediction) **Mmp2** FBgn0033438 5.81 8.00E-06 −1.68 5.50E-05 [@ppat.1003720-Stuart1] **Dscam** FBgn0033159 3.96 1.10E-05 −2.75 1.80E-05 [@ppat.1003720-Watson1] l(1)G0232 FBgn0028341 3.10 2.00E-06 −1.58 0.000193 [@ppat.1003720-StroscheinStevenson1] yip2 FBgn0040064 3.01 7.00E-05 −1.70 0.000388 [@ppat.1003720-Stuart1] CG8449 FBgn0038129 2.74 5.00E-06 −1.59 0.000972 [@ppat.1003720-Stuart1] CG3638 FBgn0261444 2.67 2.00E-06 −1.74 2.40E-05 [@ppat.1003720-StroscheinStevenson1] **crq** FBgn0015924 2.60 5.30E-05 −1.75 4.40E-05 [@ppat.1003720-Franc2], [@ppat.1003720-Stuart3] mv FBgn0265988 2.51 0.00043 −2.55 3.00E-06 [@ppat.1003720-Rahman1] CG16791 FBgn0038881 2.39 3.00E-06 −1.60 0.000161 [@ppat.1003720-StroscheinStevenson1] flr FBgn0260049 2.27 5.00E-06 −1.44 0.000183 [@ppat.1003720-StroscheinStevenson1] cpb FBgn0011570 2.25 5.00E-06 −1.48 0.00065 [@ppat.1003720-Philips1] RhoGAP71E FBgn0036518 2.19 8.00E-06 −1.48 0.000137 [@ppat.1003720-Ramet1] Rac2 FBgn0014011 2.03 9.00E-06 −1.43 0.000417 [@ppat.1003720-Stuart1], [@ppat.1003720-Sampson1], [@ppat.1003720-Philips1], [@ppat.1003720-StroscheinStevenson1], [@ppat.1003720-AvetRochex2] Arpc5 FBgn0031437 1.93 6.80E-05 −1.44 0.000494 [@ppat.1003720-Philips1], [@ppat.1003720-StroscheinStevenson1] Snx3 FBgn0038065 1.93 1.80E-05 −1.35 0.00068 [@ppat.1003720-StroscheinStevenson1] CG13887 FBgn0035165 1.92 1.90E-05 −1.41 0.000501 [@ppat.1003720-Stuart1] RN-tre FBgn0020620 1.87 3.30E-05 −1.69 0.000108 [@ppat.1003720-Stuart1], [@ppat.1003720-Ramet1] not FBgn0013717 1.84 6.20E-05 −1.41 0.000369 [@ppat.1003720-Stuart1] **PGRP-LC** FBgn0035976 1.81 6.50E-05 −1.75 0.000359 [@ppat.1003720-Ramet1] Mekk1 FBgn0024329 1.81 4.10E-05 −1.46 0.000132 [@ppat.1003720-Stuart1] kuz FBgn0259984 1.79 4.80E-05 −1.36 0.000536 [@ppat.1003720-StroscheinStevenson1] SCAR FBgn0041781 1.63 1.00E-04 −1.34 0.00084 [@ppat.1003720-StroscheinStevenson1] Vps35 FBgn0034708 1.57 0.00015 −1.37 0.000462 [@ppat.1003720-Stuart1] Traf4 FBgn0026319 1.57 0.00059 −2.26 3.00E-06 [@ppat.1003720-StroscheinStevenson1] shark FBgn0015295 1.53 0.00012 −1.36 0.00073 [@ppat.1003720-Doherty1], [@ppat.1003720-Ziegenfuss1] mib2 FBgn0086442 1.53 0.00017 −1.45 0.000534 [@ppat.1003720-Stuart1] put FBgn0003169 1.51 0.00013 −1.39 0.000902 [@ppat.1003720-Stuart1] Among the genes significantly up-regulated in control hemocytes at metamorphosis (p\<0.001; linear FC and *p* values indicated on the table) and significantly down-regulated in EcRB1DN-expressing hemocytes (p\<0.001; linear FC and *p* values indicated on the table), 35 genes were annotated by Flybase with the GO term phagocytosis or found in S2 cell-based RNAi screens for phagocytosis phenotype [@ppat.1003720-Stuart1], [@ppat.1003720-Ulvila1], [@ppat.1003720-Philips1]--[@ppat.1003720-StroscheinStevenson1]. FC: linear Fold Change; in bold: genes for which expression pattern was confirmed by qPCR. All p values are \<0.001, no FC threshold was applied. Discussion {#s3} ========== Here we have demonstrated the cell-autonomous regulation of *Drosophila* phagocytes by a hormonal signal at metamorphosis. Our study provides *in vivo* evidence that the steroid hormone ecdysone regulates both developmental and immune functions of hemocytes, such as phagocytosis of dying cells and bacteria, and the acquisition of wound responsiveness at metamorphosis. Most importantly, non-activated hemocytes are unable to efficiently fight infections, reflecting the functional importance of this hormonal regulation for the organism. Using an *in vivo* genetic approach to block EcR signaling specifically in hemocytes, we have shown that ecdysone directly regulates their cell shape. Moreover, our data indicates that ecdysone regulates the onset of hemocyte motility and dispersal at metamorphosis, reflecting its function in border cell motility during oogenesis [@ppat.1003720-Bai1]. Our microarray data reveal that EcR up-regulates the expression of several genes functioning in cell motility or cell shape regulation [@ppat.1003720-Kiger1], [@ppat.1003720-Rohn1], which could account for these phenotypes ([Table S3](#ppat.1003720.s011){ref-type="supplementary-material"} and [S4](#ppat.1003720.s004){ref-type="supplementary-material"}). Arguably, migration of hemocytes between tissues is required for clearing dying larval tissues during the pupal period. We show that hemocytes expressing the *EcRDN* construct do not engulf dead cells, which is potentially a consequence of impaired phagocytosis, motility, or a combination of both, although we cannot distinguish between these possibilities. Ecdysone has previously been shown to induce the expression in the hemocyte-derived *mbn2* cell line of *croquemort* (*crq*; [@ppat.1003720-Dimarcq1]), a gene encoding a receptor for apoptotic cells in the embryo [@ppat.1003720-Franc1]. *crq* was identified in our microarray analysis as showing EcR-dependent up-regulation at metamorphosis, and this was confirmed by qPCR, where *crq* expression is almost completely suppressed in *EcRDN*-expressing pupal hemocytes ([Fig. 6B](#ppat-1003720-g006){ref-type="fig"}). The impaired expression of *crq* in *EcRDN* hemocytes likely contributes to their deficiency in apoptotic cell phagocytosis. Functionally, the regulation of hemocytes by ecdysone, which is the coordinator of larval tissue apoptosis, may be a smart way for the fly to synchronize its macrophage scavenging activity with the moment it is most needed, at metamorphosis. Surprisingly, we did not observe any gross developmental consequences of the loss of this function, whereby *HmlΔGal4\>EcRB1DN* individuals completed metamorphosis without delay. This is in agreement with studies showing that under sterile conditions, pupae lacking hemocytes altogether progress normally through metamorphosis [@ppat.1003720-Charroux1], [@ppat.1003720-Defaye1]. It suggests that dead cells might be engulfed by other, non-professional phagocytes (e.g. neighbor cells as reported for tumorigenesis), cleared up by other unidentified means, or simply tolerated, in the absence of functional hemocytes. Furthermore, we show that the activation of hemocyte motility at metamorphosis also correlates with a change in their response to induced epithelial damage. While in the larva hemocytes are passively recruited to wounds from circulation [@ppat.1003720-Babcock1], we demonstrate that in the pupa they actively migrate to damaged tissues. Induction of epithelial wounds at different times APF demonstrated that active wound responsiveness is progressively acquired at metamorphosis. In agreement with previous *ex vivo* analysis [@ppat.1003720-Sampson1], our data highlights an intriguing plasticity of hemocytes to adapt their migratory activity and their response to wounds throughout development: chemotaxis in embryos and pupae ([@ppat.1003720-Wood1] and this study) versus passive circulation and 'capture' to wounds in larvae [@ppat.1003720-Babcock1]. This correlates with the observation that, although the heart is beating in a 20 h APF-old pupa, hemocytes are not propelled in the hemolymph by the heartbeat, but maintain a slow, steady, active migration on tissues (video 7). Most importantly, our study provides the first *in vivo* evidence of hormonal regulation of the *Drosophila* cellular response to bacterial challenge. With both *ex vivo* and *in vivo* data, we have demonstrated an important role for EcR in the up-regulation of hemocyte phagocytic activity at metamorphosis ([Fig. 5](#ppat-1003720-g005){ref-type="fig"}). How does ecdysone signaling regulate phagocytosis? Previous studies in hemocyte-derived cell lines have shown that ecdysone treatment increases the transcription of some immune-related genes encoding AMPs and immune receptors such as Crq [@ppat.1003720-Dimarcq1], [@ppat.1003720-Flatt1], [@ppat.1003720-Zhang1]. Here, using a tissue-specific, whole genome transcriptomic approach, we demonstrate that many genes are regulated by ecdysone signaling in hemocytes at metamorphosis. This analysis reveals the molecular regulation behind the observed phenotypes and allows for the identification of candidate effector genes. For example, 35 genes up-regulated by EcR at metamorphosis have been previously attributed a function in phagocytosis. These genes encode proteins involved in different steps of the phagocytosis process, such as recognition (e.g. the receptors *PGRP-LC*, *croquemort*, and *Nimrod* family members, *Dscam* and *scab*), or cytoskeletal rearrangements required for the engulfment step (e.g. *RhoGAP71E*, *Rac2*, *Arpc5* and *SCAR*). Interestingly, PGRP-LC (FC 1.8 by microarray, 3.9 by qPCR) was recently shown to be induced in ecdysone-treated S2 cells [@ppat.1003720-Rus1]. It appears that ecdysone can regulate the phagocytosis process at different levels, which may be necessary to co-ordinate the ability of hemocytes to recognize and engulf their target. Moreover, genes regulated by ecdysone signaling can be implicated in more than one process, for example phagocytosis and AMP expression (e.g. PGRP-LC), or phagocytosis and cell migration (e.g. SCAR); this may contribute to synchronisation of different hemocyte immune functions. The functional relevance of increased cellular immune activity at metamorphosis is an intriguing question. Recent studies of the contribution of cellular immunity to *Drosophila* defenses have revealed that flies in which hemocytes are genetically ablated present a high lethality at metamorphosis [@ppat.1003720-Charroux1], [@ppat.1003720-Defaye1]. This is likely the result of opportunistic bacterial infections, as feeding antibiotics was sufficient to restore wild-type viability. We did not observe such lethality under normal conditions when expressing *EcRDN* in hemocytes; however in our hands *Phago^less^* lethality in absence of infection is also lower than that previously described [@ppat.1003720-Defaye1]. This suggests that our fly strains and fly food do not harbor the same bacterial types as those used in previous studies, leading to distinct opportunistic infection scenarios. Nevertheless, our data indicate a significant lethality of *HmlΔ\>EcRDN* pupae not only after septic injury with *E. faecalis* or *E. carotovora*, but also after oral infection at larval stages with *E. carotovora*, a bacterium that is not usually lethal in wild-type individuals. This lethality is quite dramatic considering only hemocytes express the transgene, and is similar to the lethality in hemocyte-ablated individuals ([@ppat.1003720-Charroux1], [@ppat.1003720-Defaye1] and this study). It indicates that ecdysone regulation is essential for hemocyte immune functions and survival after infection. Metamorphosis may represent a stage of predisposition to opportunistic oral infection, as the larval midgut is replaced by the adult intestinal epithelium. We speculate that histolysis of the gut could release bacteria from the lumen into the body cavity; active hemocytes may be required to limit the spreading of bacteria from temporary weak points in the epithelium. We have shown that *HmlΔ\>EcRDN* prepupae induce a normal intestinal and systemic humoral immune response after being orally infected at larval stage ([Fig. S6](#ppat.1003720.s006){ref-type="supplementary-material"} and [S7](#ppat.1003720.s007){ref-type="supplementary-material"}). In the case of both septic injury and oral infection, it is therefore likely that the main cause of decreased survival in *HmlΔ\>EcRDN* pupae is their striking hemocyte phagocytosis phenotype, possibly in combination with lack of motility, inability to chemotax to damaged tissue or other potential uncharacterized hemocyte defects. The synchronization of multiple processes is a fundamental requirement for successful development, and likely to rely on hormonal signaling. Altogether, our data reveal the importance of steroid hormone signaling in the synchronization of development and immunity in *Drosophila*, by ecdysone-dependent activation of hemocytes at pupariation. Rus *et al*. have recently shown [@ppat.1003720-Rus1], that ecdysone signaling affects the humoral response through regulation of PGRP-LC expression. Interestingly, they observed an impact of this regulation on the ability of adult flies to survive infection, indicating that ecdysone regulation of immunity extends beyond metamorphosis. In humans, hormonal activation of macrophages underpins various cancer pathologies [@ppat.1003720-Harkonen1] and is therefore highly relevant in clinical terms. It is also generally accepted that steroid hormones impact immunity in mammals. For example, glucocorticoids are commonly used in pharmacology for their anti-inflammatory properties [@ppat.1003720-Necela1]. However, their regulation of the immune response is complex, as they can also enhance the immune response [@ppat.1003720-Dhabhar1], [@ppat.1003720-Goulding1]. More generally, steroid hormones\' specific action on monocytes is still not very well documented, mainly due to the complexity of mammalian systems and experimental limitations. Elucidating mechanisms for steroid hormone regulation of cellular immunity (e.g. [@ppat.1003720-Routley1]) will be essential for a full understanding of sex differences in immunity and inflammation. Here, in a simpler model system, we used an approach that allows *in vivo* and cell autonomous analysis of the hormonal regulation of innate immune cells. We think that in the future, the potential for genetic manipulation in the *Drosophila* model combined with the *in vivo* assays and transcriptomic data developed in this study should aid in deciphering the molecular mechanisms at work behind phagocyte activation by steroids, both in terms of cell migration and immunity *per se*. Materials and Methods {#s4} ===================== *Drosophila* stocks {#s4a} ------------------- We used the following stocks from Bloomington Stock Centre: *w^1118^*; *HmlΔ-GAL4,UAS-GFP* (30142); *UAS-EcRA-DN* (9451, 9452); *UAS-EcRB1-DN* (6869, 6872); *UAS-EcRB2-DN* (9449, 9450); *UAS-EcR-RNAi* (9326, 29374); *USP-RNAi* (27258); and *UAS-Bax/CyO* from [@ppat.1003720-Defaye1]. Flies were kept on standard fly food (VDRC recipe). [Results](#s2){ref-type="sec"} obtained with DN^F645A^ are shown in all figures except when specified. Imaging {#s4b} ------- Pupae were mounted and hemocytes imaged live through cuticle as previously described [@ppat.1003720-Moreira1]. Images were acquired using a confocal laser line-scanning (LSM 5 Live; Carl Zeiss) or point-scanning (LSM710; Carl Zeiss) microscope (40× oil objective). Mean velocity/min of hemocytes tracked manually over 10 min using the ImageJ plugin 'Manual Tracker' was used as a measure of motility. At least 40 cells were tracked per condition, corresponding to at least 4 independent experiments. For tracking plots, trajectories of 20 cells were transformed and analysed using the ImageJ plugin 'Chemotaxis Tool' as previously described [@ppat.1003720-Goulding1]. Laser wounding {#s4c} -------------- Wounds were made in prepupae in the dorsal thoracic epithelium using a UV laser controlled by galvanometric mirrors. Laser was emitted at 355 nm as a 20 second train of 1 second square pulses and was fired 9 times to form a square wound of approximately 40×40 µm. The wound was made in a region lateral to the sessile patches, just below the pupal epithelium and always included a tracheal branch. Hemocytes were tracked and hemocytes recruited to the wound over time were counted; the rate of recruitment was calculated as the number of hemocytes recruited to the wound over 150 min after wounding. Morphometric analysis by flow cytometry {#s4d} --------------------------------------- Hemocytes were bled into PBS and analysed for their FSC-Area and SSC-Area with a CyAn ADP flow cytometry Analyzer (Dako Cytomation/Beckman Coulter). The hemocyte population was defined as GFP positive cells using hemocytes from *w^1118^* as a negative control to establish the threshold. Infection, survival and phagocytosis assay {#s4e} ------------------------------------------ For septic injury, pupae were pricked in the lower abdomen with a needle dipped in bacterial culture. *E.carotovora* (strain *E.c.c.15*) were grown overnight in Luria Bertani (LB) at 29°C and adjusted to O.D. 100. *E. faecalis* were grown overnight in LB at 37°C and adjusted to O.D. 5. Clean pricking with sterilized needle was used as a control. Oral infection of larvae was performed as described ([@ppat.1003720-Romeo1] and see SI for more details) by feeding larvae with a mixture of banana and *E.c.c.15*. Pupae were maintained at 29°C and eclosion of adults was scored after 5 days. For [Fig. 3A](#ppat-1003720-g003){ref-type="fig"}, the stage of death was determined by careful examination and classification between four approximate stages: early pupae (from approx. 12 h to 36 h), intermediate 1 (from approx. 36 h to 54 h), intermediate 2 (from approx. 36 h to 72 h) and late pupal stage (from approx. 72 h to 90 h), corresponding to pupae with the appearance they adopt just before eclosion. For *ex vivo* phagocytosis assay, late L2/young L3 larvae, or white prepupae, were bled into 400 µL PBS (pH 7.4) and 8 µL pHRodo Red *E. coli* BioParticles (1 mg/mL; molecular Probes) were added. Phagocytosis was assayed 30 min later by flow cytometry. For *in vivo* assay, 80 nL of pHRodo particles were injected in prepupae 1h30 APF, using a nanoinjector (Nanoliter 2000, World Precision Instruments). Prepupae were either imaged 1 h after injection, or bled into PBS and analyzed by flow cytometry with a modular Flow Cytometer (MoFlo; Dako Cytomation), using 561 nm and 488 nm lasers. pHRodo particles in solution, *w^1118^* non-injected flies, *HmlΔ\>GFP* non-injected flies and *w^1118^* flies injected with pHRodo particles were used to define the thresholds for GFP and phagocytosed red-particle emission. The experiment was repeated three times, using at least 10 prepupae (roughly 10 000 hemocytes) per genotype and experiment. For *in vivo* phagocytosis assay with live bacteria, 50.6 nL of *E. coli-RFP* (at O.D. 2 after dilution in PBS) were injected in prepupae 1h30 APF. After 10 min recovery, prepupae were mounted, and imaged at 20 min post-injection. The central-most sessile patch was imaged at 40× to a depth of 30 µm. Phagocytosis was scored by eye from z-stacks; *E.coli-RFP* were scored as either attached or internalized and phagocytic index was calculated using total hemocyte number. Phalloidin and DAPI staining; measure of dead cell phagocytosis {#s4f} --------------------------------------------------------------- Circulating hemocytes were retrieved by bleeding animals into 20 µL of PBS. After 20 min, they were fixed in a 4% formaldehyde solution for 15 min, then washed in PBST (0.1% Triton X-100 in PBS) and incubated during 30 min at room temperature in a blocking solution (PBST +1% Bovine Serum Albumin). Bleeds were incubated with phalloidin (1∶200 µL; Invitrogen) and DAPI (1 µm/mL) for 30 min, then washed for 1 h in PBS and mounted in mounting medium. Quantification of hemocytes phagocytosing muscle tissue was measured as the percentage of hemocytes positive for phalloidin-positive vacuoles among total hemocyte number. Statistical analysis {#s4g} -------------------- Statistical significance was defined by pair-wise comparison to controls using the Mann--Whitney U test (non parametric), *t*-test (parametric), or Wilcoxon test (survival experiments). Indicated *p*-values are two-tailed. Calculations and graphs were produced using Excel (Microsoft) and Prism (Graphpad). Symbols in the figures: \**p\<0.05*; \*\**p\<0.01*; \*\*\**p\<0.001*; ns: not significant. FACS-sorting, RNA isolation, target synthesis and hybridization to Affymetrix Array Strips {#s4h} ------------------------------------------------------------------------------------------ Hemocytes were isolated by FACS from 3rd instar larvae (at the late feeding stage) and prepupae (1 h to 2 h APF) corresponding to two different genotypes: *w; HmlΔGal4, UAS-GFP/+* (as control), *w; HmlΔGal4; UAS-GFP/UAS-EcRB1DN^W650^*. For each of the four conditions we performed three biological replicates. Hemocytes were sorted based on their fluorescence and directly into the lysis buffer prior to total RNA extraction using the RNeasy Plus Micro Kit (Qiagen). Concentration and purity was determined by spectrophotometry and integrity was confirmed using an Agilent 2100 Bioanalyzer with a RNA 6000 Nano Assay (Agilent Technologies). RNA was processed for use on Affymetrix (Santa Clara) Drosophila Gene 1.1 ST Array Strips using the Ambion WT Expression Kit (Life Technologies) and Affymetrix GeneChip WT Terminal Labeling Kit, according to the manufacturer\'s protocols. Briefly, 100 ng of total RNA containing spiked in Poly-A RNA controls (GeneChip Expression GeneChip Eukaryotic Poly-A RNA Control Kit; Affymetrix) was used in a reverse transcription reaction (Ambion WT Expression Kit) to generate first-strand cDNA. After second-strand synthesis, double-stranded cDNA was used in an *in vitro* transcription (IVT) reaction to generate cRNA (Ambion WT Expression Kit). 15 µg of this cRNA was used for a second cycle of first-strand cDNA synthesis (Ambion WT Expression Kit). 5.5 µg of single stranded cDNA was fragmented and end-labeled (GeneChip WT Terminal Labeling Kit; Affymetrix). Size distribution of the fragmented and end-labeled cDNA, respectively, was assessed using an Agilent 2100 Bioanalyzer with a RNA 6000 Nano Assay. 3.5 µg of end-labeled, fragmented cDNA was used in a 150-µl hybridization cocktail containing added hybridization controls (GeneAtlas Hybridization, Wash, and Stain Kit for WT Array Strips, Affymetrix), of which 120 µl were hybridized on array strips for 20 h at 48°C. Standard post hybridization wash and double-stain protocols (GeneAtlas Hybridization, Wash, and Stain Kit for WT Array Strips, Affymetrix) were used on an Affymetrix GeneAtlas system, followed by scanning of the array strips. Microarray data analysis {#s4i} ------------------------ The 12 scanned arrays were analyzed first with Affymetrix Expression Console software using RMA to obtain expression values and for quality control. Control probe sets were removed and log2 expression values of the remaining 15391 transcripts were imported into Chipster 2.4 [@ppat.1003720-Kallio1]. Differential expression was determined by empirical Bayes two-group test [@ppat.1003720-Smyth1] with Benjamini-Hochberg multiple testing correction and a p-value cut-off of 0.001. Quantitative PCR (qPCR) {#s4j} ----------------------- Hemocytes were sorted and RNA extracted as described for microarrays. RNA concentration was measured with a Nanodrop 1000 spectrophotometer. Complementary DNA was synthesized using Transcriptor First Strand cDNA Synthesis kit (Roche). For quantitative PCR, kit from Applied Biosystems was used (ViiA 7 System, Applied Biosystems). See supplementary information ([Text S1](#ppat.1003720.s014){ref-type="supplementary-material"}) for primer sequences and protocol for qPCR analysis of AMP expression in gut and whole pupa. Supporting Information {#s5} ====================== ###### **Ecdysone-dependent evolution of hemocyte granularity and size over metamorphosis.** A. Forward scatter (FSC)--Area/side scatter (SSC)--Area plots reflecting size (x axis) and granularity (y axis) of hemocytes retrieved from L3W (blue), 8 h APF (orange) and 18 h APF (brown) control and *HmlΔ\>EcRB1DN* animals. Note, this is the same data represented in [Fig. 1E and F](#ppat-1003720-g001){ref-type="fig"}, separated into distinct plots for each stage for clarity. (B--C) FSC-A (indicative of cell size; B) and SSC-A (indicative of cell granularity; C) histograms for populations of control hemocytes (top) and EcRB1DN-expressing hemocytes (bottom) at different time points before (L3W larvae) and after puparium formation. Control hemocytes present a clear shift in FSC-A and SSC-A over development while the EcRB1DN-expressing hemocytes retain a larval size and morphology. These data encompass the data presented in [Fig. 1E and F](#ppat-1003720-g001){ref-type="fig"} and in [Fig. S1A](#ppat.1003720.s001){ref-type="supplementary-material"}. (TIF) ###### Click here for additional data file. ###### **Hemocytes expressing EcRB1 DN proliferate more actively in late 3^rd^ instar larvae.** L3W larvae were bled and proliferation was evaluated by measurement of the percentage of cells positive for a Phospho-histone H3 (PH3) staining among the hemocyte population (t-test; P\<0.0001). (TIF) ###### Click here for additional data file. ###### **Neither embryonic hemocyte dispersal nor recruitment of larval hemocytes to the gut proventriculus are affected by the expression of EcRB1DN.** (A--B) Similar dispersal of hemocytes in *w; srp^hemo^Gal4, UAS-GFP/+* embryo (A) or *w; srp^hemo^Gal4, UAS-GFP/UAS-EcRB1DN* embryo (B). The *serpent^hemo^Gal4* (*srp^hemo^*) driver was chosen for its early expression in hemocytes. (C--D) Similar numbers of hemocytes are recruited to the proventriculus in *w; HmlΔGal4, UAS-GFP/+* (C) and *w; HmlΔGal4, UAS-GFP/UAS-EcRB1DN* (D) larvae. In all pictures, anterior is up. Scale bars represent 20 µm. (TIF) ###### Click here for additional data file. ###### **Expression of** ***EcRB1DN*** **less than 18 h before puparium formation is sufficient to affect hemocyte motility.** We used a temperature sensitive (ts) Gal80, a Gal4 inhibitor, to control the expression of the *EcRB1DN* transgene in time. *HmlΔGal4, UAS-GFP; tub-Gal80^ts^/+* (*Gal80^ts^* control) and *HmlΔGal4, UAS-GFP; tub-Gal80^ts^/EcRB1DN* (*Gal80^ts^ EcRB1DN*) larvae were grown at 18°C (permissive for Gal80~ts~), transferred to 29°C (restrictive for Gal80*^ts^* - *EcRB1DN* is expressed) at late larval stage and hemocyte motility was measured 18 h later in 1 h APF-prepupae. P\<0.001. Mean and SEM are displayed. (TIF) ###### Click here for additional data file. ###### **Pupae in which hemocytes express EcRB1DN survive metamorphosis and are not delayed in their pupal development.** (A) Lethality at metamorphosis was very low and similar between control individuals and individuals which hemocytes express EcRB1DN (Wilcoxon test; P = 0.4042). Survival over metamorphosis is represented as the percentage of prepupae giving rise to adults. (B) The time needed for metamorphosis was not affected by expression of EcRB1DN in hemocytes (Wilcoxon Test; P = 0.7792). The curve represents the percentage of prepupae eclosed at different time points APF. Experiments were performed at 29°C. (TIF) ###### Click here for additional data file. ###### **Expression of EcRB1DN in hemocytes does not affect the humoral systemic immune response to oral infection by** ***E. carotovora*** **.** *HmlΔ\>GFP* (control) and *HmlΔ\>GFP, EcRB1DN* L3 larvae were fed on banana mixed with LB medium (as a non-infected control) or on banana mixed with a culture of *E.carotovora*. Prepupae at two stages -- early (light prepupae, 0--3 h APF) and late (dark prepupae, 4--8 h APF) were assessed for the induction of the humoral systemic immune response by RNA extraction from whole prepupae and quantitative PCR on the AMPs *Diptericin, Attacin, Metchnikowin* and *Drosomycin* (see [Text S1](#ppat.1003720.s014){ref-type="supplementary-material"} for details). In both genetic contexts, the transcription of all AMPs was strongly induced after infection except for *Drosomycin* (an antifungal AMP); a stronger induction was observed in late (dark) prepupae. Importantly, no significant differences in expression of any of the AMPs tested were observed between control and EcRB1DN, except for Attacin, which expression was significantly higher in *HmlΔ\>GFP, EcRB1DN* late prepupae (p\<0.05). These data correspond to three independent biological repeats. Mean and SEM are displayed. (TIF) ###### Click here for additional data file. ###### **Expression of EcRB1DN in hemocytes does not affect the local epithelial immune response to oral infection by** ***E. Carotovora*** **.** *HmlΔ\>GFP* (control) and *HmlΔ\>GFP, EcRB1DN* L3 larvae were fed on banana mixed with LB medium (control) or on banana mixed with a culture of *E.carotovora*. Prepupae were assessed for the induction of the local epithelial immune response by RNA extraction from guts dissected at 1--4 h APF, and quantitative PCR on the AMP *Diptericin* (see [Text S1](#ppat.1003720.s014){ref-type="supplementary-material"}). In both genetic contexts, the immune response was strongly induced after infection and *HmlΔ\>GFP, EcRB1DN* prepupae induced expression to a similar extent as controls. This graph corresponds to two independent biological repeats. *Dpt* expression is normalized to *Ecc15*-infected control and mean and data range are displayed. (TIF) ###### Click here for additional data file. ###### **An hemocyte has internalized an** ***E.coli-RFP*** **bacteria.** This image corresponds to an orthogonal cut of a still from [video S5](#ppat.1003720.s019){ref-type="supplementary-material"}, last time point (t = 31), showing a red *E.coli-RFP* bacteria inside a green hemocyte (GFP). (TIF) ###### Click here for additional data file. ###### **Genes for which expression is significantly changed at metamorphosis in an EcR-dependent manner.** From the microarray data, we selected genes for which expression was significantly changed (up or down-regulated; p\<0.001) at metamorphosis in control hemocytes. We then restricted our list to genes for which expression at pupal stage was significantly different between control and EcRDN-expressing hemocytes (p\<0.001). (XLSX) ###### Click here for additional data file. ###### **Microarray data.** Data obtained from the microarrays, with four conditions (control and EcRDN-expressing hemocytes, at larval and pupal stage) including three biological repetitions each. The table also presents the FC and p values for the following comparisons: control pupa vs control larva, EcRDN pupa vs EcRDN larva, EcRDN larva vs control larva and EcRDN pupa vs control pupa. (XLSX) ###### Click here for additional data file. ###### **EcR transcriptionally regulates genes involved in cell motility.** Genes for which expression is significantly changed at metamorphosis in an EcR-dependent manner ([Table S1](#ppat.1003720.s009){ref-type="supplementary-material"}) were cross-referenced with genes associated with a GO term related to cell motility (GO terms used: cell migration, cell motility, cell chemotaxis). (XLSX) ###### Click here for additional data file. ###### **EcR transcriptionally regulates genes involved in cell shape regulation.** Genes for which expression is significantly changed at metamorphosis in an EcR-dependent manner ([Table S1](#ppat.1003720.s009){ref-type="supplementary-material"}) were cross-referenced with the results of screens performed on S2 cells for changes in cell morphology (published in [@ppat.1003720-Kiger1], [@ppat.1003720-Rohn1]). (XLSX) ###### Click here for additional data file. ###### **List of primers used for RT-qPCR.** (DOC) ###### Click here for additional data file. ###### **Supplementary material and methods.** Material and methods used to perform the experiments presented as supplementary figures. (DOC) ###### Click here for additional data file. ###### **Hemocytes from a control prepupa acquire motility upon metamorphosis.** Hemocytes visible under the dorsal epithelium in a control prepupa (*HmlΔ\>GFP*) were imaged for 3 h, from 1 h APF to 4 h APF. The hemocytes can be observed changing morphology, and disperse by migrating away from their original location. Time indicated on the film is time APF. Images were analyzed by time-lapse confocal microscopy using a laser-scanning confocal microscope (LSM510; Carl Zeiss). Frames were taken every minute and are displayed at a rate of 7 frames/second. The scale represents 20 µm. (MOV) ###### Click here for additional data file. ###### **Hemocytes blocked in ecdysone reception do not acquire motility at metamorphosis.** Hemocytes visible under the dorsal epithelium in an *HmlΔ\>GFP, EcRB1DN* prepupa were imaged for 3 h, from 1 h APF to 4 h APF. These hemocytes, which express a dominant negative form of the EcRB1, barely move and do not change morphology. Time indicated on the film is time APF. Images were analyzed by time-lapse confocal microscopy using a laser-scanning confocal microscope (LSM510; Carl Zeiss). Frames were taken every minute and are displayed at a rate of 7 frames/second. The scale represents 20 µm. Stills from this video are presented in [figure 2](#ppat-1003720-g002){ref-type="fig"}. (MOV) ###### Click here for additional data file. ###### **Prepupal 'sessile' hemocytes are recruited to wound sites.** A wound (indicated by a box) was made with a laser in the vicinity of a sessile patch in a control prepupa (*HmlΔ\>GFP*), at 1h30 APF. Hemocytes were imaged for 3h20 after wounding. Time indicated on the film is time APF; the film starts immediately after wounding. The hemocytes respond to the wound by chemotaxing towards it, the majority originating from the lateral and longitudinal sessile patches. Images were analyzed by time-lapse confocal microscopy using a laser-scanning confocal microscope (LSM510; Carl Zeiss). Frames were taken every minute and are displayed at a rate of 14 frames/second. The scale represents 20 µm. Stills from this video are presented in [figure 3](#ppat-1003720-g003){ref-type="fig"}. (MOV) ###### Click here for additional data file. ###### **Hemocytes insensitive to ecdysone are impaired in their recruitment to wounds.** A wound (indicated by a box) was made with a laser in the vicinity of a sessile patch in a *HmlΔ\>GFP*, *EcRB1DN* prepupa, at 90 min APF. Hemocytes were imaged for 2h20 after wounding, from 1h30 APF to 3h50 APF. Time indicated on the film is time APF; the film starts immediately after wounding. Very few hemocytes can be visualised being recruited to the wound. Images were analyzed by time-lapse confocal microscopy using a laser-scanning confocal microscope (LSM510; Carl Zeiss). Frames were taken every minute and are displayed at a rate of 7 frames/second. The scale represents 20 µm. (MOV) ###### Click here for additional data file. ###### **Example of** ***in vivo*** **live imaging of phagocytosis.** A hemocyte in a control prepupa was filmed performing attachment, engulfment and internalization of an *E.coli-RFP* bacteria. One can see other hemocytes in the act of phagocytosis, or with already internalized bacteria, as well as free bacteria propelled by the hemolymph circulation. Hemocytes are filmed live through the cuticle of the prepupa. Images were acquired every 15 seconds. The film is displayed at 10 fpm and corresponds to approximately 8 minutes. (AVI) ###### Click here for additional data file. ###### **3D-image of an hemocyte imaged live in the act of engulfing a bacteria.** This 3D image corresponds to the time point 17 of [video S5](#ppat.1003720.s019){ref-type="supplementary-material"}. The hemocytes is in green (GFP), the bacteria in red (*E.coli-RFP*). 3D projection was realized with the corresponding ImageJ plugin. (AVI) ###### Click here for additional data file. ###### **Pupal hemocytes are not carried in the hemolymph by the movement of the heart.** A pupa is filmed over several minutes, where heartbeat pulses can be observed. Hemocytes are attached to surrounding tissues and do not move with the pulses. (MOV) ###### Click here for additional data file. We thank N. Matova and S. Prag for discussion, M. Crozatier, A. Vincent, A. Gontijo, T. Koyama, E. Bolukbasi, F. Leulier, P.Liehl and F. Cabreiro for helpful comments on the manuscript, C. Moreira for assistance with plotting tracking graphs, João Sobral and Jörg Becker (Instituto Gulbenkian de Ciênca) for the microarray service and discussion, the flow cytometry service of Instituto Gulbenkian de Ciênca and S. Ponte for technical assistance. We would also like to thank the anonymous reviewers whose constructive criticisms improved the original manuscript. [^1]: The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: AZR JCR AJ. Performed the experiments: AZR JCR ASB ABL ARMD. Analyzed the data: AZR JCR ASB ABL AJ. Contributed reagents/materials/analysis tools: ABL ES. Wrote the paper: AZR JCR. [^3]: Current address: Institute of Healthy Ageing, University College London, London, United Kingdom. [^4]: Current address: INSA-Lyon, INRA, UMR203 BF2I, Biologie Fonctionnelle Insectes et Interactions, Villeurbanne, France.

Introduction {#s1} ============ Tuberculosis (TB) is caused by the pathogenic species, *Mycobacterium tuberculosis (Mtb)*; together with human immunodeficiency virus (HIV/AIDS) infection, TB is among the most prevalent and severe of the infectious diseases worldwide. In 2019, an estimated 10 million people developed active tuberculosis in association with 1.6 million deaths ([@B1]). Infection with *Mtb* triggers an immune response, however *Mtb* can survive and grow by circumventing the host immune detection. One of the pathological characteristics of the successful infection with *Mtb* is the formation of granulome, which are organized cellular structures that include a variety of innate and adaptive immune cells that surround the *Mtb*-infected phagocytes ([@B2]--[@B5]). During the formation of granulome, intricate host-*Mtb* interactions occur at the infectious site and this pathogen can escape various host immune responses, which ultimately prevent *Mtb* elimination by these systems. Once *Mtb* enters the host, its cell wall components and proteins are detected by Toll-like receptors (TLRs), primarily by TLR2 and TLR4. *Mtb* is engulfed by professional phagocytic cells such as a macrophage, dendritic cell (DC), or neutrophil, and becomes incorporated into the subcellular organelle formed by the fusion of the phagosome and lysosome to create the phagolysosome, however *Mtb* is able to manipulate the endocytic pathway by suppressing fusion of the phagosome containing the bacteria with lysosomes. Infected macrophages synthesize and release both inflammatory and antimicrobial genes and molecules, including interleukin (IL)-1β, IL-6, IL-12, tumor necrosis factor (TNF), inducible nitric oxide synthase/nitric oxide synthase 2 (iNOS/NOS2), and chemokines which activate both the innate and adaptive immune systems. Activated immune cells secrete protective molecules to the extracellular space to promote recruitment of other immune cells to form a granuloma ([@B4], [@B6]). Interestingly, endogenous proteins expressed by *Mtb* serve to perturb the formation of phagolysosome, the permitting its survival and proliferation within macrophages. For preventing excessive lung damage during *Mtb* infection, *Mtb* also elicits the production of protective factors that promote its survival including anti-inflammatory mediators such as IL-4, IL-10, IL-13, and transforming growth factor β (TGF-β) ([@B7]--[@B9]) and several human TB studies show that these factors has been shown to be increased in the active TB patients ([@B10], [@B11]). These immunosuppressive factors play key roles in limits effective the immune defense to *Mtb* ([@B12], [@B13]). *Mtb* will persist and exacerbate pathophysiological manifestations within the granulome; this will ultimately result in progression of disease and dissemination to the other hosts ([@B5], [@B14]). As a major focus of this disease process, mycobacterial granulome have been the subject of intense scrutiny mainly focused on mechanisms of formation, function, maintenance, and evolution. Recently, there has been an increasing appreciation of the important relationship that exists between essential metabolism and immune cell function. Metabolic reprogramming in immune cells, a phenomenon known as immunometabolism, focuses on unique cellular functions that are essential for the immune response. During TB infection, host cells undergo profound metabolic change, which results in differential control of various cytokines and chemokines associated with inflammation, clearance, inhibition, and progression of *Mtb* infection ([@B15], [@B16]). Specifically, a shift in the use of pathways promoting glucose and lipid metabolism can be an important feature for directing host cell function to promote mycobacterial survival with the granulome ([@B17]). At homeostasis, cells in "resting" condition utilize oxidative phosphorylation (OXPHOS) to produce ATP from NADH and FADH2 by facilitating transfer of protons and electrons. Cells typically switch from OXPHOS to glycolysis in order to generate ATP under oxygen-depleted or hypoxic conditions ([@B18]). Similarly, glycolysis is main form of metabolism in immune cells that promote the inflammatory response in the immune system. This observation--that immune cells utilize glycolysis even in the presence of adequate concentrations of oxygen (i.e., aerobic glycolysis)-- is known as the "Warburg effect." To date, the Warburg effect has been explored primarily with respect to cancer metabolism. Although aerobic glycolysis generates fewer ATP molecules per cycle than does OXPHOS, this pathway is capable of rapid generation of ATP required by immune cells. Additionally, aerobic glycolysis requires a number of specific precursors, including nucleotides, amino acids, and lipids ([@B19]). Because metabolic reprogramming is essential for immune cell function, studies that explore this phenomenon in also provide new insight into the relationship between host immune cells and infection with *Mtb*. Furthermore, predisposing factors for TB, including diabetes, and HIV also related to immunometabolism against TB pathogenicity. Diabetes mellitus (DM) is a mainly risky factor for occurring active TB ([@B20]--[@B22]). In DM, innate immune cells undergo activation for releasing cytokines, recruiting neutrophils, upregulate T cell activation and antigen recognition ([@B23], [@B24]). Metabolism of DM is characterized by increasing glucose production and impairing glucose uptake. Expression of glucose transporter and glycolytic enzymes is elevated in DM ([@B25]). In DM, High glucose level increased IL-10 production, impaired macrophage phagocytic ability for promoting better milieu for survival and proliferation of TB ([@B26], [@B27]). Additionally, HIV is also other pathogen to be associated with pathogenicity of TB ([@B28]--[@B30]). In HIV-1-infected primary CD4^+^ T cells, glycolytic metabolism is induced with high pro-inflammatory response and increased production of virus ([@B31], [@B32]). Interestingly, glycolytic metabolism is regulated by HIV-1 infection in macrophage alleviated Warburg effects ([@B33]). These factors promote the activation of TB by reprogramming the metabolism. A variety of antibiotics have been introduced for promoting eradication of *Mtb* infection, including 6--9 months courses of isoniazid, rifampicin, ethambutol, and pyrazinamide. However, the emergence of multidrug-resistant TB (MDR-TB) or extensively drug-resistant TB (XDR-TB) has become a major challenge toward designing effective treatments and for eradication of this disease ([@B34], [@B35]). Among the approaches to this challenge, host-directed therapy (HDT) has been introduced as a means to potentiate and to amplify the effectiveness of current treatments used for TB ([@B36]). A clear understanding of the molecular interactions between host cell metabolism and accommodations made to *Mtb* may provide new strategies to combat infection. Here we review the current understanding of the metabolic relationship between the host and the *Mtb* pathogen. We also suggest several new strategies that may enhance host metabolic pathways and thereby promote protective antimicrobial functions in the setting of TB infection. Metabolic Reprogramming in TB {#s2} ============================= Warburg Effect in Immune Cells ------------------------------ Immune cells provide critical protection and maintain homeostasis in the mammalian host. There are currently many studies that suggest that the functions of immune cells are largely reliant on specific aspects of host metabolism. These studies, which have generated a field known as immunometabolism, have provided us for a new focus for understanding how and why immune cells exist or persist in a specific metabolic state in order to support or direct functional changes. Several recent reports suggest that different metabolic signatures have a direct impact on specific effector functions characteristic of the innate and adaptive immune systems ([@B37]). As such, among the primary functions of immune cells, there are those that generate an inflammatory response, actions typically undertaken by M1-polarized macrophages, DCs, neutrophils, and effector T cells, and those that promote an anti- inflammatory response, which include M2-polarized macrophages, as well as regulatory and memory T cells. The basic metabolic profiles of these cells differ significantly from one another. Inflammatory immune cells generate energy in the form of ATP mainly via glycolytic metabolism; by contrast, immune cells that promote anti- inflammatory activities generate ATP via oxidative phosphorylation and fatty acid oxidation ([@B38]--[@B43]). These observations have been best characterized for polarized macrophages. The predominant phenotypes of macrophages are known as M1 and M2 ([@B44], [@B45]). M1 macrophages, activated by lipopolysaccharide (LPS) and IFN-γ, promote pro-inflammatory and antibacterial functions in immune system, and they produce nitric oxide (NO) and reactive oxygen species (ROS) which are fundamental components of the pathways used to eradicate bacteria. The main metabolic pathway used by these cells is glycolysis, which results in rapid production of ATP via inhibition of the trichloroacetic acid (TCA) cycle and OXPHOS in mitochondria; this is a critical factor due to the fact that M1 macrophages require rapid generation of ATP to activate inflammation. By contrast, M2 macrophages promote anti-inflammatory responses and tissue repair; these cells mainly utilize OXPHOS and fatty acid oxidation in order to generate ATP; this takes place via efficient pathways localized in the mitochondria ([@B46]--[@B51]). In T cells, metabolic state is reprogrammed according to T cell subsets. Naïve T cells mainly use OXPHOS for generating energy. Upon TCR stimulation, glycolytic metabolism is upregulated for differentiating into activated T cell. Th1, TH2, and Th17 effector cells mainly depend on aerobic glycolysis. While, regulatory and memory T cells use fatty acid oxidation and OXPHOS for differentiation and functions ([@B52], [@B53]). Mammalian target of rapamycin (mTOR) and AKT signaling is essential for regulating metabolism of T cells and cytokine responses ([@B54]). Recently, cyclophililn D (CypD) related to necrosis is a factor for regulating metabolic state and functions in T cells ([@B55]). Pro-inflammatory immune cells generate ATP in high concentrations via glycolysis even when functioning in aerobic conditions; the phenomenon of aerobic glycolysis is also known as the "Warburg effect" ([@B56]). Hypoxia and inflammation are inherently linked to one another; upon activation, immune cells undergo considerable metabolic reprogramming to sustain energy needs and thus switch to predominantly aerobic glycolysis. Hypoxia-induced factor 1 (HIF-1), the main mediator of the Warburg effect, is expressed in response to hypoxia and controls expression of numerous glycolytic enzymes. HIF-1 has two subunits, α and β; regulation of HIF-1 is dependent on the α subunit. Post-translational regulation of HIF- 1 is modulated via the expression and stability of HIF-1α ([@B56]--[@B58]). Members of the nuclear factor-κB (NF-κB) family of transcription factors comprise the signaling pathway that is most closely involved in Hif-1α/HIF-1A expression ([@B59], [@B60]). Under conditions of physiologic oxygenation, prolyl hydrolases (PHD) degrade HIF-1α and target it for proteasome-mediated degradation. Inhibiting HIF (FIH) is an aspariginyl hydroxylase that also determines the level of active HIF-1α. Overall, hypoxia-inducible genes encode proteins involved in a myriad of cellular pathways that mediate cell survival, apoptosis, erythropoiesis, angiogenesis, glucose metabolism, and that regulate acid-base balance ([@B61]). HIF-1α is expressed in primary innate immune cells, including macrophages, DCs, neutrophils, and Th17 cells. Additional roles for HIF-1α in promoting macrophage differentiation and function have also been demonstrated. Most notably, HIF-1α-mediated metabolic reprogramming plays a significant role in modulating macrophage polarization toward the M1 or M2 phenotype ([@B62]). Glycolysis Metabolism in TB --------------------------- When the host is infected by bacteria, immune cells are activated; the characteristic immune response occur concomitant with a switch to glycolytic metabolism ([Figure 1](#F1){ref-type="fig"}). Several recent studies that have focused on transcriptome data from mouse and rabbit lung as well as granulome from the lungs of TB patients suggest that the metabolic state of the TB-infected host includes modulation of glucose metabolism ([@B63]--[@B66]). The general metabolic characteristics in TB infection included enhanced expression of genes related to the Warburg effect including HIF-1α, glycolytic enzymes, the pentose phosphate pathway, and H^+^-ATPase. Additionally, ^1^H-NMR-based metabolomics profiled the increased accumulation of lactate due to the increased levels of glycolysis in the lungs of *Mtb*-infected mice ([@B67]). Likewise, host immune cells responded to *Mtb* infection with increased expression of pro- inflammatory and antimicrobial-related genes associated with the Warburg effect. These results highlighted the importance of metabolic reprogramming due to glycolysis and its relationship to protection against *Mtb* infection. Furthermore, analysis of the transcriptomes of bone marrow-derived macrophages (BMDM) infected with one of two clinical strains of *Mtb* (the immunogenic strain CDC1551 or the hypervirulent strain HN878) included elevated levels of expression of genes associated with the Warburg effect. Given that these two clinical strains are known for differential activation of immune responses during the course of BMDM infection, different metabolic responses were anticipated ([@B64]). Interestingly, BMDMs infected with each strain promoted upregulation of genes encoding enzymes associated with the Warburg effect together with HIF-1α-associated signaling, although specific differences were observed. Of note, at 6 h post-infection, the induction of the gene encoding 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3) a member of the of phosphofructokinase (PFK)-2 family, was more prominent in CDC1551-infected BMDMs ([@B65]). *Pfkfb3* has the highest activity among the PFK-2 members, and fructose-2,6-diphosphate (F-2,6-BP), which is the product of Pfkfb3-mediated phosphorylation, is an essential component promoting regulation of glycolysis ([@B68]). CDC1551-infected BMDMs in a state of elevated glycolysis respond with a vigorous early pro-inflammatory response. By contrast, relatively limited activation of the Warburg effect together with high levels of glucose uptake were observed in response to *Mtb*. Furthermore, HN878-infection of BMDMs may result in dysregulated host cell lipid metabolism. Specifically, one study compared gene expression in response to *Mtb* H37Ra or H37Rv infection of human alveolar macrophage revealed strain-specific differences. Gene expression associated with inflammation, general metabolism, and lipid metabolism was downregulated in H37Rv infected macrophages ([@B69]). As suggested by the responses to infection with HN878, a virulent strain can have an impact on host metabolism gene by downregulating inflammatory responses that results in diminished the inflammation and prolonged *Mtb* survival. Another study compared the metabolic states elicited by macrophage challenge with *Mtb*, with the vaccine strain *M. bovis* BCG or with killed *Mtb*. Each strain promoted a unique pattern of energy modulation, as determined by XF (extracellular flux) analysis. Total metabolism in response to challenge with live *Mtb* including glucose utilization and OXPHOS is lower than that observed in response to BCG or dead *Mtb* ([@B70]). Also, CD8^+^ T cell showed similar results in *Mtb* or BCG infection. Through RNA-seq, glycolytic metabolism is upregulated by challenging *Mtb* in early and late phase. Surprisingly, *Mtb* triggered mitochondrial dysfunction, which downregulates OXPHOS metabolism, while upregulates mtROS, but metabolism is recovered against BCG ([@B71]). Thus, infection with live, virulent *Mtb* decelerated the shift to glycolytic and OXPHOS bioenergetics, and thereby limited the development of inflammatory effector functions. {#F1} The switch to glycolytic metabolism resulted in the accumulation of several TCA intermediates that themselves function as a metabolic signal to link metabolism and immunity ([Figure 2](#F2){ref-type="fig"}). Succinate, a prominent TCA intermediate, drives IL-1β production, inhibits the production of anti-inflammatory cytokines, and enhances HIF-1α activity by inhibiting HIF-1α prolyl hydrolases ([@B72]--[@B74]). The succinate-induced pro-inflammatory response is directly dependent on the activity of succinate dehydrogenase (SDH). Inhibition of SDH activity via hydrolysis of dimethyl malonate to produce malonate, results in an attenuation of the activity of LPS-induced IL-1β, and likewise a boost in IL-10 production in BMDMs generated from C57BL/6 mice ([@B75]). In *Mtb*-infected murine macrophages, *Sdh* expression is downregulated; this leads to the induction of HIF-1α, the Warburg effect, and characteristic pro-inflammatory responses ([@B76]). Itaconate, a metabolite derived from the TCA cycle intermediate cis-aconitate, also regulates SDH activity in C57BL/6 BMDMs ([@B77], [@B78]). Breakdown of TCA cycle results in downregulation of mitochondrial isocitrate dehydrogenase (*Idh*)2 immediately following formation of itaconate. Aconitate decarboxylase 1 (*ACOD1*), is also known as immune-responsive gene (Irg)1; production of this mediator is related to generation of itaconate. *ACOD1* is upregulated in *Mtb*-infected murine macrophages and lung tissue. Itaconate has antimicrobial functions via its capacity to inhibit isocitrate lyase, the essential enzyme in the glyoxylate shunt that is critical for bacterial growth. Itaconate inhibits SDH activity which results in the accumulation of succinate. Additionally, itaconate modulates pro- inflammatory responses in macrophage; *Irg1*^−/−^ BMDMs from C57BL/6 mice maintain higher HIF-1α mRNA and protein levels, and produce more pro-inflammatory cytokines and antimicrobial factors including IL-6, IL-12, IL-1β, and NO in response to lipopolysaccharide (LPS)-mediated activation ([@B79]). Thus, itaconate may be a critical link between the Warburg effect induced by *Mtb* infection, and the generation of anti-inflammatory responses to prevent damage to host cells. {#F2} Upregulated expression of HIF-1α, the enhanced Warburg effect, and the antimicrobial response to *Mtb* infection of host immune cells are all linked to the actions of the glycolytic regulatory protein, pyruvate kinase M2 (PKM2). Expression of PKM2, one of the two Pkm/PKM gene products, is upregulated in response to macrophage activation. In the cytoplasm, PKM2 maintains an enzymatically inactive state via its phosphorylation; the PKM2 dimer is transferred into the nucleus where it interacts with HIF-1α to activate target genes, including those encoding glycolytic enzymes and IL- 1β. In LPS-activated macrophages, small molecules such as TEPP-46 modulate PKM2 activation by preventing PKM2 translocation into the nucleus; consequently, results in a diminished Warburg effect and limited production of IL-1β. Inhibition of PKM2 translocation also promotes production of IL-10 and a decreased antimicrobial response in an *S. typhimurinum* infection model ([@B80]). In transcriptome analysis studies, upregulation of Pkm2/PKM2 was detected in *Mtb*- infected murine macrophages and in mouse lung tissue ([@B65]). These results suggest that, similar to itaconate, PKM2 promotes the HIF-1α-mediated Warburg effect and the associated antimicrobial response during *Mtb* infection. CypD, mitochondrial matrix protein, is regulator of metabolism in *Mtb* infection via upregulating mtROS in T cells. CypD-deficient T cells showed higher OXPHOS than wild-type T cells and more susceptible to *Mtb* ([@B55]). In summary, metabolism in *Mtb*-infected host cells undergoes a switch from OXPHOS to glycolysis and generates a Warburg effect. The HIF-1α induced Warburg effect in the setting of TB infection plays an essential role in promoting upregulation of pro-inflammatory cytokine and antimicrobial effector gene expression, both factors underlying the acute immune response. However, host immune responses were different depending on the virulence or avirulence of the *Mtb*-infecting strain. How and why immune responses are modulated by different strains of *Mtb* are not fully understood. Arginine Metabolism in TB ------------------------- Arginine, the key substrate for production of NO and other reactive nitrogen species, and also serves as a substrate for arginase. Arginine plays a distinct role in the host immune response. iNOS promotes one pathway that results in the generation of NO; the other pathway is via the arginase-mediated production of ornithine ([@B16]). iNOS is one of three NO synthase enzymes and the major isoform involved in immune cell functions. iNOS is inducible in immune cells, and is a prominent antimicrobial effector molecule produced by activated macrophages ([@B81]). The balance of arginine metabolism between the two competing pathways constitutes an important regulatory mechanism that modulates the polarization states of M1 and M2 macrophages. In M1 macrophages, arginine is in demand for protein synthesis, for production of NO, and for its antimicrobial roles; by contrast, in M2 macrophages, arginine is used for production of polyamines and proline. The iNOS pathway is in direct competition with the arginase pathway ([@B82], [@B83]). Two arginase isoforms exist in the cells. Cytosolic arginase ARG1 and mitochondrial arginase ARG2 are encoded by different genes and have different subcellular distributions ([@B84], [@B85]). ARG1 is mainly detected in murine myeloid cells, DCs, and granulocytes. ARG1 inhibits NO production from iNOS/NOS2 which is among the mechanisms used by *Mtb* for immune evasion. *Mtb*-infected *Arg1* conditional gene-deleted mice were characterized with a diminished bacterial burden; Arg1-deficient macrophages were more capable of killing *Mtb* compared to their wild-type counterparts ([@B86]). ARG1 and iNOS are distributed in distinct patterns in human TB-associated granulome; expression of iNOS was highest in the central region, and ARG1 was more prominent at the periphery ([@B87]). The role of ARG1 in mediating immune cell function is directly dependent on the stage of *Mtb* infection. At initial stages of infection, the Mtb pathogen takes advantage of ARG1 activity by limiting macrophage immunity via competition with iNOS/NOS2. During the late stages of infection, ARG1 contributes to control of prolonged hyperinflammation; ARG1 also plays a role in regulating the progression of lung immunopathology in *Mtb*-infected, Nos2-deficient mice ([@B87]). Lipid Metabolism in TB ---------------------- Once glycolytic metabolism has been activated, the genes encoding pro- inflammatory mediators are synthesized, together with the synthesis of fatty acids and phospholipids. The TCA cycle and OXPHOS are inhibited, and several intermediates of the TCA cycle accumulate *in situ* ([@B88]). Similar to what has been observed for glucose metabolism, including the TCA cycle and OXPHOS, host lipid metabolism is also regulated in *Mtb* infection ([Figure 2](#F2){ref-type="fig"}). There are master regulators that mediate lipid metabolism including the peroxisome proliferator-activated receptors (PPARs), liver X receptor (LXR), sterol regulatory element binding proteins (SREBPs) and HIF ([@B89]--[@B93]). These factors work together to regulate processes including fatty acid uptake, lipid synthesis, the activities of lipolytic enzymes, and lipid droplet (LD) biogenesis ([@B94]). The activation of TLR signaling upregulates expression of several enzymes that promote synthesis of triglycerides and/or cholesterol ester, including fatty acid synthase (FASN), diacylglycerol O- acyltransferases (DGAT-1 and DGAT-2), and acyl-CoA:cholesterol O-acyltransferases (ACAT1 and ACAT2) ([@B95]--[@B97]). During lipid accumulation, increased expression of lipid uptake and transport-related genes is observed, and expression of genes involved in lipolysis is decreased. Perilipin-2 (Plin2) and Perilipin-3 (Plin3) are the main structural proteins of LDs that serve to promote lipid accumulation ([@B96], [@B98], [@B99]). These proteins are essential for the biogenesis and assembly of LDs ([@B100]). PPARs are members of the ligand-activated transcription factor family ([@B101]). PPARs can have a direct impact on LD formation via the regulation of Plin2 expression. PPARs also regulate proteins associated with *de novo* lipogenesis, including fatty acid synthase and gene regulatory factors LXR and SREBPs ([@B94]). PPAR-γ is important for regulating lipid and glucose metabolism and other cellular process including inflammation ([@B102]). Host immune cells which are infected by *Mtb* exhibit increased PPAR-γ gene expression; this results in downregulation of NF-κB signaling and increases in production of prostaglandin (PG) E2; overall, this results in suppression of pro- inflammatory cytokines and Th1 responses ([@B103], [@B104]). Increased PPAR-γ expression in *Mtb*-infected macrophages is also associated with LD formation ([@B105]). Formation of LDs is critical for bacterial survival; the accumulated lipids in these infected cells provide nutrients and promote bacterial growth in host. Additionally, infection with *M. bovis* BCG results in upregulated expression and activation of PPAR-γ and the induction of lipid-loaded macrophages. In BCG-infected TLR2-deficient mice, production of TNF-α undergoes significant downregulation ([@B104], [@B106]). Taken together, these findings suggest that PPAR-γ accelerates intracellular lipid accumulation by modulating the expression of genes that modulate lipid absorption as well as those that promote fatty acid synthesis in response to *Mtb* infection. PPAR-α is another isoform of the PPAR family. It is a transcription factor that modulates the expression of several genes involved in lipid oxidation and glucose metabolism ([@B107]). PPAR-α enhances fatty acid oxidation and ketogenesis while inhibiting fatty acid synthesis and glycolysis ([@B108]). As such, activation of PPAR-α may prevent lipid accumulation in *Mtb*-infected cells. PPAR-α activation also results in the upregulation of transcription factor EB (TFEB) and promotes host innate immunity and autophagy against *Mtb* infection. The induction of TFEB also promotes lipid catabolism which inhibited intracellular growth of *Mtb* growth in bone marrow-derived macrophages ([@B109]). Metabolic HDT in TB {#s3} =================== In recent years many researchers have demonstrated that changes in dynamic immunometabolism take place in response to infection with microbes; as such, studies focused on immunometabolism are important so as to provide a larger understanding of their role in promoting pathogenesis in host ([@B110]). Current clinical trials have limitations with respect to the elimination of *Mtb* infection, including the need for long-term use, severe side effects, and the emergence of drug-resistant strains ([@B111]). As noted above, *Mtb* infection can induce a Warburg effect in host immune cells, similar to that described in tumor tissue ([@B65]). *Mtb* exploits host metabolism in order to escape immune surveillance and modulates various responses to subvert their activities toward promoting its survival and longevity. We expect HDT to be a clinically-feasible approach toward readjusting uncontrolled immune responses in patients with infectious disorders. We discuss HDT drugs currently in use or under development that target host metabolism. We will also suggest novel candidate HDT pathways and agents that might be effective toward eradicating *Mtb* ([Table 1](#T1){ref-type="table"}). ###### Host-directed therapies that regulate host metabolism in TB. --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **HDT in glucose metabolism** ------------------------------- ------------------------------- ------------------------------------------------------------------------------------------------ ---------------------------- **Name** **Target** **Result** **References** 2-deoxyglucose Hexokinase Inhibition of glycolysis Suppression of IL-1β ([@B73], [@B112]) 3-bromopyruvate Hexokinase Inhibition of glycolysis ([@B113]) Ritonavir Glucose transporter Inhibition of glycolysis ([@B114]) Dichloroacetate Pyruvate dehydrogenase kinase Inhibition of glycolysis ([@B115]) FX11 Lactate dehydrogenase Inhibition of glycolysis\ ([@B116]) Downregulation of cytokines and iNOS TEPP46 Pyruvate kinase M2 Inhibition of HIF-1α Suppression of IL-1β ([@B80]) Rapamycin mTOR Inhibition of glycolysis Upregulation of antimicrobial effect ([@B117], [@B118]) Loperamide mTOR Inhibition of glycolysis\ ([@B119]) Upregulation of antimicrobial effect **HDT in lipid metabolism** Metformin AMP kinase Increased fatty acid oxidation. Inhibition antibacterial activity Reduced gene of inflammation ([@B120], [@B121]) AICAR AMP kinase Increased antibacterial activity\ ([@B122]) Induced mitochondrial biogenesis and energy metabolism Inhibition of lipid synthesis C75 Fatty acid synthase Inhibition of fatty acid synthesis Reduced the inflammation and oxidative stress\ ([@B123]--[@B125]) Switch M2 to M1 Downregulation of NLRP3 inflammasom Cerulenin Fatty acid synthase Inhibition of fatty acid synthesis Downregulation of NLRP3 inflammasome ([@B125]) GW9662 PPARγ Modulation of lipid metabolism, inflammation and pathogenesis of bacteria ([@B95]) Sirtuins PGC-1α Inhibition of NF-κB signaling and proinflammatory response\ ([@B76], [@B126]--[@B128]) Upregulation of fatty acid oxidation and anti-inflammation --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- *HDT in glucose metabolism HDT in lipid metabolism*. HDT in Glucose Metabolism ------------------------- In TB infection, metabolism switches to glycolysis in order to protect the host against early-phase *Mtb* responses. HIF-1-dependent glycolysis promotes various immune effector functions including production and release of pro-inflammatory cytokines and NO. As noted earlier, virulent *Mtb* perturbs the glycolytic metabolism and thereby inhibits antimicrobial functions. These results suggest metabolic reprogramming to aerobic glycolysis is essential component of the anti-TB response. On the other hand, persistent inflammation can result in hyperinflammation and ultimately damage host cells and tissues. Among the featured mechanisms of HDT in TB, there is a focus on inhibition of glycolysis as well as modulation of mTOR and AMP-activated protein kinase (AMPK) pathways. For example, 2-deoxyglucose (2-DG) and 3-bromopyruvate suppress activity of hexokinase which is a critical enzyme that catalyzes the first step of glycolysis ([@B113]). In LPS-activated macrophages, 2-DG suppresses the production of IL-1β and results in the accumulation of succinate ([@B73]). Additionally, LPS-induced acute lung injury is reduced by 2-DG-dependent inhibition of glycolysis ([@B112]). Among others under consideration is the HIV-protease inhibitor, ritonavir, which is an antagonist of glucose transporters ([@B114]), dichloroacetate, an inhibitor of pyruvate dehydrogenase kinase ([@B115]), and FX11, a specific inhibitor of lactate dehydrogenase. In LPS-activated RAW 264.7 mouse macrophages, FX11-mediated inhibition of lactate dehydrogenase resulted in the downregulation of cytokine and iNOS production ([@B116]). Likewise, TEPP46 is small molecule that inhibits the activity of pyruvate kinase M2; this inhibitor attenuates activation of PKM2 in LPS-induced macrophage *in vivo* and results in suppression of IL-1β production ([@B80]). Induction of autophagy can be potential defense strategy used by cells to eradicate *Mtb* infection. The enzyme, mTOR kinase, negatively regulates autophagy; as such, mTOR kinase inhibitors may be potent candidates for HDT for the elimination of *Mtb* infection. Other mTOR inhibitors including rapamycin and torin serve to limit the increased levels of lactate detected in *Mtb*-infected macrophages ([@B54]). Rapamycin-mediated activation of autophagy results in acidification of mycobacterial phagosomes and thus decreased survival of BCG ([@B117]). Loperamide induces mTOR-independent autophagy and likewise controls intracellular *Mtb* burden in lung macrophages ([@B119]). However, the use of these inhibitors has several limitations. For example, rapamycin-induced autophagy resulted in enhanced intracellular bacterial replication in HIV/H37Rv co-infected cells ([@B118]). Therefore, pharmacological induction of autophagy should be carefully evaluated among the candidate drugs to be used for HDT. HDT in Lipid Metabolism ----------------------- *Mtb* exploits host lipid or fatty acid metabolism to promote its own survival and growth. Foamy macrophages are recruited to granulome where and are included in the barrier that forms around *Mtb*-infected phagocytic cells to which they provide support and nutrition. Toward this end, infection with *Mtb* induces the synthesis of LDs and fatty acids in host cell. Targeting the lipid synthesis may be a good strategy for initial HDT with the goal of eliminating *Mtb*. 5\' AMPK is a highly conserved master regulator which can restore the energy balance by shifting cellular metabolism from one that consumes ATP to a catabolic mechanism that generates ATP ([@B129]). AMPK and other metabolic energy sensors are critical in maintaining various functions of *Mtb*-infected host immune cells, including autophagy, fatty acid β- oxidation, and metabolic reprogramming; the AMPK pathway also plays multi-faceted roles in promoting host defense against viral and bacterial infection. As such, molecules that are targeted by AMPK-targeted are considered to be effective adjuvant agents used to combat *Mtb* infection ([@B130], [@B131]). Metformin, a drug that is clinically-approved for the treatment of type 2 diabetes functions by activating the AMPK-mediated signaling pathway ([@B121]). Treatment with metformin can limit intracellular *Mtb* growth in macrophages via induction mitochondrial ROS and can thereby reduce activation of inflammatory-related gene expression. Also, metformin shows some synergy with conventional anti-TB drugs, including isoniazid or ethionamide when evaluated in *Mtb*-infected mice. Metformin treatment also decreases the incidence of latent TB ([@B120]). AICAR (5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside) is another agent that activates AMPK; AICAR activates autophagy pathways in macrophages and thus promotes antibacterial activity against *Mtb*. AICAR-mediated AMPK-activation also results in the activation of the PPARGC1 (peroxisome proliferator-activated receptor gamma, coactivator 1) pathway; this latter pathway regulates mitochondrial biogenesis and energy metabolism in macrophages and in *Drosophila melanogaster* infected with *M. marinum* ([@B122]). Factors that suppress lipid synthesis can limit inflammation and balance the inflammatory state of the host. Among several candidate molecules, C75 and cerulenin inhibit fatty acid synthase. C75 effectively lowers free fatty acid accumulation in mice with sepsis and limits inflammation and oxidative stress ([@B123]). Additionally, C75-mediated inhibition of lipid-derived droplet formation results in a switch from M2 to M1 macrophage polarization, resulting in enhanced production of both ROS and NO generation ([@B124]). Additionally, inhibition of fatty acid synthase by C75 and cerulenin results in downregulated uncoupling protein (UCP2)- mediated NLRP3 inflammasome activation ([@B125]). GW9662, an antagonist of PPARγ, acts as a key modulator of lipid metabolism, inflammation, and pathogenesis in BCG-infected macrophages; this result suggests that regulation of lipid metabolism may be a strong potential host target for novel TB therapy ([@B91]). Likewise, sirtuins (SIRTs) have been recognized as potential targets for anti-TB therapeutics. Sirtuins are enzymes with deacetylase activity that modulate cellular process by inhibiting NF-κB signaling; this results in a downregulation of the pro-inflammatory response and upregulation of fatty acid oxidation and anti- inflammatory response by targeting Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α) ([@B126], [@B127]). SIRT1 expression is diminished in *Mtb*-infected THP-1 macrophages and in whole mouse lung tissue. SIRT1 promotes inflammatory resolution by downregulating the expression of the RelA/p65 unit of NF-κB ([@B128]). SIRT6 also suppress pro-inflammatory and antimicrobial responses at the early stages of *Mtb* infection ([@B76]). Conclusion {#s4} ========== Immunometabolism is among the critical features that define the intimate relationship between host and the *Mtb* pathogen; a clear understanding of these interactions will be essential for limiting the progression of the TB. Metabolic reprogramming from OXPHOS to glycolysis in *Mtb* infection results in the upregulated expression of numerous pro-inflammatory cytokines and antimicrobial effector molecules. Further investigation will be needed in order to understand more fully the relationship between *Mtb* and host metabolism. How and when *Mtb* exploit the host metabolism is not clearly understood at this time; clarification will be critical in order to identify the most appropriate candidates for HDT. Among those currently under consideration is *Mtb*-mediated modulation of glucose and/or lipid metabolism. Glucose metabolism might be targeted at the early stage, which would ultimately provide a boost to the Warburg effect. Thus, more efficient elimination of *Mtb* bacteria; by contrast, targeting glucose metabolism at a later stage may result in a much needed- alleviation of hyperinflammation. A better understanding of metabolic reprogramming in TB will provide further insights toward novel therapeutic strategies. Author Contributions {#s5} ==================== J-SK, Y-RK, and C-SY designed, conceptualized, and wrote the manuscript. All authors contributed to the article and approved the submitted version. Conflict of Interest {#s6} ==================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. We would like to thank all members of the Infection Biology Lab for critical reading and discussion of the manuscript. **Funding.** This work was supported by the NRF grant funded by the Korea government (MSIP) (2016R1D1A1A02937312 and 2019R1I1A2A01064237); a grant from the KHIDI, funded by the Ministry of Health & Welfare, Republic of Korea (HI16C1653). *Mtb* : *Mycobacterium tuberculosis* TB : Tuberculosis HDT : Host-directed target TLRs : Toll-like receptors DC : Dendritic cell IL : Interleukin TNF : Tumor necrosis factor iNOS/NOS2 : inducible nitric oxide synthase/nitric oxide synthase 2 TGF-β : Transforming growth factor β OXPHOS : Oxidative phosphorylation DM : Diabetes mellitus MDR-TB : Multidrug-resistant TB XDR-TB : Extensively drug-resistance TB NO : Nitric oxide ROS : Reactive oxygen species HIF-1 : Hypoxia-induced factor 1 NF-κB : Nuclear factor-κB CypD : Cyclophililn D PHD : Prolyl hydrolases FIH : Factor inhibiting HIF PFKFB3 : 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 F-2,6-BP : Fructose-2,6-diphosphate SDH : Succinate dehydrogenase LPS : Lipopolysaccharide ACOD1 : Aconitate decarboxylase 1 Irg1 : Immune-responsive gene1 PKM2 : Pyruvate kinase M2 ARG : Arginase PPARs : Peroxisome proliferator-activated receptors LXR : Liver X receptor SREBPs : Sterol regulatory element-binding proteins LD : Lipid droplet FASN : Fatty acid synthase DGAT : Diacylglycerol O-acyltransferase ACAT : Acyl-CoA:cholesterol O- acyltransferase Plin : Perilipin TFEB : Transcription factor EB mTOR : Mammal target of rapamycin AMPK : AMP-activated protein kinase 2-DG : 2-deoxyglucose PPARGC1 : Peroxisome proliferator-activated receptor gamma, coactivator 1 AICAR : 5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside UCP2 : Mitochondrial uncoupling protein 2 SIRTs : Sirtuins PGC-1α : Peroxisome proliferator-activated receptor gamma coactivator 1-alpha. [^1]: Edited by: Anca Dorhoi, Friedrich Loeffler Institute, Germany [^2]: Reviewed by: Arshad Khan, McGovern Medical School at UTHealth, United States; Elsa Anes, University of Lisbon, Portugal [^3]: This article was submitted to Microbial Immunology, a section of the journal Frontiers in Immunology

There are errors in the Funding section. The correct funding information is as follows: This study was supported by the National Cancer Institute of the National Institutes of Health under award number K08CA155035 and the Melanoma Research Alliance. The authors are also grateful to Timothy Dattels for his generous support. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Introduction ============ Diabetes mellitus is a common endocrine metabolic disease estimated to affect 629 million individuals in 2045 according to the International Diabetes Federation. T2DM is the extremely prevalent form of diabetes that accounts for 90% of diagnosed cases and is associated with insulin resistance and chronic hyperglycemia ([@B13]). Many clinical studies reported a broad spectrum of lower urinary tract symptoms in diabetic patients ([@B16]), accounting for 90--95% of all diabetes cases ([@B17]). DBD is a major lower urinary tract complication of diabetes and was first described by [@B26]. Such complication is traditionally described as a triad of increased capacity, decreased sensation, and poor emptying ([@B11]) and has affected over 50% of diabetic patients ([@B12]; [@B28]). DBD development is divided into two phases: the compensated phase, which occurs in the early phase and is characterized by an OAB; and the decompensated phase, which occurs in the late phase and is characterized by an atonic bladder ([@B36]; [@B24]). The pathogenesis of DBD is multifactorial and accompanied by the structural and functional impairments of the bladder ([@B37]). The bladder structural remodeling of DBD, such as the increase in bladder capacity, total BWT, and smooth muscle content, was observed in STZ-induced diabetic mice. Such remodeling may be a physical alteration to increase the urine volume ([@B22]). The two major functions of bladder are urine storage and urine disposal, and the uncoordinated contraction in the OAB of a diabetic greatly affects the urine storage ability of this organ ([@B10]). Bladder contraction is mainly mediated by purinergic and cholinergic pathways ([@B23]). In particular, ATP is the purinergic messenger released from varicosities or bulbous nerve endings of neurons, and the contractile responses mediated by ATP play a key role in DBD ([@B44]). Solute carrier family 17 member 9 (SLC17A9) is a member of the solute-carrier protein family that plays an indispensable role in the vesicular storage of ATP ([@B31]; [@B20]). The translocation of neurotransmitter-filled vesicles to the varicose terminal is the first step in the release of vesicular neurotransmitters, followed by the merger of vesicles with the membrane of the varicose terminal and the precise and rapid release of their contents into the synaptic cleft ([@B33]). In addition, the motor of vesicles for transportation to the varicose membrane in the cells is mainly provided by myosin motors, particularly myosin Va ([@B2]). Several studies found that the purinergic inhibitory neurotransmission was impaired in myosin Va-deficient mice ([@B6]). Such finding suggested that myosin Va played an important role in purinergic neurotransmission ([@B3]). Diabetic bladder dysfunction, particularly OAB, is not life threatening to humans. However, this dysfunction seriously affects the quality of the life of patients ([@B24]). The treatment methods for DBD changed when the phase progresses. Anticholinergic drugs, such as tolterodine and solifenacin, are the main treatment options for DBD patients with OAB. However, many side effects, including dry mouth, dry eyes, and memory loss, occurs after the treatment with anticholinergic drugs, thus rendering poor life quality for the patients. In the late phase, surgical intervention was the only therapeutic method for patients who did not benefit from pharmacological and behavioral treatments ([@B45]). However, pharmacological and surgical interventions were largely ineffective in clinics ([@B12]; [@B40]). Therefore, new effective treatments for DBD are urgently needed. In the treatment of diabetic OAB, traditional Chinese medicine and natural plant components have recently attracted increasing attention due to their safeness, few side effects, and excellent activity ([@B41]). SQW is a traditional Chinese herbal formula that was first recorded on *Fu Ren Liang Fang* in the Southern Song Dynasty (between 1127 and 1279 CE). This medicine is a mixture of three Chinese medicines: *roots of Lindera aggregata (Sims) Kosterm. (Lauraceae), roots of Alpinia oxyphylla Miq., (Zingiberaceae), and rhizomes of Dioscorea oppositifolia L. (Dioscoreaceae)* at a 1:1:1 ratio ([@B14]). SQW has been used to treat lower urinary tract symptoms, such as nocturia, urgency, and child bedwetting for hundreds of years ([@B4]). We have recently reported that SQW had therapeutic effects on the OAB of bladder outlet obstruction rat models by modulating the TRPV1 expression ([@B18]). In China, SQW is often used in the clinical treatment of diabetic OAB. However, its mechanism remains unclear, and its therapeutic effect has not been investigated in animal studies. Therefore, we designed experiments to explore the effects and therapeutic mechanisms of SQW in diabetic OAB mouse model. Materials and Methods {#s1} ===================== Reagents and Materials ---------------------- Suo Quan Wan was purchased from Hunan Hansen Pharmaceutical Co., Ltd. (China), and the quality control was provided by the company based on Chinese Pharmacopeia employing by high performance liquid chromatography (HPLC) technology from SQW samples ([@B9]). Three Chinese herbals were ground and mixed evenly at a 1:1:1 ratio and appropriate volumes of distilled water were used to make these powders to SQW compound. The doses were adopted according to the Experimental Methodology of Pharmacology, based on clinical usage, the Bios method ([@B39]). SQW H was 2.208 g/kg, SQW M was 1.104 g/kg, and SQW L was 0.552 g/kg. The tolterodine dose for the positive group was 0.82 mg/kg. Streptozotocin was purchased from TOKU-E Co., Ltd. (Japan). HFD (45% fat) and control diet were purchased from Guangdong Medical Laboratory Animal Center (China). Tolterodine was purchased from Chengdu Dikang Pharmaceutical Co., Ltd. (China). Roche dynamic Bg meter was purchased from Hoffmann-La Roche Inc. (Switzerland), and carbachol was obtained from Shandong Bausch & Lomb Freda Pharmaceutical Co., Ltd. (China). α,β-methylene ATP was purchased from Tocris Bio-Techne Ltd. (United Kingdom). FastQuant RT Kit (with gDNAse) and Talent qPCR PreMix (SYBR Green) were purchased from TIANGEN Biotech (Beijing) Co., Ltd. (China). TRIzol reagent was purchased from Thermo Fisher Scientific (United States). RIPA lysis buffer and protease inhibitor cocktail (100×) were obtained from CoWin Biosciences (China). All other reagents used were of analytical grade. Preparation and HPLC Conditions of SQW -------------------------------------- Suo Quan Wan samples were weighted 0.3 g and extracted with 25 mL of methanol-hydrochloric acid solution using heating reflux method and then cool the solution. Finally, the solution was filtered through 0.45 μm nylon membranes before injection. According to the Chinese pharmacopoeia 2015, the content of norisoboldine should be more than 0.4 mg/0.3 g, and the content of allantoin is more than 0.48 mg/0.3 g. The HPLC conditions and gradient elution were shown as [Tables 1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}. ###### Chromatographic condition for norisoboldine. Column C18 (25°C) ------------ ----------------------------------------- ------- Solvent A Acetonitrile Solvent B 0.5% formic acid and 0.1% triethylamine Flow rate 1.0 mL/min Wavelength 280 nm Time (min) A (%) B (%) 0 10 90 13 22 78 30 22 78 ###### Chromatographic condition for allantoin. Column C18 (25°C) ------------ ------------ ------- Solvent A Methanol Solvent B H~2~O Flow rate 1.0 mL/min Wavelength 191 nm Time (min) A (%) B (%) 0 8 92 10 10 90 20 10 90 Animal Model and Treatment -------------------------- All experimental protocols and animal procedures complied with the ethical principle guidelines of the National Research Council. A total of 100 male C57BL/6J mice (18--22 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. and housed in the Experimental Animal Center of Guangzhou University of Chinese Medicine (No. S2017051, Guangzhou, China) under room temperature and exposed to a 12 h/12 h light--dark cycle, with free access to food and water. The animals were fed with normal diet for 3 days and then divided into two groups, namely, diabetic (*n* = 85) and control (*n* = 15) groups. The mice in the diabetic group were fed with HFD, whereas those in the control group received normal diet. After 4-week feeding, the mice in the diabetic group were injected with STZ at 100 mg/kg dissolved in citrate buffer for four times (0.05 M, pH 4.3--4.5). The mice in the control group were treated with an equal volume of vehicle (0.05 M citric acid, pH 4.3--4.5). Fasting Bg (FBG) was measured using an ACCU-CHEK advantage Bg monitoring system (Roche, Indianapolis, IN, United States) through the tail vein 72 h after the last injection. The mice with FBG levels above 11.1 mmol/L were considered diabetic and selected for subsequent experiments. The mice in the diabetic group were divided into five groups: model (*n* = 13), positive (tolterodine, 0.82 mg/kg, *n* = 13), high-dose (SQW H, 2.208 g/kg, *n* = 13), medium-dose (SQW M, 1.104 g/kg, *n* = 13), and low-dose (SQW L, 0.552 g/kg, *n* = 13) groups. After 3-week feeding, the six mouse groups individually received oral gavage of distilled water (control and model group mice), tolterodine, SQW H, SQW M, and SQW L for 3 weeks. During the experiment, the mice in the control group were given normal diet, whereas those in the other groups were continually fed with HFD. FBG Test and Oral Glucose Tolerance Test (OGTT) ----------------------------------------------- Fasting blood glucose test and OGTT were conducted after the 3-week SQW treatment. All animals were fasted overnight, and the Bg concentration was measured using a glucometer (ACCU-CHEK active) through a drop of tail blood. All the mice were then given with glucose (2 mg/g body weight) by gavage, and tail blood samples were obtained at 0, 15, 30, 60, 90, and 120 min to measure the Bg concentration. The area under the curve of the Bg time course from 0 to 120 min (AUC~0-2~ ~h~) was calculated according to the following formula: AUC 0 − 2 h = \[ ( Bg 0 \+ Bg 15 ) \] × 0.5 \] ÷ 4 \+ \[ ( Bg 15 \+ Bg 30 ) × 0.5 \] ÷ 4 \+ \[ ( Bg 30 \+ Bg 60 ) × 0.5 \] ÷ 2 \+ \[ ( Bg 60 \+ Bg 120 ) × 0.5 \] Measurement of Water Intake, Urine Output, and Frequency -------------------------------------------------------- The mice were placed individually in metabolic cages for 24 h with food and water *ad libitum*. Water intake was measured based on the water consumption for 24 h. Urine output and micturition frequency were analyzed through the VSOP test. Urine output was measured by evaluating the volume of urine in the collector after the mice were placed in the cages for 5 h. Micturition frequency was measured by visualizing and analyzing the collected papers, which were placed under the metabolic cage for 5 h under ultraviolet light to identify the area of urine spots ([@B38]). The sizes of the urine spots were divided into two levels, namely, bigger volume (\>50 μL) and smaller volume (\<50 μL) voids, to measure the micturition frequency. Urodynamic Test --------------- The urodynamic test was performed using a micro-injection pump and urodynamic measuring device (Laborite Delphis 94-R01-BT, Canada). All mice were anesthetized by the intraperitoneal injection of 25% urethane (2.0 mg/kg). A ventral midline incision was made to expose the bladder, and a 25-gauge needle was inserted into the bladder dome and fixed with silk suture. The needle was connected through a three-way adapter, which was connected with urodynamics at one end and a micro-injection pump at the other. After the bladder was emptied, 0.9% saline solution was injected into the bladder through the micro-injection pump at a rate of 3 mL/h. Pumping was stopped immediately when urine was observed at the external urethra. The bladder pressure line was automatically recorded with a computer. Urodynamic test parameters included the frequency of NVC (higher than 4 cm H~2~O spontaneous bladder contraction that did not result in urination before first urination) frequency, MBC (the volume of saline pumped before first urination), maximum voiding pressure (MP, the maximum peak pressure reached during micturition), RV (manually drained and measured with 1 mL syringe), MV (calculated as MBC - RV), VE (calculated as \[(MBC - RV)/MBC\] × 100%), and BC \[calculated as (MBC/MP) × 100%\]. The mice were euthanized at the end of the experiment through cervical dislocation ([@B19]). Histological Test ----------------- After the mice were euthanized, the bladders were excised and fixed using 4% paraformaldehyde solution for approximately 24 h at room temperature. After fixation, the bladders were conventionally dehydrated and embedded in paraffin. The tissues were cut into 6 μm thickness and stained with hematoxylin and eosin (HE) and Masson's trichrome. The color segmentation of Masson's trichrome was used to identify the whole cross-sectional area and the tissue areas that were stained "pink" (urothelium), "blue" (collagen), and "red" (smooth muscle). The HE images (100×) were used to determine the BWT, whereas the Masson's trichrome stained images (100×) to measure the smooth-muscle-to-collagen ratio. The stained bladder sections were examined under a light microscope, and representative images were photographed with a digital camera mounted on the microscope. All the images were analyzed using image analysis software (Image Pro 6.0). Assessment of Detrusor Smooth Muscle Contractility Study *in vitro* ------------------------------------------------------------------- The mice were executed, and the bladders were excised at the bladder neck. Full-thickness longitudinal DSM strips (0.7--1 mm × 5 mm) were obtained and mounted in a 5 mL organ bath filled with Krebs-Henseleit solution (NaCl, 118 mM; KCl, 4.75 mM; MgSO~4~, 1.18 mM; NaHCO~3~, 24.8 mM; KH~2~PO~3~, 1.18 mM; CaCl~2~ 2.5 mM; and C~6~H~12~O~6~⋅H~2~O, 10 mM; pH = 7.4) bubbled with a mixture of 5% carbon dioxide and 95% oxygen at 37°C. One side of the strip was attached to the hook with silk suture, and the other side was connected to the force signal transducer ([@B38]). The passive tension was loaded at 0.5 g, and the tissues were equilibrated for 60 min before the experiments. The forced change signals of the DSM strips were recorded with a PowerLab recorder. Purinergic agonist, α,β-methylene ATP (100 μM) was added to the organ bath twice (30 min between each assay) to measure the difference in the contractile responses. The contraction of bladder tissue to electrical field stimulation (EFS, 1, 2, 4, 8, 16, 32, and 64 Hz; 40 V; and 0.5 ms pulse duration for 10 s) was also measured. Furthermore, tests for dose--response curve to carbachol (10^-8^--10^-5^ M) and the contractile response to KCl (120 mM) were performed in the DMS strips. At the end of the experiments, the weight and length of each detrusor strip were recorded. Real-Time RT-PCR ---------------- The total RNA from the whole bladder samples were extracted using TRIzol Reagent (Invitrogen, United States). The absorption of the samples at 260 and 280 nm was used to estimate the RNA quality. A260/A280 was used to check the purity, and A260 values confirmed the concentration of RNA (Shimadzu BioSpec-nano, Japan). The total RNA was reverse transcribed into cDNA using a PrimeScript RT Reagent Kit with gDNA eraser (TIANGEN, China). Real-time PCR analysis was performed using SYBR Green (TIANGEN, China) according to the manufacturer's instructions. Synthetic oligonucleotide primers were designed to amplify the cDNA for the genes encoding the myosin Va, SLC17A9, and β-actin. [Table 3](#T3){ref-type="table"} shows the primer pairs. The reaction program was presented as follows: 95°C for 3 min, followed by 39 cycles at 95°C for 5 s and 55°C for 10 s. Results were recorded and analyzed using complementary software, and the gene expression levels were calculated by 2^-ΔΔCt^ method. The target gene expression levels were individually normalized according to the β-actin expression. ###### Primer sequences of myosin Va, SLC17A9, and β-actin. Gene Primers (5′--3′) ----------------- ------------------------ *myosin Va - F* AGCTCAACTCCTTCCACTC *myosin Va - R* ACACACCCCTTTATCCTTCC *SLC17A9 - F* GCTTCCTCAAGGCTATGATCTT *SLC17A9 - R* AGGTCCTGAATGTTGACTGAAA *β-actin - F* CTACCTCATGAAGATCCTGACC *β-actin - R* CACAGCTTCTCTTTGATGTCAC Western Blot Analysis --------------------- The bladder tissues were homogenized using tissue grinders (Shanghai Jingxin, Shanghai, China) at 65 Hz for 2 min to extract the total protein. BCA protein assay Kit (Beyotime Biotechnology, China) was used to measure the protein concentration. Equivalent proteins (20 μg) were subjected to 10% or 8% SDS-PAGE at 80 V for 30 min or 120 V for 60 min, respectively, to separate the proteins of different molecular weights and transfer to the PVDF membranes using the transblotting apparatus (Bio-Rad Laboratories, Hercules, CA, United States) for 55 or 110 min, respectively, at 300 mA. The PVDF membranes were blocked with 5% (w/v) non-fat milk buffer at room temperature for 2 h and incubated with a primary antibody in TBST \[Myosin Va (1:1000, Santa Cruz), SLC17A9 (1:1000, MBL), or β-actin (1:1000, 4A Biotech)\] overnight at 4°C. The immune-labeled membranes were washed three times with TBST for 15 min each time, and then conjugated with a secondary antibody (1:5000, 4A Biotech) at room temperature for 2 h. After the non-binding secondary antibodies were washed away, the target protein bands were visualized using a chemiluminescent reagent (Millipore, United States). Data were processed using ImageJ, and the immunoblot protein expression levels of myosin Va and SLC17A9 were normalized using β-actin. The antibodies used in the present study are listed in [Supplementary Table 1](#SM1){ref-type="supplementary-material"}. Statistical Analysis -------------------- All data were expressed as mean ± SD. Statistical analyses were performed using SPSS 19.0 (SPSS, United States). The amplitude of contractile responses to stimulus was recorded in tension (Newton) and normalized by the weight (g) of the detrusor strips ([@B5]). Western blot analysis data were processed using ImageJ. Histological test images were analyzed using Image-Pro Plus 6.0, and one-way ANOVA was used for data analysis. *P* \< 0.05 was considered statistically significant. Results ======= HPLC Analysis of SQW -------------------- For quality assessment of SQW, HPLC analysis was conducted. The detection wavelength of norisoboldine was set at 280 nm and the allantoin was set at 191 nm. The retention times of norisoboldine and allantoin were detected at approximately 17.960 and 11.632 min, respectively ([Figure 1A](#F1){ref-type="fig"}--[D](#F1){ref-type="fig"}). According to the chromatograms results, the contents of norisoboldine and allantoin in SQW sample were 0.72 mg/0.3 g and 0.73 mg/0.3 g, respectively, indicating that the SQW samples meet the requirement. {#F1} General Characteristics of the Diabetic Model --------------------------------------------- Compared with the mice in the control group, the T2DM mice exhibited diabetes characteristics, including significantly reduced weight (*P* \< 0.01) and increased water intake (*P* \< 0.01), urine volume (*P* \< 0.01), Bg levels \[high FBG (*P* \< 0.01), OGTT (*P* \< 0.01), and AUC~0-2h~ (*P* \< 0.01)\]. No considerable differences in these parameters were observed among the mice in SQW and model groups ([Figure 2A](#F2){ref-type="fig"}--[F](#F2){ref-type="fig"}). ![Effects of SQW on the general characteristics of diabetic model after the 3-week treatment (*n* = 8). **(A)** Weight, **(B)** water intake, **(C)** 5 h urine volume, **(D)** FBG, **(E)** OGTT, and **(F)** area under the curve of AUC~0-2~ ~h~ (calculated according to the following formula: AUC~0-2h~ = \[(Bg0 + Bg15) × 0.5\] ÷ 4+ \[(Bg15 + Bg30) × 0.5\] ÷ 4+\[(Bg30 + Bg60) × 0.5\] ÷ 2+\[(Bg60 + Bg120) × 0.5\]). Data represent the means ± SD (model vs. control group, ^∗∗^*P* \< 0.01 or ^∗^*P* \< 0.05).](fphar-10-00552-g002){#F2} VSOP and Urodynamic Tests ------------------------- The representative urodynamic response curves of each group are presented in [Figure 3](#F3){ref-type="fig"}. The urinary voiding patterns showed that the frequencies of bigger volume voids (\>50 μL) and smaller volume voids (\<50 μL) were higher in diabetic mice than in the controls [Figure 4A](#F4){ref-type="fig"}. Treatment with SQW M markedly decreased both frequencies (*P* \< 0.05), whereas treatments with SQW H and SQW L reduced the frequencies of smaller volume and bigger volume voids, respectively. In addition, urodynamic test revealed that compared with the controls, the diabetic mice had significantly increased NVC, MBC, RV, and BC (*P* \< 0.01) but markedly decreased VE (*P* \< 0.01), thereby showing typical DBD in the early compensated phase ([Figure 4B](#F4){ref-type="fig"}--[F](#F4){ref-type="fig"}). SQW M treatment remarkably decreased the NVC, MBC, RV, and BC (*P* \< 0.01 or *P* \< 0.05) but significantly increased the VE of the mice (*P* \< 0.01). Furthermore, treatments with SQW H and SQW L remarkably decreased the NVC of the mice (*P* \< 0.05). No significant differences in MP were found among the control, SQW-treated, and model mice ([Figure 4G](#F4){ref-type="fig"}). {#F3} {#F4} Morphometric Analysis --------------------- The bladder weight (absolute and relative to body weight) was increased in the diabetic mice (*P* \< 0.01) but decreased in the SQW M- and SQW L-treated mice (*P* \< 0.05) compared with that in the controls ([Figure 5A,B](#F5){ref-type="fig"}). The results of the morphometric analysis were consistent with the bladder weight. The BWT was significantly increased in diabetic mice (*P* \< 0.01), but SQW treatment effectively inhibited this alteration ([Figure 5C](#F5){ref-type="fig"}). No substantial differences in the smooth-muscle-to-collagen ratio were observed among the control, SQW-treated, and model mice ([Figure 5D](#F5){ref-type="fig"}). {#F5} Contractility Studies *in vitro* -------------------------------- We found that the DMS strips of diabetic mice exhibited significantly higher amplitudes of spontaneous activity that those of the controls (*P* \< 0.01). SQW H and SQW M treatments markedly decreased this alteration (*P* \< 0.01) ([Figure 6A](#F6){ref-type="fig"}--[C](#F6){ref-type="fig"}). α,β-methylene ATP (100 μM), which is the P2X receptor agonist, caused higher contractions in the DMS strips of diabetic mice compared with those of the controls (*P* \< 0.01). α,β-methylene ATP (100 μM) was added twice to activate the bladder strip. The responses evidently increased in the first reaction but markedly decreased in the second response in the diabetic mice compared with those in the controls (*P* \< 0.05). SQW H treatment markedly reverted all these alterations (*P* \< 0.01 or *P* \< 0.05), and SQW M treatment decreased the ATP-induced contractions ([Figure 6D,E](#F6){ref-type="fig"}). In addition, the contractions caused by KCl (120 mM), EFS (1--64 Hz) were higher in the diabetic mice than in the controls (*P* \< 0.01 or *P* \< 0.05). The cumulative concentration -response curve of carbachol (10^-8^--10^-5^ M) was also higher in the diabetic mice than in the controls (*P* \< 0.01 or *P* \< 0.05). The contractions of the DSM strips were markedly decreased due to the treatment with SQW (*P* \< 0.01 or *P* \< 0.05) ([Figure 6F](#F6){ref-type="fig"}--[I](#F6){ref-type="fig"}). No significant differences in pEC50 were found among the control, SQW-treated, and model mice. The Emax of diabetic mice exhibited significantly higher than the controls (*P* \< 0.01). Positive and SQW-M treatments markedly decreased (*P* \< 0.01) ([Table 4](#T4){ref-type="table"}). ![DSM strips of diabetic mice exhibited high amplitudes of spontaneous activity and increased bladder contractions to stimuli, and SQW treatment inhibited the changes (*n* = 5). **(A)** Representative spontaneous contractions of the bladder detrusor strips from the mice in the six groups. Quantification of the **(B)** amplitude and **(C)** frequency of spontaneous contraction; **(D)** DSM strip contractions induced by α,β-methylene ATP (100 μM). **(E)** ATP ratio (calculated as \[before-cons -- after-cons\]/pro-cons), α,β-methylene ATP (100 μM)-induced contractions (two times added α,β-methylene ATP); **(F)** DSM strip contractions induced by KCl (120 mmol/L); and **(G)** DSM strip contractions induced by EFS (1--64 Hz); and **(H)** carbachol (10^-8^--10^-5^ M); **(I)** The cumulative concentration--response curves of carbachol. Data represent the means ± SD (model vs. control group, ^∗^*P* \< 0.05 or ^∗∗^*P* \< 0.01; treatment vs. model group, ^\#^*P* \< 0.05 or ^\#\#^*P* \< 0.01).](fphar-10-00552-g006){#F6} ###### The pEC50 and Emax of carbachol (means ± SD, *n* = 5). Group Dose pEC50 Emax ---------- ------------ ------------- -------------------- Control -- 5.68 ± 0.38 15.07 ± 4.51 Model -- 5.93 ± 0.20 27.58 ± 7.76^∗∗^ Positive 0.82 mg/kg 6.23 ± 0.21 18.14 ± 8.11^\#\#^ SQW-H 2.208 g/kg 5.97 ± 0.65 22.95 ± 7.78 SQW-M 1.104 g/kg 5.94 ± 0.32 16.86 ± 4.68^\#\#^ SQW-L 0.552 g/kg 6.15 ± 0.24 24.59 ± 6.60 Model vs. control group, ∗∗ P \< 0.01; treatment vs. model group \#\# P \< 0.01. Real-Time RT-PCR Analysis ------------------------- According to the results of RT-PCR analysis, the mRNA expression levels of myosin Va and SLC17A9 were significantly increased in the diabetic mice compared with those in the control mice (*P* \< 0.01 or *P* \< 0.05). The 3-week SQW M treatment, markedly decreased the mRNA expression levels of myosin Va and SLC17A9 (*P* \< 0.01 or *P* \< 0.05), whereas the SQW H and SQW L treatments reduced the myosin Va mRNA expression level only (*P* \< 0.01 or *P* \< 0.05) ([Figure 7A,B](#F7){ref-type="fig"}). {#F7} Western Blot Analysis --------------------- The protein expression levels of myosin Va, SLC17A9, and β-actin were evaluated through Western blotting. The results showed that the protein expression levels of myosin Va and SLC17A9 were significantly increased in the bladder tissues of diabetic mice compared with those in the controls (*P* \< 0.01). After the 3-week SQW treatment, SQW M treatment markedly decreased the protein expression levels of myosin Va and SLC17A9, whereas SQW H and SQW L treatments significantly reduced the protein expression of myosin Va only (*P* \< 0.01) ([Figure 8A,B](#F8){ref-type="fig"}). {#F8} Discussion ========== Diabetic bladder dysfunction is a major lower urinary tract complication of diabetes, but its molecular mechanism remains unknown. Several studies reported that rat models with STZ-induced T2DM usually exhibited major clinical urodynamic alterations ([@B11]; [@B22]; [@B38]). We have previously reported that TCM, namely SQW, had therapeutic effects on many lower urinary tract diseases, including operation-induced OAB and outlet obstruction ([@B18]). Two components (norisoboldine and allantoin) were detected as quality standards in SQW, and the results of HPLC showed that the active ingredient content met the standard. SQW is an oral traditional Chinese formula for the treatment of lower urinary tract diseases, and the oral dose is 5.4 g/day. The high dose in our study was double the clinical dose, and the middle dose was equation, while the low dose was half. Also, we calculated the equivalent doses of mice administration according to the equation ([@B39]). In the present study, we explored the effects and potential mechanism of SQW on diabetic OAB by using STZ-induced T2DM mouse model. The results of our research showed that the model mice were characterized by a range of T2DM symptoms, including abnormal fat, carbohydrate metabolism, and high Bg levels, thereby indicating the successful establishment of the T2DM model. Our VSOP and urodynamic tests showed significant alternations in the micturition of diabetic mice as characterized by increased frequency of voids (smaller micturition was apparent), NVCs, MBC, BC, and RV and markedly reduced VE after 6 weeks of hyperglycemia status. These results indicated that the diabetic mice entered the stage of OAB, which is consistent with those of previous studies ([@B22]; [@B38]). SQW treatment effectively improved the bladder function of the T2DM mice but did not change the high Bg status. In healthy bladders, the contractions caused by ATP were limited ([@B22]). However, in the early phase of DBD, the purinergic-induced contractions account for up to 150% compared with those in healthy individuals ([@B1]). The transfer of ATP-filled vesicles to the varicose membranes and the effects of DBD on the transfer mechanism were poorly explored. Recent studies have provided key information about the contributions of SLC17A9 and myosin Va in the storage and exocytosis of ATP in secretory vesicles ([@B31]; [@B6]). SLC17A9 is a vesicular nucleotide transporter that plays an essential role in the specific transport of ATP inside purinergic vesicles ([@B30]). This transporter was discovered in various invertebrates and vertebrates, indicating that the molecular mechanisms of purinergic transmission are common in animals ([@B31]). Accurate localization of the neurotransmitter-filled vesicles in varicose membranes is indispensable for exocytosis ([@B29]). The local transport of organelles, including purinergic vesicles, requires energy that is provided by molecular motors, such as myosin Va. Myosin Va is a subtype of the unconventional myosin V and is primarily expressed in the central and peripheral nervous systems and melanocytes ([@B35]). The structure of myosin Va is provided for the continuous forward transport of intracellular cargoes ([@B32]), and its movement changes with cargo binding ([@B15]). Recent studies reported that myosin Va played a key role in purinergic vesicular transport and was closely associated with ATP-containing vesicles by binding directly to SLC17A9 ([@B6]). These findings provided us the research direction to explore the effects of SQW and the alterations in purinergic vesicular transport in DBD mice. The results *in vitro* contractility study on the bladder were in accordance with the "temporal theory" of DBD. In the early phase, the DSM of diabetic mice exhibited markedly high amplitudes of spontaneous activity and increased responsiveness to stimuli, such as α,β-methylene ATP, KCL, EFS, and carbachol ([@B11]). *In vitro* studies the results showed that the DMS of diabetic mice exhibited markedly high amplitudes of spontaneous activity compared with those of the controls, and the waves were disordered. The frequency of spontaneous activity remained stable, which was consistent with a previous report ([@B38]). Compared with those of the controls, the contractile responses of the DSM strips of diabetic mice to α,β-methylene ATP were substantially increased during the first response but markedly decreased during the second time. These results suggested that the DSM strips of diabetics exhibited increased responsiveness to exogenous purinergic agonists, and the bladder remained high responsiveness after the first stimulate in diabetic mice. In agreement with the above findings, the contractions of DSM strips to KCL, EFS, and carbachol were high in the diabetic mice. After the 3-week SQW treatment, the mice in the SQW treatment groups displayed varying degrees of reduction for their contractile responses to all stimulus, especially to α,β-methylene ATP, as compared with the models. This result was consistent with those of VSOP and urodynamic test, thus further confirming the therapeutic effects of SQW in the OAB of diabetic mice. The hyper-responsiveness of diabetic DSM to α,β-methylene ATP, KCL, EFS, and carbachol reflects the changes at the neurotransmitter level and/or beyond the neurotransmitters related to the alterations in the upstream vesicular nucleotide transporters. Therefore, we evaluated the expression levels of protein and mRNA for myosin Va and SLC17A9. In the bladders of the models, we found high expression levels of protein and mRNA for myosin Va and SLC17A9, indicating that the increased expression of these genes is the potential pathogenic mechanism of diabetic OAB. Moreover, we found that the expression of levels of proteins and mRNA of myosin Va and SLC17A9 were significantly decreased after SQW treatment, suggesting that the downregulation of expression levels of myosin Va and SLC17A9 contributes to the therapeutic effect of SQW in DBD. In this study, SQW was demonstrated to have effective treatment on DBD, and the possible mechanisms were also explored. However, the active ingredients of SQW are unclear, since the complexity of its component. To further explore the major effective ingredient in SQW is our goal in the future. We want to explore weather a single herbal has effects in DBD. In the present study, we explored the effects and potential mechanism of SQW on diabetic OAB by using STZ-induced T2DM mouse model. SQW is one of the most commonly traditional Chinese formula to treat various urinary system diseases in China for thousands of years ([@B18]), such as urinary incontinence, nocturnal enuresis and OAB symptom syndrome ([@B19]; [@B21]). And diabetic OAB has the same symptom like nocturia and urgency, SQW also was used to treat DBD or combined with other drugs in clinical ([@B25]; [@B8]). Studies has already found some mechanisms, such as β receptor, P2X, TRPV1 ([@B19]; [@B42]), but its complicated. We intend to conduct further research, learn its potential herbal components. Study found that radix linderae extracts have effects on OAB and diabetic bladder ([@B34]; [@B43]). The main components in radix linderae is norisoboldine and ursolide ([@B7]; [@B27]). We intend to conduct further research on efficacy and mechanism using the component in radix linderae. Conclusion ========== In summary, our study revealed that HFD with STZ-induced T2DM model mice showed OAB symptoms after 6 weeks of hyperglycemia status. In addition, the traditional Chinese formula, SQW, exhibited therapeutic effects on the OAB of model mice. SQW directly targeted the bladder, rather than improving the Bg levels. The mechanism was related to the inhibition of the transmission of purinergic neurotransmitters in the bladder of diabetic mice by downregulating the expression levels of myosin Va and SLC17A9. Ethics Statement ================ This study was carried out in accordance with the ethical principle guidelines of the National Research Council. The experimental protocol was approved by the Committee on Ethics of Guangzhou University of Chinese Medicine. Author Contributions ==================== H-yC conceived and designed the study. PH directed the experiments. JW, X-fY, W-jL, RW, L-yT, W-kR, L-jF, F-jC, and D-wL constructed the animal model. JW, X-fY, and Y-fX analyzed the data and drafted the manuscript. D-wL made a great contribution in the revision of the later articles. All authors read and approved the final manuscript. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. **Funding.** This study was supported by the National Natural Science Foundation of China (Grant No. 81673676) entitled "Effect and Mechanism of SQW on Neurotransmission Abnormality in the Treatment Recovery of Diabetic Cystopathy" and the Science and Technology Bureau (Grant No. 2019622101002). The authors thank the School of Fundamental Medical Science, Guangzhou University of Chinese Medicine for technical support and the Hunan Hansen Pharmaceutical Co., Ltd. for providing quality control materials. Supplementary Material ====================== The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fphar.2019.00552/full#supplementary-material> ###### Click here for additional data file. ANOVA : analysis of variance BC : bladder compliance Bg : blood glucose BWT : bladder wall thickness DBD : diabetic bladder dysfunction DSM : detrusor smooth muscle EFS : electrical-field stimulation FBG : fasting blood glucose HFD : high-fat diet MBC : maximum bladder capacity MV : micturition volume NVC : non-voiding contraction OAB : overactive bladder OGTT : oral glucose tolerance test RV : residual volume SD : standard deviation SQW : Suo Quan Wan SQW H : high-dose SQW SQW L : low-dose SQW SQW M : medium-dose SQW STZ : streptozotocin T2DM : type 2 diabetes mellitus VSOP : voided stain on paper VE : voided efficiency [^1]: Edited by: Adolfo Andrade-Cetto, National Autonomous University of Mexico, Mexico [^2]: Reviewed by: Geng Wenye, Fudan University, China; Fabiola Zakia Mónica, Campinas State University, Brazil [^3]: ^†^These authors have contributed equally to this work [^4]: This article was submitted to Ethnopharmacology, a section of the journal Frontiers in Pharmacology

Introduction {#sec1} ============ In contrast to bulk silver, nanometric silver materials exhibit many extraordinary properties such as a high surface-to-volume ratio,^[@ref1],[@ref2]^ quantum tunneling effects,^[@ref3],[@ref4]^ an abundance of free electrons,^[@ref5]^ surface plasmon resonance,^[@ref2],[@ref6]−[@ref8]^ and antibacterial behaviors.^[@ref9],[@ref10]^ Because of these unique properties, noble silver nanomaterials are widely applied in diverse areas, including thermotherapy,^[@ref5],[@ref11]^ medicine,^[@ref2],[@ref5],[@ref12]^ sensors,^[@ref13]−[@ref16]^ surface-enhanced spectroscopy,^[@ref6],[@ref17]−[@ref20]^ biology,^[@ref2]^ catalysis,^[@ref21]−[@ref28]^ and electronics.^[@ref29]−[@ref31]^ Among these many applications, the catalysis of the reduction of nitroarenes to aromatic amines is increasingly attracting attention because of pharmaceutical needs and the importance of this industry.^[@ref25],[@ref26],[@ref32]−[@ref38]^ Various strategies have been proposed to reduce nitroarenes more efficiently and more rapidly using silver nanomaterials. These strategies include depositing silver nanoparticles (AgNPs) on supports used as heterogeneous catalysts,^[@ref24],[@ref26],[@ref39]−[@ref42]^ combining AgNPs with reduced graphene oxide or graphene oxide as catalysts,^[@ref43]−[@ref47]^ and using silver nanocolloids as a quasi-homogeneous nanocatalyst.^[@ref25],[@ref33],[@ref48]−[@ref54]^ All of the aforementioned catalytic approaches are efficient and selective. However, the reaction rate for heterogeneous catalysis is rather low and quasi-homogeneous catalysis suffers from possible aggregation of the nanocatalyst. Furthermore, the procedures for preparing such nanocatalysts are somewhat complex and time-consuming. Applications of magnetic nanoparticles have also been extensively investigated in recent years because they feature many notable characteristics, such as a high surface-to-volume ratio, easy attraction and redispersion, and paramagnetism. Because of these crucial properties and advantages, the combination of AgNPs and magnetic nanoparticles has become one of the most favorable approaches for the catalytic reduction of nitroarenes.^[@ref55]−[@ref60]^ In this study, a simple but facile method was applied in a single step to prepare silver-doped magnetic nanoparticles (AgMNPs) for the catalytic reduction of nitroarenes through spontaneous oxidation--reduction and coprecipitation. When mixing Fe^2+^ with Ag^+^, a spontaneous reaction is caused by the difference in standard reduction potential between the ionic species. When Ag^+^ is reduced to Ag^0^, an equivalent number of moles of Fe^2+^ ions are simultaneously oxidized to Fe^3+^. After the addition of precipitation agents, AgNPs were coprecipitated with iron oxide magnetic nanoparticles, which led to the formation of AgMNPs. The proposed preparation can be achieved in a single step, and the prepared AgMNPs can subsequently be utilized as nanocatalysts for the reduction of *o*-nitroaniline (*o*-NA). The parameters (pH, temperature, and amount of nanocatalyst) that affect the morphology and composition of the prepared AgMNPs and efficiencies of the catalytic reduction were systematically studied to gain a greater understanding of the characteristics of the AgMNPs prepared using the method proposed in this study. Additionally, the catalytic activity of the AgMNPs prepared for the reduction of other nitroarenes and their recyclability were investigated to fully evaluate their potential for practical applications. Results and Discussion {#sec2} ====================== Effect of Oxidation--Reduction Time on AgMNP Preparation {#sec2.1} -------------------------------------------------------- [Figure S1](http://pubs.acs.org/doi/suppl/10.1021/acsomega.7b01987/suppl_file/ao7b01987_si_001.pdf) shows the typical measured hysteresis loops of the prepared AgMNPs, which confirmed that the prepared AgMNPs were paramagnetic and usable for further applications. [Figure [1](#fig1){ref-type="fig"}](#fig1){ref-type="fig"} depicts the transmission electron microscopy (TEM) images of AgMNPs obtained using various reaction times. From the images, dark-sphere-like Ag nanoparticles were mixed with light-colored Fe~3~O~4~ NPs because Ag has a higher electron density that allows fewer electrons to transmit.^[@ref61],[@ref62]^ The AgNPs formed after a 10 min reaction time were larger than those formed after a 2 min reaction time. Notably, the size of the Fe~3~O~4~ NPs was mostly unaffected by the reaction time. ![TEM images of prepared AgMNPs with oxidation--reduction times of (a) 2 min, (b) 8 min, and (c)--(f) 10 min, where the \[Fe^2+^\]~0~ to \[Ag^+^\]~0~ ratios are (a)--(c) 3:1, (d) 2:1, (e) 4:1, and (f) 6:1. \[Fe^2+^\]~0~ are all 12 mM, and the magnifications of the images are all 100 000×. The yellow arrows indicate the examples of AgNPs for each sample.](ao-2017-019876_0008){#fig1} [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}a shows the evolution of the UV--vis spectra for the reduction of *o*-NA catalyzed by AgMNPs over time. The absorbance peak at 412 nm, which corresponds to the characteristic *o*-NA peak,^[@ref63]^ decreased as the reaction proceeded. The variations of the spectra indicated that *o*-NA was reduced to 1,2-phenylenediamine (1,2-PPD).^[@ref41],[@ref64]^ The relative concentration (*C*~t~/*C*~0~) of *o*-NA was obtained by dividing the absorbance recorded at 412 nm at the specified time (*C*~t~) by the absorbance at 412 nm before the addition of AgMNPs (*C*~0~). The results in [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}b were plotted using various AgMNPs prepared using various reduction durations. In [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}b, the catalytic efficiency of the AgMNPs shows no significant difference between when the reaction time was 4--10 min and when more than 95% of *o*-NA was reduced within 240 s. This finding suggests remarkable catalytic activity. By contrast, achieving the same conversion percentage required more than 500 s when the reaction time to prepare the AgMNPs was 2 or 12 min. Because the catalytic efficiency of nanocatalysts depends on their size and the amount of catalyst loaded,^[@ref26]^ we concluded that the reaction time to achieve the optimal morphology and catalyst loading was 10 min. For comparison, the results of an experiment conducted in parallel, where AgMNPs were replaced with Fe~3~O~4~ NPs, are also plotted in [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}b, showing that the reduction of *o*-NA can proceed only when AgNPs are doped. In the absence of AgNPs, the reduction of *o*-NA is suspended. {ref-type="fig"}](#fig1){ref-type="fig"}c. (b) *C*~t~/*C*~0~ of 1 mM *o*-NA (412 nm) versus the catalytic reduction time in the presence of 30 mM NaBH~4~ and AgMNPs prepared with (■) 2 min, (●) 4 min, (▲) 6 min, (▼) 8 min, (◆) 10 min, and (★) 12 min of oxidation--reduction time during the preparation, where the other conditions are the same as described in (a). Parallel experiment (□) uses 20 mg of Fe~3~O~4~ NPs as nanocatalysts, where the other conditions are the same as described as (a).](ao-2017-019876_0001){#fig2} Effect of Fe^2+^/Ag^+^ on AgMNP Preparation {#sec2.2} ------------------------------------------- As described in the previous section, catalytic efficiency is related to the size and amount of doped AgNPs. In addition, we expected the morphology and amount of doped AgNPs to be affected by the ratio of initial concentration of Fe^2+^ to Ag^+^ because AgNO~3~ acts as the oxidation agent in the formation of AgMNPs. [Figure [1](#fig1){ref-type="fig"}](#fig1){ref-type="fig"} shows the TEM images of the AgMNPs with different ratios of \[Fe^2+^\]~0~ to \[Ag^+^\]~0~. The images reveal that the morphologies of the AgNPs are similar. [Figure S2](http://pubs.acs.org/doi/suppl/10.1021/acsomega.7b01987/suppl_file/ao7b01987_si_001.pdf) shows the typical X-ray diffraction (XRD) spectra of the prepared AgMNPs and illustrates that the locations and intensity of the diffraction peaks were consistent with the standard patterns for JCPDS card no. (79-0417) magnetite and JPCPDS card no. (4-0783) standard Ag crystal. The size of AgNPs can by estimated using the Scherrer equation, and all AgNPs are approximately 20 nm.^[@ref10]^ To further explore the effect of the initial Fe^2+^ to Ag^+^ concentration ratio, energy-dispersive X-ray spectroscopy (EDS) analysis was performed and the typical spectra for the AgMNPs are shown in [Figure S3A](http://pubs.acs.org/doi/suppl/10.1021/acsomega.7b01987/suppl_file/ao7b01987_si_001.pdf). The atomic percentages of Ag, Fe, and O for various initial concentration ratios of Fe^2+^ to Ag^+^ obtained through EDS are plotted in [Figure S3B](http://pubs.acs.org/doi/suppl/10.1021/acsomega.7b01987/suppl_file/ao7b01987_si_001.pdf), which shows that the atomic percentage of Ag increased as the initial concentration ratio of Fe^2+^ to Ag^+^ decreased. When the initial concentration ratio of Fe^2+^ to Ag^+^ was 2.0, the atomic percentage of Ag in the AgMNPs was 8.23%. The atomic percentage decreased to 0.76% when the initial concentration ratio of Fe^2+^ to Ag^+^ was 6.0. [Figure [3](#fig3){ref-type="fig"}](#fig3){ref-type="fig"} plots the *C*~t~/*C*~0~ of *o*-NA as a function of the reduction time in the presence of AgMNPs prepared with various ratios of \[Fe^2+^\]~0~ to \[Ag^+^\]~0~. The figure shows that the AgMNPs prepared with a smaller ratio of \[Fe^2+^\]~0~ to \[Ag^+^\]~0~ had a higher catalytic efficiency. Through correlation with the EDS results, we concluded that the highest catalytic efficiency was obtained for the highest AgNP loading. Notably, although the catalytic efficiency of the AgMNPs was even higher when the initial concentration ratio of Fe^2+^ to Ag^+^ was 1.5, the yield of AgMNPs was very low because of weak magnetization and most particles formed after the coprecipitation stage could not be collected by the magnet. Similar results were observed when the initial concentration ratio of Fe^2+^ to Ag^+^ was 6.0. Therefore, we can conclude that the initial concentration ratio of Fe^2+^ to Ag^+^ should be between 2 and 3 to ensure that a sufficient amount of Fe^2+^ is oxidized so that the ratio of Fe^2+^ to Fe^3+^ is close to 2 before the coprecipitation agent is added to form magnetic Fe~3~O~4~ NPs.^[@ref65]^ ![*C*~t~/*C*~0~ of 1 mM *o*-NA versus catalytic reduction time in the presence of 30 mM NaBH~4~ and AgMNPs prepared with \[Fe^2+^\]~0~ to \[Ag^+^\]~0~ ratios of (■) 6:1, (●) 4:1, (▲) 3:1, (▼) 2:1, and (◆) 1.5:1, where the other conditions are the same as described in [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}a.](ao-2017-019876_0002){#fig3} Effects of pH and Temperature on the Catalytic Reaction {#sec2.3} ------------------------------------------------------- The acceleration of the reduction reaction by AgMNPs originates in a relay between the nucleophile and electrophile.^[@ref39]^ Therefore, the catalyzed reduction rate is affected by the abundance of electrons in the reaction system. To observe the relationship between the electron abundance and reaction rate, the catalysis reaction was performed at various pH values. The relative *o*-NA concentration (*C*~t~/*C*~0~) was recorded as a function of reaction time at various pH values, and the results are plotted in [Figure [4](#fig4){ref-type="fig"}](#fig4){ref-type="fig"}a. The pH played a key role in the catalysis reaction; when the pH was 9.8, the reduction of *o*-NA was completed within 4 min. The reduction rate significantly decreased when the pH was set to 8.8 and less than 85% of *o*-NA was reduced after approximately 5 min of reaction. When the pH was lower than 8, almost no conversion of *o*-NA could be observed. Because the acceleration of the reduction reaction by AgMNPs originates in a relay between the nucleophile and electrophile,^[@ref39]^ alkaline conditions enriched the electron densities on the AgNP surfaces by adsorbing more OH^--^, which promoted the reduction of *o*-NA. When the pH was set to 10.8, the reduction of *o*-NA decelerated, possibly because of the formation of yellow-colored 2,3-diaminophenazine under extremely alkaline conditions.^[@ref66]^ {ref-type="fig"}](#fig2){ref-type="fig"}a. (b) *C*~t~/*C*~0~ of 1 mM *o*-NA versus catalytic reduction time in the presence of 30 mM NaBH~4~ under different temperatures of (■) 0 °C, (●) 25 °C, and (▲) 40 °C, where the other conditions are the same as described in [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}a.](ao-2017-019876_0003){#fig4} The reaction rate was affected by the temperature of the reaction system because the reactants had more kinetic energy at higher temperatures and were able to cross the activation state more easily. [Figure [4](#fig4){ref-type="fig"}](#fig4){ref-type="fig"}b depicts the relationship between *C*~t~/*C*~0~ and reaction times at various reaction temperatures. As shown in [Figure [4](#fig4){ref-type="fig"}](#fig4){ref-type="fig"}b, the higher the temperature of the reaction system, the faster was the observed reduction of *o*-NA. Furthermore, when the temperature was 0 °C, the reduction rate of *o*-NA was similar to that at 25 °C in the first minute and *C*~t~/*C*~0~ was almost unchanged after the first minute, suggesting that the reduction was almost interrupted after 1 min at 0 °C. Effect of the Catalyst Amount on the Catalytic Reaction {#sec2.4} ------------------------------------------------------- As described in the previous sections, the conversion efficiency is related to the amount of AgNP loading. Accordingly, the conversion efficiency can also be related to the amount of AgMNPs used per experiment. [Figure [5](#fig5){ref-type="fig"}](#fig5){ref-type="fig"} plots the relationship between the *C*~t~/*C*~0~ of *o*-NA and reduction time when various amounts of AgMNPs were used. As shown in [Figure [5](#fig5){ref-type="fig"}](#fig5){ref-type="fig"}, when 1 mg of AgMNPs was used per experiment, approximately 20% of *o*-NA was reduced after 5 min of conversion and the conversion efficiency increased with the amount of AgMNPs used. When 20 mg of AgMNPs was used per experiment, the conversion of *o*-NA was almost 100% within 250 s at room temperature, which is promising for further applications. {ref-type="fig"}](#fig2){ref-type="fig"}a.](ao-2017-019876_0004){#fig5} Activity of AgMNPs for the Catalytic Reduction of Other Nitroarenes {#sec2.5} ------------------------------------------------------------------- After studying the characteristics and catalytic properties of the prepared AgMNPs, we investigated the catalytic reduction of other nitroarenes, including *m*-NA, *p*-NA, and *p*-NP, through the same procedures to investigate the ability of the AgMNPs to accelerate the reduction of other nitroarenes. [Figure S4](http://pubs.acs.org/doi/suppl/10.1021/acsomega.7b01987/suppl_file/ao7b01987_si_001.pdf) plots the evolution of the UV--vis spectra for reducing *m*-NA, *p*-NA, and *p*-NP in the presence of AgMNPs over time. The absorption maxima, which were located at 360 nm for *m*-NA, 380 nm for *p*-NA, and 400 nm for *p*-NP, decreased as the catalytic reduction proceeded. The relationship between *C*~t~/*C*~0~, where *C*~0~ is the absorbance at the initial time and *C*~t~ is the absorbance after the specific reaction time, for the nitroarenes tested and the reaction time under optimal conditions is plotted in [Figure [6](#fig6){ref-type="fig"}](#fig6){ref-type="fig"}. As indicated in the figure, the reductions of the four nitroarenes examined in this study were all completed within 4 min, which suggested that the AgMNPs prepared in this study were capable of catalytically reducing various nitroarenes. {ref-type="fig"}](#fig2){ref-type="fig"}a.](ao-2017-019876_0005){#fig6} Furthermore, as reported in other works, the catalytic reduction of nitroarenes follows the pseudo-first-order reaction kinetics.^[@ref26],[@ref43],[@ref51]^ The linear relations between ln(*C*~0~/*C*~t~) of nitroarenes examined in this study against reaction time were obtained and the rate constant (*k*) can be estimated by calculating the slopes of the obtained lines. The calculated *k* values at room temperature are 0.0192, 0.0145, 0.0185, and 0.0196 s^--1^ for the catalytic reduction of *o*-NA, *m*-NA, *p*-NA, and *p*-NP, respectively. To compare the results obtained in this study to those reported recently, [Table [1](#tbl1){ref-type="other"}](#tbl1){ref-type="other"} tabulates the catalytic activities of various AgNP-based catalytic systems. As can be observed in the table, the rate constants of AgMNPs prepared in this study for the reduction of nitroarenes are as good as other reported nanocatalysts. Moreover, because the effects of temperatures have also been studied in the previous sections, the thermodynamic parameters for the catalytic reduction of *o*-NA by AgMNPs prepared in this study can be calculated by following the Arrhenius and Eyring equations.^[@ref67],[@ref68]^ The calculated activation energy (*E*~a~) is 39.88 kJ/mol, activation enthalpy (Δ*H*) is 37.34 kJ/mol, and activation entropy (Δ*S*) is −123.29 J/(mol K). These results suggest that the AgMNPs prepared in this study are excellent nanocatalysts for the reduction of nitroarenes. The catalytic reduction of nitroarenes by metal nanoparticles is generally explained by the Langmuir--Hinshelwood mechanism, where both reactants are adsorbed on nanocatalyst surfaces and reaction occurred after the adsorption.^[@ref67],[@ref68]^ As a result of conversion, products are formed and then desorb from nanocatalyst surfaces. According to the Langmuir--Hinshelwood mechanism, the rate of catalytic reduction depends on the surface coverage reducing agent and nitroarene molecules.^[@ref69]−[@ref71]^ This mechanism rationalizes the large rate constant and small activation energy obtained in this study because the high surface-to-volume ratio and the quasi-homogeneous reaction conditions increase the surface coverage of reactants significantly. ###### Comparison of Catalytic Activities of Several AgNP-Based Systems catalysts nitroarene (final concentration; mM) concentration of NaBH~4~ (mM) temperature apparent rate constant (s^--1^) ref ------------------------------------------ -------------------------------------- ------------------------------- ------------------------------------ --------------------------------- ------------ AgNPs/polydopamine/anodic aluminum oxide *o*-NA (1.13) 400 RT[a](#t1fn1){ref-type="table-fn"} 0.0013 ([@ref40]) biogenic AgNPs *p*-NP (0.20) 10 [b](#t1fn2){ref-type="table-fn"} 0.00406 ([@ref41]) AgNPs/partially reduced graphene oxide *p*-NP (0.10) 13 RT 0.0374 ([@ref43]) AgNPs on porous glass filters *o*-NA (1.00) 30 50 °C 0.0094 ([@ref26])   *p*-NA (1.00) 30 50 °C 0.0071 ([@ref26]) AgNPs in microgels *p*-NP (0.08) 24 22 °C 0.0153 ([@ref50]) AgNPs in microgels *o*-NA (0.09) 18 22 °C 0.0067 ([@ref51])   *p*-NA (0.09) 18 22 °C 0.0101 ([@ref51])   *p*-NP (0.09) 18 22 °C 0.0052 ([@ref51]) AgNPs on fibrous nanosilica *o*-NA (0.17) 22 RT 0.0043 ([@ref52])   *p*-NP (0.099) 83 RT 0.01 ([@ref52]) Fe~3~O~4~\@SiO~2~/Ag nanocomposite *p*-NP (0.06) 6 25 °C 0.00767 ([@ref59]) AgNPs/HLaNb~2~O~7~ *p*-NP (0.091) 18 [b](#t1fn2){ref-type="table-fn"} 0.00301 ([@ref53]) this study *o*-NA (1) 30 RT 0.0192     *p*-NA (1.00) 30 RT 0.0185     *p*-NP (1.00) 30 RT 0.0196   Room temperature. Not mentioned. Recyclability of the Ag Nanocatalysts {#sec2.6} ------------------------------------- The recyclability of the AgMNPs prepared in this study was evaluated by consecutively reusing the nanocatalysts for the catalytic reduction of *o*-NA. As shown in [Figure [7](#fig7){ref-type="fig"}](#fig7){ref-type="fig"}, only approximately 40% of *o*-NA was reduced to 1,2-PPD after 10 min of reaction in the second consecutive test. This is possibly related to the adsorption of *o*-NA or 1,2-PPD on the surface of the silver nanocatalysts, which consequently reduced the electron transferability.^[@ref26]^ To reactivate the silver nanocatalysts, we soaked the used AgMNPs in an aqueous solution at pH 3 for 20 min and rinsed them with neutral water before the next use. As shown in [Figure [7](#fig7){ref-type="fig"}](#fig7){ref-type="fig"}, more than 90% of the *o*-NA was reduced within 5 min and more than 95% was reduced within 8 min. Furthermore, after treating the reused AgMNPs with an aqueous acidic solution, the AgMNPs could be recycled and their performance was similar to that in the previous run. Therefore, the AgMNPs prepared in this study can be recycled after treatment with an acidic solution, which significantly extends the practical applicability of these silver nanocatalysts. {ref-type="fig"}](#fig2){ref-type="fig"}a.](ao-2017-019876_0006){#fig7} Conclusions {#sec3} =========== In this study, a simple but facile approach to prepare silver-doped magnetic hybrid nanoparticles was proposed based on a chemical reduction and coprecipitation method. The nanoparticles prepared through this method were used as nanocatalysts for the reduction of *o*-NA. Using the AgMNPs prepared in this study as nanocatalysts exploits the advantages of quasi-homogeneous reaction conditions and enables the easy removal of nanocatalysts from the solution with a magnet. The results indicated that the composition of the AgMNPs prepared can be tuned by adjusting the ratio of \[Fe^2+^\]~0~ to \[Ag^+^\]~0~ and the chemical reduction time during the production of AgMNPs. During the catalytic reduction of *o*-NA, the pH and temperature of the system affect the reduction rate, which is also affected by the amount of nanocatalyst used in the reaction. Furthermore, the prepared AgMNPs were applicable to the catalytic reduction of other nitroarenes. Finally, the silver-doped magnetic nanocatalysts proposed in this study have several advantages, namely, easy preparation, significant catalytic activity at room temperature, high conversion ability, and recyclability, all of which enhance their usefulness for real applications. Experimental Section {#sec4} ==================== Materials {#sec4.1} --------- Ferrous sulfate and ferric chloride were obtained from Showa Chemical (Tokyo, Japan). Silver nitrate, *o*-NA, *m*-nitroaniline (*m*-NA), *p*-nitroaniline (*p*-NA), and *p*-nitrophenol (*p*-NP) were purchased from Alfa Aesar (Ward Hill, MA). Sodium borohydride was obtained from Acros Organics (Geel, Belgium). Ammonium hydroxide (28--30%, v/v) and nitric acid were purchased from Fisher Scientific (Hampton, NH). All chemicals were of reagent grade and used as received without further purification. Deionized Milli-Q water (Simplicity, Millipore, Burlington, MA) was used throughout this study. Preparation of AgMNPs {#sec4.2} --------------------- The preparation of AgMNPs was based on a chemical reduction and coprecipitation method. Briefly, 100 mL of 12 mM ferrous aqueous solution was mixed with various volumes of 200 mM silver nitrate aqueous solution under vigorous stirring for a specified amount of time. During stirring, a spontaneous oxidation--reduction reaction occurred between Ag^+^ and Fe^2+^. Ag^+^ was reduced to Ag^0^ and an equivalent number of moles of Fe^2+^ ions were oxidized to Fe^3+^. After the specified reaction time, 50 mL of 1.44 M ammonia solution, which acted as the precipitating agent, was rapidly added to the solution under vigorous stirring for 10 min to complete the coprecipitation process. After 3 h in storage, the formed nanoparticles were collected with a magnet and washed three times with distilled water and ethanol. Finally, the washed AgMNPs were dried in an oven at 140 °C for 8 h before further use. As an alternative for comparison, magnetite nanoparticles (Fe~3~O~4~ NPs) without silver doping were prepared following previous reports.^[@ref72]^ We conducted transmission electron microscopy (TEM) with a Hitachi HT-7700 microscope operated at 100 kV, energy-dispersive X-ray spectroscopy (EDS) analysis with a Hitachi SU-8010 microscope at an accelerating voltage of 15.0 kV, and powder X-ray diffraction (PXRD) with a Siemens D5000 XRD system to characterize the morphologies and compositions of the prepared AgMNPs. Hysteresis loops of the prepared AgMNPs were recorded at room temperature with a Quantum Design MPMS 3 SQUID vibrating sample magnetometer system. Reduction of Nitroaniline Catalyzed by AgMNPs {#sec4.3} --------------------------------------------- The catalytic efficiency of the AgMNPs was evaluated using the nanoparticles as prepared for the catalytic reduction of *o*-NA. A specific amount of AgMNPs was mixed with 15 mL of an aqueous solution consisting of 1 mM *o*-NA and 30 mM NaBH~4~ at room temperature. Ultraviolet--visible (UV--vis) spectra of the solution were recorded at chosen intervals. All UV--vis spectra in this study were measured using a Thermo Fisher Scientific Genesys 10S Bio UV--Vis spectrometer with a 1 nm resolution. The spectra were recorded within a wavelength range of 250--550 nm. The optical path of the UV--vis cell was 3 mm. The Supporting Information is available free of charge on the [ACS Publications website](http://pubs.acs.org) at DOI: [10.1021/acsomega.7b01987](http://pubs.acs.org/doi/abs/10.1021/acsomega.7b01987).Hysteresis loops, XRD spectrum, and EDS analysis results of the AgMNPs, and the UV--vis spectra of the reduction of *m*-NA, *p*-NA, and *p*-NP in the presence of AgMNPs with increasing times ([PDF](http://pubs.acs.org/doi/suppl/10.1021/acsomega.7b01987/suppl_file/ao7b01987_si_001.pdf)) Supplementary Material ====================== ###### ao7b01987_si_001.pdf The authors declare no competing financial interest. The authors acknowledge the financial support from the Taiwan Ministry of Science and Technology under grant MOST106-2113-M-037-016. This work was also supported by the Kaohsiung Medical University Research Foundation under grant KMU-M106016.

INTRODUCTION {#s1} ============ Whilst oncogenesis is driven by a multitude of complex, non-programmed molecular events, there are a number of key features of this process, not least of which is the aberrant activation of genes that would normally be silenced in a given tissue context \[[@R1]\]. The so called cancer/testis (CT) or cancer germline (CG) genes are one such group of genes that are frequently activated in a range of different human cancer types \[[@R2]-[@R4]\]. These genes have expression normally restricted to the human germline, many being testis-specific \[[@R2]-[@R4]\]. They have come under intense scrutiny since their original identification as the immunological privilege of their normal germline setting means that the proteins they encode can elicit an immunological response when aberrantly produced in cancers and so have exceptional potential in immunotherapeutics \[[@R5]\]; for example, the *NY-ESO-1* gene product has been successfully targeted in an adoptive therapeutic approach to melanoma therapy \[[@R6]\]. Despite this interest, remarkably little is known about the normal germline function of most CT genes. Moreover, it has been demonstrated that germline genes in *Drosophila melanogaster* are required for the oncogenic process and that the human orthologues of these *Drosophila* genes have up-regulated expression in a range of human cancers, although the functional implications for oncogenesis of this up-regulation remains unclear \[[@R7],[@R8]\]. Interestingly, down-regulation of a number of CT genes in human cancer cells results in perturbation of cellular proliferative potential \[for example, see [@R9],[@R10]\]. These findings open up the exciting possibility that CT genes might encode functions that are required for tumour homeostasis and it has recently been proposed that tumours become 'addicted' to these germline factors \[[@R11],[@R12]\], and recently, meiotic factors have been shown to contribute to telomere maintenance in cancer cells via the ALT pathway \[[@R13], [@R14]\]. The full extent of germline gene requirement is unclear, but these findings expose a new therapeutic opportunity by directly targeting the tumour-associated function of the CT gene products. Additionally, a number of studies have revealed another clinically important feature of CT genes; their expression appears to drive drug resistance as depletion of the gene products results in enhanced sensitization to anti-cancer drugs \[for example, see [@R15]\] expanding the therapeutic potential of this important class of cancer genes. Germline gene expression profiling has also recently been demonstrated to have applications in prognostics and patient stratification. In a seminal study, Rousseaux and co-workers demonstrated that expression of a sub-set of germ line genes in some lung cancers delineated patients with aggressive, metastasis prone tumours with poor prognosis \[[@R16]\]; they extended this by indicating that this cohort of patients might benefit from a drug therapeutic regime that had previously been dismissed for more general use in lung cancer patients, indicating that profiling patients for expression of a small sub-set of germline genes could be used in therapeutic decision making. Understanding germline gene expression is also critical as drug-induced augmentation of expression has also been postulated to be a potential enhancer of immunotherapeutics, the rationale being that further up-regulation of a tumour-specific antigen will result in enhanced immunological targeting of the tumour \[for example, see [@R17]\]. Taking all these factors together reveals the importance of understanding the regulatory mechanisms for somatic germline gene silencing and their aberrant activation in tumours. To date, the regulation of a number of CT genes has been studied and it has been demonstrated that DNA methylation of regulatory elements, such as promoter-associated CpG islands plays a fundamental role in the somatic silencing of these genes and the hypomethylation of these regulatory DNA regions in cancers is linked to gene activation \[for example, see [@R18]-[@R23]\], whereas gene body hypomethylation has been linked to gene down regulation in cancers \[[@R24]\]. Expression of these genes also becomes activated or further up-regulated upon enforced hypomethylation by the DNA methyltransferase inhibitor 5-aza-2′-deoxycytidine (5-aza-CdR), and to date, all CT genes studied have up-regulated expression in response to this chemotherapeutic agent, indicating a commonality in the mechanistic pathway for somatic CT gene silencing \[for example, see [@R18]-[@R23]\]. To date, most of the CT genes whose expression has been studied are located on the X chromosome (X-CT genes) and belong to large paralogous gene families \[[@R2]-[@R4]\]. Recently, a computational pipeline combining expressed sequence tag and microarray meta-analyses of the human orthologues of mouse spermatocyte-specific genes revealed a large cohort of new CT genes that were expressed in a broad spectrum of cancer types \[[@R25]-[@R29]\]. Unlike the X-CT genes, the majority of these genes are autosomally encoded and are single copy. To date, the clinical potential of these genes remains largely unexplored. In this current study, analysis of the expression of a small sub-set of these genes reveals a novel feature of CT genes, which indicates that some have a unique mechanism for somatic transcriptional silencing. This is a significant finding as these genes and their associated gene products have an increased prominence in clinical applications and hence the sub-classification of CT genes will play an important role in diagnostics, stratification and therapeutics. RESULTS {#s2} ======= All CT genes studied to date (mostly X-CT genes) require hypermethylation of regulatory DNA sequences for somatic silencing and are activated by the hypomethylating agent 5-aza-CdR. Given the clinical potential of enhanced up-regulation of immunogenic CT antigens, we set out to explore whether a similar DNA hypermethylation silencing mechanism was operating in the recently identified autosomally encoded CT genes \[[@R25],[@R27]\]. To do this, we selected a small sub-group of these genes that remained transcriptionally silenced in the colorectal cancer cell lines HCT116 and SW480 (*ARRDC5, C4orf17, C20orf201, DDX4, NT5C1B, STRA8, TDRD12*). We also selected two previously characterized CT genes (both X-CT genes) that remained transcriptionally silenced in these two cell lines to serve as exemplar controls for hypermethylation regulated CT genes, *SSX2* and *GAGE1*. To determine whether the novel CT genes are silenced via hypermethylation mediated mechanisms, similar to the characterized X-CT genes, we treated the two cell lines with the DNA methyltransferase inhibitor 5-aza-CdR to determine whether inhibition of DNA methyltransferase activity can activate these genes. Following 5-aza-CdR treatment of HCT116 and SW480 we made cDNA and carried out RT-PCR and agarose gel electrophoresis analysis of the products. The two X-CT genes were activated from the silent state with relatively low levels of 5-aza-CdR (0.1 μM; Figure [1](#F1){ref-type="fig"}; Figure [2](#F2){ref-type="fig"}). Some of the novel, autosomally encoded CT genes were similarly activated (*C20orf201, DDX4, STRA8, TDRD12*), although *C20orf201* and *DDX4* required a slightly higher 5-aza-CdR concentration for activation (0.5 μM; Figure [1](#F1){ref-type="fig"}; Figure [2](#F2){ref-type="fig"}). Additionally, activation of *STRA8* requires slightly higher concentrations of 5-aza-CdR in SW480 (Figure [2](#F2){ref-type="fig"}) than HCT116 (Figure [1](#F1){ref-type="fig"}), which indicates subtle regulatory differences between tumour cell types. However, surprisingly, three genes (*ARRDC5, C4orf17, NT5C1B*) remained tightly transcriptionally silenced, even at high concentrations of 5-aza-CdR in both cell lines (15.0 μM; Figure [1](#F1){ref-type="fig"}; Figure [2](#F2){ref-type="fig"}). This unexpected result reveals an important distinction in the way CT gene silencing is epigenetically regulated, revealing a hypermethylation-independent pathway. Interestingly, the X-CT genes (*GAGE1, SSX2*) remained activated for a prolonged period following removal of the hypomethylating agent, as did the autosomally encoded CT genes that were activated with the lowest concentration of 5-aza-CdR (*STRA8, TDRD12*) (Figure [3](#F3){ref-type="fig"}); however, the other two autosomally encoded CT genes, *C20orf201* and *DDX4*, which required slightly higher concentrations of 5-aza-CdR for activation, reverted to the silent state relatively soon after removal of the hypomethylating agent (Figure [2](#F2){ref-type="fig"}). This indicates a much greater transcriptional elasticity to the methylation-dependent silencing mechanisms for some CT genes. ![A sub-group of germline genes remain refractory to activation by epigenetic modulating agents\ RT-PCR was used to analyse activation of a group of germline genes that are normally silenced in the cancer cell line HCT116 (an additional colorectal cell line gives similar results \[see Supplementary Figure S1)\]. Whilst a cohort of known and newly identified germline genes become activated at low doses of the demethylating agent 5-aza-CdR (*GAGE1, SSX2, STRA8, TDRD12*) and others become activated with slightly higher levels of 5-aza-CdR (*C20orf201, DDX4*), some remain tightly silenced, even at high concentrations of 5-aza-CdR (*ARRDC5, C4orf17, NT5C1B*) (left column). The histone deacetylase inhibitor trichostatin A (TSA) has little activating potential (other than for *GAGE1* and *STRA8*, indicating the primary epigenetic regulation is mediated by DNA methylation (right column). Untreated and DMSO treated cells exhibit no activation of any of the genes analysed for expression activation. The chromosomal location of each gene is provided in parentheses to the right of the gene name. RT-PCR of β*ACT* shows uniform sample quality and loading.](oncoscience-01-0745-g001){#F1} {#F2} {#F3} To determine whether the silencing of hypermethylation-independent genes (*ARRDC5, C4orf17, NT5C1B*) was mediated via histone deacetylation we also treated the HCT116 and SW480 cells with the histone deacetylase (HDAC) inhibitor trichostatin A (TSA) (Figure [1](#F1){ref-type="fig"}; Figure [2](#F2){ref-type="fig"}) or a combination of 5-aza-CdR and TSA (data not shown). Remarkably, all three genes (*ARRDC5, C4orf17, NT5C1B*) remained tightly silenced under these highly transcriptionally permissive conditions. DISCUSSION {#s3} ========== CT antigens are potentially powerful targets for therapeutics, including immunotherapeutics. However, intratumour CT antigen gene expression is often heterogeneous and so there will be a lack of uniformity for any targeting strategy. To overcome this, it has been demonstrated that treatment of tumours with agents that deregulate epigenetic silencing, such as agents that result in DNA hypomethylation can generate a uniform expression of CT antigen genes within a tumour \[for example, see [@R18]-[@R23]\]. However, to date, the epigenetic regulation mechanisms for CT gene silencing has been restricted to a limited number of X-CT genes, all of which are activated by hypomethylating agents. Here we extended the analysis of epigenetic regulation of clinically important biomarkers and reveal that there is a cohort of CT genes that is not activated in response to hypomethylating agents (or HDAC inhibitors). This regulation is not simply due to a lack of methylation target CpG islands within the promoter regions as at least two of the genes (*ARRDC5* and *NT5C1B*) have reported CpG islands in their transcriptional promoter regions \[[https://genome.ucsc.edu/](http://genome.ucsc.edu/)\]. These observations indicate that there is a very broad range of mechanisms controlling CT gene regulation. This has implications for CT gene selection for clinical targeting strategies. Moreover, the mechanistic regulatory pathways might indicate sub-groups of CT genes that are co-regulated, which has implications for the study of these genes both as biomarkers, potential oncogenes and/or encoders of drug targets. Additionally, it has been demonstrated that some CT genes are required for tumour cell proliferation. Turning off these genes could reduce the proliferation-mediated burden of tumours, restricting their disease effect and/or enhancing other therapeutic approaches. MATERIALS AND METHODS {#s4} ===================== Maintenance and culturing of human colorectal cell lines HCT116 and SW480 {#s4_1} ------------------------------------------------------------------------- HCT116 and SW480 cell lines were obtained from the European Collection of Cell cultures. Both lines are tested for authenticity once per annum by LGC StandardsTM (authentication tracking number 710236782). HCT116 cells were grown in McCoy\'s 5A medium with GLUTAMAX^TM^ (Invitrogen, GIBCO 36600) and SW480 cells were grown in Dulbecco\'s modified Eagle\'s medium with GLUTAMAX^TM^ (Invitrogen, GIBCO 61965). Both media types were supplemented with 10% foetal bovine serum (Invitrogen; GIBCO 10270). Cells were incubated in humidified incubators at 37°C in a 5% CO~2~ atmosphere. Cells cultures were tested for mycoplasma infection using the LookOut^TM^ Mcycoplasma PCR Detection kit (Sigma Aldrich, MP0035). Epigentics modulating agents were added to the concentrations required as indicated in the main text. Treatment with 5-aza-CrD and TSA was for 48 hours (72 hour treatment yielded identical results). RNA extraction, cDNA synthesis and polymerase chain reaction {#s4_2} ------------------------------------------------------------ Total RNA was isolated using Trizol reagent (Invitrogen; 15596-026). Confluent cells were homogenised in Trizol (1 ml Trizol / 5×10^6^ cells) and held at room temperature (RT) for 5 minutes. Chloroform (200 μl per 1 ml of Trizol) was added to each sample and the homogenate was vigorously shaken for 15 seconds, followed by incubation for 5 minutes at RT. Samples were then centrifuged at 12,000 *g* for 15 minutes at 4°C. The aqueous layer was then removed to a new Eppendorf tube and 500 μl of isopropanol was added. After incubation at RT for 10 minutes, the samples were centrifuged again at 12,000 *g* for 20 minutes. The supernatant was removed and the pellet was washed with 70% ethanol and re-centrifuged at 7,500 *g* for 5 minutes at 4°C. The supernatant was discarded again and the cell pellet was left to dry at RT for 5-10 minutes, and then 100 μl RNase free water containing 2 μl DNase I (Sigma; D5319) was added to each RNA preparation sample. The samples were incubated at 37°C for 10 minutes and then at 75°C for 10 minutes. RNA quality and concentration was measured with a NanoDrop (ND 1000) spectrophotometer. Total RNA from normal human testis tissues was supplied by Clontech (Catalogue number; 636643) {#s4_3} ---------------------------------------------------------------------------------------------- Total RNA was used to synthesise cDNA using a SuperScript III First Strand Synthesis Kit (Invitrogen; 18080-051). Samples (1-2 μg) of total RNA were used according to the manufacturer\'s protocol. PCR using β*ACT* primers was used to check the cDNA quality. Gene sequences were obtained from the National Center for Biotechnology Information (<http://www.ncbi.nlm.nih.gov>). Primers were designed to span more than one intron where possible. Primers were designed using Primer 3 software (<http://primer3.ut.ee/>). Primer sequences are provided in the Supplementary Materials. For PCR amplification, 2 μl of diluted cDNA was supplemented with 25 μl of BioMixTM Red (Bioline; BIO-25006) and 1 μl each of the forward and reverse primer, and the final volume was adjusted with ddH~2~O to 50 μl. PCR for samples was initiated with a pre-cycling melting step at 96°C for 5 minutes, followed by 40 cycles of denaturing at 96°C for 30 seconds, an annealing step was carried out between 58-62°C for 30 seconds (specific annealing temperatures are provided in the [table](#T1){ref-type="table"} below), extension at 72°C for 30 seconds and the final extension temperature was 72°C for 5 minutes. All PCR products were evaluated on 1% agarose gels stained with ethidium bromide ###### PCR Primer sequence Gene Forward primer sequence (5′-3′) Reverse primer sequence (5′-3′) PCR annealing Temp. (°C) ------------- --------------------------------- --------------------------------- -------------------------- *βACT* TGCTATCCCTGTACGCCTCT CGTCATACTCCTGCTTGCTG 58.0 *GAGE1* TAGACCAAGGCGCTATGTAC CATCAGGACCATCTTCACAC 58.4 *SSX2* CAGAGATCCAAAAGGCC CTCGTGAATCTTCTCAGAGG 58.4 *ARRDC5* CAACAAGGCAGACTACGTGC GCGAGTGTGCATGATCTCAC 60.5 *C4orf17* CCTCATCCCAGAAGAGTCTG CTGCTGCTGGTTCCATTGAG 60.5 *C20orf201* ATCTGCTCTTCGGCGACCTG ACACTCTCAGTCGCCGTCAC 60.0 *DDX4* GTGCTACTCCTGGAAGACTG CCAACCATGCAGGAACATCC 60.5 *NT5C1B* CGGCAGGAAAATCTACGAGC CTGTAACCAGGTAGGTCCTG 60.5 *STRA8* TGGCAGGTTCTGAATAAGGC GAAGCTTGCCACATCAAAGG 58.4 *TDRD12* GAGCTAAAGTGCTGGTGCAG CTGAGGTCACCGACAATACC 60.5 AA was funded by the Government of the Kingdom of Saudi Arabia. JF was supported by the National Institute of Social Care and Health Research (grant HS/09/008). RJM, EGV and JAW were funded by Cancer Research Wales. RJM was funded by North West Cancer Research (project grants CR888 and CR950).

Introduction ============ Gingival reactive lesions like pyogenic granuloma have frequent occurrence around natural dentition, however, their association with dental implants is not common. The causes of pyogenic granuloma (PG) in relation to dental implants are not clear mainly due to few published cases ([@B1]-[@B8]). Tooth-related PG is a result of tissue response to minor injury or chronic low-grade irritation ([@B9]-[@B16]). Clinically, oral PG is characterized as a soft mass of smooth or lobulated appearance that could be sessile or pedunculated and frequently presents ulceration. The lesion grows rapidly for a few weeks and the colour ranges from pink to red purple and haemorrhage may occur either spontaneously or after minor trauma ([@B8]). Its incidence is relatively common and accounts for 3.81-7% of all biopsies harvested from the oral cavity ([@B13]-[@B16]). Microscopically, the lesion is characterized by prominent capillary growth in hyperplastic granulation tissue, which suggests a strong activity of angiogenesis. The blood vessels often show a clustered or medullary pattern separated by less vascular fibrotic septa, leading some authorities to consider PG as a polypoid form of capillary hemangioma ([@B17]). The lesions of PG may be found in the oral cavity or extraorally. The most frequent intraoral localization is the gingiva (about 60-70%), but lesions can occur on the lips (14%), tongue (9%), buccal mucosa (7%) and palate (2%) ([@B18]-[@B24]). Possible treatment methods are excision, curettage, cryotherapy, sclerotherapy, chemical and electrical cauterization, cryotherapy and the use of lasers with the carbon dioxide (CO2) or argon ([@B25]-[@B29]). Conservative local excision is the preferred form of treatment and recurrence rates after excision range from 0% to 16% ([@B29]). However, to the best of the authors' knowledge, only 5 cases of pyogenic granuloma in association with a dental implant have been reported in the international literature ([@B1]-[@B3],[@B7],[@B8]). Within the context of the scarce information available on these lesions, the aim of the present study was to report 10 novel clinical cases of pyogenic granuloma in association with titanium dental implants and to elucidate potential risk factors. Finally, the presence of marginal bone loss was evaluated. Material and Methods ==================== Patients charts at the service of oral medicine of Anitua's Dental Clinic (Alava, Spain) were revised from 1991 to 2011. Patients selection was based on the following inclusion criteria: • Treatment of pyogenic granuloma. • The presence of histopathological diagnosis. • Lesion in relation to dental implants. All patients who did not fulfill all inclusion criteria were excluded from the study. Data were collected to report on patient age, gender, patient´s disease, lesion site, type of dental implant (surface and morphology), predisposing factors (trauma, prosthesis type, poor oral hygiene), clinical and radiographic features, diagnosis, treatment and recurrence. Orthopantomography (OPG) of all lesions were examined to compare the presence or absence bone resorption around dental implants. A descriptive statistical analysis of all variables were performed. Then the relationship between PG and marginal bone loss was analyzed by nonparametric Spearman correlation. The effect of surface type on marginal bone loss was also analyzed with one-way ANOVA and Levene post hoc test. The statistical significance was set at *p*-value \< 0.05. All the statistical analyses were performed using the SPSS v15.0 for Windows statistical software package (SPSS Inc., Chicago, IL, USA). Results ======= Ten patients with pyogenic granuloma in relation to dental implants had been identified. They were 2 males and 8 females. Patients' age ranged from 21 to 92 years and all were non-smokers. Five of the ten patients (50%) had systemic disorders: cardiac arrhythmia (1 patient), hypertension (2 patients), atrial fibrillation (2 patients), Type II diabetes mellitus (2 patients), hepatitis C (1 patient ), hypothyroidism (1 patient). Within the group of patients with systemic disease, 3 of them were using 1 to 2 drugs daily, whereas the remaining patient took more than 2 drugs. With regard to oral hygiene habits, 20% of patients reported to brush once a day, 50% did twice daily and 30% brushed three times a day. A 90% of the patients received professional prophylaxis twice a year and the other 10% once a year. In the use of hygiene products the obtained results were as follows: a) use of mouthwash: only was used by 3 patients (37.5%), b) use of dental floss: only one patient (12.5%), and c) interproximal brushes: 3 patients (37.5%). The distribution of PG lesions was even between maxilla and mandible (50% for each region), and the most common oral site affected by PG was the area of tooth 41 (2 cases). The development of PG was related to only accumulation of dental plaque (one patient), bad prosthetic design (one patient), and both factors (one patient). In 4 patients, there had been a combination of tissue pressure by the prosthesis and poor oral hygiene. However, no etiological factor could be related to the development of PG in 3 patients. The clinical size of the lesions ranged from 1.1 x 0.6 mm to 36 x 19 mm. The mean diameter was 7.2 mm. All the lesions were excised and sent for histological examination. The defects were covered with a autologous fibrin membrane (Anitua's protocol). During the first week after the operation, all patients were given analgesic and 0.2% chlorhexidine gluconate mouthwash. During the follow-up period (range two months to 10 years), there were no recurrences. The histopathological reports indicated the diagnosis of PG and the description of highly vascular proliferation that resembles granulation tissue (Fig. [1](#F1){ref-type="fig"}). Figure 1Histological images of the pyogenic granuloma showing an appearance similar to granulation tissue. The histological type of the pyogenic granuloma is non-lobular capillary hemangioma. Arrow heads label blood vessels surrounded by connective tissue. The surfaces of the implants associated with the lesion were smooth (2 implants), machined (3 implants) and rough (5 implants). In no case there was a natural tooth adjacent to the implants related to the lesion. The characteristics of diameters and lengths of the implants studied can be seen in figure [2](#F2){ref-type="fig"}. The average load time of the implants studied was 115 months (SD = 67.5), ranging from a range of 9 to 184 months. Oral rehabilitation was performed with complete prosthesis in 9 patients. The mean mesial bone loss was 2.14 mm (range 0 to 6.50 mm, SD = 2.07) and the mean of distal bone was 1.66 mm (range 0 to 3.75 mm, SD = 1.21. Figure 2Diameter and length of dental implants related to the pyogenic granuloma. There were no statistically significant association between the PG area and the marginal bone loss. However the smooth implant surface showed a significant influence on bone loss (Anova: *p* = 0.001) (Fig. [3](#F3){ref-type="fig"}). Figure 3Peri-implant bone loss grouped by type of surface. The bone loss was the highest for implants with smooth surface. Discussion ========== The clinical and histopathological findings have confirmed the diagnosis of pyogenic granuloma in 10 patients. The present study is the one with the highest number of implant-related PG lesion that are available until now in the scientific literature. These PG lesions have been diagnosed as non-lobular capillary hemangioma. There are two histological types of PG. The first type is characterized by proliferating blood vessels that are organized in lobular aggregates. This histological type of PG was called lobular capillary hemangioma (LCH type). The second type (non-LCH type) consist of highly vascular proliferation that resembles granulation tissue ([@B1],[@B4],[@B11]). Literature data indicated that PG is rarely associated with dental implants, as there are only five cases reported ([@B1]-[@B3],[@B7],[@B8]). However, other reactive lesions such as gingival hyperplasia caused by phenytoin, allergy to titanium abutments or peripheral giant cell granulomas have been reported in the international literature. Causes of conventional oral pyogenic granulomas are not clear, although it has been shown that different stimuli irritants that can trigger them, such as repeated trauma, poor oral hygiene and hormonal problems ([@B1]-[@B20]). About 30-50% of patients with PG have a history of local trauma ([@B9]). Considering PG, in the case reported by Dojcinovic *et al.* ([@B1]), the inappropriate healing cap has resulted in dental plaque accumulation and chronic inflammation of the peri-implant tissues, triggering the development of a PG. However this was not the cause for PG in the case reported by Olmedo *et al.* ([@B2]). The authors have pointed out to the presence of "metal-like" particles and have postulated that these particles could be the result electrochemical phenomena, corrosion, friction, or a synergistic combination of these events ([@B4],[@B5]). Once released, these particles may trigger an inflammatory response mediated by cytokines and macrophages ([@B5]). This inflammatory reaction could perpetuate the pseudo-periodontal pocket that generates the lesion around the implant ([@B5]). In the case reported by Etöz *et al.* ([@B3]), the presence of a gap between the alveolar bone and implant surface could be associated with the occurrence of pyogenic granuloma. Although bone splitting technique was adequately performed, trauma from the upper dentition and lack of adequate keratinized mucosa could result in soft tissue invasion and may have been responsible for PG development ([@B3]). Kang *et al.* ([@B8]) have stated that the causes of the occurrence of PG was unclear however, the antithrombotic therapy may have some involvement in the development of the lesion. In this study, PG was related to only accumulation of dental plaque (one patient), bad prosthetic design (one patient), and to both factors (one patient). The mean age of patients was 74.5 years and most of them have decreased manual dexterity. The bad prosthetic design with flanges could difficult the maintenance of good oral hygiene and could predispose the development of PG around dental implants. In 4 cases, there has been a combination of tissue pressure by the prosthesis and poor oral hygiene. However, no etiological factor could be related to the development of PG in 3 patients where implants have a smooth surface. Previously published studies have reported the association of PG to implant with roughened surface ([@B1]-[@B3],[@B7],[@B8]). There are no published data on smooth or machined surfaces. In the histological analysis, the presence of metal-like particles were searched for. Such particles could not be found in any of the 10 biopsies and thus could not be related to the development or progression of PG. The marginal bone loss has shown no association with the presence of PG. However, Implant's surface has affected significantly the marginal bone loss around dental implants. This results can indicate that PG showed no predilection to specific surface type. With the data obtained from this study and others in the literature we can conclude that pyogenic granuloma in association with dental implants seems to respond to the same stimuli that triggers tooth-related PG. This lesion should be included in the differential diagnosis of soft mass growth around dental implants. PG had no significant correlation with the marginal bone loss around dental implants.

Introduction {#sec1} ============ In total hip arthroplasty (THA), optimum component position is critical for long-term success of the operation by decreasing rates of wear, aseptic loosening, and dislocation \[[@bib1], [@bib2], [@bib3], [@bib4]\]. Recognizing the importance of acetabular component position, Lewinnek et al published a "safe zone" of 5° to 25° for anteversion and 30° to 50° for acetabular abduction based on their experience with dislocations after posterior THA \[[@bib5]\]. This still serves as the standard for ideal acetabular component position but has been called into question given the importance of the spinopelvic relationship \[[@bib6],[@bib7]\]. In addition to acetabular component position, emphasis has also been placed on femoral component positioning. Historically, surgeons have identified the importance of keeping the femoral component out of varus because of increased rates of failure with varus cemented femoral stems \[[@bib8],[@bib9]\]. With the use of cementless femoral fixation, varus positioning of the femoral stem has not been shown to lead to the same increased failures \[[@bib10]\]. However, as compared with cemented femoral stems, many cementless femoral stems provide less ability to adjust the version of the component as a stable press-fit requires the stem to adapt to the proximal geometry of the native femur. Consequently, recent attention has been given to the combined version (CV) of the acetabular and femoral components, with the goal of improving impingement-free range of motion and decreasing instability \[[@bib11], [@bib12], [@bib13]\]. The concept of CV was originally introduced by Ranawat, and he described the use of the "Ranawat sign" to determine CV intraoperatively when using the posterior approach \[[@bib14]\]. While no optimum femoral version has been described, Dorr proposed a CV safe zone of 25°-50° based on previous anatomical studies and his experience with decreased instability in this range \[[@bib15]\]. More recent studies have attempted to quantify a combined anteversion that minimizes impingement \[[@bib16],[@bib17]\]. Native femoral anteversion can vary a great deal, and intraoperative judgment of femoral component version can be difficult. Using preoperative computerized tomography (CT) scans of a group of 46 patients scheduled for primary THA, Bargar et al \[[@bib18]\] found a large range of native femoral version from 6° of retroversion to 33° of anteversion. Dorr et al compared the surgeon's estimate of femoral component anteversion in the posterior approach with the postoperative CT measurement of version and found a poor precision of the surgeon\'s estimate with a correlation coefficient of only 0.688 \[[@bib19]\]. In addition, this study found that only 45% of the femoral stems landed within the desired range of 10°-20° of anteversion. In direct anterior total hip arthroplasty (DA-THA), femoral component broaching and insertion occurs while the patient is positioned supine with the leg fully extended, and the leg below the knee is often draped from the surgeon\'s view. Despite published results of comparable patient outcomes from the DA-THA with other THA approaches, some have questioned the ability to appropriately orient the femoral component with respect to femoral anteversion, via this approach \[[@bib20], [@bib21], [@bib22], [@bib23]\]. Previous studies have reported on improved acetabular component positioning in DA-THA \[[@bib24],[@bib25]\]. However, no prior study, to our knowledge, has examined the combined anteversion of the femoral and acetabular components in DA-THA. This study aims to analyze the combined femoral and acetabular anteversion with cross-sectional imaging and quantify this relative to the CV "safe zone" described by Dorr. Materials and methods {#sec2} ===================== After obtaining institutional review board approval, patients were approached for enrollment in the study. An patient who was undergoing a primary DA-THA from the senior author (JBM) was a candidate for enrollment. Patients with femoral or acetabular hardware were excluded from this study. Thirty consecutive patients were enrolled in the study. Four blinded observers independently recorded the measurements (2 fellowship-trained arthroplasty surgeons, one hip and knee fellow, and one orthopaedic resident). All implants were positioned using intraoperative fluoroscopy based on preoperative templating. A CORAIL femoral stem (DePuy, Warsaw, IN) and a PINNACLE acetabular cup (DePuy, Warsaw, IN) were used for all the cases. The senior author standardized intraoperative images by matching the anteroposterior (AP) pelvis fluoroscopic view with the preoperative AP pelvis standing radiograph. One month after surgery, all patients had a standing AP pelvis and a cross-table lateral radiograph taken, which were used for acetabular component position measurement. Abduction and anteversion measurements of the acetabulum were made from the digital radiograph using the TraumaCad (Voyant Health, Columbia, MD) hip abduction measurement tool. Femoral component position measurements were taken from limited supine CT scan of the hip and knee with 2.5-mm cuts (General Electric BrightSpeed, Fairfield, CT). CT was not selected for acetabular component position to minimize patient radiation exposure. Angular measurements were calculated using the axis of the top of neck of the femoral stem relative to both the posterior condylar axis (PCA) and the transepicondylar axis (TEA). The CV was then calculated for the TEA and the PCA by adding the femoral anteversion calculated from the CT scan with the anteversion measured from the standing AP pelvis radiograph. Statistical analysis {#sec2.1} -------------------- Measurements from the 4 observers were combined, and the mean and standard deviation were calculated. The Pearson correlation coefficient was also measured for each observer, with the kappa values reported, and compared with the group for all measurements. The mean for each measurement was used to determine the number of components placed in the "safe zone." Statistical analysis was performed with the use of SAS software (SAS Institute, Raleigh, NC). Results {#sec3} ======= Of the 30 enrolled patients, 29 had an appropriate CT scan obtained. One patient had a CT scan performed without adherence to the protocol precluding reference of femoral version to the axes of the knee and was excluded from the results. The mean acetabular abduction and anteversion were 39.3° (standard deviation \[SD\] = 4.2°) and 27.2° (SD = 4.7°), respectively. The mean stem anteversion was 17.5° (SD = 10.8°) from the TEA and 21.7° (SD = 11.3°) from the PCA. Ten of the 30 cups were placed inside of the "safe zone" of Lewinnek for acetabular anteversion, but all cups were within the "safe zone" for abduction ([Fig. 1](#fig1){ref-type="fig"}).Figure 1The Lewinnek "safe zone." Ten of the 30 cups were placed inside of the "safe zone" of Lewinnek for acetabular anteversion, but all cups were within the "safe zone" for abduction. Combined femoral and acetabular component anteversion from the TEA resulted in 79% (23 of 29) of patients within the "safe zone" of 25°-50° with accurately oriented components ([Fig. 2](#fig2){ref-type="fig"}).Figure 2Combined anteversion. Combined femoral and acetabular component anteversion. Pearson correlation coefficients were high for both stem anteversion from the TEA (R = 0.96) and the PCA (R = 0.98); however, the kappa coefficient for interobserver reliability for combined component anteversion was greater for the TEA (kappa = 0.83 vs 0.65). Discussion {#sec4} ========== Component positioning has been recognized as an important factor in the long-term survival of THA \[[@bib5],[@bib12],[@bib26]\]. Muller et al. \[[@bib27]\] suggested a cup anteversion of 10°-15° and femoral anteversion of 10° to be ideal. Lewinnek et al. \[[@bib5]\] followed with their study that found a lower dislocation rate when the acetabular components were positioned in a safe zone of 30°-50° of inclination and 5°-25° of anteversion. A study by Biedermann et al. \[[@bib28]\] found the lowest dislocation rates with acetabular components positioned at 45° of inclination and 15° of anteversion. More recently, Dorr proposed a CV safe zone of 25°-50° based on previous anatomical studies and his experience with decreased instability in this range \[[@bib15]\]. Many authors have begun to appreciate the importance of the combined femoral and acetabular anteversion on dislocation rates and impingement \[[@bib3],[@bib11],[@bib12],[@bib29]\]. Ranawat and Maynard \[[@bib6]\] suggested the importance of the combination of femoral and acetabular anteversion and recommended 45° for women and between 20° and 30° for men. Jolles et al. \[[@bib12]\] found that when the combined anteversion was outside of a range of 40°-60°, the patient's dislocation was 6.9 times higher. Hisatome and Doi \[[@bib29]\] examined combined anteversion in a mathematical model to find the optimum positions to avoid neck impingement with different sized components. They recommended an ideal position, while not accounting for patient's pelvic inclination, of 45° of cup abduction, 25° of cup anteversion, and 25° of stem antitorsion. Other studies have examined the combined component anteversion after lateral or posterior approach THA. Reikerås and Gunderson \[[@bib30]\] utilized postoperative CT scans in 91 patients after either posterior or lateral approach THA and found that only 60.4% of their patients had a combined anteversion within the acceptable safe zone. Wassilew et al. \[[@bib31]\] evaluated THAs performed using an anterolateral approach with navigation, and they found that 88% of their patients were within the CV safe zone of 25°-50°. A study by Nogler et al. \[[@bib24]\] examined the ability of the DA approach to position the femoral and acetabular components with and without navigation. However, the present study is the first to analyze CV in the DA-THA. We used the TraumaCad (Voyant Health, Columbia, MD) software tool, which was found to have good intraobserver and interobserver reliability but can underestimate acetabular anteversion by as much as 12° \[[@bib32]\]. Despite the potential for underestimation of anteversion, we believe that a standing radiograph more accurately represents the patient\'s functional anteversion and accounts for the patient\'s lumbar or pelvic tilt because of the difficulties in the estimation of tilts with supine radiographs \[[@bib33], [@bib34], [@bib35]\]. This notion is supported by the work of Hayakawa et al that found a statistically significant difference between intraoperative and postoperative radiographs in 100 consecutive patients for both anteversion and vertical tilt \[[@bib35]\]. Impingement can lead to abnormal wear patterns or dislocation. Our study found that 17% of patients had CV greater than 50°, outside of the safe zone. However, when the femoral and acetabular components were analyzed independently, 20 of 30 patients had "excessive" anteversion of the acetabular component. Intraoperative stability assessments did not identify any impingement. There was less variability in acetabular component version than in femoral version. With uncemented femoral components, femoral anteversion is largely dictated by proximal femoral geometry. Therefore, some surgeons have recommended a femur-first technique to better address this variability and "fine-tune" version on the acetabular implant \[[@bib36]\]. Our results are comparable with those of similar series that included the posterior or lateral approach THA with and without navigation, with 88% and 60.4% of the components within the safe zone for CV, respectively \[[@bib30],[@bib31]\]. We are unable to conclude that our results are superior based on the small sample size in our study, but this should be investigated further with a larger series. We used both the PCA and the TEA for measurements to determine femoral component anteversion. Interestingly, we found a slightly higher interobserver reliability with the TEA measurements than with the PCA. This may reflect the difficultly of locating the point of maximal posterior bone in the condyle with a fine-cut CT scan. The differential radii of the femoral condyles and the extremity orientation relative to the CT scanner may influence the appearance of the most posterior projection of the condyles. This could cause the most posterior condylar projection to be on separate cuts of the CT scan. Authors may consider using the TEA primarily for femoral version measurements with axial imaging or 3-dimensional imaging techniques. There are several potential limitations for this study. First, the number of patients in this study was 30. A small sample size may not reflect the variability of patient anatomies. The single-surgeon cohort may limit the generalizability of these results. The use of plain radiographs instead of CT to assess acetabular component position may be a limitation. The CT may be more accurate in determining acetabular anteversion and abduction. However, most surgeons who perform the DA-THA utilize the standing AP pelvis radiograph to position the acetabular component. It has been previously shown that the supine position of the acetabular component varies from supine to standing radiographs, and therefore, we chose to utilize a standing pelvis AP radiograph to determine the implant position. Many accepted modern studies have relied on AP radiographs to determine both anteversion and abduction angles for acetabular components \[[@bib37], [@bib38], [@bib39]\]. Finally, this study did not assess the spinopelvic relationship. Abdel et al. recently demonstrated that most dislocations occurred within the "safe zone" \[[@bib39]\] and it is possible that other factors are as important for stability as implant position. As this study represents the first to examine CV in DA-THA with postoperative axial imaging, we found that the DA approach allowed for placement of components with CV within the acceptable range for most patients. In this study, excessive anteversion of the acetabular component based on Lewinnek's "safe zone" was the most common component orientation error. However, this may simply represent a single surgeon's component orientation preference and may not necessarily be generalizable. Even with limited visualization of the proximal femur, the DA approach can reproducibly yield a high percentage of THA components in the "ideal" position for CV. Conflict of interest {#sec5} ==================== J.B. Mason receives royalties from DePuy and A Johnson & Johnson Company, is a paid consultant for DePuy and A Johnson & Johnson Company, receives other financial or material support from DePuy and A Johnson & Johnson Company, receives royalties, financial or material support from the Journal of Arthroplasty, and is a board member for the Publication Committee AAHKS; J.R. Martin is a paid consultant for DePuy and A Johnson & Johnson Company; J.L. Masonis receives royalties from Medacta, Smith & Nephew, and Zimmer, is a paid consultant for Smith & Nephew and Zimmer, holds stock or stock options in Orthogrid, receives research support a principal investigator from DePuy, A Johnson & Johnson Company, Smith & Nephew, and Zimmer, receives royalties, financial or material support from Medacta, and is a board member for the Anterior Hip Foundation; J.B. Jackson is a board member for AAOS and AOFAS committee. Appendix A. Supplementary data {#appsec1} ============================== Conflict of Interest Statement for ChristalConflict of Interest Statement for JacksonConflict of Interest Statement for MartinConflict of Interest Statement for MasonConflict of Interest Statement for MasonisConflict of Interest Statement for Springer

1. Historical Origins {#sec1-viruses-12-00132} ===================== At a meeting of the Fellows of the National Institute of Agricultural Botany in Cambridge, UK, on 14 November 1924, Dr. Redcliffe Salaman gave a lecture entitled "Degeneration of the Potato---An Urgent Problem" \[[@B1-viruses-12-00132]\]. He reported that "potato degeneration", namely the decrease in yield when potatoes were grown year after year from tubers, rather than from true seed, cost the UK between five and ten million pounds sterling each year. He noted that the condition was first reported in 1778 at a meeting in Manchester, and called "potato curl". It was worse in lowland crops and in the Southern UK than in crops grown on higher ground and in the north, and although some thought it was caused by disease, perhaps insect-borne, others believed it was a form of senility resulting from repeated vegetative reproduction! Salaman concluded that that degeneration was caused by a complex of tuber-borne pathogens. Salaman's talk was successful, as it induced the Ministry of Agriculture to found the Potato Virus Research Station in Cambridge and appoint him as director, and in turn he appointed Kenneth Smith as entomologist, who soon separated some of the components of potato curl and identified the viruses he called potato virus X and potato virus Y (PVY) \[[@B2-viruses-12-00132]\]. Other viruses similar to PVY were soon reported, for example, henbane mosaic virus \[[@B3-viruses-12-00132]\], which was like PVY in causing mosaic symptoms, being transmitted by sap, although relatively unstable in it, and also by being transmitted by aphids in short feeds. These viruses, which became known as potyviruses, short for "potato virus Y group viruses" \[[@B4-viruses-12-00132]\], were among those included in early attempts to devise biological taxonomies of plant viruses \[[@B5-viruses-12-00132]\] based on the length of their filamentous particles \[[@B6-viruses-12-00132]\]. They were also distinguished from other plant viruses by having serologically distinct virions and, biologically, by having distinct host ranges and causing distinct symptoms, and by their properties in infective sap, such as dilution end point, thermal inactivation point, and longevity in vitro. Sixteen different potyviruses had been described in 1959. Subsequently, in this pre-sequencing era, a combination of techniques, including sucrose density gradient centrifugation, analytical ultracentrifugation, ultraviolet spectrophotometry, and polyacrylamide gel electrophoresis were also included to establish the sedimentation coefficients and buoyant densities of virions, and the molecular weights of protein subunits and % nucleic acid contents, as all these properties provided additional distinguishing characteristics when novel viruses were being described \[[@B7-viruses-12-00132]\]. Virus identification and taxonomy were transformed later, when methods for sequencing genes were invented in the 1970s and applied to plant viruses \[[@B8-viruses-12-00132],[@B9-viruses-12-00132]\], and it was established that hierarchical groupings based on viral protein and gene sequences, including those of potyviruses, confirmed and extended those that had been devised previously by using phenotypic characters, serological tests, etc. As a result, 57 potyviruses had been identified by 1991 \[[@B10-viruses-12-00132]\], using sequences of the "part NIb-CP" region of their genomes, as this was bracketed by convenient primer sites \[[@B11-viruses-12-00132],[@B12-viruses-12-00132]\]. By 2000, over 1000 potyvirus sequences were recorded in the GenBank database, and there are now more than 26,000. The potyviruses now form a family, the *Potyviridae* \[[@B13-viruses-12-00132]\], containing at least eight genera of which the aphid-transmitted potyviruses, including the first described, PVY, make up the largest genus, *Potyvirus*. This large plant virus genus is one of the most important economically because of the yield and quality losses it causes in a wide range of crops worldwide. Moreover, some of its members currently endanger food security in developing countries by causing devastating diseases in tropical and subtropical food crops \[[@B14-viruses-12-00132]\]. 2. The Origins of the Potyviridae {#sec2-viruses-12-00132} ================================= The potyvirids are distinguished from other viruses by specific molecular differences, together with a combination of phenotypic properties \[[@B13-viruses-12-00132],[@B15-viruses-12-00132]\]. All potyviruses infect plants; most are transmitted in nature by arthropods---mostly by aphids---though bymoviruses are transmitted by root-infecting plasmodiophorids, which are cercozoan amoebae. Some potyviruses are seed-borne \[[@B16-viruses-12-00132]\]. Potyvirid virions are flexuous filaments, 680--900 nm long and 11--20 nm in diameter. Each is helically constructed from 1400 to 2140 subunits of a coat protein (CP), and a positive-sense, single-stranded RNA genome (usually monopartite, but bipartite in the genus *Bymovirus*) of 8--11 kb in total, which is wound into a groove within the CP subunits. The ancestry and origins of the potyvirids is being revealed by studies of the structure and sequences of their genes and proteins. These show that their genomes are polyphyletic in origin, as there are significant similarities between three of their genes and those of viruses in three otherwise-unrelated virus genera; two detected by the protein sequence similarity of the helicase region of the CI protein and the RdRp region of the NIb protein, respectively, and the third by the structure of the CP \[[@B17-viruses-12-00132],[@B18-viruses-12-00132]\]. A BLASTp search of the GenBank protein sequence database (Sept 2019), using the eight main motif regions of the polyprotein of PVY (NC_001616), and excluding matches with sequences from the *Potyviridae*, found only significant matches between the "DEAD helicase-helicase C" region of the CI protein and that of classical swine fever and hog cholera pestiviruses (chance probability, 1e-12\_-16; 30% identity, 13% indels), and no others. Likewise, there were significant similarities between the PVY RdRp region and that of astroviruses (see below). Structural studies have only been reported for the CP protein of one potyvirus, watermelon mosaic virus (WMV), and reveal a structure that is closely similar to those of other viruses with flexuous filamentous virions, including two potexviruses, and also the enveloped flexuous nucleoproteins of orthomyxoviruses and bunyavirids, which include tomato spotted wilt tospovirus \[[@B19-viruses-12-00132],[@B20-viruses-12-00132],[@B21-viruses-12-00132]\]; all these CPs have a core domain rich in alpha helices. Each CP subunit interacts with five nucleotides (nts), and has 8.8 subunits per turn in a left-handed helix, with a pitch of 34.5--35 Å. Its N-terminus is external to the virion, and its C-terminus internal. The terminal regions interact with adjacent subunits and provide flexibility to the virion. In serological studies, the N-terminus is dominant and, in potyviruses, is also involved with aphid interactions \[[@B22-viruses-12-00132],[@B23-viruses-12-00132]\]. The sequences of their RNA-dependent RNA polymerases (RdRps) place the potyvirids in the "Picornavirus Supergroup" ([Figure 1](#viruses-12-00132-f001){ref-type="fig"}) \[[@B19-viruses-12-00132],[@B24-viruses-12-00132]\], where the potyvirids are outsiders, as most of the others have icosahedral virions made of eight-stranded antiparallel beta-barrel proteins, the so-called 'jelly roll' proteins. In the Wolf et al. \[[@B24-viruses-12-00132]\] taxonomy of RdRps ([Figure 2](#viruses-12-00132-f002){ref-type="fig"}), the potyvirid RdRps form a cluster that is sister to an RdRp found in a metagenome, bufivirus UC1-gp2, isolated from "wastewater" collected in San Francisco. The sister clade to the potyvirid/bufivirus clade of RdRps are mostly those of the astroviruses, a group of gut-infecting viruses that are found in a wide range of animals, mostly mammals or birds. They have 28--35 nm diameter isometric virions (<https://en.wikipedia.org/wiki/Astrovirus> (accessed July 2019)). Sister to the RdRp clade of potyviruses/bufivirus/astroviruses are the RdRps of hypoviruses, amalgaviruses, partitiviruses, and picobirnaviruses, many of them metagenomes, including one from *Phytophthora infestans* \[[@B25-viruses-12-00132]\] and two from leeches \[[@B26-viruses-12-00132]\]. None of the motifs identified in potyvirus proteins, other than the RdRp, match those encoded by astroviruses; the nonstructural protein of a human astrovirus (NP_059443) was found to have the RdRp motif, but, in addition, only a trypsin-like peptidase, a restriction enzyme, and a motif of unknown function. The bufivirus metagenome includes a 3′ terminal S domain (jelly-roll) capsid protein gene indicating that it, like most of the picornavirus supergroup, including astroviruses, probably has isometric virions. Most of the genome of all potyvirids encodes a single polyprotein, which is post-translationally hydrolyzed into ten proteins \[[@B15-viruses-12-00132],[@B27-viruses-12-00132]\]. It also encodes another protein (P3N-PIPO) in the −1 reading frame, and a second (P1N-PISPO) in a few potyvirids \[[@B28-viruses-12-00132]\]. From the N-terminus to the C-terminus, the ten potyvirus proteins are named as follows: P1-Pro, HC-Pro, P3, 6K1, CI, 6K2, NIa-Pro, Nib, and CP. The P1-Pro protein is a serine protease (S30) that self-hydrolyses its own C-terminal cleavage site. Next is the HC-Pro protein, which is a cysteine protease (C6) that also hydrolyses its own C-terminal cleavage site. The other eight proteins have seven cleavage sites hydrolyzed by NIa-Pro, the cysteine protease (C4), encoded by the eighth region. The eight largest motifs were found in all potyvirus sequences by using the motif-matching facility Pfam, in all three rymoviruses (agropyrum mosaic, ryegrass mosaic, and hordeum mosaic viruses), and in the two most closely related potyvirids, namely reed chlorotic stripe virus and blackberry virus Y. The genomes of bymoviruses and macluraviruses have smaller sets of the enzyme motifs as some of their N-terminal motifs are missing; Pfam (<https://pfam.xfam.org/> (accessed July 2019)) found only the C-terminal four motifs of the polyprotein in the complete polyproteins of barley yellow mosaic bymovirus, cardamom mosaic macluravirus, sweet potato mild mottle ipomovirus, and wheat streak mosaic poaceaevirus. A Pfam analysis of celery latent celavirus (CLV) found only the helicase C and RdRp genes found in other potyvirids. Thus, the RdRp or the helicase proteins are probably most appropriate for inferring the phylogeny of the potyvirids. The phylogeny of RdRp genes of all named potyvirids shows that they form at least eleven genera, of which by far the largest is the potyviruses ([Figure 2](#viruses-12-00132-f002){ref-type="fig"}). The RdRp of CLV \[[@B32-viruses-12-00132],[@B33-viruses-12-00132],[@B34-viruses-12-00132]\] is the sister of the RdRps of all other potyvirids, but those of the bymoviruses and macluraviruses are closest to that of bufivirus UC1 and the astroviruses as they are on the shortest branches. The virions and genome of CLV have all the features of a potyvirid, not a bufivirus. Its virions are flexuous filaments around 900 nm in length and contain a single genome of 11,519 nts \[[@B34-viruses-12-00132]\]. This confirms that their structure is likely to be closely similar to those of WMV as, using the known parameters of WMV virions, the 11,519 nts of the CLV genome will assemble with 2.304 CP subunits and form a helix 903--916 nm long \[[@B20-viruses-12-00132]\]. The similarities and differences between CLV and other potyvirids may indicate the properties of their shared ancestor. CLV was first reported from Europe. It is sap-transmitted to several dicotyledonous plants (dicots) from several different families, but there is no record of tests of monocotyledonous plants (monocots) as hosts, and like many potyvirids, it is readily seed-borne in two plant species. CLV was not transmitted by five species of aphids. Its genome has one major open reading frame (ORF), and a minor overlapping ORF, P3N-PIPO, in the -1 frame. It has some, but not all, of the motifs found in potyvirids, but not those associated with aphid-transmission in potyvirus genomes. CLV's one unique feature is a signal peptide at the 5′ terminus of its genome. In summary, although the potyviruses and astroviruses share an RdRp ancestor \[[@B24-viruses-12-00132]\], none of their other genes are related, and no more potyvirus-like or astrovirus-like intermediate ancestors are known at present. Similar conclusions can be drawn about their shared helicase and CP genes. Nothing is known at present of the origins of the other potyvirus proteins, although their diversity and likely relationships suggest that they, and especially the P1 protein, have helped generate the extraordinary diversity of the potyvirids \[[@B35-viruses-12-00132]\]. Those unique to potyviruses may have arisen de novo, or by overprinting \[[@B36-viruses-12-00132],[@B37-viruses-12-00132],[@B38-viruses-12-00132]\], or may have come from other organisms of which the genes have not been sequenced yet. The only safe conclusion is that the potyvirids are polyphyletic in origin. 3. The Potyvirus: Rymovirus Divergence {#sec3-viruses-12-00132} ====================================== The potyviruses and rymoviruses are sister taxa ([Figure 2](#viruses-12-00132-f002){ref-type="fig"}); they diverged from a common potyvirid ancestor that, judging from their phylogenetic distances, probably had blackberry virus Y and reed chlorotic stripe virus as successive sister viruses, although the genomic sequences of both of these viruses differ significantly from those of rymoviruses and potyviruses. Although most of the differences in the genomic sequences of potyviruses and rymoviruses, or the proteins they encode, are small, they are responsible for their phenotypic differences, including their transmission by different vector types, aphids and mites. Govier and Kassanis \[[@B39-viruses-12-00132],[@B40-viruses-12-00132]\] first reported that, for transmission by aphids, potyviruses required virions and a "helper component" present in infected plants. This was shown to be the protein now called HC-Pro, which was subsequently found to have several additional functions (reviewed by \[[@B15-viruses-12-00132]\]), including the ability to suppress RNA silencing. Potyvirus proteins have at least three motifs associated with aphid transmission \[[@B15-viruses-12-00132],[@B41-viruses-12-00132]\] including the DAG- motif at the N-terminus of the CP, and the -KITC- and -PTK- motifs of the HC-Pro protein; although the first two of these are not found in the homologous sites of rymovirus polyproteins, the last is. When the HC-Pro gene of mite-transmitted wheat streak mosaic tritimovirus (WSMV) was replaced by that of aphid-transmitted turnip mosaic potyvirus (TuMV), it was no longer transmitted by its mite vector \[[@B42-viruses-12-00132]\], but there is no report of similar experiments with rymoviruses or potyviruses. Gibbs and Ohshima \[[@B43-viruses-12-00132]\] suggested that the divergence giving rise to the proto-potyvirus and proto-rymovirus is likely to have occurred in an infected Eurasian monocot. This conclusion was based on the fact that the primary hosts (i.e., the host from which they were first isolated) of all three rymoviruses are all Eurasian monocots; *Agropyron* (Eurasian; <https://en.wikipedia.org/wiki/Agropyron> (accessed 30 October 2019)), *Hordeum* (Eurasian, African, Americas; <https://en.wikipedia.org/wiki/Hordeum> (accessed 1 December 2019)), and *Lolium* (Europe, Asia, N. Africa; <https://en.wikipedia.org/wiki/Lolium> (accessed 18 December 2019)), and many of the basal potyviruses are too ([Figure 3](#viruses-12-00132-f003){ref-type="fig"}a,b). The list of potyvirus hosts closest to the rymoviruses depends on whether ORF or polyprotein sequences are used for estimating their patristic distances from the rymoviruses, but 16 of the nearest 20 are shared. Of these, 11 are from Eurasian plants (seven monocots and four dicots), and single hosts are from the Americas, Australia, Madagascar, and South Africa, and one is cosmopolitan. The possible dates of that divergence and of some of these invasions of other continents are discussed below. 4. Potyvirus Diversity {#sec4-viruses-12-00132} ====================== 4.1. Phylogenetics {#sec4dot1-viruses-12-00132} ------------------ The evolution of potyviruses has been studied by using the two strategies widely used for investigating evolutionary rates and processes. Firstly, phylogenetics \[[@B44-viruses-12-00132]\], which is based on the premise that organisms evolve by mutation and selection, so that the resulting successive divergences can be represented as a tree \[[@B45-viruses-12-00132]\], and revealed computationally and quantified by comparing their properties, especially, nowadays, those of their gene sequences and the proteins they encode. Secondly, methods of population genetics can be used (see below in [Section 4.2](#sec4dot2-viruses-12-00132){ref-type="sec"}). The ML phylogeny ([Figure 3](#viruses-12-00132-f003){ref-type="fig"}a,b) shows that most, but not all, of the lineages proposed in earlier published phylogenies (e.g., Gibbs and Ohshima \[[@B43-viruses-12-00132]\]), are confirmed. The relationships between the different potyviruses in the phylogeny are also closely similar to those in a published neighbor joining (NJ) tree phylogeny \[[@B13-viruses-12-00132]\], despite differing in relative branch lengths. Several of the virus lineages in the ML phylogeny are evident because they are subtended by long branches (Lineages 4, 5, 7, and 9); ML trees often define clusters more clearly than NJ trees, as their basal branches are relatively longer than their tip branches. Their long branches represent periods of the past that have only one known survivor, namely the lineage progenitors. Some lineages are additionally defined by the relationships of their primary hosts, namely the hosts from which they were first isolated, when the hosts are grouped at the "Order" level of the angiosperms (<https://en.wikipedia.org/wiki/APG_IV_system> (accessed 16 September 2019). Most viruses of Lineages 1 and 9 were first isolated from lilioids, most of Lineage 2 from commelinids, Lineage 4 and 7 from rosids, and Lineage 5 and 8 from asterids, whereas the hosts of some large lineages were from two plant clades, such as Lineage 3 which was from monocots (both alismatid and lilioid) or rosids, but rarely asterids, and Lineage 6, which was from a mixture of lilioids or asterids, but not rosids. These specificities are interesting as early attempts "to find some logic in the confusing issues of experimental host ranges of plant viruses" \[[@B46-viruses-12-00132]\] were not resolved by more data of the experimental hosts \[[@B47-viruses-12-00132],[@B48-viruses-12-00132],[@B49-viruses-12-00132]\]. However, the potyvirus lineages of primary hosts shown in [Figure 2](#viruses-12-00132-f002){ref-type="fig"} and [Figure 3](#viruses-12-00132-f003){ref-type="fig"}a,b, and the even more exact correlations shown by the tobamoviruses and their primary hosts \[[@B50-viruses-12-00132]\], indicate that there are phylogenetically influenced components of both virus and host that control their biochemical compatibility. It might be that the greater molecular repertoire of monocots, asterids, and rosids, which have resulted from repeated genome duplication \[[@B51-viruses-12-00132]\], allow viruses to form functioning relationships with them. It is noteworthy that few potyviruses have been reported from caryophyllids and very few from basal angiosperms, such as the ranunculids, although hibbertia virus Y is from a Gunnerid, and catharanthus mosaic virus was isolated from *Welwitschia mirabilis* (a gymnosperm; \[[@B52-viruses-12-00132]\]). Two poty-like metagenomes have recently been reported from unexpected hosts (snails and a "dipteran mix") and are discussed below. Overall, the primary 'host preferences' of potyviruses among green plants is strikingly similar to the host preferences of their principal vectors, which are heteroecious aphids, that alternate between woody and herbaceous hosts, which are mostly rosids, asterids, and commelinids (grasses), but not caryophyllids \[[@B53-viruses-12-00132]\]. Genetic recombination is common in potyvirus populations \[[@B54-viruses-12-00132]\] and in potyvirus experiments \[[@B55-viruses-12-00132]\]. The populations of four potyviruses discussed in [Section 4.2](#sec4dot2-viruses-12-00132){ref-type="sec"} below, had 35% to 64% recombinants, and most reports conclude that recombination is an important factor driving the evolution of potyvirus populations \[[@B56-viruses-12-00132]\]. However, there have been few reports that potyvirus species are recombinants involving other described viruses. Therefore, RDP version 4.95 was used to analyze the 152 ORF sequences used for [Figure 3](#viruses-12-00132-f003){ref-type="fig"}, but only five recombinants were found. Hubei poty-like virus (Lineage 2; NC_032912) was the recombinant with greatest statistical support, but is probably an in silico contaminant, which will be discussed below (see Metagenomes, [Section 6.10](#sec6dot10-viruses-12-00132){ref-type="sec"}). The analysis confirmed that WMV (Lineage 3; NC_006262) is a recombinant with soybean mosaic virus (SbMV; NC_002634; Lineage 3) as a major parent with a minor 5′-terminal region closest to bean common mosaic virus (BCMV; NC_003397; nts 1-c.770), as previously reported by Desbiez and Lecoq \[[@B57-viruses-12-00132]\], and that Sudan watermelon mosaic virus (Lineage 4; NC035459) is also a recombinant with parents from Lineage 4, namely zucchini shoestring virus (NC_043172) and a minor 5′-terminal region close to wild melon vein banding virus (NC_035458; nts 1-c.550), as reported by Desbiez et al. \[[@B58-viruses-12-00132]\]. Two other recombinants are novel but less certain. The ORF of calla lily latent virus (CLLV; EF105298) is mostly related to that of SbMV, but has a 5′-terminal region distantly related (74.9% ID) to that of konjac mosaic virus (KMV; NC_007913; nts 1-c.570), and likewise most of the ORF of vanilla mosaic virus (VMV) is related to that of BCMV but it has a 5′-terminal region (nts 1-c.665) that is also distantly related (61.8% ID) to that of KMV; these links were not resolved by direct nBLAST searches. These results indicate that, at most, only four of the 152 distinct potyviruses, we compared, were generated by recombination. The ORF phylogeny in [Figure 3](#viruses-12-00132-f003){ref-type="fig"}a,b is closely similar to a ML tree of the polyprotein sequences encoded by the ORFs, although a graph comparing their patristic distances ([Figure 4](#viruses-12-00132-f004){ref-type="fig"}) shows that there are differences of up to 15% in individual branch lengths. The patristic distances graph ([Figure 4](#viruses-12-00132-f004){ref-type="fig"}) also shows that the relationship between the ORF and polyprotein patristic distances is broadly linear except at the smallest axial values. This is perhaps evidence of mutational saturation, but not translational saturation; the rate changes may reflect the time-dependent bias in rate estimates of nt sequence change \[[@B59-viruses-12-00132],[@B60-viruses-12-00132]\]. 4.2. Population Genetics {#sec4dot2-viruses-12-00132} ------------------------ As mentioned above in [Section 4.1](#sec4dot1-viruses-12-00132){ref-type="sec"}, the evolution of populations of organisms can be studied not only by phylogenetics, but also by using the methods of population genetics (popgen) \[[@B61-viruses-12-00132],[@B62-viruses-12-00132],[@B63-viruses-12-00132]\]. These describe features of gene populations, using mathematical models, and compare observed features with those likely to result from sequential random changes. Popgen analyses have been used to study potyvirus populations within plant populations (e.g., Achon \[[@B64-viruses-12-00132]\]; Li et al. \[[@B65-viruses-12-00132]\]; Wang et al. \[[@B66-viruses-12-00132]\]; Hajizadeh et al. \[[@B67-viruses-12-00132]\]) and, increasingly, virus populations within individual plants (e.g., Cuevas et al. \[[@B68-viruses-12-00132]\]; Domingo and Perales \[[@B69-viruses-12-00132]\]; Dunham et al. \[[@B70-viruses-12-00132]\]; Kutnjak et al. \[[@B71-viruses-12-00132]\]; Rousseau et al. \[[@B72-viruses-12-00132]\]; Seo et al. \[[@B73-viruses-12-00132]\]). Such studies have shown that the effective populations of potyviruses are all around 10,000 \[[@B74-viruses-12-00132]\], and the "The high potential for genetic variation in plant viruses need not necessarily result in high diversity of virus populations. There is evidence that negative selection results in virus-encoded proteins being not more variable than those of their hosts and vectors. Evidence suggests that small population diversity, and genetic stability, is the rule" \[[@B75-viruses-12-00132]\]. We have investigated whether popgen analyses of the genome sequences now available for five potyviruses with contrasting biologies can reveal features that correlate with their contrasting biological differences. Note that we use the terms phylogroups (PVYs), strains (PPVs), and groups (TuMV) somewhat interchangeably, as these are the terms used in quoted publications. The viruses we examined are as follows: ### 4.2.1. Potato Virus Y {#sec4dot2dot1-viruses-12-00132} PVY, which originated in South America \[[@B29-viruses-12-00132]\] from a lineage of potyviruses ([Figure 3](#viruses-12-00132-f003){ref-type="fig"}b; Lineage 5) found mostly in solanaceous weeds and crops, and asteroid species in the Americas \[[@B76-viruses-12-00132],[@B77-viruses-12-00132]\]. Potatoes (*Solanum tuberosum*) were domesticated in South America and first taken to Europe in the 16th century. They became a major international vegetatively propagated crop after the mid-19th century. PVY, which is mostly spread by locally migrating aphids and planting infected seed tubers, is now common in the potato crops of the world, where recombinant PVYs have become particularly damaging. ### 4.2.2. Turnip Mosaic Virus {#sec4dot2dot2-viruses-12-00132} TuMV, which is the only dicotyledon-infecting member of a lineage of potyviruses of lilioid monocotyledons ([Figure 3](#viruses-12-00132-f003){ref-type="fig"}b). It diverged about a millennium ago from a virus of European orchids, currently referred to as the OM TuMV outgroup \[[@B78-viruses-12-00132]\]. TuMV probably emerged to infect its crop and weed hosts during the development of agriculture in Eurasia in recent centuries. It is common in crops of several annual brassicas (canola, turnips, radish, etc.), that are grown from seed. It is also common in various perennial domesticated and weed *Brassicaceae* \[[@B79-viruses-12-00132]\]. TuMV has been reported as being seed-borne in some hosts \[[@B80-viruses-12-00132],[@B81-viruses-12-00132]\], but this has not been confirmed, and it is probably mostly spread by aphids migrating within and between populations of its perennial and annual hosts. ### 4.2.3. Plum Pox Virus {#sec4dot2dot3-viruses-12-00132} Plum pox virus (PPV), which is a potyvirus of long-lived woody domesticated and wild *Prunus* fruit trees and shrubs. Its host populations are therefore much smaller and longer lived than the host populations of PVY and TuMV. PPV is usually considered to be a single species ([Figure 3](#viruses-12-00132-f003){ref-type="fig"}b), even though it has seven distinct strains. Its center of emergence was probably Eurasia \[[@B31-viruses-12-00132]\]. It is spread by aphids within a wide range of wild and domesticated *Prunus* species and non-*Prunus* weeds. PPV is also graft transmitted, and most of its crop hosts have been propagated by grafting for the last three millennia. ### 4.2.4. Sweet Potato Potyviruses {#sec4dot2dot4-viruses-12-00132} The SwPVs, or sweet potato potyviruses ([Figure 3](#viruses-12-00132-f003){ref-type="fig"}b; Lineage 8), which have been isolated from sweet potato (*Ipomoea batatas*) crops in all continents, except Antarctica. The SwPVs are usually considered to be five species \[[@B28-viruses-12-00132]\], but as all have only been isolated from sweet potato, we included them as it is possible that they are a single diverse mega-species, and it is of interest to check whether or not population genetics methods agree. ### 4.2.5. Hardenbergia Mosaic Virus {#sec4dot2dot5-viruses-12-00132} Hardenbergia mosaic virus (HarMV) is a potyvirus found in the wild only in southwestern Australia where it is widespread in the perennial native legume *Hardenbergia comptoniana* \[[@B82-viruses-12-00132]\]. It is a member of the BCMV lineage ([Figure 3](#viruses-12-00132-f003){ref-type="fig"}a; Lineage 3), which mostly radiated in South-East Asia \[[@B83-viruses-12-00132]\], and may have been carried to Australia by Austronesian boat people. This lineage probably arrived in Australia long before Europeans arrived in 1788, colonized the continent, and subsequently developed large-scale agriculture. Thus, HarMV is a virus with smaller known host and vector populations than the other four potyviruses considered here. Its biology is discussed in more detail below. A summary of the results of the popgen analyses of the non-recombinant (n-rec) ORF sequences of these five viruses is shown in [Figure 5](#viruses-12-00132-f005){ref-type="fig"}, which includes an outline ML phylogeny of the ORF sequences of the five viruses, and parameters from a graph comparing their ORF and polyprotein patristic distances, comparable to [Figure 4](#viruses-12-00132-f004){ref-type="fig"}. The pairwise nt diversity for each virus (π) is, as expected, related to the maximum pairwise ORF patristic distance (max ORF-dist), and the horizontal width of the corresponding collapsed cluster in the tree. The ω values (dN/dS) for four of the viruses are in the range of 0.065 (PVY) to 0.086 (PPV) confirming that all are under strong negative selection; however, the value of 0.260 for the SwPVs indicate that these viruses are under less stringent negative selection and/or are not evolving as a single population, i.e., providing evidence that they are separate species \[[@B84-viruses-12-00132]\]. The slope (b) of the linear regression in a patristic graph comparing ORF and polyprotein trees ([Figure 4](#viruses-12-00132-f004){ref-type="fig"}) correlates with ω. The slope varied from 0.243 to 0.450 for the five individual viruses ([Figure 4](#viruses-12-00132-f004){ref-type="fig"}). The slope was even greater for between-virus comparisons, and, for the 152 potyviruses represented in [Figure 3](#viruses-12-00132-f003){ref-type="fig"}a,b, it increased from c. 0.6 at the smallest axis values to c. 1.3 ([Figure 4](#viruses-12-00132-f004){ref-type="fig"}). The popgen comparisons of the different phylogroups/groups/strains of PVY, TuMV, SwPVs, and HarMV isolates are given in [Tables S1 to S4](#app1-viruses-12-00132){ref-type="app"} and summarized in [Figure 6](#viruses-12-00132-f006){ref-type="fig"}. The data for PPV are from Hajizadeh et al. \[[@B31-viruses-12-00132]\]. All the groupings are monophyletic; the BRs group of TuMV is normally subdivided into a basal paraphyletic "basal-BR" group with a subgroup called "Asian-BR". It can be seen that around three-quarters of the groups are represented by 10 or more sequences. The diversity of each group (π) correlates with their collapsed cluster sizes with PPV groups having the smallest diversities and HarMV group/species having the largest. The ω ratio estimates for individual genes ([Figure 7](#viruses-12-00132-f007){ref-type="fig"}) are considered to assess the strength of selection against translational change in the encoded protein. These fall into three groups; the proteins encoded by the n-rec genomes that are most strongly selected (smallest ω) are the HC-Pro, CI, Nia, and NIb proteins (ω = 0.047--0.055), an intermediate group is of the P1, P3, VPg, and CP proteins (ω = 0.101--0.229), and PIPO is least conserved (ω = 0.526), whereas the rec genomes give ω values that are around 10% greater, and in the same groupings except that VPg is one of the most strongly conserved proteins. Comparisons of the popgen results for the five viruses are instructive. There are five major phylogroups in the world PVY population ([Figure 6](#viruses-12-00132-f006){ref-type="fig"}), three of them (O, C, and N) are mostly n-rec, and two populations, R1 and R2, are based on single and double OxN recombinants. Commonest isolates from most potato-growing areas, and the most frequently sequenced, are O, R1, and R2 isolates. ORF sequences of the C phylogroup, which mostly have been isolated from a range of non-potato solanaceous hosts, are much more diverse ([Table S1.1](#app1-viruses-12-00132){ref-type="app"}; π = 0.113) than those of the O and N phylogroups (π = 0.027 and 0.043), which were mostly isolated from potatoes, and this diversity is also shown in the number of segregating sites (S), mutations within segregating sites (η), and the average number of differences between sequences from the same populations (k). The ω ratios for the individual ORFs ([Table S1.1](#app1-viruses-12-00132){ref-type="app"}), and for each of their genes separately ([Table S1.4](#app1-viruses-12-00132){ref-type="app"}), are less than one, indicating that their genomes are under dominant negative (purifying) selection; those of the n-rec isolates under stronger selection than those of the rec isolates (average ω 0.147 compared with 0.208), and with the PIPO gene ([Figure 7](#viruses-12-00132-f007){ref-type="fig"}; yellow bars) under least selection. The phylogroups are genetically distinct and there is no evidence of gene flow between them ([Table S1.3](#app1-viruses-12-00132){ref-type="app"}; Ks, Z, Snn, and Fst tests), confirming that after the primary recombination events which established the R1 and R2 lineages there was no significant gene flow between them. The Tajima's D test ([Table S1.1](#app1-viruses-12-00132){ref-type="app"}) gave a significant negative value for the O, R1, and R2 populations, confirming that they had recently expanded after a population bottleneck, and this reflects the adoption of potatoes as a major international crop after the mid-19th century with the O phylogroup and R1 and R2 necrogenic recombinants, but not the N and C phylogroups, as major pathogens of it. The major expansion evident in the Tajima's D scores was confirmed in popgen analyses of most of the individual PVY proteins ([Table S1.4](#app1-viruses-12-00132){ref-type="app"}), especially the HC-Pro, P3, CI, and VPg proteins. The evidence of less-stringent negative selection of the recombinant isolates (R1 and R2 phylogroups) is shown mostly in their P1 and PIPO genes. TuMV, which is common in cultivated and weed brassicas worldwide, has four major groups: world-B, Iranian, BRs (basal-BR and Asian-BR), and basal-B. It also has the orchid outgroup found only in Europe, as mentioned above in [Section 4.2.2](#sec4dot2dot2-viruses-12-00132){ref-type="sec"}, TuMV-OM \[[@B78-viruses-12-00132]\], and although currently included with TuMV, this is clearly distinct according to its genetic differentiation coefficient ([Table S2.3](#app1-viruses-12-00132){ref-type="app"}; mean Fst 0.81; \[[@B83-viruses-12-00132],[@B84-viruses-12-00132]\]. It was therefore excluded from [Figure 6](#viruses-12-00132-f006){ref-type="fig"}. Most diverse, and basal to the brassicas infecting isolates of TuMV proper, are the basal-B and Iranian populations, and although Tajima's D test gave negative values with both the world-B and BRs (basal-BR and Asian-BR) groups, neither was statistically significant. The TuMV population is more genetically diverse than the PVY population ([Figure 5](#viruses-12-00132-f005){ref-type="fig"}), but all TuMV genes, especially PIPO, are under stronger negative selection than those of PVY ([Figure 7](#viruses-12-00132-f007){ref-type="fig"}). These clear popgen differences between the PVY and TuMV populations probably reflect biological differences that have affected translational selection; the PVY population has recently expanded its range rapidly, mostly by trade, and into monoculture crops, whereas the TuMV population has diverged as vector aphids have moved it between and within annual crops and perennial weeds, and this greater diversity of hosts may produce greater negative selection. The PPV population has a smaller mean nt diversity than the PVY and TuMV populations; π for PPV is 0.121 ([Figure 5](#viruses-12-00132-f005){ref-type="fig"}) and a mean of 0.020 for its strains ([Figure 6](#viruses-12-00132-f006){ref-type="fig"}), but it has similar ω values. These differences probably result from its smaller population living in a long-lived woody host and transmission by grafting. Four of its major strain populations (D, M, CR, and C) have negative Tajima's Ds, but only that of the D strain is statistically significant, and this is the strain that spread during the 20th century across Europe and from there to the Americas and East Asia, whereas the others are mostly confined to Eurasia, which is probably the "center of divergence" of PPV. The ω values for the genes of different PPV strains ([Figure 7](#viruses-12-00132-f007){ref-type="fig"}), especially those with larger sample sizes, are dominated by large values for PIPO, but it can be seen that the conserved genes of PPV (HC-Pro, CI, NIa, Nib, and CP) are, like those of PVY, less conserved than those of TuMV. The popgen results for the SwPVs indicate that they are a lineage of separate viruses found in a single host species, not a single megaspecies. The mean nt diversity ([Figure 5](#viruses-12-00132-f005){ref-type="fig"}) of the SwPVs (π = 0.258) is greater than that of the other four viruses (π = 0.064--0.139), and the diversities of its component species range from 0.020 to 0.057 ([Figure 6](#viruses-12-00132-f006){ref-type="fig"}). They are also distinguished as species by the genetic differentiation coefficient (Fst \> 0.872) \[[@B85-viruses-12-00132],[@B86-viruses-12-00132]\]. The n-rec sequences of four of the SwPV populations are found worldwide, but that of sweet potato latent virus (SPLV) is not (see below), and all gave negative results in the Tajima's D test, but only that of SPVG was statistically significant; SPLV was excluded from these analyses as only three ORF sequences of it were available. Of interest too, is that all the SwPVs have a PIPO gene, like other potyviruses, but, in addition, all, except SPLV, have a second overlapping gene, PISPO, and the ω values ([Table S3.4](#app1-viruses-12-00132){ref-type="app"}) indicate that the PISPO genes of sweet potato feathery mottle virus (SPFMV) and sweet potato virus G (SPVG) are under positive selection (ω = 1.162 and 1.336) as is also the PIPO gene of sweet potato virus C (SPVC) (ω = 1.253). The HarMV population was represented by a single population of 10 ORF sequences. It is the only 'wild plant' potyvirus which we examined. The results are remarkably similar to those of the other four viruses we checked. The HarMV sequences have a nt diversity (π) similar to that of TuMV and a ω value (0.078) similar to those of PVY and PPV but less than that of the SwPVs ([Figure 6](#viruses-12-00132-f006){ref-type="fig"}, [Figure 7](#viruses-12-00132-f007){ref-type="fig"} and [Table S4.1](#app1-viruses-12-00132){ref-type="app"}). The ω values for individual HarMV genes ([Figure 7](#viruses-12-00132-f007){ref-type="fig"}) have a pattern within those of the other viruses, and with no evidence from Tajima's D tests of a major recent population expansion. The HarMV population was more variable than that of Mediterranean ruda potyvirus (MeRV), which has a similar ecology \[[@B87-viruses-12-00132]\]. In summary, the results of the popgen analyses complement and confirm inferences obtained from phylogenetic analyses of the same ORF sequences (see above in [Section 4.1](#sec4dot1-viruses-12-00132){ref-type="sec"}). They also suggest that various popgen estimates, such as ω, Fst, and b, might be used as indicators of whether a population of similar viruses is one or more species. This could provide a more theory-based way of defining which viruses form natural potyvirus species, and hence could replace the present unsatisfactory method based on arbitrary maximum %ID values (\<76% nt identity and \<82% amino acid identity; \[[@B88-viruses-12-00132]\]). Estimates of various popgen parameters indicate that the SwPVs are probably separate species, as too are the brassica-infecting and orchid-infecting TuMVs. The possibility of using popgen parameters to define species was tested further by checking whether the fast-evolving isolates of the C phylogroup of PVY \[[@B29-viruses-12-00132]\] were a separate species. These 28 extra sequences were added and the popgen factors recalculated, but only marginally changed the popgen parameters, indicating that PVY is still a single species. Further tests were also made using the ORF sequences of 16 isolates of narcissus yellow stripe-like (NYSLV) viruses, which fall into five distinct clusters, although biologically indistinguishable. Ohshima et al. \[[@B89-viruses-12-00132]\] reported that it was uncertain whether the NYSLVs should be considered one species or more, as their ORF sequence diversity is at the %ID limit set by the ICTV for delineating potyvirus species. The ω values for the four largest clusters, each of three sequences, had a mean of 0.055 (range 0.054 - 0.058), but as the clusters were progressively clustered, as in their ML tree \[[@B89-viruses-12-00132]\] ([Figure 2](#viruses-12-00132-f002){ref-type="fig"}) the value of ω increased to 0.117 \> 0.129 \> 0.146 \> 0.154 and finally 0.157 with the addition of the ORF of wild onion symptomless virus. These comparisons suggest that the NYSLVs are five independently evolving viruses, and that a calculated ω value greater than 0.1 is a useful indication that a population of potyviruses consists of more than one species. 5. Potyviruses in Space and Time {#sec5-viruses-12-00132} ================================ Attempts to add the dimensions of time and space to phylogenies of organisms, especially viruses, are currently a very active area of research \[[@B9-viruses-12-00132]\]. Although virus genomes can be recovered from preserved plant specimens, the maximum age of such recoveries is only around 1000 years \[[@B90-viruses-12-00132]\]. However, it is also possible to determine the age of virus populations that are evolving at a measurable rate by comparing the gene sequences of isolates collected "heterochronously" over the longest possible time period and analyze them by using Bayesian Monte Carlo coalescent methods in the BEAST packages \[[@B91-viruses-12-00132]\], or by regression methods such as "Least Squares Dating" \[[@B92-viruses-12-00132]\]. Bayesian methods also now extend to assessing virus migration pathways, using SPREAD \[[@B93-viruses-12-00132]\], etc. These methods have been used for elegant analyses of the spread of maize streak mastrevirus \[[@B94-viruses-12-00132]\], tomato yellow leaf curl begomovirus \[[@B95-viruses-12-00132],[@B96-viruses-12-00132]\], and of rice yellow mottle sobemovirus \[[@B97-viruses-12-00132],[@B98-viruses-12-00132]\], all of which, importantly, corroborate their conclusions using historical records of the spread of the diseases they cause. Studies on migration pathways for potyviruses are limited to TuMV \[[@B99-viruses-12-00132],[@B100-viruses-12-00132]\]. Studies of timescales have been made with several potyviruses. CP genes were used in early estimates of the "time to most recent common ancestor" (TMRCA) of the zucchini yellow mosaic virus (ZYMV) population \[[@B101-viruses-12-00132]\], the papaya ringspot virus (PRSV) population \[[@B102-viruses-12-00132]\], and of the initial radiation of all potyviruses \[[@B103-viruses-12-00132]\]. These studies gave compatible dates of 408 years before present (YBP), 2250 YBP, and 6600 YBP, respectively. CP genes were also used to obtain a recent estimate \[[@B104-viruses-12-00132]\] of 129--169 YBP as the age of the present narcissus late season yellows virus population. However, most dating studies of potyviruses have been of PVY and TuMV populations ([Table 1](#viruses-12-00132-t001){ref-type="table"}) and are based on gene sequences varying from full length ORFs of more than 9000 nts to the VPg gene of only 564 nts \[[@B29-viruses-12-00132],[@B30-viruses-12-00132],[@B100-viruses-12-00132],[@B105-viruses-12-00132],[@B106-viruses-12-00132]\]. The results show an interesting and unexpected positive correlation between the length of the sequence and its estimated TMRCA ([Table 1](#viruses-12-00132-t001){ref-type="table"}); the longer the sequence analyzed, the older the apparent age of the common ancestor. External corroborative evidence for the PVY and TuMV dating studies is imprecise. TuMV studies \[[@B99-viruses-12-00132]\] estimated that the virus arrived in Australia and New Zealand from Europe, at dates that were consistent with records of its first appearance in those countries, and the timescale obtained from the complete ORFs of PVY \[[@B29-viruses-12-00132]\] dates major divergences of its two main lineages to the mid-19th century, and hence they coincide with a major European effort to breed potato late blight (*Phytophthora infestans*)-resistant potatoes and a significant increase in European potato cropping. The estimated TMRCA of the population of one potyvirus can be extrapolated to those of other potyviruses, if their gene sequences align, and the alignment is used to calculate and compare them in a single ML tree; Fuentes et al. \[[@B29-viruses-12-00132]\] in their [Figure 4](#viruses-12-00132-f004){ref-type="fig"} showed that dates obtained from a ML tree and a Bayesian maximum clade credibility tree are linearly related. This is the basis of the "sub-tree comparison" method first used by Mohammadi et al. \[[@B107-viruses-12-00132]\], who showed that the TMRCA of the known world beet mosaic virus (BtMV) population is compatible with a 19th-century emergence of BtMV from wild beet rather than an ancient infection of chard or leaf beet. It was also used to show that the TMRCA of PPV was probably around 820 (range 865--775) BCE \[[@B31-viruses-12-00132]\], which is compatible with the invention of fruit tree grafting at the beginning of the first millennium BCE \[[@B108-viruses-12-00132]\]. When aligned representative sequences of TuMV, PVY, PPV, and the basal groups of the potyvirus tree were used to calculate an ML tree ([Figure S1](#app1-viruses-12-00132){ref-type="app"}), the divergence of the potyviruses and rymoviruses was dated, using the TuMV TMRCA ([Table 1](#viruses-12-00132-t001){ref-type="table"}), as 14,206 YBP, and using the PVY TMRCA, as 30,192 YBP. The later date is interesting as it suggests a possible route by which the PVY lineage invaded the Americas (see below, [Section 6.5](#sec6dot5-viruses-12-00132){ref-type="sec"}). 6. Evolutionary Vignettes from Potyvirus Studies {#sec6-viruses-12-00132} ================================================ 6.1. Super-Adapters {#sec6dot1-viruses-12-00132} ------------------- The conclusion that the potyviruses originated in Eurasia only a few tens of thousands of years ago is important, as it helps define the conditions that have allowed potyviruses to migrate and become the important and speciose genus it is nowadays. The hallmark property of potyviruses seems to be their ability to infect and maintain populations in a wide range of plants and to switch hosts often, but be constrained to particular clades of eudicotyledons. The divergence of potyviruses has occurred during the period when humans and their activities have come, increasingly, to dominate the world, and the ability of potyviruses to exploit those conditions have allowed them to become numerous and often damaging. Although humans originated in Africa around 300,000 YBP, they did not migrate beyond Africa's immediate environs until c. 80,000 YBP, then migrating to Asia and Australia, and eventually the Americas. Trade is an ancient human activity (<https://en.wikipedia.org/wiki/Trade#Prehistory> (accessed Dec 6 2019)), and seed-borne pathogens, such as potyviruses, are likely to have been carried along the 'Spice Trade' routes of the Indian Ocean, the Silk Road of Central Asia for many centuries. However, arguably the most important factor contributing to the success of the potyviruses and geminiviruses has been worldwide marine trade, which started with the exploration of the Americas by Columbus in the 15th century and resulted in the major exchange of plants and animals between the Americas and other regions of the world known as the Columbian Exchange (<https://en.wikipedia.org/wiki/Columbian_exchange> (accessed 6 December 2019)). Most recently, travel and trade by air has become important in the worldwide movement of viruses and their hosts despite an increased awareness of the importance of quarantine. 6.2. Australian Potyviruses {#sec6dot2-viruses-12-00132} --------------------------- A large number of potyviruses have been found on the isolated continent of Australia, with many of them being found only there. They provide insights into the timing of potyvirus migration, as Australia was biologically isolated for many millions of years (<https://en.wikipedia.org/wiki/Wallace_Line> (accessed 16 December 2019)). Australia was first colonized by the Aborigines at least 40,000--65,000 years ago \[[@B109-viruses-12-00132]\] (<https://en.wikipedia.org/wiki/History_of_Indigenous_Australians> (accessed 10 December 2019)). It was linked to Papua New Guinea by a land bridge during the last Ice Age, forming the continent of Sahul, until this bridge was flooded 6500--8000 years ago \[[@B110-viruses-12-00132]\]. The Austronesian voyagers from Asia visited the north coast of Sahul during 3000--4000 YBP, but not until the 17th century did the Macassans collect trepang or bêche-de-mer and camp on the beaches of Northern Australia. Europeans explored the entire coastline of the continent around the same time and established permanent colonies in the 18th century. Regular trade between Australia and other parts of the world started about two centuries ago and has increased ever since. The spreading of potyviruses depends on the 'nonpersistent' transmission by migrating aphids, especially those of the dioecious *Aphidini*, which, although much more ancient \[[@B111-viruses-12-00132]\], have also been favored by the development of agriculture. Australia had, until recently, a very small population of potential aphid vectors of potyviruses (i.e., *Aphidoideae*), with no more than 20 endemic species; however, there are now around 200 other species of the 4700 mostly recorded in the temperate Holarctic regions of the world \[[@B112-viruses-12-00132],[@B113-viruses-12-00132],[@B114-viruses-12-00132],[@B115-viruses-12-00132]\]. Potyviruses constitute the largest virus group with representatives known from only Australia, with most being represented only by single partial nt sequences in GenBank. They fall cleanly into one or other of two groups. There are those commonly found in crops, and also found in crops in other regions of the world, and to which they are closely related. These potyviruses most probably entered Australia recently in trade. There is also a large number of potyviruses found in native or introduced weed species, but these are more distantly related to potyviruses from other regions of the world. Most of the latter are members of the BCMV group (Group 3, [Figure 3](#viruses-12-00132-f003){ref-type="fig"}a), whereas the recent migrants are from various groups. The Australian BCMV group potyviruses form two clusters. One consists of ceratobium virus Y (CerVY), dianella chlorotic mottle virus (DiCMV), euphorbia ringspot virus (ERV), glycine virus Y (GVY), kennedya virus Y (KVY), passiflora foetida virus Y (PfoVY), passiflora virus Y (PaVY), pleione virus Y (PleVY), pterostylis virus Y (PtVY), rhopalanthe virus Y (RhoVY) and sarcohilus virus Y (SarVY), and the other of clitoria chlorosis virus (ClCV), clitoria virus Y (CliVY), diuris virus Y (DiVY), eustrephus virus Y (EustVY), HarMV, hibbertia virus Y (HibVY), passiflora mosaic virus (PaMV), passionfruit woodiness virus (PWV), and siratro viruses 1 and 2 (Sir1VY and Sir2VY) \[[@B11-viruses-12-00132],[@B82-viruses-12-00132],[@B116-viruses-12-00132],[@B117-viruses-12-00132],[@B118-viruses-12-00132],[@B119-viruses-12-00132],[@B120-viruses-12-00132],[@B121-viruses-12-00132],[@B122-viruses-12-00132]\]. This suggests that there were two near-simultaneous incursions into Australia of viruses of the BCMV group, and the fact that PaVY is also found in New Guinea \[[@B120-viruses-12-00132]\] supports the suggestion \[[@B83-viruses-12-00132]\] that the early BCMV group viruses were brought from Asia to the north coast of Sahul by Austronesian speakers who were not only explorers but also colonizers; they traveled in outrigger canoes with various live domesticated plants and animals so that they could establish viable communities when they found habitable islands. Recent studies using high throughput sequencing of plants belonging to the Southwest Australian native flora detected three additional potyviruses of orchids: blue squill virus A (BSVA), donkey orchid virus A (DOVA), and caladenia virus A (CalVA) \[[@B123-viruses-12-00132],[@B124-viruses-12-00132],[@B125-viruses-12-00132]\]. BSVA group with HarMV and PWV within the BCMV potyvirus group \[[@B125-viruses-12-00132]\], whereas CalVA and DOVA do not \[[@B103-viruses-12-00132],[@B125-viruses-12-00132]\], and their date and mode of entry to Australia is unknown. 6.3. Papaya Ringspot Virus {#sec6dot3-viruses-12-00132} -------------------------- The spread of PRSV around the world is also an informative story. It was first recorded in Hawaii, by Jensen \[[@B126-viruses-12-00132]\], and has been reported since then from most of the world's tropical and subtropical regions. PRSV is a typical potyvirus, as it is transmitted non-persistently by several aphid species \[[@B127-viruses-12-00132]\], and also occasionally in seeds of papaya (*Carica papaya*) \[[@B128-viruses-12-00132]\] and *Robinia pseudoacacia* \[[@B129-viruses-12-00132]\]. In 1984, it was shown to be closely related to one of the potyviruses common in wild and cultivated cucurbit crops worldwide, and it was called watermelon mosaic virus 1 at that time \[[@B127-viruses-12-00132]\]. These viruses are now called the PRSV-P (papaya) and PRSV-W (watermelon) biotypes of a single species; PRSV-W infects cucurbits, but not papaya, whereas PRSV-P infects both papaya and cucurbits. Cucurbitaceous weeds are natural reservoirs of infection for crops of both PRSV biotypes \[[@B130-viruses-12-00132],[@B131-viruses-12-00132]\]. PRSV is closely related to several other viruses of cucurbits in the Middle East and Africa (Group 4, [Figure 3](#viruses-12-00132-f003){ref-type="fig"}a; \[[@B58-viruses-12-00132],[@B102-viruses-12-00132],[@B103-viruses-12-00132],[@B118-viruses-12-00132]\]), one of which is recorded to infect papaya in the Congo, causing serious disease including ringspots \[[@B132-viruses-12-00132]\]. PRSV probably arose in South Asia and spread from there to the remainder of the world, including the Americas \[[@B58-viruses-12-00132]\]. An isolated outbreak of papaya ringspot disease in Australia was particularly informative as phylogenetic studies of the CP gene sequences of PRSV from the outbreak \[[@B133-viruses-12-00132],[@B134-viruses-12-00132]\] grouped with sequences from the long established local PRSV-W population, rather than others from overseas. Thus, the local PRSV-W population was the likely source of the PRSV-P outbreak. Olarte-Casillo et al. \[[@B102-viruses-12-00132]\] questioned this conclusion when they used Mesquite 4.7 \[[@B135-viruses-12-00132]\] to make an "ancestral reconstruction" of the PRSV population, using all known CP sequences, and concluded that "ancestral state could be either of the two biotypes". Ancestral reconstruction assumes, however, that the data used are a representative sample of the world PRSV population, and c. 70% of the sequences they analyzed were of PRSV-P biotype, whereas surveys of 22 Pacific Islands \[[@B136-viruses-12-00132]\], where papaya and various cucurbits are commonly grown for food, found PRSV-W on 17 different islands, but PRSV-P on only four. More recently, Maina et al. \[[@B137-viruses-12-00132]\] found evidence that only one PRSV-W introduction, or multiple introductions of very similar isolates, has occurred to Australia since agriculture commenced following the arrival of European colonists. Thus, PRSV-W genomic sequences from the entire northern coastline of Australia most resembled those from Papua New Guinea (PNG), but differed from those from East Timor (ET) and elsewhere. It seems that the PRSV-W to PRSV-P conversion occurs very infrequently. It was suggested \[[@B138-viruses-12-00132]\] that this host shift is controlled by the VPg-NIa-NIb region of the genomes of experimentally constructed PRSV-W x PRSV-P recombinants. Comparisons showed that simultaneous changes in three amino acids in the NIa were most likely involved: an Ala (W) to Val (P), an Asp (W) to Lys (P), and a Val (W) to Ile (P), all involving one or more first or second codon position mutations, which is an unlikely event. 6.4. The Columbian Exchange {#sec6dot4-viruses-12-00132} --------------------------- It is clear that the host ranges and distribution of potyviruses have been greatly affected by the Columbian Exchange, mentioned above, in [Section 6.1](#sec6dot1-viruses-12-00132){ref-type="sec"}. For example, PRSV, as described above, in [Section 6.3](#sec6dot3-viruses-12-00132){ref-type="sec"}, is part of a lineage of Old World cucurbit viruses. It was reported by Olarte-Castillo et al. \[[@B102-viruses-12-00132]\] to have a TMRCA of 2250 (95%CI 9800-250) YBP, which is congruent with the report of Desbiez et al. \[[@B58-viruses-12-00132]\] that the TMRCA of the PRSV cluster is 3600 YBP. However, papaya is a native of tropical Southern Mexico and Central America and was confined to that area until about 1550 CE (470 YBP), when Spaniards carried seeds to the Philippines, and from there to Malacca, India, and to Naples in 1626 CE, so that now papaya is commonly grown in all tropical regions of the world \[[@B139-viruses-12-00132]\]. These facts indicate that PRSV-W is the pre-Columbian version of this virus of cucurbits, and, only recently, has papaya been spread around the world by mankind and, occasionally, where they have met "papaya ringspot disease" has been generated. There are several other examples of "new encounter" potyvirus diseases involving a crop species from one region of the world that has been taken for the first time by mankind into the territory of a potyvirus, resulting in disease, often severe \[[@B140-viruses-12-00132]\]. Sweet potato, for example, was domesticated in Central America \[[@B141-viruses-12-00132]\]. It probably spread by natural means to Oceania, where it was found and exploited by the Austronesian explorers \[[@B141-viruses-12-00132]\]. Iberian galleons took sweet potatoes and other New World crops first to the Philippines and then to China and Okinawa by the early 17th century, and later to Korea. Sweet potato arrived in Europe across the Atlantic and was recorded in England in 1604 (<https://en.wikipedia.org/wiki/Sweet_potato> (accessed on 10 December 2019)). There is evidence that the five related potyviruses (Lineage 8, [Figure 3](#viruses-12-00132-f003){ref-type="fig"}b) found in sweet potato throughout the world may have originated outside the Americas, long before sweet potato encountered them. SPFMV, SPVC, SPVG, and sweet potato virus 2 (SPV2) now have a worldwide distribution, judging from GenBank records, which reveal that the sequenced isolates came from Africa, the Americas, Asia, Europe, and Oceania (SPFMV, 34 countries/ 443 records; SPVC, 16/69; SPVG 19/150 and SPV2 9/22). By contrast, SPLV, which is the basal sister to the others, has 92 records, but only from China, South Korea, Taiwan, and Tibet, and all the most closely related viruses (PPV, hyacinth mosaic virus, and asparagus virus 1) are Asian or European, suggesting that the center of divergence of Group 8 is Asia, not the Americas. 6.5. The Major Lineages: BCMV and PVY {#sec6dot5-viruses-12-00132} ------------------------------------- The BCMV and PVY lineages are the largest in the potyvirus phylogeny ([Figure 3](#viruses-12-00132-f003){ref-type="fig"}a,b; Groups 3 and 5). Both are clearly delineated by long branches, indicating a period in their history that produced no other known survivors. The BCMV lineage mostly radiated in SE Asia and Australia \[[@B83-viruses-12-00132]\], and its primary hosts are mostly rosids and monocots, and viruses of the group have caused most damage in crop species that originated in other areas of the world: common beans (*Phaseolus vulgaris*) from central America, passion fruit (*Passiflora edulis*) from South America, cowpea (*Vigna unguiculata*) from central Africa, only soybean (*Soja max*) being from SE Asia. By contrast, most of the primary hosts of the PVY lineage are asterids, and Fribourg et al. \[[@B76-viruses-12-00132]\] noted that all but two of the 27 viruses of the PVY lineage were isolated in the Americas, 17 of them having never been found anywhere else. Thus, it is likely that the PVY lineage diversified in the Americas, although how and when it originally migrated there from the Old World are interesting questions. If the TMRCA of PVY in [Table 1](#viruses-12-00132-t001){ref-type="table"} is correct, then [Figure S1](#app1-viruses-12-00132){ref-type="app"}, a ML tree of 38 representative ORF sequences, indicates that the long basal branch of the lineage covered the period 15.4--18.8 thousand YBP. This is the period human progenitors of the Amerindians are most likely to have migrated to the Americas (<https://en.wikipedia.org/wiki/Settlement_of_the_Americas> (accessed 17 December 2019); <https://en.wikipedia.org/wiki/Indigenous_peoples_of_the_Americas> (accessed 23 Dec 2019)) from Beringia, which was, at that time, at the eastern end of the mammoth steppe biome. This biome stretched from Iberia, across the north of Eurasia to Beringia (<https://en.wikipedia.org/wiki/Mammoth_steppe> (accessed 11 December 2019)). Around 16 thousand YBP, ice retreated enough to open one, possibly two, routes from Beringia into the Americas \[[@B142-viruses-12-00132]\]. The mammoth steppe was sub-Arctic and dominated by willow shrubs, grasses, and herbs, including many *Artemisia* spp. \[[@B143-viruses-12-00132],[@B144-viruses-12-00132]\], which are asterids, like many primary hosts of the PVY lineage viruses. It seems that no one has surveyed the viruses of *Artemisia* yet, and to do so might be very instructive. 6.6. Genetic Connectivity {#sec6dot6-viruses-12-00132} ------------------------- This is a biosecurity term that refers to situations where genetically similar nt sequences occur among populations of the same virus obtained from infected plants growing in different countries \[[@B14-viruses-12-00132]\]. In studies with potyvirus isolates from infected crops in Northern Australia and nearby countries ET and PNG, genetic connectivity was demonstrated in three instances. These were between one location in Northern Australia and ET for ZYMV \[[@B145-viruses-12-00132]\], two locations in Northern Australia and ET for SPFMV \[[@B146-viruses-12-00132]\], and, as mentioned above in [Section 6.3](#sec6dot3-viruses-12-00132){ref-type="sec"}, locations spanning the entire region of Northern Australia and PNG for PRSV \[[@B137-viruses-12-00132]\]. Such findings indicate that important potyviruses of economically important crops are crossing the sea separating Northern Australia from nearby countries to the north. A possible explanation is that this connectivity has arisen through trade in either direction. Alternatively, it may have resulted from spread by viruliferous insect vectors blown across the sea by annual monsoonal wind currents. These findings emphasize the need for improved biosecurity measures to protect against potentially damaging international virus movements occurring by natural means in addition to those resulting from international trade in plant materials \[[@B137-viruses-12-00132],[@B145-viruses-12-00132],[@B146-viruses-12-00132]\]. 6.7. Seed Transmission {#sec6dot7-viruses-12-00132} ---------------------- Several of the most important crop potyviruses have been shown to be seed-borne \[[@B16-viruses-12-00132],[@B147-viruses-12-00132]\]; however, many other potyviruses have not. Regardless, there are many records of crop potyviruses being found that could only be logically explained by seed transmission; indeed, it is likely that potyviruses are more frequently seed-borne than published records suggest perhaps because inadequate numbers of seeds were tested \[[@B148-viruses-12-00132]\]. There are many examples of potyviruses being spread from one part of the world to another by the international seed trade \[[@B14-viruses-12-00132],[@B16-viruses-12-00132]\]. Moreover, sowing potyvirus-infected seeds results in multiple infection foci consisting of infected seedlings scattered at random throughout crops. This constitutes a critical source from which aphid vectors can acquire a potyvirus and spread it throughout a crop resulting in serious virus disease epidemics \[[@B149-viruses-12-00132],[@B150-viruses-12-00132],[@B151-viruses-12-00132]\]. 6.8. Potyvirus Emergence from Natural Ecosystems {#sec6dot8-viruses-12-00132} ------------------------------------------------ There is evidence of agriculture being practiced by Australia's aboriginal population well before the island continent was first colonized by Europeans in 1788. The crops they grew consisted of Australian native plants, such as wild yams and grasses grown for their storage roots and seeds, respectively \[[@B152-viruses-12-00132]\]. Interfaces between natural vegetation and introduced crop species only arose after 1788, and, in many parts of the continent, are much more recent than that. Such interfaces are therefore well suited to studies on potyvirus emergence from native vegetation to infect introduced crop plants and weeds, and vice versa \[[@B140-viruses-12-00132]\]. Two such studies documented the spread of HarMV from infected *H. comptoniana* plants to nearby plants of introduced lupin species \[[@B153-viruses-12-00132],[@B154-viruses-12-00132]\]. PWV may have spread from its indigenous Australian host *Passiflora aurantia* (golden passion flower) to cause widespread infection in recently introduced crop, forage and naturalized weed legumes and *Passiflora* spp., which originated in South and Central America, or East Asia \[[@B103-viruses-12-00132],[@B121-viruses-12-00132],[@B122-viruses-12-00132],[@B140-viruses-12-00132],[@B155-viruses-12-00132]\]. PaMV and PaVY have been found infecting the passionfruit crop and naturalized introduced weeds so presumably also spread to them from an as yet unknown native plant host. Similarly, the indigenous Australian potyviruses of *Clitoria* and siratro (*Macroptilium atropurpureum*) have so far only been isolated from naturalized introduced weeds but presumably spread from unknown native plant hosts to introduced species after European colonisation \[[@B82-viruses-12-00132],[@B118-viruses-12-00132],[@B155-viruses-12-00132]\]. CerMV, DiVY, EustVY, GVY, HibVY, KVY, PleVY, and PtVY, have so far only been found infecting native plant species \[[@B118-viruses-12-00132],[@B155-viruses-12-00132]\], whereas RhVY and SarVY have only been isolated from imported orchids \[[@B118-viruses-12-00132]\]. In situations where a stable mixed plant population infected with viruses has co-evolved with native wild plants over a significant period in a given world region, populations of its isolates collected over a limited geographic range are likely to be diverse (e.g., Spetz et al. \[[@B156-viruses-12-00132]\]). Isolates of the most studied Australian indigenous potyviruses, HarMV, PaVY and PWV, provide excellent examples of this phenomenon \[[@B82-viruses-12-00132],[@B122-viruses-12-00132]\]. The localized natural distribution of HarMV reflects that of its Australian principal native host plant *H. comptoniana*. PWV and PaVY occur most commonly in the warmer tropical and subtropical regions of the Australian continent, where, as described above in [Section 6.2](#sec6dot2-viruses-12-00132){ref-type="sec"}, they infect a range of hosts. When ClVY, HarMV, PaVY, and PWV were inoculated to Australian native plant species, some that became infected produced surprisingly severe systemic symptoms. This observation did not support the suggestion that indigenous viruses are likely to be harmless when they infect wild plants. Instead, it indicates that, when they encounter native plant species they are not adapted to, they will likely behave in the wild just like crop viruses behave when they infect new crops \[[@B157-viruses-12-00132]\]. 6.9. Historical Potyvirus Specimens {#sec6dot9-viruses-12-00132} ----------------------------------- Historical collections of plant viruses are proving extremely valuable in linking the pre- and post-sequencing eras of potyvirus research. This applies not only to providing sequence data for dating studies described above in [Section 5](#sec5-viruses-12-00132){ref-type="sec"}, as these require comparisons between the sequences of old and recent potyvirus isolates covering the largest possible timespan, but also in revealing whether potyviruses studied in the post-sequencing era are named correctly, and if new sequences from unpublished potyvirus isolates from historical collections can be linked with those of established viruses. Studies on historical isolates preserved since the 1970s and 1980s from the Andean region of South America and Europe provide examples. With PVY, a dating study without these historical isolates suggested the "time to most common ancestor" (TMRCA) of PVY was the 16th century \[[@B105-viruses-12-00132]\], whereas a similar study which included sequences from European isolates from the 1980s \[[@B158-viruses-12-00132]\] placed its TMRCA around 1000 CE \[[@B106-viruses-12-00132]\]. When new PVY isolates from the Andean potato domestication center were included in dating analyses, along with the 1980s European sequences, the TMRCA was estimated to be around 156 CE \[[@B29-viruses-12-00132]\]. This study concluded that, although PVY was first taken to Europe in the 16th century in potato tubers after the Spanish conquest of Peru, as mentioned above in [Section 5](#sec5-viruses-12-00132){ref-type="sec"}, most of the current PVY diversity developed in the mid-19th century, when additional potato breeding lines were imported to help develop potatoes resistant to the potato late blight pandemic then occurring. Historical samples were also valuable when the sequence of a historical arracacha virus Y (ArVY) isolate collected in Peru in 1976 was compared with other potyvirus sequences, as it had 79% nucleotide identity with a 2013 Brazilian isolate of the subsequently described arracacha mottle virus, showing the latter was actually ArVY \[[@B159-viruses-12-00132]\]. Similarly, when a potyvirus sequence obtained from a 33-year-old mashua (*Tropaeolum tuberosum*) sample was sequenced and compared with other potyvirus sequences, it was found to be a distinct potyvirus to which the name mashua virus Y (MasVY) was given; MasVY's relationships to three other potyviruses (Tropaeolum mosaic virus, Tropaeolum virus 1 and Tropaeolum virus 2), described previously from the same host, are unknown, as the genomes of these viruses are still not sequenced \[[@B160-viruses-12-00132]\]. Finally it is noteworthy that two isolates of a potyvirus collected from pepino (*Solanum muricatum*) in Peru in 1976, and kept in a historical collection thereafter, have been recently shown, along with five new Peruvian pepino potyvirus isolates, to be isolates of wild potato mosaic virus, a virus previously known only from the wild potato *Solanum chancayense* \[[@B76-viruses-12-00132],[@B161-viruses-12-00132]\]. 6.10. Metagenomes, A New Frontier? {#sec6dot10-viruses-12-00132} ---------------------------------- Recently, new methods of gene sequencing have permitted surveys of virus-like genes and genomes in environmental samples (soil, water, air, etc.) and in bulk cellular material (blended invertebrates, etc.). The resulting metagenomic sequences have greatly expanded the size of the known virosphere \[[@B24-viruses-12-00132],[@B162-viruses-12-00132]\]. Two metagenomes appeared in the genomic sequences downloaded from GenBank, using the search term "Potyviridae", and both were reported from China. One is of Wuhan poty-like virus 1 (KX884573), which groups with the macluraviruses ([Figure 2](#viruses-12-00132-f002){ref-type="fig"}). It was isolated from the blended remains of 12 Chinese land snails (*Mastigeulota kiangsinensis*), collected in Wuhan, and has 73% ID with 95% cover of the genome of broad-leaved dock virus A (NC_038560) from *Rumex obtusifolius* from New Zealand. It is more distantly related to yam chlorotic necrosis virus (MH341583) from *Dioscorea alata* from India. Thus, Wuhan poty-like virus is more likely to be a constituent of the snails' last supper than a virus infecting the snails. The other metagenomic potyvirid is Hubei poty-like virus 1 (NC_032912), which was obtained from a "dipteran mix" of insects collected around Hubei. It is a recombinant with a major parent (nts 1-7630, 98% ID; 84% of the sequence) closest to an unusual isolate of sugarcane mosaic virus from *Canna* sp., in China (KY548507), and a minor parent (nts 7631-9094) closest to an unusual isolate of bean yellow mosaic virus (DQ060521; 95% ID) also isolated from "naturally planted" Canna sp., in China. This again suggests a contaminant virus, not an infection of the dipterans in the source material. Thus, both potyvirus-like metagenomes are more likely to be from contaminating plant materials than from the tissue of the animals being tested. Furthermore, Hubei poty-like virus 1 is probably an in silico recombinant, as its likely recombination site is adjacent to the region encoding the -GNNSGQP- motif, which is a conserved potyvirid motif \[[@B11-viruses-12-00132],[@B12-viruses-12-00132]\], and both 'parents' detected by BLASTn had been isolated from Canna plants. 7. Conclusions {#sec7-viruses-12-00132} ============== The major value of the esoteric studies of potyviruses described in this review is to provide a coherent evolutionary framework in time and space within which it is possible to plan how best to minimize the damage done by potyviruses to plants valued by humankind. Such a framework also contributes to the understanding of how potyvirus infections interact with plant populations within managed and natural ecosystems. Moreover, knowledge of the relationships of each virus, especially a newly emerged one, can provide valuable predictions of its likely behavior by extrapolating the known properties of near relatives and adding a timescale to those inferences greatly enhances their value and likely accuracy. The story in our review started in 1931, with PVY, one of the first two described viruses of potato, but now, nine decades later, at least 49 more viruses of potato have been recorded \[[@B163-viruses-12-00132]\]. The group of viruses, of which PVY is the type, is now represented in GenBank by genomic sequences of more than 150 different potyviruses and a third as many other potyvirids. Their discovery represents a considerable international effort as an analysis of the 2610 'country' records ([Table S5](#app1-viruses-12-00132){ref-type="app"}) shows that they came from 82 different countries from all regions of the world. The pioneers of potyvirology, Redcliffe Salaman, Kenneth Smith, Fred Bawden, Basil Kassanis, et al. discovered the first of a fascinating lineage of viruses that has provided us and others with an enormous amount of interesting and valuable work with no sign that the lode will 'run out' anytime soon. Apart from the four authors, nobody else helped with this study. ###### Click here for additional data file. The following are available online at <https://www.mdpi.com/1999-4915/12/2/132/s1>. Tables S1--S4. Popgen results; Table S5 (XLS file); The 'country' records of 2610 potyvirid genome sequences; Figure S1. A ML phylogeny of 38 representative potyviruses making subtree comparisons of node dates. R.A.C.J. initiated the project; A.J.G., M.H., and K.O. analyzed the data; all authors contributed to writing the paper. All authors have read and agreed to the published version of the manuscript. In Japan, this work was supported by the Japan Society for the Promotion of Science Grant numbers 16K14862 and 18K05653. It received no external funding from Australia or Iran. The authors declare that they have no conflicts of interest. ![The phylogeny of the RdRps of selected potyvirids, and of the other most closely related viral RdRps (redrawn from \[[@B24-viruses-12-00132]\]).](viruses-12-00132-g001){#viruses-12-00132-f001} ![A maximum likelihood (ML) phylogeny of 166 potyvirids, calculated using the same methods as \[[@B29-viruses-12-00132],[@B30-viruses-12-00132],[@B31-viruses-12-00132]\], from the protein sequences encoded by the RdRp regions of all genomes with GenBank Reference Sequence Accession Codes (July 2019), together with any others that had unique names. The most closely related non-potyvirid, bufivirus UC1, is included as an outlier. Branch colors indicate the major angiosperm "Order" from which each virus was isolated (i.e., its "primary host"); eudicotyledon rosid (red), asterid (blue), or caryophyllid (yellow) and monocotyledon (green) alismatid "A", lilioid "L", or commelinid "C". The RdRps of potyviruses branching most closely to the rymoviruses are shown, and the phylogeny of the other 103 is collapsed. The nodes with a red disk have \>0.9 SH support.](viruses-12-00132-g002){#viruses-12-00132-f002} ###### An ML phylogeny of 149 potyvirus ORF sequences with three rymovirus ORFs as outgroup using the same methods as \[[@B29-viruses-12-00132],[@B30-viruses-12-00132],[@B31-viruses-12-00132]\]. The ORFs are from all 126 potyviruses represented in GenBank in mid-2019 by Reference Sequences, together with single representative sequences of all 23 other potyviruses in GenBank with unique names. Note that (**a**) shows the three rymoviruses and Lineages 1--4, whereas (**b**) shows Lineages 5--9. The "Order" of the primary host, namely the plant from which each was first isolated, and often given in the name of the virus, is shown by the branch color (and letter), eudicotyledonous rosid (red), asterid (blue), or caryophyllid (yellow) and monocotyledonous alismatid "A", lilioid "L", or commelinid "C".   {ref-type="fig"}a,b and the ML tree of the protein sequences encoded by these ORFs; 149 potyviruses and three rymoviruses are included.](viruses-12-00132-g004){#viruses-12-00132-f004} {#viruses-12-00132-f005} {#viruses-12-00132-f006} {#viruses-12-00132-f007} viruses-12-00132-t001_Table 1 ###### Timescale analyses of potato virus Y and turnip mosaic virus. -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Parameter PVY TuMV ------------------------------------------------------------ ------------------------------------------------- ----------------------------------------------------- --------------------------------------------- ---------------------------------------------- ---------------------------------------------- ---------------------------------------------- ---------------------------------------------- \[[@B105-viruses-12-00132]\] \[[@B29-viruses-12-00132]\] and this study \[[@B106-viruses-12-00132]\] \[[@B100-viruses-12-00132]\] Method BEAST 1.7.2 BEAST 1.8.2 BEAST 1.8.4 BEAST 1.8.2 Number of sequences; 28 162 177 106 329 369 369 Sequence length (nucleotides) 9723 8913 564 9432 927 873 855 Region Polyprotein Polyprotein VPg Polyprotein Partial HC-Pro Partial P3 Partial NIb Sampling date range 1982--2010 1943--2016 1983--2015 1968--2012 1968--2012 1968--2012 1968--2012 Best-fit substitution model GTR+I+G GTR+I+Γ~4~ HKY+G~4~ GTR+I+G GTR+I+G GTR+I+G GTR+I+G Best-fit clock model UCLD, UCED, SC UCLD UCLD UCED UCED UCED UCED Best-fit demographic model ND EG, BSP BSP CZ CZ CZ CZ TMRCA (95% CI)\ 161--619 (UCLD), 123--436 (UCED), 525--970 (SC) 1841 (1157--2622) (EG),\ 158 (71--269) 1201 (468--2150) 951 (326--1291) 758 (274--1548) 758 (274--1548) Year before present (YBP) 1879 (1192--2659) (BSP) TMRCA effective sample size ND 238 (EG), 261 (BSP) ND ND ND ND ND Substitution rate (nt/site/year) Unknown 9.30 × 10^−5^ (6.79 × 10^−5^--1.18 × 10^-4^) (EG),\ 5.6 × 10^−4^ (3.35 × 10^−4^--8.17 × 10^-4^) 8.89 × 10^−4^ (6.87 × 10^−4^--1.30 × 10^−3^) 1.41 × 10^−3^ (1.09 × 10^−3^--1.78 × 10^−3^) 1.46 × 10^−3^ (1.25 × 10^−3^--1.87 × 10^−3^) 1.37 × 10^−3^ (1.04 × 10^−3^--1.73 × 10^−3^) 9.16 × 10^−5^ (6.90 × 10^−5^--1.15 × 10^−4^) (BSP) Substitution rates effective sample size ND \>200 \>200 ND ND ND ND Date (DRT) or cluster-based date (CDRT) randomization test Not passed DRT passed CDRT passed DRT passed DRT passed DRT passed DRT passed -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Best-fit clock model: uncorrelated relaxed lognormal distribution, UCLD; uncorrelated relaxed exponential distribution, UCED; strict clock, SC; expansion growth, EG; Bayesian skyline plot, BSP; constant size, CZ; exponential population growth, EPG. YBP, see sampling date range; ND, not done.

All relevant data are within the paper and its Supporting Information files. Introduction {#sec005} ============ Globally, an estimated 225 million women have unmet need for family planning \[[@pone.0153907.ref001]\]. Increasing acceptance and use of family planning requires more than increasing access to health services: truly effective family planning programs must address the social and gender norms that present critical barriers to sexual and reproductive health. Use of family planning is powerfully shaped by social norms, including perceived acceptability of family planning, social pressure for large families, and perceived opposition to family planning by religious and community leaders and spouses \[[@pone.0153907.ref002]\]. Rigid gender roles and unequal power between men and women inhibit couple communication and joint decision-making about family planning: power dynamics in couples have been found to influence the use of family planning and other health services \[[@pone.0153907.ref003]--[@pone.0153907.ref005]\]. Fear of and experience with intimate partner violence or gender-based violence are barriers to contraceptive use \[[@pone.0153907.ref006]--[@pone.0153907.ref009]\]. Women who report intimate partner violence are at higher risk for unintended pregnancy than women who do not report violence \[[@pone.0153907.ref008]\]. Analyses of data from Demographic and Health Surveys (DHS) in a number of African countries suggest that key dimensions of gender equity and women's empowerment---including equitable beliefs about gender roles, women's ability to negotiate sexual activity, and women's control over household economic decision-making---are associated with higher contraceptive use \[[@pone.0153907.ref010]\]. Several studies have also found a positive association between spousal communication and the use of contraception \[[@pone.0153907.ref011]--[@pone.0153907.ref014]\]. Despite evidence demonstrating the important relationships between gender, women's empowerment, and family planning, there is little consensus about how to best measure these complex social constructs. Researchers have attempted to measure specific domains of women's empowerment, and several scales have been developed including measures of mobility \[[@pone.0153907.ref015]\], power in relationships \[[@pone.0153907.ref003]\], gender norms and decision-making autonomy \[[@pone.0153907.ref016]\], control over assets \[[@pone.0153907.ref017]\], and social capital \[[@pone.0153907.ref018],[@pone.0153907.ref019]\]. The lack of a standard set of widely used measures for women's empowerment, however, hampers our ability to compare program effects across studies, populations, and cultural contexts. The objective of the research described in this article was to evaluate the impact of a CARE intervention that catalyzed community-level dialogue about gender, sexuality, and family planning on household-level gender dynamics and reported use of family planning among men and women in Kenya. For this evaluation, we used several scales from WE-MEASR (Women's Empowerment-Multidimensional Evaluation of Agency, Social Capital, and Relations) (see [S1 File](#pone.0153907.s001){ref-type="supplementary-material"}), a new tool that CARE developed following multi-year research into the effects and impact of our women's empowerment programming \[[@pone.0153907.ref020]\]. Methodology {#sec006} =========== Study Setting {#sec007} ------------- According to the 2008--09 Kenya DHS, approximately one in four women has an unmet need for family planning \[[@pone.0153907.ref021]\]. A 2013 analysis of DHS data found that Kenyan women who do not use family planning report method-related concerns, especially fear of harmful side effects, as the most common reason for non-use (43%); opposition to family planning (both the perceived religious prohibition of family planning and opposition by husbands/partners) is the second most commonly cited reason (16%) \[[@pone.0153907.ref022]\]. CARE conducted the research described here as part of the Family Planning Results Initiative (FPRI), which we implemented in Siaya County, Nyanza province, Kenya from July 2009 through December 2012. The majority of the county's population of 842,304 is under 30 years old, and 46.1% are between 0 and 14 years of age \[[@pone.0153907.ref023]\]. Over 90% of Siaya County is rural and, like the rest of Nyanza, falls behind the national average on various indicators of development, gender inequality, poverty, education, and health, including infant and under-five mortality, antenatal care coverage, and HIV prevalence \[[@pone.0153907.ref021]\]. The contraceptive prevalence rate among married women in Nyanza province is among the lowest in Kenya: 37.3% of women use any method (compared to 45.5% nationally), and only 32.9% use a modern method (39.4% nationally) \[[@pone.0153907.ref021]\]. The Intervention {#sec008} ---------------- CARE used its *Social Analysis and Action* approach \[[@pone.0153907.ref024]\] to shape FPRI's central intervention about 150 community-based facilitators were trained and held ongoing community dialogues about gender, sexuality, and family planning over three and a half years (July 2009-December 2012). At any one time, an average of 65 facilitators were actively convening dialogues. Monthly planning meetings were held to plan activities and ensure adequate location coverage and topic variation over the entire study area. On a quarterly basis, CARE observed and provided feedback on dialogues and completed progress reviews with facilitators. Based on our monitoring data, CARE organized 759 dialogues between July 2009 and December 2012. Dialogues were held in a variety of venues and were promoted by community leaders or organized around pre-scheduled community meetings. Based on proximity to a respective venue, they were attended by participants from several villages, and, many times, were included in villages' development plans. The dialogues were designed to normalize communication about sensitive topics like gender norms and FP, and to catalyze participants' critical analysis of how gender norms and dynamics restrict family planning acceptability and use. During dialogues, community leaders and satisfied family planning users acted as role models and overtly expressed support for family planning, equitable gender norms, open communication and shared decision-making regarding family planning between spouses. Trained, community-based facilitators convened dialogues in an array of settings, such as markets, churches, women's groups and village meetings. Local theatre groups often performed during the dialogues. Facilitators included community health care workers, religious leaders, local government officials, and teachers. CARE provided ongoing support to facilitators in the form of training and supervision, and also provided transport reimbursement and lunch/refreshments on days when dialogues or trainings were conducted; no incentives were provided for participation in intervention activities to any other community members. All of CARE's activities were coordinated with the District Health Medical Team (DHMT.) During the intervention period, the DHMT coordinated several interventions designed to increase availability and access to contraceptives, including in-service training for providers, strengthening of systems for ordering contraceptives, and some community outreach activities (FP counseling and method provision). To our knowledge, no other similar FP dialogue interventions or activities were implemented by other organizations during the intervention period. Theory of Change {#sec009} ---------------- The intervention aimed to challenge and shift key community norms about gender dynamics and family planning, with the ultimate goal of creating a social environment that is more supportive of equitable gender relations and the use of family planning. Our research goals were to determine whether and how the ongoing dialogues shifted social norms, and whether and how these shifts at the community level influenced communication, decision-making, and family planning use at the couple or household level. Our theorized pathway of change is shown in [Fig 1](#pone.0153907.g001){ref-type="fig"}. {#pone.0153907.g001} Evaluation Design {#sec010} ----------------- We used both qualitative and quantitative methods to evaluate the intervention. At baseline (February 2009) and again at endline (December 2012), we conducted household surveys to collect data about men and women's socio-demographic characteristics, pregnancy intentions, family planning knowledge, beliefs, attitudes, and use. At endline only, we measured key gender-related beliefs and behaviors and exposure to the intervention, which enabled us to determine whether these were important predictors of family planning use. Also at endline, we conducted in-depth, qualitative interviews with purposively selected couples from the intervention area (described below) to explore changes in communication, decision-making, and gender roles over the previous four years and to identify specific enablers and barriers to family planning use. Ethics Statement {#sec011} ---------------- CARE obtained approval to conduct this research from the Institutional Research and Ethics Committee at Moi University in Eldoret, Kenya. Quantitative Data Collection {#sec012} ---------------------------- Through independent, cross-sectional household surveys, married men (20--49 years) and women (18--45 years) were interviewed by trained interviewers using standardized questionnaires at baseline and endline. Although the two samples were entirely independent (no attempt was made at endline to include or exclude respondents who participated at baseline), both were drawn from the same Siaya county using an identical two-stage cluster sampling approach. Separately at baseline and endline, thirty sub-counties were randomly selected from Siaya's 130 sub-counties, and 60 villages were randomly selected from this sub-set of sub-counties (2 villages from each sub-county,). Households were then randomly selected for interviews using a systematic approach, selecting every 3^rd^ household for women and every 5^th^ household for men. Respondents were eligible for interview if they were married or in union and of reproductive age, as defined above. If household member listing revealed more than one woman/man meeting the eligibility criteria, one interviewee was randomly selected in each household. A total of 1,240 women (n = 650 at baseline, n = 590 at endline) and 590 men (n = 305 baseline, n = 285 endline) completed interviews. Data were double-entered and cleaned by trained clerks under the supervision of a data manager. Qualitative Data Collection {#sec013} --------------------------- To further explicate the quantitative findings, we used qualitative methods to explore gender dynamics and influences on family planning use at the couple level. Data were collected using face-to-face, in-depth interviews, and a visual timeline tool to elicit relationship histories and to explore changes in communication, decision-making, and gender roles during the previous four years. FPRI project staff identified men and women from married couples in the intervention county, and conducted interviews until saturation on key themes was reached (n = 10 couples). In each couple, one or both partners were between the ages of 18 and 49 and had participated in a minimum of three intervention dialogues between 2009 and 2012. Interviews were conducted separately with the man and woman of each couple; each was interviewed by a trained, gender-matched interviewer in the Luo language. Interviews were recorded, detailed notes were taken by a trained observer during the interview, and recordings were used to transcribe, review, and confirm completeness and accuracy of notes. To help ensure data integrity, interviewers completed written summaries on the same day as data collection. Two senior researchers confirmed completeness and clarity of notes from each interview. Narratives from each of the men's and women's interviews were then analyzed and triangulated to create a narrative summary of each couple's relationship and history of family planning use. After each couples' story was analyzed, content analysis was conducted across all narrative summaries to identify key themes related to the theory of change. Quantitative Measures {#sec014} --------------------- ### Family Planning Measures (baseline and endline) {#sec015} At both baseline and endline, men and women were asked identical questions about family planning knowledge, beliefs, attitudes, and use in order to measure changes in family planning over time (see [S2 File](#pone.0153907.s002){ref-type="supplementary-material"}). ### Family Planning Knowledge, Beliefs and Attitudes (Independent Variables) {#sec016} Men and women were asked to report the total number of family planning methods known, and if they knew where to obtain a method of family planning. Beliefs about family planning were measured using a 4-item index, in which men and women reported whether they disagreed (scored as 1), were unsure (scored as 2), or agreed (scored as 3) with four common family planning-related myths. The index score range is 4--12, and a higher score indicates less accurate knowledge about family planning. An index assessed men's and women's approval of family planning use in 5 different circumstances, with each response scored as 1 if the respondent approves and 0 if he/she does not approve. The index was constructed by summing the item scores. The index score range is 1‐5, and a higher score indicates a higher level of approval for family planning. Men and women were also asked if they believed it was acceptable for women to use a family planning method without their partner/husbands' permission, and women (but not men) were asked if they believed they could suggest using a condom to their partner/husband. ### Family Planning Use (Dependent Variable) {#sec017} Men and women were asked if they or their spouse had ever used a method to avoid or delay pregnancy; those who answered 'yes' were asked to list all the methods they had ever used. Men and women were also asked if they or their spouse were currently using a method to avoid or delay pregnancy; those who answered 'yes' were asked to list all the methods they were currently using. These data were used to create two dependent variables: *current use of any contraceptive method* and *current use of a modern contraceptive method* (oral contraceptives, male/female condoms, vasectomy and tubal ligation, intrauterine device, injectable contraception, hormonal implants). ### Exposure, Pregnancy Intentions and Gender Measures (Endline only) {#sec018} In order to measure key predictors of family planning use, at endline we asked men and women questions about their exposure to the intervention as well as their pregnancy intentions. We also used several indices and scales from CARE's WE-MEASR tool to measure key dimensions of gender equity and women's empowerment in the endline survey: men's belief in traditional gender roles (asked of men only), self-efficacy to use family planning (asked of women only), spousal communication and women's decision-making power (asked of both sexes). The WE-MEASR scales used in the evaluation were tested and confirmed to be statistically reliable for use with both men and women in this sample. ### Exposure to the intervention {#sec019} We asked men and women about their exposure to the intervention between 2009 and 2012. To ensure that reported exposure did represent participation in the intervention, FPRI managers suggested unique indicators of exposure, including a mix of dialogue topics unique to the intervention. The three exposure variables were: exposure to any dialogues (yes/no), the number of topics exposed to during the dialogues (\#), and exposure to dialogues specifically focused on family planning (yes/no). ### Pregnancy intentions {#sec020} At endline, we asked men and women whether they would like to have a child/another child, no more children, could not become pregnant, or were unsure. Those who reported wanting another child or being unsure were asked if they would like to get pregnant in the next 12 months, after 1--2 years, after more than 3 years or don't know. Men and women who did not want to have children or were not able to have children (due to sterilization or infertility) were not asked this question. Women/couples who were currently pregnant were also asked if they wanted to have another child after the child they were expecting now, wanted no more children or were unsure. Those who reported wanting another child or being unsure were asked how long they wanted to wait to have another child: within 12 months of the birth of the child they were expecting now, after 1--2 years, after more than 3 years, or don't know. ### Men's beliefs about gender (men only) {#sec021} We used a 9-item scale to measure how much men agreed or disagreed with statements about the roles and entitlements of men and women. We used a 5-point Likert scale, where Strongly Agree = 5, Agree = 4, Neither Agree Nor Disagree = 3, Disagree = 2, and Strongly Disagree = 1. The scale score was the sum of item scores divided by the number of items. The scale score range was 1‐5, and a higher scale score indicated a higher support for traditional gender roles (male dominance). Scale reliability in this sample was good (Chronbach's alpha = .66). ### Spousal communication (men and women) {#sec022} We used a 5-item scale with both men and women to measure interspousal communication across a range of topics. Men and women were asked to report how often they communicated about: what had happened during their day, their worries and fears, their household finances, their fertility intentions/desired family size, and use of family planning. We used a 5‐point Likert scale, where Always = 5, Often = 4, Sometimes = 3, Seldom = 2 and Never = 1. The scale score was the sum of item scores divided by the number of items. The scale score range was 1‐5, and a higher scale score indicated a higher level of interspousal communication. Scale reliability for both samples was good (men: Chronbach's alpha = .63; women: Chronbach's alpha = .75). ### Women's decision-making power (men and women) {#sec023} We measured women's decision-making power in the household by asking both men and women who usually made a range of key household decisions. The scale was not intended to measure a woman's decision‐making power over a specific kind of decision (e.g. household finances), but rather measure a woman's influence over a range of key decisions that affect her life. We used slightly different scales for men and women, as detailed below. We used a 12-item scale to ask women who usually made decisions about her health care, large household purchases, household purchases for daily needs, when she will visit family/relatives/friends, when the whole household will visit family/relatives/friends, how to use the money she brings into the household, how to use the money her spouse brings into the household, when to sell a large asset (e.g. cow), when to sell a small asset (e.g. chicken), whether she can work to earn money, when she and her husband have sex, and whether the woman and her husband use family planning. Item response options included: wife alone, wife and husband together, husband alone, mother‐ or father‐in‐law, someone else, and mother or father. Responses of wife alone or wife and husband together were scored as a 2 and all other responses were scored as a 1. The scale score was the sum of item scores divided by the number of items. The scale score range was 1‐2, and a higher scale score indicated women's self-perceived higher decision-making power. Reliability of this scale for women was good (Chronbach's alpha = .74). With men, we used a similar, 10-item scale that included all items described above, except whether women can work to earn money and who makes decisions about women's health care, The scale score range was 1‐2, and a higher scale score indicated men's perceptions of women's higher decision-making power. The reliability of this scale for men was also good (Chronbach's alpha = .79). ### Women's self-efficacy to use family planning (women only) {#sec024} We used a 4-item scale to measure women's self-efficacy. Women were asked how much they agreed or disagreed with four statements about how sure they felt that they could: discuss family planning with their husband, tell their husband they wanted to use family planning, use family planning, and use family planning without their spouses' approval. We assessed responses based on a 5‐point Likert scale, where Completely Sure = 5, Somewhat Sure = 4, Neither Sure/Unsure = 3, Somewhat Unsure = 2, and Not at all Sure = 1. The scale score was the sum of item scores divided by the number of items. The scale score range was 1‐5, and a higher scale score indicated higher self‐efficacy to discuss and use family planning. We found this to be a reliable scale (Chronbach's alpha = .75). Analyzing Changes in Family Planning Use {#sec025} ---------------------------------------- Sample characteristics were compared between baseline and endline using chi-squared tests for proportions and t-tests for means. We found several significant differences in the characteristics of the samples drawn at baseline and endline, and therefore used propensity scores to establish comparable samples at the two time points. Propensity scoring is a method of matching that uses available information on the characteristics of the study population to establish, in our case, matched pairs of baseline and endline survey respondents. To derive the propensity score used for matching, logistic regression models with survey round (endline vs. baseline) as the primary outcome variable were estimated separately for men and women using the following covariates: respondent's age, age difference between respondent and partner, number of living children, education, partner's education, whether the respondent works for cash, whether the woman makes decisions regarding use of any cash earned, pregnancy intentions, the number of family planning methods known, the summed score on the index assessing family planning beliefs, and the summed score on index assessing approval of family planning use. Only in the estimation models for men, we included knowledge of where to obtain family planning and score on a scale assessing beliefs about gender roles; and, only for women, we included the score on a family planning self-efficacy scale and a variable to assess whether they believed they could suggest the use of condoms to their partners. The predicted probabilities of being interviewed at endline vs. baseline from the models served as propensity scores, and the scores were then used with Stata's psmatch2 command to examine the average difference in family planning use (any family planning method and modern family planning methods) between baseline and endline. After assessing the use of different propensity score matching techniques, we chose kernel matching due to its efficiency. Only propensity score values included in the common support region (the area in which the distribution of propensity scores for being interviewed at endline overlapped with the distribution of propensity scores for being interviewed at baseline) were used in the analysis. This excluded respondents who were least likely to produce a reliable baseline-endline match based on the observed characteristics, specifically, 35 of 1,240 women and 12 of 590 men in our samples. Characteristics of the study population before and after propensity score matching were assessed to ensure that the data were fully balanced with respect to the covariates used to derive the propensity scores. Analyzing Predictors of Family Planning Use at Endline {#sec026} ------------------------------------------------------ To explore associations between exposure to the intervention and use of family planning at endline, we used the endline survey data and fitted separate logistic regression models for two key family planning outcomes: current use of any family planning method and current use of a modern family planning method. The key exposure covariates of interest used in three separate models were: exposure to the intervention (yes/no), the number of discussion topics exposed to during the intervention, and exposure to family planning discussions during the intervention (yes/no). Our multivariate models also explored the associations between current use of family planning (any and modern methods) and other key potential predictors of family planning use, including: childbearing intentions, family planning knowledge and beliefs, men's belief in traditional gender roles, men's and women's reports of spousal communication, men's and women's reports about women's household decision-making power, and women's self-efficacy to use family planning. The multivariate analysis controlled for several background factors, including age of respondent (less than 25 years, 25--29 years, 30--34 years, 35 years or older), years of school completed by respondent and by partner, religion (Protestant, Catholic, or other), and number of living children. We also included a variable measuring difference in spousal age (0--5 years, 5--9 years, 10--14 years, and more than 15 years' difference) which we hypothesized might magnify decision-making power differences between spouses. Analyses were conducted using Stata version 12.0. Results {#sec027} ======= Quantitative Results {#sec028} -------------------- ### Characteristics of Survey Sample {#sec029} [Table 1](#pone.0153907.t001){ref-type="table"} summarizes characteristics of the survey samples. A total of 1,240 women aged 15--45 (n = 650 at baseline, n = 590 at endline) and 590 men aged 18--49 (n = 305 baseline, n = 285 endline) completed interviews. At baseline, 55.6% of women interviewed were under the age of 29, as were 63.9% of women interviewed at endline. By contrast, only 34.7% of men at baseline and 44.9% of men at endline were under the age of 29. Most respondents did not report a large age difference between them and their spouses: 71.1% of women at baseline and 64.9% of women at endline reported age differences of fewer than 10 years. 10.1371/journal.pone.0153907.t001 ###### Sample characteristics. {#pone.0153907.t001g} Characteristics Women Men -------------------------------------------------------------------------------------------------- ------------ ------------ --------- -------------- -------------- --------- **Socio-demographic** Age (years; %) \<25 34.5 34.8 \<0.001 13.1 17.9 0.042 25--29 21.1 29.1 21.6 27.0 30--34 17.9 19.0 21.3 21.8 ≥35 26.6 16.4 43.9 33.3 Age difference between male and female partners (years; %)[^c^](#t001fn003){ref-type="table-fn"} \<5 30.3 31.0 0.030 41.0 41.4 0.962 5--9 40.8 33.9 38.4 39.7 10--14 18.9 21.0 14.4 13.3 ≥15 10.0 14.1 6.2 5.6 Number of living children Mean (std dev)[^d^](#t001fn004){ref-type="table-fn"} 3.5 (2.1) 3.3 (1.9) 0.025 3.4 (2.2) 2.8 (2.0) 0.001 Religion (%) Protestant 52.3 48.8 0.107 51.5 49.5 0.750 Catholic 27.1 25.6 30.5 31.9 Other 19.2 24.8 16.1 17.5 None/missing 1.4 0.9 2.0 1.1 Education (years) Mean (std dev) 7.0 (3.0) 8.2 (3.0) \<0.001 8.0 (3.2) 8.9 (3.4) \<0.001 Partner education (years) Mean (std dev) 8.4 (2.9) 9.1 (3.3) \<0.001 7.2 (2.9) 8.3 (2.8) \<0.001 Works outside home for cash (%) Yes 39.1 32.4 0.014 61.3 60.4 0.057 **Women's empowerment / Gender beliefs /Relationship with partner/pregnancy intentions** Woman makes decisions about spending the cash she earns (%) No cash earned 60.9 67.6 0.084 64.6 61.8 \<0.001 All cash 27.2 21.7 29.5 11.9 Some cash 10.6 9.8 5.9 16.5 Missing 1.2 0.9 0.0 9.8 Men's gender beliefs scale score[^e^](#t001fn005){ref-type="table-fn"} Mean (std dev) 2.0 (0.04) 2.0 (0.03) 0.820 Participation in household decision-making scale score[^f^](#t001fn006){ref-type="table-fn"} Mean (std dev) 1.6 (0.2) 1.6 (0.3) Inter-spousal communication scale[^g^](#t001fn007){ref-type="table-fn"} Mean (std dev) 2.9 (0.3) 3.4 (0.8) Pregnancy intentions (%) Wants pregnancy in next 12 months 30.1 35.1 0.001 60.0 64.2 0.026 Wants to delay pregnancy for ≥ 1 year 21.2 26.6 16.17 9.8 Wants no more children 23.9 21.7 4.3 8.4 Does not think about/not sure/not applicable 24.8 16.6 19.7 17.5 **Family planning knowledge/attitudes/subjective norms/ self-efficacy** Family planning beliefs score[^h^](#t001fn008){ref-type="table-fn"} Mean (std dev) 9.1 (1.6) 8.4 (1.9) \<0.001 8.8 (0.1) 8.4 (0.1) 0.005 \# methods known[^i^](#t001fn009){ref-type="table-fn"} Mean (std dev) 3.2 (1.3) 3.9 (1.5) \<0.001 3.4 (0.1) 3.4 (0.1) 0.784 Believe woman can use family planning w/o partner's permission (%) Agree 66.3 11.6 Believe woman can suggest use of condoms (%) Agree 19.4 15.4 0.067 Disagree/unsure 80.6 84.6 0.067 Score on family planning use approval questions[^j^](#t001fn010){ref-type="table-fn"} Mean (std dev) 11.1 (2.0) 11.9 (2.0) \<0.001 11.63 (0.12) 10.82 (0.10) \<0.001 Family planning self-efficacy scale score[^k^](#t001fn011){ref-type="table-fn"} Mean (std dev) 4.2 (1.0) Knows where to get family planning (%) Yes 45.4 **Exposure to intervention** Exposure to intervention, all topics (%) Yes 64.2 66.3 Number of topics exposed to during intervention Mean (std dev) 1.5 (1.5) 2.0 (1.8) Exposure to FP discussions during intervention (%) Yes 60.9 57.9 ^a^We restricted our sample to women aged 15--45 and men 18--49 in both survey rounds, in order to compare women and men of the same age range at the two time points. Thus, only 590 of the 617 women who completed surveys at endline were included in our analysis, and only 285 of the 302 men who completed surveys at endline were included in our analysis. Some information was collected only at endline ^b^Chi-squared tests used to compare percentages and t-tests used to compare means ^c^Difference between age of male and female partners ^d^Std dev: Standard deviation ^e^Range 1--3 ^f^Range 1--2 ^g^Range 1--5 ^h^Range 4--12 ^i^Range 1--12 ^j^Range 5--15 ^k^Range 1--5 Women reported completing fewer years of education than men at both time points, and endline samples comprised slightly more educated men and women. About half of men and women in both samples reported being Protestant. At both interview times, about a third of women and three fifths of men reported working outside the home for cash. Fewer men and women at endline than at baseline reported that women made decisions about the cash they earned: 28.4% vs 35.4%, respectively, among men and 31.6% vs 37.8%, respectively, among women. As measured at endline, men in this sample did not report strong support for traditional gender roles (mean score of 2.0 on a scale of 1--5, where 5 indicates stronger agreement with traditional gender roles). Both sexes reported that women had moderate to high household decision-making power (mean of 1.6 for both men and women on a scale ranging between 1 and 2, where a higher score indicates higher women's decision-making power). Also, men and women both reported moderate to high levels of spousal communication (mean for women 2.9, and for men, 3.4, on a scale of 1--5 where 5 indicates high levels of spousal communication.) At endline, only 11.6% of men as compared to 66.3% of women reported approving of women using family planning without their partners' permission. Very few women (15.4%) believed they could suggest the use of condoms to their husbands. Men and women also reported widely incongruent pregnancy intentions: 30.2% of women interviewed at baseline and 35.1% at endline reported wanting to get pregnant at the time of their last pregnancy, but 60.0% and 64.2% of men reported so at baseline and endline, respectively. Notably, over 20% of women at both time points reported wanting to limit childbearing compared to only 4.3% and 8.4% of men at baseline and endline, respectively. ### Changes in Family Planning Use from Baseline to Endline {#sec030} [Table 2](#pone.0153907.t002){ref-type="table"} shows unmatched and propensity score-matched differences in use of any and of modern methods of family planning between baseline and endline for men and women. Based on propensity score-matched results, 36.5% and 34.0% of women used any method and a modern method, respectively, at baseline, while 51.8% and 51.2%, respectively, did so at endline. Similarly, 33.7% and 27.9% of men used any method and a modern method, respectively, at baseline, and 53.8% and 52.2% did so, respectively, at endline. 10.1371/journal.pone.0153907.t002 ###### Changes in family planning use between baseline and endline. {#pone.0153907.t002g} Outcome Baseline (%) Endline (%) \% difference (std dev) ---------------------------------------------------------------- -------------- ------------- ------------------------- **Women** **Current use of any method** Unmatched 31.7 51.7 20.0 (2.7) Propensity score-matched[^a^](#t002fn001){ref-type="table-fn"} 36.5 51.8 15.3 (3.4) **Current use of modern methods** Unmatched 29.4 51.0 21.6 (2.7) Propensity score-matched 34.0 51.2 17.3 (3.2) **Men** **Current use of any method** Unmatched 29.5 54.0 24.5 (4.0) Propensity score-matched 33.7 53.8 20.1 (5.0) **Current use of modern methods** Unmatched 24.3 52.6 28.3 (3.9) Propensity score-matched 27.9 52.2 24.3 (4.8) ^a^Propensity scores (kernel matching technique) derived from logistic regression models with survey round (endline vs baseline) as outcome variable adjusted for the following covariates: respondent's age, age difference between respondent and his/her partner, number of living children, education, partner's education, whether the respondent works for cash, whether the woman makes decisions regarding spending of any cash earned, pregnancy intentions, the number of family planning methods known, the summed score on the index assessing family planning beliefs, the summed score on index assessing approval of family planning use, knowledge of where to obtain family planning (men only,) score on a scale assessing beliefs about gender roles (men only), score on family planning self-efficacy scale (women only) and whether they believed they could suggest the use of condoms to spouse (women only). At baseline, 19.5% of women in the sample were using injectables, 3.3% condoms, 2.9% oral contraceptives, 2.6% had a tubal ligation, 1.4% used implants or IUDs, and 2.0% traditional methods. Notably different at endline, 29.7% of women in the sample were using injectables, 8% used oral contraceptives and 6.8% condoms. Among men, 18.5% and 27.2% reported use of condoms at baseline and endline, respectively; while 4.7% and 3.7% reported that their partners used injectables and oral contraceptives at baseline, 18.3% and 10.4% did so, respectively, at endline (data not shown). ### Predictors of Family Planning at Endline {#sec031} Our fully adjusted regressions models identified significant associations between participation in the intervention and both current use of any method and current use of a modern method of family planning for women ([Table 3](#pone.0153907.t003){ref-type="table"}), but not for men ([Table 4](#pone.0153907.t004){ref-type="table"}). As expected, the strongest associations were found among women exposed to discussions specifically about family planning: these women were 1.78 times more likely to use any method or a modern method than women who were not exposed to these discussions. Yet exposure to any topic covered during the intervention increased the odds of women using any method of family planning by 62% and the odds of using a modern method by 60%. 10.1371/journal.pone.0153907.t003 ###### Key determinants of and use of family planning among women at endline. {#pone.0153907.t003g} Characteristics Current use any method Current use modern method --------------------------------------------------------------------------------------------------------------------------------- ----------------------------- ----------------------------- ----------------------------- ----------------------------- ----------------------------- ----------------------------- **Socio-demographic** Age (years; \<25 = ref) 25--29 **1.89 (1.15, 3.10)** **1.89 (1.15, 3.11)** **1.86 (1.13, 3.06)** **1.84 (1.12, 3.01)** **1.85 (1.13, 3.03)** **1.81 (1.11, 2.97)** 30--34 **1.82 (1.01, 3.27)** **1.80 (1.00, 3.24)** **1.83 (1.01, 3.29)** **1.87 (1.04, 3.34)** **1.86 (1.04, 3.33)** **1.88 (1.05, 3.37)** ≥35 0.79 (0.40, 1.57) 0.77 (0.39, 1.53) 0.78 (0.39, 1.56) 0.78 (0.39, 1.53) 0.76 (0.38, 1.50) 0.77 (0.39, 1.53) Age difference between partners (years; \<5 = ref)[^d^](#t003fn004){ref-type="table-fn"} 5--9 0.99 (0.62, 1.56) 1.02 (0.64, 1.62) 0.99 (0.62, 1.57) 0.91 (0.57, 1.44) 0.94 (0.59, 1.49) 0.91 (0.57, 1.43) 10--14 1.12 (0.65, 1.91) 1.13 (0.66, 1.93) 1.11 (0.65, 1.89) 1.10 (0.65, 1.88) 1.12 (0.66, 1.91) 1.09 (0.64, 1.86) ≥15 1.02 (0.56, 1.85) 1.02 (0.56, 1.86) 1.02 (0.56, 1.85) 1.00 (0.55, 1.81) 1.01 (0.56, 1.84) 1.00 (0.55, 1.82) Number of living children **1.21 (1.06, 1.37)** **1.21 (1.07, 1.38)** **1.20 (1.05, 1.36)** **1.20 (1.06, 1.36)** **1.21 (1.06, 1.37)** **1.19 (1.05, 1.36)** Religion (Christian Protestant = ref)[^e^](#t003fn005){ref-type="table-fn"} Catholic 0.89 (0.57, 1.30) 0.88 (0.56, 1.38) 0.90 (0.57, 1.41) 0.91 (0.58, 1.42) 0.90 (0.58, 1.41) 0.92 (0.59, 1.44) Other 0.88 (0.55, 1.42) 0.87 (0.54, 1.39) 0.90 (0.56, 1.44) 0.90 (0.57, 1.44) 0.89 (0.56, 1.42) 0.92 (0.57, 1.47) Education (years) 1.00 (0.94, 1.07) 1.00 (0.94, 1.07) 0.99 (0.93, 1.06) 0.99 (0.93, 1.06) 1.00 (0.93, 1.07) 0.99 (0.92, 1.06) Partner education (years) 1.05 (0.99, 1.12) 1.05 (0.98, 1.11) 1.05 (0.98,1.11) 1.04 (0.98, 1.11) 1.04 (0.98, 1.10) 1.04 (0.98, 1.10) **Women's empowerment / Gender beliefs /Relationship with partner/pregnancy intentions** Woman makes decisions about spending the cash she earns (women with no cash earned = ref)[^e^](#t003fn005){ref-type="table-fn"} All cash 1.46 (0.91, 2.33) 1.50 (0.93, 2.39) 1.49 (0.93, 2.38) **1.59 (1.00, 2.54)** **1.63 (1.02, 2.60)** **1.62 (1.02, 2.59)** Some cash 1.27 (0.67, 2.40) 1.22 (0.64, 2.31) 1.29 (0.68, 2.44) 1.24 (0.66, 2.32) 1.19 (0.63, 2.23) 1.26 (0.67, 2.36) Participation in household decision-making scale score 0.71 (0.28, 1.79) 0.72 (0.29, 1.79) 0.72 (0.29, 1.81) 0.66 (0.26, 1.63) 0.65 (0.26, 1.63) 0.66 (0.26, 1.65) Inter-spousal communication scale **1.32 (1.05, 1.65)** **1.30 (1.04, 1.63)** **1.30 (1.04, 1.64)** **1.30 (1.04, 1.63)** **1.29 (1.03, 1.61)** **1.29 (1.03, 1.61)** Pregnancy intentions (wants pregnancy in next 12 months at time of survey = ref) Wants to delay pregnancy for ≥ 1 year 0.91 (0.57, 1.47) 0.93 (0.58, 1.50) 0.90 (0.56, 1.45) 0.93 (0.58, 1.49) 0.95 (0.59, 1.52) 0.91 (0.57, 1.47) Wants no more children 0.98 (0.57, 1.67) 0.98 (0.57, 1.69) 0.96 (0.56, 1.66) 1.01 (0.59, 1.72) 1.01 (0.59, 1.72) 0.99 (0.58, 1.69) Has not thought about/not sure/not applicable 0.64 (0.37, 1.13) 0.62 (0.35, 1.09) 0.62 (0.35, 1.09) 0.66 (0.38, 1.15) 0.63 (0.36, 1.11) 0.63 (0.36, 1.10) **Family planning knowledge/attitudes/subjective norms self-efficacy** Family planning beliefs score 0.99 (0.90, 1.10) 0.99 (0.89, 1.09) 0.99 (0.90, 1.09) 0.98 (0.89, 1.09) 0.98 (0.89, 1.08) 0.98 (0.89, 1.08) \# methods known 1.10 (0.95, 1.26) 1.11 (0.96, 1.27) 1.09 (0.95, 1.26) 1.05 (0.91, 1.20) 1.06 (0.92, 1.21) 1.04 (0.91, 1.20) Woman can use family planning without partner's permission (no = ref) *1*.*48 (0*.*98*, *2*.*21)* *1*.*50 (0*.*99*, *2*.*25)* *1*.*50 (0*.*99*, *2*.*25)* *1*.*45 (0*.*97*, *2*.*18)* *1*.*48 (0*.*99*, *2*.*21)* *1*.*47 (0*.*98*, *2*.*21)* Woman can suggest use of condoms (disagree/unsure = ref) 0.96 (0.56, 1.65) 0.96 (0.56, 1.65) 1.00 (0.58, 1.72) 0.98 (0.57, 1.68) 0.97 (0.57, 1.67) 1.02 (0.59, 1.74) Score on family planning use approval questions 0.99 (0.89, 1.09) 0.99 (0.89, 1.09) 0.99 (0.89, 1.09) 0.99 (0.89, 1.09) 0.99 (0.89, 1.09) 0.99 (0.89, 1.09) Family planning self-efficacy scale score **1.39 (1.12, 1.73)** **1.40 (1.12, 1.74)** **1.36 (1.09, 1.69)** **1.39 (1.12, 1.72)** **1.39 (1.12, 1.72)** **1.35 (1.09, 1.68)** **Exposure to intervention** Exposure to intervention (no = ref) **1.62 (1.08, 2.41)** **1.60 (1.07, 2.38)** Number of topics exposed to during intervention **1.15 (1.02, 1.30)** **1.16 (1.02, 1.31)** Exposure to FP discussions during intervention (no = ref) **1.78 (1.20, 2.65)** **1.78 (1.20, 2.63)** ^a^ Models adjusted for all factors shown; Model I adjusted for exposure to the intervention (yes/no); Model II adjusted for the number of topics of exposure; Model III adjusted for exposure to FP discussions ^b^OR: Odds ratio: statistically significant ORs at p\<0.05 are shown in bold; statistically significant ORs at p\<0.10 are shown in italic. ^c^CI: Confidence Interval ^d^Difference between age of male and female partners ^e^Category for missing data indicator included in the model. 10.1371/journal.pone.0153907.t004 ###### Key determinants of FP use among men at endline. {#pone.0153907.t004g} Characteristics Current use any method Current use modern method -------------------------------------------------------------------------------------------------------------------------- ----------------------------- ----------------------------- ----------------------------- ----------------------- ----------------------- ----------------------- **Socio-demographic** Age (years; \<25 = ref) 25--29 0.56 (0.22, 1.41) 0.55 (0.22, 1.39) 0.56 (0.22, 1.42) 0.64 (0.26, 1.59) 0.63 (0.25, 1.57) 0.64 (0.26, 1.60) 30--34 0.87 (0.32, 2.34) 0.85 (0.31, 2.29) 0.88 (0.33, 2.36) 0.99 (0.37, 2.64) 0.97 (0.36, 2.59) 1.00 (0.37, 2.65) ≥35 0.82 (0.26, 2.63) 0.81 (0.26, 2.58) 0.83 (0.26, 2.65) 0.92 (0.29, 2.90) 0.92 (0.29, 2.87) 0.93 (0.30, 2.93) Age difference between partners (years; \<5 = ref)[^d^](#t004fn004){ref-type="table-fn"} 5--9 **2.72 (1.39, 5.32)** **2.79 (1.42, 5.49)** **2.71 (1.38, 4.31)** **2.62 (1.35, 4.10)** **2.68 (1.37, 4.22)** **2.62 (1.35, 4.09)** 10--14 **2.82 (1.06, 5.53)** **2.81 (1.05, 6.51)** **2.82 (1.05, 4.52)** 2.15 (0.82, 4.49) 2.13 (0.82, 4.57) 2.14 (0.82, 4.49) ≥15 2.32 (0.53, 6.04) 2.24 (0.52, 5.68) 2.21 (0.51, 5.51) 2.45 (0.57, 5.98) 2.38 (0.56, 5.90) 2.37 (0.55, 5.87) Number of living children 0.99 (0.82, 1.20) 0.99 (0.82, 1.20) 0.99 (0.82, 1.20) 1.00 (0.82, 1.20) 1.00 (0.82, 1.20) 1.00 (0.82, 1.21) Religion (Christian Protestant = ref)[^e^](#t004fn005){ref-type="table-fn"} Catholic 1.29 (0.66, 2.50) 1.29 (0.67, 2.51) 1.28 (0.66, 2.47) 1.27 (0.66, 2.45) 1.27 (0.66, 2.46) 1.25 (0.65, 2.41) Other 1.79 (0.80, 3.99) 1.88 (0.83, 3.22) 1.72 (0.80, 3.58) 1.49 (0.67, 3.30) 1.55 (0.70, 3.47) 1.48 (0.67, 3.27) Education (years) 1.04 (0.94, 1.16) 1.04 (0.94, 1.16) 1.04 (0.94, 1.16) 1.03 (0.92, 1.15) 1.03 (0.92, 1.15) 1.03 (0.92, 1.15) Partner education (years) *1*.*12 (0*.*98*, *1*.*29)* *1*.*13 (0*.*98*, *1*.*29)* *1*.*12 (0*.*98*, *1*.*28)* 1.12 (0.97, 1.28) 1.12 (0.98, 1.29) 1.11 (0.97, 1.28) Man works outside home for cash (no = ref) 1.64 (0.86, 3.13) 1.65 (0.86, 3.16) 1.65 (0.87, 3.15) 1.68 (0.89, 3.19) 1.69 (0.89, 3.21) 1.69 (0.89, 3.20) **Gender beliefs and roles /relationship with partner/pregnancy intentions** Woman makes decisions re spending cash she earns (women with no cash earned = ref)[^e^](#t004fn005){ref-type="table-fn"} All cash 1.85 (0.74, 4.34) 1.80 (0.72, 3.51) 1.83 (0.73, 3.59) 1.91 (0.77, 3.76) 1.87 (0.75, 3.33) 0.90 (0.76, 3.72) Some cash 1.02 (0.36, 2.87) 0.69 (0.27, 1.82) 0.72 (0.28, 1.89) 0.55 (0.21, 1.41) 0.54 (0.21, 1.40) 0.56 (0.22, 1.45) Male dominance scale score **0.55 (0.31, 0.98)** **0.52 (0.29, 0.94)** **0.54 (0.30, 0.97)** **0.57 (0.32, 1.00)** **0.55 (0.31, 0.98)** **0.57 (0.32, 1.00)** Participation in household decision-making scale score 1.69 (0.45, 3.38) 1.69 (0.45, 4.38) 1.71 (0.45, 4.04) 1.77 (0.47, 4.15 1.76 (0.47, 4.13) 1.78 (0.47, 4.18) Inter-spousal communication scale 1.08 (0.74, 1.59) 1.06 (0.72, 1.56) 1.08 (0.74, 1.59) 1.08 (0.74, 1.59) 1.07 (0.73, 1.56) 1.09 (0.74, 1.59) Pregnancy intentions (wants pregnancy in next 12 months at time of survey = ref) Wants to delay pregnancy for ≥ 1 year 1.63 (0.61, 3.35) 1.70 (0.63, 3.55) 1.58 (0.59, 3.26) 1.78 (0.67, 3.73) 1.85 (0.70, 3.90) 1.75 (0.66, 3.68) Wants no more children **3.19 (1.04, 6.74)** **3.34 (1.08, 7.27)** **3.30 (1.08, 7.10)** 2.58 (0.87, 5.43) 2.66 (0.89, 5.60) 2.64 (0.90, 5.60) Has not thought about/not sure/not applicable 0.97 (0.43, 2.19) 0.97 (0.43, 2.21) 0.96 (0.42, 2.18) 1.06 (0.47, 2.38) 1.06 (0.47, 2.40) 1.06 (0.47, 2.38) **Family planning knowledge/attitudes/subjective norms self-efficacy** Family planning beliefs score **1.16 (1.00, 1.36)** **1.17 (1.00, 1.36)** **1.16 (1.00, 1.36)** 1.13 (0.97, 1.31) 1.13 (0.97, 1.31) 1.12 (0.97, 1.31) \# methods known 1.02 (0.85, 1.23) 1.00 (0.82, 1.21) 1.01 (0.84, 1.22) 1.01 (0.84, 1.22) 0.99 (0.82, 1.20) 1.01 (0.84, 1.21) Woman can use family planning without partner's permission 1.48 (0.60, 3.63) 1.50 (0.61, 3.65) 1.43 (0.59, 3.46) 1.30 (0.52, 3.19) 1.31 (0.54, 3.18) 1.26 (0.52, 3.04) Score on family planning use approval questions **1.53 (1.25, 1.88)** **1.53 (1.25, 1.89)** **1.53 (1.24, 1.87)** **1.45 (1.19, 1.77)** **1.45 (1.19, 1.77)** **1.44 (1.18, 1.76)** Knows where to get family planning (no = ref) 0.34 (0.06, 1.85) 0.34 (0.06, 1.84) 0.34 (0.06, 1.86) 0.15 (0.02, 1.36) 0.15 (0.02, 1.35) 0.15 (0.02, 1.36) **Exposure to intervention** Exposure to intervention (no = ref) 1.26 (0.67, 2.40) 1.22 (0.65, 2.30) Number of topics of exposed to during intervention 1.12 (0.94, 1.33) 1.10 (0.93, 1.30) Exposure to family planning discussions during intervention (no = ref) 1.32 (0.72, 2.40) 1.23 (0.68, 2.23) ^a^ Models adjusted for all factors shown; Model I adjusted for exposure to the intervention (yes/no); Model II adjusted for the number of topics of exposure; Model III adjusted for exposure to FP discussions ^b^OR: Odds ratio: statistically significant ORs at p\<0.05 are shown in bold; statistically significant ORs at p\<0.10 are shown in italic. ^c^CI: Confidence interval ^d^Difference between age of male and female partners ^e^Category for missing data indicator included in the model We also found that women were significantly more likely to report use of any method or of a modern method of family planning if they reported more spousal communication and higher self-efficacy to use family planning. Reported control over their own cash earnings was significantly associated with women's use of modern methods. Beliefs that women can use family planning without their spouses' permission also appears to be positively associated with current use of any or of modern methods, but these associations were not statistically significant (at p\<0.05). Women's younger age and higher number of living children were also significant positive predictors of use of both any and modern family planning methods. As noted, none of the intervention exposure measures was significantly associated with men's reports of current use of any or modern methods of family planning. However, men were significantly more likely to use any or a modern method if they had higher approval scores for use of family planning. Further, for both any and modern methods, men who reported more traditional beliefs about gender roles were less likely to use family planning. Men who reported accurate knowledge about family planning were more likely to use modern methods, but this relationship was not statistically significant. Finally, men who reported that they did not want to have any more children were three times as likely to use any method as those who wanted a pregnancy in the future. Qualitative Results {#sec032} ------------------- ### Description of the Sample {#sec033} Data for this analysis were drawn from ten married couples. All couples were determined to be 'real' marriages or current unions based on the triangulation of key events and stories in the relationship histories from both partners. All couples reported living together in Siaya County during the intervention timeframe. At the time of data collection, couples had been married for between 5 and 32 years. Women were between the ages of 21 and 49 years; four were over the age of 35. Men's ages ranged between 23 and 68 years. Wives were younger than their husbands in all of the relationships, and in six of the 10 couples spouses were within 5 years of age of each other. All couples reported a pregnancy within the first two years of marriage, and at least half of the women reported becoming pregnant prior to marriage. Seven of the ten women interviewed reported having dropped out of school due to pregnancy, lack of school fees, or both. Women reported having between 2 and 5 living children, and one was pregnant at the time of the interview. All respondents reported using at least one method of family planning at some point during their marriage, including calendar, condoms, pills, IUD, injections, implants, and bilateral tubal ligation. Five women reported using a family planning method in secret at some point during their marriage. Two of the relationships were polygamous, and in five relationships the husband was known or suspected to have had sex with someone else during the marriage. There were no reports of wives having other sexual partners than their husbands during the marriage. All couples reported one or both partners having been tested for HIV. In five couples, one or both partners reported incidents of inter-partner violence, all of which involved husbands physically abusing their wives. There was variability in how couples were introduced to the intervention. In some cases, initial exposure occurred by accident (encountered a drama group performing in the community) and in others by invitation (a friend invited them to join a group for couples). Likewise, sometimes the husband participated in dialogues, sometimes the wife participated, and sometimes both husband and wife from the same couple participated in the same dialogues. Men and women from two of the couples interviewed reported actively sharing their stories about family planning during a dialogue convened through the intervention. ### Experience of Participation in the Intervention {#sec034} Several participants discussed attending dialogues about family planning at *barazzas* (community meetings) or events. > *"I was invited to these meetings by a friend*, *and during these meetings there were discussions about role-sharing at home and also \[use of\] family planning*!*"* Men and women reported learning about family planning in community dialogues. One woman recalled her husband's report upon returning from a *barazza*. > *"Today the discussion at the Chief's* barazza *was very heated*. *Imagine*, *we discussed about family chores and family planning*. *I have never known that family planning issues were real*.*"* The intervention offered exposure to discussion of the benefits of family planning: one woman recounted how her peers not only described, but provided 'living proof,' of positive experiences. Both men and women reported learning by contrasting real examples of health and economic stability between family planning users and non-users in the community. Others described feeling comfortable publicly sharing their own family planning stories when they perceived that the social environment was becoming more openly supportive of couples' use of family planning. One man, who had previously opposed family planning, described how learning of others' positive experiences dispelled his beliefs that it caused infertility and birth defects. Dialogues appear to have increased the acceptability of not only talking about, but also using family planning, especially when key opinion leaders (e.g. chiefs and religious leaders) both advocated for and modeled open communication with spouses and use of family planning. One woman described how community members openly supported others to use family planning during the dialogues; she and her husband reported that their ongoing involvement in dialogues about family planning and gender roles influenced how they communicated about these issues at home. ### Shifting Gender Norms {#sec035} Women described shifts towards more equitable household roles, with husbands beginning to help with household duties. Some men also described helping with chores that were traditionally considered 'women's work.' For example, one man spoke of learning the importance of sharing household responsibilities at a dialogue. He cited collecting firewood, going to the *posho* mill, cooking and fetching water as examples of things he now does. One woman relayed a story of how she and the children were shocked when, unprompted, her husband cooked for the family. She recalled her children telling her: > *"Mom*! *Guess what happened in our house today? Dad cooked us lunch!"* Some couples described starting to manage household budgets together and making more joint decisions about household purchases and assets. One woman said that one of the best moments of her marriage was when she and her husband "*made decisions together as man and wife*." Together, they decided to sell the scrap metal her husband had accumulated and bought a cow; when the cow was old enough, they sold it and bought materials for the new house they are now planning to build. One man said: > *"Before*, *what was ours was mine*. *Now*, *what's ours is ours*. *"* Despite such reported changes, both men and women stated that women still did the majority of household work, and that men maintained the majority of household decision-making power. ### Couple Communication {#sec036} Men and women reported more open and equitable communication about family planning. Several women who had previously used a method in secret reported less opposition towards family planning from their partners, and described how they now openly communicated and reached mutual agreement to use a method. One man reported that he reminded his wife to get regular injections and offered to accompany her to the clinic. Another man described extensive communication on selecting a family planning method: the couple jointly decided against oral contraceptives because of the need for daily adherence, and instead chose injectable contraception. Men reported that couple concordance about fertility intentions and desired family size helped them resist external pressure for large families. Previously, some women had experienced conflict about the ideal number of children and family planning use, including questioning and pressure from husbands about getting pregnant. Several men and women reported that improved communication with their spouses both decreased conflict about family planning and increased the "*love and harmony*" in their relationship overall. One woman cited her husband's support for family planning as an enabler for her use of family planning: it boosted her morale and brought harmony to her marriage. Previously she had used family planning without her husband's consent or knowledge, and this had contributed stress and conflict in the relationship. While women described the many benefits of having partner support for using family planning, including increased relationship harmony, they also discussed the critical importance of being able to independently access family planning methods when faced with partner opposition. One woman recalled using family planning without her husbands' permission as a time when she was able to control something about the relationship, and was happy to be free of worry that she would become pregnant. Another recalled that she had learned about family planning methods in the privacy of a health care facility, and started using a method in secret. Family planning use, she said, had helped ensure her children were adequately spaced and that she had children when she had the resources to support them, but she also described experiencing significant stress because her husband continued to pressure her for a baby and was suspicious when she did not get pregnant. She also spoke of the stress of managing side effects on her own. Subsequently, she and her husband came to agreement about the use of family planning, and she has recently switched to a more effective method with her husband's support. However, she stated that she would still be using a method in secret if her husband had not changed his attitude. She emphasized that her previous use of family planning had "*cost her harmony...but it was worth it*" to be able to control her fertility. Discussion {#sec037} ========== Our results suggest that an intervention that supports open and public dialogue about gender, sexuality, and family planning may increase use of family planning among couples in settings where the prevalence of use is relatively low. First, self-reported family planning use increased significantly for both women and men in our study samples between baseline and endline. Reports of family planning use by women at baseline closely matched use estimates for Nyanza Province in the 2008--2009 Kenya DHS \[[@pone.0153907.ref021]\]. In the analysis of endline data, use of family planning by women was significantly associated with participation in the intervention, particularly in specific dialogues on family planning. The intervention's effects were significant for use of both any and modern contraceptive methods and after adjusting for key socio-demographic factors. While similar interventions may have been implemented in similar settings, there are limited evaluation data of such interventions in the published literature. Yet, in line with results from our study, published evaluation data show positive effects of similar community-level interventions in addressing social factors and shifting social norms with regard to reproductive health issues. For example, in 2004, CARE implemented and evaluated a 3-year adolescent reproductive health project to address related behavior changes and local social norms in a rural district of the Republic of Georgia. Community engagement strategies included promoting community support for adolescent reproductive health and using \'Theatre for Development\' to promote community dialogue about social norms. Project evaluation data demonstrated improved knowledge, attitudes, behavior about FP and evidence of shifts in gender norms. \[[@pone.0153907.ref025]\] More recently, Campbell et al. explored pathways between community participation and HIV prevention, treatment and impact mitigation in Zimbabwe, reviewing six qualitative studies in Manicaland \[[@pone.0153907.ref026]\]. These found that community group membership is often associated with decreased HIV incidence and that participation in formal community groups and informal local networks provides opportunities for critical dialogue about HIV/AIDS, often facilitating renegotiation of harmful social norms, sharing of personal experiences of HIV/AIDS, formulation of positive action plans and solidarity to action them. We found that a greater level of spousal communication, and higher self-efficacy to discuss and use family planning, were significant predictors of use for women at endline. As our theory of change suggests, couple communication and self-efficacy are two of three key variables we expected to be positively associated with exposure to the intervention. However, the third key variable---women's greater participation in household decision-making---was not a significant predictor of family planning use. The comprehensive household decision-making measure we used assessed decisions across a wide range of domains. While it may provide a more robust measure of women's overall decision-making power in the household, it is not narrowly focused on decisions about family size, contraceptive use, or healthcare seeking, and therefore may be less predictive of such outcomes; future studies could compare various proposed measures of household decision-making power to compare their predictive power on a range of maternal health outcomes, including family planning. Moreover, it may be that self-efficacy to use family planning is a better measure of women's power to enact these behaviors, and may have accounted for most of the variance in this domain. Reported control over their own cash earnings was also significantly associated with use of modern methods, suggesting that economic empowerment may be associated with access to and/or use of FP. Finally, greater age and a higher number of children were significantly associated with women's family planning use, yet women's reproductive intentions were not a significant predictor of use. For men, approval of family planning, more equitable (less traditional) beliefs about gender roles, and a desire to not have more children were the strongest predictors of family planning use. Neither the scale measuring women's decision-making power in the household, nor any of the intervention exposure variables, were significant predictors of men's current use of any or of modern methods. Of note, the two areas with the most striking differences in men's and women's responses at endline were reproductive intentions (60.0% of men and 30.2% of women wanted another pregnancy at the time of the interview or within the next 12 months) and approval of women's use of family planning without her partners' permission (66.3% of women and only 11.6% of men approved of such use). Given these differences between men and women, an intervention that supports more open communication about reproductive decision-making and family planning seems appropriate. Our findings suggest that improved communication may have enabled more couple acceptance of family planning, and thus increased use of family planning over time. In addition, men's desire not to have any more children was a strong predictor of family planning use among men, whereas women's reproductive intentions were not a predictor of family planning use among women even though women were much more likely to want to delay or limit childbearing than were men. Rather than indicate that men's childbearing desires are more important to a couple's use of family planning use than women's, these results might suggest that when men's intentions are more closely aligned with women's, the couple is more likely to use family planning. Our qualitative results provide additional insight into the ways in which the intervention may have contributed to changes in couple communication and family planning behavior--the mixed-method approach we used is one of the strengths of our analysis. Public dialogues may have increased the perceived social acceptability of family planning and of the benefits of open couple communication about family planning, and these normative shifts at community level may have enabled more equitable communication and decision-making at the couple level. Ongoing, public discourse---often explicitly supported or led by community leaders---may have normalized discussions about family planning and influenced family planning acceptability at the community level. The participation of prominent male opinion leaders in these dialogues may have helped to legitimize men's participation in communication and decision-making about family planning, and may have positively influenced their approval of family planning, thus increasing use. Our quantitative results support this: men's approval of family was an important predictor of family planning use. Public discourse precipitated conversations on family planning in the household as couples discussed what they heard and learned. Men described a change in their acceptance of family planning, and women reported that men began to demonstrate more shared decision-making, better communication, and increased support for family planning. Male support of and male involvement in family planning also appear to be important factors for helping with method adherence. Women reported partners accompanying them to get methods and reminding them to take their pills daily. Increased couple communication may also have contributed to increased harmony and a decrease in conflict. Relationship harmony was highly valued by both men and women. Limitations {#sec038} =========== Our survey data are cross-sectional without a comparison group and we used a two-stage cluster sampling approach, thus the sample obtained does not cover the population as evenly as in the case of simple random sampling. Both our quantitative and qualitative results are subject to self-report bias. Propensity scores were used to reduce selection bias by equating the baseline and endline survey samples based on key measured covariates, yet propensity score matching only accounts for observed covariates: imbalances may remain even after propensity score adjustment if relevant survey respondent characteristics were not measured or were measured imprecisely. Thus, the larger estimated differences in baseline-endline family planning use based on unmatched results may be real. Importantly, self-selection to participation in the intervention is a limitation of our analyses examining associations between exposure to the intervention and family planning use at endline. Yet, some reassurance is provided since the intervention was widely advertised and supported by community leaders, and since 65% of the respondents were exposed. In addition, we cannot account for potential effects of other sources of family planning information (e.g. health providers, HIV/STI testing and counseling, domestic violence counseling, media) to the observed differences in family planning use between baseline and endline surveys and the associations observed at endline. However, to our knowledge, no other similar interventions took place in the study area during the intervention period. Also, the intensity of the intervention was measured as the number of topics covered during intervention activities attended and not as the amount of time of exposure to the intervention. Our qualitative interview guide elicited more reports on the positive than the negative aspects of the intervention and family planning use; this may be the result of social desirability bias or of the nature of our intervention activities (i.e. involvement of community leaders, theatre groups, other forms of entertainment). Finally, we do not have information on women who refused to be interviewed or completed only part of the questionnaire. Conclusion {#sec039} ========== Our results suggest that an intervention that encourages and supports dialogue and communication about gender norms and sexuality can shift gender relations and positively influence family planning use, especially for women. Participation in the intervention and greater spousal communication, in addition to variables indicative of more empowerment (self-efficacy for family planning and control over household assets), all contributed to greater use of family planning for women. For men, approval of family planning, more equitable beliefs about gender roles, and a desire to not have more children were the strongest predictors of family planning use. These findings are consistent with other research that suggests more equitable gender norms and spousal communication are linked to contraceptive use and positive reproductive health outcomes, and provide an effective intervention model for achieving those goals through community-based dialogues. While not definitive, our promising evaluation results now allow for smaller scale studies to examine the impact of specific types of dialogues and experience sharing on specific outcomes. Future studies should consider the local context and, if possible, use a similar mixed-method approach. Supporting Information {#sec040} ====================== ###### WE MEASR Scales and Indices. (DOC) ###### Click here for additional data file. ###### Family Planning Measures. (DOC) ###### Click here for additional data file. ###### Men's Data. (XLS) ###### Click here for additional data file. ###### Women's Data. (XLSX) ###### Click here for additional data file. ###### Codebook for Women's and Men's Datasets. (DOC) ###### Click here for additional data file. We would like to recognize all the CARE staff who contributed to the successful implementation of the FPRI project, especially Luis Ortiz-Echevarria, Marcie Rubardt and Mary Yetter of CARE USA, and Rosemary Mbaluka, Ferdinand Moseh, Rhodah Litoroh, Cynthia Muhambe, Henry Anyona, Mary Nyakomitta and Jude Otogo of CARE Kenya. We would also like to acknowledge Katina Pappas-DeLuca and Colleen Smith for their contributions to the endline program evaluation. Our special thanks go to the team of community-based facilitators who partnered with us throughout the project, and to the District Health Medical Health Team in Siaya for their ongoing support and coordination. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: CW CG. Performed the experiments: CW EW. Analyzed the data: CW AC CG. Contributed reagents/materials/analysis tools: CW CG EW. Wrote the paper: CW AC CG EW.

Background ========== The prevalence of type 2 diabetes rapidly rose over the past three decades. However its burden is not homogeneously distributed: racial and ethnic minorities usually experience higher prevalence than their non-minority counterparts \[[@B1]\] and are at higher risk of developing diabetes-related complications such as blindness, kidney damage, or depression, impacting both quality of life and mortality rates \[[@B2],[@B3]\]. Self-management, defined as the patient's ability to manage not only the symptoms inherent to a chronic condition but also its treatment and associated lifestyle changes \[[@B4]\], has become increasingly important in the treatment of type 2 diabetes \[[@B5],[@B6]\]. However, because adequate diabetes self-management (DSM) may require considerable lifestyle changes to several domains, namely having a healthy diet, exercising, or glucose monitoring, not all the patients are able to properly follow the self-management plans agreed with their healthcare professionals or advised by clinical guidelines. Racial and ethnic minorities are less likely to engage in DSM behaviors than other population groups, which partially explains observed disparities in health outcomes \[[@B7]\]. Some of the barriers faced by these groups for achieving an adequate DSM are related to characteristics of the groups (such as health literacy or health beliefs) and also of the healthcare system (namely accessibility of culturally sensitive information) \[[@B8],[@B9]\]. Several review studies have assessed the effect of DSM educational programs on the general population \[[@B6],[@B10]-[@B17]\]. Those studies have established that DSM educational programs can improve glycemic control \[[@B11]-[@B16]\], and identified the key characteristics for improving glycemic control, including face-to-face delivery, teaching methods based on cognitive reframing \[[@B11]\], and higher contact time between participant and educator \[[@B16]\]. A smaller number of studies have reviewed the evidence of the effect of these programs on racial/ethnic minorities \[[@B18]-[@B20]\]. Only one study \[[@B18]\] examined their impact on glycemic control, observing a reduction on glycated hemoglobin of 0.32%. There were however some limitations underlying that study, as some of the included interventions combined educational programs with other types of quality improvement strategies, making it difficult to disentangle the individual effect of the educational components. More importantly, no previous meta-regression study has identified the key common characteristics of successful educational programs targeted to racial/ethnic minority groups. This represents a considerable gap in the literature, as programs for racial/ethnic minority groups should have different components from those targeting other population groups, namely addressing the cultural idiosyncrasies associated with each group. Therefore it is not reasonable to assume that the same type of successful program will be equally successful when applied to racial/ethnic minority groups. Additionally, in the past years several clinical trials examining the impact of educational programs to improve diabetes self-management on racial/ethnic minorities have been published, making now possible a more detailed review and analysis of the available evidence regarding the effectiveness of these interventions. In this work we systematically reviewed DSM educational interventions specifically targeted to racial/ethnic minority groups. We studied the characteristics and costs of the interventions, and analyzed their impact on diabetes knowledge, self-management behaviors, and clinical outcomes. Whenever data were available, we performed meta-analyses to examine the short and long-term effects of the interventions, and meta-regressions to identify common characteristics of the interventions associated with better results. Methods ======= The review and its procedures were planned, conducted, and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines \[[@B21]\]. Data sources and searches ------------------------- A comprehensive core search strategy was developed for Medline through Ovid (combining MeSH terms and keywords) and then adapted and implemented in EMBASE and CINAHL (search strategy available in Additional file [1](#S1){ref-type="supplementary-material"}: Table S1). Gray literature and additional articles were searched in twelve more bibliographic sources (Additional file [2](#S2){ref-type="supplementary-material"}: Table S2). The search was not restricted by language or publication date. For all the references selected to be included in the review, backward and forward citation searches were performed in ISI Web of Knowledge. All searches were conducted in October 2012. A bibliographical database was created using EndNote X6, and used to store and manage the retrieved references. Study selection --------------- We included studies analyzing the effectiveness of DSM educational programs targeted to racial/ethnic minority groups with type 2 diabetes. We only included those studies in which at least 90% of the participants pertained to a racial/ethnic minority group considered to be at a higher risk for diabetes complications than the majority population group. Racial/ethnic minority group was defined as a population group with a race or ethnicity different from that of the majority population group of the host country. Groups at higher risk of diabetes complications were identified based on available literature. Interventions had to be exclusively educational, without including any other component such us financial incentives, clinician education or case management. In order to avoid possible comparisons between programs carried out in very heterogeneous settings, with very different health systems and population needs, we restricted this review to those interventions conducted in countries that were members of the OECD \[[@B22]\], when study selection was conducted (November 2012). Eligible designs were randomized controlled trials (RCTs), including cluster randomized controlled trials; controlled trials, including quasi-randomized trials; controlled before-after studies; and non-controlled before-after studies. Studies including a control group were only eligible in case the intervention was compared with care as usual. Titles and abstracts were screened for eligibility, and those fulfilling the inclusion criteria were included in the next stage, where the full texts of the selected articles were retrieved and assessed. Those that met the inclusion criteria were included for data extraction. Two reviewers independently screened citations, and any disagreements were solved by consensus with a third reviewer. Data extraction and quality assessment -------------------------------------- We designed and used structured forms to extract information of interventions' characteristics and their effectiveness. We used a previously developed taxonomy of DSM educational programs to characterize the interventions \[[@B23]\]. The following information was extracted: setting, ethnic group, administration formats, teaching methods, educational contents, educators' background, use of peer educators, and duration. Information of interventions' cost and cost-effectiveness was also extracted. We critically appraised the studies using the Quality Assessment Tool for Quantitative Studies \[[@B24]\], which enables the assessment of both internal and external validity, classifying them into three categories (good, fair or poor) depending on six aspects: selection bias, study design, confounders, blinding, data collection and withdrawals/ dropouts. Two reviewers independently extracted all the information and critically appraised the studies. Disagreements were solved through discussion with a third reviewer. When necessary we contacted the authors of the studies to request additional information. Data synthesis and analysis --------------------------- The effectiveness of the interventions was assessed in terms of their impact on 1) diabetes knowledge, 2) diabetes self-management behavior, and 3) clinical outcomes. Diabetes knowledge was ascertained by measures reflecting the theoretical knowledge the patients had about their condition. Diabetes self-management behavior measured the performance of specific activities related to adequate DSM (diet, exercise, glucose control, foot self-examination, etc.). Clinical outcomes included hemoglobin A1c (HbA1c), body mass index (BMI), or blood pressure, amongst others. All outcomes in all the studies were examined and classified as measuring one of these three domains. Variables that measured other domains were not included in the analysis. Additionally, we conducted independent meta-analyses to analyze short and long-term (six months post-intervention) effect of the interventions on glycemic control. Eligibility criteria for the meta-analyses included randomized controlled trials comparing the interventions with usual care. The mean (standard deviation) of HbA1c levels in each study were extracted. This information was transformed into weighted mean difference and 95% confidence intervals (CI) were calculated and combined using random-effects models. We imputed unreported standard deviations by use of established methods \[[@B25]\]. Heterogeneity was quantified by the I^2^ statistic, where I^2^ ≥ 50% was considered evidence of substantial heterogeneity \[[@B26]\]. Sources of heterogeneity were investigated by a Galbraith plot. Publication bias was quantitatively assessed with Egger test. We used bivariate meta-regression to explore relationships between effect size (ES) and interventions characteristics. The number of included studies was insufficient to perform a multivariate regression analysis. We conducted a sensitivity analysis, excluding the studies with higher risk of bias. All analyses were conducted with Stata, version 12.0 (StataCorp, College Station, Texas). For all the analyses, statistical significance was accepted at *p* \< 0.05. Results ======= Article identification ---------------------- Figure  [1](#F1){ref-type="fig"} reports the screening process. A total of 1,988 unique references were retrieved. 1,386 references were excluded based on title and abstract, resulting in 602 references being included in the next stage. Full text articles were retrieved and assessed, with 24 studies meeting the eligibility criteria. Backward and forward search of these 24 articles identified thirteen additional studies, resulting in thirty seven articles being included in the review \[[@B27]-[@B63]\]. Thirty five of them reported one single intervention, whereas two articles reported two distinct interventions per article. Overall, thirty seven articles were identified, which analyzed the effectiveness of thirty nine different interventions. {#F1} Characteristics of the studies and interventions ------------------------------------------------ Table  [1](#T1){ref-type="table"} shows the aggregated characteristics of the interventions and of the studies used to assess their effectiveness. Most of the studies identified were conducted in the US (92%), and published from 2001 onwards (84%). Almost two thirds of the studies were RCTs (73%), and approximately two thirds presented moderate or low risk of bias (65%). ###### Characteristics of the studies and interventions   **Number (N)\*** **Percentage (%)** ----------------------------------------------------------- ------------------------------ -------------------- ---- **Characteristics of the studies (n = 37)**     Country where the intervention took place       United States 34 92   United Kingdom 2 5   Netherland 1 3 Design       Randomized controlled trial 27 73   Quasi-experimental 10 27 Years of publication       Before 2001 6 16   2001-2006 16 43   2007-2012 15 41 Risk of bias       Low 8 22   Moderate 16 43   High 13 35 Number of participants 159 (117)† 7-529‡ **Characteristics of the interventions (n = 39)**     Ethnicity       African American 13 33   Latinos 16 41   Asiatic 2 5   Alaskan-Eskimos 1 2   Multiethnic group 7 18 Place where intervention took place       General Practice 16 44   Community Center 12 33   Home 3 8   Hospital 1 3   Other 4 11 Setting       One-on-one 13 33   Group 17 44   Both 9 23 Family invited to take part in the intervention       Yes 12 31   No 27 69 Delivery       Face-to-face 27 69   Telecommunication 5 13   Both 7 18 Number of teaching methods       1 7 18   2 24 62   3 or more 8 26 Teaching method\*       Didactic 32 82   Goal-setting, dictated 8 21   Goal-setting, negotiated 10 26   Situational problem solving 23 59   Cognitive reframing 7 18   Other 2 5 Number of educational contents       One 7 18   Two 5 13   Three 5 13   Four of more 22 56 Educational Content       Diet 33 85   Exercise 24 65   Self-monitored blood glucose 23 59   Basic diabetes knowledge 21 54   Medication adherence 11 28   Psycho-social 19 49   Other 11 28 Number of educators       One 18 46   Two 14 36   Three or more 4 10 Educators       Nurse 15 39   Dietitian 21 54   Psychologist 2 5   Physician 2 5   Research team or staff 4 10   Other 16 41   Unclear/unspecified 2 5 Peer provider       Yes 12 31   No 26 67 Culturally adapted       Yes 33 85   No 4 10   Unclear/unspecified 2 5 Length of the interventions (months) 8.2 (8.4)† 0.25-48‡ Length of the interventions (number of sessions) 13.1 (13.6)† 1-52‡ Length of the sessions (minutes) 90 (52.9)† 14-240‡ Length of the interventions (total hours of intervention) 23.3 (36.7)† 0.25-180‡ Intensity (total hours of intervention per month) 4.9 (6.9)† 0.2-36‡ \*N is not equal to 37 for all variables, as some characteristics were not reported for all the interventions. †mean (sd); ‡min-max. In relation to the characteristics of the interventions, most of them targeted African American (33%) or Latinos (41%) and took place either in General Practices (41%) or in Community Centers (31%). Face to face was the most common format of delivery (70% of the interventions). Almost half of the interventions followed a group format (44%), whereas one third was delivered individually. Patients were encouraged to bring their relatives in nearly a third of the interventions (31%). On average programs lasted eight months and included 13 sessions, with each session lasting 90 minutes. Most of the programs included multiple teaching methods and multiple contents. Approximately half of the programs were delivered by multidisciplinary educators (54%). Effectiveness of the interventions ---------------------------------- Additional file [3](#S3){ref-type="supplementary-material"}\'; Table S3 shows the characteristics of each study and their impact on diabetes knowledge, self-management behaviors and clinical outcomes. Fifteen studies analyzed the impact of the interventions on diabetes knowledge \[[@B27],[@B29],[@B32],[@B34],[@B37]-[@B39],[@B44],[@B47],[@B48],[@B50],[@B53],[@B56],[@B60],[@B62]\]. In the majority of studies (nine out of fifteen), diabetes knowledge was only measured immediately after the intervention program was finished. Different types of instruments were used to measure outcomes such as diabetes knowledge and its complications or how to conduct adequate diabetes self-management. Eleven of these studies observed that the interventions significantly improved patients' knowledge, whereas the remaining four did not observe significant effects. Health beliefs were additionally analyzed in two studies, without observing a significant improvement. Twenty studies examined the potential of the interventions to improve self-management behaviors \[[@B27],[@B28],[@B30]-[@B32],[@B34]-[@B36],[@B43]-[@B45],[@B50],[@B54],[@B55],[@B57]-[@B62]\]. Behavioral outcomes were heterogeneous, being related in most instances to dietary or physical activity behaviors, but also to behaviors related to blood glucose testing, foot care, or medication adherence. Fifteen out of twenty studies (75%) observed that the interventions produced improvements in behavioral outcomes. Interventions were more successful in improving dietary behaviors than in promoting physical activity or medication adherence. Thirty-one studies assessed the impact of the interventions on clinical outcomes \[[@B27],[@B29]-[@B33],[@B37]-[@B44],[@B46]-[@B60],[@B62],[@B63]\]. The most frequent clinical outcome was HbA1c, but blood pressure, fasting blood glucose or BMI were also included in a substantial number of studies. Twenty two studies (71%) observed that the educational programs produced statistically significant improvements in clinical outcomes. Educational programs more frequently improved fasting blood glucose, HbA1c and blood pressure (improved in 71%, 59%, and 57% of the studies, respectively) than other clinical outcomes such as lipid profile (40%), weight/BMI (28%) or waist circumference (25%). Costs were reported in only two studies \[[@B33],[@B40]\]. The cost per patient per year was approximately \$280 in the intervention developed by Banister et al. \[[@B33]\] and \$461 in the intervention by Culica et al. \[[@B40]\]. No study included a formal cost-effectiveness analysis. Effectiveness of the intervention on glycated hemoglobin -------------------------------------------------------- Twenty-one interventions employed HbA1c measures and were included in an initial meta-analysis that assessed possible baseline HbA1c differences between intervention and control groups. No statistically significant differences were found (HbA1c mean difference = -0.02% \[95% CI -0.22% to 0.18%\]). A second meta-analysis was conducted to estimate the pooled difference in HbA1c between the intervention and control group immediately after the intervention was completed, observing a significant reduction in the overall HbA1c of -0.47% (95% CI -0.76% to -0.17%). Although heterogeneity was high (I^2^ = 66.3%), it was mainly associated with one intervention \[[@B55]\], and once that intervention was excluded, heterogeneity was reduced to 0%. Twenty interventions were therefore included in the final meta-analysis, reporting on 3,094 patients (1,551 in the intervention and 1,543 in the control group). The combined effect of the intervention produced a significant reduction in the overall HbA1c of -0.31% (95% CI -0.48% to -0.14%) (Figure  [2](#F2){ref-type="fig"}). Egger test indicated the absence of publication bias (*p* = 0.22). One of the studies included in the meta-analysis presented high risk of bias \[[@B44]\]. We conducted a sensitivity analysis excluding it, obtaining very similar results. {#F2} Only three studies measured HbA1c at six months post-intervention \[[@B30],[@B45],[@B52]\]. A meta-analysis of these three studies observed a reduction in pooled HbA1c of -0.47%, although no significant differences were observed (*p* = 0.13). Intervention characteristics associated with treatment effects -------------------------------------------------------------- We conducted bivariate meta-regressions of the 20 RCTs included in our meta-analysis in order to identify the characteristics associated with increased short-term HbA1c reduction (results reported in Table  [2](#T2){ref-type="table"}). Interventions delivered face to face obtained better results than those interventions supported by telecommunication. Also, those delivered individually performed better than those delivered in a group format. As for the teaching methods, both interventions that employed cognitive reframing techniques, as well as those including only one teaching method, were associated with better outcomes. Finally, those interventions that included at least one peer educator produced significantly better effects than those not including any peer educator. ###### Meta-regression of the effect of intervention´s characteristics on pooled glycated hemoglobin (N = 20)   **Number interventions**^**a**^ **SMD** **95% CI** **Residual I**^**2**^ ---------------------------------------------- ----------------------------------------------- ---------------------------------------------- ------------- ----------------------- ---------------- -------------- --- Country       0.00%   US 18 -0.28\*\* -0.47; -0.09     Netherland 1 -0.82 -1.85; 0.21     UK 1 -0.06 -0.88; 0.77   Year       1.74%   Prior 2001 2 -0.51 -1.29; 0.26     2001-2006 6 0.32 -0.52; 1.15     2007-2012 12 0.18 -0.60; 0.97   Setting       0.00%     Primary Care 9 -0.23 -0.58; 0.11       Community center 6 -0.03 -0.48; 0.41       Other 5 -0.24 -0.74; 0.26   Mean HbA1c at baseline in intervention group       0.08%     HbA1c \< 9% 10 -0.39\*\* -0.96; 0.17       HbA1c ≥ 9% 10 0.06 -0.32; 0.43   Target population       3.33%   Ethnic minority 10 -0.30 -0.58; -0.04     Rural ethnic minority 3 0.31 -0.41; 1.03     Women from ethnic minority group 2 -0.23 -1.04; 0.58     Other (elderly or low income or low literacy) 5 -0.05 -0.47; 0.36   Ethnic minority group       0.00%     African-American 7 -0.10 -0.48; 0.28       Latinos 7 -0.31 -0.80; 0.17       Asiatic 2 -0.54 -1.28; 0.20       Multi-ethnic 4 -0.15 -0.66; 0.35   Individual Vs Group delivered       0.00%     Individual 7 -0.45 -0.75; -0.14       Group 7 0.13 -0.29; 0.56       Both 4 0.32 -0.14; 0.79   Patients accompanied by family       0.00%     No 5 -0.23\*\* -0.44; -0.04       Yes 15 -0.38 -0.84; 0.07   Face to face Vs. telecommunication       0.00%     Exclusively face to face 12 -0.37\*\* -0.62; -0.12       Exclusively telecommunication 3 -0.08 -0.53; 0.37       Combining face to face and telecommunication 5 0.33 -0.11; 0.77   Teaching methods       0.00%     Didactic       0.39%       Yes 14 -0.30\*\* -0.53; -0.07         No 6 -0.03 -0.42; 0.35       Goal setting negotiated       0.00%       Yes 8 -0.29\*\* -0.57; -0.01         No 12 -0.03 -0.41; 0.34       Goal setting dictated       0.00%       Yes 7 -0.26 -0.55; 0.02         No 13 -0.09 -0.48; 0.29       Situational problem solving       0.00%       Yes 10 -0.28\*\* -0.50; -0.06         No 10 -0.08 -0.47; 0.30       Cognitive reframing       0.00%       Yes 4 -0.47\*\* -0.91; -0.03         No 16 0.20 -0.29; 0.68       Number of teaching methods used       0.00%       1 4 -0.58\*\* -1.04; -0.12         2 10 0.37 -0.16; 0.92         3 or more 4 0.27 -0.27; 0.80   Content             Diet       0.00%       Yes 18 -0.35\* -0.54; -0.15         No 1 -0.74 -2.00; 0.50       Exercise       0.00%       Yes 15 -0.33\* -0.54; -0.12         No 4 -0.20 -0.68; 0.28       Blood glucose self-monitoring       0.00%       Yes 13 -0.26\* -0.48; -0.03         No 6 -0.39 -0.81; 0.04       Basic diabetes knowledge       0.00%       Yes 8 -0.28\* -0.53; -0.02         No 11 -0.21 -0.60; 0.17       Medication adherence       0.00%       Yes 9 -0.25\* -0.49; -0.05         No 10 -0.19 -0.58; 0.20       Psycho-social       0.00%       Yes 11 -0.18\* -0.40; 0.03         No 9 -0.38 -0.76; 0.01     Number of contents       5.17%     1 or 2 4 -0.36 -0.84; 0.12       3 or 4 9 0.02 -0.55; 0.58       5 or more 7 0.84 -0.48; 0.65   Educators             Nurse       2.59%       Yes 5 -0.02 -0.43; 0.39         No 14 -0.29\* -0.52; -0.06       Dietician       0.77%       Yes 11 -0.10 -0.27; 0.47         No 8 -0.35\* -0.62; -0.08       Psychologist       2.47%       Yes 1 0.05 -0.65; 0.76         No 18 -0.30\* -0.50; -0.11       Physician       0.00%       Yes 2 0.21 -0.29; 0.72         No 17 -0.33\* -0.54; -0.13       Research team       2.13%       Yes 2 -0.10 -0.77; 0.58         No 17 -0.29\* -0.49; -0.09   Number of types of educators       0.00%     One 11 -0.30\* -0.55; -0.04       Two 6 -0.41 -0.90; 0.07       Three or more 2 0.17 -0.34; 0.70   Peer provider       0.00%     Yes 7 -0.54 -0.93; -0.15\*       No 13 0.29 -0.15; 0.74   Duration of the intervention (months) 8.4 (1.4)† -0.02 -0.05; 0.02   Number of sessions 12.1 (2.1)† -0.01 -0.02; 0.01 0.00% Average duration of each session (hours) 1.5 (0.2)† -0.01 -0.25; 0.23 0.00% Total hours of intervention 21.9 (7.0)† -0.01 -0.01; 0.01 0.00% Intensity (number of hours/month) 4.6 (1.3)† -0.01 -0.09; 0.07 1.60% SMD = standardized mean difference; I^2^ = Variation in standardized mean difference attributable to heterogeneity; 95% CI = 95% confidence interval; HbA1c = glycated hemoglobin. ^a^N is not equal to 20 for all variables, as some characteristics were not reported for all the interventions. \**p* \< 0.05; \*\**p* \< 0.001, † = mean (SE). No statistically significant differences were observed for the total duration of the intervention, the number of sessions included, the duration of each session, the total number of hours of intervention or its intensity (number of hours per month). Discussion ========== In this systematic review we identified and characterized 39 DSM educational programs specifically targeted to racial/ethnic minority groups. Most programs produced some benefits over care as usual in improving diabetes knowledge, self-management behaviors, and clinical outcomes. Furthermore, meta-analyses indicated that these interventions decreased HbA1c, which was significant both from statistical and clinical perspectives. Larger reductions in HbA1c were observed in those interventions delivered individually and face to face, involving peer educators, based on cognitive reframing techniques, and employing a lower number of teaching methods. Long-term effects and cost-effectiveness were rarely assessed. The estimated 0.31% reduction in HbA1c observed in our meta-analysis is modest but clinically significant, as evidence suggests that every percentage point decrease in HbA1c over 10 years is associated with a risk reduction of 21% for deaths related to diabetes, 14% for myocardial infarctions, and 37% for microvascular complications \[[@B64],[@B65]\]. A substantial body of evidence for the effectiveness of educational interventions to improve glycemic control in general population has been generated \[[@B11],[@B12],[@B15],[@B16]\], observing similar effects to the one obtained by our meta-analysis for racial/ethnic minority groups. Our results also reiterate those obtained in a previous meta-analysis of interventions targeting racial/ethnic minority groups (-0.32%) \[[@B18]\]. This is the first meta-regression study analyzing the effect of specific characteristics of educational programs targeted to racial/ethnic minorities. Moreover, meta-regressions of programs targeted to non-minority groups have explored a more reduced number of characteristics \[[@B11],[@B16]\]. Educational programs delivered face to face produced a greater improvement in glycemic control than those delivered using telecommunication based formats. The comparative effectiveness of these two formats of administration is currently a topic of substantial interest, and there is no previous evidence in the context of self-management education in ethnic minorities. Although telecommunication programs have the potential to improve attrition rates, as they can help to overcome barriers such as competing responsibilities and distance to the service, they can represent an additional barrier to patients from racial/ethnic minority groups, who are more likely to have decreased access to information technologies and lower digital literacy \[[@B66]\]. Additionally, our meta-regression suggested that interventions delivered individually produced better results than those delivered in a group format. Previous research on general population has specifically explored this issue, without observing differences between individual and group administration \[[@B10],[@B64]\]. Both the lower maintenance costs and the potential for promoting patient-patient interactions \[[@B67]\] make group-based interventions very appealing. Individual education, however, can more efficiently address patients' individual characteristics and needs, producing better patient engagement. More research is needed to confirm our results. Most of the educational programs included in our review were based on traditional didactic methods, either alone or in tandem with other educational techniques. However, those interventions based on cognitive reframing techniques produced better results. Similar results were obtained in a previous meta-regression of educational programs in the general population \[[@B11]\]. Furthermore, they corroborate previous findings that knowledge of lifestyle guidelines is a necessary but not the only factor required to facilitate the appropriate behavioral changes \[[@B15],[@B17]\], suggesting that a patient's inability to adhere to an adequate self-management might be grounded in motivational factors. The importance of motivational factors to promote the adherence to lifestyle interventions has been previously acknowledged and included in interventions targeting racial and ethnic minorities \[[@B68],[@B69]\]. The involvement of peer providers also produced better results. The benefit of including peer educators has been previously suggested \[[@B70],[@B71]\], and partially explained by the fact that peers can provide a more credible source of information, empower those involved, and reinforce learning through ongoing contact \[[@B72]\]. In order to estimate the complexity of the interventions, we calculated the number of teaching methods and educational contents included in each program. Contrary to our expectations, we observed an inverse association with HbA1c, indicating that less complex interventions led to greater improvements in glycemic control. This is the first study analyzing the relation between complexity and effectiveness, and more research is needed to confirm this potential dilution effect. Strengths and weaknesses ------------------------ The main strength of this study is the comprehensiveness of the searches. Systematic and manual searches were performed in the most relevant bibliographic databases for biomedical research, as well as in specific sites of gray literature. We also examined the effect of a high number of intervention characteristics with the potential to produce better effects, some of which has not previously explored, namely the complexity of the programs or its intensity. Additional strengths are that we specifically focused our review on exclusively educational interventions (i.e., excluding those interventions with additional components such as case management, financial incentives or health provider education) and included sensitivity analysis excluding those studies with higher risk of bias. Our review also has some limitations. First, although we attempted to identify studies conducted in OECD countries, a vast majority of the interventions were conducted in the US, which limits the external validity of our results. Second, our meta-analysis and meta-regression was restricted to glycemic control. Although we attempted to conduct meta-analyses on other relevant outcomes such as diabetes knowledge, they were not consistently available or uniformly measured. Finally, although formal tests on publication bias seemed to exclude its presence, we cannot completely rule out its existence. Remaining gaps in knowledge --------------------------- More than 90% of the studies included in this review were conducted in the US, which limits the external validity of our results. Ethnic/racial inequalities in rates of diabetes-related complication have been observed in multiple countries and ethnic minorities \[[@B3]\]. Therefore, the effectiveness of interventions specifically targeting minorities needs to be assessed. This review also found that there is a considerable knowledge gap regarding the long-term effects of these interventions. Only about a fourth of the studies included had a post-test assessment, the majority of occurred within six months after the intervention ended. Given that type 2 diabetes is a chronic condition, it is crucial to understand not only that self-management educational programs can produce a discrete impact, but also whether the impact is sustained in the long term. Also importantly, a quarter of the interventions included in this review were evaluated through quasi-experimental studies. Some of these studies did not include a randomization element in the design, whereas other did not include a control group (non-controlled before-after studies). Moreover, a significant proportion of the studies (35%) presented a high risk of bias, which included small sample sizes, relevant confounders not adequately being controlled for, and participants not blinded to the intervention. Notwithstanding the difficulties underlying the execution of this type of complex clinical trials, larger and methodologically more robust trials are very much needed to confirm the findings of the present review, and to further identify characteristics of successful programs. Finally, only a small proportion of studies included cost-effectiveness estimation, which constitutes another important area for future research. Conclusions =========== In this systematic review we identified and analyzed DSM educational programs specifically targeted to racial/ethnic minority groups, observing that most of them can improve diabetes knowledge, self-management behavior, and clinical outcomes. Interventions producing higher improvements in glycemic control are those delivered individually and face to face, involving peer educators, based on cognitive reframing techniques, and employing a lower number of teaching methods. The long-term effects on patient-centered and clinically important outcomes, as well as cost effectiveness, remain unknown. Competing interests =================== The authors declare that they have no competing interests. Authors' contribution ===================== IRC and IRP designed the study. DCG, ARG and GP selected the articles and extracted relevant data. MRB conducted the statistical analysis. IRC drafted article. All authors provided input during the preparation of the manuscript, and approved the final version. IRC is the guarantor of this article. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6823/14/60/prepub> Supplementary Material ====================== ###### Additional file 1: Table S1 Search strategy in Medline (Ovid). ###### Click here for file ###### Additional file 2: Table S2 Registry of the Bibliographic Searches. ###### Click here for file ###### Additional file 3: Table S3 Characteristics and Effectiveness of the Diabetes Self-management Educational Interventions. ###### Click here for file Acknowledgments =============== The authors thank Victor Sarmiento (Andalusian Agency for Health Technology Assessment, Andalusia, Spain) for designing the bibliographic searches. Funding source -------------- National Institute of Health Carlos III (Study PS09/00747). The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.

1.. Introduction ================ Heart rate has been an important factor in indicating patients' health for quite a long time. As people are paying more and more attention on their health, long-term heart rate monitoring is of more importance to everyone as it's a valuable indicator for early diagnosis of dozens of diseases. Furthermore, long-term heart rate monitoring is also essential for sports enthusiasts and professional athletes. According to these requirements, it is necessary to design and fabricate a device which is suitable for long-term heart rate monitoring and free from external disturbance. The wearing comfort, the reliability and the cost become three key issues for long-term heart rate monitoring for daily use. Several conventional methods have already been developed to measure heart rate, most of which are based on electrocardiograms (ECGs) \[[@b1-sensors-15-03224]\], photoplethysmography (PPG) \[[@b2-sensors-15-03224]\], and the piezoelectric effect \[[@b3-sensors-15-03224]--[@b5-sensors-15-03224]\]. Other principles in heart rate measurement have also been verified by researchers \[[@b6-sensors-15-03224]--[@b8-sensors-15-03224]\]. Among these methods for heart rate monitoring, ECG is the most widely used. It is the standard diagnosis technique for heart disease in hospitals but the specialized equipment needed is clearly not suitable for household applications. Household heart rate monitors based on ECG technique have also been commercialized, but chest strap transmitters are usually needed, which is not appropriate for long-term use. Strapless wrist band ECG heart rate monitors have been developed these days. However, the contradiction between reliable electric connection and the comfort of wearing still has not been settled in a satisfactory way. Techniques for PPG require hard photoelectric modules which are tightly attached to the tissue with a certain penetration depth, for instance, the fingers. For the reasons above, these two kinds of heard rate monitor system are not appropriate for long-term ordinary daily use because of the inconvenience in user experience, especially at night. Monitors based on piezoelectric pressure sensor also have the similar limits. 2.. System Design ================= In this work, a comfort, reliable heart rate monitor for long-term household use is proposed. Unlike the previous ECG, PPG and piezoelectric devices, a flexible pressure sensor with the capability of measuring heart rate by sensing the pulsation of the artery is used in the system. This solution is adopted based on the following considerations. Since the sensor of the monitoring system is the only component which needs to inevitably be in contact with the human skin, trying to have a flexible sensor is crucial in realizing a comfortable, simple heart rate monitor. In this work, a robust, low-cost novel flexible pressure sensor has been proposed and fabricated. The polymer-based flexible sensor has no hard components which it can ensure its comfort and fitness for wearing all day in any situation. Furthermore, to distinguish the tiny pressure changes caused by the artery pulse, the flexible pressure sensor's sensitivity needs to be improved. Based on previous works in this area, a highly sensitive structure is proposed. The conventional structure with one piezoresistive layer is replaced by a novel structure with two piezoresistive layers. The contact interface of the two layers is modified with micro structures. These microstructures are realized with modern soft lithography technology and are treated as the key features for the sensor's high sensitivity. In addition to that, for the specific morphology of the surface microstructures utilized, the pressure sensor device also has a perfect linearity performance. However, due to the high sensitivity of the pressure sensor, noises may be induced by the wearer's muscle movements. This inevitable daily movement can cause pressure changes between the wearer's skin and the device. Such noises can easily make the monitor fail when heart rate counting. In order to solve this problem while maintaining a relatively low cost in signal processing, an Analog Signal Processing (ASP) system is proposed to process the signal and extract the heart rate information from it. The whole system is composed of a sensor subsystem and the ASP subsystem. The sensor subsystem is an elastic belt with the flexible pressure sensor attached on it. The tightness of the belt can be tuned in order to provide a proper pressure for measurement. The system also includes an ASP subsystem to process the pressure signals, a counter, an indicator, and a timing circuit which is used to stop the measurement. 2.1.. Flexible Pressure Sensor ------------------------------ To realize the characteristic of flexibility in addition to the pressure sensing capability, a flexible piezoresistive carbon black/silicone rubber nanocomposite is adopted as the functional material \[[@b9-sensors-15-03224]--[@b12-sensors-15-03224]\]. For its stretchable, flexible properties and piezoresistive effect, this nanocomposite material has already been used in tactile sensors for robots to provide contact or grasping force feedback \[[@b13-sensors-15-03224],[@b14-sensors-15-03224]\]. They are also some other sensors for mechanics that adopt the material for their piezoresistive properties and energy absorbing capabilities in preventing mechanical collisions \[[@b15-sensors-15-03224],[@b16-sensors-15-03224]\]. In addition to that, all other materials used in device are flexible. [Figure 1](#f1-sensors-15-03224){ref-type="fig"} shows a simple schematic diagram demonstrating the assembly and packaging steps used in fabricating the device. Two layers of surface-modified piezoresistive polymer nanocomposite films are placed face-to face inside the device as the pressure sensing layer. These polymer films are packaged by another two layer of patterned Flexible Copper Clad Laminate (FCCL) films which act either as the package material or as the electrode of the device. To isolate this device from external moisture and contamination, another layer of PI bond-ply is used as the adhesion layer to ensure these four layers of material a good firm bond. These bond-ply layers have also been previously patterned by a laser cutting machine to save the space for the pressure sensing layer. The scale of the device's sensing part is 15 × 30 mm^2^. Since the human artery pulse in the wrist is weak and the excited pressure variation is not easy to sense, the flexible device' sensitivity needs to be further improved. Several previous efforts have been made to promote the devices' sensitivity in this research area \[[@b17-sensors-15-03224]--[@b23-sensors-15-03224]\]. A pressure sensor with record sensitivity with highly sensitive material with hollow polymer sphere in it has been proposed by Bao's group \[[@b17-sensors-15-03224]\]. Another type of research mainly utilizea tiny features' very sensitive contact state to realize high sensitivity \[[@b18-sensors-15-03224],[@b19-sensors-15-03224]\]. They usually have polymer-based micrometer scale pyramid structures to make the device electrically sensitive to external pressure loading. Devices of this type usually have a very sensitive electric response at the beginning when an array of unique pyramids gets contact to the opponent electrode. However, as the pressure goes higher, device's response become dull as the pyramidal surface get saturated, as shown in [Figure 2b](#f2-sensors-15-03224){ref-type="fig"}. A certain degree of pre-pressure needs to be maintained between the sensor and the human skin in order to ensure effective contact state and signal acquisition and the wrist artery pulse induced pressure varies (range \< 3 kPa) based on this pre-pressure bias. Furthermore, the pre-pressure also changes with the differences between users when they fasten on the wrist belt. Consequently, the real effective span for the heart rate measurements will be around 8--18 kPa. Several previous works on flexible pressure sensors have reported sensitivity high enough for heart rate measurement. However, these high sensitivities only exist within a limited range around 0 Pa. As shown in [Figure 2b](#f2-sensors-15-03224){ref-type="fig"}, the slope of the device response decreases as pressure goes higher, like the sensitivity. When the pressure reaches the effective range, devices becomes so dull that it can't respond to the artery pulse. As to the proposed novel device with full-scale high sensitivity ([Figure 2a](#f2-sensors-15-03224){ref-type="fig"}), artery pulse-induced pressure variation will make it respond intensively compared with previous ones. To have this full-scale sensitive pressure response, a new flexible pressure sensing device have been proposed in this work. The surface of the nanocomposite film has been modified with micrometer scale bump structures randomly distributed on surface with certain variation in height and lateral size. [Figure 2c](#f2-sensors-15-03224){ref-type="fig"}--d shows the SEM photos of the microbumps from the top view and side view, respectively. Unlike the typical pyramid structures of the same size, the distributed tiny structures' size makes them gradually touch with the opponent conductive surface along the whole pressure loading process so as to have a full scale high sensitivity. By surface profile measurement, the overall surface obeys the Gaussian Random distribution. By numerical simulation with MATLAB, the device response is expected to be linear. Furthermore, the micrometer scale piezoresistive bumps on the surface are very sensitive to external pressure loading as the pressures are concentrated on these tiny structures. Large deformations of these piezoresistive structures lead to dramatic conductance variations of the device. Besides that, for the distributed bump size, the conducting paths are established gradually. The capability in conductance variation is even higher. Based on above statement, the device sensitivity is expected to be higher. The devices' sensing capabilities were tested using a device test platform composed of a micropositioner, force gauge and the electric measurement part. All three of these parts are connected to a computer and controlled by a Labview software routine. Then, the device is tested for 100 cycles and the test results are plotted in [Figure 3](#f3-sensors-15-03224){ref-type="fig"}. The device responds linearly to the external pressure with the sensitivity around 13.4 kPa^−1^. A certain degree of hysteresis has been observed and the relative hysteresis does not exceed 9% over the full measurement scale. The device also shows an extraordinary stability and repeatability. [Figure 4](#f4-sensors-15-03224){ref-type="fig"} shows the statistical distribution of the measured conductance result at five different pressure levels. The temperature dependence of the fabricated device has also been tested, as shown in Supplementary file. However, as the readout circuit mechanism mentioned below, the temperature variation will not have an obvious influence on the system's performance. 2.2.. Materials and Methods --------------------------- The piezoresistive polymer is prepared with the 107 silicone rubber (hydroxyl end-blocked dimethylsiloxane) and the super conductive carbon black (HG-3F: average diameter-12 nm) with the weight ratio of 100:8. An extra amount of *n*-hexane solvent is added into the mixture to promote the dispersion of carbon black particles in the silicone rubber. A subsequent process of 10 h ultrasonic treatment along with the fast stirring is carry out to ensure the carbon black's sufficient dispersion. Before using, a proper amount of curing agent is added into the mixture (carbon grease) with a certain degree of stirring and ultrasonic treatment. Then, the carbon grease is spun coated on the GRD surface mould with the spin speed of 1000 rpm. After about 48 h of curing process under room temperature, the piezeresistive film on the mould is peeled off by hand and tailored to the proper shape and size to match the device design. 2.3.. The Signal Processing Circuit ----------------------------------- To make full use of the piezoresistive pressure sensor, an operational amplifier (op-amp) other than a Wheastone bridge \[[@b24-sensors-15-03224]\] is used to transform the variation of the resistance into a voltage signal. This system is designed with the following principles. Firstly, it uses basic low-cost analog electronic components to lower the over-all cost. Secondly, it is designed to reduce noises and interferences caused by different sources. High-frequency noise is induced by the electromagnetic interference of the environment. The movement of human body, either intended or not, can also cause significant pressure changes in the sensor. This will result in undesired dc-drift of the signal which can be obviously observed before ASP circuits. Finally, after reducing the noises, the system converts the signal into a level signal, which is transferred into a low-cost digital counter. The flow chart of the ASP system is shown in [Figure 5](#f5-sensors-15-03224){ref-type="fig"}. The first op-amp transforms the pressure on the sensor into signal A. Each beat of the heart causes a short time of pressure change on the sensor, which leads to the change of resistance. This results in the variation of signal A. Signal A mainly contains heart rate information and high frequency noise and dc-drift. High frequency noise is introduced by the electromagnetic interference, and dc-drift is caused by human motion. Here the signal processing system is used to remove these two types of interference and extract heart rate information. An analog signal processing system is introduced, first two low pass filter with different cutoff frequency is used to process signal A. Low pass filter 1 generates signal B1. This filter is designed to filter noise with frequency much higher than the heart rate frequency (about 1--2 Hz) in order to reduce noise while preserving the heart rate information. This filter assures that B1 maintains the dc-drift and heart rate information. Low pass filter 2 generates signal B2 by removing the high frequency signal and add a voltage bias to signal A in order to compare with signal B2. It has a low cutoff frequency to only preserve the dc-drift of the original signal. Comparator 1 compare signal B1 with B2 and generate signal C. These series of processes make sure that signal C contains most of the heart rate information while staying nearly undisturbed by dc-drift. However, signal C still has too many glitches and cannot be used to count heart rate directly. Low pass filter 3 is used to reduce the glitches of signal C and generates signal D. Comparator 2 is introduced in order to compare signal D with a constant voltage. Signal E is generated as a result of the comparison of signal D and the constant voltage. Finally E only contains heart rate information and is used to count heart rate. 3.. Integration and Test ======================== The schematic of the device and the test setup in this experiment is shown in [Figure 6](#f6-sensors-15-03224){ref-type="fig"}. The whole system consists of an elastic belt which is used to attach the sensor onto the wearer's wrist, the flexible pressure sensor and a module which contains the ASP system and the counter. This module is used to indicate the heart rate, and power input. The artery's pulsation is transmitted to the human skin, normally this pressure change can be easily sensed if a fingertip is put onto a human's wrist. Under this application situation, an elastic belt is used to fasten the pressure sensor on the wrist above the artery, so that the pressure sensor can be firmly attached to the wrist while sensing the pressure change. The elastic belt and the pressure sensor are both flexible, which ensures that the heart rate monitor is comfortable to wear. The change of the applied pressure will cause the resistance change of the pressure sensor. An amplifier is used to convert the change of the resistance into a voltage signal. Then the voltage signal is processed by the ASP subsystem in the module, as discussed previously. The ASP subsystem generates a level signal which contains heart rate information. Finally, this level signal is transmitted into the counter directly to measure the heart rate. The power input is used to power up the whole system. 4.. Results =========== The heart rate monitor was tested on several researchers. An oscilloscope is used to capture the wave generated by the ASP subsystem while it is processing the signal. The waveform corresponding to the signal in [Figure 5](#f5-sensors-15-03224){ref-type="fig"} is shown in [Figure 7](#f7-sensors-15-03224){ref-type="fig"}. Signal A shows the voltage signal which is converted from the change of the resistance of the pressure sensor. There are eight small concave shapes on the waveform which indicate eight pulses, however this subtle heart rate signal cannot be used to count heart rate directly. B2 is the output of the low pass filter 2 with dc bias. This indicates that the dc-drift of the original signal is quite large, and even larger than pulse itself. However, testing results reveal that our ASP system can successfully remove the interference of the dc-drift. B1 is the result of low pass filter 1. Signal C is the compare result of B1 and B2. D is the output of the low pass filter 3. E is the result of the comparator 2, it is compressed proportionately and put into the counter. This result shows the monitor has good immunity to high level dc drift. Several tests have been made to confirm the reliability of the monitor, see [Table 1](#t1-sensors-15-03224){ref-type="table"}. Reference results are obtained from a commercial electronic sphygmomanometer (HEM-7052, OMRON, Kyoto, Japan). The average error is less than 3%. 5.. Conclusions =============== A flexible, low-cost, small heart rate monitor comfortable enough for full-day wear is demonstrated. It is based on the pressure changes on the skin caused by the artery pulse. A sensitive flexible pressure sensor is used to sense the pressure change and an effective ASP system is used to extract heart rate information from the signal. This heart rate sensor has also been tested on several testees, and the results shows that this sensor is highly precise. The ASP system also shows its effectiveness in reducing the noise. Our work shows that based on this novel flexible pressure sensor and ASP system, a new way for heart rate monitoring can be realized. This work shows the possibility of full-day long-term heart rate monitoring. Supplementary Material ====================== This work was supported by National Natural Science Foundation (61434001), 973 Program (2015CB352100), National Key Project of Science and Technology (2011ZX02403-002), and Special Fund for Agroscientic Research in the Public Interest (201303107) of China. Y. Shu and C. Li conceived the project and contributed equally to this work. Y. Shu conducted the design, fabrication and tesing of the flexible device. C. Li mainly focused on the design, fabrication of the circuit. Z. Wang, W. Mi, Y. Li conducted the test of the system and analysed the data. Y. Shu and C. Li wrote the manuscript. T. Ren supervised the project. All authors discussed the results and commented on the paper. The authors declare no conflict of interest. {#f1-sensors-15-03224} {#f2-sensors-15-03224} {#f3-sensors-15-03224} {#f4-sensors-15-03224} {#f5-sensors-15-03224} {#f6-sensors-15-03224} {ref-type="fig"}.](sensors-15-03224f7){#f7-sensors-15-03224} ###### Result of the heart rate sensor and commercial electronic sphygmomanometer. **Measurement Equipment** **Heart Rate Sensor** **Electronic Phygmomanometer** --------------------------- ----------------------- -------------------------------- Normal 66 64 68 66 70 70 After exercise 90 96 [^1]: These authors contributed equally to this work. [^2]: Academic Editor: Vittorio M.N. Passaro

Introduction {#s1} ============ The intestinal mucosa is a critical effector site for elimination of enteric pathogens. *Toxoplasma gondii*, a ubiquitous protozoan parasite, is a prime example of such a pathogen. Mammals are infected with *T. gondii* primarily by the ingestion of tissue cysts from undercooked meat or oocysts excreted in the feces of felines, which are the sole definitive hosts. Upon infection, the parasite induces a potent Th1 immune response that is characterized by high levels of IL-12 and IFN-γ [@ppat.1003706-Denkers1], [@ppat.1003706-Dupont1]. Initial IL-12 production is largely the result of MyD88-dependent Toll-like receptor (TLR) signaling in dendritic cells, and the parasite profilin molecule has been identified as a ligand for TLR11 and TLR12 [@ppat.1003706-Scanga1]--[@ppat.1003706-Sukhumavasi1]. IL-12 activates natural killer (NK) cells to initiate IFN-γ production and promotes T-cell differentiation towards a Th1 program. Ultimately IFN-γ is the critical cytokine involved in controlling *Toxoplasma*. While *in vitro* experiments suggest that macrophages activated by this cytokine acquire anti-*Toxoplasma* activity through upregulation of immunity-related GTPase (IRG) molecules that mediate destruction of the parasitophorous vacuole [@ppat.1003706-Zhao1]--[@ppat.1003706-Khaminets1], the *in vivo* function of IFN-γ is less clear. Inflammatory monocytes are an important component of defense against microbial pathogens, including *Toxoplasma* [@ppat.1003706-Dunay1]. These cells express high levels of Ly6C/G (Gr-1) and are recruited from the bone marrow via chemokine (C-C motif) receptor 2 (CCR2) [@ppat.1003706-Geissmann1]. During *Listeria monocytogenes* infection, inflammatory monocytes are recruited from the bone marrow to the spleen and liver where they differentiate into TNF-α- and nitric oxide (NO)-producing DCs (Tip-DCs). There they are essential for bacterial clearance and mouse survival [@ppat.1003706-Shi1], [@ppat.1003706-Serbina1]. Likewise, CCR2-dependent inflammatory monocytes are recruited to the lung during *Mycobacteria tuberculosis* infection where they protect mice from disease by recruiting and activating T cells and by producing NO [@ppat.1003706-Peters1], [@ppat.1003706-Peters2]. Mucosal defense against *T. gondii* has also recently been shown to require CCR2-dependent inflammatory monocytes [@ppat.1003706-Dunay1]. Upon recruitment to the small intestine, these cells control the parasite either indirectly by production of IL-12 and TNF-α or directly through production of NO and IRG proteins [@ppat.1003706-Yarovinsky1]--[@ppat.1003706-Andrade1], [@ppat.1003706-Zhao1], [@ppat.1003706-Taylor1], [@ppat.1003706-Dunay1], [@ppat.1003706-Ling1]. While CCR2 enables recruitment of inflammatory monocytes to sites of infection, the factors that coordinate their activation and acquisition of effector function are not known. CXCR3 is a Th1-associated chemokine receptor, and cells expressing this receptor respond to the IFN-γ-inducible chemokines CXCL9, 10, and 11 [@ppat.1003706-Groom1]. The receptor is expressed predominantly by T cells and NK cells and is rapidly upregulated upon cell activation. There is evidence that CXCR3 expression enables T-cell entry into sites of infection, although the outcome of recruitment varies among pathogens. In the case of *Leishmania major*, recruitment is protective as CXCR3-expressing T cells are required for the resolution of cutaneous lesions [@ppat.1003706-Rosas1]. However, in the case of *Plasmodium berghei* ANKA, CXCR3 is pathogenic because it allows entry of proinflammatory cells into the CNS, resulting in cerebral malaria [@ppat.1003706-Campanella1]. Here we determined the role of CXCR3 in the intestinal immune response to *Toxoplasma*. We found that loss of CXCR3 negatively affected host survival against oral infection. This was associated with diminished recruitment of CD4^+^ T cells to the lamina propria (LP), decreased T cell IFN-γ secretion, impaired inflammatory monocyte effector function, and inability to control the parasite in the intestinal mucosa. Reconstitution with CXCR3-competent CD4^+^ T cells restored inflammatory monocyte function, resulting in improved survival against the parasite. Protective effects of adoptively transferred CD4^+^ T lymphocytes depended upon their ability to produce IFN-γ, but occurred independently of CD4 expression of CD40L. Our data show that CXCR3 enables Th1 recruitment to the intestinal LP, where these cells instruct activation of CCR2-dependent inflammatory monocytes, in turn controlling infection. These results establish CXCR3 as a major determinant orchestrating communication between effectors of innate and adaptive immunity, enabling effective host defense against infection. Results {#s2} ======= CXCR3 and its ligands CXCL9 and CXCL10 are upregulated during acute toxoplasmosis {#s2a} --------------------------------------------------------------------------------- Because CXCR3 and its chemokine ligands are strongly associated with Th1 responses, we asked whether this proinflammatory axis was induced during *Toxoplasma* infection in the intestinal mucosa. Accordingly, mice were orally inoculated with cysts, and relative levels of CXCR3, CXCL9 and CXCL10 mRNA expression were measured over the course of acute infection. We found strong upregulation of CXCR3 and its specific chemokine ligands as early as Day 4 post-infection in both the ileum and mesenteric lymph nodes (MLN) ([Fig. 1A](#ppat-1003706-g001){ref-type="fig"}). Overall, peak CXCR3 mRNA levels were attained by Day 6 post-inoculation. {#ppat-1003706-g001} In order to examine CXCR3 expression in more detail, we utilized *Cxcr3* eGFP reporter (CIBER) mice, a bicistronic reporter strain in which cells expressing CXCR3 also express eGFP [@ppat.1003706-Oghumu1]. We found a large increase in CXCR3 populations of both CD4^+^ and CD8^+^ T lymphocytes in MLN, spleen (SPL) and LP compartments following infection ([Fig. 1B](#ppat-1003706-g001){ref-type="fig"}). In general, CXCR3 upregulation was most pronounced in the CD4^+^ population. For example, in the MLN there was a 6-fold increase in CD4^+^CXCR3^+^ cells but only a 2-fold increase in CD8^+^CXCR3^+^ lymphocytes. NK cells are also known to express CXCR3 and are an important source of early IFN-γ during *T. gondii* infection [@ppat.1003706-Qin1], [@ppat.1003706-Sher1]. However, while CXCR3-GFP expression was relatively high on naïve NK cells, the GFP expression was in fact reduced during infection, suggesting lack of a role for CXCR3^+^ NK cells during intestinal infection ([Fig. S1](#ppat.1003706.s001){ref-type="supplementary-material"}). We next examined expression of the activation marker CD27 on GFP^+^ and GFP^−^ T lymphocytes in infected mice. CD27 was significantly lower in CXCR3^−^GFP^−^ cells, suggesting an altered maturation state of the CXCR3^−^ T cells ([Fig. S2A and B](#ppat.1003706.s002){ref-type="supplementary-material"}). Likewise, there was a lower percentage of CD27^+^ cells amongst CXCR3-GFP^−^ CD8^+^ T lymphocytes, although the decreases in expression were not as striking as with the CD4^+^ lymphocytes ([Fig. S2C and D](#ppat.1003706.s002){ref-type="supplementary-material"}). *Cxcr3^−/−^* mice are increased in susceptibility and are prone to severe intestinal damage following *T. gondii infection* {#s2b} --------------------------------------------------------------------------------------------------------------------------- To further examine the role of CXCR3 during *T. gondii* infection, mice deficient in CXCR3 were orally inoculated with low virulence ME49 cysts, and the outcome of infection was monitored. While all wild-type (WT) mice survived acute infection with 30 cysts, *Cxcr3^−/−^* animals displayed increased susceptibility with nearly 75% of mice dying by 2 weeks post-infection ([Fig. 2A](#ppat-1003706-g002){ref-type="fig"}). When the cyst dose was increased to 50, all CXCR3 knockout (KO) mice rapidly succumbed to infection, but some WT mice also died ([Fig. 2B](#ppat-1003706-g002){ref-type="fig"}). Interestingly, when WT and KO mice were infected by intraperitoneal injection, lack of CXCR3 did not affect survival, indicating that the effect of CXCR3 is specific to the mucosal response ([Fig. S3A](#ppat.1003706.s003){ref-type="supplementary-material"}). To further examine the overall response in orally infected mice, we examined the gross appearance of the small intestine of WT and *Cxcr3^−/−^* mice after 30-cyst infection. The small intestines of the KO mice were strikingly damaged as demonstrated by massive hemorrhage compared to WT ([Fig. 2C](#ppat-1003706-g002){ref-type="fig"}). Consistent with intestinal shortening associated with increased damage [@ppat.1003706-Heimesaat1]--[@ppat.1003706-Ueno1], the length of the small intestine was reduced in the KO mice during infection ([Fig. 2D](#ppat-1003706-g002){ref-type="fig"}). Increased damage was further confirmed by H&E staining of small intestinal sections. WT mice displayed minor villus blunting accompanied by moderate to severe inflammatory cell recruitment in the submucosa ([Fig. 2E and G](#ppat-1003706-g002){ref-type="fig"}). In contrast, *Cxcr3^−/−^* mice displayed severe villus blunting, fusion, epithelial necrosis, sloughing of villus tips, and vascular congestion and hemorrhage ([Fig. 2F and H](#ppat-1003706-g002){ref-type="fig"}). Blind scoring of H&E sections revealed a significant decrease in inflammation scores in the absence of CXCR3 ([Fig. 2I](#ppat-1003706-g002){ref-type="fig"}), but when parameters of intestinal damage were quantitated, *Cxcr3^−/−^* mice scored significantly higher than WT counterparts ([Fig. 2J](#ppat-1003706-g002){ref-type="fig"}). This damage was infection-dependent as intestines from non-infected WT and *Cxcr3^−/−^* mice both had normal architecture with few inflammatory cells ([Fig. S3B](#ppat.1003706.s003){ref-type="supplementary-material"}). Increased epithelial damage in the absence of CXCR3 was further verified by loss of epithelial surface-associated Muc1 compared to infected WT animals, suggesting epithelial cell sloughing ([Fig. S3C](#ppat.1003706.s003){ref-type="supplementary-material"}). Despite the overall decreased inflammatory score, *Cxcr3^−/−^* mice consistently displayed an influx of neutrophils into the LP compartment compared to WT mice, suggesting a role for these cells in causing damage, as argued by others [@ppat.1003706-Shi1], [@ppat.1003706-Serbina1], [@ppat.1003706-Dunay2] ([Fig. S3D and E](#ppat.1003706.s003){ref-type="supplementary-material"}). {#ppat-1003706-g002} *Cxcr3^−/−^* mice are unable to control parasite replication in the small intestine {#s2c} ----------------------------------------------------------------------------------- Genetic knockout of cytokines such as IFN-γ results in susceptibility to *T. gondii* through the inability to control parasite replication, whereas the deletion of anti-inflammatory mediators such as IL-10 results in susceptibility due to cytokine pathology [@ppat.1003706-SchartonKersten1], [@ppat.1003706-Gazzinelli1]. Based on decreased inflammation scores, we hypothesized that the *Cxcr3^−/−^* mice were more likely to be succumbing from uncontrolled parasite replication rather than immune-mediated damage. To examine this, intestinal tissues were stained for parasite antigen by immunohistochemistry. Sections from WT mice displayed minimal parasite infiltration within the LP ([Fig. 3A](#ppat-1003706-g003){ref-type="fig"}). Conversely, *Cxcr3^−/−^* mice contained numerous large foci of parasite throughout the length of the small intestine that often coincided with areas of severe damage ([Fig. 3B](#ppat-1003706-g003){ref-type="fig"}). Surprisingly, this difference was restricted to the LP and submucosa of the small intestine because Peyer\'s patches (PP) in WT and *Cxcr3^−/−^* mice contained similar levels of parasite antigen ([Fig. 3C and D](#ppat-1003706-g003){ref-type="fig"}). Differences in parasite burden between WT and *Cxcr3^−/−^* mice in the MLN, spleen, and lung were also indiscernible by IHC analysis (data not shown). These results were further confirmed by quantitative PCR. Thus, while lung, liver, spleen, MLN and PP contained similar levels of parasite genomes regardless of CXCR3 expression, there was an approximately 50-fold increase in parasite levels in the absence of CXCR3 in intestinal tissues ([Fig. 3E](#ppat-1003706-g003){ref-type="fig"}). These data suggest that increased susceptibility to *Toxoplasma* in *Cxcr3^−/−^* mice was due to a localized inability to control parasite replication within the LP of the small intestine. {#ppat-1003706-g003} CD4^+^ T cells are recruited to the small intestine via CXCR3 {#s2d} ------------------------------------------------------------- The dominant effector cells required for elimination of *T. gondii* following oral infection are inflammatory monocytes. These cells express Ly6C/G (Gr-1), produce TNF-α, IL-12, and are likely to kill parasites via activation of IFN-γ-inducible p47 GTPases that assemble at the parasitophorous vacuole membrane and mediate its destruction [@ppat.1003706-Zhao1], [@ppat.1003706-Dunay1]. Consistent with others [@ppat.1003706-Dunay1], we observed these cells in the LP of infected mice ([Fig. 4A](#ppat-1003706-g004){ref-type="fig"}). Inflammatory monocytes are dependent upon CCR2 for exit from the bone marrow, but we wondered whether CXCR3 might be involved in recruiting these cells to the LP in response to *T. gondii*. Therefore, we examined CXCR3-GFP expression by intestinal inflammatory monocytes during infection. Inflammatory monocytes in the small intestinal LP of infected reporter mice did not express any GFP as compared to inflammatory monocytes isolated from infected *Cxcr3^−/−^* mice ([Fig. 4B](#ppat-1003706-g004){ref-type="fig"}). In stark contrast, approximately 50% of LP CD4^+^ T cells expressed high levels of GFP ([Fig. 4C](#ppat-1003706-g004){ref-type="fig"}). Furthermore, *Cxcr3^−/−^* mice displayed unaltered total numbers of LP inflammatory monocytes compared to wild-type controls (defined as CD11b^+^Ly6C^+^Ly6G^−^) ([Fig. 4D](#ppat-1003706-g004){ref-type="fig"}). We next assessed the kinetics by which CD4^+^ T cells and inflammatory monocytes were recruited to the lamina propria during infection. Between days 4 and 7 of infection, there was a significant increase in the total numbers of CD4^+^CXCR3^−^GFP^+^ T cells and inflammatory monocytes ([Fig. 4E](#ppat-1003706-g004){ref-type="fig"}). However, the total number of CD4^+^CXCR3^−^GFP^−^ cells remained unchanged, further indicating that infection promotes the recruitment of CD4^+^CXCR3^+^ T cells ([Fig. 4E](#ppat-1003706-g004){ref-type="fig"}). Few NK cells were observed in the lamina propria, but there was a small increase in their number during infection. This was attributable to an increase in CXCR3^−^ NK cells (data not shown). Consistent with these results, there was an influx of CD4^+^ T cells in WT small intestines that was diminished in *Cxcr3^−/−^* mice ([Fig. 4F--H](#ppat-1003706-g004){ref-type="fig"}). These findings demonstrate that CD4^+^ T cells fail to effectively traffic to the intestinal compartment in the absence of CXCR3, but the presence of LP inflammatory monocytes does not require this chemokine receptor. {#ppat-1003706-g004} Lamina propria CD4^+^ T cells display impaired IFN-γ production {#s2e} --------------------------------------------------------------- We next asked if expression of CXCR3 affected the ability of T cells to secrete the Th1 cytokine IFN-γ. Initial experiments on bulk populations of splenocytes and mesenteric lymph node (MLN) cells from Day-11 infected WT and KO revealed no differences in the amount of IFN-γ, TNF-α or IL-10 produced during *in vitro* culture ([Fig. S4](#ppat.1003706.s004){ref-type="supplementary-material"}). To specifically examine functional outcomes in intestinal cells, WT and *Cxcr3^−/−^* lamina propria leukocytes were harvested 4 and 6 days post-oral infection, stimulated *ex vivo*, and IFN-γ production was examined by flow cytometry. CD4^+^ T cells from both WT and *Cxcr3^−/−^* displayed enhanced IFN-γ production over time. However, in the absence of CXCR3, CD4^+^ T cells produced significantly less IFN-γ compared to WT cells at both examined time points ([Fig. 5A--E](#ppat-1003706-g005){ref-type="fig"}). This was confirmed by measuring IFN-γ from the supernatants of Day-6 WT and *Cxcr3^−/−^* intestinal biopsy cultures, where IFN-γ was lower in the absence of CXCR3 ([Fig. 5F](#ppat-1003706-g005){ref-type="fig"}). This effect was specific to the CD4^+^ T cell subset, as IFN-γ production by lamina propria CD8^+^ T cells and NK cells was unchanged between WT and knockout animals ([Fig. S5A](#ppat.1003706.s005){ref-type="supplementary-material"}--F). Further confirming that this loss of IFN-γ production was specific to CD4^+^ T lymphocytes in the small intestine, and consistent with the bulk splenocyte culture experiments, splenocytes isolated from infected WT and *Cxcr3^−/−^* mice secreted equivalent levels of IFN-γ upon *ex vivo* stimulation with PMA and ionomycin ([Fig. S5G](#ppat.1003706.s005){ref-type="supplementary-material"}--I). Together, these results indicate an intestine-specific defect in presence of CD4^+^ Th1 cells in the absence of CXCR3 as measured by the capacity to produce IFN-γ. {#ppat-1003706-g005} Inflammatory monocyte responses in the intestinal mucosa are defective in the absence of CXCR3 {#s2f} ---------------------------------------------------------------------------------------------- Although IFN-γ, IL-10, and TNF-α responses remained intact in the MLN and spleen late during infection of CXCR3-deficient mice, a significant decrease in IL-12 production was observed in the MLN ([Fig. 6A](#ppat-1003706-g006){ref-type="fig"}) and spleen ([Fig. 6B](#ppat-1003706-g006){ref-type="fig"}) of *Cxcr3^−/−^* mice. Defective IL-12 responses in the CXCR3 KO strain were infection dependent, because parasite antigen stimulated equivalent amounts of IL-12 in noninfected WT and KO splenocytes ([Fig. 6C](#ppat-1003706-g006){ref-type="fig"}). This response, known to derive from resident splenic CD8α^+^ DC [@ppat.1003706-Mashayekhi1], may account for equivalent Th1 priming in secondary lymphoid organs, despite lower IL-12 levels during late acute infection. {#ppat-1003706-g006} Since IL-12 is also a characteristic cytokine of inflammatory monocytes, we investigated the impact of CXCR3 deletion on intestinal inflammatory monocyte function. Indeed, *in vitro* culture of intestinal biopsy samples revealed decreased production of IL-12 ([Fig. 6D](#ppat-1003706-g006){ref-type="fig"}). To further identify the source of the defective IL-12, we examined the production of IL-12 from inflammatory monocytes. While the total numbers of LP inflammatory monocytes were equivalent in WT and CXCR3 KO mice ([Fig. 4B](#ppat-1003706-g004){ref-type="fig"}), the population of CD11b^+^Gr-1^+^ cells co-expressing IL-12 and TNF-α was dependent upon CXCR3 ([Fig. 6E and F](#ppat-1003706-g006){ref-type="fig"}). Further confirming impaired inflammatory monocyte function, iNOS expression was significantly decreased in inflammatory monocytes ([Fig. 6G and H](#ppat-1003706-g006){ref-type="fig"}). These findings strongly suggest that inflammatory monocytes are functionally impaired in the absence of CXCR3. Interestingly, neutrophils in the LP of KO mice produced significantly higher levels of TNF-α compared to WT neutrophils ([Figure S6A and B](#ppat.1003706.s006){ref-type="supplementary-material"}). Adoptive transfer of WT CD4^+^ T lymphocytes rescues inflammatory monocyte function and restores resistance in *Cxcr3^−/−^* mice {#s2g} -------------------------------------------------------------------------------------------------------------------------------- Given the data so far, we hypothesized that CD4^+^ T cells were unable to effectively home to the small intestine and prime inflammatory monocyte function in the absence of CXCR3, resulting in susceptibility to *Toxoplasma*. We therefore tested whether reconstitution with CXCR3-competent CD4^+^ T cells would allow *Cxcr3^−/−^* mice to overcome susceptibility and restore inflammatory monocyte function. Accordingly, CD4^+^ T cells from naïve WT spleens were enriched to 90--95% purity by magnetic bead separation ([Fig. 7A](#ppat-1003706-g007){ref-type="fig"}) and injected i.v. into *Cxcr3^−/−^* recipients. Mice were orally challenged with *T. gondii* 24 hours post-adoptive transfer, and survival was monitored. Knockout mice that did not receive WT cells began to die 10 days post-challenge, while all KO mice that received CD4^+^ T cells and all WT controls survived the acute phase of infection ([Fig. 7B](#ppat-1003706-g007){ref-type="fig"}). To confirm that the CD4-dependent survival was not an artifact of the transfer, *Cxcr3^−/−^* CD4^+^ T cells were adoptively transferred into KO recipients. The knockout cells were unable to protect against susceptibility, demonstrating that protection is dependent on CXCR3 expression by CD4^+^ T cells ([Fig. S7A](#ppat.1003706.s007){ref-type="supplementary-material"}). {#ppat-1003706-g007} To assess the functional impact of WT CD4 adoptive transfer, parasite burden and cytokine production in the LP were assessed by flow cytometry in WT, *Cxcr3^−/−^*, and *Cxcr3^−/−^* +CD4 mice. While *Cxcr3^−/−^* mice had a clear increase in *Toxoplasma* infected cells relative to WT, upon adoptive transfer of WT CD4^+^ T cells, parasite levels were reduced to WT ([Fig. 7C](#ppat-1003706-g007){ref-type="fig"}). Likewise, expression of IL-12/TNF-α by inflammatory monocytes was significantly reduced in KO mice, but these cytokines returned to WT levels upon transfer of WT CD4^+^ T cells ([Fig. 7D](#ppat-1003706-g007){ref-type="fig"} and [Fig. S7B](#ppat.1003706.s007){ref-type="supplementary-material"}). In addition to cytokine responses, intestinal damage was also alleviated by adoptive transfer of WT CD4^+^ T cells, as the intestinal lengths of the transferred mice were restored to WT ([Fig. 7E](#ppat-1003706-g007){ref-type="fig"}). Possibly as a result of improved monocyte function and parasite clearance, neutrophil levels and neutrophil TNF-α secretion were also restored to WT levels following the transfer of WT CD4^+^ T cells ([Fig. S7C and D](#ppat.1003706.s007){ref-type="supplementary-material"}). CD4+ T-cell rescue is dependent on IFN-γ but independent of CD40L {#s2h} ----------------------------------------------------------------- To identify the mechanism behind the rescue of *Cxcr3^−/−^* susceptibility by WT CD4^+^ T cells, we performed adoptive transfer experiments utilizing T cells derived from knockout animals. Inasmuch as CXCR3 is a Th1 chemokine receptor, we began by asking whether reversal of susceptibility was dependent on CD4-derived IFN-γ Therefore, CD4^+^ T cells were isolated from *Ifnγ* mice and adoptively transferred into CXCR3-deficient recipients. Unlike IFN-γ-competent CD4^+^ T cells ([Fig. 7B](#ppat-1003706-g007){ref-type="fig"} and [Fig. 8B](#ppat-1003706-g008){ref-type="fig"}), transfer of IFN-γ KO CD4^+^ T lymphocytes failed to provide significant protection ([Fig. 8A](#ppat-1003706-g008){ref-type="fig"}). It has been shown that CD40L contributes to inflammatory responses in the intestinal mucosa during oral *Toxoplasma* infection [@ppat.1003706-Li1]. Therefore, we performed the adoptive transfer using *Cd40l^−/−^* CD4^+^ T cells and assessed survival. As expected, CXCR3-deficient animals were highly susceptible to infection. However, *Cxcr3^−/−^* mice receiving *Cd40l^−/−^* CD4^+^ T cells survived the infection, indicating that CD40L does not mediate CXCR3^+^CD4^+^-dependent protection ([Fig. 8B](#ppat-1003706-g008){ref-type="fig"}). {ref-type="fig"} legend (WT: n = 10; KO: n = 11; *Ifn*-γ*^−/−^*: n = 11; *CD40L^−/−^*: n = 5).](ppat.1003706.g008){#ppat-1003706-g008} Discussion {#s3} ========== Effective control of pathogens such as *T. gondii* requires the coordinated action of cells of innate and adaptive immunity. Orchestration of the response is governed by an underlying network of chemokines and chemokine receptor-expressing cells in both the hematopoietic and non-hematopoietic compartments. In this study, we demonstrate a central role for chemokine receptor CXCR3 in empowering Th1 trafficking to the small intestine, in turn enabling inflammatory monocyte activation and concomitant control of infection. While the importance of IFN-γ-secreting CD4^+^ T cells in resistance to *Toxoplasma* is well known to researchers in the field [@ppat.1003706-SchartonKersten1], [@ppat.1003706-Gazzinelli2], [@ppat.1003706-Suzuki1], and while the importance of anti-microbial inflammatory monocytes has recently become clear in the context of *Toxoplasma* and other infections [@ppat.1003706-Dunay1], the present study is the first to reveal the functional link between CXCR3^+^ T-cell effectors, IFN-γ and inflammatory monocyte activation in tissues of the intestinal mucosa. Although Th1 effectors depend upon CXCR3 to reach the site of infection, inflammatory monocytes require chemokine receptor CCR2 for optimal trafficking. In the latter case, inflammatory monocytes fail to exit the bone marrow in *Ccr2^−/−^* mice, resulting in a decreased level of this population in the periphery, in turn resulting in inability to control *T. gondii* infection [@ppat.1003706-Dunay1]. Monocytes have also been suggested to promote the systemic dissemination of *T. gondii* to the brain [@ppat.1003706-Courret1]. While brain parasite loads were not examined in this study, it is unlikely that altered parasite shuttling is a mechanism by which the knockout animals are succumbing to *Toxoplasma* because peripheral parasite loads were not affected by absence of CXCR3 ([Fig. 3B](#ppat-1003706-g003){ref-type="fig"}). Our data argue that the basis for increased parasites specifically in the intestine is the result of defective regional control by inflammatory monocytes that lack CXCR3-dependent activation signals. Recent data indicate that inflammatory monocyte expression of CCR1 enables a response to IL-15-dependent CCL3-secreting innate lymphoid cells, resulting in CCR1-dependent recruitment to the intestinal mucosa of *Toxoplasma* infected mice [@ppat.1003706-Schulthess1]. Taking these data and ours collectively, we propose that CXCR3, CCR2 and CCR1 act together as a control axis of innate and acquired immunity in intestinal immunity, ensuring coordinated recruitment of inflammatory monocytes and Th1 effectors to inflamed tissues. It was recently demonstrated that NK cell-derived IFN-γ controls the differentiation of circulating monocytes into inflammatory dendritic cells during i. p. *T. gondii* infection, and is thus required for an optimal IL-12 response [@ppat.1003706-Goldszmid1]. In our model we did not see a dependence on CXCR3 for NK cell recruitment, as NK cells appeared to lose CXCR3-GFP expression over the course of infection ([Fig. S1](#ppat.1003706.s001){ref-type="supplementary-material"}), and production of NK cell IFN-γ was equivalent in WT and KO mice ([Fig. S5D](#ppat.1003706.s005){ref-type="supplementary-material"}--F). This difference may be attributable to the alternative routes of infection used, as our model incorporates the intestinal response, while the i.p. route bypasses the intestinal mucosa. Consistent with this idea, absence of CXCR3 did not affect the ability of mice to survive i. p. infection ([Fig. S3A](#ppat.1003706.s003){ref-type="supplementary-material"}). While our study focuses on the early response to *Toxoplasma* infection in the intestinal mucosa, others have examined the role of CXCR3 and its ligands in additional tissues and at different stages of infection. Antibody-mediated depletion of CXCL10, a major CXCR3 ligand, increases susceptibility and blocks influx and expansion of T cells in the liver and spleen that accompanies *T. gondii* infection [@ppat.1003706-Khan1]. Additionally, a study of ocular toxoplasmosis revealed that T cells infiltrating the eye during infection express CXCR3 and produce IFN-γ. Depletion of CXCL10 in this model reduced the number of infiltrating T lymphocytes during chronic infection, resulting in increased parasite replication and ocular damage [@ppat.1003706-Norose1]. Recently, CXCL10 was shown to impact CD8^+^ T-cell mobility in the brain of chronically infected mice, enhancing their ability to control the parasite by increasing contact with infected cells [@ppat.1003706-Harris1]. Our results for the first time highlight the importance of CXCR3 and its impact on CCR2-dependent monocytes in the initial protective response to the parasite in the intestine. It has been shown that TGF-β production by intestinal IEL protects against *T. gondii*-induced damage by down-modulating inflammation [@ppat.1003706-BuzoniGatel1]. While we did not examine intraepithelial lymphocytes (IEL) in this study, it is possible that CXCR3 expression could also affect trafficking and function of this cell type, thereby contributing to resistance in this model. Future studies will allow us to determine whether CXCR3-expressing IEL play a role in immunity during intestinal *T. gondii* infection. CXCR3 has also been assessed for its role in immunity to other protozoan pathogens, including *Leishmania* and *Plasmodium*. Interestingly, the function of CXCR3 differs depending upon the parasite, the route of infection and the site examined. For example, *Cxcr3^−/−^* mice exhibit impaired IFN-γ production and increased lesion development during cutaneous *L. major* infection, but the knockout mice are not more susceptible to hepatic *L. donovani* infection [@ppat.1003706-Rosas1], [@ppat.1003706-Barbi1]. Furthermore, CXCR3 and its chemokines promote cerebral inflammation and mortality during experimental malaria infection [@ppat.1003706-Campanella1], [@ppat.1003706-Nie1]. Overall, CXCR3 and its chemokine ligands function as double-edged swords, inasmuch as they make an important contribution to protective immunity, but when dysregulated they are the cause of deleterious immunopathology. In addition to its role in cell recruitment, CXCR3 has recently been suggested to be important for priming CD4^+^ T cells in the lymph node to become Th1 cells by promoting long-lasting interactions between T cells and CXCL10-expressing dendritic cells. In the absence of CXCR3, T cells fail to fully differentiate into IFN-γ-producing cells and are defective during subsequent lymphocytic choriomeningitis virus (LCMV) infection [@ppat.1003706-Groom2]. We found no evidence for defective Th1 responses in secondary lymphoid organs during *Toxoplasma* infection of *Cxcr3^−/−^* mice, a result that is supported by similar findings during *L. major* infection [@ppat.1003706-Rosas1]. However, our results are consistent with impaired recruitment of Th1 cells in the absence of CXCR3, as lamina propria CD4^+^ T cells from *Cxcr3^−/−^* mice exhibited defective IFN-γ production. We conclude that the function of CXCR3 in promoting Th1 differentiation versus, or in addition to, homing to inflammatory sites is likely to be a context-dependent phenomenon. Although we found no differences in T-cell activation in CXCR3 negative T cells of reporter mice or in CXCR3 KO animals, as measured by expression of CD25, CD69 and CD44 (data not shown), we observed a consistent decrease in CD27 expression amongst CXCR3-negative T cells of reporter mice. Overall levels of CD27 expression were also lower in T cells from *Cxcr3^−/−^* mice (data not shown). CD27 is a member of the TNF receptor super family that has been implicated as a T-cell costimulatory molecule [@ppat.1003706-Hendriks1]. CD27 expression is thought to characterize naïve or memory T cells, whereas loss of CD27 represents terminal differentiation [@ppat.1003706-Hamann1]. Accordingly, it is possible that in addition to controlling T-cell recruitment to sites of infection, *Cxcr3^−/−^* T cells may undergo terminal differentiation and, possibly, premature Th1 effector death in tissues prior to mediating IFN-γ dependent inflammatory monocyte activation. This is consistent with studies showing that CD27^low^ cells are more susceptible to apoptosis and that accumulation of influenza-specific T lymphocytes was impaired in the lungs of *Cd27^−/−^* mice during infection [@ppat.1003706-Hendriks1], [@ppat.1003706-Kapina1]. In our study, CD27 expression was not rescued by adoptive transfer of WT CD4^+^ T cells (data not shown), which may suggest that increasing the T-cell pool allows the system to cross a certain threshold of T-cell levels in order to activate inflammatory monocytes without reversing CD27 expression. In the absence of CXCR3, we found that lamina propria neutrophil levels were increased during infection, as was their activation status as measured by TNF-α expression. Furthermore, adoptive transfer of WT CD4^+^ T lymphocytes into the KO strain reversed increased levels of PMN as well as their production of TNF-α. Because this inflammatory cytokine has been linked to intestinal damage during *Toxoplasma* infection [@ppat.1003706-Liesenfeld1], it seems likely that CXCR3-dependent effects on neutrophils are likely to be secondary to loss of inflammatory monocyte function. In this scenario, defects in monocyte-mediated parasite killing would result in damage to the intestinal mucosa. Translocation of luminal gut flora, known to contribute to emergence of parasite-induced intestinal lesions [@ppat.1003706-Heimesaat2]--[@ppat.1003706-Egan1], would in turn be expected to result in local neutrophil recruitment. Indeed, based on neutrophil depletion studies it has been suggested that these cells mediate damage to the intestinal mucosa during *T. gondii* infection [@ppat.1003706-Dunay2]. The cellular and molecular basis for this effect is not at present known, but both the IL-17/IL-23 axis and CXCL8 have been shown to promote neutrophil accumulation in infected tissues suggesting involvement of one or both of these mediators [@ppat.1003706-DelRio1]--[@ppat.1003706-Murphy1]. The results of this study extend our understanding of immunity in the intestinal mucosa, which has become increasingly important as inflammatory bowel diseases (IBD), such as ulcerative colitis and Crohn\'s disease, become more common in developed regions of the world. In this regard, abnormally high levels of CXCR3 are associated with dysregulated intestinal responses in human IBD patients, underscoring the potential hazards of unbalanced inflammatory responses [@ppat.1003706-Papadakis1]--[@ppat.1003706-Schroepf1]. While CXCR3 can be pathogenic by recruiting effector cells to otherwise healthy tissue, as in IBD or cerebral malaria, we show here that CXCR3-expressing T cells play an essential protective role in host defense by enabling defense against pathogenic organisms. Antibodies against CXCL10 have been suggested as potential therapeutic agents against IBD [@ppat.1003706-Nishimura1], [@ppat.1003706-Singh1]. The results from this study, however, highlight the possible harm of inhibiting CXCR3-expressing cells into sites of inflammation during infection with an enteric pathogen. As the first study to demonstrate a protective role for CXCR3^+^CD4^+^ T cells in the intestinal immune response, we have shown here that failure to appropriately recruit these T cells results in impaired inflammatory monocyte activation, accumulation of intestinal parasites, and subsequent recruitment of potentially pathogenic TNF-α-secreting neutrophils. Our results reveal CXCR3 as a critical chemokine receptor of the adaptive immune system that ensures appropriate placement of T cells in inflamed tissue, enabling inflammatory monocytes of innate immunity to acquire effector functions and mediate effective host defense. Materials and Methods {#s4} ===================== Ethics statement {#s4a} ---------------- All experiments in this study were performed strictly according to the recommendations of the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocols were approved by the Institutional Animal Care and Use Committee at Cornell University (permit number 1995--0057). All efforts were made to minimize animal suffering during the course of these studies. Mice and infections {#s4b} ------------------- Female Swiss Webster mice (6--8 weeks of age) were purchased from the Jackson Laboratory (Bar Harbor, ME). *Cxcr3^−/−^* and *Cxcr3* eGFP knock-in reporter mice [@ppat.1003706-Oghumu1] were established as breeding colonies in the Transgenic Mouse Facility at the Cornell University College of Veterinary Medicine. The CXCR3 internal ribosomal entry site bicistronic eGFP reporter (CIBER) mice were generated as described [@ppat.1003706-Oghumu1]. This strain possesses a functional CXCR3 receptor, and all CXCR3 positive cells also express intracellular eGFP. Mouse infections were initiated by oral inoculation of cysts of the type II *T. gondii* ME49 strain. Cysts were isolated from chronically infected Swiss Webster mice by homogenization of whole brain in sterile PBS. Unless stated otherwise, mice were infected at 8--12 weeks of age with 30 cysts. Tissue staining and immunofluorescence microscopy {#s4c} ------------------------------------------------- Intestines were excised, flushed with 10% neutral-buffered formaldehyde, and embedded in paraffin for sectioning. Sections were stained for hematoxylin and eosin for assessment of pathological changes. Sections were also stained for parasite antigen by immunohistochemistry at the Cornell Animal Health Diagnostic Center. Frozen sections were obtained by embedding 1 cm lengths of intestine in OCT. Sections of 6--8 µm were cut on a cryostat, fixed in ice-cold acetone, and blocked with PBS containing 2× casein and goat serum. To examine T-cell infiltration, sections were incubated with rat anti-CD4 antibody (GK1.5) (ATCC, Manassas, VA) or rat IgG at 4°C overnight followed by goat anti-rat Alexa-647 secondary antibody (Life Technologies, Grand Island, NY). Sections were then mounted in DAPI-Prolong antifade (Life Technologies) and imaged by confocal microscopy. ImageJ software was used to analyze fluorescence of independent channels. Cell isolation {#s4d} -------------- Spleens and mesenteric lymph nodes were excised, crushed between sterile slides, and passed through a 70-µm filter (BD, Franklin Lakes, NJ). Red blood cells from splenocyte suspensions were lysed with ACK lysis buffer (Life Technologies). For LP leukocyte isolation, the small intestine was removed, cleaned of mesentery, flushed with sterile PBS, and cleared of Peyer\'s patches. The intestine was opened longitudinally and the mucosal layer was scraped with a blunt scalpel to remove epithelial cells. The tissue was cut into 5 mm sections and vigorously washed with Dulbecco\'s modified Eagle\'s media (Cellgro, Manassas, VA) and 5 mM EDTA (Life Technologies). Cells were liberated from the intestinal tissue by digestion with 10 mg/ml collagenase (Sigma, St. Louis, MO) at 37°C and subsequently passed through a 70-µm filter. Cytokine measurement {#s4e} -------------------- Secretion of IFN-γα, TNF-α, and IL-10 was assayed by ELISA (eBioscience, San Diego, CA) following manufacturer\'s instructions in the presence of soluble tachyzoite antigen (STAg) prepared as previously described [@ppat.1003706-Denkers2]. IL-12p40 was quantitated using an in-house ELISA [@ppat.1003706-Butcher1]. For ileum biopsy cultures, 1 cm intestinal sections were flushed with PBS, opened longitudinally, and cultured overnight in complete Dulbecco\'s modified Eagle\'s media supplemented with 10% bovine growth serum (Hyclone), 0.05 mM β-mercaptoethanol (Sigma), 1 mM sodium pyruvate, 0.1 mM nonessential amino acids, 10,000 U/ml penicillin, 10,000 µg/ml streptomycin, and 30 mM HEPES (reagents from Life Technologies). Supernatants were collected and assayed for cytokine by ELISA. Measurement of mRNA by quantitative PCR {#s4f} --------------------------------------- RNA was isolated from MLN and intestinal tissue from mice over a time course of infection. Tissue was initially disrupted with a tissue homogenizer and subjected to RNA isolation following manufacturer\'s instructions (Tissue RNA Kit, Omega Biotek, Norcross, GA). RNA was converted to cDNA (Quantas Biosciences, Gaithersburg, MD) and assayed for gene expression by SYBR green technology (Quanta Biosciences). Primers were designed to span exons by Integrated DNA Technologies. GAPDH was used as a housekeeping gene. Gene expression from each timepoint was normalized to uninfected control samples. Flow cytometry {#s4g} -------------- Single cell suspensions were pelleted and resuspended with primary antibodies (BioLegend, San Diego, CA: anti-CD4 PerCP, anti-CD8α APC-Cy7, anti-CD45 Alexa-488, anti-CD11b APC-Cy7, anti-CD11b APC, anti-Ly6G FITC; eBioscience, San Diego, CA: anti-CD25 PE, anti-CD45 APC or FITC, anti-CD69 PE, anti-Ly6C/G APC; BD Biosciences, San Jose, CA: anti-Ly6G PE-Cy7, anti-Ly6C V420, anti-CD45 PE, anti-Ly6C/G PerCP, anti-CD44 APC) in ice-cold FACS buffer (1% bovine serum albumin/0.01% NaN~3~ in PBS) for 30 min. For IFN-γ staining, cells were incubated for 6 hrs with Brefeldin-A (eBioscience; 10 ug/ml), PMA (Sigma; 10 ng/ml), and ionomycin (Sigma; 1 ug/ml), then fixed with 4**%** paraformaldehyde and subsequently incubated with primary antibodies resuspended in the FoxP3/transcription factor buffer staining set (eBioscience). For IL-12 and TNF-α staining, cells were incubated for 6 hrs with Brefeldin-A only (eBioscience). Intracellular staining experiments used 10^6^ cells. Antibodies used for intracellular staining included anti-IFN-γ PE-Cy7, anti-TNF-α PE-Cy7 (Biolegend); anti-IL-12 PE, anti-TNF-α APC (BD Biosciences), and anti-*Toxoplasma* p30 (Argene, Shirley, NY). Cell fluorescence was measured using a FACS Canto (BD Biosciences). Data was analyzed using FlowJo software (FlowJo, Ashland, OR). Quantitative PCR for parasite burden {#s4h} ------------------------------------ DNA was isolated from whole intestinal tissue using a Tissue DNA kit following manufacturer\'s instructions (Omega Biotek, Norcross, GA). DNA was amplified by quantitative PCR as described previously using primers against the *T. gondii* B1 gene and the host argininosuccinate lyase (ASL) gene [@ppat.1003706-Butcher2]. Ten-fold serial dilutions of genome copy standard curves were created using known quantities of host (splenocytes) and parasite (tachyzoites) cells based on DNA quantity, Avogadro\'s number, and genome size. To quantify parasite burden, the generated values for host and parasite genome copies from the DNA preparations were expressed as a ratio of parasite (B1) to host (ASL) genomes. Adoptive transfer {#s4i} ----------------- Splenocytes from naïve mice were harvested and subjected to CD4 positive selection by magnetic bead sorting following manufacturer\'s instructions (Stem Cell Technologies, Vancouver, British Columbia). Cells were purified to ∼90--95% purity and were transferred by intravenous retro-orbital injection into *Cxcr3^−/−^* recipients at 5×10^6^ CD4^+^ cells per mouse. Twenty-four hours post transfer, mice were challenged with 30 cysts of the *T. gondii* ME49 strain. In some experiments mice were left to assess survival following cell transfer. In other experiments intestinal tissue was collected at day 9 post-infection for intracellular cytokine analysis. Pathology scoring {#s4j} ----------------- Swiss rolls of the intestines were histopathologically scored by an investigator that was double-blinded to sample identity. Intestines were scored on an ascending 0--4 scale as previously described [@ppat.1003706-Egan1], [@ppat.1003706-Johnson1]. Briefly, scores of 0 were normal, scores of 1 indicated mild focal lesions, scores of 2 indicated moderate focal lesions, scores of 3 indicated moderate multifocal lesions, and scores of 4 indicated severe multifocal lesions. Histopathological features scored included: inflammation of the intestinal submucosa (lamina propria), inflammation extending throughout all histological layers of the intestine (transmural inflammation), sloughing of intestinal epithelium, intestinal villus fusion and blunting, and necrosis of villi. Sections of complete small intestines from 13 KO and 14 WT infected mice were scored. The data are plotted as mean scores of individual mice. Statistics {#s4k} ---------- Differences between groups were analyzed using student\'s *t*-test. Differences between 3 or more groups were analyzed using one-way Anova with Newman-Keuls post-test. P-values less than 0.05 are considered significant and are designated by \* p\<0.05, \*\* p\<0.01, or \*\*\* p\<0.001. Pathology scores were analyzed using the Mann-Whitney *t*-test. Supporting Information {#s5} ====================== ###### **CXCR3-GFP expression on NK cells decreases during infection.** Lamina propria and mesenteric lymph node (MLN) leukocytes were isolated from noninfected (NI) and Day 7-infected (INF) CIBER reporter mice, and CXCR3-GFP expression was measured on NK1.1^+^ cells. Numbers in each panel indicate the percent of NK1.1^+^ cells falling within the CXCR3-eGFP positive quadrant. (TIF) ###### Click here for additional data file. ###### **CXCR3-GFP^+^ and CXCR3-GFP^−^ fractions of CD4^+^ T cells differentially express activation markers.** Splenocytes (n = 4), MLN (n = 5), and Peyer\'s patch (n = 2) cells were isolated from Day 11-infected CIBER reporter mice and stained for surface markers CD27 as well as CD4 (A and B) and CD8 (C and D). Representative mice are shown in A and C. Bar graphs (B and D) represent averages of multiple mice, and significance is represented by \* p\<0.05, and \*\*\* p\<0.001. (TIF) ###### Click here for additional data file. ###### ***Cxcr3^−/−^*** **mice are more susceptible to oral infection with** ***Toxoplasma*** **.** (A) WT and KO mice (n = 5 per group) were infected by i.p. injection of 30 ME49 cysts, and survival was monitored. (B) Small intestines of naïve WT and KO mice (n = 3 per strain) were harvested, fixed in formaldehyde, embedded in paraffin, and sections were stained with H&E. (C) Frozen sections of Day 10 WT and *Cxcr3^−/−^* intestines were stained for Muc1 followed by anti-rabbit Alexa-488 (red). Sections were counter-stained with DAPI (blue). (D) Lamina propria leukocytes from Day 9-orally infected WT and *Cxcr3^−/−^* mice were stained for neutrophil markers Ly6C/G (Gr-1) and Ly6G (1A8). (E) Neutrophil levels were assessed in individual mice (n = 5 per group). The graph shows mean +/− SEM (\*\* p\<0.01). (TIF) ###### Click here for additional data file. ###### **Cytokine responses in WT and KO mice.** Splenocytes (A, C and E) and MLN (B, D and F) were harvested from Day 11-infected WT and *Cxcr3^−/−^* mice and cultured in the presence of soluble tachyzoite antigen (STAg) for 72 hr. Supernatants were collected, and IFN-γ (A and B), TNF-α (C and D), and IL-10 (E and F) were measured by ELISA. (TIF) ###### Click here for additional data file. ###### **T cell and NK cell production of IFN-γ in the presence and absence of CXCR3.** Lamina propria leukocytes were isolated from WT (A) and *Cxcr3^−/−^* (B) mice 6 days post-infection, cultured for 6 hr in the presence of PMA, ionomycin, and Brefeldin-A, and stained for CD8 and IFN-γ. (C) Means and standard errors of individual mice. Lamina propria leukocytes were isolated from WT (D) and *Cxcr3^−/−^* (E) mice, cultured in the presence of PMA, ionomycin, and Brefeldin-A and stained for NK1.1 and IFN-γ. (F) Mean and standard error of individual mice (WT: n = 9; KO: n = 6). Splenocytes were harvested from WT (G) and *Cxcr3^−/−^* (H) Day 7-infected mice, cultured in the presence of PMA, ionomycin, and Brefeldin-A, and stained for CD4 and IFN-γ. (I) Mean and standard error of individual mice (WT: n = 2; KO: n = 2). (TIF) ###### Click here for additional data file. ###### **Lamina propria neutrophils secrete elevated TNF-α in the absence of CXCR3.** Lamina propria leukocytes were harvested from WT and *Cxcr3^−/−^* mice and cultured in the presence of Brefeldin-A for 6 hr. Cells were then stained for neutrophil markers CD11b, Ly6C/G (Gr-1), and Ly6G (1A8), fixed, permeabilized, and intracellularly stained for TNF-α. Cytokine production was assessed by flow cytometry (A). Shown are the mean +/− SEM of individual mice (B) (n = 5 per group, \*\* p\<0.01). (TIF) ###### Click here for additional data file. ###### **Adoptive transfer of CXCR3^+^CD4^+^ T cells into** ***Cxcr3^−/−^*** **recipients protects against oral** ***Toxoplasma*** **infection.** (A) Splenic CD4^+^ T cells were isolated from naive *Cxcr3^−/−^* mice, and 5×10^6^ cells were adoptively transferred intravenously into *Cxcr3^−/−^* recipients. Mice were orally challenged 24 hr later with 30 ME49 cysts and assessed for survival (n = 5 mice per group). (B) Lamina propria leukocytes were harvested from infected WT and *Cxcr3^−/−^* mice and cultured in the presence of Brefeldin-A for 6 hr. Cells were surface stained for CD11b, Ly6C/G (Gr-1), and Ly6G to identify inflammatory monocytes. Cells were then fixed, permeabilized, and intracellularly stained for IL-12 and TNF-α. Cytokine production by inflammatory monocytes was analyzed by flow cytometry. Shown are representative FACS plots of individual mice. (C) Lamina propria leukocytes were surface stained for CD11b, Ly6C/G (Gr-1), and Ly6G (1A8) to identify neutrophils. (D) Cells were then fixed, permeabilized, and intracellularly stained for TNF-α. Cytokine production by neutrophils was analyzed by flow cytometry. Shown are the mean +/− SEM of individual mice (n = 5 mice per group; \* p\<0.05, \*\* p\<0.01). (TIF) ###### Click here for additional data file. We thank M. Hossain and A. Lee for expert technical assistance and B. Butcher for useful discussions. [^1]: The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: SBC EYD. Performed the experiments: SBC CEE. Analyzed the data: SBC KJM CEE ARS EYD. Contributed reagents/materials/analysis tools: SO ARS. Wrote the paper: SBC EYD. [^3]: Current address: Center for Comparative Medicine and Research, Dartmouth College, Lebanon, New Hampshire, United States of America [^4]: Current address: Department of Surgery, Division of Pediatric Surgery, Children\'s Hospital of Pittsburgh of UPMC and the University of Pittsburgh, Pittsburgh, Pennsylvania, United States of America

Our individual data points can not be publicly deposited as supporting information due to ethical restrictions imposed by Nagoya University International Bioethics Committee, since the data contains the sensitive personal information. Information regarding our data can be requested at the following e-mail address: Kazunori Hashimoto (e-mail: <khashim@med.nagoya-u.ac.jp>). Introduction {#sec001} ============ Exposure to arsenic (As) via drinking water is a health risk for humans \[[@pone.0198743.ref001]--[@pone.0198743.ref003]\]. Previous studies showed that exposure of humans to As was associated with cancer of the bladder, kidney, skin, prostate, lung and liver \[[@pone.0198743.ref004]\] and with cardiovascular disease \[[@pone.0198743.ref005]\]. It was also shown that exposure to As was associated with neurological diseases including cognitive impairment in children \[[@pone.0198743.ref006], [@pone.0198743.ref007]\] and adults \[[@pone.0198743.ref008]\] and with diabetes mellitus \[[@pone.0198743.ref009]--[@pone.0198743.ref012]\]. Levels of toxic elements in noninvasive biological samples including nails, hair and urine have been determined to investigate associations with hearing loss in humans \[[@pone.0198743.ref013], [@pone.0198743.ref014]\]. A univariate analysis showed a significant correlation of As levels in fingernails with amplitude of distortion product otoacoustic emission (DPOAE) at 2 kHz, which reflects activity levels of outer hair cells, in 30 subjects aged 9--78 years residing in the gold mining community of Bonanza, Nicaragua \[[@pone.0198743.ref015]\]. Thus, it is possible that As in nails is a reliable index for As-mediated hearing loss in humans. However, there is no direct evidence showing correlations between As levels in toenails and hearing levels. In our previous study using pure tone audiometry (PTA), an association of oral exposure to As via drinking water with hearing loss in young people was shown by multivariate analysis with adjustments for confounders including age, smoking, sex and BMI \[[@pone.0198743.ref016]\]. However, multivariate analysis has not been performed to demonstrate the association of As levels in nails with hearing loss in humans, although age and smoking are known to be strong factors affecting hearing levels in humans \[[@pone.0198743.ref014], [@pone.0198743.ref017]--[@pone.0198743.ref019]\]. In experimental studies, oral exposure to As has been shown to cause several health risks including carcinogenesis \[[@pone.0198743.ref020]\] and cardiovascular diseases \[[@pone.0198743.ref021]\]. Levels of toxic elements in inner ears, which are known to be the sensory organ for hearing, have been determined to demonstrate the correlation with hearing loss \[[@pone.0198743.ref022], [@pone.0198743.ref023]\]. Exposure of guniea pigs to As at 200 mg/L via intraperitoneal injection for 2 months resulted in morphological impairments of the inner ear \[[@pone.0198743.ref024]\]. Our previous study has showed that exposure of young mice to As via drinking water caused accumulation of As in inner ears, resulting in hearing loss \[[@pone.0198743.ref016]\]. However, it is not clear whether there is a correlation between As levels in inner ears and nails in mice. In this study, we epidemiologically and experimentally investigated whether As levels in nails are reliable as non-invasive indexes for As-mediated hearing loss in humans and whether As levels in nails are associated with As levels in inner ears in mice. Materials and methods {#sec002} ===================== Epidemiological study {#sec003} --------------------- We obtained information on age, sex, smoking history, weight and height in 145 healthy subjects aged from 12 to 55 years by self-reported questionnaires in Bangladesh as previously described \[[@pone.0198743.ref014]\]. We obtained informed consent in written form from all of the subjects. For subjects aged 12 to 18 years, we obtained consent in written form from their parents. The subjects were only Bangladeshi people who lived mainly in rural areas. The subjects living in one area drink tap water and the rest of the subjects living in another area drink tube well water contaminated with As. The subjects included students, housewives and businessmen. The subjects did not have portable music players with earphones and had no occupational exposure to wielding fumes. None of the subjects had a habit of drinking alcohol. We excluded only subjects who had a history of ear diseases or illness at the time of the survey. We calculated body mass index (BMI) by diving weight (kg) by the square of height (m^2^) and used definitions of underweight (\< 18.5 kg/m^2^), normal weight (18.5--24.9 kg/m^2^) and overweight (≥ 25 kg/m^2^) set by the WHO \[[@pone.0198743.ref025]\]. We set the mean value of age (29.6 years) of the subjects in this study as the cut-off value for age. We collected 0.1--2 cm samples of toenails from each subject by using a ceramic nail clipper. We also collected urine samples from each subject and stored the samples in 50 ml sterile tubes at -80°C until measurements. We determined the hearing levels at 1, 4 and 8 kHz in addition to 12 kHz of the participants by PTA, since hearing level at 12 kHz was shown to be sensitive to environmental stresses \[[@pone.0198743.ref014], [@pone.0198743.ref017], [@pone.0198743.ref026]\]. We used an iPod with earphone-type headphones (Panasonic RP-HJE150) in a soundproof room as described previously \[[@pone.0198743.ref027], [@pone.0198743.ref028]\]. We output pure tones at each frequency to a subject until the subject recognized the sound. We stood behind the subject and provided the subject with an initial stimulus of 5 decibels (dB) and then increased the sound level by 5 dB step-by-step. The subject raised their hand when they recognized the sound. We evaluated the sound level recognized by each subjects as hearing threshold. Pure tones at each frequency from the earphones in the soundproof room were verified by using a noise level meter (Type 6224 with an FFT analyzer, ACO CO., LTD, Japan). The epidemiological study was approved by Nagoya University International Bioethics Committee following the regulations of the Japanese government (approval number: 2013--0070) and the Faculty of Biological Science, University of Dhaka (Ref. no. 5509/Bio.Sc). Experimental study {#sec004} ------------------ Hairless mice having the C57BL/6J background (1 month old, female, body weight of 10--15 g) were procured from Hoshino Laboratory Animal, Inc. Three or four mice were housed in a cage under super pathogen-free (SPF) conditions with a standard temperature of 23 ± 2°C and a 12-h light/dark cycle. The mice were fed a standard rodent chow (Clea Rodent Diet CE-2). Neither randomization nor blinding investigation were used in this animal study. In brief, we exposed mice (n = 7) to 22.5 mg/L of sodium arsenite (NaAsO~2~, Sigma-Aldrich) dissolved in distilled water for 2 months via drinking water and changed the drinking water every week as previously described \[[@pone.0198743.ref016]\]. The exposure dose was based on the As exposure for mice at 10 and 100 ppm via drinking water in a previous study \[[@pone.0198743.ref029]\]. The control group (n = 6) was given just distilled water. After exposure for 2 months, mice were anesthetized with isoflurane and sacrificed by decapitation. Nails and inner ears were collected and kept in a 1.5 ml tube. For the collection of inner ears, we first identified temporal bones at the bottom of the skull and then carefully removed the cranial nerves and tissues using standard forceps. The inner ears were dislodged by pushing down on the posterior semicircular canal with the thumb of a hand and fixing the tip region of the otic bone capsule with standard forceps. After carefully removing extra tissues adhered to the inner ears, we used the inner ears for ashing. The experimental study was approved by the Institutional Animal Care and Use Committee in Nagoya University (approval number: 28251) and followed the Japanese Government Regulations for Animal Experiments. Measurement of As levels in biological samples {#sec005} ---------------------------------------------- As levels in biological samples including toenails and urine were measured by inductively coupled plasma mass spectrometry (ICP-MS; Agilent 7500cx) as described previously \[[@pone.0198743.ref014], [@pone.0198743.ref022], [@pone.0198743.ref023]\]. In brief, toenails were washed with detergent water. One or two drops of acetone were added and then the samples were kept at room temperature until starting the ashing. Samples were ashed by incubation in 3 ml of HNO~3~ at room temperature overnight followed by incubation at 80°C for 3 hours. Samples were further incubated in 1--1.5 ml of H~2~O~2~ at 80°C for 3 hours. After cooling, Milli-Q water was added to the samples to adjust the final volume to 5 ml. In the case of urine, 4 ml of urine was incubated in 1 ml of HNO~3~ at room temperature overnight followed by incubation at 80°C for 24--72 hours. After cooling, the ashed urine samples were centrifuged at 2,000 rpm (= 269 g) for 1 min and Milli-Q water was added to the samples to adjust the final volume to 5 ml. Levels of As in urine were normalized by specific gravity using the following formula: SG-corrected concentration = raw hormone concentration × \[(SG~target~− 1.0)/(SG~sample~---1)\], where SG~target~ is a population mean SG \[[@pone.0198743.ref030]\]. In this study, the mean SG was 1.012 for the subjects. In the case of murine samples, nails were rinsed with Milli-Q water 3 times and air-dried at room temperature. Ashing of the samples was then performed with the same protocol as that described above. Levels of As in biological samples were determined by ICP-MS. Statistical analysis for epidemiological study {#sec006} ---------------------------------------------- All statistical analyses were performed by JMP software (version 11.0.0). In univariate analysis, the Mann-Whitney *U* test and Steel-Dwass test were used to detect significant differences in hearing levels between the groups because the hearing levels were discontinuous variables. In multivariate analysis, binary logistic regression analysis was performed with adjustments for age, sex, smoking and BMI as confounding factors. We used the method in a previous study \[[@pone.0198743.ref026]\] to categorize subjects with an auditory threshold higher than the cut-off value at each frequency as hearing loss. We set hearing thresholds \[1 kHz (≥ 7 dB), 4 kHz (≥ 10 dB), 8 kHz (≥ 24 dB), and 12 kHz (≥ 45 dB)\] with the mean values of hearing levels at each frequency to divide the subjects into two groups. In this study, values of p \< 0.05 were considered statistically significant. Statistical analysis for experimental study {#sec007} ------------------------------------------- In our experimental study, Spearman's correlation coefficients were used to evaluate the association between As levels in inner ears and those in nails. Values of p \< 0.05 were considered statistically significant. Results {#sec008} ======= Characteristics of the study participants {#sec009} ----------------------------------------- [Table 1](#pone.0198743.t001){ref-type="table"} shows the characteristics of the subjects, which were described in our previous report \[[@pone.0198743.ref014]\]. The hearing thresholds in the older group (≥ 29.6 years old, n = 68) were significantly higher than those in the younger group (\< 29.6 years old, n = 77) at 1 kHz (p = 0.0098), 4 kHz, 8 kHz and 12 kHz (p \< 0.0001) ([Table 2](#pone.0198743.t002){ref-type="table"}). The hearing thresholds in females (n = 76) were significantly higher than those in males (n = 69) at 4 kHz (p = 0.0257), 8 kHz (p = 0.0004) and 12 kHz (p = 0.0066) ([Table 2](#pone.0198743.t002){ref-type="table"}). No significant differences were found in hearing thresholds among the three BMI groups ([Table 2](#pone.0198743.t002){ref-type="table"}). Smokers (n = 31) had higher hearing thresholds than those in non-smokers (n = 114) at 1 kHz (p = 0.0057), 4 kHz (p \< 0.0001), 8 kHz (p = 0.0002) and 12 kHz (p \< 0.0001) ([Table 2](#pone.0198743.t002){ref-type="table"}). 10.1371/journal.pone.0198743.t001 ###### Characteristics of the study participants. {#pone.0198743.t001g} ------------------------------------------------------------------ Characteristics mean ± SD Variables Participants\ \[n (%)\] ----------------- ------------- ------------------ --------------- Age 29.6 ± 11.0 \< 29.6 77 ≥ 29.6 68 Sex Male 69 Female 76 Smoking history No 114 Yes 31 BMI 22.0 ± 3.4 \< 18.5 23 18.5 ≤ BMI \< 25 94 ≥ 25 28 ------------------------------------------------------------------ 10.1371/journal.pone.0198743.t002 ###### Associations between hearing levels and confounding factors including age, BMI, sex and smoking. {#pone.0198743.t002g} Hearing levels (mean ± SD) ------------------ ---------------------------- --------------- --------------- --------------- Age \< 29.6 6.10 ± 3.50 7.01 ± 4.39 19.29 ± 9.79 35.39 ± 15.64 ≥ 29.6 7.43 ± 4.69 13.31 ± 9.60 29.93 ± 12.53 56.69 ± 22.15 p value 0.0098 p \< 0.0001 p \< 0.0001 p \< 0.0001 Sex Male 6.59 ± 3.88 8.84 ± 6.81 21.23 ± 12.08 40.94 ± 21.61 Female 6.84 ± 4.38 10.99 ± 8.76 27.04 ± 11.98 49.41 ± 21.13 p value 0.7697 0.0257 0.0004 0.0066 Smoking No 6.40 ± 3.85 8.42 ± 5.70 22.06 ± 10.56 40.70 ± 18.38 Yes 7.90 ± 4.96 15.65 ± 11.74 32.42 ± 14.94 62.58 ± 24.49 p value 0.0057 p \< 0.0001 0.0002 p \< 0.0001 BMI \< 18.5 6.52 ± 4.38 7.17 ± 3.64 19.78 ± 5.93 42.17 ± 16.91 18.5 ≤ BMI \< 25 6.49 ± 3.35 10.05 ± 6.66 24.26 ± 12.11 44.95 ± 21.61 ≥ 25 7.68 ± 6.01 11.96 ± 12.72 28.04 ± 15.65 49.46 ± 25.40 p value 0.6148 0.1797 0.1346 0.6595 Association between As levels in biological samples and hearing levels by univariate analysis {#sec010} --------------------------------------------------------------------------------------------- Concentrations of As (mean ± SD) in toenails and urine were 1.38 ± 1.17 μg/g and 90.27 ± 103.04 μg/L, respectively ([Table 3](#pone.0198743.t003){ref-type="table"}). In this study, we set the cut-off values based on the receiver operating characteristic (ROC) curve and the highest Youden index \[[@pone.0198743.ref031]\]. We categorized the subjects into two groups at 0.60 μg/g in toenails and 76.12 μg/L in urine ([Table 3](#pone.0198743.t003){ref-type="table"}). The mean ages of subjects in the high and low As groups were 31 years and 27 years, respectively. We found that hearing thresholds were significantly higher in the high As group in toenails (n = 97) at 4 kHz (mean = 10.48 dB; p = 0.0023), 8 kHz (mean = 26.71 dB; p \< 0.0001) and 12 kHz (mean = 51.64 dB; p \< 0.0001) than those in the low As group (n = 48) at 4 kHz (mean = 9.44 dB), 8 kHz (mean = 21.81 dB) and 12 kHz (mean = 39.03 dB) ([Fig 1A](#pone.0198743.g001){ref-type="fig"}). We also found that the group with high As levels in urine (n = 73) had significantly higher hearing thresholds at 4 kHz (mean = 10.41 dB; p = 0.0200), 8 kHz (mean = 25.96 dB; p = 0.0104) and 12 kHz (mean = 50.07 dB; p = 0.0015) than those in the low As group (n = 72) at 4 kHz (mean = 9.51 dB), 8 kHz (mean = 22.57 dB) and 12 kHz (mean = 40.63 dB) ([Fig 1B](#pone.0198743.g001){ref-type="fig"}). {#pone.0198743.g001} 10.1371/journal.pone.0198743.t003 ###### As levels in biological samples from humans. {#pone.0198743.t003g} Mean ± SD Range Cut-off value ----------------- ---------------- -------------- --------------- Toenails (μg/g) 1.38 ± 1.17 0.06--6.22 0.60 Urine (μg/L) 90.27 ± 103.04 4.17--1013.5 76.12 Association between hearing thresholds and As levels in biological samples {#sec011} -------------------------------------------------------------------------- We next performed logistic regression analysis with adjustments for age, sex, smoking history and BMI to determine the risk of hearing loss in subjects with high As in biological samples ([Table 4](#pone.0198743.t004){ref-type="table"}). In this study, we followed the method used in a previous study \[[@pone.0198743.ref016]\] to define subjects with hearing loss as subjects with hearing thresholds more than the cut-off values at each frequency. Levels of As in toenails were significantly associated with hearing loss at 4 kHz \[odds ratio (OR) = 4.27; 95% confidence interval (CI): 1.56, 12.70\], 8 kHz (OR = 3.91; 95% CI: 1.50, 10.77) and 12 kHz (OR = 5.58; 95% CI: 2.21, 15.07) ([Table 4](#pone.0198743.t004){ref-type="table"}). No significant correlations were found between As levels in urine and hearing loss at any frequency ([Table 4](#pone.0198743.t004){ref-type="table"}). We further shifted the cut-off values of the independent variables dichotomizing As levels in biological samples to verify the models. We found that the significance of ORs in toenails remained when the cut-off values in toenails were shifted from 0.30 to 0.70 μg/g, while As levels in urinary samples did not show significant ORs even when the cut-off values were shifted from 10 to 380 μg/L. 10.1371/journal.pone.0198743.t004 ###### Adjusted ORs (95% CI) for hearing loss and As levels in biological samples (n = 145)[^a^](#t004fn002){ref-type="table-fn"}. {#pone.0198743.t004g} -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 1 kHz\ 4 kHz\ 8 kHz\ 12 kHz\ (≥ 7 dB) (≥ 10 dB) (≥ 24 dB) (≥ 45 dB) -------------------- -------------- ---------------------------------------------- ---------------------------------------------- ---------------------------------------------- **As in toenails** Low Reference Reference Reference Reference High 1.28\ 4.27[\*\*](#t004fn003){ref-type="table-fn"}\ 3.91[\*\*](#t004fn003){ref-type="table-fn"}\ 4.15[\*\*](#t004fn003){ref-type="table-fn"}\ (0.43--3.83) (1.51--12.05) (1.47--10.38) (1.55--11.09) **As in urine** Low Reference Reference Reference Reference High 0.69\ 1.89\ 1.48\ 1.15\ (0.29--1.63) (0.83--4.29) (0.65--3.34) (0.52--2.56) -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- OR, odds ratio; CI, confidence interval. References mean 1.00. ^a^Adjusted for age, sex, smoking history and BMI. \*\**p* \< 0.01. Mice orally exposed to As showed an association between As levels in inner ear and nails {#sec012} ---------------------------------------------------------------------------------------- We finally performed an experimental study to determine the correlation between As levels in nails and inner ears in mice. After exposure of mice to As via drinking water for 2 months, we measured As levels in nails and inner ears from the exposed group and the control group. We found that there was a significant correlation (r = 0.8113, p = 0.0014) between As levels in inner ears and nails ([Fig 2](#pone.0198743.g002){ref-type="fig"}). {#pone.0198743.g002} Discussion {#sec013} ========== As level in toenails is a possible biomarker associated with hearing loss {#sec014} ------------------------------------------------------------------------- In our epidemiological study, As levels in toenails were shown to be significantly correlated with hearing loss at 4, 8 and 12 kHz in humans by multivariate analysis with adjustments for age, sex, smoking and BMI. Our experimental study also showed that As level in nails is a reliable index to assess As level in inner ears. Thus, the results of our combined study suggest that As level in toenails is a possible biomarker associated with hearing loss at 4, 8 and 12 kHz in humans. Association between chronic exposure to As and hearing loss {#sec015} ----------------------------------------------------------- The levels of an element in nails and hair are indexes reflecting chronic exposure status, while those in blood and urine samples are regarded as indexes of acute exposure \[[@pone.0198743.ref032], [@pone.0198743.ref033]\]. In our multivariate analysis, As levels in urine were not significantly correlated with hearing loss. Correspondingly, previous studies in which multivariate analyses with adjustments for age, sex and smoking were performed showed that As levels in urine were not associated with hearing loss \[[@pone.0198743.ref034]--[@pone.0198743.ref036]\]. Therefore, our results suggest that chronic exposure to As, but not acute exposure, is required for hearing loss in humans. In this study, we determined associations between As levels in biological samples and hearing levels in Bangladeshi only. Previous studies showed that races are associated with hearing levels \[[@pone.0198743.ref037], [@pone.0198743.ref038]\]. Thus, further study is needed to verify the associations between As levels in toenails and hearing loss in other countries. Association between As levels in hair and hearing loss at 12 kHz {#sec016} ---------------------------------------------------------------- The association between As levels in hair and hearing loss in humans aged 12--55 years was also analyzed in this study ([S1 Table](#pone.0198743.s002){ref-type="supplementary-material"}). As levels in hair were significantly associated with hearing loss at 12 kHz (OR = 2.94; 95% CI: 1.20, 7.20) but not with hearing loss at 4 or 8 kHz ([S1 Table](#pone.0198743.s002){ref-type="supplementary-material"}). Thus, our multivariate analysis showed that the significant association of As levels in hair with hearing loss was limited to an extra-high frequency (12 kHz), which is not necessary for daily communication in humans. A previous study showed that children living in an As-polluted area had hearing losses at 125, 250, 500, 1,000 and 8,000 Hz \[[@pone.0198743.ref039]\]. The As level in hair was 3.26 μg/g (mean value) in the previous study, while the As level in hair was 0.50 μg/g (mean value) in this study, about 6.5-times lower than that in the previous study. Therefore, it is possible that different levels of As in hair are associated with hearing loss at different frequencies. Relative contributions of As in toenails and confounders to hearing loss {#sec017} ------------------------------------------------------------------------ Age and smoking are known to be strong confounders for hearing loss. In this study, multivariate analysis showed that age was significantly associated with hearing loss at 4 kHz (OR = 4.25; 95% CI: 1.83, 9.89), 8 kHz (OR = 4.37; 95% CI: 1.84, 10.39) and 12 kHz (OR = 3.31; 95% CI: 1.46, 14.51). Smoking also had significant associations with hearing loss at 4 kHz (OR = 3.78; 95% CI: 1.17, 12.15) and 12 kHz (OR = 4.49; 95% CI: 1.39, 14.51). We then used the McFadden's Pseudo *R*^*2*^ values \[[@pone.0198743.ref040]\] to determine the relative contributions (%) of As in toenails and the other confounders including age to hearing loss at each frequency in the multivariate analysis ([S2 Table](#pone.0198743.s003){ref-type="supplementary-material"}). The relative contribution of As in toenails (17.33%) to hearing loss at 12 kHz was higher than the relative contributions of age (16.53%) and smoking (13.78%), while the relative contributions of age to hearing loss (20.75% at 4 kHz and 20.41% at 8 kHz) were higher than those of As in toenails (14.54% at 4 kHz and 13.70% at 8 kHz), smoking (9.44% at 4 kHz) and sex (12.09% at 8 kHz). Thus, our multivariate analysis suggests that As level in toenails is the largest contributor to hearing loss at 12 kHz among the confounders including age and smoking, while As level in toenails is the second-largest contributor after age to hearing loss at 4 and 8 kHz. On the other hand, hearing levels in males are generally known to be worse than those in females \[[@pone.0198743.ref041]\]. In this study, hearing thresholds in females were significantly higher than those in males. As levels in toenails (mean ± SD) in females (1.58 ± 1.00 μg/g) were significantly higher than those in males (1.17 ± 1.30 μg/g; p \< 0.0001). Therefore, it is possible that there is an association between higher As levels in females and higher hearing thresholds than those in males. It would be worthwhile to investigate the reason for the gender difference in As levels. Possible route of exposure to As for the subjects in this study {#sec018} --------------------------------------------------------------- The major route of exposure to As in the subjects in this study is not clear, but in our experimental study in which mice were orally exposed to As, there was a correlation between As levels in nails and inner ears. In our epidemiological study, we found a significant correlation (r = 0.5826; p \< 0.0001) between As levels in toenails and duration of drinking tube well water ([S1 Fig](#pone.0198743.s005){ref-type="supplementary-material"}). Therefore, it is likely that the route of exposure to As for the subjects in this study was drinking well water. Tube well water polluted with As is known to contain other elements. In our previous studies, barium was detected at a level similar to that of As in well drinking water in Bangladesh \[[@pone.0198743.ref042]\] and was to be associated with hearing loss in humans \[[@pone.0198743.ref014]\]. In this study, significant correlations between As levels in toenails and hearing loss remained in multivariate analysis adjusted with barium in addition to the confounders ([S3 Table](#pone.0198743.s004){ref-type="supplementary-material"}). Thus, the results suggest that As level in toenails is independently associated with hearing loss at 4, 8 and 12 kHz in humans. Further study is needed to investigate the association between hearing loss and As levels in well water worldwide, since more than 137 million people drink tube well water polluted by As worldwide \[[@pone.0198743.ref043]\]. Study limitations {#sec019} ----------------- This pilot study has several limitations. First, we used an ipod with headphones as the screening method for the field work in Bangladesh, since there is no clinical audiometer in rural areas. In addition, we did not perform otoscopy and tympanometry to check middle-ear problems. Second, our cross-sectional analysis was useful for determining the association between As levels in biological samples and hearing loss, but a the causal relationship could not be established. Cohort studies will be needed to determine the causality. Third, there was no information about noise background, though the subjects lived in rural areas and did not have portable music players with earphones. Fourth, the sample size of the pilot study in Bangladesh was small. Additional study is needed to analyze the association between As and hearing loss with consideration of the above limitations in larger sample sizes in other areas. Conclusion {#sec020} ---------- In conclusion, our combined experimental study and epidemiological study showed that As levels in nails were significantly associated with hearing loss in humans and that As levels in nails were significantly associated with those in inner ears in mice. Our epidemiological study also suggests the possible thresholds of As ranging from 0.30 μg/g to 0.70 μg/g in toenails that increase the risk for hearing loss in humans. Exposure to As is a worldwide health risk for humans. Our study provides new information that As level in toenails is a reliable index to predict As-mediated hearing loss in humans living in As-polluted areas. Supporting information {#sec021} ====================== ###### Determination of As levels in hair samples. (DOC) ###### Click here for additional data file. ###### Adjusted ORs (95% CI) for hearing loss and As levels in hair samples (n = 145)^a^. (DOC) ###### Click here for additional data file. ###### Hearing loss on McFadden's pseudo *R*^*2*^ for each factor. (DOC) ###### Click here for additional data file. ###### Adjusted ORs (95% CI) for hearing loss and As levels in biological samples (n = 145)^a^. (DOC) ###### Click here for additional data file. ###### Correlation between As levels in toenails and duration of drinking tube well water. Correlation between As levels in toenails and duration of drinking tube well water (years) was determined by spearman correlation coefficients. (TIF) ###### Click here for additional data file. We would like to thank the laboratory member for their helpful discussion. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.

{#sp1 .67}

Introduction {#s1} ============ Some of the most diverse and bizarre organelle genomes of all eukaryotes come from the Volvocales, which is a large order of predominantly freshwater green algae, belonging to chlorophycean class of the Chlorophyta. Volvocalean mitochondrial and plastid DNAs (mtDNAs and ptDNAs) show an impressive array of architectures, nucleotide landscapes, and coding compositions ([Table 1](#pone-0057177-t001){ref-type="table"})-- and see Leliaert et al. [@pone.0057177-Leliaert1] and Lee and Hua [@pone.0057177-Lee1] for additional compilations. Moreover, certain volvocalean species, particularly those within the "*Reinhardtinia* clade" *sensu* Nakada et al. [@pone.0057177-Nakada1], have proven to be excellent systems for testing contemporary hypotheses on the evolution of organelle genome expansion and linearization [@pone.0057177-Smith1], [@pone.0057177-Michaelis1], [@pone.0057177-Smith2]. 10.1371/journal.pone.0057177.t001 ###### Completely sequenced organelle genomes from volvocalean green algae. {#pone-0057177-t001-1} Species Clade (lineage) Organelle genome architecture -------------------------------- ---------------------------- ------------------------------- -------- ---- -------- ---- ------ ------------------- MITOCHONDRIAL DNA * Chlamydomonas* *reinhardtii* *Reinhardtinia*(volvocine) Linear 16--19 55 67--82 7 0--3 EU306617--23 * Chlamydomonas* *moewusii* *Xenovolvoxa* Circular 23 65 54 7 9 AF008237 * Chlorogonium* *elongatum* *Caudivolvoxa* Circular 23 62 53 7 6 Y13643--4,Y07814 * Dunaliella salina* *Caudivolvoxa* Circular 28 66 42 7 18 GQ250045 * Gonium pectorale* *Reinhardtinia*(volvocine) Circular 16 61 73 7 1 AP012493 * Polytomella capuana* *Reinhardtinia* Linear 13 43 82 7 0 EF645804 * Polytomella parva* *Reinhardtinia* Linear 16 59 66 7 0 AY062933-4 * Polytomella* sp.SAG 63--10 *Reinhardtinia* Linear 16 58 66 7 0 GU108480-1 * Volvox carteri* *Reinhardtinia*(volvocine) Circular 35 66 \<40 7 3 EU760701,GU084821 PLASTID DNA * Chlamydomonas* *reinhardtii* *Reinhardtinia*(volvocine) Circular 204 66 44 66 7 FJ423446 * Dunaliella salina* *Caudivolvoxa* Circular 269 68 35 66 \>35 GQ250046 * Gonium pectorale* *Reinhardtinia*(volvocine) Circular 223 70 44 66 3 AP012494 * Volvox carteri* *Reinhardtinia*(volvocine) Circular 525 57 \<20 66 9 GU084820 Note: Values rounded to whole numbers. Clade names are based on Nakada et al. [@pone.0057177-Nakada1]. Percent coding includes all annotated protein-, rRNA-, and tRNA-coding regions as well as non-standard ORFs, such as the *rtl* gene in the *C. reinhardtii* mtDNA. Gene number includes standard protein-coding genes, but does not include intronic or nonstandard ORFs, like *rtl*. Duplicate genes and introns were counted only once. Genome statistics for *P. parva* and *P. piriformis* are based on the concatenation of the two mitochondrial chromosomes; those for *V. carteri* should be considered as approximations as the mtDNA and ptDNA contain assembly gaps due to unresolved repeats. For *C. reinhardtii*, the mitochondrial genome size, intron number, and coding content can vary because of optional introns. Most of our understanding of volvocalean mitochondrial and plastid genomes is limited to unicellular species, such as the model organisms *Chlamydomonas reinhardtii* and *C. globosa* (previously misidentified as *C. incerta*; see Nakada et al. [@pone.0057177-Nakada2]) [@pone.0057177-Michaelis1], [@pone.0057177-Popescu1], the colorless and wall-less *Polytomella capuana*, *P. parva*, and *P. piriformis* [@pone.0057177-Smith2], [@pone.0057177-Fan1], [@pone.0057177-Smith3], and the halotolerant β-carotene-rich *Dunaliella salina* [@pone.0057177-Smith4]. Surprisingly little is known about the organelle genomes of colonial and multicellular volvocaleans, which are found within the volvocine lineage of the *Reinhardtinia* clade ([Figure S1](#pone.0057177.s001){ref-type="supplementary-material"}). Volvocine algae are preeminent models for studying the evolution of multicellularity [@pone.0057177-Kirk1], [@pone.0057177-Sachs1], and span the gamut of cellular complexity, from simple 4-celled species (e.g., *Tetrabaena*), to 8-64-celled colonial forms (e.g., *Gonium*), all the way to highly complex spheroidal taxa, with more than 500 cells (e.g., *Volvox*) [@pone.0057177-Nozaki1], [@pone.0057177-Herron1]. It is estimated that multicellular volvocine species last shared a common unicellular ancestor ∼200 million years ago [@pone.0057177-Herron1]. Of the 8 different volvocalean algae for which complete mtDNA and/or ptDNA sequences are available [@pone.0057177-Smith4], only one is multicellular: *Volvox carteri*, which is comprised of ∼4,000 cells. The organelle genomes of this species are distended with repetitive noncoding DNA, and similar palindromic repeats are located in both the mitochondrial and plastid compartments [@pone.0057177-Smith5]. Moreover, the *V. carteri* ptDNA, at ∼525 kb, is among the largest plastomes ever observed (from any eukaryote) [@pone.0057177-Smith1], dwarfing that of *C. reinhardtii*, which is 204 kb [@pone.0057177-Maul1]. Although smaller than its plastid counterpart, the ∼35 kb mtDNA of *V. carteri* is still larger than any the other completely sequenced volvocalean mitochondrial genome. It is hypothesized that the expanded organelle genomes of *V. carteri* are a consequence of a low organelle mutation rate and/or a small effective population size [@pone.0057177-Smith1]. The *V. carteri* mtDNA assembles as a circular molecule, contrasting the linear (or linear fragmented) architectures of all other well-studied *Reinhardtinia*-clade mitochondrial genomes, including those of *C. reinhardtii*, *Polytomella* spp., and the multicellular *Pandorina morum* [@pone.0057177-Michaelis1], [@pone.0057177-Smith4], [@pone.0057177-Moore1]. These linear mtDNAs have evolved complex terminal structures [@pone.0057177-Michaelis1], [@pone.0057177-Smith3], called mitochondrial telomeres, which form long palindromic repeats at the genome ends. The origin and number of times that linear mitochondrial chromosomes have evolved within the *Reinhardtinia* is unknown, but it has been argued that they arose only once [@pone.0057177-Smith1]. If true, this would imply that in a recent ancestor of *V. carteri*, the mtDNA reverted from a linear to a circular form. To learn about organelle genome architecture within multicellular volvocine algae and to gain insight into ptDNA expansion and the origin of linear mtDNAs, we sequenced the mitochondrial and plastid genomes of *Gonium pectorale*--an 8- or 16-celled freshwater colonial alga, occupying a phylogenetic position closer to that of *V. carteri* than *C. reinhardtii* within the volvocine line [@pone.0057177-Nozaki1], [@pone.0057177-Herron1], [@pone.0057177-Nozaki2] ([Figure S1](#pone.0057177.s001){ref-type="supplementary-material"}). Materials and Methods {#s2} ===================== The organelle genomes described here come from *Gonium pectorale* K3-F3-4 (mating type *minus*), which was one of the F3 backcross strains to K41 (mating type *plus*) (originating from K41×K32 \[F1 strains of Kaneko3×Kaneko4\]) [@pone.0057177-Hamaji1], [@pone.0057177-Mogi1] and is available as NIES-2863 from the Microbial Culture Collection at National Institute for Environmental Studies, Tsukuba, Japan (<http://mcc.nies.go.jp/>). *G. pectorale* was grown in 200--300 mL VTAC medium [@pone.0057177-Nozaki3], [@pone.0057177-Kasai1] at 20°C on a 14∶10 h light-dark cycle, under cool-white fluorescent lamps (165--175 µmol m^−2^ s^−1^ intensity). Total DNA was extracted based on the protocol of Miller et al. [@pone.0057177-Miller1]. Sequencing libraries were prepared from *G. pectorale* K3-F3-4 genomic DNA using the GS FLX Titanium Rapid Library Preparation Kit (F. Hoffmann-La Roche, Basel, Switzerland) and the TruSeq DNA Sample Prep Kit (Illumina Inc., San Diego, CA, USA), and were run on a GS FLX (F. Hoffmann-La Roche) and a MiSeq sequencer (Illumina Inc.), respectively. The GS FLX reads were assembled with Newbler v2.6. A fosmid library (23,424 clones) was constructed from *G. pectorale* K3-F3-4 genomic DNA using fosmid vector pKS300, which was developed in-house. End sequencing of the fosmid library and the BAC library of *G. pectorale* Kaneko3 (18,048 clones, Genome Institute (CUGI), Clemson Univ., Clemson, SC, USA) was carried out using a BigDye terminator kit ver3 (Life Technologies, Carlsbad, California, USA) and was run on automated ABI 3730 capillary sequencers (Life Technologies). The GS FLX contig sequences, which were derived from mitochondrial and chloroplast genomes, and the BAC/fosmid end-sequences were assembled using the Phrap/Consed systems. Gap closing and re-sequencing of low-quality regions in the assembly were performed by shotgun sequencing of the corresponding BAC/fosmid clones, PCR, primer walking, and direct sequencing of fosmid clones. The MiSeq sequence reads were mapped against the assembly sequences using the BWA program [@pone.0057177-Li1] after passing through the quality filter. The errors on each GS FLX assembly sequence were also corrected. The assembling delineated one circular mtDNA and two ptDNA isoforms (A and B), a common feature of plastid genomes with inverted repeats [@pone.0057177-Stein1], [@pone.0057177-Harris1] ([Figure S2](#pone.0057177.s002){ref-type="supplementary-material"}). The annotated *G. pectorale* mtDNA and ptDNA (isoform A) sequences are deposited in the DDBJ database under accession numbers AP012493 and AP012494, respectively. Phylogenetic analyses were performed under maximum likelihood (ML) using RAxML [@pone.0057177-Stamatakis1] and PhyML 3.0 [@pone.0057177-Guindon1] with 100 bootstrap replicates. Maximum parsimony (MP) bootstrap analyses (based on 10 random replications of the full heuristic search with the tree bisection--reconnection branch-swapping algorithm) were performed in PAUP 4.0b10 [@pone.0057177-Swofford1] with 1,000 replications. MtDNA protein phylogeny was based on the deduced *nad5*, *cox1*, and *cob* amino acid sequences ([Table S1](#pone.0057177.s010){ref-type="supplementary-material"}), which were aligned using Clustal X [@pone.0057177-Thompson1]. Intron phylogenies were based on the deduced and aligned amino acid sequences of the *nad5* and *psaB* intronic open reading frames (ORFs), which gave data matrices of 205 and 256 amino acids with 9 and 14 operational taxonomic units (OTUs), respectively ([Tables S2](#pone.0057177.s011){ref-type="supplementary-material"}, [S3](#pone.0057177.s012){ref-type="supplementary-material"}). Intron secondary structure maps were constructed as previously described [@pone.0057177-Nozaki4]. Results and Discussion {#s3} ====================== The *Gonium pectorale* mtDNA: A Compact Circular Mapping Chromosome {#s3a} ------------------------------------------------------------------- The mitochondrial genome of *G. pectorale* has a conservative architecture: it is small (16 kb), circular-mapping, AT rich (61%), compact (73% coding), contains very few repeats, and has only a single intron ([Figure 1](#pone-0057177-g001){ref-type="fig"}, [Table 1](#pone-0057177-t001){ref-type="table"}, [Figure S3](#pone.0057177.s003){ref-type="supplementary-material"}). It lacks the eccentricities that often characterize the mtDNAs of other volvocalean species, such as a high GC content (e.g., *P. capuana*), a linear or linear-fragmented conformation (e.g., *P. parva*), a large intron density (e.g., *D. salina*), non-standard genes (e.g., *C. reinhardtii*), and/or a bloated repeat-rich structure (e.g., *V. carteri*) [@pone.0057177-Smith4]. The *G. pectorale* mtDNA is gene poor, encoding 7 proteins, 2 rRNAs, and 3 unique tRNAs, representing methionine, glutamine, and tryptophan ([Figure 1](#pone-0057177-g001){ref-type="fig"}). Two copies of *trnM* were identified adjacent to one another in the genome. Both have similar sequences and cloverleaf structures, and appear to have a role in elongation rather than initiation, as suggested for the *trnM* of other volvocalean algae. When ignoring non-standard genes and duplicate tRNAs, the *G. pectorale* mitochondrial gene repertoire mirrors those from all other available volvocalean algae, with the exception of *Polytomella* species, which lack *trnW* and *trnQ*. The *G. pectorale* mitochondrial large and small subunit (LSU and SSU) rRNA genes, like those from other available *Reinhardtinia* algae, are fragmented and scrambled throughout the genome into 8 and 4 coding modules, respectively. In *V. carteri* the eighth LSU module has been invaded by palindromic repeats, splitting it into two segments (L8a and L8b) [@pone.0057177-Smith5]; in *G. pectorale*, however, the L8 module is intact. {#pone-0057177-g001} The sole intron of the *G. pectorale* mtDNA, located in *nad5*, is of group ID affiliation [@pone.0057177-Michel1] ([Figure S3](#pone.0057177.s003){ref-type="supplementary-material"}) and encodes a putative intronic endonuclease. Other volvocaleans contain a *nad5* group I intron (with the same insertion site), but none are from the *Reinhardtinia* clade. Our phylogenetic analyses of various volvocalean intronic ORFs ([Figure S4](#pone.0057177.s004){ref-type="supplementary-material"}) suggest that the *G. pectorale nad5* intron either was acquired through horizontal transmission from a volvocalean closely related to *Chlamydomonas moewusii* or *Chlorogonium elongatum* or that it was present in the ancestor of the Volvocales and preserved in *G. pectorale*. Linear mitochondrial chromosomes are widespread throughout the *Reinhardtinia* clade, occurring in all explored taxa [@pone.0057177-Laflamme1], with the exception of *V. carteri*, which has a circular mtDNA map, but rare possible linear forms of the genome have been observed [@pone.0057177-Smith1], [@pone.0057177-Smith5] ([Table 1](#pone-0057177-t001){ref-type="table"}). Our *de novo* and mapping assemblies of the *G. pectorale* mtDNA gave an unambiguous circular-mapping chromosome (see Materials and Methods), and although such a map could represent a circularly permuted, linear-type structure, various features of the *G. pectorale* mitochondrial genome support the idea that it is circular. For instance, all twelve of the *G. pectorale* mtDNA genes have the same transcriptional polarity--a trait that is also found in *V. carteri* and available volvocalean species with circular mitochondrial genomes. Conversely, in all of the sequenced linear mtDNAs from the Volvocales, the genes are divided into two transcriptional polarities, proceeding outward towards the ends of the chromosome [@pone.0057177-Smith2].Furthermore, our Southern blot analysis of the *G. pectorale* mtDNA, cut with restriction enzymes, demonstrates that it is a circular molecule ([Figure S5](#pone.0057177.s005){ref-type="supplementary-material"}). Our evidence for a circular mitochondrial genome in *G. pectorale* raises interesting questions about the origin of linear mtDNAs within the *Reinhardtinia* clade. There is little doubt that the ancestral volvocalean mtDNA was circular, and it is argued that there was a single shift from a circular to a linear mtDNA structure in the ancestor that gave rise to *Reinhardtinia* algae [@pone.0057177-Smith2]. Within the *Reinhardtinia* clade, *V. carteri* and *G. pectorale* belong to a monophyletic colonial or multicellular volvocalean group from which unicellular members are separated [@pone.0057177-Nozaki1], [@pone.0057177-Herron1] ([Figure 2](#pone-0057177-g002){ref-type="fig"}), but the multicellular volvocalean *Pandorina morum* has a linear mtDNA [@pone.0057177-Moore1]. Moreover, *V. carteri* and *P. morum* belong to the monophyletic Volvocaceae from which *G. pectorale* is excluded [@pone.0057177-Nozaki1], [@pone.0057177-Herron1], [@pone.0057177-Nozaki2] ([Figure S1](#pone.0057177.s001){ref-type="supplementary-material"}). Thus, the appearance of circular mitochondrial genome maps in both *V. carteri* and *G. pectorale* suggests that the mtDNAs of these species independently reverted from a linear to a circular conformation in the two separate ancestors of *G. pectorale* and *V. carteri* ([Figure S1](#pone.0057177.s001){ref-type="supplementary-material"}) or alternatively that there were multiple origins of linear mitochondrial genomes in the *Reinhardtinia* clade, in the ancestors of *Polytomella*, *C. reinhardtii*, and *P. morum* ([Figure S1](#pone.0057177.s001){ref-type="supplementary-material"}). Studies of mtDNA structure from other volvocine species, such as *Tetrabaena* and *Yamagishiella*, are needed to further investigate these hypotheses. ![MtDNA protein phylogeny of seven species belonging to *Reinhardtinia* clade and three outgroup species from the Volvocales.\ The tree was constructed under the RAxML (with WAG+I+4G model) method using the concatenated sequences of the deduced *nad5*, *cox1*, and *cob* amino acid sequences. Left, middle, and right bootstrap values (≥50%) obtained using the RAxML, PhyML (with LG+I+4G model), and MP analysis, respectively. The amino acid sequences of the three proteins were aligned by Clustal X [@pone.0057177-Guindon1], and ambiguously aligned and highly variable regions were removed to construct a multiprotein data matrix of 909 amino acids from the 10 operational taxonomic units ([Table S1](#pone.0057177.s010){ref-type="supplementary-material"}).](pone.0057177.g002){#pone-0057177-g002} The *Gonium pectorale* ptDNA Shows Moderate Genome Expansion {#s3b} ------------------------------------------------------------ Volvocalean plastid genomes are big and that of *G. pectorale*, at 222.6 kb, is no exception. Of the approximately 300 complete (or almost complete) ptDNAs in GenBank, as of 1 August 2012, fewer than ten have a length \>200 kb, all but one of which are from chlorophyte green algae, including the volvocaleans *C. reinhardtii* (204 kb), *D. salina* (269 kb) and *V. carteri* (∼525 kb) [@pone.0057177-Smith1], [@pone.0057177-Smith4], [@pone.0057177-Maul1]. The large size of volvocalean ptDNAs is not a product of an inflated gene number, but a consequence of having an abundance of noncoding nucleotides, often represented by repetitive elements and introns. This is also true for the *G. pectorale* ptDNA, which is 56% (∼125 kb) noncoding. Almost all of these noncoding nucleotides are AT rich (average = 71%) and found in intergenic regions. The coding regions also have a high AT content (68%) and encompass a total of 98 unique genes, encoding 67 proteins, 3 rRNAs, 27 tRNAs, and a single misc RNA (*tscA*) ([Figure 3](#pone-0057177-g003){ref-type="fig"}). Six of these genes (*psbA*, *rrnL*, *rrnS*, *rrnF*, *trnA,* and *trnI*) are duplicated, being located in a pair of 14.8 kb inverted repeats, which divide the *G. pectorale* ptDNA into a large (99.6 kb) and a small (93.5 kb) single-copy region ([Figure 3](#pone-0057177-g003){ref-type="fig"}). This gene complement and inverted-repeat arrangement is almost identical to those of *C. reinhardtii* and *V. carteri* ([Figure 3](#pone-0057177-g003){ref-type="fig"}, [Figure S6](#pone.0057177.s006){ref-type="supplementary-material"}). {#pone-0057177-g003} Although some volvocalean algae harbour many ptDNA introns ([Table 1](#pone-0057177-t001){ref-type="table"})--*V. carteri* has 9 and *D. salina* has \>35--*G. pectorale* harbours just three: one located in *psaB*, which appears to be of group IA affiliation [@pone.0057177-Michel1] ([Figure S3](#pone.0057177.s003){ref-type="supplementary-material"}), and encodes a putative endonuclease-like protein, and two short group II introns (117 and 176 bp) found upstream of *psaA* exons 2 and 3 ([Figure 3](#pone-0057177-g003){ref-type="fig"}). Phylogenetic analysis of the *G. pectorale* intron ([Figure S7](#pone.0057177.s007){ref-type="supplementary-material"}) show that it is closely related to the *psaB* group I intron of the chlorophycean (but non-volvocalean) green alga *Stigeoclonium helveticum* [@pone.0057177-Blanger1]; moreover, both introns have the same insertion site within the *psaB* gene. *V. carteri* also has a *psaB* intron, but it is of group II affiliation [@pone.0057177-Smith1]. In fact, there is not a single homologous pair of either group I or group II introns among the *G. pectorale*, *V. carteri*, and *C. reinhardtii* plastid genomes ([Figure S6](#pone.0057177.s006){ref-type="supplementary-material"}), suggesting that rapid horizontal intron transfer and loss occurred within the colonial Volvocales. The *G. pectorale* plastid genome, like its *V. carteri* and *C. reinhardtii* counterparts, contains hundreds of short repetitive elements, distributed throughout the intergenic regions, as demonstrated by the dotplot similarity matrix ([Figure S8](#pone.0057177.s008){ref-type="supplementary-material"}). Many of the *V. carteri* ptDNA repeats are palindromes, and can be folded into hairpin structures [@pone.0057177-Smith5]. The same is true for the *G. pectorale* ptDNA, which contains ∼135 short (13 nt) palindromic repeats (including eight in the coding regions) with the motif: 5′- TCCCCNNNGGGGA-3′ ([Figure S9](#pone.0057177.s009){ref-type="supplementary-material"}). This is fewer repeats than found in the *V. carteri* ptDNA, which contains over a thousand palindromic elements. The *G. pectorale* ptDNA is slightly more expanded (by ∼19 kb) than that of *C. reinhardtii*, but much smaller than those of the unicellular *D. salina* (269 kb, ∼65% noncoding) and the multicellular *V. carteri* (∼525 kb, \>80% noncoding) ([Table 1](#pone-0057177-t001){ref-type="table"}). What has led to such a wide spectrum of ptDNA expansion within the Volvocales? One contemporary--and controversial [@pone.0057177-Daubin1], [@pone.0057177-Sloan1]--hypothesis for the evolution of genome size, called the mutational hazard hypothesis [@pone.0057177-Lynch1], argues that genome expansion is a product of a low effective population size (*N~e~*) (which results in increased random genetic drift) and/or a low mutation rate (μ), which reduces the burden of harbouring excess DNA. The *V. carteri* ptDNA is estimated to have a very low *N~e~*μ [@pone.0057177-Smith1], about twenty times lower than that of the *C. reinhardtii* ptDNA [@pone.0057177-Smith6], which may explain why it is so bloated. We do not know the value of *N~e~*μ for the *G. pectorale* ptDNA--this will require sequencing the plastid genomes of several additional *G. pectorale* isolates. However, given that this species is ∼10 times larger than *C. reinhardtii* (16 cells vs a single cell) and a hundred times smaller than *V. carteri* (16 cells vs 4,000 cells), and that all three of these algae are found in a similar environment (freshwater ponds)--unlike *D. salina*, which is marine--one might expect the effective population size of *G. pectorale* to be similar or marginally smaller than that of *C. reinhardtii*, and much larger than that of *V. carteri*. If true, this may have contributed to *G. pectorale* having a ptDNA architecture comparable to that of *C. reinhardtii* but much different than that of *V. carteri*. Under this hypothesis, it can therefore be predicted that as more volvocine organelle DNAs are sequenced, species with large cell numbers and presumably low effective population sizes will have more bloated genomes than those with small cell numbers and large effective population sizes. Supporting Information {#s4} ====================== ###### Simplified diagram for phylogenetic relationships of selected taxa of the unicellular, colonial and multicellular vovlocaleans. (TIF) ###### Click here for additional data file. ###### Diagrams of possible isoforms of ptDNA of *Gonium pectorale*. A. Two isoforms as found in other ptDNAs with a typical inverted repeat. B. Two additional isoforms that were not rejected based on assembling of our sequence data. (TIF) ###### Click here for additional data file. ###### Secondary structures of group I introns within the *Gonium pectorale* organelle DNAs. A. Mitochondrial *nad5* group ID intron. B. Chloroplast *psaB* group IA intron. (TIF) ###### Click here for additional data file. ###### Phylogeny of *Gonium pactorale nad5* group I intronic ORF. The tree was constructed under the RAxML (with WAG+4G model) method using 8 additional, related amino acid sequences selected based on the topology of the distance tree provided by blastp research of NCBI (<http://www.ncbi.nlm.nih.gov/>). Numbers on the left, middle and right at branches represent bootstrap values (≥50%) obtained using the RAxML, PhyML (with LG+4G model), and MP analysis, respectively. The amino acid sequences were aligned by Clustal X, and ambiguously aligned and highly variable regions were removed to construct a data matrix of 205 amino acids from the 9 operational taxonomic units ([Table S2](#pone.0057177.s011){ref-type="supplementary-material"}). (TIF) ###### Click here for additional data file. ###### Southern blot analysis of *Gonium pectorale* mtDNA with four restriction enzymes that cut the genome once (SacI and StuI) or twice (SacII and EcoRI). Genome map coordinates are based on the *G. pectorale* mtDNA DDBJ accession (AP012493). SacI and StuI digestions each gave single genome-sized bands (∼16 kb), and the SacII and EcoRI reactions each gave two bands. These data are consistent with the *G. pectorale* mtDNA being a circular molecules. Probe DNA was amplified by PCR with two specific primers (Gopec-mito-F 5′-CGGGCAAAGCATAATTAGTGTAG-3′ and Gopec-mito-R 5′-ACGAACAAGAGGAAGACCTAAC-3′). (TIF) ###### Click here for additional data file. ###### Venn diagram comparing the gene repertoires of three volvocalean chloroplast genomes (AP012494, GU084820 and FJ423446). 102 genes (single asterisk) shared by the three genomes include 12 genes distributed in IRA and IRB and *trnI* (cau), which was previously annotated as one of the triplicated *trnM* in *C. reinhardtii* and *V. carteri*. Double asterisks represent one of the duplicated genes in *G. pectorale* and *C. reinhardtii*. Triple asterisks exhibit one of the duplicated genes in *C. reinhardtii*. Note that all intronic ORFs in *G. pectprale* (1^\#^) and *V. carteri* (6^\#^) are unique for each genome and considered "non-coding" in the text. (PDF) ###### Click here for additional data file. ###### Phylogeny of *Gonium pactorale psaB* group I intronic ORF. The tree was constructed under the RAxML (with WAG+4G model) method using 13 additional, related amino acid sequences selected based on the topology of the distance tree provided by blastp research of NCBI (<http://www.ncbi.nlm.nih.gov/>). Numbers on the left, middle and right at branches represent bootstrap values (≥50%) obtained using the RAxML, PhyML (with LG+G model), and MP analysis, respectively. The amino acid sequences were aligned by Clustal X, and ambiguously aligned and highly variable regions were removed to construct a data matrix of 256 amino acids from the 14 operational taxonomic units ([Table S3](#pone.0057177.s012){ref-type="supplementary-material"}). (TIF) ###### Click here for additional data file. ###### Dotplot similarity matrix of the *Gonium pectorale* plastid genome. The X- and Y-axes each represent the *G. pectorale* plastid genome (222.6 kb). Dots in the nucleotide similarity matrix represent regions of sequence similarity. The matrix was generated using JDotter, with a sliding-window size of 50. The inverted repeats are highlighted in red in the matrix. (TIF) ###### Click here for additional data file. ###### Distribution of short (13 nt) palindromic repeats (including seven \[red arrows\] in five coding regions \[blues arrows\]) with the motif: 5′- TCCCCNNNGGGGA-3′ in ptDNA of *Gonium pectorale*. The repeats were examined by using Serial Cloner 2.5 (<http://serialbasics.free.fr/Serial_Cloner.html>). (JPG) ###### Click here for additional data file. ###### Amino acid alignment and origin of the data used for [Figure 2](#pone-0057177-g002){ref-type="fig"}. (DOC) ###### Click here for additional data file. ###### Amino acid alignment and origin of the data used for [Figure S4](#pone.0057177.s004){ref-type="supplementary-material"}. (DOC) ###### Click here for additional data file. ###### Amino acid alignment and origin of the data used for [Figure S7](#pone.0057177.s007){ref-type="supplementary-material"}. (DOC) ###### Click here for additional data file. We thank all the technical staff of the Comparative Genomics lab at National Institute of Genetics for their assistance. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: HT DS H. Noguchi AT AF H. Nozaki. Performed the experiments: TH AT MS HKT AF IN TM BO H. Nozaki. Analyzed the data: TH DS H. Noguchi AT AF H. Nozaki. Contributed reagents/materials/analysis tools: TH IN TM BO H. Nozaki. Wrote the paper: TH DS H. Noguchi AT AF H. Nozaki.

Introduction {#s1} ============ Mitochondria perform a dual role in the life and death of the cardiomyocyte. When functioning normally they generate the energy required for normal cellular processes and survival. However, in situations of cellular stress such as during acute myocardial ischemia-reperfusion injury (IRI), they can become dysfunctional and be the arbitrators of cardiomyocyte death. Therefore, new treatment strategies which are capable of preventing mitochondrial dysfunction during acute IRI may reduce myocardial injury, preserve cardiac function and improve clinical outcomes in patients with ischemic heart disease. In this regard, the mitochondrial serine-threonine protein kinase, PTEN (phosphatase and tensin homologue on chromosome 10)-induced kinase 1 (PINK1), may provide a novel therapeutic target for cardioprotection [@pone.0062400-Siddall1]. Mutations in the PINK1 gene are responsible for the autosomal recessive PARK6 inherited form of early onset Parkinson disease, a neurodegenerative disorder characterized by the loss of dopaminergic neurons in the substantia nigra [@pone.0062400-Valente1]. Genetic ablation of PINK1 in neurons results in mitochondrial dysfunction characterized by: mitochondrial membrane depolarization [@pone.0062400-WoodKaczmar1], [@pone.0062400-Gandhi1], reduced mitochondrial respiration and ATP levels [@pone.0062400-Park1], increased oxidative stress [@pone.0062400-WoodKaczmar1], [@pone.0062400-Gandhi1], [@pone.0062400-Clark1]--[@pone.0062400-Dagda1], mitochondrial calcium overload [@pone.0062400-Gandhi1], and enhanced susceptibility to mitochondrial permeability transition pore (MPTP) opening [@pone.0062400-Gandhi1]. In contrast, wild-type PINK1 has been reported to protect neurons from mitochondrial dysfunction [@pone.0062400-Valente1], reduce mitochondrial cytochrome C release and caspase 3 and 9 activation [@pone.0062400-Petit1], [@pone.0062400-Wang2], and attenuate apoptotic cell death [@pone.0062400-Valente1], [@pone.0062400-Petit1]. Interestingly, PINK1 protein is highly expressed in the myocardium [@pone.0062400-Unoki1] but its role in the heart, is not clear [@pone.0062400-Siddall1], [@pone.0062400-Siddall2]. Given its beneficial effects on mitochondrial function and neuroprotective properties, we investigated whether PINK1 could also protect the heart against acute IRI. We find that the loss of PINK1 increases the heart\'s vulnerability to ischemia-reperfusion injury and this may be by worsening mitochondrial function. Materials and Methods {#s2} ===================== Animal experiments were conducted in strict accordance with the *Animals* (*Scientific Procedures*) *Act 1986* published by the UK Home Office and the *Guide for the Care and Use of Laboratory Animals* published by the US National Institutes of Health (NIH Publication No. 85--23, revised 1996). Approval has been granted by a local ethics review board at University College London. All efforts were made to minimize suffering. HL-1 Cell Culture and PINK1 Over-expression {#s2a} ------------------------------------------- HL-1 cells are an adherent murine atrial cell line that spontaneously beat in culture (the cells were obtained from Claycomb) [@pone.0062400-Claycomb1]. Cells were cultured in tissue culture flasks pre-coated for 2--3 hrs with 10 µg/ml fibronectin (diluted in 0.02% gelatin). Growth medium (Claycomb media supplemented with 10% FBS, 2 mM L glutamine (Invitrogen, Gibco), 0.1 mM norepinephrine (prepared in 30 mM ascorbic acid), 500 IU penicillin and 500 µg streptomycin (PAA Laboratories)) was changed every 1--2 days and cells were maintained at 37°C in 95%O~2~/5%CO~2~ with 90% humidity. A similar vector expressing PINK1 under the control of the CMV promoter (Addgene plasmid 13315: pcDNA-DEST53 PINK1) from Addgene Inc., Cambridge, MA was used to over-express PINK1. HL-1 cells were seeded onto fibronectin coated glass cover slips and upon reaching 50--60% confluence were transfected for 24 hours using Fugene6 ® (Roche, UK) according to manufacturer's instructions. The pEGFP expression plasmid (Clontech) was included for identification of successfully transfected cells, at a ratio of 1∶2. The vector control group was designated as cells transfected with an empty plasmid expression vector (RcCMV). A similar vector expressing PINK1 under the control of the CMV promoter (Addgene plasmid 13315: pcDNA-DEST53 PINK1) from Addgene Inc., Cambridge, MA was used to over-express PINK1 in a separate set of cells. The pEGFP expression plasmid (Clontech) was included for identification of successfully transfected cells, at a ratio of 1∶2. The transfection efficacy was 60--70% of cells. Culture media containing transfection components was replaced with fresh growth medium and cells were incubated overnight. Unfortunately due to a lack of a specific commercially available PINK1 antibody were not able to demonstrate PINK1 protein expression or localization. Simulated Ischemia-reperfusion Injury in HL-1 Cells Over-expressing PINK1 {#s2b} ------------------------------------------------------------------------- In order to determine the effect of PINK1 over-expression on the susceptibility to simulated ischemia-reperfusion injury (SIRI), HL-1 cells were subjected to a sustained episode of lethal hypoxia and reoxygenation [@pone.0062400-Lim1], [@pone.0062400-Smith1]. Culture medium was removed and replaced with hypoxic buffer (comprising in mM: KH~2~PO~4~ 1.0, NaHCO~3~ 10.0, MgCl~2~.6H~2~O 1.2, NaHEPES 25.0, NaCl 74.0, KCl 16, CaCl~2~ 1.2 and NaLactate 20 at pH 6.2, bubbled with 100% nitrogen) and then placed in an airtight custom-built hypoxic chamber kept at 37°C for 12 hours to simulate ischemia. Following the period of simulated ischemia, the cells were removed from the hypoxic chamber and placed in normoxic Claycomb medium (containing 3 µM propidium iodide) and returned to a tissue culture incubator, to simulate reperfusion. Following 1 hour simulated reperfusion at 37°C, the percentage of GFP-transfected cells stained with propidium iodide was determined using a Nikon Eclipse TE200 fluorescent microscope in order to calculate the percentage cell death in each treatment group. For each treatment group 80 cells were counted, taken from four randomly-selected fields of view. This experiment was repeated on at least four separate occasions giving a total of 320 cells per treatment group. For a time-matched normoxic control group, HL-1 cells were placed in normoxic buffer (comprising in mM: KH~2~PO~4~ 1.0, NaHCO~3~ 10.0, MgCl~2~.6H~2~O 1.2, NaHEPES 25.0, NaCl 98.0, KCl 3, CaCl~2~ 1.2, d-glucose 10.0, Na pyruvate 2.0 at pH 7.4, bubbled with 5% CO~2~/95% O~2~) for the total 13 hours duration of the experiment and the percentage cell death was determined. ROS-induced MPTP Opening in HL-1 Cells Over-expressing PINK1 {#s2c} ------------------------------------------------------------ To determine the effect of PINK1 over-expression on the susceptibility to MPTP opening, a previously validated cell model of MPTP opening was utilized [@pone.0062400-Zorov1]. Confocal laser-stimulation of the fluorophore tetra methyl rhodamine methyl (TMRM), which accumulates into mitochondria, generates reactive oxygen species (ROS) within mitochondria. This cell model can be used to simulate the events occurring at reperfusion, in which the production of ROS results in MPTP opening and the collapse of the mitochondrial membrane potential [@pone.0062400-Zorov1], [@pone.0062400-Davidson1]. The collapse of mitochondrial membrane potential in this cell model has previously been verified as indicating MPTP opening as it coincides with the redistribution of calcein from the mitochondria to the cytosol [@pone.0062400-Hausenloy1]. Culture medium was removed and replaced with Krebs imaging buffer. Cells were then loaded with the 3 µM TMRM for 15 min at 37°C and washed with Krebs imaging buffer. The time taken to induce mitochondrial membrane depolarization is recorded as a measure of susceptibility to MPTP opening. This was defined as the time taken to reach half the maximum TMRM fluorescence intensity. Twenty transfected cells were randomly selected for the induction and detection of MPTP opening from each treatment group, and this was repeated in four independent experiments giving a total of 80 cells per treatment group. As a positive control and in order to confirm that mitochondrial membrane depolarization was indicative of MPTP opening, following TMRM loading, a group of cells were pre-treated for 10 minutes with the MPTP inhibitor, cyclosporin A (0.2 µM) [@pone.0062400-Lim1], [@pone.0062400-Hausenloy2], [@pone.0062400-Davidson2]. The time taken to induce to MPTP opening was recorded. HL-1 cells were visualized using a Leica TCS SP5 CLSM confocal microscope equipped with HCX PL APO 40×/1.25 oil objective lens using the 488-nm of an Argon laser and the 543-nm emission line of a HeNe laser. Time scans were recorded with simultaneous excitation at 488 nm (for GFP) and 543 nm (for TMRM), collecting fluorescence emission at 500--536 nm and 585--680 nm, respectively. For these MPTP experiments, all conditions of the confocal imaging system (laser power, confocal pinhole - set to give an optical slice of 1 micron -- pixel dwell time, and detector sensitivity) were identical to ensure comparability between experiments. Images were analyzed using the LAS AF Version: 2.0.0 Build 1934 software program. PINK1 Knockout Mice {#s2d} ------------------- PINK1 knockout mice were a kind gift from Dr Luis Miguel Martins and Dr Nicoleta Moisoi of the MRC Toxicology Unit at the University of Leicester. These were bred in-house and PINK1+/+, +/− and −/− genotypes were generated. Genotyping was performed by extracting DNA from mouse ear biopsies using a Qiagen DNeasy kit with a proteinase K and spin column extraction method (Qiagen, UK) as previously described [@pone.0062400-WoodKaczmar1]. Myocardial Infarction in PINK1 Knockout Mouse Hearts {#s2e} ---------------------------------------------------- Hearts from PINK1+/+, PINK1+/− and PINK1−/− mice (10--15 weeks, 20--30 g) were isolated and perfused using a Langendorff constant pressure system as described previously [@pone.0062400-Siddall1]. Mice were given 500 IU of heparin by intraperitoneal injection before being culled by cervical dislocation. Hearts were rapidly excised and retrogradely perfused via the aorta on a Langendorff-apparatus (AD Instruments, UK) at 100 mmHg, with oxygenated Krebs-Henseleit buffer containing NaCl 118 mM, NaHCO~3~ 24 mM, KCl 4 mM, NaH~2~PO~4~ 1 mM, CaCl~2~ 1.8 mM, MgCl~2~ 1.2 mM and glucose 10 mM. Myocardial temperature was maintained at 37.0±0.5°C. Isolated perfused hearts were subjected to a 30 minute stabilization period followed by 35 minute global normothermic ischemia and 30 minute reperfusion. Infarct size was measured by perfusing a 1% triphenyltetrazolium chloride (TTC) solution retrogradely through the aorta and incubating the hearts at 37.0°C for 10 min before storing at −20.0°C. Subsequently, hearts were sliced (\<1 mm slices), destained in formalin, photographed and planimetered using the NIH Image 1.63 software package (National Institutes of Health, Bethesda, MD, USA). Infarct size was calculated as the percentage of the whole myocardium at risk (I/R%). Isolation of Adult Murine Cardiomyocytes {#s2f} ---------------------------------------- As described previously [@pone.0062400-Lim1] PINK1+/+ and PINK1−/− mice (10--15 weeks, 20--30 g) were injected (i.p) with 500 IU of heparin 30 minutes prior to an 0.01 mg/g i.p injection of anaesthetic, (10 mg/ml Ketamine, 2 mg/ml Xylazine and 0.06 mg/ml Atropine) as a terminal procedure. Hearts were excised and immediately placed in ice cold calcium free perfusion buffer (113 mM NaCl, 4.7 mM KCL, 0.6 mM KH~2~PO~4~, 0.6 mM Na~2~HPO~4~, 1.2 mM MgSO~4~.7H~2~O, 12 mM NaHCO~3~, 10 mM KHCO~3~, 30 mM Taurine, 10 mM HEPES, 11 mM Glucose and 10 mM 2,3-Butanedione monoxime). Within 3 minutes the heart was secured to a 22 gauge cannula via the aorta and attached to a perfusion apparatus. The heart was retrogradely perfused at 3 ml/min, with pre-warmed (37.0°C) oxygenated (95%O~2~/5%CO~2~) calcium-free perfusion buffer for 4 minutes. The heart was then perfused for 10 minutes with pre-warmed oxygenated digestion buffer (220 U/ml of type 2 Collagenase (Worthington, UK) and 55 U/ml Hyaluronidase (Sigma, UK) dissolved in the calcium free perfusion buffer and supplemented with 12.5 µM CaCl~2~). The ventricles were collected in 10 mlof digestion buffer and gently teased apart for additional tissue disruption. The tissue was digested further by incubating the mixture with 95%O~2~/5%CO~2~ in a shaking incubator (180 rpm) at 37°C for 10 minutes. The supernatant was collected and the remaining tissue pellet was incubated with an additional 10 ml digestion buffer followed by 10 minutes incubation with 95%O~2~/5%CO~2~ in a shaking incubator at 37°C. Under sterile conditions each ventricular cell suspension was transferred to a fresh tube and 5% FBS was added. The cells were centrifuged at 600 relative centrifugal force (RCF) for 3 minutes to separate the larger, cardiomyocyte pellet. The smaller fibroblasts and remaining connective tissue in the supernatant were discarded. The cardiomyocyte pellet was re-suspended in a low calcium buffer, which consisted of the calcium free perfusion buffer supplemented with 12.5 µM CaCl~2~. Small volumes of calcium were re-introduced to the cardiomyocyte suspension every 4 minutes to gradually reach a final concentration of 1 mM. The cells were then centrifuged at 600 RCF for 3 min and re-suspended in 1--2 ml of plating media (Medium-199)(Sigma, UK) supplemented with 2 mg/ml bovine serum albumin, 0.66 mg/ml creatine, 0.66 mg/ml taurine, 0.32 mg/ml carnitine, 50 U/ml penicillin, 5 µg/ml streptomycin and 25 µM blebbistatin (Calbiochem, Nottingham, UK). The cell suspension was then added to a glass cover slip, 22 mm diameter (VWR, Lutterworth, UK) pre-coated with laminin (1 mg/ml) to aid cell adhesion. Cells were left to adhere for 1 hr at 37°C, 95%O~2~/5%CO~2~ and 90% humidity. Finally, each cover slip was washed with plating media and the adhered cardiomyocytes were incubated with 1 ml of fresh plating media without blebbistatin. Measuring Mitochondrial Membrane Potential in PINK1−/− Cardiomyocytes {#s2g} --------------------------------------------------------------------- Primary adult cardiomyocytes were isolated from the myocardium of PINK1+/+ and −/− mice, as described above and loaded with 50 nM TMRM diluted in imaging buffer (which consisted of low calcium perfusion buffer without BDM and supplemented with 10 mM HEPES and 1.2 mM CaCl~2~ at pH 7.4) for 30 minutes. This was used to measure the mitochondrial membrane potential [@pone.0062400-Gandhi1], [@pone.0062400-Davidson1]. The isolated cardiomyocytes were mounted onto the confocal apparatus. The HeNe laser (543 nm) used to excite the TMRM was set to 2% power to prevent bleaching. Images were captured and the fluorescent intensity of each cell was recorded using the Leica application suite for Advanced Fluorescence (LAS AF Leica TCS SP5). Measuring Oxygen Consumption in Intact Cardiomyocytes and Isolated Mitochondria {#s2h} ------------------------------------------------------------------------------- To measure respiration rate in intact cells, approximately 2 × 10^6^ cells were suspended in HBSS in a Clark-type oxygen electrode thermostatically maintained at 37°C. The oxygen electrode was calibrated with air-saturated water, assuming 406 nmol O atoms/ml at 37°C. Oxygen consumption was measured over time with addition of oligomycin (final concentration 2 µg/ml) and 1 µM FCCP. To measure respiratory control ratio, intact mitochondria were isolated from hearts of WT and PINK1 KO mice by a method of differential centrifugation [@pone.0062400-Rosca1] and resuspended in medium containing 135 mM KCl, 10 mM NaCl, 20 mM HEPES, 0.5 mM KH~2~PO~4~, 1 mM MgCl~2~, 5 mM EGTA at pH 7.1. Oxygen consumption was measured in a Clark-type oxygen electrode thermostatically maintained at 25°C. Glutamate (5 mM) and malate (5 mM) were added to measure Complex I-linked respiration, succinate (5 mM) with rotenone (5 µM) were added to measure Complex II-linked respiration. All data were obtained using an Oxygraph Plus system with Chart recording software. Measuring Time to Contracture in PINK1−/− Cardiomyocytes {#s2i} -------------------------------------------------------- Primary adult cardiomyocytes were isolated from PINK1+/+ and −/− hearts, as described above, and, while imaging the cells, 10 µM carbonyl cyanide m-chlorophenyl hydrazone (CCCP) was added to the imaging buffer. In the presence of this uncoupling agent, the F~0~F~1~ATPase runs in "reverse" mode, consuming ATP in order to pump protons out of the mitochondria and maintain the membrane potential. When ATP decreases to a threshold level, adult cardiomyocytes undergo rigor contracture. The time to contracture can therefore be used as an indirect measure of basal levels of ATP or the activity of F~0~F~1~ATPase [@pone.0062400-Li1]--[@pone.0062400-Kaminishi1]. Sequential images of cells incubated with CCCP were taken at intervals of one minute using a standard light microscope affixed with a SPOT camera (Diagnostics Instruments, USA) connected to SPOT imaging software version 4.6 (Diagnostic Instruments, USA). The time to contracture was recorded and compared in the isolated PINK1+/+ and −/− cardiomyocytes. Measuring ROS during SIRI in PINK1−/− Cardiomyocytes {#s2j} ---------------------------------------------------- The generation of ROS following SIRI was investigated in isolated adult cardiomyocytes isolated from PINK1+/+ and PINK1−/− mouse hearts. Cardiomyocytes were subjected to 45 min simulated ischemia followed by 30 min re-oxygenation (SIRI) in the presence of 2 µM dihydroethidium (DHE, Molecular Probes, Invitrogen, UK), which is oxidized in the presence of superoxide to become fluorescent [@pone.0062400-Robinson1]. The fluorescence intensity, reflecting ROS levels, using SPOT imaging software version 4.6 (Diagnostic Instruments, USA) and NIH-Image, values were normalized to PINK1+/+ normoxic control. Statistical Analysis {#s2k} -------------------- Values are mean ± SEM. Data were analyzed using either the Student\'s *t*-test or one-way analysis of variance (ANOVA), followed by Bonferroni's multiple comparison post hoc test. P\<0.05 was considered significant. Results {#s3} ======= PINK1 Over-expression Protects HL-1 Cells against SIRI {#s3a} ------------------------------------------------------ PINK1 over-expression significantly reduced HL-1 cardiac cell death following SIRI: 49.0±2.4% in the vector control to 29.0±5.2% with PINK1 ([Figure 1](#pone-0062400-g001){ref-type="fig"}; P\<0.05). The proportion of dead cells in the time-matched normoxic control conditions was \<5.0% and this was not significantly altered by transgene expression. {#pone-0062400-g001} PINK1 Over-expression in HL-1 Cells Decreases Susceptibility to MPTP Opening {#s3b} ---------------------------------------------------------------------------- PINK1 over-expression in HL-1 cells delayed the time to MPTP opening by 1.3±0.3 fold when compared to control values (P\<0.01; [Figure 2](#pone-0062400-g002){ref-type="fig"}). The delay in MPTP opening was similar to that conferred by the known MPTP inhibitor, CsA (1.4±0.1 fold;P\<0.05; [Figure 2](#pone-0062400-g002){ref-type="fig"}). {#pone-0062400-g002} Myocardial Infarct Size is Increased in PINK1 Knockout Mice {#s3c} ----------------------------------------------------------- PINK1−/− mice developed significantly larger myocardial infarcts following an episode of sustained ischemia-reperfusion injury compared to PINK1+/+ mice, with PINK1+/− mice sustaining an intermediate sized infarct (25.1±2.0% in PINK1+/+ hearts versus 38.9±3.4% in PINK1+/− hearts; P\<0.01) and (25.1±2.0% in PINK1+/+ hearts versus 51.5±4.3% in PINK1−/− hearts (P\<0.001)(see [Figure 3](#pone-0062400-g003){ref-type="fig"}). {#pone-0062400-g003} Mitochondrial Membrane Potential and Time to Contracture in PINK1−/− Cardiomyocytes {#s3d} ----------------------------------------------------------------------------------- Since PINK1 has been implicated in mitochondrial function we evaluated mitochondrial membrane potential in primary adult cardiomyocytes under basal conditions. This was found to be lower in PINK1−/− cardiomyocytes compared to PINK1+/+ cardiomyocytes ([Figure 4](#pone-0062400-g004){ref-type="fig"}), as evidenced by decreased TMRM fluorescence (In arbitrary units: 12.1±2.8 in PINK1−/− cardiomyocytes versus 17.9±2.6 in PINK1+/+ cardiomyocytes; P\<0.05; [Figure 5](#pone-0062400-g005){ref-type="fig"}). {#pone-0062400-g004} {#pone-0062400-g005} Oxygen Consumption in Isolated Intact Cardiomyocytes {#s3e} ---------------------------------------------------- The basal oxygen consumption rate was significantly reduced in the PINK1 KO cardiomyocytes (0.48±0.04 nmol/O~2~/min/10^6^ cells, N = 4 experiments; [Figure 6](#pone-0062400-g006){ref-type="fig"}) compared to control cells (0.81±0.01 nmol O/min/10^6^ cells, N = 4 experiments, P\<0.001). Oligomycin (2 µg/ml) inhibited the respiration coupled to oxidative phosphorylation in control cells (to 0.11±0.01 nmol O/min/10^6^ cells, P\<0.05) but significantly less in PINK1 KO cardiomyocytes (0.37±0.03 nmol O~2~/min/10^6^ cells, compared to basal 0.48±0.04 nmol/O~2~/min/10^6^ cells; P\<0.001; [Figure 6](#pone-0062400-g006){ref-type="fig"}). 1 µM FCCP accelerated respiration to maximal levels in control cells, but to a lesser extent in PINK1 KO cardiomyocytes (4.2±0.19 vs. 3.4±0.03 nmol/O~2~/min/10^6^ cells; P\<0.001; [Figure 6](#pone-0062400-g006){ref-type="fig"}). This data suggest a generalised impairment of respiration in PINK1 KO heart cells. To identify the mechanism underlying the impaired mitochondrial respiration we investigated isolated mitochondria. {#pone-0062400-g006} Oxygen Consumption in Isolated Mitochondria {#s3f} ------------------------------------------- We evaluated the effect of PINK1 deficiency on oxygen consumption in isolated mitochondria. Compared to WT, oxygen consumption in PINK1 deficient mitochondria in the presence of the substrate of Complex I (5 mM Malate/5 mM Glutamate) was not significantly different (N = 6 experiments; [Figure 7a,c](#pone-0062400-g007){ref-type="fig"}). Application of Complex II substrate (succinate in the presence of rotenone; N = 6 experiments; [Figure 7b,c](#pone-0062400-g007){ref-type="fig"}) activated oxygen consumption equally in PINK1-KO and WT heart mitochondria. The respiratory control ratio (RCR), the ratio of state 3 (ADP-stimulated) to state 4 respiration (no ADP present), is an indication of the degree of coupling of the mitochondrial respiratory chain activity to oxidative phosphorylation. The RCR was unchanged in PINK1 KO, when compared to WT mitochondria. {#pone-0062400-g007} Time to Uncoupler-induced Hypercontracture {#s3g} ------------------------------------------ The time taken to hypercontracture following the administration of mitochondrial uncoupler, CCCP was recorded in cardiomyocytes. This time was significantly decreased in PINK1−/− cardiomyocytes (22.8±1.8 min in PINK1+/+ cells to 13.3±1.6 min in PINK1−/− cells; P\<0.01; [Figure 7](#pone-0062400-g007){ref-type="fig"}) indicating either a significantly lower basal level of cellular ATP or decreased function of the F~0~F~1~ATPase. However, it must be important to bear in mind that this is an indirect measure of cellular ATP levels. Oxidative Stress Post-SIRI is Increased in PINK1−/− Mice {#s3h} -------------------------------------------------------- At baseline, there was no significant difference in oxidative stress in isolated PINK1+/+ and PINK1−/− cardiomyocytes ([Figure 8](#pone-0062400-g008){ref-type="fig"}). However, following SIRI, the PINK1−/− cardiomyocytes exhibited a significantly greater level of superoxide production when compared to PINK1+/+ cardiomyocytes (145±12% in PINK1+/+ cardiomyocytes compared to 216±27% in PINK1−/− cardiomyocytes following SIRI (P\<0.05; [Figure 9](#pone-0062400-g009){ref-type="fig"}). {#pone-0062400-g008} {#pone-0062400-g009} Discussion {#s4} ========== The major findings of the current study are as follows: (1) Over-expressing PINK1 in the HL-1 cardiac cell line delayed the time taken to induce MPTP opening and reduced cell death following SIRI; (2) Mice lacking PINK1 sustained larger myocardial infarcts compared to wild type littermates (PINK1+/+), suggesting that the absence of PINK1 in the heart makes it more vulnerable to IRI. Interestingly, mice heterozygous for PINK1 sustained myocardial infarcts which were intermediate in size; (3) The increased susceptibility to IRI of PINK1−/− mice may be due to impaired mitochondrial function, as evidenced by a lower mitochondrial potential under basal conditions, impaired mitochondrial respiration, increased susceptibility to rigor contracture, and enhanced oxidative stress production during SIRI. Our study suggests a role for mitochondrial PINK1 as a target for cardioprotection in the heart. This mitochondrial protein appears to be required for endogenous protection against SIRI, as its genetic ablation resulted in an enhanced susceptibility to myocardial infarction, a response which was graded according to whether one or both PINK1 alleles were knocked-out. Our data suggests that the loss of PINK1 increases the heart\'s vulnerability to ischemia-reperfusion injury by worsening mitochondrial function. Its absence was associated with mitochondrial membrane depolarization, impaired mitochondrial respiration, increased susceptibility to rigor contracture, and more oxidative stress production during SIRI. In contrast, the over-expression of PINK1 was able to prevent the opening of the MPTP in the HL-1 cardiac cell line. A minor limitation of our study is that we did not investigate whether MPTP opening susceptibility was increased in adult cardiomyocytes deficient in PINK1. A recently published experimental study has also reported detrimental effects on mitochondrial function and cardiomyocyte homeostasis in mice lacking PINK1, but the effects of PINK1 on susceptibility to IRI was not explored in that study [@pone.0062400-Billia1]. Furthermore, these authors reported cardiac hypertrophy and cardiac fibrosis from 2 months of age [@pone.0062400-Billia1]. Despite many studies implicating PINK1 as a neuroprotective mitochondrial protein kinase, there have been a limited number of studies investigating the protective effect of PINK1 against ischemic neuronal injury. Shan and co-workers [@pone.0062400-Shan1] have reported that PINK1 can protect neonatal rat cortical neurons against simulated ischemia. Emerging studies have suggested that a functional impairment in activity of complex I of the electron transport chain underlies the mitochondrial dysfunction central to Parkinson's disease [@pone.0062400-Gandhi1], [@pone.0062400-Morais1]. The involvement of complex I inhibition in the pathogenesis of Parkinson's disease has long been appreciated with reduced complex I activity noted in patients with Parkinson's disease [@pone.0062400-Schapira1], and the use of complex I inhibitors such as rotenone to reproduce robust animal models of Parkinson's disease [@pone.0062400-Betarbet1]. Complex I inhibition may also explain some of the other known features of Parkinson's mitochondrial dysfunction including mitochondrial membrane depolarization [@pone.0062400-WoodKaczmar1], [@pone.0062400-Gandhi1], [@pone.0062400-Morais1], increase production of oxidative stress from complex I [@pone.0062400-WoodKaczmar1], [@pone.0062400-Gandhi1], [@pone.0062400-Clark1]--[@pone.0062400-Dagda1], reduced mitochondrial respiration and ATP depletion [@pone.0062400-Park1] and predisposition to MPTP opening [@pone.0062400-Gandhi1]. Interestingly, in our studies, we found that in isolated PINK1 deficient mitochondria there were no changes in the rate of mitochondrial respiration in the presence of substrates for Complex I and Complex II. Furthermore, PINK1 deficiency had no effect on uncoupling of oxidative phosphorylation as the respiratory control ratio remained unchanged. These findings are in agreement with experimental data, which have been performed in other cell types [@pone.0062400-Gandhi1], [@pone.0062400-Yao1]. However, in intact cardiomyocytes, PINK1 deficiency did significantly depress oxygen consumption, suggesting that the absence of PINK1 does not impair respiratory complex function but affects the delivery of substrates to the mitochondria (such as inhibition of glycolysis or the TCA cycle or inhibition of glucose uptake) [@pone.0062400-Gandhi1]. The MPTP is a non-selective high-conductance channel of the inner mitochondrial membrane whose opening results in cell death by uncoupling oxidative phosphorylation [@pone.0062400-Hausenloy2], [@pone.0062400-Halestrap1], [@pone.0062400-Di1]. In the absence of PINK1, a number of factors can result in an increased susceptibility to MPTP opening including mitochondrial membrane depolarization [@pone.0062400-Morais1], reduced mitochondrial calcium retention [@pone.0062400-Gandhi1], increased oxidative stress [@pone.0062400-WoodKaczmar1], [@pone.0062400-Gandhi1], [@pone.0062400-Clark1]--[@pone.0062400-Dagda1] particularly from complex I inhibition [@pone.0062400-Batandier1]. Of course there may be other beneficial effects on mitochondrial function which may explain the cardioprotective effects of PINK1. PINK1 has been reported to phosphorylate TNF receptor-associated protein 1 (TRAP1), a mitochondrial chaperone protein, which protects against oxidative stress-induced apoptotic cell death [@pone.0062400-Pridgeon1]. It has been proposed that PINK1 promotes the translocation of Parkin (an E3 ubiquitin ligase) to dysfunctional mitochondria, where outer mitochondrial membrane Mitofusins are ubiquinated, to provide a signal for mitophagy (the autophagic removal of dysfunctional mitochondria) [@pone.0062400-Poole1], [@pone.0062400-Ziviani1]. Changes in mitochondrial morphology can impact on a variety of cellular functions including metabolism, development and more recently cardioprotection [@pone.0062400-Ong1]. However, the effect of PINK1 on mitochondrial morphology has been variable and inconclusive [@pone.0062400-Bueler1]. Whether any of these other effects of PINK1 on mitochondrial function occur in the heart remains to be determined. Conclusions {#s4a} ----------- In conclusion, our data suggests that the loss of PINK1 increases the heart\'s vulnerability to ischemia-reperfusion injury, which may be due, in part, to worsened mitochondrial function. The mechanism underlying its cardioprotective effect appears to be mediated at the level of the mitochondria with improved mitochondrial function, less oxidative stress, and reduced susceptibility to MPTP opening. Therefore, the discovery of novel pharmacological activators of PINK1 may provide a novel therapeutic strategy for cardioprotection. Furthermore, its importance for endogenous cardioprotection lends further support to the importance of mitochondria in cardioprotection. We thank Luis Miguel Martins and Nicoleta Moisoi from the MRC Toxicology Unit, University of Leicester for donating the PINK1 knockout mice. For assistance in breeding the PINK1 knockout mice we would like to thank Abdul Mokit from the Biological Service Unit, University College London. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: HKS DMY SBO UAM NB ARH PRA MHRL ED SMD MMM DJH. Performed the experiments: HKS SBO UAM NB ARH PRA MHRL ED SMD DJH. Analyzed the data: HKS DMY SBO UAM NB ARH PRA MHRL ED SMD MMM DJH. Contributed reagents/materials/analysis tools: HKS DMY SBO UAM NB ARH PRA MHRL ED SMD MMM DJH. Wrote the paper: HKS DMY SBO UAM NB ARH PRA MHRL ED SMD MMM DJH.

Introduction {#tca13044-sec-0001} ============ Nivolumab, a PD‐1 antibody, is an immune checkpoint inhibitor (ICI) that has been proven to be active in patients with several different tumor types. Nivolumab has been shown to have an overall response rate (ORR) of approximately 20% in patients with previously treated non‐small cell lung cancer (NSCLC).[1](#tca13044-bib-0001){ref-type="ref"}, [2](#tca13044-bib-0002){ref-type="ref"} However, nivolumab is not effective in more than 40% of NSCLC patients who experience disease progression, despite nivolumab treatment. To improve the efficacy of this promising immunotherapy, additional modalities such as chemotherapy, radiotherapy (RT), or other ICIs may also be administered to patients in whom ICIs have not been completely effective. Interestingly, RT stimulates a systemic immune response and causes the release of tumor‐related antigens.[3](#tca13044-bib-0003){ref-type="ref"} Recent preclinical studies have demonstrated a synergistic tumor response with RT and the blockade of PD‐1.[4](#tca13044-bib-0004){ref-type="ref"}, [5](#tca13044-bib-0005){ref-type="ref"}, [6](#tca13044-bib-0006){ref-type="ref"} It is possible that tumor‐specific immunity is induced by radiation. Although RT plays an important role in the local control and elimination of tumors, it also contributes to the induction of antitumor immune responses, and the immunosuppressive and immunostimulatory effects of RT.[6](#tca13044-bib-0006){ref-type="ref"} Radiation‐induced cell death causes a release of danger signals such as HMGB1, ATP, and HSP70, and the dendritic cells can stimulate activated CD8 T cells and tumor‐specific T cells.[6](#tca13044-bib-0006){ref-type="ref"} In addition to immune activation, RT induces transforming growth factor‐β (TGF‐β), an immunosuppressive cytokine. To reduce the immunosuppressive functions, a combination of RT and TGF‐β inhibitors has been identified as a valuable option in preclinical settings.[6](#tca13044-bib-0006){ref-type="ref"} A recent experimental study demonstrated that fractionated RT increases PD‐L1 surface expression on tumor cells, suggesting a key rationale for the combination of RT with ICIs.[7](#tca13044-bib-0007){ref-type="ref"} Of the 98 patients registered in the KEYNOTE‐001 phase I trial, Shaverdian *et al.* reported that the duration of progression‐free survival (PFS) with pembrolizumab was significantly longer in patients previously administered RT than in those not treated with RT.[8](#tca13044-bib-0008){ref-type="ref"} Fiorica *et al.* also reported that nivolumab treatment after hypofractionated RT improved the outcome in 35 patients with pretreated or metastatic NSCLC.[9](#tca13044-bib-0009){ref-type="ref"} These results suggest that previous RT clinically improves tumor response and immune reaction to ICIs, such as nivolumab or pembrolizumab. However, the synergistic effect of ICIs and previous RT was not fully elucidated in these studies. As the former study was a phase I trial, pembrolizumab was administered at different doses and treatment deliveries.[8](#tca13044-bib-0008){ref-type="ref"} The latter study was limited by a small sample of only 35 patients.[9](#tca13044-bib-0009){ref-type="ref"} Analysis of these studies shows that immunotherapy after previous RT prolongs survival;[8](#tca13044-bib-0008){ref-type="ref"}, [9](#tca13044-bib-0009){ref-type="ref"} however, neither the ORRs of ICIs after previous RT nor the populations that might benefit from ICI treatment were included. Little detailed clinical data of the effect of ICI administration after previous RT has been reported; therefore, we attempted to elucidate the potential synergistic antitumor effect of nivolumab after RT in patients with previously treated NSCLC. Methods {#tca13044-sec-0002} ======= Patient eligibility and data collection {#tca13044-sec-0003} --------------------------------------- The eligibility criteria for our retrospective analysis were: histologically or cytologically proven advanced NSCLC with stage III or IV disease or recurrence after surgical resection; age \> 20 years; patients with disease progression after at least one prior cytotoxic chemotherapy treated with nivolumab; *EGFR* mutation‐positive patients administered EGFR‐tyrosine kinase inhibitors prior to any cytotoxic chemotherapy; and patients with available ORR data of nivolumab according to Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. Patients were excluded if they had: a concomitant serious illness, such as myocardial infarction in the previous three months; uncontrolled angina pectoris, heart failure, uncontrolled diabetes mellitus, uncontrolled hypertension, interstitial pneumonia, or lung disease; an infection or other disease contraindicating chemotherapy; or were pregnant or breastfeeding. This study was approved by the institutional ethics committee of the Saitama Medical University International Medical Center. Treatment and efficacy evaluation {#tca13044-sec-0004} --------------------------------- Nivolumab was intravenously administered at 3 mg/kg every two weeks. A complete blood cell count, differential count, routine chemistry measurements, physical examination, and toxicity assessment were performed on a weekly basis. Acute toxicity was graded according to Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. The tumor response was evaluated according to RECIST version 1.1.[10](#tca13044-bib-0010){ref-type="ref"} Statistical analysis {#tca13044-sec-0005} -------------------- *P* \< 0.05 indicated statistical significance. Fisher\'s exact tests were conducted to examine the association between the categorical variables. The Kaplan--Meier method was used to estimate survival as a function of time, and survival differences were analyzed by log‐rank tests. PFS was defined as the time from the initiation of nivolumab therapy to tumor recurrence or death from any cause, while overall survival (OS) was defined as the time from the initiation of nivolumab therapy to death from any cause. Statistical analyses were performed using GraphPad Prism 4 and JMP 8.0. Results {#tca13044-sec-0006} ======= Patient demographics {#tca13044-sec-0007} -------------------- From February 2016 to December 2017, 152 patients with pretreated NSCLC were administered nivolumab. Twenty‐eight patients were excluded because of inadequate medical information or the absence of an evaluable target lesion. Thus, a total of 124 patients (*n* ~males~ = 93, *n* ~females~ = 31; median age: 69 years; range: 31--85 years) were eligible for analysis. Patient characteristics are listed in Table [1](#tca13044-tbl-0001){ref-type="table"}. A total of 99 patients had a smoking history. Clinical staging indicated that 27 patients had stage III disease, 77 had stage IV disease, and 20 patients developed recurrence after surgical resection. The patients were divided into RT and non‐RT groups. ###### Comparison of demographics in patients treated with or without RT before nivolumab ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Variables All patients\ Patients administered any RT before Nivo (*n* = 66) Patients not administered RT before Nivo (*n* = 58) *P* (*n* = 124) ---------------------------------------------------------- --------------- ----------------------------------------------------- ----------------------------------------------------- ---------- Age ≦ 69/\> 69 65/59 36/30 29/29 0.71 Gender Male/female 93/31 51/15 42/16 0.54 Smoking Yes/no 99/25 52/14 47/11 0.82 ECOG PS 0/1--3 59/65 31/35 18/40 0.09 Stage III/IV 27/97 20/46 7/51 0.71 T factor T1--2/T3--4 68/56 35/31 33/25 0.58 N factor N0/N1‐3 20/104 12/54 8/50 0.62 Histology Adeno/non‐adeno 65/59 31/35 34/24 0.21 *EGFR* mutation status Mutant/wild 14/104 10/51 4/53 0.15 Nivo response CR or PR/SD or PD 35/89 24/44 11/47 **0.04** White blood cells[†](#tca13044-note-0002){ref-type="fn"} High/low 65 / 59 32/34 33/25 0.37 Neutrophils^1^ High/low 64 / 60 33/33 31/27 0.72 Lymphocytes^1^ High/low 62 / 62 26/40 36/22 **0.01** ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Bold values indicates statistically significance. Laboratory findings before nivolumab administration. IV, stage IV including recurrence after surgical resection; adeno, adenocarcinoma; CR, complete response; ECOG PS, Eastern Cooperative Oncology Group performance status; Nivo, nivolumab; non‐adeno, non‐adenocarcinoma; PD, progressive disease; PR, partial response; RT, radiotherapy; SD, stable disease; WBC, white blood cell. The percentages of patients with an Eastern Cooperative Oncology Group performance score (PS) of 0--1 in the RT and non‐RT groups were 75% (50/66) and 82% (48/58), respectively, without significant difference. There were 65 patients with adenocarcinoma (AC), 38 with squamous cell carcinoma (SQC), and 21 with other histologies. *EGFR* mutation analysis was performed: 104 patients had wild‐type *EGFR*, 14 harbored mutant *EGFR*, and 6 patients had unknown *EGFR* status. Table [1](#tca13044-tbl-0001){ref-type="table"} shows a comparison of the groups prior to nivolumab administration. The patient demographics in both groups were well balanced, except for the lymphocyte count. Sixty‐six patients were administered any RT prior to nivolumab treatment. Of these 66 patients, 24 were treated with concurrent platinum‐based chemoradiotherapy (50--60 Gy), 16 with palliative thoracic RT (30--40 Gy), 14 with palliative bone RT (8--30 Gy), and 11 with cranial RT (30--50 Gy). In terms of systemic chemotherapy prior to nivolumab treatment, 118 patients were treated with platinum‐based regimens and 6 with non‐platinum regimens. In the 66 patients administered any previous RT, 52 were treated with extracranial RT and 40 with thoracic RT. Patients were subdivided into three groups for further analysis: any previous RT (n = 66), extracranial RT (*n* = 52), and thoracic RT (*n* = 40). Palliative RT after nivolumab administration was administered to 25 of 66 patients (37.8%) who had undergone any previous RT (2 in thoracic sites, 13 in bone sites, 2 in lymph node metastases, and 8 in brain metastases) and 23 (of 58 patients 36.6%) without previous RT (4 in thoracic sites, 11 in bone sites, 2 in lymph node metastases and 6 in brain metastases), without statistical significance (37.8% vs. 36.6%; *P* = 0.85). Treatment delivery and response rate {#tca13044-sec-0008} ------------------------------------ The median number of nivolumab cycles was 4 (range: 1--43). The ORR and disease control rate (DCR) of nivolumab were 28.0% (95% confidence interval \[CI\] 20.1--35.9%) and 58.4% (95% CI 49.8--67.0%), respectively. Furthermore, analysis of all patients according to the number of lymphocytes showed an ORR of nivolumab treatment of 32% (20/62) in patients with low lymphocytes and 27% (17/62) in patients with high lymphocytes (*P* = 0.69). The median duration of follow‐up after RT in the 66 patients administered any RT prior to nivolumab treatment was 314 days (range: 12--3768). We used this median value of 314 days as a cutoff, and found that the ORR and DCR in 35 patients with a follow‐up of \< 314 days were 42% and 71%, and those in 31 patients with a follow‐up of \> 314 days were 29% and 64%, respectively, demonstrating no significant difference between the two groups. The ORR (36.4%, 95% CI 24.8--48.0%) in patients treated with previous RT was significantly higher than in patients without previous RT (19%, 95% CI 8.9--29.1%). The ORRs and DCRs of nivolumab in patients with or without previous RT are listed in Table [2](#tca13044-tbl-0002){ref-type="table"}. There was no statistically significant difference in the ORRs and DCRs among patients administered previous RT, extracranial RT, or thoracic RT. In the analysis according to histology, a statistically significant difference in the DCR, but not the ORR, was observed between AC and non‐AC histologies among patients with any previous RT, extracranial RT, and thoracic RT. No statistically significant differences in the ORRs and DCRs were observed between patients with stage III and IV. However, the ORR of patients with non‐AC histology and wild‐type *EGFR* seemed to be higher than in patients with AC histology, but the difference was not statistically significant. Among patients with SQC histology, the ORRs and DCRs were 43.4% and 78.2% in patients administered any previous RT (*n* = 23), 52.6% and 84.2% in patients administered extracranial RT (*n* = 19), and 52.6% and 89.4% in patients administered thoracic RT (*n* = 19), respectively. ###### Response of nivolumab in patients treated with or without previous RT Variables CR PR SD PD ORR (%) DCR (%) ----------------------------- ---- ---- ---- ---- --------- --------- Any previous RT (*n* = 66) 1 23 21 21 36.4% 68.2% Adeno (*n* = 31) 1 8 7 15 29.0% 51.6% Non‐adeno (*n* = 35) 0 15 14 6 42.8% 80.0% *EGFR* wild type (*n* = 51) 1 22 16 12 45.1% 76.4% *EGFR* mutant (*n* = 10) 0 1 1 8 10.0% 20.0% Stage III (*n* = 20) 0 8 6 6 40.0% 70.0% Stage IV (*n* = 37) 0 12 11 14 32.4% 62.2% Extracranial RT (*n* = 52) 1 19 17 15 38.4% 71.2% Adeno (*n* = 21) 1 4 5 11 23.8% 47.6% Non‐adeno (*n* = 31) 0 15 12 4 48.3% 87.1% *EGFR* wild type (*n* = 44) 1 19 12 12 45.5% 72.7% *EGFR* mutant (*n* = 4) 0 0 1 3 0.0% 25.0% Stage III (*n* = 19) 0 8 5 6 42.1% 68.4% Stage IV (*n* = 24) 0 8 8 8 33.3% 66.7% Thoracic RT (*n* = 40) 0 17 14 9 42.5% 77.5% Adeno (*n* = 11) 0 3 2 6 27.2% 45.5% Non‐adeno (*n* = 29) 0 14 12 3 48.3% 86.9% *EGFR* wild type (*n* = 35) 0 17 10 8 48.6% 77.1% *EGFR* mutant (*n* = 1) 0 0 0 1 0.0% 0.0% Stage III (*n* = 19) 0 8 6 5 42.1% 73.7% Stage IV (n *n* = 14) 0 6 5 3 42.8% 78.6% No previous RT (*n* = 58) 0 11 17 30 18.9% 46.5% Adeno (*n* = 34) 0 6 10 18 17.6% 47.1% Non‐adeno (*n* = 24) 0 5 7 12 20.8% 50.0% *EGFR* wild type (*n* = 53) 0 11 14 28 20.7% 47.2% *EGFR* mutant (*n* = 4) 0 0 2 2 0.0% 50.0% Stage III (*n* = 7) 0 0 2 5 0.0% 28.6% Stage IV (*n* = 40) 0 6 12 22 15.0% 45.0% Adeno, adenocarcinoma; CR, complete response; DCR, disease control rate; non‐adeno, non‐adenocarcinoma; ORR, overall response rate; PD, progressive disease; PR, partial response; RT, radiotherapy; SD, stable disease. Survival analysis and toxicity {#tca13044-sec-0009} ------------------------------ The median PFS and OS rates after nivolumab administration in all patients were 132 and 561 days, respectively. Of the 124 patients, 64 died and 101 experienced recurrence after initial nivolumab treatment. The median PFS of patients administered any previous RT (*n* = 66), extracranial RT (*n* = 52), thoracic RT (*n* = 40), and no previous RT (n = 58) were 204, 206, 233, and 79 days, respectively, and the median survival times were 562 days, not reached (NR), NR, and 524 days, respectively (Fig [S1](#tca13044-supitem-0001){ref-type="supplementary-material"}). Univariate and multivariate analyses were performed in all patients (Table [3](#tca13044-tbl-0003){ref-type="table"}). In univariate analysis, gender, smoking, histology, any previous RT, extracranial RT, and thoracic RT were identified as significant prognostic markers for PFS, while PS and neutrophil count were significant predictors for OS. In multivariate analysis, we included variables with *P* \< 0.05 in univariate analysis. Multivariate analysis confirmed that any previous RT and smoking were independent prognostic factors for poor PFS, whereas PS was the only significant prognostic marker for OS (Table [3](#tca13044-tbl-0003){ref-type="table"}). Figure [1](#tca13044-fig-0001){ref-type="fig"} shows the Kaplan--Meier survival curves according to any previous RT, extracranial RT, and thoracic RT. ###### Univariate and multivariate survival analyses Number of patients PFS OS ------------------------------------------------------------------- -------------------- --------------------------- ----------------------------- ---------------------------- ----------------------------- Age ≦ 69/\> 69 65/59 103/140 days (0.49) ‐ 709/524 days (0.68) Gender Male/female 93/31 184/68 days (**0.01**) 1.03(0.73--1.41) (**0.85**) 709/445 days (0.27) Smoking Yes/no 99/25 184/60 days (**\< 0.01**) 1.44(1.01--2.05) (**0.04**) 709/284 days (0.24) ECOG PS 0/1--3 59/65 177/81 days (0.12) --- 709/179 days (**\< 0.01**) 1.49(1.07--2.00) (**0.01**) Stage III/IV 27/97 139/176 days (0.89) --- NR/528 days (0.40) LN metastasis Yes/no 20/104 139/245 days (0.49) --- 524/NR days (0.19) Histology Adeno/non‐adeno 65/59 85/195 days (**0.02**) 1.12(0.96--1.39) (0.29) 528/709 days (0.38) Any RT Yes/no 66/58 204/79 days (**0.02**) 1.30(0.06--1.59) (**0.01**) 562/524 days (0.48) Thoracic RT Yes/no 40/58 233/79 days (\< **0.01**) --- NR/528 days (0.41) Extracranial RT Yes/no 52/58 206/79 days (**0.01**) --- NR/528 days (0.29) White blood cells[†](#tca13044-note-0004){ref-type="fn"} High/low 65/59 128/144 days (0.39) --- 478/NR days (0.12) Neutrophils[†](#tca13044-note-0004){ref-type="fn"} High/low 64/60 118/177 days (0.15) --- 406/778 days (**0.05**) 1.25(0.97--1.62) (0.07) Lymphocytes[†](#tca13044-note-0004){ref-type="fn"} High/low 62/62 180/129 days (0.35) --- 778/409 days (0.14) Laboratory findings before nivolumab administration. IV, stage IV including recurrence after surgical resection; CI, confidence interval; adeno, adenocarcinoma; ECOG PS, Eastern Cooperative Oncology Group performance status; HR, hazard ratio; LN, lymph node; M, months; MST, median survival time; Nivo, nivolumab; non‐adeno, non‐adenocarcinoma; OS, overall survival; PFS, progression‐free survival; PR, partial response; RT, radiotherapy. ) RT(+) (*n* = 66) and () RT(−) (*n* = 58), (**b**) extracranial RT () RT(+) (*n* = 52) and () RT(−) (*n* = 58), and (**c**) thoracic RT, () RT(+) (*n* = 40) and () RT(−) (*n* = 58). A statistically significant difference in PFS was observed between patients treated with and without RT. Kaplan--Meier survival curves of overall survival (OS) according to (**d**) any previous RT () RT(+) (*n* = 66) and () RT(−) (*n* = 58), (**e**) extracranial RT () RT(+) (*n* = 52), and () RT(−) (*n* = 58) and (**f**) thoracic radiotherapy () RT(+) (*n* = 40) and () RT(−) (*n* = 58). No statistically significant difference in OS was observed between the patients treated with and without RT.](TCA-10-992-g001){#tca13044-fig-0001} Figure [2](#tca13044-fig-0002){ref-type="fig"} shows a forest plot of PFS and OS according to RT administration prior to nivolumab treatment for each variable. Compared to no RT, previous RT was significantly linked to favorable PFS in patients with wild‐type *EGFR* and stage IV disease, while extracranial RT and thoracic RT yielded significantly better PFS in patients with non‐AC histology and wild‐type *EGFR*, and better OS in patients with non‐AC histology (Figs [2](#tca13044-fig-0002){ref-type="fig"}, [3](#tca13044-fig-0003){ref-type="fig"}). {#tca13044-fig-0002} ) RT(+) (*n* = 31) and () RT(−) (*n* = 28). The one and two‐year OS rates were 77% and 65%, respectively.](TCA-10-992-g003){#tca13044-fig-0003} In the analysis of pulmonary toxicities, 8 (20%) patients treated with previous thoracic RT experienced treatment‐related pulmonary toxicities compared to 12 (14%) patients not administered thoracic RT, without any statistical significance. Furthermore, no statistically significant difference in the incidence of grade 3 or higher pulmonary adverse events was observed between patients with or without a history of thoracic RT. Discussion {#tca13044-sec-0010} ========== This study was a retrospective evaluation of the efficacy of nivolumab treatment according to a history of previous RT in patients with previously treated NSCLC. We found that any previous RT was an independent prognostic marker of a favorable prognosis with nivolumab administration and could markedly improve the response rate and outcome of nivolumab treatment. The ORR of nivolumab was particularly improved in patients with non‐AC histology and wild‐type *EGFR,* with an increase of more than 40% if any previous RT was performed prior to nivolumab administration, whereas that of nivolumab in patients without previous RT was similar to that observed in previous phase III studies.[1](#tca13044-bib-0001){ref-type="ref"}, [2](#tca13044-bib-0002){ref-type="ref"} In addition, the frequency of low lymphocytes was significantly higher in patients administered previous RT; however, no statistically significant difference in the ORR of nivolumab was observed between patients with and without any previous RT. Therefore, we consider that the number of lymphocytes does not bias the efficacy of nivolumab, although it remains unclear why there is a trend of low lymphocytes in the RT group. Our detailed survival analysis also revealed that nivolumab increased PFS and OS in patients with non‐AC histology who were administered extracranial RT prior to nivolumab therapy. Of the 59 patients with non‐AC histology, SQC histology was noted in 38 (64.4%). There was no significant difference in OS between patients with and without previous RT in our study; however, any sequential treatment after nivolumab may have biased these results. In addition, the frequency of palliative RT administration after nivolumab treatment was not significantly different between patients with and without previous RT; thus, it did not affect the survival difference between the groups. We believe that any previous RT contributes to prolonged OS after the initiation of nivolumab. Based on these results, further investigation in prospective studies evaluating the efficacy of nivolumab following any previous RT is warranted. Extracranial RT seemed to increase the efficacy of nivolumab administration more than any RT, including for brain metastases, although the mechanism remains unclear. Shaverdian *et al.* reported that any previous treatment with RT in patients with advanced NSCLC receiving pembrolizumab was associated with longer PFS and OS.[9](#tca13044-bib-0009){ref-type="ref"} They analyzed the clinical features of a subset of 97 patients administered pembrolizumab in the phase I KEYNOTE001 trial. Forty‐two (43%) of the 97 patients were administered any previous RT before the initiation of pembrolizumab, 38 (39%) were administered extracranial RT, and 24 (25%) were administered thoracic RT. The PFS (6.3 months) and OS (11.6 months) rates of patients who underwent extracranial RT were significantly longer than in patients who did not undergo extracranial RT (2.0 and 5.3 months, respectively). In their analysis according to the type of previous RT administered, extracranial RT seemed to improve prognosis after pembrolizumab compared to any RT. This phenomenon suggests that previous RT, except to the brain, strongly contributes to the synergistic effect of ICIs, which is similar to the results of our study. Fiorica *et al.* also confirmed the synergistic effect of RT on nivolumab against advanced NSCLC.[8](#tca13044-bib-0008){ref-type="ref"} Their study included 15 patients previously administered RT and 20 patients never administered RT, and PFS and OS after nivolumab treatment were compared. The outcome of patients administered RT prior to nivolumab treatment was markedly better than that of patients who were not. However, the relationship between the response rate of ICIs and previous RT remains unclear. To our knowledge, our study is the first to verify the improvement in the ORR of nivolumab after any previous RT. Nivolumab administration after RT increased the response rate nearly two‐fold (36.4% vs. 18.9%); a favorable trend was also observed in patients with non‐AC histology and wild‐type *EGFR*. Patients with SQC histology achieved an ORR of 52.6% and a DCR of 89.4% with nivolumab treatment. We found that previous RT yields a different synergistic effect in the response to nivolumab treatment according to histological type. The combined sequence of RT and nivolumab may be a promising treatment in patients with non‐AC histology, particularly SQC histology. Further study is warranted to elucidate the additive effect of radiation on the efficacy of nivolumab according to the histological type in advanced NSCLC. Recently, Britschgi *et al.* reported the existence of abscopal effects induced by RT and nivolumab in patients with metastatic NSCLC, suggesting that this represented clinical determination of the synergistic effect of local RT and ICIs.[11](#tca13044-bib-0011){ref-type="ref"} An abscopal effect is a phenomenon wherein untreated tumor lesions regress after local treatment, such as RT. Theoretically, radiation‐triggered antitumor T cells are thought to kill tumor cells outside the irradiated tumor sites.[12](#tca13044-bib-0012){ref-type="ref"} However, this depends on many factors, such as whether tumor‐specific T cells are induced by RT and are effective for tumor control. In a preclinical study, Zhang *et al.* reported that the synergistic local and abscopal effects of hypofractionated RT and anti‐PD‐1 treatment are different in different tumor cell lines.[13](#tca13044-bib-0013){ref-type="ref"} Yuan *et al.* presented a case of lung SQC with systemic tumor regression by RT even after nivolumab had failed.[14](#tca13044-bib-0014){ref-type="ref"} Their result suggests that RT stimulated immune activation in nivolumab‐refractory circumstances and activated T cells killed tumor cells, even outside irradiated sites. Moreover, Meng *et al.* reported that cells in the immune microenvironment, such as CD4+, CD8+, and FOXP3+ tumor‐infiltrating lymphocytes, are likely different between non‐SQC and SQC patients.[15](#tca13044-bib-0015){ref-type="ref"} Although it is unknown whether the synergistic effect of RT and ICIs is stronger in SQC than in AC tumors, our results suggest that nivolumab has a stronger synergistic effect after previous RT in patients with SQC histology than those with AC histology. Studies have evaluated the adverse events in patients treated with both ICIs and thoracic RT.[8](#tca13044-bib-0008){ref-type="ref"}, [9](#tca13044-bib-0009){ref-type="ref"}, [16](#tca13044-bib-0016){ref-type="ref"}, [17](#tca13044-bib-0017){ref-type="ref"} No statistically significant difference in the frequency of grade 3 or higher pulmonary toxicities was observed in patients with or without a history of previous thoracic RT prior to ICI treatment. Our results are consistent with these findings. Therefore, RT prior to nivolumab administration is acceptable in terms of safety. Our study has several limitations. First, as this was a retrospective study with a small sample, bias may be present in our results. Second, the patients in our study received different total doses and schedules of RT according to the stage or extent of their tumor, therefore we could not confirm whether the efficacy of nivolumab varies according to the total RT dose. However, an optimal trend between the presence of previous RT and the efficacy of nivolumab was identified. Finally, it remains unclear why previous RT enhances the response rate of nivolumab. Although many experimental studies have explored the relationship between radiation and immune reaction, little is known about the clinical significance of RT as a sensitizer to immunotherapy. Our investigation verified that previous RT enhances the response rate of nivolumab and contributes to the prolongation of disease‐free survival. In the randomized phase III PACIFIC study, the anti‐PD‐L1 antibody durvalumab significantly improved the survival duration of patients with locally advanced NSCLC administered platinum‐based concurrent chemoradiotherapy.[18](#tca13044-bib-0018){ref-type="ref"} Therefore, the prognostic significance of administering an anti‐PD‐L1 antibody following thoracic radiation has been established in patients with locally advanced disease. Currently, there are planned or ongoing prospective clinical studies investigating RT combined with ICIs in patients with NSCLC.[17](#tca13044-bib-0017){ref-type="ref"} In conclusion, RT prior to nivolumab therapy influences the efficacy of nivolumab and is intricately linked with favorable prognosis after immunotherapy. Extracranial RT has been clinically identified as a better modality to improve the efficacy of nivolumab compared to any other RT, including for brain metastases. Further investigation is required to establish the promising sequential strategy of nivolumab followed by RT. Disclosure {#tca13044-sec-0011} ========== OY, AM, KK, and HK received research grants and speaker honorarium from Ono Pharmaceutical Company and Bristol‐Myers Company. All remaining authors report no conflict of interest. Supporting information ====================== ###### **Figure S1.** The median progression‐free survival rates of patients administered any previous radiotherapy (RT, *n* = 66), extracranial RT (*n* = 52), thoracic RT (*n* = 40), and no previous RT (*n* = 58) were 204, 206, 233, and 79 days, and the median survival times were 562, not reached (NR), NR, and 524 days, respectively. ###### Click here for additional data file.

Cedell[@bib1] described talus fracture in 4 young active sportsmen in 1974. These injuries are rare, and from then until now, only a few cases have been reported. The patient may arrive in the emergency department with pain and tenderness in the posteromedial region of the talus. However, most of the time, the injury\'s similarity to an ankle sprain on normal anteroposterior and lateral ankle radiographic views may contribute to misdiagnosis or delayed diagnosis.[@bib2] Different treatments of this injury are supported in the literature: conservative treatment of nondisplaced or minimally displaced fractures, open reduction--internal fixation for displaced fractures, or fracture excision for malunion of displaced fractures that cause posteromedial ankle impingement.[@bib2], [@bib3] Technique {#sec1} ========= The diagnosis can be made using anteroposterior and lateral ankle radiographs ([Fig 1](#fig1){ref-type="fig"}) and computed tomography ([Fig 2](#fig2){ref-type="fig"}).Fig 1Preoperative anteroposterior and lateral radiographs of a right ankle showing a posteromedial talus fracture (arrow).Fig 2Preoperative 3-dimensional right ankle reconstruction showing a displaced posteromedial talus fracture with malunion (arrow). Preoperative Setup {#sec1.1} ------------------ The patient is placed in the prone position with application of a thigh tourniquet to provide a bloodless surgical field. All bony prominences are padded. A small support is situated under the lower leg, making it possible to move the ankle freely. The ankle should be kept in a plantar-flexed position to relax the neurovascular bundle. The operative leg is prepared and draped in standard fashion. No traction is required. Portal Placement {#sec1.2} ---------------- According to the recommendations of Van Dijk et al,[@bib4] the posterolateral portal is created through an incision at the level of or slightly above the tip of the lateral malleolus, just lateral to the Achilles tendon. Careful palpation of the Achilles tendon before portal insertion reduces the risk of damaging the tendon. Blunt dissection to the level of the joint is carried out with a small mosquito clamp directed anteriorly, pointing in the direction of the interdigital web space between the first and second toes. It is exchanged with a 4.5-mm arthroscope shaft (Arthrex, Naples, FL) with a blunt trocar. The posteromedial portal is made just medial to the Achilles tendon, at the same level as the posterolateral portal. A mosquito clamp is introduced and directed toward the arthroscope shaft at 90° to touch this shaft; it should pass the neurovascular bundle without a problem. A 4.0-mm 30° arthroscope (Arthrex) is used through the posterolateral portal. The posterolateral portal is used as the viewing portal, and the posteromedial portal serves as the working portal ([Fig 3](#fig3){ref-type="fig"}). If there is any difficulty in determining whether the arthroscope is inserted correctly, confirmation of its position by fluoroscopy is recommended.Fig 3The patient is in the prone position. Portal placement is shown for the right ankle. Placement of the posterolateral portal is performed just lateral to the Achilles tendon at the level of the tip of the malleolus. A 4-mm 30° arthroscope is in position. The posteromedial portal is made just medial to the Achilles tendon, at the same level as the posterolateral portal. A shaver is introduced and directed toward the arthroscope shaft. Surgical Correction of Posterior Ankle Impingement {#sec1.3} -------------------------------------------------- Before the surgeon addresses the pathology, it is paramount to identify the flexor hallucis longus (FHL) and confirm that it moves with passive motion of the hallux. A base loop is passed around the tendon. The FHL is used as the medial border of the working area, which helps prevent injury to the neurovascular bundle. Then, debridement starts with an arthroscopic shaver (Arthrex) and radiofrequency device (Arthrex) inserted through the posteromedial portal.[@bib4] When soft-tissue debridement is complete, the fracture malunion is clearly defined ([Fig 4](#fig4){ref-type="fig"}). Then, it is time to perform resection. This may include partial or total resection to correct the impingement ([Fig 5](#fig5){ref-type="fig"}). We perform partial resection of the fragment with a motorized 4.0-mm burr (Arthrex) to restore the normal shape of the posterior talus. We check, under arthroscopic control, that range of ankle motion is completely restored and posterior ankle impingement is corrected ([Fig 6](#fig6){ref-type="fig"}).Fig 4Arthroscopic view of the posteromedial aspect of the right ankle from the posterolateral portal. The fracture malunion (yellow arrow) and posterior ankle impingement are visualized. A shaver is introduced through the posteromedial portal. The flexor hallucis longus is on the medial side with a base loop around it (green arrow).Fig 5Arthroscopic view of the posteromedial aspect of the right ankle from the posterolateral portal. Partial resection of the fracture malunion (arrow) is performed. A burr is introduced through the posteromedial portal.Fig 6Arthroscopic view of the posteromedial aspect of the right ankle from the posterolateral portal. Final reshaping of the posterior talus is performed, and posterior ankle impingement (red arrow) is corrected. The flexor hallucis longus is on the medial side with a base loop around it (green arrow). Complications {#sec1.4} ------------- Although we have not encountered any complications with this procedure, the major complication is injury to the tibial neurovascular bundle. It is recommended to keep the ankle in plantar flexion during the procedure and keep the instruments lateral to the FHL. Postoperative Rehabilitation Protocol {#sec1.5} ------------------------------------- The patient is discharged on the day after surgery using crutches. Weight bearing is allowed. A rehabilitation program to gain full mobility and strength is followed. A control computed tomography scan is ordered 1 month after surgery ([Fig 7](#fig7){ref-type="fig"}). Return to sports is gradually permitted, based on functional demands, with the most demanding activities being avoided until 3 months after ankle arthroscopy.Fig 7Three-dimensional reconstruction 1 month after arthroscopic resection of a malunion of a posteromedial talus fracture showing correction of posterior ankle impingement (arrow). A step-by-step summary of our technique is provided in [Table 1](#tbl1){ref-type="table"}. Pearls and pitfalls are presented in [Table 2](#tbl2){ref-type="table"}, and advantages and disadvantages are listed in [Table 3](#tbl3){ref-type="table"}. Key steps of the procedure are shown in [Video 1](#appsec1){ref-type="sec"}.Table 1Step-by-Step Summary of Arthroscopic Treatment of Malunion of Posteromedial Talus Fracture1.Position the patient in the prone position with application of a thigh tourniquet to provide a bloodless surgical field.2.Keep the ankle in a plantar-flexed position to relax the neurovascular bundle.3.Establish the posterolateral and posteromedial portals.4.Use a 4-mm 30° arthroscope inserted through the posterolateral portal.5.Identify the FHL and confirm that it moves with passive motion of the hallux.6.Start debridement with an arthroscopic shaver and radiofrequency device inserted through the posteromedial portal.7.Identify the fracture malunion.8.Perform partial resection of the fragment with a motorized 4.0-mm burr.9.Check, under arthroscopic control, that range of ankle motion is completely restored and posterior ankle impingement is corrected.[^1]Table 2Pearls and PitfallsPearlsPitfallsUse the prone position; pad all bony prominences.Pressure sores and lateral femoral cutaneous nerve neuropathyNote that no traction is required.No free movement of ankleKeep the ankle in a plantar-flexed position.Neurovascular bundle injuryPerform careful palpation of the Achilles tendon.Achilles tendon injuryIdentify the tip of the lateral malleolus for correct portal placement.Incorrect portal placementUse the arthroscope shaft as a guide to place the posteromedial portal.Neurovascular bundle injuryUse fluoroscopy to confirm the position and direction of the arthroscope if necessary.Incorrect portal placementIdentify the FHL by passive motion of the great toe and pass a base loop around the tendon.Neurovascular bundle injuryPerform a detailed examination of posterior ankle impingement.Incorrect portal placementIdentify the fracture malunion.No identification of fracture malunionReshape the posterior talus.No impingement correctionAvoid excessive bone resection.Excessive bone resection leading to ankle instabilityCheck complete ankle motion and correction of posterior ankle impingement.No improvement in clinical resultsTable 3Advantages and DisadvantagesAdvantages The procedure allows excellent access to the posterior ankle compartment. The procedure is minimally invasive. The procedure allows posterior ankle impingement to be corrected. The recovery time is shorter.Disadvantages The technique is challenging. Neurovascular bundle injury can occur. An experienced arthroscopist is required. The operative time is longer. Discussion {#sec2} ========== Posteromedial talus fracture has a potential for delayed diagnosis.[@bib2], [@bib3], [@bib5] First, it is a rare injury. Second, it shows similarity to an ankle sprain on standard ankle radiographs. To avoid future morbidity, it is important to diagnose this fracture at the time of the initial presentation. Careful clinical and radiographic evaluation is required to obtain a prompt diagnosis. Tenderness over the deltoid ligament on physical examination should raise suspicion. If no abnormal findings are observed on routine trauma views of the ankle, a more precise examination should be scheduled. Ebraheim et al.[@bib6] described, in a cadaveric study, the utility of a 30° external rotation view of the ankle as a radiologic method to diagnose a posteromedial tubercle fracture of the talus. They recommended its use with the 3 routine trauma views of the ankle. A computed tomography scan and magnetic resonance imaging should be requested if radiographic evaluation findings are normal. Malunion of a posteromedial fracture might be the cause of posteromedial ankle impingement and persistent posteromedial ankle pain. Bone excision is recommended in these cases.[@bib2], [@bib3], [@bib7] Open surgical treatment through a posteromedial approach has the risk of damaging the neurovascular structures. If it is necessary to improve visualization of the ankle and subtalar joints, an external fixator should be placed intraoperatively. Often, open surgical treatment is followed by a short period of immobilization. The arthroscopic approach has many advantages. It is a surgical procedure with less morbidity. It offers good access to the posterior ankle compartment. Malunion of a posteromedial fracture can be visualized and treated, and no immobilization is required.[@bib2], [@bib4] We recommend this surgical procedure to perform a partial resection of a malunion of a posteromedial fracture to correct posterior ankle impingement. Supplementary Data {#appsec1} ================== Video 1Arthroscopic treatment in the right ankle of a 23-year-old male patient with a posteromedial fracture malunion causing posteromedial ankle impingement. Radiographs and computed tomography show the fracture malunion. The patient is placed in the prone position. The procedure is carried out using the 2 standard posterior portals. The posterolateral portal is used as the viewing portal, and the posteromedial portal serves as the working portal. The flexor hallucis longus (FHL) is identified on passive motion of the great toe, and a base loop is passed around it. Debridement is performed. The fracture malunion is identified. The dynamic impingement is identified with passive motion of the ankle. Partial resection of the fragment is performed. Full ankle motion is checked under arthroscopic control. Posterior ankle impingement correction is checked under arthroscopic control.ICMJE author disclosure forms The authors report that they have no conflicts of interest in the authorship and publication of this article. Full ICMJE author disclosure forms are available for this article online, as [supplementary material](#appsec1){ref-type="sec"}. [^1]: FHL, flexor hallucis longus.

Introduction {#Sec1} ============ Organic light-emitting diodes (OLEDs) are being widely applied in displays for mobiles and televisions owing to their self-emitting characteristics, excellent colour gamut, high-speed operation, and applicability in flexible or stretchable devices^[@CR1]^. Practical OLEDs function based on phosphorescence owing to their nearly 100% internal phosphorescence efficiency^[@CR2],[@CR3]^. However, the external quantum efficiency of OLEDs is \~20% because the surface plasmon polariton modes at the metal-organic interface, the waveguide mode in the indium tin oxide (ITO)/organic layer, and the substrate mode in the glass substrate produce light loss^[@CR4]--[@CR8]^. Generally, light loss in the substrate mode is larger than that in the waveguide mode due to small differences in refractive index between the ITO/glass interface and the glass/air interface^[@CR9]^. In addition, microcavity top-emitting organic light-emitting diodes (TEOLED) without the loss of the substrate mode and having the advantage of improved colour purity and aperture ratio have been applied to mobile displays. However, microcavity TEOLED originally had a problem associated with viewing angles due to the blue shift that occurred with the variation of the sight angle, even though the light efficiency along the normal direction was amplified. Therefore, researchers have attempted to find solutions for improving the light extraction and viewing angle of OLEDs^[@CR10]^. Different approaches have been adopted with the aim of optimizing light extraction and viewing angle, including micro lens array (MLA)^[@CR5],[@CR11]--[@CR16]^, dielectric Bragg gratings^[@CR17]--[@CR19]^, surface plasmons^[@CR20]^, buckling patterns^[@CR21]^, periodic corrugation^[@CR8]^, low-index grids^[@CR22]^, and photonic crystals^[@CR23]^. Methods that modify the internal or external surfaces of the substrate of OLEDs minimize the total internal reflection. Such surface shapes are produced by different technologies which traditionally involve photo-lithography or printing, moulding, and embossing methods^[@CR24]--[@CR28]^. Therefore, these technologies employ at least the photo-lithographic step or utilize lithographic templates once. They generally end up being considerably complex or expensive for mass production. On the other hand, other researchers have reported a simple scattering layer synthesized by competitive and scalable methods. In these methods, particles or voids having different refractive indices were embedded in a polymer matrix and utilized for the scattering layer^[@CR29]--[@CR36]^. However, because the particles or voids are micro-sized or the pitch between the nearest particles or voids is micro-sized, these techniques reveal a small effect of diffraction that is difficult to control. In addition, to prevent the loss of the substrate mode, the most representative MLA has larger visibility than a nano-sized array because of its size. Alignment is a problem in the case of high-resolution displays with very small pixels. In this paper, we propose a simple method to fabricate random nanoscale rods (RNRs) as scattering layers for enhancement of light extraction and improvement of viewing angle characteristics in OLEDs. In our device, there is virtually no spectral distortion or shift, which are issues observed in 2D photonic crystal-based OLEDs, and the spectral shift according to the change in viewing angle is reduced. This layer is fabricated by a cost-effective and scalable procedure that consists of coating and etching the polymer at low temperatures (below 100 °C) without photomasks or templates. Additionally, the RNRs can control the diffraction as well as scattering effects in the visible wavelength range owing to the distance between closest rods and are compatible with high-resolution displays for virtual reality because the process of alignment between the individual pixels in the panel and the scattering layer is unnecessary. The RNRs can also be applied in flexible displays and lighting owing to the use of etched polymers. OLEDs equipped with RNRs exhibit superior optical properties such as high external quantum efficiency (EQE), high luminance efficiency (LE), and colour variation with changes in viewing angle, to control OLEDs. Results and Discussion {#Sec2} ====================== To investigate the optical effect of RNRs on glass, the electrical field distribution was simulated by the finite-difference time domain (FDTD) method. The boundary conditions for the simulation were set for a perfect optically matched layer to avoid the reflection of electromagnetic waves at the edges of the structure on all the sides, except for the metal cathode layer. The simulated structure consisted of an aluminium cathode, *N*,*N*'-bis(naphthalen-1-yl)-*N*,*N*'-bis(phenyl)-benzidine (NPB; refractive index n = 1.81), tris(8-hydroxy-quinolinato)aluminium (Alq~**3**~; n = 1.72), ITO (n = 1.9), glass (n = 1.52), SU-8 polymer, and RNRs (n = 1.59). A dipole source of visible wavelength was generated in the Alq~**3**~ layer, and one transverse electric (TE) and two transverse magnetic (TM) modes were used^[@CR37]^. To reflect the randomness of RNRs in the simulation, their height was varied from 1000 nm to 1200 nm and width from 50 nm to 150 nm (Supplementary Information Fig. [S1](#MOESM1){ref-type="media"}). In the reference structure, most of the light incident at an angle above the critical angle became a waveguide because of the difference in refractive index between glass and air (Fig. [1a](#Fig1){ref-type="fig"}). In order to confirm the optical effect of the SU-8 polymer, a simulation using SU-8 without the pattern was performed (Fig. [1b](#Fig1){ref-type="fig"}). As a result, the optical path was almost unchanged, and the difference between the refractive indices of glass and air slightly increased; this difference seemed to be slightly adversely affected for external light extraction. In contrast, in the RNRs present on the glass structure, some portions of the waveguide light were allowed to escape from the glass substrate, and light extraction increased below the critical angle (Fig. [1c](#Fig1){ref-type="fig"}). Optical phenomena such as refraction, diffraction, and interference were produced by the RNRs because the differences in refractive index at the interface between glass and air changed and the random periodic rods revealed distances between each other that were similar to the visible wavelength. Figure [2e](#Fig2){ref-type="fig"} illustrates the light path in OLEDs. The photons trapped (rays 2 and 3) within the glass are scattered by the RNRs, increasing the probability of them being emitted from the structure. Furthermore, as a result of simulation, in which scattering occurs several times in one rod, it can be inferred that scattering occurs more frequently when the height of the rod is greater.Figure 1FDTD simulation of E-field distribution induced by transverse magnetic and electric dipoles: (**a**) Reference (w/o RNRs), (**b**) unpatterned SU-8 on a glass structure, and (**c**) RNRs on the glass structure.Figure 2Cross-sectional scanning electron microscopy images of random nanoscale rods (RNRs) obtained under different plasma etching conditions: (**a**) RNRs 1, (**b**) RNRs 2, (**c**) RNRs 3, and (**d**) RNRs 4, and (**e**) schematic illustration of OLED outcoupling enhancement by the RNR scattering layer. It is desirable to fabricate a nanostructure of proper height and periodicity that is comparable to the wavelength of the visible range in order to enhance light extraction. For manufacturing RNRs, SU-8 polymer was chosen because it is highly transparent to visible light and has a higher refractive index than glass. Figure [2](#Fig2){ref-type="fig"} shows the cross-sectional scanning electron microscopy images of the RNRs obtained under different plasma etching conditions. Obtained by using only an oxygen plasma for a duration of 9 min (denoted as RNRs 1), the structure was random columnar-like with high aspect ratio and density^[@CR38]^. The height and density of RNRs 1 were about 1200 nm and 9.7 ea/µm^2^, respectively (Fig. [2a](#Fig2){ref-type="fig"}, Supplementary Information Fig. [S2a](#MOESM1){ref-type="media"}). The width of the individual rods was observed to be approximately 50 nm. To control the height and density, we applied an additional argon plasma treatment at higher powers and lower process pressures than the oxygen plasma. The low pressure led to an increase in the mean free path of argon ions and reduced the energy loss of these ions as they approached the RNRs. In general, the etching process using argon plasma carried out in a reactive ion etcher resulted in anisotropic and directional features due to the utilization of the physical bombardment energy^[@CR38]^. As a result of these features, the height and density were adjusted as much as a target to the little changed width of the RNRs. The three treated RNRs shown in Fig. [2b--d](#Fig2){ref-type="fig"} corresponded to argon plasma treatment durations of 3, 6, and 9 min (denoted as RNRs 2, RNRs 3, and RNRs 4, respectively), while the conditions of the oxygen plasma process remained unchanged. As the duration of the argon plasma treatment increased, the height decreased to 950, 580, and 440 nm, respectively, and the density of the RNRs reduced (Supplementary Information Fig. [S2](#MOESM1){ref-type="media"}). Figure [3](#Fig3){ref-type="fig"} shows the total transmittance and haze as functions of wavelength for the different RNRs. According to the material data sheet provided by Micro-Chem., SU-8 is highly transparent with little absorption in the visible range. The optical properties of unpatterned SU-8 fabricated by spin coating showed that the total transmittance was over 95% while the haze was less than 1%. All the RNRs exhibited total transmittance values above 90%. Therefore, the RNRs were thought to be appropriate films for light extraction, with minimal light loss through absorption and back scattering. The haze slightly increased from 30.1% (RNRs 1) to 31.5% (RNRs 2) at 520 nm for an additional argon plasma time of 3 min and decreased to 19.4% (RNRs 3) and 17.8% (RNRs 4) for treatment times of 6 min and 9 min. Haze was calculated using equation ([1](#Equ1){ref-type=""}).$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Haze=\frac{Total\,transmittance-specular\,transmittance}{Total\,transmittance}$$\end{document}$$Figure 3Total transmittance and haze as functions of visible wavelength of different RNRs and unpatterned SU-8. In the case of RNRs 1, haze was affected by the reduced total transmittance, which was attributed to increases in back scattering and the amount of trapped light in RNRs when the height of the rod was greater than a specific thickness. Besides, as the shapes of rods were longer, it was not perfectly straight, and therefore, back-scattering and light trapping seemed to occur more than in the cases of the other RNRs. The density of the rods also had an effect on the total transmittance, and RNRs 1, with the highest density, seemed to be one of the reasons for the decrease in the total transmittance. In the cases of RNRs 3 and RNRs 4, the lower heights and densities led to decreases in haze due to lower scattering probabilities. Considering the total transmittance and haze, the treatment condition of RNRs 2 was expected to produce the optimal height and density of rods for enhanced light extraction of OLEDs. The electrical characteristics of the device with and without the RNRs were almost identical since they were placed, where they do not affect the devices (Fig. [4a](#Fig4){ref-type="fig"}). The OLED with unpatterned SU-8 showed the identical electrical properties and a slightly lower external quantum efficiency (EQE; Supplementary Information Fig. [S3](#MOESM1){ref-type="media"}). The lower EQE is attributed to the fact that the refractive index of the SU-8 film is higher than that of the glass and that weak interfacial reflections occur between the glass and the SU-8 film. However, the OLEDs with the RNRs exhibited higher efficiencies than those without the RNRs in terms of overall current density. The OLED treated with oxygen plasma for 9 min and argon plasma for 3 min (RNRs 2) showed the maximum EQE of 1.42%, while the reference device exhibited an EQE of 1.26%. The enhancement in the maximum EQE was 13% in the normal direction. Considering the scattering characteristics, the films with RNRs 1 and RNRs 2 were similar, but it was estimated that the difference in total transmittance caused a gap in the EQE. The maximum EQEs of RNRs 1, 3, and 4 were 1.4%, 1.32%, and 1.34%, respectively, which increased relative to the reference device. The current and power efficiencies of RNRs 2 were improved by 3.89 cd/A and 1.63 lm/W, respectively, relative to the values of 3.39 cd/A and 1.42 lm/W of the reference device. RNRs 1, 3, and 4 showed values of 3.81 cd/A, 3.58 cd/A, 3.64 cd/A and 1.71 lm/W, 1.50 lm/W, 1.52 lm/W, respectively.Figure 4EL characteristics of OLEDs with different RNRs: (**a**) J-V and L-V characteristics, (**b**) external quantum efficiencies as functions of current density, and (**c**) current and power efficiencies as functions of luminance. Table [1](#Tab1){ref-type="table"} shows the densities and average pitches of the RNRs calculated using image processing software (Image-Pro Plus 4.5, Media Cybernetics, Inc.). The densities of RNRs 1, 2, 3, and 4 were 9.7 ea/μm^2^, 7.1 ea/μm^2^, 5.1 ea/μm^2^, and 3.5 ea/μm^2^, respectively, and decreased with the increase in argon plasma time. The reduction in density from RNRs 1 to RNRs 2 was 2.6 ea/μm^2^, whereas that from RNRs 3 to RNRs 4 was 1.6 ea/μm^2^. It was presumed that this was due to the initial removal of RNRs that were not straight-shaped among the RNRs etched by the oxygen plasma. The average pitch was derived from the density and calculated using the distance between the centre of a certain rod and the centre of the nearest rod, assuming that the rod was circular in shape. Though the distances between the RNRs were not always identical due to random periods, the calculated average distance between closest rods suggested that RNRs 2 more likely produced diffraction than the other rods^[@CR21]^.Table 1Density, calculated average pitch, total transmittance, haze, and the enhancement in EQE for different RNRs.RNRs 1RNRs 2RNRs 3RNRs 4Density (ea/μm^2^)9.77.15.13.5Calculated Average Pitch (nm)361423500603Total Transmittance (%)92.294.594.895.8Haze (%)30.131.519.417.8Enhancement in EQE (%)21.023.49.212.5 Figure [5](#Fig5){ref-type="fig"} shows that the OLEDs with RNRs exhibited improved luminance intensities with changes in viewing angle from 0° to 70° due to scattering effect. The luminance was measured with a constant current of 7 mA, and the reference device had a luminance of 578 cd/m^**2**^ in the normal direction. Since the luminance in the normal direction differed according to the RNR condition, it was normalized based on luminance in the normal direction. Under the conditions of RNRs 1 and RNRs 2, the normalized luminance intensity increased by 13.6% and 14.1%, respectively, with respect to the reference device at the angle of 40°, and these two conditions revealed nearly similar viewing angle characteristics from 0° to 70°. The devices with RNRs 3 and RNRs 4 also exhibited enhanced luminances of 9.0% and 7.5%, respectively, compared to the reference device at 40°. It was observed that this viewing angle characteristic showed a similar tendency to the haze values of the RNRs. For the best conditions (RNRs 2), the improvement in the EQE was 23.4% relative to the reference device. The viewing angles of RNRs 1, RNRs 3, and RNRs 4 also increased by 21.0%, 9.2%, and 12.5%, respectively. Therefore, it was confirmed that the efficiency and viewing angle improved by reducing the loss of the substrate mode based on the scattering characteristics of the RNRs. As with the haze properties, the unpatterned SU-8 exhibited almost identical viewing angle characteristics as the reference device (Supplementary Information Fig. [S4](#MOESM1){ref-type="media"}).Figure 5Normalized angular luminance distributions of OLEDs operating at 7 mA between 0° and 70°. We also investigated the light extraction efficiency of green phosphorescent OLEDs to re-verify the effect of RNRs due to the low EQE of fluorescent OLEDs with NPB/Alq~3~. The green phosphorescent OLEDs without RNRs, with unpatterned SU-8, and with RNRs 2 (optimal condition) were fabricated. The electrical characteristics of the devices were almost identical (Fig. [6a](#Fig6){ref-type="fig"}). However, the OLED with RNRs 2 exhibited higher efficiencies than that without the RNRs or with unpatterned SU-8 in terms of overall current density. The maximum EQE of the OLED with RNRs 2 was 20.92%, while those of the reference and the OLED with unpatterned SU-8 were 17.62% and 17.94%, respectively. The enhancement in the maximum EQE was 18.7% relative to that of the reference device in the normal direction. In addition, the EQEs of the OLEDs with RNRs 2, without RNRs and with unpatterned SU-8 were evaluated as 12.5%, 11%, and 10.8%, respectively, at 20 mA/cm^2^. The enhancement in the EQE was 13.6%, which was similar to the 13% enhancement observed in the case of fluorescent OLEDs. Moreover, the EQE of the OLED with unpatterned SU-8 was similar to that of the reference device, indicating that the SU-8 film did not contribute to the light extraction effect.Figure 6EL characteristics of green phosphorescent OLEDs with RNRs 2, without RNRs (reference), and with unpatterned SU-8: (**a**) J-V and L-V characteristics, (**b**) external quantum efficiencies as functions of current density, and (**c**) current and power efficiencies as functions of luminance. Figure [7](#Fig7){ref-type="fig"} shows the variations in the luminance intensities of the green phosphorescent OLEDs for three different conditions according to the changes in viewing angle from 0° to 70°. As with fluorescent OLEDs, the different intensities were normalized to the luminance in the normal direction, and the reference device displayed a luminance of 906 cd/m^2^ at the constant current of 0.9 mA. In the case of the green phosphorescent OLED with RNRs 2, the luminance improved relative to that of the reference at all viewing angles due to the scattering effect of RNRs 2. However, in the case of unpatterned SU-8, the change in luminance characteristics according to viewing angle was almost the same as that of the reference. These results show that the scattering effect of the unpatterned SU-8 film is almost non-existent. Considering the viewing angle, the improvement in the EQE of the green phosphorescent OLED with RNRs 2 was 31% relative to that of the reference device. The difference between the EQE improvements of the fluorescence (23.4%) and phosphorescence (31%) OLEDs is presumed to be due to a deviation in the low EQE of the fluorescence device. Furthermore, as a result of the maximum difference in the colour coordinates in CIE 1931 for viewing angles varying from 0° to 70°, the reference device exhibited a variation of Δ(x, y) = (0.020, 0.034), whereas the RNRs 2 device revealed a variation of Δ(x, y) = (0.007, 0.014) (Supplementary Information Fig. [S5](#MOESM1){ref-type="media"}). This result also proved that the RNRs played appropriate roles as scattering layers.Figure 7Normalized angular luminance distributions of green phosphorescent OLEDs operating at 0.9 mA. We demonstrated a simple method for fabricating RNRs as a scattering layer at low temperatures without any mask. In particular, the height and density could be controlled without changing the width of the rods by utilizing the anisotropic etching characteristics of argon plasma. The optimal RNRs showed a haze of 31.5% despite the total transmittance being 94.5%. As a result, a 31% enhancement in the EQE of the OLED was achieved. FDTD simulation and efficiency improvement results indicated that the RNRs were suitable scattering layers that reduced the loss of the substrate mode. Moreover, the variations in light intensity and colour coordinates with viewing angle could be alleviated by using these RNRs. We thus believe that this study will open a new and practical approach to improving the performance of high-resolution displays and flexible lightings. Methods {#Sec3} ======= Numerical simulation {#Sec4} -------------------- The optical effects of OLEDs with RNRs were simulated using a FDTD software (Lumerical Solutions, Inc.). Each layer had a different refractive index that was measured by a thin-film analyser (F-20, Filmetrics, Inc.). The simulation domain condition involved a perfectly matched layer on all sides, except for the metal cathode layer, where the metal boundary condition was used. The three dipoles (x-, y-, and z-polarized) were placed inside the emitting layer. The optical enhancement image of the OLEDs with RNRs relative to the reference structure was calculated as the ratio of the integrated electric field intensities of the two devices that were measured by a fixed far-field monitor. Fabrication {#Sec5} ----------- The fabrication process of the RNRs as the scattering layer is shown in Fig. [8](#Fig8){ref-type="fig"}. The polymer SU-8 (SU-8 2010, Micro-Chem.) was mixed with a thinner in the ratio 1:1 by weight to achieve the target thickness. The mixed material was spin-coated, baked at 95 °C using a hot plate, and exposed at 130 mJ/cm^2^ to conventional UV radiation (350 nm--400 nm). As a result, the thin film was highly transparent to visible light^[@CR39]^. The film was etched in a reactive ion etcher (RIE) by using oxygen plasma at the RF power of 150 W, process pressure of 70 mtorr, and gas flow of 50 sccm. After the first etching of the polymer, the structures were random columnar-like, with high aspect ratios and densities^[@CR40]^. A second plasma treatment with argon gas at the high RF power of 200 W and process pressure of 25 mtorr was applied to the structure to reduce the aspect ratio and density. To investigate the properties of the RNRs, organic light-emitting devices were fabricated. To obtain devices with different properties, the duration of argon plasma treatment was altered while the other conditions were unchanged. After the RNRs were formed on glass substrate, ITO was deposited on the opposite side by radio frequency sputtering. The following layers were then thermally evaporated onto the ITO under vacuum conditions (2 × 10^−6^ Torr): a 60 nm thick NPB layer for hole transport, an 80 nm thick Alq~**3**~ layer for light emission, a 0.7 nm thick lithium fluoride layer for electron injection, and a 100 nm thick aluminium layer that acted as the cathode. In the case of green phosphorescent OLEDs, the following layers were utilized: a 40 nm thick NPB(N,N'- bis(naphthalen-1-yl)-N,N'-bis(phenyl)-benzidine) layer and a 10 nm thick TCTA (4,4′,4′′-Tris(carbazol-9-yl)triphenylamine) layer for hole transport, a 30 nm thick CBP(4,4′-bis(carbazol-9-yl)biphenyl) doped with 10 wt% Ir(ppy)3(Tris(2-phenylpyridine)iridium(III)) served as the phosphorescent green emitting layer; and 55-nm-thick B3PyMPM(bis-4,6-(3,5-di-3-pyridylphenyl)-2-methylpyrimidine) as used as the electron-transport layer, a 0.4 nm thick lithium fluoride layer for electron injection, and a 100 nm thick aluminium layer that acted as the cathode.Figure 8Schematic of the process flow for fabricating OLEDs with RNRs. Measurements {#Sec6} ------------ The surface morphologies and cross-sections of the RNRs were measured by scanning electron microscopy (S-4800, HITACHI High Technology Inc.). The haze and total transmittance were evaluated by UV-vis-NIR spectroscopy (Cary 5000, Agilent Technologies Inc.). The current--voltage characteristics were measured using a Keithley 237 High Voltage Source-Measure Unit (Keithley Instruments, Inc.) and the electroluminescence intensity was measured in a dark box using a spectroradiometer (PR-670 Spectra Scan, Photo Research, Inc.). Electronic supplementary material ================================= {#Sec7} Supplementary Information **Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Electronic supplementary material ================================= **Supplementary information** accompanies this paper at 10.1038/s41598-018-32538-4. This research was supported by a grant (No. 2016R1A2B4014073) from the National Research Foundation of Korea (NRF), funded by the Korean Government (MSIP), the Ministry of Education (No. NRF-2017R1D1A1B03036520), and the Industry Technology R&D program (10048317, Development of red and blue OLEDs with external quantum efficiency over 20% using delayed fluorescent materials) funded by MOTIE/KEIT. J.H.K. conceived the idea and performed the fabrication, measurements, and data analysis. J.C. performed the optical simulations. C.P. and H.H. measured the optical properties of the RNR films and analysed the data. B.K.J. and Y.W.P. directed the research project. All the authors discussed the results and commented on the manuscript. Competing Interests {#FPar1} =================== The authors declare no competing interests.

Introduction {#ss1} ============ Throughout Europe birth-rates have declined over the last decades, and in most countries they are well below the replacement level ([@CIT0001]). In Sweden, the number of children being born per woman has fluctuated greatly since the mid-1970s but has now been predicted to stabilize at a level around 1.8 ([@CIT0002]). The declining birth-rate can partly be explained by an increasing age at first birth ([@CIT0003]); furthermore, today there is clear empirical evidence of the postponement of the first child in several countries ([@CIT0004; @CIT0005; @CIT0006]). Higher education has been associated with delayed childbearing ([@CIT0007]), and the association between female education and the age of becoming a first-time parent has been well documented ([@CIT0008; @CIT0009]), showing that highly educated women are more likely to pursue careers and postpone having children ([@CIT0005]). Women\'s labour force participation has been associated with postponement largely due to the incompatibility of caring for children and participation in the paid labour force. However, it has been possible to combine female employment and childbearing when the reduction in work--family conflict was facilitated by state or policy interventions, such as in some Scandinavian countries ([@CIT0010; @CIT0011]). It seems, however, that postponed childbearing is a result of several different reasons. For example, efficient and reliable oral contraception has had an impact on family planning in many modern societies ([@CIT0012; @CIT0013]). A US study found that multiple partnerships before marriage, higher levels of non-marital cohabiting, and difficulty in finding a suitable partner might contribute to later births ([@CIT0014]). It has also been reported that young adults delay childbearing until they are financially secure, so they can afford to support children ([@CIT0015]). In addition, the emergence and development of assisted reproductive technologies (ARTs) have been suggested to promote perceptions that childbearing can be resumed at a later and more convenient phase of life ([@CIT0016]). Surveys concerning attitudes towards future parenthood and awareness about fertility among undergraduate and postgraduate students in Sweden have shown that these young women and men had largely positive attitudes towards having children, but they were not sufficiently aware of the limitations associated with ageing ([@CIT0017; @CIT0018; @CIT0019]). Similar results have been reported in a recent study among Finnish university students, in which one-third of the women and more than half of the men believed that a marked decline in female fertility begins after the age of 45 ([@CIT0020]). In a Canadian survey among female undergraduate students, it was found that women were aware that fertility declines with age, but they overestimated the chances of pregnancy at any age and were not aware of the steep rate of fertility decline with age ([@CIT0021]). However, knowledge about the influence of female age on childbearing success was not predictive of their childbearing intentions. The adverse effects of ageing on reproduction are complex and multifactorial. It is known that women\'s ability to conceive starts to decline in their late 20s and rapidly falls by the mid-30s, mainly due to reduced quantity and quality of ovarian follicles ([@CIT0003; @CIT0022]). In addition, after the age of 35 an increase in pregnancy complications and prenatal maternal morbidity, as well as impaired prenatal and postnatal outcome of the child has been reported ([@CIT0023; @CIT0024; @CIT0025]). There is also increasing evidence that paternal age, at least over the age of 40, is associated with lower fertility-related outcome and anomalies in the offspring ([@CIT0026; @CIT0027]). Societal attitude towards family and children also forms the context in which the subjective preferences and assessments regarding family formation are made. The tendency towards individualism, the endeavour for self-fulfilment, increasing freedom of choice, and greater tolerance towards new lifestyles and reproductive behaviours have also been suggested to influence the decline in births and the postponement of childbearing ([@CIT0028]). Most research concerning postponed childbearing has focused on women and has mostly been based on standardized questionnaires or register data ([@CIT0029]); however, the information gathered from the viewpoint of highly educated young people themselves is more random. Thus, the aim of the present study was to gain a more comprehensive understanding of how young, highly educated women and men without children, who had started their professional career, reflect on fertility and postponed parenthood. Materials and methods {#ss2} ===================== We performed individual interviews with 22 women and 18 men between 24 and 38 years of age. The criteria for inclusion were 4 years or more of university education, having started a professional career, and not yet having children. The participants were recruited using advertisements placed in different work-places where the majority of staff had a higher education. Those who were interested in participating were requested to phone or send an e-mail to the project leaders for further information and arrangement of a suitable time for the interview. The interviews took place in a convenient room offered by the work-place, or outside the office. The interviews were audiotaped and lasted for about half an hour. The interview guide covered two main areas: 1) personal attitudes towards having children in the future and 2) reflections on fertility and postponement of parenthood. The findings related to personal attitudes towards future parenthood have previously been published ([@CIT0030]). All interviews were transcribed verbatim and analysed according to qualitative content analysis ([@CIT0031]). The transcripts were read several times to gain a sense of the whole. Constellations of words, sentences, or paragraphs related to the study aim were identified and condensed. Thereafter, codes were created, and finally the codes were sorted into seven subcategories and two categories according to their content and meaning. The process of developing categories is illustrated in [Tables I](#T1){ref-type="table"} and [II](#T2){ref-type="table"}. ###### Examples of condensed meaning units, codes, subcategories, and categories. Meaning units Codes Subcategory Category ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------ --------------------------------------------------------- ------------------------------------------------------------------ 'In a larger city one has so many other things to keep you busy anyway, so children come significantly later. Yes, but yeah, maybe it\'s time, I thought a little like that, that it\'s about time\' City life offers many alternatives to family formation A consequence of contemporary lifestyles Postponed parenthood---a rational adaptation to societal changes 'To delay having children is probably due to the increased fear of growing up, to seriously be an adult because it\'s so nice to have to worry only about yourself\' A sign of extended immaturity, or egocentricity 'Today it is expected to not only get an education but also to travel, learn a new language, live overseas and much more, and parenthood doesn\'t fit so well in the story\' Expectations of further learning and striving for experiences and adventures A consequence of competing priorities 'There are more women who study at universities, more women than men that graduate, and maybe the birth of a child is viewed as an obstacle to one\'s career, that must be the explanation\' Women\'s increased investment in higher education 'To have a child before 25 is not exactly something that is encouraged and maybe also not having children after 40 either\' Socially mediated perceptions about the timing of childbearing A consequence of prevailing discourses about parenthood 'There is such an incredibly negative view about having a child when you are young, and among my friends there are many that see it as completely abnormal to have a child before 30\' Early parenthood is unfavourable ###### Women\'s and men\'s reflections on fertility and postponed parenthood. Subcategories Categories --------------------------------------------------------- ------------------------------------------------------------------ Unconsidered and taken for granted Fertility---an imperceptive and retrievable capacity Unpredictable and different for women and men Restorable by medical technology Replaceable by alternatives to a biological child A consequence of contemporary lifestyles Postponed parenthood---a rational adaptation to societal changes A consequence of competing priorities A consequence of prevailing discourses about parenthood Results {#ss3} ======= In the following section, the two categories 'Fertility---an imperceptible and retrievable capacity\' and 'Postponed childbearing---a rational adaptation to societal changes\' are presented. Quotations from the interviews are used for illustration. Fertility---an imperceptible and retrievable capacity {#ss4} ----------------------------------------------------- The category 'Fertility---an imperceptible and retrievable capacity\' consisted of four subcategories: 'Unconsidered and taken for granted\', 'Unpredictable and different for women and men\', 'Restorable by medical technology\', and 'Replaceable by alternatives to a biological child\'. ### Unconsidered and taken for granted {#ss5} This subcategory consisted of statements showing that even though most informants wanted to have children in the future, some had never reflected on their own reproductive capacity. It also included comments that revealed perceptions about fertility as something natural. Some referred to a healthy body, while others put confidence in their genetic heritage. In addition, some held the opinion that there is no point in reflecting too much about one\'s fertility as it might cause unfounded worries. > My own fertility, nah, I have not thought about that; actually, I have never done that. (w12) > I presume that I can deliver what\'s required when it comes to that. (m17) > I have never had any diseases, and my body has always functioned as expected, so, I have no reason to doubt my fertility. (w1) > I don\'t think that you should plan and think about it so much, and maybe start worrying about it unnecessarily. (m5). ### Unpredictable and different for women and men {#ss6} The second subcategory involved expressions describing fertility as a capacity that could not be taken for granted. It also included comments revealing an awareness that difficulties could arise. For some, this was based on their own experiences, while others referred to relatives or friends. Difficulties related to fertility were foremost described as being associated with age and sex, and especially the reproductive ageing process in women. In relation to this, some mentioned miscarriage, preterm delivery, and having an unhealthy child as risks associated with advanced maternal age, while others had vague ideas about increased risks. On the other hand, some men accentuated that the perception of an insignificant role of paternal age might be more based on myth rather than on scientific evidence. In addition, the subcategory also included expressions of emotional distress, primarily related to the fear of finally being proved infertile. > When I was young, I took it for granted that everyone who wanted to could have a child, but now I realize that it isn\'t so easy to have all that work as it should, and when and how you want it to. (m18) > It is mostly my own age that I think about because the older you get, the harder it becomes to get pregnant, and that is the risk one takes when one waits. (w17) > Before one thought that maybe it was just women who bore the risk of having sick children or not having children at all; but in recent times, it has become known that even men\'s ability to reproduce declines with age, the quality of sperm deteriorates. (m4) > I have thought about my own fertility a lot, and I sometimes think that maybe I have 'missed the boat\', in that case, it would be an enormous disappointment. (w19) ### Restorable by medical technology {#ss7} The third subcategory included comments on the perception that most of the difficulties related to reproduction could be solved with medical interventions. Some informants emphasized that even though the process of investigation and treatment of involuntary childlessness might be an expensive, time-consuming, and draining procedure, the possibility to have a biological child was important. Even though many of the informants expressed that all accessible interventions or technologies to 'treat\' involuntary childlessness are justified, some of the men said that donor insemination was an exception. In addition, the subcategory also included statements indicating that most of the informants would make use of prenatal diagnosis in case of a future pregnancy. > Anyway, one can do an investigation first and then it depends on what the problem is, but I know that there is good medical help available, even if you end up being a little older. (w3) > All methods that are available to have a biological child are worth a try, I think, then all the assisted ways makes no difference. (m16) > If it should be me that is the missing link and for some reason can\'t get my partner pregnant, it would still feel very strange if she were to be fertilized with sperm from some other man because then it\'s not my child. (m6) > It, maybe, sounds cold but as I feel now, I would have an abortion if I found out that the foetus was not fully healthy, and then try again. (w1) ### Replaceable by alternatives to a biological child {#ss8} The last subcategory consisted of statements revealing that even though many of the women and men initially trusted and would use reproductive technologies, some considered adoption to be the most ethical and gender-equal choice in case of involuntary childlessness. It also included comments indicating that childlessness did not necessarily have to be a tragedy and could be compensated by other qualities in life. Moreover, some expressed that a rewarding professional career, or having an animal, could replace having children. > There are so many children that have no parents, so adoption is the most ethical choice. (w8) > I think right away about adoption, maybe because it would anyway feel more, yes, more equal in some way; and in addition, there is already a child there who can have a better life. (m9) > If it ended up that I couldn\'t have children, it would be bad but not some huge catastrophe, I can be happy with being a great uncle instead. (m2) > I feel like I have my child substitute in my horse, of course, it doesn\'t work to compare them, and I know that, but I feel completely satisfied with it. (w21) Postponed parenthood---a rational adaptation to societal changes {#ss9} ---------------------------------------------------------------- The category 'Postponed parenthood---a rational adaptation to societal changes\' consisted of three subcategories: 'A consequence of contemporary lifestyle\', 'A consequence of competing priorities\', and 'A consequence of prevailing discourses about parenthood\'. ### A consequence of contemporary lifestyle {#ss10} The first subcategory consisted of statements describing postponed parenthood as a city phenomenon in that living in a city offers many alternatives to family life. The postponement of childbearing was also perceived as a consequence of postponing the establishment of a stable relationship. In addition, 'a consequence of contemporary lifestyle\' also included comments revealing reflections on postponed parenthood as a result of increased individualization in society. Some viewed it as a sign of extended immaturity, others as a sign of increased egocentricity. Still, others expressed the opposite view, and perceived postponed parenthood as a sign of extended responsibility: > In small towns, you find your place much faster than in larger towns because it is, one can say, just two alternatives: start a family or move. (m15) > Today, you wait longer to have children, but it is really that you also wait longer, for the most part, to find a good partner and many live this life of 'Sex and the City\' for a very long time. (w4) > I think that we have become more comfortable and more selfish; there is so much more focus on what I, myself, must do, what I should be and what I should achieve. (w6) > I think, if anything, that one is more caring towards one\'s child and takes parenthood more seriously; you don\'t have children only because you want to, but also to be able to have something to offer a child. (m7) ### A consequence of competing priorities {#ss11} The second subcategory included expressions describing postponed parenthood as a result of the wide range of possibilities that young women and men have during early adulthood. A commonly described reason concerned education and the fact that more people today move on to higher education. In particular, women\'s investment in higher education was perceived as weighty with regard to the timing of parenthood. It also involved comments describing expectations of further learning and striving for experiences and adventures. > You are expected to have so much more than, for example, when my parents started their family in the 70s, if you then had housing and like some sort of security, it was quite enough. (w20) > There are, of course, more women than men that study at universities, more women that graduate, and maybe they feel that childbearing would be an obstacle to their career. (m12) > Today it is expected to not only get an education but also have time to travel, learn a new language, live overseas and much more, and parenthood and the family don\'t fit so well in the story, that is anyway what I think. (w10) ### A consequence of prevailing discourses about parenthood {#ss12} The third subcategory included comments describing postponed childbearing as a corollary of socially mediated perceptions about the timing of parenthood. In relation to this, the feature of 'early\' parenthood was perceived as unfavourable. It also involved expressions exposing perceptions about 'early\' parenthood as a sign of low ambitions. On the other hand, the subcategory also contained statements revealing that changed attitudes related to the timing of parenthood might be on the way. > To have a child before 25 is not exactly something that is encouraged and maybe also not having children after 40 either. (w13) > There is such an incredibly negative view about having a child when you are young, and among my friends there are many that see it as completely abnormal to have a child before 30. (m9) > People have children for many different reasons, but sometimes it feels like that some do it because they cannot think of anything else to do in life. (w8) > Before, I had thought that you had children around age 30, but now, I have so many friends in my age group that have children, so this trend to postpone having children is maybe going to change and that feels kind of good. (m6) Discussion {#ss13} ========== This study shows that highly educated, young women and men in contemporary Sweden have many competing priorities when planning and setting goals for their lives; whether and when to have children is only one of them. They describe that societal changes and expectations related to lifestyle and parenthood influence decisions relating to the timing of childbearing. The participants were fairly aware that there are limitations on human reproduction; however, they also believed that fertility problems could be restored or replaced by medical technology or by other alternatives to a biological child. The finding that fertility may be unconsidered and taken for granted has also been shown in other studies ([@CIT0017; @CIT0032; @CIT0033]); furthermore, previous research has found that women and men highly value the capability of reproduction ([@CIT0017; @CIT0018; @CIT0019; @CIT0020; @CIT0029; @CIT0034]). However, with efficient and successful contraception together with voluntary postponement of childbearing, this ability may remain 'unused\' until ages when the reproductive capacity has started to decline. This can, at least, have two consequences: the link between sexuality and reproduction becomes more vague, and relationships do not necessarily include becoming parents. Family planning services have so far been equal with contraceptive services, with the main focus of helping women to avoid unwanted pregnancies. Not much focus is placed on male responsibility for reproduction and on concrete planning for a future family with regard to both women and men. Women\'s and men\'s consultations with health care providers (for example, for family planning, testing for sexually transmitted infection, or cervical cancer screening) could be used to convey important pre-conception messages, for example, the age-related decline in human fertility. A potentially useful tool in this respect is a reproductive life plan (RLP), which is a set of goals related to having or not having children. The aim with an RLP is that women and men should reflect upon their reproductive intentions within the overall context of personal life goals and values ([@CIT0035]). Preliminary tests of the RLP have shown positive attitudes among patients ([@CIT0036]). Options for enhancing fertility have grown with the development of sophisticated techniques of assisted reproduction. However, conventional *in vitro* fertilization treatments (IVF) are not a guarantee for getting pregnant at advanced ages. The success rate for both natural fertility and assisted reproductive technologies is lower for women in their late 30s and 40s ([@CIT0037]). For most couples, third-party conception becomes an option once treatment with their own gametes has failed or if the couple possesses a genetic disorder they do not wish to pass on to their children. Thus far, oocyte donation is the only technique that has had a high success rate in women at advanced ages, which makes the treatment attractive for those who desire to bear a child even if the genetic link between the mother and child is lacking ([@CIT0038]). Internationally, assisted reproductive treatments differ between countries due to tradition and legislation. Additional treatments such as oocyte cryopreservation for storage and use later in life has become a more successful alternative in case of, for example, cancer during a woman\'s younger years; however, at the moment it is not a common option for the public. In Sweden surrogacy, embryo donation, and assisted reproductive treatment for single women are not yet permitted. The participants described the postponement of parenthood as a rational adaptation to societal changes. The socially accepted window for having children seems to have narrowed. Before becoming parents, certain life events, such as education, experiences and adventures, and the start of a professional career, should ideally have been accomplished. Early parenthood was considered unfavourable and a sign of low ambition, a finding consistent with an earlier Swedish study ([@CIT0039]). This is also in line with recent research showing that Swedish women in their mid-20s regard childbearing as a future project, as it would impede their free and active life since it demands structure, stability, and space ([@CIT0040]). Even if Sweden has generous parental allowances and a well-established child-care system, young women and men seem to consider parenthood as a major personal undertaking, and the responsibility of providing emotional as well as practical/economic support for the child is heavily dependent on the parents. To postpone parenthood until one\'s educational goals are fulfilled and a reasonable economy is secured, therefore, becomes crucial, especially since the parental allowance in Sweden is based on the level of income before pregnancy. Strengths and limitations {#ss14} ------------------------- We were able to recruit as many as 40 participants, which resulted in rich data. However, since the participants knew the topic for discussion beforehand, it is possible that those who were not planning to have children were less motivated to participate. The interviewers were midwives with a clear pre-understanding of reproductive health issues, which could represent a limitation but also a strength. It was obvious that the participants appreciated the midwifery expertise and took the opportunity to discuss reproductive health issues with the interviewers after finalization of the formal interview. The purpose with qualitative studies is to gain a deeper understanding of people\'s lived experiences, so striving for generalization is neither desirable nor possible. However, we believe that our findings could be transferred to similar populations and contexts, bearing in mind that only highly educated women and men were invited to participate. To ensure credibility, we described the entire process in detail and inserted quotations to make it possible to judge the credibility of our findings. Dependability was created by recruiting women and men of different ages, with different occupations and civil status, and by using the same interview guide. To establish confirmability, the researchers continuously discussed the interpretation of the data, until consensus was reached. Conclusion {#ss15} ========== This study indicates that highly educated young women and men in contemporary Sweden have many competing priorities when planning and setting goals for their lives, and whether or not to have children as well as when to have them is an example of this. They describe fertility as an imperceptible and retrievable capacity, and, in case of problems, they believe fertility could be restored or replaced by medical technology or by alternatives to a biological child. They also describe postponed parenthood as a rational adaptation to societal changes, including socially mediated perceptions about 'early\' childbearing as unfavourable. These findings suggest that increased information about the limitations of human reproduction is needed, but also that societal arrangements making it possible to have children at younger ages are of utmost importance. We would like to thank all informants. ***Declaration of interest:*** The study was funded by the Medical Faculty at Uppsala University, Sweden. The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. T.T. had the idea for the study. T.T. and C.E. conducted the interviews. C.E. carried out the analysis in collaboration with M.L., T.T., and A.S.S. All authors contributed to the final manuscript.

Introduction {#Sec1} ============ Pre-diabetes (pre-DM), typically defined as blood glucose levels above normal but below diabetes thresholds, has been increasing globally and has a high chance of developing diabetes mellitus (DM)^[@CR1]^. It is estimated that there were 318 million adults suffering from impaired glucose tolerance in the world by 2015^[@CR2]^. As to the Chinese adult population, a cross-sectional survey in 2010 reported that the estimated prevalence of pre-DM was 50.1%^[@CR3]^. Previous studies^[@CR4]^ have shown that about 5--10% of pre-DM progressed to DM every year, and persistent hyperglycemia leads to the complications that are the major source of morbidity, mortality, and cost. Nowadays, There is a common conception that this natural history is not inevitable^[@CR5]^. Randomized controlled trials showed that lifestyle intervention or glucose-lowering medications could delay and even reverse the natural course of pre-DM^[@CR6]--[@CR8]^. The American Diabetes Association (ADA) recommends at least annual screening via testing fasting plasma glucose (FPG), 2-hr 75 g oral glucose tolerance test (OGTT), or hemoglobin A1c (HbA1c) in those with pre-DM^[@CR9]^. However, clinicians are far from satisfied with the presently used method of monitoring glycemic level because of its hysteresis. FPG, OGTT and HbA1c are used more as a diagnostic than a predictive marker. Interestingly, multiple cross-sectional and prospective cohort studies have revealed the metabolism of impaired branched-chain amino acid (BCAA), aromatic amino acid (AAA), free fatty acid (FFA), acylcarnitines and glycerophospholipid are associated with insulin resistance, and many metabolites were considered as biomarkers for the prediction of pre-DM and DM^[@CR10]--[@CR13]^. However, little is known about the potential metabolic biomarkers of different glycemic prognoses among subjects with pre-DM. More importantly, building a metabolic model for predicting the transition from pre-DM to NGR or DM would be helpful for the early prevention and treatment among individuals with pre-DM. Metabolomics provides a snapshot of the metabolic dynamics that reflects the response of living systems to pathophysiological stimuli and/or genetic modifications and surrounding environment. Furthermore, in many ways, tanscriptomic, genomic, and proteomic changes are upstream of the final physiology of cells, whereas the metabolic profile is likely closer in response to the disease process^[@CR14]^. Ultra-performance liquid chromatography-quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS) of biofluids can easily detect hundreds of individual species in a single clinical sample, reflecting the biochemical fingerprint of the organism^[@CR15]^. Characterized by sensitivity and high mass accuracy, the technique has been employed in identifying novel biomarkers for cancers^[@CR16],[@CR17]^, metabolic disorders^[@CR18]--[@CR20]^, drug toxicity and function^[@CR21],[@CR22]^, and so on. With the current study, we characterized the metabolic profiles of fasting plasma samples among the 108 pre-DM at baseline with different outcomes ten years later utilizing untargeted UPLC-QTOF-MS analysis. 23 and 22 metabolites were identified as biomarkers for transition to NGR and DM from pre-DM, respectively. And the underlying biochemical pathways leading to different prognoses were investigated. Results {#Sec2} ======= Demographic and Clinical Characteristics {#Sec3} ---------------------------------------- 108 participants with pre-DM from a longitudinal cohort study were followed up for ten years and were divided into 3 groups according to different glycemic outcomes. 20 participants progressed to DM, 20 regressed to NGR, and 68 remained at pre-DM, respectively. At baseline, there were no significant differences in ages, gender, body mass index, blood glucose, lipid profile, blood pressure as well as general health conditions among these 3 groups (Supplemental Table [S1](#MOESM1){ref-type="media"}). At the end-point of the study, no significant differences in the biochemical characteristics were found among the three groups, except for fasting glucose, 2-h glucose and HbA1c (Table [1](#Tab1){ref-type="table"}).Table 1Characteristics of the study participants at end-point.NGR (n = 20)Pre-DM (n = 68)DM (n = 20)Age, years57.4 ± 8.860.4 ± 8.957.9 ± 10.1Gender, N, male/female7/1325/437/13Body mass index, kg/m^2^24.0 ± 3.124.8 ± 2.926.1 ± 3.9Waist circumference, cm83.3 ± 7.587.1 ± 8.488.3 ± 11.9Waist-hip ratio0.88 ± 0.040.90 ± 0.060.92 ± 0.08Hypertension, %45.057.470.0Family history of DM, %35.033.845.0Current smoker, %5.022.115.0Current drinker, %15.013.20.0**Physical activity**Inactive, %5.011.810.0Medium, %70.073.575.0Active, %25.014.715.0SBP, mmHg132.5 ± 19.4138.9 ± 14.8138.7 ± 17.5DBP, mmHg77.4 ± 8.880.5 ± 9.580.4 ± 10.4Fasting glucose, mmol/L5.2 ± 0.2^\*^5.6 ± 0.58.0 ± 2.9^\*\*\#^2-h glucose, mmol/L6.0 ± 0.8^\*\*^7.6 ± 1.813.1 ± 4.7^\*\*\#^HbA1c, %5.3 ± 0.5^\*\*^5.8 ± 0.37.2 ± 1.4^\*\*\#^Fasting insulin, pmol/mL8.4 ± 3.79.2 ± 5.011.3 ± 3.8^\*a^2-h insulin, pmol/mL38.9 ± 23.571.2 ± 65.458.7 ± 36.8HDL, mmol/L1.4 ± 0.31.5 ± 0.41.5 ± 0.4LDL, mmol/L3.1 ± 0.73.0 ± 0.93.5 ± 1.1TC, mmol/L5.2 ± 1.25.5 ± 1.86.2 ± 1.3^b^TG, mmol/L1.4 ± 0.72.0 ± 2.62.8 ± 2.3^\*\#^BUA, umol/L311.9 ± 69.7338.7 ± 88.5321.3 ± 79.4CR, umol/L81.1 ± 18.672.8 ± 29.062.6 ± 25.7ALT, U/L20.8 ± 9.023.0 ± 16.833.1 ± 32.7AST, U/L24.5 ± 6.527.3 ± 17.828.1 ± 16.8γGT, U/L24.0 ± 21.431.8 ± 30.531.8 ± 17.2Values are mean ± SD or %. \*p \< 0.01, \*\*p \< 0.001 compared to pre-DM and ^\#^p \< 0.001 compared to NGR. ^a^Exact significance (2-tailed), p = 0.015 compared to NGR; ^b^p = 0.017 compared to pre-DM. SBP: systolic blood pressure; DBP: diastolic blood pressure; HbA1c: hemoglobin A1c; HDL: high density lipoprotein; LDL: low density lipoprotein TC: total cholesterol; TG: triglyceride; BUA: blood uric acid; CR: creatinine; ALT: alanine transaminase; AST: aspartate transaminase; γGT: γ glutamyl transferase. Quality control {#Sec4} --------------- The robustness and stability of the method was assessed by repeat analysis of a representative pooled quality control (QC) sample during sample runs. The overlapped total ion current chromatograms of the QC sample demonstrated the repeatability of our UPLC-QTOF-MS system (Supplemental Fig. [S1A](#MOESM1){ref-type="media"}). The principal component analysis (PCA) performed on QC and other groups revealed that QC samples were clustered in the PCA scores plot (Supplemental Fig. [S2A](#MOESM1){ref-type="media"}). The percentage coefficient of variation (CV%) of peak intensity was estimated as 4.1--18.6%. These results collectively indicated good repeatability, reliability, and stability of this method for metabolite analysis. Plasma metabolite profile and makers for NGR and DM {#Sec5} --------------------------------------------------- Representative base peak intensity (BPI) chromatograms of plasma samples indicated that the sample metabolites attained suitable separation. Typical single UPLC-QTOF/MS base peak intensity chromatograms of a healthy control, a patient with pre-DM and a patient with DM are presented in Supplemental Fig. [S1B](#MOESM1){ref-type="media"}. Multivariate statistical analysis was performed to determine whether the plasma metabolic profiles were different among participants progressed to DM, regressed to NGR and remained at pre-DM. The PCA score plots of the three groups, NGR vs DM, NGR vs pre-DM and pre-DM vs DM are presented in Supplemental Fig. [S2B--D](#MOESM1){ref-type="media"}. The R2X values of PCA analysis were \> 0.5 (0.651, 0.604, 0.631, respectively). Partial least squares discriminant analysis (PLS-DA) results of pair-wise groups indicated separations in the three groups (Fig. [1A](#Fig1){ref-type="fig"}) and sub-comparison groups (Fig. [1B--D](#Fig1){ref-type="fig"}) with valid model fits (R2Y(cum) \> 0.7 and Q2(cum) \> 0.4)^[@CR23]^. Variable importance in projection (VIP) values were obtained from the models and pair-wise statistical test for difference was performed. Variables with VIP value \> 1.0 and statistical p value \< 0.05 were selected and verified by loading plots. Finally, a metabolite was annotated according to MS information, structure information, accurate mass, retention time, fragmentation pattern and standards.Figure 1PLS-DA score plots of different groups based on plasma spectral data of UPLC-QTOF-MS positive ion mode. One point stands for one subject. (**A**) PLS-DA score plot of the NGR vs pre-DM vs DM groups. (**B**) PLS-DA score plot of the NGR vs DM groups. (**C**) PLS-DA score plot of the NGR vs Pre-DM groups. (**D**) PLS-DA score plot of the pre-DM vs DM groups. Based on the above steps, a panel of 23 (Table [2](#Tab2){ref-type="table"}) metabolites distinctively discriminated NGR and DM groups as well as NGR and pre-DM groups with a same trend, which were considered to be the potential metabolic biomarkers for the prognosis to NGR. Similarly, a panel of 22 (Table [3](#Tab3){ref-type="table"}) metabolites discriminated NGR and DM groups as well as DM and pre-DM groups with a same trend were considered to be metabolic markers for the prognosis to DM. Compared to the pre-DM group, 18 biomarkers were up-regulated and 5 biomarkers were down-regulated in the NGR group, while 5 biomarkers increased and 17 biomarkers declined in the DM group. Of the substances that were screened in this study, several substances appear to be unknown. We will attempt to identify these unknown substances in future studies.Table 2Discriminative metabolites between NGR and pre-DM (ESI+ mode).MetabolitesMeasured mass, DaCalculated mass, DaMass accuracy, ppmQuasi-molecular ionVIP^\*^FC^‡^p^†^Identification2,3-Epoxymenaquinone347.1612347.1618−1.7283M + Na1.361.500.0005HMDBPc(14:1/16:1)740.4650740.46332.2959M + K1.521.320.0152STD5-methoxytryptamine213.1025213.10097.5081M + Na1.463.560.0041STDN(6)-(octanoyl)lysine311.1732311.1737−1.6068M + K1.311.310.0204STD3-Phenylbutyric acid356.1910356.19061.12302 M + 3H2O + 2 H1.311.290.0337HMDBLysyl-Tyrosine327.2046327.20285.5011M + NH41.291.220.0401HMDBN6-Acetyl-L-lysine399.2205399.2214−2.25442 M + Na1.301.230.0468STDPantetheine301.1185301.1198−4.3172M + Na1.351.460.0013HMDBS-(hydroxymethyl)glutathione355.1262355.1287−7.0398M + NH41.281.520.0042STD3-Ethylphenol283.1110283.10955.29832 M + K1.111.340.0085HMDBLysoPE(20:5/0:0)532.2992532.3034−7.8490M + CH3OH + H1.261.180.0437HMDBDelta 8,14 -Sterol487.2746487.27391.4366M +  + 2 K + H1.261.230.0488HMDBCaprylic acid306.2666306.26398.8158M + NH41.081.150.0319HMDB1-Stearoylglycerophosphoglycerol576.3257576.3272−2.5767M + ACN + Na1.331.630.0028HMDBEndomorphin-1649.2475649.2535−9.2415M + K1.571.730.0026HMDB20-Hydroxy-leukotriene E4473.2654473.2680−5.4937M + NH41.281.280.0004HMDBLysopc(18:3)518.3237518.3241−0.8007M + H1.860.290.0020STDLysopc(20:5)542.3238542.3241−0.5808M + H1.730.110.0000STDcis-13,16-Docosadienoic acid354.3349354.3372−6.4910M + NH41.541.600.0029HMDBLysopc(20:4)544.3356544.3398−7.6515M + H1.600.700.0397STDL-palmitoylcarnitine400.3410400.3421−2.8351M + H2.112.250.0075STDPc(18:3/20:3)806.5683806.5694−1.4022M + H1.090.380.0336HMDBLysoPC(20:3)546.3487546.3554−12.2925M + H1.620.140.0013STD^\*^Variable importance in the projection (VIP) was obtained from PLS-DA with a threshold of 1.0. ^†^p values were calculated from tests of statistical difference. Difference was considered statistically significant when p \< 0.05. ^‡^Fold change (FC) was calculated from the arithmetic mean values of NGR and pre-DM groups. Fold change with a positive value indicates a relatively higher concentration present in NGR patients while negative indicates lower. Table 3Discriminative metabolites between DM and pre-DM (ESI+ mode).MetabolitesMeasured mass, DaCalculated mass, DaMass accuracy, ppmQuasi-molecular ionVIP^\*^FC^‡^p^†^Identifi-cationPc(16:0/14:0)782.4430782.4499−8.8121M + 2 K + H1.160.820.0351HMDB2-Pyrroloylglycine191.0420191.0427−3.7165M + Na1.190.840.0070STDDityrosine383.1181383.1219−9.9186M + Na2.431.230.0188STDKynuramine203.0561203.0581−9.9529M + K3.111.250.0007HMDBL-lysine188.1400188.13943.4496M + ACN + H1.270.740.0025STDL-threonine164.0296164.02941.2010M + 2Na-H2.330.590.0002STD5-hydroxy-2-oxo-4-ureido-2,5-dihydro-1h-imidazole-5-carboxylate220.0656220.0676−9.2791M + NH41.850.690.0002HMDB1,3,7-trimethyluric acid274.0925274.09115.2719M + ACN + Na1.981.520.0071STDBetaine118.0858118.0863−3.8531M + H1.280.710.0006HMDBIso-valeraldehyde104.1077104.10751.9211M + NH42.160.740.0000HMDBL-carnitine162.1132162.11301.2337M + H1.190.730.0005STD2-ketobutyric acid103.0397103.03951.9410M + H1.390.720.0003HMDB3,5-dihydroxybenzoic acid187.0605187.06012.1490M + CH3OH + H1.770.610.0010HMDBUric acid169.0358169.03561.0885M + H1.970.200.0001STDLysope(16:0/0:0)454.2908454.2928−4.4355M + H2.070.240.0006HMDBPantetheine301.1181301.1198−5.6456M + Na1.150.740.0336HMDBPalmitic amide256.2629256.2635−2.3062M + H1.420.410.0279HMDB3-dehydroxycarnitine184.0736184.0740−2.1730M + K1.631.300.0061STDLysopc(18:1)544.3363544.3403−7.3484M + H1.060.710.0184STDLinoleic acid341.3053341.30500.8174M + H1.141.190.0212HMDBLysopc(18:0)524.3706524.3711−0.8887M + H1.340.420.0398STDPc(18:0/18:2)786.5968786.6007−4.9975M + H1.290.090.0054HMDB^\*^Variable importance in the projection (VIP) was obtained from PLS-DA with a threshold of 1.0. ^†^p values were calculated from tests of statistical difference. Difference was considered statistically significant when p \< 0.05. ^‡^Fold change (FC) was calculated from the arithmetic mean values of DM and pre-DM groups. Fold change with a positive value indicates a relatively higher concentration present in DM patients while negative indicates lower. Metabolic pathways {#Sec6} ------------------ Pathway analysis carried out by IPA software revealed that three metabolic pathways including glycerophospholipid metabolism, Lipoate Biosynthesis and Incorporation II, and Melatonin Degradation II were found to contribute in the process from pre-DM to NGR (Fig. [2A](#Fig2){ref-type="fig"}). While L-carnitine Biosynthesis, Superpathway of Methionine Degradation, Mitochondrial L-carnitine Shuttle Pathway, and Choline Degradation I were associated with the development of DM (Fig. [2B](#Fig2){ref-type="fig"}).Figure 2Biological network and canonical pathways related to the identified metabolites in NGR (**A**) and DM (**B**). Molecules are represented as nodes, and the biological relationship between two nodes is represented as a line. Red symbols represent up-regulated metabolites; green symbols represent down-regulated metabolites. The solid lines and dotted lines show direct and indirect functional relationships, respectively. Verification and Optimization of Potential Biomakers {#Sec7} ---------------------------------------------------- Receiver operating characteristic (ROC) curves for each potential biomarker were established to test the probability of 'single biomarkers'. Area under curve (AUCs) and their CV% values in QC samples are showed in Supplemental Tables [S2](#MOESM1){ref-type="media"} and [S3](#MOESM1){ref-type="media"}. To assess how multiple metabolites collectively classify the NGR/DM and the Pre-DM groups, we built logistic regression model using stepwise selection on the 23/22 metabolites for the samples, respectively.Figure 3(**A**) ROC analysis for discrimination of pre-DM and NGR groups by logistic regression model combining 20-Hydroxy-leukotriene E4, Lysopc(20:4), 5-methoxytryptamine, Endomorphin-1, Lysopc(20:3). (**B**) ROC analysis for discrimination of pre-DM and DM groups by logistic regression model combining Iso-valeraldehyde, linoleic acid, Lysopc(18:1), 2-Pyrroloylglycine, Dityrosine. From pre-DM to NGR, five metabolites entered the multiple regression model: 20-Hydroxy-leukotriene E4 (standardized \[std\] β = 2.135, p = 0.004), Lysopc(20:4) (std β = −1.423, p = 0.027), 5-methoxytryptamine (std β = 0.606, p = 0.047), Endomorphin-1 (std β = 4.685, p = 0.018),and Lysopc(20:3) (std β = −19.176, p = 0.002) with an overall correct percentage of 86.0%. We calculated the sensitivity and specificity based on estimates of the final model built on the samples (logitP1 = −1.847 + 2.135(20-Hydroxy-leukotriene E4) − 1.423(Lysopc(20:4) + 0.606 (5-methoxytryptamine) + 4.685 (Endomorphin-1) − 19.176 (Lysopc(20:3)), and the model fit very well (AUC = 0.910, 95% CI \[0.838, 0.983\], sensitivity = 84.2%, specificity = 91.0%) with cut-off value −0.08 and p value \< 0.001. Meanwhile, five metabolites entered the multiple regression model from pre-DM to DM: Iso-valeraldehyde (standardized \[std\] β = −1.544, p = 0.049), linoleic acid (std β = 2.194, p = 0.019), Lysopc(18:1) (std β = −4.769, p = 0.013), 2-Pyrroloylglycine (std β = −2.922, p = 0.044) and Dityrosine (std β = 0.285, p = 0.007) with an overall correct percentage of 95.4%. The final model (logitP2 = −1.119--5.144(Iso-valeraldehyde) + 2.194(linoleic acid) − 4.769(Lysopc(18:1) − 2.922(2-Pyrroloylglycine) + 0.285 (Dityrosine)) showed satisfactory fitness (AUC = 0.976, 95% CI \[0.943, 1.000\], sensitivity = 90.0%, specificity = 98.5%) with cut-off value −0.33 and p value \< 0.001. The ROC curves of the combined biomarkers are shown in Fig. [3](#Fig3){ref-type="fig"}. Both of them were statistically different from single metabolites as they showed a higher lower bound of 95% CI of AUCs than most of the upper bounds of 95% CI of single metabolites, and were further validated by comparison of AUCs with MedCalc Statistical Software with a statistical p value \< 0.05^[@CR24]^. Relative concentrations of these metabolites in the logistic regression equations are presented in Fig. [4](#Fig4){ref-type="fig"}. These data strongly support the robustness of UPLC-QTOF-MS to identify metabolic differences in the plasma samples of pre-DM patients with different prognoses (NGR or DM).Figure 4Box plots of mean intensity of ten representative metabolites in plasma samples of NGR, pre-DM and DM patients. Discussion {#Sec8} ========== Early diagnosis in hyperglycemia, prior to developing into diabetes, can improve the living quality of those suffering from pre-DM. The rapid rise in pre-DM worldwide raises urgent needs to develop effective prognostic biomarkers and prognosis evaluation methods based on the easily accessible materials, such as blood and urine. Using UPLC-QTOF-MS based metabolic profiling combined with pattern recognition techniques on plasma samples, we identified molecular markers that discriminate the prognoses to NGR and to DM in 108 pre-diabetic patients in a longitudinal study with more than ten years of follow-up. In this study, a total of 23 metabolites involved in different biochemical metabolic pathways with high statistical significance were associated with the outcome of NGR, while 22 metabolites with the outcome of DM. Of these perturbed metabolic pathways, particular importance should be given to glycerophospholipid metabolism. Earlier studies have demonstrated that diabetes is intimately associated with metabolic disorders of lipids, especially phospholipids^[@CR25],[@CR26]^. Phosphocholine metabolites, including PC and LysoPC, are key components of the biomembranes of cells, as well as participating in various biological pathways^[@CR27],[@CR28]^, especially in cellular signaling and metabolism. Correspondingly, significant changes in phosphatidylcholine metabolism were observed in our study. PC (16:0/14:0) and PC (18:0/18:2) showed a reduction in level from pre-DM to DM, while PC (14:1/16:1) increased and PC (18:3/20:3) decreased from pre-DM to NGR. This is in line with a previous prospective cohort study based on 866 participants with seven years' follow-up that a series of PC such as C32:1, C36:1, C38:3, and C40:5 were independently associated with increased risk of T2D and C34:3, C40:6, C42:5, C44:4, and C44:5 with decreased risk^[@CR29]^. However, it was also demonstrated that the type of linkage between phospholipid core and fatty acid residue may be the key factor contributing to the antithetical association between two phosphatidylcholine subclasses and T2DM risk^[@CR29]^. Therefore, the specific structural formula of the PCs detected needs further validation. What's more, some studies have reported that changed concentrations of lysoPCs are associated with the risk of T2DM, especially lysoPC(18:2), which was significantly altered in patients with impaired glucose tolerance (IGT) and was identified as an IGT-specific biomarker^[@CR30]^. LysoPCs can mediate many cell-signaling pathways in monocytes/macrophages^[@CR29],[@CR31]^ and specific receptors^[@CR32]^, and therefore participate in the inflammatory response. In our results, several lysoPC species (LPC C20:4, LPC C18:3, LPC C20:5, LPC C20:3) were down-regulated from pre-DM to NGR group, while the plasma level of LPC C18:1 were higher and LPC C18:0 were lower in DM than pre-DM patients, possibly owing to altered activity of phospholipase A2 which catalyzes PC hydrolysis to lysoPC^[@CR33]^. In mouse models, Yea, K *et al*. have reported that the blood glucose lowering effect of LPC were found to be sensitive to variations in LPC acyl chain length^[@CR34]^, which may elucidate the divisive findings mentioned above. In recent years, 5-methoxytryptamine (5-MT) has been revealed to play a pivotal role in the alternative melatonin synthetic pathway, in which serotonin is first O-methylated to 5-MT and, thereafter, 5-MT is N-acetylated to melatonin (N-acetyl-5-methoxytryptamine)^[@CR35]^. Melatonin was reported as a free radical scavenger and an antioxidant for protection from the oxidative stress^[@CR36]--[@CR38]^, especially reducing oxidative damage to lipids, which is important for the maintenance of mitochondrial homeostasis. Letra-Vilela, R. *et al*. observed that removal of the N-acetyl group enhances the antioxidant and neuroprotective properties of the maletonin^[@CR39]^. The elevated level of plasma 5-MT concentrations contributed to scavenge radicals and radical products and functioned as antioxidant against reaction oxygen species (ROS). Meanwhile, an increased level of plasma S-(hydroxymethyl)glutathione concentration was observed in patients regressed to NGR. S-(hydroxymethyl)glutathione interconverts with glutathione (GSH), later of which is a key antioxidant and marker for conditions with oxidative stress^[@CR40]^. Therefore, we can conclude that a balance between oxidation and antioxidation was achieved in patients regressed to NGR. Caprylic acid, also named octanoic acid, is a medium-chain fatty acid which down-regulates a number of key adipogenic genes including peroxisome proliferator activated receptor (PPAR), CCAAT/enhancer binding protein alpha. Rats fed on diet rich in medium-chain fatty acids had smaller fat pads, reduced adipose tissue lipoprotein lipase activity and improved insulin sensitivity and glucose tolerance^[@CR41]^. Thus the elevated level of caprylic acid may be another contributor to glucose tolerance recovery. In this study, a major alteration observed in patients progressed to DM was amino acid metabolism with a decreased level of L-lysine, L-threonine, Betaine, Iso-valeraldehyde, 2-ketobutyric acid, and 2-Pyrroloylglycine. 2-Ketobutyric acid is a substance that is involved in the metabolism of many amino acids, such as glycine, methionine, valine, leucine, serine, threonine and isoleucine. It is also one of the degradation products of threonine, which can be converted to propionyl-CoA (and subsequently methylmalonyl CoA, which can be converted to succinyl CoA, a citric acid cycle intermediate), and thus enter the citric acid cycle. Our research is consistent with the report that 2-ketobutyric acid is associated with both prevalent diabetes (OR: 1.43, 95% CI (1.06,1.92)) and incident diabetes (OR: 1.81, 95% CI (1.35, 2.42)^[@CR42]^ while glycine is found inversely associated with diabetes risk^[@CR43],[@CR44]^. This suggest that increased activity in the L-threonine-2-ketobutyric acid-citric acid cycle and decreased glycine, which plays an important role in metabolic regulation and anti-oxidative reactions^[@CR45]^ may account for diabetes risk. Betaine, also glycine betaine, has been reported to be disturbed in diabetes and it is regarded as a marker of diabetes in plasma and urine samples^[@CR46]--[@CR48]^. It is inversely associated with several components of metabolic syndrome including obesity, hypertension, and hyperlipemia^[@CR49]^. Mouse model studies suggest increased betaine metabolism in diabetes, which could be expected to lower plasma betaine^[@CR50]^. In agreement with these findings, a reduction of betaine level is an important contribution to DM development. A decreased level of L-carnitine and an increased level of dityrosine and 3-dehydroxycarnitine were detected in patients progressed to DM. These results are consistent with an interesting study investigating the relationship between the consumption of red meat and the development of associated diseases^[@CR51]^. L-carnitine is known to be a long-chain fatty acid transporter in the "Mitocondrial L-carnitine Shuttle Pathway"^[@CR52]^, and 3-dehydroxycarnitine has been identified as an intermediate metabolite in the intestinal bacterial catabolism of L-carnitine^[@CR53]^, which may give a hint the involvement of gut microbiota in DM development. Dityrosine has been proposed as a biomarker of oxidative stress under a variety of conditions and biological systems including aging, exposure to oxygen free radicals, nitrogen dioxide, and lipid hydroperoxides^[@CR54]^. Increases in dityrosine levels have been associated with pathologies such as atherosclerosis, Alzheimer's Disease, and so on^[@CR55],[@CR56]^. Together, these findings reflect impaired mitochondrial β-oxidation and perturbed fatty acid metabolism in the development of insulin resistance. To make it better to apply the identified biomarkers for prediction of pre-DM prognosis in the long run in clinical setting, the key of this study was to select and combine several specific biomarkers for establishing a noninvasive and accurate predict method for prognosis of pre-DM. The candidate biomarker selection rationale was as follows: first, the biomarkers must be confirmed by standards; second, the biomarkers with high VIPs in the pattern recognition analysis and significant discrepancy between groups; and last, the biomarkers in the logistic regression equation with higher AUCs and predictive sensibility and specificity. As a result, five biomarkers (20-Hydroxy-leukotriene E4, Lysopc(20:4), 5-methoxytryptamine, Endomorphin-1, Lysopc(20:3)) and five biomarkers (Iso-valeraldehyde, linoleic acid, Lysopc(18:1), 2-Pyrroloylglycine, Dityrosine) were included in the predictive equation of NGR and DM, respectively. In summary, our study comprehensively captured alterations in the human metabolome associated with different glucose tolerance outcomes of pre-DM in a longitudinal cohort study. A decreased glycerophosphoslipid metabolism and balanced oxidation and antioxidation contributed to the transition to NGR, while impaired amino acid metabolism and perturbed mitochondrial β-oxidation were associated with the development of DM. Targeting the pathways that involve in these newly prognosis biomarkers would be beneficial for the regression to NGR and the early prevention of DM among participants with pre-DM. Subjects and Methods {#Sec9} ==================== Ethics Statement {#Sec10} ---------------- This study was approved by Ruijin Hospital Ethics Committee (approval no. 2014--114). Written informed consent signed by each of participants was provided before blood samples were taken. All methods were carried out in accordance with the relevant guidelines and regulations. Study design and Subjects {#Sec11} ------------------------- This study was from a population-based prospective cohort study of 2132 men and women aged 18--76 years, from November 2002 to January 2003, among whom 778 participants were pre-DM at baseline. The follow-up visit was conducted from July 2013 to October 2014 and 526 participants who were pre-DM at baseline were followed, among whom 334 individuals both answered questionnaires and had plasma glucose measurement during an oral glucose tolerance test (OGTT). Serum lipid profile and liver function were also assayed. After excluded the individuals on anti-diabetes medication or with serious liver, renal dysfunction and cancer, the remaining 108 individuals were included in our final analysis. This study design has been described previously^[@CR57]--[@CR59]^. According to different glycemic outcomes at follow up, the 108 participants were divided into 3 groups. 20 participants progressed to diabetes DM, 20 regressed to NGR, and 68 remained at pre-DM. Details of the study population are presented in Supplemental Fig. [S3](#MOESM1){ref-type="media"}. Venous blood samples were collected at baseline and follow-up. The glucose level was measured by means of glucose oxidase method. Pre-DM and diabetes were diagnosed according to American Diabetes Association (ADA) 2010 Guidelines^[@CR60]^. Pre-DM refers to subjects with impaired fasting glucose (fasting glucose ranging from 5.6 to \< 7.0 mmol/L, as well as 2-hour glucose \< 7.8 mmol/L) and subjects with impaired glucose tolerance (2-hour glucose ranging from 7.8 to \< 11.1 mmol/L). Both impaired fasting glucose and impaired glucose tolerance were confined to non-diabetic fasting and 2-hour concentrations. Fasting plasma samples were used to analyze biochemical indexes and metabolomics. Sample collection and preparation {#Sec12} --------------------------------- Fasting blood samples were drawn under sterile conditions from an antecubital vein of all the study participants between 6:30 and 9:30 after a 12-hour overnight fast, and were collected directly into heparinized tubes. The tubes were centrifuged at 12000 g for 10 min and the supernatant (plasma sample) was aspirated and stored at −80 °C until analysis. The plasma sample (100 μl) was thawed at 4 °C. 100 μl of plasma was spiked with 300 μl mixed solution (methanol: acetonitrile = 3:2) and vigorously vortexed for 30 seconds. The sample solution was centrifuged at 12000 × *g* for 10 min at 4 °C, and the supernatant was analyzed using UPLC-QTOF-MS. Study samples were analyzed in random order using a random-number generator in Excel 2015 (Microsoft, Redmond, Washington). QC samples were prepared by mixing equal volumes (10 μl) of different individual plasma samples and one QC sample was run after every ten study sample injections throughout the analytical workflow. UPLC-QTOF-MS conditions {#Sec13} ----------------------- In this study, a Waters ACQUITY^TM^ ultra performance liquid chromatography system (Waters Corp., Milford, USA) coupled with aSynaptG2 quadrupole time-of-flight (Q/TOF) tandem mass spectrometer (Waters, Milford, MA) was used to perform the analysis of plasma samples. The set-up parameters for the UPLC-QTOF-MS analysis were as follows: A T3 C18 chromatographic column (Waters, 2.1 mm × 100 mm, 1.7 μm) was used to separate metabolites contained in plasma with column temperature set at 45 °C. The eluted solution was 0.1% formic acid combined with 5 mM ammonium acetate in water (A) and acetonitrile (B) with a flow rate of 300 μl/min. The gradient elution program for analysis of plasma samples was as follows: 0--1 min, A: 98%; 1--3 min, A: 98--50%; 3--8 min, A: 50--45%; 8--12 min, A: 45%; 12--17 min, A: 45%−10%; 17--20 min, A: 10--98%. The MS parameters were set up as follows: the electrospray ionization source (ESI) interface operated with a positive mode, capillary voltage of 3000 V, sample cone voltage of 40 V, extraction cone voltage of 4.0 V, desolvation gas flow of 650 L/h at 450 °C, source temperature of 120 °C, and cone gas flow of 50 L/h. Centroid data were collected under a scan time of 0.25 s and an inter scan delay of 0.02 s condition in continuum mode which ranged from m/z 100 to m/z 1200 Da. To avoid possible contamination and keep the signal stable, the Q-TOF mass spectrometer system was tuned for optimum accuracy and reproducibility using leucine-enkephalin (m/z 556.2771) as the lock mass in all analyses at a concentration of 0.5 μg/mL. The lock spray frequency was set at 5 s and the lock mass data were averaged over 10 scans. MS^E^ was applied for the MS2 analysis with the low collision energy of 5 eV and the high collision energy of 30 eV. Data extraction and multivariate statistics {#Sec14} ------------------------------------------- The raw data produced by UPLC-QTOF-MS were initially processed using MarkerLynx Applications Manager version 4.1 (Waters Corp., Manchester, UK). The data were peak-detected and noise-reduced so that only true peaks are further processed by the software. The data were presented with the ion intensities corresponding the retention time and m/z for each peak. The main parameters were set as follows: retention time window 0.5--16.5 min, mass range 100--1200 Da, XIC window 0.02 min, automatically calculate peak width and peak-peak baseline noise, use the raw data during the deconvolution procedure, marker intensity threshold (count) 1000, mass window 0.02 Da, retention time windows 0.2 min, noise elimination level 6.0, and retain the isotopic peaks. The internal standard was used for data quality control and data normalization (reproducibility). The ion peaks generated by the internal standard were removed and the metabolites were filtered by the QC samples. The 80% rule was applied to treat the missing values^[@CR61]^ and a data matrix that consisted of the ion intensities corresponding the retention time and m/z for each peak was generated and then exported to Simca-P software (v13.0, Umetrics, Umea, Sweden) followed by a series of pattern recognition (PR) methods. Multivariate statistical analyses, including principal component analysis (PCA) and a partial least squares discriminant analysis (PLS-DA) were carried out using SIMCA-P 13.0 software. The score plots from PLS-DA showed the differentiation of metabolic profiles of different groups. In addition, loading plots indicated the variables contributing to the classification. The quality of the model was described by the cross-validation parameter Q2 (cum), and R2Y, which represents the total explained variation for the X matrix. After the analysis of the three groups, pair-wise analysis (NGR vs pre-DM, DM vs pre-DM and NGR vs DM) was conducted to searching for the discrepant metabolites between groups. Identification of potential biomarkers and metabolic pathway analysis {#Sec15} --------------------------------------------------------------------- The metabolites responsible for the separation of metabolic profiles of the pair-wise groups were obtained based on a variable importance in projection (VIP) threshold (VIP \> 1 represented higher influence on the classification)^[@CR17]^ from PLS-DA models accompanied with loading plots and a statistical test for difference (*p* \< 0.05 was considered significant). A two-tailed t test or a nonparametric Mann-Whitney test was used for significance evaluation following data normality test with Shapiro-Wilk tests, which was performed with the R statistical software 3.3.2 for Windows. A metabolite was detected and identified based on accurate mass, retention time, MS information and metabolite structure information from related databases: METLIN (<https://metlin.scripps.edu/index.php>) and HMDB (<http://www.hmdb.ca/>). Some of the metabolites were confirmed by comparison of retention time and fragmentation pattern with authentic standards. The pathway analysis and network of potential biomarkers contributing to the classification between groups was carried out by IPA software (IPA, Ingenuity^s^ Systems, <http://www.ingenuity.com>). The schematic flow chart of the metabolic profiling and biomarker identification and optimization strategy used in the study is shown in Fig. [5](#Fig5){ref-type="fig"}.Figure 5Schematic flow chart of the metabolomics analysis in the study. NGR: normal glucose regulation; Pre-DM: pre-diabetes; DM: diabetes; UPLC-QTOF-MS: Ultra-performance liquid chromatography-quadrupole time-of-flight mass spectrometry; VIP: variable importance in the projection; ROC: receiver operating characteristic curve; AUC: area under the curve. Statistical analysis {#Sec16} -------------------- Demographic and biochemical characteristics were presented as the mean (±SD) for continuous variables and % for categorical variables. Data normality and homogeneity of continuous variables were confirmed with Shapiro-Wilk tests and Levene tests, respectively. If data were normally distributed and variances were equal, data were analyzed by means of one-way analysis of variance (ANOVA) with LSD test; Otherwise, Kruskal-Wallis test with a nonparametric two-tailed Mann-Whitney test was used (alpha was adjusted to 0.05/3 here). And categorical variables were analyzed by χ2 test or Fisher's exact test. Difference was considered statistically significant when p \< 0.05. Multiple logistic regression analysis of the potential metabolites was performed and receiver operating characteristics (ROC) analysis was used to evaluate predictive ability of potential metabolic biomarkers. Area under the curve (AUC), best cut-off point, sensitivity and specificity were determined using the maximum value of the Youden index. The analyses were performed using SPSS software version 23.0 (IBM Corp., USA). Electronic supplementary material ================================= {#Sec17} Supplementary information Hailuan Zeng and Renchao Tong contributed equally to this work. A correction to this article is available online at <https://doi.org/10.1038/s41598-017-14856-1>. **Electronic supplementary material** **Supplementary information** accompanies this paper at doi:10.1038/s41598-017-06309-6 **Change History:** A correction to this article has been published and is linked from the HTML version of this paper. The error has been fixed in the paper. **Publisher's note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. We are indebted to staff and participants of the study for their important contributions. We gratefully acknowledge grants supported by the National Natural Science Foundation of China (No. 81270935 and 81573581), Transform Medicine Innovation Foundation of Shanghai Jiao Tong University School of Medicine (15ZH2001), the fund of the Key Laboratory of Stem Cell Biology of Chinese Academy of Sciences (No. 201601), Shanghai Municipal Health Bureau Foundation (20114301), the Program of Shanghai Subject Chief Scientist (17XD1403500) and Shanghai Science and Technology Commission Project (15401902400). J.T. and L.Y. designed the study. H.Z., R.T., W.T., Q.Y., M.Q., A.X. performed experiments. H.Z. and R.T. analyzed data and wrote the paper. L.D., S.S., H.Z., J.T. and L.Y. reviewed and edited the manuscripts. All authors reviewed and approved the final article. Competing Interests {#FPar1} =================== The authors declare that they have no competing interests.

1. Introduction {#sec1-pharmaceutics-10-00021} =============== The discovery of vascular endothelial growth factor (VEGF) \[[@B1-pharmaceutics-10-00021],[@B2-pharmaceutics-10-00021]\] and the subsequent recognition of its critical role in the pathogenesis of several chorioretinal vascular conditions constitute the most important advances in ophthalmology over the past 30 years. Strong evidence correlates the development of both neovascularization and macular edema in the two most common causes of blindness in industrialized nations---neovascular age-related macular degeneration (nAMD) and diabetic retinopathy (DR)---with the upregulation of VEGF \[[@B3-pharmaceutics-10-00021]\]. Furthermore, disease severity frequently correlates with intraocular VEGF concentrations, thereby making VEGF a logical target for therapeutic intervention. Soon after VEGF was discovered and sequenced, the production of inhibitory molecules began \[[@B4-pharmaceutics-10-00021]\]. Thus far, five VEGF-neutralizing molecules (pegaptanib, Macugen^®^, Bausch & Lomb, Bridgewater, NJ, USA; ranibizumab, Lucentis^®^, Genentech, S. San Francisco, CA, USA/Roche, Basel, Switzerland; aflibercept, Eylea^®^, Regeneron, Tarrytown, NY, USA; conbercept, Chengdu Kanghong Pharmaceutical Group, Chengdu, China; and bevacizumab, Avastin^®^, Genentech, S. San Francisco, CA, USA/Roche, Basel, Switzerland) have been used to treat ophthalmologic conditions, though only the first three have received United States Food and Drug Administration (US FDA) approval for intraocular use. Intravitreal therapy usually begins with monthly injections (in accordance with package labeling) but most physicians will attempt to extend the time between injections as much as possible with either monthly *pro re nata* (PRN) or treat and extend strategies \[[@B5-pharmaceutics-10-00021]\]. Treatment intervals for many patients cannot be extended beyond eight weeks \[[@B6-pharmaceutics-10-00021]\], resulting in a large group of patients who require frequent injections for long periods of time. This large number of intravitreal injections burdens physicians and their staffs, and challenges patients' compliance. Therefore, new, longer acting anti-VEGF medications and drug delivery systems are needed to improve outcomes, optimize compliance, and reduce the total cost of care. This manuscript discusses extended duration anti-VEGF therapies that have been recently introduced, as well as those that are in various stages of development. 2. Vascular Endothelial Growth Factor (VEGF) Physiology and Pharmacokinetics {#sec2-pharmaceutics-10-00021} ============================================================================ VEGF was discovered independently by two research groups in 1989 \[[@B1-pharmaceutics-10-00021],[@B2-pharmaceutics-10-00021]\] and its important role in both physiologic angiogenesis and pathological neovascularization was realized almost immediately. VEGF is actually a group of molecules that segregate into seven closely related families: VEGF-A, VEGF-B, VEGF-C, VEGF-D, VEGF-E, VEGF-F, and placental growth factor (PlGF) \[[@B7-pharmaceutics-10-00021]\]. Each of the families is characterized by common, critical binding sequences, and most families contain multiple isoforms that share similar binding properties and biological actions. VEGF-A synthesis is upregulated in eyes with chorioretinal vascular conditions, including nAMD, diabetic macular edema (DME), and retinal vein occlusion (RVO) \[[@B3-pharmaceutics-10-00021]\], and is believed to play a central role in the development of these conditions. Several in vivo models show that VEGF-A promotes the growth of choroidal neovascular membranes \[[@B8-pharmaceutics-10-00021]\] and produces retinal vascular lesions that resemble DR \[[@B9-pharmaceutics-10-00021]\]. Evidence suggests that VEGF~165~ may be the most biologically active isoform because of its high tissue concentrations and 10-fold potentiation of activity through its interaction with the transmembrane co-receptor neuropilin-1 \[[@B10-pharmaceutics-10-00021]\]. Most VEGF inhibitory molecules block the receptor binding region (amino acids 81--92) of VEGF-A isoforms, whereas pegaptanib interacts with the heparin binding region (amino acids 110--165) of VEGF~165~. Research suggests that VEGF-B, VEGF-C, VEGF-D, and PlGF may also contribute to pathologic ocular angiogenesis in humans but their relative contribution is not known \[[@B11-pharmaceutics-10-00021],[@B12-pharmaceutics-10-00021]\]. Increased VEGF synthesis by vascular endothelial cells, glia, pericytes, Müller cells, retinal pigment epithelium (RPE) cells, and invading leukocytes \[[@B13-pharmaceutics-10-00021],[@B14-pharmaceutics-10-00021]\] results from tissue ischemia and inflammation \[[@B15-pharmaceutics-10-00021],[@B16-pharmaceutics-10-00021]\]. Cells throughout the retina and choroid respond to increased VEGF concentrations but the primary targets are retinal and choroidal vascular endothelial cells \[[@B17-pharmaceutics-10-00021]\]. VEGF-A has a short half-life of 30 min in the eye and serum, and homeostatic concentrations are generally low (approximately 9 ng/mL) \[[@B18-pharmaceutics-10-00021]\]. Some systemic conditions increase serum VEGF concentrations but chorioretinal vascular conditions produce insufficient VEGF to meaningfully change serum levels. 3. Currently Available Therapies {#sec3-pharmaceutics-10-00021} ================================ Several anti-VEGF drugs have been developed exclusively for ocular use or, in the case of bevacizumab, are used off-label for chorioretinal vascular conditions. Peak clinical efficacies of these drugs (except for pegaptanib) are similar and though product labels describe different injection intervals (monthly or every two months) the differences in their duration of action are on the order of only days. Currently available drugs, recently failed therapies, and drugs and systems under development are listed in [Table 1](#pharmaceutics-10-00021-t001){ref-type="table"}. 3.1. Pegaptanib {#sec3dot1-pharmaceutics-10-00021} --------------- Pegaptanib (molecular weight (MW) of 50 kDa), an aptamer to VEGF, was the first ocular drug approved for the intravitreal treatment of neovascular age-related macular degeneration (nAMD). Clinicians hoped that q6week treatment with pegaptanib would improve best corrected visual acuity (BCVA) but in most eyes it only decreased the rate of vision loss by approximately one half \[[@B19-pharmaceutics-10-00021]\]. Its use dropped significantly when more potent anti-VEGF drugs were introduced and pegaptanib is rarely used today. 3.2. Bevacizumab {#sec3dot2-pharmaceutics-10-00021} ---------------- Bevacizumab is a full-length, recombinant, humanized, monoclonal antibody (MW of 149 kDa) that binds all isoforms of VEGF-A. It was developed and approved for the intravenous treatment of several advanced solid tumors (colorectal carcinoma, non-small cell lung carcinoma, renal cell carcinoma, glioblastoma, and breast cancer, though this approval was rescinded in 2011) \[[@B20-pharmaceutics-10-00021]\]. Single injections of bevacizumab were first given to patients with nAMD and macular edema due to a central retinal vein occlusion (CRVO) in 2005 \[[@B39-pharmaceutics-10-00021],[@B40-pharmaceutics-10-00021]\], and within six months off-label use of bevacizumab had become the accepted standard-of-care treatment of chorioretinal vascular conditions. Hundreds of ocular disease studies have established bevacizumab's efficacy and safety, though the best evidence comes from the Comparison of Age-related Macular Degeneration Treatment Trials (CATT) for nAMD and the Diabetic Retinopathy Clinical Research Network Protocol T trial for DME \[[@B6-pharmaceutics-10-00021],[@B21-pharmaceutics-10-00021]\]. The use of bevacizumab varies among countries due to regulatory restrictions, reimbursement policies, and availability of safely compounded drug. Because physicians have accumulated extensive clinical experience with bevacizumab and are able to acquire it inexpensively, bevacizumab remains the most commonly used anti-VEGF drug in the United States. 3.3. Ranibizumab {#sec3dot3-pharmaceutics-10-00021} ---------------- Ranibizumab is a recombinant, humanized, monoclonal antibody fragment (Fab with MW of 48 kDa) that binds all isoforms of VEGF-A \[[@B4-pharmaceutics-10-00021]\]. It has been approved by the United States Food and Drug Administration (USFDA) for the treatment of nAMD (2006), DME, DR, macular edema due to vein occlusions, and choroidal neovascular membranes (CNVM) associated with high myopia \[[@B22-pharmaceutics-10-00021]\]. Following completion of the phase III MARINA and ANCHOR trials, ranibizumab was approved for the monthly treatment of nAMD and subsequently for PRN treatment. The CATT trial reported that PRN treatment is non-inferior to monthly treatment for nAMD \[[@B6-pharmaceutics-10-00021]\] though pooled data from CATT and IVAN suggest that PRN is inferior to monthly injections. Ranibizumab is approved for the monthly treatment of DME, but after one year of intensive treatment in the Diabetic Retinopathy Clinical Research (DRCR).net Protocol I trial, less frequent injections are needed during subsequent years \[[@B41-pharmaceutics-10-00021]\]. Because ranibizumab was the first approved intravitreal anti-VEGF drug (after pegaptanib), it became the standard against which other drugs have been compared in most randomized, controlled trials. These trials have included CATT \[[@B6-pharmaceutics-10-00021]\], IVAN \[[@B42-pharmaceutics-10-00021]\], and the other national AMD trials; the VIEW 1 and 2 trials (nAMD) \[[@B43-pharmaceutics-10-00021]\]; CEDAR and SEQUOIA (nAMD); and DRCR.net Protocol T (DME) \[[@B21-pharmaceutics-10-00021]\]. 3.4. Aflibercept {#sec3dot4-pharmaceutics-10-00021} ---------------- Aflibercept is a recombinant fusion protein (MW of 115 kDa) consisting of the natural (all human) extracellular ligand binding sequences of VEGFR1 (domain 2) and VEGFR2 (domain 3) attached to the Fc portion of an IgG molecule \[[@B24-pharmaceutics-10-00021]\]. Aflibercept is approved for the treatment of nAMD, DME, DR, and macular edema due to RVO \[[@B23-pharmaceutics-10-00021]\]. The three-dimensional configuration of aflibercept enables it to simultaneously bind both sides of the VEGF dimer in a "two-fisted grasp". This results in a higher binding affinity for VEGF~165~ (*k~D~* = 0.45 pM) compared to ranibizumab (*k~D~* = 46--172 pM) and bevacizumab (*k~D~* = 58--1100 pM) \[[@B44-pharmaceutics-10-00021]\]. Rabbit studies suggest that aflibercept has a slightly longer intravitreal half-life that either bevacizumab or ranibizumab but head-to-head human studies have not been performed \[[@B45-pharmaceutics-10-00021]\]. Peak efficacy of aflibercept in patients with nAMD is similar to that of ranibizumab but the duration of action is slightly longer \[[@B46-pharmaceutics-10-00021]\]. Though aflibercept is approved for q8week dosing (compared to monthly for ranibizumab), its duration of action exceeds that of ranibizumab by only five to seven days. So despite the fact that the phase III trials suggested that aflibercept could be equally effective with only half the dosing frequency of ranibizumab, clinical use suggests that the difference is considerably shorter. Ziv-aflibercept (Zaltrap^®^, Regeneron, Tarrytown, NY, USA) is the intravenous formulation of aflibercept that is used to treat advanced colorectal carcinoma. Small series of patients with nAMD, DME, and RVOs have responded well to intravitreal ziv-aflibercept with excellent improvements in macular morphology and visual acuity \[[@B47-pharmaceutics-10-00021]\]. Head-to-head studies with aflibercept have not been performed, but the two molecules will likely perform comparably, though the lower dose of ziv-aflibercept (1.25 mg vs. 2 mg) may provide a slightly shorter duration of action. 3.5. Conbercept {#sec3dot5-pharmaceutics-10-00021} --------------- Conbercept (KH902, Chengdu Kanghong Biotech Co., Sichuan, China) is a recombinant, fusion protein (MW of 143 kDa) that contains the second immunoglobulin (Ig) binding domain from VEGFR1, the third and the fourth binding domains from VEGFR2, and the Fc region of human IgG. Like aflibercept, conbercept acts as a soluble, decoy receptor \[[@B24-pharmaceutics-10-00021],[@B25-pharmaceutics-10-00021]\] that binds all isoforms of VEGF-A, VEGF-B, and placental growth factor. Conbercept has a high affinity for VEGF~165~ (*k~D~* = 0.77 pM) because the fourth Ig domain of VEGFR2 enhances the association rate of VEGF to the receptor \[[@B25-pharmaceutics-10-00021]\]. At concentrations between 100 ng/mL and 100 µg/mL, conbercept is not cytotoxic to cultured human retinal vascular endothelial cells (hRVACs). Conbercept significantly suppresses glucose-induced migration and sprouting of hRVACs by downregulating the expression of phosphoinositide 3-kinase and inhibiting the activation of Src, Akt1, and Erk1/2 \[[@B48-pharmaceutics-10-00021]\]. Four weeks after intravitreal injection, conbercept-treated diabetic rats had better retinal electrophysiological function, less retinal vessel leakage, and lower levels of PlGF, VEGFR2, PI3K, Akt, p-Akt, p-ERK and p-SRC than did Pbs or bevacizumab-treated rats \[[@B49-pharmaceutics-10-00021]\]. The distribution of claudin-5 and occludin in the retinal vessels of diabetic rats treated with conbercept was smoother and more uniform than those of diabetic rats treated by Pbs or bevacizumab. Conbercept is approved in China for the treatment of nAMD and a phase III trial evaluating the efficacy of conbercept for the treatment of DME is currently enrolling patients. Conbercept trials within the United States are now being planned. 4. Therapies under Development {#sec4-pharmaceutics-10-00021} ============================== The currently available anti-VEGF drugs have significantly advanced our treatment of chorioretinal vascular conditions and have benefitted hundreds of thousands of patients, but injections must usually be administered every four to eight weeks and treatment often continues for years. The extended durations of action that were promised by the newer drugs have not concretized, since a wealth of clinical experience shows us that the differences among the drugs are far shorter than are suggested by the packaging labels. Nevertheless, research continues with new drugs and delivery methods that developers hope will extend the clinical duration of action. Several of the most promising drugs and some of the recent failures are discussed below. 4.1. Abicipar Pegol {#sec4dot1-pharmaceutics-10-00021} ------------------- Abicipar pegol is a designed ankyrin repeat protein (DARPin) that binds all isoforms of VEGF-A. Its small size (MW = 34 kDa) would suggest a brief intraocular half-life, but pegylation (binding to a poly(ethylene) glycol moiety) may give it the pharmacokinetic characteristics of a much larger molecule (approximately 250--350 kDa) \[[@B50-pharmaceutics-10-00021]\]. Abicipar has an intravitreal half-life of six days in rabbits and, in a small DME study of four eyes, of 13.4 days in humans \[[@B26-pharmaceutics-10-00021]\]. Its strong binding affinity to VEGF~165~ (*k~D~* = 2 pM) also favors a long duration of action. In dose escalation trials, a maximum tolerated dose of 4.2 mg was found, so investigators elected to develop the two-milligram dose. In the phase II PALM DME trial, abicipar injections every 8 or 12 weeks were non-inferior to monthly ranibizumab \[[@B51-pharmaceutics-10-00021]\]. In the ongoing phase III nAMD CEDAR (NCT02462928) and SEQUOIA (NCT02462486) trials, q8week and q12week abicipar is being compared to monthly ranibizumab. 4.2. Brolucizumab {#sec4dot2-pharmaceutics-10-00021} ----------------- Brolucizumab is a single-chain, high binding affinity (*k~D~* for VEGF~165~ = 1.6 pM), antibody fragment currently being developed by Alcon/Novartis (Ft. Worth, TX; Basel, Switzerland) for the treatment of nAMD \[[@B52-pharmaceutics-10-00021]\]. Its small size (MW = 26 kDa) allows for the injection (six milligrams) of 12--24 times as many molecules as with the other anti-VEGF drugs \[[@B27-pharmaceutics-10-00021]\]. A phase II clinical trial compared brolucizumab to aflibercept in patients with nAMD. The trial's primary objective was to compare the efficacy of six-milligram brolucizumab against two-milligram aflibercept with the primary endpoint being the mean change in BCVA from baseline to 12 weeks. Patients continued receiving q8week treatment until week 40, though brolucizumab patients were eligible for two q12week cycles. At week 12, BCVA gains with brolucizumab (+5.75 letters) were similar to those with aflibercept (+6.89 letters). Approximately 50% of brolucizumab patients were stable during the q12week cycles \[[@B53-pharmaceutics-10-00021]\]. The phase III nAMD clinical trials, HAWK and HARRIER, were initiated in December 2014, with an enrollment goal of 1700 patients in more than 50 countries. These two-year, double-masked, multi-center trials randomize patients with untreated nAMD to one of two dosage intervals of brolucizumab, or aflibercept bimonthly. At the 48-week primary endpoint, mean BCVA gains in both brolucizumab arms were non-inferior to aflibercept. The majority of patients receiving six milligrams brolucizumab (57% and 52%) were maintained exclusively on q12week dosing \[[@B28-pharmaceutics-10-00021]\]. 4.3. Ranibizumab Port Delivery System {#sec4dot3-pharmaceutics-10-00021} ------------------------------------- A refillable ranibizumab port delivery system is being co-developed by Genentech and ForSight Vision 4 to reduce the need for repeated intravitreal anti-VEGF injections. The preloaded implant is surgically implanted beneath the conjunctiva through a 3.2 mm scleral incision over the pars plana. The reservoir tip can be accessed easily in the office and refilled through the conjunctiva as needed. The device continuously releases ranibizumab into the vitreous between refills. A phase I trial for patients with nAMD was performed in Latvia \[[@B29-pharmaceutics-10-00021]\]. At baseline, the reservoir was implanted and eyes were given 500 µg of ranibizumab, 250 µg into the vitreous and 250 µg into the reservoir for sustained release. Additional refills were performed when indicated by optical coherence tomography (OCT) evaluation of disease activity. The primary endpoint was 12 months with an observation period that extended through 36 months. The primary objective of the study was safety assessment, with secondary objectives that included functional measurements. Four of the patients suffered significant or serious adverse events (endophthalmitis, vitreous hemorrhage (2), and traumatic cataract) but three of these four had improved vision by the study's endpoint. The average visual acuity gains for the cohort were +10 letters, 10 eyes (50%) gained at least three lines, and two (10%) lost at least three lines. The mean number of refills through 12 months was 4.8 per patient. The multicenter, randomized, treatment-control, phase II LADDER trial will include 220 patients at 55 U.S. sites. Patients will be randomized 3:3:3:2 to receive one of three different ranibizumab implant doses or monthly 0.5 mg ranibizumab injections. Study enrollment was completed in October 2017 \[[@B30-pharmaceutics-10-00021]\]. 4.4. Gene Therapy {#sec4dot4-pharmaceutics-10-00021} ----------------- Avalanche Biotechnologies developed a viral delivery system (AVA-101) to induce long-term anti-VEGF receptor synthesis by the outer retina. An adenovirus vector inserts the DNA for a naturally occurring sFLT-1 (soluble VEGF receptor-1) into RPE cells. Infected cells synthesize and excrete the soluble VEGF inhibitory protein into the outer retina and choriocapillaris. In a phase IIa trial, 21 patients with nAMD received AVA-101, with 0.5 mg ranibizumab injected both at baseline and one month, and as rescue therapy when needed. Patients underwent core vitrectomy and subretinal injection of AVA-101 adjacent to the macula at day seven. Evaluations were performed monthly and patients were eligible for rescue ranibizumab therapy based on pre-specified criteria. Eleven control patients received only 0.5 mg ranibizumab monthly. At the 52-week endpoint, mean improvement in BCVA was +2.2 letters in the AVA-101 group compared to −9.3 letters in the control group \[[@B31-pharmaceutics-10-00021]\]. These differences were statistically significant. Mean center point thickness improved by −27 µm in the AVA-101 group and −85 µm in the control group. There were no serious ocular adverse events in the AVA-101 group and no systemic safety signals were noted. All patients in the AVA-101 group that were phakic at baseline developed cataracts and three (14%) developed moderate vitreous hemorrhages. Gene therapy was well tolerated by patients but the technology failed to provide a complete or durable anti-VEGF response. Though AVA-101 produced superior BCVA changes compared to the control group, the overall performance of the AVA-101 group was disappointing. Soon after the phase IIa trial results were announced, Avalanche decided not to proceed with phase IIb trials \[[@B54-pharmaceutics-10-00021]\]. 4.5. Encapsulated Cell Technology {#sec4dot5-pharmaceutics-10-00021} --------------------------------- Encapsulated cell technology (ECT) uses immortalized RPE cells that have been programmed to over-synthesize a specified biochemical product, and packages them in a cylinder lined by semi-permeable membranes that allow ingress of nutrients and egress of the synthesized product. The membrane prevents outward migration of the modified cells while shielding them from the body's immune system. The 10 mm long cylinder is surgically implanted through the pars plana and is sutured to the sclera. Trials with ciliary neurotrophic factor (CNF) production have been completed in eyes with retinitis pigmentosa and atrophic AMD \[[@B32-pharmaceutics-10-00021]\]. Pharmacokinetic analyses showed that the half-life of CNF production was 54 months and the ECT cylinder was well tolerated. Unfortunately, the trials failed to meet their primary therapeutic endpoints. Phase I trials with a cylinder that produces a high-affinity VEGF binding protein similar to aflibercept have been completed. A multi-center phase II trial compared a higher dose, anti-VEGF implant against ranibizumab therapy. The trial was discontinued early because a larger number of patients than expected required intravitreal rescue injections \[[@B33-pharmaceutics-10-00021]\]. No further nAMD trials have been announced but Neurotech continues to develop the platform for other retinal vascular conditions such as macular telangiectasia. 4.6. Colloidal Carriers {#sec4dot6-pharmaceutics-10-00021} ----------------------- Injections of some liposomal drug formulations have shown promise including early work with anti-VEGF agents. In experimental models, in vitro release of ranibizumab from negatively charged liposomes was exhausted at two days, whereas ex vivo transport across sclera (simulating a subconjunctival injection) occurred in a linear manner for seven days \[[@B34-pharmaceutics-10-00021]\]. This suggests that sclera acts as a classic membrane that allows the diffusion of liposomal-formulated ranibizumab and raises the possibility that subconjunctival injections could serve as long-acting depots. These results differ from those reported by Kim et al. \[[@B55-pharmaceutics-10-00021]\] in which poly lactic-co-glycolic acid nanoparticles and liposomes do not facilitate drug diffusion across sclera. A steep concentration gradient created by the thick sclera, Bruch's membrane-choroid, and retinal pigment epithelium results in low drug concentrations within the retina. 4.7. Pump Delivery {#sec4dot7-pharmaceutics-10-00021} ------------------ Microelectromechanical system (MEMS) technology is a miniaturized system that is currently used in insulin pumps to deliver drug to tissues. The Posterior MicroPump Drug Delivery System (PMP, Replenish Inc., Pasadena, CA, USA) using MEMS technology is implanted on the sclera, similar to placement of a glaucoma drainage device, to deliver drug into the eye. Long-term safety after implantation into animal eyes has been demonstrated \[[@B35-pharmaceutics-10-00021],[@B36-pharmaceutics-10-00021]\] as the PMP reliably delivered 100 programmed doses of an anti-VEGF drug (equivalent to over eight years of therapy). The PMP was well tolerated by 11 patients with DME over three months, with no cases of endophthalmitis or strabismus \[[@B37-pharmaceutics-10-00021]\]. 4.8. Topical Therapy {#sec4dot8-pharmaceutics-10-00021} -------------------- PanOptica, Inc. is developing a topical anti-VEGF medication (PAN-90806) for the treatment of nAMD and proliferative diabetic retinopathy (PDR). In animal models, pharmacokinetic measurements show excellent drug concentrations in the central retina and choroid as long as 17 h after administration. Control of leakage and bleeding from choroidal neovascular membranes was comparable to that achievable with intravitreal anti-VEGF antibodies, but with minimal systemic exposure to the drug. In a phase II trial with 50 treatment-naïve nAMD patients, an independent panel of experts judged that PAN-90806 showed promise as a therapeutic agent \[[@B38-pharmaceutics-10-00021]\]. Approximately 45--50% of treated patients experienced improvements in vascular leakage, retinal morphology, and vision. No systemic adverse events were noted and ocular surface irritation due to the eye drops reversed when therapy was discontinued. PanOptica plans to investigate higher doses in a phase I/II nAMD trial, and a phase I trial for the treatment of PDR is underway. 5. Discussion {#sec5-pharmaceutics-10-00021} ============= The quest for longer duration anti-VEGF therapies continues along several fronts with the injectable drugs abicipar and brolucizumab most likely to achieve US FDA approval. Each drug may be shown to be effective as q12week therapy---roughly half of the brolucizumab patients were sustained on q12week injections---but the importance of such a finding is not clear. Control groups in the phase III trials were treated with q4week ranibizumab and q8week aflibercept but neither of these drugs was tested in a q12week arm. Therefore, true head-to-head comparisons of these control drugs to abicipar and brolucizumab have not been performed with similar injection frequencies. In the VIEW trials, q4week ranibizumab was compared to q8week aflibercept during the first year. Aflibercept-treated patients experienced comparable improvements in BCVA and edema at 52 weeks compared to ranibizumab and was approved for q8week therapy (compared to q4week for ranibizumab). However, when patients received PRN (with 12-week cap) injections in the second year of the trials, aflibercept-treated patients received a mean of 4.2 injections, compared to 4.7 for ranibizumab. This difference in durations of action has been estimated to be five to seven days and post-approval experience also suggests that the difference is small. It is reasonable to suspect that post-approval differences with abicipar and brolucizumab will also be disappointingly small. The quest for a single application, long-term anti-VEGF therapy has been disappointing. Encapsulated cell technology and adenovirus-mediated gene therapy are exciting technologies, but both failed to perform adequately in phase II trials and neither developer will pursue phase III anti-VEGF trials. The ranibizumab port delivery system allows for trans-conjunctival (as opposed to intravitreal) injections as needed. However, since the phase I trial required a mean of 4.8 refills over the course of 12 months, this does little to decrease the frequency of clinic visits or injections. Unless the new dosing arms in the phase II trials decrease the number of refills, many physicians will likely continue with PRN and treat and extend regimens since they have comparable treatment burdens. The use of eye drops does not constitute long-duration therapy but some patients will prefer self-administering drops when coupled with infrequent visits to the clinic. Eye drops effectively treat many anterior segment conditions and experimental CNVM in rats, but drops do not effectively treat retinal disorders in humans. Because topically delivered medications must pass through cornea, conjunctiva, sclera, uvea, and vitreous to reach the retina, the molecule must be small. Therefore, antibody-related macromolecules would be ineffective in eye drop form. We have been fortunate in identifying VEGF as a pivotal molecule in the pathogenesis of chorioretinal vascular conditions, but just as the search for additional molecular targets has been disappointing, our attempts to significantly extend the duration of action of anti-VEGF therapy has met with more failures than successes. Despite ongoing research, it remains likely that frequent injection of anti-VEGF drugs will remain the standard-of-care for several years to come. Michael W. Stewart was solely responsible for the production of this manuscript. Alkahest: Consultant; Allergan: Institutional research support; Bayer: Consultant; Regeneron: Institutional research support. pharmaceutics-10-00021-t001_Table 1 ###### This table lists the currently available anti-VEGF drugs, several that have failed clinical trials, and others that are in various stages of development. Additional information includes regulatory approvals and comments on drug characteristics, pharmacokinetics, preclinical studies, and clinical trials. AMD: age-related macular degeneration; DME: diabetic macular edema; DR: diabetic retinopathy; RVO: retinal vein occlusion; VEGF: vascular endothelial growth factor; PlGF: placental growth factor; CNVM: choroidal neovascular membrane; RPE: retinal pigment epithelium; BCVA: best corrected visual acuity. ---------------------------------------------------- ---------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **Currently Available Drugs** **Drug** **Approvals** **Comments** Pegaptanib Neovascular AMD Binds to VEGF~165~Poor efficacy \[[@B19-pharmaceutics-10-00021]\], used rarely Bevacizumab Advanced carcinomas \[[@B20-pharmaceutics-10-00021]\] Off-label for all ophthalmic use Recombinant, humanized, murine antibody to VEGF-ANational Eye Institute sponsored studies have established effectiveness for neovascular AMD \[[@B6-pharmaceutics-10-00021]\], DME \[[@B21-pharmaceutics-10-00021]\], and RVOsInexpensive dose cost after compoundingMost commonly used intraocular anti-VEGF drug in the United States Ranibizumab Neovascular AMD, DME, DR, Macular edema due to RVOs, Myopic CNVM \[[@B22-pharmaceutics-10-00021]\] Recombinant, humanized, murine antibodyfragment (Fab) to VEGF-A \[[@B4-pharmaceutics-10-00021]\]Most thoroughly studied anti-VEGF drug Aflibercept Neovascular AMD \[[@B23-pharmaceutics-10-00021]\], DME, DR, Macular edema due to RVOs Completely human, fusion protein, soluble receptor \[[@B24-pharmaceutics-10-00021]\]High affinity for VEGF-A, VEGF-B, and PlGF Conbercept Neovascular AMD (China only) Similar structure and binding affinity as aflibercept \[[@B24-pharmaceutics-10-00021],[@B25-pharmaceutics-10-00021]\]In phase III DME trialUnited States trials being planned **Therapies Under Development or Recently Failed** **Drug** **Technology** **Comments** Abicipar Designed Ankyrin Repeat Protein (DARPin) Pegylation may extend intravitreal half-life (estimated as 13.4 days in humans) \[[@B26-pharmaceutics-10-00021]\]Phase III CEDAR and SEQUOIA nAMD trials have completed enrollmentq8week and q12week experimental arms; control is q4week ranibizumab Brolucizumab Single strand, antibody fragment Small size (26 kDa) allows for injection of large quantity of drug \[[@B27-pharmaceutics-10-00021]\]Phase III nAMD trials recently completed57% and 52% of eyes sustained with q12week injection intervals \[[@B28-pharmaceutics-10-00021]\] Ranibizumab Port Delivery System Trans-scleral refillable drug reservoir Reservoir is refilled via trans-conjunctivalinjectionPhase I study showed +10 letter improvement in BCVA with average of 4.8 refills \[[@B29-pharmaceutics-10-00021]\]Phase II LADDER trial underway with three different dose treatment arms \[[@B30-pharmaceutics-10-00021]\] AVA-101 Adenovirus vector Insertion of soluble VEGF-receptor DNA Injected subretinally after vitrectomyBCVA changes were better than ranibizumab in phase II trial but both arms performed poorly with minimal decrease in edema \[[@B31-pharmaceutics-10-00021]\] NT-503 Encapsulated Cell Technology using immortalized RPE cells Ciliary neurotrophic eluting device failed in dry AMD and retinitis pigmentosa trials \[[@B32-pharmaceutics-10-00021]\]High dose (NT-503) device failed in phase II neovascular AMD trial \[[@B33-pharmaceutics-10-00021]\]Currently being tested in patients with macular telangiectasia Colloidal Carriers Liposomal formulated ranibizumab Liposomal formulation delays drug releaseRanibizumab can cross sclera after subconjunctival depot \[[@B34-pharmaceutics-10-00021]\] Posterior Micropump Delivery System Microelectromechanical Systems (MEMS) Technology Same technology as in insulin pumpsSafely delivered 100 injections in animal models \[[@B35-pharmaceutics-10-00021],[@B36-pharmaceutics-10-00021]\]Three-month DME trial in humans was well tolerated \[[@B37-pharmaceutics-10-00021]\] PAN-90806 Small molecular weight drug Formulated for eye drop deliveryIn animal models, found to produce high retinal concentrations 17 h laterJudged to show therapeutic promise in small human nAMD study \[[@B38-pharmaceutics-10-00021]\] ---------------------------------------------------- ---------------------------------------------------------------------------------------------------- -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

1. Introduction {#sec1-materials-13-03111} =============== The efficiency, stability, and lifetime of any device is strongly dependent on the stability of compounds and alloys that a given device is made of. The attempts to understand the particular state of matter are crucial points of understanding its implications for application in advanced electronic devices. This gain in importance when exotic states of matter are considered. Certain exotic behavior, namely the insulating character of the bulk and metallic character of the surface, was observed in a large group of materials, which were later classified as topological insulators \[[@B1-materials-13-03111],[@B2-materials-13-03111]\]. Among classified topological insulators the Bi~2~Te~3~ compound, considered as good material for thermoelectric \[[@B3-materials-13-03111],[@B4-materials-13-03111]\] and optoelectronic \[[@B5-materials-13-03111],[@B6-materials-13-03111]\] applications, was selected as the subject of this study. Topological insulators (TIs) are a novel type of quantum materials of which the bulk part is a band insulator whereas the surface always exhibits the presence of electronic states carrying electric current \[[@B2-materials-13-03111],[@B7-materials-13-03111]\]. This phase shows a clear bulk energy gap separation like any ordinary insulator or semiconductor, but the difference is in the presence of the gapless edge states (2D) or surface states (3D) which are protected by time-reversal symmetry. The unusual nature of the TI surface states resulted in the widespread interest in the fabrication and characterization of this class of materials. The nature of the TI class of materials resulting from its unique spin nature is gaining in importance when one realizes that TIs properties may entail generation of quasiparticles and electronic states which are not accessible in classic condensed-matter systems. Not surprisingly the topological insulators are considered as promising materials for multiple applications in next generation electronic or spintronic devices \[[@B8-materials-13-03111],[@B9-materials-13-03111]\] as well as for applications in energy conversion such as thermoelectrics \[[@B10-materials-13-03111],[@B11-materials-13-03111]\]. It is well known that defects, strain, and doping influence are important features of the TIs. Coulomb, magnetic, or disorder perturbations can modify the surface states \[[@B12-materials-13-03111],[@B13-materials-13-03111],[@B14-materials-13-03111]\]. Elemental doping and alloying in the TI materials allows for the control of the Fermi level via population with the n-type or p-type carriers which is also a crucial step towards the future junctions of topological insulators that associate two doped TI layers \[[@B15-materials-13-03111],[@B16-materials-13-03111]\]. Among doped TIs, special interest is placed on magnetic dopants, mostly due to the possible interplay between topological magnetic orders, which may lead to the implementation of exotic quantum phenomena, such as the magnetoelectric effect \[[@B17-materials-13-03111]\] and quantum anomalous Hall effect (QAHE) \[[@B18-materials-13-03111],[@B19-materials-13-03111]\], and later on the exploration of its potential device applications. Recent research is focused mostly on TI's electronic structure and their thermoelectric properties and include structures containing transition metals dopants like V, Mn, Cr, Fe, and Cu \[[@B20-materials-13-03111],[@B21-materials-13-03111],[@B22-materials-13-03111],[@B23-materials-13-03111],[@B24-materials-13-03111],[@B25-materials-13-03111]\], and rare earth dopants Gd, Ce, Y, Sm, Dy, and Eu \[[@B8-materials-13-03111],[@B9-materials-13-03111],[@B10-materials-13-03111]\]. Mn and Fe doping of the Bi~2~Te~3~ causes the opening of a small gap at the Dirac point in the surface Dirac cone \[[@B22-materials-13-03111]\]. Further, with the formation of ferromagnetic state only at the surface, a state which leads to opening the gap, has been confirmed for the Mn doped Bi~2~(Te,Se)~3~ as well as for the Cr-doped (Bi,Sb)~2~Te~3~ thin films \[[@B26-materials-13-03111]\]. Bi~2~Te~3~ doped with rare earth is discussed mostly from the point of view of modifying the thermoelectric properties \[[@B27-materials-13-03111],[@B28-materials-13-03111],[@B29-materials-13-03111],[@B30-materials-13-03111],[@B31-materials-13-03111],[@B32-materials-13-03111],[@B33-materials-13-03111]\]. Controlling the distribution of magnetic dopants within the TI's matrix resulting from sample preparation procedures and chemical potentials of the constituent atoms is still challenging and requires in-depth and careful characterizations at the atomic level. This is especially the case when considering nanoscale systems, where already a controlled fabrication, quality control, as well as subsequent preservation of the completed structure or even device requires specific attention. In our previous paper \[[@B34-materials-13-03111]\], we have studied the multilayered structures of Bi~2~Te~3~ and europium and shown that the interface between the layers is not stable and the reaction that took place at room temperature led to the decomposition of the originally layered structure and formation of new phases. This observation led us to focus on the electronic and crystallographic structures of a simplified structure where europium is doped in the Bi~2~Te~3~ thin film. Europium by itself is quite chemically active, and possesses interesting magnetic properties related to its valence state. It is worth noting that it is reasonable, at a first glance, to assume that europium atoms will partially substitute the bismuth atoms in the Bi~2~Te~3~ matrix as it happens for the Bi~2~Te~3~ doped with Gd \[[@B35-materials-13-03111],[@B36-materials-13-03111]\]. Recently, the homogeneous incorporation of Eu^2+^ into the Bi sites of the Bi~2~Te~3~ lattice was observed \[[@B37-materials-13-03111]\]. Our point of interest was strongly aimed at the electronic structure of the obtained film, especially on the valence state of europium. Europium in alloys and compounds occurs in 2+ and 3+ valence states. Eu^3+^ is non-magnetic (J = 0) while the Eu^2+^, observed for pure Eu and europium in intermetallic alloys and compounds \[[@B38-materials-13-03111],[@B39-materials-13-03111]\], has a large pure spin moment (J = 7/2). A mixed valence state has also been found in many intermetallic compounds. In the presented studies, the response of thin film of Bi~2~Te~3~ to the disorder introduced by Eu dopant is discussed. The studies were performed on thin films with an unprecedented combination of techniques such as photoemission spectroscopy, TOF-SIMS, RHEED, femtosecond spectroscopy, and LC-AFM. 2. Materials and Methods {#sec2-materials-13-03111} ======================== 2.1. Materials {#sec2dot1-materials-13-03111} -------------- The molecular beam epitaxy system was used for the growth of two thin films of Bi~2~Te~3~. The films were grown in accordance with the procedure we developed earlier \[[@B40-materials-13-03111]\] for the Bi~2~Te~3~ monocrystalline films. Freshly cleaved (110) 0.15--0.17 mm thick Muscovite mica substrates (Ted Pella, Inc., Mica grade V1) was used as a substrate for the Bi~2~Te~3~ films. In order to prevent the problem of the sample surface charging resulting from the used substrate, the mica surface was ground using silver paste. The mica was outgassed at 300 °C for 1.5 h under UVH conditions (5 × 10^−9^ mbar) and kept at 120 °C during the deposition of the Bi~2~Te~3~ films. High-purity Bi (99.999% Aldrich Chem. Co., Milwaukee, WI, USA), spectrographically standardized Te ingot (Johnson Matthey Chemicals, London, UK), and europium were evaporated by conventional effusion Knudsen cells (K-cell). The growth rate of each element, controlled by quartz microbalance, was customized to obtain assumed stoichiometry. The total growth rate of the films was about 0.2 QL min^−1^. In order to obtain a uniform distribution of deposited elements, the films were grown in the co-deposition mode. For the europium doped Bi~2~Te~3~ film, the growth rates of elements were adjusted as if Eu atoms were to substitute for bismuth in the Bi~2~Te~3~ crystal lattice. Assumed level of europium dopants was set at the level of about 2--3%. After the deposition process the films were annealed at 100 °C for 18 h. During the deposition process, the vacuum was of about 4 × 10^−9^ mbar for the undoped film and about 9 × 10^−9^ mbar for Bi~2~Te~3~ film doped with europium. Further measurements were performed on films cooled to room temperature. Additionally, pure, bulk europium was used as a reference sample for the photoelectron spectroscopy measurements. 2.2. Characterization Methods {#sec2dot2-materials-13-03111} ----------------------------- In-situ structural characterization of the deposited film was realized with the use of reflection high energy electron diffraction (RHEED, STEIB Instrument, Langenbach, Germany) and X-Ray Photoelectron Spectroscopy. The RHEED diffraction patterns were obtained using an electron gun set at 15 kV, 1.4 mA along \[100\] direction after annealing and cooling down to room temperature. The electronic structure studies were conducted in-situ after the growth process using the XPS spectrometer (Physical Electronics PHI 5700, Chanhassen, MN, USA). For the XPS measurements, the monochromatic X-ray (1486.7 eV, Al Kα) radiation was used to generate photoelectron spectra; the survey spectra and high-resolution photoelectron spectra of Bi4f, Te3d, Eu3d, Eu4d, and valence band were collected at room temperature. The transmission electron microscopy (TEM) observations were performed using JEOL JEM-3010 high-resolution (HR-TEM) machine (Tokyo, Japan) working with 300 kV acceleration voltage and equipped with a Gatan 2k × 2k Orius™ 833 SC200D CCD camera (Gatan-Pleasanton, CA, USA). Atomic force microscopy (AFM) (Omicron VT AFM/STM, Taunusstein, Germany) was performed ex-situ under UHV condition at room temperature. AFM in tapping mode was used to study the layer growth process with high lateral resolution while the contact mode (LC-AFM) with conducting tip was applied to examine the character of the local conductivity at various polarization and bias voltage values up to 50 mV. Time of flight-secondary ion mass spectrometry (TOF--SIMS) measurements were carried out with the use of a TOF-SIMS.5 (ION-TOF GmbH, Munster, Germany) reflection-type spectrometer, equipped with bismuth liquid metal ion gun and cesium gun. 2D micro-and 3D nanostructure was examined using Bi^+^ ion beam (30 kV, 0.5 pA) for the analysis, and Cs^+^ (500 V, 100 nA) for depth profiling. Time-resolved optical measurements were performed using the femtosecond pump-probe technique, based on a 80 MHz repetition rate Ti:sapphire femtosecond laser (120 fs). The details of the set-up can be found in Ref. \[[@B41-materials-13-03111],[@B42-materials-13-03111]\]. This permits to probe the impact on the Eu doping on the electron and phonon dynamics. For all ex-situ measurements, the film was measured on mica substrate without any additional preparation with the exception of TEM measurements. In that latter case, the film was partially mechanically exfoliated from the mica substrate and later placed on the copper mesh. The high-resolution images were obtained near the edge of the exfoliated film, where remaining mica support was thin or absent. 3. Results and Discussion {#sec3-materials-13-03111} ========================= 3.1. Crystalographic Structure {#sec3dot1-materials-13-03111} ------------------------------ The crystallographic structure of obtained films was studied right after the deposition process with the use of a RHEED diffractometer. The results presented in [Figure 1](#materials-13-03111-f001){ref-type="fig"} indicate monocrystalline character of the undoped ([Figure 1](#materials-13-03111-f001){ref-type="fig"}a) and Eu doped ([Figure 1](#materials-13-03111-f001){ref-type="fig"}b) films. The in-plane lattice parameter calculated for the trigonal symmetry as in the Bi~2~Te~3~ structure (space group 166, $R\overline{3}m$) was 4.5 Å for the undoped film and 4.4 Å for the film doped with europium. Those values are in good agreement with the a = 4.38 Å parameter of the Bi~2~Te~3~ compound. The europium doping caused a slight expansion of the Bi~2~Te~3~ unit cell. Moreover, analyzing the RHEED images, a blurring of the diffraction pattern that can be observed for the Eu doped film is most likely related to weakly developed monocrystalline grains. High-resolution TEM images of the film surface aligned perpendicular to the electron beam as well as obtained diffraction patterns indicate the presence of monocrystalline ordering of the atoms in the film ([Figure 2](#materials-13-03111-f002){ref-type="fig"}a). The grains have rather an irregular shape, their size varies between 8--25 nm, however, the grains are oriented towards each other at an angle of 60 degrees. Such an arrangement of the single crystalline grains is a favorable situation to obtain a RHEED pattern typical to a macroscopic single crystal as presented in [Figure 1](#materials-13-03111-f001){ref-type="fig"}. This would also explain blurring of the RHEED pattern of Eu doped film. The trigonal symmetry is perfectly confirmed when looking at the atomic scale. The in-plane lattice parameter was of the same value as given for pure Bi~2~Te~3~ \[[@B43-materials-13-03111]\] and value obtained in our previous paper \[[@B40-materials-13-03111]\]. The high-resolution TEM image shown in [Figure 2](#materials-13-03111-f002){ref-type="fig"}b represents an image obtained with the electron beam aligned parallel to the film surface. The visible stripes quite accurately show a layered structure of the film. The distance between those features is about 1 nm and is in good agreement with the thickness of the single quintuple layer \[[@B44-materials-13-03111],[@B45-materials-13-03111]\]. 3.2. Electronic Structure {#sec3dot2-materials-13-03111} ------------------------- The structural characterization was completed with the analysis of the electronic structure studied in terms of photoelectron spectroscopy. The studies were performed in-situ after the deposition process and compared to that of the undoped Bi~2~Te~3~ film and metallic Eu. Both Bi~2~Te~3~ films were obtained in the same system at the equivalent conditions. Both had a thickness of about 15 nm. Bulk Eu was prepared in the UHV conditions by scraping with a diamond file. No contamination by oxygen nor carbon was detected. The XPS core level spectra Bi 4f~5/2,~ Te 3d and Eu 4d were used to calculate the atomic composition which was Te:Bi:Eu = 55.3:41.3:3.4. The calculations were performed using Multipak v. 9.6 software, the Shirley method of background removal was applied. The ratio Te/(Bi + Eu) is slightly lower than expected but similar composition was found for high quality Bi~2~Te~3~ films \[[@B40-materials-13-03111],[@B41-materials-13-03111]\]. In the valence band region (see [Figure 3](#materials-13-03111-f003){ref-type="fig"}), the difference between the undoped and Eu doped film is mostly pronounced in the binding energy range around 1 eV, close to the top of the valence band and in the region of about 3--4 eV. That difference may be related with the contribution from the Eu 4f states which have a high sensitivity factor for photoionization and despite the low Eu content give a relative increase of intensity in the region of about 1--4 eV. The Eu 4f states in the pure Eu are visible as a clear peak with a maximum at about 2.1 eV. It is worth mentioning that Denecke et al. reported the position of the Eu 4f peak at about 1.5 eV for the (Pb,Eu)Te bulk system \[[@B46-materials-13-03111]\]. The region close to the Fermi level, where the surface states plus possibly bottom of the conduction band are located, is hardly affected by Eu doping. This is in accordance with the measured electrical properties of the film and the transient reflectivity data described below. Eu remains in the divalent state, which can be clearly seen in the exchange split core d levels and by the lack of trivalent Eu 4f contribution in the region of binding energy 5--10 eV \[[@B47-materials-13-03111]\]. The chemical shifts of Te and Bi in undoped films show the expected behavior where the chemical shift of Bi is positive and that of Te is negative with respect to pure elements. The position of the Bi 4f and Te 3d photoemission levels is the same as for the undoped film. There is a slight difference in the line width as the Bi line is broader in the doped film (see [Figure 4](#materials-13-03111-f004){ref-type="fig"}a) while the Te one is slightly narrower ([Figure 4](#materials-13-03111-f004){ref-type="fig"}b). A slight broadening of the Bi lines can be related with a modified bonding when the nearest neighbor is Eu atom. The very interesting point is the negative chemical shift of the Eu 4f level which is visible for Eu core levels. [Figure 5](#materials-13-03111-f005){ref-type="fig"} shows the comparison of the Eu 4d and 3d spectra obtained for the doped film with pure Eu. The shape of the core levels and well resolved exchange splitting confirms the divalent state indicating the presence of half occupied 4f shell. Assuming the stable ^7^S ground state for the Eu 4f shell, the negative shift may be related with the reconfiguration of the conduction electrons leading to the reduction of screening of the nuclear charge. This is surprising given the presence of Te ligands which have a much higher value of electronegativity. For the Eu 4d multiplet (see [Figure 5](#materials-13-03111-f005){ref-type="fig"}a) the general structure is the same as for Eu metal except the intensity ratio of the first exchange split lines \[[@B48-materials-13-03111]\]. The splitting, i.e., distance between the first five well resolved lines within the main 4d~5/2~ peak is of about 0.85 eV what is very close to the value for Eu metal. The first peak which can be ascribed to the ^9^D~6~ term of the 4d^9^4f^7^ configuration has an Eu metal lower intensity than expected from the quantum multiplicity of states. The same effect was found for Gd and its metallic compounds, and remains unexplained up to date \[[@B49-materials-13-03111],[@B50-materials-13-03111]\]. The intensity ratio of the 4d multiplet components (ratio of two lowest energy lines) observed in Eu doped Bi~2~Te~3~ film is thus exceptional. Similar structure was only found in early studies on bulk EuTe \[[@B51-materials-13-03111]\] and recent studies of europium incorporated in Bi~2~Te~3~ matrix \[[@B37-materials-13-03111]\]. That exceptional behavior can be related to the spin dependent relaxation rates of the photoexcited 4d states. The lowest spin dependent relaxation rate is expected for the state with the highest spin polarization and lowest binding energy within the mutiplet \[[@B49-materials-13-03111]\]. The lack of the effect in Eu doped Bi~2~Te~3~ may be related to the modified electronic structure close to the Fermi energy although the ground state of the Eu 4f level is ^7^S~0~ (divalent Eu as for Eu metal), leading to the well resolved exchange splitting and spin polarization of all core levels. 3.3. TOF-SIMS Depth Profiles {#sec3dot3-materials-13-03111} ---------------------------- After removing the (Eu, Bi)~2~Te~3~ film from the growth chamber, the film was transported under atmospheric conditions to the vacuum chambers of the mass spectrometer and atomic force microscope, the transport process took no more than a few minutes. The ex-situ TOF-SIMS characterization was focused on the analysis of the spatial distribution of layer components. It should be remembered that the TOF-SIMS technique, unlike the XPS, does not provide information on the atomic concentration of layer components, here the number of detected ions is related to the ionization probability and not to the real amount of particular element in the film. Moreover, the depth of analysis for mass spectrometry is about an order of magnitude smaller than for photoelectron spectroscopy. [Figure 6](#materials-13-03111-f006){ref-type="fig"}a presents depth profiles indicating that the distribution of film components is not uniform as it supposed to be in a homogenously doped film. Europium ions exhibit a tendency to be uniformly distributed within the film, however in the upper part of the layer large amount of them was detected. A larger amount of europium at the surface of the film is probably related to surface segregation during the film growth or to the annealing process and, to a small extent, related to the oxidation of the film surface, an effect of migration towards the sample surface and further oxidation of the Eu. Interestingly, the presence and the distribution of the EuTe^+^ ionic species indicate that a relatively thick layer containing Eu and Te was formed in the top part of the film. As this layer also contains bismuth, one can relate that part of the film to the main Bi~2~Te~3~ phase with europium substituting bismuth. This is in agreement with the XPS data for the as grown sample. Regarding the analysis of the distribution of Te and Bi atoms, we used Cs~2~Te^+^ ionic species and BiTe^+^. The cesium forms that kind of ionic species with Te and Bi and allows to increase the signal derived from secondary ions of those elements \[[@B52-materials-13-03111]\]. The distribution of tellurium represented by the Cs~2~Te^+^ ionic species is not uniform. The linear depth profile and 3D distribution show that tellurium occurs in the entire film except for the sample surface, it can be also seen that a local maximum of Te atoms is present in the middle part of the film, however, it is not in the same place where the maximum for EuTe^+^ was observed. The bismuth in the analyzed film, represented by CsBi^+^ species, is partially accumulated in the area near the mica substrate. It seems that the formation of sub layer from which EuTe^+^ ions have been emitted leads to the displacement of some amount of the bismuth atoms towards the substrate. This result together with described above results of structural analysis confirms substituting Eu in the Bi~2~Te~3~ lattice. In our previous work \[[@B34-materials-13-03111]\] we already observed similar displacement of Bi atoms in the multi-layered system. Surface oxidation, typical for rare earth doped Bi~2~Te~3~ thin films \[[@B32-materials-13-03111]\], may, as in ex-situ measurements performed by us, enhance the segregation and substitution effect. Analysis of the Al^+^ signal, representing the mica substrate, indicate that the surface of the mica is not fully flat, the features (see [Figure 6](#materials-13-03111-f006){ref-type="fig"}b, Al^+^) visible in the top part of the 3D depth profile suggest the presence of small island or crumbs on the surface of mica. That is also the reason why the interface between the mica and deposited film is not sharp as can be seen in [Figure 6](#materials-13-03111-f006){ref-type="fig"}a for the Al^+^ linear profile. 3.4. Phonon-Electron Interactions {#sec3dot4-materials-13-03111} --------------------------------- To complete the electronic and structural characterization, time-resolved optical measurements have been conducted to evaluate the electron--phonon collision time and to assess the impact of such chemical and structural disorder onto the phonon dynamics. The Eu doped Bi~2~Te~3~ film was investigated after a few weeks after the deposition process. A stable oxidation layer appears, as already evidenced earlier \[[@B41-materials-13-03111],[@B42-materials-13-03111]\]. This oxide layer thickness was estimated to be around 2 nm \[[@B41-materials-13-03111],[@B42-materials-13-03111]\]. We have then measured the transient optical reflectivity signal of two films with and without Eu doping. The signals are shown in [Figure 7](#materials-13-03111-f007){ref-type="fig"}. The optical penetration of both the pump and probe light is similar to the film thickness, so we excite and probe the entire thickness. Moreover, the light impinges on the surface with a diameter of around 5 micrometers. So, the information that such technique obtains is an average response over that photoexcited volume. We do not have in plane resolution, however, since we are able to record time of flight of phonons (acoustic) we are sensitive to in-depth inhomogeneities as we will discuss in the following \[[@B53-materials-13-03111],[@B54-materials-13-03111],[@B55-materials-13-03111]\]. The transient optical reflectivity ([Figure 7](#materials-13-03111-f007){ref-type="fig"}) exhibits a fast and sharp variation just after the femtosecond excitation due to the ultrafast excitation of the electronic cloud. Then, the signal decays in time due to the relaxation of hot carriers and to heat diffusion process. Some oscillating components superimpose on the electronic response ([Figure 7](#materials-13-03111-f007){ref-type="fig"}a) and can be evidenced more clearly once the baseline is removed ([Figure 7](#materials-13-03111-f007){ref-type="fig"}c and [Figure 8](#materials-13-03111-f008){ref-type="fig"}a). The oscillatory components come from a photo-induced coherent optical phonon (THz range as seen in [Figure 7](#materials-13-03111-f007){ref-type="fig"}c,d) and a coherent acoustic phonon (sub-THz range as seen in [Figure 8](#materials-13-03111-f008){ref-type="fig"}a,b). The assignment of these oscillations has been previously established for undoped Bi~2~Te~3~ \[[@B41-materials-13-03111],[@B42-materials-13-03111]\]. The phonons signal is extracted after removing the baseline. We have recorded the signal with a broad time window (0--30 ps) with low time resolution sufficient to record the acoustic phonon signals ([Figure 8](#materials-13-03111-f008){ref-type="fig"}). In contrast, for evidencing more clearly the high frequency phonons, we have increased the time resolution of the record (refined sampling) but we have limited the record to the first 8ps ([Figure 7](#materials-13-03111-f007){ref-type="fig"}b,c). As observed in [Figure 7](#materials-13-03111-f007){ref-type="fig"}a,b within the first 5ps, the signal decays rapidly, which is due to the hot carriers' relaxation. A time of typically 2 ps is found consistent with previous reports \[[@B41-materials-13-03111],[@B42-materials-13-03111]\]. We note that such electron-phonon relaxation time depends on the electronic structure close to the Fermi level since the photoexcited carriers are promoted towards empty states with quanta of the pump excitation of 1.5 eV and relaxes towards the initial states afterward. In [Figure 3](#materials-13-03111-f003){ref-type="fig"}, we have shown that the valence band structure down to 2 eV below the Fermi level was not too different for the doped and undoped layers. Consequently, we are not surprised to measure equivalent cooling time for Eu-doped and undoped samples. However, the doping seems to have a drastic influence on the phonon dynamics as discussed hereafter. The optical mode has a frequency close to 2THz (Raman active A~1g~ mode seen in [Figure 7](#materials-13-03111-f007){ref-type="fig"}d) in agreement with previous investigations \[[@B41-materials-13-03111],[@B42-materials-13-03111]\]. The fact that we detect the same A~1g~ frequency as the one measured in undoped Bi~2~Te~3~ materials, is an additional confirmation that the right structure has been grown consistently with a conclusion drawn before. We note that such optical phonon frequency is governed by the local interatomic potential and, for this reason, is a good local probe of a structure. However, we can see a huge difference in the lifetime of these coherent optical phonons. While for undoped sample the coherent optical phonon is visible beyond the time window of observation (8ps), its lifetime drastically decreases for the doped sample. This A~1g~ optical phonons have a very low group velocity. As a matter of fact, the observed limited lifetime cannot be attributed to the scattering at grain boundaries or at any defect present in the structure since this A~1g~ phonons do not propagate (contrary to acoustic phonon as discussed later on). Rather, the lifetime is likely associated to local fluctuations of chemical compositions that can modify the interatomic potential and hence modify the intrinsic A~1g~ frequency (slight change). Consequently, the total signal arises from contributions of slightly different A~1g~ frequency coming from unaffected and affected (due to the doping) sites of Bi~2~Te~3~ (the A~1g~ mode is associated to longitudinal displacement within the QL). At this level, we cannot detect the slight difference in the A~1g~ frequency in the time domain, but such different cosines like signals interfere thus reducing the coherence of the signal in the time domain, i.e., reducing the lifetime. This effect is well known, referred to as inhomogeneous broadening, and has been already observed when alloying some semiconductors or where slight disorder exists in a structure \[[@B56-materials-13-03111]\]. So, we can conclude that the doping creates such inhomogeneous broadening due to the perturbation of the local interatomic potential due to the insertion of Eu within the Bi~2~Te~3~ lattice. Since we have shown previously that Eu mainly substitute to Bi, such fluctuations of the interatomic potential can be reasonably attributed to that substitution. The impact of the doping on the elastic properties and scattering of acoustic phonon is another property we reveal. The acoustic phonons signals are very different ([Figure 8](#materials-13-03111-f008){ref-type="fig"}). These acoustic modes are acoustic eigenmodes as shown in previous reports \[[@B42-materials-13-03111],[@B57-materials-13-03111]\]. Said differently, these acoustic frequencies correspond to the light-induced mechanical resonances of the entire film. The mechanical resonance of a film having a thickness H and a longitudinal sound velocity V, is given by f~n~ = nV/2H. There is an important difference to be underlined. For the non-doped sample, we can evidence two eigenmodes (around f~n\ =\ 1~ = 80 GHz and f~n\ =\ 2~ = 160 GHz), as already observed previously \[[@B41-materials-13-03111]\]. We have repeated the experiment on the same sample studied in 2015 to show the stability of the sample and to confirm the acoustic phonon frequency did not change \[[@B41-materials-13-03111]\]. In contrast, for the Eu-doped sample, only one mode is clearly revealed with a frequency shifted upward (around 120 GHz compared to 80 GHz). Both films have a nominal thickness of 15 nm after the MBE growth (the thickness of undoped sample was measured by X-ray Reflectivity see \[[@B41-materials-13-03111]\]). We obtain a sound velocity of 2400 m/s for the undoped Bi~2~Te~3~ while we arrive to a value around 3600 m/s for the doped sample. This difference is considerable. We have noticed that the A1g frequencies ([Figure 7](#materials-13-03111-f007){ref-type="fig"}) are pretty identical, indicating that within the QL the interatomic potential might not change too much upon doping. The reasons for such a large variation of the acoustic phonon velocity might either be due to the reduction of the Bi~2~Te~3~ layer thickness and/or a partial segregation of the dopant or other chemical species relocated between QLs. Such reduction of thickness H, immediately induces an increase of the eigenfrequency f~n~. This scenario could be consistent with TOF-SIMS results where some layers with various chemical composition appear to exist. The other possibility, if we rule out the change of H parameter, could come from a partial segregation of the dopant or other chemical species between QLs. In that case, the sound propagation which is governed by the interatomic interaction between QLs might be affected. 3.5. Local Conductivity {#sec3dot5-materials-13-03111} ----------------------- The closing investigation allowed exploring the surface topography and the local conductivity properties on the surface of the Eu doped Bi~2~Te~3~ film for which we also reveal some interesting spatially inhomogeneous physical properties. The AFM measurements were performed on the external AFM system, and thus the sample was briefly introduced to the atmosphere. It was shown \[[@B39-materials-13-03111]\], that the short introduction to the ambient conditions had a limited impact on the surface electrical properties. The investigations showed a relatively flat surface with a number of larger islands randomly present on the surface. The size of those islands is between 100 and 200 nm, however they also can be found packed in close formations of 300 nm size on average. The maximal island height is close to 20 nm. The average RMS value in the regions without the islands was very low and close to 0.8(1) Å, otherwise (regions with islands) the RMS is close to 3.0--5.0 nm. We did not observe clear triangular-shaped islands (contrary to undoped films \[[@B40-materials-13-03111]\]), however an island-like growth mechanism still occurs. Further, we have investigated the local current maps with the use of LC-AFM--where the conducting AFM tip is in the contact with the surface under a small voltage (several mV). This allows for the investigation of the local conductivity of the surface along with its topography. The typical results for the LC-AFM are shown in [Figure 9](#materials-13-03111-f009){ref-type="fig"}. The map shows that the surface conductivity is inhomogeneous and is correlated with the topography--[Figure 9](#materials-13-03111-f009){ref-type="fig"}c). The interpretation of such conductivity can be drawn with a comparison to the undoped sample surface (see local conductivity maps in \[[@B40-materials-13-03111]\]). First, the regions between the islands are conducting poorly, which is the opposite to the expectation of a topological insulator surface (pure Bi~2~Te~3~ thin film). From the TOF-SIMS technique, one can notice that the Eu content on (or close to) the surface is high. Thus, poor conductivity in such regions can originate from that fact. Eu is very susceptible to oxidation and a layer of EuO or Eu~2~O~3~ is formed. The EuO is semiconducting \[[@B58-materials-13-03111]\], while the Eu~2~O~3~ is insulating giving again very low currents on a current map. Moreover, in the compounds with the divalent Eu, like EuO or EuTe relatively strong magnetic interactions can appear leading to the destruction of the topologically protected states and some regions on the surface become semiconducting. Hence, the relatively low voltage used in our experiment (several mV) results in very low currents on a current map. At the same time, the regions close to the large islands are conducting very well, giving low resistance in the range of 80 kΩ. The resistivity connected to the experimental setup in our LC-AFM can be estimated to 60 kΩ (which depends strongly on the tip-sample contact area). Thus, we estimate that the results of below 100 kΩ are very close to the lowest resistance that can be detected, thus can be interpreted as very large conductivity. In addition, the I--V curves show metallic conductivity, leading to an assumption that those regions conduct with a very high conductivity, which in turn is a hallmark of the topological insulator state. In addition, one can notice a number of small islands conducting poorly, especially those with a relatively low height. One may suppose that the presence of Eu rich phases can be responsible for that behavior. In summary, we assume that the large islands shown in the topography are composed of a main Bi~2~Te~3~ phase with a small Eu doping level, while the rest of the surface is covered by Eu containing phases, most likely oxides. 4. Conclusions {#sec4-materials-13-03111} ============== The impact of Eu doping on the structure and properties of Bi~2~Te~3~ was investigated. Our studies indicated that 15-nm thick film of Bi~2~Te~3~ doped with europium basically preserves the crystallographic structure of the undoped compound. The electron diffraction and TEM images, the measurement of the characteristic frequency of the A1g Raman active mode of Bi~2~Te~3,~ and the density of states probed by UPS-XPS confirm the average structure is successfully obtained. The in-situ XPS results indicate the divalent state of Eu, which means the presence of magnetic atoms within the lattice. Interestingly, the negative chemical shift of Eu core levels has been found and an exceptional structure of the exchange split Eu core levels has been observed. The magnetic interactions related to the magnetic atoms do not seem to influence the presence of surface states, as confirmed by a very high surface conductivity. However, some inhomogeneities have been revealed on the surface by AFM and LC-AFM, as well as in the body of the film thanks to TOF-SIMS and through the analysis of the coherent phonon dynamics. Even if the average structure is preserved, these inhomogeneities give rise to a complex landscape of surface conductivity and might affect some elastic properties of the layer, maybe due to some interstitial dopant between quintuple layers. The surface oxidation locally reduces the conductivity, which is probably connected to the presence of Eu oxides. Further optimization of the doped film growth process is necessary to obtain detailed knowledge concerning the effect of doping necessary to develop new classes of devices. Each of the authors participated in experimental investigations, data analysis, and in describing the results of particular experiments; thin film growth: B.W., K.B.; photoelectron spectroscopy: K.B., M.W. (Mateusz Weis.), J.S.; mass spectroscopy: K.B.; atomic force microscopy: M.W. (Marcin Wojtyniak); electron microscopy: M.Z.; femtosecond spectroscopy: P.R., R.G., M.W. (Mateusz Weis), B.W. The paper draft was prepared by K.B., and reviewed by J.S. and P.R., J.S. supervised the work and acquired funds for its realization. All authors have read and agreed to the published version of the manuscript. This work was conducted under Grant No. 2016/21/B/ST5/02531funded by the National Science Centre, Poland and under the support of Région Pays de la Loire, France and the PPI program (Femtosecond Plateform). The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. {#materials-13-03111-f001} {#materials-13-03111-f002} {#materials-13-03111-f003} {#materials-13-03111-f004} {#materials-13-03111-f005} {#materials-13-03111-f006} {#materials-13-03111-f007} {ref-type="fig"}a) for undoped and Eu doped Bi~2~Te~3~ film. (**b**) Corresponding coherent acoustic phonon spectra obtained by a fast Fourier transform (FFT).](materials-13-03111-g008){#materials-13-03111-f008} {#materials-13-03111-f009}

Introduction {#Sec1} ============ Rhinosinusitis (RS) is a condition that arises from inflammation of the nose and the paranasal sinuses resulting in two or more symptoms including nasal blockage, obstruction, congestion, and nasal discharge \[[@CR1]••, [@CR2]••\]. Additional symptoms may include facial pain and pressure, reduction or loss of smell. During diagnosis, endoscopic signs of polyps, mucopurulent discharge, and mucosal obstruction in the middle meatus and computerized tomography (CT) changes within the ostiomeatal complex may be observed \[[@CR2]••\]. RS can be classified into two major types based on the duration and nature of the disease. Acute rhinosinusitis (ARS) involves the presentation of two or more symptoms for the duration of less than 12 weeks, with the damage to the nasal passage being often reversible. On the other hand, chronic rhinosinusitis (CRS) involves remodeling of the nasal passage due to long-standing inflammation causing persistent symptoms for more than 12 weeks \[[@CR2]••, [@CR3]\]. Regardless of duration, both ARS and CRS can be caused by infection by pathogens or allergies against environmental allergens. However, in most cases, the key trigger of RS is often a viral infection that either initiates or exacerbates the symptoms \[[@CR1]••, [@CR2]••, [@CR4]••\]. Epidemiology of ARS and CRS {#Sec2} =========================== RS is a very common condition encountered in primary care, accounting for about 15% of the population in Western countries \[[@CR4]••\]. The prevalence of ARS is relatively high with an estimated prevalence rate of 6--15% worldwide according to various epidemiological studies \[[@CR2]••, [@CR5]\]. On the other hand, studies on the epidemiology and prevalence of CRS are less comprehensive compared to ARS. This is due to the fact that CRS encompasses many forms of presentation which confound the study parameters when accounting for their prevalence. A prevalence of 5--15% prevalence is estimated both in the USA and Europe based on compilation of available studies \[[@CR2]••, [@CR6]--[@CR9]\]. Pathophysiology and Mechanism of Action of Viral Infection in ARS and CRS {#Sec3} ========================================================================= The nature of the disease of RS is highly dependent on the predisposing conditions such as allergic rhinitis, nasal deformity, immune deficiency, and other environmental factors. These underlying conditions can be exacerbated following primary damage caused by viral infections and the secondary damage caused by the host response against the virus, resulting in the pathology of RS and its symptoms. In ARS, the primary and secondary damage does not result in permanent changes to the nasal passage, and the ARS symptoms arising from the acute infection typically subside within 12 weeks. On the other hand, when the host responses triggered by the viral infection cause permanent alteration of the anatomy of the nasal passage such as remodeling of the epithelial layer, long-term effects may present in the form of CRS \[[@CR2]••, [@CR3], [@CR10]\]. As the main cause of ARS is viral, the host immune response that leads to the pathophysiology of RS is mainly antiviral in nature. The antiviral immune response involves nonspecific and specific components that require the coordination between different cell types including neutrophils, macrophages, eosinophils, dendritic cells, epithelial cells, mast cells, natural killer cells, and lymphocytes \[[@CR11]••, [@CR12]•\]. An ideal scenario is when the elicited immune response is timely and culminates in early viral elimination and minimal damage to the host. However, the cascade of inflammation initiated by the epithelial cells normally lead to damage by the infiltrating cells, causing edema, engorgement, fluid extravasation, mucus production, and sinus obstruction in the process, eventually leading to ARS or exacerbating ARS \[[@CR1]••, [@CR2]••\]. The exact roles and mechanisms of respiratory viruses in ARS exacerbation are still under debate. It is often speculated that the T-helper 1 (Th1) response is initiated from the epithelial innate immune response via toll-like receptors 3, 7, and 9 (TLR 3, TLR7, and TLR9) due to the virus infection \[[@CR13], [@CR14]••\]. Depending on the type of viral pathogen, the pathogen-sensing molecules in turn activate the production and secretion of nuclear factor-κB (NF-κB), interferon-β (IFN-β), tumor necrosis factor-α (TNF-α), and interleukins-1β, 6, and 8 (IL-1 β, IL-6, and IL-8), which are potent inducers or recruiters of neutrophils and macrophages \[[@CR4]••, [@CR12]•\]. The initial action of neutrophils against virus-infected cells usually contributes to the early symptoms of an acute respiratory virus infection. Following this, the further secretion of TNF-α and interferon-γ (IFN-γ) increases the recruitment of Th1 cells and cytotoxic T-cells which leads to the clearance of the viral pathogens and viral-infected cells. The process of clearing the virus generates dead epithelial and infiltrating cells which contribute to the pathology of ARS. It also creates an environment suitable for secondary bacterial infections (such as *Staphylococcus aureus* and *Streptococcus pneumoniae*), which represent another factor that exacerbates ARS symptoms initiated by a viral infection \[[@CR11]••, [@CR12]•, [@CR15]\]. On the other hand, CRS pathology is not as straightforward as ARS as the former is associated with multiple factors and disease presentations. However, virus infection is known to contribute to and exacerbate CRS. People suffering from CRS often have predisposing factors that contribute to the inflammatory condition in the nasal passage. One factor is the remodeling of the nasal epithelium due to the extended period of inflammation. During viral infection, the similar inflammatory process causes exacerbation as with ARS. In addition, the destruction of the epithelial layer and macrophage recruitment also induce the production of matrix metalloproteinase-9 (MMP-9), a key remodeling protein that mediates repair of the epithelial and extracellular matrix \[[@CR16]\]. The remodeling of the nasal passage in turn exacerbates the symptoms of CRS even further, as normal epithelial function is further diminished due to possible hyperplasia and replacement of epithelial cells with fibroblasts following respiratory viral infection. Key Viruses Causing the Pathophysiology of ARS and CRS Exacerbation {#Sec4} =================================================================== Viruses account for at least 80 to 90% of the ARS occurrence, with rhinovirus (RV), respiratory syncytial virus (RSV), influenza virus, coronavirus (CorV), parainfluenza virus (PIV), adenovirus (AdeV), and enterovirus (EV) playing a major role in ARS exacerbation \[[@CR2]••, [@CR4]••, [@CR9]\]. These viruses induce strong Th1 responses which lead to the pathology of ARS exacerbation. In addition, RV and PIV may contribute further to the exacerbation as their infection upregulates ICAM-1, viral receptor for major type of RV \[[@CR17]--[@CR19]\] (Tan et al., unpublished observation). ICAM-1 is also a signaling protein that activates the migration and infiltration of immune cells to contribute further to the exacerbation and pathophysiology of ARS \[[@CR20]\]. RV and CorV are the most common viruses isolated from adult ARS, accounting for approximately 50% of ARS diagnosis. Geographically, there are also other viruses isolated from patients with ARS, e.g., human bocavirus is frequently isolated from ARS cases in Taiwan \[[@CR21]\]. On the other hand, most of the viruses that exacerbate ARS will also cause similar exacerbation in CRS patients. There is a high rate of viral detection from the nasal wash of CRS patients including viruses commonly found in ARS patients, with RV also being the most prevalent virus in CRS exacerbation \[[@CR22]\]. Human bocavirus and metapneumovirus (hMPV) are also found in the nasal washes of CRS patients during flare ups. Moreover, there are also certain viruses (herpes virus, Epstein-Barr virus, and human cytomegalovirus) found in CRS patients with nasal polyps, and it is currently unclear whether they contribute to exacerbation of CRS \[[@CR23]\]. Due to the multifactorial nature of CRS exacerbation, it is relatively harder to identify key viruses that exacerbate CRS compared to ARS, and thus, a direct association between CRS and viral exacerbation is still unclear. The putative underlying mechanism of viral ARS and CRS exacerbation is summarized in Fig. [1](#Fig1){ref-type="fig"} based on existing literature \[[@CR1]••, [@CR2]••, [@CR11]••, [@CR12]•, [@CR14]••\].Fig. 1Putative underlying mechanisms of viral-induced ARS and CRS exacerbation. The figure shows the chain of events leading to ARS and CRS exacerbation from common respiratory virus infections (rhinovirus and influenza virus). When a respiratory virus infects its respective host cells in the nasal epithelium, the intracellular pathogen sensors TLRs 3, 7, and 9 are activated to initiate an antiviral cellular response via the activation of STAT1/2 and NF-κB transcription factors. These activated transcription factors then induce the expression of interferons (IFNs), interferon-stimulated genes (ISGs), chemokines, and cytokines against the virus. The cascade of intracellular events leads to the recruitment of innate responders such as neutrophils, macrophages, and dendritic cells, and in turn activates Th1 adaptive responses against the invading respiratory pathogen. This cascade of inflammatory events against the virus culminates in the symptoms of ARS and its exacerbation, especially if persistent infection occurs due to insufficient viral clearance. Furthermore, the priming of the nasal epithelium against viral infection potentiates its environment to be suitable for secondary bacterial infection which may further exacerbate the symptoms. The infection, if kept unchecked, may result in continued inflammation and expression of remodeling genes that may transition into chronic symptoms of CRS, which increases the susceptibility against further viral infection, causing further exacerbation of symptoms. The putative underlying mechanisms are summarized based on existing literature \[[@CR1]••, [@CR4]••, [@CR11]••, [@CR12]•, [@CR14]••\]. *ARS* acute rhinosinusitis, *CD* cluster of differentiation, *CRS* chronic rhinosinusitis, *DCs* dendritic cells, *IFN* interferon, *IL* interleukin, *ISGs* interferon stimulated genes, *STAT* signaling transducers and activators of transcription, *TLR* toll-like receptor, *MMP* matrix metalloproteinase, *NF-κB* nuclear factor kappa-light-chain-enhancer of activated B cells, *TGF* transforming growth factor, *Th1* T-helper 1 Diagnosis of Viral Infections that Lead to ARS and CRS Exacerbation {#Sec5} =================================================================== As the cause of RS (especially ARS) is almost exclusively viral in nature, the diagnosis of specific viral infection is important to prevent the unnecessary prescription of antibiotics \[[@CR1]••\]. In clinical practice, most viral infections are usually diagnosed by symptomatic assessment, which lack objective diagnostic tools for confirmation as well as identification of the type of viral infection \[[@CR1]••, [@CR2]••, [@CR9], [@CR15]\]. However, with the development of rapid viral assays and diagnostic kits, viral detection and identification can now be performed within a point-of-care setting for sensitive and specific identification of respiratory virus infections in patients presenting with ARS or CRS exacerbation \[[@CR24]••\]. The current viral detection systems, their features, and use in clinical settings are summarized in Table [1](#Tab1){ref-type="table"}.Table 1Current viral detection systems, their features, and use in clinical settingsDetection methodPrinciple and detectable virusesAdvantagesDisadvantagesCurrent useRapid virus detection by cultureDetecting viral infection in host cells: AdeV influenza A and B PIV 1--4 hMPV RSVWide array of viruses detectedTime-consuming (48-h detection time), host cell specificityShell vial, cluster trays, and similar assays in laboratory attached to hospitalDirect fluorescence antibody testDetecting viral antigens via fluorescence antibody: AdeV influenza A and B PIV 1--3 hMPV RSVSensitive detection of virus in moderate duration (30--60-min detection time)Requires specific training and laboratory spaceImmuno-fluorescence assays conducted in laboratory attached to hospitalRapid antigen direct testDetecting viral antigens via immuno-chromatography: influenza A and B RSVRapid, the only bench-to-bedside clinical diagnosis tool (15--20-min detection time)Prone to false positives and negativesHighly variable sensitivity and specificityBinaxNOW™, Directigen™ Xpect™, and similar products. Rapid antigen test kitsMolecular detectionDetecting specific viral nucleic acid sequences: all respiratory viruses with known sequencesSensitive and specific and allows wide array of virus detection (20--80-min detection time)Requires virus sequence to be known and requires specific training and laboratory spaceMolecular assays (PCR, microarray based) conducted in laboratory attached to hospital*AdeV* Adenovirus, *hMPV* human metapneumovirus, *PIV* parainfluenza virus, *RSV* respiratory syncytial virus Current assays for rapid detection of respiratory viruses include virus detection, antigen detection, and molecular detection. Rapid virus detection culture such as shell vial or cluster trays facilitate detection of adenovirus, influenza A and B virus, PIV 1--4, hMPV, and RSV within 48 h \[[@CR25]\]. Assays for virus antigen detection involve the use of viral antigen-specific antibodies using direct fluorescence antibody test and rapid antigen test. The direct fluorescence antibody test can detect AdeV, influenza A and B virus, hMPV, PIV, and RSV within 30 to 60 min \[[@CR26]\], while the rapid direct antigen test utilizes immuno-chromatographic antibody detection kits that are simple to perform (detection in 15--20 min), but so far limited only for detecting influenza A and B virus and RSV \[[@CR27], [@CR28]\]. Lastly, molecular detection includes amplification of viral nucleic acid sequences through polymerase chain reaction of patient samples. This assay can detect respiratory viruses with high sensitivity and specificity within a short time frame of 20 to 80 min \[[@CR29]\] and are recognized to play important roles in respiratory viral diagnosis due to their versatility \[[@CR24]••, [@CR30], [@CR31]\]. In addition to the above methods, there is also serological detection such as ELISA which detects antibodies against the virus. However, while serological detection was widely used in the current diagnostic settings, it is not suitable for the diagnostics of acute infections that caused ARS and CRS exacerbations, and therefore not discussed in this review. Currently, rapid culture, direct fluorescence antibody, and molecular detection necessitate specialized laboratories attached to larger hospitals to perform the diagnosis and identification of viruses causing ARS and CRS exacerbation, due to the complexity and training to operate such assays. The point-of-care detection method is the rapid direct antigen test through immuno-chromatography, but this test suffers from sensitivity issues and may lead to false negatives and misdiagnosis \[[@CR32], [@CR33]\]. Therefore, it is imperative to continue to improve diagnostic kits to rapidly and accurately pinpoint the causative agents of ARS or CRS exacerbation to prevent inappropriate prescriptions for treatment of exacerbation episodes, especially the misuse of antibiotics for viral-induced exacerbation of ARS and CRS. Prevention and Treatment of Viral Exacerbation of ARS and CRS {#Sec6} ============================================================= The clinical management of ARS and CRS exacerbation due to virus infection should be similar for any episodes of ARS and CRS exacerbation. Treatment usually aims to resolve and relieve the symptoms while the infection resolves itself. Symptomatic treatment involves the use of nasal decongestants and topical corticosteroids to relieve nasal congestion and to reduce nasal edema \[[@CR34]\]. Nonsteroidal anti-inflammatory drugs (NSAIDs) may also be used to decrease inflammation arising from viral infection to relieve symptoms such as fever and nasal edema \[[@CR2]••\]. When secondary bacterial infection is suspected, narrow-spectrum antibiotics such as amoxicillin can be prescribed. In addition to symptomatic management and relief, treatment and prevention of ARS and CRS exacerbation can be further enhanced with advances in viral detection to identify the exact viral etiology. By using antiviral treatments, viral load in the nasal epithelium can be reduced and therefore further lowering the inflammation associated with the infection, reducing RS symptoms and its duration. Hence, it is possible to add currently available antivirals to the repertoire of preventive and therapeutic strategies to counter viral exacerbation episodes of ARS and CRS. From the perspective of exploiting antiviral agents to prevent or treat ARS and CRS exacerbation, influenza virus infection has the widest array of therapies. While the adamantane derivatives amantadine and rimantadine that blocks influenza fusion is currently not recommended as a treatment due to widespread resistance \[[@CR35], [@CR36]\], there are still multiple neuraminidase inhibitors specific against influenza viruses that are available as therapeutic agents for severe episodes of ARS and CRS exacerbation. For example, oseltamivir, laninamivir, peramivir, and zanamivir specifically inhibit the replication of influenza viruses, and thus prevent or minimize the effects of ARS and CRS exacerbation \[[@CR37]•, [@CR38]••\]. Such treatments may be harnessed as useful prophylaxis during influenza season for patients with severe episodes of ARS or CRS exacerbation. In addition to neuraminidase inhibitors, annual influenza vaccination can also serve as a means of preventing ARS and CRS exacerbation due to influenza virus infection. On the other hand, other common respiratory infections due to RV, PIV, RSV, CorV, and AdeV do not have suitable vaccines and have limited antiviral agents that can be used against them. Palivizumab is a monoclonal antibody targeting the fusion of RSV and is used as a prophylaxis for infants at high risk of morbidity or mortality from RSV complications \[[@CR39]\]. There are also broad-spectrum antivirals such as ribavirin that prevent viral RNA synthesis that can be used against influenza virus, RSV, PIV, and to a certain extent, coronavirus, but are only reserved for severe cases due to their undesirable adverse effects \[[@CR40], [@CR41]\]. Similarly, cidofovir, a DNA synthesis inhibitor is used against disseminated adenovirus infection as a last line of defense against severe adenovirus disease due to concerns on its toxicity. To date, noninfluenza antivirals against respiratory viruses are still in the early stages of development and may not yet be optimal for the prevention and management of viral ARS and CRS exacerbation \[[@CR37]•\]. Notwithstanding this, there are also investigational agents in clinical trial phases and are close to being available clinically. These treatments may be useful in the near future to treat respiratory viral infections to manage viral ARS and CRS exacerbation. Favipiravir (T-705) has so far been tested to be efficacious against influenza virus and is being evaluated for other RNA viruses and was recently approved in 2014 to be marketed in Japan for severe influenza \[[@CR42], [@CR43]\]. Fludase (DAS181), a fusion protein that prevents influenza infection, is currently in phase II clinical trials \[[@CR44], [@CR45]\]. Presatovir (GS-5806), ALS-8176, and ALN-RSV01 are all inhibitors of RSV replication that are undergoing trials as potential treatment against RSV \[[@CR46]--[@CR49]\]. If found to be efficacious with minimal side effects, these investigational agents can be exploited in the future as effective measures in managing viral exacerbation of ARS and CRS \[[@CR38]••\]. Detailed information on the current antiviral drugs against respiratory viruses and their mechanisms of action is discussed in Fig. [2](#Fig2){ref-type="fig"} and Table [2](#Tab2){ref-type="table"}.Fig. 2Current antiviral drugs against respiratory viruses and their mechanisms of action. The figure shows a representative respiratory virus infection (influenza) in the host cell. Other than vaccines, which do not act directly on viral replication in the host cell, other antivirals target specific virus components required for the replication cycle: (*1*) viral entry receptor---fludase; (*2*) viral fusion---presatovir, palivizumab; (*3*) preventing transcription/replication---ribavirin, taribavirin, cidofovir, favipiravir, ALS-8176; (*4*) preventing translation---ALN-RSV01; (*5*) preventing viral excision---oseltamivir, laninamivir, peramivir, zanamivir. The antivirals and their mechanisms are summarized based on existing literature \[[@CR37]•, [@CR38]••, [@CR39]--[@CR41]\]. *mRNA* messenger RNA, *vRNP* viral ribonucleoprotein, *siRNA* small interfering RNA, *ALS-8176* deoxy-3′,5′-di-O-isobutyryl-2′-fluorocytidine, *ALN-RSV01* asvasiran sodium Table 2Current and experimental antiviral agents against respiratory virusesManagementMarket nameRespiratory virus pathogen efficacious againstMode of actionCurrent FDA-approved usage OR at research development stageVaccination Annual influenza vaccination*Multiple*Influenza virusBoosting immune response for influenza-specific antibody productionPreventive prophylaxis during influenza seasonFusion inhibitors Amantadine*Symmetrel*®Influenza virusPrevents viral fusion and release of nucleic acidInfluenza virus infection (discontinued) Rimantadine*Flumadine*® Palivizumab*Synagis*®Respiratory syncytial virusRespiratory syncytial virus infectionNeuraminidase inhibitors Oseltamivir*Tamiflu*®Influenza virusInhibits influenza neuraminidase to prevent viral release and replicationInfluenza virus infection Laninamivir*Inavir*® (Japan) Peramivir*Rapiacta*®*PeramifluRapivap*® Zanamivir*Relenza*®Guanosine analogue Ribavirin*Copegus*®Influenza virusRespiratory syncytial virusParainfluenza virusCoronavirusBlocks viral RNA synthesisHepatitis C infection*Pegasys*® Taribavirin*Viramidine*®DNA/RNA polymerase inhibitor Cidofovir*Forvade*® *Vistide*®AdenovirusBlocks viral DNA polymeraseCytomegalovirus retinitis Favipiravir (T-705)*Avigan*® (Japan)Influenza virusInhibits viral polymeraseInfluenza virus infectionExperimental drugs Experimental drug Fludase (DAS181)NilInfluenza virusParainfluenza virusCleaves cell receptor required for viral entryPhase II clinical trial Experimental drug Presatovir (GS-5806)NilRespiratory syncytial virusInhibits fusion of virus in host cellsPhase II clinical trial Experimental drug ALS-8176NilRespiratory syncytial virusRSV polymerase inhibitorPhase II clinical trial Experimental drug ALN-RSV01NilRespiratory syncytial virussiRNA targeting RSV transcriptsPhase II clinical trial*ALS-8176* deoxy-3′,5′-di-O-isobutyryl-2′-fluorocytidine, *ALN-RSV01* asvasiran sodium Conclusion {#Sec7} ========== ARS and CRS remain a considerable economic and public health burden worldwide due to their high prevalence. As virus infections are now established as the key cause of most cases of RS and exacerbation (especially ARS), more attention should be focused on the proper study, diagnosis, prevention, management, and treatment of viral-induced exacerbation of ARS and CRS symptoms. A more detailed perspective of the virology of ARS and CRS exacerbation can enhance our current understanding of RS to improve the quality of treatment of episodes of ARS and CRS exacerbation, as well as to minimize and prevent the misuse of antibiotics in the treatment of ARS and CRS exacerbation, which may contribute to the growing spectrum of antibiotic resistance. This article is part of the Topical Collection on *Rhinosinusitis* Kai Sen Tan and Yan Yan contributed equally to this work. Kai Sen Tan, Yan Yan, Hsiao Hui Ong, Vincent T. K. Chow, Li Shi, and De-Yun Wang declare that they have no conflicts of interest relevant to this manuscript. This article does not contain any studies with human or animal subjects performed by any of the authors. This study was supported by grants from the National Medical Research Council (NMRC/CIRG/1362/2013 and NMRC/CIRG/1458/2016).

Introduction ============ The emergence of stem cells as a therapeutic for many diseases, disorders and injuries has brought excitement among scientists, clinicians and patients alike regarding the potential treatment of previously untreatable conditions. However, the implementation of many stem cell therapies in patients may still be years away. When considering translating these therapies into patients, there are two principal concerns that must be resolved: I. Can the stem cells efficiently produce the desired therapeutic outcome, albeit tissue replacement or repair, in vitro?; and II. Can the in vitro studies be replicated in vivo, both short- and long-term, with increased confidence? Much of the past research has concentrated on question one, or more appropriately, the philosophy of can we apply the method? However, to recognize stem cells as key factors in the treatment of various ailments, we need to rest assured that we can also answer question two -- Is this a viable treatment approach? These questions are aside from the ethical implications surrounding the field, which ask should we do it. In any case, stem cells will continue to be researched as a potential treatment for a multitude of diseases and disorders. Considerable progress has been made in addressing the first question stated above -- can we do it. A vast number of tissue types have been generated from both embryonic (ES cells) and adult stem cells. ES cells are pluripotent cells derived from the inner cell mass of the blastocyst, which hold tremendous potential in generating specified tissue types ([@b25-btt-2-699]). However, the potential for immune rejection, together with the possibility of tumor formation has caused their application in humans to proceed with caution ([@b25-btt-2-699]). Adult stem cells tend to be tissue-specific cells with limited differentiation potential compared with ES cells. Adult stem cells are clinically attractive therapies due to their reduced risk of tumorigenesis and ability to expand with relative ease ([@b8-btt-2-699]). Among the many types of adult stem cells, those resident to the bone marrow (BM), particularly mesenchymal stem cells (MSCs), have gained extensive interest among scientists and clinicians ([@b12-btt-2-699]). MSCs are mesodermal cells primarily resident to the adult BM, which undergo lineage-specific differentiation to generate bone, fat, and cartilage among other tissue types ([@b3-btt-2-699]). MSCs have also been reported to transdifferentiate into defined ectodermal and endodermal tissues in vitro, thus alluding to their inherent plasticity (Choi and Panayi 2001; [@b9-btt-2-699]; [@b14-btt-2-699]; [@b27-btt-2-699]; [@b19-btt-2-699]; [@b21-btt-2-699]; [@b23-btt-2-699]; [@b34-btt-2-699]). MSCs are available for autologous therapies, have a unique ability to bypass immune rejection and are inherently migratory ([@b31-btt-2-699]). These properties of MSCs make them particularly well suited when considering the second question posed earlier -- can in vitro findings be accurately recapitulated in vivo? Whereas tissues derived from ES cells or other types of stem cells may be rejected when transplanted, MSCs offer the potential for allogeneic transplantation and a readily available source of "off-the-shelf" stem cells for personalized therapies. However, the unique immune properties of MSCs do not guarantee that the cells will produce the desired therapeutic outcome or even that they will not be rejected. In vitro, a MSC's growth conditions can be closely monitored to favor stem cell growth and/or differentiation. In vivo, the transplanted MSCs are exposed to local immune cells and soluble mediators that could influence the cells' behavior, either positively or negatively regarding the desired outcome. This concept of the tissue microenvironment has become a growing concern among researchers, and may be the ultimate factor in deciding whether a stem cell therapy succeeds or fails ([@b18-btt-2-699]; [@b37-btt-2-699]; [@b17-btt-2-699]; [@b29-btt-2-699]). A prototypical example of a tissue microenvironment affecting stem cell behavior is observed among hematopoietic stem cells (HSCs) and their niche within the BM. HSCs are relatively quiescent cells located close to the BM endosteum at relatively low oxygen concentration ([@b18-btt-2-699]). As HSCs differentiate, the maturing immune cells migrate towards the central sinus of the BM under progressively higher oxygen concentrations ([@b18-btt-2-699]). The change in oxygen is a key determinant in the maturation of the immune cells before they leave the BM and migrate into the peripheral circulation ([@b18-btt-2-699]). In contrast, MSCs are located close to trabecular bone near the central sinus of the BM ([@b3-btt-2-699]). As MSCs migrate towards the endosteum under progressively lower concentrations of oxygen, the stem cells differentiate into stromal fibroblasts, which form the principal support structure for immune cell maturation ([@b3-btt-2-699]). This example demonstrates that local microenvironmental changes in variables such as oxygen concentration can drastically affect the behavior of MSCs. Since MSCs have been shown to generate a vast number of tissues, they have clinical implications in a wide array of diseases and disorders. Among possible transplantation sites are tissues such as cardiac, neural, pancreatic and bone. Each tissue provides a unique local microenvironment that can affect the success of the therapy. The problem facing researchers is accurately developing in vitro models to recapitulate the tissue microenvironment so that cellular behavior can be observed prior to transplantation. This is no easy task considering the dynamic nature of the microenvironment. Transplantation of MSCs alone will generate a local immune response and disrupt homeostasis within the tissue milieu by causing release of inflammatory mediators, such as cytokines. The anatomy of the BM is such that MSCs are in direct interaction with immune cells and form synapse-like structures with innervating nerve fibers ([@b3-btt-2-699]). MSCs express receptors for many cytokines and neurotransmitters, thus demonstrating their potential to respond to local microenvironmental changes ([@b20-btt-2-699]). Excessive cytokine release within the transplantation site could lead to the production of other soluble factors by the MSCs themselves. If these factors are immunoreactive, then other immune cells could infiltrate the tissue and cause an exacerbated immune response, rejection of the transplant or differentiation of the MSCs ([Figure 1A](#f1-btt-2-699){ref-type="fig"}). On the other hand, MSCs have been shown to be a potent source of trophic factors ([@b29-btt-2-699]). These findings indicate that MSCs could also be used to aid normal tissue repair, perhaps even more so than in cell replacement. Whether transplanted MSCs cause an immune insult or help repair injured tissues may be difficult to determine unless appropriate models are developed to better predict the outcome. However, if MSCs are found to negatively impact the host microenvironment through exposure to soluble mediators, there are still potential methods to develop effective therapeutics. When considering the example presented in [Figure 1A](#f1-btt-2-699){ref-type="fig"}, inclusion of specific cytokine receptor antagonists or inhibitors could suppress the untoward effects of the host microenvironment on the transplanted MSCs, thus leading to defined therapeutic outcomes ([Figure 1B](#f1-btt-2-699){ref-type="fig"}). Throughout the remainder of this review, we will address the feasibility of using similar pharmacologic approaches in MSC transplants, while focusing on three ubiquitous microenvironmental factors: IL-1α/β, TNFα, and SDF-1α. Specifically, we will examine how receptor antagonists or inhibitors to these factors, whether federally approved or in development, may limit the potential negative influences of the tissue microenvironment. Interleukin-1α/β ================ IL-1α and IL-1β are members of the IL-1 superfamily of cytokines. These pro-inflammatory mediators are primarily synthesized by macrophages, monocytes and dendritic cells, and are responsible for immune defense against infection ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b13-btt-2-699]). IL-1α and IL-1β are also key regulators of hematopoesis and the inflammatory process ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b13-btt-2-699]). Both cytokines are found throughout the body, thus they are expected to be present within the local microenvironment of most tissues. Our laboratory has previously demonstrated that MSCs express IL-1RI, which is the principal receptor for both IL-1α and IL-1β ([@b20-btt-2-699]). Interestingly, membrane expression of the receptor was maintained throughout the course of transdifferentiation of MSCs into functional neurons ([@b20-btt-2-699]). If IL-1α or IL-1β were found to have negative effects on MSCs, then these effects may also be seen on transplantable tissues differentiated from MSCs. These results have implications regarding the ideal stage of stem cell implantation, whether undifferentiated, partly differentiated or fully differentiated. IL-1α was found to alter the behavior of undifferentiated MSCs and neurons differentiated from MSCs ([@b20-btt-2-699]). Specifically, stimulation of MSCs with IL-1α caused production of the neurotransmitter, substance P (SP), by undifferentiated and differentiated cells ([@b20-btt-2-699]). Similar effects were not observed in cells stimulated with IL-1β ([@b20-btt-2-699]). SP has involvement in various physiological functions, such as the perception of pain and breast cancer progression, however the peptide also has a stimulatory effect on immune cell development and function ([@b18-btt-2-699]). SP was also found to stabilize IL-1RI mRNA, thus potentially forming an autocrine feedback loop whereby IL-1α present in the microenvironment continually stimulates production of SP by the MSCs or their differentiated progeny ([@b20-btt-2-699]). The excessive levels of SP could lead to immune cell infiltration and an exacerbated immune response, which may cause rejection of the transplant. Interestingly, IL-1α also had a more global effect on the behavior of MSCs, specifically their ability to transdifferentiate into neurons ([@b20-btt-2-699]). MSCs that were grown in neuronal induction media containing IL-1α showed greater expression of genes linked to neurogenesis compared to cells induced without IL-1α ([@b20-btt-2-699]). In comparison to the deleterious effects of IL-1α mentioned above, these results demonstrate what appears to be a positive effect on MSC differentiation. If the desired therapeutic value of MSCs is transdifferentiation, in this case into neurons, then this finding can be considered beneficial. However, in many cases, the maintenance of MSCs as stem cells will be desired. Hence, the influence of the microenvironment on premature MSC differentiation would be unwarranted. To counter the negative effects of IL-1α present within a tissue microenvironment, co-therapies utilizing MSCs and specific IL-1R antagonists or inhibitors may be successful. The IL-1R antagonist (IL-1ra) is naturally occurring and binds to the IL-1RI. IL-1ra competes for binding to the IL-1RI with IL-1α and IL-1β, however binding of this ligand does not result in an intracellular signal ([@b13-btt-2-699]). A commercially available IL-1RI antagonist is Kineret^®^, also known as Anakinra^®^, which is a recombinant, non-glycosolated version of human IL-1ra ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b22-btt-2-699]; [@b2-btt-2-699]; [@b5-btt-2-699]; [@b15-btt-2-699]). The drug has been used in the treatment of inflammatory conditions such as rheumatoid arthritis. Recent studies in mice have shown that MSCs have an inherent ability to counteract the deleterious inflammatory effects of IL-1α in injured tissues ([@b28-btt-2-699]). In response to bleomycin-induced inflammation and fibrosis within the lungs of mice, transplanted MSCs were shown to synthesize IL-1ra to neutralize microenvironmental IL-1α. In addition, MSCs also protected the tissue from further damage by inhibiting the production of TNFα within the lung. If, however, exogenous IL-1ra supplementation is also necessary, it may be possible to administer an IL-1ra, such as Anakinra^®^, during MSC transplantation. However, the means of accurately delivering the drug to its target is unknown. Perhaps the drug could be bound to the MSCs in such a way that the stem cells actually "piggy-back" the drug to the target. A great deal of research is still necessary to develop these types of therapeutics. Tumor necrosis factor α ======================= TNFα is a pro-inflammatory cytokine principally synthesized by macrophages, which is involved in the acute phase of systemic inflammation ([@b26-btt-2-699]). More specifically, TNFα mediates immune cell homing, proliferation and differentiation, as well as tumorigenesis and viral replication ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b26-btt-2-699]). TNFα is ubiquitously found throughout the body, and is another important factor present within local tissue microenvironments. We have previously demonstrated that MSCs express the principal receptor for TNFα, TNF-R1 ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b20-btt-2-699]). Similar to IL-1RI expression, TNF-RI levels were maintained during the entire course of MSC transdifferentiation into neurons ([@b20-btt-2-699]). However, whereas IL-1α was able to induce SP production in undifferentiated, partly differentiated and fully differentiated cells, TNFα had similar effects only in partly and fully differentiated cells ([@b20-btt-2-699]). Additionally, TNFα did not have the same enhancing effect on MSC transdifferentiation as IL-1α. These results re-emphasize that the ideal stage of differentiation for efficient transplantation is unknown. In the case of exposure to microenvironmental TNFα, little to no negative effects may be observed with undifferentiated MSCs, while an exacerbated immune response may be seen if transplanting partly or fully transdifferentiated cells. Recent studies have shown that undifferentiated MSCs incubated with TNFα have a greater ability to migrate in the presence of chemokines compared to cells incubated without TNFα ([@b30-btt-2-699]; [@b33-btt-2-699]). Increased chemotaxis of MSCs would be clinically important if proper homing to the site of tissue injury became more efficient. However, since MSCs are inherently chemotactic, increased sensitivity to chemokine gradients could cause continuous mobilization within a tissue, and impede proper homing and delivery of the desired therapeutic. To offset any deleterious effects of microenvironmental TNFα on proper homing of MSCs to the site of tissue injury, co-thereapies with existing TNFα inhibitors may be beneficial. Currently there are three approved TNFα inhibitors, infliximab (Remicade^®^), adalimumab (Humira^®^) and etanercept (Enbrel^®^), which are primarily used to treat inflammatory and autoimmune disorders ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b24-btt-2-699]; [@b11-btt-2-699]). Infliximab and adalimumab are monoclonal antibodies that bind TNFα and block signaling through the TNF-RI. Etanercept is a large molecular weight, soluble recombinant TNFα receptor fusion protein that binds TNFα and prevents signaling through membrane-bound TNF-RI. Administration of these pharmacologics in combination with MSC therapies may negate any untoward effects of TNFα on the desired therapeutic outcome. Stromal cell-derived factor-1α ============================== SDF-1α, also known as CXCL12, is a chemokine produced by stromal fibroblasts, which mediates inflammation and the immune response through modulating lymphocyte chemotaxis ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b4-btt-2-699]). Additionally, SDF-1α regulates hematopoeisis and has a role in tumor metastasis ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b4-btt-2-699]). Like IL-1α/β and TNFα, SDF-1α is ubiquitously found throughout the body, and is an important soluble mediator of the tissue microenvironment. Within the BM, the primary lineage-specific progeny of MSCs are stromal fibroblasts ([@b12-btt-2-699]; [@b3-btt-2-699]). It is not surprising then that SDF-1α is important in the biology of MSCs. Expression of the principal SDF-1α receptor, CXCR4, has been demonstrated on MSCs, where it has been shown to mediate site-directed homing of MSCs in models of tissue engineering ([@b32-btt-2-699]). In the BM, SDF-1α is vital to the hematopoietic supportive function that MSCs exert to maintain proper hematopoeisis (Van Overstraeten et al 2006). As mentioned earlier, preconditioning MSCs with cytokines such as IL-1β and TNFα increased the migratory capacity of the cells ([@b30-btt-2-699]; [@b33-btt-2-699]). However, these enhancing effects were shown to be independent of SDF-1α, and were instead mediated by other chemokines ([@b30-btt-2-699]). In general, the clinical relevance of SDF-1α in the success of MSC therapies is positive, since SDF-1α gradients help MSCs home to sites of tissue injury ([@b4-btt-2-699]). In theory, MSCs could be administered systemically and allowed to respond to SDF-1α gradients for proper delivery to the target tissue. Once at the site of injury, MSCs themselves could serve as a source of SDF-1α ([@b38-btt-2-699]). Expression of SDF-1α by MSCs has been shown to promote survival of cardiac myocytes after myocardial infarction in rats ([@b38-btt-2-699]). However, SDF-1α is also a potent lymphocyte chemoattractant ([@b4-btt-2-699]). Excess production of SDF-1α within the microenvironment could potentially lead to increased immune cell infiltration and transplant rejection. SDF-1α may be less beneficial or even deleterious in therapies transplanting cells partly or fully differentiated from MSCs. For instance, SDF-1α has been shown to increase the proliferation of neural progenitor cells dissociated from rat cortex ([@b16-btt-2-699]). Comparing these results to the example of MSC neuronal transdifferentiation discussed throughout this review, excess SDF-1α might impede the final steps of neuronal maturation necessary for therapeutic improvement. To counteract any negative influences of local microenvironmental SDF-1α on MSC therapies, there are several available pharmacologics that inhibit the SDF-1α/CXCR4 interaction. Plerixafor, also known as Mozobil^®^ or AMD3100, is a partial antagonist of CXCR4, which has recently competed Phase 3 clinical trials, but is not yet in routine clinical use ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b6-btt-2-699]). Current in vivo studies have shown that pre-treatment of MSCs with AMD3100 significantly prevented stem cell migration to the injured rat brain ([@b36-btt-2-699]). Similar approaches may be beneficial in order to prevent non-specific MSC migration. T134, a small molecule CXCR4 inhibitor, and tannic acid, a water-soluble polyphenol widely distributed within the plant kingdom that acts as a selective CXCR4 antagonist, both inhibit the SDF-1α/CXCR4 interaction and may have future application as approved pharmacologics ([Table 1](#t1-btt-2-699){ref-type="table"}) ([@b1-btt-2-699]; [@b7-btt-2-699]). Administration of these compounds would most likely have to be locally delivered to the target tissue, since systemic inhibition of SDF-1α/CXCR4 could disrupt BM homeostasis or lead to excessive HSC mobilization into the peripheral circulation. Summary ======= In this review, we address the concept of the tissue microenvironment, and examine its clinical importance in MSC therapies. By accurately developing in vitro models that mimic the local tissue milieu, the behavior of MSCs or their differentiated progeny can be observed prior to in vivo application. Through like approaches, ill effects from soluble mediators or other cell types, which can impede the desired therapeutic outcome, can be assessed. In many cases, co-therapy with a pharmacologic such as a cytokine receptor antagonist may negate the deleterious effects of the microenvironment and optimize the therapeutic potential of MSCs. Here, we have focused on three ubiquitous microenvironmental factors with known effects on MSC function, and addressed how local delivery of inhibitors to these factors could improve the MSC therapy. {#f1-btt-2-699} ###### Microenvironmental factors implicated in the outcome of MSC therapies Cytokine/chemokine Source Physiological function Receptor expression on MSCs Inhibitors/antagonists Reference -------------------- ----------------------------------------- ------------------------------------------------------------------------------------------------- ----------------------------- ----------------------------------- ----------------------------------------------------------------------- IL-1α/β macrophages, monocytes, dendritic cells immune response, inflammation, hematopoiesis yes IL-1ra, Kineret^®^ [@b5-btt-2-699]; [@b15-btt-2-699]; [@b22-btt-2-699]; [@b22-btt-2-699] TNFα macrophages immune response, inflammation, proliferation, differentiation, tumorigenesis, viral replication yes Remicade^®^, Humira^®^, Enbrel^®^ [@b24-btt-2-699]; [@b11-btt-2-699] SDF-1α tissue-specific stromal cells immune response, inflammation, hematopoiesis, chemotaxis, tumor metastasis yes Mozobil^®^, T134, tannic acid [@b6-btt-2-699]; [@b1-btt-2-699]; [@b7-btt-2-699] **Notes:** Listed microenvironmental factors are ubiquitously expressed throughout the body and have known effects on mesenchymal stem cells MSCs. Co-therapy with pharmacologics, such as receptor antagonists or specific inhibitors, may improve the desired therapeutic outcome.

INTRODUCTION {#sec1-1} ============ Hospital staffs are the most important group of health care providers. As a result of their continuous hard work, they are frequently faced with physical and psychological health problems. If these problems linger on, disappointment and boredom will settle, which in turn will lower the quality of health care services. Health care quality is highly dependent on clinical and professional ethical principles, which are under the influence of religious beliefs. An influential factor helping preservation of psychological stability is to believe in and to rely on God; closeness to God and performing religious duties brings about peace. Spirituality is a human dimension that combines our existence with concepts such as human nature, sacred mental experience, tendency to know more, elevation toward Good, and finding a meaning to life. It has also been defined as a number of values, attitudes, and hopes related to the superior existence and thus guiding us throughout life.\[[@ref1]\] The relationship between the mental health and spirituality has received much attention recently. Research shows spirituality is highly influential in physical and mental health preservation. Many researchers have concluded that spirituality has an immense effect on mental health.\[[@ref2][@ref3]\] According to WHO definition, health is total physical, mental, and social existential well-being, and mental health is the ability to establish harmonious communication with others, modify surroundings, and resolve conflicts.\[[@ref4]\] Behaviors such as reliance on God, praying, and going on pilgrimage can promote solace and mental health through peace and developing hope and positive thinking. Religious beliefs increase resistance against calamities, and thus help preserve physical and mental health, prevent infliction of diseases, and finally promote hopefulness.\[[@ref5][@ref6][@ref7][@ref8][@ref9]\] According to the cognitive-emotional-religious theory, man cannot comprehend the meaning of life without religious beliefs. This theory discusses religious beliefs and God, existence, and man, and proposes that these beliefs can be applied to treat psychological problems.\[[@ref10]\] From a clinical viewpoint, lack of appraisal of religious spiritual dimensions sets obstacles in the way of (1) acknowledging significance of these dimensions for health and (2) learning about the mechanisms involved.\[[@ref11]\] Islam as an ideology presents an impeccable and health-preserving lifestyle; its orders cover a wide range of life aspects, including personal and social ethics, interpersonal relations, and physical and mental health.\[[@ref11]\] Some studies indicate a correlation between reliance on God and high self-esteem,\[[@ref12]\] lower levels of depression, more mature social behavior,\[[@ref13][@ref14]\] and better mental predisposition.\[[@ref15]\] Kendler *et al*.\[[@ref16]\] studied a number of spiritual/religious themes and identified those factors which had a significant inverse relationship with lack of mental health. Hill *et al*. found that true religious belief acts as a motivational drive and is closely related to physical and mental health.\[[@ref17]\] Koenig *et al*. and Smith *et al*. also showed that health is a causal result of the relationship between higher levels of spirituality and reliance on God and trust in God enables people to tolerate hardships of life much better since they believe in God and entrust their lives into his hands.\[[@ref18][@ref19][@ref20]\] Also, James states that the foundation of religions, that is based on spiritual relation with God, meaningfulness of the whole universe, and meaningfulness of life is the essence of spirituality.\[[@ref21]\] Modern analyses have reported that there is a significant positive relationship between religious duties and concerns and health and life time.\[[@ref22][@ref23]\] Koenig *et al*. also found that spiritual/religious beliefs are important predictors of youth depression.\[[@ref18]\] Bahrami and Tashak\[[@ref24]\] reported a significant positive relationship between religious orientation and mental health promotion and decrease in psychological disorders. The study done by Omran-Nasab\[[@ref25]\] also indicated that there was a significant correlation between religious beliefs and mental health. Cohen, Yoon, and Johnston in a study on 168 patients with various physical disorders showed that there is generally a positive relationship between positive spiritual tolerance and mental health and a negative relationship between negative spiritual tolerance and mental health, but there was no relationship between more personal religious duties such as praying and mental health.\[[@ref26]\] Cohen and Hall also studied 1,000 aged people and found that there was a relationship between some religious beliefs (fear of God, fear of death, and belief in life after death) and feeling of well-being and it is more significant in Protestants than in Catholics and Jews.\[[@ref27]\] Krause also suggested that the relationship between calamities of life and depression symptoms in aged people reduces when they believe God knows better when to respond to their prayers.\[[@ref28]\] Keyes and Reitzes\[[@ref29]\] emphasize the importance of the impact of religious identity on dignity and depression symptoms in retired workers. Mazidi and Ostovar\[[@ref30]\] concluded that both Islam and Christianity affect mental health of Iranian youth positively. The findings of Jang *et al*., Hills *et al*., and Doolittle *et al*.\[[@ref2][@ref3][@ref31]\] showed that spirituality is beneficial to physical and mental health. Some researcher found a close relationship between religiousness and elevation of soul and health.\[[@ref32][@ref33]\] In other words, positive tendencies such as feeling of happiness, joyousness, and hope for future are related to spirituality.\[[@ref23][@ref34][@ref35]\] The favorable effect of positive constructs such as optimism and hope for future on physical and mental health has been confirmed by various studies.\[[@ref36]\] Snyder\'s theory of hope made researcher devote a lot of research to the relationship between hope and health.\[[@ref37][@ref38]\] Snyder believes hope is not a passive drive emerging only in dark moments of life, but a cognitive process through which one activity pursues goals. He states that hope is (1) a goal-determining process through which people (2) build up strategies to achieve goals and (3) are motivated enough to apply those strategies. These three components are known as goals, pathway thinking, and agency thinking.\[[@ref39]\] Thus, those staffs who are hopeful develops enough motivation to initiate and perform hard tasks, and emphasizing their capabilities, move toward their grand goals. Snyder also believes there is a positive relationship between positive emotions and purposefulness in life.\[[@ref40]\] Other studies have also indicated the relationship between hope and positive emotions is a positive relationship, and it has a negative relation with depression, anxiety, and negative emotions in general.\[[@ref40]\] Similarly, Hezarjaribi *et al*.,\[[@ref41]\] believe that there is a maximum significant relationship between hope for future and joyousness, and with the increase of the feeling of joy, hope for future is also enhanced. Many researchers such as Bandloo *et al*. believe important factors, including personality traits, cognitive components, and religious attitudes pave the way for feeling of joyousness.\[[@ref30]\] They have also reported a direct relationship between genetic, personality, cognitive, and religious factors, and anxiety on one hand and joyousness and hope for future on the other.\[[@ref40]\] In other words, religious attitudes and practices definitely help people find a meaning to life. Once hospital staff finds a meaningful life, hope for God\'s support, receive social and religious support, and feel belonging to a holy and graceful existence, they can survive hardships and calamities of life and maintain their mental health successfully; these people can work more fruitfully and provide better services.\[[@ref3]\] Thus, considering how crucial the job is, and also regarding the lack of a research on spiritual variables in Iran, this study was designed to answer this research question: Is there a relationship between religious/spiritual dimensions of one\'s character and his/her mental health and his/her hope for future? MATERIALS AND METHODS {#sec1-2} ===================== This was a correlational study. The statistical population included all state hospital\'s staffs in Shiraz-who were selected randomly using the sample size formula for correlational studies and matched with Cohen\'s table of sample size.\[[@ref42]\] Out of the 250 selected participants, 212 questionnaires were returned (85.2%). Females (156 people) formed 73.6% and males (56 people) formed 26.4% of the sample size. In six hospitals in the study, 95 staff (44.8%) were single, 112 staff (52.8%) were married, and 5 (2.4%) did not report their marital status. Regarding the educational level, 1 (0.5%) had not finished high school, 12 (5.7%) had high school diplomas, 19 (0.9%) had associate degrees, 160 (75.5%) had bachelor degrees, and 13 (6.1%) abstained from reporting. Ethical considerations were observed by reassuring subjects that their personal information would be regarded strictly confidential and by offering them to feel free to or not to answer the questionnaires. Data collection was done through using three questionnaires: Mental health questionnaire: the 12-item Goldberg and William\'s mental health questionnaire was one of the tools used in this study. The 4-point Likert scale (0 = always to 3 = Never) was applied. The total score was 36. The highest score indicated the lowest level of mental health. Goldberg *et al*.,\[[@ref43]\] have reported that the reliability of their test is 0.78 and Pearson correlation coefficient for 3 factors, and the total score is 0.57 if *P* \< 0.000. Ebadi *et al*.,\[[@ref44]\] applying the tool on a research sample of 18-25-year old Iranians, reported an internal consistency of 0.81 for the tool. They also confirmed the validity of the tool using variance analysis and Known-groups method of validity. The cut point was calculated to be 3.5 with a sensitivity rate of 0.87% and specifity of 60%.Hope for future questionnaire: The questionnaire developed by Snyder *et al*.\[[@ref37][@ref38][@ref39][@ref40][@ref41][@ref42][@ref43][@ref44][@ref45]\] was applied. It includes 12 statements and is self-administered. Three statements measure agency thinking; four statements measure pathway thinking; and four statements are distractors. Thus, the questionnaire covers two subscales, namely pathway and motivation; to respond to each statement, subjects rate them from 1 (definitely false) to 4 (definitely true). It should be mentioned that the questionnaire has been reported by Golazari^\[2007\]^ to enjoy a validity rate of 0.89%. Moreover, the hope scale is highly correlated with other psychometric tools, for instance, its correlation with Scheier and Carver\'s test, which assesses optimism, is between 50% and 60%. Also the scores of the test are negatively correlated with depression inventory scores (−0.41 and −0.51). Content validity of the test has also been confirmed by clinical experts.\[[@ref46]\]Scale of dimensions of religiosity: this is a scale developed by Haber, Jacob and Spangler, which was localized for conducting research in Iran through translation and modification of the items by Islamic scholars. The scale includes 31 items focusing on five religions spiritual components: (a) Existential well-being (six items), (b) motivation, devotion, and coping (nine items), (c) spiritual transcendence (seven items), (d) religious support (six items), and (e) religious attitudes and practices (three items). This questionnaire was first translated to Farsi by Nadi and Sajjadian\[[@ref46]\] in 2010 and its face and content validity have been confirmed. Thirty subjects were asked to check vague items on the questionnaire. Afterwards, 10 psychometrics experts judged the extent of congruity of the items and the components. Upon their approval, the final version was administered to the subjects of the study Cronbach\'s alpha coefficient confirmed the reliability of the components of the questionnaire (*r* = 0.73, 0.76, 0.70, 0.71, 0.75, respectively). The Likert 7-point Scale included responses ranging from 1 = definitely wrong to 7 = definitely right. High scores represented strong religious belief. Maximum and minimum scores of the test were 217 and 31, respectively. The advantage of the scale is that has no bias towards any religion. The time allocated for responding to the questionnaire was 15-20 min. To analyze the data, Pearson correlation\'s test and hierarchic regression test were employed on the statistical package for social science version 18. FINDINGS {#sec1-3} ======== Findings shown in [Table 1](#T1){ref-type="table"} indicate that hope for future has the highest correlation with existential well-being (*r* = 0.275) and the lowest correlation with mental health (*r* = 0.171). Furthermore, it is significantly correlated with motivation, devotion, and coping (*r* = 0.263, *P* \< 0.01). ###### Means, SDs, and correlations  However, no significant correlation existed between hope for future and religious support, spiritual transcendence, and religious attitudes and practices. Furthermore, mental health appeared to have a high correlation with religious support (*r* = 0.427) and religious attitudes and practices (*r* = 0.179) but had the lowest with hope for future (*r* = 0.171). As it is seen in [Table 2](#T2){ref-type="table"}, from among study variables, the best predictive variables for hope for future in staff are (1) motivation devotion, and coping, (2) mental health, (3) existential well-being. Based on the results of multiple regression analysis, there was a significant relationship between hope for future and (1) mental health, (2) existential well-being, and (3) motivation, devotion, and coping, Based on the same results, motivation devotion, and coping by itself accounts for 5% of the variance. With mental health combined, this reaches 6.4%, and when the third variable, existential well-being, is included, it increases by 9%. But in phase four, with inclusion of motivation, devotion, and coping together with existential well-being variance is 9%. Since the observed F at the level of *P* ≤ 0.01 is significant, regression equation could be generalized to the whole research population. [Table 3](#T3){ref-type="table"} also depicts the results of standard and non-standard regressions in this regard. ###### Adjusted R square multiple correlation coefficients for religious/spiritual components and mental health in predicting rate of hope for future in hospitals staff  ###### Standard and non standardized coefficients regression coefficients for prediction of the staff\'s hope for future  The findings suggest that considering the β coefficient (from phase three) one-unit increase in mental health increases hope for future by 0.081 unit; one-unit increase in motivation, devotion, and coping increases hope for future by 0.153 unit; and one-unit increase in existential well-being increases it by 0.195 unit. The findings also show that with the omission of mental health variable, one unit increase in motivation, devotion, and coping increases hope for future by 0.159 units and also one unit increase in existential well-being increases hope for future by 0.219 units, which is a better predictor for staff\'s hope for future. DISCUSSION {#sec1-4} ========== The study investigated two questions: (1) Is there a positive significant relationship between religiosity components, mental health, and hope for future in hospital staff? and (2) which components of mental health and religiosity can predict the staff\'s hope for future? Findings showed hope for future is positively and significantly correlated with existential well-being and motivation, devotion, and coping components. These finding are congruent with those of some previous studies\[[@ref3][@ref23][@ref38]\] Given that existential well-being is a psychological factor related to self-identify and meaning of life, mental health can bring about satisfaction and purposefulness. Therefore, a positive, strong relationship between existential well-being and hope for future seems highly plausible. Moreover, hopefulness as one of the results of general health, as we read in the Quran, is the characteristic of true believers and healthy followers.\[[@ref47]\] Therefore, a positive, strong relationship between existential well-being and hope for future seems highly plausible. Moreover, hopefulness as one of the results of general health, as we read in the Quran, is the characteristic of true believers and healthy followers.\[[@ref47]\] Findings showed there was no relationship between spiritual transcendence, religious support, and religious attitudes and practices, which was not congruent with the findings of some previous studies.\[[@ref2][@ref3][@ref4][@ref5][@ref6][@ref7][@ref8][@ref9][@ref10][@ref11][@ref12][@ref13][@ref14][@ref15][@ref16][@ref17][@ref18][@ref19][@ref20][@ref21][@ref22][@ref23][@ref24][@ref25][@ref26][@ref27][@ref28][@ref29][@ref30][@ref31][@ref48]\] This could be explained in two ways. First, previous studies did not investigate the simultaneous effects of three variables; and second, they were conducted in different cultural environments where Islamic customs and rituals were not dominant norms and values of the population. On the other hand, the positive relationship between mental health and all spiritual/religious components was in line with the findings of other previous studies. It shows monotheistic beliefs, regardless of the type of religion; promote human health through acting as a positive construct which brings about solace and life satisfaction.\[[@ref2][@ref3][@ref22][@ref23][@ref24][@ref25][@ref26][@ref27][@ref28][@ref29][@ref30][@ref31][@ref32][@ref33][@ref34][@ref49][@ref50]\] Believing that there is support could be as effective as the support *per se*, especially if one believes that an imperishable power is supporting him/her. This can guarantee mental health.\[[@ref24][@ref25]\] In a religious belief system, all personal and social activities of individuals are based on movement toward physical and mental health and reliance on Him. Thus, working is also viewed as an activity to result in God\'s satisfaction, and therefore, promoting mental and physical health. The effect of religious/spiritual components of mental health and hope for future turned out to be 10% in this study. Of course in the fourth step of the stepwise regression, mental health variable, for the presence of another variable, existential well-being and its high predictive power, was omitted, which resulted in predicting 9% of the staff\'s hope for future by existential well-being and motivation, devotion, and coping variables. As the existential well-being and motivation, devotion, coping have been studies for the first time, it is impossible to compare the findings with those of previous studies, but generally it seems, that hopeful individuals are highly devoted to religion, and, even when praying, show this tendency by using words of hope. The research by Ghanizadeh and Vahedi also confirms our findings in this regard. Pinner *et al*. showed the relationship between religious tendencies and avoidance of hedonistic practices on one hand and well-being on the other. Van Dierendonck and Mohan stated that spiritual well-being is centered on an internal source, including knowledge of self and a profound spiritual feeling.\[[@ref47]\] Such a source doubles individual\'s self-confidence and strengthens him/her to deal with hardships. As the Quran says believers find their life meaningful and hope for God\'s favor and generosity. Through Knowing God, spiritual life is formed. Based on the interpretation and impression are forms about God and his/her activities, which in turn, influence life style, problem-solving strategies, hardship\'s management, and decision making procedures. Furthermore, the findings of the present study indicate that religious beliefs can guide individuals toward personality development and perfection and guarantee mental health. Faith in God could make one potentially secure against stereos that can threaten his/her mental health. CONCLUSION {#sec1-5} ========== It could be concluded the religious belief while ensuring staff\'s mental health forms the mental construct of reliance on God based on the establishment of a logical relationship with others and trust in God. As a result, work atmosphere is filled with spirituality, and the staffs, in spite of their hard-working conditions, do their part efficiently for the sake of God\'s satisfaction. It is worth mentioning that the staff\'s participating in the study feel that religion and spirituality stand above race, and geographical boarders. It is an individual\'s identity, and as it was mentioned before, it could unify people with each other, with the environment, and with God. Through the findings of the present study, to some extent, fills the information gap regarding hospital staff, they should be cautiously generalized to other populations and environments. It is also suggested that emphasis be shifted from superficial true religious beliefs in health providing centers. Convening educational workshops could also help nurture religious mental and behavioral patterns that develop spirituality and enhance the staff\'s mental health. Finally, as a limitation of the study, it should be mentioned that because of lack of similar studies, the findings be generalized with caution. **Source of Support:** Nil **Conflict of Interest:** None declared

INTRODUCTION ============ Patient expect for improved esthetics has driven the advancement of ceramic for use with fixed partial prostheses.[@B1] Many clinical studies demonstrate excellent long-term success of ceramic restorations. In recent year, strong ceramic cores unioning esthetic veneering porcelains have become popular as all ceramic restorations which have compensated the brittleness of porcelain and unesthetic metal substructure. The clinical success of all ceramic prosthesis depends on a number of factors, such as composition of the ceramic material and the cementation procedure.[@B2] Hence, bonding to ceramic requires strict attention to detail for optimal clinical outcomes.[@B3] A vital importance is due to the adhesive strength and durability of the complex formed between the three different components: the resin cement, the ceramic surface and the tooth surface especially in anterior laminate veneers.[@B4][@B5] In some cases, significant amounts of exposed dentin is usually unavoidable during the preparation of anterior teeth,[@B1][@B6] the protection is required during the period between preparation and cementation for prevention of post-operative sensitivity and bacterial invasion.[@B2] They suggested the application of the dentin bonding agent immediately after tooth preparation. This new technique of dentin bonding agent application prevents the bacterial invasion and dentin sensitivity during the provisional states, and the technique is concerned with increased bond strength *in vitro*.[@B7] Magne (2005) also recommended application of dentin adhesive to the freshly cut dentin when a significant area of dentin has been exposed during preparation for indirect restoration. The dentin could be sealed immediately after tooth preparation with IDS prior to impression taking.[@B3] IDS is the application of dentin bonding agent to freshly cut dentin when it is exposed during tooth preparation for indirect restorations. IDS protects the dentin against bacterial leakage and tooth sensitivity before cementation of final prosthesis. An advanced protocol, IDS is devised to address the challenges of preparation, provisionalization, and the final cementation of indirect restorative procedures. The general protocol of IDS includes the use of filled three-step total etch systems, two-step total-etch systems, and two-step self-etch systems incorporating low elastic liners.[@B8] Especially prosthodontic patients, complex inlay, onlay and veneer situations may require longer periods with provisional restoration until the final ceramic restoration is delivered.[@B9] The provisional restoration must protect the pulp from thermal changes as well as from invasive microorganisms in the mouth. However, it is difficult to takestable and sealed provisionals as they detached easily during temporary states, allowing microleakage of bacteria and sensitivity.[@B10] In addition, the bond strength between resin composite and a pre-treated ceramic restoration has been described to be negatively affected by external factors such as thermocycling, fatigue and water sorption.[@B11] Also, cement breakdown can bring about results as microleakage, marginal discoloration, debonding, secondary caries, pulpal irritation, and decreased fracture load.[@B12] We pose the clinical assumption that the dentin surface after IDS procedure is exposed to the oral environment and factors that could result in fatigue may influence the physical and mechanical properties of dentin bond strength. There are no studies available concerning the dentin bond strength on exposed sealed dentin with a long delay under thermocycling. The aim of this study was to evaluate the effect of IDS on dentin bond strength of lithium disilicate ceramic (IPS Empress II, Ivoclar) under various thermocycling periods of 1, 2, 7, 14 days. MATERIALS AND METHODS ===================== For tooth preparation, freshly extracted sound human mandibular third molars stored in solution saturated with thymol were used. The midcoronal dentin surfaces were created after the removal of the occlusal half of the crown using a low-speed diamond saw (Isomet: Buehler Ltd., Lake Bluff, IL, USA). Each specimen was individually secured to a silicon mold (diameter: 25 mm, height: 15 mm) and self-curing polyester resin (CH-304, Aekyung Chemical Co., Ltd., Seoul, Korea) was poured to create a resin-embedded specimen block. The surface was wet polished to create hybrid layer with 320, 400 and 600 grit SiC abrasive papers. Total 50 specimens were prepared. The experimental groups were divided into four groups (10 specimens per group) according to the thermocycling period (1, 4, 7, 14 days). The control group consisted of 10 specimens without thermocycling. A total of 50 ceramic discs (diameter: 4 mm, height: 2 mm) of lithium disilicate ceramic (IPS Empress II ingot, shade A1, Ivoclar Vivadent AG, Schaan, Liechtenstein) were fabricated. The surfaces were etched with 9.5% hydrofluoric acid (HF) (Porcelain etchant, Bisco Inc., Schaumburg, IL, USA) for 90 seconds and one layer of silane coupling agent (Porcelain primer, Bisco Inc., Schaumburg, IL, USA) was applied and allowed to air dry for thirty seconds at room temperature. The surface treatments of the ceramic discs were according to the manufacturer\'s instruction. The materials used in this study are shown in [Table 1](#T1){ref-type="table"}. For immediate dentin sealing procedure, the dentin surface was etched with 32% phosphoric acid (H~3~PO~4~) for 15 seconds, followed by rinsing with distilled water and air drying for 5 seconds. Then, five coat of dentin bonding primer (3-step etch-rinse adhesive system; All Bond II, Bisco Inc., Shaumburg, IL, USA) with a light brushing motion for 30 seconds was applied to the surface and air thinning for 3 seconds. After one coat of Pre-bond resin, the layer was light polymerized (VIP Junior light curing unit, Bisco Inc., Schaumburg, IL, USA) for 20 seconds at 600mW/cm^2^. Ceramic disc was attached with dual-cured resin cement (Duo-link, Bisco Inc., Shaumburg, IL, USA) on IDS treated dentin surface. After then, the specimens of experimental groups were submitted to 1500 thermal cycles between 5℃ and 55℃ (dwell time of 30 seconds) in a thermal cycling machine (Thermocycling testing machine, CDM-127, CDM, Korea) for 1, 2, 7, 14 day (group 1d, 2d, 7d, 14d) ([Table 2](#T2){ref-type="table"}).[@B13][@B14] Control group (0d) was not thermocycled and the specimens were stored for 14 days in deionized water at room temperature. Ceramic discs after surface treatment with ceramic etchant and silane coupling agent were cemented to the surfaces of specimens with or without thermocycling with dual-cured resin cement (Duo-link, Bisco Inc., Shaumburg, IL, USA) ([Fig. 1](#F1){ref-type="fig"}). The excess cement was removed with a disposable microbrush, followed by light curing (VIP Junior light curing unit, Bisco Inc., Schaumburg, IL, USA) with a light intensity of 600 mW/cm^2^. The light was applied for 100 seconds (20 seconds each from occusal, buccal, lingual, mesial and distal aspects). The extrusion shear bond strength test (SBST)[@B15][@B16] represents a confinement situation for the composite, and the resulting interface would be more likely to present defects that resemble clinical conditions. After the cementation procedures, the specimens of control group and thermocycled groups were stored in distilled water for 24 hours at 37℃. The SBST was conducted in a universal testing machine (Instron, Shimadzu, Japan) at a crosshead speed of 1.0 mm/min. For the scanning electron microscopy (SEM) analysis, the dentin surfaces of specimens from each group were air dried and gold coated with a sputter coater (IB-3 ION coater, Elko Co., Tokyo, Japan) and examined under scanning electron microscope (Scanning Electron Microscopy, S-2300, Hitachi, Co., Ltd., Tokyo, Japan). The specimens were vertically sectioned using a low-speed diamond saw under water lubrication to observe the interface of dentinresin-ceramic. Bond strength values were analyzed using one-way analysis of variance and followed by Tukey\'s HSD multiple comparison tests. Statistical analysis was conducted using SAS software version 9.1 for windows (SAS Institute Inc., Cary, NC, USA). The significance level for all statistical tests was set at 0.05. RESULTS ======= The values of the shear bond strength recorded in debonding force (DF) for the specimens in all groups. The mean values and standard deviation of shear bond strength are shown in [Fig. 2](#F2){ref-type="fig"}. Two thermocycled groups had lower mean values than the control group. The analysis of variance indicated no statistically significant difference in shear bond strength between all groups, however the mean value started to decrease in group 7d, and group 14d showed the lowest mean bond strength in all groups. SEM micrograph of fractured surface of dentin side after SBST (original magnification ×100) are shown in [Fig. 3](#F3){ref-type="fig"}. SEM analysis of fractured surface in the control group, thermocycled 1d and 2d groups demonstrated mostly undamaged interface of overlaying resin cement on IDS surface ([Fig. 3A, Fig. 3B, Fig. 3C](#F3){ref-type="fig"}). It confirmed that dentin was completely covered with dentin bonding and adhesive resin cement, suggesting failure was not occurred between IDS surface and resin cement. At the group 7d ([Fig. 3D](#F3){ref-type="fig"}), two distinct areas were observed, there were exposed IDS surface and resin cement. Areas with detached resin cement has increased and founded widely. As shown in the [Fig. 3E](#F3){ref-type="fig"}, most of resin cement has been detached from IDS surface. A representative SEM images of fractured surface of dentin side after SBST (original magnification ×400) are shown in [Fig. 4](#F4){ref-type="fig"}, and showed similar pattern to SEM views for original magnification ×100 ([Fig. 3](#F3){ref-type="fig"}). The specimens of groups 0d, 1d and 2d showed very small portion of detached resin cement and there were undamaged interface of overlaying resin cement on IDS surface ([Fig. 4A, Fig. 4B, Fig. 4C](#F4){ref-type="fig"}). While in group 7d ([Fig. 4D](#F4){ref-type="fig"}), showed a mixed failure surface, areas with remained resin cement and exposed IDS surface were found at the same time. In group 14d ([Fig. 4E](#F4){ref-type="fig"}), extensive area of IDS surface was found. Most of resin cement has been detached from IDS surface, suggesting failure between existing resin cement and IDS surface. Especially, group 7d and 14d ([Fig. 4D, Fig. 4E](#F4){ref-type="fig"}), showed distinct exposed dentin, which suggests adhesion failure between dentin and adhesive. In [Fig. 5](#F5){ref-type="fig"}, adhesive-resin cement (b) was found between ceramic disc (a) and dentin covered with adhesive (c). The specimens of groups 0d, 1d and 2d, uniform and thin adhesive-resin cement layers were observed ([Fig. 5A, Fig. 5B, Fig. 5C](#F5){ref-type="fig"}). An almost imperceptible transition seemed to take place at the ceramic disc-resin cement-dentin interface, where no gap was observed despite of some stresses generated by thermocycling. In group 7d ([Fig. 5D](#F5){ref-type="fig"}), as compared with above three groups, thickening and peeling of adhesive-resin cement layer (b) was found. In group 14d ([Fig. 5E](#F5){ref-type="fig"}), partially visible discontinuity of interface and linear detachment between IDS surface and resin cement was observed. In [Fig. 6](#F6){ref-type="fig"}, adhesive-resin cement (b) was found between porcelain disc (a) and dentin covered with adhesive (c). Higher magnification examination shows similar pattern to SEM views for original magnification ×100. Between group 0d (control group) and thermocycled group 1d and 2d, no differences were found between adhesive-resin cement layers (b) with regard to form and thickness ([Fig. 6A, Fig. 6B, Fig. 6C](#F6){ref-type="fig"}). Adhesive interface remained intact and complete continuity was observed. In group 7d ([Fig. 6D](#F6){ref-type="fig"}), perceptible changes appeared in adhesion interface, in group 14d ([Fig. 6E](#F6){ref-type="fig"}), obvious gap between adhesive-resin cement layer (b) and dentin (c) was found. It showed adhesive failure between the IDS surface and resin cement. DISCUSSION ========== Over the past twenty years, the field of esthetic prosthodontics has substantially expanded to satisfy the needs of the patients.[@B8] Bonded ceramic restorations are a predictable and durable treatment option to re-establish the strength and function of the teeth as well as esthetic.[@B3] Especially, the success of the ceramic veneer is highly dependent on the strength and durability of the bond formed between the three different components of the bonded veneer complex, as there are; the tooth surface, the luting composite and the ceramic veneer.[@B9] There have been no studies regarding the dentin bond strength and SEM analysis of immediate sealed dentin that is exposed with an extended delay under thermocycling. Phosphoric acid etching causes dissolution of interprismatic material in the enamel, producing a roughened enamel surface, and forms enamel resin tags. In recent years, the research of increasing the bond strength between the dental materials and dental hard tissues have been proceeded, although most adhesive systems have provided clinically acceptable bond strengths.[@B11] Mechanical adhesion requires roughening to create sufficient surface activation. Common surface treatment methods include sandblasting with alumina, grinding, and abrasion with diamond instrument,[@B10][@B12] and acid etchingwith HF solution or ammonium bifluoride.[@B17][@B18] The sequential application of silane coupling agents to previously HF-etched ceramic surface provides a chemical bonding. Chemical bonding of silane is shown to be in correlation with increased durability of long-term bond potential and less with immediate mechanical bond strengths.[@B19][@B20] The preferred procedure for predictable bonding to silicabased ceramics is acid etching with HF and the subsequent application of silane coupling agents. With the improvement of esthetic restoration material, newly developed resin cement which can provide both mechanical and chemical bonding has been introduced.[@B13][@B21][@B22][@B23] Application of a silane coupling agent to HF-etched ceramic surfaces provides a strong chemical covalent bond between the hydroxyl group of silane coupling agent and that of the silica of ceramic. When the silane primer is applied onto the porcelain surface, two hydroxyl groups (one from and silane and one from the porcelain) will react slowly while liberating a water molecule to form a stable siloxane. The most commonly used medium and high leucite-containing feldspathic porcelains is etched with HF and makes preferential etching of the glass phase of ceramic, leaving the less soluble crystalline phase of leucite composition. This forms micromechanically retentive surfaces on porcelain. The preferential etching provides many hydroxyl (OH) groups that can be coupled with silane coupling agent for improved chemical bonding with resin materials.[@B24] When preparing teeth for indirect bonded restorations, IDS with a 3-step etch-and-rinse DBA, before impression taking, leads to increased microtensile bond strength compared to delayed dentin sealing (DDS). The failure site was observed with microscope (×30) and failure mode of cementation was determined. If the failure was occurred entirely between the adhesive and dentin or continued from the adhesive into either the composite resin or dentine, it was classified as an interfacial failure. And if the failure occurred exclusively within the resin composite or dentin, it was classified as a substrate failure. IDS groups showed interfacial failure that was generally two areas as typically mixed, with both areas of failed adhesive resin and areas of cohesively failed dentin. For DDS specimens, it was hard to decide the exact nature of interfacial failure as the similar nature of the filled adhesive and hybrid layer.[@B1] Magne (2005) demonstrated that several reasonable motives supporting IDS. First, immediately exposed dentin after preparation has an ideal surface for dentin bonding.[@B1] Confoun-ding results have been reported when the shear bond strength of dentin bonding agents was measured on dentin that had been contaminated with provisional cements prior to the bonding procedure.[@B2] Freshly cut dentin exists only immediately after tooth preparation. Secondly, Precuring of the DBA results in increased bond strength. In recent studies on bond strength of DBA, the infiltrating resin and adhesive layer are generally polymerized first (precuring), before composite resin is incremented, which represents to generate increased bond strength. If exposed dentin is sealed without delay, the DBA being covered and cured immediately after the finish of tooth preparations, before the final impression, generated higher bond strength[@B25][@B26][@B27] and fewer gap formations.[@B28][@B29] IDS allows stress-free dentin bond development. When indirect bonded restorations proceeded with IDS, because of postponed placement of the restoration (intrinsic to indirect techniques), the dentin bonding strengths can be improved without stress, resulting in considerably developed restoration adaptation.[@B30] In addition, IDS reduced bacterial infiltration and sensitivity to dentin during provisional restoration. Based on the fact that provisional restorations enable bacterial leakage and dentin sensitivity, previous researcher suggested dentin sealing in crown preparations.[@B31] This suggestion demonstrates even more advantages when using bonded ceramic restoration (e.g., veneers) given the specific difficulty to obtain sealed and stable provisionals. An important element in IDS is the advancement of an efficient resin-to resin bond between the new luting composite resin and the adhesive resin coating. This condition is somewhat similar to a composite resin-composite resin repair.[@B5] For years, many believed that an oxygen-inhibited layer is essential prior to bonding additional layers of composite. The general understanding regarding the oxygen-inhibited layer has been documented. Suh[@B32] reported that an oxygen-inhibited layer is not necessary for bonding with composite resin and Papachini[@B33] identified that oxygen inhibition layer is not crucial to the success of composite-composite bond. A previously placed resin matrix is naturally cohesive with monomers in freshly placed composite resin.[@B32][@B33] Magne *et al*.,[@B5] reported that the bond strength of IDS teeth was not affected by up to 12 weeks of elapsed time prior to placement of the definitive restoration. In this present study, there was no statistically significant difference in shear bond strength between all groups. As demonstrated by the shear strength bond results, the mean shear bond strength varied from 82.7 to 93.93 N. The mean value started to decrease at day 7 and markedly decreased at day 14. In most of the studies regarding IDS, teeth were restored with provisional restoration material and left in place for a certain period of time. However, in this study, we hypothesized that all-ceramic restoration such as laminate veneer crown is likely to be contaminated with a wide variety of things including temporary cement residues, saliva and bacteria. We attempt to mimic extreme clinical conditions when the provisional restoration is fractured or dislodged before the delivery of the final restoration. In our study, after immediate dentin sealing, teeth were immersed in saline solution without provisional restoration. Thus, the result of shear bond strength in this study shows difference from that of previous studies. In conventional glass-ceramic restorations for prosthesis, the adhesive technique is crucial for successful bonding. Considering limited flexural strength and the brittleness of glass ceramics, final adhesive cementation with luting composite resin should be used to improve the fracture resistance. In this present study, IPS Empress II was tested which is widely used as a ceramic material of laminate veneer crown. IPS Empress II (Ivoclar Vivadent, Schaan, Liechtenstein) is a lithium-disilicate glass ceramic (SiO~2~-Li~2~O) manufactured through a combination of the lost-wax and heat-compression techniques. A glass-ceramic ingot is plasticized at 920℃ and pressed into an investment mold under pressure and vacuum. Its predecessor, IPS Empress (Ivoclar Vivadent), is a leucite-reinforced glass ceramic (SiO~2~-Al~2~O~3~-K~2~O) and is restricted in use of single unit complete-coverage crown restoration due to its limited strength. IPS Empress 2 has increased flexural strength by a factor of 3 over IPS Empress, and enables the use of 3-unit FPDPs in the anterior area, and the second premolar area.[@B34][@B35][@B36] One way of evaluating the physical and mechanical properties of the material is to subject them to an *in vitro* simulation of the aging that would occur in the medium in which they are used, by means of anaccelerated aging test by thermocycling. Chang *et al*.[@B37] reported that bond strength of various adhesive systems is influenced by thermocycling. The adverse effect of thermal cycling on the adhesive combination can be explained by the fact that materials with different lineal thermal expansion coefficients (LTEC) also presented different degrees of shrinkage and expansion. That process facilitates the fatigue phenomenon of the materials, leading to rupture of the union between them.[@B38] Mechanical fatigue alone could not deteriorate the interface between the cements and the ceramic materials as much as it could for the humid condition with the presence of stress corrosion.[@B39] Sobrinho *et al*. proved a reduction in strength after fatigue in a wet environment for In-Ceram and IPS Empress ceramic in comparison to the dry condition.[@B40] Considering that final restoration is delivered one or two weeks after the tooth preparation, the specimens were submitted to thermocycling until 14 days. At the groups 1d, 2d, 7d and 14d, the specimens were thermocycled 1,500 cycles (1 day), 3,000 cycles (2 days), 10,500 cycles (7 days), 21,000 cycles (14 days) in baths at 5C and 55C, with 30-second dwell times. The results of the present study demonstrated that thermal cycling negatively influenced the quality of adhesive resistance, promoting the decrease in union values when the number of thermal cycles was increased.[@B14][@B38][@B41][@B42][@B43][@B44] Qualtrough *et al*. reported that there was a deterioration in marginal quality at the porcelain inlay/composite resin interface as a result of thermocycling.[@B45] In the present study, cross-sectional SEM analysis of dentin surface in magnification of ×100 and ×1000 demonstrated that the thermocycled group until 2 days did not show a difference from the control group with regard to shape and morphology: uniform and thin layer of adhesive-resin cement layer was observed. In thermocycled group 7d, thickening and peeling of adhesive-resin cement layer was appeared, and in thermocycled group 14d, distinctive failure on IDS surface were observed. SEM analysis of dentin surface showed dentin is almost completely covered with adhesive-resin cement layer in thermocycled groups 1d and 2d, and small fragment of detached resin was observed. In thermocycled group 7d, largely detached resin and extensive area of dentin were observed, and in thermocycled group 14d, most of the adhesive resin was detached. The limitation of this study is that quantitative assessment of the size of the ceramic discs was not performed. In this study, a large standard deviation of shear bond strength indicates that there exists a delicate difference in adhesive areas of porcelain discs due to little difference in the size of porcelain discs. Further studies should be conducted to evaluate shear bond strength per unit area. CONCLUSION ========== In conclusion, among the specimens of thermocylced groups (1 day, 2 days, 7 days, 14 days) after IDS procedure, the bond strength started to decrease in the thermocycled 7 days group and noticeably decreased in the thermocycled 14 days group. The SEM micrographs on dentin surface of adhesively failed specimens revealed that the thermocycled 7 days group showed a mixed failure surface areas with remained resin cement and exposed IDS surface, and also the largest area of detached resin cement and exposed dentin were identified in the thermocycled 14 days group. The SEM micrographs of cross-section demonstrated partial detachment and thickening of resin cement layer in thermocycled 7 days group, and distinct linear gap between resin cement layer and the dentin in the thermocycled 14 days group. Therefore, the bond of definitive ceramic restorations such as inlay/onlay and laminate veneer should be bonded within one week after IDS procedure. This work was supported by the Basic Science Research Program of the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2013R1A1A1061129). {#F1} {#F2} {#F3} {#F4} {#F5} {#F6} ###### Materials used in this study  Materials Product name Composition Manufacturer ----------------- ------------------- ----------------------------------------------- -------------------------------------------- DBA system All Bond II 4^th^ Generation: 3-step total etching system Bisco Inc., Shaumburg, USA Resin cement Duo-link Dual cure resin cement Bisco Inc., Shaumburg, USA Etchant UNI-ETCH 32% Phosphoric acid (15 sec) Bisco Inc., Shaumburg, USA Ceramic disc IPS Empress II Lithium disilicate glass-ceramic Ivoclar Vivadent AG, Schaan, Liechtenstein Ceramic etchant Porcelain etchant 9.5% Hydrofluoric acid (90 sec) Bisco Inc., Shaumburg, USA Silane Porcelain primer Coupling agent (30 sec) Bisco Inc., Shaumburg, USA ^\*^DBA: Dentin Bonding Agent ###### Experimental design and groups used in this study (d; day)  Thermocycling time Group -------------------- -------------- 0 day (immediate) 0d (control) 1 day 1d 2 days 2d 7 days 7d 14 days 14d

INTRODUCTION ============ For decades, high ligation and saphenous vein stripping with phlebectomy of varices were the gold standard procedures in the treatment of the incompetence of the great saphenous vein (GSV) \[[@b1-vsi-30-102],[@b2-vsi-30-102]\]. An alternative method was cryostripping which is less invasive than traditional stripping \[[@b3-vsi-30-102]\]. Because of the need for minimal invasiveness, the endovenous procedures spread worldwide in the treatment of the incompetence of the saphenous trunks in the last decade \[[@b4-vsi-30-102],[@b5-vsi-30-102]\]. At the end of the last century, Milleret and Le Pivert \[[@b6-vsi-30-102]\] described cryosclerosis as a safe, efficient and feasible endovenous cryoablation technique to treat the incompetence of the GSV. The segmental freezing of the saphenous trunk was performed with a flexible cryoprobe by inguinal approach. In 1987, Le Pivert \[[@b7-vsi-30-102]\] published the experiences of 350 patients treated by this method. In 1994, Garde \[[@b8-vsi-30-102]\] concluded that the technique was safe and efficient based on the results of a randomized controlled trial on 800 patients. We have completed a prospective non-randomized trial to compare cryosclerosis and classical stripping, and regarding the short term results, this endovenous procedure seems to be effective. The electron microscopic examination proved that immediate ultrastructural changes were formed in the GSV \[[@b9-vsi-30-102]\]. CASE ==== We present the histological changes of the GSV at 2 years after cryoablation. The examined vein piece was harvested from the proximal femoral part of the GSV from a 31-year-old female patient who underwent cryosclerosis 23 months ago. This procedure was performed by inguinal approach, after high ligation a cryoprobe was introduced into the GSV and segmental freezing of the vessel wall was carried out. The patient was admitted again to our department of surgery because of recurrent varicosity that caused pain and limb swelling at the end of the day. Ultrasonography showed the recanalization of the GSV. Saphenofemoral reflux and an incompetent posteromedial perforating vein were detected as possible reasons. High ligation and cryostripping with phlebectomy of varices were performed under spinal anesthesia. The GSV was very crusted and segmental strictures were found when the cryoprobe was introduced into the vessel lumen. A small vein piece was harvested from the proximal femoral part of the GSV for histological examination. The patient healed without any postoperative complications. Hematoxyllin-eosin, Picro-sirius Red, van Gieson and immunohistochemical stains (CD34, smooth muscle actin) were performed. The anatomical structure of the vein couldn't be recognized microscopically. The wall was collagenized and padded by endothelium that was caused by progressive fibrosis ([Fig. 1](#f1-vsi-30-102){ref-type="fig"}). The patient consented to this trial and agreed to apply her records. DISCUSSION ========== Generally, the role of the endovenous ablation techniques (endovenous laser ablation \[EVLA\], radiofrequency ablation \[RFA\], ultrasound guided foam sclerotherapy, mechanochemical ablation, steam ablation) has increased in the treatment of the incompetence of the GSV. These are widely studied, safe and efficient procedures. The most frequently used methods are EVLA and RFA. The long term results are favorable, but the recanalization of the GSV could be observed some years after the treatment that may cause recurrent varicosity with symptoms \[[@b1-vsi-30-102],[@b2-vsi-30-102],[@b4-vsi-30-102],[@b5-vsi-30-102]\]. In this short report, the presented method is cryosclerosis, which is the endovenous cryoablation of the GSV. This procedure is not widely known. The reported patient had recurrent varicosity with symptoms. Cryosclerosis was ineffective 2 years after the treatment; however, segmental stenoses of the GSV were observed during the operation. The incompetence of the saphenofemoral junction and the posteromedial perforating vein could cause the recurrent disease. The role of tumescent fluids during cryosclerosis is not clear and it is not routinely used, but it might be useful since by decreasing the volume of the intraluminal blood, vein wall destruction can be achieved more efficiently \[[@b10-vsi-30-102],[@b11-vsi-30-102]\]. A prospective trial is necessary to assess the effect of injecting warming liquid around the GSV before cryosclerosis. The histomorphological effects of the endovenous ablation methods have been previously investigated, mostly from animal experiments \[[@b12-vsi-30-102]--[@b14-vsi-30-102]\]. Their long term effects have not been described yet. It's typical for all procedures that the vein wall is destroyed by thermal, mechanical or chemical effects \[[@b1-vsi-30-102],[@b4-vsi-30-102],[@b6-vsi-30-102],[@b9-vsi-30-102],[@b10-vsi-30-102]\]. Heger et al. \[[@b12-vsi-30-102]\] summarized in their review on EVLA that the heat induced thrombosis and the thermal damage in the vein wall results in inflammation and activation of the immune system (fibroblast migration). Finally, the vein wall becomes collagenized and results in occlusion or stenosis. The presented histomorphological examination showed that the whole vein treated by cryosclerosis underwent typical remodeling, thus the anatomical structure couldn't be recognized. The wall became crusted because of collagen deposition by fibroblast activity. However, the GSV wasn't occluded yet segmental strictures were observed ([Fig. 1](#f1-vsi-30-102){ref-type="fig"}). Cryosclerosis seems to have the same effect as the other familiar endovenous ablation techniques \[[@b1-vsi-30-102],[@b2-vsi-30-102],[@b4-vsi-30-102],[@b6-vsi-30-102]--[@b9-vsi-30-102]\]. The reason for recanalization of the GSV is not clear, with recurrent incompetence of the saphenofemoral junction and the posteromedial perforating vein being one of the reasons, but it is also possible that the remodeling process of the GSV was not effective enough. The pathogenesis of recanalization after the endovenous procedures and recurrent varicosity has been studied, but more evidence is needed \[[@b5-vsi-30-102]\]. Considering that this article is only a case report, it has serious limitations. More cases are necessary to assess the real effect of cryosclerosis on the GSV, but the procedure seems to be efficient in obliterative remodeling of the vein. Conflict of interest: None. {#f1-vsi-30-102}

Introduction {#S1} ============ Porcine circovirus has long harmed the sound development of pig husbandry, and especially the PCV3 that has appeared in recent years has caused tremendous economic losses for pig husbandry ([@B17]; [@B21]). Current research indicates that PCV is mainly divided into three genotypes ([@B8]). PCV1 is generally considered non-pathogenic ([@B27]). PCV2 is widely prevalent worldwide, previous studies have shown that PCV2 is the main pathogenesis of post-weaning multisystemic wasting syndrome (PMWS) and swine dermatitis and nephrotic syndrome (PDNS) ([@B1]; [@B30]). However, in recent years, some researchers have detected PCV3 from PDNS piglets ([@B22]; [@B13]). Studies have shown that PCV3 and PCV2 are mixed infections and have become popular in many countries. It has been reported that PCV3 may cause reproduction disorder in sows and PDNS in adult pigs ([@B22]). Similar to PCV2, PCV3 is often mixed with PRRSV, PCV2 ([@B2]). However, unlike PCV2, there are currently commercial vaccines to prevent PCV2 infection ([@B23]), and there is still a shortage of vaccines and related drugs to prevent PCV3. Obtaining standard virus strains is the basis for the development of PCV3 vaccines and related biological products. However, no experimental report on the successful isolation of PCV3 has been reported. With the development of genetic engineering technology, a variety of virus strains have been constructed through reverse genetics ([@B34]; [@B10]). With the continuous analysis of the genome structure and function of PCV2 and PCV3, researchers have obtained infectious clones by constructing eukaryotic expression vectors of PCV2 and PCV3 ([@B28]; [@B11]). According to the genome structure of PCV, the porcine circovirus genome is a single-stranded negative-stranded DNA, the virions are only 14--17 nm, and there is no capsule on the surface of the virus capsid ([@B18]; [@B32]). The size of the PCV1 and PCV2 genome research surface is 1767--1768 bp and contains 11 open reading frames (ORFs), of which ORF1 is a virus replication--related protein (Rep) and is a necessary element of virus replication ([@B3]). ORF2 encodes the viral capsid protein (Cap) ([@B20]) and is commonly used in the study of subunit vaccines and diagnostic reagents ([@B16]). Unlike the genomes of PCV1 and PCV2, the full length of the PCV3 genome is 2000 bp ([@B24]). Some researchers have predicted the PCV3 genomic DNA, and currently they have predicted a total of three ORFs ([@B22]). One Rep protein composed of 297 amino acids was encoded by ORF1 ([@B33]). Besides, another Cap protein covering 214 amino acids by ORF2 was replicated in the opposite direction and a protein with an unknown function and containing 231 amino acids by ORF3 ([@B5]). Whether PCV3 has other open reading frames and their functions still needs further study. The above research is to construct infectious clones under the premise of grasping the genome structure of the virus. However, for some newly discovered circular DNA viruses, the genome structure and function are undefined yet, so it is hard to construct infectious clones with the application of the eukaryotic expression vector. This study intends to use the biological characteristics of PCV3 circular DNA to construct PCV3 infectious clones and construct a Kunming mouse infection model without the help of exogenous expression vectors. The results showed that the PCV3 strain can infect the myocardium and lung of mice. This study will provide a method for the construction of infectious clones of other circular DNA viruses and lay a foundation for the study of the pathogenic mechanism of PCV3. Materials and Methods {#S2} ===================== Cells and Cultures {#S2.SS1} ------------------ The 3D4/21 cell line (iCell Bioscience Inc., Shanghai, China) was cultivated in Dulbecco's minimum essential medium (MEM, Gibco) supplemented by 10% fetal bovine serum at 37°C in a humidified 5% CO~2~ incubator. The cloning and construction of recombinant expression plasmids was carried out in *E. coli* strain DH5α cells (Takara Bio, Dalian, China). The prokaryotic expression vector pET-32a (+) and *E. coli* BL21 (DE3) cells were harvested from the stocks of our laboratory. The SP2/0 cells were also acquired from the stocks. Viral Gene and Primer Synthesis {#S2.SS2} ------------------------------- The PCV3 gene sequence (GenBank accession number: [MH107162.1](MH107162.1)) was taken to form a loop on Snap-Gene software, and the position of the unique restriction site was adopted to open the sequence, termed as the rearranged linear PCV3 gene sequence. Since the PCV3 is circular, the cyclization of the original linear DNA sequence was considered, the *Hin*dIII restriction site was recruited as the reopening site. When the opening process was completed, the *Hin*dIII restriction site was added to the end of the linear sequence. The newly rearranged sequence was synthesized by Sangon (Shanghai, China). [Figure 1](#F1){ref-type="fig"} illustrates the gene pattern of the novel PCV3 and presents the design pattern of the circular PCV3 amplification primer and detection primer. Primers listed in [Table 1](#T1){ref-type="table"} were designed in accordance with the rearranged PCV3 gene sequence and PCV3 Cap gene sequence; the designed primers were synthesized by Sangon (Shanghai, China). ###### Primers used for the construction and identification of the recombinant virus. Primer Sequence --------------------------------------------------------------------- PCV3 M1: 5′-CCC[AAGCTT]{.ul}GTGCGGATGCGGCTGCGCG-3′; PCV3 L2: 5′- CCC[AAGCTT]{.ul}CCCGCGTTTTCCCACAACC-3′; PCV3 L3: 5′- GGTTGTGGGAAAACGCGGG-3′; PCV3 M2: 5′-CGCGCAGCCGCATCCGCACAAGCTT-3′; PCV3 Cap *Bam*HI: 5′-CGC[GGATCC]{.ul}ATGAGACACAGAGCTATATTCAGAA-3′; PCV3 Cap *Hin*dIII: 5′-CCC[AAGCTT]{.ul}TTAGAGAACGGACTTGTAACGAATC-3′ The underlined sequences represent restriction enzyme sites introduced. PCV3 M1, PCV3 L2: PCV3 complete genome amplification primers; PCV3 L3 is a reverse complementary sequence of PCV3 L2; PCV3 M2 is the nucleotide sequence next to the primer binding site of PCV3 L2; PCV3 Cap BamHI, PCV3 Cap HindIII: The design of the primer is the reverse complement of the negative chain of PCV3 with the Cap gene sequence, and specific primers were then designed according to the reverse complementary DNA sequence. {#F1} Amplification and Cyclization of the PCV3 Genome {#S2.SS3} ------------------------------------------------ To amplify the newly synthesized rearranged PCV3 gene, PCV3 M1, PCV3 L2 primers, and KFX-101 high fidelity enzymes (TOYOBO, Japan) were adopted. The final volume of reaction fluid reached 100 μL. The PCR cycling conditions included: pre-denaturation at 94°C for 2 min, 30 cycles at 98°C for 10 s, 58°C for 30 s, and 68°C for 2 min, as well as the final extension for 10 min at 68°C. PCR products were preserved at 16°C for subsequent experiments. The amplified PCR products of rearranged PCV3 gene were purified following the instructions of the DNA Purification Kit (Takara Bio, Dalian, China). Subsequently, the purified PCR product was digested by *Hin*dIII restriction endonuclease (NEB, Beijing, China) and then incubated in a water bath at 37°C for 5 h. Then, the digested products of *Hin*dIII restriction endonuclease were purified following the manufacturer's instructions (Takara Bio, Dalian, China). Next, the purified DNA fragments were connected based on T4 DNA ligase instructions (NEB, Beijing, China) at 16°C for 12 h. Last, the cyclized PCV3 DNA was harvested. Transfection of Cyclized PCV3 DNA {#S2.SS4} --------------------------------- The cyclized PCV3 DNA was transfected into 3D4/21 cells (60% confluency) in each well of the 6-well plates. Next, cells were transfected with of 10 μg cyclic PCV3 DNA with the use of Lipofectamine 3000 (Invitrogen, Carlsbad, CA, United States) following the manufacturers' protocol. Afterward, the mixture was gently blended with a pipette and cultivated in a 5% carbon dioxide incubator at 37°C for 7--10 days. Detection of PCV3 Infectious Clone by PCR {#S2.SS5} ----------------------------------------- Nucleic acid was extracted from cyclized PCV3 DNA transfected cells and normal cell supernatant in accordance with the instructions of the Takara virus DNA/RNA Extraction Kit (Takara Bio, Dalian, China) to perform PCR and RT-PCR. PCV3 L3, PCV3 M2 primers and Premix Taq^TM^ (Takara Bio, Dalian, China) were adopted for the common PCR detection, and PCV3 Cap *Bam*HI and PCV3 Cap *Hin*dIII primers and a PrimeScript^TM^ One Step RT-PCR Kit (Takara Bio, Dalian, China) were employed to perform the RT-PCR analyses. The schemes of PCR cycling included: 94°C for 4 min, 30 cycles at 94°C for 30 s, 58°C for 30 s, and 72°C for 2 min, as well as the final extension for 10 min at 72°C. The RT-PCR procedure included: reverse transcription at 50°C for 30 min, pre-denaturation at 94°C for 4 min, 30 cycles at 94°C for 30 s, 58°C for 35 s and 72°C for 2 min, as well as the final extension at 72°C for 10 min. Preparation of Specific Monoclonal Antibody Against PCV3 Cap {#S2.SS6} ------------------------------------------------------------ To produce monoclonal antibodies (mAbs) against PCV3 Cap, recombinant PCV3 Cap protein was expressed with the prokaryotic expression system. In brief, PCV3 Cap gene was amplified and then cloned into pET-32a (+) vector; subsequently, it was expressed in *E. coli* and then purified with a His-tag protein purification kit (CWBio, Beijing, China). Likewise, PCV1 and PCV2 Cap protein were prepared and then purified. Five 6-week-old female BALB/c mice underwent the subcutaneous injection of 50 μg recombinant PCV3 Cap and 50 μL Freund's complete adjuvant in a final volume of 100 μL. Three booster immunizations, each of which had the equivalent dosage of antigen plus Freund's incomplete adjuvant, were administered at the intervals of 2 weeks. Five days after the final booster injection, blood samples were collected via the tail vein. Subsequently, the mice were euthanized, and the spleen was obtained and then fused with SP2/0 myeloma cells to screen out the positive hybridoma cells as previously described ([@B19]). The mice were injected in their abdominal cavity with positive hybridoma cells secreting anti-PCV3 Cap antibodies. The ascites of the mice were collected after the occurrence of abdominal bulge; then, they were filtered and purified. The anti-PCV3 Cap mAb was obtained and named after the hybridoma cell line. PCV3 Cap, PCV2 Cap, PCV1 Cap, and His-tagged proteins acted as detection agents, and hybridoma cells were screened by indirect ELISA. Western Blot {#S2.SS7} ------------ The reactivity of mAbs to recombinant PCV3 Cap and rescued PCV3 was ascertained by a Western blotting assay. The virus and protein were fractionated on a 10% Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis (SDS-PAGE) gel; next, they were transferred to PVDF membrane and then blocked with PBST buffer supplemented by 5% skim milk powder for 2 h. Subsequently, PCV3 Cap mAb (1/1000 dilution in PBST) was adopted as the primary antibody, and goat anti-mouse HRP-IgG (1/5000 dilution in 5% skim milk blocking solution) acted as the secondary antibody after rinsing five times with PBST. Lastly, the color development of PVDF membranes was achieved following the instructions of DAB color development kit (CWBio, Beijing, China). Indirect Immunofluorescence Assay {#S2.SS8} --------------------------------- 3D4/21 cells were cultivated in 6-well cell culture plates. When 70% of the cells were fused, 200 μL rescued PCV3 virus solution was added, and the cells without inoculation were set as the control. After 72 h cell culture, the supernatant was abandoned, and the cells were subsequently rinsed with PBS. In each well, 500 μL 4% paraformaldehyde (PFA) was added to fix cells at ambient temperature for 10 min. Then, 200 μL 0.1% Triton X-100 was applied for rinsing at ambient temperature for 10 min, and 500 μL 3% BSA solution was applied for blocking at ambient temperature for another 1 h. The ascites identified as positive by a Western blotting assay acted as the first antibody (1/200 dilution in PBS), and goat anti-mouse TRITC-IgG acted as the secondary antibody (1/1000 dilution in PBS) to incubate cells. Lastly, a DAPI staining kit (Beyotime, Shanghai, China) was employed to observe the cell morphology as irradiated by the red excitation light. Viral Titration for Serial Passages {#S2.SS9} ----------------------------------- Virus titers in cell cultures for each passage were measured by IFA, and TCID~50~/ml was recorded with Reed--Muench method, and then draw virus growth curve. The viral cultures were serially diluted 10-fold in DMEM medium supplemented by 5% FBS and antibiotics. The rescued PCV3 virus dilution was inoculated into 96-well plates covering 100 μL 3D4/21 cells in suspension. After the incubation for 2 h at 37°C, the liquids in 96-well plates were removed, and DMEM with 2% FBS was added to the wells. Subsequently, the plates were incubated for an additional 72 h at 37°C. Animal Experimental Design {#S2.SS10} -------------------------- Two experiments were designed to analyze the pathogenesis of PCV3 infection in mice. Ten 6-week-old KM mice were randomly split into two groups. The mice in the experimental group were intraperitoneally injected with the 0.5 mL rescued PCV3 virus, respectively. The mice in the control group were inoculated with the identical dose of PBS. Each group of mice was housed in an individual room while fed sterile food and water. Mice were subject to euthanasia 21 days after infection, and then tissue samples were dissected and collected. If the mice died in the experiment, they would be immediately dissected and their tissue samples would be collected. Clinical and Pathological Examination {#S2.SS11} ------------------------------------- The clinical symptoms of mice infected with virus were observed every morning and evening. After 21 days of the infection, the mice were dissected, and the pathological variations of various tissues and organs were recorded. The tissues of heart, liver, spleen, lung, kidney, brain, and lymph nodes were fixed with 4% formaldehyde solution. Paraffin sections were produced routinely, stained with HE, and subsequently observed under a microscope. Immunohistochemical Staining {#S2.SS12} ---------------------------- To detect PCV3 viral antigen, the histopathological tissue samples were stained with IHC. The tissue samples were harvested from PCV3-inoculated mouse. First, the tissue samples were fixed with 4% formaldehyde to develop paraffin blocks. The slices were cut into 3--4 μm thick slices and then incubated overnight at 37°C. Subsequently, the slides were de-waxed and then blocked with 3% H~2~O~2~ for 10 min at ambient temperature. After rinsing with PBS five times, the slides were incubated with normal goat serum (1/20 dilution in PBS) at 37°C for 15 min and subsequently with PCV3 Cap mAb (1/100 dilution in PBS) at 4°C overnight. Afterward, the slides were rinsed with PBS and then cultivated with goat anti-mouse HRP-IgG (1/200 dilution in PBS) at 37°C for 1 h before being counterstained with hematoxylin for 10 s. Last, the slides were analyzed under a microscope. Statistical Analysis {#S2.SS13} -------------------- All the experiments data are expressed as the mean ± Standard Deviation (SD). The statistical analysis was conducted by Student's *t* test, and *p* value \< 0.05 was considered as to be of statistical significance. Results {#S3} ======= Construction of PCV3 Infectious Clone {#S3.SS1} ------------------------------------- The PCV3 infectious clone was achieved by self-cyclization of the genomic DNA followed digestion of recombinant genomic DNA by *Hin*dIII, gel extraction and ligation at the created *Hin*dIII site by T4 DNA ligase ([Figure 2A](#F2){ref-type="fig"}). The identified cyclized PCV3 DNA was transfected into the 3D4/21 cell line with Lipofectamine 3000 (Invitrogen) in accordance with the manufacturer's protocol. {#F2} Nucleic acid was extracted from the rescued viruses analyzed by PCR and RT-PCR. The full-length (2000 bp) PCV3 PCR products were amplified from the DNA extracted from the rescued cells ([Figure 2B](#F2){ref-type="fig"}). The PCV3 Cap gene length of the RT-PCR products amplified by primers PCV3 Cap *Bam*HI and PCV3 Cap *Hin*dIII reached 645 bp ([Figure 2C](#F2){ref-type="fig"}). The identical bands of the expected sizes were achieved; it is therefore revealed that the cyclized PCV3 DNA was detected, and the PCV3 Cap gene was replicated and then transcribed. Identification and Potency Analysis of PCV3 Monoclonal Antibody {#S3.SS2} --------------------------------------------------------------- The figure reveals the successful building of recombinant plasmid pET32a-PCV1 Cap, pET32a-PCV2 Cap, pET32a-PCV3 Cap ([Figure 3A](#F3){ref-type="fig"}). Meantime, recombinant PCV1, PCV2, and PCV3 Cap proteins exhibited successful expression using prokaryotic expression system; then, they were purified ([Figure 3B](#F3){ref-type="fig"}). {#F3} The indirect ELISA method was employed to ascertain the titer of mouse serum and anti-PCV3 Cap protein mAb. As suggested from the results, the titer of mouse serum Number 1 was the optimal, reaching over 1:64,000. The spleen cells of that mice were fused. Seven days later, unfused cells showed the lysis and death, and grape clusters were formed in fused cells. Last, through reactions with PCV1, PCV2, and PCV3 Cap proteins, two specific PCV3 Cap hybridoma cell lines were achieved after triple cloning, purification and screening. The cell lines were nominated as B6 and D17. [Table 2](#T2){ref-type="table"} lists the indirect ELISA results for the supernatants of hybridoma cells. ###### Screening results of ELISA for positive hybridoma cells. Number PCV3 Cap PCV2 Cap PCV1 Cap His -------- ---------- ---------- ---------- ------- 1 0.11 0.052 0.04 0.732 2 0.021 0.039 0.07 0.052 3 0.057 0.037 0.053 0.867 4 0.62 0.509 0.317 0.017 5 0.029 0.022 0.019 0.016 6 0.851 0.052 0.033 0.025 7 0.071 0.026 0.015 0.055 8 0.509 0.221 0.156 0.027 9 0.341 0.309 0.281 0.011 10 0.015 0.016 0.012 0.067 11 0.023 0.015 0.025 0.033 12 0.029 0.028 0.033 0.867 13 0.047 0.029 0.023 0.015 14 0.038 0.024 0.021 0.032 15 0.019 0.028 0.014 0.917 16 0.044 0.036 0.023 0.017 17 0.717 0.016 0.011 0.018 18 0.016 0.012 0.001 0.835 19 0.016 0.023 0.015 0.649 20 0.044 0.036 0.023 0.017 The blue font is the result of an ELISA reaction with only the PCV3 Cap. As suggested from SDS-PAGE results of the two purified antibodies, the antibody was well purified, and the antibody had obvious light and heavy chains ([Figure 3C](#F3){ref-type="fig"}). Meantime, the antibody titer results suggested that the B6 mAb titer was 2 × 10^5^ ([Figure 3D](#F3){ref-type="fig"}), and the D17 mAb titer was 6 × 10^5^ ([Figure 3E](#F3){ref-type="fig"}). According to the results of antibody subclass identification, the B6 mAb was IgG1, and the D17 mAb was IgG2b ([Table 3](#T3){ref-type="table"}). ###### Analysis of monoclonal antibody subtypes. Number 6 17 -------- ------- ------- M 0.051 0.038 G1 0.155 0.001 G2a 0.018 0.027 G2b 0.009 0.230 G3 0.016 0.018 A 0.016 0.015 κ 0.045 0.031 λ 0.034 0.033 Blue font is positive reaction results. Western Blotting Assay {#S3.SS3} ---------------------- The results of the Western blotting assays of anti-PCV3 Cap protein mAb demonstrated that the purified recombinant PCV3 Cap protein displayed a specific band at 38 kDa. The anti-PCV3 Cap protein mAb could react with PCV3 Cap to reach the target band ([Figure 4A](#F4){ref-type="fig"}), suggesting that the PCV3 Cap obtained here exhibited prominent immunogenicity. The rescued PCV3 virus was tested by Western blotting assays with D17 mAb. The size of bands on the PVDF membrane were nearly 60 kDa obtained using the DAB kit, whereas no specific bands were detected in the supernatant of normal cells ([Figure 4B](#F4){ref-type="fig"}). The results revealed that rescued viruses could be identified by PCV3 Cap mAb. {#F4} Immunofluorescence Detection {#S3.SS4} ---------------------------- The 3D4/21 cells were transfected with the cyclized PCV3 DNA, and the morphology was observed within 72 h. It was reported that the built infectious cloning of PCV3 may lead to the pathological variations in 3D4/21 ([Figure 5A](#F5){ref-type="fig"}). By the immunofluorescence assay, viral antigens were identified by the PCV3 specific antibodies. D17 mAb was observed to be able to recognize PCV3 in cells, represented as red under a fluorescence microscope ([Figure 5B](#F5){ref-type="fig"}), while no fluorescence was exhibited by control cells transfected in parallel as the mock control. The results revealed the presence of rescued PCV3 as determined. {#F5} The nuclei of PCV3-infected cells and normal cells were stained with a DAPI kit (Beyotime, Shanghai, China). The results of nuclear staining are presented in [Figure 5C](#F5){ref-type="fig"}. By overlapping the pictures of the same location ([Figure 5D](#F5){ref-type="fig"}), the PCV3 Cap was reported to be primarily expressed in the nucleus of cells. Viral Titration {#S3.SS5} --------------- To assess the replication capabilities of the rescued PCV3 virus. 3D4/21 cells were inoculated the PCV3 that has been constructed from above for six passages. The virus titers increased during the second passage and reached up to 10^--4.8^ TCID~50~/ml at the sixth passage ([Figure 6](#F6){ref-type="fig"}). {#F6} Rescued PCV3 Virus Pathogenicity Analysis {#S3.SS6} ----------------------------------------- During the experiment, no clinical variations were observed in the control group of mice. On the third day after infection, the infected group of mice was slightly depressed. Individual mice were reluctant to move, and no other symptoms occurred apparently. The tissues and organs of the control group and the experimental group were basically normal, and no significant variations were observed ([Figure 7A](#F7){ref-type="fig"}). {#F7} Histopathological examination showed slight lesions in the tissues and organs of the control mice. In the mice in the infected group, significant changes in lung and heart tissues, the alveolar epithelial cells were proliferated in the local area of the lungs, and congestion took place at the edge of the local lobules, and other tissues varied slightly ([Figure 7B](#F7){ref-type="fig"}). PCV3 Antigen Detection in Tissues {#S3.SS7} --------------------------------- For 6-week-old KM mice, lung, liver, kidney, spleen, and heart samples of all the mice subsequently underwent immunohistochemical staining of PCV3 antigen. Various tissues and organs developed similar distributions of PCV3 antigen. As shown in [Figure 8](#F8){ref-type="fig"}, in the lung, the bronchial epithelial cell surface, interstitial vascular contents, alveolar exudate, dust cells, and septal cells all had robust positive reactions. Myocardial fibers showed varied positive reactions, and necrotic tissues and vascular contents had significantly positive reactions. {#F8} Discussion {#S4} ========== Previous studies have shown that PCV2 is the main pathogen causing PWMS and PDNS ([@B25]; [@B9]). However, since first reported in 2016 ([@B22]), increasing epidemiological data ([@B15]; [@B26]; [@B36]; [@B6]; [@B7]; [@B13]; [@B14]) suggested that PCV3 has been extensively distributed in pig farms worldwide. Researchers found that PCV3 can also cause PDNS ([@B35]), and is often infected in combination with PCV2 ([@B29]; [@B31]). As indicated, PCV3 was a potential pathogenic factor leading to an increase in the morbidity and mortality of swine; it has posed threats to the swine industry, causing significant economic losses. However, currently there is a lack of commercial vaccines for PCV3, the infection mechanism of PCV3 has not been clarified, and the gene structure and function of PCV3 have not been fully analyzed. Separating and obtaining standard virus strains is the key to solving the above problems. Although researchers have constructed PCV3 infectious clones ([@B11]), the methods for preparing PCV3 infectious clones mostly rely on the application of eukaryotic expression vectors, and use PK-15 cell line for virus packaging and rescue. It has been reported that there is a long-term presence in the PCV1 group ([@B4]), suggesting that there are certain risks in the current use of porcine cells and their use as animal models. Accordingly, PCV3 infectious clones prepared with PK-15 cells might be a mixed-gene PCV and an existed risk of contamination of foreign virus genomes. This study based on the clinical report that PCV3 can cause bronchial interstitial pneumonia in pigs ([@B22]), 3D4/21 cell line was used for the first time as the packaging cell line of PCV3, confirming that PCV3 can be replicated and packaged in 3D4/21 cells. According to the genome structure of circovirus and its replication mechanism, it can be known that the circovirus itself has a replication initiation element, and the circovirus does not need to resort to an exogenous promoter when infecting the body. Therefore, according to the biological characteristics of circovirus, this research obtained PCV3 circular DNA *in vitro* to prepare PCV3 infectious clones. This will provide practical experience for the construction of other circular viruses and the construction of unknown circular virus infectious clones in the future. The experimental results of cyclization demonstrated that large loops might be synthesized from the linearized PCV3 DNA by connecting several fragments, whereas there were also some end-to-end connections between the linearized PCV3 DNA, thereby leading to the formation of a closed loop. The PCV3 DNA containing several closed rings was transfected into cells, and the large rings in the PCV3 DNA might not be transcribed and expressed in cells or digested by intracellular digestive enzymes. In the present study, specific primers were designed, covering the PCV3 M1 and PCV3 L2 primers, the PCV3 complete genome amplification primers; PCV3 L3 primer was the reverse complementary sequence of PCV3 L2, and PCV3 M2 primer referred to the nucleotide sequence next to the primer binding site of PCV3 L2. The PCV3 Cap *Bam*HI and PCV3 Cap *Hin*dIII primers were designed as the reverse complement of the negative chain of PCV3 Cap gene. Subsequently, specific primers were conceived following the reverse complementary DNA sequence. The experimental results proved that cyclized PCV3 DNA was transfected into cells and then transcribed and duplicated after several blind passages. In order to reduce the infection of PCV3 by the PCV1 virus carried by the pigs themselves, this study looked for other animal models to evaluate the pathogenicity of PCV3 infectious clones. Studies have shown that PCV3 has the ability of cross-host transmission of infected mice ([@B12]), so this study uses SPF Kunming mice as a model of PCV3 infection. The results show that PCV3 can infect myocardium and lung tissue of Kunming mice, which is consistent with the current report that PCV3 can cause myocarditis in pigs ([@B24]). The molecule that PCV3 causes myocarditis in piglets still needs further exploration. Studies have shown that PCV3 can cause the expression level of IFN-γ to be up-regulated in piglet, and IFN-γ has been found to be related with viral myocarditis in the study of Coxsackie B3 virus (CVB3) causing myocarditis ([@B37]). This provides a reference for further study on the mechanism of PCV3 induced myocarditis in piglets. Conclusion {#S5} ========== In this study, our results demonstrated that infectious cloning of PCV3 were successfully obtained, rescued PCV3 virus could infect the heart muscle and lung of KM mice, and mAbs against PCV3 Cap were produced. The mentioned results can lay a theoretical and practical foundation for the accurate diagnosis of PCV3, the development of antibody drugs and subunit vaccines, as well as the analysis of relevant pathogenic sites. However, subsequent studies should still assess the significance of the findings to the biology and immunopathology of PCV3 infection. Data Availability Statement {#S6} =========================== The read sequences for this study were deposited to the NCBI Sequence Read Archive (SRA), accession number: [MH107162.1](MH107162.1). Ethics Statement {#S7} ================ The animal study was reviewed and approved by the Ethics Committee of South China Agricultural University, Guangzhou, China. Author Contributions {#S8} ==================== ZJ, GZ, and ML designed the study. JW, MJ, and WB performed the experiments. YX and CL analyzed the data. ZJ and JW wrote the manuscript. ML and YX revised the manuscript. All authors reviewed the manuscript. Conflict of Interest {#conf1} ==================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. **Funding.** This study was supported by the National Key R&D Programmes, Ministry of Science and Technology of China, Grant/Award Number 2017YFF0210202. [^1]: Edited by: Jianming Qiu, The University of Kansas Medical Center, United States [^2]: Reviewed by: Wei Zou, University of Michigan, United States; Jun Han, China Agricultural University, China [^3]: ^†^These authors have contributed equally to this work [^4]: This article was submitted to Virology, a section of the journal Frontiers in Microbiology

Background {#Sec1} ========== The development of resistance to cytotoxic agents represents a major concern in cancer chemotherapy. Multi-drug resistance (MDR) is associated with over-expression of transmembrane glycoprotein (P-gp) which functions as a drug efflux pump, reducing the intracellular levels of cytotoxic drugs (Juranka et al. [@CR29]). P-gp belongs to the ATP-binding cassette (ABC) transport proteins, which also include the multi-drug resistance associated protein 1 (MRP1) (Shen et al. [@CR63]; Biedler and Spengler [@CR4]; Efferth et al. [@CR15]), or the breast cancer resistance protein (BCRP/ABCG2) (Shen et al. [@CR63]). The oncogene epidermal growth factor receptor (EGFR) (Biedler and Spengler [@CR4]; Efferth et al. [@CR15], [@CR16]) and the deletions or inactivation of tumor suppressor gene p53 (el-Deiry [@CR17]) have also been involved in MDR mechanism of cancer cells. Overcoming this resistance requires a permanent search of new antineoplastic agents. In the past, natural products from plant kingdom have revealed a high potential as cytotoxic drug reservoir (Kuete and Efferth [@CR33]). According to the World Health Organization, about 80 % of the population of developing countries relies on traditional medicines, mostly plant drugs, for their primary health care needs (FAO [@CR20]). It has also been reported that modern pharmacopoeia still contain at least 25 % drugs derived from plants and many others which are synthetic analogues (FAO [@CR20]). Therefore, fighting cancers with botanicals represents a very promising alternative, especially regarding the diversity of plant's secondary metabolites. African flora has previously been found to be very fruitful in the search of antiproliferative molecules. Many compounds including xanthones: 8-hydroxycudraxanthone G, morusignin I, cudraxanthone I (Kuete et al. [@CR38]), and xanthone V1 (Kuete et al. [@CR35]), benzophenones: guttiferone E and isogarcinol (Kuete et al. [@CR39]), quinone: 2-acetylfuro-1,4-naphthoquinone (Kuete et al. [@CR35]), flavonoids: 4-hydroxylonchocarpin, 6,8-diprenyleriodictyol (Kuete et al. [@CR36]), 2′,4′-dihydroxy-3′,6′-dimethoxychalcone and 4′-hydroxy-2′,6′-dimethoxychalcone (Kuete et al. [@CR41]; Dzoyem et al. [@CR13]) and alkaloids: isotetrandrine (Kuete et al. [@CR44]) and montrofoline (Kuete et al. [@CR45]) displayed good antiproliferative effects against various cancer cell lines. In a collaborative research programme between the Council for Scientific and Industrial Research (CSIR) in South Africa and the National Cancer Institute (NCI) in the USA, several South African plant extracts exhibited anticancer activity against a panel of three human cell lines (breast MCF7, renal TK10 and melanoma UACC62) (Fouche et al. [@CR21], [@CR22]). African medicinal plants such as *Fagara heitzii* (Dzoyem et al. [@CR14]), *Echinops giganteus*, *Xylopia aethiopica*, *Piper capense*, *Imperata cylindrica* (Kuete et al. [@CR37]), *Beilschmiedia acuta*, *Clausena anisata* (Kuete et al. [@CR40]) also displayed good cytotoxicity towards drug-sensitive and drug-resistant cancer cell lines. In our ongoing search of anticancer products from African medicinal flora, we designed the present study to investigate the cytotoxicity of 11 plants traditionally used to manage cancer or disease states bearing relevance to cancer or cancer-like symptoms, such as immune and skin disorders, inflammatory, infectious, parasitic and viral diseases (Kuete et al. [@CR44]). The study was extended to the evaluation of the ability of the three most active extracts from two medicinal plants, *Annona muricata* Lin. (Annonaceae) and *Passiflora edulis* Sims (Passifloraceae) to alter the cell cycle distribution, caspases activity, mitochondrial membrane potential (MMP) and to increase the production of reactive oxygen species (ROS) in leukemia CCRF--CEM cells. Methods {#Sec2} ======= Plant material and extraction {#Sec3} ----------------------------- All medicinal plants tested are traditionally used in the management of cancer or disease states with symptoms related to cancer. Plants were collected in different regions of Cameroon in January 2012. They included *Pachypodanthium staudtii*, *Alchornea floribunda*, *Annona muricata*, *Canarium schweinfurthii*, *Hibiscus esculentus*, *Colocasia esculenta*, *Moringa oleifera*, *Triumphetta pentandra*, *Xanthosoma mafaffa*, *Euphorbia prostata* and *Passiflora edulis*. The plant parts investigated are shown in Table [1](#Tab1){ref-type="table"}. The plants were identified at the National Herbarium (Yaoundé, Cameroon), where voucher specimens were deposited under the reference numbers shown in Table [1](#Tab1){ref-type="table"}. The air-dried and powdered plant material (50 g) was soaked in methanol (200 mL) for 48 h, at room temperature. The methanol extract was concentrated in vacuum under reduced pressure at 68 °C to give the crude extract. This extract was completely dried at room temperature, then conserved at 4 °C until further use.Table 1General information and reports on evidence of biological activities and chemistry of the studied plantsSpecies (family); voucher number^a^Traditional usesParts used (%yield)^b^Bioactive or potentially bioactive componentsBioactivity of crude extract*Alchornea floribunda* Müll. Arg. (Euphorbiaceae)\ 4595/HNCTreatment of [bacterial]{.ul} and parasitic infections, painful urination in children (Adjanohoun et al. [@CR2]; Jiofack et al. [@CR28]), urinary, respiratory and intestinal problems, pains in the heart, diarrhoea, ovarian problems, stomach ailments and intestinal disorders (Siwe Noundou et al. [@CR66]), [trypanosomiasis]{.ul}, urinary, respiratory and intestinal disorders (Musuyu Muganza et al. [@CR49]; Mesia et al. [@CR47]), [inflammation]{.ul} (Okoye et al. [@CR58])Bark (18.91 %) and leaves (4.56 %)Eugenol, cadinol, nanocosaine, ethyl iso-allocholate, 3-acetoxy-7,8-epoxylanostan-1-ol (Okoye et al. [@CR58])Antibacterial activities of crude against *Bc*, *Ef*, *Ec*, *Sa*, *Kp*, *Mc*, *Pm*, *Ss* (Siwe Noundou et al. [@CR66]); topical anti-inflammatory effects (Okoye et al. [@CR58])*Annona muricata* Lin. (Annonaceae); 18681/SRF/CamTreatment of wounds and insomnia; [antiparasitic]{.ul}, insecticidal (Rajeswari [@CR60])Leaves (4.50 %), seeds (9.15 %), pericarp (5.17 %)Epomuricenins-A and B, montecristin, cohibins-A and B, muridienins-1 and 2, muridienins-3 and 4, muricadienin and chatenaytrienins-1, 2 and 3 and sabadelin, murihexol, donhexocin, annonacin A and Annonacin B (Rajeswari [@CR60])Antimicrobial activities of aqueous, ethanol and methanol extracts against *Sa*, *Vc*, *Ec*, *Se*, *Lv* and *On* (Vieira et al. [@CR68]) and *Pv*, *Sp*, *Bs*, *St*, *Kp*, *Ea* (Rajeswari [@CR60]), *Lb*, *Lp*, *Hv* (Rajeswari [@CR60]), *Ec*, *Ea*, *Kp*, *Ps* (Dzotam et al. [@CR11])*Canarium schweinfurthii* Engl. (Burceraceae); 16929/SRF/CamTreatment of [malaria]{.ul}, constipation, diarrhea, rheumatism and [sexually transmitted diseases]{.ul} (Koudou et al. [@CR32])Fruits (0.78 %)Saponins, cardiac glycosides, tannins, flavonoids and steroids (Ngbede et al. [@CR52])Antimicrobial activities of EO against *Bc*, *Ef*, *Ec*, *Li*, *Se*, *Sd*, *Sa*, *Pm*, *Sc* and *Ca* (Obame et al. [@CR55])*Colocasia esculenta* (L.) Schott (Araceae); 42352/HNCTreatment scorpion and snake bite (Nakade et al. [@CR50]), [infectious diseases]{.ul} (Dzotam et al. [@CR11])Leaves (6.25 %)Quinones, alkaloids, saponins, tannins, phenols, terpenoids, glycosides and steroids (Nakade et al. [@CR50])Antimicrobial activities of ethyl acetate extract against *St*, *Kp*, *Pa*, *Sp*, *Bs*, *Pv*, *Ec* (Nakade et al. [@CR50]) aqueous and methanolic extracts: (Q) *Vspp* (Lee et al. [@CR46])*Euphorbia prostata* W. Ait. (Euphorbiaceae) 33585/HNCTreatment of bronchial ashma, diarrhea, [skin diseases]{.ul} (Shrama and Tripathi [@CR64])Whole plant (13.82 %)Flavonoids, tannins and phenolic acid; gallic acid, apigenin, luteolin (Gupta [@CR25])Crude extract has cardiac depresent and hypotensive actions (Shrama and Tripathi [@CR64]), showed effects on early grades of hemorrhoids (Gupta [@CR25])*Hibiscus esculentus*L. (Tiliaceae); 8537/SRF/CamTreatment of [cancer]{.ul}, i[nflammation]{.ul}, ulcer, analgesic, hyperglycemia (Daly [@CR9]; Uraku et al. [@CR67])Fruits (2.98 %)Alkaloids, polyphenols, flavonoids, triterpenes, sterols (Dzotam et al. [@CR11])Antimicrobial activities of crude extract on *St*, *Shigella* and *Ec*, *Ea*, *Kp*, *Ps* (Dzotam et al. [@CR11]; Nwaiwu et al. [@CR54])*Moringa oleifera* Lam. (Moringaceae); 49178/HNCTreatment of [cancer]{.ul}, dental caries, [syphilis]{.ul}, typhoid, diarrhea, epilepsy (Fuglie [@CR23]), [fever]{.ul}, [HIV-AIDS]{.ul} (Abrams et al. [@CR1])Leaves (3.95 %)4-(4′-*O*-acetyl-α-[l]{.smallcaps}-rhamnopyranosyloxy)benzylisothiocyanate, 4-(-[l]{.smallcaps}-rhamnopyranosyloxy)benzylisothiocyanate, niazimicin, pterygospermin, benzylisothiocyanate and 4-(α-[l]{.smallcaps}-rhamnopyranosyloxy)benzylglucosinolate (Fahey [@CR18])Antimicrobial activities of aqueous and ethanol extracts of seeds against *Sa*, *Vc*, *Ec*, *Se*, *Lv* and *On* (Viera et al. [@CR69])*Pachypodanthium staudtii* Engl & Diels (Annonaceae), 23170 SFR/CamTreatment of [cancer]{.ul}, Chest pain (Irvine [@CR27]); bronchitis (Bouquet and Debray [@CR6]) and oedema (Ngadjui et al. [@CR51]).Leaves (10 %), bark (9.4 %) and roots (6.25 %)Pachypodol, 2,4,5-Trimethoxystyrene, Pachypophyllin, pachypostaudins A and B (Ngadjui et al. [@CR51]); Sabinene, β-elemene, *E*-β-caryophyllene, β-selinene, β-bisabolene, δ-cadinene, 2,4,5-trimethoxy-1-vinylbenzene (Yapi et al. [@CR70]).Methanol extract against *Ec*, *Ea*, *Ecl*, *Kp*, *Ps* (Fankam et al. [@CR19])*Passiflora edulis* Sims (Passifloraceae); 65104/HNCTreatment of [cancer]{.ul}, [fungal infections]{.ul}, [inflammation]{.ul}, insomnia and anxiety, antihypertensive (Ichimura et al. [@CR26]), gastric trouble (Silva et al. [@CR65]), antioxidant (Kannan et al. [@CR30])Fruit (3.92 %); fruit pericarp (2.73 %)Ionone-I, ionone-II, megastigma-5,8-dien-4-1, megastigma-5,8(*Z*)-diene-4-1, 4,4*a*-Epoxy-4, 4*a*-dihydroedulan, 3-hydroxyedulan, edulan-I, edulan-II, passifloric acid methyl ester (Kannan et al. [@CR30])Antimicrobial activities of methanol extract against *Ec*, *Kp*, *Ea*, *Pa*, *Ps*, *Sa*, *Ef*, *Bs*, *Ec*, *Pv* and *St* (Kannan et al. [@CR30])*Triumphetta pentandra* A.Rich. (Tiliaceae); 9014/SRF/CamInduce fertility and implantation of the fetus (Okoli et al. [@CR57]; Ngondi et al. [@CR53]), treat [infectious diseases]{.ul} (Dzotam et al. [@CR11])Leaves (5.50 %)Triumfettamide, triumfettoside, heptadecanoic acid, β-sitosterol glucopyranoside, friedeline, lupeol, betuline, maslinic acid, 2-hydroxyoleanolicacid and the mixture of stigmasterol and β-sitosterol (Sandjo et al. [@CR62]; Sandjo and Kuete [@CR61])Antimicrobial activities of methanol extract against *Ec*, *Ea*, *Kp*, *Ps* (Dzotam et al. [@CR11])*Xanthosoma mafaffa* (L.) Schott (Araceae); 18675/SRF/CamTreatment of [infectious diseases]{.ul}; osteoporosis (Dzotam et al. [@CR12]; Cancer in Africa [@CR7])Leaves (4.30 %)Polyphenols, coumarins, tannins, triterpenes, sterols, saponins (Dzotam et al. [@CR12])Antimicrobial activities of methanol extract against *Ec*, *Ea*, *Kp* (Dzotam et al. [@CR12])EO: essential oil; *Bc*: *Bacillus cereus*; *Bs*: *Bacillus subtilis*; *Ca*: *Candida albicans*; *Ec*: *Escherichia coli*; *Ea*: *Enterobacter aerogenes*; *Ecl*: *Enterobacter cloacae*; *Ef*: *Enterococcus faecalis*; HIV-AIDS: human immunodeficiency virus-acquired immuno deficiency syndrome; *Hv*: *Herpes virus*; *Kp*: *Klebsiella pneumoniae*; *Lb*: *Leishmania braziliensis*; *Lp*: *Lieshmaniapanamensis*; *Lv*: *Litopenaeusvannmaei*; *Mc*: *Moraxella catarrhalis*; *On*: *Oreochromis nicoticus*; *Pa*: *Pseudomonas aeruginosa*; *Li*: *Listeria innocua*; *Pm*: *Proteus mirabilis*; *Pv*: *Proteus vulgaris*; *Ps*: *Providencia stuartii*; *Sa*: *Staphylococcus aureus*; *Sc*: *Staphylococcus camorum*; *Sd*: *Shigelladysenteriae*; *Se*: *Salmonella enterica*; *Ss*: *Staphylococcus saprophyticus*; *Sp*: *Streptococcus pyogenes*; *St*: *Salmonella typhi*; *Vc*: *Vibrio cholerae*; *Vspp*: Vibrio species; underline: disease states bearing relevance to cancer or cancer-like symptoms^a^(HNC): Cameroon National Herbarium; (SRF/Cam): Société des Réserves Forestières du Cameroun^b^Yield calculated as the ratio of the mass of the obtained methanol extract/mass of the plant powder Chemicals {#Sec4} --------- Doxorubicin 98.0 % and vinblastine ≥96 % from Sigma-Aldrich (Munich, Germany) were provided by the University Pharmacy of the Johannes Gutenberg University (Mainz, Germany), dissolved in phosphate buffer saline (PBS; Invitrogen, Eggenstein, Germany) at a concentration of 10 mM and used as positive control drugs. Geneticin \>98 % (Sigma-Aldrich), stored at a stock concentration of 72.18 mM was used to maintain the resistance patterns of MDR carcinoma cell lines. Cell cultures {#Sec5} ------------- The cell lines used in the present study included drug-sensitive leukemia CCRF--CEM and multidrug-resistant P-glycoprotein-over-expressing subline CEM/ADR5000 cells (Efferth et al. [@CR15]; Kimmig et al. [@CR31]; Gillet et al. [@CR24]), breast cancer MDA-MB-231-pcDNA3 cells and its resistant subline MDA-MB-231-*BCRP* clone 23 (Doyle et al. [@CR10]), colon cancer HCT116 (*p53*^+*/*+^) cells and its knockout clone HCT116 (*p53*^−*/*−^), glioblastoma U87MG cells and its resistant subline U87MG.Δ*EGFR* (Kuete et al. [@CR38], [@CR39]; Dzoyem et al. [@CR14]). Leukemia CCRF--CEM and CEM/ADR5000 cells were maintained in RPMI 1640 medium (Invitrogen) supplemented with 10 % fetal calf serum in a humidified 5 % CO~2~ atmosphere at 37 °C. Sensitive and resistant cells were kindly provided by Dr. J. Beck (Department of Pediatrics, University of Greifswald, Greifswald, Germany). Breast cancer cells transduced with control vector (MDA-MB-231-pcDNA3) or with cDNA for the breast cancer resistance protein *BCRP* (MDA-MB-231-*BCRP* clone 23) were maintained under standard conditions as described above for CCRF--CEM and CEM/ADR5000 cells. Human wild-type HCT116 (*p53*^+*/*+^) colon cancer cells as well as knockout clones HCT116 (*p53*^−^*/*^−^) derived by homologous recombination were a generous gift from Dr. B. Vogelstein and H. Hermeking (Howard Hughes Medical Institute, Baltimore, MD). Human glioblastoma multiforme U87MG cells (non-transduced) and U87MG cell line transduced with an expression vector harboring an epidermal growth factor receptor (*EGFR*) gene with a genomic deletion of exons 2 through 7 (U87MG.Δ*EGFR*) were kindly provided by Dr. W. K. Cavenee (Ludwig Institute for Cancer Research, San Diego, CA). MDA-MB-231-*BCRP*, U87MG.Δ*EGFR* and HCT116 (*p53*^−^*/*^−^) were maintained in DMEM medium containing 10 % FBS (Invitrogen) and 1 % penicillin (100 U/mL)-streptomycin (100 μg/mL) (Invitrogen) and were continuously treated with 800 ng/mL and 400 µg/mL geneticin, respectively. The multidrug resistance profile of these cell lines has been reported (Doyle et al. [@CR10]). Human liver hepatocellular carcinoma HepG2 and the AML 12 normal heptocytes were obtained from ATCC (USA). The above medium without geneticin was used to maintained MDA-MB-231, U87MG, HCT116 (*p53*^+*/*+^), HepG2 and AML 12 cell lines. The cells were passaged twice weekly. All experiments were performed with cells in the logarithmic growth phase. Resazurin reduction assay {#Sec6} ------------------------- The cytotoxicity of the tested samples was performed by resazurin reduction assay as previously described (Kuete et al. [@CR39]; O'Brien et al. [@CR56]). The assay is based on reduction of the indicator dye, resazurin, to the highly fluorescent resorufin by viable cells. Non-viable cells rapidly lose the metabolic capacity to reduce resazurin and thus produced no fluorescent signal. Briefly, adherent cells were detached by treatment with 0.25 % trypsin/EDTA (Invitrogen) and an aliquot of 1 × 10^4^ cells was placed in each well of a 96-well cell culture plate (Thermo Scientific, Germany) in a total volume of 200 µL. Cells were allowed to attach overnight and then were treated with different concentrations of the studied sample. For suspension cells, aliquots of 2 × 10^4^ cells per well were seeded in 96-well-plates in a total volume of 100 µL. The studied sample was immediately added in varying concentrations in an additional 100 µL of culture medium to obtain a total volume of 200 µL/well. After 24 or 48 h, 20 µL resazurin (Sigma-Aldrich, Germany) 0.01 % w/v in ddH~2~O was added to each well and the plates were incubated at 37 °C for 4 h. Fluorescence was measured on an Infinite M2000 Pro™ plate reader (Tecan, Germany) using an excitation wavelength of 544 nm and an emission wavelength of 590 nm. Each assay was done twice, with six replicates each. The viability was evaluated based on a comparison with untreated cells. IC~50~ values representing the sample's concentrations required to inhibit 50 % of cell proliferation were calculated from a calibration curve by linear regression using Microsoft Excel (Kuete et al. [@CR35]; Dzoyem et al. [@CR13]). In a preliminary step, all samples were tested against the sensitive CCRF--CEM cells at various concentrations ranging from 0.16 to 80 µg/mL (crude extracts) or 0.08 to 10 µg/mL (doxorubicin), and samples displaying IC~50~ values below 20 µg/mL were further investigated in 8 other tumor cell lines as well as in normal AML12 hepatocytes. Doxorubicin was used as positive control, while dimethylsulfoxide (DMSO) used to dissolve the samples was used as negative control. The highest concentration of DMSO was less than 0.4 %. Flow cytometry for cell cycle analysis and detection of apoptotic cells {#Sec7} ----------------------------------------------------------------------- Extracts from *Passiflora edulis* fruit (PEF), *Annona muricata* leaves (AML), *Annona muricata* seeds (AMS) that displayed the best cytotoxicity as well as doxorubicin were used to treat CCRF--CEM cells (1 × 10^6^) at their IC~50~ values. Thus, CCRF--CEM cells were cultured in RPMI medium as described above, in the presence of each sample at a concentration corresponding to the IC~50~ values obtained in the cell line. The cell cycle was then analyzed after incubation for 24, 48 and 72 h. All reagents, experimental conditions and apparatus were identical to those previously reported (Kuete et al. [@CR38]; Dzoyem et al. [@CR14]). Briefly, cell cycle analysis was performed by flow cytometry using Vybrant^®^ DyeCycle™ (Invitrogen, Darmstadt, Germany). Cells were measured after Vybrant^®^ DyeCycle™ Violet staining (30 min at 37 °C) on a LSR-Fortessa FACS analyzer (Becton--Dickinson, Heidelberg, Germany) using the violet laser. Vybrant^®^ DyeCycle™ Violet stain was measured with 440 nm excitation. Cytographs were analyzed using FlowJo software (Celeza, Switzerland). All experiments were performed at least in triplicate. Caspase-Glo 3/7, caspase-Glo 8 and caspase-Glo 9 assay {#Sec8} ------------------------------------------------------ The influence of extracts on caspase 3/7, caspase 8 and caspase 9 activity in leukemia CCRF--CEM cell line was detected using Caspase-Glo 3/7, Caspase-Glo 8 and Caspase-Glo 9 Assay kits (Promega, Germany). Cells cultured in RPMI medium were seeded in 96-well plates and treated with the sample (2 × IC~50~; IC~50~) or DMSO (solvent control). After 6 h incubation in a humidified 5 % CO~2~ atmosphere at 37 °C, 100 µL of caspase reagent were added to each well, mixed and incubated for 1 h at room temperature. Luminescence was measured using well Infinite M2000 Pro™ instrument (Tecan). Caspase activity was expressed as percentage relative to the untreated control (Kuete et al. [@CR42]). Analysis of mitochondrial membrane potential (MMP) {#Sec9} -------------------------------------------------- The effects of extracts on the MMP were analyzed by 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbocyanine iodide) (JC-1; Biomol, Germany) staining (Kuete et al. [@CR40]). JC-1 is a dye that can selectively enter into mitochondria and exhibits an intense red fluorescence in healthy mitochondria with normal membrane potentials. In cells with reduced MMP, the red fluorescence disappears. Briefly, 1 × 10^6^ CCRF--CEM cells treated at different concentrations with PEF, AML, AMS or vinblastine for 24 h were incubated with JC-1 staining solution according to the manufacturer\`s protocol for 30 min. Subsequently, cells were measured in a LSR-Fortessa FACS analyzer (Becton--Dickinson). For each sample, 1 × 10^4^ cells were counted. The JC-1 signal was measured with 561 nm excitation (150 mW) and detected using a 586/15 nm bandpass filter. The samples signal was analyzed with 640 nm excitation (40 mW) and detected using a 730/45 nm bandpass filter. All parameters were plotted on a logarithmic scale. Cytographs were analyzed using FlowJo software (Celeza, Switzerland). All experiments were performed in triplicate. Measurement of reactive oxygen species (ROS) by flow cytometry {#Sec10} -------------------------------------------------------------- The 2′,7′-Dichlorodihydrofluorescein diacetate (H~2~DCFH-DA) (Sigma-Aldrich, Germany) is a probe used for the highly sensitive and quantifiable detection of ROS. The non-fluorescent H~2~DCFH-DA diffuses into the cells and is cleaved by cytoplasmic esterases into 2′,7′-dichlorodihydrofluorescein (H~2~DCF) which is unable to diffuse back out of the cells. In the presence of hydrogen peroxide, H~2~DCF is oxidized to the fluorescent molecule dichlorofluorescein (DCF) by peroxidases. The fluorescent signal emanating from DCF can be measured and quantified by flow cytometry, thus providing an indication of intracellular ROS concentration (Kuete et al. [@CR37]; Bass et al. [@CR3]; Cossarizza et al. [@CR8]). Briefly, 2 × 10^6^ CCRF--CEM cells were resuspended in PBS and incubated with 2 µM H~2~DCFH-DA for 20 min in the dark. Subsequently, cells were washed with PBS and resuspended in RPMI 1640 culture medium containing different concentrations of PEF, AML, AMS or DMSO (solvent control), or hydrogen peroxide (H~2~O~2~; positive control). After 24 h of incubation, cells were washed and suspended in PBS. Subsequently cells were measured in a FACSCalibur flow cytometer (Becton--Dickinson, Germany). For each sample 1 × 10^4^ cells were counted. DCF was measured at 488 nm excitation (25mW) and detected using a 530/30 nm bandpass filter. All parameters were plotted on a logarithmic scale. Cytographs were analyzed using FlowJo software (Celeza, Switzerland). All experiments were performed in triplicate. Results {#Sec11} ======= Cytotoxicity of the studied samples {#Sec12} ----------------------------------- In this study, we first screened the cytotoxicity of 17 crude extracts belonging to 11 plants towards drug-sensitive CCRF--CEM leukemia cells. The results are shown in Table [2](#Tab2){ref-type="table"}. All tested extracts had IC~50~ values below 80 µg/mL. Ten extracts from five plants including *Alchornea floribunda* bark (AFB), *Annona muricata* fruit pericarp (AMP), leaves (AML) and seeds (AMS), *Euphorbia prostata* whole plant (EPW), *Pachypodanthium staudtii* bark (PSB), leaves (PSL) and roots (PSR), and *Passiflora edulis* fruit pericarp (PEP) and fruit (PEF) displayed IC~50~ values below 20 µg/mL in CCRF--CEM cells (Table [2](#Tab2){ref-type="table"}). These extracts were further selected for IC~50~ determination towards a panel of sensitive and MDR cell lines. The results summarized in Table [3](#Tab3){ref-type="table"} indicate that all selected extracts were also active against P-glycoprotein-over-expressing CEM/ADR5000 leukemia cells with IC~50~ values below 40 µg/mL. IC~50~ values ranged from 10.13 µg/mL (towards CEM/ADR5000 cells) to 72.01 µg/mL (on resistant colon carcinoma HCT116 (*p53*^−*/*−^) cells) for PSR, from 14.97 µg/mL (on CEM/ADR5000 cells) to 65.68 µg/mL (against HCT116 (*p53*^−*/*−^) cells) for PSB, from 18.21 µg/mL (against CEM/ADR5000 cells) to 65.21 µg/mL (on HCT116 (*p5*^+*/*+^) cells) for PSL and from 0.11 µg/mL (towards CCRF--CEM cells) to 108 µg/mL (against CEM/ADR5000 cells) for doxorubicin in the 8 other cancer cell lines studied. Apart from extract from *P. staudtii*, other extracts were less active on carcinoma cells including normal AML12 hepatocytes, with IC~50~ values above 80 µg/mL. Collateral sensitivity (or hypersensitivity: higher toxicity to resistant than to sensitive cells with a degree of resistance below 1) (Kuete et al. [@CR38]) was observed in CEM/ADR5000 cells to PSB (degree of resistance of 0.87-fold) and PSR (0.59-fold) (Table [3](#Tab3){ref-type="table"}). Hypersensitivity of resistant carcinoma cells was also recorded in many cases to PSL, PSB or PSR even though they were moderately active. However, if cross-resistance of CEM/ADR5000 cells to the tested extracts were observed, the degrees of resistance were in all cases lower than that of doxorubicin (Table [3](#Tab3){ref-type="table"}). AMS, AML and PEF had IC~50~ values below 1 and 10 µg/mL in sensitive CCRF/CEM cells and it resistant subline CEM/ADR5000 cells respectively; they were subsequently selected for mechanistic studies.Table 2IC~50~ values of the tested plant extracts towards leukemia CCRF--CEM cells and as determined by the resazurin assayTested plant and partsIC~50~ values (µg/mL)PlantsParts*Alchornea floribunda*Bark (AFB)*18.88* ± *1.65*Leaves46.00 ± 4.26*Annona muricata*Fruit pericarp (AMP)*4.58* ± *0.25*Leaves (AML)*0.57* ± *0.02*Seeds (AMS)*0.36* ± *0.03Canarium schweinfurthii*Fruit38.62 ± 3.69*Colocasia esculenta*Leaves38.19 ± 4.39*Euphorbia prostata*Whole plant (EPW)*18.59* ± *1.12Hibiscus esculentus*Fruit60.79 ± 7.04*Moringa oleifera*Leaves29.79 ± 1.26*Pachypodanthium staudtii*Bark (PSB)*17.22* ± *1.16*Leaves (PSL)*13.59* ± *1.12*Roots (PSR)*17.62* ± *1.18Passiflora edulis*Fruit pericarp (PEP)*3.41* ± *0.55*Fruit (PEF)*0.69* ± *0.13Triumphetta pentandra*Leaves36.28 ± 2.84*Xanthosoma mafaffa*Leaves43.20 ± 0.99Doxorubicin*0.11* ± *0.03*In italics: significant cytotoxic effectTable 3Cytotoxicity of the tested extracts and doxorubicin towards sensitive and drug-resistant cancer cell lines and normal cells as determined by the resazurin assayCell linesSamples, IC~50~ values in µg/mL and degrees of resistance^a^ (in bracket)DoxorubicinAFBAMLAMSAMPEPWPSLPSBPSRPEFPEPCEM/ADR500029.49 ± 1.77 (1.56)*5.25* ± *0.38* (9.29)*6.65* ± *0.22* (18.47)23.70 ± 1.64 (5.17)37.00 ± 2.17 (1.99)*18.21* ± *1.45* (1.34)*14.97* ± *0.97* (0.87)*10.13* ± *0.88* (0.59)*8.20* ± *1.02* (11.88)*18.40* ± *1.42* (5.40)108.00 ± 7.92 (975.60)MDA-MB-231-*pcDNA*\>80\>80\>80\>80\>8052.08 ± 4.9852.66 ± 6.0337.19 ± 2.74\>80\>80*0.61* ± *0.15*MDA-MB-231-*BCRP*\ Degree of resistance\>80\>80\>80\>80\>8061.98 ± 4.31 (1.19)47.27 ± 3.76 (0.90)46.92 ± 4.89 (1.26)\>80\>80*4.33* ± *0.26* (7.12)HCT116 (*p53* ^+*/*+^)\>80\>80\>80\>80\>8065.21 ± 7.1534.35 ± 1.9928.66 ± 1.62\>80\>80*0.78* ± *0.16*HCT116 (*p53* ^−*/*−^)\ Degree of resistance\>80\>80\>80\>80\>8056.97 ± 4.09 (0.87)65.68 ± 4.80 (1.91)72.01 ± 5.26 (2.51)\>80\>80*2.25* ± *0.04* (2.88)U87MG\>80\>80\>80\>80\>8065.21 ± 5.7952.46 ± 5.2224.80 ± 1.36\>80\>80*0.59* ± *0.08*U87MG.Δ*EGFR*\ Degree of resistance\>80\>80\>80\>80\>8068.65 ± 3.48 (1.05)58.70 ± 3.67 (1.12)46.91 ± 3.01 (1.89)\>80\>80*3.38* ± *0.32* (5.76)HepG238.69 ± 3.08 (\>802.07)\>80\>80\>80\>8046.98 ± 3.17 (\>1.70)36.39 ± 3.08 (0.62)37.56 ± 2.17 (0.80)\>80\>80*2.12* ± *0.52* (\>37.74)AML12\ Degree of resistance\>80\>80\>80\>80\>80\>80\>80\>80\>80\>80\>80^a^The degree of resistance was determined as the ratio of IC~50~ value in the resistant divided by the IC~50~ in the sensitive cell line; CEM/ADR5000, MDA-MB-231-*BCRP*, HCT116 (*p53* ^−*/*−^), U87MG.Δ*EGFR* and AML12 were used as the corresponding resistant counterpart for CCRF--CEM (Table [1](#Tab1){ref-type="table"}), MDA-MB-231-*pcDNA*, HCT116 (*p53* ^+*/*+^), U87MG and HepG2 respectively; the tested methanol extracts were from AFB: *Alchornea floribunda* bark; AML: *Annona muricata l*eaves; AMS: *Annona muricata* seeds; AMP: *Annona muricata* fruit pericarp; EPW: *Euphorbia prostata* whole plant; PSL: *Pachypodanthium staudtii* leaves; PSB: *Pachypodanthium staudtii* bark; PSR: *Pachypodanthium staudtii* roots; PEF: *Passiflora edulis* fruit; PEP: *Passiflora edulis* fruit pericarp; in italics: significant activity Cell cycle distribution and apoptosis {#Sec13} ------------------------------------- The best extracts (AMS, AML and PEF) as well as doxorubicin were used to treat CCRF--CEM cells at their IC~50~ values, and the cycle distribution was analyzed. Results depicted in Fig. [1](#Fig1){ref-type="fig"} show dose-dependent and significant modifications of the cell cycle phases after treatment of cells with all samples. Both PEF and AML induced cell cycle arrest in G0/G1 phase while AMS induced cell cycle arrest in S-phase. After treatment with these three extracts, CCRF--CEM cells underwent apoptosis with dose-dependent increases in sub-G0/G1 phase. The percentages of cells in sub-G0/G1 phase varied from 9.31 % (in 24 h) to 48.69 % (72 h), from 8.87 % (in 24 h) to 33.98 % (72 h) and from 11.03 % (24 h) to 21.63 % (72 h) after PEP, AML and AMS treatments respectively, while doxorubicin increased apoptosis in a range of 6.02 % (24 h) to 51.87 % (72 h). The highest percentage of sub-G0/G1 phase in non-treated cells was only 6.42 % after 72 h.Fig. 1Cell cycle distribution of CCRF--CEM leukemia cells treated with extracts from PEF, AML and AMS or doxorubicin. PEF, AML and AMS were tested at 0.69, 0.57 and 0.36 and 8.02 µg/mL respectively while doxorubicin was tested at 0.11 µg/mL corresponding to their IC~50~ Effects on the activity of caspases, MMP and ROS {#Sec14} ------------------------------------------------ After treating CCRF--CEM cells for 6 h at different concentrations of PEF, AML and AMS, no changes of caspase 3/7, caspase 8 and caspase 9 activities were observed. No increase in ROS production was also not found in CCRF--CEM cells treated with the three extracts (data not shown). PEF, AML and AMS induced significant MMP loss in the respective ranges of 35.3 % (1/2-fold IC~50~ treatment) to 46.7 % (2-fold IC~50~), 28.2 % (1/2-fold IC~50~) to 53.8 % (2-fold IC~50~) and 36.6 % (1/2-fold IC~50~) to 51.0 % (2-fold IC~50~) (Fig. [2](#Fig2){ref-type="fig"}). A 48.6 % loss of MMP at 2-fold IC~50~ of vinblastine was previously reported under similar experimental conditions in CCRF--CEM cells (Kuete et al. [@CR38]).Fig. 2Effect of PEF, AML and AMS on the mitochondrial membrane potential in CCRF--CEM cells. *C* control; PEF was tested at 24 h at 0.35 µg/mL (PEF1), 0.69 µg/mL (PEF2), and 1.38 µg/L (PEF3) while AML was tested at 0.29 µg/mL (AML1), 0.57 µg/mL (AML2), and 1.14 µg/mL (AML3) and AMS was tested at 0.18 µg/mL (AML1), 0.36 µg/mL (AML2), and 0.72 µg/mL (AML3) corresponding to 1/2-fold, IC~50~ and 2-fold IC~50~. Data for the positive control, vinblastine in similar experimental conditions were previously reported (Kuete et al. [@CR39]); Loss of MMP (Q1), intact cells (Q2), ruptured cell membrane (Q3 and Q4) Discussion {#Sec15} ========== According to the U.S. National Cancer Institute (NCI) plant screening program, plant extracts with IC~50~ values below of 20 µg/mL following incubation between 48 and 72 h (Boik [@CR5]) are recognized as potential cytotoxic substances. In the present study, multi-factorial drug-resistant cancer cell lines such as leukemia CEM/ADR5000 cells over-expressing P-gp, breast adenocarcinoma MDA-MB-231-*BCRP* clone 23 expressing BCRP, EGFR-transfected U87MG.Δ*EGFR* glioblastoma cells and p53 knockout HCT116 (*p53*^−*/*−^) colon cancer cells (Efferth et al. [@CR15]; Kuete et al. [@CR38], [@CR39], [@CR43]; Kimmig et al. [@CR31]; Gillet et al. [@CR24]; Doyle et al. [@CR10]) were used to determine the cytotoxicity the selected plant extracts. In the first step of the investigations, we carried out a preliminary assays with the sensitive leukemia CCRF--CEM cells. In regards to the NCI threshold, AFB, AMP, AML, AMS, EPW, PSB, PSL, PSR, PEP and PEF (Table [2](#Tab2){ref-type="table"}) displaying IC~50~ values below 20 µg/mL were selected and further tested on a panel of 8 other cell lines. Interestingly, the P-gp over-expressing leukemia CEM/ADR5000 was also sensitive to most of the extracts with IC~50~ value below 20 µg/mL obtained with AML, AMS, PSL, PSB, PSR, PEF and PEP. This suggests that these extracts can be used to manage hematological cancers including resistant phenotypes. Data obtained with AML, AMS and PEP are very interesting as they displayed IC~50~ values below 10 µg/mL in the resistant CEM/ADR5000 cells and even below 1 µg/mL in its sensitive counterpart CCRF--CEM cells. Nonetheless, they were not active in carcinoma cells, clearly indicating their selectivity to leukemia cells. Alteration of MMP has been reported as a mode of apoptosis induction of plant extracts (Kuete and Efferth [@CR34]). AML, AMS and PEP induced MMP loss but no caspase activation nor increase ROS production. Hence, MMP is the main mode of induction of apoptosis of AML, AMS and PEP in CCRF--CEM cells as observed in this study. To the best of our knowledge, the cytotoxicty of the five most active plants, *Alchornea floribunda*, *Annona muricata*, *Euphorbia prostata*, *Pachypodanthium staudtii* and *Passiflora edulis* towards the cell line panel tested in this study is being reported for the first time. Nevertheless, the leaves ethanol extract of *Annona muricata* was reported to have antiproliferative effect against leukemia HL-60 cells with an IC~50~ value of 14 µg/mL, and also induced apoptosis through the loss of MMP with G0/G1 phase cell arrest (Pieme et al. [@CR59]). This is in accordance with data reported herein. The ethyl acetate extract of the leaves of this plant harvested in Malaysia was also found active against colon carcinoma HCT-116 and HT-29 cells with the respective IC~50~ values of 11.43 and 8.98 µg/mL (Zorofchian Moghadamtousi et al. [@CR71]). In the present study, IC~50~ were not detected at up to 80 µg/mL, either indicating that the active constituents of the plant against carcinoma cells might not be well extracted with methanol or that the geographic distribution influences the cytotoxic potential of the plant. Also the methanol extracts of the leaves and fruits of *Passiflora edulis* harvested in Egypt were screened at 100 µg/mL against HCT-116 cells, HepG2 cells as well as against the breast carcinoma MCF-7 cells and lung carcinoma A-549 cells; As results, less than 50 % growth inhibition was recorded (Moustafa et al. [@CR48]), coroborating the low activity obtained with various parts of this plant against carcinoma cells. Conclusions {#Sec16} =========== In this study, ten extracts from five medicinal plants, *Alchornea floribunda*, *Annona muricata*, *Euphorbia prostata*, *Pachypodanthium staudtii* and *Passiflora edulis* had good cytotoxicity against CCRF--CEM leukemia cells and its resistant subline CEM/ADR5000 cells. Their selectivity to these two cell lines, indicates that they can be sources for the development of novel anticancer drugs to fight leukemia. AML, AML and PEF were the most cytotoxic extracts and induced apoptosis in CCRF--CEM cells mediated by loss of MMP. Further phytochemical investigations of these extracts will be done to isolate their active constituents. ABC : adenosine triphosphate-binding cassette AFB : *Alchornea floribunda* bark AML : *Annona muricata l*eaves AMP : *Annona muricata* fruit pericarp AMS : *Annona muricata* seeds BCRP : breast cancer resistance protein DCF : dichlorofluorescein DMSO : dimethylsufoxide EGFR : epidermal growth factor receptor EPW : *Euphorbia prostata* whole plant H~2~DCFH-DA : 2′,7′-dichlorodihydrofluorescein diacetate IC~50~ : inhibitory concentration 50 % JC-1 : 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbocyanine iodide MDR : multi-drug resistant MMP : mitochondrial membrane potential PBS : phosphate buffer saline PEF : *Passiflora edulis* fruit PEP : *Passiflora edulis* fruit pericarp P-gp : P-glycoprotein PSB : *Pachypodanthium staudtii* bark PSL : *Pachypodanthium staudtii* leaves PSR : *Pachypodanthium staudtii* roots ROS : reactive oxygen species VK, JKD, IKV and AGF carried out the study; VK wrote the manuscript; TE supervised the work; VK and TE designed the experiments, TE provided the facilities for the study. All authors read the manuscript and approved the final version. All authors read and approved the final manuscript. Acknowledgements {#FPar1} ================ Authors acknowledge the Cameroon National Herbarium (Yaoundé) for the plant identification. VK is very grateful to the Alexander von Humboldt Foundation for the 2015--2018 Linkage program and 18 months' fellowship in Germany through the ''Georg Foster Research Fellowship for Experienced Researcher'' program. Competing interests {#FPar2} =================== The authors declare that they have no competing interests.

Introduction {#section1-2050313X19827737} ============ Congenital diaphragmatic eventration is an abnormal diaphragmatic elevation caused by insufficient or absent muscularization of the pleuroperitoneal membrane.^[@bibr1-2050313X19827737]^ It is difficult to assess the exact incidence of this abnormality because it is rare and generally diagnosed incidentally on chest radiography. Symptoms of diaphragmatic eventration vary from asymptomatic, mild gastrointestinal disease to life-threatening diaphragm rupture.^[@bibr2-2050313X19827737]^ Both induction and emergence from anesthesia should be smooth to avoid abdominal pressure increase, as this may cause diaphragmatic rupture.^[@bibr3-2050313X19827737]^ If the diaphragm ruptures, it should be rapidly diagnosed and treated. Lung ultrasonography can be used to monitor diaphragmatic movement on a real-time basis.^[@bibr4-2050313X19827737]^ This case report describes the anesthetic management of a pediatric patient with congenital diaphragmatic eventration and perioperative observation of diaphragmatic motion using lung ultrasonography. Case report {#section2-2050313X19827737} =========== A 28-month-old male (height: 87.8 cm, weight: 11.3 kg) was scheduled for excisional biopsy of osteochondroma on the right distal ulna. The patient had a family history of osteochondroma, and the surgery was planned to prevent further deformity of the right arm. The boy was born at a gestational age of 29 weeks and 2 days with a birth weight of 1400 g. He was transferred to the neonatal intensive care unit (NICU) after birth and remained there for 54 days. Haziness was noted in the right lower lung on chest radiography 30 days after the infant's admission to the NICU ([Figure 1(a)](#fig1-2050313X19827737){ref-type="fig"}). Based on this finding, fluoroscopy was performed and the patient was diagnosed with eventration, which persisted until his discharge. At the time of discharge, no abnormal diaphragmatic movement was observed on fluoroscopy despite right hemidiaphragm elevation. After the patient was discharged from the hospital, his clinical course was uneventful and similar to that of other babies of the same age. As part of the pre-operative evaluation, chest radiography showed an abnormal finding indicating eventration of the right diaphragm ([Figure 1(b)](#fig1-2050313X19827737){ref-type="fig"}). Since no pulmonary symptoms were present, the possibility of atelectasis or pneumonic consolidation was ruled out. The patient was already diagnosed with diaphragmatic eventration by fluoroscopy and ultrasonography. Blood test and electrocardiogram results were all within normal ranges. {#fig1-2050313X19827737} Because the patient was non-cooperative as a result of his young age, anesthesia was induced with ketamine 1.5 mg/kg before surgery. With the ventilator, spontaneous tidal volume was about 100 mL before intubation. After the administration of 0.8 mg/kg rocuronium as a muscle relaxant, bag-valve-mask ventilation was initiated with pressure less than 15 cmH~2~O and tidal volume under 100 mL. Intubation was performed with a cuffed endotracheal tube (size 4.0). Lung sounds were clear on the upper and lower left side. However, on the right side, lung sounds were auscultated only on the upper side, but not on the lower side. Anesthesia was maintained with sevoflurane and remifentanil. Bronchoscopy (3.1 mm diameter, Olympus America, Brooklyn Park, MN, USA) was used to identify three openings on the right bronchus. Initially, volume-controlled ventilation was set at tidal volume 100 mL, respiratory rate 25, fraction of inspired oxygen (FiO~2~) 50%, and flow 3 L. With these initial ventilator settings, peak pressure was 25 cmH~2~O. Before the initiation of surgery, the zone of apposition was examined using ultrasonography. To prevent an increase in abdominal pressure, the probe was placed on the lateral chest wall instead of the abdomen.^[@bibr5-2050313X19827737]^ Lung sliding and pleural edge were observed in the 8th through the 10th intercostal spaces on the left side and in the 5th through the 7th intercostal spaces on the right side. At about 15 min after the initiation of surgery, the peak pressure suddenly rose to 28--30 mmHg. Lung ultrasonography was repeated to determine whether the rise in pressure was due to single lung ventilation and diaphragmatic rupture. In both lung fields, the level of lung sliding motion remained the same as that observed during the initial examination. The diaphragm demonstrated good movement. The sevoflurane concentration was increased to 2.5% to heighten the depth of anesthesia. To prevent diaphragmatic rupture, the tidal volume was lowered to 90 mL. Consequently, the peak pressure dropped to 22 cmH~2~O while maintaining end-tidal CO~2~ (ETCO~2~) at about 38 mmHg. Following surgery, the motion of the diaphragm was examined with real-time lung ultrasonographic imaging in the post-anesthesia care unit. It was difficult to observe slow movement of lung sliding because the child was crying. However, the pleural edge moved rapidly in the caudal direction in the zone of apposition when the patient cried, creating effects similar to a sniffing test (see [Supplemental Video](https://journals.sagepub.com/doi/suppl/10.1177/2050313X19827737)). Thus, the possibility of diaphragmatic rupture was excluded. Discussion {#section3-2050313X19827737} ========== Diaphragmatic eventration is so rare that it is difficult to measure the exact incidence. Although there are some reports of the anesthetic management of adult patients with this condition, cases of pediatric patients are limited.^[@bibr6-2050313X19827737]^ The reported incidence of pediatric congenital diaphragmatic eventration is 1/1400,^[@bibr7-2050313X19827737]^ and only one case of spontaneous rupture of a congenital diaphragmatic eventration in an infant has been reported.^[@bibr8-2050313X19827737]^ However, the exact incidence of the condition is unknown in the general population. Partial diaphragm elevation has generally been found on the anteromedial right hemidiaphragm, while complete elevation has been observed on the left hemidiaphragm.^[@bibr9-2050313X19827737]^ While hypoplastic lung-related diaphragmatic eventration is diagnosed using radiography, fluoroscopy, and computed tomography (CT),^[@bibr7-2050313X19827737]^ it is difficult to count the number of lung lobes on CT images. However, three lobe openings on the right lung were confirmed on bronchoscopy in the present case during general anesthesia. In addition, diaphragmatic motion was noted on fluoroscopy, although congenital eventration of the diaphragm is normally associated with inadequate development of muscles or absence of the phrenic nerve. Extra precautions are required when administering general anesthesia in patients with diaphragmatic eventration. A sudden increase in intra-abdominal pressure may cause diaphragmatic rupture, especially in patients with an abnormal diaphragm. Therefore, the prevention of severe coughing and bucking in patients should be ensured. If diaphragmatic rupture occurs, cardiac output will decrease with the migration of intra-abdominal organs from the intra-abdominal space to the intra-thoracic space, resulting in compression of the heart, aorta, and vena cava. Thus, it is important to maintain sufficient anesthetic depth during induction and emergence. In addition, low-volume bag-valve-mask ventilation is necessary to prevent peak pressure increase. Moreover, total intravenous anesthesia is preferred to balanced anesthesia with inhalation as the latter causes hypoxic pulmonary vasoconstriction. Peak pressure may rise upon single lung ventilation or migration of abdominal organs to the thoracic area after diaphragmatic rupture. When the peak pressure does rise, it is important to determine its cause. In addition, lung ultrasonography can be useful to evaluate diaphragm function, although there is no consensus on the sensitivity of ultrasound in this assessment.^[@bibr10-2050313X19827737]^ Abnormal diaphragmatic motion during breathing can be examined in M-mode ultrasonography.^[@bibr4-2050313X19827737]^ Normal diaphragmatic movement during inspiration shows movement toward the transducer when the transducer is on the right below the normal diaphragm position. When the diaphragm ruptures, the diaphragm may appear to be floating or invisible, or a subphrenic fluid collection may appear on ultrasonography.^[@bibr11-2050313X19827737]^ Ultrasonography may show herniation of the solid abdominal contents, such as the liver, omentum, or a bowel segment with peristaltic activity. If the liver sliding that is hepatic parenchymal movement is shown on the right upper trunk instead of the lung parenchyma, it may indicate liver herniation.^[@bibr12-2050313X19827737]^ Comparison of the amplitude of diaphragmatic movement and the alteration of changes in diaphragm thickness with the contralateral side is also important. In this case, liver and lung diaphragm were found at the locations expected from the results of pre-operative chest posteroanterior (PA) imaging of the right side: five or six rib levels higher than on the left side. When the patient cried, normal diaphragmatic excursion was observed. The fact that diaphragmatic movement in the zone of apposition was observed from the lateral wall of the chest does not necessarily mean the movement was pleural. This is because the movement in the zone of apposition may indicate paradoxical motion induced by the opposite lung.^[@bibr5-2050313X19827737]^ However, paradoxical movement due to the opposite side lung was excluded because caudal moving of the pleural edge and inspiration during snappy crying were observed at the same time. In cases of diaphragmatic rupture, the movement of the diaphragm would be flattening or limited at the zone of apposition. However, in the present case, there was no such abnormal diaphragmatic movement. Indeed, the boundary between the abdominal compartment contents, such as the liver and bowel, and the chest compartment, was in the same position as during the initial examination. The thickness of the diaphragm during contraction was also normal. Thus, we thought the possibility of diaphragmatic rupture was low. Regional anesthesia can be a safe choice for certain surgical lesions.^[@bibr13-2050313X19827737]^ It is known that peripheral nerve blocks are safe when performed on sedated pediatric patients.^[@bibr14-2050313X19827737]^ For the patient in the present, brachial plexus block (BPB) could not be performed after sedation as it may induce respiration depression, thereby accelerating desaturation owing to the small lung volume. In addition, manual ventilation or airway intubation in an emergency may increase the likelihood of diaphragmatic rupture. Moreover, BPB performed above the supraclavicular level may trigger phrenic nerve palsy. An axillary approach for BPB was also inappropriate in this case as it requires deep sedation to avoid puncture of the surrounding vessels. Conclusion {#section4-2050313X19827737} ========== When performing mechanical ventilation on patients with congenital diaphragmatic eventration, it is important to prevent the rise of abdominal pressure and peak pressure, as this may lead to diaphragmatic rupture. In addition, perioperative lung ultrasonography may be useful for the rapid diagnosis of diaphragmatic rupture. Supplemental Material {#section5-2050313X19827737} ===================== ###### supple_description -- Supplemental material for Ultrasound-guided perioperative management of 28-month-old patient with congenital diaphragmatic eventration ###### Click here for additional data file. Supplemental material, supple_description for Ultrasound-guided perioperative management of 28-month-old patient with congenital diaphragmatic eventration by Hyemin Kang, Sangmin Lee, Hyunwoo Park, Yeojung Kim, YoungKwon Ko, Yoon-Hee Kim and Boohwi Hong in SAGE Open Medical Case Reports The authors thank the nurses in the Department of Anesthesiology and Pain Medicine who are devoted to patients. **Declaration of conflicting interests:** The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. **Ethical approval:** Our institution does not require ethical approval for reporting individual cases or case series. **Funding:** The author(s) received no financial support for the research, authorship, and/or publication of this article. **Informed consent:** Written informed consent was obtained from a legally authorized representative(s) for anonymized patient information to be published in this article. **ORCID iD:** Boohwi Hong  <https://orcid.org/0000-0003-2468-9271> **Supplemental material:** Supplemental material for this article is available online.

Introduction {#s1} ============ Diabetic dyslipidaemia is characterised by hypertriglyceridaemia, low high-density lipoprotein (HDL) cholesterol (c) and normal low-density lipoprotein-cholesterol (LDLc) but preponderance of small-dense, highly atherogenic particles. The increase in free fatty acids (FFAs) as degradation products of triglycerides (TGs) is associated with the development of insulin resistance [@pone.0027208-Wilding1]. Cholesteryl Ester Transfer Protein (CETP) and Hepatic Lipase (HL) are central enzymes in the metabolism of HDL particles and reverse cholesterol transport. CETP is responsible for an exchange of cholesteryl ester (CE) for triglycerides (TGs) between LDL and HDL and TG rich-lipoprotein particles [@pone.0027208-Morton1]. The result is an enrichment of HDL and LDL particles in TGs, which makes them good substrates for HL [@pone.0027208-Thuren1]. The latter catalyses the hydrolysis of the TGs and phospholipids present in several lipoprotein subclasses, leading to changes in the size and density of lipoproteins [@pone.0027208-Thuren1]. The increased activity of either enzyme results in lower HDLc levels and a predominance of small, dense HDL and LDL particles [@pone.0027208-Lagrost1], [@pone.0027208-Zambon1]. The variations in the *CETP* gene, which lead to changes in enzyme function, have consequences on lipoprotein composition. *CETP* deficiency in humans is characterized by increases in HDLc, whereas increases in its activity are associated with an enrichment of HDL particles in TGs and a decrease in HDLc levels [@pone.0027208-Ikewaki1]. The most extensively studied polymorphism in *CETP* is Taq1B (rs708272) [@pone.0027208-Kondo1]. The G allele, also called B1, is associated with higher enzymatic activity, higher CETP mass and lower HDLc levels [@pone.0027208-Noone1], [@pone.0027208-Boekholdt1]. It has been estimated that this polymorphism is responsible for 5.8% of the variation in HDLc levels [@pone.0027208-Corella1]. Studies in transgenic mice demonstrate that the over-expression of the gene encoding HL, *Lipc*, leads to a marked decrease in plasma HDLc levels[@pone.0027208-Isaacs1] , an observation supported by human studies showing an inverse correlation between HL activity and HDLc concentrations [@pone.0027208-Blades1]. The -G250A *LIPC* polymorphism (rs2070895) [@pone.0027208-Todorova1], located in the promoter region of the gene, has been extensively studied in relation to enzyme activity and lipid metabolism. The minor allele (A) is associated with a reduction of transcriptional activity *in vitro* [@pone.0027208-Deeb1] and a 15--45% reduction in enzymatic activity [@pone.0027208-Tahvanainen1]. In humans, the minor allele has also been associated with an increased HDLc concentration and more buoyant LDL particles [@pone.0027208-Tahvanainen1], [@pone.0027208-Zambon2]. Studies assessing the association of this variant with T2D show conflicting results [@pone.0027208-Zacharova1]. Diabetes is often preceded and even predicted, by the presence of dyslipidemia [@pone.0027208-Todorova1]. Thus, mechanisms involved in the development of diabetic dyslipidemia may also play a role in the pathogenesis of T2D. The effects of the mentioned polymorphisms in *CETP* and *LIPC* on HDLc concentrations are well established, but their relation with the risk of T2D is less known. Therefore, the aim of our study was to analyze the relationship between polymorphisms in these two genes and the presence of diabetes and insulin resistance in a Canarian population. Methods {#s2} ======= Study population {#s2a} ---------------- The Telde study is a cross-sectional population-based study on the prevalence of diabetes and cardiovascular risk factors in Telde, a city located on the island of Gran Canaria, Spain. The study population and design of this survey has been previously described [@pone.0027208-Boronat1]. An oral glucose tolerance test (OGTT) was performed and the subjects were classified (using ADA 1997 criteria) as diabetic (n = 115) and pre-diabetic (n = 116) if they had impaired fasting glucose, impaired glucose tolerance or both. A total of 226 subjects with a normal OGTT were selected, after matching for gender and age with the other two groups. All participants gave their written informed consent for participation in the study, which was carried out according to the declaration of Helsinki and approved by the local ethics committee. Genetic analyses {#s2b} ---------------- The biochemical analyses and insulin resistance parameters have been described previously [@pone.0027208-Novoa1]. Genomic DNA was extracted from whole blood (n = 457) using a salting-out method. The Taq1B *CETP* polymorphism was amplified by PCR-RFLP as described by June Hsieh Wu [@pone.0027208-Wu1] and the *G-250A LIPC* polymorphism was analyzed by AMRS-PCR (Amplification Refractory Mutation System-Polymerase Chain Reaction) [@pone.0027208-Ye1]. Two pairs of primers were used, one which amplifies a fragment of 366 bp, common to both alleles (outer primers: 5′-CTT TTC TTT TTC TTT GGG CTT AGG CT-3′ and 5′-AAG ACT GCC CAT TAA TAA TTA ACC TCT CAA-3′) and another pair specific for the SNP (inner primers): 5′-CAA GGT CAG AGT TCC AAA TTA ATC CAC-3′ for the G allele and 5′-TTC CAA ACA CAA CAC AGT AGC TTT CAA-3′ for the A allele. The primers were designed *"in silico"* in a free access web (<http://cedar.genetics.soton.ac.uk>, accessed in August 2007) and then checked for specificity (<http://blast.ncbi.nlm.nih.gov/Blast.cgi>, accessed in August 2007). The PCR reaction was carried out in a total volume of 25 µl containing 1∶5 ratio of outer to inner primer concentration. The annealing temperatures were 70°C during 15 sec for the outer primers and 58°C during 25 sec for the inner primers with a 30 sec extension at 72°C. PCR products were mixed with 2 µl of loading buffer and run on 2.5% agarose gel stained with Ethidium Bromide. This resulted in 3 DNA fragments: one of 366 bp, one of 234 bp for the A allele and one of 185 bp for the G allele ([figure 1](#pone-0027208-g001){ref-type="fig"}). {#pone-0027208-g001} Statistical analysis {#s2c} -------------------- Statistical analyses were performed with SPSS for WINDOWS, version 13 (SPSS Inc., Chicago, IL). The quantitative variables are described as mean ± standard deviation (S.D). Before further analyses, variable distribution was checked with the Kolmogorov--Smirnov test. A logarithmic transformation was performed for variables not following a Gaussian distribution. Differences between groups were analyzed using either analysis of variance or analysis of covariance, both with the Bonferroni post hoc correction test, after adjusting for age, gender, Body Mass Index (BMI) and waist. The categorical variables were compared using Fisher\'s exact test for 2×2 tables and chi-squared or the Mantel-Haenszel test for linear association. The independent contribution of each polymorphism to DM2 risk was analyzed by a multinomial logistic regression model, which included age, gender, BMI and waist. All tests were considered significant if p was \<0.05. Regarding the effect of the interaction of both polymorphisms on the risk of diabetes, the reference category was defined by the non-B1B1 genotype, regardless of the -250G/A *LIPC* genotypes (nonB1B1*CETP* genotype). A second group included B1B1 and non-GG genotypes (B1B1CETP/non-GGLIPC) and a third, B1B1 CETP and *LIPC* GG genotypes (B1B1CETP/GGLIPC). Results {#s3} ======= Patient description {#s3a} ------------------- The anthropometric, clinical and genetic characteristics of the whole population and their classification according to the OGTT are shown in [table 1](#pone-0027208-t001){ref-type="table"}. The frequencies of the B1B1, B1B2 and B2B2 genotypes of the Taq1B *CETP* polymorphism in the whole population were 46.38%, 41.57% and 12.03% respectively, while the frequencies of GG, AG and AA genotypes of the -250G/A *LIPC* polymorphism were 49.23% 43.54% and 7.22%, respectively The distribution and the allele frequency of both polymorphisms followed the Hardy-Weinberg equilibrium. Due to very low frequencies of the 2 genotypes the B2B2 of the Taq1B *CETP* and the AA of the -250G/A *LIPC* were analyzed in the same category as the corresponding heterozygotic genotype, namely as non-B1B1 (B2 carriers) and non-GG (A carriers), respectively. 10.1371/journal.pone.0027208.t001 ###### Main features of the study population. {#pone-0027208-t001-1} WHOLE POPULATION (n = 457) CONTROLS (n = 226) PRE-DIABETIC (n = 116) DIABETIC (n = 115) ----------------------------- ---------------------------- -------------------- ------------------------ -------------------- Age (y) 55.02±11.09 54.48±11.86 52.25±12.36 58.84±10,60 Sex (male/female) 226/234 104/124 57/59 65/51 BMI (kg/m2) 29.62±5.00 28.60±4.45 30.30±5.03 30.97±5.44 Cholesterol (mmol/L) 5.65±1.02 5.65±1.00 5.64±0.91 5.66±1.16 HDL-C (mmol/L) 1.38±0.32 1.45±0.32 1.35±0.30 1.29±0.31 LDL-C (mmol/L) 3.55±0.87 3.58±0.87 3.53±0.79 3.53±0.96 TG (mmol/L) 1.55±0.91 1.34±0.69 1.65±0.91 1.84±1.18 HbA1c (%) 5.76±4.89 5.51±6.76 5.12±0.45 6.90±1.80 Fasting glucose (mmol/L) 6.04±2.29 4.92±0.52 5.64±0.66 8.66±3.22 Glucose after OGTT (mmol/L) 7.13±2.78 5.50±1.21 8.28±1.76 12.11±3.22 Insulin (pmol/L) 75.4±65.3 59.7±38.7 76.7±45.7 104.9±103.1 HOMA 3.19±4.17 1.91±1.32 2.75±1.61 5.71±5.70 *CETP* Allele B1 67% 65% 67% 71% Allele B2 33% 35% 33% 29% LIPC Allele A 29% 30% 28% 26% Allele G 71% 70% 72% 74% Continuous variables are expressed as mean and S.D. Allele frequencies are expressed as percentages. The polymorphisms and the biochemical variables {#s3b} ----------------------------------------------- In the whole population, the B1B1 genotype carriers showed significantly lower HDLc concentrations than the B2-allele carriers (1.33±0.30 mmol/L vs. 1.44±0.33 mmol/L, p\<0.001), as well as higher glucose levels after the OGTT. In the control population, there was a significant difference in HDLc levels (1.38±0.31 mmol/L vs. 1.50±0.31 mmol/L, p = 0.004; B1B1 vs. B2-carriers, respectively) and an almost significant difference in glucose levels after the OGTT (5.64±1.08 mmol/L vs. 5.39±1.29 mmol/L, p = 0.069). There were no significant differences between the groups in the pre-diabetic and diabetic subjects (data not shown). Regarding the *LIPC* polymorphism, the GG genotype carriers showed significantly higher fasting glucose concentrations than the A allele carriers (non-GG genotype) in the whole population (6.30±2.67 mmol/L vs. 5.79±1.83 mmol/L, p = 0.02). However, only the diabetic subjects showed significantly different glucose levels after the OGTT depending on their genotype (13.03±2.72 mmol/L vs. 11.11±2.94 mmol/L, p = 0.025, respectively). Given the fact that the HOMA index is not a reliable estimation of insulin resistance in diabetic subjects, this subgroup was not analyzed separately in the analysis of variance or covariance. The groups analyzed were: (1) the whole population, (2) the whole population without diabetes (healthy subjects and pre-diabetic subjects) and (3) the healthy control group. [Table 2](#pone-0027208-t002){ref-type="table"} displays the main results obtained. In summary, B1B1 carriers of the *CETP* polymorphism showed lower HDLc levels and higher HOMA and post-OGTT glucose in all groups and higher insulin levels in the whole population even after adjusting for confounding variables. On the other hand, the GG genotype carriers of the *LIPC* polymorphism showed higher fasting glucose levels than A-allele carriers. 10.1371/journal.pone.0027208.t002 ###### The Effect of Taq 1B *CETP* (A) and --G250A *LIPC* (B) polymorphisms on several glucose homeostatic parameters for each population studied. {#pone-0027208-t002-2} A\) WHOLE POPULATION CONTROLS NON-DIABETIC POPULATION ---------------- --------- ------------------- ---------- ------------------------- ------- ------------------- ------- FastingGlucose NonB1B1 5.64 (5.45--5.85) 4.86 (4.78--4.95) 5.08 (4.99--5.17) (mmol/L) B1B1 5.91 (5.69--6.13) 0.079 4.95 (4.85--5.05) ns 5.19 (5.08--5.29) ns Glucoseafter NonB1B1 6.38 (6.08--6.68) 5.18 (4.98--5.39) 5.92 (5.68--6.17) OGTT(mmol/L B1B1 6.94 (6.60--7.31) 0.017 5.56 (5.31--5.83) 0.023 6.37 (6.08--6.68) 0.021 Insulin NonB1B1 56.7 (52.7--61.0) 46.7 (42.7--50.9) 51.8 (48.0--55.8) (pmol/L) B1B1 63.2 (58.5--68.4) 0.047 50.9 (48.2--59.5) 0.052 57.5 (52.8--62.6) 0.068 HOMA NonB1B1 2.05 (1.51--2.23) 1.45 (1.32--1.60) 1.68 (1.55--1.82) B1B1 2.39 (2.18--2.62) 0.017 1.69 (1.51--1.89) 0.038 1.90 (1.74--2.08) 0.039 HDL-C NonB1B1 1.39 (1.36--1.43) 1.46 (1.41--1.52) 1.42 (1.38--1.47) (mmol/L) B1B1 1.30 (1.26--1.33) 0.000 1.33 (1.28--1.39) 0.002 1.31 (1.27--1.36) 0.001 B\) WHOLE POPULATION CONTROLS NON--DIABETIC POPULATION ---------------- -------- ------------------- ---------- -------------------------- ---- ------------------- ------- FastingGlucose Non-GG 5.62 (5.42--5.82) 4.88 (4.79--4.98) 5.06 (4.97--5.16) (mmol/L) GG 5.92 (5.71--6.14) 0.044 4.92 (4.83--5.02) ns 5.19 (5.09--5.29) 0.067 Glucoseafter Non-GG 6.67 (6.35--7.00) 5.33 (5.11--5.55) 6.08 (5.82--6.35) OGOTT(mmol/L) GG 6.61 (6.28--6.95) ns 5.36 (5.13--5.60) ns 6.19(5.92--6.47) ns Insulin Non-GG 58.9 (54.6--63.5) 47.9 (43.6--52.7) 51.8 (47.9--56.0) (pmol/L) GG 60.3 (55.8--65.1) ns 51.5 (46.6--56.9) ns 56.8 (52.4--61.5) ns HOMA Non-GG 2.12 (1.93--2.31) 1.49 (1.35--1.65) 1.67 (1.54--1.82) GG 2.28 (2.08--2.50) ns 1.62 (1.46--1.80) ns 1.88 (1.73--2.05) 0.055 HDL-C Non-GG 1.35 (1.31--1.39) 1.41 (1.36--1.47) 1.38 (1.33--1.42) ns (mmol/L) GG 1.35 (1.31--1.39) ns 1.40 (1.34--1.46) ns 1.37 (1.33--1.42) ns P values from analysis of covariance with the Bonferroni post hoc test, after adjusting for age, gender, Body Mass Index (BMI) and waist. Association between CETP and LIPC polymorphisms and the prevalence of type 2 diabetes {#s3c} ------------------------------------------------------------------------------------- [Figure 2](#pone-0027208-g002){ref-type="fig"} shows the frequency of each polymorphism for the different categories of glucose tolerance (diabetic, pre-diabetic and control groups). The frequency of the B1B1 genotype of *CETP* increased with worsening glucose metabolism, whereas the GG genotype of *LIPC* did not show a significant difference, although a similar trend was found. {#pone-0027208-g002} The independent effect of each polymorphism on the prevalence of diabetes was assessed using a multinomial logistic regression model adjusting for age, gender, BMI and waist. The B1B1 genotype was associated with an increased risk of diabetes (OR (IC95%): 1.81 (1.12--2.91); p = 0.002), but not pre-diabetes (OR (IC95%): 1.11 (0.70--1.77); p = ns). On its own, the *LIPC* polymorphism was not significantly associated with the risk of diabetes. Finally, we analyzed the effect of the interaction of both polymorphisms on the risk of diabetes. We observed that among the B1B1 *CETP* carriers, the presence of the GG *LIPC* genotype increased the risk of having the disease ([table 3](#pone-0027208-t003){ref-type="table"}). 10.1371/journal.pone.0027208.t003 ###### Multinomial logistic regression model assessing the combined effects of *CETP* and *LIPC* genotypes on the risk of T2D. {#pone-0027208-t003-3} VARIABLE P O.R 95% CI -------------------------- ------- --------- ------------ NonB1B1 *CETP* 1(ref.) B1B1 *CETP*/nonGG *LIPC* 1.32 0.71--2.44 B1B1 *CETP*/GG *LIPC* 0.036 2.42 1.30--4.50 OR: odds ratio, 95% CI: 95% confidence interval. Discussion {#s4} ========== The roles of CETP and HL on lipid metabolism and reverse cholesterol transport have been extensively described [@pone.0027208-Zambon1]. In this study, we investigated a variant present on each gene encoding these enzymes in relation to the risk of diabetes in a Canarian population. Our results show that the B1B1 genotype of the Taq1B *CETP* polymorphism is associated with more insulin resistance, higher post-OGTT glucose levels and an increased risk of T2D. On other hand, the -250G/A *LIPC* polymorphism is associated with higher fasting glucose levels, but does not seem to confer a risk of T2D by itself. However, the interaction between both polymorphisms does have an effect on the risk of diabetes. The Taq 1B CETP polymorphism (rs708272) {#s4a} --------------------------------------- Previous studies have shown that the Taq1B *CETP* polymorphism is associated with increased enzyme activity, TG-enriched LDL and HDL particles and low HDLc levels [@pone.0027208-Boekholdt1], [@pone.0027208-Gudnason1]. Besides, high CETP activity has been demonstrated in obese and diabetic subjects [@pone.0027208-Dullaart1], [@pone.0027208-Smaoui1]. Previous studies have also shown an association between this polymorphism and the metabolic syndrome [@pone.0027208-Sandhofer1], independently of the well-known effect on HDLc concentrations and insulin resistance [@pone.0027208-LpezRos1], a fact that suggests a possible role of CETP on glucose metabolism. However, to our knowledge, this is the first study to show an association between this polymorphism in *CETP* and the risk of T2D. The frequency of the B1B1 genotype increases with worsening glucose tolerance ([figure 2](#pone-0027208-g002){ref-type="fig"}) and its presence is associated with the risk of diabetes even after adjusting for other confounding factors such as age, BMI, waist and TG. In addition, we found lower HDLc levels in B1B1 genotype carriers, as well as higher insulin, HOMA and post-OGTT glucose in non-diabetic B1B1 carriers, further supporting its role in the development of T2D. Previously, we proposed that the contribution of the Taq1B *CETP* polymorphism on insulin resistance could be mediated by an increased flux of FFAs from HDL particles to the liver. Since homozygotes for the B1 allele have an increased CETP activity, they should have an increased TG content in their HDL particles, which in turn would become a good substrate for HL. Thus, individuals with the B1B1 genotype would have an increased flux of free fatty acids to the liver from HDL that would decrease the hepatic sensitivity to insulin [@pone.0027208-LpezRos1]. The --G250A LIPC polymorphism (rs2070895) {#s4b} ----------------------------------------- HL catalyzes the hydrolysis of TG and phospholipids in TG-enriched HDL and LDL particles, giving rise to smaller, denser particles[@pone.0027208-Lambert1]. In fact, HL activity shows an inverse correlation with HDLc concentration. Two (SNPs) have been described in the promoter region of the gene (-514 C\<T, rs1800588 and -250G\>A, rs2070895) [@pone.0027208-Andersen1], which are in almost complete linkage disequilibrium. The minor allele in the -250A/G polymorphism is associated with low HL activity [@pone.0027208-Lindi1], an increased HDLc concentration[@pone.0027208-Isaacs1] and more buoyant LDL particles. In fact, the effect abdominal obesity has on HL activity is cushioned by this allele [@pone.0027208-Carr1]. Its frequency in our population was similar to that found by others [@pone.0027208-Todorova1], [@pone.0027208-Zacharova1], [@pone.0027208-Jackson1]. However, unlike other authors [@pone.0027208-Zacharova1], [@pone.0027208-Despres1], we did not find an influence of the *LIPC* genotype on HDLc concentrations, nor on the risk of T2D, but it was associated with higher glucose concentrations. The latter followed the same direction as the results from the Finnish Diabetes Prevention Study [@pone.0027208-Todorova1], which showed that the GG genotype doubles the risk of progression to T2D. On the other hand, a large Danish cross-sectional study, which included 3082 cases and 4882 controls, was negative in this aspect [@pone.0027208-Grarup1]. Taq 1B CETP and --G250A LIPC polymorphisms {#s4c} ------------------------------------------ The epistatic effect of *CETP* and *LIPC* on HDLc concentrations [@pone.0027208-Isaacs2], [@pone.0027208-Soyala1] and atherosclerosis[@pone.0027208-Soyala1] has been previously reported. The authors observed a marked increase in HDLc levels in carriers of both minor frequency genotypes [@pone.0027208-Soyala1]. However, to our knowledge, this is the first time the effect of genetic interaction between *CETP* and *LIPC* is assessed on the risk of T2D. Since the minor allele at -250G/A *LIPC* is associated with a decrease in HL activity and a reduction of TG catabolism from HDL and LDL and the TG content in HDL and LDL is, to a certain extent, the result of an increased CETP activity associated to the B1B1 genotype, we propose that the presence of the A allele of -250G/A *LIPC* reduced the risk of T2D among B1B1 Taq1B *CETP* genotype carriers, although the *LIPC* polymorphism was by itself not associated with T2D. One of the strengths of this study is its population-based design including well-characterized subjects diagnosed using an OGTT [@pone.0027208-Dullaart1], [@pone.0027208-Smaoui1]. In addition, to our knowledge, this is the first time the effects of *CETP* and *LIPC* are assessed in relation to the risk of T2D. We are also aware of some limitations of the study. Its cross-sectional nature does not allow us to infer causality. Furthermore, our failure to identify an effect of the *LIPC* polymorphism on diabetes risk might be due to sample size, as is suggested by its association with glycemia and its significant interaction with *CETP*. However, the effects we did find are likely to be true positive, biologically plausible effects, with stringent corrections for multiple analyses. We suggest that while the risk of T2D associated to the B1B1 *CETP* genotype is a consequence of hypertriglyceridaemia by the increase of the flux of FFAs from HDL and LDL particles to the liver, a decrease activity of HL associated to -250 A *LIPC* allele reduces that flux of FFAs, and the risk of T2D amongst B1B1 carriers. In summary, the Taq1B *CETP* polymorphism was significantly associated with HDLc levels and the presence of T2D and, although we did not find the same association between the -250A/G *LIPC* polymorphism and HDLc levels or T2D, the presence of the A allele appears to exert a protective effect in B1B1 genotype carriers in our population. Nevertheless, larger studies performed in different populations are needed to confirm our findings. The authors express their gratitude for excellent technical assistance to Angelines Jiménez, the Research Unit from Hospital Dr Negrín (Gran Canaria) and Instituto Canario de Investigación del Cáncer (ICIC, Spain). **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work has been supported in part by grants from Fundación Canaria de Investigación y Salud (FUNCIS, Spain; ref: Canarias Bioregión). LLR is being supported by a post-doctoral fellowship from ULPGC. FJN and AMW receive funding from Ministerio de Ciencia e Innovación (PI08/1113 and PI10/2310) and the European Foundation for the Study of Diabetes (EFSD/JDRF/Novo Nordisk Programme 2008). No additional external funding was provided for this study. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. [^1]: Conceived and designed the experiments: LLR FJN RC MBC AMW. Performed the experiments: LLR MBC FV. Analyzed the data: LLR RC MBC AMW. Contributed reagents/materials/analysis tools: FJN RC FV. Wrote the paper: LLR RC AMW.

Background ========== Diatoms are important primary producers in the ocean \[[@B1]\], contributing approximately 40% to global marine productivity. Although diatoms often dominate phytoplankton communities in nutrient-rich ecosystems, members of this diverse group are also adapted to survive and persist in nutrient-limited conditions. The development of large diatom blooms upon nutrient resupply demonstrates the metabolic plasticity inherent to their ability to recover rapidly from nutrient limitation. Iron is an essential nutrient for all organisms and in particular for photoautotrophic organisms. It functions as a powerful electron carrier in iron-sulfur- and heme-containing proteins and as such is a required component of the photosynthetic apparatus. Solubility of iron in seawater is low and phytoplankton growth in marine habitats is often limited by iron availability. This is best illustrated in high-nitrate low-chlorophyll (HNLC) regions, remote oceanic areas that lack any form of regular iron supply and suffer from a persistent shortage of this micronutrient. Here, although other commonly limiting nutrients like nitrate or phosphate are present at high concentrations, primary productivity - and biomass as a whole - is low \[[@B2]\]. Numerous large-scale iron fertilization experiments have confirmed that iron is the limiting nutrient in HNLC regions \[[@B3]\]. Phytoplankton blooms induced by iron fertilization were dominated by diatoms and carbon export to the deep-sea floor could be observed in some cases. The strong response of diatoms to the input of iron in HNLC regions has been a motivation for exploring large-scale iron fertilization as a possible bioengineering strategy to sequester CO~2~into the ocean in HNLC regions, which are otherwise rich in macronutrients. Genome projects on the model organisms *Thalassiosira pseudonana*\[[@B4]\] and *Phaeodactylum tricornutum*\[[@B5]\] have already generated a wealth of insights into the metabolic complexity of diatoms \[[@B6]\], a consequence of the secondary endosymbiosis event that gave rise to this group \[[@B7]\]. This secondary endosymbiosis brought together the benefits of a heterotrophic host and the \'red\'-type photosynthesis of red alga cells, which already have an elemental composition low in iron \[[@B8]\]. The impact of iron availability on phytoplankton growth has led to the evolution of strategies to counteract iron limitation. Well established parts of the low-iron response found in diverse phytoplankton species are the reduction of the chloroplast system, the corresponding development of a chlorotic phenotype, compensation mechanisms (replacement of iron-rich elements with iron-poor substitutes) and the activation of high-affinity iron-uptake systems \[[@B9]\]. The substitution of ferredoxin by flavodoxin \[[@B10]\], the use of plastocyanin instead of cytochrome c~6~\[[@B11]\] and a variant stoichiometry of photosynthetic complexes \[[@B12]\] are notable adaptive strategies to facilitate diatom growth in low-iron conditions. Oceanic and neritic phytoplankton species can be distinguished from each other by their growth characteristics and their tolerance to nutrient limitation \[[@B13]\]. Unlike many other *Thalassiosira*species that are predominantly found in coastal waters, *Thalassiosiraoceanica*is adapted to oligotrophic conditions and is highly tolerant to iron limitation in particular. Therefore, we chose *T. oceanica*CCMP1005 as a model for a comprehensive analysis of its low-iron response in the context of genomic information. Here, we explore the complex cellular response of *T. oceanica*to low-iron conditions with genomics, transcriptomics and proteomics approaches complemented by reverse transcription-quantitative PCR (RT-qPCR) analyses. We present a metabolic reconstruction of the iron limitation response based on the transcriptomics data from cells grown under iron-limited versus iron-replete conditions. A metabolic isotope labeling approach using ^14^N/^15^N was established for *T. oceanica*and showed the response to iron limitation at the protein expression level in a marine diatom for the first time. General characteristics of the \'diatom\' low-iron response and its ecological implications are discussed, as well as the constraints for species-specific adaptations to low-iron environments. Results ======= Characteristics of the *T. oceanica*genome ------------------------------------------ The genome of the centric diatom *T. oceanica*CCMP1005 (Figure [1](#F1){ref-type="fig"}) was *de novo*assembled from 725 Mb of Roche 454 sequence read information, generated using nuclear genomic DNA (gDNA) of an axenic clonal culture as substrate \[[@B14]\]. The current assembly version comprises 51,656 contigs of total size 92.15 Mb at N50 = 3,623 (that is, 50% of the genomic sequence information is present as contigs ≥3,623 bases). From a median 8.7-fold coverage of long contigs (≥10 kb) we estimated a true haploid nuclear genome size of 81.6 Mb, suggesting some redundancy in the current assembly. This estimate is in good agreement with the 159 Mb measured by van Dassow *et al*. \[[@B15]\] for the diploid G1 DNA content. The gene finder tool AUGUSTUS \[[@B16]\] predicts 37,921 protein gene models that cluster into a non-redundant set of 29,306 models including pseudogenes and short ORFs; 10,109 models have BLAST hits to the National Center for Biotechnology Information (NCBI) nr protein database at a conservative E-value cutoff of 1.0E-30 and thus are more indicative of the expected true protein-coding gene number (that is, expressed genes excluding pseudogenes and short ORFs). Best BLAST hits are listed in Additional file [1](#S1){ref-type="supplementary-material"}. In Table [1](#T1){ref-type="table"} we present an overview of the most abundant Clusters of Orthologous Groups (COG) domains in *T. oceanica*. The abundances of diverse groups of ATPases were overall very similar to those for other diatoms. A group of 19 chitinases is shared between the two centric *Thalassiosira*species. {#F1} ###### Most abundant protein domains in diatom genomes COG To Tp Pt Fc -------------------------- ------------------------------------------------------------------------------------------------------------------------- ----- ----- ----- ----- ATPases  COG0515 SPS1, serine/threonine protein kinase 115 132 119 137  COG0464 SpoVK, ATPases of the AAA+ class 90 43 38 44  COG1132 MdlB, ABC-type multidrug transport system, ATPase and permease components 54 44 47 51  COG1222 RPT1, ATP-dependent 26S proteasome regulatory subunit 50 41 37 42  COG0465 HflB, ATP-dependent Zn proteases 49 37 35 39  COG1223 Predicted ATPase (AAA+ superfamily) 44 39 34 41  COG3899 Predicted ATPase 42 48 11 2  COG2274 SunT, ABC-type bacteriocin/lantibiotic exporters, contain an amino-terminal double-glycine peptidase domain 42 52 50 61  COG5265 ATM1, ABC-type transport system involved in Fe-S cluster assembly, permease and ATPase components 40 33 30 33  COG4618 ArpD, ABC-type protease/lipase transport system, ATPase and permease components 39 41 42 43  COG4987 CydC, ABC-type transport system involved in cytochrome bd biosynthesis, fused ATPase and permease components 31 50 46 56  COG4988 CydD, ABC-type transport system involved in cytochrome bd biosynthesis, ATPase and permease components 29 52 49 60  COG0488 Uup, ATPase components of ABC transporters with duplicated ATPase domains 29 50 46 53  COG1131 CcmA, ABC-type multidrug transport system, ATPase component 22 56 52 65  COG0661 AarF, predicted unusual protein kinase 21 21 22 26  COG0474 MgtA, cation transport ATPase 12 19 19 18 Basic cellular functions  COG0513 SrmB, superfamily II DNA and RNA helicases 46 48 44 54  COG0553 HepA, superfamily II DNA/RNA helicases, SNF2 family 35 27 24 36  COG5059 KIP1, kinesin-like protein 24 25 15 14  COG1643 HrpA, HrpA-like helicases 21 14 9 20  COG0443 DnaK, molecular chaperone 18 14 9 10  COG5021 HUL4, ubiquitin-protein ligase 15 7 8 8  COG5022 Myosin heavy chain 14 11 9 9  COG1249 Lpd, Pyruvate/2-oxoglutarate dehydrogenase complex, dihydro-lipoamide dehydrogenase (E3) component, and related enzymes 12 14 13 20 Chitinases  COG3325 ChiA, chitinase 19 19 1 0 Protein domains are listed based on Clusters of Orthologous Groups COG with an E-value threshold of 1.0E-10. To, *Thalassiosira oceanica*; Tp, *Thalassiosira pseudonana*; Pt, *Phaeodactylum tricornutum*; Fc. *Fragilariopsis cylindrus*. The chloroplast genome has been published previously \[[@B17]\]. The mitochondrial genome encodes 31 protein genes and is represented by two contigs at a total of 35.3 kb (excluding the characteristic mitochondrial repeats). The current genome assembly, AUGUSTUS protein gene models, ESTs and proteomics peptides as well as updated versions thereof are publicly accessible with the *Thalassiosira oceanica*Genome Browser \[[@B18]\]. With an estimated haploid size of approximately 80 Mb, the genome of *T. oceanica*is significantly larger than those of *T. pseudonana*(approximately 34 Mb) or *P. tricornutum*(approximately 28 Mb), and rather comparable to that of *Fragilariopsis cylindrus*(approximately 80 Mb) \[[@B19]\]. The genome expansion has occurred by DNA recruitment from both internal and external DNA sources. A best BLAST hit analysis indicated a putative vertical inheritance for greater than 95% of the 10,109 predicted genes (that is, any genes that have not been acquired by a horizontal transfer event), with most of the genes (88%) having a match in the genome of *T. pseudonana*, the most closely related species for which a sequenced genome is available (Figure [2](#F2){ref-type="fig"}). However, a significant fraction (10%) of the genes mapped to *P. tricornutum*instead. This could have resulted from frequent gene loss/replacement events in the genome of *T. pseudonana*, thereby reflecting the overall high capacity for horizontal gene transfer in diatoms \[[@B5]\]. Alternatively, the small genome size of *T. pseudonana*may have arisen from reductional trends in this species. {#F2} Further, the best BLAST screening revealed 530 genes whose best hits are assigned to taxa of diverse sources, indicating a putative lateral acquisition for these genes. The taxonomic distribution of the best BLAST hits at a conservative E-value cutoff of 1.0E-30 is presented in Figure [2](#F2){ref-type="fig"}. More refined phylogenetic analyses were obtained for 198 of these 530 genes (Additional file [2](#S2){ref-type="supplementary-material"}). Of the 198 cases we examined, 180 had sister groups that contained no stramenopiles. The sister groups for the remaining 18 cases contained a heterogenous mix of taxa, suggesting frequent transfer between taxa for the respective genes (Figure S1 in Additional file [3](#S3){ref-type="supplementary-material"}). Accordingly, a minimum of 1.8% of the 10,109 AUGUSTUS genes were confirmed to be from lateral gene transfer (LGT) based on the phylogenetic analyses. However, this may rise to 5% as more sequence information becomes available for the remaining \>300 genes for which the limited number of homologous sequences did not permit construction of phylogenetic trees. The group of genes for which phylogenetic trees are available indicated that genes from LGT could be assigned as prokaryotic (35%) and eukaryotic (59%) with approximately 10% of questionable taxonomic assignments. Among the eukaryotic taxa are several expected to be present in the ecological niche of *T. oceanica*, like the green algal genuses *Micromonas*and *Ostreococcus*. Genomic expansion originating from internal DNA sources may happen from genomic duplication events or transposon activity. In *T. oceanica*we observe several paralogous gene pairs that could be the result of either mechanism (Figure S2 in Additional file [3](#S3){ref-type="supplementary-material"}). Notably, several iron-regulated genes have either been duplicated (for example, the *ISIP1*genes *ISIP1A*and *ISIP1B*and the flavodoxin genes *FLDA1*and *FLDA2*) or contain domain duplications (for example, *CREGx2*) as discussed below. Physiology of the low-iron response: Fe(-) versus Fe(+) ------------------------------------------------------- The variable to maximal fluorescence ratio F~v~/F~m~, an indicator of Fe-limitation in the laboratory \[[@B20]\], was used as a rapid measure of the physiological status in Fe-replete and Fe-limited cultures of *T. oceanica*harvested in late exponential growth phase. The growth rate of iron-limited cells in exponential phase was accordingly much smaller than for iron-replete cells (Table [2](#T2){ref-type="table"}), and cellular protein content was 50% lower. ###### Physiology of the *T. oceanica*low-iron response Fe(+) Fe(-) ------------- ----------------------------------------- ------------- 0.5 - 0.6 F~v~/F~m~ 0.2 - 0.3 0.73 ± 0.01 Growth rate (µ)(day^-1^) 0.28 ± 0.02 4 Chloroplasts/cell 2 409 ± 48 Chlorophyll a/cell (fg) 58 ± 27 122 ± 3 Cell surface area (µm^2^) 140 ± 5 100 ± 4 Cell volume (µm^3^) 80 ± 5 15.3 ± 0.9 Single chloroplast surface area (µm^2^) 12.0 ± 0.8 4.7 ± 0.3 Single chloroplast volume (µm^3^) 3.5 ± 0.2 Cellular dimensions and physiological parameters are compared between nutrient replete Fe(+) and iron-limited Fe(-) cells of exponentially growing *T. oceanica*cultures. The cell volume of iron-limited *T. oceanica*was smaller than that of iron-replete cells (Table [2](#T2){ref-type="table"}), whereas the cells had a larger surface area to volume ratio at low-iron due to a smaller diameter (4.7 ± 0.1 versus 5.9 ± 0.1 µm) and a larger length (7.0 ± 0.4 versus 5.5 ± 0.2 µm). This imposed an elongated phenotype on the cells, but at the same time increased the surface/volume ratio by 43% (1.75 versus 1.22). The increase in surface/volume ratio is expected to favor the uptake of nutrients (that is, iron) into the cell \[[@B21]\]. The intracellular space of iron-limited cells exhibited increased vesiculation (Figure [3](#F3){ref-type="fig"}). {#F3} Under low-iron conditions *T. oceanica*cells show a severe decrease in chlorophyll content (Table [2](#T2){ref-type="table"}). This chlorosis response of iron-limited *T. oceanica*is further accompanied by a decrease in cellular chloroplast volume and in total cellular chloroplast surface area (Figure [3](#F3){ref-type="fig"}). Iron-limited cells have reduced the number of chloroplasts to two instead of four in the iron-replete counterpart, and these are also smaller in size. Total chloroplast dimensions for individual cells were distributed over a range spanning the two-fold increase in volume and surface, thereby reflecting chloroplast duplication during cellular growth. Transcriptomics --------------- For an in-depth analysis of the *T. oceanica*low-iron response, we focused on approximately 300 genes that were identified from a log-likelihood ratio test statistic \[[@B22]\] as significantly differentially regulated and that could be assigned a specific function (Figure [4a](#F4){ref-type="fig"}). Some additional genes for paralogous proteins were added. These were selected on the basis of their involvement in substitution between related proteins under iron-limited and replete conditions, or as members of a protein family exhibiting a differential response to low-iron conditions. In such cases, the response of a specific gene is better understood in the context of its respective group or family. The complement of organellar genes (encoded by the chloroplast and mitochondrial genome) was added as representative for the two well-defined and important pathways of photosynthetic and respiratory electron transport, or as proxy for organellar activity, respectively. A list of abbreviations for the genes discussed in this work can be found in Additional file [4](#S4){ref-type="supplementary-material"}. All sequences of the selected proteins are provided in Additional file [5](#S5){ref-type="supplementary-material"}, and the corresponding annotation is provided in Additional file [6](#S6){ref-type="supplementary-material"}. {#F4} To determine the major metabolic differences found in iron-limited compared to iron-replete growth conditions, all annotated gene products together with their respective expression data were mapped on a cellular scheme. The major cellular trends that could be deduced are summarized in Figure [4c](#F4){ref-type="fig"}. Identifier and detailed information on the discussed proteins (Additional file [7](#S7){ref-type="supplementary-material"}) are given in Additional file [8](#S8){ref-type="supplementary-material"}. In the following, proteins are referred to as exemplary (HSF1, p271) with HSF1 reflecting the gene name (or shortcut) and p271 being the identifier of its respective manually improved protein model (Additional file [5](#S5){ref-type="supplementary-material"}). Under stress conditions, maintaining cellular integrity is crucial to survival. During iron limitation, the electron flow through the impaired photosynthetic machinery leads to enhanced production of reactive oxygen species that damage biomolecules located near the thylakoid membranes \[[@B23]\]. The need for protein repair and refolding induces an \'oxidative stress response\' that is presumably coordinated by up-regulated heat shock factors (HSF1, p271; HSF2, p256). While all other chloroplast-encoded transcripts were down-regulated in the course of the general chlorosis response, the chloroplast chaperones dnaK and clpC were up-regulated. Additionally, an LHCSR (light harvesting complex stress responsive subunit) ortholog (LI818, p170), belonging to the FCP (fucoxanthin-chlorophyll a/c-binding protein) family of light-harvesting proteins and implicated in efficient non-photochemical quenching \[[@B24],[@B25]\], showed an increased transcript level. The development of a chlorotic phenotype and the corresponding retrenchment of the chloroplast system is the most pronounced cellular response to low iron. Accordingly, we find substantial changes in organellar transcript levels, which suggests that major functions related to the cellular energy metabolism are adopted by the mitochondrial system instead (\'metabolic shift\'). Chloroplast transcript levels decreased (2,026 Fe(-) versus 14,931 Fe(+) total chloroplast reads), while mitochondrial transcripts showed a two-fold increase (31,261 Fe(-) versus 18,136 Fe(+) total mitochondrial reads). Much of this effect can be attributed to the organellar rRNA operons, whose transcription is indicative of organellar translational activity (Figure S3 in Additional file [3](#S3){ref-type="supplementary-material"}). In parallel, diverse nuclear-encoded but chloroplast-targeted gene products were down-regulated. These included genes coding for enzymes involved in chlorophyll biosynthesis and the Calvin cycle, as well as components of the light reaction, such as photosystem (PS) subunits and several FCPs. Conversely, components of the mitochondrial respiratory chain, like cytochrome c oxidase, cytochrome b and several subunits of the NADH dehydrogenase, were up-regulated. This was also seen for a mitochondrial ATP/ADP-translocase (p242) involved in the transport of energy equivalents. Cellular retrenchment (that is, the reduction of cellular biomass and activity) and decreased growth rates are general responses of nutrient-limited cells \[[@B13]\]. While chloroplast reduction was readily observable in iron limitation due to the visual predominance of these organelles in the cells, we also saw indications of a general cellular retrenchment in the transcriptional response. The expression level of the 18S rRNA gene (represented by 1,154 Fe(-) versus 2,691 Fe(+) reads) suggests a lower translation rate under iron limitation. Though such inferences must be taken with care, this would be in agreement with the decreased growth rate and lower biomass, as cellular rRNA correlates with cellular biomass. The strong up-regulation of mitochondrial isocitrate lyase (ICL, p419) and glutamine synthetase (GS, p302) suggests biomass recycling strategies to avoid losing fixed carbon and nitrogen during the metabolite conversions associated with enhanced respiration. The isocitrate lyase bridges the two decarboxylation steps of the mitochondrial citric acid cycle (carried out by isocitrate dehydrogenase and α-ketoglutarate dehydrogenase), thereby preserving carbon as glyoxylate. The glutamine synthetase reincorporates free ammonium, preserving nitrogen as glutamine. Under low-iron conditions, utilization of ammonium is energetically advantageous due to the high iron requirements of the nitrate assimilation pathway \[[@B26]\]. The concerted action of cellular retrenchment and biomass recycling allows for prolonged growth despite reduced carbon assimilation, thereby increasing the probability of cell survival. Diverse genes, whose products are targeted to the secretory pathway, are up-regulated under iron limitation, suggesting extensive cell-surface remodeling as also observed for iron-limited *P. tricornutum*\[[@B20]\]. Many of these genes are assigned adhesive or degradative functions. An enhanced capacity for adhesion favors recruitment of organic matter to the cell. As organic matter can be a rich and complex source for various nutrients, including iron, its recruitment to the cellular surface represents a required first step in iron uptake. Besides providing a source of iron, the bound organic matter could also serve as a source for other nutrients like nitrogen or phosphorus in the context of facultative mixotrophy. Example genes assignable to such a hypothetical scenario and highly responsive to low iron are given in Figure [5](#F5){ref-type="fig"} and include *Adhesin 1*(p329), *CB*(Carbohydrate-binding 1, p230), *CHIT*(chitinase, p88), *M-Phosphoesterase*(p323), *M-Protease*(p279), *Redox 1*(p232). However, under the photoautotrophic experimental conditions, the cultures lacked any external organic carbon source except the essential vitamins. {#F5} A straightforward strategy to survive in low-iron conditions is to lower cellular iron requirements by replacing components that are rich in iron with iron-free substitutes that are functionally equivalent, like the substitution of the chloroplast electron carrier ferredoxin with flavodoxin \[[@B10]\]. The genome of *T. oceanica*encodes two cytochrome c~6~genes and one plastocyanin gene. While the cytochrome c~6~genes *CYTC6A*and *CYTC6B*are found to be weakly expressed, the plastocyanin gene *PETE*shows high expression under high-iron conditions with a characteristic decrease in low-iron conditions as seen from many constitutively expressed chloroplast genes in the course of the chlorosis response. This suggests a constitutive use of plastocyanin (PETE, p175) instead of cytochrome c~6~for photosynthetic electron transport and is consistent with prior findings \[[@B11],[@B12]\]. Constitutive expression of plastocyanin could certainly be regarded as a specific adaptation to low-iron regimes, although the retention of the cytochrome c~6~genes suggests that these may play a role under specific environmental conditions. Fructose-bisphosphate aldolase (FBA) genes are redundant in some diatoms and have recently been described in more detail for *P. tricornutum*\[[@B27]\]. The *T. oceanica*genome, too, was found to encode several FBA enzymes, with the cytosol, the chloroplast stroma and the chloroplast pyrenoid harboring two FBA enzymes each (FBA1, p380, and FBA3, p153, in the chloroplast pyrenoid; FBA2, p381, and FBA5, AUG_g19407, in the chloroplast stroma; FBA6, AUG_g24977, and FBA4, p154, in the cytosol). As is the case in *Phaeodactylum*, one of the *T. oceanica*FBAs from each compartment (FBA1, FBA2, FBA6) appears to act through metal catalysis (class II) while the second (FBA3, FBA5, FBA4) is predicted to use Schiff-base catalysis (class I) instead. While the metal cofactor of different class II FBAs was found to be Mn^2+^\[[@B28]\], Zn^2+^\[[@B29]\] or Cd^2+^\[[@B30]\] in *Escherichia coli*, the orthologous FBAs of *T. oceanica*apparently are differentially regulated through the availability of iron, suggesting the involved metal in these enzymes might be Fe^2+^, and implying a pairwise substitution by class I enzymes. An essential part of iron-uptake systems are ferric reductases (FREs) and ferrous oxidases (MUCOX proteins) that act on the interconversion of the two ionic species Fe^3+^and Fe^2+^. In the iron-limited transcriptome we find an up-regulated putative ferric reductase (FRE1, p157) and an up-regulated multicopper oxidase (MUCOX2, p67) that shows characteristics of a ferrous oxidase. Their differential regulation with respect to iron availability makes them candidates for iron-specific reductase and oxidase involved in iron uptake (Figure [5](#F5){ref-type="fig"}). Iron uptake requires initial binding of iron and/or iron complexes. The involved receptors are presently unknown, though a number of genes, exclusively expressed under iron limitation, are targeted to the cell surface, making them candidates for iron-binding receptors. The low-iron-responsive gene *ISIP1*(Iron-starvation induced protein) was first identified in *P. tricornutum*, but has conserved orthologs in *T. oceanica*(*ISIP1A*, p159) and *F. cylindrus*. We provide further evidence for a role of the ISIP1 protein as a putative receptor below. Additional members in this group are ISIP2 (p160) and ISIP3 (p161), both represented by orthologs in *P. tricornutum*as well. Further, we list some proteins that contain duplicated domains known from *P. tricornutum*low-iron-responsive genes, like an eight-fold duplicated ISIP2-like subdomain (ISIP2x8, p84) or a duplicated CREG-like domain (CREGx2, p90). Duplication of iron-binding domains would directly enhance the capacity for iron binding and enable increased uptake kinetics \[[@B26]\]. Non-ribosomal peptide synthases (NRPSs) \[[@B31]\] are responsible for the production of peptide antibiotics or - in some cases - siderophores that are capable of binding iron \[[@B32]\]. In addition to a conserved fungal NRPS (NRPS1, p174) with orthologs in *T. pseudonana*and *P. tricornutum*, we find a putatively cytosolic NRPS of bacterial origin (NRPS2, p173) up-regulated in low-iron conditions. Co-regulated with this bacterial NRPS is a multidrug resistance-associated protein (MRP, p57) that might be involved in the export of the respective peptide products. The up-regulation of NRPSs likely indicates a defense mechanism in response to enhanced competition (either for iron or, under the premise of facultative mixotrophy, for organic matter). We observe the induction of a reverse transcriptase (RT, p222) and a CRE-like recombinase (CRE, p321), potentially indicating an activation of mobile elements under iron limitation. These enzymes might also be involved in gene and/or domain duplication events through reverse transcription and genomic integration of cellular mRNA copies. Thereby, this molecular system may provide a link between environmental stresses and the structural dynamics of the diatom genome. Proteomics ---------- The transcriptomic data of *T. oceanica*unveils extensive changes in cellular transcript levels in response to iron limitation. Although informative, transcript abundances do not necessarily reflect cellular protein levels \[[@B33]\]. We therefore supplemented the transcriptomic data with proteomic data to determine the protein complement in action under the defined iron-replete and iron-limited growth conditions. Figure [4b](#F4){ref-type="fig"} illustrates the dynamic range of differential abundances for all proteins detected by liquid chromatography-tandem mass spectrometry (LC-MS/MS) relative to equal amounts of total cellular protein for both conditions. The induction of flavodoxin is a hallmark of iron-deficiency responses in many diatoms and cyanobacteria (see above). In accordance with the transcriptome response, flavodoxin as well as ISIPs or class I FBAs could only be found under iron limitation. The extent of correlation between proteomics and transcriptomics data was assessed through plotting the relative abundance data from peptides (proteomics P) against the relative abundance data from their corresponding transcripts (transcriptomics T) (Figure S4 in Additional file [3](#S3){ref-type="supplementary-material"}). A stretched cluster along the y-axis indicates a high dynamic range of the transcriptomics data, while the proteomics data is more uniform for this group. Both transcriptomics and proteomics data are biased towards highly abundant transcripts/proteins. Especially the proteomics data, despite its relatively high number of signals, could resolve only a subset of the protein complement. Accordingly, we interpret the complement of differentially regulated genes and proteins recovered from both approaches as complementary in the information that they provide, and we do not expect them to show a complete overlap. However, the overlap in the response for the specifically induced proteins ISIP1 and class I FBAs shows that the data from both approaches are, in general, in good agreement with each other. In the proteomics data it is of specific interest to have a closer look at proteins of the photosynthetic machinery. Chloroplast ribosomal proteins provide an appropriate internal reference for the regulation of chloroplast proteins and indicate a down-regulation of the ribosomes at a ratio of 0.8 relative to the iron-replete proteome. Protein subunits of PS I were reduced about two-fold under low iron conditions (0.45), except PsaL, which was only found under iron limitation. In cyanobacteria, PsaL, generally important for trimer formation, facilitates the formation of IsiA (iron stress induced protein A) rings around PS I monomers under iron-deprivation \[[@B34]\]. We speculate that PsaL might be involved in the organization of PS I light-harvesting structures specifically formed under low-iron conditions and/or oligomerization of PS I in iron-limited *T. oceanica*. Subunits of the iron-containing cytochrome b~6~/f (cyt b~6~/f) complex, were down-regulated, with ratios of 0.2 and 0.32. In contrast, PS II subunits PsbB, PsbC, PsbE, PsbH and PsbV remained almost constant, with ratios at about 1.1. While the PS II core complex seems to be retained to some extent, the labile D1 protein is down-regulated at 0.7, probably reflecting a proportional decrease in functional PS II. The differential regulation of the two photosystems (0.45 for PS I versus 0.7 for PS II D1 protein) supports an adaptive significance for the remodeling of the photosynthetic architecture under iron limitation, in contrast to earlier findings \[[@B12]\]. While PS I and cyt b~6~/f complexes were down-regulated two- to threefold, it was still possible to detect the iron-rich mitochondrial complexes under iron limitation. Relative protein quantification was possible for subunits of complex III, complex IV and the ATPase with low-to-high iron ratios ranging from 0.95 for QOR2 (a NADPH-dependent quinone oxidoreductase) to 1.7 for the beta subunit of the mitochondrial cytochrome c oxidase (Figure [4b](#F4){ref-type="fig"}). This is in agreement with the transcriptomic data and supports the idea that mitochondrial electron transfer protein complexes are preserved under iron limitation relative to photosynthetic electron transfer protein complexes. While the magnesium chelatase, involved in chlorophyll synthesis, is down-regulated at 0.35, the numerous FCP light-harvesting proteins showed very diverse responses to iron limitation (Figure S5 in Additional file [3](#S3){ref-type="supplementary-material"}). Some FCPs showed down-regulation under low-iron whereas others were up-regulated. In particular, LHCSR-like FCPs, involved in photoprotection, were highly abundant under iron limitation, corroborating the transcriptome analysis. Notably, the xanthophyll cycle enzyme violaxanthin de-epoxidase showed significant up-regulation at 3.1, suggesting a possible linkage to the group of FCPs, which accumulate under iron limitation. Comparative genomics reveals extensive genomic plasticity in *T. oceanica* -------------------------------------------------------------------------- We used the genome information of *T. oceanica*, *T. pseudonana*, *P. tricornutum*and *F. cylindrus*to investigate central issues of the diatom low-iron response in a comparative genomics approach. ### Taxonomic distribution of iron-regulated genes We screened the four diatom genomes known to date (*T. oceanica*, *T. pseudonana*, *P. tricornutum*and *F. cylindrus*) for the highly conserved iron-regulated *ISIP1*, *ISIP3*, *PETF*, *FLDA*, *CYTC6*, *PETE*and class I and II *FBA*genes (Table [3](#T3){ref-type="table"}; Additional file [9](#S9){ref-type="supplementary-material"}). Phylogenetic trees for the important groups of flavodoxin \[[@B35]\] and FBA proteins are provided in Figures S6 and S7 in Additional file [3](#S3){ref-type="supplementary-material"}. ###### Presence and copy number of iron-regulated genes in the genomes of ecologically distinct diatoms Gene Product Destination Mutual substitution at low-iron Putative role in iron uptake To Tp Pt Fc ------------------------ ----------------------------------------- ------------------------ ---------------------------------------------- ------------------------------ ---- ---- ---- ---- *PETF* Ferredoxin CP Ferredoxin → flavodoxin (short) 1 1 1 1 *FLDA*(s) Flavodoxin (short) CP Ferredoxin → flavodoxin (short) 2 0 1 1 *FLDA*(l) Flavodoxin (long) SP (ER?) None (distinct functional context) 1 1 1 1 *CYTC6*(type A) Cytochrome c6 CP Cytochrome c6 (type A) → plastocyanin 2 1 1 1 *CYTC6*(type B) Cytochrome c (?) SP (ER?) None (distinct functional context) 1 1 1 0 *PETE*/*PCY* Plastocyanin CP Cytochrome c6 (type A) → plastocyanin 1 0 0 1 Class II *FBA*(type A) Class II fructose-bisphosphate aldolase CP pyrenoid (Pt FBAC1) Class II FBA (type A) → class I FBA (type A) 1 1 1 1 Class II *FBA*(type B) Class II fructose-bisphosphate aldolase CP stroma (Pt FBAC2) Class II FBA (type B) → class I FBA (type B) 1 1 1 1 Class II *FBA*(type C) Class II fructose-bisphosphate aldolase Cytosolic (Pt FBA3) Class II FBA (type C) → class I FBA (type C) 1 1 1 1 Class I *FBA*(type A) Class I fructose-bisphosphate aldolase CP pyrenoid (Pt FBAC5) Class II FBA (type A) → class I FBA (type A) 1 0 1 1 Class I *FBA*(type B) Class I fructose-bisphosphate aldolase CP stroma Class II FBA (type B) → class I FBA (type B) 1 0 0 1 Class I *FBA*(type C) Class I fructose-bisphosphate aldolase Cytosolic (Pt FBA4) Class II FBA (type C) → class I FBA (type C) 1 1 1 1 *ISIP1* Iron starvation induced protein 1 Cell surface Receptor (?) 2 0 1 3 *ISIP3* Iron starvation induced protein 3 Cell surface Co-receptor (?) 1 1 1 2 The coastal diatom species *T. pseudonana*(Tp) lacks several genes that are found in the genomes of diatoms with high tolerance to low-iron conditions (*T. oceanica*(To), *P. tricornutum*(Pt), *F. cylindrus*(Fc)). Listed are also the respective counterparts whose products are subject to substitution under iron-limited conditions. The conserved paralogous genes of *FLDA*and *CYTC6*are predicted to contain a signal peptide and are assumed to act in a different functional context. CP chloroplast; ER endoplasmic reticulum; SP, secretory pathway. The short flavodoxin isoform, plastocyanin and the class I FBAs are known or assumed to replace iron-containing counterparts under low-iron conditions. The two oceanic diatoms *T. oceanica*and *F. cylindrus*, which have some of the highest tolerance to low-iron conditions, both contain all five of the respective genes while *P. tricornutum*lacks two of them. The typical coastal species *T. pseudonana*lacks all except the gene for the cytosolic class I FBA, while at the same time having the highest requirement for iron in the group of diatoms for which genome information is currently available. Further, we find multiple copies of the *ISIP1*gene in *T. oceanica*and *F. cylindrus*, while this gene is absent in *T. pseudonana*. The presence or copy number of these genes in the tested diatom genomes suggests an adaptive significance with respect to the low-iron conditions found in oceanic waters. ### Domain duplications of iron-regulated cell-surface proteins While differentially regulated genes for cell-surface proteins, identified from the low-iron response of *P. tricornutum*\[[@B20]\], like *ISIP1*, *ISIP2*, *FLDA*or *CREG*, represent single-copy genes encoding well-defined single-domain proteins, the situation in *T. oceanica*is different (Figure S2 in Additional file [3](#S3){ref-type="supplementary-material"}). Here, we find additional paralogous versions of several iron-regulated genes (*ISIP1*, *FLDA*), as well as diverse examples of domain duplications (*CREGx2*, *ISIP2x8*). In the case of iron-binding proteins the duplication of domains might provide benefits under iron limitation through a higher density of exposed domains, thereby increasing the affinity for iron at the cell surface \[[@B26]\]. With respect to the selective pressure encountered in the low-iron open ocean the duplication of complete genes may provide a possible mechanism for adaptation on the molecular level, in that it allows one of the two gene copies to vary, improve and optimize its iron-binding themes/motifs. This may potentially result in more efficient iron uptake. RT-qPCR allowed us to distinguish iron-regulated genes from their closely related paralogs (Figure S8 in Additional file [3](#S3){ref-type="supplementary-material"}). ### Iron uptake and the cell-surface protein ISIP1 Conservation between the predicted protein orthologs of ISIP1 in *T. oceanica*, *P. tricornutum*and *F. cylindrus*was high, and the orthologs exhibited identical secondary structure predictions (Figure [6](#F6){ref-type="fig"}). We found an amino-terminal signal peptide targeting the protein to the secretory pathway, while a carboxy-terminal transmembrane domain anchors the protein to a membrane. The major part of the protein is represented by a domain rich in β-strands that likely folds into a β-propeller-like structure. A clue to the structure and function of ISIP1 could be the low-density lipoprotein receptor LDLR, an important cell-surface receptor in humans \[[@B36]\]. Although its extracellular domains differ from the single β-propeller domain of ISIP1, the remainder of the protein is strikingly similar with regard to amino acid composition and secondary structure prediction. Hence, we may transfer the respective LDLR annotation to the ISIP1 protein model. {#F6} Accordingly, the ISIP1 protein would represent a cell-surface receptor, anchored to the plasma membrane by a carboxy-terminal transmembrane helix. A small carboxy-terminal tail without well-defined secondary structure contains a conserved endocytosis motif responsible for endocytotic cycling. An α-helical region amino-terminal from the transmembrane helix is predicted to be O-glycosylated and would thereby serve to expose the large β-propeller as a putative receptor domain to the extracellular space. An alignment of the ISIP1 proteins from *T. oceanica*, *P. tricornutum*and *F. cylindrus*illustrates that the extracellular β-propeller domain contains a cysteine-rich center (Figure [6](#F6){ref-type="fig"}) whose pattern is reminiscent of cysteines found in Fe-S cluster proteins and might be involved in binding Fe. The cysteine-rich center is not found in the orthologous p130B of *Dunaliella salina*, which is thought to have undergone an evolutionary change in function and interacts with transferrin-like proteins \[[@B37]\]. ### A conserved promoter motif associated with diverse iron-regulated genes At the core of an organism\'s low-iron response are transcription factors (repressors or enhancers) and their respective binding sites (specific promoter motifs) that mediate the cellular response at the gene expression level. From pairwise promoter comparisons between the exclusively iron-regulated *ISIP1*, *FLDA*and *FBA3*genes of *T. oceanica*, *P. tricornutum*and *F. cylindrus*using dotlet \[[@B38]\] and MEME \[[@B39]\], we identified a conserved palindromic motif \'ACACGTGC\' located around position -200 from the translation start. Upon genome-wide screening, a total of 45 gene models contained the complete motif (perfect match) at a position of 150 to 250 bases before the translation start. Functional assignments for genes with positive matches were rarely possible (mostly hypothetical genes of unknown function and without significant regulation). However, the accumulation of low-iron responsive genes (*ISIP1*and three *FBA*genes) in this group is remarkable. In Figure S9 in Additional file [3](#S3){ref-type="supplementary-material"} we present only those genes whose orthologs in other species carry the motif in their promoters. The complexity of the identified motif (A~2~T~1~C~3~G~2~) is high; its palindromic structure suggests binding of a homo- or heterodimeric protein factor. The remarkable conservation of this motif and its position (-200) relative to the translation start across three diatom species reinforces the suggestion that this motif plays a prominent role in iron-dependent gene regulation. Discussion ========== The cellular response to iron limitation ---------------------------------------- As the most prominent part of the complex low-iron response of *T. oceanica*we observe clear indications for an extensive cellular retrenchment, best seen in the reduced number and size of the chloroplasts. The proteomics and qPCR results indicate that not only is the iron-rich photosynthetic machinery affected, but that other cellular components also encounter a large-scale reduction, resulting in decreased growth rates. In addition, low-iron cells have a significantly lower protein biomass at roughly half that of iron-replete cells. We speculate that during the transformation from a high-iron/high-biomass cell to a low-iron/low-biomass cell the cellular biomass itself may serve as a supplementary energy source to compensate for the decrease in photosynthetic performance, that is, carbon fixation and generation of ATP. This is in line with the observation of increased cellular vesiculation in low-iron conditions (light microscopy) and increased lipid metabolism (transcriptomics and proteomics). It is also consistent with the relative increase of the mitochondrial respiratory machinery as deduced from the transcriptomics and proteomics data. As phototrophy, and thereby chloroplast energy metabolism, is largely impaired in low-iron conditions, mitochondrial respiration may provide a more constant and robust energy source that is retained, that is, excluded from the observed cellular retrenchment. Hence, we can describe the principal cellular changes observed at iron limitation as a metabolic shift with a gradual take-over of the energy metabolism by the mitochondrial system. Cellular maintenance under iron-limited conditions is further supported from biomass recycling through the action of isocitrate lyase and glutamine synthetase. Moreover, changes in the photosynthetic machinery are likely a consequence of a coordinated remodeling process, indicating an intricate regulatory network that adjusts cellular energy demand in response to the availability of iron. The active remodeling as a response to low iron is an unexpected result since it was proposed earlier that photosynthesis of *T. oceanica*is constitutively adapted to a low-iron environment \[[@B12]\]. The observation of distinct cellular phenotypes in iron-limited versus iron-replete cultures may be due to the differing iron levels used in iron-replete control cultures, with approximately 60 nM applied by Strzepeck and Harrison \[[@B12]\] compared to the saturating 10 µM FeCl~2~used in this work, though not expected to ever occur under natural conditions. Nevertheless, our data demonstrate that *T. oceanica*possesses the potential to remodel its bioenergetic pathways in response to iron availability. While the cultures in the present experimental setup were held under axenic photoautotrophic conditions, the situation in the natural context of the open ocean is very different. A major difference found between the artificially induced iron limitation of photoautotrophically growing cultures and the iron limitation encountered by diatoms in their natural habitats is the ubiquitous presence of particulate and dissolved organic matter in the latter, albeit dilute as in the case of the oligotrophic Sargasso Sea \[[@B40]\]. Some of the strongest up-regulated transcripts under low-iron conditions were found to be targeted to the secretory pathway, that is, with the cell surface or the vacuolar system as their destination. The functional annotation of the respective protein models reveals a complex suite of molecules capable of adhesion or degradative functions, suggesting a possible role in a mixotrophic context (Figure [5](#F5){ref-type="fig"}). Observations of diatom mixotrophy have been reported for five decades \[[@B41]\]. This characteristic has recently been explored for biotechnological application \[[@B42]\]. A principle metabolic competency for heterotrophic nutrition was demonstrated in transgenic *P. tricornutum*that were able to grow on sugar upon expression of a transgenic sugar transporter \[[@B43]\]. A hypothetical capability for mixotrophic nutrition and the conjoined ability to feed on dissolved and/or particulate organic matter would place diatoms in close proximity to the microbial loop responsible for recycling organic matter in the marine food web. Moreover, this would contribute to resolving the paradox of diatom survival in a low-iron world, as particulate and dissolved organic matter can be expected to be a relevant iron source. On the other hand, utilization of organic carbon/iron sources in low-iron conditions will immediately put the cell into competition with the bacterial community of diverse and specialized heterotrophs. In line with this scenario we find an up-regulation of both identified NRPS enzymes (Figure S8 in Additional file [3](#S3){ref-type="supplementary-material"}) that may mediate a bacterial defense response under iron-limited conditions. Evolutionary roots and the impact of genome plasticity on adaptation to low iron -------------------------------------------------------------------------------- Diatoms are the outcome of an endosymbiotic event that brought together two nutritional modes, each somewhat pre-adapted to low-iron conditions: phototrophy was contributed by the red-type plastids of the red algal endosymbiont, whose elemental composition shows a reduced iron requirement relative to the green-type plastids \[[@B8]\]; the host (related to ancient oomycetes) contributed an efficient heterotrophic machinery that might be retained to some extent in today\'s diatoms (compare \[[@B44]\]), thereby enabling exploitation of particulate and dissolved organic matter as a supplementary carbon and iron source. Survival under iron-limited conditions would have benefited from both the ancestral host\'s and symbiont\'s characteristics. Additional adaptive strategies to the low-iron environment have probably also originated by means of horizontal gene transfer, with possibly as much as 5% of the conserved genes encoded in the *T. oceanica*genome being assigned to diverse taxonomic groups. This is consistent with the findings from the *Phaeodactylum*genome project \[[@B5]\] and appears to be a recurrent topic in diatom genomics. A prerequisite for integration of foreign DNA into a genome is its uptake, and an explanation for how lateral gene transfer might occur in heterotrophic cells has been proposed \[[@B45]\]. Accordingly, heterotrophic organisms would continuously take up and integrate new genes along with the organic matter they feed on, thereby creating some genetic redundancy that eventually leads to the permanent replacement of the genuine counterparts as long as no disadvantage is encountered. We take the proposed mechanism of lateral gene transfer, together with the extent of laterally acquired genes observed in *T. oceanica*, as additional evidence for mixotrophic potential in *T. oceanica*. In low-iron conditions we observe the joint up-regulation of a reverse transcriptase and a CRE-like recombinase, thereby providing an appropriate mechanistic basis for genome rearrangements via transposon mobilization. Moreover, these are activated at a time where enhanced DNA input through increased uptake of organic matter might be expected under natural conditions. Stress-induced transposon activation has been reported for higher organisms as well - for example, \[[@B46]\]. The observed plasticity of diatom genomes clearly has ecological implications, as bacterial inventions such as genes that are beneficial in a competitive context might quickly find their way into diatom species and strains. We therefore state that the hypothesized close integration of diatoms within the microbial loop (due to mixotrophy) and their remarkable genomic plasticity (as seen from lateral gene transfer) are keys to diatoms\' ecological success: while mixotrophy opens up complex sources for carbon, energy and nutrients, the high capacity for lateral acquisition of genetic material facilitates adaptation in the context of the resulting competition with bacteria for organic matter, nutrients and iron. Iron uptake, cellular iron requirements and adaptation of species to low iron ----------------------------------------------------------------------------- For a better characterization of the complex cellular low-iron response it is necessary to distinguish the iron-specific aspects directed at counteracting the shortage of the limiting nutrient (substitution of iron proteins, induction of high-affinity iron uptake) from the rather general stress response directed at the situation of impaired growth and cellular starvation (cellular retrenchment, chlorosis, metabolic shift). With regard to the cellular iron economy we can define some constraints for beneficial adaptations to low-iron conditions \[[@B26]\]. Each evolutionary innovation that lowers the cellular iron requirements will favor survival under iron-limited conditions. Besides general biomass retrenchment - that is, the cellular adoption of a state with decreased biomass - the cell can actively lower its iron requirements through substitution of iron-containing proteins like ferredoxin or other metal enzymes that use iron as a cofactor. In addition to the well-known substitution pair ferredoxin/flavodoxin \[[@B10]\], we present evidence for the replacement of three metal-containing FBA enzymes by substitutes that use an amino acid-based Schiff-base catalysis instead. Differential regulation of diatom FBA genes was recently described for *P. tricornutum*and appears to be an evolutionarily conserved feature of several diatoms \[[@B27]\]. In *T. oceanica*we even find the apparently permanent functional replacement of an iron-rich counterpart as in the case of plastocyanin, though at the same time the genome harbors two genes encoding cytochrome c~6~whose function and regulation remain unknown. Cytochrome c~6~might be retained in the genome for replacing plastocyanin under specific circumstances like copper limitation, though this could not yet be observed \[[@B11]\]. Benefits may also arise from improved regulation of mutual substitution pairs - for example, as in the case of the transfer of the usually organellar ferredoxin gene *petF*to the nuclear compartment as found in *T. oceanica*\[[@B17]\]. Further, any improvement of the cellular iron-affinity and/or iron-uptake system will improve competitive fitness under low-iron conditions. The activation of a specific high-affinity uptake system as observed in yeast is expected to occur in diatoms as well. It has been found that reduction of Fe^3+^to Fe^2+^represents an essential step in uptake of organically complexed iron \[[@B47]\], suggesting that iron is extracted from its complexes prior to uptake. Accordingly, major players involved in iron uptake can be expected to be iron-complex-binding receptors, redox enzymes needed for extracting the iron from its complexes, possibly also specific iron-binding cell-surface molecules for short-term iron storage (like *D. salina*transferrins), ferric reductases and ferrous oxidases responsible for interconversion of the iron redox species +III and +II, and finally iron permeases for iron import. In this work on *T. oceanica*we were able to identify several candidate elements for the above groups. However, critical for iron-uptake kinetics is the overall iron-binding capacity of the cell surface, which directly depends on the sheer amount of iron-binding sites exposed to the cellular exterior \[[@B26]\]. A straightforward strategy to enhance the cellular capacity for iron binding is seen in the remarkable extent of domain duplications in iron-regulated cell-surface proteins. From the strong and exclusive expression in iron-limited conditions we speculate that ISIP1/ISIP3 are part of a specialized high-affinity iron-uptake system, with ISIP1 as the putative receptor (Figure [6](#F6){ref-type="fig"}). Carboxy-terminal endocytosis motifs as seen in ISIP1 (Figure [6](#F6){ref-type="fig"}) can also be found in other iron-regulated proteins (*T. oceanica*CREGx2, *P. tricornutum*ISIP2) and biochemical work is needed to confirm the location and proposed receptor function for these components. How far such features are species-specific adaptations or rather common to diatoms can only be clarified by a comparative genomic approach. Notably, screening for highly conserved iron-regulated genes in the genomes of *T. oceanica*, *T. pseudonana*, *P. tricornutum*and *F. cylindrus*revealed a correlation between the extent of a diatom\'s tolerance to low-iron and the presence of *ISIP1*and *PETE*, which directly impact cellular iron economy and uptake. Conclusions =========== From their evolutionary roots diatoms already appear to be pre-adapted to low-iron conditions through the endosymbiotic acquisition of a \'red\'-type photosynthetic machinery. While they have retained an implicit capacity for mixotrophy, the main contribution to cellular growth under conditions where iron or other nutrients limit the build-up of biomass, like in the iron-limited southern ocean or the oligotrophic Sargasso Sea, can be expected to stem from photosynthetic carbon assimilation. The combined efforts in genomics, transcriptomics and proteomics reveal an unexpected metabolic flexibility in response to iron availability for *T. oceanica*CCMP1005. These responses include an extensive cellular retrenchment, the pronounced remodeling of bioenergetic pathways and an intrinsic shift to a mixotrophic life style. As a consequence of iron deprivation, the photosynthetic machinery undergoes remodeling into a photo-protected mode to cope with the overall decrease in photosynthetic electron transfer complexes. From the genomic and transcriptomic data we identify candidate components of a diatom high-affinity iron-uptake system, and we present a novel cellular strategy to enhance iron economy of phototrophic growth with the iron-regulated mutual substitution of three metal-containing FBA proteins. The enormous genomic plasticity of *T. oceanica*, as seen from the large fraction of genes acquired through horizontal gene transfer, provides a platform for complex adaptations to the iron-limited ocean. The inferred dynamic exchange of genes between marine microbes suggests the genome of *T. oceanica*may not be an exceptional evolutionary invention, but rather that it may be seen as one possible outcome from a larger metagenomic gene pool. The future comprehensive characterization of this gene pool constitutes the ultimate challenge in appreciating the solutions that marine life found for defying the persistent shortage of iron in the open ocean. Materials and methods ===================== Strains, cultures and physiology -------------------------------- The sequenced strain *T. oceanica*(Hustedt) Hasle et Heimdal CCMP1005 \[[@B48]\] was obtained from the Center for Culture of Marine Phytoplankton \[[@B49]\]. *T. oceanica*cells were grown in 8 l batch cultures using iron-free f/2 nutrients in artificial seawater medium (ASW) \[[@B50]\] at 100 µE, 25°C and a 14/10 h light/dark cycle. Iron-replete cultures (\'control\') were supplied in excess of other essential nutrients (10 µM FeCl~3~); no iron was added to the iron-limited cultures (\'stress\'), except for residual iron from the ASW salts, promoting iron-limited growth. Cells were harvested at late exponential growth phase by filtration on 5 µm polycarbonate filters of 47 mm diameter, resuspended into a small volume of media, followed by centrifugation at 4°C for 10 minutes at 11,000 rpm. Cell pellets were frozen in liquid N~2~and stored at -80°C. For iron-limited cultures, iron-free techniques were applied as follows. Culture bottles were composed of plastic material, washed and incubated for some days with 1 N HCl and rinsed with ultrapure MilliQ water. All additional supply for iron-free work was washed with 1 N HCl and stored in closure bags until use. Iron-limited cultures were best achieved from a minimal inoculation volume of 10 to 20 µl or less than 10,000 cells. Throughout this work we compare cells from late exponential growth phase, though we recommend to use iron-limited cells from late stationary phase when working on specifically low-iron responsive genes (for example, from compensation pairs or involved in iron uptake). For these, the expression level at the late stationary phase is found to be even higher than in the late exponential phase. Total cellular protein was determined for iron-replete and iron-limited cells as follows. First, 85.5 million cells each were concentrated by centrifugation. The resulting pellets were frozen in liquid N~2~and stored at -80°C. For protein determination, pellets were re-dissolved and lysed in 200 to 300 µl SDS/CO~3~buffer with additional application of ultrasonication. Cell debris was precipitated by 4 minutes of centrifugation at room temperature and maximum rpm; 5 to 10 µl of the supernatant served as input for bicinchoninic acid (BCA) protein assay (Thermo Fisher Scientific, Waltham, MA, USA). SDS/CO~3~buffer was composed of 4% SDS, 68 mM Na~2~CO~3~and 0.4 mM phenylmethylsulfonyl fluoride (the latter was dissolved in EtOH to obtain a 100 mM stock solution). The variable chlorophyll fluorescence F~v~/F~m~of *T. oceanica*cells was measured from fresh cultures with a PhytoPAM (PHYTO-PAM Phytoplankton Analyzer, Heinz Walz GMBH, Effeltrich, Germany) \[[@B51]\] upon 5 minutes of dark incubation. Comparative genomics was done with genome data available at the Joint Genome Institute for *T. pseudonana*(Hustedt) Hasle & Heimdal CCMP1335 \[[@B4]\], *P. tricornutum*Bohlin CCAP1055/1 \[[@B5]\] and *F. cylindrus*(Grunow) Krieger CCMP1102 \[[@B19]\]. Microscopy and confocal microscopy ---------------------------------- *T. oceanica*was imaged *in vivo*using confocal laser scanning microscopy LSM 510 (Zeiss). Chlorophyll autofluorescence was excited at 488 nm (1% laser intensity), and emission recorded with a long pass (LP) 650 nm filter. Images were made using a Plan-Neofluar 40×1.3 oil objective (Zeiss). Z-section image series were captured with LSM 510 v3.2 software (Zeiss). Three-dimensional reconstructions of the chlorophyll fluorescence signal were made using the cell surface area-/cell volume-analyzing \'Surpass\' program module in Imaris 7.1.1 (Bitplane, Zürich, Switzerland). Images were segmented using consistent threshold values. Surface area grain size was set at 0.1 µm. In 20 cells from both iron-replete and iron-limited cultures, cell dimensions were calculated from transmission image measurements based on a cylinder model. Chloroplast dimensions were calculated from three-dimensional chlorophyll autofluorescence signal reconstructions. Despite imaging in a narrow time window (12:00 ± 2 h), growth in the *T. oceanica*cultures was not synchronized. Thus, the observed values for the parameters with a growth-dependent variability are accordingly distributed over a characteristic growth range. The representative cellular and chloroplast dimensions provided in Table [2](#T2){ref-type="table"} were therefore determined from the statistical mean of 20 cells by calculating back to a cell at the beginning of its cell cycle. For this purpose we used a regular cylinder that expands through a gradual two-fold increase of its height as a model for the diatom cell. Around the statistical mean from all cells (of a variable parameter like cell volume) we created a range whose higher end differs from the lower end by a factor of 2, thereby representing a two-fold increase of that parameter during growth of the diatom cell. The lower end of that range is given in Table [2](#T2){ref-type="table"} as a representative value for a \'cellular unit\', that is, a freshly divided cell at the beginning of its cell cycle. Nucleic acid extraction and sequencing -------------------------------------- *T. oceanica*CCMP1005 was grown as axenic clonal culture from a single cell isolate obtained from serial dilutions of a stock culture to extinction. Nuclear gDNA for sequencing of the *T. oceanica*genome was extracted from nutrient-replete cells and separated from organellar gDNA in a CsCl gradient (Figure [1](#F1){ref-type="fig"}) using the alternative cetyltrimethlyammonium bromide protocol for algae \[[@B52]\].The quality of nucleic acids was assessed from NanoDrop UV absorption profiles and agarose gel electrophoresis. Second generation sequencing technology was applied to the gDNA as follows. After mechanical shearing by nebulization, followed by end-repair, specific sequencing adaptors were ligated. The genomic DNA fragments were shotgun sequenced using massive parallel pyrosequencing \[[@B53]\] on a Roche 454 GS-FLX instrument (Roche, Penzberg, Germany) according to the manufacturer\'s protocol. Total RNA for transcriptome sequencing was extracted from frozen pellets of iron-replete and iron-limited *T. oceanica*cells from late exponential growth phase using the QIAGEN (Hilden, Germany) RNeasy kit. RNA quality was assessed from NanoDrop UV absorption profiles and agarose gel electrophoresis. For reverse transcription of total RNA the SMART cDNA synthesis kit from Clontech (Mountain View, CA, USA) was used with 1 µg input material and 15 rounds of amplification. The size distribution of the obtained cDNA libraries were controlled with agarose gel electrophoresis and then subjected to Roche 454 sequencing as described above for gDNA. Transcriptomics --------------- Global gene expression was assessed through Roche 454 massive parallel pyrosequencing of cDNA libraries prepared from total RNA extracted from iron-limited and iron-replete cultures. The 2× 95,000 sequence reads from both libraries were pooled, cleaned from adapter ends and processed in a combined assembly revealing 11,264 contigs (that is, transcript fragments) that map to approximately 6,500 distinct AUGUSTUS gene models (Additional file [10](#S10){ref-type="supplementary-material"}). The differential read contribution from the Fe(-) and Fe(+) libraries to each contig is taken as a measure for the differential transcription of the respective gene. For the purpose of statistically evaluating the gene expression level across our two cDNA libraries, we applied a log-likelihood ratio test statistic as described in Stekel *et al*. \[[@B22]\]. Differentially regulated genes were first screened with T-ACE \[[@B54]\], a transcriptome database browser that plots the assembled transcript fragments according to their differential regulation (Figure [4a](#F4){ref-type="fig"}) and provides information from BLAST analyses against the NCBI nr protein (NR) database and the Conserved Domain Database. Models from selected genes were manually curated and annotated for protein function and location as described above (Additional file [6](#S6){ref-type="supplementary-material"}). RT-qPCR ------- RT-qPCR was done as in \[[@B17]\] and is described in the Supplementary Methods in Additional file [3](#S3){ref-type="supplementary-material"}. Primers are listed in Additional file [11](#S11){ref-type="supplementary-material"}. The differential regulation between high and low iron conditions with respect to 18S and RPB1 (threshold level) is shown in a ΔΔC~T~plot as calculated below: $$\Delta\Delta\textsf{C}_{\textsf{T}} = \left\lbrack {\Delta\textsf{C}_{\textsf{T}}@\textsf{Fe(~+~)~-~}\Delta\textsf{C}_{\textsf{T}}@\textsf{Fe(~-~)}} \right\rbrack$$ where ΔC~T\ =~(C~T~gene 1 - C~T~gene 2) and represents the difference between the qPCR threshold cycle values (C~T~) of gene 1 (the gene of interest) and gene 2 ( the house-keeping gene, either the 18S rRNA or RPB1). (Figure S8 in Additional file [3](#S3){ref-type="supplementary-material"}). Proteomics ---------- The proteomes of iron-replete and iron-limited cells were analyzed in a mass spectrometry approach. Differentially labeled iron-sufficient and iron-deficient cells were mixed at equal protein concentration and fractionated by SDS-PAGE. Protein bands were excised, digested in-gel with trypsin and subjected to LC-MS/MS analyses. Identification and quantification of peptides and proteins was performed using Proteomatic data evaluation pipelines \[[@B55]\] as follows. To provide candidate peptides in the database search step using OMSSA \[[@B56]\], several sequence sources were used: (1) the AUGUSTUS gene models, (2) Genomic Peptide Finder peptides \[[@B57]\], (3) high quality chloroplast protein models, (4) a set of manually curated protein models. Resulting peptide/spectral matches were filtered with a hit distinctiveness filter, using a threshold of 2. Peptide/spectral matches were further filtered with a dynamically determined E-value threshold to achieve an estimated false discovery rate of 1% \[[@B58]\]. Finally, all peptide/spectral matches with a precursor mass deviation \>5 ppm were discarded. A total of 1,695 peptides could be identified from two independent biological experiments and assigned to 767 protein groups. All identified peptides were subsequently quantified using qTrace \[[@B59]\], resulting in the quantification of 633 protein groups and an additional set of 88 quantified peptides, which were identified exclusively via Genomic Peptide Finder. For the determination of Fe(-)/Fe(+) protein ratios, all resulting combinations of peptide, SDS-PAGE band and charge state were grouped and all group ratios were combined into a total protein group ratio by calculating the median and interquartile range. Data for differential protein expression as revealed from the mass spectrometry approach refers to equal amounts of total protein. For relating the observed regulation to a \'cellular unit\', the ratio of cellular protein biomass \[Fe(-)/Fe(+)\] as determined from a BCA (bicinchoninic acid) protein assay needs to be taken into account. Bioinformatics -------------- Bioinformatic analyses are described in the Supplementary Methods in Additional file [3](#S3){ref-type="supplementary-material"}. Data access ----------- The *T. oceanica*CCMP1005 whole genome shotgun assembly is registered as bioproject 36595, the *T. oceanica*CCMP1005 transcriptome shotgun assembly is registered as bioproject 73029. The genomic and transcriptomic Roche 454 GS FLX sequence reads from this study have been submitted to the NCBI Sequence Read Archive \[[@B60]\] under accession numbers SRA045826 and SRA045825, respectively. The whole genome shotgun project has been deposited at DDBJ/EMBL/GenBank under the accession AGNL00000000. The version described in this paper is the first version, AGNL01000000. AUGUSTUS gene models deduced from the genome assembly have been assigned the gene loci accession numbers THAOC_00001 to THAOC_37921. The transcriptome assembly has been submitted to the NCBI Transcriptome Shotgun Assembly database \[[@B61]\] under accession numbers JP288099 to JP297710. The proteomics mass spectrometry mzML data associated with this manuscript may be downloaded from the Tranche repository \[[@B62]\] using the following hash: \'T9vKohxPxffmhOBsgb9kTlBKCrQIQYziH8hdonm9scou13EAFv57Uo+XYTj4d8XHbLRxR03+6WeDRSp2yhpp348wzWsAAAAAAABjUg==\'. The data from this study can be accessed in an integrated form with the *Thalassiosira oceanica*Genome Browser \[[@B18]\]. Abbreviations ============= CCAP: Culture Collection of Algae and Protozoa; CCMP: Culture Collection of Marine Phytoplankton; EST: expressed sequence tag; FBA: fructose-bisphosphate aldolase; FCP: fucoxanthin-chlorophyll a/c-binding protein; Fe(-): iron deplete; Fe(+): iron replete; gDNA: genomic DNA; HNLC: high-nitrate low-chlorophyll; HSF: heat shock factor; ISIP: iron starvation induced protein; LC-MS/MS: liquid chromatography-tandem mass spectrometry; LDLR: low-density lipoprotein receptor; LGT: lateral gene transfer; LHCSR: light harvesting complex stress responsive subunit; NCBI: National Center for Biotechnology Information; NRPS: non-ribosomal peptide synthase; ORF: open reading frame; PCR: polymerase chain reaction; PS: photosystem; RT-qPCR: reverse transcription-quantitative PCR. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= ML prepared the *T. oceanica*RNA used for transcriptomics, was involved in genomics, transcriptomics and proteomics data analysis, performed manual gene modeling and annotation, contributed the comparative genomics section, prepared the figures, and drafted the manuscript. MS developed the bioinformatic tools for analysis of the proteomics data and prepared Figure [4b](#F4){ref-type="fig"}. ML and MS conducted the gene prediction on the *T. oceanica*genome assembly using AUGUSTUS. ASR cultured the algae, carried out the RT-qPCR work and commented on the manuscript. LK performed the assembly of the Roche 454 reads and a general blast and domain annotation. RA set up the web-server and Generic Model Organism Database, annotated gene models, and helped with genome and proteome analysis. MAG carried out the confocal microscopy and the three-dimensional rendering of chlorophyll autofluorescence signals. JW and SVB carried out the proteomics experiment and participated in the proteomics analysis. MBS prepared the gDNA libraries, performed the Roche 454 sequencing and generated the initial assemblies. UCK prepared the cDNA libraries for transcriptomics. RGB and RA contributed the phylogenetic analysis of LGT genes. PR coordinated the sequencing and contributed to manuscript writing. MH coordinated the proteomics and contributed to manuscript writing. JLR coordinated the study, isolated the *T. oceanica*gDNA and contributed to manuscript writing. All authors read and approved the final manuscript. Supplementary Material ====================== ###### Additional file 1 **Supplemental table - BLAST analysis of AUGUSTUS-predicted protein models versus the NCBI Non-redundant Protein (nr) database and the Conserved Domain Database**. The table lists the best BLAST hits from a BLASTP analysis of AUGUSTUS-predicted protein models against NCBI nr protein and Conserved Domain Database. The file is in .xls format (compressed to .zip). ###### Click here for file ###### Additional file 2 **Phylogenetic trees for *T. oceanica*LGT candidate genes**. The file comprises 254 unrooted, Newick-formatted trees containing candidate LGT genes as described in the main manuscript. Trees and corresponding support values were generated using FastTree. Novel *T. oceanica*genes are identified in the trees by a reference number followed by \'\_TO\'. Homologs from the \'nr\' database identified using BLAST are given in the form \'Genus_GI\', except for matches to unnamed genera, which are shown as RefSeq GI number followed by \'X\'. Visual inspection of trees was performed by importing this file into FigTree (Andrew Rambaut 2007) and assigning a root as described in the Supplementary Methods in Additional file 3. The file is in .tre format (compressed to .gz). ###### Click here for file ###### Additional file 3 **Supplementary Methods and Figures**. The file provides a description of the RT-qPCR as well as of the bioinformatics methods (genome assembly and analysis, setup of the *T. oceanica*genome browser and phylogenetic analyses). It further contains supplementary Figures S1 to S9. Figure S1: taxonomy of genes acquired by lateral gene transfer (2). An overview of the taxonomic assignments for LGT candidate genes as revealed from a refined phylogenetic analysis. Figure S2: duplications of genes and domains occurring in genes whose regulation is dependent on the cellular supply with iron. Figure S3: metabolic shift. The change in relative abundance of organellar RNA between *T. oceanica*cells subjected to high or low iron. Figure S4: correlation plot P versus T. The extent of correlation between proteomics (P) and transcriptomics (T) data. Figure S5: FCP proteomics. The variation in the response of different FCP light-harvesting proteins to iron limitation. Figure S6: phylogenetic tree of flavodoxin proteins. A neighbor-joining tree of diatom flavodoxin proteins. Figure S7: phylogenetic tree of FBA proteins. A neighbor-joining tree of diatom fructose bisphosphate aldolase proteins. Figure S8: qPCR ΔΔC~T~. The differential regulation of genes in response to low iron. Data represent the change in transcript level for iron-limited *T. oceanica*when compared to an iron-replete control. Figure S9: promoter motif. The conservation across diatom species of a motif found in the promoters of several iron-regulated genes. The file is in .pdf format. ###### Click here for file ###### Additional file 4 **Gene abbreviations list**. The file lists gene abbreviations and gene product descriptions for all genes mentioned in this paper. The file is in .xls format. ###### Click here for file ###### Additional file 5 **Supplemental sequences - curated protein models**. The file lists 436 manually curated protein sequences from nuclear genes (with custom identifiers x1 to x12 and p1 to p455) together with the 158 proteins encoded by the organellar genomes in FASTA format. ###### Click here for file ###### Additional file 6 **Supplemental table - manual transcriptome annotation**. The table lists annotation information for all protein sequences provided in Additional file 5. The file is in .xls format. ###### Click here for file ###### Additional file 7 **Supplemental sequences - selected protein models discussed in the manuscript**. The file lists sequences of proteins that we explicitly refer to in the discussion of the low-iron response. The sequences are provided in FASTA format. ###### Click here for file ###### Additional file 8 **Supplemental table - overview of low-iron responsive genes**. The table contains more detailed annotation information for all protein sequences provided in Additional file 7. Deviation of a custom curated model from its respective official \'t1\'-AUGUSTUS prediction is indicated as \'mod\'. The file is in .xls format. ###### Click here for file ###### Additional file 9 **Supplemental table - comparative genomics of selected genes with a putative adaptive relevance**. The table contains all official database identifiers for the gene products listed in Table [3](#T3){ref-type="table"}. The file is in .xls format. ###### Click here for file ###### Additional file 10 **Supplemental table - BLAST mapping of *T. oceanica*transcript fragments versus NCBI databases and other diatom genomes**. The table provides a comprehensive overview for all 11,264 transcript fragments with read statistics, best BLASTX hits to NCBI nr and Conserved Domain Database, and additional best TBLASTX hits to the genomes of *T. pseudonana*, *P. tricornutum*and *F. cylindrus*, notably including worldwide web link-outs to the respective orthologous genes of these species. The file is in .xls format (compressed to .zip). ###### Click here for file ###### Additional file 11 **Supplemental table - primers used for RT-qPCR**. The table lists the sequences of all primers used in the RT-qPCR experiments. The file is in .xls format. ###### Click here for file Acknowledgements ================ We thank Prof. Stefan Rose-John (Institute of Biochemistry, Christian-Albrechts-University Kiel, Kiel, Germany) for advice in the isolation of nuclear genomic DNA from *T. oceanica*and access to his laboratory and equipment, and Prof. Stefan Schreiber (Institute of Clinical Molecular Biology, Christian-Albrechts-University Kiel, Kiel, Germany) for extensive help in building up sequencing resources in Kiel. Prof. Thomas Bosch (Institute of Zoology, Christian-Albrechts-University Kiel, Kiel, Germany) and Dr Georg Hemmrich-Stanisak (Institute of Clinical Molecular Biology, Christian-Albrechts-University Kiel, Kiel, Germany) provided help with the initial contig assembly. We thank Tania Klüver (Leibniz Institute of Marine Sciences at Kiel University IFM-GEOMAR, Kiel, Germany) for help with the laboratory experiments and culturing of the algae. Dr Dhwani Desai (Leibniz Institute of Marine Sciences at Kiel University IFM-GEOMAR, Kiel, Germany) helped with bioinformatics analyses and setup of the genome browser. The upper left light micrograph in Figure 1 showing *T. oceanica*in valve view is courtesy of CCMP. Sequence data from *F. cylindrus*were produced by the US Department of Energy Joint Genome Institute \[[@B63]\] in collaboration with the user community. This work was supported in part by a DFG grant to JLR (RO2138/6-1) and by funding from the DFG Cluster of Excellence \'Future Ocean\' (EXC 80) to JLR and PR. MH and JLR acknowledge funding from the BMBF \'BIOACID\' (03F0608N) program. RA received funding from EC FP7/2007-2011 under grant agreement number PITN-GA-2008-215157 and EC FP7 Grant \#205419 (ECOGENE).

Handled by Yu Xue Introduction {#s0005} ============ Cancer somatic mutations and viral oncogenes can generate tumor-specific protein sequences that are entirely absent from normal human cells. These novel proteins may result in the formation of tumor-specific antigens (TSAs) [@b0005]. As an important type of TSAs, neoantigens are generated by tumor-specific proteins, and presented by major histocompatibility complexes (MHCs) on cell surfaces through antigen presentation, where they can be recognized by T-cell receptors (TCRs) [@b0010], [@b0015]. Recently, neoantigens have attracted a large amount of attention, because they are potential biomarkers to distinguish tumor cells from normal cells. Neoantigens are of critical importance for cancer immunotherapy in the following two aspects. First, the neoantigen burden and quality can be used to predict therapeutic effects for immune checkpoint blockade therapy, such as blockage of programmed death-1 (PD-1) and cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) [@b0020], [@b0025], [@b0030]. Second, neoantigens can be used as potential targets for cancer immunotherapy, such as personalized cancer vaccines [@b0035], [@b0040] and adoptive cell therapy (ACT) [@b0045]. Therefore, there is an urgent need to identify neoantigens accurately for cancer patients. With the progress of cancer immunogenomics, several kinds of integrated software have been developed for tumor-specific neoantigen detection, such as TSNAD [@b0050] and pVAC-seq [@b0055]. The most critical function of such software is to predict the binding affinities between mutant peptides and human leukocyte antigen (HLA) alleles. To achieve this, a lot of well-acknowledged and popular tools, such as NetMHC [@b0060], NetMHCpan [@b0065], sNebula [@b0070], and HLA-CNN [@b0075], can be used. In addition, several databases can provide necessary information for the development of tools to predict the affinities between peptides and HLA alleles. For example, the Immune Epitope Database (IEDB) is an important immune-related database, providing a large amount of valuable and experimentally-validated information of immune epitopes [@b0080]. The International Immunogenetics Information System (IMGT) offers information about antibodies, TCRs, MHCs, and so on [@b0085]. Taking advantage of existing neoantigen prediction software, several neoantigen-related databases have been built. For example, TRON Cell Line Portal (TCLP) presents potential neoantigens of 1082 cancer cell lines [@b0090]. The Cancer Immunome Atlas (TCIA) presents the relationship between tumor genotypes and immunophenotypes based on 20 solid cancers, and provides a quantitative index for immunotherapy response [@b0095]. With the rapid growth of cancer genomics data, researchers are able to discover potential shared neoantigens across tumor patient populations [@b0100], [@b0105]. In this study, we developed a tumor-specific neoantigen database (TSNAdb v1.0) from pan-cancer immunogenomic analyses. Based on the 7748 tumor samples of 16 tumor types from The Cancer Genome Atlas (TCGA), we predicted the binding affinities between mutant/wild-type peptides and HLA class I molecules. Datasets we used include somatic mutation data of tumor samples from TCGA and the corresponding HLA allele data from TCIA. Two different versions of NetMHCpan, v2.8 [@b0065], and v4.0 [@b0110], were used for prediction. Furthermore, we also conducted extensive analyses and presented detailed information of potential neoantigens generated by somatic mutations, utilizing the related filtering tools embedded in TSNAD [@b0050]. In addition, we employed the recurrent missense mutations in combination with the highly frequent HLA alleles to predict and analyze potential shared neoantigens. Our study would provide a platform to discover putative targets for neoantigen-based cancer immunotherapy. Database content and usage {#s0010} ========================== Data source {#s0015} ----------- We collected somatic mutations and HLA alleles of 7748 tumor samples across 16 tumor types from TCGA (Release7.0, <https://portal.gdc.cancer.gov>) and TCIA (<https://tcia.at/home>), respectively. These tumor samples carry 972,187 missense mutations, among which 18,897 were found recurrently (at least three occurrences in all tumor samples). We selected the top 100 HLA alleles (frequency \>0.5%) of 7748 tumor samples and combined them with the recurrent missense mutations to predict potential shared neoantigens. Moreover, we also extracted 13,459 recurrent missense mutations from 9155 samples derived from the International Cancer Genome Consortium (ICGC) (Release20, <https://icgc.org/>) and 16 highly frequent HLA alleles (frequencies \>5%) from the 1000 Genome Project [@b0115] for the prediction of potential shared neoantigens. Neoantigen prediction {#s0020} --------------------- We took the information on somatic mutations and HLA alleles of each tumor sample and employed NetMHCpan v2.8 [@b0065] and NetMHCpan v4.0 [@b0110] for neoantigen prediction, using the filtering tools embedded in our previously-developed software TSNAD [@b0050]. All the peptides with 8--11 amino acids that contain missense mutations were extracted as mutant peptides, and the corresponding wild-type peptides were extracted as references. We collected the mutant peptides and HLA alleles with binding affinity IC~50~ \< 500 nM (including strong binding with IC~50~ \< 150 nM and weak binding with 150 nM \< IC~50~ \< 500 nM), without consideration of the binding level between their corresponding wild-type peptides and HLA alleles. We then clustered prediction results based on tumor types and calculated the frequencies of shared neoantigens. Compared with NetMHCpan v2.8, NetMHCpan v4.0 is trained based on both binding affinity data and mass spectrometry data, thus adopting stricter criteria for binding prediction. Consequently, 3,707,562 and 1,146,961 neoantigens were predicted by NetMHCpan v2.8 and v4.0, respectively, among which, 716,876 neoantigens were found in both predictions. The potential shared neoantigens based on recurrent mutations and highly frequent HLA alleles were predicted in the similar way. Web interface {#s0025} ------------- To facilitate the utilization of TSNAdb, we have established a web interface to browse and analyze neoantigens. The web interface comprises five main pages ([Figure 1](#f0005){ref-type="fig"}A): (i) Home, (ii) Browse, (iii) Search, (iv) Validation, and (v) Download. In the following context, we exemplify the usage of TSNAdb with the results predicted by NetMHCpan v2.8.Figure 1**An overview of the TSNAdb web interfaceA.** TSNAdb comprises five main components: (i) Home; (ii) Browse; (iii) Search; (iv) Validation, and (v) Download. The distribution of HLA alleles (**B)** and somatic missense mutations (**C)** extracted from TCGA and TCIA are listed. **D.** The average neoantigen loads across different tissues. **E.** Top 20 genes with predicted neoantigens in 7748 samples. **F.** The result of 'Search' page under the selection of "NetMHCpan2.8, *KRAS*, and TCGA". **G.** The result of 'Search' page under the selection of "NetMHCpan2.8, *KRAS*, and ICGC". **H.** Example of validation data from IEDB in 'Validation' page (top) and partial information on 'Download' page (bottom). The predicted binding level 'Strong' indicates strong binding with IC~50~ \< 150 nM. TCGA, The Cancer Genome Atlas; TCIA, The Cancer Immunome Atlas; HLA, human leukocyte antigen; TSNAD, tumor-specific neoantigen detector; ICGC: International Cancer Genome Consortium; IEDB, Immune Epitope Database; WT, wild type. In the 'Home' page, TSNAdb provides a statistical table about the database, including the number of projects, samples, HLA alleles ([Figure 1](#f0005){ref-type="fig"}B), mutations ([Figure 1](#f0005){ref-type="fig"}C), and predicted neoantigens of each tumor type. The table presents 3,707,562 potential neoantigens from the 7748 tumor samples across 16 tumor types. The average predicted neoantigen loads vary across different tumor types ([Figure 1](#f0005){ref-type="fig"}D). For instance, the average predicted neoantigen load for uterus cancer is 1957, which is 21 for thyroid cancer. Besides, users can get the detailed information about the predicted neoantigens of each tumor type by clicking on the tissue name in the table. The content of each tumor type includes top 20 HLA alleles and top 20 genes with predicted neoantigens in this tumor type, which are shown as two figures. All predicted neoantigens in this tumor type would be displayed below figures. To further understand the distribution of neoantigens at gene level, users can retrieve neoantigens by feeding in a gene name in the 'Browse' page ([Figure 1](#f0005){ref-type="fig"}E). The retrieval results provide more functions than that indicated in each tumor type, such as sorting and searching within results, making it easier to search the most frequent neoantigens generated by each gene. For instance, the potential neoantigen based on *BRAF* V600E mutation and HLA-A03:01 exists in 117 tumor samples, and corresponding neoantigens based on several mutations of *KRAS* and *PIK3CA* are also shared in more than 40 tumor samples ([Table 1](#t0005){ref-type="table"}).Table 1**Top 10 shared neoantigens of 7748 tumor samples from TCGAGeneMutationHLA alleleWT peptideWT affinity (nM)MT peptideMT affinity (nM)Frequency***BRAF*V600EA03:01KIGDFGLATVK94.09KIGDFGLATEK125.24117/7748*KRAS*G12DA02:01KLVVVGAGGV520.08KLVVVGADGV213.8282/7748*KRAS*G12VA02:01KLVVVGAGGV520.08KLVVVGAVGV111.8771/7748*KRAS*G12VA02:01KLVVVGAG17,690.28KLVVVGAV162.9771/7748*BRAF*V600EA11:01KIGDFGLATVK53.27KIGDFGLATEK45.2068/7748*PIK3CA*H1047RC07:01AHHGGWTTKM6742.50ARHGGWTTKM248.5762/7748*PIK3CA*H1047RC07:02AHHGGWTTKM2596.23ARHGGWTTKM217.7656/7748*PIK3CA*E545KA03:01STRDPLSEITE28,265.76STRDPLSEITK321.1954/7748*BRAF*V600EB57:01FGLATVKSRW128.34FGLATEKSRW246.2341/7748*BRAF*V600EB57:01LATVKSRW73.82LATEKSRW124.6141/7748[^8] There would always be some newly discovered tumor patients with combinations of somatic mutations and HLA alleles absent from 'Browse' page. We thus provide neoantigen prediction results of all possible combinations (Cartesian product) of recurrent mutations (occurring at least three times in all samples) and highly frequent HLA alleles in 'Search' page. The data used for neoantigen prediction include 18,897 recurrent missense mutations and the top 100 HLA alleles (frequency \>0.5%) of 7748 tumors from TCGA. Furthermore, we also employed the 13,459 recurrent missense mutations from ICGC and 16 HLA alleles with frequency \>5% in the population collected in the 1000 Genome Project [@b0115], for the prediction of potential shared neoantigens. Compared with the prediction results from real tumor samples, the frequencies of shared neoantigens predicted on recurrent mutations and highly frequent HLA alleles are similar ([Table 2](#t0010){ref-type="table"}). The distribution of predicted shared neoantigens is displayed as shown in [Figure 1](#f0005){ref-type="fig"}F and G.Table 2**Frequency of the top 10 shared neoantigens predicted by recurrent mutations in combination with highly frequent HLA alleles from TCGAGeneMutationHLA alleleExpected frequencyObserved frequency***BRAF*V600EA03:011.55%1.51%*KRAS*G12DA02:011.01%1.06%*PIK3CA*H1047RC07:010.73%0.80%*PIK3CA*E545KA03:010.68%0.70%*PIK3CA*E542KA03:010.44%0.44%*TP53*R248WA02:010.33%0.34%*TP53*R273CA02:010.29%0.31%*TP53*R248QC07:020.25%0.23%*TP53*Y220CA02:010.24%0.19%*PIK3CA*R88QC07:020.16%0.17%[^9] Besides, we present experimentally-validated data for the predicted neoantigens in the 'Validation' page ([Figure 1](#f0005){ref-type="fig"}H), according to the binding level between peptides and HLA alleles. Limited by the availability of binding data between mutant peptides and HLA alleles, all the validation data derived from IEDB is for wild-type peptides and HLA alleles [@b0080]. TSNAdb v1.0 (<http://biopharm.zju.edu.cn/tsnadb/>) is freely available for all academic users. Users can download data from the 'Download' page ([Figure 1](#f0005){ref-type="fig"}H), according to tumor types and the prediction tools chosen. Case study {#s0030} ---------- The major function of our database is to provide potential neoantigens of various tumor types and shared neoantigens across tumor patient populations. Therefore, we further provide statistical analyses of neoantigen prediction results in each tumor type. Here, we take the results of bladder cancer predicted by NetMHCpan v2.8 as an example to demonstrate the utilization of TSNAdb. There are 408 tumor samples for bladder cancer, with 106 different HLA alleles and 49,537 missense mutations. From these tumor samples, we obtain 182,756 predicted neoantigens. We present the top 20 HLA alleles, top 20 genes, and detailed neoantigen information in the web page ([Figure 2](#f0010){ref-type="fig"}A--C). According to the number of predicted neoantigens presented by each HLA allele, the top three HLA alleles are A02:01, A11:01, and C03:04, which account for 19.5%, 5.6%, and 4.7% of the total HLA alleles, respectively. These three HLA alleles also show \>5% frequency in the 1000 Genome Project [@b0115]. According to the number of predicted neoantigens generated by each gene, top three genes are *TTN*, *MUC16*, and *TP53*, which have 584, 318, and 270 neoantigens, respectively. In these genes, *TTN* and *MUC16* encode large proteins with numerous random mutations, whereas *TP53* is the most famous tumor suppressor gene with lots of recurrent mutations. The most recurrent mutation of *TP53* is R248Q, which exists in 17 out of 408 bladder cancer patients. The mutant peptide arising as a consequence of *TP53* R248Q mutation could bind to HLA-C07:02 and be presented as a potential neoantigen in four patients, which is the most frequent neoantigen in bladder cancer. If a bladder cancer patient carries the same mutation and HLA allele with existing patients, such as *TP53* R248W and HLA-A02:01, the corresponding neoantigen can be retrieved from the 'Browse' page directly. And the potential neoantigens can be used for experimental validation, which would facilitate the following cancer immunotherapy. If the combination of HLA allele and *TP53* mutation of this patient is absent in the existing samples, users can try to retrieve it in the 'Search' page. For instance, the combination of *TP53* R273H and HLA-A02:01 is absent in existing bladder cancer patients but can be retrieved in the 'Search' page, which provides the predicted neoantigens generated by all combinations of recurrent *TP53* mutations and highly frequent HLA alleles ([Figure 2](#f0010){ref-type="fig"}D). There are 155 types of recurrent *TP53* mutations that can generate at least one potential neoantigen presented by highly frequent HLA alleles, and 130 of these mutations could be presented by at least ten highly frequent HLA alleles. For instance, peptides generated by *TP53* G105C mutation are predicted to bind to 47 different HLA alleles. The most frequent potential neoantigen is generated by *TP53* R273H (0.86%) and HLA-A02:01 (41.4%), which shows the frequency of 0.36%.Figure 2**Example applications of predicted neoantigens for bladder cancer and the gene *TP53***Top 20 HLA alleles (**A**) and genes (**B**) with predicted neoantigens are displayed in the page using bladder cancer as an example, with the detailed neoantigen information listed (**C**). The binding level 'Strong' indicates strong binding with IC~50~ \< 150 nM, 'Weak' indicates weak binding with 150 nM \< IC~50~ \< 500 nM, '-' indicates non-binding with IC~50~ \> 500 nM. **D.** Distribution of the predicted neoantigens for all combinations of recurrent mutations of *TP53* and the highly frequent HLA alleles according to the TCGA dataset. The color gradient indicates the frequencies of potential shared neoantigens for the specific combinations of somatic mutations and HLA alleles. All the data shown are predicted by NetMHCpan v2.8. HLA, human leukocyte antigen; TCGA, The Cancer Genome Atlas; WT, wild type; MT, mutant. Perspectives and concluding remarks {#s0035} =================================== In this study, we developed a comprehensive database named TSNAdb for tumor-specific neoantigens based on 7748 tumor samples of 16 tumor types from TCGA. This database provides detailed affinity information between mutant/wild-type peptides and HLA alleles, and the frequencies of neoantigens shared by tumor samples of each tumor type and pan-cancer. Furthermore, this database also provides potential shared neoantigens generated from all possible combinations of recurrent mutations and highly frequent HLA alleles. The information provided by the database could facilitate the subsequent experimental design and validation and the discovery of potential targets for cancer immunotherapy. Compared with other existing neoantigen-related databases, such as TCIA, TSNADB provides the HLA binding information of both mutant peptides and wild-type peptides, which could be used for evaluating the differential agretopicity index (DAI), the difference of HLA binding affinity between mutant and wild-type peptides [@b0120]. Besides, users could search neoantigens at the gene level and obtain the potential shared neoantigens generated from all possible combinations of recurrent mutations and highly frequent HLA alleles from TSNAdb, which makes the database more user-friendly and comprehensive. In the future, we would expand our work from the following three aspects. In terms of the data, we would collect more samples from not only TCGA, but also other cancer databases such as ICGC or published literatures, for more comprehensive combination of HLA alleles and somatic mutations. In terms of the methods, we would apply more state-of-the-art methods on the neoantigen prediction and update our prediction software to improve the accuracy. In terms of the evaluation metrics, we would employ more well-acknowledged metrics to evaluate predicted neoantigens, *e.g.*, neoantigen quality indicating the probability for TCR recognition [@b0125], as well as DAI [@b0120]. Authors' contributions {#s0040} ====================== ZZ and SC conceived of the idea and supervised the study. ZS and XG participated in the design of the study. JW constructed and maintained the database and web interface, performed the data analysis. ZZ wrote the program and WZ designed the system architecture. WZ participated in the data analysis; ZS participated in the data acquisition; BZ participated in the statistical analysis. JW, BZ, and ZZ wrote the manuscript. All authors have read and approved the final manuscript. Competing interests {#s0045} =================== The authors have declared no competing interests. This work was supported by the National Key Research and Development Program of China (Grant No. 2017YFC0908600), the National Natural Science Foundation of China (Grant No. 31501021), and the Fundamental Research Funds for the Central Universities of China. The authors gratefully acknowledge the clinical contributors and data producers from the TCGA Research Network for referencing the TCGA datasets and the TCIA for referencing HLA-type data of TCGA samples. Peer review under responsibility of Beijing Institute of Genomics, Chinese Academy of Sciences and Genetics Society of China. [^1]: ORCID: 0000-0003-4554-9155. [^2]: ORCID: 0000-0003-1834-0348. [^3]: ORCID: 0000-0002-9141-8474. [^4]: ORCID: 0000-0002-4881-1238. [^5]: ORCID: 0000-0001-9845-6986. [^6]: ORCID: 0000-0002-2730-5483. [^7]: ORCID: 0000-0002-0792-3735. [^8]: *Note*: WT, wild type; MT, mutant. Amino acid residue changes caused by somatic mutations are indicated in red. [^9]: *Note*: Expected frequency indicates the frequency of shared neoantigens predicted by recurrent mutations in combination with highly frequent HLA alleles. Observed frequency, the frequency of shared neoantigens in 7748 tumor samples.

INTRODUCTION {#s1} ============ Pediatric glioblastoma (GBM) is a relatively rare primary brain tumor in children \[[@R1]\]. Maximum surgical resection is considered the key and prognosis-determining factor in the treatment, followed by radiotherapy \[[@R1], [@R2]\]. Unfortunately, pediatric GBM is poorly studied at the molecular and genomic levels. Radiotherapy plays a critical role in eradicating the post-surgical residual microtumor \[[@R3]\]. Yet the molecular and genomic changes post-radiation in pediatric GBM have not been well examined. We have recently identified many radiation-responsive genes in adult radioresistant GBM cells that explain the radioresistance and increased malignant features of recurrent GBM \[[@R4]\]. However, adult and pediatric GBMs are distinct from each other at both molecular and genetic levels \[[@R2]\]. We are presenting a full RNA sequencing profile of both the radiation-naïve pediatric GBM (SJ-GBM2) cell line and stable radioresistant pediatric GBM (SJ-GBM2-10gy) cell line that we recently developed \[[@R5], [@R6]\]. Our data demonstrated that radiation perturbed the expression of many genes related to many different known pathways in cancer biology. The irradiated cells exhibited an enhanced growth rate, overexpressed protease cathepsin B, and both subunits of the rate-limiting enzyme of DNA synthesis ribonucleotide reductase (RR). In this study, we shed light on the irradiation responsive mRNA changes that transform the tumor cells toward a more aggressive and resistant form, for which treatment choices are limited. This study opens the door to further examining the possibility of targeting these modified pathways as a therapeutic strategy to block GBM tumor recurrence and progression. RESULTS {#s2} ======= Radioresistant irradiated pediatric GBM cells exhibited higher growth rate than control cells {#s2_1} --------------------------------------------------------------------------------------------- The *in-vitro* growth rate of the SJ-GBM2 and SJ-GBM2-10gy cells was evaluated using an MTT growth assay over a period of 10 days. SJ-GBM2-10gy cells showed a superior divergent growth starting from day 3, having the difference in growth maximized on day 7 when the control cells significantly slowed down their proliferation rate (Figure [1](#F1){ref-type="fig"}). The control cells SJ-GBM2 reached about 7.4 growth fold in 10 days, while the irradiated cells reached a 10.5 fold from their baseline. This represents an estimated 30% increase in growth (Figure [1A](#F1){ref-type="fig"}). {#F1} Irradiated radioresistant pediatric GBM overexpresses ribonucleotide reductase {#s2_2} ------------------------------------------------------------------------------ To probe for a mechanism promoting the superior growth rate in irradiated cells, the expression of both ribonucleotide reductase (RR) subunits was measured. The RR enzyme, specifically the RRM2 subunit, has been reported to be essential for proliferation and invasion of GBM cells \[[@R7]\]. Immunoblotting of control and irradiated cells revealed an increase in the expression of RRM1 subunit by 2-fold, and an increase in the RRM2 subunit by 3.5-fold in irradiated cells relative to control cells (Figure [1B](#F1){ref-type="fig"}). This increase in cellular expression of RRM2 was confirmed by immunofluorescence probing of intact cells, demonstrating the distribution of RRM2 in the cytoplasm of the irradiated cells and was greater than the control cells (Figure [1B, 1C](#F1){ref-type="fig"}). Irradiated radioresistant pediatric GBM overexpresses pro-cathepsin B {#s2_3} --------------------------------------------------------------------- We were interested in evaluating whether protease may play of role in promoting invasion and progression of irradiated radioresistant GBM. Cathepsin B, a cysteine protease, has been shown to play a role in tumor growth and invasion \[[@R8], [@R9]\]. We probed for the differential expression level of pro-cathepsin B in control and irradiated cells. Western blot of cell lysates and immunofluorescence of intact cells revealed 3-fold overexpression of pro-cathepsin B in irradiated cells over control cells (Figure [1B, 1C](#F1){ref-type="fig"}). In addition to being localized to the cytoplasm, pro-cathepsin B was present in the processes of the irradiated cells, a feature that may be important in the invasiveness of GBM into the surrounding tissue (Figure [1C](#F1){ref-type="fig"}). Irradiation of pediatric GBM cells induces differential expression of 1192 radiation-responsive genes {#s2_4} ----------------------------------------------------------------------------------------------------- Total mRNAs from SJGBM2 and SJGBM2-10gy cells were harvested and subjected to further analysis. To screen for global mRNA changes following irradiation, we profiled transcriptomes of the control SJ-GBM2 cell line and its derivative radioresistant irradiated SJ-GBM2-10gy cells by RNA sequencing (Table [1](#T1){ref-type="table"} and [Supplementary Table 1](#SD2){ref-type="supplementary-material"}). The criteria for differentially expressed genes were 2-fold or greater than statistically significant values (*P* \< 0.05). We identified 1192 radiation responsive genes. Among these 1192 radiation responsive genes, 584 were upregulated and 608 were downregulated ([Supplementary Tables 2](#SD3){ref-type="supplementary-material"} and [3](#SD4){ref-type="supplementary-material"}). ###### Enriched gene ontology categories of differentially expressed genes following irradiation based on sets of statistically significant (more than 2-fold) upregulated and downregulated genes (*P* \< 0.05) Differentially expressed Category *P*-value -------------------------------------------------------- --------------------------------------- ----------- **Upregulated** GO:0048863: Stem cell differentiation 6.60E-03 GO:0010628: Positive regulation of gene expression 7.70E-03 GO:0002040: Sprouting angiogenesis 3.20E-02 GO:0008284: Positive regulation of cells proliferation 2.40E-02 GO:0070848: Response to growth factor 7.40E-02 GO:0001558: Regulation of cell growth 7.60E-02 GO:0055114: Oxidation reduction process 3.00E-02 GO:0071356: Cellular response to tumor necrosis factor 3.60E-02 GO:0006954: Inflammatory response 4.10E-02 GO:0016055: Wnt signaling pathway 8.40E-02 GO:0043066: Negative regulation of apoptotic process 9.40E-02 GO:0004222: Metalloendopeptidase activity 2.40E-02 **Downregulated** GO:0007155: Cell adhesion 4.90E-03 GO:0043065: Positive regulation of apoptosis 1.30E-04 GO:008285: Negative regulation of cell proliferation 4.10E-02 GO:0002020: Protease binding 1.20E-03 Experiments were performed in triplicate. Upregulation of genes promoting tumor growth and aggressiveness following irradiation {#s2_5} ------------------------------------------------------------------------------------- Gene ontology analysis of the mRNA data was utilized to categorize genes into functional groups. It revealed that upregulated genes were enriched in positive regulation of stem cells differentiation, angiogenesis, cell proliferation, cell growth, inflammatory response, positive regulation of the Wnt signaling pathway, response to hypoxia, metalloendopeptidase activity, cellular response to tumor necrosis factor and negative regulation of apoptotic process (Table [1](#T1){ref-type="table"} and [Supplementary Table 2](#SD3){ref-type="supplementary-material"}, and [Supplementary Table 4](#SD5){ref-type="supplementary-material"}). The upregulated genes were enriched in positive regulation of gene expression and tumor cell proliferation such as KIT, connective tissue growth factor CTGF and ID1, ID2, and TLE1-FOXG1 transcriptional factors that have been reported to enhance growth and proliferation of GBM cells \[[@R10]--[@R13]\]. G protein-coupled receptor kinase 5 (GRK5) plays an important role in tumor cells' proliferation \[[@R14]\]. Fibroblast growth factor 4 (FGF4) also has been correlated with a greater malignancy profile in high-grade gliomas \[[@R15]\].The radioresistant cells had upregulated expression of many anti-apoptotic genes, including BCL2, CD74, and WT1, which regulates GBM cells proliferation and apoptosis \[[@R16]\]. A significant upregulation (10-fold) of the AIM2 gene, a tumor-associated antigen, was found. This gene upregulation was observed in GBM patients and in malignant cell lines as well \[[@R17]\]. Proteolytic enzymes such as ADAM28, MMP12 and MMP17, which can enhance extracellular invasion and expansion of tumor volume, were also upregulated in the irradiated SJGBM2-10gy cells ([Supplementary Table 4](#SD5){ref-type="supplementary-material"}). Downregulated genes were enriched in the apoptotic process, tumor suppression, protease binding and cell adhesion {#s2_6} ----------------------------------------------------------------------------------------------------------------- The gene ontology analysis was conducted to analyze differentially expressed genes. Compared with control cells, downregulated genes in the irradiated SJ-GBM2-10gy cells were enriched in growth inhibitor, transcription DNA-templated, cell adhesion, apoptotic process, and protease binding (Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}, [Supplementary Table 3](#SD4){ref-type="supplementary-material"}). Tumor irradiation has been shown to cause silencing of many tumor suppressor genes, which in turn enhances tumor aggressive growth and invasion \[[@R18]\]. Many of the well-known tumor suppressor genes were found downregulated in the irradiated SJGBM2-10gy including DAB2IP, ING2, interleukin 1 beta, and MEG3 (Table [2](#T2){ref-type="table"}). Downregulation of DAB2IP induces radioresistance by accelerating DNA double strand repair after radiation and evasion of apoptotic process in prostate carcinoma \[[@R19]\]. Inhibitor of Growth Family member 2 (ING2) is a tumor suppressor that is involved in activation of p53/Tp53-dependent apoptosis \[[@R20]\]. Interleukin 1 beta is known to suppress GBM cells through promoting hypoxia-induced apoptosis by downregulation of HIF1 \[[@R21]\]. Maternally expressed gene 3 (MEG3) plays a role in activation of p53-dependent apoptosis and is found downregulated in the glioma cells compared with normal brain cells \[[@R22]\]. Interestingly, the irradiated cells showed significant repression of the main cellular metalloprotease inhibitor proteins called TIPM4 and alpha2 macroglobulin (A2M) \[[@R23]--[@R26]\]. Downregulation of protease inhibitors causes unopposed digestion of the extracellular matrix around the tumor by metalloproteases, which were also upregulated in irradiated cells, and therefore, facilitating tumor invasion of the extracellular space. ###### Downregulated genes of selected enriched gene ontology categories following irradiation of SJ-GBM2 cells are shown based on sets of statistically significant changes (*P* \< 0.05) Gene ontology group *P*-value Fold changes Gene symbol Gene description ------------------------------------------- ----------- --------------------------------------------------------------- ------------- -------------------------------------- Cell adhesion 4.90E-03 0.3390 CD33 CD33 molecule 0.4851 EDIL3 EGF like repeats and discoidin domains 3 0.3426 EPHA4 EPH receptor A4 0.4796 F11R F11 receptor 0.0456 L1CAM L1 cell adhesion molecule 0.3634 SPOCK1 SPARC/osteonectin, cwcv and kazal like domains proteoglycan 1 0.1247 TNFAIP6 TNF alpha induced protein 6 0.3202 ALCAM activated leukocyte cell adhesion molecule 0.4053 ADGRB1 adhesion G protein-coupled receptor B1 0.4680 BCAM basal cell adhesion molecule (Lutheran blood group) 0.1871 CDH3 cadherin 3 0.3202 CDH8 cadherin 8 0.4662 COL5A1 collagen type V alpha 1 chain 0.2467 COL8A1 collagen type VIII alpha 1 chain 0.0839 CNTN2 contactin 2 0.1860 CNTN4 contactin 4 0.4905 CNTNAP3 contactin associated protein-like 3 0.4111 EMP2 epithelial membrane protein 2 0.2666 FAP fibroblast activation protein alpha 0.3239 LGALS3BP galectin 3 binding protein 0.3272 ITGA11 integrin subunit alpha 11 0.3245 ITGA9 integrin subunit alpha 9 0.2359 LAMA3 laminin subunit alpha 3 0.2795 PTPRK protein tyrosine phosphatase, receptor type K 0.1867 ROBO2 roundabout guidance receptor 2 0.2710 TNC tenascin C Positive regulation of apoptotic process 1.30E-04 0.4109 AKAP13 A-kinase anchoring protein 13 0.3966 DAB2IP DAB2 interacting protein 0.3180 FGD4 FYVE, RhoGEF and PH domain containing 4 0.3340 KLF11 Kruppel like factor 11 0.2253 ARHGEF16 Rho guanine nucleotide exchange factor 16 0.1708 ARHGEF4 Rho guanine nucleotide exchange factor 4 0.3533 TNFRSF8 TNF receptor superfamily member 8 0.4738 ALDH1A2 aldehyde dehydrogenase 1 family member A2 0.1373 EEF1A2 eukaryotic translation elongation factor 1 alpha 2 0.4137 GAL galanin and GMAP prepropeptide 0.3390 ING2 inhibitor of growth family member 2 0.2705 IFIT2 interferon induced protein with tetratricopeptide repeats 2 0.4339 IRF5 interferon regulatory factor 5 0.4589 JMY junction mediating and regulatory protein, p53 cofactor 0.0767 NGFR nerve growth factor receptor 0.0816 PNMA2 paraneoplastic Ma antigen 2 0.3719 PAWR pro-apoptotic WT1 regulator 0.4911 STK3 serine/threonine kinase 3 Negative regulation of cell proliferation 4.10E-02 0.2946 CEBPA CCAAT/enhancer binding protein alpha 0.3390 CD33 CD33 molecule 0.3966 DAB2IP DAB2 interacting protein 0.4179 DLC1 DLC1 Rho GTPase activating protein 0.3340 KLF11 Kruppel like factor 11 0.2957 RERG RAS like estrogen regulated growth inhibitor 0.3533 TNFRSF8 TNF receptor superfamily member 8 0.4053 ADGRB1 adhesion G protein-coupled receptor B1 0.4738 ALDH1A2 aldehyde dehydrogenase 1 family member A2 0.4351 CHD5 chromodomain helicase DNA binding protein 5 0.1004 F2R coagulation factor II thrombin receptor 0.3390 ING2 inhibitor of growth family member 2 0.4605 IRF1 interferon regulatory factor 1 0.4571 IL1B interleukin 1 beta 0.3036 LDOC1 leucine zipper down-regulated in cancer 1 0.4785 LIF leukemia inhibitory factor 0.0953 MEG3 maternally expressed 3 (non-protein coding) 0.2795 PTPRK protein tyrosine phosphatase, receptor type K 0.4911 STK3 serine/threonine kinase 3 0.4718 SLIT3 slit guidance ligand 3 Protease binding 1.20E-03 0.0331 CD70 CD70 molecule 0.1539 TIMP4 TIMP metallopeptidase inhibitor 4 0.2020 A2M alpha-2-macroglobulin 0.2780 CSTA cystatin A 0.1772 CST6 cystatin E/M 0.2666 FAP fibroblast activation protein alpha 0.4143 IL1R1 interleukin 1 receptor type 1 0.1026 RYR1 ryanodine receptor 1 0.2443 SERPINA1 serpin family A member 1 0.3182 SERPINA5 serpin family A member 5 0.4862 SERPINE1 serpin family E member 1 Experiments were performed in triplicate. DISCUSSION {#s3} ========== Tumor radiation plays a critical role in pediatric brain tumors after gross total resection \[[@R1]--[@R3]\]. Unlike brain tumors in adults, the benefit of adjuvant chemotherapy is limited. As a result, it is typical that solely radiotherapy is utilized to eradicate any remnant tumor cells after surgery \[[@R1]--[@R3]\]. Ironically, irradiation of tumors is known to transform tumor cells into a more aggressive, radioresistant form, for which the treatment options are limited \[[@R27]\]. However, post-radiation transcriptome changes of the pediatric GBM have not been closely studied. The pediatric GBM cell line SJ-GBM2 is a common line that has been used for *in vitro* studies and is considered a radiation-naive cell line as it has not previously been irradiated \[[@R28], [@R29]\]. We previously described a stable radioresistant pediatric GBM model of irradiated SJ-GBM2 cells \[[@R5], [@R30]\]. In the current study, we further characterized these radioresistant cells and examined mRNA changes induced by irradiation. The results showed that the irradiated cells were more aggressive and possessed a higher proliferation rate when compared with their progenitors. SJ-GBM2 cells showed 7.4-fold growth on day 10 of incubation, while the irradiated SJ-GBM2-10gy cells grew up to 10.5-fold during the same period, suggesting that irradiation promotes a higher cell proliferation rate in radioresitant cells (Figure [1A](#F1){ref-type="fig"}). This rapid growth was paralleled by the increase in the expression of RR subunits, the main enzymes involved in DNA synthesis during cell division (Figure [1B, 1C](#F1){ref-type="fig"}). RR activity is critical for tumor cell growth \[[@R31]\]. The RRM2 subunit specifically has been linked to DNA repair capacity after radiation \[[@R32]\]. Our data suggested that RR overexpression in the irradiated cells may contribute to their ability to grow after radiation. In order to understand the changes in gene expression induced by irradiation, we performed a complete RNA sequencing of SJ-GBM2 and SJ-GBM2-10gy cells. Of the 32998 genes sequenced, a relatively small number (3.6%) of genes following irradiation were differentially expressed by meeting the criteria of a more than two-fold change ([Supplementary Table 1](#SD2){ref-type="supplementary-material"}). The upregulated genes such as *KIT, ID1, ID2, GRK5, CTGF, LEF1*, *NTRK3* and *PGF* are considered oncogenes that promote tumor proliferation, invasion and gene expression, in GBM or other types of cancers \[[@R11], [@R12], [@R14], [@R33]--[@R36]\]. *KIT* is an oncogene that enhances GBM proliferation and growth and it was found to be upregulated in patient-derived GBM samples \[[@R10]\]. *BMP1* is involved in many signaling pathways in GBM and correlate with poor prognosis in glioma patient \[[@R37]\]. Among the upregulated genes there was a significant increase in the expression of glutathione peroxidase 3 (*GPX3*), which is known to be an oxidative stress-induced antioxidant \[[@R38]\]. This increase of *GPX3* may protect cells against reactive oxygen species produced after radiation. Gene ontology showed that many anti-apoptotic and anti-inflammatory genes were upregulated in the irradiated cells (Table [1](#T1){ref-type="table"} and [Supplementary Table 4](#SD5){ref-type="supplementary-material"}). The downregulated genes were enriched in adhesion molecules (Table [2](#T2){ref-type="table"}). High-grade gliomas alter their expression of extracellular matrix adhesion proteins for tumor progression and invasion into the normal surrounding brain tissue \[[@R39]\]. Irradiation of cells depressed the expression of many adhesion molecules such as *CHD3, CHD8, FAP, BCAM, L1CAM, TNC, ITGA11*, and *ITGA9*. Eighteen pro-apoptotic genes, along with 21 tumor suppressor genes, were downregulated following irradiation (Table [2](#T2){ref-type="table"}). This explains the acquired radioresistant feature of irradiated cells and the increased malignant nature that was observed in irradiated cells (Figure [1A](#F1){ref-type="fig"}). Specifically, the *DAB2IP* gene was found to play a tumor suppressor role in medulloblastoma and a lower expression of *DAB2IP* caused resistance to radiation in prostate carcinoma \[[@R19], [@R40]\]. *P53* is believed to be the regulator of many radiation responsive genes that alter tumor cells sensitivity to radiation \[[@R41]\]. Any loss or mutation in the *P53* function leads to radioresistance \[[@R41]\]. Our results revealed that the downregulation of genes involved in radiation responsive activation of *P53*-apoptotic pathway such as *ING2, IL2B* and *MEG3*, enabled cells to acquire radioresistance. In addition, irradiation also altered metalloprotease activity. Although irradiated cells have lower expression of protease inhibitors *TIPM4* and A2M, the expression of metalloproteases such as *MMP12, MMP17* and cathepsin B was augmented. The net effect is the enhancement in metalloprotease activity, allowing the invasion and expansion of GBM after radiation. Alpha-2 macroglobulin (A2M), a large plasma glycoprotein natural inhibitor of cathepsin B and many other growth factors especially epidermal growth factors, exhibits the capability of abrogating the malignant potential of astrocytoma cells such as cell proliferation, invasion, and migration, and, as such, it can be exploited for therapeutic purposes \[[@R25], [@R26], [@R42]--[@R46]\]. Alpha-2 macroglublin has been proposed as a molecule conferring cancer resistance to the long-lived (30-year lifespan) naked mole rat, whose A2M transcript level is 140-fold higher than that of the control, and no tumor has ever been observed in these rats \[[@R47]--[@R49]\]. In this study, we report the alterations of gene expressions in response to radiation, which might play a critical role in the acquisition of radioresistance by irradiated pediatric GBM cells. The results indicated that irradiated cells were not only radioresistant, but they also transformed into a more aggressive, malignant GBM. This parallels the clinical observation that recurrent GBM is more aggressive and malignant after irradiation. This malignant transformation emphasizes the importance of developing a treatment regimen consisting of a multiple-agent cocktail that acts on multiple implicated pathways to effectively target irradiated pediatric GBM. An alternative to radiation or perhaps a new therapy- targeting differentially expressed genes, such as metalloproteases, growth factors, and oncogenes- that aims to minimize oncogenic changes following radiation is necessary to improve recurrent pediatric GBM survival. As a main cellular inhibitor of all tumor-associated metalloproteinases, and a sequester of many growth factors, alpha 2-macroglobulin might be exploited as a novel therapy to modulate pediatric GBM growth and invasion \[[@R42]--[@R46]\]. MATERIALS AND METHODS {#s4} ===================== Reagents and cells {#s4_1} ------------------ The pediatric glioblastoma cell line (SJ-GBM2) was obtained from the Children's Oncology Group (COG), (Dallas, TX, USA). Cell lines were cultured in Eagle's minimum essential medium containing 10% (v/v) fetal bovine serum, and supplemented with 1% sodium pyruvate and 0.1% gentamicin. Culture medium materials were obtained from Life Technologies, Inc. (Grand Island, NY, USA). Generation of the stable pedicatric radioresistant GBM model {#s4_2} ------------------------------------------------------------ We previously generated and described a stable radioresistant GBM model \[[@R5]\]. Briefly, to generate the SJ-GBM2-10gy, SJ-GBM2 cell lines were grown to confluence and then irradiated with a Pantak HF320 X-ray machine (Agfa NDT Ltd., Reading, UK) operating at 300 kV at a dosage of 2.09 Gy/min to a total radiation dose of 10 Gy. Over the course of weeks, most cells died and less than ∼1% of cells survived the irradiation. These radioresistant SJGBM2-10gy cells were allowed to grow to confluence and were perpetuated for experiments. MTT viability and Western blot assays {#s4_3} ------------------------------------- SJ-GBM2 and SJ-GBM2-10gy cell growths were measured by MTT assay. Cells were plated in an initial density of 0.05 × 10⁶ cell/ml in 96-well micro-well plates and incubated for 1 to 10 days. A daily readout of the cellular viability was recorded for growth rate measurement. Cell lysates from both irradiated and control cell lines were analyzed for the expression of ribonucleotide reductase subunits and cathepsin B by Western blot. Antibodies used were mouse anti RRM1 (Santa Cruz Biotechnology, sc-377415), mouse anti-RRM2 (Santa Cruz Biotechnology, sc-376973), mouse anti pro-cathepsin B (Life Technologies, Inc., 414800) and mouse anti b-actin (Santa Cruz Biotechnology, sc-47778). Secondary antibody used was HRP goat anti mouse (Li-Cor, 926-80010). Immunofluorescence {#s4_4} ------------------ SJ-GBM2 and SJ-GBM2-10gy cells were grown on chamber slides for 24 h. Cells were fixed with 3.7% paraformaldehyde in PBS, washed with ice-cold phosphate-buffered saline (PBS), permeabilized with 0.2% Triton 100 and incubated in 1% bovine serum albumin (BSA) in PBS blocking buffer. Next, cells then were incubated in 37° C with anti-pro-cathepsin B and anti-RRM2 for 1 h. Cells then were washed with 0.01% Tween PBS, and incubated in the dark with correspondent Texas horse anti mouse (Vector, TI2000) secondary antibodies for 1 h. Cells then were washed and counterstained with DAPI (Invitrogen, D1306) for 5 min. Confocal images were captured with a Carl Zeiss LSM510 microscope (Jena, Germany) equipped with a plan-apochromat 20×/0.8 NA or 40×/1.2NA lens available at the Imaging Core Facility at the Children's Research Institute, Medical College of Wisconsin (Milwaukee, WI, USA). Red fluorophore (A546) was excited with a diode laser (561 nm) and DAPI was excited with a 405 laser. Laser power to the sample was controlled with an acousto-optic tunable filter (AOTF) and the appropriate dichroics and filters for each fluorophore were used during image acquisition. The images were corrected for pixel saturation using with photomultiplier detector gain and amplifier offset controls as per the manufacturer's recommendations. Images were taken four to eight times and collected using Aim 4.2 software. RNA library preparation and sequencing {#s4_5} -------------------------------------- RNA sequencing was done as previously described \[[@R4]\]. In brief, RNA-sequencing libraries were prepared using the TruSeq Stranded mRNA Library Prep Kit (Illumina, Inc., San Diego, CA, USA) according to the manufacturer's protocol. PolyA mRNA from an input of 500 hg high quality total RNA (RINe \> 8) was purified and fragmented. First strand complementary deoxyribonucleic acid (cDNA) syntheses were performed using random hexameres and ProtoScript II Reverse Transcriptase (New England BioLabs Inc., Ipswich, MA, USA). The 3′ ends of the cDNA were adenylated and then indexing adaptors were ligated. Polymerase chain reactions were used to selectively enrich those DNA fragments that have adapter molecules on both ends and to amplify the amount of DNA in the library. The libraries were quantified using the Promega QuantiFluor dsDNA System on a Quantus Fluorometer (Promega, Madison, WI, USA). The size and purity of the libraries were analyzed using the High Sensitivity D1000 Screen Tape on an Agilent 2200 TapeStation instrument. The libraries were normalized, pooled, and subjected to cluster, and pair read sequencing was performed for 150 cycles on a HiSeq4000 instrument (Illumina, Inc., San Diego, CA, USA), according to the manufacturer's instructions. Gene ontology analysis {#s4_6} ---------------------- The gene ontology enrichment analysis was performed using DAVID Bioinformatics Resources 6.7, NIAIS/NIH (<http://david.abcc.ncifcrf.gov/>). SUPPLEMENTARY MATERIALS TABLES {#s5} ============================== We would like to thank Lydia Washechek of the Center for Imaging Research for editing the manuscript. **CONFLICTS OF INTEREST** None declared. **FUNDING** Musella Foundation Grant, Department of Neurosurgery Larson Endowment Grant. GBM : Giloblastoma PBS : phosphate-buffered SJ-GBM2 : radiation-naïve pediatric GBM RR : ribonucleotide reductase SJ-GBM2-10gy : stable radioresistant pediatric GBM

1. Introduction =============== Blastic plasmacytoid dendritic cell neoplasm (BPDCN) is a rare and aggressive hematologic malignancy derived from precursors of plasmacytoid dendritic cells. This disease entity was recognized in the 2008 World Health Organization (WHO) classification of tumors of hematopoietic and lymphoid tissues, where it was separately included in the group of acute myeloid leukemia (AML) and related precursor neoplasm.^\[[@R1]\]^ This disease almost always presents with cutaneous involvement as the 1st manifestation, with subsequent or concurrent spread to bone marrow and peripheral blood.^\[[@R2]--[@R4]\]^ Although it is extremely rare, a minority but significant proportion of patients present without skin lesions. Furthermore, BPDCN present at other sites has not been yet reported. To date, nasal cavity lesion as the 1st manifestation in BPDCN has not been reported yet. Here we report 2 cases of BPDCN preseting as masses of nasal cavity and nasopharynx with leukemic manifestation without skin lesion in adolescent patients. In addition, we briefly reviewed previous cases of BPDCN without skin manifestation. 2. Case reports =============== 2.1. Case 1 ----------- The 1st patient was a 16-year-old girl who presented with recurrent epistaxis. She had no significant medical history or family history of cancer or known genetic disorders. On sinonasal computed tomography (CT), a 2.9-cm sized, polypoid mass was noted in the nasal cavity. Cutaneous examination was unremarkable. Biopsy of this mass was performed. Histologically, the nasal mucosa diffusely expanded. It was infiltrated by atypical lymphoid infiltrates. Infiltrative tumor cells were diffuse, monomorphic medium-sized cells with fine chromatin, irregular nuclei, and scanty cytoplasm, showing blastic morphology. Mucosal glands often became widely spaced and lost. An angiocentric and angiodestructive growth pattern were not identified. Mucosal ulceration and necrosis were not identified either (Fig. [1](#F1){ref-type="fig"}A, B). Immunohistochemically, atypical lymphoid cells were positive for CD2, CD4, CD56, and CD123 with focal weak staining for TCL-1, but negative for CD20, CD3, TdT, MPO, and EBV-encoded small RNA (Fig. [1](#F1){ref-type="fig"}C--F). No clonal TCRG or IgH gene rearrangement was detected. Peripheral blood work-up revealed pancytopenia while bone marrow biopsy revealed involvement of neoplastic cells, similar to histology and immunohistochemical findings of nasal cavity mass. {#F1} The patient was treated with induction chemotherapy with Berlin-Frankfurt-Münster regimen used for acute lymphoblastic leukemia. She achieved complete remission. After the 1st remission, she received allogenic peripheral blood stem-cell transplant (PBSCT). No relapse was observed at 14 months after transplantation. Interestingly, she had no skin lesions at initial diagnosis or during the course of their illness. 2.2. Case 2 ----------- The 2nd patient was a previous healthy 17-year-old female who presented with nasal obstruction and voice change for a month. CT scans revealed a large enhancing nasopharyngeal mass involving adenoid and several small indeterminate lymph nodes at the neck. Biopsy of the nasopharyngeal mass was performed. Microscopically, the nasopharyngeal mucosa was entirely replaced by diffuse atypical lymphoid cells with blastoid morphology (Fig. [2](#F2){ref-type="fig"}A, B). Immunohistochemically, these atypical lymphoid cells were positive for CD4, weak CD56, CD123, TCL1, and TdT, but negative for CD20, CD3, CD8, and CD1a (Fig. [2](#F2){ref-type="fig"}C--F). Peripheral blood count results were as follows: WBC, 4890/μL; Hb, 11 g/dL; and platelet, 127/μL. Blast was measured 13% of WBCs. Bone marrow biopsy showed infiltration of blastic tumor cells and demonstrated increase of CD4+, CD56+, TdT+, CD10−, and CD34− blasts up to 95% of total nucleated cells. Abdominal scan revealed mild hepatosplenomegaly while PET scan suggested hypermetabolism at nasopharynx, systemic lymph nodes, breast, liver, spleen, and bone marrow. Based on these findings, the diagnosis was most consistent with BPDCN for the 2 cases. {#F2} The patient was treated with AraC/Idarubicin (AId) induction chemotherapy. However, persistent blasts (32.5% of total nucleated cells) were observed in bone marrow biopsy. She is now taking Cladribin/Ara-C/G-CSF (CLAG) reinduction chemotherapy. After the remission, she received allogenic PBSCT. No relapse was observed at 11 months after transplantation. Interestingly, she also had no skin lesions at initial diagnosis or during the course of their illness. 3. Discussion ============= In this study, we report 2 cases of BPDCN in adolescent patients who had unusual extracutaneous manifestation without skin lesion. Clinicopathologically, the differential diagnosis of our cases included extranodal NK/T-cell lymphoma, acute leukemia of ambiguous lineage, and NK lymphoblastic leukemia/lymphoma. Nasal cavity lesion as the 1st clinical manifestation and CD56-positive tumor cells raised the possibility of extranodal NK/T-cell lymphoma. However, absence of angioinvasion, no expression of cytoplasmic CD3, and cytotoxic granule proteins such as granzyme B and no association with EBV ruled out the diagnosis of extranodal NK/T-cell lymphoma. According to the 2017 WHO criteria, tumors that express some immunophenotypic features of BPDCN but not all immunohistochemical markers may be better classified as "acute leukemia of ambiguous lineage."^\[[@R5]\]^ At present, NK lymphoblastic leukemia/lymphoma is considered a provisional entity. It should be diagnosed after ruling out BPDCN. Blastic cells expressing CD56 and CD2 raised the possibility of NK lymphoblastic leukemia/lymphoma. However, CD4 positivity made it doubtful for such diagnosis. In such cases, immunohistochemical analysis including the most characteristic and reliable marker is essential for the diagnosis of BPDCN. BPDCN was initially characterized by the expression of CD4, CD56, and the lack of B cells, T cells, myeloid or monocytic cells, and NK cell markers. More specific plasmacytoid dendritic cell markers (CD123, CD303, and TCL1) have been recently used to diagnose BPDCN.^\[[@R6],[@R7]\]^ Since they are concomitantly expressed in only 46% of patients, it has been proposed that diagnosis of BPDCN can be made when 4 of these 5 markers (CD4, CD56, CD123, CD303, and TCL1) are expressed.^\[[@R8]\]^ Although tumor cells of the 1st case showed focal positive for TCL1, both of 2 cases showed all 5 markers except CD303 which was not performed in our institution. Therefore, our 2 cases were histologically diagnosed with BPDCN. The BPDCN without cutaneous lesion is exceedingly rare to diagnose. Patients without cutaneous involvement have been described in the literature. Table [1](#T1){ref-type="table"} presents a summary of 39 published cases of BPDN without skin involvement. Bone marrow involvement was observed in the majority of patients at diagnosis. Through hematopathology consultation service at the National Institutes of Health, Jegalian et al^\[[@R6]\]^ have evaluated 55 BPDCN cases. Among them, 9 (16%) patients lacked cutaneous disease at presentation. A retrospective multicenter study of 43 patients (the GIMEMA study) presenting with leukemic manifestation was reported in 2012.^\[[@R14]\]^ Among 43 patients, 8 (19%) cases had no cutaneous manifestations.^\[[@R14]\]^ In these patients lacking skin involvement, other extracutaneous and extramedullary sites in lymph node, spleen, and liver are most commonly observed. Rauh et al^\[[@R13]\]^ have demonstrated that patients with BPDN without skin involvement and leukemic presentation show adverse prognosis than those with skin involvement. Interestingly, no case of BPDCN presenting with nasal cavity mass has been reported. It is of note that we identified nasal cavity as the unusual site of BPDCN. ###### Summary of 39 published cases of BPDCN without skin involvement in the literatures.  Lack of traditional lineage-specific markers for B cells, T cells, myeloid, or monocytic cells with the absence of cutaneous manifestation has diagnostic challenge. The diagnosis of BPDCN is usually suspected in patients with skin lesion. Despite the absence of skin lesions and tumor involvement of unusual site, the diagnosis of BPDCN should not be ruled out since a minority of cases present with leukemia without skin involvement. Accurate recognition of BPDCN is important because of its different clinical course and outcome as well as treatment strategy compared to other differential diagnoses. In this regard, our 2 cases are significant as they have unusual presentation with leukemia in the absence of characteristic cutaneous manifestations. 4. Conclusion ============= In conclusion, our 2 cases demonstrate an atypical presentation without skin manifestation, characteristic of BPDCN. Although BPDCN without skin lesion is extremely rare, it should be considered in the differential diagnosis of blastic leukemia with an undifferentiated and ambiguous immunophenotype. Author contributions ==================== **Conceptualization:** Seung Eun Lee, Yoon Kyung Jeon, Wook Youn Kim. **Supervision:** Yoon Kyung Jeon, Wook Youn Kim. **Visualization:** Dohee Kwon, Ha Young Park. **Writing -- original draft:** Seung Eun Lee. **Writing -- review & editing:** Seung Eun Lee, Yoon Kyung Jeon, Wook Youn Kim. Abbreviations: AML = acute myeloid leukemia, BPDCN = blastic plasmacytoid dendritic cell neoplasm, CT = computed tomography, WHO = World Health Organization. Written informed consent was obtained from the patient for publication of this case report and its accompanying images. The authors have no funding and conflicts of interest to disclose.

1. Introduction =============== *Mycobacterium tuberculosis* is counted with justification among the most dangerous and successful microorganisms in today's world, especially in developing countries which have become the reservoir of resistant strains (the most burdened countries are in South Africa and East Asia) \[[@B1-molecules-19-00651]\]. These strains are causing the biggest number of problems connected to tuberculosis (TB) treatment \[[@B2-molecules-19-00651]\]. The main problem is resistance. It can be divided into three groups. The first one is called multidrug-resistant TB (MDR-TB) and these microbes are resistant to all first-line antituberculotic drugs (pyrazinamide - PZA, isoniazid - INH, rifampicin - RIF, ethambutol - ETH, streptomycin - STR). The second group is more treacherous. This type of resistance is called extensively or extremely drug-resistant TB (XDR-TB) with the resistance to the first-line anti-TB agents isoniazid and rifampicin together with the resistance to any fluoroquinolone used in the therapy and to at least one of three injectable second-line antituberculotic drugs (amikacin, kanamycin or capreomycin) \[[@B3-molecules-19-00651]\]. The last and the latest category was the most dreaded and was called totally drug-resistant TB (TDR-TB) and the first case was recorded in India \[[@B4-molecules-19-00651]\]. These mycobacterial strains were resistant to all current known therapy. Nowadays, this group has disappeared with the approval of bedaquiline for therapy of resistant forms of tuberculosis. Another problem is connected with the HIV pandemic. These two infectious diseases are influencing each other in a synergic way so this has resulted in efforts to develop new anti-tubercular agents. This work deals with a microwave-assisted synthesis of pyrazinamide analogues with potential antimycobacterial activity. It is caused by the fact that pyrazinamide is counted among the first-line anti-tuberculosis drugs used in current therapy. Its unique ability to kill the dormant forms of *Mycobacterium tuberculosis* is crucial in shortening the time needed for the treatment, so PZA has sterilizing activity especially in combination with rifampicin \[[@B5-molecules-19-00651]\]. PZA itself has multiple mechanisms of action. The first described was the activation of this prodrug *via* the enzyme pyrazinamidase (EC 3.5.1.19) to form pyrazinoic acid (POA). This metabolite causes a lowering of the inner compartment pH in mycobacterial cells. This leads to inhibition of membrane transport and then to cellular death \[[@B6-molecules-19-00651],[@B7-molecules-19-00651],[@B8-molecules-19-00651]\]. The gene encoding this enzyme is called *pncA* gene and its mutation is responsible for the origin of mycobacterial resistance to PZA \[[@B9-molecules-19-00651]\]. The second mechanism of action is connected with fatty acid synthase I (FAS I) (EC 2.3.1.85). It is suggested that the disruption of metabolism can be caused by inhibition of the cell membrane synthesis, which is essential for the survival of *Mycobacteria*, but this mechanism was mainly rejected for PZA itself by Boshoff because there is only low inhibition \[[@B10-molecules-19-00651]\]. On the other hand, the PZA analogues such as 5‑chloropyrazinamide, esters of pyrazinoic acid and esters of 5‑chloropyrazinoic acid were proven to act in this way \[[@B11-molecules-19-00651],[@B12-molecules-19-00651],[@B13-molecules-19-00651]\]. Recent research has suggested a novel mechanism of action of PZA - inhibition of *trans*-translation. This process is vital for survival and virulence of *Mycobacteria* and its inhibition leads to blockage of the proteosynthetic apparatus in ribosomes and to cellular death. These assumptions were proven by Zhang *et al.* \[[@B14-molecules-19-00651]\]. Although PZA is the first-line antituberculotic drug, it was found that this molecule has exhibited other interesting biological activities such as antifungal, antibacterial, antiviral and antineoplastic effects \[[@B15-molecules-19-00651],[@B16-molecules-19-00651],[@B17-molecules-19-00651],[@B18-molecules-19-00651],[@B19-molecules-19-00651],[@B20-molecules-19-00651]\]. There is another application of PZA derivatives that can be used in agriculture. The most successful pyrazine derivative diquat-dibromide (6,7-dihydrodipyrido\[1,2-a:2\',1\'-c\]pyrazinediium-dibromide), a non-selective, contact herbicide, which has been used to control many submerged and floating aquatic macrophytes, was found to interfere with the photosynthetic process by releasing strong oxidizers that rapidly disrupt and inactivate cells and cellular functions (at present banned in many EU countries) \[[@B21-molecules-19-00651]\]. Many structural variations of pyrazine compounds with herbicidal properties can be found in the patent literature \[[@B22-molecules-19-00651],[@B23-molecules-19-00651],[@B24-molecules-19-00651],[@B25-molecules-19-00651]\]. However, several pyrazine derivatives were also described as inhibitors of Hill reaction which inhibit photosynthetic electron transport (PET) in photosystem (PS) 2 \[[@B18-molecules-19-00651],[@B26-molecules-19-00651]\]. The site of action of these PET inhibitors in the photosynthetic apparatus was situated predominantly on the donor side of PS2, in the section between oxygen evolving complex and intermediate D^·^, *i.e.*, tyrosine radical (Tyr~D~•) occurring on the 161^st^ position in D~2~ protein. Consequently, these compounds can be considered as PS2 herbicides which could have ultimately adverse effect on plant growth. In general, the PET-inhibiting effectiveness of pyrazine derivatives depends on compound lipophilicity and σ Hammett constants of individual substituents. Hosseini *et al.* studied the electronic and structural descriptors, which are the main factors for the cytotoxicity in the series of substituted *N*-phenylpyrazine-2-carboxamides \[[@B27-molecules-19-00651]\]. This study is focused on preparation of *N*-substituted structural and functional derivative of PZA (5-chloro-6-methylpyrazine-2,3-dicarbonitrile) that was treated with ring-substituted benzylamines using the advantages of a microwave reactor. It should be stressed that this type of syntheses has become popular due to its higher yields, shorter reaction times or solvent savings in comparison with conventional organic syntheses \[[@B28-molecules-19-00651]\]. One of the main advantages is the heating. It is uniform through the volume of the sample and the microwaves usually interact with molecules themselves not vessel sides. Another benefit is connected with the temperature reached by the solvent used. The final temperature is usually far higher than the standard boiling point of the solvent when using over- pressurized systems. It is reached and bypassed in seconds. Improved heating usually leads to higher yields and shorter reaction times. There is one limitation for choosing the conditions. It is the polarity of the solvent when the non-polar solvents cannot be used in the way the polar ones can be. If the polar solvent is used in the reaction, there is a direct coupling of microwaves with molecules. More polar solvents have greater ability to interact with microwave radiation. Using the solvents with low polarity (low absorbers) leads to longer times of heating and reaction. On the contrary, if the reagents themselves are polar it could lower the disadvantages of non-polar solvents. Finally there are new approaches to microwave accelerated methods using ionic liquids or solid phase reactions (adsorption on mineral oxides, phase transfer catalysis, neat reactions) \[[@B29-molecules-19-00651]\]. Microwave assisted condensation in polar solvent is used in this work to accelerate the aminodehalogenation reaction. The conditions for the synthesis were proven experimentally. Antimycobacterial activity of the all prepared compounds was determined and compounds were evaluated also in relation to inhibition of photosynthetic electron transport (PET) in spinach (*Spinacia oleracea* L.) chloroplasts. The structure-activity relationships between the chemical structure and *in vitro* biological activities of evaluated compounds are discussed. 2. Results and Discussion ========================= 2.1. Chemistry -------------- The starting compound 5-chloro-6-methylpyrazine-2,3-dicarbonitrile and the final compounds **1**--**15** were synthesized according to the general procedure shown in [Scheme 1](#molecules-19-00651-f006){ref-type="scheme"}. The aminodehalogenation reaction of this starting compound and ring-substituted benzylamines yielded a series of 15 secondary amines of which 14 were novel. 5-(Benzylamino)-6-methylpyrazine-2,3-dicarbonitrile (**6**) was previously synthesised by Takematsu *et al.* and the reported melting point was 118--119 °C \[[@B30-molecules-19-00651]\]. The compound we obtained melted at 128.7--130.7 °C. This difference can be caused by the mode of crystallization. All reactions were done using microwave reactor with focused field and yields were in the range between 17% and 62%. Lower yields were caused by the purification using preparative chromatography and recrystallization. It is also known that 3-nitro substitution, for which the yield was the lowest, is counted among the electron-withdrawing groups reducing the basicity of the amine nitrogen. Obtained analytical data were fully consistent with the proposed structures. [Table 1](#molecules-19-00651-t001){ref-type="table"} shows the substituents and other data of the synthesized compounds. ![Synthesis of starting compound \[[@B30-molecules-19-00651]\] and microwave assisted aminodehalogenation reaction resulting in a series of compounds **1**--**15**.](molecules-19-00651-g006){#molecules-19-00651-f006} molecules-19-00651-t001_Table 1 ###### Experimentally determined values of lipophilicity log *k*, calculated values of log *P*, electronic Hammett's σ parameters and π parameters, 50% inhibition concentration IC~50~ \[μmol/L\] values related to PET inhibition in spinach chloroplasts in comparison with 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU) standard and *in vitro* antimycobacterial activity against *M. tuberculosis* H37Rv (minimal inhibition concentration (MIC) \[μg/mL\]) of compounds **1**--**15** compared to pyrazinamide (PZA) and isoniazid (INH) standards. Compound R log *P* log *k* σ π IC~50~ \[μmol/L\] MIC *M.tuberculosis* H37Rv \[μg/mL\] ---------- ------------ --------- --------- ------- --------- ------------------- -------------------------------------- 1 2-CH~3~ 3.41 0.4668 −0.17 0.1674 114.0 \>100 2 3-CF~3~ 3.84 0.5369 0.43 0.2375 37.7 12.5 3 3,4-Cl 4.04 0.7538 0.60 0.4544 16.4 6.25 4 4-CH~3~ 3.41 0.5157 −0.17 0.2163 104.7 25 5 4-OCH~3~ 2.8 0.2820 −0.27 −0.0174 464.6 \>100 6 H 2.92 0.2994 0 0 \- 25 7 4-NH~2~ 2.12 −0.2014 −0.15 −0.5008 \- 25 8 3-Cl 3.48 0.5172 0.37 0.2178 57.4 12.5 9 2-Cl 3.48 0.4663 0.22 0.1669 79.0 6.25 10 2-F 3.08 0.3042 0.06 0.0048 195.6 12.5 11 4-CF~3~ 3.84 0.5626 0.51 0.2632 39.6 6.25 12 2-CF~3~ 3.84 0.4865 0.51 0.1871 71.6 12.5 13 2,4-OCH~3~ 2.67 0.3699 −0.55 0.0705 \- 25 14 3-NO~2~ 2.71 0.1808 0.71 −0.1186 487.4 12.5 15 4-Cl 3.48 0.5384 0.23 0.2390 86.5 12.5 PZA \- 12.5 INH \- 1.5625 DCMU 1.9 \- π = log *k*~(substituted)~ -- log *k*~(unsubstituted).~ 2.2. Calculated and Experimentally Set Lipophilicity ---------------------------------------------------- Lipophilicity is one of the most important factors that can affect the biological effect of the compound. It is connected with the membrane transport and other biological processes and it is also connected with solubility in the media. This physical-chemical property can be set experimentally. In this work we have used the Reversed Phase - High Performance Liquid Chromatography (RP-HPLC) methodology for measuring the capacity factors *k* with the calculation of log *k* that can be correlated to calculated values of lipophilicity log *P* (resp. Clog *P*). These calculations were carried out by using PC program CS ChemBioDraw Ultra 13.0. The results of these measurements are shown in the [Table 1](#molecules-19-00651-t001){ref-type="table"}. The lowest lipophilicity was shown by compound **7** (R = 4-NH~2~) and on the contrary, compound **3** (R = 3,4-Cl) was the most lipophilic compound of this series. Lipophilicity, based on log *k* values, increased for substituents in the benzyl part of the molecule as follows: 4-NH~2~ \< 3-NO~2~ \< 4-OCH~3~ \< H \< 2-F \< 2,4-OCH~3~ \< 2-Cl \< 2-CH~3~ \< 2-CF~3~ \< 4-CH~3~ \< 3-Cl \< 3-CF~3~ \< 4-Cl \< 4-CF~3~ \< 3,4-Cl The dependence of the measured log *k* parameters on the calculated log *P* values showed an approximate linearity, which is shown in [Figure 1](#molecules-19-00651-f001){ref-type="fig"}, and the corresponding correlation can be expressed by the following regression equation: {#molecules-19-00651-f001} 2.3. Biological Assays ---------------------- ### 2.3.1. Antimycobacterial *In Vitro* Screening All prepared compounds were tested against four strains of *Mycobacterium*. These were *M. tuberculosis* and three non-tuberculosis strains. The most active substances against *M. tuberculosis* were compounds **3**, **9** and **11**. Their activity expressed as minimal inhibition concentration (MIC) was 6.25 µg/mL and the activities of standards were 12.5 µg/mL for PZA and 1.5625 µg/mL for INH. Nearly the whole series showed activity against *M. tuberculosis* H37Rv, which was in the range from 25 to 6.25 µg/mL, but compounds **11** and **12** were also active against other strains and both of them were active against *M. avium* 152. These strains are usually resistant or unsusceptible to pyrazinamide. This is the reason why INH was chosen as second standard. Obtained results can be compared with other synthesised compounds and a comparison can be drawn between more mycobacterial strains not only *M. tuberculosis*. As shown in [Figure 2](#molecules-19-00651-f002){ref-type="fig"}, dependence of antimycobacterial activity on lipophilicity expressed by π constant (log *k*) as well as on the σ constant of the R substituent was observed. The most active compounds (MIC = 6.25 µg/mL) were **3** (R = 3,4-Cl), **9** (R = 2-Cl) and **11** (R = 4-CF~3~) and were mentioned above. All of these compounds are situated in the right upper quadrant of the Craig's plot. Further groups are represented by compounds **2** (R = 3-CF~3~), **8** (R = 3-Cl), **10** (R = 2-F), **12** (R = 2-CF~3~), **14** (R = 3-NO~2~) and **15** (R = 4-Cl) showing moderate activity with MIC = 12.5 µg/mL and compounds **4** (R = 4-CH~3~), **6** (R = H), **7** (R = 4-NH~2~) and **13** (R = 2,4-OCH~3~) with MIC = 25 µg/mL. Low antimycobacterial activity with MIC \>100 µg/mL was exhibited by compounds **1** (R = 2-CH~3~) and **5** (R = 4‑OCH~3~). It can be stated that the activity is more dependent on π constant than on σ constant. On the other hand this dependence is not unambiguous. However, it is clear that the lipophilicity is important for antimycobacterial activity and the most suitable ring substituents are electron-withdrawing groups such as halogen or trifluoromethyl moieties. {#molecules-19-00651-f002} ### 2.3.2. Antimycobacterial *In Vitro* Screening Against *M. smegmatis* This evaluation was performed against fast-growing *Mycobacterium smegmatis* using isoniazid and ciprofloxacin as standards. None of the compounds showed activity against this mycobacterial strain. This may be caused by the fact that this fast-growing mycobacterium is less susceptible to antibiotic treatment. ### 2.3.3. Antifungal and Antibacterial *In Vitro* Screening This evaluation was performed in order to obtain results for antifungal and antibacterial activity against eight fungal strains and eight bacterial strains. None of the prepared compounds showed antibacterial activity against the tested strains. On the contrary, compounds **7** (MIC = 500 µmol/L), **10** (MIC = 500 µmol/L) and **12** (MIC = 125 µmol/L) exhibited activity against *Trichophyton mentagrophytes* but in comparison to standards, this activity was negligible. It can be also caused by the bigger susceptibility of this fungal strain being evaluated. ### 2.3.4. Herbicidal Activity of Prepared Compounds The studied compounds inhibited photosynthetic electron transport in spinach chloroplasts ([Table 1](#molecules-19-00651-t001){ref-type="table"}). The PET inhibiting activity of studied compounds **1**--**15** expressed by IC~50~ value varied from 16.4 μmol/L (**3**) to 487.0 μmol/L (**14**). The inhibitory activity of the most active compounds **3** was 8.6 times lower than that of the standard DCMU (IC~50~ = 1.9 μmol/L), a well-known PS2 herbicide. Compounds **6** (R = H), **7** (R = 4-NH~2~) and **13** (R = 2,4-OCH~3~) were inactive and IC~50~ values related to PET inhibition could not be determined. It can be stated that the inhibitory activity increases linearly with increasing lipophilicity of the compounds based on the dependence of PET inhibiting activity of studied dicarbonitriles **1**--**15** on log *P* ([Figure 3](#molecules-19-00651-f003){ref-type="fig"}A) and log *k* ([Figure 3](#molecules-19-00651-f003){ref-type="fig"}B). {#molecules-19-00651-f003} Also if compound **14** (R = 3-NO~2~) was excluded, the dependence of PET inhibiting activity on the Hammett constant σ of R substituent showed again linear dependence ([Figure 4](#molecules-19-00651-f004){ref-type="fig"}). A linear increase of PET-inhibiting activity was observed in the range of σ from -0.15 (**7**; R = 4-NH~2~) to 0.6 (**3**; R = 3,4-Cl). {#molecules-19-00651-f004} The correlation between log (1/IC~50~) \[mol/L\] and log *P* or between log (1/IC~50~) \[mol/L\] and σ constant of R substituent could be expressed by the following equations: The use of both the above mentioned descriptors (log *P* and σ) in a multilinear correlation did not improve the results of statistical analysis: Similarly, better results of statistical analysis were obtained for the correlation between log (1/IC~50~) \[mol/L\] and log *k* than for multilinear correlation using two descriptors (log *k* and σ). pI 50 = 2.703 (±0.414) + 2.835 (±0.816) log k r = 0.934, s = 0.144, F = 61.8, n = 11 pI 50 = 2.997 × (± 0.331) + 2.034 × (± 0.738) log k \+ 0.481 (0.303) σ r = 0.976 s = 0.093 F = 80.2 n = 11 These results indicate that lipophilicity of the compound is determining for PET-inhibiting activity of the studied *N*-substituted 5-amino-6-methylpyrazine-2,3-dicarbonitriles. Lee *et al.* focused on the search for the minimum structural requirements for herbicidal evaluation of 5-(R^1^)-6-(R^2^)-*N*-(R^3^-phenyl)-pyrazine-2-carboxamide analogues as a new class of potent herbicides in a previous paper \[[@B31-molecules-19-00651]\]. Based on IC~50~ values of 19 pyrazine derivatives related to PET inhibition in spinach chloroplasts which were published by Dolezal *et al.* \[[@B15-molecules-19-00651]\], Lee derived and discussed quantitatively 3D-QSARs models between the substituents (R^1^--R^3^) changes of the analogues and their herbicidal activity using comparative molecular field analysis (CoMFA) and comparative molecular similarity indice analysis (CoMSIA) methods. It was predicted that the herbicidal activity increases when large steric substituents were introduced to one part of the *ortho*- and *meta*- positions on the *N*-phenyl ring as R^3^- substituent and small steric substituents on the other part. The same 19 pyrazine derivatives with herbicidal activity were also subjected to the two dimensional quantitative structure activity relationships studies using Vlife Molecular Design 3.0 module which contains various combinations of thermodynamic, electronic, topological and spatial descriptors \[[@B32-molecules-19-00651]\]. It was found that decreasing of number of hydrogen bond acceptor atoms and reduction the any atoms (single, double or triple bonded) separated from any oxygen atom by a seven bond distance in a molecule could be helpful for designing more potent herbicidal agents. Because the studied *N*-substituted 5-amino-6-methylpyrazine-2,3-dicarbonitriles were found to inhibit the Hill reaction, they can be considered as photosystem 2 (PS2) inhibitors, *i.e.*, PS2 herbicides which ultimately adversely affect growth of weeds as well as agricultural plants. The PS2 inhibitors can act on the donor and/or the acceptor side of PS2. Using EPR spectroscopy it was found that *N*-phenylpyrazine-2-carboxamides interacted with the D^·^ intermediate which is situated at 161^st^ position in D~2~ protein occurring on the donor side of PS2. Due to interaction of these pyrazine derivatives with this part of PS2, the photosynthetic electron transport from the oxygen evolving complex to the reaction centre of PS2 was impaired and consequently, the electron transport between PS2 and PS1 was inhibited \[[@B33-molecules-19-00651]\]. Use of an artificial electron-donor, 1,5-diphenylcarbazide (DPC), acting in the Z^·^/D^·^ intermediate is suitable to estimate whether the studied PET inhibitor acts only on the donor or also on the acceptor side of PS2. Complete restoration of PET after DPC addition to chloroplasts activity of which was inhibited by an inhibitor, indicates that PET between the core of PS2 (P680) and the secondary quinone acceptor Q~B~ was not affected by this inhibitor. Consequently, its site of inhibitory action is situated on the donor side of PS2. However, upon addition of DPC to the studied dicarbonitriles only partial restoration of PET was observed (up to 85% of the control) and therefore it could be assumed that the studied compounds block PET not only by interaction with proteins occurring in the section between the oxygen evolving complex (OEC) and the Z^·^/D^·^ intermediate but to PET inhibition contributes also their interaction with some constituents of photosynthetic apparatus on the acceptor side of PS2. Similar results were obtained previously with 5-*tert*-butyl-*N*-(3-hydroxy-4-chlorophenyl)pyrazine-2-carboxamide and 5-*tert*-butyl-6-chloro-*N-*(3-fluorophenyl)pyrazine-2-carboxamide \[[@B33-molecules-19-00651]\]. Compounds **1**--**15** affected the chlorophyll *a* (Chl*a*) fluorescence in spinach chloroplasts. As shown in [Figure 5](#molecules-19-00651-f005){ref-type="fig"}, the intensity of the Chl*a* emission band at 686 nm belonging to the pigment--protein complexes in photosystem 2 decreased in the presence of compound **3** \[[@B34-molecules-19-00651]\]. This finding indicates a perturbation of the Chl*a*--protein complexes in the thylakoid membrane caused by the tested compound. Similar Chl*a* fluorescence decrease in spinach chloroplasts was observed previously with several PET inhibitors, *i.e.*, *N-*benzylpyrazine-2-carboxamides \[[@B35-molecules-19-00651]\], ring-substituted 3-hydroxynaphthalene-2-carboxanilides \[[@B36-molecules-19-00651]\] and 1-hydroxynaphthalene-2-carboxanilides \[[@B37-molecules-19-00651]\], 2-hydroxynaphthalene-1-carboxanilides \[[@B38-molecules-19-00651]\] and ring-substituted 4-arylamino-7-chloroquinolinium chlorides \[[@B39-molecules-19-00651]\]. {#molecules-19-00651-f005} 3. Experimental =============== 3.1. General ------------ All the chemicals used for preparation of starting compound and final products were purchased from Sigma-Aldrich (Sigma-Aldrich, St. Louis, MO, USA) and were reagent or higher grade of purity. Starting compound was prepared according to proven methodology of conventional organic synthesis. The final aminodehalogenation reactions were performed in CEM Discover microwave reactor with focused field (CEM Corporation, Matthews, NC, USA) connected to the Explorer 24 autosampler (CEM Corporation) and this equipment was running under CEM's Synergy™ software for monitoring the progress of reactions. The reaction progress was checked by Thin Layer Chromatography (TLC) (Alugram^®^ Sil G/UV254, Machery-Nagel, Postfach, Germany) using 254 nm wavelength UV detection. All the obtained products were purified by crystallization or by preparative flash chromatography (CombiFlash^®^ Rf, Teledyne Isco Inc., Lincoln, NE, USA), using gradient elution with hexane (LacheNer, Neratovice, Czech Republic) and ethyl acetate (Penta, Prague, Czech Republic) as mobile phases. Silica gel (0.040--0.063 nm, Merck, Darmstadt, Germany) was used as the stationary phase. NMR spectra were recorded on Varian Mercury--VxBB 300 (299.95 MHz for ^1^H and 75.43 MHz for ^13^C) or Varian VNMR S500 (499.87 MHz for ^1^H and 125.71 MHz for ^13^C) spectrometers (Varian Corporation, Palo Alto, CA, USA). Chemical shifts were reported in ppm (δ) and were applied indirectly to tetramethylsilane as a signal of solvent (2.49 for ^1^H and 39.7 for ^13^C in DMSO-*d*~6~). Infrared spectra were recorded with spectrometer FT-IR Nicolet 6700 (Thermo Scientific, Waltham, MA, USA) using attenuated total reflectance (ATR) methodology. Melting points were assessed by SMP3 Stuart Scientific (Bibby Scientific Ltd., Staffordshire, UK) and were uncorrected. Elemental analyses were measured with EA 1110 CHNS Analyzer (Fisons Instruments S. p. A., Carlo Erba, Milano, Italy). Calculation of electronic Hammett's σ parameters was carried out on the software ACD/Percepta ver. 2012 (Advanced Chemistry Development, Inc., Toronto, ON, Canada). 3.2. Starting Compound and Final Products Synthesis --------------------------------------------------- The starting compound 5-chloro-6-methylpyrazine-2,3-dicarbonitrile was synthesized in a two-step reaction according to the reported methodology \[[@B30-molecules-19-00651]\]. The first step was a condensation reaction between diaminomaleonitrile (0.025 mol) and pyruvic acid (0.025 mol), which were dissolved in methanol (60 mL), and hydrochloric acid (10 mL, 15%) was added dropwise. It takes two hours to react at room temperature. After the evaporation of two thirds of the solvent, hot water (80 mL) was added and then the rest of methanol was evaporated *in vacuo*. The whole mixture was cooled to 5 °C to initiate the crystallization. The product was collected by suction, dried overnight and subsequently chlorinated with phosphoryl chloride. Product (0.015 mol) was again dissolved in POCl~3~ (0.060 mol) and cooled to 0 °C. Pyridine (0.020 mol) was added dropwise and after the termination of the exothermic reaction, the mixture was heated to 90 °C for 2 h. Excess POCl~3~ was evaporated *in vacuo* and the rest of product was extracted into toluene three or four times, the toluene was evaporated and then the crude product was recrystallized from chloroform. The starting compound (1.12 mmol) was finally treated with 15 variously ring-substituted benzylamines (2.24 mmol) and all these aminodehalogenation reactions took place in the microwave reactor. The conditions used for microwave syntheses were as follows---150 °C, 30 min, 120 W, methanol as a solvent, pyridine as a base and were set experimentally. The reaction was monitored by TLC using hexane/ethyl acetate 2:1 mixture as mobile phase. The final compounds were purified using flash column chromatography with gradient elution using a hexane/ethyl acetate system and if necessary recrystallization from a mixture of ethanol and water. 3.3. Analytical Data of the Prepared Compounds ---------------------------------------------- *5-Methyl-6-\[(2-methylbenzyl)amino\]pyrazine-2,3-dicarbonitrile* (**1**). Light orange crystalline solid. Yield 47.9%; M.p. 152.1--152.9 °C; IR (ATR-Ge, cm^−1^): 3381~m~ (-NH-), 2226~m~ (-CN), 1570~vs~, 1521~s~, 1403~vs~, 1349~m~ (arom.); ^1^H-NMR (500 MHz) δ 8.61 (1H, bs, NH), 7.25--7.10 (4H, m, H3\', H4\', H5\', H6\'), 4.56 (2H, s, NCH~2~), 2.47 (3H, s, CH~3~), 2.33 (3H, s, CH~3~); ^13^C-NMR (125 MHz) δ 152.9, 147.5, 136.0, 135.6, 130.2, 130.1, 127.5, 127.2, 125.9, 117.4, 115.7, 114.8, 42.5, 21.3, 19.0; Elemental analysis: calc. for C~15~H~13~N~5~ (MW 263.12): 68.42% C, 4.98% H, 26.60% N; found 68.22% C, 5.17% H, 26.44% N. *5-Methyl-6-{\[3-(trifluoromethyl)benzyl\]amino}pyrazine-2,3-dicarbonitrile* (**2**). Yellow-orange crystalline solid. Yield 51.6%; M.p. 158.5--159.7 °C (decomp.); IR (ATR-Ge, cm^−1^): 3377~m~ (-NH-), 2227~m~ (-CN), 1570~vs~, 1521~s~, 1402~vs~, 1353~m~, 1324~vs~ (arom.), 1128~vs~ (-C-F); ^1^H-NMR (300 MHz) δ 8.75 (1H, bs, NH), 7.77--7.60 (1H, m, H2\'), 7.68--7.50 (3H, m, H4\', H5\', H6\'), 4.69 (2H, d, *J* = 4.6 Hz, NCH~2~), 2.46 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 152.9, 147.7, 139.6, 131.8, 130.0, 129.6, 129.2 (q, *J* = 31.5 Hz), 124.5 (q, *J* = 4.0 Hz), 124.4 (q, *J* = 272.3 Hz), 124.0 (q, *J* = 4.0 Hz), 117.7, 115.6, 114.8, 44.0, 21.2; Elemental analysis: calc. for C~15~H~10~F~3~N~5~ (MW 317.27): 56.78% C, 3.18% H, 22.07% N; found 56.59% C, 3.30% H, 21.93% N. *5-\[(3,4-Dichlorobenzyl)amino\]-6-methylpyrazine-2,3-dicarbonitrile* (**3**). Light yellow crystalline solid. Yield 60.9%; M.p. 154.4--156.2 °C; IR (ATR-Ge, cm^−1^): 3329~m~ (-NH-), 2227~m~ (-CN), 1568~vs~, 1516~s~, 1466~m~, 1402~vs~ (arom.); ^1^H-NMR (300 MHz) δ 8.69 (1H, bs, NH), 7.61 (1H, d, *J* = 1.9 Hz, H2\'), 7.57 (1H, d, *J* = 8.2 Hz, H5\'), 7.33 (1H, dd, *J* = 8.2 Hz, *J* = 1.9 Hz, H6\'), 4.59 (2H, s, NCH~2~), 2.46 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 152.8, 147.8, 139.4, 131.1, 130.6, 130.0, 129.8, 129.6, 127.9, 117.8, 115.6, 114.8, 43.3, 21.2; Elemental analysis: calc. for C~14~H~9~Cl~2~N~5~ (MW 318.16): 52.85% C, 2.85% H, 22.01% N; found 52.71% C, 3.05% H, 21.88% N. *5-Methyl-6-\[(4-methylbenzyl)amino\]pyrazine-2,3-dicarbonitrile* (**4**). Light yellow crystalline solid. Yield 60.5%; M.p. 140.8--142.3 °C; IR (ATR-Ge, cm^−1^): 3370~s~ (-NH-), 2223~m~ (-CN), 1576~vs~, 1519~s~, 1438~m~, 1400~vs~, 1351~s~ (arom.); ^1^H-NMR (300 MHz) δ 8.68 (1H, bs, NH), 7.24--7.18 (2H, m, AA\', BB\', H2\', H6\'), 7.14--7.09 (2H, m, AA\', BB\', H3\', H5\'), 4.55 (2H, bs, NCH~2~), 2.44 (3H, s, CH~3~), 2.26 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 152.8, 147.5, 136.4, 135.0, 130.1, 128.1, 127.5, 117.3, 115.7, 114.9, 44.1, 21.2, 20.8; Elemental analysis: calc. for C~15~H~13~N~5~ (MW 263.30): 68.42% C, 4.98% H, 26.60% N; found 68.30% C, 5.17% H, 26.43% N. *5-\[(4-Methoxybenzyl)amino\]-6-methylpyrazine-2,3-dicarbonitrile* (**5**). Ochre crystalline solid. Yield 61.3%; M.p. 153.0--154.2 °C (decomp.); IR (ATR-Ge, cm^−1^): 3345~m~ (-NH-), 2948~w~ (-OCH~3~), 2223~m~ (-CN), 1585~s~, 1570~vs~, 1511~vs~, 1462~m~, 1403~vs~ (arom.); ^1^H-NMR (300 MHz) δ 8.66 (1H, bs, NH), 7.29--7.23 (2H, m, AA\', BB\', H2\', H6\'), 6.90--6.84 (2H, m, AA\', BB\', H3\', H5\'), 4.52 (2H, bs, NCH~2~), 3.71 (3H, s, OCH~3~), 2.43 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 158.6, 152.8, 147.4, 130.1, 129.9, 129.0, 117.3, 115.7, 114.9, 113.9, 55.2, 43.8, 21.2; Elemental analysis: calc. for C~15~H~13~N~5~O (MW 279.30): 64.51% C, 4.69% H, 25.07% N; found 64.35% C, 4.81% H, 24.84% N. *5-(Benzylamino)-6-methylpyrazine-2,3-dicarbonitrile* (**6**) \[[@B30-molecules-19-00651]\]. Yellow-orange crystalline solid. Yield 53.3%; M.p. 128.7--130.7 °C (described in the literature 118--119 °C); IR (ATR-Ge, cm^−1^): 3404~m~ (‑NH-), 2228~m~ (-CN), 1564~vs~, 1508~vs~, 1454~m~, 1400~vs~, 1357s (arom.); ^1^H-NMR (300 MHz) δ 8.72 (1H, bs, NH), 7.37--7.20 (5H, m, H2\', H3\', H4\', H5\', H6\'), 4.61 (2H, d, *J* = 3.3 Hz, NCH~2~), 2.45 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 152.9, 147.5, 138.1, 130.1, 128.5, 127.5, 127.2, 117.4, 115.7, 114.9, 44.3, 21.2; Elemental analysis: calc. for C~14~H~11~N~5~ (MW 249.27): 67.46% C, 4.45% H, 28.10% N; found 67.21% C, 4.63% H, 27.95% N. *5-\[(4-Aminobenzyl)amino\]-6-methylpyrazine-2,3-dicarbonitrile* (**7**). Brown crystalline solid. Yield 37.2%; M.p. 154.9--156.5 °C; IR (ATR-Ge, cm^−1^): 3413~m~ (-NH~2~), 3376~m~ (-NH-), 2234~m~ (-CN), 1572~vs~, 1519~vs~, 1440~m~, 1402~vs~, 1352~s~ (arom.); ^1^H-NMR (300 MHz) δ 8.54 (1H, bs, NH), 7.04--6.94 (2H, m, AA\', BB\', H2\', H6\'), 6.54--6.45 (2H, m, AA\', BB\', H3\', H5\'), 4.97 (2H, bs, NH~2~), 4.41 (2H, d, *J* = 3.8 Hz, NCH~2~), 2.41 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 152.7, 148.0, 147.3, 130.2, 128.7, 124.7, 117.0, 115.8, 114.9, 113.9, 44.2, 21.2; Elemental analysis: calc. for C~14~H~12~N~6~ (MW 264.29): 63.62% C, 4.58% H, 31.80% N; found 63.48% C, 4.45% H, 31.65% N. *5-\[(3-Chlorobenzyl)amino\]-6-methylpyrazine-2,3-dicarbonitrile* (**8**). Yellow crystalline solid. Yield 60.2%; M.p. 151.0--151.8 °C; IR (ATR-Ge, cm^−1^): 3306~m~ (-NH-), 2248~m~ (-CN), 1567~vs~, 1513~vs~, 1468~m~, 1401~vs~, 1351~s~ (arom.); ^1^H-NMR (300 MHz) δ 8.70 (1H, t, *J* = 5.9 Hz, NH), 7.43--7.39 (1H, m, H2\'), 7.38--7.27 (3H, m, H4\', H5\', H6\'), 4.61 (2H, d, *J* = 5.9 Hz, NCH~2~), 2.46 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 152.9, 147.7, 140.7, 133.2, 130.4, 130.0, 127.4, 127.2, 126.2, 117.7, 115.6, 114.9, 43.8, 21.2; Elemental analysis: calc. for C~14~H~10~ClN~5~ (MW 283.72): 59.27% C, 3.55% H, 24.68% N; found 59.08% C, 3.71% H, 24.57% N. *5-\[(2-Chlorobenzyl)amino\]-6-methylpyrazine-2,3-dicarbonitrile* (**9**). Dark orange crystalline solid. Yield 24.7%; M.p. decomp.; IR (ATR-Ge, cm^−1^): 3381~m~ (-NH-), 2227~m~ (-CN), 1570~vs~, 1521~s~, 1474~m~, 1440~m~, 1401~s~, 1351~s~ (arom.); ^1^H-NMR (300 MHz) δ 8.68 (1H, bs, NH), 7.51--7.24 (4H, m, H3\', H4\', H5\', H6\'), 4.66 (2H, s, NCH~2~), 2.49 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 153.0, 147.7, 134.9, 132.3, 130.0, 129.4, 129.1, 128.9, 127.4, 117.8, 115.6, 114.8, 42.4, 21.2; Elemental analysis: calc. for C~14~H~10~ClN~5~ (MW 283.72): 59.27% C, 3.55% H, 24.68% N; found 59.08% C, 3.68% H, 24.90% N. *5-\[(2-Fluorobenzyl)amino\]-6-methylpyrazine-2,3-dicarbonitrile* (**10**). Dark yellow crystalline solid. Yield 60.8%; M.p. 134.8--136.3 °C; IR (ATR-Ge, cm^−1^): 3377~m~ (-NH-), 2231~m~ (-CN), 1573~vs~, 1523~s~, 1491~m~, 1402~s~, 1354~s~ (arom.); ^1^H-NMR (300 MHz) δ 8.68 (1H, t, *J* = 5.3 Hz, NH), 7.42--7.10 (4H, m, H3\', H4\', H5\', H6\'), 4.63 (2H, d, *J* = 5.3 Hz, NCH~2~), 2.46 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 160.0 (d, *J* = 244.8 Hz), 152.9, 147.6, 129.8 (d, *J* = 20.4 Hz), 129.8, 129.4 (d, *J* = 8.3 Hz), 124.6 (d, *J* = 17.8 Hz), 124.5 (d, *J* = 6.6 Hz), 117.7, 115.5 (d, *J* = 9.2 Hz), 115.2, 114.8, 38.3, 21.2; Elemental analysis: calc. for C~14~H~10~FN~5~ (MW 267.26): 62.92% C, 3.77% H, 26.20% N; found 62.71% C, 3.59% H, 26.38% N. *5-Methyl-6-{\[4-(trifluoromethyl)benzyl\]amino}pyrazine-2,3-dicarbonitrile* (**11**). Light yellow crystalline solid. Yield 38.3%; M.p. 148.4--149.6 °C; IR (ATR-Ge, cm^−1^): 3397~m~ (-NH-), 2232~m~ (-CN), 1570~vs~, 1520~s~, 1422~m~, 1399~s~, 1324~vs~ (arom.), 1114~vs~ (-C-F); ^1^H-NMR (300 MHz) δ 8.54 (1H, t, *J* = 5.3 Hz, NH), 7.71--7.64 (2H, m, AA\', BB\', H2\', H6\'), 7.38--7.52 (2H, m, AA\', BB\', H3\', H5\'), 4.69 (2H, d, *J* = 5.3 Hz, NCH~2~), 2.51 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 152.9, 147.7, 143.1, 130.0, 128.2, 127.9 (q, *J* = 31.5 Hz), 125.4 (q, *J* = 3.7 Hz), 124.5 (q, *J* = 272.0 Hz), 117.8, 115.6, 114.8, 44.0, 21.2; Elemental analysis: calc. for C~15~H~10~F~3~N~5~ (MW 317.27): 56.78% C, 3.18% H, 22.07% N; found 56.59% C, 2.99% H, 21.91% N. *5-Methyl-6-{\[2-(trifluoromethyl)benzyl\]amino}pyrazine-2,3-dicarbonitrile* (**12**). Yellow-orange crystalline solid. Yield 49.1%; M.p. 157.8--159.4 °C (decomp.); IR (ATR-Ge, cm^−1^): 3408~m~ (-NH-), 2230~m~ (-CN), 1569~vs~, 1522~s~, 1430~m~, 1409~s~, 1398~s~, 1368~s~, 1314~vs~ (arom.), 1114~vs~ (-C-F); ^1^H-NMR (300 MHz) δ 8.76 (1H, bs, NH), 7.77--7.71 (1H, m, H3\'), 7.64--7.43 (3H, m, H4\', H5\', H6\'), 4.77 (2H, bs, NCH~2~), 2.51 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 153.0, 147.7, 136.3, 132.9, 130.0, 128.3, 127.8, 126.4 (q, *J* = 30.3 Hz), 126.1 (q, *J* = 6.0 Hz), 124.7 (q, *J* = 274.0 Hz), 118.0, 115.5, 114.7, 41.1, 21.2; Elemental analysis: calc. for C~15~H~10~F~3~N~5~ (MW 317.27): 56.78% C, 3.18% H, 22.07% N; found 56.62% C, 3.08% H, 22.00% N. *5-\[(2,4-Dimethoxybenzyl)amino\]-6-methylpyrazine-2,3-dicarbonitrile* (**13**). Ochre crystalline solid. Yield 62.3%; M.p. 151.6--152.9 °C; IR (ATR-Ge, cm^−1^): 3336~m~ (-NH-), 2924~w~ (-OCH~3~), 2224~m~ (-CN), 1609~s~, 1574~vs~, 1509~vs~, 1467~m~, 1440~s~, 1405~vs~ (arom.); ^1^H-NMR (300 MHz) δ 8.45 (1H, t, *J* = 5.7 Hz, NH), 7.07 (1H, d, *J* = 8.2 Hz, H6\'), 6.55 (1H, d, *J* = 2.5 Hz, H3\'), 6.43 (1H, dd, *J* = 8.2 Hz, *J* = 2.5 Hz, H5\'), 4.47 (2H, d, *J* = 5.7 Hz, NCH~2~), 3.80 (3H, s, OCH~3~), 3.72 (3H, s, OCH~3~), 2.44 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 160.1, 158.1, 153.0, 147.4, 130.2, 128.8, 117.3, 117.2, 115.8, 114.9, 104.6, 98.5, 55.7, 55.4, 39.5, 21.3; Elemental analysis: calc. for C~16~H~15~N~5~O~2~ (MW 309.32): 62.13% C, 4.89% H, 22.64% N; found 62.36% C, 5.01% H, 22.52% N. *5-Methyl-6-\[(3-nitrobenzyl)amino\]pyrazine-2,3-dicarbonitrile* (**14**). Orange-red crystalline solid. Yield 17.2%; M.p. decomp.; IR (ATR-Ge, cm^−1^): 3394~m~ (-NH-), 2231~m~ (-CN), 1571~vs~, 1547~m~, 1524~vs~, 1478~m~, 1430~m~, 1396~s~, 1366~s~, 1348~vs~ (arom.); ^1^H-NMR (300 MHz) δ 8.79 (1H, t, *J* = 5.8 Hz, NH), 8.23 (1H, s, H2\'), 8.11 (1H, d, *J* = 7.8 Hz, H4\'), 7.81 (1H, d, *J* = 7.8 Hz, H6\'), 7.62 (1H, t, *J* = 7.8 Hz, H5\'), 4.73 (2H, d, *J* = 5.8 Hz, NCH~2~), 2.47 (3H, s, CH~3~); ^13^C-NMR (75 MHz) δ 152.9, 148.0, 147.8, 140.6, 134.4, 130.0, 130.0, 122.5, 122.3, 117.9, 115.6, 114.8, 43.8, 21.2; Elemental analysis: calc. for C~14~H~10~N~6~O~2~ (MW 294.27): 57.14% C, 3.43% H, 28.56% N; found 56.93% C, 3.64% H, 28.56% N. *5-\[(4-Chlorobenzyl)amino\]-6-methylpyrazine-2,3-dicarbonitrile* (**15**). Brown crystalline solid. Yield 30.8%; M.p. 152.8--154.8 °C (decomp.); IR (ATR-Ge, cm^−1^): 3378~m~ (-NH-), 2230~m~ (-CN), 1570~vs~, 1521~s~, 1487~m~, 1441~m~, 1403~s~, 1347~s~ (arom.), 804~s~ (-C-Cl); ^1^H-NMR (500 MHz) δ 8.42 (1H, bs, NH), 7.39--7.33 (4H, m, H2\', H3\', H5\', H6\'), 4.59 (2H, s, NCH~2~), 2.45 (3H, s, CH~3~); ^13^C-NMR (125 MHz) δ 152.8, 147.6, 137.2, 131.8, 130.0, 129.4, 128.5, 117.6, 115.6, 114.8, 43.7, 21.2; Elemental analysis: calc. for C~14~H~10~ClN~5~ (MW 283.72): 59.27% C, 3.55% H, 24.68% N; found 59.16% C, 3.65% H, 24.62% N. 3.4. Lipophilicity HPLC Determination and Calculations ------------------------------------------------------ Experimental lipophilicity parameter log *k* was ascertained using an Agilent Technologies 1200 SL HPLC system with a SL G1315C Diode-Array Detector, ZORBAX XDB-C18 5 μm, 4 × 4 mm, Part No. 7995118-504 chromatographic pre-column and ZORBAX Eclipse XDB-C18 5 μm, 4.6 × 250 mm, Part No. 7995118-585 column (Agilent Technologies Inc., Colorado Springs, CO, USA). The separation process was controlled by Agilent ChemStation, version B.04.02 extended by spectral module (Agilent Technologies Inc.). A solution of MeOH (HPLC grade, 70%) and H~2~O (HPLC-Milli-Q Grade, 30%) was used as mobile phase. The total flow of the column was 1.0 mL/min, injection 20 μL, column temperature 30 °C. Detection wavelength λ= 210 nm and monitor wavelength λ= 270 nm were chosen for this measurement. The KI methanol solution was used for the dead time (TD) determination. Retention times (TR) of synthesized compounds were measured in minutes. The capacity factors *k* were calculated using Microsoft Excel according to formula *k* = (TR − TD)/TD, where TR is the retention time of the solute and TD denotes the dead time obtained *via* an unretained analyte. Log *k*, calculated from the capacity factor *k*, is used as the lipophilicity index converted to log *P* scale. Values of log *P* and Clog *P* were calculated with the PC programme CS ChemBioDraw Ultra 13.0 (CambridgeSoft, Cambridge, MA, USA). 3.5. Biological Assays ---------------------- ### 3.5.1. Antimycobacterial *In Vitro* Screening Mycobacterial screening was performed against four mycobacterial stems (*M. tuberculosis* H37Rv CNCTC My 331/88, *M. kansasii* CNCTC My 235/80, *M. avium ssp. avium* CNCTC My 80/72, *M. avium* CNCTC My 152/73 (Czech National Collection of Type Cultures, National Institute of Public Health, Prague, Czech Republic)) using isoniazid and pyrazinamide (Sigma-Aldrich) as standards. Culturing medium was Sula's semisynthetic medium (Trios, Prague, Czech Republic) with pH 6.0. Tested compounds were dissolved in dimethylsulfoxide (DMSO) and diluted with medium to final concentrations 100, 50, 25, 12.5, 6.25, 3.125 and 1.5625 μg/mL. The method used for this assay was microdilution broth panel method. The final concentration of DMSO did not exceed 1% (v/v) and did not affect the growth of *Mycobacteria*. The cultures were grown in Sula's medium at 37 °C in humid dark atmosphere. The antimycobacterial activity was determined using Alamar Blue colouring after 14 days, resp. 6 days against *M. kansasii,* of incubation as MIC (µg/mL). This evaluation was done in cooperation with Department of Clinical Microbiology, University Hospital in Hradec Kralove, Hradec Kralove, Czech Republic. ### 3.5.2. Antimycobacterial *In Vitro* Screening Against *M. smegmatis* This assay was focused on the activity against fast growing *Mycobacterium smegmatis* MC2155 (CIT Collection, Cork Institute of Technology, Cork, Ireland). The method used there was also microdilution broth panel method and as the medium was used Middlebrook 7H9 Broth with 10% of OADC supplement (Sigma-Aldrich). Tested compounds were dissolved in DMSO and medium and the final concentrations were set as 1000, 500, 250, 125, 62.5, 31.25, 15.625 and 7.8125 µg/mL. The final concentration of DMSO did not exceed 2% (v/v) and did not affect the growth of *M. smegmatis*. The standards used for determination of activity were isoniazid and ciprofloxacin. MIC was read after 48 h of incubation at 37 °C, addition of Alamar Blue stain was followed by 12 h of additional incubation with this reagent. This screening was performed under the patronage of the Department of Biological Sciences, Cork Institute of Technology, Cork, Ireland. ### 3.5.3. Antifungal and Antibacterial *In Vitro* Screenings Antibacterial evaluation was made using the microdilution broth method in plates M27A-M1 (200+10) against eight bacterial stems from the Czech Collection of Microorganisms (Brno, Czech Republic) or clinical isolates from Department of Clinical Microbiology, University Hospital in Hradec Kralove (Hradec Kralove, Czech Republic) (*Staphylococcus aureus* CCM 4516/08, *Staphylococcus aureus* H 5996/08 methicillin resistant, *Staphylococcus epidermidis* H 6966/08, *Enterococcus* sp. J14365/08, *Escherichia coli* CCM 4517, *Klebsiella pneumoniae* D 11750/08, *Klebsiella pneumoniae* J 14368/05 ESBL positive, *Pseudomonas aeruginosa* CCM 1961). Mueller Hinton broth was used for the cultivation that was done in humid atmosphere at 35 °C. The readings were made after 24 and 48 h and MIC was set as 80% inhibition of control. The standards were neomycin, bacitracin, penicillin G, ciprofloxacin and phenoxymethylpenicillin and tested products were dissolved in DMSO (final concentration of DMSO did not exceed 1% (v/v)) \[[@B40-molecules-19-00651]\]. Antifungal evaluation was also accomplished with microdilution broth method. On the contrary, there was used RPMI 1640 broth with glutamine as medium and conditions were humid and dark atmosphere, pH 7.0 (buffered with 3-morpholinopropane-1-sulfonic acid) and 35 °C. Eight fungal strains were used (*Candida albicans* ATCC 44859, *Candida tropicalis* 156, *Candida krusei* E28, *Candida glabrata* 20/I, *Trichosporon asahii* 1188, *Aspergillus fumigatus* 231, *Absidia corymbifera* 272, *Trichophyton mentagrophytes* 445) together with 4 antimycotic standards amphotericin B, voriconazole, nystatin and fluconazole. The MIC was set as 80% inhibition of control and readings were made after 24 and 48 h (50% IC, 72 and 120 h for fibrous fungi) and tested compounds were also dissolved in DMSO (final concentration of DMSO in medium did not exceed 2.5% (v/v)) \[[@B41-molecules-19-00651]\]. ### 3.5.4. Study of the Inhibition of Oxygen Evolution Rate in Spinach Chloroplasts Chloroplasts were prepared from spinach (*Spinacia oleracea* L.) according to Masarovicova and Kralova \[[@B42-molecules-19-00651]\]. The inhibition of photosynthetic electron transport (PET) in spinach chloroplasts was determined spectrophotometrically (Genesys 6, Thermo Scientific, Madison, WI, USA) using an artificial electron acceptor 2,6-dichlorophenol-indophenol (DCPIP) according to Kralova *et al.* \[[@B43-molecules-19-00651]\] and the rate of photosynthetic electron transport (PET) was monitored as a photo-reduction of DCPIP. The measurements were carried out in a phosphate buffer (0.02 mol/L, pH 7.2) containing sucrose (0.4 mol/L), MgCl~2~ (0.005 mol/L) and NaCl (0.015 mol/L). The chlorophyll content was 30 mg/L in these experiments and the samples were irradiated (\~100 W/m^2^) from a 10 cm distance with halogen lamp (250 W) using a 4 cm water filter to prevent warming of the samples (suspension temperature 4 °C). The studied compounds were dissolved in DMSO due to their limited water solubility. The applied DMSO concentration (up to 4% (v/v)) did not affect the photochemical activity in spinach chloroplasts (PET). The inhibitory efficiency of the studied compounds was expressed as the IC~50~ values, *i.e.*, molar concentration of the compounds causing 50% decrease in the oxygen evolution relative to the untreated control. The comparable IC~50~ value for a selective herbicide 3-(3,4-dichlorophenyl)-1,1-dimethylurea (Diurone^®^, DCMU) was about 1.9 μmol/L \[[@B44-molecules-19-00651]\]. ### 3.5.5. Study of Fluorescence of Chlorophyll *a* in Spinach Chloroplasts The fluorescence emission spectra of chlorophyll *a* (Chl*a*) in spinach chloroplasts were recorded on fluorescence spectrophotometer F-2000 (Hitachi, Tokyo, Japan) using excitation wavelength λ~ex~ = 436 nm for monitoring fluorescence of Chl*a*, excitation slit 20 nm and emission slit 10 nm. The samples were kept in the dark 2 min before measuring. The phosphate buffer used for dilution of the chloroplast suspension was the same as described above. Due to low aqueous solubility the compounds were added to a chloroplast suspension in DMSO solution. The DMSO concentration in all samples was the same as in the control (10% (v/v)). The chlorophyll concentration in chloroplast suspension was 10 mg/L. 4. Conclusions ============== A series of 15 pyrazinamide derivatives (14 of them novel) was synthesized by aminodehalogenation reactions focusing on microwave assisted synthesis. All the final compounds were characterized with IR, NMR and other analytical data and then subjected to *in vitro* evaluation in order to discover their potential antimycobacterial, antifungal, antibacterial and herbicidal activities. The lipophilicity was measured using RP-HPLC methodology and also calculated or predicted with the PC program CS ChemBioDraw Ultra 13.0. These values were compared and the dependence between log *k* and log *P* was linear. In antimycobacterial screening compounds **3** (R = 3,4-Cl), **9** (R = 2-Cl) and **11** (R = 4-CF~3~) showed good activity against wild strain *M. tuberculosis* H37Rv (MIC = 6.25 µg/mL) compared to the standards pyrazinamide (MIC = 12.5 µg/mL) and isoniazid (MIC = 1.56 µg/mL). Compounds **11** (R = 4-CF~3~) and **12** (R = 2-CF~3~) were active against the non-tuberculosis strains *M. kansasii* and *M. avium* as well. Although the majority of synthesized compounds were active, there is no clear dependence between lipophilicity and antimycobacterial activity, but it can be stated that the most potent substances were also from the group of most lipophilic compounds and the most favourable substitutions are the electron-withdrawing groups such as chlorine or trifluoromethyl. Activity against fast growing *Mycobacteria* was also determined but no active substances were identified. No interesting results were observed in antibacterial and antifungal screenings. Three compounds (**7** (R = 4-NH~2~), **10** (R = 2-F), and **15** (R = 4-Cl)) showed insignificant activity against the fungus *Trichophyton mentagrophytes*, which was found to be worse compared to the standards. The rest of substances showed no *in vitro* antibacterial or antifungal activity. On the contrary, *N*-substituted 5-amino-6-methylpyrazine-2,3-dicarbonitriles were found to inhibit the Hill reaction in spinach chloroplasts which indicated that these compounds act as PS2 inhibitors.The IC~50~ values related to PET inhibition varied in the investigated set in the range from 16.4 μmol/L (**3**; R = 3,4-Cl) to 487.0 μmol/L (**14**; R = 3-NO~2~). The lipophilicity of the compounds was determinant for PET‑inhibiting activity. The site of inhibitory action of studied compounds in the photosynthetic apparatus is situated both on the donor and on the acceptor side of PS2. Perturbation of the Chl*a*-protein complexes in the thylakoid membranes caused by the tested compounds was documented by a decrease of the Chl*a* emission band intensity at 686 nm belonging to the pigment‑protein complexes in photosystem 2. Based on the obtained results it can be assumed that the antifungal and antibacterial activities of studied compounds are not directly dependent on lipophilicity. This conclusion cannot be applied for the herbicidal activity because there is a linear dependence between activity and lipophilicity. Dependence between antimycobacterial activity and the benzyl substituents was found. This reliance was expressed by the lipophilicity parameter log *k* resp. π constant and showed the importance of the ring-substituted benzyl moiety. The publication is co-financed by the European Social Fund and the state budget of the Czech Republic. Project No. CZ.1.07/2.3.00/20.0235, the title of the project: TEAB. This study was also supported by the Ministry of Health of Czech Republic (IGA NZ 13346), Grant Agency of Charles University B-CH/710312, the Slovak Grant Agency VEGA, Grant No. 1/0612/11, by the Project APVV-0061-11 and by Sanofi-Aventis Pharma Slovakia. The authors also wish to thank Ida Dufkova for performing and evaluating the antifungal and antibacterial assays and Barbora Servusova and Jan Zitko for English revisions. *Sample Availability*: Samples of the compounds are available from the authors. ###### Click here for additional data file. The authors declare no conflict of interest.