Datasets:

suolyer

/

pile_pubmed-central

like 1

Methicillin-resistant *Staphylococcus aureus* (MRSA) emerged in the 1960s and has since become a major cause of illness and death in the healthcare setting ([@R1]*,*[@R2]). Risk factors for infection with healthcare-associated MRSA (HA-MRSA) include hospitalization, residence in a long-term care facility, older age, invasive devices (e.g., catheters, feeding tubes), and exposure to antimicrobial agents. HA-MRSA isolates are often resistant to several antimicrobial drug classes in addition to β-lactams ([@R3]). In the 1990s, investigators began describing serious MRSA infections among persons who did not have typical risk factors for infections with this organism ([@R2]*,*[@R4]*--*[@R8]). These community-associated MRSA (CA-MRSA) infections affected young, healthy persons ([@R4]*,*[@R5]*,*[@R7]) and were associated with factors such as participating in contact sports, sharing towels or athletic equipment, using illegal intravenous drugs, and living in crowded or unsanitary areas (e.g., prisons, hurricane evacuee centers) ([@R9]*,*[@R10]). Pulsed-field gel electrophoresis (PFGE) demonstrated that MRSA strains causing these community-associated infections (USA300 and USA400) were different than those causing healthcare-associated infections (USA100 and USA200) ([@R11]). USA300 and USA400 MRSA strains typically have the staphylococcal cassette chromosome (SCC) *mec* type IV, not the SCC*mec* type II carried by most USA100 and USA200 isolates ([@R12]). In addition, USA300/400 isolates usually carry the gene that encodes the Panton-Valentine leukocidin (pvl), a bicomponent (*luk*F-PV and *luk*S-PV) pore-forming leukotoxin ([@R8]*,*[@R13]*--*[@R15]). Currently, the role of PVL in the pathogenesis of infections caused by USA300/400 isolates is controversial. Epidemiologic studies and a study by **Labandeira-Rey** et al. suggest that PVL is associated with virulence and causes the necrosis characteristic of infections with these strains ([@R16]). In contrast, a study by Voyich et al. found no difference in virulence between the wild-type parent strains and the isogenic knockout strains that did not produce PVL ([@R17]). A recent multicenter study by Moran et al. showed that USA300 MRSA is now the most common cause of skin and soft tissue infections (SSTIs) among adults seeking treatment in emergency departments in 11 large metropolitan areas ([@R15]). USA300 also causes serious invasive infections such as necrotizing pneumonia, bloodstream infections, and surgical site infections, some of which are acquired in hospitals ([@R18]*--*[@R22]). Although most USA300 and USA400 isolates are currently resistant to fewer classes of antimicrobial drugs than are HA-MRSA isolates ([@R13]), a recent paper by Han et al. identified a USA300 subtype that is resistant to erythromycin, clindamycin (constitutive), tetracycline, mupirocin, and fluoroquinolones ([@R23]). Most epidemiologic studies of CA-MRSA have examined isolates from SSTIs infections ([@R7]*,*[@R15]*,*[@R18]), and most studies that evaluated patients with invasive disease have involved single healthcare facilities ([@R21]*,*[@R24]) or isolates obtained primarily from large urban areas ([@R22]). We describe the molecular epidemiology of invasive infections caused by USA300 and USA400 in a rural state. We characterized invasive MRSA from 1999--2005 (select isolates) and in 2006 (all isolates) submitted to the statewide surveillance system in Iowa for invasive MRSA infections. Methods ======= As part of a statewide surveillance system, the Iowa Department of Public Health has mandated since 1999 that clinical microbiology laboratories submit invasive isolates of MRSA to the University Hygienic Laboratory (UHL), Iowa's public health laboratory ([@R25]*,*[@R26]). After performing antimicrobial drug susceptibility testing on all isolates, we further characterized (by PFGE, PVL detection, and SCC*mec* typing) all isolates from 1999--2005 that were resistant to [\<]{.ul}3 non--β-lactam antimicrobial drug classes (i.e., consistent with USA300/400) and all 343 isolates from unique patients with invasive infections submitted to the UHL during 2006. Antimicrobial Drug Susceptibility Testing ----------------------------------------- All invasive MRSA isolates during 1999--2006 were tested for antimicrobial drug susceptibility by the broth dilution method described by the Clinical and Laboratory Standards Institute ([@R27]). Invasive isolates were defined as any organism from a normally sterile body site such as blood, cerebrospinal fluid, pleural fluid, joint fluid, or fluid from a liver abscess. Isolates from urine were excluded. Isolates were tested for susceptibility to tetracycline, erythromycin, clindamycin, trimethoprim/sulfamethoxazole, gentamicin, levofloxacin, moxifloxacin, linezolid, daptomycin, vancomycin, and rifampin. Multidrug-resistant isolates were defined as MRSA isolates that were resistant to more than 3 of 8 representative antimicrobial drug classes: macrolides (erythromycin), lincosamides (clindamycin), quinolones (levofloxacin or moxifloxacin), tetracyclines, sulfa drugs (trimethoprim/sulfamethoxazole), aminoglycosides (gentamicin), glycopeptides (vancomycin), and rifampin. Molecular Typing and PCR to Assess SCC*mec* Type and Presence of the PVL Gene ----------------------------------------------------------------------------- PFGE was performed as previously described ([@R28]). Each gel accommodated bacteriophage Lambda ladders (at 3 places on the gel), DNA from 17 isolates, type strains for USA300 and USA400 from the Centers for Disease Control and Prevention (Atlanta, GA, USA), and *S. aureus* NCTC-8325 (at 3 places on the gel). The gel images were saved as TIFF files and BioNumerics computer software (Biosystematica, Llandysul, Wales, UK) was used to perform cluster analysis. Isolates were classified as the same strain if cluster analysis indicated that they were [\>]{.ul}80% similar. PFGE patterns for clinical isolates were compared visually and by computer-assisted gel analysis with the type strains for USA300 and USA400. We defined CA-MRSA as MRSA isolates in either the USA300 or the USA400 pulsetypes. Multiplex PCR was performed, as previously described, to type the SCC*mec* A ([@R29]) and to detect the PVL genes ([@R30]). Epidemiologic Data Collection ----------------------------- Epidemiologic data on the isolates were obtained from UHL. These data were age, sex, race/ethnicity, inpatient status, intensive care unit status, long-term-care facility status, hospital admission date, specimen type, specimen collection date, the hospital code number, and the Iowa Reporting Region. Isolates were considered to have been acquired nosocomially if the specimen culture date minus the admission date was [\>]{.ul}2 days. Statistical Methods ------------------- PFGE patterns and antimicrobial drug susceptibility test results were merged with the demographic data. These data were analyzed with SAS version 9.1 (SAS Institute, Cary, NC, USA) to assess trends in the frequency of USA300/400 in Iowa and to identify possible risk factors for invasive infections with these strains. We used χ^2^ and adjusted χ^2^ tests to analyze categorical data and linear regression and logistic regression to analyze continuous data. Alpha was set at 0.05 and all reported p values were 2-tailed. Seasonality of infections was analyzed by χ^2^ analysis. Winter was defined as December 22 to March 19, spring as March 20 to June 20, summer as June 21 to September 22, and fall as September 23 to December 21. The relationships between CA-MRSA and potential risk factors were assessed by univariate analysis. Subsequently, stepwise logistic regression was used to identify factors independently associated with invasive CA-MRSA infection. Results ======= Patients infected by USA300/400 MRSA were younger than those infected by other strains (p\<0.0001 for both time periods; [Tables 1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}). During 2006, more males than females were infected with USA300/400 (p = 0.06). Most isolates during both time periods were obtained from blood cultures and the distribution of strains did not vary by body site. Most patients were hospitalized for their infections and the proportion of patients admitted to intensive care units did not vary by strain (p = 0.27 and p = 0.35). However, the proportion of MRSA infections that met the definition of nosocomial decreased significantly from 26.1% from 1999--2005 to 16.6% in 2006 (p = 0.0003). During 2006, patients infected with other MRSA strains were more likely than those infected with USA300/400 to have infections that met the definition of nosocomial (p = 0.0006). ###### Descriptive epidemiology of invasive MRSA in Iowa, USA, 1999--2005\* Characteristic† Total no. (%), N = 1,323 USA type p value ---------------------- -------------------------- ------------------ ------------------ ---------- Mean age, y 67.8 (SD = 17.6) 46.0 (SD = 22.0) 68.2 (SD = 17.2) \<0.0001 Female gender 550 (41.6) 9 (34.6) 541 (42.7) 0.549 Inpatient stay 1,124 (85.0) 24 (92.3) 1,100 (84.8) 1.000 ICU admission 221 (16.7) 4 (15.4) 217 (16.7) 0.764 Nosocomial infection 346 (26.2) 5 (19.2) 341 (26.3) 0.306 Specimen type \<0.0001 Blood 1,256 (94.9) 25 (96.2) 1,231 (94.9) CSF 9 (0.7) 0 9 (0.7) Joint fluid 33 (2.5) 1 (3.9) 32 (2.5) Pleural fluid 8 (0.6) 0 8 (0.6) Other 6 (0.5) 0 (0.0) 6 (0.5) Iowa region 0.054 1 32 (2.4) 1 (3.9) 31 (2.4) 2 370 (28.0) 10 (38.5) 360 (27.8) 3 335 (25.3) 2 (7.7) 333 (25.7) 4 272 (20.6) 4 (15.4) 268 (20.7) 5 140 (10.6) 5 (19.2) 135 (10.4) 6 63 (4.8) 4 (15.4) 59 (4.5) PVL ND ND ND ND SCC*mec* IV ND ND ND ND \*MRSA, methicillin-resistant *Staphylococcus aureus*; ICU, intensive care unit; CSF, cerebrospinal fluid; PVL, Panton-Valentine leukocidin; SCC*mec* IV, staphylococcal chromosomal cassette *mec* type IV; ND, not done for all isolates.
†The number of patients missing data on specific variables: age = 13; gender = 11; inpatient = 85; ICU = 356; nosocomial = 358; specimen type = 11; Iowa Department of Public Health Reporting Region = 11.
‡Of the subset of isolates that were typed (N = 180), 173 (96%) were USA100. The remainder clustered with USA200 ([@R3]), USA500 ([@R2]), or did not match an existing USA type. ###### Descriptive epidemiology of invasive MRSA in Iowa, USA, 2006\* Characteristic† Total no. (%), N = 343 USA type p value ---------------------- ------------------------ ------------------ ------------------ ---------- Mean age, y 66.3 (SD = 17.0) 50.6 (SD = 21.2) 69.2 (SD = 14.4) \<0.0001 Female gender 135 (39.4) 14 (25.9) 121 (41.9) 0.059 Inpatient stay 278 (81.0) 50 (92.6) 228 (78.9) 0.271 ICU admission 56 (16.3) 8 (14.8) 48 (16.7) 0.348 Nosocomial infection 57 (16.6) 1 (1.9) 56 (19.4) 0.0006 Specimen type 0.0021 Blood 322 (93.9) 45 (83.3) 276 (95.0) CSF 0 0 0 Joint fluid 13 (3.8) 5 (9.3) 8 (2.9) Pleural fluid 2 (0.6) 2 (3.7) 0 Other 6 (1.7) 2 (3.7) 5 (1.4) Iowa region 0.268 1 10 (2.9) 0 10 (3.5) 2 93 (27.0) 13 (24.1) 80 (27.7) 3 49 (14.2) 12 (22.2) 37 (12.8) 4 105 (30. 5) 16 (29.6) 88 (30.5) 6 20 (5.8) 5 (9.3) 15 (5.2) PVL 54 (15.7) 52 (96.3) 2 (0.7)§ \<0.0001 SCC*mec* IV 68 (19.8) 54 (100.0) 13 (4.5) \<0.0001 \*MRSA, methicillin-resistant *Staphylococcus aureus*; ICU, intensive care unit; CSF, cerebrospinal fluid; PVL, Panton-Valentine leukocidin; SCC*mec* IV, Staphylococcal chromosomal cassette *mec* type IV.
†The number of patients missing data on specific variables: age = 10; gender = 12; inpatient = 31; ICU = 122; nosocomial = 101; specimen type = 0; Iowa Department of Public Health Reporting Region = 6; PVL = 3; SCC*mec* IV = 3.
‡Of the subset of isolates that were typed (N = 272) 94% were USA100. The remainder clustered with USA200 (5), USA500 (5), USA600 (1), USA800 (4), or did not match an existing USA type.
§Both isolates clustered with USA100 and were SCC*mec*II. The antimicrobial susceptibility of 54 invasive USA300/400 isolates is shown in [Table 3](#T3){ref-type="table"}. None of the USA300 or USA400 isolates had a multidrug-resistant phenotype (e.g., all were resistant to [\<]{.ul}3 non--β-lactam classes). Specifically, none of the USA300 isolates from Iowa demonstrated the multidrug-resistance pattern described by Diep et al. that is mediated by the multidrug-resistance plasmid pUSA03 ([@R31]). All isolates were typeable when the DNA was digested with *Sma*I. We did not identify any invasive MRSA infections caused by USA300/400 between 1999 and 2002. USA300 caused 3 (1.5%) of 195 infections in 2003, 14 (5.8%) of 243 infections in 2004, 7 (2.5%) of 275 infections in 2005, and 51 (14.9%) of 343 infections in 2006. USA400 caused 2 (0.7%) of 275 infections in 2005 and 3 (0.9%) of 343 infections in 2006. The proportion of MRSA isolates from invasive infections that were CA-MRSA (either USA300 or USA400) increased significantly from 1999--2005 to 2006 (p\<0.0001; [Figure 1](#F1){ref-type="fig"}). ###### Antimicrobial drug susceptibility of 54 invasive MRSA USA300/400 isolates, Iowa, USA, 2006\* Antimicrobial agent \% Susceptible ------------------------------- ---------------- Erythromycin 9 Levofloxacin 57 Clindamycin† 93 Tetracycline 93 Mupirocin 98 Rifampin 98 Trimethoprim/sulfamethoxazole 100 Vancomycin 100 Gentamicin 100 Daptomycin 100 Linezolid 100 \*MRSA, methicillin-resistant *Staphylococcus aureus*.
†Includes D-testing of all erythromycin-resistant, clindamycin-susceptible isolates. {#F1} Reporting region 4, which had the third largest population of the 10 regions, submitted the most isolates; region 1, which had the fourth smallest population, submitted the fewest isolates. We did not find significant differences between the type of MRSA causing infections and the reporting region during 2006. Incidence of infections caused by CA-MRSA varied by season during 2006 (p = 0.0004); a total of 47.3% of these infections occurred during the summer ([Figure 2](#F2){ref-type="fig"}). {#F2} The full model for predicting invasive infection with CA-MRSA compared with HA-MRSA included age (young vs. old), seasonality, hospital exposure, and specimen type. However, the only significant predictors of CA-MRSA infection compared with HA-MRSA were age \<69 years, which was associated with increased risk (odds ratio \[OR\] 5.1, 95% confidence interval \[CI\] 2.06--12.64), and hospital exposure (OR 0.07, 95% CI 0.01--0.51), which was associated with decreased risk. Discussion ========== The current study was unique because it evaluated invasive MRSA isolates from a statewide surveillance system in a rural area. Most prior studies of the epidemiology of CA-MRSA have focused on SSTI among patients in urban areas. The published literature documents that incidence of CA-MRSA has increased over time in large urban areas. For example, Kaplan et al. found that incidence of CA-MRSA at Texas Children's Hospital increased each year from August 1, 2001 to July 31, 2004 ([@R32]). The percentage caused by USA300 increased from ≈50% in 2000 to \>90% in 2003. Of these infections, 95.6% were SSTI and 4.4% were invasive. EMERGEncy ID NET reported that USA300 caused 97% of the MRSA SSTIs seen in emergency rooms in 11 US metropolitan areas during August, 2004 ([@R15]). Seybold et al. demonstrated that by 2004 USA300 had become a common cause of MRSA healthcare-associated bloodstream infections (28%) and of nosocomial MRSA bloodstream infections (20%) at Grady Memorial Hospital in Atlanta ([@R21]). In contrast, the number of CA-MRSA (primarily USA300) isolates causing invasive infections did not increase substantially in Iowa until 2006. Klevens et al. published a study of 8,987 invasive MRSA infections reported from the 9 sites in the Active Bacterial Core surveillance (ABCs/Emerging Infections Program Network) from July 2004 through December 2005 ([@R22]). The investigators conducted PFGE on 864 (11.3%) of the 7,648 isolates submitted from 8 sites. Of these isolates, 29% were USA300 (16% of the healthcare-associated, hospital-onset infections, 22% of the healthcare-associated, community-onset infections, and 67% of the community-associated, community-onset infections); \<0.1% were USA400. In our study, 4.5% of all isolates we typed and 14.9% of isolates from 2006 were USA300, which suggested that the incidence of invasive infections caused by USA300 remains lower in Iowa than in the urban centers studied by Klevens et al. Unlike the findings of Seybold et al. ([@R21]) and Klevens et al ([@R22]) from studies conducted in urban areas, USA300/400 rarely caused invasive nosocomial infections in Iowa, a rural state, during the study period. However, unpublished data from the University of Iowa Hospitals and Clinics indicate that these strains are becoming more common causes of invasive nosocomial infections, suggesting that the epidemiology of MRSA may be changing more slowly in rural areas than in large urban areas. Diep et al. published a follow-up study of previous observations by Han et al. ([@R23]) about multidrug-resistant USA300 isolates ([@R31]). These investigators found multidrug-resistant isolates in Boston and in San Francisco and they identified male to male sex, past MRSA infection, and use of clindamycin to be risk factors for multiresistant USA300. A multidrug resistance plasmid (pUSA03) mediated these drug resistances. Fortunately, we did not identify this resistance phenotype among our USA300 isolates from Iowa. However, given the rapidity with which plasmid-mediated antimicrobial drug resistance can spread, and given the epidemic nature of USA300, we will continue surveillance for this and other antimicrobial resistances among USA300 isolates in Iowa. Investigators in the Netherlands, Denmark, and Canada have found nontypeable MRSA among swine ([@R33]*--*[@R37]) and persons caring for swine ([@R33]*--*[@R36]). Strain ST398, which is not typeable by PFGE after DNA is digested with *Sma*I, has been found in all of these countries. Iowa produces ≈25% of the swine in the United States. However, we did not identify this strain among the invasive MRSA isolates submitted to the UHL. Our data did not include information about preceding influenza infections, but we noted that CA-MRSA was isolated twice from the pleural space; other strains of MRSA were not isolated from this site. This finding suggests that CA-MRSA may have caused serious pulmonary infections in a few persons in Iowa. CA-MRSA, particularly USA300, has caused severe infections after influenza (or influenza-like) infections ([@R20]*,*[@R24]). During the influenza pandemic of 1918, Chickering and Park noted that many patients acquired severe secondary *S. aureus* pneumonias following influenza infections ([@R38]). Their observations suggest that coincident outbreaks of pandemic influenza and USA300 pneumonia could be catastrophic. CA-MRSA, particularly USA300, is a public health concern for several other reasons. As noted previously, USA300 is rapidly replacing other strains of MRSA in the community ([@R15]*,*[@R31]) and has become an important nosocomial pathogen ([@R21]). Moreover, the types of infections caused by USA300 and the epidemiology of this strain are changing rapidly ([@R39]). Currently, most USA300 and USA400 isolates have fewer co-resistances than do HA-MRSA isolates ([@R13]). However, selective pressures can cause genetic drift in favor of more resistances; papers by Han et al. ([@R23]) and Diep et al. ([@R31]) documented that USA300 can acquire additional drug resistance determinants. If USA300 and USA400 become resistant to oral antimicrobial agents and the proportion of invasive MRSA infections caused by CA-MRSA continues to increase, many more patients will need parenteral vancomycin therapy, which will increase the difficulty and cost of treating these infections. Furthermore, as the incidence of CA-MRSA infections increases, horizontal transmission of these strains could increase in hospitals, making control of MRSA in healthcare settings even more difficult ([@R6]*,*[@R40]). Nosocomial bloodstream infections, ventilator-associated pneumonia, and surgical site infections caused by these strains could be devastating given the necrotizing nature of the infections. Our study had several limitations. First, the surveillance system was passive. Consequently, demographic data and data about race/ethnicity and exposure to long-term-care facilities were incomplete and data about prior antimicrobial drug exposure and underlying diseases were not available. Additionally, we could identify the region where the specimen originated but not the specific city or county. Moreover, a validation study in Iowa found that hospital laboratories submit only 37% of the invasive MRSA isolates that they identify (D. Dufficy, pers. comm.). However, underreporting affected all regions equally. Furthermore, we used different selection criteria for typing invasive MRSA isolates submitted from 1999--2005 than we did for those submitted in 2006. We typed isolates from 1999--2005 only if they had [\<]{.ul}3 non--β-lactam coresistances, but we typed all invasive MRSA isolates from 2006. Thus, we may have introduced selection bias that would predispose the incidence of USA300/400 during 1999--2005 toward the null hypothesis (i.e., the annual proportion of MRSA isolates that were USA300/400 did not change during1999 to 2005) but away from the null hypothesis for the incidence of USA300/400 isolates during 1999--2006 (i.e., H~0~: The annual proportion of MRSA isolates that were USA300/400 was the same in 1999--2005 and in 2006). However, given that all the USA300/400 isolates identified during 2006 would have been detected using the 1999--2005 coresistance selection criterion, we believe that we typed all of the invasive USA300/400 isolates obtained during 1999--2005. Some might argue that the incidence of invasive infections caused by CA-MRSA increased artificially because physicians were more aware of these organisms in 2006 than they were previously. CA-MRSA certainly has become a hot topic. CA-MRSA was initially identified in the mid 1990s, and many articles about these organisms have been published since then. However, the incidence of invasive infections caused by CA-MRSA in Iowa did not begin rising until 2006. Moreover, Iowa requires laboratories to send all invasive MRSA isolates to the UHL, and laboratory personnel are unlikely to know the details of each patient's infection. Thus, laboratories probably would not submit isolates of 1 strain preferentially. Finally, many hospitals submitted isolates to the UHL, which suggests that submission bias was unlikely in 2006. In conclusion, the number of invasive MRSA infections reported in Iowa and the number of invasive infections caused by CA-MRSA increased in Iowa from 1999--2005 to 2006. The increase of CA-MRSA (particularly USA300) poses a unique public health threat. Our study demonstrated that CA-MRSA no longer causes only SSTIs but now causes an increased proportion of invasive infections in a rural state. This finding is particularly disconcerting in light of the severity of these infections and the reports of necrotizing pneumonia caused by USA300 after influenza or influenza-like illness. Surveillance systems must continue to monitor the number and incidence of infections caused by USA300 and to monitor these isolates for changes in antimicrobial susceptibility. The relationship between seasons and CA-MRSA warrants further study. *Suggested citation for this article*: Van De Griend P, Herwaldt LA, Alvis B, DeMartino M, Heilmann K, Doern G, et al. Community-associated methicillin-resistant *Staphylococcus aureus*, Iowa, USA. Emerg Infect Dis \[serial on the Internet\]. 2009 Oct \[*date cited*\]. Available from <http://www.cdc.gov/EID/content/15/10/1582.htm> We thank Patricia Quinlisk and Deborah L. Dufficy for the work performed by the Iowa Department of Public Health to create and validate the surveillance system. We also thank Michael Pentella for facilitating this study and for critiquing the manuscript. Finally, we thank Richard Hollis, Linda Boyken, Sam Messer, Jennifer Kroeger, and Cassie Dohrn for their assistance with the laboratory methods used in this study. This study was funded in part by a grant from the Centers for Disease Control and Prevention and was approved by the institutional review board of the University of Iowa College of Medicine. Mr Van De Griend has an MPH degree from the University of Iowa College of Public Health and is currently a medical student at the University of Iowa Carver College of Medicine. His research interest is molecular epidemiology of staphylococcal infections.

1. Introduction {#s0005} =============== Dental caries, a persistent infectious disease, affects billions of people with large individual differences in numbers of caries lesions and activity ([@bb0220]; [@bb0190]; [@bb0110]). At least half of school children and the vast majority of adults worldwide experience accordingly caries, and the economic burden of caries and dental diseases represents about 4.6% of global health expenditures ([@bb0140]). However, in Sweden and countries with a low caries prevalence, many children are either free of caries or have a low disease level while 15--20% have a high caries burden ([@bb0220]; [@bb0100]). These high caries cases are poorly explained by life style-related variables, such as sugar consumption, oral hygiene, or fluoride use (*i.e.* relative risks 0.9--1.2) and seem to be largely unaffected by traditional prevention based on the same factors ([@bb0100]). Accordingly, life style, saliva, and bacteria are poor predictors of caries development ([@bb0220]; [@bb0190]), and better etiological models and diagnostic and preventive tools are needed. In spite of many advances in etiological and biochemical mechanisms related to caries disease during the last decades ([@bb0220]; [@bb0170]), dental caries is still generally considered a life style condition in which plaque acidification from sugar consumption shifts the oral ecology toward aciduric and acid-producing species; of these, *Streptococcus mutans* is the most well-known ([@bb0220]; [@bb0230]; [@bb0040]; [@bb0195]; [@bb0005]). The bacterium was early identified as the primary caries pathogen and vaccine candidate, but the inability of *S. mutans*, or any other species, to match or predict individual caries development has hampered its use in individualized oral care ([@bb0220]; [@bb0190]). *S. mutans* (serotypes *c* \> \> *e* \> *f* and *k*) infects the oral cavity of 40--80% of subjects depending on age, ethnicity and disease prevalence and colonize individuals with a dominant and largely unique genotype transmitted from parent to child through saliva ([@bb0120]; [@bb0065]). Cariogenic properties besides acid production that dissolves enamel are oxygen tolerance, bacteriocin production and adhesion and colonization at tooth surfaces ([@bb0040]; [@bb0195]). Together, sucrose-independent adhesion of SpaP or Cnm adhesins to host salivary agglutinin/DMBT1 and collagen, respectively, and sucrose-dependent adhesion of glycosyltransferases to bacterial polysaccharides allow *S. mutans* to colonize naked and cavitated tooth surface and promote plaque growth ([@bb0170]). Salivary agglutinin, originally identified by its ability to agglutinate *S. mutans*, is identical to DMBT1 or gp340 ([@bb0205]). Salivary agglutinin or DMBT1 is a pattern-recognition receptor composed of multiple domains designed to bind *S. mutans* and a wide array of microbes along with many innate and adaptive immunity factors ([@bb0155]; [@bb0145]; [@bb0150]). DMBT1 thus modulates innate and adaptive immunity, including complement activation, NF-κb signaling via Toll receptors and cellular proliferation ([@bb0155]). A 6.2 kb *dmbt1* deletion variant has been associated with cancer ([@bb0155]) and inflammatory bowel disease by increased NF-κb mediated inflammation in human cases ([@bb0215]). The corresponding salivary size variant (designated gp340 or DMBT1 size variant I) coincides with increased levels of caries and *S. mutans* adhesion in children ([@bb0095]). The AgI/II adhesin SpaP of *S. mutans* and AgI/II orthologs in various oral streptococci have been extensively explored for structural organization and interaction with its salivary receptor DMBT1 ([@bb0025]). AgI/II adhesins have an unusual tertiary structure where a central variable domain (V-domain) is presented at the tip of a stalk formed by intertwined, flanking alanine- and proline-rich regions ([@bb0125]). The carboxy-terminal domain (C-domain) to which a small N-terminal domain is bound is attached to the cell-wall via a cell-wall anchoring region ([@bb0080]). The SpaP binding sites for the DMBT1 agglutinin localize to the V-domain and the C-domain ([@bb0075]), and SpaP holds a supramolecular functional architecture at the cell surface ([@bb0085]). The SpaP adhesin harbors variants A, B and C (also referred to as A, B~1~, and B~2~) with clustered amino acid substitutions and different DMBT1 binding levels despite similar levels of SpaP expression ([@bb0065]). The interaction of SpaP and AgI/II orthologs with DMBT1 depends on whether DMBT1 is in the fluid- or surface-bound form and also depending on the *S. mutans* strain ([@bb0145]; [@bb0075]), suggesting that SpaP polymorphisms may modulate adhesion and aggregation by DMBT1 and consequently caries activity. *S. mutans* also harbors collagen-binding Cnm and Cbm adhesins in 15% and 3% of clinical isolates, respectively, and more frequently in serotype *e*, *f* and *k* than in *c* isolates ([@bb0015]). Cnm/Cbm are highly homologous and consist, similar to collagen-binding proteins in *Staphylococcus aureus* and other bacteria ([@bb0105]; [@bb0255]), of an N-terminal collagen-binding domain presented on a stalk formed by several threonine-rich repeat domains and a cell wall anchoring region ([@bb0175], [@bb0180]). Whereas *S. mutans* (serotypes *c*, *e*, *f* and *k*) may cause infective endocarditis ([@bb0090]), serotypes *f* and *k* in which Cnm and Cbm are more frequent coincide with inflammatory bowel disease ([@bb0115]) and Cnm phenotypes with hemorrhagic stroke ([@bb0165]; [@bb0250]). The Cnm/Cbm phenotypes also increase strain virulence in endocarditis ([@bb0185]). As potential virulence mechanisms in these so-called extra-oral infections, Cnm/Cbm mediate invasion of endothelial cells ([@bb0010], Review), formation of thrombus or heart valve vegetations or inhibition of platelet aggregation and wound healing ([@bb0165]; [@bb0015]). The aim of this study was to clarify the role of *S. mutans* as a caries pathogen by matching sucrose-independent adhesin types SpaP A, B, C and Cnm/Cbm with individual differences in caries development. We analyzed 452 Swedish children for the presence of *S. mutans* adhesin types and related them to baseline caries and 5-year increment and to cariogenic properties. 2. Methods {#s0010} ========== 2.1. Study Participants and Registration of Caries {#s0015} -------------------------------------------------- A total of 452 12-year-old children were collected as two independent samples (*n* = 218, *n* = 234) from 13 clinics in the northern county of Västerbotten, Sweden (Fig. S1). Included in the first sample were children born in 1996 and caries cases (≥ 1 Decayed and Filled Surfaces, DFS, in the permanent dentition) and caries-free controls in a 1:1 ratio; and for the second sample, they were born in 1998 and caries cases (≥ 2 DFS) and controls in a 2.1 ratio, and receiving ordinary dental care at Public Dental Service Clinics. The exclusion criterion was unwillingness to participate in the study. Both samples were re-examined after 5 years, for a total of 390 examined children with 14% drop-out rate (62/452) for having moved out of the area (20 children) or repeatedly missed the examination (42 children). The children received operative treatment and prevention of caries between 12 and 17 years of age, and 15% of the children orthodontic treatment with multibrackets after 12 years of age (as established from dental records), according to ordinary routines and policies at the clinics. The study was approved by the Ethics Committee for Human Experiments at Umeå University, Sweden, and informed consent was obtained from the children and their parents before participation. All parents signed consent to participate. Caries was recorded by three dentists (intra- and inter-examiner kappa ≥ 0.979) by a mirror, probe and two bitewing radiographs and mean number of [D]{.ul}ecayed ([e]{.ul}namel caries included), [F]{.ul}illed Surfaces in the permanent dentition (DeFS) was the primary caries outcome measure. The 5-year caries increment (ΔDeFS) was calculated by subtracting latest from earliest DeFS, dividing that value by the number of observed years and multiplying the result by 5. The 1:1 ratio (first sample) and increased 2:1 ratio (second sample) of caries cases *versus* controls and DeFS index generated a continuous gradient of discriminatory caries DeFS scores in the entire sample at baseline (Fig. S1, Table S1). 2.2. Genotyping of *S. Mutans* Adhesin Types in Whole Saliva {#s0020} ------------------------------------------------------------ Whole saliva, collected by chewing on paraffin for 5 min that was stored frozen at − 80 °C, was genotyped for *cnm*, *cbm* and *spaP A*, *B* and *C* status using bacterial DNA prepared with the GenElute™ bacterial genome DNA kit (Sigma-Aldrich, Sweden) and quantitative polymerase chain reaction (qPCR) using the KAPA SYBR Fast Universal qPCR kit (Tectum, Sweden). The primers for *cnm* and *cbm* were as described in Table S2 and did not cross-react in between or with other templates. The primers for *spaP A*, *B*, and *C* were from the *spaP* sequences (Table S2) and selected from prior testing and lack of cross-reactivity between *A*, *B*, and *C* or with DNA from oral streptococci with *spaP* analogs. The genotyping used internal standards and quantitative calibration curves based on dilutions of DNA purified from a reference genotype of each adhesin type, and cut-off values for *A* (3000 pg), *B* (3000 pg), *C* (6000 pg), *cnm* (3000 pg), and *cbm* (1000 pg). 2.3. Quantification of *S. Mutans* in Whole Saliva {#s0025} -------------------------------------------------- *S. mutans* in whole saliva was quantified by culturing and qPCR ([@bb0260]). Serial dilutions of whole saliva were cultured on MSB agar plates and counted for colony-forming units of *S. mutans* (designated as ms). Plaque DNA purified from whole saliva samples was measured by qPCR using the KAPA SYBR FAST Universal qPCR kit (Sigma-Aldrich, Sweden) and Corbett Rotor-Gene 6000 and *S. mutans* specific primers (Table S2). Quantitative calibration curves from DNA prepared from serial dilutions of *S. mutans* strain Ingbritt ([@bb0065]) were used to transform the qPCR responses into colony-forming units. 2.4. Isolation and Typing of *S. Mutans* Strains from Plaque {#s0030} ------------------------------------------------------------ Plaque was collected from buccal surfaces of teeth 34--36 (premolars and the first molar of the left lower jaw) in caries-free children or from caries lesion in children with tooth decay ([@bb0065]). Strains of *S. mutans* were isolated from the plaque samples by culturing on MSB agar plates and typed by Rapid ID 32 STREP Kit (Bio Merieux, La Balme les Grottes, France) metabolic tests ([@bb0090]). A total of 321 strains from 214 out of 217 infected children, one strain from each of the 214 children and additional isolates from 70 extreme or 26 *cnm*-positive children thereof, were isolated and phenotypically characterized (Fig. S1, Table S3). Of these isolates, one strain was mutated. The isolates were genotyped using qPCR and DNA purified by the GeneElute Bacterial Genomic DNA Kit (Sigma-Aldrich, Sweden). Typing of *cnm* and *cbm* status relied on specific primers and KAPA SYBR Fast Universal qPCR kit as described (Table S2) ([@bb0180]), and typing of serotypes *c*, *e*, *f* and *k* was performed using specific primers and the PCR polymerase kit (GE Healthcare, Sweden) as described (Table S2, [@bb0225]; [@bb0160]). 2.5. Sequencing of *SpaP* and *Cnm* Genes {#s0035} ----------------------------------------- The full length of *spaP* and *cnm* genes were amplified by PCR using iProof High fidelity DNA polymerase (Bio-Rad, Sweden) and specific primers for *spaP* (Table S2) ([@bb0065]) and *cnm* (Table S2) ([@bb0175]). Amplified fragments were sequenced both forward and reverse by dideoxy chain termination/cycling sequencing on an ABI 3730 XL machine (Eurofins Genomics, Germany). Primers for *spaP* and *cnm* sequencing are listed in Table S2. Assembly and analyses of the full gene sequences were done with DNA Sequence Polymorphism DnaSP software ([@bb0135]) and Molecular Evolutionary Genetics Analysis MEGA 5 software ([@bb0240]). SpaP protein models were developing using the known crystal structures of *S. mutans* SpaP (Protein Data Bank IDs 3IOX and 3QE5) and CCP4MG software ([@bb0200]). 2.6. Sequence Typing of *S. Mutans* Isolates {#s0040} -------------------------------------------- Sequence typing of *S. mutans* isolates (*n* = 144) was done by sequencing of housekeeping gene segments of *map*, *sod*, *rpoB*, *ppaC*, *pfl*, *pyk*, and *tuf* and inclusion of the *spaP* sequences ([@bb0065]; [@bb0020]; [@bb0050]). DnaSP software ([@bb0135]) was used to generate sequence characteristics (Table S4) and sequence types of the concatenated sequences, and MEGA 5 software ([@bb0240]) to generate neighbor joining trees from concatenated sequences. The neighbor joining tree of all concatenated sequences was more stable (upon bootstrapping) than that of *spaP* or housekeeping genes alone and appeared to cluster isolates with host caries status better ([@bb0050]). The eBURST software was used to group isolates into clonal complexes based on sharing of at least six out of eight alleles with at least one other member in the group ([@bb0070]). 2.7. Acid Tolerance {#s0045} ------------------- Isolates were screened for differences in acid tolerance by growth on Todd Hewitt medium agar plates at buffered pH values under aerobic conditions ([@bb0235]). The isolates (5 μL inoculum, OD~495~ 2.0) were grown in 5% CO~2~ at 37 °C for 5 d, and acid tolerance was scored as the presence or absence of colony-forming units at different pHs from 5.0 to 7.0. All isolates yielded colony-forming units at pH 7.0, but these failed to grow at pH 6.0, 5.5, and 5.25 for isolates of low, moderate, and high acid tolerance, respectively. Isolates of *S. mutans* were also tested for acid tolerance by quantitative growth curves in BHI broth at pH 7.0 and pH 5.0 at 37 °C under aerobic and anaerobic conditions. 2.8. Hydroxyapatite Adhesion Assay {#s0050} ---------------------------------- Adhesion of metabolically ^35^S-Met-labeled *S. mutans* isolates to saliva-coated hydroxyapatite beads (Bio-Rad Laboratories) was measured in 96 well microtiter plates as described ([@bb0065]; [@bb0145]; [@bb0095]). Briefly, after hydration of the hydroxyapatite beads in adhesion buffer at 4 °C overnight (5 mg beads in 125 μL per well), the beads were coated with 125 μL parotid saliva diluted in adhesion buffer 1:1 or with DMBT1 (2.5 μg/mL) or collagen type I (Sigma-Aldrich, Sweden) for 1 h and washed. The parotid saliva was collected by acid stimulation and Lashley cups and pooled from three donors before use ([@bb0065]), and DMBT1 was purified biochemically from parotid saliva of several donors as described ([@bb0145]; [@bb0060]). The coated beads were incubated with ^35^S-labeled bacteria (125 μL, 10^8^ cells/ml; 1000 cells/cpm) for 1 h under agitation. After repeated washes (3 × 200 μL adhesion buffer), the beads were counted with a scintillation counter (MicroBeta^2^, PerkinElmer) and the proportion of bound bacteria out of the total amount of added bacteria (percent adhesion) was calculated. 2.9. Microtiter well Adhesion Assay {#s0055} ----------------------------------- Adhesion of isolates to collagen was also measured in 96 well microtiter plates ([@bb0165]). Briefly, wells (NUNC, Sigma-Aldrich, Sweden) were incubated overnight at 4 °C with 100 μL collagen type I (30 μg/mL, Sigma-Aldrich, Sweden, C224910) in 10 mM phosphate buffered saline, (PBS), pH 7.2. The wells were washed three times with PBS and incubated for 3 h at 37 °C with a suspension of *S. mutans* (100 μL, 1.0 OD~600~ in PBS) from overnight cultures in Jordan broth. Adherent cells after washes with PBS were stained by adding 200 μL 0.05% crystal violet in water for 5 min, washed and measured by absorbance at 595 nm after addition of 96% ethanol. 2.10. Construction of *SpaP* and *Cnm* Knockout Strains {#s0060} ------------------------------------------------------- To construct *spaP* and *cnm* knockout mutants in *S. mutans* wild type strain 49 (no. 23 in [Fig. 2](#f0010){ref-type="fig"}, Fig. S2), a PCR overlapping strategy was used involving insertion of a selective marker (*ermAD* or *add9*) within the target gene. The DNA was purified using GeneElute Bacterial Genomic DNA Kit (Sigma-Aldrich, Sweden), PCR amplifications were done with iProof High fidelity DNA polymerase (Bio-Rad, Sweden) and all primers (Table S2) were from Eurofins Genomics (Germany). Briefly, the 5′- and 3′-regions of the target gene were fused to the selective marker by end homology PCR and sub cloned into ZeroBlunt TOPO vector and transformed into a TOPO10 *E. coli* strain (Bio-Rad, Sweden). Correct generated disruption fragments were confirmed by sequencing (Eurofins Genomics, Germany). For construction of knockout mutants the cloned disruption fragments were amplified using T7 and Sp6 primers, transformed into isolates of *S. mutans* strain 49 by using a competence-stimulating peptide (SGSLSTFFRLFNRSFTQALGK) ([@bb0130]). Correctly integrated fragments were confirmed by PCR and sequencing (Eurofins Genomics, Germany) (see also Fig. S5). 2.11. Multivariate Partial Least Squares (PLS) Statistics {#s0065} --------------------------------------------------------- Multivariate PLS regression was performed with Simca^+^ P12.01 (Umetrics AB, Umeå, Sweden) ([@bb0055]). We used PLS, which relates two data matrices X and Y to each other by a linear multivariate prediction model, to identify amino acid substitutions of SpaP (the X-matrix) that modify adhesion and caries (the Y-matrix). The PLS model gives the X variable\'s ability to explain (R^2^) and -- *via* cross-validation -- predict (Q^2^, \< 10% reflects a weak model) the variation in Y along with regression coefficients (CoeffCS) and the relative variable importance in projection (VIP) of each X variable (VIP-values \> 1.0 indicate influential X variables). 2.12. Univariate Statistics {#s0070} --------------------------- Data were expressed as means ± standard deviations (SD) and 95% confidence intervals (95% CI). The statistical analyses were performed with Chi-squared, Fisher\'s exact and Mann-Whitney *U* tests, or Spearman\'s rank correlation using SPSS or GraphPad softwares. All statistical analyses, univariate or multivariate, used independent isolates or subjects and two-tailed tests and *p* values \< 0.05 were considered significant. 3. Results {#s0075} ========== 3.1. Relationship Between Infection with *S. Mutans* and Caries Development {#s0080} --------------------------------------------------------------------------- Infection by *S. mutans* was measured in 452 12-year-old children by qPCR and culturing of whole salivas and related to caries scores at 12 and 17 years of age and 5-year caries increment (Fig. S1, [Table 1](#t0005){ref-type="table"}). About half of the children were infected with *S. mutans* at baseline (48%) and experienced a \~ 2-fold increased caries DeFS scores at 12 and 17 years of age and 5-year caries increment compared to non-infected children ([Table 1](#t0005){ref-type="table"}). The *S. mutans* infection load, however, did not coincide with baseline or prospective caries ([Table 1](#t0005){ref-type="table"}). Thus, children infected with *S. mutans* had 1.8-fold increased caries increment.Table 1Stability of infection by *S. mutans* and adhesin types and their association with caries in 452 Swedish children.Table 1Group or adhesin type[a](#tf0005){ref-type="table-fn"}, [b](#tf0010){ref-type="table-fn"}DeFS-12y[c](#tf0015){ref-type="table-fn"}DeFS-17y[c](#tf0015){ref-type="table-fn"}ΔDeFS-5y[d](#tf0020){ref-type="table-fn"}Colonization stability, %[g](#tf0035){ref-type="table-fn"}n%Mean ± SD*p*[e](#tf0025){ref-type="table-fn"}nMean ± SD*p*[e](#tf0025){ref-type="table-fn"}nMean ± SD*p*[e](#tf0025){ref-type="table-fn"}452 children*S. mutans* positive[f](#tf0030){ref-type="table-fn"}217483.6 ± 3.08 × 10^− 13^1859.2 ± 8.84 × 10^− 8^1855.5 ± 7.13 × 10^− 5^78*S. mutans* negative234521.7 ± 2.1Ref.2044.9 ± 5.7Ref.2043.1 ± 4.9Ref.

*S. mutans* positiveHigh \> 250,000 cfu49233.9 ± 3.00.2844310.3 ± 9.80.544436.4 ± 7.80.502Low \< 250,000 cfu168773.5 ± 3.0Ref.1428.8 ± 8.5Ref.1425.3 ± 6.8Ref.

SpaP A111513.0 ± 2.8Ref.968.4 ± 8.2Ref.965.2 ± 7.1Ref.82SpaP B71334.2 ± 3.50.0345511.4 ± 11.10.045556.7 ± 8.80.11088SpaP C1993.3 ± 1.80.365187.5 ± 4.80.746184.3 ± 4.10.92989Cnm26124.4 ± 2.60.0122010.4 ± 8.70.182206.0 ± 7.20.46099Cbm732.9 ± 3.70.63153.8 ± 2.80.20152.4 ± 1.60.599100

SpaP B^+^71254.2 ± 3.53 × 10^− 6^5511.4 ± 11.15 × 10^− 4^556.7 ± 8.88 × 10^− 4^SpaP B^−^363752.2 ± 2.4Ref.3196.0 ± 6.5Ref.3193.8 ± 5.4Ref.[^2][^3][^4][^5][^6][^7][^8] 3.2. Colonization of Specific *S. Mutans* Adhesin Types in Children in a 5-year Period {#s0085} -------------------------------------------------------------------------------------- To establish the colonization pattern of specific *S. mutans* adhesin types in the 452 children, SpaP (A, B, or C) and collagen-binding (Cnm or Cbm) adhesin types were determined in the 452 children at 12 and 17 years of age using qPCR ([Fig. 1](#f0005){ref-type="fig"}A, B; [Table 1](#t0005){ref-type="table"}). The *S. mutans*-positive children (*n* = 217) were either infected with a single dominant SpaP A (51%), B (32%), or C (8.5%) adhesin type or a mixture of two SpaP adhesin types (8.5%) at baseline. A portion of *S. mutans*-positive but none of the negative children were either Cnm-positive (26 children, 12%) or Cbm-positive (7 children, 3%) at baseline irrespective of SpaP A, B, or C status ([Table 1](#t0005){ref-type="table"}). *S. mutans* and its adhesin types generally colonized the children at both 12 and 17 years with a colonization stability ranging from 78% to 100% ([Table 1](#t0005){ref-type="table"}). Thus, children were stably colonized by specific SpaP A, B or C adhesin types that in some children also were Cnm or Cbm positive.Fig. 1Determination of *S. mutans* adhesin types in whole saliva. Children infected (+) or not infected (−) with A) a dominant or mixed SpaP A, B or C adhesin types and B) a Cnm or Cbm adhesin type of *S. mutans* by quantitative PCR of whole salivas (pg DNA responses, 95% CI). C) Total colony-forming units, CFU, of *S. mutans* (ms counts) in whole saliva of infected children as shown by box plots (\**p* = 0.003, Mann-Whitney *U* test).Fig. 1 3.3. Relationship of Adhesin Types to Individual Caries Development {#s0090} ------------------------------------------------------------------- We next matched the *S. mutans* adhesin types with individual differences in caries at 12 and 17 years and in the 5-year caries increment ([Table 1](#t0005){ref-type="table"}). The presence of *S. mutans* SpaP B and Cnm coincided with 1.4- and 1.5-fold increased caries scores at baseline compared to SpaP A (*p* = 0.034, *p* = 0.012), which was most prevalent and together with SpaP C and Cnm coincided with low caries scores at baseline. The increased association of SpaP B and Cnm with caries and comparably low caries scores of SpaP A, C, and Cbm remained at 17 years of age and for the 5-year caries increment ([Table 1](#t0005){ref-type="table"}). In addition, the differential association of the adhesin types with caries did not coincide with differences in numbers of *S. mutans* in saliva except for a slight reduction in Cbm-positive children ([Fig. 1](#f0005){ref-type="fig"}C). Thus, the SpaP B and Cnm adhesin types coincided with increased baseline and prospective caries development although the children experienced adolescence and puberty and were treated for caries by operative and preventive means (Table S1) during the study period. 3.4. Multilocus Sequence Typing (MLST) of *S. Mutans* Using Housekeeping and *SpaP* Gene Sequences {#s0095} -------------------------------------------------------------------------------------------------- To explore further the adhesin types for genetic relatedness and grouping of specific subtypes with caries cases, we isolated and typed *S. mutans* strains from all infected children as SpaP A, B, or C in good agreement with the saliva typing (Table S3) and subjected a subset of 70 isolates from 35 caries and 35 caries-free extremes to MLST using *spaP* and housekeeping genes ([Fig. 2](#f0010){ref-type="fig"}, Fig. S2; Table S4). The discriminatory ability of sequence typing related to host variables may increase with genes encoding bacterial proteins subject to host adaptation ([@bb0050]), and inclusion of *spaP* improved the grouping of isolates with caries cases and controls ([Fig. 2](#f0010){ref-type="fig"}, [Table 2](#t0010){ref-type="table"}).Fig. 2*S. mutans* biotypes A, B and C with corresponding SpaP adhesin subtypes of high (B-1) and low (A-1) cariogenicity. Clustering of *S. mutans* isolates from 35 caries (+) or 35 caries-free (−) extremes of infected children with caries status (+, − or DeFS at 12 years of age) based on neighbor joining and clonal complex analyses of *spaP* and housekeeping gene sequences. The isolates grouped both into biotypes A, B, and C with distinct SpaP A, B, and C adhesin types and clonal complexes C1--C9, and into genetically related subtypes that coincided with high caries (*i.e.* B-1, B-2 and C6, C7) or low caries (*i.e.* A-1 and C1) cases. The high- and low-cariogenic types also differed in acid tolerance and adhesion to DMBT1 and saliva. The clonal complexes share at least six identical alleles with at least one member in the group, all reflecting housekeeping alleles except for C8 which also shared the same *spaP B* allele. A total of 55 sequence types (ST) occurred among the 70 children. Numbers in parentheses mark STs shared in 15 children or identical STs in 25 children from which 74 additional strains were analyzed (mean 3, range 2 to 8 isolates/child).Fig. 2Table 2Association of *S. mutans* SpaP subtypes with caries in 35 caries and 35 nearly caries-free extremes of children.Table 2Group[a](#tf0040){ref-type="table-fn"}SpaP typeSub-typeDMBT1 bindingDeFS-12y[b](#tf0045){ref-type="table-fn"}DeFS-17y[b](#tf0045){ref-type="table-fn"}ΔDeFS-5y[c](#tf0050){ref-type="table-fn"}nMean ± SD*p*[d](#tf0055){ref-type="table-fn"}nMean ± SD*p*[d](#tf0055){ref-type="table-fn"}nMean ± SD*p*[d](#tf0055){ref-type="table-fn"}nMean ± SD*p*[d](#tf0055){ref-type="table-fn"}SpaP AA3913.3 ± 3.60.862393.2 ± 4.20.518349.0 ± 10.10.204345.5 ± 6.50.282AA-1913.2 ± 1.70.71292.7 ± 4.00.88285.3 ± 7.60.80183.2 ± 5.20.803AC11012.9 ± 1.8Ref.102.4 ± 3.9Ref.94.7 ± 7.5Ref.92.9 ± 5.0Ref.SpaP BB2717.8 ± 6.60.002273.8 ± 4.40.3202211.3 ± 5.00.114227.3 ± 10.40.104BB-11021.8 ± 8.3\< 0.001105.7 ± 4.30.074712.3 ± 10.60.03378.4 ± 7.30.029BB-2619.8 ± 5.30.00266.7 ± 3.90.041419.0 ± 10.90.023412.4 ± 7.20.019BC6217.7 ± 2.60.05327.0 ± 2.80.119228.0 ± 1.40.030218.2 ± 0.10.029SpaP CCC-1412.8 ± 3.10.39642.5 ± 2.50.48034.3 ± 2.30.29931.0 ± 0.10.925

Adhesion[e](#tf0060){ref-type="table-fn"}SpaP BBHigh1720.7 ± 6.5\< 0.001174.8 ± 4.30.0731516.1 ± 16.0\< 0.0011510.3 ± 11.50.003BLow1012.8 ± 2.1102.2 ± 4.271.0 ± 1.270.9 ± 1.0SpaP AAHigh1616.6 ± 1.5\< 0.001161.9 ± 3.50.090136.1 ± 8.00.353133.1 ± 3.80.260ALow2310.9 ± 1.5234.2 ± 4.42110.8 ± 10.9216.9 ± 7.4

Acid tolerance[f](#tf0065){ref-type="table-fn"}SpaP BBHigh134.4 ± 4.60.5581317.8 ± 16.70.0011311.6 ± 11.80.001SpaP ABLow143.3 ± 4.392.0 ± 2.291.1 ± 1.3AHigh75.3 ± 4.20.115612.8 ± 13.10.58567.8 ± 9.00.820ALow322.8 ± 4.1288.1 ± 9.4284.9 ± 6.0[^9][^10][^11][^12][^13][^14] Quantitative neighbor joining and qualitative clonal complex methods were used to analyze the sequence variations and grouped the SpaP A, B, and C adhesin types into corresponding biotypes A, B, and C that each had a unique housekeeping allele profile ([Fig. 2](#f0010){ref-type="fig"}). Accordingly, each biotype A, B, or C had a unique panel of clonal complexes C1--C9 in which each clonal complex shared allelic profile: C1--C5 in biotype A; C6--C8 in biotype B; and C9 in biotype C. Thus, biotypes A, B, and C in *S. mutans* may differ in adhesion-related events because of corresponding SpaP A, B, and C adhesin types and in other properties because of different housekeeping gene profiles. To identify SpaP A and B adhesin biotypes with markedly low and high cariogenicity, we further matched their specific subtypes with caries status of the strain donor ([Fig. 2](#f0010){ref-type="fig"}, [Table 2](#t0010){ref-type="table"}). A subset of biotype B isolates that grouped together in the neighbor joining (*i.e.* B-1, B-2) and clonal complex (*i.e.* C6, C7) analyses also clustered with children of high caries experience and increment. Also, genetically related isolates of biotypes A (*i.e.* A-1, C1) and C (*i.e.* C9) grouped with low-caries children. Children infected with B-1 isolates had a 2.6-fold increased 5-year caries increment compared to children with A-1 (C1) isolates (*p* = 0.029) ([Table 2](#t0010){ref-type="table"}), compared to the overall 1.3-fold difference between the biotype B and A isolates (*p* = 0.11) ([Table 1](#t0005){ref-type="table"}). Thus, subtypes of SpaP adhesin biotype B display markedly high cariogenicity and subtypes of SpaP adhesin biotype A markedly low cariogenicity. 3.5. Association of SpaP Adhesin Biotype B Isolates with DMBT1-Binding and Acid Tolerance {#s0100} ----------------------------------------------------------------------------------------- Having found subtypes of the SpaP adhesin biotypes A, B and C with different cariogenicity and *spaP* and housekeeping allele profiles, including genes potentially involved in acid tolerance ([@bb0045]), we next investigated the subtypes for binding to DMBT1 and saliva and acid tolerance ([Fig. 2](#f0010){ref-type="fig"}, Fig. S3). SpaP B isolates bound DMBT1 and saliva better than SpaP A and C isolates and high-cariogenicity SpaP B-1 isolates bound better than low-cariogenicity A-1 (C1) isolates ([Fig. 2](#f0010){ref-type="fig"}, Figs. S3, S4; [Table 2](#t0010){ref-type="table"}). High *versus* low DMBT1-binding of SpaP B isolates coincided with a 11.4-fold increased 5-year caries increment in children ([Table 2](#t0010){ref-type="table"}), and binding of SpaP B isolates to DMBT1 and saliva correlated with individual numbers of caries lesions ([Fig. 3](#f0015){ref-type="fig"}A). In contrast, the binding level in SpaP A isolates did not coincide with caries increment or individual caries scores (*r* = 0.06, *p* = 0.75) ([Table 2](#t0010){ref-type="table"}).Fig. 3SpaP B- and Cnm-mediated binding to DMBT1 correlates with individual caries scores. A) Correlation of binding of SpaP B to DMBT1 with individual numbers of caries DeFS lesions (binding to saliva generated virtually identical results, data not shown). B) Binding of Cnm-positive and Cnm-negative isolates to collagen, DMBT1, and saliva. Each isolate is marked by a dot and mean adhesion by a black line (Mann-Whitney *U* test). C) Correlation of binding of Cnm isolates (no. 1--12) to DMBT1 and collagen (*r* = 0.62, *p* = 0.031, respectively; data not shown). D) Deduced Cnm protein sequences of isolates no. 1--12. Substitutions are indicated with (\|), translational stop codons with (\*), and numbers of β-repeats with (o). LP = leading peptide. CWA = cell-wall anchoring motif.Fig. 3 Biotype B and subtype B-1 isolates had a high relative acid tolerance compared to biotype A and subtype A-1 (C1) isolates ([Fig. 2](#f0010){ref-type="fig"}, Fig. S3), which coincided with a 10.5-fold increased 5-year caries increment as compared to isolates with a low acid tolerance ([Table 2](#t0010){ref-type="table"}). 3.6. Association of Substitutions in SpaP B Isolates with DMBT1-Binding and Caries {#s0105} ---------------------------------------------------------------------------------- Having found an association between adhesion and caries in SpaP adhesin B subtypes and because they differ in sequence mainly outside the V-domain ([@bb0065], [Fig. 4](#f0020){ref-type="fig"}AB), we hypothesized that single amino acid substitutions in the C- and A-domains may specify the high-cariogenicity subtype B-1 isolates. We therefore used PLS analyses to screen single SpaP substitutions in SpaP B (and SpaP A) isolates for association with adhesion and caries ([Fig. 4](#f0020){ref-type="fig"}CD, Table S5). Substitutions that coincided with caries and adhesion were associated with high-cariogenicity SpaP adhesin B-1 (C6, C7) isolates, and 10 out of the 14 caries-associated substitutions in SpaP correlated positively with DMBT1-binding. SpaP A adhesin biotype A isolates behaved differently (Fig. S5, [Table 2](#t0010){ref-type="table"}). In addition, the clustered substitutions that distinguish SpaP A, B, and C all localized on opposite sides of a potential binding pocket in the DMBT1-binding V-domain ([Fig. 4](#f0020){ref-type="fig"}B).Fig. 4Adhesins SpaP A, B and C of different cariogenicity. A) Domain (A, V, P and C) structure and sequence identity of SpaP A, B and C and orthologs in various streptococci (reference SpaP A). The SpaP A, B and C adhesins differ by clustered substitutions in the V-domain and by single substitutions in the A-, P- and C- domains. LP = leading peptide, CWA = cell wall anchoring region. B) The clustered amino acid substitutions specific to SpaP A, B, and C localize on opposite sides of a pocket in the V-domain; A as reference, B and C share 29 substitutions (side chains in red), C holds 32 unique substitutions (side chains in yellow). C) SpaP structure with a tip-localized V-domain, intertwined P and A domains followed by C- and N-terminal complexes. Marked are all substitutions associated positively with caries and adhesion (red) or with caries alone (blue) in SpaP B isolates. D) Amino acid substitutions in SpaP B isolates (marked by a row) that coincide positively with caries and adhesion (to DMBT1 and saliva) or with caries alone upon PLS analyses. The substitutions are enriched in B-1 and B-2 isolates. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 4 3.7. Association of Cnm Isolates with DMBT1-Binding and Acid Tolerance {#s0110} ---------------------------------------------------------------------- Knowing that Cnm and Cbm isolates of serotypes *f* and *k* have been associated with systemic infection risk, we next explored all *S. mutans* isolates for adhesin (Cnm or Cbm) and serotype status and binding to collagen (Fig. S4, Table S3). Cnm and Cbm status, which coincided with collagen binding, was frequent with serotype *k* (100%) and *f* (50%) reported in systemic infections, but less common in serotype *e* (11%) and *c* (4%) isolates (Table S3). We next explored whether the caries association of Cnm phenotypes similarly could be referred to binding to DMBT1 and saliva and to acid tolerance ([Fig. 3](#f0015){ref-type="fig"}B--D). Of the 26 *cnm*-positive children, 12 had *S. mutans* isolates with a dominant Cnm-positive phenotype (Table S3), and these isolates bound markedly better to DMBT1, saliva and collagen than Cnm-negative isolates ([Fig. 3](#f0015){ref-type="fig"}B). Binding of Cnm to DMBT1 was indicated by a *cnm* knockout mutant (Fig. S5), and by the more avid binding of Cnm than SpaP isolates to saliva and DMBT1 ([Fig. 3](#f0015){ref-type="fig"}B). The binding of the isolates to DMBT1 and collagen correlated positively with individual numbers of caries lesions ([Fig. 3](#f0015){ref-type="fig"}C), and the isolates had unique Cnm sequences or truncated Cnm proteins with weak or no binding ([Fig. 3](#f0015){ref-type="fig"}D). In addition, Cnm isolates showed low acid tolerance in aerobic but high acid tolerance under anaerobic conditions (Fig. S6). 4. Discussion {#s0115} ============= The present findings provide the first evidence for specific adhesin types of *S. mutans* that match and predict individual caries development in people and thus provide a rationale for individualized oral care. The presence of SpaP B and Cnm subtypes coincided with more cross-sectional caries but also predicted increased 5-year caries increment in 452 Swedish children. In addition, binding of those strains to salivary DMBT1 correlated with individual caries scores. A lack of such evidence for a specific *S. mutans* caries pathogen has hampered its clinical use and favored the idea of caries resulting from an ecological shift toward various acid-tolerant species by plaque acidification from sugar consumption. However, because children not infected with *S. mutans* also developed caries, although to lesser extent, both life style and non-mutans bacteria may also contribute to caries development. We assume a heterogeneous etiology of caries in which both life style and host susceptibility influence the role of the microbiota and its components in caries development. Moreover, both SpaP and glycosyltransferases, two long standing vaccine candidates involving host-microbe and bacteria-bacteria adhesion events, respectively, may play important roles in caries development depending on *S. mutans* biotype, disease prevalence, and life style profile in different populations and individuals. Our present demonstration in *S. mutans* of specific biotypes A, B, and C with corresponding SpaP A, B, and C adhesin types, distinct housekeeping alleles and behaviors may in part explain previous difficulties in distinguishing *S. mutans* strains of low to high cariogenicity. The biotypes apparently differ in adhesion and in acid tolerance, and plausibly in as-yet-unknown cariogenic effects. However, similar to the *H. pylori* BABA and *Streptococcus parasanguinis* Fap1 adhesin structures ([@bb0030]; [@bb0210]), the SpaP protein structure may adapt to pH and together with housekeeping genes, such as superoxide dismutase and elongation factor Tu, directly participate in acid tolerance ([@bb0045]). Moreover, the different V-domains of SpaP A, B, and C may reflect an adaptation to host DMBT1 variants associated with different individuals and caries susceptibility ([@bb0095]) or microbial partner receptor configurations or epitopes, whereas differences in acid tolerance may match different biofilm conditions in various *Actinomyces*-*Streptococcus* communities. Of note, AgI/II adhesins participate in intra and inter generic interactions between *Actinomyces*-*Streptococcus* species ([@bb0170]). Biotypes A, B and C could accordingly represent members of distinct biofilm communities. The present findings provide a rationale for individualized oral care in terms of risk assessment and prevention. Infection with SpaP B (16% prevalence), B-1 (3.5%) or Cnm (6%) subtypes and absence of infection with *S. mutans* (52%) would identify reasonable portions of high- *versus* low-risk children, respectively. The fact that *S. mutans* infection *per se* and type, but not infection load in terms of gross numbers of the organism, predicted 5-year caries increment highlights the importance of plaque quality. The distinct adhesin and biotype behaviors could also identify subjects and caries subtypes that benefit from different treatments. The children infected with SpaP B and Cnm types may benefit from anti-microbial treatment with chlorhexidine or antibiotics and novel replacement with low-cariogenic types, such as A-1 isolates, and Cnm phenotypes may be inhibited by high oxygen conditions, improved oral hygiene, or other means. The high agreement between adhesin type and cross-sectional and prospective caries, although the children received caries prevention care and experienced puberty and adolescence as well as orthodontic treatment during the study, strengthens the adhesin-specific differences in cariogenicity. Moreover, the high colonization stability and persistence of the adhesin types in individual carriers over 5 years, and that children received multibrackets after 12 years of age, support this conclusion. However, it is difficult to estimate strain cariogenicity in absolute terms because of these experiences, along with the fact that treatment of caries with fillings reduces surfaces at future risk and that children differ in life style, socio-economic factors and genetic susceptibility. Moreover, the high *S. mutans* genetic diversity generated a repertoire of SpaP B-1 and B-2 subtypes of different cariogenic potentials and frequencies. The relative cariogenicity and natural history of the adhesin types should accordingly be replicated and further explored in longitudinal studies on initiation and progression of caries at different surfaces in the primary and permanent dentitions. Binding strength of SpaP B- and Cnm-positive isolates for DMBT1 correlated with individual caries scores, and the high cariogenicity B-1 isolates were enriched in substitutions that correlated with both adhesion and caries. The localization of these substitutions in the C and A domains suggests that SpaP polymorphisms outside the V-domain play a more important role in strain binding and cariogenicity than previously anticipated. Of interest, the collagen-binding affinity of the *S. aureus* CNA adhesin specifies strain differences in virulence in experimental arthritis ([@bb0255]). The Cnm and SpaP B adhesin subtypes, which both display increased saliva binding, may evoke different adhesion, aggregation, and innate immunity events by DMBT1 that favor surface-close colonization and tooth demineralization rather than overall plaque colonization of *S. mutans* ([@bb0145]; [@bb0075]). The mechanism underlying the correlation between binding affinity and individual caries scores may involve *S. mutans* adaptation to individual conditions or receptor modifications during life-long chronic caries infection and inflammation. We thus assume that variation in DMBT1-binding reflects a direct virulence mechanism and not a DMBT1-mediated pattern recognition of pathogenicity markers such as leucine-rich repeats ([@bb0035]). Accordingly, *S. mutans*, SpaP and Cnm/Cbm, as opposed to many other bacteria, bind to DMBT1 *via* non-leucine-rich repeats ([@bb0150]). The stable colonization of collagen binding Cnm or Cbm adhesin types of serotype *f* or *k* in \~ 1% children could lead to an increased risk for such diverse extra-oral infections as inflammatory bowel disease, hemorrhagic stroke and endocarditis. Of note, inflammatory bowel disease commonly debuts in adolescence and involves polymorphisms in the *S. mutans* receptor DMBT1. Moreover, although the prevalence of Cbm-positive children is too low for firm conclusions, the 7 children with Cbm-positive saliva had markedly low caries levels. This emphasizes that it remains to explore if extra-oral infections by *S. mutans* SpaP/Cnm/Cbm adhesins and serotype settings involve the same adhesin and biotype heterogeneity as observed in caries development. It also remains to be explored if expression of Cnm represses SpaP function and phenotype characteristics and if serotype *c* and *e* strains with Cnm expression are as virulent as their serotype *f* and *k* counterparts. Of note, Cnm isolates seemed to have both increased acid tolerance under anaerobic conditions and saliva/DMBT1 binding regardless of SpaP A, B or C background. The present study emphasizes that careful sequence analysis and evaluation of *S. mutans* genotypes in concert with dental caries incidence serves as a useful example and model to better understand strain variation and disease association in relation to bacterial virulence in chronic infections. These results highlight the importance of developing novel approaches to diagnose high-risk patients and improve prevention and treatment of chronic infectious disease. Our findings may also have relevance beyond dental caries and translate to improved systemic health. For example, missing teeth as a consequence of caries and periodontitis is a reported risk factor for cardio-vascular disease ([@bb0165]; [@bb0245]). Therefore, chronic caries infections and associated inflammation from carriage of high virulence strains would likely contribute to poor outcomes related to systemic disease risk as well as to diminished oral health. Appendix A. Supplementary data {#s0135} ============================== Supplementary materialImage 1 Acknowledgments {#s0120} =============== This work was supported by the Swedish Research Council (10906), the Faculty of Medicine of Umeå University, and Västerbotten County Council. We acknowledge financial support provided through a regional agreement between Umeå University and Västerbotten County Council in the field of Medicine, Odontology, and Health (377341) and through a Fund for Cutting-Edge Medical Research grant from the County Council of Västerbotten (135041). We also thank the Swedish dental Society and local foundations at Umeå University. The work was in part performed at Umeå Centre for Chronic Infectious Disease. Ingmarie Bernhardsson, Ewa Strömqvist-Engbo, Ulla Öhman and Prof. Lennart Hammarström, KI Stockholm, are acknowledged for assistance. We thank all families and public health care personnel who participated in the study. The funders had no role in study design; data collection, analysis or interpretation; writing of the report. A patent application has been filed; no. PCT/SE/2017/000028. Conflicts of Interest {#s0125} ===================== The authors have no competing interests to declare. Author Contributions {#s0130} ==================== N.S. designed research; A.E., N. Sheng performed most of the research; R.C., K.P. performed research; N.S., L.M., N. Sheng collected the clinical samples; A.E., N. Sheng, L.M., R.C., K.P.,T.B., N.S. analyzed data; N.S. wrote the paper. Supplementary data to this article can be found online at <https://doi.org/10.1016/j.ebiom.2017.09.027>. [^1]: Contributed equally to this work. [^2]: Children infected with *S. mutans* and specific adhesin types or being negative as determined by qPCR of whole saliva; data missing from one child. [^3]: Infected children are either spaP A, B or C and Cnm distribute in equal portions in A-, B- or C-positive children. [^4]: DeFS, Decayed, enamel included, Filled Surfaces at 12 and 17 years of age. [^5]: ΔDeFS-5y = 5 year caries increment from 12 to 17 years of age. [^6]: Two-sided *p*-value from Mann-Whitney *U* test compared to a reference (Ref.). [^7]: Cut off value of 10,000 cfu. [^8]: Colonization stability of *S. mutans* and adhesin types marks the same type at both 12 and 17 years of age at an individual level. [^9]: Children infected with *S. mutans* SpaP types and subtypes from sequencing and functional tests of isolates in 35 caries and 35 nearly caries-free extremes of children. [^10]: Caries DeFS (Decayed, enamel included, Filled Surfaces) at 12 and 17 years of age. [^11]: ΔDeFS (5 y) = 5-years prospective caries increment from 12 to 17 years of age. [^12]: 2-sided *p*-value from Mann-Whitney *U* test. [^13]: Highest (50%) and lowest (50%) adhesion phenotypes among the 70 isolates. [^14]: Low (0) and high (1, 2) acid tolerance.

Background {#Sec1} ========== Single-cell RNA-sequencing (scRNAseq) analysis plays an important role in investigating tumour evolution, and is more powerful to characterize the intra-tumor cellular heterogeneity \[[@CR1], [@CR2]\]. Compared with traditional RNA sequencing (i.e. bulk RNAseq) which measures the specific gene expression level within a cell population, scRNAseq quantifies the specific gene expression level within only an individual cell \[[@CR3], [@CR4]\]. scRNAseq is more likely to understand the detailed biological processes of cell developmental trajectories and cell-to-cell heterogeneity, providing us fresh insights into cell composition, dynamic cell states, and regulatory mechanisms \[[@CR5]--[@CR8]\]. However, there are still several big challenges we have to carefully deal with before analyzing scRNAseq data \[[@CR9], [@CR10]\]. The first challenge is that the scRNAseq data is easy to involve some unwanted variables \[[@CR11], [@CR12]\], e.g. batch effects, confounding factors, etc. Moreover, the scRNAseq data set has their own characterizes, such as gene expression matrix is extremely sparse because of the quite small number of mRNAs represented in each cell \[[@CR13]\]; current sequencing technologies, e.g. CEL-Seq2 \[[@CR14]\] and Drop-seq \[[@CR15]\], etc, do not have enough power to quantify the actual concentration of mRNAs (i.e. well-known "dropout events") \[[@CR16]\]; the heavy amplifications may result into strong amplification bias \[[@CR17]\]; cell cycle state, cell size or other unknown factors may contribute to cell-cell heterogeneity even within the same cell type \[[@CR18]\]. The second important feature of the scRNAseq data set is of count nature \[[@CR19]\]. In most RNA sequencing studies, the number of reads mapped to a given gene or isoform is often used as an intuitive estimate of its expression level. To account for the count nature of the RNA sequencing data, and the resulting mean-variance dependence, most statistical methods were developed using discrete distributions in differential expression analysis, i.e., PQLseq \[[@CR20]\], edgeR/DESeq \[[@CR21], [@CR22]\], and MACAU \[[@CR23]\]. Therefore, a nature choice of analyzing scRNAseq data is to develop count-based dimensionality reduction methods. Although several dimensionality reduction techniques have been already applied to scRNAseq data analysis, such as principal component analysis (PCA) \[[@CR24]\]; independent components analysis (ICA) \[[@CR25]\], and diffusion map \[[@CR26]\]; partial least squares (PLS) \[[@CR27], [@CR28]\]; nonnegative matrix factorization (or factor analysis) \[[@CR29], [@CR30]\], gene expression levels are inherently quantified by counts, i.e., count nature of scRNAseq data \[[@CR31], [@CR32]\]. Therefore, developing the bespoke scRNAseq dimensionality reduction method has been triggered within the last two years. The first factor analysis method, ZIFA, is trying to model the drop-out events via the zero-inflated model, but the method does not take into account the count nature of the data \[[@CR33]\]; pCMF is trying to build sparse Gamma-Poisson factor model within the Bayesian framework, but such method does not include the covariates \[[@CR34]\]; ZINB-WaVE is trying to involve both gene-level and sample-level covariates via a hierarchical model, but the method is really time-consuming when sample size is large \[[@CR35], [@CR36]\]. Here, in this paper, we propose a fast and efficient count-based matrix factorization method that utilizes the negative binomial distribution to account for the over-dispersion problem of the count nature of scRNAseq data, single-cell Negative Binomial-based Matrix Factorization, scNBMF. The reason of choosing negative binomial model instead of zero-inflated negative binomial model is that not only the most scRNAseq data sets do not show much technical contribution to zero-inflation (Fig. [1](#Fig1){ref-type="fig"}a), but also can largely reduce the computation burden in estimating drop-out parameters for each gene. With the stochastic optimization method Adam \[[@CR37]\] implemented within TensorFlow framework, scNBMF is roughly 10 -- 100 times faster than the existing count-based matrix factorization methods, such as pCMF and ZINB-WaVE. To make the proposed method scalable, we apply scNBMF to analyze three publicly available scRNAseq datasets. The results demonstrate that scNBMF is more efficient and powerful than other matrix factorization methods. Fig. 1A simple example to show the parameter effect or optimizer effect of NMI and ARI in scRNA-seq data on clustering. **a** This figure shows the relationship between mean gene expression levels and dropout rates. The black line indicates observed value, which is computed by the number of unexpressed cells divided by the number of cells; The red line represents expected value, which is calculated by negative binomial distribution with mean gene expression levels and dispersion parameter *ψ*(*ψ*=*mean*(*ψ*~*i*~))**b** This figure shows how optimizers affect the performance of different methods on NMI and ARI. **c**-**d** These two figure indicate how the number of factors affect the NMI and ARI, respectively Materials and methods {#Sec2} ===================== scNBMF: model and algorithm {#Sec3} --------------------------- scNBMF is to fit the logarithm likelihood function of negative binomial model-based matrix factorization. Given *n* cells and *p* genes, we denote *Y* as a gene expression matrix, and its element *y*~*ij*~ is the count of gene *i* and cell *j*. To account for the over-dispersion problem, we model the gene expression level *y*~*ij*~ as a random variable following the negative binomial distribution with parameters *μ*~*ij*~ and *ϕ*~*i*~, i.e., $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$y_{ij} \sim NB(\mu_{ij},\phi_{i}) $$ \end{document}$$ where the rate parameter *μ*~*ij*~ denotes the mean expression level for gene *i* and cell *j*; the parameter *ϕ*~*i*~ represents variance of gene expression, typically means gene-specific over-dispersion; *NB* is the negative binomial distribution, i.e. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$ \text{Pr}_{NB}(y_{ij}|\mu_{ij},\phi_{i}) \,=\, \left(\! \begin{array}{c} y_{ij} + \phi_{i} - 1\\ y_{ij} \end{array} \!\!\right) \!\!\left(\frac{\mu_{ij}} {\mu_{ij} + \phi_{i} } \!\right)^{y_{ij}}\!\! \left(\frac{ \phi_{i}} {\mu_{ij} + \phi_{i}} \!\right)^{\phi_{i}}\!. $$ \end{document}$$ For the rate parameter *μ*~*ij*~, we consider the following regression model $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$log(\mu_{ij}) = log(N_{j}) + {\sum}_{k = 1}^{K} W_{ik} X_{kj}. $$ \end{document}$$ where *N*~*j*~ is the total read count for the individual cell *j* (a.k.a read depth or coverage); *W*~*ik*~ is the loadings while *H*~*kj*~ is the factors represents the coordinates of the cells, which can be used to identify cell type purpose; *K* is the pre-defined number of components; When all *ϕ*~*i*~→0, the negative binomial distribution will reduce to the standard Poisson distribution. Therefore, the log-likelihood function for gene *i* and cell *j* is $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$\begin{array}{*{20}l} \mathcal{L}_{NB}(\mu,\phi| Y) = &\sum\limits_{i = 1}^{p} \sum\limits_{j = 1}^{n} log \text{Pr}_{NB} \left(y_{ij}|\mu_{ij},\phi_{i}\right)\\ = &\sum\limits_{i = 1}^{p} \sum\limits_{j = 1}^{n} y_{ij} log (\mu_{ij}) + \phi_{i} log(\phi_{i}) \\&- (y_{ij} + \phi_{i}) log (\mu_{ij}+\phi_{i}) \\ & + log\left(\begin{array}{c} y_{ij} + {\phi_{i}} - 1\\ y_{ij} \end{array} \right). \end{array} $$ \end{document}$$ where *μ* denotes the mean gene expression matrix and its element $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mu _{ij}=e^{log(N_{j}) + {\sum }_{k = 1}^{K} W_{ik} X_{kj} }$\end{document}$; *ϕ* is a *p*-vector, and its element *ϕ*~*i*~ represents the over-dispersion parameter for gene *i*. To make our model more interpretation for the biological applications, we introduce a sparse penalty (LASSO) on loading matrix *W* since some genes are expressed while some are not in real-world biological processes. Therefore, the objective function of optimization problem becomes $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$\mathcal{L} = \mathcal{L}_{NB}(\mu,\phi|Y) + \lambda \sum\limits_{i = 1}^{p} \left\| W_{i} \right\|_{1} $$ \end{document}$$ where ∥·∥~1~ is a *l*~1~-norm (i.e. LASSO penalty); *λ* denotes the penalty parameter. In the above model, we are interested in extracting the factor matrix *H* for detecting the cell type purposes. We first estimate the dispersion parameter *ϕ*~*i*~) for each gene via edgeR \[[@CR21]\] with default parameter settings, then fit the above model using Adam optimizer within TensorFlow. For deep learning model, we set the learning rate of the network as 0.001 and maximum iteration as 18000. Compared methods and evaluations {#Sec4} -------------------------------- To make scNBMF scalable, we compared seven existing methods, i.e. PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF, and ZINB-WaVE, in the experiments. Since PCA and ZIFA are only for normalized gene expression data, we normalized raw count data following previous recommendations \[[@CR38]\]. Typically, we transformed the count data using base 2 and pseudo count 1.0, i.e., *log*~2~(*Y*+1.0), into continuous data. The performance of each method was evaluated by the normalized mutual information (NMI), defined in \[[@CR39]\] $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$ NMI(L_{e}, L) = \frac{\sum\limits_{k = 1}^{K}\sum\limits_{t = 1}^{K_{e}} \frac{n_{kt}}{n}log \left(\frac{n_{kt}}{n} \right) - \sum\limits_{k = 1}^{K}\frac{n_{k}}{n}log \left(\frac{n_{k}}{n} \right)- \sum\limits_{t = 1}^{K_{e}}\frac{n_{t}}{n}log \left(\frac{n_{t}}{n} \right)} {\sqrt{\sum\limits_{k = 1}^{K}\frac{n_{k}}{n}log \left(\frac{n_{k}}{n} \right)* \sum\limits_{t = 1}^{K_{e}}\frac{n_{t}}{n}log \left(\frac{n_{t}}{n}\right)} }. $$ \end{document}$$ and the adjusted rand index (ARI), defined in \[[@CR40]\] $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$ ARI(L_{e}, L) = \frac{{{\sum}_{kt} {\left(\begin{array}{l} n_{kt}\\ 2 \end{array} \right) - \left({{\sum}_{k} {\left(\begin{array}{l} n_{k}\\ 2 \end{array} \right) {\sum}_{t} {\left(\begin{array}{l} n_{t}\\ 2 \end{array} \right)}} } \right)/\left(\begin{array}{l} n\\ 2 \end{array} \right)}}}{{\frac{1}{2}\left({{\sum}_{k} {\left(\begin{array}{l} n_{k}\\ 2 \end{array} \right) + {\sum}_{t} {\left(\begin{array}{l} n_{t}\\ 2 \end{array} \right)}} } \right) - \left({{\sum}_{k} {\left(\begin{array}{l} n_{k}\\ 2 \end{array} \right){\sum}_{t} {\left(\begin{array}{l} n_{t}\\ 2 \end{array} \right)}} } \right)/\left(\begin{array}{l} n\\ 2 \end{array} \right)}}. $$ \end{document}$$ where *L*~*e*~ and *L* are the predicted cluster labels and the true labels, respectively; *K*~*e*~ and *K* are the predicted cluster number and the true cluster number, respectively; *n*~*k*~ denotes the number of cells assigned to a specific cluster *k* (*k*=1,2,⋯,*K*); similarly *n*~*t*~ denotes the number of cells assigned to cluster *t* (*t*=1,2,⋯,*K*~*e*~); *n*~*kt*~ represents the number of cells shared between cluster *k* and *t*; and *n* is the total number of cells. Public scRNAseq data sets {#Sec5} ------------------------- Three publicly available scRNAseq data sets were collected from three studies: The first scRNAseq data set was collected from human brain \[[@CR41]\]. There are 420 cells in eight cell types after excluded hybrid cells including, fetal quiescent cells (110 cells), fetal replicating cells (25 cells), astrocytes cells (62 cells), neuron cells (131 cells), endothelial (20 cells) and oligodendrocyte cells (38 cells) microglia cells(16 cells), and (OPCs, 16 cells), and remain 16,619 genes to test after filtering out the lowly expressed genes. The original data was downloaded from the data repository Gene Expression Omnibus (GEO; GSE67835);The second scRNAseq data set was collected from human pancreatic islet \[[@CR42]\]. There are 60 cells in six cell types after excluding undefined cells including alpha cells (18 cells), delta cells (2 cells), pp cells (9 cells), duct cells (8 cells), beta cells (12 cells) and acinar cells (11 cells),and 116,414 genes to test after filtering out the lowly expressed genes. The original data was downloaded from the data repository Gene Expression Omnibus (GEO; GSE73727);The third scRNAseq data set was collected from the human embryonic stem \[[@CR43]\]. There are 1018 cells which belong to seven known cell subpopulations that include neuronal progenitor cells (NPCs, 173 cells), definitive endoderm derivative cells (DEDs), endothelial cells (ECs, 105 cells), trophoblast-like cells (TBs, 69 cells), undifferentiated H1(212 cells) and H9(162 cells) ESCs, and fore-skin fibroblasts (HFFs, 159 cells), and contains 17,027 genes to test after filtering step. The original data was downloaded from the data repository Gene Expression Omnibus (GEO; GSE75748). Results {#Sec6} ======= Model selection {#Sec7} --------------- Our first set of experiments is to select the optimization method for the log-likelihood function of negative binomial matrix factorization model. Without loss of generality, we choose the human brain scRNAseq data set. Five optimization methods were compared to optimize the neural networks, i.e., Adam, gradient descent, Adagrad, Momentum and Ftrl. The results show that the Adam significantly outperforms other optimization methods regardless of what criteria we choose (Fig. [1](#Fig1){ref-type="fig"}b). Specifically, for NMI, Adam, gradient descent, Adagrad, Momentum, and Ftrl achieve 0.8579, 0.0341, 0.0348, 0.4859, and 0.1251, respectively. Therefore, in the following experiments, we will choose the Adam method to optimize the neural networks. Our second set of experiments is to select the number of factors in the low dimensional structure of cell types. Without loss of generality, we still choose the human brain scRNAseq data set. We varied the number of factors (*k* = 4, 6, 10, 15, and 20). The results demonstrate that the number of factors does not impact PCA (Fig. [1](#Fig1){ref-type="fig"}c and d; bule line). The other four methods show an increasing pattern when the number of factors varied from 4 to 20 (Fig. [1](#Fig1){ref-type="fig"}c and d). Therefore, we choose the top 20 factors in the following experiments. Public scRNAseq data sets {#Sec8} ------------------------- Our third set of experiments is to apply scNBMF to three scRNAseq real data sets, human brain, human pancreas islet, and human embryonic stem. The cell type information of the three data sets were reported by the original studies. For the comparison, we compared seven other methods, PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF and ZINB-WaVE. For the evaluation, we extracted the low dimensional structure with top 10 factors, and used *k*-means clustering method in an unsupervised manner, repeated 100 times to test how well each method can recover the cell type assignments on NMI and ARI in the studies. The first biological data application is performed on the human brain scRNAseq data set. Figure [2](#Fig2){ref-type="fig"} demonstrates the comparison results of tSNE with respect to seven compared clustering methods. scNBMF shows the clearly cell type patterns with the annotated cell type (Fig. [1](#Fig1){ref-type="fig"}h). Also, we carried out the same analysis using PCA (Fig. [2](#Fig2){ref-type="fig"}a), Nimfa (Fig. [2](#Fig2){ref-type="fig"}b), NMFEM (Fig. [2](#Fig2){ref-type="fig"}c), tSNE (Fig. [2](#Fig2){ref-type="fig"}d), ZIFA (Fig. [2](#Fig2){ref-type="fig"}e), pCMF (Fig. [2](#Fig2){ref-type="fig"}f), and ZINB-WaVE (Fig. [2](#Fig2){ref-type="fig"}g). For NMI and ARI, scNBMF outperforms the other methods. Specifically, for NMI criterion, PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF, ZINB-WaVE and scNBMF achieve, 0.582, 0.494, 0.456, 0.712, 0.797, 0.787, 0.892, and 0.901, respectively (Fig. [2](#Fig2){ref-type="fig"}i and Table [1](#Tab1){ref-type="table"}); while for ARI criterion, PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF, ZINB-WaVE and scNBMF achieve, 0.339, 0.258, 0.264, 0.544, 0.721, 0.788, 0.916, and 0.933, respectively (Fig. [2](#Fig2){ref-type="fig"}i and Table [1](#Tab1){ref-type="table"}). Fig. 2Performance evaluation on human brain scRNA-seq data. In this data set there are 420 cells in eight different cell types after the exclusion of hybrid cells. Each kind of color represent a kind of cell type. **a**-**h** These eight figures display the clustering output of two dimension of tSNE using eight matrix factorization methods(PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF, ZINB-WaVE, and scNBMF). **f** This figure shows NMI and ARI values which are from eight compared methodsTable 1Clustering comparison of the matrix factorization-based methods in terms of Normalized Mutual information (NMI) and Adjusted Random Index (ARI)MethodBrainEmbryoPancreasNMIARINMIARINMIARIPCA0.5820.3390.3660.1870.6300.368Nimfa0.4940.2580.4140.1730.4560.114NMFEM0.4560.2640.7410.6140.4350.175tSNE0.7120.5440.6580.538**0.7930.652**ZIFA0.7970.7210.8880.7480.6410.429pCMF0.7870.7880.8220.6590.5470.334ZINB-WaVE0.8920.9160.8880.7210.5180.342scNBMF**0.9010.9330.9080.763**0.7160.472The number with bold indicates the best performance method and the number with grey represents the second best performance method The second biological data application is to investigate the character of human pancreas islet scRNAseq data set. This data set has a smaller number of cells - only 60 cells in six cell types. Since all methods do not have enough power to detect the cell type clustering patterns, we did not show the tSNE plots for this data set. For NMI and ARI, tSNE shows the highest performance, while scNBMF achieves the second best performance (Table [1](#Tab1){ref-type="table"}). Specifically, tSNE achieves 0.973 and 0.652 on NMI and ARI, respectively; while scNBMF is 0.716 and 0.472 on NMI and ARI respectively. The third biological data application is to investigate lineage-specific transcriptomic features at single-cell resolution. To elucidate the distinctions between different lineages, we performed eight matrix factorization methods, i.e., PCA (Fig. [3](#Fig3){ref-type="fig"}a), Nimfa (Fig. [3](#Fig3){ref-type="fig"}b), NMFEM (Fig. [3](#Fig3){ref-type="fig"}c), tSNE (Fig. [3](#Fig3){ref-type="fig"}d), ZIFA (Fig. [3](#Fig3){ref-type="fig"}e), pCMF (Fig. [3](#Fig3){ref-type="fig"}f), ZINB-WaVE (Fig. [3](#Fig3){ref-type="fig"}g), and scNBMF (Fig. [3](#Fig3){ref-type="fig"}h). scNBMF demonstrates more clearly their respective cell-type patterns compared with other methods. The cell type H1 and H9 show the tight overlapping pattern to indicate the relative homogeneity of human ES cells, such results are also consistence with the previous results \[[@CR43]\]. For NMI and ARI, scNBMF outperforms other methods (Fig. [3](#Fig3){ref-type="fig"}i and Table [1](#Tab1){ref-type="table"}). Specifically, for NMI, PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF, ZINB-WaVE and scNBMF achieve, 0.366, 0.414, 0.741, 0.658, 0.888, 0.822, 0.888, and 0.908, respectively; For ARI, PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF, ZINB-WaVE and scNBMF achieve, 0.187, 0.173, 0.614, 0.538, 0.748, 0.659, 0.721, and 0.763, respectively. Fig. 3Performance evaluation on human embryonic stem scRNA-seq data set, which contains 1018 cells in seven cell types. Different colors also represent different cell types. **a**-**h** These five figure display the clustering output of two dimension of tSNE using five matrix factorization methods(PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF, ZINB-WaVE, and scNBMF). **f** This figure shows NMI and ARI values which are from eight compared methods Computation time {#Sec9} ---------------- The last set of experiments is to compare the computation time of PCA, Nimfa, NMFEM, tSNE, ZIFA, pCMF, and ZINB-WaVE. Without loss of generality, we use human brain data set to show the computation time of the compared methods (Table [2](#Tab2){ref-type="table"}). Nimfa, NMFEM, ZIFA, pCMF, and ZINB-WaVE are the bespoke scRNAseq methods. Compared with the count-based methods, ZINB-WaVE and pCMF, scNBMF is roughly 100 folds faster than ZINB-WaVE, and 10 folds faster than pCMF. Even comparing the non-count based methods, ZIFA, Nimfa, and NMFEM, scNBMF is still the fastest method. Table 2Computation times (second) of the matrix factorization-based methods on human brain scRNAseq data set, *k* represents the number of factorsMethod*k*=2*k*=4*k*=6*k*=10*k*=15*k*=20PCA11.5411.5511.7011.3511.3711.59Nimfa639.151990.662260.132490.052705.422924.87NMFEM1471.391628.21913.112248.182659.233027.5tSNE1.8514.4132.1156.0177.20101.25ZIFA5331.255831.046347.086987.527338.267722.33pCMF12391.613517.1214260.2615111.5515978.4417158.42ZINB-WaVE71053.179402.1790118.3101072.9115379.7126575.2scNBMF456.12478.90541.31717.881053.221563.75 Conclusion {#Sec10} ========== With rapid developing sequencing technology, a large amount of scRNAseq data sets is easily obtained via different sources. Therefore, computation time is one of these big issues for downstream analysis. On the other hand, scRNAseq data have their own characterizes, i.e., count nature, noisy, and sparsity, etc. These have been triggered the development of a fast and efficient count-based matrix factorization method. In this paper, we proposed a count-based matrix factorization (scNBMF) method to model the raw count data, prevent losing information from normalizing raw count data. On three public biological scRNAseq data sets, scNBMF provides powerful performance compared with other seven methods in terms of NMI, ARI, and computation time. Zero-inflated distribution is more appropriate method to account for dropouts, e.g. ZIFA and ZINB-WaVE. In current study, we did not consider the zero-inflated model because the tested data sets do not show too much dropouts. However, this is a necessary step in analyzing some scRNAseq data sets. Therefore, we will add the zero-inflated distribution in the future version of the scNBMF. Biologically, if we incorporate all genes in scRNAseq data analysis, probably it would be able to involve some unwanted variables because not all genes are expressed in biological processes. An interesting direction to improve the performance of scNBMF is to select some informative genes first, this step can largely reduce unwanted variables, and exclude some redundancy genes \[[@CR44], [@CR45]\] in the downstream analysis. In addition, because gene expression levels are highly affected by other gene specific annotations, such as GC-content, gene length, and chromatin states \[[@CR46]\]. If some interesting variables in the statistical model, such as "drop-out" parameter, can be inferred by annotation information, the method probably will significantly improve the power of detecting cell types from scRNAseq data. ARI : Adjusted rand index DESeq : Differential expression edgeR : Empirical analysis of digital gene expression data in R ICA : Independent components analysis MACAU : mixed model association for count data via data augmentation NMI : Normalized mutual information PCA : Principal component analysis pCMF : Probabilistic count matrix factorization PLS : Partial least squares PQLseq : Penalized quasi-likelihood scNBMF : Single-cell negative binomial matrix factorization scRNAseq : Single-cell RNA sequencing tSNE : t-distributed stochastic neighbor embedding ZIFA : Zero-inflated factor analysis ZINB-WaVE : Zero-inflated negative binomial-based wanted variation extraction No applicable. Funding {#d29e2761} ======= Publication of this artical was sponsored by the Top International University Visiting Program for Outstanding Young scholars of Northwestern Polytechnical University; Fundamental Research Funds for the Central Universities (Grant: 3102017OQD098); Natural Science Foundation of China (NFSC; Grants: 61772426 and 61332014). Availability of data and materials {#d29e2766} ================================== scNBMF was implemented by R and Python, and the source code are freely available at <https://github.com/sqsun>. The three publicly scRNAseq datasets are available at <https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE67835><https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE73727><https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE75748> About this supplement {#d29e2788} ===================== This article has been published as part of *BMC Systems Biology Volume 13 Supplement 2, 2019: Selected articles from the 17th Asia Pacific Bioinformatics Conference (APBC 2019): systems biology*. The full contents of the supplement are available online at <https://bmcsystbiol.biomedcentral.com/articles/supplements/volume-13-supplement-2>. SS, YC, YL and XS conceived and wrote the manuscript. SS and YC implemented the software and analyzed the data. All authors read and approved the final manuscript. Ethics approval and consent to participate ========================================== No applicable. Consent for publication ======================= No applicable. Competing interests =================== The authors declare that they have no competing interests. Publisher's Note ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Competing interest statement ============================ Conflict of interest: the authors declare no potential conflict of interest. Introduction {#sec1-1} ============ As dramatic progress has been made in the therapeutic strategies for multiple myeloma (MM), in particular the development of proteasome inhibitors (PIs) and immunomodulatory drugs (IMiDs), extramedullary disease (EMD) is observed with increasing frequency.^[@ref1]^ To date, visceral organ involvement of spleen, liver, kidneys, lymph nodes, and cutaneous tissue has been reported.^[@ref2]^ The prognosis of MM patients with extramedullary relapse is very poor.^[@ref3]^ EMD is often associated with high serum lactate dehydrogenase levels,^[@ref4]^ plasmablastic cell morphology, and complex cytogenetic abnormality.^[@ref5]^ In the largest retrospective analysis of relapse patterns following autologous hematopoietic stem cell transplantation for MM, Alegre *et al*. reported that increased EMD was evident both at diagnosis and after therapy.^[@ref6]^ Also they found that the risk of extramedullary relapse was not significantly increased after the use of PIs and IMiDs.^[@ref6]^ However, it has been suggested that PIs and IMiDs may increase the incidence of EMD by attenuating the biology of MM.^[@ref7]^ The mechanism of EMD development is elusive, but it is suggested that oncogenes,^[@ref10],[@ref11]^ instability of chromosomes,^[@ref12]^ cell adhesion molecules,^[@ref7],[@ref8],[@ref13]^ and tumor microenvironment^[@ref14]^ may have important role. Here, we report a surprisingly large tumor mass occupying the left upper limb in the course of MM treatment including PIs and IMiDs. EMD of this large size has not been reported before, and this case may provide a clue to understand the features of EMD in MM. Case Report {#sec1-2} =========== A 66-year-old woman developed pain in her left shoulder and was diagnosed with pathological fracture. Her serum IgG was elevated to 3811 mg/dL and IgG-lambda type M-protein was detected by serum immunoelectrophoresis assay. She had 19.6% of bone marrow plasmacytosis with normal cytogenetics. The diagnosis of ISS stage 1 IgG-lambda type MM was established. No other bone lesion, anemia or kidney injury was found. Her left upper extremity was treated with 8 Gy single fraction using a 4MV photon beam by parallel opposed portals, prior to the conventional vincristine/adriamycin/dexamethasone (VAD) induction therapy. After the three courses of VAD treatment, she was switched to a bortezomib/dexamethasone (BD) regimen because of a Helicobacter cinaedi bacteremia developed in the third course of VAD. After the completion of 3 courses of BD regimen, her bone marrow plasmacytes was decreased to 0.4%, however, a tumor of 1×1 cm large developed in her left arm. With a clinical diagnosis of plasmacytoma, second course of radiotherapy with 8 Gy irradiation (left upper extremity was treated with 8 Gy single fraction using a 4MV photon beam by parallel opposed portals), was performed, after which the patient underwent lenalidomide/dexamethasone therapy. Lenalidomide/dexamethasone was effective, and her EM nodule decreased in size to visually undetectable level; however, 6 months later, after 4 courses of Lenalidomide/dexamethasone treatment the nodule enlarged again. There were 15×15 mm tumor in flexor side of her left arm and 20×20 mm tumor in her extensor side of forearm. MRI revealed those tumors were not connected to cortical bones, in addition, no other tumors in her left arm were found. A needle biopsy of a tumor in her left arm was performed and an accumulation of atypical plasmacytes were detected. As a third line therapy, 2 courses of bortezomib/cyclophosphamide/dexamethasone regimen, followed by melphalan/thalidomide/prednisolone (MPT) was administered, however, during the 6th course of MPT, the extramedullary plasmacytoma occupied her left upper limb. Then, pomalidomide/dexamethasone as a fourth line was started, which was initially effective; the size of the tumor mass decreased, and the vessels on the surface of the bulk of the tumor appeared to be reduced. However, in the 3rd course of pomalidomide treatment, the EMD enlarged again and extended to her left forearm and back of the hand ([Figure. 1](#fig001){ref-type="fig"}). On the contrary, there were only scarce MM cells in the bone marrow (3.6%). Those MM cells were morphologically plasmablastic and harbored complex cytogenetic abnormality. She died of severe respiratory failure. Pleural and pulmonary tumor infiltration was suspected. Post-mortem examination revealed extensive MM involvement of multiple organs, including not only the left upper limb, but also lung, liver, kidney, stomach, and thyroid ([Figure 2](#fig002){ref-type="fig"}). However, the bone marrow had only scattered patchy myeloma cells. Discussion {#sec1-3} ========== We demonstrate a remarkably large plasmacytoma developing in the left arm of a patient with refractory MM. The presence of this large extramedullary plasmacytoma may be rare; so far, no cases of EMD of a similar size have been reported. Consistent with prior reports, the patient had plasmablastic MM cells, complex karyotype, elevated LDH, and presented a very aggressive course. Although MM cells were seen to invade multiple organs, including lung, liver, and kidney, there was no evidence of plasma cell leukemia, and only 3.6% bone marrow plasmacyotosis was detected. This dissociation between bone marrow and peripheral blood findings and aggressive visceral organ invasion may be significant. The difference between marrow and visceral organ involvement has been reported. For example, in the largest retrospective analysis of relapse patterns following autologous hematopoietic stem cell transplantation for MM, Alegre *et al*. found that 14% of cases that relapsed were EMD and were not associated with BM infiltration.^[@ref6]^ Also this dissociation may indicate the heterogeneity of MM; for a long time, it has been recognized that not all tumor cells are identical.^[@ref15]^ For example, tumors are heterogeneous in their proliferative ability and only a subset of tumor cells have long-term renewal potential.^[@ref15]^ In addition, it has been noticed that clonal evolution, in which tumor cells accumulate mutations, some of which confer increased fitness and survival advantage can occur.^[@ref16]^ Furthermore this dissociation may imply there are biological difference between myeloma homing in bone marrow and myeloma forming EMD. Regarding the biology of EMD formation, genetic abnormalities^[@ref10],[@ref11],[@ref12]^ decreased cell adhesion molecules expression, down regulation of chemokine receptors,^[@ref7],[@ref8],[@ref13]^ and tumor microenvironment^[@ref14]^ may have important role. In the courses of the pomalidomide treatment, the vessel pattern in the bulk of the tumor had reduced. This is compatible with the widely recognized phenomenon that IMiDs antagonize tumor angiogenesis. Conclusions {#sec1-4} =========== We discussed an extremely large EMD developing in a patient with refractory MM. MM cells extensively invaded multiple organs without significant bone marrow infiltration. This case may provide clues for a better understanding of the EMD. {#fig001} {#fig002} [^1]: Contributions: JT and SH committed the clinical practice and wrote the paper. KO and TI evaluated the post-mortem examination. DN and KN contributed essential advice regarding clinical practice and writing the article. SH and KN supervised the study.

1. Introduction {#sec1-ijms-20-01218} =============== Mitochondrial transcription factor A (mtTFA, mtTF1, TFAM) is critical in regulating mitochondrial DNA (mtDNA) transcription, packaging and copy number \[[@B1-ijms-20-01218],[@B2-ijms-20-01218],[@B3-ijms-20-01218]\]. It was pointed out that suppressing the expression of TFAM inhibited the proliferation of cancer cells \[[@B4-ijms-20-01218],[@B5-ijms-20-01218]\], increased the sensitivity of cancer cells to chemotherapeutic drugs or ionizing irradiation \[[@B6-ijms-20-01218],[@B7-ijms-20-01218],[@B8-ijms-20-01218]\], and triggered apoptosis \[[@B9-ijms-20-01218]\]. The increased expression of TFAM can be considered as a prognosis marker for poor clinical outcomes of specific type of cancer \[[@B10-ijms-20-01218]\]. At the same time, the decreased expression of TFAM and mitochondrial dysfunction was associated with the pathogenesis of brain and bone diseases \[[@B11-ijms-20-01218],[@B12-ijms-20-01218]\]. TFAM can be up-regulated by ionizing radiation in cancer cell lines, to decrease radiation induced cell death \[[@B13-ijms-20-01218]\]. However, the mechanisms on how TFAM is regulated in irradiated tumor cells remain largely known. Cyclooxygenase-2 (COX-2), encoded by the *PTGS2* gene, plays important roles in tumorigenesis and inflammation \[[@B14-ijms-20-01218],[@B15-ijms-20-01218],[@B16-ijms-20-01218]\]. The increased expression of COX-2 is considered as a marker for the proliferation of tumor cells \[[@B17-ijms-20-01218]\]. COX-2 plays a critical role in the production of prostaglandin E2 (PGE2). Previous studies showed that COX-2-derived PGE2 induced Id1-dependent radiation resistance and self-renewal in experimental glioblastoma \[[@B18-ijms-20-01218]\]. Other studies have confirmed that the inhibition of COX-2 expression increases the sensitivity of cancer cells to radiation, and COX-2 signaling is a potential therapeutic target for consolidating cancer treatment \[[@B19-ijms-20-01218],[@B20-ijms-20-01218],[@B21-ijms-20-01218]\]. It was reported that a majority of COX-2 in tumor cells were co-localized with heat shock protein-60 in mitochondria, and the mitochondrial localization of COX-2 might confer resistance to apoptosis in different cancer cell lines \[[@B22-ijms-20-01218]\]. Dynamin-related protein 1 (DRP1), a key mediator of mitochondrial fragmentation, is encoded by the *DNM1L* gene \[[@B23-ijms-20-01218]\]. Recent studies have shown that radiation-induced the localization of DRP1 to the mitochondria, and accelerated mitochondrial fragmentation \[[@B24-ijms-20-01218]\]. Preventing mitochondrial fragmentation impaired mitochondrial functions, and led to the loss of mitochondrial DNA \[[@B25-ijms-20-01218]\], indicating that the potential association between mitochondrial morphologies and TFAM was involved in the regulation of mitochondrial biogenesis \[[@B3-ijms-20-01218],[@B26-ijms-20-01218],[@B27-ijms-20-01218]\]. Both TFAM and COX-2 contribute to the resistance of cancer cells to radiation, and they are considered as potential targets for improving the efficacy of radiation treatment in cancers. Besides, they are mitochondrial proteins, and affect mitochondrial functions. Therefore, in this research, we aimed at exploring the interconnections between TFAM and COX-2 in irradiated cancer cells. We identified that COX-2 derived PGE2 enhanced the activation of p38-MAPK, which further stimulated DRP1-mediated up-regulation of TFAM. Our results provided new information on the mechanisms for how COX-2 affects mitochondrial functions, and its implications in increasing the sensitivity of cancer cells to radiation during therapy. The results are described in the following section. 2. Results {#sec2-ijms-20-01218} ========== 2.1. Concomitant Up-Regulation of TFAM and COX-2 in Irradiated Tumor Cells {#sec2dot1-ijms-20-01218} -------------------------------------------------------------------------- TFAM-knockdown U-2 OS and Hep G2 cells were established by transfecting short hairpin RNA (shRNA) plasmids targeting human *TFAM* ([Figure 1](#ijms-20-01218-f001){ref-type="fig"}A). In TFAM knockdown cells, radiation induced elevation of mtDNA copy number was suppressed ([Figure 1](#ijms-20-01218-f001){ref-type="fig"}B). Clonogenic survival assay was applied to test the role of TFAM in sensitizing tumor cells to γ-ray irradiation. As shown in [Figure 1](#ijms-20-01218-f001){ref-type="fig"}C, plots were fitted according to the linear quadratic model, S = exp (−α × *D* − β × *D*^2^), where S is the surviving fraction, *D* is the radiation dose (Gy), and α and β are the fitting parameters. According to the surviving fraction curves, for U-2 OS cells transfected with scramble shRNA plasmid, the 10% survival dose (*D*~10~) and 37% survival dose (*D*~0~) were 5.32 Gy and 2.7 Gy. The *D*~10~ and *D*~0~ for U-2 OS cells transfected with shTFAM1 and shTFAM2 plasmids were 3.64 Gy, 1.54 Gy and 4.18 Gy, 1.85 Gy, respectively. For Hep G2 cells transfected with scramble shRNA plasmid, the 10% survival dose (*D*~10~) and 37% survival dose (*D*~0~) were 5.62 Gy and 2.9 Gy, respectively. The *D*~10~ and *D*~0~ for Hep G2 cells transfected with shTFAM1 and shTFAM2 plasmids were 4.38 Gy, 1.83 Gy, and 3.49 Gy, 1.44 Gy respectively. These results indicated that the knockdown of TFAM expression increased the sensitivity of U-2 OS and Hep G2 cells to radiation. Since COX-2 has been reported to be a pro-survival protein in a wide range of tumor cells, we next detected the expression levels of TFAM and COX-2 after γ-ray irradiation. As shown in [Figure 1](#ijms-20-01218-f001){ref-type="fig"}D, 12 h post-4 Gy γ-ray radiation, the expression levels of both TFAM and COX-2 in U-2 OS, HeLa, and MCF7 cells displayed notable enhancement. We then checked the radiation dose dependency of TFAM and COX-2 expression in U-2 OS cells, 12 h post radiation. As shown in [Figure 1](#ijms-20-01218-f001){ref-type="fig"}E, the expression levels of TFAM and COX-2 were enhanced, with the dose increasing from 1 Gy to 8 Gy, to around 3- and 3.5-fold, compared to those observed in the non-irradiated cells. Next, the time-course for the expression of TFAM and COX-2 were investigated in U-2 OS cells ([Figure 1](#ijms-20-01218-f001){ref-type="fig"}F). The expression levels of TFAM and COX-2 increased within 12 h after radiation, reaching around 2.8- and 3-fold, compared to the non-irradiated controls respectively. Also, due to TFAM up-regulation decreasing the sensitivity of tumor cells to radiation, and because it was concomitantly up-regulated with COX-2, we then aimed at determining whether COX-2 affected the expression of TFAM in irradiated cells. 2.2. Activation of COX-2 Up-Regulates TFAM in Irradiated Cells {#sec2dot2-ijms-20-01218} -------------------------------------------------------------- To test whether COX-2 contributed to the up-regulation of TFAM or not, the selective COX-2 chemical inhibitor NS-398 was added into cell culture medium 6 h before 4 Gy γ-radiation at a final concentration of 20 μmol/L. At 6 and 12 h post-radiation, the expression levels of TFAM in U-2 OS and HeLa cells were detected. As displayed in [Figure 2](#ijms-20-01218-f002){ref-type="fig"}A, the addition of NS-398 obviously inhibited the induction of TFAM by radiation. Since NS-398 functions in blocking the enzymatic activity of COX-2, which is desired for the synthesis of prostaglandin, we therefore detected whether prostaglandin E2 (PGE2), the major form of physiological prostaglandin, stimulated the expression of TFAM. As shown in [Figure 2](#ijms-20-01218-f002){ref-type="fig"}B, in U-2 OS and HeLa cells, PGE2 treatment resulted in the elevation of TFAM expression by over 60% at 1 ng/mL, and by over 100% at a final concentration of 10 ng/mL. Further, we tested the effect of NS-398 on the radio-sensitivity of U-2 OS cell under 4 Gy of radiation. As shown in [Figure 2](#ijms-20-01218-f002){ref-type="fig"}C, 4 Gy radiation decreased the cell-surviving fraction to around 24%, compared to that of the non-irradiated cells. NS-398 pre-treatment resulted in a further decrease of the surviving fraction to around 8%. This confirmed the role of COX-2 in stimulating TFAM expression. 2.3. Mitochondrial Fragmentation Mediated by DRP1 Up-Regulates TFAM in Irradiated Cells {#sec2dot3-ijms-20-01218} --------------------------------------------------------------------------------------- Mitochondrial fragmentation is associated with the biogenesis of mitochondria. DRP1 mediates mitochondrial outer membrane fission in mammalian cells, and it is an important regulator of mitochondrial fragmentation. Given these facts, we next investigated whether DRP1 was involved in radiation-induced TFAM up-regulation. Firstly, the expression of DRP1 in irradiated U-2 OS cells was investigated. As shown in [Figure 3](#ijms-20-01218-f003){ref-type="fig"}A,B, radiation enhanced DRP1 expression in a dose-dependent manner within the range of 0 to 8 Gy. Besides, DRP1 expression was enhanced with post-radiation time elongation within 24 h, reaching a level of around 1.5-folds at the 6 h time point compared, to that in the non-irradiated cells. Then, the *DRP1* small interfering RNA (siRNA) and the selective chemical inhibitor Mdivi-1 were used to check the effect of DRP1 on radiation-induced TFAM expression. A final concentration of 20 μmol/L Mdivi-1 was added into the U-2 OS, Hep G2, and MCF7 culture medium 1 h before 4 Gy radiation. Twelve hours after radiation, total cell lysates were collected for analysis. As in [Figure 3](#ijms-20-01218-f003){ref-type="fig"}C, radiation-induced TFAM expression was largely blocked. Similar results were observed in *DRP1* siRNA-treated Hep G2 and U-2 OS cells ([Figure 3](#ijms-20-01218-f003){ref-type="fig"}D). The indicated radiation-induced DRP1 expression contributed to TFAM up-regulation. As mentioned above, DRP1 is an important regulator of mitochondrial fragmentation, we next investigated the association of mitochondrial morphology with TFAM expression. Plasmid pcDNA3.1 with an insertion of DNA sequence coding mitochondria-localization peptide fused turbo-green fluorescent protein (GFP) was transfected into U-2 OS cells, to visualize mitochondria in living cells under a fluorescence microscope. As shown in [Figure 3](#ijms-20-01218-f003){ref-type="fig"}E, around 85% of cells bore tubular mitochondria, and 5% of the cells bearing fragmented mitochondria if the cells were not irradiated. Three and 6 h after 4 Gy radiation, the cells bearing tubular mitochondria decreased to around 33% and 21%, respectively. However, the mitochondrial fragmentation caused by radiation was inhibited by Mdivi-1 pretreatment. Around 72% and 63% of the irradiated cells 3 or 6 h post-radiation displayed tubular morphology. To confirm the effect of DRP1 on TFAM expression, we tested the effect of Mdivi-1 on the radio-sensitivity of U-2 OS cells under 4 Gy radiation. As shown in [Figure 3](#ijms-20-01218-f003){ref-type="fig"}F, 4 Gy radiation decreased the cell-surviving fraction to around 25%, compared to that of the non-irradiated cells. Also, Mdivi-1 pre-treatment resulted in a further decrease of the surviving fraction to around 1%. Taken together with the results of DRP1 and TFAM expression, DRP1-mediated mitochondria fragmentation was the reason for TFAM induction in the irradiated cells. 2.4. Activation of p38-MAPK Enhances DRP1/TFAM Expression {#sec2dot4-ijms-20-01218} --------------------------------------------------------- p38-MAPK was reported to be involved in radiation-induced biological responses. To determine the upstream regulator of DRP1/TFAM, the phosphorylation of p38 was firstly estimated. As shown in [Figure 4](#ijms-20-01218-f004){ref-type="fig"}A, the phosphorylation of p38 showed an increasing tendency following 4 Gy radiation within 12 h in U-2 OS cells. We next investigated the association between the activation of p38 and the expression of DRP1/TFAM. Selective p38 inhibitor SB203580 was used to pre-treat U-2 OS and HeLa cells 1 h prior to radiation exposure at a final concentration of 20 μmol/L. Six and 12 h post-4 Gy radiation, the expression levels of DRP1 and TFAM were analyzed by Western blotting. As displayed in [Figure 4](#ijms-20-01218-f004){ref-type="fig"}B, radiation induced expression of both TFAM and DRP1 were attenuated by the addition of SB203580 in the two cell lines, indicating that the activation of p38 by radiation was an upstream regulator of the enhanced expression of DRP1 and TFAM. 2.5. COX-2/PGE2 Promotes p38 Phosphorylation and DRP1 Up-Regulation {#sec2dot5-ijms-20-01218} ------------------------------------------------------------------- Since we proved that both COX-2/PGE2 and DRP1 up-regulation mediated by p38 contributed to the enhanced expression of TFAM in irradiated cells, we next investigated whether COX-2/PGE2 affected the phosphorylation of p38 and DRP1 expression. Firstly, we checked the influence of the COX-2 inhibitor NS-398 on radiation-induced phosphorylation of p38. As shown in [Figure 5](#ijms-20-01218-f005){ref-type="fig"}A, NS-398 pre-treated U-2 OS and HeLa cells showed attenuated levels of phosphorylated p38 at 6- and 12-hr time points post-4 Gy radiation, compared to the corresponding groups that were not pre-treated with NS-398, indicating that COX-2 contributed to the activation of p38. This was further confirmed by the results that 1 ng/mL and 10 ng/mL PGE2 could lead to obvious enhancement of p38 phosphorylation ([Figure 5](#ijms-20-01218-f005){ref-type="fig"}B). We next explored whether COX-2/PGE2 also promoted DRP1 expression in U-2 OS and HeLa cells. As shown in [Figure 5](#ijms-20-01218-f005){ref-type="fig"}B,C respectively, DRP1 expression was attenuated by NS-398 pretreatment in cells irradiated by 4 Gy γ-rays, and this was enhanced by PGE2 treatment. To further confirm the effect of COX-2 on DRP1, the mitochondrial fragmentation induced by 4 Gy radiation under the presence of NS-398 was measured with U-2 OS cells bearing plasmid-producing mitochondria-localized turbo-GFP. As shown in [Figure 5](#ijms-20-01218-f005){ref-type="fig"}D, around 85% of the non-irradiated cells bore tubular mitochondria. Three and 6 h after 4 Gy radiation, the cells bearing tubular mitochondria decreased to around 35% and 20%, respectively. However, NS-398 pre-treatment resulted in recovery of the percentage of the irradiated cells displayed tubular mitochondria, to around 64% and 52%. This indicated that COX-2 spurred DRP1-mediated mitochondrial fragmentation. Together with the above results, it was identified that radiation-induced COX-2 facilitated the activation of p38-MAPK, which further enhanced DRP1 expression and its mediated mitochondrial fragmentation. 3. Discussion {#sec3-ijms-20-01218} ============= Previous reports have shown that the expression of TFAM was up-regulated, together with the increase of the mtDNA copy number after α-particle irradiation in human lung adenocarcinoma A549 cells \[[@B7-ijms-20-01218]\]. In the present study, we demonstrated that γ-irradiation up-regulated the expression of TFAM in U-2 OS, Hep G2, HeLa, and MCF7 cells, indicating that the enhanced expression of TFAM is widespread in tumor cells that are exposed to ionizing irradiation. Xie \[[@B9-ijms-20-01218]\] reported that the down-regulation of TFAM inhibited lung cancer cell tumorigenesis, leading to increased apoptotic cell death. Yao \[[@B4-ijms-20-01218]\] determined that microRNA-200a inhibits cell proliferation by targeting TFAM in breast cancer. Han \[[@B5-ijms-20-01218]\] indicated that the over-expression of TFAM enhanced the growth of cancer cell lines, whereas the down-regulation of TFAM inhibited cell growth. Fan \[[@B6-ijms-20-01218]\] found that MiR-199a-3p enhanced breast cancer cell sensitivity to cisplatin by downregulating TFAM. Mei \[[@B28-ijms-20-01218]\] indicated that reduced mtDNA copy numbers caused by TFAM knockdown sensitized the tumor cell lines HEp-2, HNE2, and A549 to chemotherapeutics. Taken with our result that TFAM knockdown sensitized U-2 OS and Hep G2 to ionizing radiation, it could be inferred that the enhanced expression of TFAM was a response of tumor cells to stressed conditions, and that the inhibition of TFAM might be a way for increasing the efficacy of tumor therapy. COX-2 can be induced by oncogenes, growth factors, and proinflammatory mediators in various types of cells \[[@B29-ijms-20-01218]\]. COX-2 is over-expressed in numerous types of cancer, and it mediates the production of prostaglandins, which stimulate cancer growth and protect the cells against damage by cytotoxic agents \[[@B30-ijms-20-01218],[@B31-ijms-20-01218]\]. Dandekar \[[@B32-ijms-20-01218]\] reported that the COX-2 inhibitor celecoxib augmented chemotherapeutic drug-induced apoptosis by enhancing the activation of caspase-3 and -9 in prostate cancer cells. Liou \[[@B22-ijms-20-01218]\] showed that a majority of COX-2 expression in tumor cells were co-localized with heat shock protein-60 in mitochondria. Another study indicated that mitochondrial-located COX-2 was involved in protecting tumor cells via impairing mitochondrial damage and apoptosis \[[@B33-ijms-20-01218]\]. Chang \[[@B34-ijms-20-01218]\] determined that WNT signaling controlled the radiosensitivity of head and neck cancer cell lines via the COX-2 mediated expression of the DNA repair protein Ku. Besides, further reports showed that the enhanced sensitivity of tumor cells to ionizing radiation under the presence of selective COX-2 inhibitors involved the inhibition of cellular repair from radiation damage and cell cycle redistribution \[[@B21-ijms-20-01218]\]. To our knowledge, there have been no previous reports on whether COX-2 could affect TFAM-mediated biogenesis, to help cells to resist stress conditions. Our results confirmed this hypothesis by providing data to show that selective COX-2 inhibitor NS-398 treatment blocked the induced expression of TFAM in irradiated tumor cells. Prostaglandin E2 (PGE2) is the most abundant prostanoid, and it is produced predominantly from arachidonic acid by tightly regulated cyclooxygenases and prostaglandin E synthases. COX-2-derived PGE2 induced Id1-dependent radiation resistance and self-renewal in experimental mouse glioblastoma \[[@B18-ijms-20-01218]\]. Our results indicated that PGE2 could stimulate TFAM expression in tumor cells without being irradiated. This provided new information on the mechanisms for the regulation of mitochondrial functions by COX-2 under extrinsic stimuli. Mitochondria are double-membrane-enveloped organelles. They are highly dynamic, with their normal functions being maintained by continuous morphological changes \[[@B35-ijms-20-01218],[@B36-ijms-20-01218]\]. DRP1 is the master regulator of mitochondrial fragmentation \[[@B37-ijms-20-01218]\]. Parone \[[@B25-ijms-20-01218]\] reported that preventing mitochondrial fragmentation by down-regulating the expression of DRP1 in mammalian cells led to the loss of mitochondrial DNA. Zhang \[[@B13-ijms-20-01218]\] reported that the knockdown of TFAM resulted in the suppression of radiation-induced elevation of mtDNA copy number. In this research, we demonstrated that radiation fragmented mitochondria and up-regulated TFAM. Also, when radiation induced mitochondrial fragmentation was suppressed, TFAM up-regulation was subsequently attenuated, indicating that the morphological changes of mitochondria contributed to radiation-induced TFAM expression. Our work provided results to show that the inhibition of COX-2 by NS-398 decreased the expression of DRP1 and mitochondrial fragmentation in the irradiated tumor cells, which was in line with the findings of Zhou \[[@B38-ijms-20-01218]\], and that the down-regulation of mitochondrial COX-2 inhibited the stemness of nasopharyngeal carcinoma by decreasing the activity of DRP1. Combined with our above results, we identified a signaling mechanism where COX-2 stimulated TFAM expression in irradiated tumor cells via DRP1-mediated mitochondrial fragmentation. p38 is a member of the MAPKs, and it responds to a variety of stress stimuli, including cytokines, radiation, and heat shock. Its activation in irradiated cells is associated with radiation-induced cell cycle arrest and apoptosis \[[@B39-ijms-20-01218],[@B40-ijms-20-01218]\]. Kim \[[@B41-ijms-20-01218]\] proved that p38 contributes to radiation-induced innate immune responses through inducing the expression of pro-inflammatory cytokines such as TNF-α, IL-6 and IL-12p40. Besides, p38 is involved in the regulation of mitochondrial functions and dynamics. Cryptotanshinone promoted mitochondrial biogenesis in C3H10T1/2 mesenchymal stem cells via AMPK and p38 signaling \[[@B42-ijms-20-01218]\]. Debattisti \[[@B43-ijms-20-01218]\] proved that reactive oxygen species (ROS) activated p38 to affect mitochondrial distribution and motility. It was also reported that p38 chemical inhibitors treatment down-regulated Drp1 in rat brain tissue collected from experimental stroke \[[@B44-ijms-20-01218]\]. In this research, our results showed that p38 was activated in irradiated tumor cells and that its inhibitor repressed radiation-induced up-regulation of DRP1, as well as TFAM. This further confirmed the regulatory role of p38 on mitochondria, and indicates the impacts of mitochondrial morphological changes on TFAM expression and mitochondrial biogenesis. Since we proved that COX-2 and p38 positively regulates DRP1/TFAM expression in irradiated tumor cells, we subsequently checked the contribution of COX-2 to p38 phosphorylation. Our results showed that NS-398 treatment decreased radiation-induced p38 phosphorylation, while PGE2 treatment could stimulate it. Kim \[[@B45-ijms-20-01218]\] reported that COX-2 inhibitor SC-236 suppressed nuclear factor-kappa B activation and the phosphorylation of p38 in human mast cell line cells. Zhao \[[@B46-ijms-20-01218]\] found that in HeLa cells overexpressing connexin-26, the knockdown of COX-2 resulted in the attenuated phosphorylation of p38. In addition, Jin \[[@B47-ijms-20-01218]\] determined that PGE2 stimulates the phosphorylation of ERK and p38 to inhibit the ROMK-like small-conductance K (SK) channels, and Ca^2+^-activated big-conductance K channels (BK) in mouse cortical collecting duct cells of. Our results that PGE2 stimulated p38 phosphorylation were consistent with these results, and further confirmed that COX-2 contributes to DRP1/TFAM expression via p38. Besides, Tessner \[[@B48-ijms-20-01218]\] reported that p38 is critical for the enhanced transcription and expression of COX-2 in γ-ray-irradiated human epithelial cells, showing that there was feedback regulation between COX-2 and p38 in irradiated cells. In conclusion, our current work provides evidence to show that COX-2 contributes to the up-regulation of TFAM, which further helps tumor cells to resist ionizing radiation. We also clarify that this is mediated by COX-2-stimulated p38 activation, and the subsequent enhancement of DRP1 expression and mitochondrial fragmentation. The results of our research put forward a mechanism for the regulation of TFAM in irradiated tumor cells, which may be considered as a candidate sensitization target in cancer radiotherapy. 4. Materials and Methods {#sec4-ijms-20-01218} ======================== 4.1. Cell Cultures and γ-Irradiation {#sec4dot1-ijms-20-01218} ------------------------------------ Human tumor cell lines U-2 OS, Hep G2, HeLa, and MCF7 were from ATCC (Manassas, VA, USA) and cultured in Dulbecco's modified Eagle's medium/F12 with 10% fetal bovine serum (Clark Bioscience, Richmond, VA, USA) at 37 °C in a humidified 5% CO~2~ incubator. The γ-ray emitter was a Biobeam GM gamma irradiator (Leipzig, Germany), which contained a ^137^cesium source, with a dose rate of 3.27 Gy/min. 4.2. Chemicals and Reagents {#sec4dot2-ijms-20-01218} --------------------------- The following primary antibodies were utilized: TFAM (1:500; sc-376672; Santa Cruz Biotechnology, Dallas, TX, USA); COX-2 (1:1000; 160112; Cayman Chemical Company, Ann Arbor, MI, USA); P-p38 (1:2000; 612280; BD Biosciences, Franklin Lakes, NJ, USA); DRP1 (1:1000; 12957-1-AP; ProteinTech, Rosemont, IL, USA); β-actin (1:1000; sc-8432; Santa Cruz Biotechnology, Dallas, TX, USA). HRP-linked goat anti mouse or rabbit IgG were purchased from Jackson ImmunoResearch (West Grove, PA, USA). SB203580 was purchased from Selleck Chemicals (Houston, TX, USA). NS-398 was purchased from Abcam (Cambridge, UK) and Mdivi-1 was purchased from Sigma-Aldrich (Merck KGaA, Darmstadt, Germany). PGE2 was purchased from Santa Cruz Biotechnology (Dallas, TX, USA). In this work, SB203580, NS-398, Mdivi-1, and PGE2 were respectively dissolved in DMSO, to prepare 1000-fold stock solutions. Upon usage, equal volumes of vehicle were added into the control groups. 4.3. Western Blot Analysis {#sec4dot3-ijms-20-01218} -------------------------- Cell samples were washed with PBS, and the cell lysate was prepared using RIPA buffer containing a protease inhibitor cocktail (Roche Diagnostics GmbH, Germany) and a protein phosphatase inhibitor cocktail (Sigma, Merck KGaA, Darmstadt, Germany). Protein concentrations were identified by using a BCA kit (Sangon Biotech Co., Ltd., Shanghai, China). The cell lysate (50 μg) was resolved using 10% or 12% SDS-PAGE, and transferred onto polyvinylidene fluoride (PVDF) membrane. Following blocking in 1% skim milk, the PVDF membrane was incubated with the primary antibody at 4 °C overnight, according to the product datasheets. Subsequently, the membrane was washed with PBST (PBS buffer containing 0.1% Tween-20) or TBST (NaCl, Tris-HCL, 0.1% Tween-20) and incubated with the corresponding HRP-conjugated secondary antibody for 2 h at room temperature. Protein bands were visualized using a chemiluminescence substrate (Boster, Wuhan, China) and band density was analyzed with ImageJ software (National Institutes of Health, Bethesda, MD, USA). 4.4. Quantitative Real-Time PCR {#sec4dot4-ijms-20-01218} ------------------------------- Cellular DNA was extracted using DNAiso reagent (Takara, Shiga, Japan). The relative mitochondrial DNA (mtDNA) copy number was quantified with a SYBR Green quantitative PCR kit (Takara) using the relative standard curve method. mtDNA content was determined by amplification of 12S ribosomal DNA (rDNA) coded by mtDNA with primers 5′TAACCCAAGTCAATAGAAGCC and 5′CTAGAGGGATATGAAGCACC. Nuclear DNA (nDNA) content was determined by the amplification of the β-actin coding sequence with the primers 5′GAGCGGGAAATCGTGCGTGAC and 5′GGAAGGAAGGCTGGAAGAGTG. The mtDNA/nDNA ratio was used to estimate the relative mtDNA copy number. The PCR conditions were: 95 °C for 30 s, 52 °C for 30 s, and 72 °C for 30 s, 30 cycles. 4.5. Clonogenic Survival Assay {#sec4dot5-ijms-20-01218} ------------------------------ Post irradiation, cells were trypsinized, diluted, and seeded into cell culture dishes. After incubation for three weeks, the dishes were washed with pre-warmed PBS, fixed with methanol and acetic acid (*v*/*v* = 9:1), and stained with crystal violet for 15 min. Colonies containing more than 50 cells were counted and plotted. 4.6. RNA Interference {#sec4dot6-ijms-20-01218} --------------------- Human *TFAM* shRNA (shTFAM) constructs were purchased from OriGene (Rockville, MD, USA). Cells with stable knockdown of TFAM were isolated by transfecting the shTFAM construct and then performing selection under the presence of puromycin. The siRNA oligonucleotide targeting human *DRP1* (GCUACUUUACUCCAACUUAUUTT) was synthesized in GenePharma (Shanghai, China). To knock down DRP1, the cells were grown to 70% confluence in a cell culture dish. Then, the DRP1 siRNA was mixed with Lipofectamine 2000 (Thermo Fishier, Carlsbad, CA, USA) and added into the cell culture. After incubation for 48 h, the cells were subjected to radiation treatment and analysis. 4.7. Mitochondrial Morphology Analysis {#sec4dot7-ijms-20-01218} -------------------------------------- To analyze the effects of radiation on the dynamic changes of mitochondrial morphology, the DNA segment coding turbo-GFP with the N-terminal fused to the mitochondria-localization sequence of cytochrome C oxidase subunit IV was inserted into pcDNA3.1 plasmid, according to the reading frame. The recombinant plasmid was then transfected into U-2 OS cells. Cells stably expressing mito-tGFP were isolated for mitochondrial morphology analysis. After radiation or chemical inhibitor treatment, at the desired time points, the mitochondrial morphologies were captured under a fluorescence microscope. Mitochondrial morphologies were divided into three types. ''Tubular'' means that over 70% of cellular mitochondria showed tubular morphology. ''fragmented'' meant that over 70% of cellular mitochondria showed fragmented morphology. Others were classified as being ''tubular + fragmented''. At least 200 cells were analyzed for each sample. 4.8. Statistical Analysis {#sec4dot8-ijms-20-01218} ------------------------- All data were presented as the mean ± standard deviation from at least three independent experiments performed in triplicate. Statistical significance between the two groups was evaluated using Student's t-test with GraphPad Prism 5 (GraphPad Software, Inc., San Diego, CA, USA). Statistical significance between multiple groups was evaluated using one-way analysis of variance with SPSS 12.0 software (SPSS, Inc., Chicago, IL, USA). *p* \< 0.05 was considered to indicate a statistically significant difference. Conceptualization, J.W.; methodology, F.T. and R.Z.; writing, F.T. and J.W. This research was funded by National Natural Science Foundation of China, grant number 31370842, 11575232, 11835014. The International Partnership Program of the Chinese Academy of Sciences, grant number 116134KYSB20160084. The Innovative Program of the Development Foundation of Hefei Center for Physical Science and Technology, grant number 2016FXCX005. The APC was funded by the National Natural Science Foundation of China. The authors declare no conflict of interest. {#ijms-20-01218-f001} {#ijms-20-01218-f002} {#ijms-20-01218-f003} {#ijms-20-01218-f004} {#ijms-20-01218-f005} [^1]: These authors contributed equally.

Introduction {#S1} ============ Multiple sclerosis (MS) is a chronic inflammatory, demyelinating disease of the central nervous system (CNS) with axonal injury, characterized by varying clinical course, pathology, and inflammatory patterns ([@B1]). It develops in susceptible hosts after interaction with environmental factors which trigger the disease by promoting the activation of myelin-specific T cells that normally circulate in the peripheral lymph organs of all individuals ([@B2]). It has been suggested that some infectious agents, in particular viruses, may be potential triggers of MS ([@B2]--[@B4]). Among different infective agents, Epstein--Barr virus (EBV) has been mostly associated with increased MS risk ([@B5]). Recently, it has been shown increased CD8^+^ T cell response to EBV lytic antigens in active MS and also in relapses ([@B6]). Infection with murine gamma herpesvirus 68 (γHV-68), the murine homolog to EBV, polarizes the adaptive immune response and heightens CNS pathology following experimental autoimmune encephalomyelitis (EAE) induction and likely, influences MS pathogenesis ([@B7]). Experimental autoimmune encephalomyelitis is the experimental model of MS, induced in susceptible animals by active immunization with myelin antigens mixed with adjuvant ([@B8]). Immunized mice develop ascending paralysis with CD4^+^ T cells and macrophages in infiltrations in the white matter of the spinal cord, and with minimal brain inflammation in the majority of experimental models. However, in MS, the vast majority of myelin lesions are found within the brain parenchyma with infiltrations that contain equivalent numbers of CD8^+^ T and CD4^+^ T cells ([@B9], [@B10]). Despite these differences, EAE is considered as a valuable tool for research of MS pathogenesis. Moreover, several therapeutics that are now being used to treat MS were developed in EAE ([@B11]). BALB/c mice are found partially or completely resistant to the induction of EAE with encephalitogenic peptide, myelin oligodendrocyte glycoprotein (MOG~35--55~). Cytomegalovirus (CMV) classified within the *Betaherpesvirinae* subfamily establishes life-long latent infections in 70--100% of the human population ([@B12]). After a primary infection of fibroblasts, epithelial, endothelial, and smooth muscle cells ([@B6]), mostly asymptomatic in the immunocompetent host, CMV persists in myeloid precursor cells ([@B7]). During latency periodic asymptomatic reactivations occur ([@B13]). CMV contains a large number of latent and lytic genes, many of which code proteins that have the role in immunoregulation ([@B5]). When monocytes that carry CMV enter visceral parenchyma and differentiate into macrophages and myeloid dendritic cells, virus reactivates and through expression of different genes can modulate the immune response of the host ([@B14]). Data on the role of CMV infection in etiopathogenesis of MS are controversial. CMV has been found in demyelinating plaques and the cerebrospinal fluid of MS patients ([@B15]) and causes demyelination mainly in the CNS of immunocompromised hosts ([@B16]). Further, enhancement of numbers of EBV and CMV-specific CD8^+^ T cells among T cells in chronic inflammatory lesions of brain of MS patients was reported ([@B17]). Several studies involving human subjects indicate correlation between CMV infection and MS development, greater rate of relapses and greater brain atrophy ([@B18]--[@B20]). Other studies indicate that CMV seropositivity is associated with a decreased MS risk and predicts a better clinical and radiological outcome in MS patient ([@B21]), suggesting a protective effect of CMV on autoimmune neuropathology ([@B22]). Furthermore, CMV encodes multiple factors that trigger immunomodulatory or evasion mechanisms, which can decrease the immune response in MS patients ([@B23], [@B24]). We have recently shown that deletion of an immunoregulatory pathway, IL-33/ST2 axis, may enhance susceptibility to EAE in resistant BALB/c strain ([@B25], [@B26]). The present study was done with the aim to explore whether infection with murine CMV (MCMV) create "fertile field" ([@B27], [@B28]) that facilitates the expansion and activation of encephalitogenic cells leading to autoimmune disease of CNS. Here, we show that MCMV infected and MOG~35--55~ immunized BALB/c mice develop very pronounced neuroinflammation with extensive infiltrations in brain and spinal cord parenchyma containing large proportion of CD8^+^ cells in infiltrates in addition to accentuation of Th1 and Th17 immune response and skewing microglia to M1 phenotype. Our results are compatible with the notion that MCMV abrogates inherent resistance of BALB/c mice to EAE induction with MOG~35--55~ peptide through enhancement of inflammatory dendritic cells in the periphery, M1 type of microglia and recruitment of MOG~35--55~ responsive CD8^+^ T cells in the CNS. Thus, CMV-induced inflammatory environment may enhance autoimmunity in CNS. Materials and Methods {#S2} ===================== Infection, Induction, and Scoring of EAE {#S2-1} ---------------------------------------- Female 8-week-old BALB/c mice were used throughout this study. Mice were infected subcutaneously (footped) with 10^5^ plaque-forming units of tissue culture MCMV, strain MW97.01 ([@B29]). EAE was induced 10 days after infection by subcutaneous administration of 200 µL suspension at two sites over the hind flanks. Depletion of CD4^+^ lymphocytes, where indicated, was performed with intraperitoneal injection of 100 µg of anti-CD4 mAb, 1 day prior to and 5 days after MOG~35--55~ immunization. The suspension consisted of 300 µg MOG~35--55~ peptide (Sigma-Aldrich, Germany) in 100 µL of PBS, emulsified with 100 µL complete Freund's adjuvant (Sigma-Aldrich, Germany) with 0.7 mg heat-inactivated *Mycobacterium tuberculosis* (strain H37 RA; Difco Laboratories, Detroit, MI, USA). Each mouse was immediately thereafter, injected intraperitoneally and 48 h later with 300 ng pertussis toxin (List Biological Laboratories, Campbell, CA, USA) in 100 µL 0.9% NaCl. Clinical signs of EAE were assessed daily by the following scoring system: grade 0, no signs; grade 1, paralyzed tail; grade 2, ataxic; grade 2.5, one hind leg paralyzed; grade 3, both hind legs paralyzed; grade 3.5, three legs paralyzed; grade 4, both hind legs and front limbs completely paralyzed; grade 5, moribund as previously described ([@B30], [@B31]). Mice were monitored daily with fluid administration and mashed chow on the base of cages for all mice displaying a clinical score of 3. Mice were maintained in our animal facilities in a temperature-controlled environment with a 12-h light/12-h dark cycle and were administered standard laboratory food and water *ad libitum*. All experiments were approved by and conducted in accordance with the Guidelines of the Animal Ethics Committee of Faculty of Medical Sciences, University of Kragujevac, Serbia. Endangered animal species were not used in this study. Isolation of Mononuclear Cells from CNS and Lymph Nodes {#S2-2} ------------------------------------------------------- At day 15 post-EAE induction (mean clinical score of 3 for MCMV EAE mice), mice were perfused with PBS, and brain and spinal cord were carefully removed. The mononuclear cells from CNS were isolated as described previously ([@B25]). Briefly, the brains and spinal cords were homogenized in RPMI 1640 (Sigma-Aldrich) with 10% FBS and 1 mg/mL collagenase type I (Sigma-Aldrich) and incubated at 37°C for 60 min. After digestion, the tissue was passed through a 40-µm mesh, pelleted, resuspended in 10 mL 30% Percoll (Sigma-Aldrich), overlaid onto 5 mL 70% Percoll, and centrifuged at 390 *g* for 20 min. The myelin layer was removed, and the mononuclear cells accumulated in the intermediate phase were collected, washed twice in PBS, and resuspended in RPMI 1640 containing 10% FBS. Total cell numbers were determined by counting on a hemocytometer, and viability was assessed by trypan blue exclusion. Lymph nodes were minced in RPMI 1640 (Sigma-Aldrich) and forced gently through 40-µm cell-strainer nylon mesh (Falcon) using a sterile syringe plunger and centrifuged at 400 *g* for 5 min. Pellet from lymph nodes was resuspended in RPMI 1640 containing 10% FBS. Flow Cytometry {#S2-3} -------------- Single-cell suspensions of brain and spinal cord tissue were prepared according to standard protocols. For cytofluorometry, following antibodies were used: CD4, CD8, CD45, CCR6, CXCR3, TCRβ, CD11c, CD11b, CD49b, CCR2, CD86, T-bet, RORγt, IL-17, IFN-γ, TNF-α, and IL-12 with conjugated fluorochromes (BD Biosciences). Antibodies were incubated with cells in PBS with 2% FBS for 30 min at 4°C, and then cells were analyzed. For intracellular staining of cytokines, cells were stimulated for 4 h in RPMI 1640 containing 10% FBS (Gibco), 10 ng/mL phorbol 12-myristate 13-acetate (Sigma-Aldrich), and 500 ng/mL ionomycin (Sigma-Aldrich) with addition of Brefeldin A (BD Biosciences). Antibodies for the cell surface markers were added to the cells in PBS with 2% FBS for 30 min on ice. After wash, cells were resuspended in Cytofix/Cytoperm buffer (BD Biosciences) for 20 min on ice, washed twice, and incubated with Abs for intracellular antigens (cytokines) in Perm buffer (30 min, on ice). For staining of transcriptional factors, unstimulated cells were used. Data were acquired using FACSCalibur (BD Biosciences) and analyzed with FlowJo software (Tree Star). Tetramer Staining {#S2-4} ----------------- Immune cells were isolated as described above. Cells isolated from CNS were incubated with H-2L(d)/IE-1/pp89 (168--176 YPHFMPTNL) and H-2D(d)/m164 (257--265 AGPPRYSRI) tetramers provided by NIH tetramer core facility. Cells stained with tetramer and anti-CD8 and anti-CD3 antibodies were incubated for 30 min at room temperature and then washed. Data were acquired using a FACSCalibur (BD Biosciences) and analyzed with FlowJo software (Tree Star). Immunohistochemistry and Evaluation of Brain and Spinal Cord Pathology {#S2-5} ---------------------------------------------------------------------- Brains and spinal cords were fixed in 4% buffered formalin fixative overnight. Paraffin wax embedded sections (5 µm) were stained with hematoxylin and eosin and CD3 (ab699; Abcam) immunohistochemical staining. The slides were analyzed on light microscope (BX51; Olympus), and digital images were acquired by digital camera. The level of infiltration was graded using the following score: 0, no inflammatory cells; 1, a few scattered inflammatory cells; 2, organization of inflammatory infiltrates into perivascular cuffs; 3, extensive perivascular cuffing with extension into adjacent subarachnoid space and CNS parenchyma, and 4, extensive perivascular cuffing with increasing subarachnoid and parenchymal inflammation ([@B32]). Slides were analyzed on Olympus BX51 microscope, and digital images were acquired by Olympus digital camera (DP71). Interferon-γ Assay {#S2-6} ------------------ Mononuclear cells isolated from CNS, 10^5^ in 100 µL complete media were put on 96-well plate in duplicates, and 100 µL of media, or MOG~35--55~ (1 μg/well) were added. After incubation for 1 h on 37°C, 0.2 µL of Brefeldin A in 10 µL of medium was added to each well and incubated for 4 h on 37°C. Cells were washed and then incubated with anti-CD8 and anti-CD4 antibodies on +4°C for 15 min. After washing, cells were resuspended in Cytofix/Cytoperm buffer for 30 min on ice, washed twice, and incubated with anti-IFNγ antibodies diluted in Perm wash buffer (30 min, on ice) and resuspended in FACS media. Statistical Analysis {#S2-7} -------------------- All statistics were carried out using SPSS 18.0 for Windows software. Results were analyzed using the Student's *t*-test or Mann--Whitney test and ANOVA or Kruskal--Wallis. Data in this study were expressed as the mean + SEM. Values of *P* \< 0.05 were considered significant. Results {#S3} ======= MCMV Infection in Adult Life Abrogates Resistance to EAE in BALB/c Mice {#S3-1} ----------------------------------------------------------------------- BALB/c mice immunized with MOG~35--55~ did not develop clinical signs of EAE, while BALB/c mice infected with MCMV 8 weeks after birth and 10 days later challenged with MOG~35--55~ in CFA and pertussis toxin (MCMV + MOG~35--55~) developed clinical signs that correspond to EAE manifestations seen in C57BL/6 mice (Figure [1](#F1){ref-type="fig"}A). Based on evaluation of clinical course (Figure [1](#F1){ref-type="fig"}A) and mean maximal clinical score (Figure [1](#F1){ref-type="fig"}B), MCMV-infected BALB/c mice developed disease indistinguishable from disease in susceptible C57BL/6 mice. Infiltration in CNS of MCMV + MOG~35--55~, expressed by mean histological score (Figure [1](#F1){ref-type="fig"}D) and total cell number (Figure [1](#F1){ref-type="fig"}C), was significantly higher compared with BALB/c mice immunized with MOG~35--55~ only mice. BALB/c mice infected with MCMV and 10 days later immunized with MOG~35--55~ developed subarachnoid and perivascular infiltrations in the brain cortex, perivascular infiltrations in brainstem and cerebellum with spreading to parenchyma (Figure [1](#F1){ref-type="fig"}E), and white matter spinal cord infiltrations (Figure [1](#F1){ref-type="fig"}F). Single-cell infiltrates were detected in brains and spinal cords of MOG~35--55~-immunized mice only, and mild perivascular infiltrations were detected in brainstem of MCMV-infected mice (Figures [1](#F1){ref-type="fig"}E,F). Immunostaining of spinal cord sections showed presence of CD3^+^ cells in the infiltrates (Figure [1](#F1){ref-type="fig"}G). {#F1} CNS Infiltrates of MCMV + MOG~35--55~ Mice Contain Higher Amounts of T1/T17 CD4^+^ and CD8^+^ T Cells {#S3-2} ----------------------------------------------------------------------------------------------------- Further analysis showed significantly higher number of CD4^+^ and CD8^+^ T cells in the infiltrates of MCMV + MOG~35--55~ mice compared with MOG~35--55~ mice (Figure [2](#F2){ref-type="fig"}A). In the CNS of MCMV + MOG~35--55~ mice, there was higher number of CD4^+^ T cells then CD8^+^ T cells, similar to typical EAE in C57BL/6 mice. MCMV + MOG~35--55~ mice had increased percentage (Figure [2](#F2){ref-type="fig"}B) and number (Figure [2](#F2){ref-type="fig"}C) of both Th1 and Th17 cells, as well as Tc1 and Tc17 cells compared to MOG~35--55~ mice. Similarly, significantly higher number of CD4^+^ cell- and CD8^+^ cell-expressing transcriptional factors, T-bet and RORγt (Figure [2](#F2){ref-type="fig"}D), and chemokine receptors CCR6 and CXCR3 (Figure [2](#F2){ref-type="fig"}E) was noticed in CNS of MCMV + MOG~35--55~ mice compared with MOG~35--55~ mice. {#F2} CD8^+^ Cells Take a Role in Autoimmune Neuropathology in BALB/c Mice with MCMV Infection {#S3-3} ---------------------------------------------------------------------------------------- Depletion of CD4^+^ cells in MCMV-infected mice abrogated susceptibility to MOG~35--55~-induced disease (Figure [3](#F3){ref-type="fig"}A) indicating autoimmune disease. Given the significant number of CD8^+^ cells in the infiltrates that were not seen in classical EAE, we explored in more details these cells found in CNS. There was significantly higher percentage and number of CD8^+^ cell-expressing markers of cytolytic activity in the CNS of MCMV + MOG~35--55~ BALB/c mice compared to MOG~35--55~ mice (Figure [3](#F3){ref-type="fig"}B). To indirectly determine the percentage of MOG~35--55~-specific CD4^+^ and CD8^+^ cells, mononuclear cells were isolated from CNS of MCMV + MOG~35--55~ BALB/c, MOG~35--55~ BALB/c, and MOG~35--55~ C57BL/6 mice and *ex vivo* restimulated with MOG~35--55~ peptide, and IFN-γ^+^ cells were enumerated. Significantly higher percentage of CD4^+^ and CD8^+^ cells from CNS of MCMV + MOG~35--55~ mice responded to *ex vivo* restimulation compared with MOG~35--55~ only BALB/c mice (Figure [3](#F3){ref-type="fig"}C). Further, significantly higher percentage of CD8^+^ cells isolated from CNS of MCMV + MOG~35--55~ BALB/c mice contained IFN-γ after *ex vivo* restimulation with MOG~35--55~ compared to C57BL/6 mice immunized with MOG~35--55.~ Even more importantly, IFN-γ-containing CD8^+^ cells in MCMV + MOG~35--55~ BALB/c mice after restimulation with MOG~35--55~ were not specific for viral epitopes pp89 and m164, suggesting that inflammatory CD8^+^ cells in the CNS are autoimmune (Figure [3](#F3){ref-type="fig"}D). {#F3} Chronic Non-Productive MCMV Infection Also Facilitates EAE Development in BALB/c Mice {#S3-4} ------------------------------------------------------------------------------------- In order to test whether the chronic non-productive MCMV infection could facilitate EAE development, we immunized BALB/c mice with MOG~35--55~ peptide 3 months after MCMV infection in adult life. As shown in Figure [4](#F4){ref-type="fig"}, infected mice developed clinical signs of EAE while age-matched mice immunized with encephalitogen only, did not. Mice with chronic non-productive MCMV infection started to manifest signs of EAE 6 days after immunization with MOG~35--55~ peptide; maximal clinical score reached 15 days after immunization and had very mild signs of the disease 60 days after immunization (Figure [4](#F4){ref-type="fig"}A). Chronic disease was confirmed with histological analysis. Perivascular infiltrates were detected in the spinal cords of mice with latent MCMV infection 2 months after challenge with MOG~35--55~ peptide (Figure [4](#F4){ref-type="fig"}B). Among cells isolated from brains of MCMV-infected BALB/c mice 2 months after MOG~35--55~ immunization, higher percentage of CD4^+^ and CD8^+^ cells contained inflammatory cytokines IL-17 and IFN-γ, after *in vitro* restimulation with MOG~35--55~ peptide compared to stimulated cells isolated from MOG~35--55~-immunized mice (Figure [4](#F4){ref-type="fig"}C). Our findings indicate that BALB/c mice with latent MCMV infection develop disease with long-lasting infiltrates in the CNS that contains lymphocytes specific for neuroantigens. {#F4} MCMV Infection Induces Inflammatory Phenotype of Antigen-Presenting Cells in Periphery and in CNS {#S3-5} ------------------------------------------------------------------------------------------------- It is known that viral infection induces antiviral immune response mediated by NK cells, CD8^+^ and CD4^+^ lymphocytes ([@B33]). Such inflammatory microenvironment in peripheral lymph organs can affect activation of antigen-presenting cells and thus indirectly contribute to development of inflammatory lymphocytes. Therefore, we explored possible influence of viral infection on changes of phenotype of antigen-presenting cells. To this end, mononuclear cells were isolated from inguinal lymph node 12 days after MCMV was administered in foot pad and compared with cells isolated from mice immunized with MOG~35--55~ only. Lymph nodes of MCMV-infected and MCMV + MOG~35--55~ mice had significantly higher percentage of CD11c^+^ dendritic cells and CD11c^+^PDCA1^+^plasmocitoid dendritic cells in inguinal lymph nodes compared to MOG~35--55~-immunized mice (Figure [5](#F5){ref-type="fig"}A). Also higher percentage of dendritic cell-expressing CCR2 chemokine receptor was found in both groups of MCMV-infected mice (Figure [5](#F5){ref-type="fig"}B). More importantly, lymph nodes of MCMV-infected and MCMV + MOG~35--55~ mice contained higher percentage of activated CD86^+^ and CD40^+^ (Figure [5](#F5){ref-type="fig"}C) and inflammatory IL-12^+^ dendritic cells (Figure [5](#F5){ref-type="fig"}E). Higher expression of markers of activation, CD86^+^ and CD40^+^ was noticed in MCMV-infected mice compared with MOG~35--55~-immunized mice (Figure [5](#F5){ref-type="fig"}D). These data suggest that inflammatory phenotype of dendritic cells is achieved in BALB/c mice with viral infection but not with encephalitogenic challenge only, as it was seen in C57BL/6 mice. {#F5} It is known that systemic inflammation in mice causes activation of microglia that persists for months ([@B34], [@B35]). Since MCMV infection in BALB/c mice causes systemic inflammation, we wanted to see effect of MCMV infection on phenotype of antigen-presenting cells in CNS. We analyzed expression of markers of classical (CD11c) and alternative activation (CD206) in the population of microglia (CD45^int^CD11b^+^) in healthy mice, MOG~35--55~-immunized mice, MCMV-infected, and MCMV + MOG~35--55~ mice. As shown in Figure [5](#F5){ref-type="fig"}F, microglia of mice with viral infection in adult life, with and without EAE, had proinflammatory, M1 phenotype. Significantly higher percentage of CD11c-expressing cells in microglia population was found in both groups of infected mice compared to healthy and MOG~35--55~-immunized mice. On the other hand, the highest percentage of (type 2) CD206-expressing microglia was found in MOG~35--55~ BALB/c mice. Higher percentage of M2 microglia was also found in MCMV + MOG~35--55~-immunized mice compared to healthy and MCMV-infected mice but lower compared to MOG~35--55~-immunized mice. Thus, high percentage of M2 microglia in mice at the peak of EAE could be also the compensatory mechanism that precedes disease attenuation. Discussion {#S4} ========== Here, we provide the first evidence that MCMV infection results in breaking resistance of BALB/c mice to EAE induction with MOG~35--55~ peptide, as indicated by typical clinical manifestations and massive inflammatory infiltration in the CNS (Figure [1](#F1){ref-type="fig"}). The role of MCMV infection in EAE is not studied, while its significance in MS is controversial. There are several prospective clinical studies that indicate protective effect of CMV infection on MS risk ([@B21], [@B22], [@B36]), while recent well-powered meta-analysis found no significant difference in the rate of CMV seropositivity between MS patients and healthy controls based on pooled samples from all studies to date ([@B37]). On the other hand, CMV has been found in demyelinating plaques and the liquor of MS patients ([@B15]), and several clinical studies support the role of CMV in MS pathogenesis ([@B18]--[@B20]). Additionally, CMV infection induces expansion of inflammation-seeking/proinflammatory effector-memory CD4^+^CD28^null^ T cells that are attracted to MS lesions *via* a CX3CL1 gradient ([@B38], [@B39]) and are mostly found in MS patients ([@B40]). Stimulation of these cells with myelin autoantigens results in their proliferation and release of cytotoxic granules, and thus may contribute to MS pathology ([@B41]). Some experimental animal studies support the role of CMV in EAE development. Cross-reactivity between CMV~981--1,003~ and MOG~35--55~ peptides was found in Lewis rats immunized with MOG~35--55~ ([@B42]), while cross-reactivity between CMV~981--1,003~ and MOG~34--56~ was found in rhesus monkeys ([@B43]). Immunization of rhesus monkeys with human CMV~981--1,003~ peptide induced expansion of MOG~34--56~-specific T cells ([@B43]). Female SJL/J mice primed with vaccinia virus that contain PLP gene and later challenged with MCMV-developed lesions in white matter regions in the brains ([@B28]). Other studies on primates also support the role of CMV in MS pathogenesis ([@B44], [@B45]). Recently, Juranic Lisnic et al. demonstrated that MCMV infection of murine fibroblasts induced highest expression of interferon β, transcriptional factor T-bet, chemokine CXCL10 ([@B46]), and the role of these markers of Th1 cells, in EAE pathogenesis is well known ([@B47]--[@B49]). However, there was no evidence that MCMV infection may directly facilitate EAE. Here, we show that adult MCMV infection overcomes resistance of BALB/c mice to induction of EAE with MOG~35--55~ peptide. The disease is characterized by typical clinical signs (Figure [1](#F1){ref-type="fig"}A) seen in susceptible C57BL/6 mice and massive brain and spinal cord infiltrations (Figures [1](#F1){ref-type="fig"}E,F). It should be noted that brain infiltrates are more significant in BALB/c mice treated with encephalitogen then in "classical" EAE in C57BL/6 mice. It appears that the disease after infection + MOG~35--55~ challenge in otherwise resistant mice is more similar to MS than classical EAE ([@B50]). While in C57BL/6 mice, CD4^+^ cells dominate in CNS infiltrates in the experiments presented here, there was similar number of CD4^+^ and CD8^+^ cells in the infiltrates (Figure [1](#F1){ref-type="fig"}). Encephalitogenity of CD4^+^ T cells in the infiltrates in BALB/c mice is further documented by their expression of chemokine receptors CXCR3 (Figure [2](#F2){ref-type="fig"}E), whose blockade during EAE induction attenuated the disease ([@B51]), and CCR6 receptor known to have the key role in development of initial autoimmune infiltration in the CNS ([@B52], [@B53]). While γHV-68 infection in C57BL/6 mice with EAE leads to almost exclusive infiltration with Th1 cells ([@B7]), in the CNS of MCMV-infected BALB/c mouse with EAE there is almost equal participation of Th1 and Th17 cell-expressing IFN-γ and T-bet (Th1) and IL-17 and RORγt (Th17) (Figures [2](#F2){ref-type="fig"}C,D). Moreover, inflammatory infiltrates in MCMV-pretreated BALB/c mice immunized with MOG~35--55~ contained CD8^+^ cell-expressing T1 and T17 transcriptional factors and corresponding cytokines TNF-α and IFN-γ (Tc1) and IL-17 (Tc17 cells) (Figures [2](#F2){ref-type="fig"}C,D). Interestingly, it was suggested that Tc17 cells are required for Th17 accumulation and development of MS ([@B54]). Patients with early-stage MS harbor a greater number of Tc17 cells in the cerebrospinal fluid than in peripheral blood that contribute to the initiation of CNS autoimmunity ([@B54]). Since in the CNS of BALB/c mice with neonatal MCMV infection but without immunization with MOG~35--55~ dominates Tc1 cells (IFN-γ and T-bet^+^) ([@B55]), it appears that the newly developing autoimmune process attracts a different population of CD8^+^ cells (IL-17 producing) that contribute to autoimmune process ([@B54]). Significant accumulation of CD8^+^ cells in the infiltrates and their role in autoimmune pathogenesis is documented by clinical signs and typical pathology complemented by specificity of infiltrating cells. *Ex vivo* restimulation with MOG~35--55~ leads to significant increase of CD8^+^ cell-producing IFN-γ in MCMV + MOG~35--55~ BALB/c mice (Figure [3](#F3){ref-type="fig"}C). This finding is at variance with our finding in C57BL/6 mice. CD8^+^ cells from EAE C57BL/6 mice did not responded to restimulation with MOG~35--55~ (Figure [3](#F3){ref-type="fig"}C)~.~ Further, MOG~35--55~ responsive CD8^+^ cells in MCMV + MOG~35--55~ BALB/c mice were not specific for viral epitopes pp89 and m164 (Figure [3](#F3){ref-type="fig"}D) implicating that CD8^+^ cell contributes to autoimmune process in this model of EAE. This finding is in line with previous report that initiation of autoimmune process in CNS with CD4^+^ T cells is followed with spreading to myelin-specific CD8^+^ T cells that are capable of direct recognition of oligodendrocytes and contribute to tissue damage ([@B56]). Similarly, heighten EAE in C57BL/6 mice infected with γHV-68 is accompanied with infiltration of brain parenchyma with CD8^+^IFN-γ^+^granzyme^+^ cells ([@B7]). However, specificity for autoantigen of these inflammatory and cytolytic CD8^+^ cells in γHV-68-infected C57BL/6 mice was not studied ([@B7]). Finding that mice with depletion of CD4^+^ cells after MCMV infection and before MOG~35--55~ immunization did not develop the disease (Figure [3](#F3){ref-type="fig"}A) contributes to the conclusion that MCMV-infected BALB/c mice developed autoimmune neuropathology. Persistence of CNS infiltrations and MOG~35--55~-specific CD4^+^ and CD8^+^ cells in CNS of BALB/c mice 2 months after MOG~35--55~ immunization that were infected 3 months previously (Figure [4](#F4){ref-type="fig"}) also proving autoimmune nature of the disease. Thus although CD4^+^ cells are required, it appears that CD8^+^ cells are the main effector cells. MCMV infection significantly increases proportion of dendritic cells (CD11c^+^), plasmocitoid dendritic cells (CD11c^+^PDCA1^+^) in peripheral lymph nodes (Figure [5](#F5){ref-type="fig"}A) compared to immunization with MOG~35--55~. Further higher percentage of dendritic cells in lymph nodes in MCMV-infected mice is accompanied with higher percentage of CCR2^+^ dendritic cells (Figure [5](#F5){ref-type="fig"}B). It is known that MCMV encodes proinflammatory factor (MCK-2), analog of chemokine CCL2 ([@B57]) that enhances monocyte recruitment and viral dissemination ([@B58]). Then, higher percentage of dendritic cells in inguinal lymph nodes of MCMV-infected mice could be the consequence of CCL2 analog production, since CCL2 goes to lymph nodes where is presented on the surface of high endothelial venules for recruitment of monocytes ([@B59]). Although it was previously shown that MCMV attracts monocytes that have the immunosuppressive role ([@B60]), here, we found higher percentage of dendritic cell-expressing markers of activation CD86 and CD40 (Figure [5](#F5){ref-type="fig"}C) and containing Th1 promoting cytokine, IL-12 (Figures [5](#F5){ref-type="fig"}D,E). Our results indicate that MCMV infection of BALB/c mice induces increase of inflammatory dendritic cells in peripheral lymph nodes and thus enables development of encephalitogenic T cells. This finding is in correlation with previous report that MCMV-infected mice are resistant to bacterial infection due to prolonged production of the antiviral cytokine IFN-γ and systemic activation of macrophages ([@B61]). Significantly, there was an increase of classically activated microglia (CD45^med^CD11b^+^CD11c^+^) in the CNS of BALB/c mice 25 days after MCMV infection compared to MOG~35--55~-immunized mice without previous infection that had mostly alternatively activated microglia (CD45^med^CD11b^+^CD206^+^) (Figure [5](#F5){ref-type="fig"}F). Lower percentage of alternatively activated microglia in MCMV + MOG~35--55~ mice observed on day 15 after immunization and day 25 after infection in compared to MOG~35--55~ treated mice, and prevalence of M1 microglia in virus-infected mice may contribute to chronic disease in MCMV-infected MOG~35--55~-immunized BALB/c mice. Previously, it was shown that systemic MCMV infection elicited a significant increase in the number of microglia with morphological signs of activation and M1 phenotype ([@B62]). Thus, it appears that at least one level of resistance of BALB/c mice to EAE is the inability to convert microglia into M1 phenotype. In summing up, we report here that MCMV infection may promote autoimmune neuropathology and convert resistant mice into susceptible to EAE induction. This was achieved by activation of antigen-presenting cells and promoting M1 phenotype of microglia as well as participation of CD8^+^ encephalitogen-specific T cells in the autoimmune pathogenesis. Author Contributions {#S5} ==================== Conceived and designed the experiments: JM, MM, NA, SJ, and ML. Performed the experiments: JM, BP, MM, AA, BS, and DK. Analyzed the data: JM, BP, MM, DK, IT, and AK. Wrote the paper: JM and ML. Conflict of Interest Statement {#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. This work was funded by grants from the Serbian Ministry of Science and Technological Development (Grants No. ON175071, ON175069, and ON175103), Serbia, and the Faculty of Medical Sciences, University of Kragujevac (MP 01/14 and MP 02/14). The authors thank Milan Milojevic for excellent technical assistance. [^1]: Edited by: Luigi Daniele Notarangelo, Harvard Medical School, USA [^2]: Reviewed by: Silva Markovic-Plese, University of North Carolina at Chapel Hill, USA; Anne Kathrin Mausberg, Essen University Hospital, Germany [^3]: Specialty section: This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Immunology

1. Introduction {#sec1-biomedicines-08-00069} =============== Replacement therapy using plasma-derived Factor IX (FIX) products is a life-saving treatment for patients with hemophilia B. Both recombinant and plasma-derived FIX show high efficacy in clinical trials \[[@B1-biomedicines-08-00069]\]. Production of FIX normally involves multiple steps. High purity FIX is obtained from prothrombin complex concentrate (PCC), which is a mixture of vitamin K-dependent clotting factors, e.g., factor II (prothrombin), V, VII, IX, and X, and clotting inhibitors, e.g., protein C, Z, and S \[[@B2-biomedicines-08-00069]\]. PCC preparation is a highly complex mixture of proteins and may contain up to 50% of proteins other than FIX \[[@B3-biomedicines-08-00069]\]. Both highly purified FIX and PCC can be used for hemophilia B treatment \[[@B4-biomedicines-08-00069]\]. Also, PCC preparation may be useful for prevention of bleeding due to overdose of oral anticoagulants and liver dysfunctions \[[@B2-biomedicines-08-00069],[@B3-biomedicines-08-00069]\]. As it is with all plasma-derived products, the viral safety of FIX-rich PCC is a critical issue \[[@B5-biomedicines-08-00069]\]. According to current regulations, at least two orthogonal virus clearance steps must be implemented to ensure viral safety of the final product \[[@B6-biomedicines-08-00069]\]. The steps to mitigate virus contamination of FIX and PCC products include donor screening for known blood-borne viruses, i.e., HIV 1--2, HBV, HCV, HAV, and parvo B19; virus inactivation, such as solvent/detergent, mixed chemical inactivation (tri-n-butyl phosphate) and detergent (nonionic, polysorbate, and polyethylene oxide) treatment; and incubating intermediate product in controlled temperature (usually 6 h at 25 °C) \[[@B3-biomedicines-08-00069],[@B7-biomedicines-08-00069],[@B8-biomedicines-08-00069]\]. Dry heat treatment may also be used after lyophilization, e.g., 100 °C for 1 h or 80 °C for 72 h. In the past, steam treatment at 60 °C (190 mbar) for 10 h or 80 °C (375 mbar) for 1 h was reported \[[@B3-biomedicines-08-00069],[@B7-biomedicines-08-00069],[@B8-biomedicines-08-00069]\]. Presently, virus removal nanofiltration has become widely used as a robust and reliable method for ensuring viral safety. Nanofiltration is attractive because it is capable of physically removing all types of viruses from protein solution as opposed to virus inactivation. Several authors have described the application of virus removal nanofiltration for FIX industrial products using Planova 15/20/35N \[[@B9-biomedicines-08-00069],[@B10-biomedicines-08-00069],[@B11-biomedicines-08-00069],[@B12-biomedicines-08-00069],[@B13-biomedicines-08-00069],[@B14-biomedicines-08-00069],[@B15-biomedicines-08-00069]\], Viresolve NFP \[[@B11-biomedicines-08-00069],[@B16-biomedicines-08-00069],[@B17-biomedicines-08-00069],[@B18-biomedicines-08-00069],[@B19-biomedicines-08-00069]\], and Ultipor DV50 filters \[[@B2-biomedicines-08-00069]\]. The conclusion of these studies is that filtration of plasma-derived FIX-rich products is challenging due to the presence of large molecular weight impurities and protein aggregates. It is known that the levels of different impurities are highly dependent on the type of chromatographic separation that was used during plasma treatment \[[@B3-biomedicines-08-00069],[@B18-biomedicines-08-00069],[@B20-biomedicines-08-00069]\]. The most extensively described impurities in vitamin K-dependent clotting factors include inter-α-trypsin-inhibitor (ITI), complement 4b binding protein (C4BP), and vitronectin (VN). C4BP is a large glycoprotein of 570 kDa \[[@B21-biomedicines-08-00069]\] in the shape of an octopus that consists of seven α-chains connected to a single β-chain by disulfide bonds \[[@B22-biomedicines-08-00069],[@B23-biomedicines-08-00069],[@B24-biomedicines-08-00069]\]. The α-chains are responsible for binding C4b, while the β-chain has high-affinity for VN (protein S), forming large complexes \[[@B22-biomedicines-08-00069],[@B25-biomedicines-08-00069]\]. The molecular conformation of C4BP is highly dependent on the surrounding medium composition \[[@B11-biomedicines-08-00069],[@B19-biomedicines-08-00069]\]. In the charged state the α-chains repel each other, thereby occupying much larger volume than the same molecule in the uncharged state. In its open structure form, C4BP would have a diameter of approximately 66 nm since each α-chain arm is 33 nm long \[[@B11-biomedicines-08-00069],[@B22-biomedicines-08-00069]\]. Varying salt concentrations have been shown to affect the compactness of C4BP molecule and thereby the flux properties of the virus removal filter \[[@B19-biomedicines-08-00069]\]. Another large Mw impurity present often in FIX products is ITI. ITI (225 kDa) is a large complex that consists of one light and several heavy chains (H1-H3) covalently linked by a chondroitin sulfate chain. The heavy chains of ITI proteins function as hyaluronic acid (HA) binding proteins, whereas the light chain, also called bikunin, functions as a serine protease inhibitor upon activation \[[@B26-biomedicines-08-00069]\]. The third extensively described impurity is VN \[[@B27-biomedicines-08-00069]\]. The monomeric form of VN has a cryptic hydrophobic pocket, which upon exposure and conformational changes exhibits heparin- and C5b-7 binding activity \[[@B28-biomedicines-08-00069]\]. Normally, only 2% of VN in plasma shows heparin-binding activity but its fraction increases manifold during coagulation \[[@B29-biomedicines-08-00069]\]. VN also presents a free thiol group capable of disulfide bonding \[[@B30-biomedicines-08-00069]\]. When unfolded VN is highly prone to polymerization and may form aggregates with Mw up to 1000 kDa \[[@B28-biomedicines-08-00069]\]. Studies on nanofiltration of FIX products where VN aggregates were detectable confirmed its role as a filter foulant \[[@B27-biomedicines-08-00069]\]. Extensive coverage of various impurities at different intermediate stages during FIX manufacturing is discussed elsewhere \[[@B15-biomedicines-08-00069],[@B18-biomedicines-08-00069]\]. So far successful virus removal filtration of FIX products has been described in the literature only for a limited number of commercial filters. A novel type of virus removal filter paper was developed at Uppsala University, which is produced by adapting traditional paper making technology and consists of 100% cellulose nanofibers \[[@B31-biomedicines-08-00069],[@B32-biomedicines-08-00069]\]. The pore size and flux properties of the filter paper can be controlled, which opens new opportunities to model fundamental aspects of bioprocessing \[[@B32-biomedicines-08-00069],[@B33-biomedicines-08-00069],[@B34-biomedicines-08-00069]\]. The filter paper was previously validated in numerous studies to remove several large and small-size model viruses, including retroviruses (xMuLV, 100 nm) \[[@B35-biomedicines-08-00069]\], parvoviruses (MVM, 20 nm) \[[@B32-biomedicines-08-00069],[@B36-biomedicines-08-00069]\], and model phages (ΦX174, 28 nm) \[[@B37-biomedicines-08-00069],[@B38-biomedicines-08-00069],[@B39-biomedicines-08-00069]\]. Recently, it was shown that this nanocellulose-based virus removal filter paper is useful for bioprocessing human plasma-derived IgG \[[@B40-biomedicines-08-00069]\]. In this article, for the first time the filtration of FIX-rich PCC using a nanocellulose-based virus removal filter paper is described. Furthermore, a two-step size-exclusion nanofiltration process is developed to remove foulants and ensure efficient virus removal filtration of FIX-rich PCC using nanocellulose-based virus removal filter paper. FIX-rich PCC was used as a model for a highly complex plasma-derived product to simulate industrial bioprocesses where impurities may greatly affect product yield and biosafety. 2. Materials and Methods {#sec2-biomedicines-08-00069} ======================== 2.1. Materials {#sec2dot1-biomedicines-08-00069} -------------- Cladophora cellulose was provided by FMC Biopolymer (batch 3095-10; Newark, DE, USA). FIX-rich PCC was provided by National Center for Hematology, Moscow, Russia, as lyophilized powder. Coliphages ΦX174 (ATCC 13706-B1™) and PR772 (BAA-769-B1), and the host bacteria Escherichia coli (Migula) Castellani and Chalmers C (ATCC 13706) and K12 J53-1(R15) \[HER 1221\] (BAA-769) strains were obtained from ATCC (Manassas, VA, USA). Agar (214530) was obtained from BD (Franklin Lakes, NJ, USA). Tryptone (LP0042B) and yeast extract (Oxoid) (LP0021) were obtained from Thermo Fisher Scientific. Phosphate-buffered saline (P4417), 2-mercaptoethanol (M3148), sodium chloride (S5886), sodium phosphate dibasic (71640) and 2-mercaptoethanol (M3148) were purchased from Sigma-Aldrich (Saint Louis, MO, USA). Any kD™ Mini-PROTEAN^®^ TGX Stain-Free™ protein gels (4568125), tris/glycine/SDS running buffer (1610732), 4x Laemmli Sample Buffer (1610747), and Precision Plus Protein™ unstained protein standards (1610363) were purchased from Bio-Rad (Hercules, CA, USA). 2.2. Filter Preparation {#sec2dot2-biomedicines-08-00069} ----------------------- Filters of different thickness were prepared from Cladophora cellulose dispersion (0.1 wt.%) made by microfluidization with 200 μm (twice) and 100 μm hole sized chambers at 1800 bar using LM20 Microfluidizer (Microfluidics, Westwood, MA, USA). Furthermore, the wet cake was made by draining the dispersion over a membrane (Durapore, 0.65 µm hydrophilic PVDF "DVPP", Merck Millipore, Burlington, MA, USA) in a funnel, driven by vacuum. Obtained cellulose cakes were dried at 140 °C to produce pre-filters and 80 °C for filter papers using hot-press (Carver Model 4122CE, Carver, Wabash, IN, USA). 2.3. Dissolution of Factor IX-rich PCC {#sec2dot3-biomedicines-08-00069} -------------------------------------- Lyophilized FIX-rich PCC samples were reconstituted by dissolving in phosphate-buffered saline (PBS). No visible particles could be seen after reconstitution, and the solution was clear and transparent. Upon dissolution, the conductivity and pH values were 15.4 mS cm^−1^ and 7.4, respectively. 2.4. Filtration Setup {#sec2dot4-biomedicines-08-00069} --------------------- Pre-filtration and filtration steps were performed in a 47 mm diameter Advantech KST 47 filter holder. Prior to filtration, the pre- and filter papers were wetted in order to extrude the air by running 20 mL of PBS. The pre-filtration steps with 6 and 11 μm pre-filters were performed at 1 bar transmembrane pressure, and the filtrations with 33 μm filters were carried out at 1 or 3 bar. The permeate solutions were collected and for filtrations of the larger volume, permeate was collected in one or three fractions and saved. 2.5. Scanning Electron Microscopy (SEM) {#sec2dot5-biomedicines-08-00069} --------------------------------------- For top-view images samples were fixed onto aluminum stubs with double-adhesive carbon tape, and for cross-section images the samples were mounted onto aluminum sample holders with screw. Imaging was performed using Zeiss Merlin FEG-SEM instrument (Jena, Germany). To reduce charging effects samples were sputtered with Au/Pd with a sputter coater (Polaron, Ashford, UK) was used. The sputtering settings were 4 × 10^2^ mbar and 35 mA, and the sputtering time was 30 s. 2.6. Cryoporometry by Differential Scanning Calorimetry {#sec2dot6-biomedicines-08-00069} ------------------------------------------------------- Filter paper samples (1.5--2 mg) were soaked into deionized water overnight at room temperature. Water was decanted, and the samples were placed in aluminum crucibles with a lid. Samples were cooled down to 248.15 K (−25 °C) at a rate of 10 K min^−1^ followed by heating to 277.15 K (4 °C) at a rate of 0.7 K min^−1^. Measurements were performed in five replicates. The pore size was calculated according to Landry \[[@B41-biomedicines-08-00069]\]:$$\Delta Τ_{~} = - \frac{19.082}{r_{p} - 1.12} - 0.1207~~$$ where *r~p~* is the radius of pore (nm) and *ΔT* is the difference between the peak maximum for melting of pore-confined water and peak value for melting of bulk water, experimentally determined at 0.6 ± 0.01 °C. 2.7. Dynamic Light Scattering {#sec2dot7-biomedicines-08-00069} ----------------------------- Particle size distribution was obtained from dynamic light scattering (DLS) using a Zetasizer Nano ZS (Malvern, UK) instrument. All experiments were performed in triplicates. 2.8. Polyacrylamide Gel Electrophoresis {#sec2dot8-biomedicines-08-00069} --------------------------------------- Protein separation was performed by reducing polyacrylamide gel electrophoresis (SDS-PAGE). Samples were diluted (1:20 *v*/*v*) with PBS and Laemmli buffer, and boiled for 10 min. Electrophoretic separation was carried out at 270 V with Mini-PROTEAN Tetra Vertical Electrophoresis Cell (Bio-Rad, Hercules, CA, USA). Protein bands were detected with Gel Doc™ EZ System (Bio-Rad, Hercules, CA, USA), and quantified using Image Lab 6.0 analysis software (Bio-Rad). 2.9. Analytical SEHPLC {#sec2dot9-biomedicines-08-00069} ---------------------- Samples were analyzed by size-exclusion high-pressure liquid chromatography using Hitachi Chromaster HPLC-UV system with bioZen 1.8 μm SEC-3 (Phenomenex, Torrance, CA, USA) analytical column. Chromatography was performed with 100 mM sodium phosphate, pH 6.8 mobile phase at 0.3 mL min^−1^ flow rate for 20 min. 2.10. LCMS {#sec2dot10-biomedicines-08-00069} ---------- Equal amounts (20 µg) of protein samples were taken out for digestion. After reduction and alkylation, the proteins were on-filter digested by trypsin using 3 kDa centrifugal filters (Millipore Tullagreen, Ireland) according to a standard operating procedure. Obtained peptides were dried using a speedvac system. Pellets were resolved in 60 μL of 0.1% formic acid and further diluted four times prior to nano-LCMS/MS. Tandem mass spectrometry was performed by applying HCD in the QEx-Orbitrap mass spectrometer (Thermo Finnigan, San Jose, CA, USA), equipped with a reversed-phase C18-column by 35 min long gradient. Database searches were performed using the Sequest algorithm, embedded in Proteome Discoverer 1.4 (Thermo Fisher Scientific, Waltham, MA, USA) against Homo Sapience proteome extracted from Uniprot, Release June 2019 with 95% confidence level per protein. 2.11. Bacteriophage Filtration and Titration {#sec2dot11-biomedicines-08-00069} -------------------------------------------- Coliphages PR772 and ΦX174 were spiked to the pre-filtered solutions to obtain final titer about 10^6^ plaque forming units (PFU) mL^−1^ before filtration was performed. Bacteriophage titer was determined by PFU assay by double agar overlay method. Briefly, ten-fold serially diluted bacteriophage samples were mixed with host *E. coli* strains and melted soft agar, and poured on the surface of prepared hard agar plate, followed by incubation at 37 °C for 5 h. Bacteriophage titer was calculated using Equation (1):$$log_{10}\left( {PFU~{mL}^{- 1}} \right) = log_{10}\left( \frac{N~}{V \cdot {d~}} \right)$$ where *N* is the number of plaques, V is the volume (typically 0.1 mL) of added virus and d is the dilution factor. The virus retention was expressed as *log*~10~ reduction value (*LRV*):$$LRV = log_{10}\frac{PFU_{feed}}{PFU_{permeate}}$$ 3. Results {#sec3-biomedicines-08-00069} ========== 3.1. One-Step 33 μm Filtration of FIX-Rich PCC {#sec3dot1-biomedicines-08-00069} ---------------------------------------------- When the FIX-rich PCC at 20 L m^−2^ volumetric load was filtered through the 33 μm mille-feuille filter paper, a rapid flux decline was observed, e.g., from about 80 L m^−2^ h^−1^ to about 10 L m^−2^ h^−1^ at 3 bar overhead pressure. DLS analysis of the feed and permeate samples revealed that the feed sample showed widely distributed fraction of protein impurities above 70 nm, which were not detectable after filtration as shown in [Figure 1](#biomedicines-08-00069-f001){ref-type="fig"}. Notably, these large-size impurities could not be detected in the volume distribution profiles of the feed sample but only in the intensity distribution plots, which suggests that the original amount of the aggregates is small. In the permeate sample, no particle fractions above 40 nm were detected by DLS. To investigate if significant changes were recorded in the protein molecular weight distribution in the permeate sample, SDS-PAGE analysis was performed, as shown in [Figure 2](#biomedicines-08-00069-f002){ref-type="fig"}. Additional proteomics analysis of the detected bands was not performed as it was outside of the scope of the present work. It is seen in [Figure 2](#biomedicines-08-00069-f002){ref-type="fig"} that all major fractions in the permeate sample were reduced compared to the feed. The observed decrease in total protein fraction is concordant to that reported earlier for PCC product filtered through Planova 15N filter \[[@B14-biomedicines-08-00069]\]. In all, it appears that the large molecular weight protein fractions are the main reason for the observed fouling. 3.2. Development and Validation of Two-Step Size-Exclusion Bioprocess for FIX-Rich PCC Nanofiltration {#sec3dot2-biomedicines-08-00069} ----------------------------------------------------------------------------------------------------- The nanocellulose-based filter paper platform provides the possibilities to relatively easily tailor the pore-size distribution of the filter paper to a specific cut-off value. This could be achieved for instance by varying the thickness of the filter paper. [Figure 3](#biomedicines-08-00069-f003){ref-type="fig"} shows the SEM images of the filters with varying thickness, including their top-view and cross-section. It is seen from the images of the cross-sections of the filters that they indeed feature varying thicknesses. To derive information about the pore size of the filter cryoporometry analysis was performed. [Figure 4](#biomedicines-08-00069-f004){ref-type="fig"} shows the typical CP-DSC curves of the studied samples and the boxplots of the derived pore width modes. Cryoporometry analysis has the benefit that it probes the pores in the wet state, and it is a relatively quick and highly automated and reliable method. In this method, the samples are first frozen to −40 °C and then slowly thawed. As the ice crystals start to melt, there is a detectable endotherm peak. When the water is located inside mesopores (i.e., 2--50 nm pore width), there will be a melting point depression as opposed to bulk water, present outside pores or in macropores (above 50 nm). In our experiments, bulk water melts at around 0.6 °C. The larger the melting point depression, the smaller are the pores. As seen from [Figure 4](#biomedicines-08-00069-f004){ref-type="fig"} there is a trend of decreasing pore width mode with increasing thickness. To assess the particle rejection cut-off for each filter, model probes with 2 different particle sizes were used in the form of bacteriophages, i.e., PR772 (70 nm) and ΦX174 (28 nm) phages, see [Table 1](#biomedicines-08-00069-t001){ref-type="table"} and [Table 2](#biomedicines-08-00069-t002){ref-type="table"}, respectively. These probes provide a highly sensitive tool for assessing the size-dependent rejection capability of the filters with varying thickness, i.e., 6, 11, and 33 μm. The 33 µm mille-feuille filter paper shows the lowest hydraulic flux, i.e., 38 L m^−2^ h^−1^ bar^−1^, and the highest virus removal capacity for both small- and large-size viruses, i.e., LRV ≥5.7. The 6 µm filter in the series exhibits the fastest flux, i.e., 405 L m^−2^ h^−1^ bar^−1^, but poor virus removal capacity, i.e., LRV \<1 and \<2 for 28 nm and 70 nm model phages. The flux and virus removal properties of 11 µm filter are intermediate to the other two filters, wherein the 11 µm filter paper shows high clearance towards 70 nm virus, i.e., \>5.7, and moderate clearance toward 28 nm one, i.e., LRV 3.5--4.5, and hydraulic flux of 125 L m^−2^ h^−1^ bar^−1^. Interestingly, the small-size virus removal capacity of 11 µm filter decreased with increasing load volume, whereas that of 33 µm filter remained unaffected under the experimental conditions. The latter could probably be due to redistribution of flow through the larger pores when the smaller pores become clogged in 11 µm filter paper. 3.3. Two-Step 6 μm/33 μm Filtration of FIX-Rich PCC {#sec3dot3-biomedicines-08-00069} --------------------------------------------------- [Figure 5](#biomedicines-08-00069-f005){ref-type="fig"} shows the permeate flux through the 6 μm/33 μm filtration sequence at 1 bar. Rapid flux decline was observed for the permeate after initial plateau. Please note that the flux of pre-filtrate was so fast that it was not recorded. [Figure 6](#biomedicines-08-00069-f006){ref-type="fig"} shows the DLS results for pre-filtrate and permeate samples for the 6 μm/33 μm filtration sequence at 1 bar. It was observed the fraction of large colloids was not removed by 6 μm pre-filtration. However, no aggregates were observed in the permeate sample after filtration through 33 μm filter. [Figure 7](#biomedicines-08-00069-f007){ref-type="fig"} shows the results of SDS-PAGE analysis of the collected samples. It is seen from the graph that all bands showed decreasing intensity. Even after 6 μm pre-filtration, some decline in the band intensity was observed. The bands for lower Mw fractions, i.e., bands 4--6, were reduced to a greater extent after pre-filtration than those of the larger Mw, i.e., bands 1--3. In the permeate sample all band intensities were further decreased. LCMS analysis suggested that key coagulation factors IX, X, V as well as prothrombin were not removed following the two-step 6 μm/33 μm filtration sequence, as shown in [Appendix](#app1-biomedicines-08-00069){ref-type="app"} [Table A1](#biomedicines-08-00069-t0A1){ref-type="table"}, [Table A2](#biomedicines-08-00069-t0A2){ref-type="table"} and [Table A3](#biomedicines-08-00069-t0A3){ref-type="table"}. Overall, the results from the filtration with 6 μm/33 μm sequence suggest that the large Mw impurities were not removed during the pre-filtration step and, subsequently, caused filter fouling and thereby low product yield during the second step. 3.4. Two-Step 11 μm/33 μm Filtration of FIX-Rich PCC {#sec3dot4-biomedicines-08-00069} ---------------------------------------------------- In another set of experiments, the pre-filtration was performed using 11-μm filter paper followed by filtration with 33 μm filter at 1 bar. [Figure 8](#biomedicines-08-00069-f008){ref-type="fig"} shows the flux data of permeate for 11 μm/33 μm filtration sequence. Increasing the thickness of the pre-filter from 6 to 11 μm significantly affected the results. The flux values for pre-filtration indicated rapid fouling as observed above for 33 μm filtration. However, in the second step of 11 μm/33 μm filtration sequence, i.e., through 33 μm filters, stable flux was observed for the entire processed volume, [Figure 7](#biomedicines-08-00069-f007){ref-type="fig"}. The results contrast starkly those observed for 6 μm/33 μm filtration sequence as shown in [Figure 4](#biomedicines-08-00069-f004){ref-type="fig"}. [Figure 9](#biomedicines-08-00069-f009){ref-type="fig"} shows the results of DLS analysis of the pre-filtrate and permeate samples. It is seen that the fraction of large-size impurities, which was clearly visible in the feed solution, was absent both in the pre-filtrate and permeate fractions of 11 μm/33 μm filtration sequence. The latter suggests that pre-filtration with 11 μm filter paper efficiently removes the large-size impurities, unlike pre-filtration with 6 μm filter paper. Additional SEHPLC analysis was performed on these samples as shown in [Figure 8](#biomedicines-08-00069-f008){ref-type="fig"}. It is seen in the graph that the peak retention times and relative intensities are similar in all three samples except for the early peak at 0.5 min in the feed sample. This peak, which corresponds to the largest protein fraction was not detectable in pre-filtrate and permeate samples. The results of the SDS-PAGE analysis for 11 μm /33 μm filtration sequence are summarized in [Figure 10](#biomedicines-08-00069-f010){ref-type="fig"}. It is seen from the graph that the band intensities were reduced in the pre-filtrate and permeate samples as compared to the feed. It should be noted that in general the band intensities were reduced to a greater extent after pre-filtration with 11 μm filter than with 6 μm filter. The decrease of band intensity levels in the permeate sample passed through the 33 μm filter after 11 μm filtration was much less drastic than that for 6 μm/33 μm filtration sequence. In particular, no significant changes were observed for bands 1, 2, and 4. For bands 3, 5, and 6 some intensity reduction was further detected in the permeate sample. LCMS analysis suggested that key coagulation factors IX, X, V as well as prothrombin were not removed following the two-step 11 μm/33 μm filtration sequence (for details see [Appendix](#app1-biomedicines-08-00069){ref-type="app"} [Table A1](#biomedicines-08-00069-t0A1){ref-type="table"}, [Table A4](#biomedicines-08-00069-t0A4){ref-type="table"} and [Table A5](#biomedicines-08-00069-t0A5){ref-type="table"}). Based on the above results, it was concluded that pre-filtration with 11 µm pre-filter removes the aggregates, which in turn greatly enhances the yield of the 33 µm filtration. To confirm the high virus removal capacity of 33 µm filter, the filter paper was loaded with much larger volume than that tested earlier, i.e., 175 L m^−2^. [Figure 11](#biomedicines-08-00069-f011){ref-type="fig"} shows the result of the large load filtration. Following the filtration, no abrupt filter fouling was detected for the entire processed volume, although some flux decline could be observed ([Figure 11](#biomedicines-08-00069-f011){ref-type="fig"}A). Under the experimental conditions, it is estimated that Vmax of the process will be roughly around 500 L m^−2^, which is a drastic improvement from 20 L m^−2^ when filtering in a single-step process through 33 µm filter paper. Furthermore, the filter paper showed high model small-size virus removal capacity, wherein LRV was ≥ 5 in all collected fractions ([Figure 11](#biomedicines-08-00069-f011){ref-type="fig"}B). In particular, no detectable PFUs were observed at all up to 90 L m^−2^ load volume. In the last fractions only residual breakthrough (1--2 PFUs per agar plate, corresponding to 0.7 PFU mL^−1^) was detected. Thus, it was confirmed that the two-step 11 µm/33 µm filtration provides enhanced throughput and good capacity to remove small-size virus without abrupt fouling even when challenged with a relatively large load. 4. Discussion {#sec4-biomedicines-08-00069} ============= In this article, the filtration of a highly challenging hematologic product was investigated. Considering that FIX-rich PCC inherently consists of many bioactive components and some impurities, the virus removal filtration of this product is difficult without fouling. The virus removal filtration of PCC was previously reported using Ultipor DV50 filters, which are dedicated for removal of large-size viruses but do not ensure viral safety against parvoviruses \[[@B2-biomedicines-08-00069]\]. Filtration of PCC through small-size virus removal filters, e.g., Planova 15N, resulted in nearly 39% total protein loss and reduced FIX and FII activity, which was ascribed to presence of large-size complexes between clotting factors and high molecular weight impurities \[[@B14-biomedicines-08-00069]\]. It was further reported in the same study that filtration of highly purified FIX through Planova 15N not only did not result in the decrease of FIX activity but also improved its purity \[[@B14-biomedicines-08-00069]\]. In this work, in order to achieve high virus removal capacity combined with reduced fouling, a tailored two-step process of filtration with nanocellulose-based filter paper was developed. In particular, sacrificial pre-filters with a thickness of 6 and 11 μm were tested. The increased thickness of the filters resulted in tighter pore structure as detected by cryoporometry. The observed effect is explained as follows and illustrated in [Figure 12](#biomedicines-08-00069-f012){ref-type="fig"}. The mille-feuille filter paper consists of a stratified 3-dimensional network of cellulose nanofibers, producing a mesh-like stricture. The layered structure is illustrated in the side-view panel of [Figure 12](#biomedicines-08-00069-f012){ref-type="fig"}. Considering that the nanofibers are randomly distributed in each layer, the pores, which percolate throughout the entire depth of the filter, become tighter with increasing number of layers. The latter is reflected, e.g., in improved virus clearance properties with increased thickness or enhanced aggregate removal properties. Based on the results of PFU titrations of 27 and 70 nm phage particles, it was concluded that the tested filters show varying particle size rejection threshold as the thickness of the filter is increased. Thus, the observed effect is due to the combination of the receding pore size and depth effects (increased tortuosity). The latter enables using pre-filters with tailored cut-off to remove protein aggregates, which eventually results in improved flux through the dedicated virus removal filter. Overall, the two-step approach presented here is based on the size-exclusion principles and is therefore robust. It could thus further be adapted in the manufacturing of other protein-based pharmaceutics, too, including recombinant proteins wherein impurities in the form of host cell proteins may greatly affect the final yield of the biologics during virus removal nanofiltration. 5. Conclusions {#sec5-biomedicines-08-00069} ============== A two-step process was developed to both enhance filtration capacity (25-fold) and achieve high clearance of small-size viruses (LRV \>5) using appropriate pre-filter paper. Large-size aggregates were the main foulants in the feed solution, and by tailoring the properties of the pre-filters the foulants were efficiently removed. In particular, 11 μm/33 μm filtration was found most suitable. The presented approach could potentially be applied for bioprocessing other protein-based drugs, both derived from plasma and produced by recombinant approaches. The article further provides new insights regarding the mechanism of virus removal in the nanocellulose-based filter paper, highlighting the combined effect of size exclusion and tortuosity of pore network. For LCMS analysis, the services of the Mass Spectrometry Based Proteomics Facility in Uppsala were used. The authors thank Ganna Shevchenko, Alexander Falk, and Jonas Bergquist for assistance in performing LCMS analysis. For research articles with several authors, a short paragraph specifying their individual contributions must be provided. The following statements should be used "Conceptualization, A.M. (Albert Mihranyan) and M.R.; methodology, A.M. (Albert Mihranyan), L.M., A.M. (Athanasios Mantas); supply of material, E.G., A.K.; formal analysis, L.M.; investigation, L.M, A.M. (Athanasios Mantas); resources, A.M. (Albert Mihranyan); data curation, L.M. and A.M. (Athanasios Mantas); writing---original draft preparation, L.M., A.M. (Albert Mihranyan); writing---review and editing, A.M. (Albert Mihranyan) L.M., R.M., A.K.; visualization, L.M.; supervision, A.M. (Albert Mihranyan); project administration, A.M. (Albert Mihranyan); funding acquisition, A.M. (Albert Mihranyan). All authors have read and agreed to the published version of the manuscript. The project was funded by Knut and Alice Wallenberg Foundation (Bridge grant 2018.01141), Swedish Research Council \[Vetenskapsrådet No. 2016-05715\], and EU EIT Health Innovation by Idea program \[VIREPAP 19104\]. EIT Health is supported by the European Institute of Innovation and Technology (EIT), a body of the European Union that receives support from the European Union' s Horizon 2020 Research and innovation program. The corresponding author (A.M. (Albert Mihranyan)) is the inventor behind the IP pertaining to virus removal filter paper. biomedicines-08-00069-t0A1_Table A1 ###### LCMS analysis of FIX-PCC feed solution. Accession Description Score Coverage MW \[kDa\] calc. pI ----------- ---------------------------------------------- --------- ---------- ------------ ---------- P0C0L5 Complement C4-B 1920.89 76.38 192.6 7.27 P0C0L4 Complement C4-A 1898.90 76.03 192.7 7.08 P00734 Prothrombin 1767.89 75.56 70.0 5.90 P19823 Inter-alpha-trypsin inhibitor heavy chain H2 917.12 53.59 106.4 6.86 P19827 Inter-alpha-trypsin inhibitor heavy chain H1 685.72 55.76 101.3 6.79 Q06033 Inter-alpha-trypsin inhibitor heavy chain H3 257.18 40.22 99.8 5.74 P02760 Protein AMBP 199.11 27.56 39.0 6.25 P00740 Coagulation factor IX 134.15 47.29 51.7 5.47 P00742 Coagulation factor X 59.86 38.52 54.7 5.94 P02768 Serum albumin 48.49 28.41 69.3 6.28 P49747 Cartilage oligomeric matrix protein 41.93 14.27 82.8 4.60 P67936 Tropomyosin alpha-4 chain 32.67 39.11 28.5 4.69 P01857 Immunoglobulin heavy constant gamma 1 32.57 35.15 36.1 8.19 P01834 Immunoglobulin kappa constant 30.33 39.25 11.8 6.52 P07225 Vitamin K-dependent protein S 26.15 12.72 75.1 5.67 P0DOY3 Immunoglobulin lambda constant 3 21.63 78.30 11.3 7.24 P51884 Lumican 21.32 19.53 38.4 6.61 B9A064 Immunoglobulin lambda-like polypeptide 5 19.49 41.12 23.0 8.84 P04004 Vitronectin 19.13 14.02 54.3 5.80 P01861 Immunoglobulin heavy constant gamma 4 19.09 20.18 35.9 7.36 Q08380 Galectin-3-binding protein 18.16 9.74 65.3 5.27 P07359 Platelet glycoprotein Ib alpha chain 17.06 8.28 71.5 6.29 P35443 Thrombospondin-4 16.88 6.66 105.8 4.68 P01876 Immunoglobulin heavy constant alpha 1 15.74 24.93 37.6 6.51 P01859 Immunoglobulin heavy constant gamma 2 13.53 21.78 35.9 7.59 P01871 Immunoglobulin heavy constant mu 8.22 4.64 49.4 6.77 Q14515 SPARC-like protein 1 7.09 6.93 75.2 4.81 P04070 Vitamin K-dependent protein C 5.36 7.16 52.0 6.28 O95810 Caveolae-associated protein 2 4.07 4.71 47.1 5.21 P16070 CD44 antigen 3.37 2.83 81.5 5.33 Q99436 Proteasome subunit beta type-7 2.23 8.30 29.9 7.68 P12259 Coagulation factor V 1.95 0.81 251.5 6.05 P07237 Protein disulfide-isomerase 1.69 3.35 57.1 4.87 P10909 Clusterin 1.53 4.01 52.5 6.27 biomedicines-08-00069-t0A2_Table A2 ###### LCMS analysis of FIX-PCC pre-filtrate following filtration with 6 μm filter. Accession Description Score Coverage MW \[kDa\] calc. pI ----------- ----------------------------------------------- --------- ---------- ------------ ---------- P00734 Prothrombin 2584.88 71.86 70.0 5.90 P0C0L5 Complement C4-B 2441.20 77.69 192.6 7.27 P0C0L4 Complement C4-A 2393.34 74.89 192.7 7.08 P19827 Inter-alpha-trypsin inhibitor heavy chain H1 1154.64 54.34 101.3 6.79 P19823 Inter-alpha-trypsin inhibitor heavy chain H2 1136.27 55.50 106.4 6.86 P02760 Protein AMBP 337.33 25.57 39.0 6.25 Q06033 Inter-alpha-trypsin inhibitor heavy chain H3 332.27 43.15 99.8 5.74 P00740 Coagulation factor IX 151.88 54.66 51.7 5.47 P00742 Coagulation factor X 81.34 41.19 54.7 5.94 P02768 Serum albumin 60.16 37.44 69.3 6.28 P01834 Immunoglobulin kappa constant OS=Homo sapiens 48.66 81.31 11.8 6.52 P51884 Lumican 42.56 35.50 38.4 6.61 P49747 Cartilage oligomeric matrix protein 41.14 16.12 82.8 4.60 P01857 Immunoglobulin heavy constant gamma 1 40.75 41.21 36.1 8.19 P67936 Tropomyosin alpha-4 chain 31.98 36.29 28.5 4.69 P01861 Immunoglobulin heavy constant gamma 4 30.25 28.13 35.9 7.36 P01860 Immunoglobulin heavy constant gamma 3 28.46 27.06 41.3 7.90 P04004 Vitronectin OS=Homo sapiens 25.83 17.99 54.3 5.80 P0DOY2 Immunoglobulin lambda constant 2 23.29 67.92 11.3 7.24 P07225 Vitamin K-dependent protein S 21.00 11.83 75.1 5.67 P07359 Platelet glycoprotein Ib alpha chain 20.03 8.28 71.5 6.29 Q08380 Galectin-3-binding protein 17.21 12.82 65.3 5.27 P01876 Immunoglobulin heavy constant alpha 1 15.47 19.55 37.6 6.51 P35443 Thrombospondin-4 14.04 4.79 105.8 4.68 P01859 Immunoglobulin heavy constant gamma 2 12.71 27.91 35.9 7.59 P01024 Complement C3 7.05 1.56 187.0 6.40 P12259 Coagulation factor V 6.08 1.57 251.5 6.05 Q14515 SPARC-like protein 1 5.71 6.93 75.2 4.81 P16070 CD44 antigen 5.36 2.70 81.5 5.33 P13591 Neural cell adhesion molecule 1 5.24 2.56 94.5 4.87 P04070 Vitamin K-dependent protein C 5.14 5.86 52.0 6.28 P07900 Heat shock protein HSP 90-alpha 4.64 4.92 84.6 5.02 P61981 14-3-3 protein gamma 3.79 7.29 28.3 4.89 P63104 14-3-3 protein zeta/delta 3.48 6.94 27.7 4.79 Q99436 Proteasome subunit beta type-7 2.11 8.30 29.9 7.68 P25786 Proteasome subunit alpha type-1 1.66 7.98 29.5 6.61 biomedicines-08-00069-t0A3_Table A3 ###### LCMS analysis of FIX-PCC permeate following two-step filtration with 6 μm/33 μm filters. Accession Description Score Coverage MW \[kDa\] calc. pI ----------- ---------------------------------------------- --------- ---------- ------------ ---------- P00734 Prothrombin 2759.75 73.31 70.0 5.90 P0C0L5 Complement C4-B 2411.85 78.61 192.6 7.27 P0C0L4 Complement C4-A 2379.43 75.80 192.7 7.08 P19823 Inter-alpha-trypsin inhibitor heavy chain H2 1480.56 56.77 106.4 6.86 P19827 Inter-alpha-trypsin inhibitor heavy chain H1 1167.61 54.45 101.3 6.79 Q06033 Inter-alpha-trypsin inhibitor heavy chain H3 352.03 43.60 99.8 5.74 P02760 Protein AMBP 263.80 27.56 39.0 6.25 P00740 Coagulation factor IX 216.46 54.66 51.7 5.47 P00742 Coagulation factor X 90.46 37.09 54.7 5.94 P02768 Serum albumin 62.34 36.45 69.3 6.28 P01857 Immunoglobulin heavy constant gamma 1 47.83 39.09 36.1 8.19 P04004 Vitronectin 45.60 23.64 54.3 5.80 P49747 Cartilage oligomeric matrix protein 43.78 19.82 82.8 4.60 P51884 Lumican OS=Homo sapiens 43.31 42.90 38.4 6.61 P01834 Immunoglobulin kappa constant 40.16 49.53 11.8 6.52 P67936 Tropomyosin alpha-4 chain 39.83 40.32 28.5 4.69 P07359 Platelet glycoprotein Ib alpha chain 31.44 12.12 71.5 6.29 P0DOY2 Immunoglobulin lambda constant 2 29.10 75.47 11.3 7.24 Q08380 Galectin-3-binding protein 28.60 18.80 65.3 5.27 P01861 Immunoglobulin heavy constant gamma 4 28.41 22.02 35.9 7.36 P01876 Immunoglobulin heavy constant alpha 1 23.24 25.78 37.6 6.51 B9A064 Immunoglobulin lambda-like polypeptide 5 22.01 38.79 23.0 8.84 P07225 Vitamin K-dependent protein S 20.55 14.35 75.1 5.67 P01859 Immunoglobulin heavy constant gamma 2 19.80 24.23 35.9 7.59 Q14515 SPARC-like protein 1 19.72 9.19 75.2 4.81 P35443 Thrombospondin-4 14.73 4.79 105.8 4.68 P12259 Coagulation factor V 8.97 2.34 251.5 6.05 P13591 Neural cell adhesion molecule 1 8.03 4.55 94.5 4.87 P12814 Alpha-actinin-1 7.56 3.48 103.0 5.41 P0CG38 POTE ankyrin domain family member I 7.35 3.07 121.2 6.21 P01871 Immunoglobulin heavy constant mu 7.06 8.61 49.4 6.77 P04070 Vitamin K-dependent protein C 5.93 10.63 52.0 6.28 P16070 CD44 antigen 5.80 2.70 81.5 5.33 P61981 14-3-3 protein gamma 5.53 10.93 28.3 4.89 P07900 Heat shock protein HSP 90-alpha 3.08 3.42 84.6 5.02 P01023 Alpha-2-macroglobulin 2.56 1.76 163.2 6.46 Q14185 Dedicator of cytokinesis protein 1 1.99 1.66 215.2 7.56 Q562R1 Beta-actin-like protein 2 1.78 7.71 42.0 5.59 biomedicines-08-00069-t0A4_Table A4 ###### LCMS analysis of FIX-PCC pre-filtrate following filtration with 11 μm filter. Accession Description Score Coverage MW \[kDa\] calc. pI ----------- ---------------------------------------------- --------- ---------- ------------ ---------- P00734 Prothrombin 2458.17 67.20 70.0 5.90 P0C0L5 Complement C4-B 2389.47 80.68 192.6 7.27 P0C0L4 Complement C4-A 2352.05 79.59 192.7 7.08 P19823 Inter-alpha-trypsin inhibitor heavy chain H2 1268.03 56.87 106.4 6.86 P19827 Inter-alpha-trypsin inhibitor heavy chain H1 1122.57 59.50 101.3 6.79 Q06033 Inter-alpha-trypsin inhibitor heavy chain H3 352.56 40.00 99.8 5.74 P02760 Protein AMBP 285.56 27.56 39.0 6.25 P00740 Coagulation factor IX 182.33 57.27 51.7 5.47 P00742 Coagulation factor X 80.33 41.19 54.7 5.94 P01834 Immunoglobulin kappa constant 54.45 54.21 11.8 6.52 P49747 Cartilage oligomeric matrix protein 53.77 24.31 82.8 4.60 P01857 Immunoglobulin heavy constant gamma 1 45.54 43.33 36.1 8.19 P02768 Serum albumin 44.26 25.78 69.3 6.28 P07359 Platelet glycoprotein Ib alpha chain 33.83 16.56 71.5 6.29 P01860 Immunoglobulin heavy constant gamma 3 33.53 31.83 41.3 7.90 P0DOY3 Immunoglobulin lambda constant 3 33.11 70.75 11.3 7.24 P01861 Immunoglobulin heavy constant gamma 4 30.81 23.85 35.9 7.36 P04004 Vitronectin 27.21 23.64 54.3 5.80 P51884 Lumican 26.62 35.50 38.4 6.61 P01876 Immunoglobulin heavy constant alpha 1 25.67 29.18 37.6 6.51 Q08380 Galectin-3-binding protein 25.15 17.78 65.3 5.27 P67936 Tropomyosin alpha-4 chain 25.00 44.35 28.5 4.69 P01859 Immunoglobulin heavy constant gamma 2 17.12 25.77 35.9 7.59 P07225 Vitamin K-dependent protein S 16.75 9.62 75.1 5.67 P35443 Thrombospondin-4 16.66 5.52 105.8 4.68 P12259 Coagulation factor V 10.18 1.98 251.5 6.05 P04070 Vitamin K-dependent protein C 7.36 5.86 52.0 6.28 P16070 CD44 antigen 5.64 2.70 81.5 5.33 P13591 Neural cell adhesion molecule 1 5.50 1.75 94.5 4.87 P22105 Tenascin-X 4.42 2.33 458.1 5.17 Q99436 Proteasome subunit beta type-7 4.03 8.30 29.9 7.68 P12814 Alpha-actinin-1 3.68 2.35 103.0 5.41 P27348 14-3-3 protein theta 3.53 6.53 27.7 4.78 P07900 Heat shock protein HSP 90-alpha 1.73 3.14 84.6 5.02 biomedicines-08-00069-t0A5_Table A5 ###### LCMS analysis of FIX-PCC permeate following two-step filtration with 11 μm/33 μm filters. Accession Description Score Coverage MW \[kDa\] calc. pI ----------- ---------------------------------------------- --------- ---------- ------------ ---------- P0C0L5 Complement C4-B 2266.18 76.26 192.6 7.27 P0C0L4 Complement C4-A 2219.29 76.26 192.7 7.08 P00734 Prothrombin 2011.89 71.54 70.0 5.90 P19823 Inter-alpha-trypsin inhibitor heavy chain H2 1452.50 55.81 106.4 6.86 P19827 Inter-alpha-trypsin inhibitor heavy chain H1 1205.49 54.23 101.3 6.79 Q06033 Inter-alpha-trypsin inhibitor heavy chain H3 256.28 46.97 99.8 5.74 P02760 Protein AMBP 213.99 25.57 39.0 6.25 P00740 Coagulation factor IX 114.46 45.55 51.7 5.47 P00742 Coagulation factor X 74.31 36.07 54.7 5.94 P02768 Serum albumin 46.51 27.91 69.3 6.28 P01857 Immunoglobulin heavy constant gamma 1 45.77 43.33 36.1 8.19 P49747 Cartilage oligomeric matrix protein 34.16 20.08 82.8 4.60 P01834 Immunoglobulin kappa constant 32.23 49.53 11.8 6.52 P67936 Tropomyosin alpha-4 chain 27.70 37.90 28.5 4.69 P01860 Immunoglobulin heavy constant gamma 3 27.37 25.99 41.3 7.90 P51884 Lumican 25.59 39.94 38.4 6.61 P01861 Immunoglobulin heavy constant gamma 4 24.73 26.61 35.9 7.36 P01859 Immunoglobulin heavy constant gamma 2 23.14 32.82 35.9 7.59 P04004 Vitronectin 22.83 23.43 54.3 5.80 P07359 Platelet glycoprotein Ib alpha chain 21.45 9.82 71.5 6.29 P07225 Vitamin K-dependent protein S 20.38 10.65 75.1 5.67 P0DOY2 Immunoglobulin lambda constant 2 17.99 37.74 11.3 7.24 Q08380 Galectin-3-binding protein 17.44 15.73 65.3 5.27 P01876 Immunoglobulin heavy constant alpha 1 17.06 25.78 37.6 6.51 P35443 Thrombospondin-4 9.76 4.79 105.8 4.68 P12259 Coagulation factor V 8.36 2.11 251.5 6.05 P04070 Vitamin K-dependent protein C 7.60 8.89 52.0 6.28 P61981 14-3-3 protein gamma 5.93 10.53 28.3 4.89 P16070 CD44 antigen 4.96 2.70 81.5 5.33 Q99436 Proteasome subunit beta type-7 4.05 8.30 29.9 7.68 Q14515 SPARC-like protein 1 4.05 3.77 75.2 4.81 P07900 Heat shock protein HSP 90-alpha 2.46 2.60 84.6 5.02 P01023 Alpha-2-macroglobulin 1.97 1.49 163.2 6.46 P22105 Tenascin-X OS=Homo sapiens 1.66 0.66 458.1 5.17 P13591 Neural cell adhesion molecule 1 0.00 1.75 94.5 4.87 Q5UIP0 Telomere-associated protein RIF1\] 0.00 1.62 274.3 5.52 {#biomedicines-08-00069-f001} {#biomedicines-08-00069-f002} {#biomedicines-08-00069-f003} {#biomedicines-08-00069-f004} {#biomedicines-08-00069-f005} {#biomedicines-08-00069-f006} {#biomedicines-08-00069-f007} {#biomedicines-08-00069-f008} {#biomedicines-08-00069-f009} {#biomedicines-08-00069-f010} {#biomedicines-08-00069-f011} {#biomedicines-08-00069-f012} biomedicines-08-00069-t001_Table 1 ###### LRVs for 70 nm (PR772) bacteriophages filtered through 6, 11, and 33 μm filter papers. The results represent the virus clearance data of virus-spiked PBS. Green color code denotes high virus clearance LRV \> 5; Yellow denotes moderate virus clearance (2 \< LRV \< 5); and pink denotes low virus clearance (LRV \< 1). Thickness (μm) Load Volume ---------------- ------------- ------------- ------------- 6 1.6 ± 0.2 1.3 ± 0.1 1.1 ± 0.3 11 \>5.5 ± 0.2 \>5.5 ± 0.2 \>5.5 ± 0.2 33 \>5.7 ± 0.2 \>5.7 ± 0.2 \>5.7 ± 0.2 biomedicines-08-00069-t002_Table 2 ###### LRVs for 28 nm (ΦΧ174) bacteriophages filtered through 6, 11, and 33 μm filter papers. The results represent the virus clearance data of virus-spiked PBS. Green color code denotes high virus clearance LRV \> 5; Yellow denotes moderate virus clearance (2 \< LRV \< 5); and pink denotes low virus clearance (LRV \< 2). Thickness (μm) Load Volume ---------------- ------------- ------------- ----------- 6 0.7 ± 0.3 0.8 ± 0.3 0.8 ± 0.2 11 4.5 ± 0.5 3.7 ± 0.4 3.5 ± 0.7 33 \>5.7 ± 0.4 \>5.7 ± 0.4 5.7 ± 0.4

Introduction {#Sec1} ============ Silage is one of the most common feedstuff for ruminants in Europe. In 2015, the area of land harvested in Poland for this purpose exceeded 500.000 ha, and is still increasing each year (Central Statistical Office of Poland [@CR7]). Silage is mostly composed of maize, grass, clover, sugar beet tops, alfalfa and milo (Storm et al. [@CR40]). The ensiling process allows for preservation of fodder for livestock for longer periods of time, without degradation, and with minimum loss of nutrients (Tangni et al. [@CR42]). It enables its use as forage during the periods of feed scarcity (Alonso et al. [@CR3]). However, silage can become contaminated with toxigenic fungi, either pre-harvest (e.g. *Alternaria* spp. and *Fusarium* spp.), post-harvest (e.g. *Penicillium* spp.) (Rasmussen et al. [@CR34]) or at both times (e.g. *Aspergillus* spp*.*). The occurrence of these fungal contaminants depends on many factors, such as climate, storage conditions and agricultural practice (Storm et al. [@CR40]). Under specific conditions, growth of toxigenic moulds can result in the production of mycotoxins. The intake of these secondary metabolites can exert several adverse effects on livestock animals (Scudamore and Livesey [@CR38]). Therefore, the occurrence of mycotoxins in livestock animals is one of the most serious health threats in agriculture. Moreover, production of feedstuff without any mycotoxins is very difficult (Wambacq et al. [@CR44]). Hundreds of mycotoxins are known of (Berthiller et al. [@CR4]), but European Union regulation on feed has so far been established only for aflatoxins (AFB~1~, AFB~2~, AFG~1~, AFG~2~) by Directive 32/2002 (European Communities [@CR23]), additional "guidance values" have been published by the European Commission for several other compounds, namely deoxynivalenol (DON), fumonisins (FB~1~, FB~2~), ochratoxin A (OTA), zearalenone (ZEN) (European Commission [@CR20]) and for T-2 and HT-2 toxins (European Commission [@CR22]). Because there are no specific regulations on mycotoxins in silage (e.g. grass silage, only for maize-based product guidance value is available), currently recommended levels for animal feed could also be considered as guidelines for silage (Cheli et al. [@CR8]). Regarding to DON and ZEN is recommended not to exceed 12 mg/kg and 3 mg/kg, respectively. In recent years, researchers have additionally paid more attention to the presence of "emerging mycotoxins" in food and feed, especially for the enniatins (ENNs) and beauvericin (BEA). Data on the toxicity and occurrence of emerging mycotoxins are limited, and further investigation of these compounds is needed for a proper risk assessment. Nevertheless, there have been some studies describing their potential implications for food safety (EFSA [@CR16]). Based on recent scientific opinion of European Food Safety Authority (EFSA), some "opinion" papers about the risk to human and animal health to the presence of regulated, modified and emerging mycotoxins have been published (EFSA [@CR13], [@CR14], [@CR15], [@CR17], [@CR18], [@CR19]). The determination of mycotoxins in silage is also an analytical challenge. Silage has a complex matrix that contains many compounds, such as organic acids, sugars, chlorophyll and others, that are difficult to remove using sample extract preparation (Rasmussen et al. [@CR34]). Hence, it is necessary to develop a suitable method of analysis for mycotoxins in silage. Altogether, several multi-analyte methods for the simultaneous determination of mycotoxins in silage do exist, and have been described in the literature so far, mostly based on liquid chromatography with tandem mass spectrometry (LC-MS/MS). This core technique can provide the highest sensitivity and specificity, enabling detection of low levels of mycotoxins in various samples and reducing the number of sample preparation steps and analysis time (Wang et al. [@CR45]). Until now, most researchers paid attention on mycotoxins occurrence in grains, cereals (Monbailu et al. [@CR32]; Schenzel et al. [@CR36]; Kovalsky et al. [@CR29]; Abdallah et al. [@CR1]) and maize silage (Dagnac et al. [@CR9]; Gallo et al. [@CR24]; Grajewski et al. [@CR26]; Kosicki et al. [@CR28]; Storm et al. [@CR41]; Zachariasova et al. [@CR46]). Only few studies on the occurrence of mycotoxins in grass silage have been published (Driehuis et al. [@CR10]; McElhinney et al. [@CR31]). In the aforementioned studies, the research was focused mainly on regulated mycotoxins, with *Fusarium* toxins the most frequently detected compounds. Data on emerging toxins are scarce, and further surveys are needed for a proper risk assessment. Therefore, the aim of this study was to assess the contamination levels of silage in Poland, and to study possible correlations between different toxins. Materials and methods {#Sec2} ===================== Chemicals and reagents {#Sec3} ---------------------- Acetonitrile (ACN, analytical grade), methanol (MeOH, LC-MS grade), acetic acid and C18 bulk sorbent were purchased from J.T. Baker of Avantor Performance Materials (Netherlands). Formic acid and ammonium acetate (LC/MS grade) were supplied by Sigma-Aldrich (Germany). Magnesium sulphate was obtained from Chempur (Poland). Water was purified by a Milli-Q apparatus (USA). Standard solutions {#Sec4} ------------------ From Sigma-Aldrich (Germany), the standards were obtained for AFB~1~, AFB~2~, AFG~1~, AFG~2~, 3-acetyldeoxynivalenol (3-AcDON), 15-acetyldeoxynivalenol (15-AcDON), citrinin (CIT), beauvericin (BEA), diacetoxyscirpenol (DAS), DON, enniatins A (ENN A), A~1~ (ENN A~1~), B (ENN B), B~1~ (ENN B~1~), FB~1~ and FB~2~, fusarenon X (FUS-X), nivalenol (NIV), OTA, sterigmatocystin (STC), HT-2, T-2, ZEN and β-zearalenol (β-ZEL). All standards were stored according to their manufacturer's recommendations. Primary standard stock solutions were prepared: in acetonitrile for 3-AcDON, 15-AcDON, DON, FUS-X, HT-2, STC, T-2 and ZEN; in methanol for AFs, ENNs, NIV, OTA and β-ZEL and in 50% solution of ACN in H~2~O for FB~1~ and FB~2~. The stock solutions were used to prepare working standard solutions containing the 24 analytes in concentrations corresponding to the lowest regulatory levels or guidance levels (GL) in feedstuffs (Supplementary Table [S2](#MOESM1){ref-type="media"}). Samples {#Sec5} ------- One hundred twenty visibly mould-free samples of silage, consisting of maize (87) and grass (33), were collected from 16 provinces (voivodeships) of Poland, with eight samples coming from each region (Fig. [1](#Fig1){ref-type="fig"}). Samples were collected between July and December of 2015 by the Veterinary Inspection officers working with feed manufacturers. The types of silage sampled were representative of the different regions of Poland, and were taken in compliance with European regulations (European Commission [@CR21]), as part of a national monitoring programme. The samples, weighing about 5 kg each, were divided separately into 1-kg subsamples, frozen and chopped (grass). Silage was homogenised (using a Waring Blender 8010EB, USA), and stored in the dark at − 20 °C until the date of analysis.Fig. 1Map of Poland illustrating localization of surveyed samples Sample preparation {#Sec6} ------------------ The protocol for sample preparation was adapted from a previous study (Jedziniak et al. [@CR27]), with some modifications. Five grams of sample was placed into a glass tube and extracted using a 20 ml of mixture consisting of acetonitrile:water:formic acid (79:20:1, *v*/*v*/*v*), with a homogeniser (Polytron PT 3000, Switzerland) for 2 min (2240 × g). The sample was then put into storage for 12 h at 4 °C. Subsequently, the whole sample was put into a 50-ml polypropylene tube and shaken vertically (200 cycles/min) for 30 min. The sample was centrifuged (2643 × g, 15 min), and 2 ml of supernatant was transferred to a plastic test tube containing MgSO~4~ (150 mg) and C18 bulk sorbent (50 mg). The aliquot was immediately shaken vigorously for 30 s then centrifuged (2643 × g, 15 min). The extract (1 ml) was evaporated (40 °C) to dryness in a gentle stream of nitrogen. The dry residue was reconstituted with 500 μl of mobile phase A and 500 μl of mobile phase B (see the "[LC-MS/MS conditions](#Sec7){ref-type="sec"}" section), and then sample was transferred to a 1.5-ml polypropylene tube for centrifuging for 30 min (16,602 × g). The extract was put into an autosampler vial and 5 μl was injected for UPLC-MS/MS analysis. For each analysis series, matrix-matched calibration curves were prepared at three levels (0.5 × GL, 1 × GL, 2 × GL), for both the maize and grass silage. LC-MS/MS conditions {#Sec7} ------------------- Chromatographic separation was performed for 16 min on a Kinetex BiPhenyl column (100 × 2.1 mm; particle size 2.6 μm), coupled with a BiPhenyl security guard cartridge (Phenomenex, Torrance, CA, US). The column oven temperature was set to 40 °C. The gradient elution had flowrate of 0.3 ml/min. Mobile phase A consisted of 10 mM ammonium acetate and 0.1% acetic acid/MeOH (95:5, *v*/*v*). Mobile phase B consisted of 10 mM ammonium acetate and 0.1% acetic acid/MeOH (5:95, *v*/*v*), was used as follows: (1) linear gradient to 95% solvent B 0--9 min; (2) 95% solvent B held from 9 to 13 min and (3) column reconditioning with the initial composition of the mobile phase at 13--15.9 min. The analyses were performed with a Nexera X2 system, coupled with a LCMS-8050 triple quadrupole mass spectrometer (Shimadzu, Japan), equipped with an electrospray and operated in positive (ESI +) and negative (ESI −) modes using fast polarity switching, controlled by LabSolution 5.60 SP2 software. Two multiple reaction monitoring (MRM) transitions for each analyte were monitored (Supplementary Table [S1](#MOESM1){ref-type="media"}). The time of analysis was divided by time segments (retention time ± 2 min), each acquiring different MRM transitions. The following parameters were used: resolution Q1 and Q3 unit; nebulising gas flow, 2 L/min; heating gas flow, 10 L/min; drying gas flow, 10 L/min; interface temperature, 300 °C; desolvation line temperature, 250 °C and heat block temperature, 400 °C. Method validation {#Sec8} ----------------- During the validation process, the following parameters were determined for the maize and grass silage: linearity range; limit of detection (LOD, μg/kg); limit of quantification (LOQ, μg/kg); recovery (REC, %) and repeatability expressed as relative standard deviation (RSD, %). LOD and LOQ were calculated based on a signal-to-noise (S/N) ratio of peaks (LOD, S/N = 3; LOQ, S/N = 10). The results were checked by analysis of the background noise of 20 different pseudo-blank silage samples (Schaechtele and Robouch [@CR35]). Repeatability was determined using pseudo-blank samples of maize and grass silage spiked at 1 GL (Supplementary Table [S2](#MOESM1){ref-type="media"}), with a working solution in six repetitions over two different days. For the recoveries study, the same samples were used and compared with concentration of standard solution. For the linearity range, five-point matrix-match calibration curves were prepared by spiking pseudo-blank samples at different levels (0.25 GL, 0.5 GL, 1 GL, 2 GL, 5 GL). Additionally, according to Matuszewski et al. ([@CR30]), the matrix effect (ME) for each mycotoxins was evaluated as a ratio of the concentration of pre-spiked and post-spiked samples (in three repetitions) at 1 GL. Statistical analysis {#Sec9} -------------------- The correlations between mycotoxin concentrations were calculated using STATISTICA, version 10 (StatSoft, Inc. 2011), with a Spearman correlation test used for this purpose. The correlation was considered significant at a level of *p* = 0.05. To assess the significance of the differences in the results between concentrations of mycotoxins in maize and grass silage, a Mann-Whitney rank sum test was performed (*p* value of \< 0.05 was regarded as significant). Results and discussion {#Sec10} ====================== Method validation {#Sec11} ----------------- The LC-MS/MS method was successfully validated for all analytes for maize and grass silage (Supplementary Table [S3](#MOESM1){ref-type="media"}). The obtained results demonstrated sufficient linearity, with *R*^2^ above 0.98 for most of the analytes in both matrixes. Calculated recoveries ranged between 70 and 120% for 19 and 13 out of compounds for maize and grass silage, respectively. The RSD measured during the repeatability study did not exceed 30%. LODs and LOQs ranged from 0.06 to 15.0 μg/kg and 0.20 to 50.0 μg/kg, respectively. The lowest LOQ was obtained for ENNs, STC and BEA, both for maize and grass silage. In the field of ubiquitous contaminants, blank samples may not always be available. As an alternative, low contaminated samples (pseudo-blanks) were used in this study. The evaluation of ME for both matrixes demonstrated large variation between analytes (Supplementary Fig. [S1](#MOESM1){ref-type="media"}). In the case of maize silage, 75% of the compounds fell into a range of 70--120%, and in turn, in the grass silage, only 45% of the analytes were in this range. For example, DON in maize showed a slight enhancement effect (109%), while grass had a matrix suppression effect (62%). The observed results (recovery and ME) indicated the need to quantitate mycotoxins in various commodities, by preparing matrix-matched calibration curves for both matrixes, in each analysis. This has also been pointed out by other researchers (Rasmussen et al. [@CR34]; Dzuman et al. [@CR11]; Dagnac et al. [@CR9]). Frequency and levels of mycotoxins in the silage {#Sec12} ------------------------------------------------ The relatively high frequency of mycotoxins as obtained in our study was because of the low LOQs obtained by the LC-MS/MS method used for analysis. In general, the results are generally in line with other surveys' results in our region (Grajewski et al. [@CR26]; Zachariasova et al. [@CR46]). Overall, the data revealed the presence of 15 and 12 different analytes in maize and grass silage, respectively. BEA, DON, HT-2, ENNs (ENN A, ENN A~1,~ ENN B, ENN B~1~,) NIV and ZEN (Fig. [2](#Fig2){ref-type="fig"}) were frequently found in all samples. Detailed information on the concentrations and prevalence of the detected toxins is compiled in Supplementary Table [S4](#MOESM1){ref-type="media"}. With respect to maize by-products, all of the regulated mycotoxins were below the EU guidance values (European Commission [@CR20]). The *Fusarium* toxins DON and ZEN were amongst the most frequently encountered mycotoxins in maize silage and were found in 82% and 57%, of the samples, with average concentrations at levels of 447 μg/kg (DON) and 82.4 μg/kg (ZEN). Nearly half of the positive samples contained less than 200 μg/kg and 100 μg/kg, respectively (Supplementary Fig. [S2](#MOESM1){ref-type="media"}). Similar results were obtained by Kosicki et al. ([@CR28]) who stated that DON and ZEN were the most frequent toxins, and detected in 86% and 88% of positive-maize samples, respectively. In contrast, the results of Storm et al. ([@CR41]) are contrary to our data, as the authors determined DON in only 6% of examined samples, compared to the 82% reported in our study.Fig. 2Box plot presentation of mycotoxins levels in maize and grass silage Maize and grass silage showed qualitative and quantitative differences with regard to mycotoxin contamination. Except for STC, all compounds were more often found in the maize silage than in grass silage (Fig. [3](#Fig3){ref-type="fig"}). This is probably a result of the fact that fungi and other pathogens can easily survive on the maize crops, which are richer in necessary proteins and polysaccharides than the grass (Zachariasova et al. [@CR46]). The differences between concentrations of mycotoxins in these two types of silage were statistically significant (*p* \< 0.05) for BEA, DON, ENNs, FB1, HT-2, NIV, STC and ZEN. In our study, DON and ZEN were detected in grass silage at a frequency of 37% and 3%, respectively. In the literature, information on the occurrence of mycotoxins in grass silage is scarce, especially for *Fusarium* toxins. In the studies conducted by Skladanka et al. ([@CR39]), the content of several mycotoxins in grass silage was assessed, with maximum content of DON at 167 μg/kg and ZEN at 66.9 μg/kg. These findings support our observation that DON could be present in grass silage and is probably produced by *Fusarium* species during their growth in the field. Moreover, Cavallarin et al. ([@CR6]) suggest that *Fusarium* mycotoxins could be produced within silage, as they detected ZEN in high concentration in grass silage (above 300 μg/kg), whereas Driehuis et al. ([@CR10]) detected ZEN in 6% of surveyed samples of grass silage. The authors verified that the occurrence of toxins in maize silage is higher than in grass or wheat silage. Low level of ZEN was also quantified in grass silage in the study (McElhinney et al. [@CR31]) (mean 53 μg/kg) and was the only EU-regulated mycotoxin detected in surveyed samples. ZEN was found in the 43% of unfermented hay in the German study (Schollenberger et al. [@CR37]).Fig. 3Frequency of mycotoxins in maize and grass silage FUS-X and NIV are 8-ketotrichothecenes structurally related to DON. The mean level of NIV was significantly higher in grass than in maize silage, namely 4473 μg/kg versus 544 μg/kg, respectively. In four samples, the concentration of NIV exceeded a value of 5000 μg/kg; however, these samples constituted only a small percentage of silage (3.33% of positives). Several authors have previously reported the occurrence of NIV in maize silage (Oldenburg and EIIner [@CR33]; Schollenberger et al. [@CR37]; Storm et al. [@CR41]). The average level in these studies ranged from 263 to even 1612 μg/kg. In hay, Schollenberger et al. ([@CR37]) reported a frequency of 4% with an average NIV concentration level of 131 μg/kg. In our survey, FUS-X was detected both in maize and grass silage at similar average concentration 92 and 59 μg/kg, respectively. Lastly, Zachariasova et al. ([@CR46]) noted that the mean concentration of FUS-X in maize silage was 77 μg/kg, while in grass was not determined. The T-2 and HT-2 toxins (type A trichothecenes) were observed only in maize silage, with a frequency of 3% and 51% in mycotoxin-positive samples, respectively. Moreover, concentrations of T-2 toxin constituted only 10% of HT-2 toxin's average content (5.90 and 43.2 μg/kg, respectively). Generally, our results are in line with Grajewski et al. ([@CR26]), especially in cases of high content of HT-2 toxin (45.6 μg/kg) compared to a lower T-2 toxin content (3.10 μg/kg). This could be explained by the fact that T-2 toxin is rapidly metabolised to HT-2 toxin during the fermentation process in silage. Regarding AFs and OTA, these toxins were not present in any sample, similarly to the findings of Driehuis et al. ([@CR10]) and Zachariasova et al. ([@CR46]). AFs and OTA are produced by storage fungi (*Aspergillus* or *Penicillium* fungi) under favourable storage conditions, such as excessive humidity. Ensiling maize limits the available oxygen and water content, which probably prevents storage moulds from growing. However, in Europe, a few authors have reported the occurrence of AFB~1~ in maize silage (Garon et al. [@CR25]; Tsiplakou et al. [@CR43]). The differences in occurrence of the main mycotoxins detected in central and eastern Europe (DON, NIV and ZEN) could be caused by a climatic differences. The warmer, more humid Mediterranean climate creates favourable conditions for the growth of moulds producing, for example, AFB~1~. Of the so-called emerging mycotoxins, BEA was the most commonly detected, with a presence in 108 samples (87%) and average and maximum concentrations of 35.8 μg/kg and 1309 μg/kg, respectively. Nevertheless, the mean content of this toxin in positive samples was low---less than 50 μg/kg in 85% of the samples contaminated with BEA (Supplementary Fig. [S2](#MOESM1){ref-type="media"}). Finally, ENN A, ENN A~1~, ENN B and ENN B~1~ were the most prevalent toxins in the investigated silage, with a presence in 66%, 71%, 89% and 78% of tested samples, respectively. ENNs were two times often detected in maize than grass silage (88% and 43%, respectively). The highest concentrations were found for ENN B (344 μg/kg). However, in many cases, ENNs were found in low concentrations, mostly less than 10 μg/kg (Supplementary Fig. [S2](#MOESM1){ref-type="media"}). Our results for BEA and the four ENNs are similar as reported by other authors (Dagnac et al. [@CR9]), who reported that the most frequently detected toxin was ENN B (51%), in an average concentration of 393 μg/kg. The high prevalence of BEA and ENNs has already been described by McElhinney et al. ([@CR31]), who state that these toxins to be the most prevalent in pit and bale silage (less than 50% and 60% of samples, respectively). Still, no information is currently available on the adverse effects of ENNs and BEA on animal health, or their possible combined effects (EFSA [@CR16]). Co-occurrence of the detected mycotoxins {#Sec13} ---------------------------------------- Statistical analysis confirmed our finding that samples with a relatively high concentration of DON (699 μg/kg) were often contaminated with FUS-X (113 μg/kg), NIV (615 μg/kg) or ZEN (444 μg/kg) (Table [1](#Tab1){ref-type="table"}). Moreover, in few cases, DON was quantified simultaneously with its analogue---3-AcDON; however, the ratio 3-AcDON/DON was below 4%. Co-contamination of maize with DON and it analogues (3-AcDON, 15-AcDON and NIV) was already reported (Oldenburg and EIIner [@CR33]). Our results are similar as that reported by Eckard et al. [@CR12], who rarely found 3-AcDON and only in samples with high total *Fusarium* toxins concentration.Table 1Co-contamination of selected silage samples having the highest overall toxin concentrationNIV \[μg/kg\]DON \[μg/kg\]BEA \[μg/kg\]ZEN \[μg/kg\]FB1 \[μg/kg\]ENA B \[μg/kg\]FB2 \[μg/kg\]FUS-X \[μg/kg\]HT-2 \[μg/kg\]ENA B1 \[μg/kg\]ENA A1 \[μg/kg\]3-AcDON \[μg/kg\]STC \[μg/kg\]T2 \[μg/kg\]ENA A \[μg/kg\]Sample 01\*123855334.650.2--15.2------9.078.92--14.8--9.14Sample 02\*198959614063.8--53.6----49.519.525.1------4.73Sample 03\*103436912.9----16.5----9.413.351.54--------Sample 04\*\*14,26215820.8----4.64------5.476.24--22.0--8.57Sample 05\*2165140435.364.891.967.965.0--16.216.29.0325.1----0.54Sample 06\*1413434711022842.466.3----78.114.211.0------2.58Sample 07\*37550321.8235--72.1----12.810.46.64------1.10Sample 08\*4749837.7875.8--29.6------6.243.83------0.34Sample 09\*46888716.91158.7928.0--79.868.97.435.42------1.30Sample 10\*13574912.02149.8438.1----22.33.251.10--------Sample 11\*23369979.397.436789.433338.891.013.78.1124.9--10.73.46Sample 12\*26622431.320.729910126732.327.357.251.2------6.68Sample 13\*955221022817388.482.746.895.187.321.717.737.2--2.855.72Sample 14\*21719759.219.626.917.6----24.67.889.44------2.77Sample 15\*22893933.617518.182.7--10010710.74.09------0.82Sample 16\*61569941.7444--36.3--1133.851.69------0.52Sample 17\*3163546.4624.9--24.5----60.34.902.54------0.67Sample 18\*26123530.4103--41.3----79.29.494.02------0.68\*maize silage, \*\*grass silage The outcome described above clearly demonstrates that the occurrence of fungi's secondary metabolites in silage is relatively high, and that some samples were co-contaminated with several toxins, although at low concentration. All of the samples contained at least one mycotoxin, 61% of the samples were contaminated with at least five toxins (Supplementary Fig. [S3](#MOESM1){ref-type="media"}). Regarding the co-occurrence of major *Fusarium* mycotoxins (BEA, DON, ENNs, HT-2, NIV and ZEN), 24 of our 120 samples contained all of these compounds (Supplementary Fig. [S4](#MOESM1){ref-type="media"}). The observation is in agreement with the studies of Zachariasova et al. ([@CR46]), who noted co-contamination of multiple mycotoxins in maize silage with DON, ENNs, BEA and ZEN. Those authors reported each tested sample to be positive for at least one mycotoxin at a quantifiable level, with the simultaneous presence of DON and ENNs in the positive-surveyed samples. In our study, a significant number of silage samples (42%) both DON and ZEN were present simultaneously and a high correlation of co-occurrence between these toxins in maize silage was also noted (*r* = 0.74, *p* \< 0.05) (Supplementary Table [S5](#MOESM1){ref-type="media"}). The positive correlation between DON and ZEN in maize silage was previously observed by Kosicki et al. ([@CR28]). Our findings on the co-occurrence between of DON and ZEN are partially in disagreement with the study conducted by Borutova et al. ([@CR5]), who observed a positive correlation in silage only between FB~2~ and ZEN, and between FB~1~ and FB~2~. This could probably be explained by the completely different weather conditions in their Asia-Oceania sample region, with preferable conditions for *Fusarium* species to produce FB~1~ and FB~2~, then DON. However, in a different matrix (raw maize samples), they found high positive correlation between DON and ZEN. In the case of the four ENNs, they were simultaneously present in 61% of the positive samples (Supplementary Fig. [S3](#MOESM1){ref-type="media"}) with high positive correlation between ENN B and ENN B~1~ (*r* = 0.90, *p* \< 0.05), the same pattern appearing in grass silage (*r* = 0.87, *p* \< 0.05) (Supplementary Table [S5](#MOESM1){ref-type="media"}). The exposure to low concentrations of several mycotoxins may be of concern in terms of their potentially additive or synergistic effects on animals. In the study conducted by Alassane-Kpembi et al. ([@CR2]), the interactions between B-type trichotecenes (DON, 3-AcDON, 15-AcDON, FUS-X and NIV) on intestinal epithelial cells were assessed. The authors reported that the combination of toxins had an additive effect. These results demonstrate that the simultaneous presence of mycotoxins, in this case, for example, DON and NIV, can be more toxic than the toxicity predicted for one mycotoxin itself. The presence of multiple mycotoxins in animal feed could be considered as a potential source of health problems; however, co-contamination of samples at levels describes in our study seems to be less important than contamination with one toxin at higher concentration (e.g. NIV at maximum level 14,262 μg/kg). The dimensions of the potential problem related to the co-occurrence of multiple mycotoxins in silage still have not been fully evaluated, especially in the case of emerging mycotoxins. Silage produced in Poland during the period of this survey was frequently contaminated with DON and ZEN, albeit at relatively low levels. It has to be emphasised that concentrations of all regulated toxins were considerably lower than the guideline values recommended by the European Commission. In some cases, concentration of emerging mycotoxins (BEA, NIV) was at possibly relevant levels. Moreover, the co-occurrence of the toxins was high, and the impact of their mixture could pose chronic problems for exposed cattle, with possible synergistic and/or additive effects. Higher frequency and concentrations for almost toxins were in maize than in grass silage. Putting these results in the context of mycotoxins exposure to animals health's suggests that grass silage could be a "safer" option as source of animal feed. Therefore, multi-toxin monitoring should be increased in order to provide the information on the occurrence of different classes of mycotoxins simultaneously in different feed commodities. Further data on the toxicity of mixtures of mycotoxins are needed, in order to establish safe limits specifically for silage. Electronic supplementary material ================================= {#Sec14} ESM 1(DOCX 387 kb) The study was funded by the KNOW (Leading National Research Centre) Scientific Consortium "Healthy Animal--Safe Food", a decision of the Ministry of Science and Higher Education No. 05-1/KNOW2/2015, and by the statutory research funds (S/255) from the Ministry of Science and Higher Education of Poland. Conflict of interest {#FPar1} ==================== None.

Background ========== Ulcerative colitis (UC) a chronic intestinal tract inflammatory disorder that is a type of inflammatory bowel disease (IBD) \[[@b1-medscimonit-26-e919530],[@b2-medscimonit-26-e919530]\]. UC usually causes bloody diarrhea, abdominal pain, and extraintestinal manifestations \[[@b3-medscimonit-26-e919530],[@b4-medscimonit-26-e919530]\]. According to data of the Fifth International Meeting on inflammatory bowel diseases, about 0.4% of people in the United States and 0.5% of people in Canada have IBD \[[@b5-medscimonit-26-e919530]\]. Moreover, the incidence of UC has increased during the past 20 years \[[@b6-medscimonit-26-e919530]\]. However, despite numerous studies on UC, the underlying mechanisms are still unclear. Nod-like receptor protein 3 (NLRP3) is a newly-found inflammation-related factor, which is involved in inflammatory response of many diseases \[[@b7-medscimonit-26-e919530]\]. It was reported that NLRP3 promoted angiogenesis in early stages of wound healing \[[@b8-medscimonit-26-e919530]\]. A recent study also showed the inhibition of NLRP3 could protect against peritonitis \[[@b9-medscimonit-26-e919530]\]. Studies showed high-mobility group box 1 protein (HMGB1) could promote the activation of severe NLRP3 bioprocesses \[[@b10-medscimonit-26-e919530]\]. Several studies also found NLRP3 and HMGB1 were abnormally expressed in IBD, and it was found that fecal HMGB1 was elevated in patients with UC and Crohn's disease \[[@b11-medscimonit-26-e919530]\]. NLRP3 was also found to be increased in a mouse UC model \[[@b12-medscimonit-26-e919530]\]. However, despite these studies, clinical evidence for NLRP3 and HMGB1 is still inadequate and few studies have reported on the relationship between NLRP3 and HMGB1 in UC patients. In the present study, we demonstrated that both serum NLRP3 and HMGB1 were elevated in UC patients. The serum levels of NLRP3 were positively correlated with serum levels of HMGB1, ET-1, IL-1β, and TNF-α, as well as severity of UC patients. This research might give more clinical evidence for NLRP3 and HMGB1 in UC and might provide potential new biomarkers for UC diagnosis. Material and Methods ==================== Patients -------- This prospective observational study included a total of 62 cases with active ulcerative colitis who came to our hospital during January 2017 to December 2018. All patients who met the inclusion criteria were consecutively enrolled into the research. The diagnosis of ulcerative colitis was confirmed by colonoscopy according to the standards of the Digestive Society of the Chinese Medical Association and British Society of Gastroenterology \[[@b13-medscimonit-26-e919530],[@b14-medscimonit-26-e919530]\]. None of the patients had received any related treatment before the study. The patients were divided into a mild/moderate group or a severe group according to Sutherland Disease Activity Index (DAI) score, in which 3\~5 represents mild, 6\~10 represents moderate, and 11\~12 represents severe UC DAI \[[@b15-medscimonit-26-e919530]\]. The following patients were excluded: patients \<18 years or \>70 years; patients with other severe intestinal disease such as Crohn's disease, local stenosis, intestinal obstruction, intestinal perforation, rectal polyps, toxic colonic dilatation, and colorectal cancer; patients with other severe liver, renal, cardiovascular, or inflammatory diseases; and patient who were pregnant. All patients received standard treatment according to the Chinese Medical Association after diagnosis. Written informed consent was obtained from all patients. The present study was approved by HwaMei Hospital, University of the Chinese Academy of Sciences. Clinical activity index and endoscopic index -------------------------------------------- The clinical activity index and endoscopic index were measured to determine the severity of patients \[[@b16-medscimonit-26-e919530]\]. The clinical activity index was recorded for the first week after admission, and the endoscopic index was recorded at colonoscopy. Measurement of serum inflammatory factors ----------------------------------------- Serum levels of inflammatory factors NLRP3 (kit: NLRP3 ELISA kit, LifeSpan Biosciences, LS-F31954), HMGB1 (kit: HMGB1 ELISA kit, LifeSpan Biosciences, LS-F26519), endothelin-1 (ET-1, kit: ET-1 ELISA kit, R&D Systems, QET00B), IL-1β (kit: IL-1β ELISA kit, Abcam, ab46052), and TNF-α (kit: IL-1β ELISA kit, Abcam, ab181421) were determined by enzyme-linked immunosorbent assay (ELISA) using commercial kits according to the manufacturer's instructions. Statistical analysis -------------------- Measurement data are expressed as mean±SD. Comparison between 2 groups was performed using the *t* test. Correlations were analyzed using Pearson's analysis. P\<0.05 was considered as statistically significant. All calculations were made using SPSS 20.0. Results ======= Basic characteristics for mild/moderate and severe UC patients -------------------------------------------------------------- In all patients, 41 cases were diagnosed as mild/moderate UC and 21 cases were diagnosed as severe UC. The mean age of all patients was 40.68±13.72, with a male: female sex ratio of 37: 25. The Sutherland DAI score, clinical activity index, and endoscopic index were all significantly higher in severe patients than in the mild/moderate group ([Table 1](#t1-medscimonit-26-e919530){ref-type="table"}, P\<0.05). No significant difference was found in age of sex. Relationship between serum NLRP3, HMGB1, and other inflammatory factors ----------------------------------------------------------------------- Serum levels of NLRP3, HMGB1, endothelin-1, IL-1β, and TNF-α were determined by ELISA. Results showed all factors were significantly higher in severe UC patients (P\<0.05, [Figure 1](#f1-medscimonit-26-e919530){ref-type="fig"}). Pearson's analysis was used to determine the correlation among factors. It was found NLRP3 level was positively correlated with HMGB1, ET-1, IL-1β, and TNF-α levels (all P\<0.05). Similar results were also found for HMGB1. Relationship between serum NLRP3, HMGB1, and clinical outcomes -------------------------------------------------------------- We used Pearson's analysis to assess whether serum levels of NLRP3 and HMGB1 were correlated with Sutherland DAI score, clinical activity index, and endoscopic index. As shown in [Figure 2](#f2-medscimonit-26-e919530){ref-type="fig"}, both NLRP3 and HMGB1 were positively correlated with Sutherland DAI score, clinical activity index, and endoscopic index, indicating both factors were positively correlated with UC severity. Discussion ========== Although there have been numerous studies on ulcerative colitis, the diagnosis of UC still needs more effective biomarkers, and the molecular mechanisms of UC remain unclear. In recent years, the NLRP3/HMGB1 axis was shown to be involved in inflammatory response in many diseases. Some studies also found NLRP3 and HMGB1 might be associated with UC \[[@b11-medscimonit-26-e919530],[@b12-medscimonit-26-e919530]\]. However, the relationship between NLRP3 and HMGB1 in UC patients has seldom been reported, and the clinical significance of NLRP3 and HMGB1 is unclear. In the present study, we further confirmed that serum levels of BLRP3 and HMGB1 were upregulated in UC patients. We found a positive correlation between NLRP3 and HMGB1, as well as between NLRP3/HMGB1 and other inflammatory factors of ET-1, IL-1β, and TNF-α. We also found NLRP3 and HMGB1 were associated with severity of UC. NLRP3 was reported to be associated with inflammatory response in many studies. Coll et al. found a kind of NLRP3 inhibitor, MCC950, and demonstrated that inhibition of NLRP3 by MCC950 could significantly improve autoinflammatory and autoimmune diseases and reduced IL-1β level \[[@b17-medscimonit-26-e919530]\]. Wu et al. demonstrated that NLRP3 was elevated in a lung inflammation model and that activation of NLRP3 contributed to lung inflammation-induced injury \[[@b18-medscimonit-26-e919530]\]. A relationship has also been reported between NLRP3 and ulcerative colitis. Itani et al. demonstrated that NLRP3 level in colon tissues of IBD patients was correlated with histological scores \[[@b12-medscimonit-26-e919530]\]. It was also found the NLRP3 inflammasome was activated in UC patients with long-standing disease \[[@b19-medscimonit-26-e919530]\]. However, the clinical evidence for NLRP3 in UC is still inadequate. In the present research, we demonstrated that the elevated NLRP3 level was positively correlated with levels of HMGB1, ET-1, IL-1β, and TNF-α, as well as UC severity. The role of HMGB1 and its relationship with NLRP3 in inflammation have also been demonstrated. Ana et al. showed that HMGB1-NLRP3 signaling was activated in Aβ-mediated microglial inflammation \[[@b20-medscimonit-26-e919530]\]. It was also found that HMGB1 can activate the NLRP3 inflammasome and further increased IL-1β production in vascular smooth muscle cells \[[@b21-medscimonit-26-e919530]\]. It was also found that HMGB1 was upregulated in stools of UC patients and could be considered as a novel biomarker for intestinal inflammation \[[@b11-medscimonit-26-e919530]\]. In a recent study, Palone confirmed that fecal HMGB1 was remarkably increased in pediatric and adult patients with Crohn's disease and UC \[[@b22-medscimonit-26-e919530]\]. However, up to now, no study has focused on the relationship between NLRP3 and HMGB1 in UC patients. In this research, we showed that HMGB1 level was positively correlated with NLRP3, as well as ET-1, IL-1β, and TNF-α, and could be used as a biomarker for UC severity. The present study has some limitations. First, the study sample size is small and from a single center. Secondly, molecular mechanisms for NLRP3 and HMGB1 in UC development are still unclear. All these topics need further research. Conclusions =========== The present prospective observational study demonstrated that NLRP3 and HMGB1 were elevated in serum of UC patients. The serum levels of NLRP3 were positively correlated with serum levels of HMGB1, ET-1, IL-1β, and TNF-α, and both NLRP3 and HMGB1 were associated with UC severity. This provides clinical evidence for the role of NLRP3 and HMGB1 in UC and might provide potential new biomarkers for UC diagnosis. **Source of support:** Departmental sources DAI : Disease Activity Index ELISA : enzyme-linked immunosorbent assay UC : ulcerative colitis IBD : inflammatory bowel disease NLRP3 : nod-like receptor protein 3 HMGB1 : high-mobility group box 1 protein {#f1-medscimonit-26-e919530} {#f2-medscimonit-26-e919530} ###### Basic clinical information for all patients. Variables Mild/moderate, n=41 Severe, n=21 P ------------------------- --------------------- -------------- ------- Age, year 39.07±14.11 43.81±12.66 0.201 Gender, Female (%) 24: 17 13: 8 0.623 Sutherland DAI score 7.05±2.12 11.48±0.51 0.000 Clinical activity index 8.24±1.79 13.57±3.61 0.000 Endoscopic index 6.41±1.77 9.38±1.91 0.000 [^1]: Study Design [^2]: Data Collection [^3]: Statistical Analysis [^4]: Data Interpretation [^5]: Manuscript Preparation [^6]: Literature Search [^7]: Funds Collection

1. Introduction {#sec1-molecules-24-04165} =============== Natural products play the central role in drug discovery \[[@B1-molecules-24-04165]\] due to their inherent biological activity and because have a wide span of structural diversity. Spirocyclic compounds have also occupied a special place in medicinal chemistry \[[@B2-molecules-24-04165]\]. Spirocycles are thought to possess a good balance of conformational rigidity and flexibility to be, on one hand, free from absorption and permeability issues characteristic of conformationally more flexible, linear scaffolds. On the other hand, spirocycles are more conformationally flexible compared to, for example, flat aromatic heterocycles and can adapt to many proteins as biological targets; thus, increasing the chances of finding bioactive hits \[[@B3-molecules-24-04165]\]. Spirocycles are distinctly three-dimensional and initial hits can be further optimized via manipulation of the molecular periphery whose three-dimensional positioning is well defined \[[@B4-molecules-24-04165]\]. We thought it worthwhile to gain insight into the structural diversity of naturally-occurring spirocyclic compounds in relation to the information of their biological activity which would provide a new angle for designing novel bioactive, druglike compounds. Modern literature features a limited number of reviews devoted to total syntheses of spirocyclic natural products \[[@B5-molecules-24-04165]\], including one for spirolactones \[[@B6-molecules-24-04165]\] and one for spirooxyindoles \[[@B7-molecules-24-04165]\]. Illustrative examples of approved natural-product drugs containing a spirocyclic motif include antifungal drug griseofulvin (**1**) and diuretic drug spironolactone (**2**). Interesting related compounds that have not achieved clinical approval include isochromanquinone antibiotic griseusin B (**3**) \[[@B8-molecules-24-04165],[@B9-molecules-24-04165]\] and spirotriprostatin (**4**) \[[@B10-molecules-24-04165]\] ([Figure 1](#molecules-24-04165-f001){ref-type="fig"}). For the purpose of the analysis presented in this review, we considered the chemical diversity of structurally unique and well characterized (i.e., those whose structures were assigned using modern analytical techniques) spirocyclic compounds registered in the ChemBL or SciFinder databases, or the Dictionary of Natural Products (DNP). The occurrence of various ring combinations (A = any atom, mostly carbon or oxygen) selected for discussion in this review is presented in [Table 1](#molecules-24-04165-t001){ref-type="table"}. Considering the uneven distribution of the ring combination occurrence statistics ([Table 1](#molecules-24-04165-t001){ref-type="table"}), the present review is structured according to the size of the \[x.y.0\] spirocyclic system. The review aims to cover either rare representatives of the spirocyclic systems that seldom occur in the natural product realm or only structurally-unique, representative compounds for those spirocyclic systems that are more widely populated with natural products reported in the literature, with an emphasis on their associated biological activities and the solid structure assignment techniques employed (structures assigned solely based on mass-spectrometric measurements are not taken into account). 2. \[2.4.0\] Spirocyclic System {#sec2-molecules-24-04165} =============================== Spirocyclic motifs containing a cyclopropane unit were found in some sesquiterpenes (**5**--**7**) which were isolated from the essential oils of South-American *Schinus terebinthifolius* fruit \[[@B11-molecules-24-04165]\] ([Figure 2](#molecules-24-04165-f002){ref-type="fig"}). In 2017, a novel condensed \[2.4.0\] spirocycle (**8**) was reported \[[@B12-molecules-24-04165]\]. It was isolated and characterized among the secondary metabolites of the *Helminthosporium velutinum* plant and was named cyclohelminthol X ([Figure 3](#molecules-24-04165-f003){ref-type="fig"}). This compound was shown to inhibit the growth of a human colon adenocarcinoma (COLO201) cell line with moderate potency (IC~50~ = 16 μM), and, much more potently (IC50 = 0.35 μM)---leukemia HL60 cell line \[[@B12-molecules-24-04165]\]. Bioassay-guided separation of *Valerianae Radix* plant extract led to the isolation and characterization of valtrate (**9**), which inhibited Rev protein mediated transport of HIV-1 from the nucleus to cytoplasm ([Figure 4](#molecules-24-04165-f004){ref-type="fig"}). This compound also inhibited p-24 production of HIV-1 virus without any notable cytotoxicity displayed against MT-4 cells. The presence of the chemically labile oxirane ring as part of the generalized \[2.4.0\] spirocyclic system is likely critical for the observed inhibition, as **9** was shown to covalently interact with cysteine \[[@B13-molecules-24-04165]\]. Additional two compounds (**31** and **32**) containing this and another (\[4.4.0\]) spirocyclic system are discussed in [Section 7](#sec7-molecules-24-04165){ref-type="sec"} of this review. 3. \[2.5.0\] Spirocyclic System {#sec3-molecules-24-04165} =============================== This group of spirocyclic natural products is represented by sesquiterpenoids illudins M and S (**10** and **11**, respectively) isolated from fungi, including the highly poisonous Jack-o′-lantern mushroom *Omphalotus illudens*. Compound **11** is currently in Phase II clinical trials against ovarian, prostate, and gastrointestinal cancers ([Figure 5](#molecules-24-04165-f005){ref-type="fig"}). Structurally analogous to illudins are sesquiterpenes **12**--**14** isolated from fungus *Agrocybe aegerita* \[[@B14-molecules-24-04165]\] also containing a \[2.5.0\] spirocyclic system ([Figure 6](#molecules-24-04165-f006){ref-type="fig"}). These compounds displayed antifungal activity against *Candida albicans* and *Candida kefyr*. An oxirane-bearing sesquiterpene (−)-ovalicin (**15**) also containing a \[2.5.0\] spirocyclic system was isolated from fungus *Pseudorotium ovalis Stolk* \[[@B15-molecules-24-04165]\]. It---and the structurally similar monoester fumagillin (**16**) displayed potent antiparasitic activities and are generally devoid of toxicity \[[@B16-molecules-24-04165]\] ([Figure 7](#molecules-24-04165-f007){ref-type="fig"}). For both compounds **15** and **16**, total syntheses have been reported \[[@B17-molecules-24-04165]\]. A \[2.5.0\] spirocyclic system is recognizable in the structure of duocarmycin SA (**17**) and duocarmycin A (**18**)---new antitumor antibiotics isolated from *streptomyces* sp. ([Figure 8](#molecules-24-04165-f008){ref-type="fig"}) \[[@B18-molecules-24-04165]\]. 4. \[3.4.0\] Spirocyclic System {#sec4-molecules-24-04165} =============================== This is an exceedingly rare type of spirocyclic motif encountered among natural products. The only compound reported in the literature to date containing such a spirocyclic system presented as a combination of a β-lactone and a pyrrolidine ring (**19**) was isolated from marine-derived *Streptomyces* strain collected in the southern area of the Korean Jeju Island \[[@B19-molecules-24-04165]\] ([Figure 9](#molecules-24-04165-f009){ref-type="fig"}). This structurally intriguing compound displayed antibacterial activity. 5. \[3.5.0\] Spirocyclic System {#sec5-molecules-24-04165} =============================== The only spirocyclic combination of a four and six-membered rings represented in natural products is rather simple achiral 1-oxaspiro\[3.5\]nonan-7-ol substituted cleroindicin A (**20**) \[[@B20-molecules-24-04165]\]. This compound was isolated from fungus *Clerodendrum japonicum* ([Figure 10](#molecules-24-04165-f010){ref-type="fig"}). 6. \[3.7.0\] Spirocyclic System {#sec6-molecules-24-04165} =============================== This intriguing spirocyclic combination of four and eight-membered rings is represented in only four closely-related sesquiterpene bis-lactones, **21**--**24** ([Figure 11](#molecules-24-04165-f011){ref-type="fig"}), isolated from poisonous plants in the *Illicium* genus grown in China \[[@B21-molecules-24-04165]\]. These structures could also be viewed as possessing a \[3.5.0\] spirocyclic motif. 7. \[4.4.0\] Spirocyclic System {#sec7-molecules-24-04165} =============================== Besides the approved diuretic spironolactone (**2**, vide supra), heteroatom-containing \[4.4.0\] spirocyclic motifs are widely represented by various lactones ([Figure 12](#molecules-24-04165-f012){ref-type="fig"}). The most structurally simple, naturally occurring spirocyclic lactone, 1,7-dioxaspiro\[4.4.0\]nonane or longianone (**25**) was isolated from higher fungi *Xylaria longiana* \[[@B22-molecules-24-04165]\]. The absolute configuration of longianone was confirmed by stereoselective total synthesis \[[@B23-molecules-24-04165]\]. Hyperolactones A (**26**) and C (**27**) isolated from *Hypericum chainens* plant \[[@B24-molecules-24-04165]\] displayed antiviral activity \[[@B25-molecules-24-04165]\]. The Nicolaou group reported a photochemical, \[2 + 2\]-cycloaddition based synthesis of a library based on natural product biyouyanagin (**28**) which allowed revising its originally reported absolute configuration \[[@B26-molecules-24-04165]\]. (+)-Crassalactone D (**29**) is a styryl-lactone isolated from the leaves of *Polyalthia crassa* plant which displayed cytotoxic properties \[[@B27-molecules-24-04165]\]. Pyrenolide D (**30**) is a highly oxygenated tricyclic spirolactone isolated from phytopathogenic fungus *Pyrenophora teres*, also displaying potent cytotoxicity \[[@B28-molecules-24-04165]\]. Sesequiterpene levantenolide (**31**) also contained a \[4.4.0\] spirocyclic lactone moiety; it was isolated from tobacco grown in Turkey \[[@B29-molecules-24-04165]\]. It exerted potent suppression of cytokine cascades and can, therefore, be considered a lead for anti-inflammatory drug development \[[@B30-molecules-24-04165]\]. Complex polycyclic alkaloids represented by compound **32** were isolated from *Stemona* genus shrubs. These compounds contain a basic cyclopenta\[1,2-b\]pyrrolo\[1,2-a\]azepine scaffold and display promising anti-cough medicinal properties \[[@B31-molecules-24-04165]\] ([Figure 12](#molecules-24-04165-f012){ref-type="fig"}). A \[4.4.0\] spirocyclic lactone moiety is found (in combination with a \[2.4.0\] spirocyclic oxirane) in limonoids **33**--**34**, which were recently isolated from *Trichilia connaroides* ([Figure 13](#molecules-24-04165-f013){ref-type="fig"}). For these compounds, some insights into a possible biosynthetic pathway have been provided. Likewise, these compounds were screened for various types of bioactivity and have been shown to inhibit NO production in a cellular model of inflammation (induced in RAW264.7 cell line with LPS) by 25.89% and 37.13% at 25 and 50 μM, respectively \[[@B32-molecules-24-04165]\]. Studies of secondary metabolite structures in endophyte fungus *Penicillium purpurogenum* unveiled a series of unique sesquiterpene lactone compounds (**35**--**37**) containing spirocyclic combinations of three five-membered rings ([Figure 14](#molecules-24-04165-f014){ref-type="fig"}). All three compounds were screened against several cancer cell lines (melanoma M14, colon cancer HCT-116, glioma U87, ovary cancer A2780, stomach cancer BG-823, hepatoma Bel-7402, and lung cancer A549) and several pathogenic microorganisms (*Mycobacterium spegmatis* (ATCC70084), *Staphylococcus aureus* (ATCC25923), and *Staphylococcus epidermidis* (ATC12228)); however, no activity was detected at 50 μM \[[@B33-molecules-24-04165]\]. Rather intriguing are the structures of curcumanolides **38**--**41**, natural \[4.4.0\] spirocyclic lactones recently isolated from *Curcuma heyneana*, a traditional medicinal plant of Indonesia ([Figure 15](#molecules-24-04165-f015){ref-type="fig"}) \[[@B34-molecules-24-04165]\]. In the course of the thorough structural investigation of a series of iridoid glycosides isolated from the *Morinda citrifolia* plant, a revised structure was assigned. In particular, dehydromethoxygaertneroside (**42**), dehydroepoxymethoxygaertnoside (**43**), and citrifolinoside A (**44**) were shown to be structurally distinct compounds, all of which, however, possessed a \[4.4.0\] spirocyclic lactone moiety ([Figure 16](#molecules-24-04165-f016){ref-type="fig"}) \[[@B35-molecules-24-04165]\]. Two diastereomers of spirophthalides, **45** and **46,** which possess a unique presentation of a \[4.4.0\] spirocyclic lactone, were recently isolated from a marine-sponge-derived fungus, *Setosphaeria* sp. ([Figure 17](#molecules-24-04165-f017){ref-type="fig"}) \[[@B36-molecules-24-04165]\]. Unique spirocyclic dihydroindole-containing \[4.4.0\] spirocyclic lactones **47** and **48**, also possessing a quinazolone substituent, were identified among mycotoxins produced by *P. aethiopicum* ([Figure 18](#molecules-24-04165-f018){ref-type="fig"}) \[[@B37-molecules-24-04165]\]. During a chemical and structural investigation of secondary metabolites of *Penicillium dangeardii*, a series of related \[4.4.0\] spirocyclic lactones (penicillactones A-C) **49**--**51** was identified. These possessed a complex molecular framework rich in carboxylate functionality and a well stereodefined substitution pattern around the spirocyclic core ([Figure 19](#molecules-24-04165-f019){ref-type="fig"}). Compounds **49**--**51** showed promise as leads for new antibiotic development. Additionally, penicillactones B and C (**50** and **52**, respectively) showed inhibition of the release of β-glucuronidase from polymorphonuclear leukocytes with ED~50~ values of 2.58 and 1.57 μM \[[@B38-molecules-24-04165]\]. Rather intriguing and unique is the structure of spirocyclic hydantoins possessing a furanose unit. One of the first representatives of these natural products (hydantocidine **52**) was isolated from *Streptomyces hygroscopicus* ([Figure 20](#molecules-24-04165-f020){ref-type="fig"}). Hydantocidine displayed herbicidal properties which were linked to its ability to inhibit adenylate succinate synthase \[[@B39-molecules-24-04165]\]. A unique presentation of a \[4.4.0\] spirocyclic system is featured in spirocyclic benzofuranones **53**--**55** isolated from ethanolic extracts of *Ganoderma Applanatum* ([Figure 21](#molecules-24-04165-f021){ref-type="fig"}) \[[@B40-molecules-24-04165]\]. Not less interesting than the spirocyclic benzofuranones discussed above are natural products possessing a spirooxyindole motif. One of the first representatives of \[4.4.0\] spirocyclic compounds reported in the literature is spirotriprostatine (**56**), possessing moderate (IC~50~ = 197.5 μM) cytotoxic activity \[[@B41-molecules-24-04165]\]. Naturally occurring spirooxyindoles were first isolated from plants *Apocynaceae* and *Rubiacae*, and from *Aspidosperma, Mitragyna, Ourouparia, Rauwolfia* and *Vinca* genera. These compounds can be further classified into two substructural classes: the tetracyclic secoyohimbane type (e.g., rhynchophylline (**57**)) and the pentacyclic heteroyohimbane type (e.g., formosanine (**58**)) ([Figure 22](#molecules-24-04165-f022){ref-type="fig"}) \[[@B42-molecules-24-04165]\]. A very interesting class of natural products containing a \[4.4.0\] spirocyclic motif includes spiropseudoindoxyl alkaloids. Microbial transformation of the alkaloid mitragynine by the fungus *Helminthosporum* sp. was reported in 1974 to yield two major metabolites. The compounds were isolated from the biological milieu and their structures were elucidated as mitragynine pseudoindoxyl (**59**) and hydroxy mitragynine pseudoindoxyl (**60**) ([Figure 23](#molecules-24-04165-f023){ref-type="fig"}) \[[@B43-molecules-24-04165]\]. These compounds were later shown to possess opioid analgesic activity by exerting mu agonism and delta antagonism while not recruiting β-arrestin-2 \[[@B44-molecules-24-04165]\]. The \[4.4.0\] spirocyclic pseudoindoxyl motif represents a rather common feature in indole alkaloids, as can be illustrated by such examples as fluorocurine (**61**) \[[@B45-molecules-24-04165]\], several diketopiperazines isolated from holothurianderived fungus *Aspergillus fumigatus* (**62a**--**d**) \[[@B46-molecules-24-04165]\], brevianamide B (**63**) \[[@B47-molecules-24-04165]\], and rauniticine pseudoindoxyl (**64**) \[[@B48-molecules-24-04165]\] ([Figure 24](#molecules-24-04165-f024){ref-type="fig"}). A structurally unique \[4.4.0\] spiroheterocyclic system is represented by a series of highly oxygenated lactone lactams (**65**--**69**) isolated from marine sediment-derived fungus *Aspergillus sydowi* D2--6 ([Figure 25](#molecules-24-04165-f025){ref-type="fig"}). Compounds **65**--**69** were shown to inhibit growth of adenocarcinoma cell line A549 with an IC~50~ value of 10 μM \[[@B49-molecules-24-04165]\]. A wide diversity of heterocyclic spirocyclic scaffolds all belonging to the generalized \[4.4.0\] system (**70**--**73**) ([Figure 26](#molecules-24-04165-f026){ref-type="fig"}) have been isolated recently. Two regioisomeric phytoalexins---erucalexin (**70**) and its regioisomer (+)-1-methoxyspirobrassinin (**71**) were isolated from the wild crucifer *Erucastrum gallicum* \[[@B50-molecules-24-04165]\]. Mycotoxins related to tryptoquialanine A (**71**) were isolated from *Penicillium* spp. and *Aspergillus clavatus* \[[@B51-molecules-24-04165]\]. For tryptoquialanines, the biosynthetic pathway has been recently elucidated \[[@B25-molecules-24-04165]\]. Another spirooxyindole lactone lactam compound **73** isolated from *Coix lachryma-jobi* L. has been recently reported and shown to possess activity against human lung cancer (A549) and colon carcinoma (HT-29 and COLO205) cell lines \[[@B52-molecules-24-04165]\]. Secondary metabolite investigation of the liquid culture of entomogenous fungus *Isaria cateniannulata* led to the identification of a new spirocyclic compound **74** containing a 1,6-dioxaspiro\[4.4\]nonane moiety ([Figure 27](#molecules-24-04165-f027){ref-type="fig"}). The compound showed weak inhibitory activity against the HeLa cancer cell line \[[@B53-molecules-24-04165]\]. Spirocyclic \[4.4.0\] tetrahydrofurans are featured in a series of twelve natural products **75a--l** dubbed bipolaricins ([Figure 28](#molecules-24-04165-f028){ref-type="fig"}). These compounds are ophiobolin-type tetracyclic sesterterpenes from a phytopathogenic *Bipolaris* sp. fungus. They were tested for HMGCoA reductase inhibition as well as anti-inflammatory and cytotoxic activities. The biological activity discovered provided the basis for considering these compounds as leads for antiinflammation and antihyperglycemic therapy developments \[[@B54-molecules-24-04165]\]. An interesting type of \[4.4.0\] spirocyclic motif is present in fredericamycin A (**76**), an antitumor antibiotic produced by *Streptomyces griseus* ([Figure 29](#molecules-24-04165-f029){ref-type="fig"}) \[[@B55-molecules-24-04165],[@B56-molecules-24-04165]\]. Summing up, the overall scaffold distribution within the general \[4.4.0\] spirocyclic system is shown in [Figure 30](#molecules-24-04165-f030){ref-type="fig"}. Spirolactones are the most widely represented motifs in the \[4.4.0\] spirocyclic systems, with over 20 examples discussed above. Spirocyclic lactams are exemplified by 10 natural products. However, \[4.4.0\] spirocyclic lactam lactones and spirooxyindoles are much less common in the natural products and are represented by only a handful of examples. In terms of biological activity, the current data are mostly limited to cytostatic and antibacterial properties. The natural products isolated within the last 1--2 years are poorly investigated with regard to their biological properties. 8. \[4.5.0\] Spirocyclic System {#sec8-molecules-24-04165} =============================== Secondary metabolite investigation of *Teucrium viscidum* led to the identification of a \[4.5.0\] spirocyclic compound (**77**) possessing a unique skeleton \[[@B57-molecules-24-04165]\]. A skeleton of similar complexity had only been featured once in the literature three decades before that \[[@B58-molecules-24-04165]\] ([Figure 31](#molecules-24-04165-f031){ref-type="fig"}). The \[4.5.0\] spirocyclic motifs are featured in many natural terpenes. Recently, new spirocyclic triterpenoids **78**--**79** were isolated from *Leonurus japonicus* fruit ([Figure 32](#molecules-24-04165-f032){ref-type="fig"}). These compounds displayed moderately potent (IC~50~ \< 10 μM) growth inhibition of five human cancer cell lines (stomach cancer BGC-823 and KE-97, hepatocarcinoma Huh-7, Jurkat T-cell limphoblasts, and breast adenocarcinoma MCF-7) \[[@B59-molecules-24-04165]\]. Another example of an all-carbon \[4.5.0\] spirocyclic system is provided by spirocarolitone (**80**), recently isolated from *Ruptiliocarpon caracolito* \[[@B60-molecules-24-04165]\] ([Figure 33](#molecules-24-04165-f033){ref-type="fig"}). Structurally novel tricyclic-iridal triterpenoids belamcandanes A and B (**81** and **82**) ([Figure 34](#molecules-24-04165-f034){ref-type="fig"}) were recently isolated from *Belamcanda chinensis* and shown to possess moderate hepatoprotective properties. A possible biosynthetic pathway has been proposed \[[@B61-molecules-24-04165]\]. New biologically active sesquiterpenoids **83**--**85** possessing an all-carbon \[4.5.0\] spirocyclic system were isolated from rhizomes of *Acorus calamus* ([Figure 35](#molecules-24-04165-f035){ref-type="fig"}). Compound **83** exhibited weak hepatoprotective activities against APAP-induced HepG2 cell damage \[[@B62-molecules-24-04165]\]. The ethyl acetate soluble fraction of a MeOH extract of the dried stems and roots of *Capsicum annum* gave several new sesquiterpenoids, among which two \[4.5.0\] spirocyclic compounds termed canusesnols (**86**--**87**, [Figure 36](#molecules-24-04165-f036){ref-type="fig"}) were identified and evaluated for their cytotoxic activities \[[@B63-molecules-24-04165]\]. Perhaps the most clinically advanced natural spirocyclic compound---spirocyclic benzofuran griseofulvin (**88**) isolated from *Penicillium griseofulvum* has been employed in clinical practice for therapy against ring worms \[[@B64-molecules-24-04165]\] and was marketed by GlaxoSmithKline under the trade name Grisovin^TM^ \[[@B65-molecules-24-04165]\] ([Figure 37](#molecules-24-04165-f037){ref-type="fig"}). Natural \[4.5.0\] spirocyclic lactones are characterized by a wide structural diversity and abundance of biological activities reported for them. These are exemplified by the mediator of mycoparasitism lambertollol C (**89**) \[[@B66-molecules-24-04165]\], glycine-gated chloride channel receptor modulator (−)-ircinianin (**90**) \[[@B67-molecules-24-04165]\], and terpenoid andirolactone (**91**) isolated from *Cedrus libanotica* \[[@B68-molecules-24-04165]\] ([Figure 38](#molecules-24-04165-f038){ref-type="fig"}). More examples of bioactive \[4.5.0\] spirocyclic lactones are provided by abyssomicins (**92a**--**c**, [Figure 39](#molecules-24-04165-f039){ref-type="fig"}), which were isolated from Actinobacteria and shown to inhibit *p*-aminobenzoate biosynthesis \[[@B69-molecules-24-04165]\]. Antibacterial and antitumor compound lactonamycin Z (**93**) was isolated from *Streptomyces sanglieri* \[[@B70-molecules-24-04165]\] and is an example of a \[4.5.0\] spirocyclic lactone embedded in a complex polycyclic system ([Figure 40](#molecules-24-04165-f040){ref-type="fig"}). In 2015, Cech and co-workers reported new antibiotic spirocyclic lactone chaetocuprum (**94**) \[[@B71-molecules-24-04165]\]. This compound was isolated from an endophyte fungus growing on the roots of wild *Anemopsis californica* plant which was traditionally used by North American tribes to treat infections and inflammation. Similarly, growing endophyte fungal parasites on the roots of *Chaetomium indicum* allowed Asai and Oshima \[[@B72-molecules-24-04165]\] to isolate both epimers of spiroindicumide A and B (**95** and **96**, respectively) which feature an unprecedented spirocyclic lactone scaffold ([Figure 41](#molecules-24-04165-f041){ref-type="fig"}). In addition to lambertollol C discussed above, two related epimeric compounds labertollol A (**97**) and B (**98**), also bearing a 4,8-dihydroxy-2,3,4-trihydronaphthalen-1-one scaffold and featuring a spirobutenolide moiety ([Figure 42](#molecules-24-04165-f042){ref-type="fig"}) were reported to possess high antifungal activity (IC~50~ = 0.5 μg/mL) \[[@B73-molecules-24-04165]\]. Traditional Chinese medicinal plant *Rehmannia glutinosa* turned out to be a rich source of \[4.5.0\] spirocyclic lactones: massarigenin D (**99**), spiromassaritone (**100**), and paecilospirone (**101**) ([Figure 43](#molecules-24-04165-f043){ref-type="fig"}) which displayed potent (IC~50~ from 0.25 to 32 μg/mL) antifungal activity \[[@B74-molecules-24-04165]\]. Perenniporide A (**102**) was the only spirocyclic lactone derivative of the naphthalenone family of natural products perenniporides A--D isolated from solid cultures of a fungus *Perenniporia* sp. inhabiting the larva of *Euops chinesis*, a phytophagous weevil with high host specificity to the medicinal plant *Fallopia japonica* ([Figure 44](#molecules-24-04165-f044){ref-type="fig"}) \[[@B75-molecules-24-04165]\]. A \[4.5.0\] spirocyclic lactone moiety is featured in secochiliolide acid **103** ([Figure 45](#molecules-24-04165-f045){ref-type="fig"}), for which antiparasite activity was reported \[[@B76-molecules-24-04165],[@B77-molecules-24-04165]\]. A rather unique \[4.5.0\] spirocyclic lactone moiety was identified in sesquiterpene abiespiroside A (**104**), which was isolated from Chinese tree *Abies dalavayi* ([Figure 46](#molecules-24-04165-f046){ref-type="fig"}). For this compound, anti-inflammatory activity was discovered \[[@B78-molecules-24-04165]\]. A \[4.5.0\] spirocyclic lactone motif is featured in pathylactone A (**105**) isolated from marine sources, which demonstrated Ca^2+^ channel antagonistic activity ([Figure 47](#molecules-24-04165-f047){ref-type="fig"}) \[[@B79-molecules-24-04165]\]. A whole series of spirolactones containing a terpenoid carane system (**106**--**110**) was reported as synthesized in enantioselective fashion ([Figure 48](#molecules-24-04165-f048){ref-type="fig"}). For these compounds, insect-feeding deterrent activity was reported \[[@B80-molecules-24-04165]\]. In addition to the abundance of \[4.5.0\] spirocyclic lactones reported in the literature, some instances of spirocyclic tetrahydrofurans can be encountered. For example, 15-methoxycyclocalopin A (**111**) and isocyclocalopin A (**112**) were reported to be isolated from *Boletus calopus* \[[@B81-molecules-24-04165]\]. Notably, compound **112** can be also considered a \[5.5.0\] spirocyclic hexahydropyran ([Figure 49](#molecules-24-04165-f049){ref-type="fig"}). The structures of these compounds are reminiscent of spirocyclic dihydrofuran 8,9-dehydrotheaspirone, both enantiomers of which (**113a**--**b**) have been reported as volatile constituents of nectarines \[[@B82-molecules-24-04165]\]. Their presence in the fruit was connected to some specific organoleptic properties of some kinds of nectarines ([Figure 50](#molecules-24-04165-f050){ref-type="fig"}) Rather intriguing labdane-type diterpenoids (**114a**--**b**), epimeric to each other, isolated from the fruit of *Vitex agnus-castus* plant feature a unique skeleton consisting of both a \[4.4.0\] and a \[4.5.0\] spirocyclic tetrahydrofuran system ([Figure 51](#molecules-24-04165-f051){ref-type="fig"}) \[[@B83-molecules-24-04165]\]. Rather unique is the structure of heliespirone **115** isolated from highly polar fractions of *Helianthus annuus* L. extract \[[@B84-molecules-24-04165]\]. In this natural product, tetrahydrofuran forms a spirocyclic motif with a quinone-like moiety ([Figure 52](#molecules-24-04165-f052){ref-type="fig"}). An oxygenated \[4.5.0\] spirocyclic framework is featured in several toxins, exemplified by arthropod toxin **116** ([Figure 53](#molecules-24-04165-f053){ref-type="fig"}) isolated from *Dinophysis acuta* and shown to potentiate erectile function \[[@B85-molecules-24-04165]\]. Another example of similarly polyoxygenated \[4.5.0\] spirocyclic tetrahydrofuran is provided by quinochalcone **117**, named saffloquinoside A, isolated from *Carthamus tinctorius* ([Figure 54](#molecules-24-04165-f054){ref-type="fig"}) \[[@B86-molecules-24-04165]\]. Compound **117** was evaluated in vitro for the inhibitory effect on the release of β-glucuronidase from rat polymorphonuclear neutrophils (PMNs) induced by the platelet-activating factor (PAF). It exhibited anti-inflammatory activity and the inhibitory rate was 54.3% (at 10^−5^ mol/L concentration). Nitrogen-containing \[4.5.0\] spirocyclic systems are a lot more scarce compared to their oxygen-containing counterparts and can be exemplified by only two examples discussed below. Alkaloid (±)-pandamarine (**118**) isolated as a major component from *Pandanus amaryllif olius* contains a \[4.5.0\] spirocyclic scaffold composed of a piperidine and a pyrollen-2-one rings ([Figure 55](#molecules-24-04165-f055){ref-type="fig"}) \[[@B87-molecules-24-04165]\]. Another example of nitrogen-containing \[4.5.0\] spirocyclic system is provided by surugatoxin (**119**) isolated from the toxic Japanese ivory shell (*Babylonica japonica*) ([Figure 56](#molecules-24-04165-f056){ref-type="fig"}). This toxin suppresses the presynaptic nervous system \[[@B88-molecules-24-04165]\]. Its total synthesis, in the racemic form, was achieved in 1994 by the Inoue group \[[@B89-molecules-24-04165]\]. A \[4.5.0\] spirocyclic system is recognizable in spirostaphylotrichins which are spirocyclic γ-lactams mainly produced by several endophytic fungal strains of *Curvularia*, *Pyrenophora*, and *Staphylotrichum*. These are exemplified by spirostaphylotrichin X (**120**), characterized as an antiinfluenza agent targeting RNA polymerase PB2 \[[@B90-molecules-24-04165]\], and spirostaphylotrichin W (**121**), investigated as a potential mycoherbicide for cheatgrass (*Bromus tectorum*) biocontrol \[[@B91-molecules-24-04165]\] ([Figure 57](#molecules-24-04165-f057){ref-type="fig"}). Summarizing this Section, the scaffold diversity stemming from the general \[4.5.0\] spirocyclic framework is comparable to that of the \[4.4.0\] spirocyclic system discussed earlier ([Figure 30](#molecules-24-04165-f030){ref-type="fig"}) and is shown in [Figure 58](#molecules-24-04165-f058){ref-type="fig"}. 9. \[4.6.0\] Spirocyclic System {#sec9-molecules-24-04165} =============================== As to the spirocyclic systems combining five and seven-memebred rings (the \[4.6.0\] spirocyclic system), spiro meroterpenoids spiroapplanatumines (**122**--**124**) isolated from the fruiting bodies of the fungus *Ganoderma applanatum* provide an eloquent example ([Figure 59](#molecules-24-04165-f059){ref-type="fig"}). Biological evaluation of the compounds disclosed that compound **124** inhibited JAK3 kinase with an IC~50~ value of 7.0 ± 3.2 μM \[[@B92-molecules-24-04165]\]. In 2003, investigation of the neutral ether extracts of the fungus *Fomes cajanderi* led to the isolation of three novel ketal lactones named fomlactones A (**125**), B (**126**), and C (**127**) ([Figure 60](#molecules-24-04165-f060){ref-type="fig"}). The compounds clearly possess a \[4.6.0\] spirocyclic lactone moiety. However, their biological potential remains to be investigated \[[@B93-molecules-24-04165]\]. A very unique spirocyclic \[4.6.0\] framework formed by a spiro\[benzofuranonebenzazepine\] skeleton is featured in natural products (±)-juglanaloid A (**128a**--**b**) and (±)-juglanaloid B (**129a**--**b**). These benzazepine alkaloids were isolated from the bark of *Juglans mandshurica*. Remarkably, both racemic natural products were successfully resolved by chiral separation and absolute configurations were unambiguously assigned ([Figure 61](#molecules-24-04165-f061){ref-type="fig"}). These enantiopure versions were screened for their in vitro inhibitory activities against self-induced Aβ~1-42~ aggregation using the Thioflavin T (Th-T) assay using curcumin as a reference compound. The compounds demonstrated promise acting as inhibitors of amyloid β aggregation \[[@B94-molecules-24-04165]\]. Furthermore, in the last 1--2 years there has been an avalanche of new \[4.6.0\] spirocyclic structures reported in the literature. For examples, lanostane-type spirolactone triterpenoids **130a**--**c** isolated from *Ganoderma applanatum* ([Figure 62](#molecules-24-04165-f062){ref-type="fig"}) were reported to possess anti-hepatic fibrosis activities \[[@B95-molecules-24-04165]\]. Interestingly, an additional \[4.5.0\] and \[2.5.0\] spirocyclic motif is recognizable in compounds **130b** and **130c**, respectively. Another recent example (reported in 2019) of a \[4.6.0\] spirocyclic system is provided by grayanane diterpenoid auriculatol A (**131**) isolated from leaves of *Rhododendron auriculatum* ([Figure 63](#molecules-24-04165-f063){ref-type="fig"}). This compound is the first example of a 5,20-epoxygrayanane diterpenoid bearing a 7-oxabicyclo\[4.2.1\]nonane motif and a *trans/cis/cis/cis*-fused 5/5/7/6/5 pentacyclic ring system. Auriculatol A showed analgesic activity in the acetic acid-induced writhing test \[[@B96-molecules-24-04165]\]. Finally another \[4.6.0\] spirocyclic lactone, seconoriridone A (isolated as a 7:1 epimeric mixture of (**132a**) and (**132b**)) was isolated in 2019 from *Belamcanda chinensis* ([Figure 64](#molecules-24-04165-f064){ref-type="fig"}). Although no biological activity was reported for this intriguing molecular structure, a plausible biosynthetic pathway was proposed \[[@B97-molecules-24-04165]\]. The \[4.6.0\] spirocyclic system is amply exemplified in the natural products domain by the gelsenium alkaloids---gelsebanine (**133**), 14α-hydroxyelegansamine (**134**), 14α-hydroxygelsamydine (**135**) \[[@B98-molecules-24-04165]\], 14-acetoxygelsenicine (**136**), 14-acetoxy-15-hydroxygelsenicine (**137**), 14-hydroxy-19-oxogelsenicine (**138**), and 14-acetoxygelseligine (**139**) \[[@B99-molecules-24-04165]\] ([Figure 65](#molecules-24-04165-f065){ref-type="fig"}). 10. \[4.7.0\] Spirocyclic System {#sec10-molecules-24-04165} ================================ Spirocyclic natural products whose scaffolds contain rings larger than six-membered, e.g., \[4.7.0\] spirocyclic systems, are exceedingly rare. An eloquent example is provided by natural sugar-containing compounds phyllanthunin (**140**) recently isolated from an ethanol extract of the fruit of *Phyllanthus emblica* ([Figure 66](#molecules-24-04165-f066){ref-type="fig"}) \[[@B100-molecules-24-04165]\]. Additionally, remarkably illustrative of the presence of \[4.7.0\] spirocyclic motifs in natural products, are portimines A (**141**) and B (**142**) isolated from the marine benthic dinoflagellate *Vulcanodinium rugosum* collected from Northland, New Zealand \[[@B101-molecules-24-04165],[@B102-molecules-24-04165]\]. In addition to a \[4.7.0\] spirocyclic system, these compounds also contain a \[4.5.0\] spirocycle ([Figure 67](#molecules-24-04165-f067){ref-type="fig"}). Portimine has also been shown to induce apoptosis and reduce the growth of a variety of cancer cell lines at low nanomolar concentrations. 11. \[5.5.0\] Spirocyclic System {#sec11-molecules-24-04165} ================================ Among natural products containing a \[5.5.0\] spirocyclic motif, new spirocyclic chamigrane sesquiterpenes, merulinols B (**143**), C (**144**), E (**145**), and F (**146**) are notable examples ([Figure 68](#molecules-24-04165-f068){ref-type="fig"}). These compounds were isolated from basidiomycetous endophytic fungus XG8D associated with the mangrove *Xylocarpus granatum* \[[@B103-molecules-24-04165]\]. The in vitro cytotoxicity of all compounds was evaluated against three human cancer cell lines, MCF-7, Hep-G2, and KATO-3. Compound **144** selectively displayed cytotoxicity against KATO-3 cells with an IC~50~ value of 35.0 μM. Highly oxygenated acylphloroglucinol, hyperbeanol C (**147**), was isolated from the methanol extract of *Hypericum beanie* \[[@B104-molecules-24-04165]\]. This compound contains an all-carbon \[5.5.0\] spirocyclic system, spiro\[5.5.0\]undec-2-ene-1,5-dione ([Figure 69](#molecules-24-04165-f069){ref-type="fig"}). The cytotoxicity of **147** against the cancer cell lines HL-60, SMMC-7721, PANC-1, MCF-7, K562, and SK-BR-3 was tested using the methyl thiazol tetrazalium (MTT) method with *cis*-platinum as the positive control. It exhibited modest cytotoxicity against K562 cells with an IC~50~ 16.9 μM. Remarkable presentation of the (*R*)-1,7-dioxaspiro\[5.5\] undecane framework is found in nor-spiro-azaphilones, thielavialides A−D (**148**--**151**), and bis-spiro-azaphilone, thielavialide E (**152**) together with bis-spiro-azaphilone pestafolide A (**153**) ([Figure 70](#molecules-24-04165-f070){ref-type="fig"}). All these compounds were isolated from the endophytic fungal strain, *Thielavia* sp. PA0001, occurring in the healthy leaf tissue of aeroponically grown *Physalis alkekengi* \[[@B105-molecules-24-04165]\]. A very similar \[5.5.0\] spirocyclic moiety can be found in the structure of pteridic acids C and F (**154** and **155**, respectively) isolated in 2017 from a culture broth of the marine-derived actinomycete *Streptomyces* sp. SCSGAA 0027 ([Figure 71](#molecules-24-04165-f071){ref-type="fig"}). While these compounds were seen as potential leads for antibacterial drug discovery, their extensive testing for antimicrobial activity against two gorgonian pathogenic fungal strains *Aspergullus versicolor* SCSGAF 0096 and *Aspergullus sydowii* SCSGAF 0035; a human pathogenic fungal strain *Candida albicans* SC5314; and two bacterial strains *Escherichia coli* and *Bacillus subtilis*, showed that the compounds had only a weak antimicrobial activity \[[@B106-molecules-24-04165]\]. A unique \[5.5.0\] spirocyclic skeleton formed by a hexahydropyran and a pyrrolo\[2,1-c\]morpholine moieties is found in pollenopyrroside A (**156**) and B (**157**) isolated from bee-collected *Brassica campestris* pollen ([Figure 72](#molecules-24-04165-f072){ref-type="fig"}). The Chinese team who reported these natural products in 2010 also proposed a biosynthetic pathway that involves a reaction of 3-deoxy-[d]{.smallcaps}-fructose and 5-oxymethyl-2-formyl-pyrrole as the key step. Biological testing of these aldehyde compounds using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) method revealed that they possess no cytotoxicity against A549, Bel7420, BGC-823, HCT-8, and A2780 cancerous cell lines at 10 μg/mL \[[@B107-molecules-24-04165]\]. Another unique \[5.5.0\] spirocyclic skeleton is noteworthy in the context of this review. Two structurally unique spirocyclic alkaloids **158** and **159** were isolated in 2007 from the halotolerant B-17 fungal strain of *Aspergillus variecolor* ([Figure 73](#molecules-24-04165-f073){ref-type="fig"}). Both compounds possessed an intriguing spirocyclic piperazin-2,5-dione moiety and exhibited cytotoxic properties \[[@B108-molecules-24-04165]\]. Remarkably, in 2018, a very similar spirocyclic piperazin-2,5-dione variecolortin B (**160**) was isolated from the marine-derived fungus *Eurotium* sp. SCSIO F452 ([Figure 74](#molecules-24-04165-f074){ref-type="fig"}). The compound exhibited different antioxidative and cytotoxic activities. Interestingly, the same species gave rise to a compound possessing an even more seldomly-occurring spirocyclic moiety; namely, \[5.6.0\] (vide infra) \[[@B109-molecules-24-04165]\]. The \[5.5.0\] spirocyclic system occurs very prominently in bioactive meroterpenoids **161a**--**e** and **162a**--**d** isolated in 2019 from mangrove-derived fungus *Penicillium* sp. ([Figure 75](#molecules-24-04165-f075){ref-type="fig"}). Several of these compounds showed growth inhibition activity against newly hatched larvae of *Helicoverpa armigera* Hubner with IC~50~ values ranging from 50 to 200 μg/mL, and some notable activity against *Caenorhabditis elegans* \[[@B110-molecules-24-04165]\]. Workers of the ant *Carebarella bicolor* collected in Panama were found to contain the histrionicotoxin class of alkaloids with unusual 2,7-disubstituted-1-azaspiro\[5.5\]undecanol structures **163a**--**i** ([Figure 76](#molecules-24-04165-f076){ref-type="fig"}) \[[@B111-molecules-24-04165]\]. 12. \[5.6.0\] Spirocyclic System {#sec12-molecules-24-04165} ================================ An interesting group of natural products representative of this spirocyclic system are periplosides (**164**), a spiro-orthoester group-containing pregnane-type glycosides discovered in the course of phytochemical investigation of the root bark of *Periploca sepium* ([Figure 77](#molecules-24-04165-f077){ref-type="fig"}). The \[5.6.0\] spirocyclic orthoester core is distinctly modified with a steroid unit on one hand (R^1^) and with an oligosaccharide moiety on the other (R^2^). The compounds were evaluated for their inhibitory activities against the proliferation of T-lymphocytes. As a result, one specific compound (periploside C), the most abundant glycoside containing a spiro-orthoester moiety found in the plant, exhibited the most favorite selective index value (SI = 82.5). The inhibitory activity and the SI value appear to depend on the constitution of the saccharide chain \[[@B112-molecules-24-04165]\]. The remarkable, from a structural perspective, spirolide G (**165**), was isolated from Danish strains of toxigenic dinoflagellate *Alexandrium ostenfeldii*. The toxicological profile of this compound was evaluated \[[@B113-molecules-24-04165]\]. Interestingly, in addition to the spirocyclic \[5.6.0\] moiety in question, spirolide G (**165**) contains two others; namely, a \[4.4.0\] and a \[4.5.0\] motif ([Figure 78](#molecules-24-04165-f078){ref-type="fig"}). Referring back to the chemical investigation of the marine-derived fungus *Eurotium* sp. SCSIO F452 discussed above in connection with compounds belonging to the \[5.5.0\] spirocyclic system, an intriguing \[5.6.0\] spirocyclic compound **166** ([Figure 79](#molecules-24-04165-f079){ref-type="fig"}) was also isolated from the same species \[[@B109-molecules-24-04165]\]. This is a case of one species giving rise to a diversity of spirocyclic frameworks, underscoring the significance of spirocycles in the natural product realm. One particular example of such spirocycle diversity derived from a single organism is discussed in [Section 13](#sec13-molecules-24-04165){ref-type="sec"} below. A \[5.6.0\] spirocyclic moiety is recognizable in the new sesquiterpene dimer vieloplain G (**167**) isolated in 2019 from the roots of *Xylopia vielana* ([Figure 80](#molecules-24-04165-f080){ref-type="fig"}). This compound showed considerable cytotoxicity against DU145 cells with an IC~50~ value of 9.5 μM \[[@B114-molecules-24-04165]\]. 13. \[6.6.0\] Spirocyclic System {#sec13-molecules-24-04165} ================================ This type of spirocyclic framework is exceedingly rare in the natural product domain, with only one example of unique 1-oxaspiro\[6.6\]tridecane **168**, a spirocyclic nortriterpenoid Spiroschincarin A isolated from the fruit of *Schisandra incarnate* ([Figure 81](#molecules-24-04165-f081){ref-type="fig"}) \[[@B115-molecules-24-04165]\]. 14. Plant Species Distinctly Rich in Diverse Spirocyclic Natural Products {#sec14-molecules-24-04165} ========================================================================= Some cases when the same plant or microorganism gave rise to secondary metabolites with several structurally-diverse spirocyclic frameworks were discussed above. However, one recent example published in 2019, describing a chemical investigation of monoterpenoid indole alkaloids isolated from the roots of *Gelsemium elegans* (also briefly discussed in [Section 8](#sec8-molecules-24-04165){ref-type="sec"} of this review), stands out from the standpoint of hitherto unprecedented skeletal diversity \[[@B116-molecules-24-04165]\]. In particular, the following spirocyclic frameworks were encountered among the natural products isolated from this species: \[4.5.0\]---featured in 19-oxogelsevirine (**169**), gelsevirine (**170**), and koumimine (**171**); \[4.7.0\]---featured in gelsedethenine (**172**); and a unique \[4.8.0\] system---featured in humantenine (**173**) and 19,20-epoxyhumantenine (**174**) ([Figure 82](#molecules-24-04165-f082){ref-type="fig"}). 15. Conclusions and Perspectives {#sec15-molecules-24-04165} ================================ Spirocyclic scaffolds are omnipresent in the natural products domain. By analyzing the diversity of spirocyclic systems reported for natural products in the literature, one can appreciate an uneven distribution of such motifs according to the spirocycle type: certain motifs are more abundant than others and some are rather scarce, exemplified by only a handful of naturally occurring compounds. The most widespread are the \[5.5.0\], \[4.5.0\], and \[4.4.0\] spirocycles. In terms of associated bioactivities discovered and reported for spirocyclic products, these are mostly limited to the usual profiling in the context of antiproliferative, anti-inflammatory, and antimicrobial activities. However, the strong connections of spirocyclic frameworks to the natural product domain and their emerging privileged motif status in the synthetic drug discovery argues in favor of the need for more thorough panel profiling of all newly-discovered natural products, as novel and hitherto unprecedented bioactivity leads could be discovered. Certain scarcely-populated areas of the spirocyclic natural product space can be specifically developed into synthetic libraries and investigated for bioactivity. More spirocycles appear to have been discovered in the last 5--10 years, with an apparent advent of plant species giving rise to several types of spirocyclic frameworks in the course of their chemical investigation. The spirocyclic natural product discovery, therefore, appears to be on the rise and is likely to inspire new scaffolds for drug design and screening library development. The authors are indebted to Dmitry Dar'in of Saint Petersburg University for helpful discussions regarding this work. This research was funded by the Ministry for Science and Education of the Russian Federation under the Federal Target Program, "High-Priority Areas for Research and Development in Science and Technology for Russia in 2014--2020," grant number (RFMEFI57718X0285). The authors declare no conflicts of interest. {#molecules-24-04165-f001} ![Sesquiterpenes from *Schinus terebinthifolius* fruit containing a \[2.4.0\] spirocyclic moiety.](molecules-24-04165-g002){#molecules-24-04165-f002} {#molecules-24-04165-f003} {#molecules-24-04165-f004} {#molecules-24-04165-f005} {#molecules-24-04165-f006} {#molecules-24-04165-f007} {#molecules-24-04165-f008} ![The only known natural product containing a \[3.4.0\] spirocyclic motif.](molecules-24-04165-g009){#molecules-24-04165-f009} {#molecules-24-04165-f010} ![Sesquiterpene bis-lactones isolated from *Illicium* plants containing a \[3.7.0\] spirocyclic motif.](molecules-24-04165-g011){#molecules-24-04165-f011} ![Various naturally occurring \[4.4.0\] spirocyclic lactones.](molecules-24-04165-g012){#molecules-24-04165-f012} ![Limonoids **31**--**32** containing both a \[4.4.0\] and a \[2.4.0\] spirocyclic system.](molecules-24-04165-g013){#molecules-24-04165-f013} ![Tricyclic spirolactones (incorporating two \[4.4.0\] spirocyclic systems) isolated from *Penicillium purpurogenum*.](molecules-24-04165-g014){#molecules-24-04165-f014} ![New spirocyclic curcumanolides possessing a \[4.4.0\] spirocyclic system each, isolated from *Curcuma heyneana*.](molecules-24-04165-g015){#molecules-24-04165-f015} {#molecules-24-04165-f016} {#molecules-24-04165-f017} {#molecules-24-04165-f018} {#molecules-24-04165-f019} {#molecules-24-04165-f020} {#molecules-24-04165-f021} {#molecules-24-04165-f022} {#molecules-24-04165-f023} ![Structures of \[4.4.0\] spirocyclic pseudoindoxyl alkaloids fluorocurine (**61**), fungus-derived diketopiperazines (**62a**--**d**), brevianamide B (**63**), and rauniticine pseudoindoxyl (**64**).](molecules-24-04165-g024){#molecules-24-04165-f024} {#molecules-24-04165-f025} ![Natural products illustrating the range of heterospirocyclic \[4.4.0\]-sized motifs.](molecules-24-04165-g026){#molecules-24-04165-f026} ![1,6-Dioxaspiro\[4.4\]nonane secondary metabolite isolated from entomogenous fungus *Isaria cateniannulata*.](molecules-24-04165-g027){#molecules-24-04165-f027} {#molecules-24-04165-f028} ![Structure of fredericamycin A possessing a \[4.4.0\] spirocyclic motif.](molecules-24-04165-g029){#molecules-24-04165-f029} ![Overall diversity of \[4.4.0\] spirocyclic scaffolds represented in the natural products domain.](molecules-24-04165-g030){#molecules-24-04165-f030} {#molecules-24-04165-f031} {#molecules-24-04165-f032} {#molecules-24-04165-f033} {#molecules-24-04165-f034} ![The \[4.5.0\] spirocyclic sesquiterpenoids from rhizomes of *Acorus calamus*.](molecules-24-04165-g035){#molecules-24-04165-f035} {#molecules-24-04165-f036} {#molecules-24-04165-f037} ![Examples of natural \[4.5.0\] spirocyclic lactones.](molecules-24-04165-g038){#molecules-24-04165-f038} {#molecules-24-04165-f039} {#molecules-24-04165-f040} {#molecules-24-04165-f041} {#molecules-24-04165-f042} {#molecules-24-04165-f043} {#molecules-24-04165-f044} {#molecules-24-04165-f045} {#molecules-24-04165-f046} {#molecules-24-04165-f047} {#molecules-24-04165-f048} {#molecules-24-04165-f049} ![Enantiomers of \[4.5.0\] spirocyclic dihydrofuran 8,9-dehydrotheaspirone reported in the literature.](molecules-24-04165-g050){#molecules-24-04165-f050} {#molecules-24-04165-f051} {#molecules-24-04165-f052} ![Erectile function-potentiating toxin featuring a \[4.5.0\] spirocyclic motif.](molecules-24-04165-g053){#molecules-24-04165-f053} ![\[4.5.0\] spirocyclic quinochalcone saffloquinoside A isolated from *Carthamus tinctorius*.](molecules-24-04165-g054){#molecules-24-04165-f054} {#molecules-24-04165-f055} {#molecules-24-04165-f056} ![Structures of representative spirostaphylotrichins possessing a \[4.5.0\] spirocyclic motif.](molecules-24-04165-g057){#molecules-24-04165-f057} ![Current diversity of \[4.5.0\] spirocyclic scaffolds.](molecules-24-04165-g058){#molecules-24-04165-f058} {#molecules-24-04165-f059} ![Fomlactones A--C possessing a \[4.6.0\] spirocyclic moiety.](molecules-24-04165-g060){#molecules-24-04165-f060} {#molecules-24-04165-f061} {#molecules-24-04165-f062} ![Structure of auriculatol A possessing a \[4.6.0\] spirocyclic motif.](molecules-24-04165-g063){#molecules-24-04165-f063} ![Structure of \[4.6.0\] spirocyclic seconoriridone A.](molecules-24-04165-g064){#molecules-24-04165-f064} ![Structures of gelsenium alkaloids possessing a \[4.6.0\] spirocyclic system.](molecules-24-04165-g065){#molecules-24-04165-f065} ![Natural product phyllanthunin possessing a \[4.7.0\] spirocyclic moiety isolated from *Phyllanthus emblica*.](molecules-24-04165-g066){#molecules-24-04165-f066} ![Portimines A and B isolated from *Vulcanodinium rugosum* containing both one \[4.7.0\] and one \[4.5.0\] spirocyclic motif.](molecules-24-04165-g067){#molecules-24-04165-f067} {#molecules-24-04165-f068} {#molecules-24-04165-f069} {#molecules-24-04165-f070} ![Pteridic acids C and F isolated from *Streptomyces* sp. SCSGAA 0027 possessing a 1,7-dioxaspiro\[5.5.0\]undecane motif.](molecules-24-04165-g071){#molecules-24-04165-f071} {#molecules-24-04165-f072} {#molecules-24-04165-f073} {#molecules-24-04165-f074} ![Bioactive \[5.5.0\] spirocyclic meroterpenoids isolated from mangrove-derived fungus *Penicillium* sp.](molecules-24-04165-g075){#molecules-24-04165-f075} {#molecules-24-04165-f076} ![General structure of \[5.6.0\] spirocyclic orthoester periplosides.](molecules-24-04165-g077){#molecules-24-04165-f077} {#molecules-24-04165-f078} ![A \[5.6.0\] spirocyclic compound isolated from marine-derived fungus *Eurotium* sp. SCSIO F452.](molecules-24-04165-g079){#molecules-24-04165-f079} ![Vieloplain G isolated from *Xylopia vielana* containing a \[5.6.0\] spirocyclic scaffold.](molecules-24-04165-g080){#molecules-24-04165-f080} {#molecules-24-04165-f081} {#molecules-24-04165-f082} molecules-24-04165-t001_Table 1 ###### Occurrence of various ring combinations in the spirocyclic natural products analyzed in this review. ---------------------------------- ---------------------------------- ---------------------------------- ---------------------------------- ---------------------------------- ---------------------------------- ---------------------------------- Ring combinations        0 0 7 7 0 0  0 0 1 1 0 4  7 1 58 50 14 4  7 1 50 27 4 0  0 0 14 4 1 0  0 4 4 1 0 0 ---------------------------------- ---------------------------------- ---------------------------------- ---------------------------------- ---------------------------------- ---------------------------------- ----------------------------------

Introduction ============ There is a paucity of information concerning side-by-side comparison of real-time 3D echocardiography (RT3DE) and cardiac computed tomography (CCT) ventricular systolic performance assessment. We sought to compare those techniques with different temporal and spatial resolution, regarding left ventricle (LV) systolic function and volumes. Methods ======= We studied 67 consecutive patients (37 males, 55 ± 11 years) by RT3DE and by 64-slice CCT. We analysed by both techniques the LVEF, LVEDV and LVESV, and by RT3DE the LV dyssynchrony percentage indexes (DI%) (6, 12, 16 segment model). RT3DE and CCT data were compared by coefficients of determination (*r*: Pearson), Bland-Altman test and linear regression, 95% CI. Results ======= RT3DE data: LVEF ranged from 30 to 78.6 (63.1 ± 7.33)%; LVEDV ranged from 44.1 to 210 (104.9 ± 29.7) ml; LVESV from 11.4 to 149 (38.9 ± 19.3) ml; 6S DI% ranged from 0.25 to 29.7 (1.92 ± 4.67)%; 12S DI% ranged from 0.29 to 26.78 (2.10 ± 5.10)%; 16S DI% ranged from 0.29 to 27.81 (2.57 ± 4.37)%. CCT data: LVEF ranged from 28 to 86 (66 ± 8.4)%; LVEDV ranged from 51 to 212 (110.3 ± 31.2) ml; LVESV from 7 to 152 (38.2 ± 19.2) ml. Correlations relative to RT3DE and CCT were: LVEF (*r*= 0.74, *P*\< 0.0001, 95% CI = 0.6169 to 0.8379); LVEDV (*r*= 0.8213, *P*\< 0.0001, 95% CI = 0.7229 to 0.8870); LVESV (*r*= 0.9096, *P*\< 0.0001, 95% CI = 0.8929 to 0.9627). RT3DE (*x*) LVEF was compared with CCT (*y*) LVEF as: *y =*19.7862 + 0.6525 *x*, *R*^2^= 0.5586, *P*\< 0.0001, coefficient slope = 0.6525; RT3DE (*x*) LVEDV was compared with CCT (*y*) LVEDV as: *y =*15.7057 + 0.7823 *x*, *R*^2^= 0.6745, coefficient slope = 0.7823, *P*\< 0.0001; and RT3DE (*x*) LVESV was compared with CCT (*y*) LVESV as: *y =*2.7997 + 0.9439 *x*, *R*^2^= 0.8828, coefficient slope = 0.9439, *P*\< 0.0001. Conclusion ========== In this series, adequate correlation was observed between real-time 3D echocardiography and cardiac computed tomography regarding ventricular systolic function and geometry assessment.

Introduction {#S0001} ============ Capillaroscopy is a non-invasive diagnostic technique designed to evaluate small vessels of the microcirculation \[[@CIT0001]\]. Nailfold capillaries were first observed in the 17^th^ century with primitive magnifying equipment, and in the early 19^th^ century the first associations between inflammation and capillary alterations were made. Beginning with the works of Maurice Raynaud\'s \[[@CIT0002]\], research in the second half of the 19^th^ and first decades of the 20^th^ century established a direct link between capillary abnormalities and certain medical conditions. In the 1930s, interest in capillaroscopy began to decline, to rise again in the 1980s and 1990s. With the advent of modern digital equipment and evidence-based methodology, at the beginning of the 21^st^ century, we can witness a renaissance of the capillaroscopic technique and widespread recognition of its significance. Principles of capillaroscopy {#S0002} ============================ Microcirculation {#S20003} ---------------- The vasculature of the microcirculation consists of the smallest blood vessels in the human body -- arterioles, capillaries and venules. The capillaries, in turn, are formed by an arterial limb, capillary loop and venous limb. This pattern is found in every tissue except liver, spleen and bone marrow, where capillaries are replaced by sinusoids. Microcirculation\'s main function is capillary exchange -- delivery of oxygen and nutrients to tissues and removal of carbon dioxide and waste products \[[@CIT0003]\]. In a systemic disease in which vascular damage is one of the pathogenetic factors, abnormalities in capillary morphology can be observed long before the onset of clinical symptoms. In patients already diagnosed with a systemic disease, damage to the capillaries may reflect the involvement of internal organs and help determine the stage of the disease \[[@CIT0004]\]. Microcirculation in capillaries is routinely evaluated within the skin of the nailfold. The entire skin abounds in capillaries; however, they run perpendicular to the skin surface, and only the tip of the loop is visible. In the nailfold, terminal rows of capillaries run parallel to the skin surface and, therefore, all morphological details and the nature of the blood flow can be examined \[[@CIT0005]\]. Performing capillaroscopy {#S20004} ------------------------- A range of optical devices can be used to perform capillaroscopic examination, for example a dermatoscope, an ophthalmoscope or a traditional microscope. However, it is best to perform the examination with equipment dedicated for capillaroscopy, i.e. a stereomicroscope or a digital videocapillaroscope. Of these two, the latter is preferred, as a hand-held probe can be easily used in every situation, e.g. for bedside examination or in patients with severe flexion contractures. In order to enhance skin transparency, a drop of immersion oil is applied to the nailfold before capillaroscopy. In a routine examination, all fingers except the thumbs are evaluated. Each finger should be examined in two magnifications: × 50, showing the general architecture of the terminal capillary row, and × 200--300, in which morphological details of a single capillary can be assessed. Digital equipment allows the obtained images to be stored and used for an objective comparison if the patient needs subsequent examinations \[[@CIT0006]\]. Patient preparation {#S20005} ------------------- Environmental factors can cause physiological constriction of capillaries, thus greatly affecting the capillaroscopic image. Before the examination, patients should be acclimatised at a temperature of 20--22°C for 15--20 minutes, and should refrain from smoking and drinking caffeine for 4 hours. Capillaroscopy should not be performed if the patient has recently (in the last 3 weeks) undergone any cosmetic procedure involving the nailfold area, since the consequent micro-traumas can give false-positive results \[[@CIT0006]\]. Important capillaroscopic parameters and normal capillary image {#S0006} =============================================================== There are several parameters which should be evaluated during capillaroscopy ([Table I](#T0001){ref-type="table"}) \[[@CIT0001], [@CIT0007], [@CIT0008]\]. A normal capillaroscopic image of a healthy control is shown in [Figure 1](#F0001){ref-type="fig"}. {#F0001} ###### Important capillaroscopic parameters Capillaroscopic parameter Normal image ------------------------------------------------------ ----------------------------------------------------- Skin transparency and visibility Transparent, capillaries clearly visible Pericapillary oedema Absent Subpapillary venous plexus Visible in up to 30% of healthy people Capillary array and architecture Straight capillaries, perpendicular to the nailfold Capillary morphology U-shaped Capillary loop diameter \< 20 µm Tortuosity Usually absent Dilated (20--50 µm) and giant (\> 50 µm) loops Absent Ramified capillaries Absent Neoangiogenesis Absent Haemorrhages, hemosiderin deposits Usually absent, may be present after local trauma Capillary density 9--13 in 1 linear millimetre Avascular areas (distance between 2 loops \> 500 µm) Absent Capillary blood flow Dynamic, no stasis or thrombosis Raynaud\'s phenomenon {#S0007} ===================== The most important indication for capillaroscopy is Raynaud\'s phenomenon (RP) \[[@CIT0008]\]. This condition usually presents in the fingers or toes, more rarely in the nose and ears, and involves a sequence of skin colour changes, attributed to vasospasm and subsequent vessel dilation. These include skin turning white (vasoconstriction), then blue (hypoxia) and finally red (reperfusion). Raynaud\'s phenomenon can be primary (idiopathic) or secondary to numerous conditions. Capillaroscopy plays a pivotal role in differentiating one from another \[[@CIT0009]\]. In primary RP capillaries of the nailfold are normal, with no apparent abnormalities. It should be stressed that diagnosis of the primary condition cannot be based solely on capillaroscopy; it also requires normal levels of acute phase reactants and absence of any other clinical symptoms. It is recommended to perform capillaroscopy every 12--24 months in primary RP, since up to 10% of these patients will develop a connective tissue disease, sometimes after decades. In RP secondary to non-rheumatic conditions, the capillaroscopic image may be normal or present nonspecific alterations. On the other hand, in connective tissue diseases, especially in scleroderma spectrum disorders, microangiopathy of the peripheral microcirculation is prominent, with the presence of specific patterns that can be attributed to particular diseases \[[@CIT0010]\]. Scleroderma spectrum disorders {#S0008} ============================== Scleroderma spectrum disorders are a heterogeneous group of connective tissue diseases linked to systemic sclerosis (SSc) and exhibiting characteristic capillaroscopic abnormalities \[[@CIT0011]\]. These include giant capillaries, loss of capillaries with areas of avascularisation, ramified capillaries with pathological neoangiogenesis and severe derangement of capillary architecture. All of these pathologies may be present to a different extent in scleroderma spectrum disorders. In SSc, however, they appear and evolve in a clearly defined sequence called the scleroderma pattern. Systemic sclerosis {#S0009} ================== Systemic sclerosis (SSc) is a rare connective tissue disease presenting with diffuse fibrosis and dysfunction of internal organs due to microangiopathy \[[@CIT0012]\]. In well over 95% of SSc patients, peripheral microangiopathy follows a typical scleroderma pattern, consisting of 'early', 'active' and 'late' phases ([Table II](#T0002){ref-type="table"}, [Fig. 2](#F0002){ref-type="fig"}) \[[@CIT0013]\]. Differentiation of these patterns is of great clinical significance, since the 'early' pattern can be detected many years before full clinical manifestation of SSc, and progression to 'active' and 'late' patterns closely corresponds to internal organ involvement \[[@CIT0014]\]. The importance of capillaroscopy is underlined by its inclusion in the current EULAR classification criteria of SSc \[[@CIT0015]\]. {#F0002} ###### Scleroderma pattern \[[@CIT0013]\] ------------------ -------------------------------------------------------- 'Early' pattern Preserved capillary architecture\ Few giant capillaries and microhaemorrhages\ No evident capillary loss 'Active' pattern Slight disorganisation of capillary architecture\ Numerous giant capillaries and microhaemorrhages\ Moderate capillary loss\ Few ramified capillaries 'Late' pattern Severe derangement of capillary architecture\ Almost absent giant capillaries and microhaemorrhages\ Significant capillary loss with avascularisation\ Numerous ramified capillaries with neoangiogenesis ------------------ -------------------------------------------------------- Capillaroscopy can be useful in predicting development of finger ulcerations in SSc patients. The capillaroscopic skin ulceration risk index (CSURI) is a tool devised by Sebastiani et al., based on the formula D × M/N^2^ (D -- diameter of the biggest giant loop, M -- number of giant loops, N -- number of all loops) ([Fig. 3](#F0003){ref-type="fig"}) \[[@CIT0016]\]. In a prospective clinical trial, CSURI correctly predicted formation of new ulcerations within 3 months following capillaroscopy, with a positive predictive value of 81% for CSURI \> 2.96 and a negative predictive value of 93% for CSURI \< 2.96 \[[@CIT0017]\]. {#F0003} Dermatomyositis and antisynthetase syndrome {#S0010} =========================================== Dermatomyositis (DM) and its variant, antisynthetase syndrome, are connective tissue diseases with prominent muscle and skin involvement \[[@CIT0018], [@CIT0019]\]. Despite diverse clinical presentations, they share to some extent a common pathogenesis with SSc, i.e. primary damage to microcirculation. Presence of severe capillary abnormalities such as giant loops and ramified capillaries with intense neoangiogenesis is common in DM patients, with a typical scleroderma pattern occurring in up to 40% of cases \[[@CIT0020]\]. Contrary to SSc, in DM the connection between the capillary image and internal organ involvement has not yet been fully studied. However, a good response to treatment and, later on, remission of the disease are often accompanied by normalization of the capillary image \[[@CIT0021]\]. Capillaroscopy can be very useful in differentiating DM and polymyositis (PM). In PM, the suggested pathogenetic mechanism is an autoimmune response of cytotoxic T cells, not microangiopathy as postulated in DM. Capillaroscopic findings support this theory, since in PM patients capillary abnormalities are rare, mild and comparable to the healthy population \[[@CIT0020]\]. Mixed connective tissue disease {#S0011} =============================== Mixed connective tissue disease (MCTD) is a clinical entity that combines symptoms of SSc, DM, rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) \[[@CIT0022]\]. The incidence and extent of capillary involvement in MCTD vary greatly, from a normal image through mild non-specific abnormalities to severe microangiopathy, rarely with a scleroderma pattern. Scleroderma-like abnormalities in MCTD often (76% of patients) accompany interstitial lung disease (ILD) \[[@CIT0023]\]. Scleromyositis {#S0012} ============== Scleromyositis (ScM) is an overlap syndrome combining features of SSc and DM/PM, associated with the presence of anti-PM/Scl antibodies. The clinical course of this disease is usually more benign than SSc, with a normal capillary image. In more severe cases, however, alterations of nailfold capillaries, especially presence of the scleroderma pattern, may predict emergence of symptoms typical of systemic sclerosis, e.g. ILD \[[@CIT0024]\]. Primary biliary cirrhosis {#S0013} ========================= Primary biliary cirrhosis (PBC) is a chronic, autoimmune liver disease, characterized by progressive destruction of bile ducts, cholestasis and presence of antimitochondrial antibodies (AMA). Incidence of PBC and SSc is strongly associated, with PBC being the most common liver disease in SSc patients. Capillaroscopy is a potential screening method in PBC patients with symptoms suggesting SSc. Similarly to ScM, scleroderma-like capillary abnormalities may correlate with ILD \[[@CIT0025]\]. Other connective tissue diseases {#S0014} ================================ Extensive research has been performed on capillary alterations in other connective tissue diseases, but no specific patterns have been found so far. In SLE, nailfold capillaries are often normal, and mild changes occur in up to 30% of patients. The most frequent alterations include increased tortuosity, elongated loops, and meandering bizarre loops. These abnormalities are often associated with discoid lesions on the skin of the fingers. Patients with RA are reported to have increased capillary tortuosity and prominent, clearly visible subpapillary venous plexus. Antiphospholipid syndrome (APS) and small-vessel vasculitides can both present as multiple microhaemorrhages and hemosiderin deposits, with no morphological changes in capillaries. However, it should be noted that such an image can be seen even after a minor trauma, so caution is advised when reporting capillaroscopic findings in patients with APS or vasculitis \[[@CIT0026]--[@CIT0028]\]. Interstitial lung disease {#S0015} ========================= Interstitial lung disease (ILD) commonly accompanies scleroderma spectrum disorders, but can also be associated with other rheumatic diseases, infection, environmental factors, or appear with no obvious cause \[[@CIT0029]\]. Capillaroscopy may be useful in equivocal cases, since patients with ILD and scleroderma-like capillary abnormalities are likely to have or soon develop SSc or an associated disease \[[@CIT0030]\]. Limitations of capillaroscopy {#S0016} ============================= It should be duly noted that apart from connective tissue diseases, more common medical conditions may be a cause of peripheral microangiopathy. In the course of both diabetes and hypertension, severe capillary alterations can be observed, with heterogeneously enlarged loops, capillary loss and neoangiogenesis. A pathological capillary image may also result from prolonged exposure to vibration, chemical agents, ionizing radiation and other occupational factors. Wide field capillaroscopy performed with a stereomicroscope can have low reproducibility, due to qualitative and subjective assessment. On the other hand, videocapillaroscopy requires skill and precision to perform properly, since even small pressure of the probe on the nailfold can disrupt capillary blood flow and influence the study results. Conclusions and summary of clinical applications {#S0017} ================================================ Capillaroscopy is an easy and safe method for diagnosing pathologies of peripheral microcirculation. Its applications are summarized in [Table III](#T0003){ref-type="table"}. With digitized equipment, capillaroscopy allows for precise qualitative and quantitative evaluation of the microcirculation and is a valuable tool in the rheumatologists' daily practice. ###### Indications for capillaroscopy -------------------------------------------------------------------------- Differential diagnosis of primary and secondary Raynaud\'s phenomenon Staging of systemic sclerosis Predicting formation of new digital ulcers in systemic sclerosis (CSURI) Response assessment of dermatomyositis treatment Differential diagnosis of dermatomyositis and polymyositis Screening for interstitial lung disease in MCTD, ScM and PBC Differential diagnosis of interstitial lung disease -------------------------------------------------------------------------- The authors declare no conflict of interest. This paper is part of the first author\'s PhD thesis.

In a unique collaboration between The Seventh Hospital of Hangzhou, the International Society for Traumatic Stress Studies (ISTSS), and the Zhejiang Behavior Medicine Association, the international conference "Posttraumatic stress: state-of-the-art research and clinical implications for China" was organized in Hangzhou, China, on 17--19 October 2014. The organizational team was led by Dr Zhonglin Tan, a psychiatrist from The Seventh Hospital of Hangzhou, who has spent a year with Professor Olff\'s research team in Amsterdam, The Netherlands. With the support of Dr Zhang, Director of The Seventh Hospital of Hangzhou, and with guidance from Professor Olff, Dr Zhonglin Tan successfully created a full program. Over 300 participants, the majority of them from China, enjoyed one full day in both English and Chinese (with simultaneous translation) and the rest of the meeting in Chinese only. The renowned board members of ISTSS (Kaysen, Stappenbeck, Rhew, & Simpson, [@CIT0005]; Kim, [@CIT0006]; Kudler, [@CIT0007]; O\'Donnell, [@CIT0008]; Olff et al., [@CIT0009]; Schnyder, [@CIT0012]) and other international experts (Jongedijk, [@CIT0004]; Wu, [@CIT0015]) had volunteered to speak at this first ISTSS conference in China, as part of ISTSS\'s *global meetings program* (see [www.istss.org](http://www.istss.org)). Recognizing that trauma is a global issue (Schnyder, [@CIT0011]), ISTSS is opening doors to parts of the world where, up till now, no ISTSS conferences have been held, being eager to spread knowledge, to collaborate, and to learn from each other, and thus to advance the field of traumatic stress worldwide. Next to the international speakers, a strong representation of Chinese PTSD experts was seen who gave keynotes on topics ranging from biological stress systems (Bao & Swaab, [@CIT0019]) web-based interventions (Wang, [@CIT0020]; Wang & Mearcker, [@CIT0014]) and post-disaster psychosocial care (Zhang S, [@CIT0017]) to the exciting field of Chinese traditional medicine for PTSD (Zhang Y-H, [@CIT0018]). Apart from keynote lectures, this collection contains the abstracts of a selection of over 100 submissions, the authors of which were awarded for their excellent contribution to the field with a certificate and abstract publication in the *European Journal of Psychotraumatology* (Cao, Wang, Wang, Qing, & Zhang, [@CIT0001]; Hall, Chen, Wu, Zhou, & Latkin, [@CIT0002]; Hong, Cao, & Efferth, [@CIT0003]; Reifels et al., [@CIT0010]; Teng, Hall, & Li, [@CIT0013]; Xu et al., [@CIT0016]). Miranda Olff Editor-in-Chief Zhonglin Tan Hangzhou Mental Health Center Hangzhou Seventh Peoples' Hospital Hangzhou, People\'s Republic of China

Background {#Sec1} ========== Recent studies have demonstrated that several major nosocomial pathogens are shed by patients and contaminate environmental surfaces at concentrations sufficient for transmission \[[@CR1]--[@CR3]\]. Such pathogens can survive for extended periods despite cleaning with chlorine-releasing disinfectants \[[@CR4], [@CR5]\] and can be transferred to the hands of healthcare workers \[[@CR6]\]. The spread of nosocomial pathogens has been linked to poor hand-hygiene practices. However, healthcare workers are more likely to contaminate their hands from touching the patient environment than from patient contact \[[@CR6]\]. Mounting evidence demonstrates that outbreak strains of methicillin-resistant *Staphylococcus aureus* (MRSA), extensively drug-resistant (XDR) *Acinetobacter baumannii*, *Pseudomonas aeruginosa* and (extended spectrum beta-lactamase)-producing *Enterobacteriaceae* survived significantly longer on environmental surfaces than non-outbreak strains, indicating a possible fitness advantage \[[@CR7]--[@CR13]\]. In healthcare settings, surfaces are usually decontaminated using liquid chemical disinfectants, often chlorine derivatives \[[@CR14]\]. However, these products have some drawbacks, they are usually toxic to humans, they display chemical reactivity, and they require long periods of contact (up to 15 min) with surfaces to kill microorganisms \[[@CR15]\]. Moreover, some nosocomial pathogens are resistant to many disinfectants \[[@CR16]--[@CR18]\]. The development of portable steam generators has made the disinfection of environments more practical \[[@CR19]--[@CR21]\]. The aim of the present study was to assess "in vitro" the ability of the overheated dry-saturated steam vapour system to kill multidrug and XDR nosocomial pathogens on surfaces, and to define the antimicrobial spectrum and the contact times required by this system. Our results were then compared with those obtained using sodium hypochlorite, an agent commonly used in clinical sanitation procedures \[[@CR14], [@CR15]\]. Methods {#Sec2} ======= Surfaces {#Sec3} -------- The in vitro tests were carried out in a microbiology laboratory. Glass surfaces were chosen because they are flat, inert, easy to contaminate and highly resistant to chemical products and to heat. The dimensions of the surfaces were 50 × 50 cm, with a thickness of 30 mm and a weight of 38 kg/m^2^. Disinfection using the overheated dry-saturated steam vapour system {#Sec4} ------------------------------------------------------------------- The steam generator device consists of a professional steam generator (Sani System Polti, Medical Division Polti s.p.a., Como, Italy) (Fig. [1](#Fig1){ref-type="fig"}). The dimensions of the unit were 47 × 45 × 90.5 cm, with a weight of 27.5 kg. The portable unit was outfitted with a hose connected to a steam dispenser. The overheated dry-saturated steam vapour was high temperature steam generated in the steel boiler that reached a pressure of 6 bar and was then further superheated in an expansion chamber to generate a dry saturated steam vapour at 180 °C. The unit was filled with tap water. Fifteen minutes before use, the unit was activated to reach the maximum operating boiler pressure (6 bar) in accordance with the manufacturer's instructions (Sani System Polti). Protection equipment, such as heat-resistant gloves, safety aprons or glasses, is not required by workers when operating this system.Fig. 1Overheated dry-saturated steam vapour disinfection device used in this study. (Sani System Polti, Medical Division Polti s.p.a., Como, Italy) Disinfection with sodium hypochlorite {#Sec5} ------------------------------------- The commercial product Decs containing sodium hypochlorite at 2.8 % (2.7 % active *chlorine)* (Lombarda H s.r.l, Albairate, Milan, Italy) was used in this study. The stock solution of the product was diluted in sterilised water to a final concentration of 5 % to obtain approximately 1400 ppm active chlorine. This is the concentration of sodium hypochlorite solution usually used for disinfection in hospitals \[[@CR22]\]. Culture methods {#Sec6} --------------- The bactericidal effects of the overheated dry-saturated steam vapour and of sodium hypochlorite were evaluated using seven environmental organisms, including Gram-positive and Gram-negative bacteria, yeast and fungi: XDR *A. baumannii* (strain 4500/2010), *P. aeruginosa* (strain 3637/2006), MRSA (strain 3582/2006), high-level aminoglycoside-resistant (HLAR) *Enterococcus faecalis* (strain 3084/2005), carbapenemase-producing *Klebsiella pneumoniae* (KPC; strain 4640/2012), *Candida parapsilosis* (strain 4093/2009) and *Aspergillus fumigatus* (strain 3430/2006). These environmental strains were isolated between January 2005 and December 2012 in the neonatal and the adult intensive care units of the University Hospital "Federico II" in Naples, Italy, during environmental microbiological investigations performed to identify sources and reservoirs of infection in the course of nosocomial outbreaks \[[@CR10], [@CR11]\]. Environmental isolates were identified by commercial systems (VITEK^®^ 2 automatic system; bioMèrieux Marcy-L'Etoile, France and Becton--Dickinson Phoenix, Phoenix Technologies Ltd, San Jose, CA, USA). All isolates were stored at −80 °C in glycerol solution. Susceptibility testing and screening {#Sec7} ------------------------------------ The strains used in our study were selected because of their antimicrobial resistance phenotypes. Antimicrobial susceptibility patterns were analysed using an automated system (BD Phoenix) and by manual methods (i.e., Kirby-Bauer disk diffusion assay, Etest and microdilution tests) and the results were interpreted according to EUCAST \[[@CR23]\]. Antimicrobial susceptibility testing revealed a multidrug-resistant antibiotype for all of the isolated microorganisms. The antimicrobial susceptibility patterns of strains included in the study are listed in Tables [1](#Tab1){ref-type="table"} and [2](#Tab2){ref-type="table"}.Table 1Antimicrobial susceptibility patterns of the strains tested in this studyAntibioticMIC value*KPC*-*K. pneumoniaeXDR A. baumanniiP. aeruginosaE. faecalis HLARC. parapsilosisA. fumigatus*Amikacin\<4\>32\>32Amoxicillin-clavulanate\>16/8\>16\>32Ampicillin\>16\>32Ampicillin-sulbactam\>16/8Aztreonam\>16Cefazolin\>16\>64Cefepime2Cefotaxime\>16\>32\>64Ceftazidime\>32\>16\>16Ceftriaxone\>32Chloramphenicol16\>16Ciprofloxacin\>2\>2Gentamicin≤2\>8\>8Gentamicin high-level\>2000Imipenem8\>8\>8Levofloxacin≤1\>2\>2Meropenem8\>8\>8Netilmicyn high-level\>2000Nitrofurantoina≤16\>512Norfloxacina≤2Piperacillin\>64\>6416Piperacillin--tazobactam≥64/4≥64/416Streptomycin high-level\>2000Tetracycline\>8\>8Trimethoprim--sulfamethoxazole≤0.5/9.5\>2/385-Flucytosine\>32^a^Fluconazole\>64^a^Itraconazole\>1^a^\>1^a^Anidulafungin\>2^a^\>2^aa^Sensititre YeastoneTable 2Disk diffusion susceptibility test and Etest values of *K. pneumoniae* strain identified as a KPC producerDouble disk test (mm zone diameter)MIC of drug Etest value (mg/L)MEM (10 µg)MEM plus boronic acid (600 µg)MEM/MEM boronic acidInterpretationErtapenemInterpretation1621\>4+\>1+ ### In vitro time-kill tests {#Sec8} The bactericidal effects of the overheated dry-saturated steam vapour and the sodium hypochlorite on multidrug-resistant nosocomial pathogens were assessed by measuring viable cell counts using the quantitative time-kill test \[[@CR24]\] as recommended by the European Committee for Standardization (CEN EN1276) with or without bovine serum albumin (BSA 0.3 g/100 mL) \[[@CR25]\]. The quantitative time-kill test was performed as previously described \[[@CR24]\]. In brief, a logarithmic-phase culture was adjusted to ca. 10^9^ CFU/mL for bacteria and 10^7^ CFU/mL for yeasts and fungi in Luria--Bertani broth. The test was carried out on glass surfaces initially contaminated with 100 μL of microbial suspension, prepared as above, with and without BSA. The surfaces were sanitised with the overheated dry-saturated steam vapour generated at 180 °C at different time points (1, 2, 3 min, up to 8 min). At each time point, the surfaces were rinsed with 1 mL of a sterile physiological solution, and 100 μL was removed from the rinsing solution and serially diluted (1:10, 1:100, 1:1000, 1:10,000) in phosphate-buffered saline (pH 7.4) and spread on Bacto D/E neutralizing agar (Becton--Dickinson). Plates were incubated at 37 °C for 24 h for bacterial strains and at 32 °C for 72 h for *C. parapsilosis* and *A. fumigatus* strains. Each experiment was performed in triplicate. Time-kill tests following hypochlorite disinfection were carried out on glass surfaces with and without BSA. The glass surfaces were contaminated at eight different points with 100 μL of each microbial suspension. The surfaces were sanitised at each inoculum point with 900 µL of disinfectant solution at room temperature (22 °C) and then 100 µL of each suspension was removed from the surfaces at different time points and spread on Bacto D/E neutralizing agar after serial dilution, as described above. Plates were incubated under the same conditions as above. The antimicrobial action of chlorine was neutralised by sodium thiosulphate \[[@CR26]\], which is contained within the D/E neutralizing agar at a concentration of 6.0 g/L. Each experiment was performed in triplicate. Viable cell counts of the bacterial, yeast and fungal strains were evaluated on agar plates (Bacto D/E neutralizing agar) after incubation under the conditions described above. The results of the in vitro tests were interpreted in accordance with the CEN standards \[[@CR25]\]. According to these standards, the bactericidal activity of disinfectant was defined as a ≥4 log 10 CFU/mL decrease in the viable count compared with the initial inoculum. Statistical analysis {#Sec9} -------------------- Data were analysed using GraphPad Prism v.5.04 software (GraphPad Software, La Jolla, CA, USA). The significance of the reduction in viable counts was analysed using the Student's *t* test and the reduction was considered significant when the P value was \<0.05. Results and discussion {#Sec10} ====================== Overheated dry-saturated steam vapour {#Sec11} ------------------------------------- The bactericidal activity of the dry saturated, steam vapour system was observed for Gram-negative bacteria at a temperature of 180 °C after 5 min of treatment without BSA. A significant reduction in *P. aeruginosa* (p = 0.0004, as determined by the Student's t test compared with untreated samples), XDR *A. baumannii* (p \< 0.0001) and KPC (p \< 0.0001), from an initial concentration of 10^9^ CFU/mL, was observed after 2 min of treatment with this system. The same results were obtained in the presence of BSA (Fig. [2](#Fig2){ref-type="fig"}). For MRSA and HLAR *E. faecalis*, bactericidal activity was observed after 5 min contact time with a significant reduction (p \< 0.0001) from the initial concentration (10^9^ CFU/mL) after 3 min of treatment, in the presence and absence of BSA (Fig. [2](#Fig2){ref-type="fig"}).Fig. 2Microbial time-kill diagrams after dry saturated steam vapour treatment with or without BSA A longer contact time (7 min) was necessary to reduce the viable count of *C. parapsilosis* and *A. fumigatus* (to 10^7^ CFU/mL) without or with BSA, although a significant reduction in the yeast (p \< 0.0001) and fungi (p = 0.0004) counts from the initial concentration (10^7^ CFU/mL) was observed after 3 min of treatment (Fig. [2](#Fig2){ref-type="fig"}). Sodium hypochlorite {#Sec12} ------------------- The in vitro tests with sodium hypochlorite at 5 % in the absence of an organic substance (i.e. BSA) resulted in an overall reduction in the viable count within 5 min for Gram-negative and Gram-positive bacteria (from 10^9^ CFU/mL) (Fig. [3](#Fig3){ref-type="fig"}). A longer contact time (7 min) was necessary under the same conditions without BSA to reduce the yeast and fungi count (from 10^7^ CFU/mL) (Fig. [3](#Fig3){ref-type="fig"}).Fig. 3Microbial time-kill diagrams after chlorine-derivative treatment with or without BSA Conversely, 5-min hypochlorite treatment in the presence of an organic substance reduced the initial viable count (10^9^ CFU/mL) to 10^5^ CFU/mL (4 log decrease) for all bacterial strains, except HLAR *E. faecalis* that showed a 2 log decrease (10^9^ to 10^7^ CFU/mL). The reduction in *C. parapsilosis* and *A. fumigatus* counts (10^7^ to 10^5^ CFU/mL) at 7 min was only 2 log units in the presence of BSA (Fig. [3](#Fig3){ref-type="fig"}). The disinfection of all multidrug-resistant pathogens was rapid and complete. This is an important result because surface-related transmission of antibiotic-resistant microorganisms is a growing threat in healthcare settings and chemical disinfectants may have the potential to select cross-resistance to antibiotics \[[@CR16], [@CR27]\]. The in vitro tests using sodium hypochlorite at 5 % in the absence of an organic substance also resulted in an overall reduction in bacterial and mycotic concentrations. In the presence of an organic substance, the antimicrobial activity of sodium hypochlorite was reduced. The use of disinfectants requires an initial cleaning step to remove the organic matter that would otherwise "consume" the oxidizing disinfectant \[[@CR14], [@CR15]\]. Therefore, in the presence of an organic substance, a large quantity of oxidising disinfectant (containing chlorine and ozone) may be required, often at higher-than-standard concentrations. The disinfecting effect of chlorine is only visible when the "requirement" of organic substances has been met \[[@CR14], [@CR15]\]. In contrast, the steam vapour device depends on heat for efficacy and the presence of organic matter does not influence the effectiveness \[[@CR28]\]. Moist heat acts by denaturation and coagulation of protein, breakage of DNA strands and loss of functional integrity of cell membranes, resulting in cell death \[[@CR29]\]. This system could help to reduce the risk of spreading nosocomial infections in healthcare facilities. Our time-kill curve studies revealed the time-dependent effect of killing by steam vapour disinfection with an overall drop in microbial counts obtained at either 5 or 7 min depending on the microorganism. These findings are in accordance with previous studies performed using saturated steam showing that this method is effective in decontaminating surfaces contaminated with high concentrations of pathogenic microorganisms, killing 100 % of bacteria under all experimental conditions \[[@CR19]--[@CR21]\]. The antimicrobial activity of the steam disinfection system has great potential in the disinfection of contaminated hospital environments, because many of the most important nosocomial pathogens, such as *P. aeruginosa* and *A. baumannii*, have natural resistance to liquid chemical disinfectants and especially to quaternary ammonium compounds \[[@CR30]\]. The saturated steam vapour disinfection system has a broad range of activity \[[@CR19]--[@CR21]\] and this technology is also considered to have great potential for eliminating biofilms, aggregates of active cells embedded within a polymeric matrix and attached to a biotic or abiotic surface \[[@CR31]\]. Conclusions {#Sec13} =========== The steam vapour system has been proven to reduce or completely eliminate microbial contamination on hard surfaces at earlier and later time points, respectively. This portable device quickly reduced and then eliminated microbial loads in the presence or absence or an organic substance in contrast to sodium hypochlorite, whose disinfectant effect is only visible when the "requirement" of the organic substance has been met. The findings of this study suggest that the portable vapour disinfection system is a viable alternative to available chemical disinfectants, including chloride derivatives, for the disinfection of hospital environmental surfaces. MB conceived and designed the study. RBuo, DM, RBel and NG performed the assays. MG and AN analysed and interpreted the results of time-kill tests. RZ and MT supervised the study. MB and RZ wrote the paper. All authors read and approved the final manuscript. Acknowledgements {#FPar1} ================ The authors would like to thank Professor V. Iaccarino from the Department of Diagnostic Imaging and Radiotherapy, Section of Vascular and Interventional Radiology AOU Federico II Napoli, for his enthusiastic support. The authors received no financial support to carry out this study. Competing interests {#FPar2} =================== The authors declare that they have no competing interests.

Problem ======= Quality and concise handover between clinicians is essential for patient safety \[[@R1]\]. At this time of increased risk during the patient\'s pathway the handover sheet is an important adjunct helping to reduce this risk \[[@R2]\]. The handover sheet within elective orthopaedic ward care takes on a variety of formats, with Microsoft (MS) Word or Excel being utilised frequently. These are susceptible to the erroneous deletion of data and on occasions the complete document. The change of team to which the patient is allocated within the speciality can also lead to data loss when being deleted from one list and added to another. Documents are rarely automated, requiring time to be spent ordering the patients by location and formatting the document. Following the amalgamation of two services into a large, new hospital it was recognised that a complete solution was required for the day to day management and handover of elective patients. Background ========== The GMC\'s Good Medical Practice states that doctors should share all relevant information with colleagues involved in their patients' care within and outside the team, including when they hand over care and go off duty \[[@R3]\]. Poor handover leaves on-call weekend doctors with busier shifts, a more stressful time prioritising patients and tasks whilst "fire-fighting" the new issues as they arise \[[@R4]\]. Clear patient handovers with concise and necessary information only lead to time-efficient, safer ward rounds. The handover sheet allows a snap shot to the patient\'s problems, the ongoing plan, outstanding jobs, and other relevant information. Keeping them up to date can be time consuming and this normally falls to the junior doctors to undertake. Their importance is evident within the current rules for working hours where most juniors rarely work a "normal" week, which leads to increased frequency of handovers \[[@R2]\]. They are also crucial when changing from the well staffed week-day rota to the weekend on-call team who have not met all of the patients. At this time the handover list and the information contained within is relied upon heavily \[[@R5]\]. As afore mentioned, IT is recognised as being an important part of patient handover, providing uniformity, continuity and an audit trail \[[@R6],[@R7]\]. Baseline measurement ==================== Research was conducted within the orthopaedic junior doctors cadre to establish any concerns as to the current documents being used to conduct patient handover (MS Word). A questionnaire was distributed among the doctors (Pre-intervention questionnaire). Replies were completed by 6 doctors. Options on the questionnaire - dependant on question type - see pre-intervention questionnaire for further details. ALWAYS / MOSTLY / RARELY / NEVER 0-2 / 3-4 / 5-6 / 7-8 / 9-10 The questionnaire demonstrated the following results (Median): 1. Handover lists were readily accessible - MOSTLY 2. Handover lists were regularly updated by the SHO covering those patients - MOSTLY 3. Handover lists were automatically arranged into sensible layouts - NEVER 4. Patient data was lost when patients were transferred from one list to another - 3-4 TIMES 5. Patients were dropped off patient lists accidentally - 5-6 TIMES 6. These patients were dropped off these lists predominantly at the WEEKEND 7. Handover sheets allowed time efficiency at the weekend - RARELY Comments from freehand notes: Only 1 person on each team could access the list at once Messy spreadsheet rarely contained ward locations There was a lack of continuity as to the layout of handover sheets between the four teams Patients who switched team due to changing consultant had been taken off a list and not added to another resulting in them not being reviewed/managed correctly See supplementary file: ds4728.docx - "Pre-intervention questionnaire" Design ====== It was deduced that the solution for this issue was within a bespoke computer interface. The author had reasonable experience with MS Access and its functions, and chose to create a system that could be implemented and serviced without the hospital IT department\'s assistance. A table was created within MS Access, on which the data was to be stored, this was to be the "foundations" of the database with the information it held being interrogated/updated indirectly by users. Queries that could sort and display relevant data according to criteria were created to subdivide by Consultant and therefore into the four teams that made up the department. A form was created to allow the information from the table, selected by a query, to be accessed by the teams and edited as required. Specific functions were embedded within these forms, allowing new patients to be added and patients to be discharged (a two stage process preventing accidental deletion). Movement between team lists was enabled using similar function enabled buttons. Reports were created allowing team lists to be generated from information within the relevant form at the click of a button, this could then be printed. The patients on the list were ordered by ward and then by surname alphabetically. A weekend report was also available which allowed patient lists to be printed by ward. This enabled better time efficiency during an already busy shift avoiding checking of multiple lists. Reports were automatically headed with the date and had a list of useful telephone numbers within the footnote. Finally the robustness of the Database was ensured, with a "Save and Close" button with a number of background functions running. It saved the document followed by closing the document. Each time the Database was closed it would save a copy of the background Table as an Excel spreadsheet to an archive folder allowing teams to check back through lists at a later date. Strategy ======== PDSA cycle 1 The Pre-implementation questionnaire was produced and handed out to members of the team. This allowed for an assessment of the current state of patients lists and their uses and downfalls, enabling a set of criteria on which to improve. These criteria also formed that basis upon which the new database was subsequently built. PDSA cycle 2 \- Following feedback from the questionnaire and initial demonstration/teaching session a number of ideas had been put forward: \- Multi-user functionality was poor and needed improving. \- A couple of program issues had been identified preventing the smooth running and failing its aim of being a robust system. \- A few spelling corrections were required! These were all addressed, with multi-user functionality being addressed by the "splitting" of the database, programming issues (navigation buttons not working correctly) were resolved. Spelling errors were corrected. PDSA cycle 3 - The database was again demonstrated to the junior doctors. On this occasion there were deemed to be no new issues. It was subsequently run in parallel with the old (MS Word) system for a two day period allowing for a safety net should any major errors arise. From this point onwards, no further formal updates to the program were required but instead as minor issues were found they were corrected on the live system. PDSA cycle 4 - The questionnaire used in cycle 1 was again distributed. This revealed a general improvement across all aspects along with patient safety. See supplementary file: ds4735.docx - "PDSA Cycles 1-4" Post-measurement ================ Whilst information for the QI project is somewhat opinion based it has been formalised into a questionnaire in order to try and quantify/qualify these opinions. Indeed there is no obvious measure that can demonstrate an improvement in handover other than the lack of adverse incidents and near misses. When patients are found on the ward having been dropped from a list the doctors instinct is not to tell the author about the case but to ensure the patients safety and catch up on anything outstanding in the patient\'s care. Following the implementation of the database the questionnaire was re-distributed with the results summarised as follows: The handover list was now always accessible and was no longed dependant on another doctor not having the document open simultaneously. The automated layout was sensible and continuous across all four teams. The incidence of patients and/or patient data being dropped from these lists had fallen dramatically with the discharge of a patient from the database now requiring a two stage process. Changing of team produced no loss in data as this was done by changing the allocated consultant within the database. The "Print all patients" button allowed for all elective inpatients to be listed by ward and then surname order allowing for a logical and time efficient ward round. See supplementary file: ds4741.pdf - "Database screenshots" Lessons and limitations ======================= When the system was taken "online" as the sole patient handover management system there were a number of issues that had to be resolved and this was done so whilst the author was in work. A problem arose a couple of weeks further down the line whilst the author was not in work. The result of an error within the spelling of a consultant\'s name had led to a patient being left off a list. This was one of the specific aims this project was seeking to resolve, thankfully the patient\'s team were aware of this and flagged it up to the author resulting in a further review of all spellings to ensure this didn\'t happen again. This kind of project needs to be future proofed, allowing someone who is less IT literate to manage its upkeep once the author has moved on. Conclusion ========== Reviewing the issues that led to the creation of this database: \- There was a lack of continuity as to the layout of handover sheets between the four teams \- Patients who switched team due to changing consultant had been taken off a list and not added to another resulting in them not being reviewed/managed correctly \- When carrying out the weekend ward round of all elective patients there was no easy way to print the patients by ward and this involved spending time studying lists to ensure no-one was missed It was felt that the database had improved continuity for the junior doctors, with one departmental solution instead of a single list for each team. Patients were far less likely to be dropped from lists. Weekend elective ward rounds were more manageable with easier navigation of the lists and subsequent improvement in economy of time. Supplementary Material ====================== ###### Web supplement ###### Web supplement ###### Web supplement **Declaration of interests:** nothing to declare

Introduction ============ Carbon nanotubes (CNTs) have attracted particular interest because of their remarkable mechanical and electrical properties \[[@B1]\]. The combination of these properties with very low densities suggests that CNTs are ideal candidates for high-performance polymer composites \[[@B2]\]. In order to increase the application range of polymers, highly conductive nanoscale fillers can be incorporated into the polymeric matrix. As CNTs present high electrical conductivity (10^3^-10^4^S/cm), they have been widely used \[[@B3]\]. Therefore, CNT/polymer composites are expected to have several important applications, namely, in the field of sensors and actuators \[[@B4]\]. However, in order to properly tailor the composite material properties for specific applications, the relevant conduction mechanisms must be better understood. The experimental percolation thresholds for CNT composites results in a wide range of values for the same type of CNT/polymer composites \[[@B5]\], being a deviation from the bounds predicted by the excluded volume theory and a dispersion for the values of the critical exponent (*t*) \[[@B6],[@B7]\]. It was demonstrated that the conductivity of CNT/polymer composites can be described by a single junction expression \[[@B8]\] and that the electrical properties also strongly depend on the characteristics of the polymer matrix \[[@B9]\]. This article explores the effects of nanotubes surface modifications in the electrical response of the composites. Experimental ============ Preparation and characterization of the modified CNT samples ------------------------------------------------------------ Commercial multi-walled CNTs (Nanocyl - 3100) have been used as received (sample CNTs). Further details on this material can be found elsewhere \[[@B10]\]. CNTs sample was functionalized by oxidation under reflux with HNO~3~(7 M) for 3 h at 130°C, followed by washing with distilled water until neutral pH, and drying overnight at 120°C (sample CNTox was obtained). The CNTox material was heat treated under inert atmosphere (N~2~) at 400°C for 1 h (sample CNTox400) and at 900°C for 1 h (sample CNTox900), to selectively remove surface groups. The obtained samples were characterized by adsorption of N~2~at -196°C, temperature-programmed desorption (TPD) and determination of pH at the point of zero charge (pH~PZC~) from acid-base titration according to the method of the literature \[[@B11]\]. The total amounts of CO and CO~2~evolved from the samples were obtained by integration of the TPD spectra. Composites preparation ---------------------- Polymer films with thicknesses between 40 and 50 μm were produced by mixing different amounts of CNT (from 0.1 to 1.0%) with *N, N*-dimethylformamide (DMF, Merck 99.5%) and PVDF (Solef 1010, supplied by Solvay Inc., molecular weight = 352 × 10^3^g/mol) according to the procedure described previously \[[@B9]\]. Solvent evaporation, and consequent crystallization, was performed inside an oven at controlled temperature. The samples were crystallized for 60 min at 120°C to ensure the evaporation of all DMF solvents. After the crystallization process, the samples were heated until 230°C and maintained at that temperature for 15 min to melt and erase all polymer memory. This procedure produced α-PVDF crystalline phase samples \[[@B12]\]. Sample characterization ----------------------- Topography of the samples and CNT distribution was performed by scanning electron microscopy (SEM, FEI - NOVA NanoSEM 200). The dielectric response of the nanocomposites was evaluated by dielectric measurements with a Quadtech 1920. Circular gold electrodes of 5-mm diameter were evaporated by sputtering onto both sides of each sample. The complex permittivity was obtained by measuring the capacity and tan δ in the frequency range of 100 Hz to 100 kHz at room temperature. The volume resistivity of the samples was obtained by measuring the characteristic *I*-*V*curves at room temperature using a Keithley 6487 picoammeter/Voltage source. Results and discussion ====================== Characterization of CNT samples ------------------------------- Oxidations with HNO~3~originate materials with large amounts of surface acidic groups, mainly carboxylic acids and, to a smaller extent, lactones, anhydrides, and phenol groups \[[@B10],[@B13],[@B14]\]. These oxygenated groups (Figure [1](#F1){ref-type="fig"}) are formed at the edges/ends and defects of graphitic sheets \[[@B15]\]. The different surface-oxygenated groups created upon oxidizing treatments decompose by heating, releasing CO and/or CO~2~, during a TPD experiment. As this release occurs at specific temperatures, identification of the surface groups is possible \[[@B10],[@B13],[@B14]\]. It is well known that CO~2~formation results from the decomposition of carboxylic acids at low temperature, and lactones at higher temperature; carboxylic anhydrides originate both CO and CO~2~; phenols and carbonyl/quinone groups produce CO \[[@B10],[@B13],[@B14]\]. {#F1} Figure [2](#F2){ref-type="fig"} shows the TPD spectra of the CNT before and after the different treatments. It is clear that the treatment with HNO~3~produces a large amount of acidic oxygen groups, such as carboxylic acids, anhydrides, and lactones, which decompose to release CO~2~. Part of these groups (carboxylic acids) is removed by heating at 400°C. A treatment at 900°C removes all the groups, so that the obtained sample is similar to the original. The total amounts of CO and CO~2~evolved from the samples, obtained by integration of the TPD spectra, are presented in Table [1](#T1){ref-type="table"}. {#F2} ###### BET surface areas obtained by adsorption of N~2~at -196°C and amounts of CO~2~and CO obtained by integration of areas under TPD spectra Sample CNTs CNTox CNTox400 CNTox900 --------------------------- ------ ------- ---------- ---------- BET surface area (m^2^/g) 254 400 432 449 pH~PZC~ 7.3 4.2 6.9 7.4 CO~2~(μmol/g) 70 778 230 24 CO (μmol/g) 193 1638 1512 204 CO/CO~2~ 2.76 2.11 6.57 8.50 All the samples release higher amounts of CO than CO~2~groups (Table [1](#T1){ref-type="table"}). The CNTox sample has the highest amount of surface oxygen. This sample also presents the lowest ratio CO/CO~2~and the lowest value of pH~PZC~, indicating that this is the most acidic sample. CNTox900 presents the highest CO/CO~2~ratio, suggesting the less-acidic characteristics, which matches well with the pH~PZC~results (Table [1](#T1){ref-type="table"}). The acidic character of the samples decreases by increasing the thermal treatment temperature, since the acidic groups are removed at lower temperatures than neutral and basic groups, as seen in previous studies \[[@B10],[@B13],[@B14]\]. The CNT samples have N~2~adsorption isotherms of type II (not shown), as expected for non-porous materials \[[@B16]\]. The surface areas of the samples, calculated by the BET method (S~BET~), are included in Table [1](#T1){ref-type="table"}. It can be observed that the oxidation treatments lead to an increase of the specific surface area. This occurs because the process opens the endcaps of CNTs and creates sidewall openings \[[@B17]\]. The specific surface areas of the samples slightly increase as the thermal treatment temperature increases, since carboxylic acids and other groups, introduced during oxidation, are removed. Composites processing and characterization ------------------------------------------ The morphology and fiber distribution of the composite samples were analyzed by SEM to evaluate the CNT dispersion in the polymeric matrix and determine how the composites influence the polymer crystallization microstructure. Figure [3](#F3){ref-type="fig"} shows the SEM images for the PVDF/CNT composites. The main relevant microstructural feature of the composite is that the CNT are randomly distributed into the polymeric matrix. The spherulitic structure characteristic of the pure PVDF is still present in all the composites samples \[[@B12],[@B18]\]. {#F3} CNT agglomerates are nevertheless more often observed for the CNTox composites samples, especially for the ones treated at the highest temperatures. With respect to the electrical properties, oxidation reduces the composite conductivity for a given concentration and shifts the percolation threshold to higher concentrations (Figure [4](#F4){ref-type="fig"}). This behavior is mainly due to the reduction of the surface conductivity of the CNTs due to the oxidation process \[[@B8]\], and is similar for all the functionalized composites. Further, the increase of surface area due to the functionalization treatment certainly causes surface defects on the CNTs that also reduced electrical conductivity. The increase of agglomerations for the treated samples should not have, on the other hand, a large influence in the electrical response \[[@B8]\]. A change of several orders of magnitude of the electrical resistivity with increasing CNTs concentration was observed for all samples, indicating a percolative behavior of the nanocomposites. In general, both in surface (not shown) and in bulk resistivity (Figure [4a](#F4){ref-type="fig"}), the percolation threshold appears between 0.2 wt.% for the original CNT samples and shifts to 0.5 wt.% CNTs for the functionalized nanocomposites. {#F4} Dielectric measurements show that the incorporation of the CNT in the PVDF matrix but leads to a gradual increase of the dielectric constant (ε\') as the amount of the filler is increased (Figure [4b](#F4){ref-type="fig"}). The increase of the ε\' is larger for the pristine CNT. A maximum for the 0.5% pristine CNT sample with ε\' 22 at a frequency of 10 kHz at room temperature was found, whereas for the functionalized nanocomposites the value is 16. The frequency behavior of the dielectric permittivity is similar to the one obtained for the pure polymer, except for an increase of the low frequency dielectric constant and dielectric loss (not shown) with increasing CNT loading due to interfacial polarization effects (Figure [4b](#F4){ref-type="fig"}). No noticeable differences have been observed for the different oxidation treatments in terms of the dielectric response. In a previous study \[[@B19]\], it was demonstrated that an increase in the dielectric constant is related with the formation of a capacitor network. Conclusions =========== The effect of surface modifications of multi-walled CNTs on the electrical response of CNT/PVDF nanocomposites has been investigated. The main effect of oxidation is a reduction of the composite conductivity for a given concentration and a shift of the percolation threshold to higher concentrations. On the other hand, no significant differences have been observed between the nanocomposites prepared with the different functionalized CNTs. The reduction of the electrical surface conductivity of the CNT due to the oxidation process, together with an increase of the surface area and defect formation, is at the origin of the observed effects. Abbreviations ============= CNT: carbon nanotubes; DMF: *N, N*-dimethylformamide; SEM: scanning electron microscopy. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= SACC performed the functionalisation and characterisation of carbon nanotubes samples and drafted the manuscript. JNP, CP, and VS participated in the nanocomposite samples processing, experimental measurements, analysis and interpretation of the results. MFRP and SL-M conceived and coordinated the research work and carried out analysis and interpretation of the experimental results. All authors read and approved the final manuscript. Acknowledgements ================ The authors thank the Fundação para a Ciência e a Tecnologia (FCT), Portugal, for financial support through the projects PTDC/CTM/69316/2006 and NANO/NMed-SD/0156/2007), and CIENCIA 2007 program for SAC. V.S. and J.N.P. also thank FCT for the SFRH/BPD/63148/2009 and SFRH/BD/66930/2009 grants.

Sir, Leiomyoma is the most common tumor of uterus and female pelvis. Leiomyosarcoma almost always arise de novo and almost it doesn't results from sarcomatous transformation of a leiomyoma. One of the most controversial concepts on the subject of uterine smooth muscle tumors is smooth muscle tumor of uncertain malignant potential (STUMP), a term first used by Kempson in 1973.\[[@ref1]\] These are a group of heterogeneous and uncommon uterine smooth muscle tumors which fulfill some but not all the diagnostic criteria for leiomyosarcoma. This makes them unclassifiable by currently available criteria as unequivocally benign or malignant.\[[@ref1]\] In these tumors, it is simply impossible with current tools to predict the behavior with certainty and this makes their management difficult.\[[@ref2]\] What makes the management more complicated is the difficulty in counselling patients with regards to the likely clinical behavior. However, data from literature suggest a low risk of recurrence and a generally good clinical outcome.\[[@ref2][@ref3]\] Since recurrence of STUMP has been reported to be regional and resectable, surgical resection has been recommended as the primary modality for the treatment of recurrence.\[[@ref2]\] Recurrence rates have been similar for patients who underwent myomectomy and those who underwent hysterectomy.\[[@ref4]\] Moreover, leiomyosarcomatous transformation likelihood is low and there is no evidence that adjuvant treatments improve long-term outcomes. As a result, most authors have recommended expectant management of STUMP in the form of close clinical observation in all patients.\[[@ref2]\] Herein, we briefly report a case of 29 years nulligravid woman presented in 2011 with a two years history of menometrorrhagia and pelvic pain. Ultra sonography revealed a pedunculated subserosal myomatous mass measuring 65 × 50 × 50 mm as well as three small intramural myomatous masses with the greatest diameter of 17 mm. The patient underwent myomectomy. On pathologic examination, one of the small intramural masses was found to be STUMP. Since the gynaecologists frequently decide not to remove small myomatous masses during myomectomy procedure, the question is which myomatous masses should be considered for surgical removal. She recovered completely without complication. The problem becomes more challenging when considering the fact that the preoperative diagnosis of STUMP is usually leiomyoma\[[@ref2]\] This question is open to more discussions and suggestions by experts in this field.

Introduction {#Sec1} ============ Ultrasonic nebulizers or atomizers are very important due to their many applications such as drug delivery, mass spectrometry, humidity control, spray pyrolysis, coating, etc. In almost all cases, the distribution of droplet sizes is an important parameter. For example, in many drug delivery systems, droplets need to be sufficiently small to reach the lower parts of the pulmonary tract, as larger droplets are deposited predominantly at the start of the airways^[@CR1]^. Despite extensive research^[@CR2]--[@CR12]^, many aspects of the nebulization process are still poorly understood, mainly because of the very complicated dynamics and the small length and time scales on which it takes place. This work is the first study that focusses primarily on the droplet size distribution for ultrasonic nebulization. We show that, where direct measurements are mostly inaccessible, the shape of the size distribution can be used as an indirect measure for the breakup mechanism. Depending on the type of nebulizer, the mechanism of nebulization can be very different. This is reflected in the droplet sizes, that unlike what is commonly observed, can be both narrowly or broadly distributed. The process of ultrasonic nebulization is mostly explained by the capillary wave mechanism. The formation of capillary waves on the surface of a fluid supported by a vibrating solid was first described by Faraday in 1831, and are therefore also referred to as Faraday waves^[@CR13]^. Later, Kelvin derived the formula for the capillary wavelength *λ*^[@CR14]^, relating it to the surface tension *σ*, density of the fluid $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\rho $$\end{document}$, fluid depth *h* and frequency *f* of the standing waves as follows:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda {[\tanh (\frac{2\pi h}{\lambda })]}^{-1/3}={(\frac{2\pi \sigma }{\rho {f}^{2}})}^{1/3},$$\end{document}$$where for a sufficiently deep liquid layer, i.e. *h*/$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda \gg 1$$\end{document}$, one can take $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tanh (2\pi h/\lambda )=1$$\end{document}$. As with many parametric oscillators, the frequency of the nonlinear standing waves is half the excitation frequency *F*, i.e. $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f=F$$\end{document}$/2. For ultrasonic nebulizers, according to the capillary wave mechanism, the amplitude of the oscillation is large enough to cause droplet pinch-offs, thereby nebulizing the fluid (Fig. [1a](#Fig1){ref-type="fig"}). Since the size of the pinch-off droplets is proportional to the capillary wavelength, the median droplet size (designated as *D*~50~) is given by$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${D}_{50}=\kappa \lambda =\kappa \,{(\frac{8\pi \sigma }{\rho {F}^{2}})}^{1/3},$$\end{document}$$where *F* is the frequency of the nebulizer and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\kappa $$\end{document}$ a proportionality constant. Lang was the first to experimentally determine this constant and found $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\kappa =0.34$$\end{document}$, where *D*~50~ is the median droplet size for the *number* (not *volume*) distribution^[@CR2]^. For all that follows, we will use *D*~50~ only to indicate the median particle size by *volume*. Although Lang's prediction works well in many cases, there are often slight and sometimes even significant differences in the proportionality constant $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\kappa $$\end{document}$^[@CR12],[@CR15]^. This and the fact that many nebulizers are distinctly different has led to the formulation of many different scaling laws and different nebulization mechanisms, one example being cavitation^[@CR4],[@CR5],[@CR10]^.Figure 1(**a**) Ideal case of standing capillary waves of wavelength *λ*. If the amplitude exceeds a certain value, equal sized ligaments are produced, leading to the break-off of monodisperse droplets of size *d* \~ *λ*. (**b**) Capillary waves in a system with maximum interference, leading to a distribution of wavelengths and amplitudes. This results in a broader distribution of droplet sizes set by the average ligament roughness and the distribution of ligament sizes. (**c**) Faraday waves are superposed on larger waves of the order of the wavelength of the chip material. The motion of these larger waves initiates breakup of the smaller superposed capillary waves. High-speed camera images show that jets of small droplets are formed at the crests of these larger waves at maximum acceleration. In studies of ultrasonic nebulizers, the focus is mostly on the median droplet size. The spread around this median is however just as important, especially for their implementation in different applications. For sprays and jets, the shape of the droplet size distribution is well understood^[@CR16]--[@CR21]^. There, it is set by the ligament corrugation and the distribution of ligament sizes. Assuming that capillary waves are the droplet formation mechanism, the droplet sizes are determined by the initial size of the waves and the roughness of the pinch-off ligaments (Fig. [1](#Fig1){ref-type="fig"}), and would therefore be comparable with the breakup of sprays. In this case, waves can be more or less spread in size, giving more or less dispersion in droplet sizes (Fig. [1b](#Fig1){ref-type="fig"}). Similarly, ligaments can be very corrugated, leading to a broad distribution, or very smooth giving a narrow size distribution. In this work we investigate three types of ultrasonic nebulizers, working at different frequencies. We find that for these devices the capillary wave hypothesis works well, with proportionality constants depending on the type of nebulizer. We further show that droplet size distributions also depend on the type of nebulizer. Distributions can be surprisingly broad, presumed to be due to large variability in wavelengths and rough pinch-off ligaments, but also very narrow as predicted by the classical picture of capillary wave breakup, where waves are similar sized with smooth pinch-offs. Nebulizers {#Sec2} ========== SAWN {#Sec3} ---- The surface acoustic wave nebulizer (SAWN) consists of two interdigital transducers (IDT) with lithium niobate as the piezoelectric substrate. Between the two IDTs there is a space (delay zone) where a droplet can be placed for nebulization (Fig. [2b](#Fig2){ref-type="fig"} with holder and Fig. [3a](#Fig3){ref-type="fig"}). For this particular chip, the resonance frequency is 9.6 MHz. The distance between the metallic electrodes of the IDT sets the wavelength of the acoustic waves at $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda =0.36\,{\rm{mm}}$$\end{document}$. The surface acoustic waves can either be created by one of the IDTs (single mode) or both (dual mode). In this work, one IDT is used to create the waves and the other IDT as a transceiver. Experiments do not show any difference between the single and dual mode other than quicker nebulization for the latter.Figure 2Nebulizer types investigated. (**a**) SAWN chip with holder. (**b**) Nebulizer chip. (**c**) Mist maker. The gold-colored part is the vibrating membrane. The transducer is placed roughly 4 cm under water.Figure 3(**a**) Picture of the nebulization of a droplet on the SAWN chip taken with a 5 *μ*s flash light. Two droplet size regimes can be observed as indicated by the red arrows. (**b**) The nebulization process for the nebulizer chip. (**c**) The mist maker producing large amounts of small droplets. As can be seen, the mist maker is the only submerged nebulizer, where for the other nebulizers small amounts of fluids are placed on top of the transducer. The SAWN has many applications of which one of the most prominent ones is in mass spectrometry, where it has several advantages over more conventional nebulization methods^[@CR22],[@CR23]^. The SAWN creates plumes of droplets with ionized molecules, which after the evaporation of the solvent, can be directly used by the mass spectrometer. The precise mechanism of the ionization is unknown, but is probably due to the high voltages present on the chip. It can also be a result of cavitation, which is known to produce sufficiently high temperatures and pressure^[@CR24]^, but is considered unlikely to occur for the low power SAWN^[@CR25]^. The droplet size distribution plays an important role for applications, as for example in mass spectrometry large droplets can lead to loss of sensitivity and if not properly desolvated may lead to vacuum fluctuations, while small droplets will not. We operate the SAWN chip with an arbitrary waveform function generator (RIGOL DG1062) amplified by a RF power amplifier (Mini-Circuits TVA-R5-13). As soon as the SAWN is operative, the droplet becomes opaque due to the waves on the surface of the droplet, but nebulization only starts after a certain waiting time. This waiting time decreases when the chip has been in operation longer, which is most likely due to heating of the chip and chip holder. This conjectured temperature-dependent onset of nebulization can be caused by lowering of the viscosity, which has a dampening effect; this is supported by the observation that a glycerol-water solution does not nebulize at the same applied power as pure water. After prolonged operation, heat (IR) camera images show that the chip and droplet can reach temperatures of around 70 °C. When the droplet starts to nebulize, a plume of micron-sized droplets is created together with much larger particles of \~50 *μ*m, visible in Fig. [3a](#Fig3){ref-type="fig"}. These larger particles are most likely due to the direct interaction of surface acoustic waves of the chip with the droplet, which have a much larger wavelength than the parametrically excited capillary waves. High-speed microscopic picture sequences of a nebulizing water droplet (Fig. [4a](#Fig4){ref-type="fig"}), suggest that the breakup of the capillary waves, which are believed to be superposed on the larger surface waves, is triggered by the rapid acceleration that these waves experience. The breakup of these droplets is therefore observed as jets (Fig. [4a](#Fig4){ref-type="fig"}) on top of the crests of these waves as illustrated by Fig. [1c](#Fig1){ref-type="fig"}. To the best of our knowledge, this is the first report of such a breakup mechanism for small droplets of the SAWN, or in fact for any type of nebulizer.Figure 4High-speed microscopic images from a nebulizing water droplet taken from the side of the SAWN chip. (**a**) Parametrically driven capillary waves are superposed on much larger waves set by the wavelength of the chip. The acceleration of these larger waves induces the sudden release of small droplets on their crests, visible here as jets, indicated by arrows. The three images are sequential, with time steps of 90.91 *μs* (11000 fps), showing that the jet containing small droplets is shot upwards. (**b**) The breakup of droplets is governed by the statics of the interference of waves, which sometimes can lead to the production of large amplitudes (ligaments) that subsequently break up to form droplets. Panel (b) shows two examples (I and II) of such waves in the case of the larger surface waves, i.e. not the parametrically excited waves which are too small to observe. Note that the two peaks shown in (**b**) are two extreme cases; most big droplets are created by waves of smaller amplitude. The breakup of droplets by Faraday waves is governed by the statistics of the interference of waves. An interesting large scale example of interference leading to very large amplitudes, are ocean freak waves or rogue waves. These waves that are much heigher than the average wave, pose a serious threat even to modern ships. Although the situation here is very different, extreme amplitudes do appear for which two examples are shown in Fig. [4b(I and II)](#Fig4){ref-type="fig"}. Note that these are the larger surface waves, not the parametrically excited waves, which are too small to be observed. Nebulizer chip {#Sec4} -------------- The nebulizer chip ('Grove' type water atomizer by Seeed Technology) is a low-frequency (105 ± 5 kHz), low-power (2 W) and low-cost type of atomizer (Fig. [2](#Fig2){ref-type="fig"}). The transducer has a coin shape with and a rubber ring for isolation and a small reservoir to contain the fluid to be nebulized. If the transducer is placed in a small bath of water, water will be absorbed from under the chip into the reservoir such that the nebulizer can work continuously. In the middle of the chip there is a small bump where the nebulization takes place (Fig. [2b](#Fig2){ref-type="fig"}). Mist maker {#Sec5} ---------- The mist maker is a widely commercial available type of nebulizer mostly used for decoration and terraria (Fig. [2a](#Fig2){ref-type="fig"}). Unlike the other nebulizers investigated, this type is submerged. We immersed a Nebler nebulizer in water, placing it about 4 cm under the water surface where it produces a small cone-shaped fountain accompanied by a large amount of micron-sized droplets, seen as mist (Fig. [3c](#Fig3){ref-type="fig"}). The depth of the transducer follows from the fact that when the nebulizer is placed near the surface, it produces a lot of splashing, while if it is placed to deep, attenuation of the waves will reduce its effectiveness. The frequency of the nebulizer is 1700 ± 50 kHz with a power of 30 W. Experiments {#Sec6} =========== All the droplet size distributions were measured by a laser diffraction method (Malvern Spraytec). Since the diffraction angle is inversely proportional to the droplet size, analysis of the scattered laser light intensity allows for the determination of the droplet size distribution, assuming a spherical droplet shape, an assumption easily met for such small droplets. With dense sprays, multiple scattering events take place for which the software accounts for. There is however a limit in the density of the fog at which the droplet size distribution can still be measured accurately. Therefore, the location of the laser beam through the mist is chosen such that the mist is dilute enough (i.e. high transmission), but not too far from the nebulization location to prevent possible changes to the size distribution due to coalescence events or evaporation. It is know for example that when droplet sizes are micron sized, even at a relative humidity of 100%, distributions can change quickly over time, causing a significant change in the median droplet size^[@CR26]^. Therefore the SAWN and nebulizer chip were placed in a glass container with a relative humidity of a 100%, with the walls of the container placed under an angle with the laser to prevent reflection of the beam on the detector. Background calibration and laser alignment are done between each measurement, thereby taking into account the walls of the glass container. The laser beam was placed 10 *cm* above the nebulizer, which appeared to be the optimum place of measurement, since it is high enough to have good transmission, yet not so far that the time between droplet creation and measurement is getting too long. For the mist maker measuring droplet sizes is slightly more complicated. The mist maker produces a large volume of droplets that do not ascent in the same manner as the other nebulizers. Directing the droplets through the beam with an airflow is likely to affect the droplet sizes by coagulation and evaporation. To introduce the least disturbance we placed the nebulizer in a beaker filled for roughly 3/4 with water, that after some operation time quickly reaches a relative humidity of a 100% above the water surface. The mist could then be guided directly through the laser beam by tilting the beaker, without the mist traveling a significant distance for any evaporation to take place. Even though the result does not show any abnormalities, this method is not optimal. No method however, can guarantee the absence of any coalescence or evaporation. For the SAWN, we explored several waveforms such as pulsed, sinusoidal, square and AM-signals. There seems to be no affect of the waveform on the nebulization process. The only essential component seems to be the total power; certain waveforms give more power, resulting in better and quicker nebulization. Droplet Size Distributions {#Sec7} ========================== What determines the droplet size distribution in different types of fluid breakup, such as spray and jets, is well understood. In fluid fragmentation, where ligaments are of equal size, the rescaled droplet size distribution is given by a Gamma function $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm{\Gamma }}(x=d/\langle d\rangle ,n)$$\end{document}$, where $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\langle d\rangle $$\end{document}$ is the mean droplet size and *n* a parameter that sets the ligament corrugation. For systems with highly corrugated ligaments, the distribution is broad and *n* = 4--5, while for very smooth ligaments $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n=\infty $$\end{document}$, which would result in a delta peak. For more complicated fluid breakup, such as in agricultural sprays, the ligaments themselves can also be distributed in size^[@CR21]^, leading to a two-parameter compound Gamma distribution:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathscr{P}}}_{m,n}(x=\frac{d}{\langle d\rangle })=\frac{2{(mn)}^{\tfrac{(m+n)}{2}}{x}^{\tfrac{(m+n)}{2}-1}}{{\rm{\Gamma }}(m){\rm{\Gamma }}(n)}{{\mathscr{K}}}_{m-n}(2\sqrt{mnx}),$$\end{document}$$where *m* sets the width of the ligament size distribution, *n* the ligament corrugation and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathscr{K}}}_{m-n}$$\end{document}$ is a Bessel function of order *m* − *n*^[@CR27]^. Here we argue that the ultrasonic atomization is similar to other fluid fragmentation processes, since here we can have a spread in ligament sizes due to a spread in wavelengths (Fig. [1b](#Fig1){ref-type="fig"}), and the pinch-off ligaments can be either smooth or rough. Figure [5](#Fig5){ref-type="fig"} shows the rescaled size distributions for the different nebulizers for three consecutive measurements. The laser diffraction method does not allow for error bars on the data points, the setup averages hundreds of measurements per run, and the standard deviation between such measurements is typically small. The agreement between the three different runs shows however there is little variability between measurements. The droplet sizes are fitted with Eq. ([3](#Equ3){ref-type=""}) to obtain the parameters *m* and *n*. Due to the form of Eq. ([3](#Equ3){ref-type=""}), exponents can become large, leading to numerical problems in evaluating the distribution function. Standard minimization algorithms therefore do not work properly, since they are sensitive to the values of the initial guess. Fit parameters were therefore obtained by manually fitting the distribution function. This however does not affect the results, since minor changes of the parameters does not alter the conclusions in any way.Figure 5Rescaled droplet size (*d*) distributions for the three different nebulizers, each of which measured three consecutive times. Solid lines are fits to a two-parameter compound Gamma distribution $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathscr{P}}}_{m,n}$$\end{document}$, Eq. ([3](#Equ3){ref-type=""}). (**a**) Size distribution of the SAWN for the smallest droplets, with fit parameters $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m=3$$\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}$$n=4$$\end{document}$. The fit parameters indicate a very large spread in pinch-off ligament sizes and very corrugated ligaments, which can be expected for the nebulization of a droplet by the violent mechanism as seen in Fig. [4a](#Fig4){ref-type="fig"} and illustrated in Fig. [1c](#Fig1){ref-type="fig"}. Median droplet size for the smallest droplets of the SAWN is 1.1 *μ*m. (**b**) Size distribution for the mist maker with fit parameters $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m=40$$\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}$$n=8$$\end{document}$ and a median droplet size of 5.6 *μ*m. These parameters indicate that the wavelengths are very similar and the pinch-off ligaments reasonably smooth, in accordance with the capillary wave mechanism of a free surface such as present in case of the the mist maker. (**c**) Size distribution of the nebulizer chip, with $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m=20$$\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}$$n=8$$\end{document}$ and a median droplet size of 9.5 *μ*m. This shows that the nebulizer chip also has similar-sized pinch-off ligaments, although less so than the mist maker. All distributions appear to be quite narrow, except for the smallest droplets of the SAWN. The values *m* and *n* for the SAWN distribution indicate large variations in wavelengths and rough ligaments. This is an expected result, considering that the smallest droplets of the SAWN are released by the rapid acceleration of the larger surface waves, thereby releasing small droplets in the form of jets as shown in Fig. [4a](#Fig4){ref-type="fig"} and schematically illustrated by Fig. [1c](#Fig1){ref-type="fig"}. Since this is a sudden and powerful ejection of droplets, there will be large variations in the ligament sizes with irregular pinch-offs. The mist maker produces capillary waves on a free surface; its operation therefore bears the most resemblance to the classical Faraday wave scenario, where the breakup is smooth with similar-sized wavelengths (Fig. [1a](#Fig1){ref-type="fig"}). The distribution obtained experimentally is indeed the most narrow of all three nebulizers, with $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m=40$$\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}$$n=8$$\end{document}$. Still, some randomness is expected; the mist maker creates a fountain of water at the center due to the driving force of the speaker, undoubtedly causing irregularities. For the nebulizer chip, a layer of fluid is placed on the center of the transducer, where the nebulization takes place. The nebulizer only operates when the liquid layer is sufficiently thin, which can be associated with the shallow wave regime. We find that the size distribution is quite narrow with $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m=20$$\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}$$n=8$$\end{document}$, which is somewhat surprising, considering that its operation appears to be similar to that of the SAWN. Still, what is distinctively different is that for the SAWN the smallest droplets are created by interaction with waves that are of a much larger length scale. Since the formation of the small and big droplets of the SAWN are interconnected, the lack of such bigger droplets implies that the nebulization mechanism for the nebulizer chip must be different. The SAWN creates not only micron-sized droplets but also much larger (\~50 *μ*m) droplets that are due to the direct interaction of the surface acoustic waves with the droplet on the chip. During nebulization, the edge of the droplet flattens out, creating a thin fluid layer. There, as a result of constructive interference, occasional peaks arise that frequently break up to form large droplets and simultaneously create the previously described jets of small droplets (Fig. [4](#Fig4){ref-type="fig"}). In contrast to what is seen for the smallest droplets, the big droplets are narrowly distributed (Fig. [6](#Fig6){ref-type="fig"}). With $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m=150$$\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}$$n=15$$\end{document}$, the waves responsible for droplet formation seem to be of equal size with reasonably smooth pinch-off ligaments. The sizes of the peaks are determined by the precise statics of the interference of such waves, which apparently are quite regular. Although Fig. [4b](#Fig4){ref-type="fig"}) shows two examples of peaks, it should be noted that these are two extreme cases of which no conclusion can be drawn about the general shape of the pinch-off. Most spikes that lead to the formation of big droplets are much smaller and barely visible due to masking by other waves and smaller droplets.Figure 6Droplet size distribution of the SAWN including the larger droplets. These larger droplets are due to waves created by the direct interaction of the surface acoustic waves and the droplet. Three measurements are shown along with the fit (solid line) according to Eq. ([3](#Equ3){ref-type=""}). Droplet Sizes {#Sec8} ============= In Table [1](#Tab1){ref-type="table"} the median droplet sizes for the different types of nebulizers are listed, together with the ratio *D*~50~/*λ*, where *λ* is the capillary wavelength. This ratio is the proportionality constant $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\kappa $$\end{document}$ in Eq. ([2](#Equ2){ref-type=""}), which Lang^[@CR2]^ found to be 0.34 in the case of a *number* median. Although the investigated devices are distinctively different, with frequencies almost 2 orders apart, we find proportionality constants in the range 0.17--0.65, close to Lang's value. The proportionality constants of the nebulizer chip and the big droplets of the SAWN are both at the lower end of this range. This could be because the assumption of a sufficient height of the fluid layer is not met for these cases: for shallow water, one expects smaller waves, resulting in a smaller proportionality constant.Table 1Frequency, median diameter and ratio of median diameter over capillary wavelength, calculated with Kelvin's formula.DeviceFrequency*D*~50~ (*μm*)*D*~50~/*λ*SAWN small droplets9.6 MHz1.10.39Nebulizer chip105 ± 5 kHz9.50.17Mist maker1700 ± 50 kHz5.60.65**DeviceFrequency*D*** ~**50**~ **(** ***μm*** **)*D*** ~**50**~ **/** $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ell $$\end{document}$SAWN big droplets---560.16For the big droplets produced by the SAWN, the wavelength $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ell $$\end{document}$ of the surface acoustic waves of the chip is used, which is set by the distance of the metallic strips of the IDT. Conclusions {#Sec9} =========== This is the first study that focuses primarily on the droplet size distribution of ultrasonic nebulizers. We provide a new view on the droplet sizes, by employing a distribution function that can be directly linked to the breakup mechanisms. This is in contrast with other studies, where distribution functions are fitted without any clear physical interpretation. Our proposed interpretation of the fit parameters of the distributions could give insight in the waves that contribute to droplet formation, something that is otherwise difficult to assess by direct measurements, due to the small length- and time-scales involved. Additional measurements on systems with larger Faraday waves, could further develop this notion and allow for a direct verification. We investigated three types of ultrasonic nebulizers and show that the droplet size distribution is analogous to that observed for other types of fluid breakup such as sprays. We argue that the shape of the distribution is set by the dispersion in wavelengths and the roughness of the ligaments. In the classical case of ultrasonic nebulization through the capillary wave mechanism, wavelengths are equal in size and the pinch-offs are smooth, leading to a narrow size distribution. Our results therefore support this classical picture of nebulization for some types of nebulizers. However, our results also reveal that droplets can be very different in size, something that has not has not been observed before. For a nebulizer based on surface acoustic waves, there are waves of two wavelengths involved: waves on the order of the surface acoustic waves and the much smaller parametrically excited capillary waves. We theorize that as a result of the fast acceleration of these larger waves, the smaller waves break apart, thereby ejecting droplets from the surface. This phenomenon is observed as jets of small droplets on top of the crests of these waves. Since according to this view droplets are shed from the surface in an abrupt way, large variations in ligament size can be expected, leading to a big dispersion in droplet sizes. For different devices that operate at different frequencies, the capillary wavelength scales well with the the median droplet size, with proportionality constants of order unity. In some cases the liquid layer is shallow, which could explain the smaller droplets we observe for these nebulizers. In this work we did not investigate the effect of the droplet size on fluid properties such as viscosity, surface tension and density, which would be necessary to systematically evaluate the relation between droplet sizes and capillary wavelength. Even though most existing scaling laws predict the fluid parameters to have only a minor effect on the median drop size, their effect on the droplet size distribution would also be of interest. **Publisher's note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This work is part of the Industrial Partnership Program Hybrid Soft Materials that is carried out under an agreement between Unilever Research and Development B.V. and the Netherlands Organization for Scientific Research (NWO). S.K., D.B., A.A. and G.C. designed the study. S.K. and D.B. wrote the manuscript. S.K. carried out the experiments. Competing Interests {#FPar1} =================== The authors declare no competing interests.

Introduction ============ Alpha (α)-asarone \[1,2,4-trimethoxy-5-\[(E)-pro-1-enyl\] benzene; Pubchem CID: 636822; **Figure [1](#F1){ref-type="fig"}**), is one of the main pharmacologically active compounds present in *Acorus calamus* Linn (Acoraceae), *Acorus tatarinowii* Schott (Acoraceae), and *Acorus gramineus* Solander (Acoraceae; [@B30]). The various neuropharmacological activities of α-asarone in numerous preclinical studies were reported in the literature including anticonvulsant ([@B11]), neuroprotective ([@B15]), anxiolytic ([@B16]), and nootropic effects ([@B15]; [@B13]). Recently, the antidepressant-like effect of an essential oil from *Acorus tatarinowii* Schott have been reported in well-validated animal models of depression such as TST and forced swim test (FST). In the same study, the acute treatment of α-asarone at lower doses (10 and 20 mg/kg, i.p.) showed an antidepressant-like activity in both established mouse models ([@B9]). Besides that, other reports claimed that α-asarone possess CNS depressant-like effect whereby mice treated with α-asarone at higher doses (≥50 mg/kg, i.p.) affected locomotor activity and potentiated the pentobarbitone-induced sleeping time, a test that is used for the screening of potential CNS depressants ([@B21]; [@B26]; [@B16]). Similarly, CNS depressant-like activity of roots, rhizome and leaf extracts of *Acorus calamus* have been reported ([@B23]; [@B27]). In our previous study, *Acorus calamus* leaf extracts significantly increased the immobility time in FST, and diazepam-induced sleeping time and significantly reduced the spontaneous locomotor activity without affecting motor coordination ([@B27]). Conversely, antidepressant-like activities of methanolic extract of rhizomes and leaves of *Acorus calamus* in FST and TST have been reported in other studies ([@B28]; [@B29]). These seemingly contradictory reports led us to try and find answers by conducting further studies to determine dose-dependent effect of α-asarone, the active phytoconstituent of *Acorus* species, at doses that are beyond those that have been reported, which is more than 20 mg/kg in the TST. Additionally, the underlying mechanism(s) involved in the antidepressant-like activity of α-asarone was examined using its interaction with noradrenergic neuromodulators such as AMPT, prazosin, and yohimbine and serotonergic neuromodulators PCPA and WAY100635 in the TST. ![**Chemical structure of α-asarone (1, 2, 4-trimethoxy-5-\[(E)-pro-1-enyl\] benzene)**.](fphar-07-00072-g001){#F1} Materials and Methods {#s1} ===================== Animals ------- Adult male, ICR mice (Institute for Cancer Research) of age 8--10 weeks, bred and supplied by the Animal Experimental Unit (AEU, Faculty of Medicine, University of Malaya, Kuala Lumpur) were used in all our experiments. The mice were housed (four mice per cage) in an individually ventilated cage at the Satellite Animal Facility (SAF), Department of Pharmacology, Faculty of Medicine, University of Malaya, Kuala Lumpur, and acclimatized for a week in a controlled environment \[22 ± 2°C, 50--70% humidity and 12 h light/dark (lights on at 7.00 am)\] with food and water available *ad libitum*. AEU and SAF have been accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC). All experimental protocols adhered to the guidelines of the National Research Council of the National Academies of the USA ("Guide for the Care and Use of Laboratory Animals," Eighth Edition; [@B7]) and were assessed and approved by the Faculty of Medicine-Institutional Animal Care and Use Committee (FOM-IACUC), University of Malaya (Ethics Approval no: 2014-10-14/PHAR/R/VP). The behavioral experiments were performed during the light cycle between 10.00 am and 6.00 pm. All efforts were made to minimize suffering in the mice, and to reduce the number of mice used in the experiments. Drugs and Treatment ------------------- The following drugs were used: α-asarone (Lot \# S18779V; Purity 98% w/w), Tween 80 (polyethylene sorbitan monooleate; Lot \# MKBP0682V; Purity ≥ 99% v/v), α-methyl-*p*-tyrosine (AMPT; Lot \# STBD4408V; Purity 98% w/w), and 4-chloro-[D-L]{.smallcaps}-phenylalanine methyl ester hydrochloride (PCPA or Fenclonine; Lot \# SHBD9164V; Purity 97% w/w) ± 8-hydroxy-2-dipropylamino tetralin hydrobromide (8-OH-DPAT; Lot \# 053M4102V; Purity ≥ 98% w/w), prazosin hydrochloride (Lot \# 129K1137V; Purity ≥ 99% w/w) and yohimbine hydrochloride (Lot \# 13CBM8231V; Purity ≥ 98% w/w; purchased from Sigma-Aldrich, St. Louis, MO, USA); bupropion hydrochloride (Lot \#2596608; Purity ≥ 99.5% w/w) and fluoxetine hydrochloride (Lot \#2597489; Purity 99.8% w/w; obtained from LKT laboratories, Inc., St. Paul, MN, USA); *N*-\[2-\[4-(2-methoxyphenyl) piperazin-1-yl\] ethyl\]- *N*-pyridin-2-ylcyclohexanecarcoxamide hydrochloride (WAY100635; Lot \# 6-GJF-12-1; Purity 98% w/w; purchased from Toronto Research Chemicals, Inc., Toronto, ON, Canada). The α-asarone was suspended in 5% v/v Tween 80 prepared in normal saline. Bupropion, fluoxetine, PCPA, WAY100635, 8-OH-DPAT, prazosin and yohimbine were dissolved in normal saline and AMPT was suspended in 10% v/v Tween 80 prepared in normal saline. All the drugs were administered i.p., whereas WAY100635 was administered s.c. on a constant dose volume of 10 mL/kg body weight of mice. The mice in the control group received the appropriate vehicle used in this study. The administration schedule and dose of drugs used in this study was chosen as reported in the published literature and standardized in our laboratory. Behavioral Procedures --------------------- ### Tail Suspension Test The TST was performed for 6 min as described previously ([@B34]). Briefly, both acoustically and visually isolated mice were suspended 25-cm above the floor by adhesive tape placed approximately 1 cm from the tip of the tail. Mice were considered immobile when they hung passively and completely motionless. The motionless hanging posture represents depression-like behavior of the animals. The experiment was recorded and monitored using a Logitech webcam (C270) connected to a personal computer and the immobility time in second was measured using a digital stop-watch during the 6 min test by an experienced observer (blinded to the experiment). Antidepressants decrease the immobility time in the TST ([@B34]). ### Horizontal Wire Test The experiment was performed as described previously ([@B12]) with slight modifications. The mouse was lifted by the tail and the forepaws were allowed to grasp the center of the horizontal metallic wire (2 mm diameter, 70 cm long) suspended in the air about 40 cm from the surface of the table and then the tail was released to let the mouse to hang with its forelimb. The ability of the mouse to actively grasp the wire within the first 10 s (grasping reflex) and hang on, or climb up within 20 s test was measured. The mouse which tends to grasp, hang or climb-up, was considered as a normal motor coordination. On the other hand, mouse which failed to grasp or fall off from the wire within 20 s was considered as mouse with impaired motor coordination. The data are expressed as % of mice with normal motor coordination. Spontaneous Locomotor Activity ------------------------------ The spontaneous locomotor activity was assessed using actimeter (Model: ACT-01, Orchid's Scientific, Nasik, India) fabricated with clear square Plexiglas arena (50 cm × 50 cm), equipped with 32-infrared sensors. The mouse was placed in the center of the arena and the locomotor activity was measured for the duration of 10 min. The data are expressed as the total light beam interruptions (locomotor counts). The floor of the apparatus was cleaned with 20% v/v ethanol between tests. Experimental Design ------------------- ### Effect of α-Asarone *per se* in the TST The mice were divided into eight groups (*N* = 10). Thirty minutes after acute treatment with the vehicle (5% v/v Tween 80) or with bupropion (20 mg/kg, i.p.) as a positive control or the test compound α-asarone (10, 15, 20, 30, 50, and 100 mg/kg, i.p.), the immobility time in seconds was measured in the TST. ### Effect of α-Asarone in the Horizontal Wire Test The mice were assessed for motor coordination in the horizontal wire test. Briefly, seven groups (*N* = 9--10) were treated with an acute dose of vehicle (5% v/v Tween 80) or with α-asarone (10, 15, 20, 30, 50, and 100 mg/kg, i.p.). Thirty minutes after acute treatment with vehicle or α-asarone, the horizontal wire test was performed. ### Effect of α-Asarone in the Spontaneous Locomotor Activity The mice were divided into seven groups (*N* = 9--10). Thirty minutes after acute treatment with vehicle (5% v/v Tween 80) or α-asarone (10, 15, 20, 30, 50, and 100 mg/kg, i.p.), the spontaneous locomotor activity was assessed for 10 min in actimeter. Investigation of the Involvement of Noradrenergic System in the Antidepressant-Like Effect of α-Asarone ------------------------------------------------------------------------------------------------------- To investigate the involvement of the noradrenergic system in the antidepressant-like activity of α-asarone, the mice were pretreated with saline, or AMPT (100 mg/kg, i.p., a catecholamine synthesis inhibitor). Four hours after AMPT administration, mice were treated with either vehicle (5% v/v Tween 80) or bupropion (20 mg/kg, i.p.) or α-asarone (20 mg/kg, i.p.). Thirty minutes after vehicle or drug treatment, TST was performed ([@B18]; [@B14]). In another study, the mice were pretreated with saline or prazosin (1 mg/kg, i.p., an α~1~-adrenoceptor antagonist) or yohimbine (1 mg/kg, i.p., an α~2~-adrenoceptor antagonist). Thirty minutes after administration of saline or prazosin, or yohimbine, mice were treated with either vehicle (5% v/v Tween 80) or α-asarone (20 mg/kg, i.p.). Thirty minutes after vehicle or α-asarone treatment, the immobility time in seconds was measured in the TST ([@B3]; [@B39]). Investigation of the Involvement of Serotonergic System in the Antidepressant-Like Effect of α-Asarone ------------------------------------------------------------------------------------------------------ To assess the involvement of serotonergic system in the antidepressant-like activity of α-asarone, the mice were pretreated with saline, or PCPA (100 mg/kg, i.p., a serotonin synthesis inhibitor, once daily for four consecutive days). On day 5 (24 h after last PCPA treatment), mice received either vehicle (5% v/v Tween 80) or fluoxetine (30 mg/kg, i.p.) or α-asarone (20 mg/kg, i.p.) 30 min prior to TST ([@B17]; [@B14]). In another study, the mice were pretreated with either saline or WAY100635 (0.1 mg/kg, s.c., a selective 5-HT~1A~ receptor antagonist). Thirty minutes after saline or WAY100635 treatment, the mice were administered with either vehicle (5% v/v Tween 80) or 8-OH-DPAT (1 mg/kg, i.p., a selective 5-HT~1A~ receptor agonist) or α-asarone (20 mg/kg, i.p.). The immobility time in seconds was measured after 30 min of vehicle or 8-OH-DPAT or α-asarone ([@B36]; [@B3]). Statistical Analysis -------------------- Values are expressed as mean ± SEM. The behavioral tests were analyzed by one-way ANOVA followed by *post hoc* Dunnett's multiple comparison test or two-way analysis of variance (two-way ANOVA) followed by *post hoc* Bonferroni test using Graphpad prism 5.03 (Graphpad Software, Inc., USA). Statistical significance was set at *p* \< 0.05. Results ======= Effect of α-Asarone *per se* in the TST --------------------------------------- As shown in (**Figure [2](#F2){ref-type="fig"}**), one-way ANOVA results revealed that the acute treatment of α-asarone at relatively lower doses (15 and 20 mg/kg, i.p.) and the reference antidepressant, bupropion (20 mg/kg, i.p.) significantly reduced the immobility time in the TST. In contrast, α-asarone at relatively higher doses (50 and 100 mg/kg, i.p.) significantly increased the immobility time in the TST as compared with vehicle control \[*F*~(6,62)~ = 15.30, *p* \< 0.0001\]. {#F2} Effect of α-Asarone *per se* in the Horizontal Wire Test -------------------------------------------------------- The results of the horizontal wire test revealed that α-asarone (10, 15, 20, 30, 50, and 100 mg/kg, i.p.) did not alter the normal motor coordination that is, it did not affect the grasping reflex, hanging or climbing behavior (data not shown). Effect of α-Asarone *per se* on the Spontaneous Locomotor Activity ------------------------------------------------------------------ As shown in (**Figure [3](#F3){ref-type="fig"}**), one-way ANOVA results revealed that α-asarone at lower doses (10, 15, 20, and 30 mg/kg, i.p.) did not significantly affect the spontaneous locomotor activity, whereas relatively higher doses of α-asarone (50 and 100 mg/kg, i.p.) significantly decreased locomotor activity when compared with the vehicle control group \[*F*~(6,60)~ = 8.363, *p* \< 0.0001\]. {#F3} Involvement of the Noradrenergic System in the Antidepressant-Like Effect of α-Asarone -------------------------------------------------------------------------------------- The results depicted in (**Figure [4A](#F4){ref-type="fig"}**) shows that the anti-immobility effect of α-asarone (20 mg/kg, i.p.) and positive control, bupropion (20 mg/kg, i.p.) was significantly blocked in AMPT (100 mg/kg, i.p.)-pretreated mice in the TST. Two-way ANOVA results revealed that there was a significant effect on AMPT pretreatment \[*F*~(2,42)~ = 12.82, *p* \< 0.0001\], α-asarone or bupropion treatment \[*F*~(1,42)~ = 72.92, *p* \< 0.0001\] and α-asarone or bupropion × AMPT interaction \[*F*~(2,42)~ = 6.20, *p* \< 0.0044\]. Moreover, mice pretreated with prazosin (1 mg/kg, i.p.) significantly reversed the anti-immobility effect of α-asarone (20 mg/kg, i.p.) in the TST. Two-way ANOVA results showed a significant effect on prazosin pretreatment \[*F*~(1,27)~ = 7.47, *p* \< 0.0109\], α-asarone treatment \[*F*~(1,27)~ = 25.76, *p* \< 0.0001\] and α-asarone × prazosin interaction \[*F*~(1,27)~ = 6.50, *p* \< 0.0168; **Figure [4B](#F4){ref-type="fig"}**\]. **Figure [4C](#F4){ref-type="fig"}** shows that yohimbine (1 mg/kg, i.p.) pretreatment significantly inhibited the anti-immobility effect of α-asarone (20 mg/kg, i.p.) in the TST. Two-way ANOVA results revealed a significant effect on α-asarone treatment \[*F*~(1,26)~ = 9.41, *p* = 0.0050\], α-asarone × yohimbine interaction \[*F*~(1,26)~ = 5.43, *p* \< 0.0278\], but not with yohimbine pretreatment \[*F*~(1,26)~ = 2.77, *p* = 0.1082\]. {#F4} Involvement of the Serotonergic System in the Antidepressant-Like Effect of α-Asarone ------------------------------------------------------------------------------------- The anti-immobility effect of α-asarone (20 mg/kg, i.p.) or positive control, fluoxetine (30 mg/kg, i.p.) in the TST was prevented in PCPA (100 mg/kg, i.p., once daily for four consecutive days)-pretreated mice (**Figure [5A](#F5){ref-type="fig"}**). Two-way ANOVA results revealed that there was a significant effect of α-asarone or fluoxetine treatment \[*F*~(1,42)~ = 17.98, *p* \< 0.0001\], α-asarone or fluoxetine × PCPA interaction \[*F*~(2,42)~ = 5.02, *p* \< 0.0111\], and not with PCPA pretreatment \[*F*~(2,42)~ = 1.38, *p* = 0.2630\]. Besides that, mice pretreated with WAY100635 (0.1 mg/kg, s.c.) reversed the anti-immobility effect of α-asarone (20 mg/kg, i.p.) or 8-OH-DPAT (1 mg.kg, i.p.; **Figure [5B](#F5){ref-type="fig"}**). The two-way ANOVA results revealed a significant effect of WAY100635 pretreatment \[*F*~(2,41)~ = 3.52, *p* = 0.0390\], α-asarone or 8-OH-DPAT treatment \[*F*~(1,41)~ = 15.45, *p* = 0.0003\], α-asarone or 8-OH-DPAT × WAY100635 interaction \[*F*~(2,41)~ = 7.44, *p* = 0.0018\]. {#F5} Discussion ========== Depression is a heterogeneous affective disorder which particularly affects the mood and is associated with high rates of recurrence, relapses, and premature deaths. Globally, up to 20% of population are affected with major depression ([@B1]). The TST is a well-validated and widely used animal model to screen potential antidepressants. It is an inexpensive, highly predictive, and is considered a high-throughput screening for the acute behavioral effects of antidepressants ([@B24]). The present TST results demonstrate that acute treatment of α-asarone at relatively lower doses (15 and 20 mg/kg, i.p.) showed an antidepressant-like effect in the mouse model of depression. This result is in accordance with the recent literature reported by [@B9] in which a significant antidepressant-like effect of α-asarone was demonstrated in both the FST and TST at doses of 10 and 20 mg/kg, i.p. However, α-asarone at relatively higher doses (50 and 100 mg/kg, i.p.) significantly increased the immobility time in the TST. Furthermore, in order to eliminate false positive results of psychostimulant-like substances in the TST, the effect of α-asarone on spontaneous locomotor activity in mice was assessed. These results revealed that α-asarone (10--30 mg/kg, i.p.) did not affect the spontaneous locomotor activity, suggesting that the anti-immobility effect of α-asarone (15 and 20 mg/kg, i.p.) in the TST could not have been mediated through stimulation of the CNS. Moreover, α-asarone at higher doses (50 and 100 mg/kg, i.p.) significantly decreased the spontaneous locomotor activity and this result is in good agreement with the previous findings in which acute treatment of α-asarone (48, 50, and 100 mg/kg, i.p.) significantly decreased the locomotor activity in mice ([@B21]; [@B26]; [@B16]). In addition, α-asarone (10--100 mg/kg, i.p.) did not affect the normal motor coordination as indicated in the horizontal wire test, thus alluding to absence of muscle relaxant property of α-asarone up to a dose of 100 mg/kg. This result corroborates with the previous report by [@B26] which demonstrated that the motor coordination of mice was not affected when the mice were pretreated with α-asarone (22 and 60 mg/kg, i.p.) in the rotarod test. Therefore, it can be suggested that the increased immobility time in the TST, and a reduction in spontaneous locomotor activity at higher doses of α-asarone (50 and 100 mg/kg, i.p.) could be attributed to its depressant-like effect and not due to any muscle relaxant effect. The present study, however, could not clarify the exact mechanism(s) involved in the depressant-like effect of α-asarone at relatively higher doses. Based on our thorough literature search, it is postulated that GABAergic mediated mechanism could be involved in the depressant-like effect of α-asarone at higher doses. In earlier studies, α-asarone at relatively higher doses showed antiepileptic activity mediated through GABAergic mechanism in pentylenetetrazole or picrotoxin-induced seizures in mice ([@B26]). Similarly, the electrophysiological studies ([@B11]; [@B37]) also confirmed the facilitatory effect of α-asarone on GABA~A~ receptors. The monoamine theory of depression was proposed by [@B33] and states that depression is caused by functional deficit of the monoamine neurotransmitters mainly noradrenaline (NA) and/or serotonin. The currently available antidepressants which are very effective in treating major depression including monoamine oxidase inhibitors (MAOIs), tricyclic antidepressants (TCAs), selective serotonin reuptake inhibitors (SSRIs), noradrenaline reuptake inhibitors (NRIs) and serotonin and noradrenaline reuptake inhibitors (SNRIs). The key mechanism(s) of action of all of these drugs are similar, that is, by enhancing the brain's noradrenergic and/or serotonergic transmissions ([@B1]; [@B22]). Thus, the monoamine theory of depression is still considered a promising tool in the novel drug discovery for the treatment of major depressive disorder. Down regulation of the central noradrenergic system and the ensuing reduction of brain noradrenaline level are the main key factors responsible for the pathophysiology of depressive disorder ([@B22]). AMPT is an inhibitor of tyrosine hydroxylase, a rate limiting enzyme in the biosynthesis of noradrenaline and dopamine ([@B38]). Mice pretreated with AMPT showed a significant reduction in the brain noradrenaline and dopamine levels without affecting the levels of serotonin ([@B20]). It has also been reported that AMPT pretreated mice demolished the antidepressant activity of bupropion (a non-selective noradrenaline and dopamine reuptake inhibitor) in the TST ([@B14]). Moreover, several studies elucidated the involvement of α~1~, and α~2~- adrenoceptors in the antidepressant-like effect of drugs in animal behavioral models of depression ([@B17]; [@B2]; [@B8]; [@B39]). Evidence suggested that the brain α~1~-adrenoceptors were desensitized in the depressed patients and that the activation of α~1~- adrenoceptors restored the normal mood ([@B35]). In addition, chronic treatment with TCAs enhanced the density of α~1~- adrenoceptors ([@B35]) and importantly, the antidepressant effect of desipramine (a TCA) was blocked in the mice pretreated with prazosin (an α~1~-adrenoceptor antagonist) in the FST ([@B4]). Furthermore, antagonism of presynaptic α~2~-adrenoceptors enhanced the noradrenaline levels and on the other hand, activation of post-synaptic α~2~-adrenoceptors facilitated the antidepressant-like effect ([@B40]). Similarly, the antidepressant-like effect of clonidine (an α~2~-adrenoceptors agonist) was inhibited by yohimbine (α~2~-adrenoceptor antagonist) in the FST ([@B25]). It has also been found that the α~2~-adrenoceptors were up-regulated in depressed patients and that chronic antidepressant therapy decreased its up-regulation ([@B6]). These evidences suggested the importance of the noradrenergic (α~1~ and α~2~ adrenoceptors) system for effective antidepressant therapy. In the present study, the mice pretreated with AMPT (a catecholamine synthesis inhibitor) prevented the antidepressant-like effect of α-asarone and in addition, the mice pretreated with prazosin (an α~1~-adrenoceptor antagonist) and yohimbine (α~2~-adrenoceptor antagonist) also abolished the antidepressant-like activity of α-asarone, indicating the involvement of the noradrenergic (α~1~ and α~2~ adrenoceptors) system in the antidepressant-like effect of α-asarone. The down-regulation of brain serotonergic system is strongly implicated in depressive disorder ([@B19]). PCPA, a selective serotonin synthesis inhibitor, inhibits tryptophan hydroxylase and depletes serotonin level in the brain without affecting the brain noradrenaline and dopamine levels ([@B31]). Several studies highlighted that the pretreatment of PCPA inhibited the antidepressant-like effect of fluoxetine (a selective serotonin reuptake inhibitor) in the TST and FST ([@B17]; [@B14]). The Positron emission tomography with \[11C\] WAY100635 in depressed patients revealed a decreased expression of 5-HT~1A~ receptors in several brain regions including frontal cortex and hippocampus ([@B5]; [@B32]). Furthermore, one of the mechanisms involved in the antidepressant-like effect of MAOI, TCAs, or SSRIs is mediated by its interaction with 5-HT~1A~ receptors ([@B10]). In another study, it was found that the 5-HT~1A~ knockout mice treated with the fluoxetine or paroxetine (SSRIs) did not decrease the immobility time whereas desipramine (TCA) decreased the immobility in the TST ([@B20]), which clearly suggested the importance of 5-HT~1A~ receptors in the antidepressant effect of SSRIs. In the present study, the mice pretreated with PCPA (a selective serotonin synthesis inhibitor) blocked the antidepressant-like effect of α-asarone. Furthermore, pretreatment of mice with WAY100635 (a 5-HT~1A~ antagonist) abolished the antidepressant-like effect of α-asarone. These results demonstrated the involvement of the serotonergic (particularly the 5-HT~1A~ receptors) system in the antidepressant-like effect of α-asarone. In summary, our study demonstrate that acute treatment of α-asarone exhibited antidepressant-like activity at relatively lower doses (15 and 20 mg/kg, i.p.) without affecting either locomotor activity or motor coordination in mice. On the other hand, α-asarone at relatively higher doses (50 and 100 mg/kg, i.p.) significantly enhanced the immobility time in the TST and diminished the spontaneous locomotor activity which indicates a depressive-like effect of α-asarone at higher doses. Moreover, the antidepressant-like effect of α-asarone was prevented in the mice pretreated with AMPT or prazosin or yohimbine or PCPA or WAY100635, thereby indicating the involvement of both the noradrenergic and serotonergic systems in the antidepressant-like effect of α-asarone. Conclusion ========== This study results indicates that acute treatment of α-asarone exhibited a biphasic effect on the immobility time in the TST, with an antidepressant-like activity at lower doses and depressive-like effect at higher doses. The antidepressant-like effect of α-asarone is mediated by its interaction with noradrenergic (α~1~ and α~2~ adrenoceptors) and serotonergic (particularly, the 5-HT~1A~ receptors) systems. Therefore, we suggest that α-asarone could be singled out and further tested as a potential drug for the treatment of major depressive disorder. Author Contributions ==================== RC designed, performed the experiments, analyzed the data, and wrote the manuscript, VP participated in the study design and critically revised the manuscript for important intellectual content. ZM critically revised the manuscript for important intellectual content. All authors read and approved the 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. This study was supported by University of Malaya Research grants \[PG008-2015A; RG495-13HTM and HIR MOHE Project No: UM.C/625/1/HIR/MOHE/MED/05 (H-20001-E000088)\]. The funding sources were not involved in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. We are grateful to the management of University of Malaya for providing financial assistance and necessary infrastructure to carry out this research. CNS : central nervous system GABA : gamma-aminobutyric acid 5-HT : 5- hydroxytryptamine i.p. : intraperitoneal injection s.c. : subcutaneous injection TST : tail suspension test [^1]: Edited by: *Luyong Zhang, China Pharmaceutical University, China* [^2]: Reviewed by: *Fang-Rong Chang, Kaohsiung Medical University, Taiwan; Manoj Gajanan Kulkarni, University of KwaZulu-Natal, South Africa* [^3]: This article was submitted to Ethnopharmacology, a section of the journal Frontiers in Pharmacology

All relevant data are within the paper. Introduction {#sec001} ============ Dopamine (DA) is a biogenic amine neurotransmitter found in both vertebrates and invertebrates that affects a wide variety of physiological and behavioral functions, including reproduction\[[@pone.0193999.ref001],[@pone.0193999.ref002]\], hormone synthesis and release\[[@pone.0193999.ref003]\], locomotion\[[@pone.0193999.ref004]\], respiration\[[@pone.0193999.ref005]\], feeding behavior\[[@pone.0193999.ref006]\], and the circadian rhythm\[[@pone.0193999.ref007]\]. The dopamine receptors can be divided into five subtypes (DA~1~-DA~5~), which all belong to the family of G protein-coupled receptors (GPCRs). According to their conserved structures, signaling mechanisms and pharmacological profiles, these receptors are further classified into two types the D1-like and D2-like receptors\[[@pone.0193999.ref001]\]. The D1-like receptors include the DA~1~ and DA~5~ subtypes, which activate adenylyl cyclase, resulting in increased levels of intracellular cyclic adenosine monophosphate (cAMP); regulate cell metabolism, including ion channel function, and desensitize GPCRs, leading to the release of neurotransmitters. D2-like receptors consist of the DA~2~, DA~3~, and DA~4~ subtypes, which inhibit adenylyl cyclase through the coupled signal transduction pathway and thus decrease cAMP; D2-like receptors can be blocked by the pertussis toxin\[[@pone.0193999.ref008]\]. In mammals, these receptors occur in the brain, peripheral nervous system, cornea of the eye, heart, kidney and lymphocytes\[[@pone.0193999.ref002]\]. However, only a few studies on the dopamine receptor in crustaceans have been reported. Using RACE technology and a degenerate PCR strategy with conventional library screening, the gene and protein sequences of DA~1α~, DA~1β~ and DA~2α~ in *Panulirus interruptus* have been obtained\[[@pone.0193999.ref008],[@pone.0193999.ref009]\]. In the sequencing of transcriptomes from the nervous systems of *Cancer borealis* and *Homarus americanus*, DA~1α~, DA~1β~ and DA~2α~were found in both decapod crustaceans\[[@pone.0193999.ref010]\]. In addition, a type 1 dopamine receptor from *Penaeus monodon* has been identified\[[@pone.0193999.ref002]\]. However, information is still lacking on gene and protein sequences of dopamine receptors in economically valuable decapod crustaceans, such as *E*. *sinensis*, a richly nutritious species with high market demand that has become economically important in Chinese freshwater aquaculture\[[@pone.0193999.ref011]\]. Light influences the growth and development of crustaceans\[[@pone.0193999.ref012],[@pone.0193999.ref013]\], such as *Macrobrachium rosenbergii* and *Portunus pelagicus*, and daily changes in dopamine synthesis and release depend on the interactions between the photoreceptors and the dopaminergic neurons, where dopamine release is induced by light\[[@pone.0193999.ref014],[@pone.0193999.ref015]\]. High levels of dopamine have been detected during light periods and low levels during dark periods\[[@pone.0193999.ref016]--[@pone.0193999.ref018]\], for this reason, it is believed that dopamine promotes light adaptation. However, constant light results in a dramatic reduction in dopamine levels in chicken retina\[[@pone.0193999.ref019]\]. Constant light and constant darkness have significant effects on survival and growth of larvae of *P*. *pelagicus*\[[@pone.0193999.ref012]\] and *M*. *rosenbergii*\[[@pone.0193999.ref013]\], but the effects of different photoperiods on dopamine receptors in crablets remain uncertain. In addition to promoting light adaptation, dopamine can also participate in feeding regulation. Exogenous injection of DA has been found to significantly decrease food intake compared to that of a control group in neonatal layer-type chickens\[[@pone.0193999.ref006]\], and cannabinoid-induced feeding behavior may be modulated by dopamine receptor 2\[[@pone.0193999.ref020]\]. However, by promoting either the initiation or cessation of feeding behavior, increased activity of DA neurons can either increase or reduce food intake\[[@pone.0193999.ref021]\]; inhibition of D1-type dopamine receptor neurons decreases food intake\[[@pone.0193999.ref022]\]. Dopamine receptors have also been found to be distributed in the intestinal tract and are considered to be involved in regulating gastrointestinal motility\[[@pone.0193999.ref023],[@pone.0193999.ref024]\]. In addition, the presence of specific receptors on the membranes of target cells is essential for dopamine to produce any physiological effects. Inhibitors of dopamine receptors can block the effects of dopamine\[[@pone.0193999.ref025]\]. Due to the diversity of dopamine receptors, light stimulation has different effects on their expression levels. An understanding of the variation of receptors will help to identify cellular targets of DA and to understand which receptors are activated for particular processes. The present study describes the molecular cloning and characterization of the dopamine receptor 2 full-length cDNA from *E*. *sinensis* and its expression profile in various tissues under different photoperiods and feeding statuses. Materials and methods {#sec002} ===================== Animals and sampling {#sec003} -------------------- In this study, the experimental animals (n = 48) were healthy crablets (exhibiting secondary sexual characteristics) with initial masses of 13.43±1.81 g, collected from the Shuxin crab base in Chongming, Shanghai (China). Crabs were housed for one week for acclimatization in clear glass aquaria (length× width× height = 120×60×40cm) with sufficient ambient medium and cyclic water flow. Crabs were fed once a day at 09:00. Thirty crabs were acclimatized to 26±1°C and assigned randomly to three groups: a control group (L:D = 12h:12h), a group held in constant darkness (L:D = 0h:24h) and a group held in constant light (L:D = 24h:0h). There were 10 crabs per group, and treatments continued for 14 days\[[@pone.0193999.ref026]\]. Then crabs from the control group were frozen on ice and dissected, and different tissues, including the gill, heart, muscle, hepatopancreas, intestine, cranial ganglia, thoracic ganglia, eyestalks, and hemolymph were harvested. At the same time, eyestalks, cranial ganglia and thoracic ganglia were harvested on ice from the constant darkness group and constant light group. For a separate group of 18 crabs, the hepatopancreas and intestine were collected on ice after a week of rearing. Tissues were collected at 08:00 (before feeding, n = 6), 10:00 (feeding period, feeding time was 09:00 to 10:00, n = 6), or 16:00 (after feeding, feces were mostly in the hind gut and the crabs began to evacuate 6h after feeding, n = 6), \[[@pone.0193999.ref027]\]. All the samples were stored at -80°C until RNA isolation. Nucleic acid extraction {#sec004} ----------------------- Total RNA was extracted from *E*. *sinensis* using the RNAiso Plus reagent (RNA Extraction Kit, TaKaRa, Japan) according to the manufacturer's instructions. Briefly, tissues were ground in a mortar with liquid nitrogen and collected in 1.5 ml centrifuge tubes. The RNAiso Plus reagent was added (1 ml), and samples were left at room temperature for 5 min. Samples were then centrifuged 5 min at 4°C and 12000 rpm, and the supernatant was collected in new 1.5 ml tubes. Chloroform (200 μl) was added, and samples were then oscillated and again left at room temperature for 5 min before a 15 min centrifugation at 4°C and 12000 rpm. The supernatant was collected in new 1.5 ml tubes and 500 μl of isopropyl alcohol was added. Samples were left at room temperature for 10 min and then centrifuged 10 min at 4°C and 12000 rpm. Pellets were washed with 1 ml of 75% alcohol and centrifuged 10 min at 4°C and 7500 rpm. The supernatant was removed, and the remaining pellets were dried and dissolved in 30 μl of DEPC-treated water. The concentration and quality of the total RNA were estimated by micro-volume ultraviolet-visible spectrophotometer (Quawell Q5000; Thmorgan, China) and agarose-gel electrophoresis, respectively. Cloning of full-length *E*. *sinensis* DA~2~ cDNA {#sec005} ------------------------------------------------- Transcriptomes sequences were obtained from the Y-organ of *E*. *sinensis*. The amino acid sequence of the EST (length: 637bp) was verified to be highly homologous to the *C*. *borealis* dopamine receptor 2 (AOG14374.1) using BlastX analysis. A pair of gene primers, DA~2~-F and DA~2~-R ([Table 1](#pone.0193999.t001){ref-type="table"}), was designed to amplify the full-length DA~2~ cDNA from *E*. *sinensis* for sequence verification. 10.1371/journal.pone.0193999.t001 ###### Primers used in cloning and characterizing the DA~2~ gene. {#pone.0193999.t001g} Primers Sequences (5'-3') Usage --------------- ------------------------------ ---------- DA~2~-F `CTAGCCATAGTTCTGGCGGCG` RT-PCR DA~2~-R `TCCTTACCGGACCACAGAACG` RT-PCR DA~2~-3'Outer `TGAACTCCTTCCTCAACCCCG` 3'RACE DA~2~-3'Inner `GTATGCCAGAGCTAGCCGGG` 3'RACE DA~2~-5'Outer `ATCTTCACTTTCTTCTGCTTCACGA` 5'RACE DA~2~-5'Inner `CTCGTCTGGCTTACGTTCTCGATCAC` 5'RACE qRT-DA~2~-F `TGCTATTATCTGGGTGGTGT` q-RT-PCR qRT-DA~2~-R `ATGATGAAGTCTGCGTTGTG` q-RT-PCR 18S-F `TCCAGTTCGCAGCTTCTTCTT` q-RT-PCR 18S-R `AACATCTAAGGGCATCACAGA` q-RT-PCR Four gene-specific primers, DA~2~-3'Outer, DA~2~-3'Inner, DA~2~-5'Outer and DA~2~-5'Inner ([Table 1](#pone.0193999.t001){ref-type="table"}), were designed based on the 637bp singlet to clone the 3'- and 5'-ends of the DA~2~ cDNA by rapid amplification of cDNA ends (RACE) using the SMARTer RACE 5'/3' Kit (Clontech, USA). The 3'- and 5'-end cDNA templates were synthesized according to the manufacturer's instructions. Specific products were obtained via touchdown PCR and nested PCR. Touchdown PCR was carried out as follows: 94°C for 5 min; 5 cycles of 94°C for 30 s and 72°C for 3 min; 5 cycles of 94°C for 30 s, 70°C for 30 s, and 72°C for 3 min; 30 cycles of 94°C for 30 s, 68°C for 30 s, and 72°C for 3 min; a final extension for 10 min at 72°C; and a cooling hold at 4°C. Nested PCR amplification conditions were as follows: 94°C for 3 min; 34 cycles of 94°C for 30 s, 66°C for 30 s, and 72°C for 2 min; 72°C for 7 min, and a cooling hold at 4°C. Amplification products were run on a 1.5% agarose gel and purified with a TIANgel Midi Purification Kit (TIANGEN, China). The DNA fragments were cloned into a pMD18-T vector (TaKaRa, Japan) and transformed into TOP10 chemically competent *E*. *coli* cells (TIANGEN, China). Bacteria were grown according to the manufacturer's instructions. The positive clones containing the inserts of the expected size were sequenced using M13±primers by Sangon Biotech (Shanghai). Sequence analysis {#sec006} ----------------- The generated sequences were verified for similarity by using the BLAST programs (<http://blast.ncbi.nlm.nih.gov/>). Then, to obtain the full-length DA~2~ cDNA, the partial fragment, and the 3'- and 5'-end sequences were assembled. After the open reading frame (ORF) was obtained using the ORF finder (<http://www.ncbi.nlm.nih.gov/gorf/gorf.html>), the coding region sequences were translated into amino acid sequences by using the sequence manipulation suite (SMS) tool (<http://www.bio-soft.net/sms/index.html>). The molecular mass and the theoretical isoelectric point of the DA~2~ protein were predicted using the Compute pI/Mw tool (<http://cn.expasy.org/tools/pi_tool.html>). The trans-membrane domains of the protein sequence were predicted by the TMHMM server (<http://www.cbs.dtu.dk/services/TMHMM>). Protein phosphorylation sites were predicted using DISPHOS 1.3 (<http://www.dabi.temple.edu/disphos/>). N-glycosylation sites were predicted using the NetNGly 1.0 server (<http://www.cbs.dtu.dk/services/NetNGlyc/>). An amino acid multiple sequence alignment was performed with the ClustalX program, and phylogenetic tree was constructed using the neighbor-joining (NJ) method with1000 bootstraps in the MEGA 5.0 program\[[@pone.0193999.ref028]\]. Real-time quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) analysis {#sec007} ----------------------------------------------------------------------------------------- The mRNA expression pattern of the DA~2~ transcript in various tissues (gill, heart, muscle, hepatopancreas, intestine, cranial ganglia, thoracic ganglia, eyestalks, and hemolymph) were measured by qRT-PCR using 18S ribosome RNA (18S) as a reference gene. We confirmed that 18S expression was stable. Total RNA was extracted from each sample, and reverse transcription was performed with equal quantities of total RNA (1μg). Relative quantification was performed using the ABI 7500 Real-Time PCR System (Life Technology, USA). Gene-specific primers, qRT-DA~2~-F and qRT-DA~2~-R ([Table 1](#pone.0193999.t001){ref-type="table"}), were designed based on the cloned DA~2~ cDNA to produce a 126bp amplicon. Real-time qPCR amplification reactions were performed in a final volume of 10 μl, which contained 5 μl of 2×SYBR Premix Ex Taq^TM^ (TaKaRa, Japan), 1 μl of diluted cDNA template, 3.4 μl of PCR-grade water, 0.2 μl of ROX Reference Dye Ⅱ, and 0.2 μl each of the forward and reverse primers. The PCR conditions used were as follows: 95°C for 30 s; 40 cycles of 95°C for 5 s, and 60 0078C for 34 s; and generation of a melting curve at 95°C for 15 s, 60°C for 1min, and 95°C for 15 s. Samples were run in triplicate, and results were normalized to the expression of the reference gene 18S. The DA~2~ cDNA expression levels were calculated by the 2^-ΔΔCt^ comparative threshold cycle (Ct) method (where ^Δ^Ct = ^Δ^Ct sample - ^Δ^Ct reference). Data were analyzed and presented as triplicate means± SE (standard error) and as n-fold differences relative to the control data. Statistical analysis {#sec008} -------------------- Statistical analysis of relative gene expression was performed using SPSS software (Chicago, USA; Version 17.0). Data are presented as the means± SE. Statistical significance was determined using one-way analysis of variance and post-hoc Duncan multiple range tests. *P*\<0.05 indicated statistical significance. Results {#sec009} ======= Cloning and identification of the DA~2~ cDNA {#sec010} -------------------------------------------- The full-length DA~2~ cDNA was isolated from the intestine of the Chinese mitten crab. The full-length cDNA (2369bp) contained a 1770bp open reading frame (ORF), which encodes a putative DA~2~ protein with 589 amino acids, a 192bp 5'-untranslated region (UTR), and a 407bp 3'-UTR with a 27bp poly (A) tail. Sequence analysis revealed that the DA~2~ protein has a theoretical isoelectric point of 8.37 and a molecular weight of 64.81kDa. The deduced DA~2~ protein has three major domains, four extracellular domains, seven transmembrane domains (TM-Ⅰ through TM-Ⅶ) and four cytoplasmic domains. The transmembrane domains consist of seven hydrophobic regions, which are highly conserved compared with those of other dopamine receptors. In contrast, the amino-terminal region, the second extracellular loop between TM-Ⅳ and TM-Ⅴ, and most of the third cytoplasmic loop between TM-Ⅴ and TM-Ⅵ display a low degree of sequence identity. Despite the low homology between the extracellular N-terminal and the third cytoplamic loops of the *E*. *sinensis* DA~2~ receptor and those of other DA~2~ receptors, consensus motifs for N-linked glycosylation sites (N-x-\[S/T\]) and consensus sites for phosphorylation by protein kinase C (PKC) (S/T-x-\[R/K\]) are found in these domains ([Fig 1](#pone.0193999.g001){ref-type="fig"}). As in the DA~2~ receptors of other species, there is a conserved DRY motif in the second intracellular loop. Sequence analysis of the DA~2~ cDNA with BLASTn and BLASTp revealed a significant sequence similarity to *C*. *borealis* and *P*. *interruptus* DA~2~ sequences found in the National Center for Biotechnology Information database. {#pone.0193999.g001} Homology analysis of the DA~2~ gene {#sec011} ----------------------------------- The degree of homology of the DA~2~ gene with other representative vertebrate and invertebrate DA~2~ amino acid sequences was investigated via multiple sequence alignment in ClustalX ([Fig 2](#pone.0193999.g002){ref-type="fig"}). The alignment indicated that the *E*. *sinensis* DA~2~ sequence shares high amino acid identity with the DA~2~ sequence of *C*. *borealis* (84%) and *P*. *interruptus* (72%). The alignment also revealed that the amino acid sequence in the transmembrane domains is highly conserved among these three species. {#pone.0193999.g002} A neighbor-joining phylogenetic tree was constructed based on the reported DA~2~ amino acid sequences using MEGA 5.0 software ([Fig 3](#pone.0193999.g003){ref-type="fig"}), with confidence in the resulting tree branch topology measured by bootstrapping through 1000 pseudo replicates. The tree provides evidence that the *E*. *sinensis* DA~2~ gene is grouped with DA~2~ genes of other species. {#pone.0193999.g003} Tissue distribution of expressed DA~2~ mRNA {#sec012} ------------------------------------------- To determine mRNA expression patterns of DA~2~ in *E*. *sinensis*, the total RNA extracted from various tissues including gill, heart, muscle, hepatopancreas, intestine, cranial ganglia, thoracic ganglia, eyestalks, and hemolymph was reverse transcribed and subjected to quantitative real-time PCR with qRT-DA~2~-F and qRT-DA~2~-R primers ([Table 1](#pone.0193999.t001){ref-type="table"}). The results ([Fig 4](#pone.0193999.g004){ref-type="fig"}) show that DA~2~ mRNA was expressed in all of these tissues, but the expression levels varied. DA~2~ mRNA expression levels were highest in the cranial ganglia and thoracic ganglia (*P*\<0.05); only low levels of expression were detected in other tissues (*P*\<0.05). {#pone.0193999.g004} Effect of photoperiod on DA~2~ mRNA expression in the cranial ganglia and thoracic ganglia {#sec013} ------------------------------------------------------------------------------------------ We determined the effect of photoperiod on DA~2~ mRNA expression in the cranial ganglia and thoracic ganglia after the crabs were cultured in different photoperiods for 14 days ([Fig 5](#pone.0193999.g005){ref-type="fig"}). The relative expression of DA~2~ in the cranial ganglia was significantly induced by constant darkness compared with the control treatment (*P*\<0.05), while there was no effect of constant light in this tissue. Although the DA~2~ mRNA expression in the thoracic ganglia appeared higher in crabs exposed to constant darkness and constant light than in crabs under control condition, there was no statistically significant difference between treatment (*P*\>0.05). As in the cranial ganglia, the expression of DA~2~ in eyestalks was significantly increased by constant darkness (*P*\<0.05), but there was no effect of constant light (*P*\>0.05). {#pone.0193999.g005} DA~2~ mRNA expression levels during different feeding statuses {#sec014} -------------------------------------------------------------- To research the relationship between DA~2~ receptors and feeding/digestion, we determined the DA~2~ expression levels in the hepatopancreas and intestines of crabs during three feeding status ([Fig 6](#pone.0193999.g006){ref-type="fig"}). In both tissues, the levels of DA~2~ expression were significantly higher after feeding than before feeding (*P*\<0.05), but there were no significant differences between the before feeding and during feeding periods (*P*\>0.05). In the hepatopancreas, the DA~2~ expression level after feeding was also significantly higher than during the feeding period (*P*\<0.05), while in the intestines, the DA~2~ expression level after feeding was not significantly different from expression during with feeding period (*P*\>0.05). {#pone.0193999.g006} Discussion {#sec015} ========== In the present study, we have characterized a DA~2~ receptor of the Chinese mitten crab, *E*. *sinensis*. The obtained sequence has considerable similarity with orthologous receptors from other invertebrates\[[@pone.0193999.ref002],[@pone.0193999.ref008]--[@pone.0193999.ref010],[@pone.0193999.ref029]\] and vertebrates\[[@pone.0193999.ref030],[@pone.0193999.ref031]\]. The encoded 589 amino acids sequence contains typical characteristics of DA~2~ receptors, such as a large third intracellular loop, a short C-terminal region, a DRY motif in the second intracellular loop and other conserved consensus sequences\[[@pone.0193999.ref032]\]. Sequence comparison and phylogenetic analysis suggest that DA~2~ is a member of the D-2 like subfamily of dopamine receptors. The DA~2~ in *E*. *sinensis*is most related to the DA~2~ of *P*. *interruptus*, while lower but still remarkable degrees of homology are detected with other arthropods' DA~2~ receptors; this suggests that the *E*. *sinensis* DA~2~ possesses a highly conserved structure typical of type 2 dopamine receptors. All residues involved in dopamine receptor activation were present in the *E*. *sinensis* DA~2~ sequence. These include several putative phosphorylation sites and 5 putative N-glycosylation sites, such as those found in mammalian DA~2~ receptors, which are highly phosphorylated and glycosylated neural receptors\[[@pone.0193999.ref031]\]. In addition, the *E*. *sinensis* DA~2~ contains the conserved DRY (Asp-Arg-Tyr) motif at the interface between TM-Ⅲ and the second intracellular loop; the DRY tripeptide is the key to the conformational changes necessary for receptor activation\[[@pone.0193999.ref033]\]. The Phe528 in TM-Ⅵ of the DA~2~ is analogous to the conserved phenylalanine in other dopamine receptors that interacts with the aromatic biogenic amine ligand. Two serine residues in TM-Ⅴ (Ser275 and Ser276) are known to be involved in the formation of hydrogen bonds with the catechol hydroxyl group of the dopamine agonist\[[@pone.0193999.ref001]\]. These interactions aid in the proper positioning of dopamine in the binding pocket of the receptor\[[@pone.0193999.ref034]\]. On the basis of our analysis, we conclude that the patterns of conservation and divergence observed in the DA~2~ will help to describe the parts of the receptor molecule that are important for proper receptor function\[[@pone.0193999.ref030]\]. A fundamental aspect of dopamine function in the whole organism is the localization of its receptors in the various areas of the nervous system or in the periphery. When studying DA~2~ gene transcript levels with qRT-PCR, the highest expression was observed in the central nervous tissues, including the cranial ganglia and the thoracic ganglia, followed by the periphery tissues. This result has been supported by several studies showing that the DA~2~ receptor is expressed in the central nervous system and in peripheral tissues. In *Oreochromis niloticus*, higher expression of DA~2~ is found in the anterior part of the brain than in other parts of the brain, and DA~2~ is also expressed in the pituitary gland, liver and gills\[[@pone.0193999.ref031]\]. high levels of DA~2~ have also been found in the pituitary gland of *Xenopus*\[[@pone.0193999.ref030]\]. These studies were thus somewhat similar to our results; however, in contrast to our results, the DA~2~ receptor is not expressed in the heart and muscle of the tilapia, *Oreochromis niloticus*. DA~2~ receptors are also expressed by amacrine, bipolar, and ganglia cells\[[@pone.0193999.ref035],[@pone.0193999.ref036]\] and possibly by the intrinsically photosensitive retinal ganglion cells\[[@pone.0193999.ref037]\], functioning as both postsynaptic receptors and autoreceptors that inhibit dopamine release\[[@pone.0193999.ref038]\]. In addition, the DA~2~ receptor is expressed at high levels in the caudate nucleus, putamen, olfactory bulb, substantia nigra, nucleus accumbens, and ventral tegmental area and is found at low levels in the hypothalamus, kidney, blood vessels, heart, septum, cortex, gastrointestinal tract and sympathetic ganglia in mammals such as rats and humans\[[@pone.0193999.ref023]\], where it is involved in reward-motivation functions, blood pressure regulation, working memory, and gastrointestinal motility. Thus, overall, DA~2~ receptor primarily present in the nervous system, but also present in peripheral tissues. In the eyestalks and intestines of *E*. *sinensis*, which contain neurons, the expression levels of DA~2~ are similar to levels in the heart, gill and hemolymph; this may be caused by spatial and temporal differences in DA~2~ gene expression. Further research on expression under different environmental conditions is needed to investigate this possibility. It is well known that DA mediates a various functions via different DA receptors. DA~1~ receptors are involved in coordinating metamorphosis in *Drosophila*\[[@pone.0193999.ref039]\], and DA~1~ and DA~2~ receptors regulate the phase change of migratory locust in two different directions\[[@pone.0193999.ref040]\]. In the inner retina, the DA~2~ receptor also plays a role in regulating the development of light responses\[[@pone.0193999.ref041]\], and in *Daphnia magna*, the DA~2~ receptor is involved in swimming behavior\[[@pone.0193999.ref042]\]. In the mushroom body of the silkworm, DA~2~ plays a role in the release of the diapauses hormone\[[@pone.0193999.ref029]\]. In the rat, the DA~3~ receptor may reinforce the effects of cocaine and may be a useful target for treating cocaine abuse\[[@pone.0193999.ref043]\]. Recent research has shown that DA~4~ activation induces the hippocampal neuronal calcium response\[[@pone.0193999.ref044]\], and activation of DA~5~ inhibits gastric cancer cell growth\[[@pone.0193999.ref045]\]. In this study, we researched, for the first time, the effect of photoperiod on DA~2~ mRNA expression. Light is one of the important environmental factors affecting crustaceans' survival, directly or indirectly influencing their growth, feeding and reproduction\[[@pone.0193999.ref013],[@pone.0193999.ref046],[@pone.0193999.ref047]\], and light environment is the major factor regulating the synthesis and metabolism of dopamine\[[@pone.0193999.ref016]\]. As a benthic animal, the Chinese mitten crab prefers a dark environment, and planting waterweed can have a shading and cooling effect that is beneficial to growth and survival\[[@pone.0193999.ref048]\]. Boosting dopamine can also promote biological growth and survival\[[@pone.0193999.ref049]\]. In this study, the effect of photoperiod on DA~2~ mRNA expression in the cranial ganglia, thoracic ganglia and eyestalks was analyzed by qRT-PCR. We found that the expression level of the DA~2~ receptor was significantly induced by constant darkness in the cranial ganglia and eyestalks but not affected by constant light in any tissues. Therefore, we speculate that a prolonged dark period is good for the growth of these crabs. Our results are in conformity with Dubocovich's finding that constant light activates the dopamine-containing retinal neurons, leading to elevated dopamine release and DA~2~ receptor down-regulation\[[@pone.0193999.ref050]\]. Furthermore, other studies have shown that in the retina, exposure to constant light induces dopamine release and dopamine receptor 2 down-regulation\[[@pone.0193999.ref019],[@pone.0193999.ref051]\], and under different light:dark cycles, retinal levels of dopamine are high during light phases and low during dark phases\[[@pone.0193999.ref016]\]. We believe that the main causes of this difference are species differences and tissue differences. However, there are few studies on the effects of constant darkness on dopamine and dopamine receptors. Therefore, this aspect remains to be thoroughly investigated. Previous studies have found that feeding status affects glucose metabolism\[[@pone.0193999.ref052]\], insulin secretion\[[@pone.0193999.ref053]\] and hypothalamic neuronal activity\[[@pone.0193999.ref054]\] in vertebrates and that food intake is accompanied by a significant decrease in DA levels in rats compared with levels before food was provided\[[@pone.0193999.ref055]\]. However, the effect of food intake on DA receptors is not clear. In a previous study, we divided feeding status into 3 categories: before feeding, during the feeding period and after feeding\[[@pone.0193999.ref027]\]. In this study, we studied the variation of DA~2~ receptor expression in tissues from crabs of different feeding statuses, and the results showed that the expression levels of the DA~2~ receptor were significantly higher after feeding than before feeding or during the feeding period, indicating that the DA~2~ receptor plays an important digestive role in the hepatopancreas and intestines\[[@pone.0193999.ref056]\]. Although, dopamine hinders the absorption rate of glucose, it seems to play a role in regulating the digestion and transport function of the enterocyte membrane in rats\[[@pone.0193999.ref057]\]. At the same time, in the hepatopancreas of *Cyrtograpsus angulatus*, DA significantly decreases lipase activity and the digestive capacity was reduced\[[@pone.0193999.ref058]\]. These findings suggest that there are differential and specific mechanisms by which DA modulates the activity of digestive enzymes in such tissues. Our research on the effects of DA on digestion is of great significance. There have been few studies investigating which types of dopamine receptors play roles in digestion, and therefore these pathway have not yet been elucidated. We are very grateful to professor Gang Wang (Shanghai Jiao Tong University, China) for his guidance on gene cloning in this experiment. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.

INTRODUCTION ============ Antibodies are the primary tools of the immune system, which participate in the protection of the organism against pathogenic microorganisms. The significance of antibodies is growing as researchers become aware of their potential not only as tools to be used in diagnostics, but in therapy as well \[[@R1]\]. Antibodies have been successfully used to treat certain forms of oncological conditions. Over the past decades, monoclonal antibodies have been widely used in diagnostics and for research purposes. Yet, the conventional methods used to obtain monoclonal antibodies, based on dealing with animal-origin cells, make difficult their use as therapeutic agents. Introduction of these monoclonal bodies into the human organism may result in the onset of an undesirable immune reaction, particularly, if used repeatedly \[[@R2]\]. In order to prevent the emergence of such an immune response, the following approaches have been developed: production of recombinant immunoglobulins in which the regions that are not responsible for antigen recognition are replaced by corresponding fragments of human origin (humanized antibodies), or removal of the domains that are not involved in antigen binding (mini-antibodies). The so-called recombinant technologies, based on the use of libraries comprising sequences from human antibodies, have found increasing application over the past decade. When constructing these libraries, variable domains of the heavy and light strands are linked in the expression vector via random screening within one reading frame via the linker sequence \[[@R3]\]. It is rather laborious to deal with cumbersome libraries of these single-stranded antibodies (scFv), and only in rare cases is a highly affine antibody finally obtained. Certain difficulties are associated with the instability of genetic constructions, the low level of product expression, and its solubility \[[@R4]\]. A significant breakthrough in this field has been the detection of non-canonical antibodies in members of the Camelidae biological family. These antibodies do not contain light strands and represent a dimer of shortened heavy strands \[[@R5], [@R6]\]. An immune response with the participation of these antibodies can be induced by conventional immunization. There are a number of advantages in using the repertoire of these non-canonical antibodies to create libraries of sequences of variable domains (for the heavy strand only). The single-domain structure of the recognizing variable domain stipulates a small size of the antigen-binding fragment (mini-antibodies), high stability, and solubility \[[@R7]\]. Thanks to their structure, mini-antibodies can be used to reveal epitopes that are hidden for the conventional immunoglobulins. The expression from a single gene simplifies genetic engineering procedures and, therefore, the work with the libraries containing the sequences of variable domains. Low immunogenicity (conditioned by the high homology of the sequences of mini-antibodies with a variable domain of heavy strands of human IgG3) and the relative simplicity of the humanization procedure open broad opportunities for the application of mini-antibodies in the design of novel pharmaceutical agents \[[@R8]\]. These features of the structure of mini-antibodies and the simplicity with which their genes can be manipulated enable efficient and economical production of large amounts of a mini-antibody, using various expression systems \[[@R9]\]. The use of the prokaryotic expression system to produce mammalian proteins has to do with the possibly low functional activity of the proteins obtained, due to the absence of a system for post-translational modification in prokaryotic cells. Moreover, no matter how thorough the purification, the final product can still be contaminated with pyrogenes. One of the promising methods for delivering genetic material to target cells is the use of viral vectors. Expression constructions bearing one or several recombinant genes are incorporated into the viral genome using methods of genetic engineering. Vectors based on the genome of the adeno-associated virus have been proposed in a number of studies \[[@R10], [@R11]\] for delivery of mini-antibody genes to target cells. Adenoviral vectors are among the most universal tools used for delivery and expression of recombinant genes in mammalian cells. It is known that recombinant adenoviruses efficiently transfer the genes of bacterial and viral antigens, cytokines, growth factors, and other proteins to the target cells, ensuring a high level and duration of target gene expression \[[@R12]\]. Adenoviral vectors are capable of transducing both dividing and postmitotic cells. Adenoviral DNA remains in its extrachromosomal form, whereas the recombinant virus is excreted from the organism within 4--5 weeks \[[@R13], [@R14]\]. The production of recombinant adenoviruses is characterized by the following feature: the virus is capable of reproducing only *in vitro* in special cell lines, which ensures the vector's safety \[[@R15]\]. The fact that recombinant adenoviral vectors can be used efficiently for the expression of antigen-binding fragments of antibodies is borne out by the example of mini-antibodies to the cell epitope (the epidermal growth factor receptor (erbB-2) and anthrax toxin component) \[[@R16], [@R17]\]. The aim of the present work is to examine how recombinant adenoviral vectors can be used for delivery and efficient expression of single-domain mini-antibodies (nanoantibodies) obtained using the novel technology of generation of special single-stranded antibodies extracted from camel. The nanoantibody earlier obtained and characterized to the cell cytokeratine-8 \[[@R18]\] was selected as the model antibody. It was subsequently used to demonstrate the fundamental possibility of expressing the single-domain antibodies obtained by immunization of members of the Camelidae family via recombinant adenoviruses. EXPERIMENTAL ============ **Enzymes** In this study, restriction endodeoxyribonucleases, T4 DNA ligase, alkaline phosphatase (CIAP) purchased from Fermentas MBI (Lithuania), and Taq-polymerase purchased from Promega (United States) were used. **Cell lines** The HEK-293 cell line (human embryonic kidney cell culture transformed by the E1-region of human adenovirus serotype 5) and Н1299 cell line (human lung cancer cells) were used. The cells were cultured in a DMEM medium containing 10% of fetal bovine serum (FBS) purchased from HyClone (United States). **Production of the cDNA clone encoding the single-domain mini-antibody (nanoantibody) which specifically recognizes the endogenous mouse cytokeratin-8** Antibody aCyK-V ~H~ H, which specifically recognizes mouse cytokeratin-8, was obtained earlier by S.V. Tillib's research group ( Institute of Gene Biology, Moscow) in collaboration with the laboratory headed by S. Muyldermans (Vrije Universiteit Brussel) and used (via binding to the fluorescent protein sequence) to obtain fluorescent nanoantibodies (or chromobodies) aimed at demonstrating the new method for tracing antigens in a living cell. It should be noted that the aCyK-V ~H~ H nanoantibody was one of the first antibodies to endogenous structural eukaryotic proteins. The first stage of its production comprised immunization of the Bactrian camel ( *Camelus bactrianus* ) with a protein extract from mouse soft tissue cells (predominantly from the liver). The subsequent selection procedure, based on the phage display method, was performed as described in the online supplement to the article \[[@R18]\]. The fundamental stage after selection of the most enriched antibody clones was the identification of the unknown antigen recognized by these nanoantibodies. The proteins from the nanoantibody-binding region upon Western blotting were additionally separated by electrophoresis to obtain individual products. Western blotting was then used to analyze the recognition of each product by a nanoantibody. The product recognized by a nanoantibody was identified using mass spectrometrical analysis of its trypsin hydrolysate. The resulting nanoantibody aCyK-V ~H~ H recognized cytokeratin-8, a fact attested to via the immunofluorescent staining of С2С12 (mouse myoblast cell line) with these antibodies, revealing the characteristic distribution of cytokeratin intermediate filaments in the cytoplasm. The nanoantibody aCyK-V ~H~ H produced in the bacterial periplasm was modified by binding an antigen-recognizing sequence of two additional small fragments, epitope of influenza virus haemogglutinine (HA-tag) and six histidine residues (His ~6~ -tag), in order to purify it and simplify its detection. **Obtaining recombinant adenovirus** Plasmids and the recombinant adenoviral vector were obtained using the gene of antibody to cytokeratin *aCyK-V ~H~ H* . The nucleotide sequence encoding the nanoantibody was obtained by chemical synthesis in "Evrogen" JSC. The AdEasy Adenoviral Vector System (Stratagene, United States) was used in order to construct the рAd-aCyK-V ~H~ H plasmid vector containing the genome of the recombinant adenovirus with E1 region deletion, and a transgene expression cassette incorporated instead of it via homologous recombination in *E. coli* cells. The recombinant adenovirus was obtained via transfection of HEK-293 cell lines with the рAd-aCyK-V ~H~ H plasmid construct linearized on the PacI site. Lipofectamine 2000 (Invitrogen, United States) was used for the transfection, according to the manufacturer's recommendations. The recombinant human adenovirus of serotype 5 with E1 region deletion and an incorporated transgene-free cassette expression (Ad-null) inserted instead of it was used as the control. To accumulate adenoviral preparations, an infected cell suspension (10 ^7^ PFU of the virus per Petri dish with a diameter of 15 cm) was coated to the HEK-293 cell monolayer with 50--70% confluence. The infected cell suspension was destroyed by three freeze-thaw cycles and clarified by centrifuging (2000 rpm, 10 min, +4°С). The titres of the specimens Ad5-aCyK-V ~H~ H and Ad-null (10 ^8^ PFU/ml) were determined by the plaque formation technique in the HEK-293 cell culture. **Infection of cells with 
a recombinant adenovirus** Approximately 10 ^6^ cells of the H1299 line were infected with recombinant adenoviruses. The cells were seeded to \~ 70% of the monolayer, cultivated for 24 h, and infected with the recombinant adenovirus (the multiplicity of infection being 100 PFU/cell) in a DMEM medium containing 2% of FBS. Two hours after the viral preparation was introduced, the medium was collected, the cell culture was washed, and a fresh DMEM medium was added. The medium from the infected cells was collected 72 h after infection and concentrated by centrifuge ultrafiltration through a membrane with a nominally intercepted molecular weight of 10 kDa. After thickening by a factor of 10, the supernatant was fractioned in a 10% polyamide gel and used for immune blotting analysis. **Antigen preparation** Homogenized mouse liver lysate (BALB/c line) was obtained via extraction with the use of a RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, protease inhibitor kit (Roche, Switzerland)). The concentration of the total protein in the specimens was determined by the Bradford method (Sigma-Aldrich, United States). The specimens with an equal protein concentration were applied to the gel to be separated by electrophoresis. **Polyacrylamide gel electrophoresis and immunoblotting** Cellular proteins were separated by polyacrylamide gel electrophoresis by the Laemmli procedure under denaturing conditions in the presence of sodium dodecyl sulphate. Protein Test Mixture 4 (Serva, Germany) was used as the molecular weight standard. After the gel electrophoresis, the proteins were placed onto a Hybond-P PVDF membrane (GE Healthcare, United States) using a TE70 Semi-Dry Transfer Unit (Hoefer Scientific, United States) in accordance with the manufacturer's recommendations. The nanoantibodies were detected using the Monoclonal Anti-HA--Peroxidase antibody (Sigma-Aldrich, United States). The immobilized proteins were detected using ECL Plus Western Blotting Detection Reagents (GE Healthcare, United States) in accordance with the manufacturer's recommendations. The chemilumenscent radiation was recorded with the aid of an Amersham Hyperfilm ECL X-ray film (GE Healthcare, United States). RESULTS ======= As a result of the earlier performed selection of the phage library of the antigen-binding domains of single-stranded antibodies, DNA from the pHEN4 phagemid with an insertion encoding the nanoantibody, which has a high affinity towards the structural cytoplasmatic mouse protein, cytokeratin-8, was collected. The data on the structure of the target protein were obtained via mass spectrometric identification. The nucleotide sequence encoding the nanoantibody was cloned in the recombinant adenoviral vector. {#F1} A leading peptide of the mouse immunoglobulin κ-chain was bound to its N-terminus, in order to ensure efficient extracellular expression of a nanoantibody. The HA-tag, which is effectively recognized by commercial antibodies, a requirement for confirming nanoantigen expression at the protein level, was bound to the C-terminus of the nanoantibody. *[Fig. 1](#F1){ref-type="fig"}* shows the scheme of the resulting construct. In order to construct the adenoviral vector, the sequence encoding the HA-tag-labelled аCyK-V ~H~ H antibody was cloned in the shuttle plasmid vector pShuttle-CMV (Stratagene, United States). This vector contains terminal fragments of the human adenovirus serotype 5 genome, the expression cassette containing the human cytomegalovirus promoter (CMV) and polyadenylation signal. The presence of the insertion and its orientation were confirmed by restriction mapping. The recombinant plasmid adenoviral vector bearing the target gene was obtained via homologous recombination in *E. coli* cells. The plasmid construct obtained contained the replication initiation site ori, the gene of antibody resistance, and a cassette with the target gene within the adenoviral genome. The main advantage of this method is the potential utilization of *E. coli* cells as the main tools for cloning, recombination, and production of adenoviral DNA in preparative amounts. The opportunity to perform the homologous recombination in *E. coli* cells makes it possible to deal with the individual clones containing plasmid constructs only with recombinant adenoviruses, which eliminates the possibility of contamination with a wild-type adenovirus. The shuttle plasmid construct bearing the expression cassettes with the nanoantibody gene was linearized on the PmeI site and introduced along with pAd-EASY (Stratagene) to *E. coli* BJ5183 cells by electroporation. Recombinant clones obtained by homologous recombination were collected on the selective kanamycin-containing medium (50 µg/ml). The presence of recombinant clones of nucleotide sequences encoding the aCyK-V ~H~ H antibody and human adenovirus serotype 5 fibre in plasmid DNA was analyzed by PCR with specific primers and via restriction mapping, using HindIII restrictase, which enables one to obtain a restriction pattern that is typical for the human adenovirus genome. HEK-293 cells were transfected with a plasmid cleaved at the PacI site and containing the recombinant adenovirus genome with E1 region deletion and the expression cassette with a transgene inserted instead of it. The resulting recombinant adenovirus Ad5-aCyK-V ~H~ H was analyzed by PCR using the primer pair that was complementary to the target gene, the hexon gene of human adenovirus serotype 5, and the E1 region of the adenovirus in order to control the possible presence of replication-competent viral particles. **Detection of the expression of the 
nanoantibody gene within the recombinant adenovirus Ad5-aCyK-V ~H~ H** {#F2} The expression of the target gene within the recombinant human adenovirus serotype 5 Ad5-aCyK-V ~H~ H was analyzed at the level of the mRNA. With this purpose in mind, the cells of the HEK-293 line that are permissive for human adenovirus serotype 5 were infected with the recombinant virus Ad5-aCyK-V ~H~ H. The total RNA of infected cells was used to produce DNA, which was analyzed by PCR with primers specific to the sequence of the nanoantibody gene to mouse cytokeratin 8, to viral DNA, and the constitutively expressed gene of glyceraldehyde-3-phosphate dehydrogenase (GAPDH). HEK-293 cell lines infected with Ad-null virus ( *[Fig. 2](#F2){ref-type="fig"}A* ) were used as the negative control. RT PCR was used to demonstrate that the recombinant adenoviral vector expresses mRNA of the nanoantibody gene to cytokeratin and can be used to analyze protein production. Nanoantibody expression at the translational level was analyzed in H1299 cells infected with the recombinant adenovirus carrying the gene of nanoantibody to aCyK-V ~H~ H tagged with HA-epitope of the influenza virus (Ad5-aCyK-V ~H~ H), and the recombinant adenovirus containing the transgene-free expression cassette (Ad-null). The presence of a nanoantibody in the culture medium containing the infected cells was measured by immunoblotting with antibodies to HA-epitope of the influenza virus conjugated with horseradish peroxidase ( *[Fig. 2](#F2){ref-type="fig"}B* ). **Biological activity** The specificity of a nanoantibody expressed by the adenoviral vector to cytokeratin was confirmed by means of comparison of the interaction between the antigene and the proteins of the cultural fluid from the cells infected with recombinant adenovirus, and the interaction between the antigene and the antibody purified from *E. coli* periplasm. The lysates of mouse liver and cerebrum cells were fractioned in a polyacrylamide gel, transferred to the PVDF membrane, which was incubated with the cultural medium of the cells infected with Ad5-aCyK-V ~H~ H. The expressed nanoantibody served as the primary antibody to the target protein (mouse cytokeratin-8, 55 kDa) detected in the total lysate. The membrane was simultaneously incubated with antibodies aCyK-V ~H~ H produced in *E. coli* periplasm. *[Fig. 3](#F3){ref-type="fig"}* shows the results of an electrophoresis of protein lysates in polyacrylamide gel and the data obtained by immunoblotting with nanoantibodies aCyK-V ~H~ H after development by secondary antibodies to the HA-epitope of the influenza virus conjugated with horseradish peroxidase. Immunoblotting results attest to the fact that the antibody expressed by the adenovirus has the same specificity as the antibody synthesized in *E. coli* periplasm, its gene being cloned in the recombinant adenovirus. DISCUSSION ========== At the time of writing, there were a number of technologies capable of producing mini-antibodies with a predetermined specificity. Only quite recently was it revealed that, in addition to the canonical antibodies, functionally active noncanonical single-stranded antibodies were produced in relatively large amounts in members of the Camelidae family. Therefore, it is now possible to obtain mini-antibodies on the basis of libraries containing the antigen-recognizing domains of single-stranded antibodies of immunized animals. Noncanonical antibodies consist of a dimer with a single shortened heavy immunoglobulin chain (containing no light chains). Single-domain mini-antibodies (nanoantibodies) are genetically engineered derivatives of the antigen-recognizing domains of these noncanonical antibodies. The selection of clones of a mini-antibody with the predetermined specificity from the library of sequences of the entire repertoire of antigene-recognizing domains of noncanonical antibodies obtained from immunized camel is based on the highly efficient procedure of functional selection of filamentous phage particles containing both an exposed mini-antibody on the surface, and the DNA encoding it within the phage particle (phage display). {#F3} Mini-antibodies produced by this technology are characterized by high stability, solubility, and low immunogenicity. Mini-antibodies can be produced (selected) to any antigens and any antigen epitopes, including conservative ones, which often cannot be produced using the conventional procedure. Since the encoding nucleotide sequence is known for each mini-antibody, it is possible to produce the corresponding protein in any of the known expression systems (prokaryotic and eukaryotic). It is economically viable to produce protein preparations of mini-antibodies in *E. coli* cells, yeast, or CHO cells. When injecting these preparations to experimental animals (or patients), their very short lifetime in the organism (less than 24 h) should be taken into account. The period of therapeutic action of preparations based on mini-antibodies can be increased using the vector systems, providing that the synthesis of the active agent takes place immediately in the infected cells of the organism. Recombinant adenoviruses are the optimal expression system for solving such problems. Their safety and efficiency has been proved in a number of clinical trials performed globally; the time needed to produce a target protein is approximately 20 days. The potential application of recombinant adenoviral vectors for the expression of the genes of the antigen-recognizing fragments of single-stranded antibodies obtained from Bactrian camel was studied in this work. It was demonstrated that expression of the nanoantibody gene using the adenoviral vector is possible. Transgene expression was confirmed at the level of the RNA transcript and protein product. The specific interaction of the nanoantibody secreted by eukaryotic cells with a target protein attests to the fact that its functional activity is retained. Further studies are necessary for a qualitative estimation of the efficiency of nanoantibody expression using a recombinant adenovirus. CONCLUSIONS =========== The delivery of the gene of a single-domain mini-antibody (nanoantibody) selected from the library containing sequences of the variable domains of specific single-stranded antibodies of immunized camel to eukaryotic cells using the recombinant adenoviral vector provides efficient expression and functioning of the nanoantibody. The results of this study can be used for the production of passive immunization agents for protection against pathogens, or for the design of new-generation immunobiological antitoxic preparations. HEK-293 : human embryonic kidney cell culture His~6~-tag - amino acid motif in proteins consisting of six histidines; НА-tag : epitope tag (YPYDVPDYA) derived from the haemagglutinin molecule PFU : plaque-forming unit PCR : polymerase chain reaction PT-PCR : reverse transcription polymerase chain reaction

Introduction ============ Starting with winter semester 2003/2004, the University of Cologne introduced the new degree course at the medical faculty called 4C (Competence-based Curriculum Concept Cologne) \[[@R1]\], \[[@R2]\], \[[@R3]\]. Apart from fulfilling the goals of the Medical Licensure Act (Approbationsordnung für Ärzte \[[@R4]\]), the specific profile of the 4C model is defined in the so-called General Teaching Principles (Leitbild Lehre) \[[@R5]\], according to which Cologne graduates: have the necessary knowledge and skills to identify important and common diseases and acutely life-threatening situations and to induce their treatment;show behavior and attitudes that are conducive to their acceptance by patients and medical staff and to improving the standing of physicians in society;are willing and capable to engage in independent and science-based CPD in general practice but also in a clinical discipline or a basic subject of their choice. These goals are primarily achieved through the following new elements in the curriculum \[[@R1]\], \[[@R2]\] which are combined with traditional subject and cross-subject teaching: interdisciplinary competence areas (88 in total, each with 5-12 hours of lectures on important and common topics of in- and out-patient treatment; competence areas 1-24 are for undergraduate students)patients treatment in parallel to the course ("Studipat") in which medical students are brought into contact with one patient each at a GP surgery for the first four years of studyorganization of week-long block placement from the 5th semester onwardsoffering electives at the end of each semester via compulsory blocks of electivesscientific qualification through two projects in which the students immerse themselves into a scientific subjectskills training in the KIS^S^ (Cologne Inter-Professional Skills Lab and Simulation Center). KIS^S^ is a central institution of the deanery of student affairs and serves to provide training and independent practice of practical medical skills in all outcome-areas as defined by Harden \[[@R6]\], \[[@R7]\]. In contrast to the traditional process-based separation of knowledge, skills and attitudes in the German-speaking world, outcome-based trainings starts by defining outcomes (elsewhere also vaguely termed "competence-based" \[[@R8]\], \[[@R9]\]. Three different levels for curriculum development are discussed (which build on each other). The first is "Doing the right thing" which refers to the knowledge level and the ability to handle the daily clinical routine. The second is "Doing the things right", the level of emotional, analytic and creative intelligence to use knowledge and skills appropriately. The third level focuses on the person and their future development - "The right one doing it", the professional role in its social and societal reality. Recently more and more curricula have been developed from this point of view and as a result, there are also calls for the adaptation to accreditation criteria from the English-speaking world \[[@R10]\]. This is somewhat surprising, especially in view of the thin to non-existent proofs of efficacy, in particular with regard to the 3rd level "Attitudes" and "Professionalism\" \[[@R11]\]. Based on these development criteria, and bearing in mind the requirements of the medical licensure act, clinical skills are grouped into three fields in Cologne: emergency skills, communication skills and technical skills. These elements transcend the entire degree course as a teaching and learning helix, with the educational objective of students reaching Clerkship Maturity after five semesters and the Internship Maturity ("PJ-Reife") after ten semesters. Table 1 [(Tab. 1)](#T1){ref-type="fig"} shows the teaching content of skills training taught up to Clerkship Maturity divided into semesters. In a subsequent formative OSCE (Objective Structured Clinical Examination, cf Harden & Gleeson \[[@R12]\]) with six stations during the 5th semester, students receive feedback on their practical skills on a German-English bilingual certificate, "Famulaturreife - Clerkship Maturity". The program running from the first semester of study to Clerkship Maturity was developed in 2003/2004 by an expert task force of the Advisory Board of Study Affairs of the Medical Faculty of the University of Cologne. Starting with the winter semester 2003/2004, and with broad participation from many different disciplines (anesthesia, psychosomatic medicine, transfusion medicine, internal medicine) and following the Delphi-model the "Leitbild Lehre" was developed \[[@R5]\]. To validate this procedure retrospectively, two questions were investigated: What skills are required by the physicians and GPs supervising students on clerkships?How frequently were students able to perform these skills as part of their clinical clerkship and which advantages did the Clerkship Maturity examination bring them from their point of view? Materials and methods ===================== To evaluate the opinions of the teaching staff, a semi-standardized, machine-readable questionnaire was designed. Some of the content is shown in Table 2 [(Tab. 2)](#T2){ref-type="fig"}^1^. The questions were primarily developed on the immediate basis of the earlier KIS^S^ planning. In addition, open questions were used for additional information. The questionnaire was tested using ten randomly selected students to ensure its comprehensibility and subsequently sent to the senior physicians of the Cologne Teaching Hospitals, the GP surgeries and the senior physicians at the University Clinic of Cologne by post. To evaluate the opinion of the students a semi-standardized, machine-readable questionnaire based on the above described questionnaire was designed to reflect the reality of education. In this context the primary concern was not an evaluation of the importance of the skills training content by the students but data on the frequency of skills used ("empirical relevance") as part of the clerkships (For Questionnaire Excerpts see Table 3 [(Tab. 3)](#T3){ref-type="fig"})^2^. The questionnaire was sent to every student of the target group (7^th^semester). Both questionnaires looked at the teaching/learning on offer in the KISs skills lab and the skills actually required during the clerkships. In addition, the form collected socio-demographic data regarding the age, sex, year of study, number of completed weeks on clerkship and additional medical training that may have been gained. The completed questionnaires were digitized using Remark^®^. The data was evaluated statistically using Microsoft Excel^®^ 2003. Mean values and standard deviation are also shown. The graphics were generated using Microsoft Excel^®^ 2003. The open questions were analyzed using qualitative, structured content analysis after Mayring \[[@R13]\]. Results ======= **Results of the teaching staff questionnaire:** 36 of 80 (45%) hospital-based physicians and 77 of 140 (55%) GPs returned completed questionnaires. There was no follow-up due to anonymisation. Figure 1 [(Fig. 1)](#F1){ref-type="fig"} shows the evaluation of the necessity/importance of different skills for Clerkship Maturity as viewed by the teaching staff. While frequently needed skills such as disinfecting hands or taking blood pressure and pulse are rated quite high, more rarely needed skills or skills deemed as premature for this stage were rated low (such as placing a stomach tube or a CVL). Significant differences between the hospital-based physicians and GPs were only seen with a few items such as taking blood samples, iv/im injections and suturing techniques (see Figure 2 [(Fig. 2)](#F2){ref-type="fig"}). It became clear that skills desirable for the ward round where rated more highly by hospital-based physicians significantly more frequently. Interpersonal skills were rated differently by hospital-based physicians and GPs too. While empathy with 64% was the most frequently mentioned criterion by GPs, only 36% of hospital-based physicians rated this skill as being important. With 20%, both groups equally mention the "ability to listen/handle patients" or "interest in the patient" as an important skill for students. Additional suggestions for skills that should be included in the KIS^S^ offer regarding Clerkship Maturity are only given sporadically (e.g. computer skills, wound management or team skills) and thus do not allow a consistent overview. **Results of the student questionnaire:**The questionnaire was sent to 140 students of which 75 were returned completed, a return rate of (54%). 24 of the 75 students who responded were male, 45 female (6 did not specify). The mean age of the students was 26 years. 59 students were attending the 7^th^ semester, 7 the 8th semester and 3 higher semesters. 7 students listed a qualification in the medical field, including 3 paramedics (most frequent mention). Only 5 students who had returned their questionnaires were studying under the regular degree course. For this reason, there is no comparison between the regular and the model degree course responses. Figure 1 [(Fig. 1)](#F1){ref-type="fig"} contrasts the student responses regarding the frequency of skills needed during the clerkship and their importance as rated by the teaching staff. It is clear that the frequency and necessity (importance) of some skills diverge rather drastically (especially in emergency care). While the teaching staff rate training in examination-skills as very important (1.4±0.9) whereas the students rate this as less important (2.9±1.7). Students suggested the following in the open questions regarding possible improvements of the skills training prior to Clerkship Maturity: better knowledge transfer (9 mentions)more opportunities for practicing (5 mentions)practicing with real patients (3 mentions)ability to practicing rather than watching the lecturer (3 mentions)smaller groups (2 mentions) Although skills in taking patient histories are rated as relevant and frequently needed, as can be seen in Figure 1 [(Fig. 1)](#F1){ref-type="fig"}, (note that 16% of the students stated they did not have the opportunity to take medical histories in their clerkship), students rate the learning effect of the course on taking histories and communication rather moderately (2.9±1.7 to 3.1±1.4, cf question 7b in Table 3 [(Tab. 3)](#T3){ref-type="fig"}). This is consistent with requests by the teaching staff for more training in taking patient histories (1.7±1.1, cf Question 10 in Table 2 [(Tab. 2)](#T2){ref-type="fig"}). The students also assessed the certificate they received following the OSCE examination during the 5th semester. The feedback on their practical skills provided by the certificate is rated rather modestly (3.4± 1.7 on a 6-point Likert-scale, 1=very helpful and 6=not helpful at all). The certificate was seen as even less helpful when it comes to applying for clerkship placements (4.7±1.8). This corresponds to statements made by the physicians of which only 6% stated to be familiar with the certificate. Discussion ========== The aim of this study was to validate the planning of training on practical skills as designed by the experts for the first 5 semesters of study in the new degree course in human medicine at the University of Cologne empirically. As a validation of training goals, both teaching staff and students were questioned and the data of these two groups was compared. The students were also asked for their subjective experience gain through the KIS^S^ training. Generally speaking, it was possible to confirm the validity of the Clerkship Maturity goal based on the data collected from the teaching staff and the students. Skills explicitly taught in the Skills Lab for achieving Clerkship Maturity are usually indeed frequently needed in clinical clerkships. Even the handling of drips which is not explicitly part of the skills training and which is taught quasi in passing through the teaching of placing winged infusion sets. Interestingly, students are confronted with the topics ultrasound and ECG rather more frequently than suspected. It might be worthwhile considering if the program needs to be amended to reflect demand in the clerkships. On the other hand, this content is taught later on in the clinical section in more detail. This also corresponds to the opinion of the physicians who rate neither as very important skills for the clerkships (ultrasound even less so than ECGs). These differences that were discovered between skills rated as necessary by the teaching staff on the one hand and the experienced frequency of emergency care components needed during the clerkships as reported by the students on the other hand can be easily explained. These skills extremely important to have in some situations (in some cases life-saving) but in reality these are rarely required of students on clerkships. There is a remarkable difference concerning the divergence between the frequency of applying skills for clinical examination as stated by the students and the perceived necessity as rated by the teaching staff. The teaching physicians rate this item in place 2, whereas the students rank this skill in place 5. The method of comparing the frequency of a clinical skill used by students on the one hand with a ranking of the importance by teaching staff on the other could be seen as problematic. But it is not unlikely that at this stage students (in the first or second clinical semester) only possess a limited ability to evaluate the necessity of a skill but are quite capable of reporting the frequency based on their own experiences. Allowing the frequency of skills used during clerkships to be rated by the teaching physicians is also questionable as such an evaluation might not be valid due to the workload and the fact that students are supervised by different members of staff. However, all physicians are able to rate the necessity of certain skills for the clinical activities of students on clerkships. The differences between the rated necessity of the skills in "taking blood samples", "iv/im injections" and "suture techniques" between hospital-based staff and GPs is not surprising, given the different fields of activity. Nevertheless, it must be borne in mind that the practical training has to prepare students both for clerkships at a hospital and in GP's surgery. Although the use of the majority of skills taught in training was confirmed by the questionnaire, a small number of skills at the bottom end of the ranking suggest room for re-organization. For example catheterizing the bladder, which seems to have a rather low relevance in the clinical routine of a student on clerkship (ranked 15th of 18 items of all skills, see Figure 1 [(Fig. 1)](#F1){ref-type="fig"}). The focus of this unit has already been changed to "working with sterile surfaces and material" using catheterizing as an example. It is interesting that the category "empathy" is given twice as frequently in the open questions when asked about the soft skills students should have with hospital-based doctors citing this item twice as often as GPs. A possible explanation of this is the fact that the qualitative content analysis summed up related words like "humbleness" (in front of the patient) under the umbrella term "empathy". In any case this should give food for thought further down the line, in particular regarding the teaching of "attitudes," as they are referred to in the General Teaching Principles (see above), corresponding to the third level of the outcome-based learning (professionalism \[[@R6]\]); because empathy is seen as a prerequisite skill for a successful interaction between the doctor and the patient. Although it was shown in the past that, as a first step to a conscious doctor-patient interaction, the "First Semester Tutorial on Medical Psychology" (now called "Introduction to Bio-psycho-social Medicine" as an introduction to teaching communicative competences) that the thinking of students could be directed towards a multidimensional understanding of health and disease \[[@R14]\], \[[@R15]\], \[[@R16]\], \[[@R17]\]. However, this does not correspond to a complex modulation of attitudes^3^ as required by the General Teaching Principles \[[@R5]\] (see also \[[@R6]\]). So the comments made by the physicians in the open answers can be interpreted as a sign of gaps in the efforts to date to teach adequate doctor-patient communication. In this context the evaluation of the model degree course by the students that they had not received enough training in doctor-patient communication can also allow the conclusion that being familiar with a procedure and its risks leads to more reflective actions. Especially because the quality of the self-evaluation given by the students is questionable as a valid source, more research is needed in this area. It should not be left unmentioned that there is a possible clue that more intensive practical training can (initially) lead to unexpected uncertainty. Because a structured curriculum for skills training in Germany is not part of standard education (and not currently required by the ÄAppO), it is not surprising that the answers to the question on the usefulness of the OSCE certificate at the end of the training phase in the 5th semester is subjectively not seems as being advantageous by the students. Based on our data we cannot clarify if this is simply down to the fact that this certificate is not sufficiently well known in hospitals and surgeries to date or if students do not even mention it in their applications. Interestingly, isolated reports of experiences by students applying for international clerkships present a different picture. Even though there is room for improvement regarding the perceived bonus of the skills training during the pre-clinical phase with the aim of Clerkship Maturity (the usefulness of the OSCE feedback could, if necessary, be increased by skills area relevant evaluations such as hygiene skills, fine motor skills, etc.), the construct itself seems valid and reliable. The success of this initiative overall and a positive view by the students can only come into its forte through the national debate on education and the nascent National Competence-based Catalog of Learning Targets \[[@R18]\]. Only time will tell if the concept of Clerkship Maturity (and the Internship Maturity and Professional Development Maturity based upon it) can be transferred. The development of medical skills training which is adapted to local needs is still in its infancy in the German-speaking countries and must be further developed in many ways. The idea of Clerkship Maturity can at least provide a conceptual framework for this. Notes ===== ^1^ In addition the questionnaire for the teaching staff included questions about the general trainability of medical skills and differences in the performance of students from the regular degree course (8^th^ ÄAppO) and the model degree course (9^th^ ÄAppO). These questions are not discussed in this article as it apparently was not possible for the teaching staff to differentiate between students of the regular and the model degree course. ^2^ The student questionnaire also asked for an evaluation of their own competences in comparison to students of the regular degree course (students following the old curriculum i.e. 8^th^ ÄAppO). The results are not discussed in this article due to the low response rate. ^3^ See the training goals developed in the Study Guidelines at the Medical Faculty of the University of Cologne, based on the General Teaching Principles (§1 Sect 3 „Attitudes") \[[@R5]\]: Readiness to apply medical ethical principles in practice and researchRespect and honesty towards patients and colleaguesRealistic estimation of one's own skills, abilities and limits and readiness, to draw appropriate conclusions from this;Readiness to take on responsibility and for accuracy. Acknowledgement =============== Special thanks to Mrs Imke Wietoska, Marlen E. Sauer and Charlotte M. Schober who contributed to the creation of the questionnaire databases as part of their scientific project. I would like to sincerely thank PD Dr. Jan Matthes for his support in bringing this Master Thesis to fruition as part of the MME-D. {#T1} {#T2} {#T3} {#F1} {#F2}

**Core tip:** This article reviews the current role of diffusion weighted imaging for various oncological and non-oncological applications in the liver. INTRODUCTION ============ Diffusion-weighted imaging (DWI) is a functional imaging technique, allowing qualitative and quantitative assessment of the diffusion properties of various types of tissues\[[@B1],[@B2]\]. Numerous studies over the past decade have validated the role of DWI in oncologic and non-oncologic applications in the body\[[@B1],[@B3]-[@B6]\]. Multiphase contrast enhanced MRI is an established technique for evaluation of a wide spectrum of liver diseases including focal lesions and diffuse parenchymal abnormalities. DWI compliments routine MRI of the liver by providing both qualitative and quantitative assessment for both focal and diffuse hepatic parenchymal processes. Factors such as the ease of acquisition and ability to obtain functional information in the absence of intravenous contrast, especially in patients with abnormal renal function, have contributed to the growing interest in exploring clinical applications of DWI. DWI improves sensitivity in detection of focal lesions, helps differentiate benign from malignant focal hepatic lesions, and also permits evaluation of treatment response to systemic and loco-regional therapies in primary and secondary hepatic malignancies. This review article focused on the basic principles, technique, current clinical applications and recent updates in DWI of the liver. DWI: BASIC PRINCIPLES AND TECHNIQUE =================================== DWI exploits the regional differences in the motion of water molecules within the extracellular/extravascular compartment of tissues. In highly cellular tissues (*e.g*., lymphoma, carcinoma and abscess), the compact nature of the extracellular space causes increased impediment to motion of water molecules and the resultant water diffusion in such tissues is said to be "restricted". On the contrary, in tissues that are necrotic or fluid filled (*e.g*., cysts), there is unrestricted motion of water molecules and water diffusion in such tissues, which is said to be "free". Therefore, the diffusion properties in different tissues provide information on tissue cellularity and the integrity of cellular membranes\[[@B1],[@B2]\]. DWI is basically a modified T2 weighted sequence where the signal intensity depicts the tissue diffusion characteristics. Single-shot spin-echo (SE) echo-planar technique is the most commonly utilized technique to acquire DW-MRI in combination with fat suppression\[[@B7]\]. To obviate the effect of motion, it can be acquired either using breath-hold or free breathing sequences with multiple signal acquisitions (in combination with respiratory and/or cardiac triggering). Free breathing sequences provide improved signal to noise ratios (SNR), thinner image sections, and higher number of b-values obtainable compared to breath-hold sequences. However, these take longer time (3-6 min) to acquire than breath hold sequences to evaluate the liver compared to free breathing EPI which takes (40-60 s)\[[@B8]\]. The free breathing technique has been shown to have better reproducibility of ADC values than other acquisition techniques like breath-hold, respiratory-triggered (RT), and navigator-triggered DWI\[[@B9],[@B10]\]. Although cardiac motion also impacts quantitative ADC measurements, cardiac triggering is not routinely used in clinical practice\[[@B11]\]. Intravoxel incoherent motion (IVIM) imaging is a technique that has been introduced to quantitatively study the effects of tissue perfusion on the signal acquired with DWI and it resolves DWI measurements into true molecular-based (*D*) and perfusion-related (*D\**, *f*) diffusion\[[@B12]\]. In patients with renal failure, gadolinium is contraindicated due to risk for developing nephrogenic systemic fibrosis (NSF)\[[@B13]\]. These patients also have a risk of worsening renal failure with iodinated CT contrast. MRI without contrast is a reasonable option for these patients but non-contrast protocols do not have a diagnostic accuracy comparable to multi-phase contrast MRI. DWI does not require administration of intravenous contrast, and because of its performance in oncological applications in general, it has generated much interest recently. The diagnostic performance of DWI has been tested in metastatic liver disease and HCC, and the results were comparable to contrast MRI\[[@B14]-[@B16]\]. CLINICAL APPLICATIONS IN LIVER ============================== Imaging of focal liver lesions ------------------------------ **Lesion detection:** Multiphase contrast enhanced-MRI is currently the state-of-the-art imaging method for liver lesion detection and characterization. DWI at high *b*-values (≥ b100) provides a low background signal from normal liver parenchyma and thereby results in increased contrast between the background liver and lesions, enhancing the detection of focal liver lesions\[[@B17]\]. DWI is especially useful in detection of small lesions around vessels and in the periphery of liver which can be challenging to detect on routine T2 weighted images\[[@B18],[@B19]\]. The DW-MRI can be particularly valuable in oncologic patients with compromised renal function who cannot get intravenous gadolinium based contrast agents\[[@B14]-[@B16]\]. DWI adds value in oncologic patients (Table [1](#T1){ref-type="table"})\[[@B15],[@B20]-[@B22]\] by depicting more metastatic liver lesions when combined with multiphase contrast enhanced-MRI protocols, and improves reader confidence in lesion detection\[[@B22]-[@B25]\]. DW-MRI alone is less sensitive than gadoxetic acid-enhanced MRI for detecting liver metastases, but increases the sensitivity of detection for liver metastases (90.6%-95.5%) when combined with multiphase contrast enhanced MRI\[[@B25]\]. A major impact has been noted in the detection of metastases measuring ≤ 10 mm\[[@B17],[@B22],[@B24]-[@B27]\] (Figure [1](#F1){ref-type="fig"}). DWI has been used in detection of metastatic liver lesions from colorectal, pancreatic and neuroendocrine primaries\[[@B25],[@B28],[@B29]\]. ###### Comparison of SSEPI diffusion-weighted magnetic resonance imaging *vs* conventional magnetic resonance sequences for detection of hepatic metastases\[[@B15],[@B20]-[@B22],[@B27]\] **Ref**. ***b* value** **(s/mm^2^)** **Compared with (Seq)** **Sensitivity of DWI *vs* other sequences** **Accuracy of DWI *vs* other sequences** **Advantages of DWI** ------------------------------------------------ ------------------------------- ------------------------------------------------------------------------------------- --------------------------------------------- --------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Bruegel et al\[[@B27]\] 50, 300, 600 5 different T2-TSE (Turbo Spin Echo) sequences 0.88-0.91 compared to 0.45-0.62 0.91-0.92 compared to 0.47-0.67 Better sensitivity and accuracy Zech et al\[[@B21]\] 50 Fat suppressed T2WI 83% *vs* 61% \- Better image quality Fewer artifacts Better sensitivity Hardie et al\[[@B15]\] 0, 50, 500 Gadolinium enhanced T1WI 66.3% *vs* 73.5% 88.2% and 88.2% for DW-MRI, 90.2% and 92.2% for CE MRI, respectively, for observers 1 and 2 Not significantly different Donati et al\[[@B20]\] 0, 150, 500 Combined (Gd-EOB-DTPA) enhanced MRI/DWI *vs* Gd-EOB-DTPA enhanced MRI and DWI alone \- Gd- EOB-DTPA/DWI: 0.84 and 0.83 *vs* 0.73 and 0.72 for DWI alone Increase in diagnostic confidence No significant increase in diagnostic accuracy Colagranade et al\[[@B22]\] 0-500 Added value of DWI for lesion detection in unenhanced and Gd-EOB-DTPA enhanced MRI -62.5% for unenhanced MRI w/o DWI -81.1% for unenhanced MRI w/o DWI DWI improved all statistical parameters in the unenhanced examinations, as for nodules either smaller or greater than 1 cm. In EOB-enhanced examinations DWI increased specificity/negative predictive value -85.0% for unenhanced MRI+ DWI -89% for unenhanced MRI + DWI -95.6% for CE MRI -92.9% for CEMRI -97.3% for CE MRI + DWI -95.5% for CE MRI + DWI DWI: Diffusion-weighted imaging; MRI: Magnetic resonance imaging. {#F1} DWI has also been found to be useful in detection of primary hepatic malignancies such as hepatocellular carcinoma (HCC) and cholangiocarcinoma both in cirrhotic and non-cirrhotic livers (Figure [2](#F2){ref-type="fig"}). A combination of DW hyper-intensity and arterial hyper-enhancement results in increased sensitivity for diagnosis of HCC as compared to traditional criteria, particularly for small HCC \< 20 mm\[[@B30],[@B31]\]. {#F2} A low cost abbreviated MRI (AMRI) protocol for HCC screening and surveillance has been proposed based on a simulation study using DWI and T1-weighted imaging obtained at the hepatobiliary phase (HBP) after gadoxetic acid injection\[[@B32]\]. The AMRI shows sensitivity and negative predictive values of 80.6% and 80% (for DWI + T1W HBP) compared to 90.3% and 94.9% for a full dynamic contrast enhanced data-set\[[@B32]\]. **Lesion characterization:** Several studies have attempted characterization of liver lesions using DW-MRI\[[@B33]-[@B38]\]. A general assumption is that ADC values are higher in benign lesions and lower in malignant liver lesions\[[@B33]-[@B36]\]. In fact, studies have found statistically significant difference in ADC values between benign and malignant liver lesions\[[@B3]\]. Different studies have reported variable success using various ADC cut-off values with high variability likely due to the difference in scanners and parameters used to obtain DW-MRI and ADC maps\[[@B39]-[@B43]\]. Moreover, there is a high degree of overlap between solid benign and malignant lesions\[[@B44],[@B45]\]. Hence, the use of absolute ADC values or ADC value cut-off for characterization of focal hepatic lesions should be avoided and DWI should always be interpreted as a complimentary technique to conventional MR sequences\[[@B42],[@B46],[@B47]\]. It is also important to note that solid benign lesions such as hemangioma, FNH and hepatocellular adenoma can also show diffusion restriction compared to normal liver parenchyma. ADC values for these lesions are intermediate, generally greater than solid malignant lesions but with a significant degree of overlap\[[@B44],[@B45]\]. Hepatic abscesses show lower ADC values than solid malignant lesions, and restriction pattern may be different from malignant lesions\[[@B42]\] (Table [2](#T2){ref-type="table"}). ###### Liver lesion characterization based on ADC values\[[@B33],[@B35],[@B44],[@B45],[@B102]\] **Ref**. **Lesion type** **Mean ADC (10^-3^mm^2^/s)** **Sample size** ***b*-values** **Conclusion** -------------------------------- ----------------- ------------------------------ ----------------- -------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Parsai et al\[[@B44]\] Cyst 2.66 2 100, 200, 500, 750, and 1000 mm^2^/s ADC cutoff value threshold of 1.6 × 10^-3^ mm^2^/s yielded higher accuracy for differentiating benign from malignant lesions. DWI is not reliable to differentiate malignant from benign solid lesions HCC 1.07 26 Metastases 1.04 39 Taouli et al\[[@B98]\] Cyst 3.63 52 0, 500 Threshold ADC value of 1.5 × 10^-3^ mm^2^/s to differentiate between benign and malignant lesions, but with a significant overlap between benign hepatocellular lesions and HCCs HCC 1.33 Metastases 0.94 Parikh et al\[[@B35]\] Cyst 2.54 211 0, 50, 500 Accuracy of 75.3% for differentiating benign from malignant, by using a threshold ADC of less than 1.60 × 10^-3^mm^2^/s . Equivalent performance of DW imaging and T2- weighted imaging for lesion characterization HCC 1.31 Metastases 1.5 Bruegel et al\[[@B33]\] Cyst 3.02 204 50, 300, 600 88% of lesions were correctly classified as benign or malignant using a threshold value of 1.63 × 10^-3^ mm^2^/s. Measurements of the ADCs of focal liver lesions on the basis of a respiratory triggered DW-SS-EPI sequence may constitute a useful supplementary method for lesion characterization HCC 1.05 Metastases 1.22 Gourtsoyianni et al\[[@B102]\] Cyst 2.55 37 0, 50, 500, 1000 Sensitivity and specificity of 100% for differentiating benign from malignant lesions using a cutoff ADC value of 1.47 × 10^-3^ mm^2^/s HCC 1.38 Metastases 0.99 HCC: Hepatocellular carcinoma; DWI: Diffusion-weighted imaging. DWI has also been used to assist in differentiation of cirrhotic hepatocellular nodules\[[@B48]\]. Lesion hyperintensity on DWI, especially in association with hypointensity on delayed hepatocellular phase images, and low lesion-to-liver ratios should raise the suspicion of HCC or high-grade dysplastic nodules\[[@B49]\]. The HCCs have a tendency for angio-invasion and can present with filling defects in the portal or hepatic veins. Angio-invasion carries a high risk of distant metastasis and recurrence after transplantation. HCC invasion into the portal vein is considered as a contraindication for liver transplantation. It is important to distinguish tumor thrombus from a bland thrombus that is also common in chronic portal hypertension and has different clinical implications. In patients with locally advanced HCC, DW-MRI has been shown to be useful in characterization of the venous thrombus as bland *vs* tumor thrombus\[[@B50]\]. The mean ADC ratio of tumor thrombus and HCC has been reported to be \< 2 (0.998) as compared to bland thrombus (2.9)\[[@B50]\]. **Tumor grade and prognostication:** Recently, there have been attempts to predict the histopathological grades of HCC using DWI. ADC values have been found to correlate with histopathological differentiation and microvascular invasion with poorly differentiated HCCs showing significantly lower ADC than well-differentiated and moderately differentiated HCCs\[[@B51]-[@B54]\]. A cut-off value of 1.175 × 10^-3^ mm^2^/s has been recommended as a predictor of microvascular invasion\[[@B52]\]. Additionally, the recurrence-free survival has been found to be significantly shorter in low-ADC group than in high-ADC group\[[@B52]\]. The association of ADC and histopathological grades has shown conflicting results in few other studies\[[@B55],[@B56]\]. This might be a result of tumor necrosis, as it can result in reduced cellularity and increased ADC in high-grade lesions. Higher signal intensity on DWI has been reported to be associated with higher pathological grades despite insignificant correlation with ADC values\[[@B54],[@B56]\]. Diffuse liver diseases ---------------------- **Evaluation of NAFLD:** Non-alcoholic fatty liver disease (NAFLD) is the most common liver disorder in western industrialized countries with a prevalence of 6%-35% worldwide\[[@B57]\]. The severe form of this disease is steatohepatitis which can progress to cirrhosis in 15% of the patients\[[@B58]\]. Currently, the diagnosis of NAFLD is established based on histopathological evaluation of liver biopsy specimens. Liver biopsy is invasive and has risks of complications and sampling error, and cannot be frequently repeated. The feasibility of DWI and IVIM was first tested in animal models with early results showing that the IVIM diffusion parameters, in particular the "*f"* values, might be potential biomarkers of NAFLD\[[@B59]\]. The correlation between histologic features of NAFLD and quantitative measures derived from IVIM-DWI was later tested in humans which showed that the true molecular diffusion was significantly decreased with steatosis\[[@B60],[@B61]\]. ADC was not found to be associated with any histological feature\[[@B60]\]. Although these early results are promising, standardization of acquisition and post-processing techniques of IVIM DW-MRI is needed. **Evaluation of liver fibrosis and cirrhosis:** Aubé et al\[[@B62]\] reported early benefits of DWI in the evaluation of diffuse liver diseases, particularly in the detection and quantification of hepatic fibrosis. Several authors thereafter have tried to find a simple, reliable and non-invasive method to detect and monitor hepatic fibrosis, thereby avoiding the existing gold standard involving liver biopsy and its complications\[[@B63],[@B64]\]. A recent meta-analysis suggests that DWI and IVIM parameters can reliably stage hepatic fibrosis\[[@B65],[@B66]\]. However, IVIM measurements and ADC values have been reported to be influenced by presence of fat or iron within the liver that can impact their accuracy for staging of fibrosis\[[@B67]-[@B69]\] and ascites\[[@B70]\]. Recent studies comparing MR elastography (MRE) and DWI in characterizing hepatic fibrosis demonstrate higher predictive ability of MRE in distinguishing stages of fibrosis compared to DWI\[[@B71],[@B72]\]. Gadoxetate disodium enhanced liver MRI is also more strongly correlated with fibrosis stage as compared to DWI\[[@B73],[@B74]\]. Considering the conflicting evidence, it can be concluded that at present, DWI cannot replace liver biopsy in liver fibrosis. Further investigations and analysis are needed to increase the reliability of the technique. Monitoring treatment response ----------------------------- There has been a lot of interest in using DWI as an imaging biomarker for monitoring treatment response to various locoregional and systemic therapies in hepatic malignancies (Table [3](#T3){ref-type="table"})\[[@B75]-[@B79]\]. In comparison to conventional morphological methods of monitoring response such as RECIST and WHO which rely on changes in tumor dimensions for quantitating tumor response, DW-MRI allows evaluation of treatment response to novel targeted therapies which cause early changes in tumor physiology prior to change in tumor size. The increase in post-treatment ADC values precedes a decrease in size of tumor which has been the traditional method of measurement for post-treatment response, especially in systemic therapy\[[@B80]-[@B82]\]. ###### Role of diffusion-weighted magnetic resonance in assessment of treatment response\[[@B75]-[@B79]\] **Ref**. **Treatment modality** **Tumor type** **DW-MR parameter evaluated** **Study results/teaching point** -------------------------- ------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Chapiro et al\[[@B79]\] TACE HCC (3D) quantitative enhancement-based and DW volumetric MR High accuracy and intermethod agreement of 3D quantitative techniques in the assessment of tumor necrosis after TACE is clinically relevant High diagnostic performance of qEASL criteria and qADC may help in triaging patients for repeat treatment after a TACE session Mannelli et al\[[@B87]\] TACE HCC ADC measured with DWI in treatment response Pre TACE ADC obtained at 0, 50, 500 s/mm^2^ *b* values before and after treatment may be used to predict HCC response to TACE Park et al\[[@B42]\] Radiotherapy HCC DW MR *vs* conventional MR for treatment response Improved detection of viable tumor when DW MR is added to conventional sequences Yu et al\[[@B76]\] Radiation therapy HCC DW MR Change in ADC value before and after RT is related to local progression free survival. Hence ADC value and RECIST may substitute for mRECIST in patients who cannot receive contrast agents Schraml et al\[[@B77]\] Radiofrequency *n* = 16 HCC, 1 = cholangiocarcinoma, and 37 = metastases (28 colorectal cancer, 3 melanoma, 3 breast cancer, 1 pancreatic cancer, 1 gastric cancer, esophageal cancer) DW MR and mean ADC values ADC-based evaluation of signal alterations adjacent to the ablation zone may contribute to the identification of local tumor progression and nontumoral post- treatment tissue changes Ablation HCC: Hepatocellular carcinoma; DW MR: Diffusion-weighted magnetic resonance; TACE: Trans-arterial chemoembolization. **Percutaneous ablation:** ADC-based evaluation of signal alterations adjacent to the ablation zone may contribute to the identification of local tumor progression and non-tumoral post-treatment tissue changes after radiofrequency ablation of hepatic primary tumors and metastases\[[@B77]\]. Early post-ablation zone may show heterogeneous signal on non-enhanced T1 and T2 weighted images due to edema, hemorrhage and inflammatory reaction. These changes resolve within 4-6 mo after ablation leaving behind a characteristic homogenous high T1 signal and low T2 signal (coagulation necrosis). Nodular enhancing foci within the ablation zone are considered as a sign of local recurrence. Low ADC values at 1 mo (\< 1.145 × 10^-3^ mm^2^/s) after RFA have been shown to be associated with an early local recurrence of HCC\[[@B83]\]. **Intra-arterial therapies**: The utility of DWI has been assessed in treatment response after trans-arterial chemoembolization (TACE) of HCC\[[@B84]-[@B87]\]. DWI has been shown to perform equally\[[@B78]\] or better than gadolinium-enhanced MRI in quantifying the area of tumor necrosis after chemoembolization\[[@B78],[@B86],[@B88]\]. Increased ADC values in non-enhancing tumors show a high correlation to the degree of tumor necrosis at pathology\[[@B86],[@B88]\]. Mannelli et al\[[@B78]\] showed excellent performance of ADC for prediction of complete tumor necrosis after TACE (sensitivity of 75% and specificity of 88%) which was comparable to 100% sensitivity, and 58%-79% specificity for contrast-enhanced MRI. Transarterial radioembolization (TARE) using yttrium-90 (^90^Y)-loaded resin microspheres is a treatment option for various liver malignancies (including liver-dominant breast metastases). Early arterial blood flow stasis with consecutive incomplete dose administration may occur in 12%-25% of resin-based radioembolization procedures. The perfusion-sensitive IVIM parameter "*f*" may predict early blood flow stasis in patients undergoing TARE for liver-dominant breast metastases\[[@B89]\]. **Image-guided radiation therapy:** Image-guided targeted external beam radiation therapy is emerging as an alternative option in the treatment of advanced unresectable HCC. Accurate post-radiation response assessment can be challenging due to the concomitant changes occuring in the radiation zone. MRI is the preferred modality for response assessment. Inclusion of DWI in the imaging protocol has been shown to significantly enhance the diagnostic accuracy (91%-97% *vs* 72%) for detection of viable tumors after radiation treatment with improved sensitivity, specificity, and negative predictive value as compared to routine MR sequences (90%-97%, 91%-97% and 91%-97% *vs* 41%-55%, 86%-97% and 67%-70%, respectively)\[[@B75]\]. ADC values have also been shown to correlate with local progression-free survival\[[@B76]\]. Another group demonstrated that ADC values correlate with local progression-free survival and proposed that ADC and RECIST criteria could be substituted for mRECIST in post-radiation evaluation of patients not amenable to receiving contrast agents\[[@B76]\]. **Systemic chemotherapy:** DWI can detect the effects of chemotherapy combined with antiangiogenetic treatment on liver metastases in patients with advanced colorectal cancer\[[@B90]\]. An increase in ADC values following systemic chemotherapy can be a sign of tumor response with non-responders showing lower ADC values than responders\[[@B91]\]. In addition to monitoring therapeutic response, DWI has also been found to be useful in prediction of response to chemotherapeutic agents\[[@B92],[@B93]\]. Limitations of DWI ------------------ Diffusion imaging has several limitations, mostly attributable to the EPI based nature of the sequence\[[@B94],[@B95]\]. SS EPI provides a limited image quality with low spatial resolution and poor SNR and is susceptible to several artifacts, including blurring, ghosting and distortions. Although modern scanners with multichannel coils, strong gradients, high magnetic fields and advanced software have been successful in reducing such effects to a great extent\[[@B96]\]. In addition, parallel imaging techniques improve SNR by allowing a decrease in acquisition time (TE)\[[@B97],[@B98]\]. 3T MRI offers an advantage due to an inherent high SNR, but suffers from several limitations. Uniform fat suppression for liver DWI has always been a challenge with 3 Tesla magnets and susceptibility artifacts are also more pronounced at 3 Tesla scanners\[[@B99]\]. The reproducibility of quantitative ADC values has also been questioned. ADC values have been reported to vary significantly depending on the hardware, human or biologic factors\[[@B100]\]. There has been considerable effort, however, to "industrialize" this important biomarker across vendor platforms\[[@B101]\]. CONCLUSION ========== DWI is useful for focal liver lesion detection and is a desirable tool in patients who cannot receive intravenous contrast. In patients receiving systemic and local therapies for hepatic malignancies, DWI acts as a clinical tool for monitoring treatment response and predicting prognosis. Its utility in the assessment of diffuse hepatic parenchymal diseases is still at a research level. Further investigation and analysis are needed to increase the reliability of the technique for these indications. DWI has certain limitations and remains an adjunct and not a replacement to conventional sequences. Manuscript source: Invited manuscript Specialty type: Gastroenterology and hepatology Country of origin: United States Peer-review report classification Grade A (Excellent): 0 Grade B (Very good): B, B Grade C (Good): C Grade D (Fair): D Grade E (Poor): 0 Conflict-of-interest statement: Nothing to disclose. Peer-review started: January 3, 2017 First decision: February 4, 2017 Article in press: June 7, 2017 P- Reviewer: Jamali R, Karthik SV, Namisaki T, Waszczuk EM S- Editor: Qi Y L- Editor: Ma JY E- Editor: Li D [^1]: Author contributions: All authors contributed equally for concept, literature search and manuscript writing. Correspondence to: Avinash Kambadakone, MD, FRCR, Assistant Professor, Harvard Medical School, Abdominal Imaging and Interventional Radiology, Massachusetts General Hospital, White 270, 55 Fruit Street, Boston, MA 02114, United States. <akambadakone@mgh.harvard.edu> Telephone: +1-617-6432009 Fax: +1-617-7264891

I. INTRODUCTION & BACKGROUND {#acm20054-sec-0001} ============================ Magnetic resonance imaging (MRI) is increasingly used in radiation oncology departments for radiation treatment planning because of the excellent contrast for soft tissues and tumors.^(^ [^1^](#acm20054-bib-0001){ref-type="ref"} ^--^ [^3^](#acm20054-bib-0003){ref-type="ref"} ^)^ MRI is also considered a primary imaging modality for Gamma Knife (GK) stereotactic radiosurgery treatment planning. Important MRI safety issues that may be new to the radiation oncology clinic include potential damage and injury to property, patients and staff from several sources: the rapid acceleration of nearby ferromagnetic objects by the high‐static magnetic field (up to 3 T for clinical units);^(^ [^4^](#acm20054-bib-0004){ref-type="ref"} ^--^ [^11^](#acm20054-bib-0011){ref-type="ref"} ^)^ the gradient fields, which have been shown to induce nerve stimulation in humans;^(^ [^12^](#acm20054-bib-0012){ref-type="ref"} ^,^ [^13^](#acm20054-bib-0013){ref-type="ref"} ^)^ the cryogenics, which can cause severe frostbite, suffocation, and substantial explosions if the pressure relief system of the cryogen containers become defective;^(^ [^14^](#acm20054-bib-0014){ref-type="ref"} ^)^ and the radiofrequency (RF) fields, which are likely the primary source of MRI‐induced thermal injury.^(^ [^9^](#acm20054-bib-0009){ref-type="ref"} ^,^ [^15^](#acm20054-bib-0015){ref-type="ref"} ^--^ [^23^](#acm20054-bib-0023){ref-type="ref"} ^)^ This study is motivated by the increasing number of reports of MRI‐induced patient thermal injuries, including burns,^(^ [^9^](#acm20054-bib-0009){ref-type="ref"} ^,^ [^13^](#acm20054-bib-0013){ref-type="ref"} ^--^ [^23^](#acm20054-bib-0023){ref-type="ref"} ^)^ and by our clinic\'s use of a dedicated 3T MR simulator^(^ [^1^](#acm20054-bib-0001){ref-type="ref"} ^,^ [^2^](#acm20054-bib-0002){ref-type="ref"} ^)^ that serves as an integral part of an active GK radiosurgery program. The causes of reported MRI‐induced thermal injuries and burns are often not well understood, are sometimes described as unknown or mysterious, and seem to originate with different heating mechanisms. In some cases, burns were associated with wires used with electronic monitoring equipment or implanted biomedical devices such as pacemakers,^(^ [^8^](#acm20054-bib-0008){ref-type="ref"} ^--^ [^11^](#acm20054-bib-0011){ref-type="ref"} ^,^ [^15^](#acm20054-bib-0015){ref-type="ref"} ^,^ [^17^](#acm20054-bib-0017){ref-type="ref"} ^,^ [^20^](#acm20054-bib-0020){ref-type="ref"} ^--^ [^24^](#acm20054-bib-0024){ref-type="ref"} ^)^ while in other cases, thermal injuries have occurred with no wires present near the patient, in the extremities or around tattoos,^(^ [^24^](#acm20054-bib-0024){ref-type="ref"} ^)^ for instance. While the specific situations leading to these injuries may be difficult to pinpoint, the heating mechanisms causing them are not mysterious. They are the well‐understood physical phenomena of electromagnetic induction and the antenna effect, both originating with the radiofrequency (RF) outputs of the MRI machine. In 2003, headframe and GK manufacturer, Elekta, (Elekta AB, Stockholm, Sweden) notified users of the availability of "insulated posts", with the stated use for "high tesla MR units and high frequency MR sequences".^(^ [^25^](#acm20054-bib-0025){ref-type="ref"} ^)^ Beginning at approximately the same time, reports to the US Food and Drug Administration (FDA) documented thermal injuries due to MR‐induced heating for patients wearing stereotactic headframes.^(^ [^26^](#acm20054-bib-0026){ref-type="ref"} ^--^ [^33^](#acm20054-bib-0033){ref-type="ref"} ^)^ The physical explanation of these reported thermal injuries has not been given. Thus, this study seeks to prevent MRI‐induced burns in GK patients by first understanding the physical mechanisms that could lead to these injuries and, subsequently, validating the technique recommended to prevent them. The manufacturer\'s recommended burn prevention technique is to replace the tapped holes at the GK headframe screw--post interface, a metal‐to‐metal junction, with snap‐in insulated nuts. The use of the insulated headframe posts is required for both 1.5 T and 3 T MRI scans.^(^ [^34^](#acm20054-bib-0034){ref-type="ref"} ^)^ The use of uninsulated posts is permitted for X‐ray‐only procedures, such as CT scans or biplanar projection angiography. A. MRI‐induced heating mechanisms {#acm20054-sec-0002} --------------------------------- Previous studies of MRI‐induced burns considered three potential heating mechanisms:^(^ [^15^](#acm20054-bib-0015){ref-type="ref"} ^,^ [^16^](#acm20054-bib-0016){ref-type="ref"} ^)^ 1) resistive heating from currents induced by direct electromagnetic induction, 2) the unlikely coincidental presence of a conducting loop arranged perpendicular to the RF field and containing the right combination of inductance and capacitance to result in a resonant frequency equal to that of the MRI RF field (a special case of electromagnetic induction), and 3) the antenna effect, which is antinodal heating at the tips of wires or other conductors of appropriate length that act as antennas. A brief review of these potential heating mechanisms is presented. The first mechanism, electromagnetic induction, is described by Faraday\'s law: $$\oint\overset{\rightarrow}{E} \cdot \overset{\rightarrow}{dl} = - d/_{dt}\left( \int{\overset{\rightarrow}{B} \cdot d\overset{\rightarrow}{a}} \right)$$ where $\overline{E}$ is the electric field, $\overline{l}$ is the distance around the loop, $\overline{B}$ is the magnetic flux density, and $\overline{a}$ is the cross‐sectional area enclosed by a conducting loop. This can be stated more simply as: $$V \propto dB/dt$$ where *V* is the voltage induced in the loop, *B* is the magnetic field, and *t* is time. In this case, the rapidly changing magnetic fields induce a current in loops of wire or other conducting material, with the area enclosed by the loop oriented perpendicular to the changing magnetic field. The voltage in turn induces a current: $$i = V/R$$ where *V* is the voltage and *R* is the resistance of the loop. The induced current is then dissipated as heat at a rate: $$P = i^{2}R$$ with the greatest heating occurring at locations with the highest resistance. These loops can be formed by wires, other conducting material such as the GK headframe posts, and/or loops of human tissue such as a patient with his arm forming a closed loop, or human tissue plus a section of wire forming a loop. The current will dissipate via resistive heating, the majority of which will occur at the position of highest electrical resistance which tends to occur at the skin--skin interface (e.g., loop formed by the arm), or the wire--skin interface (e.g., loop formed by a wire). In the study by Dempsey et al.^(^ [^16^](#acm20054-bib-0016){ref-type="ref"} ^)^ different diameter loops of copper wire were placed perpendicular to the RF field and the temperature rise was recorded. Larger diameter loops resulted in more heating, consistent with Faraday\'s law. No resistor was placed in the loop and, therefore, the resistive heating was distributed evenly around the loop, resulting in a temperature rise along the whole loop. Had a resistor been placed in the loop, most of the heat would have been dissipated locally at and near the resistor, resulting in a higher temperature rise over a much smaller region of the loop. Whether the induced currents originated with the RF field or with the gradient field was not resolved in the Dempsey study. However, direct measurement of the induced voltage waveform in a loop of wire showed that the source of the heating is primarily the RF field.^(^ [^17^](#acm20054-bib-0017){ref-type="ref"} ^)^ Subsequent measurements confirmed this result, showing that reducing the magnitude of the RF signal reduces the induced voltage in the loop of wire.^(^ [^17^](#acm20054-bib-0017){ref-type="ref"} ^)^ Resonance heating is a second possible heating mechanism. Dempsey et al.^(^ [^16^](#acm20054-bib-0016){ref-type="ref"} ^)^ found very high temperature rises of up to 61°C in loops with appropriately valued inductance and capacitance to cause resonance. The resonant frequency of the loop is given by: $$f = \frac{1}{2\pi\sqrt{\text{LC}}}$$ In practice, it is unlikely that the loops described above will by coincidence happen to have the appropriate values of inductance (L) and capacitance (C) so that the resonant frequency matches the frequency of the MRI machine; however, if they do, there will be substantial heating. The third mechanism that can lead to heating is the antenna effect, which occurs when a wire of appropriate length is exposed to the RF frequency and acts as an antenna. This effect is exploited in half‐wavelength dipole antennas for receiving radio signals. If a length of wire or other conductor is $\sim\lambda/2$, an electromagnetic oscillation (resonance) will be produced with a node in the center of the wire and an antinode at either end. The maximum amplitude will occur at the antinodes, resulting in the ends of the wire heating to high enough temperatures for thermal injury. Dempsey and Condon^(^ [^15^](#acm20054-bib-0015){ref-type="ref"} ^)^ showed that for a 1.5 T MRI machine with operating frequency of 63.87 MHz, the antenna effect occurred in a wire length of $\sim 220\,\text{cm}$, in rough agreement with the theoretical expectation of $\lambda/2$ approximately equal to 235 cm. Local temperature increases of up to 63°C were observed at the wire tip, high enough to cause burns to the experimental apparatus,^(^ [^16^](#acm20054-bib-0016){ref-type="ref"} ^)^ and certainly high enough to cause severe or ablative tissue injury. This mechanism is suspected in a number of cases of MRI induced patient burns, including pulse oximeter wires.^(^ [^24^](#acm20054-bib-0024){ref-type="ref"} ^)^ It is extremely important to realize that the length of an ideal $\lambda/2$ antenna inside the human body is reduced by an order of magnitude, because the EM wavelength is significantly reduced due to the dielectric and electrical properties of human tissues. For instance, pacemaker leads in human‐equivalent soft tissue will resonate at lengths of $\sim 20\,\text{cm}$ for 1.5 T (64 MHz), or $\sim 10\,\text{cm}$ for 3 T,^(^ [^20^](#acm20054-bib-0020){ref-type="ref"} ^)^ not the 220 to 235 cm in‐air values reported earlier.^(^ [^15^](#acm20054-bib-0015){ref-type="ref"} ^)^ The antenna effect in conductive tissues is explained as follows. The human body is conducting, and thus the behavior of incident EM waves is described by Maxwell\'s equations applied to conducting media. For this case, Ampere\'s law is: $$\overset{\rightarrow}{\nabla} \times \overset{\rightarrow}{B} = \mu\varepsilon^{d\overset{\rightarrow}{E}}/_{dt} + \mu\sigma\overset{\rightarrow}{E}$$ where $\overline{B}$ is the magnetic field, $\overline{E}$ is the electric field, μ is the permeability, ɛ is the permittivity, and σ is the conductivity. The solutions to the resulting wave equations are plane waves, but with a complex wave number $$\overset{\sim}{k} = k + \mathit{iK}$$ where $$k = \omega\sqrt{\varepsilon\mu/_{2}}\left\lbrack \sqrt{1 + {(\sigma/_{\varepsilon\omega})}^{2} + 1} \right\rbrack^{1/_{2}}$$ and $$K = \omega\sqrt{\varepsilon\mu/_{2}}\left\lbrack \sqrt{1 + {(\sigma/_{\varepsilon\omega})}^{2} - 1} \right\rbrack^{1/_{2}}$$ The wavelength, λ, in the human body is thus given by: $$\mathit{\lambda} = \frac{1}{k}$$ which depends primarily on the permittivity, ɛ, the conductivity, σ, and the frequency, ω. Another result is that the amplitude of the wave will decrease with increasing penetration depth into the human body, quantified by the skin depth, δ $$\delta = \frac{1}{K}$$ which describes the depth at which the amplitude is decreased by 1/e, or to about 37% of the surface value. For given ω, the wavelength depends on ɛ, and σ, and the wave amplitude depends on its depth below the surface, quantified by the skin depth. Accordingly, the behavior of EM waves in the human body varies within the different organs and tissue types, based on ɛ and σ values over the frequency range.^(^ [^35^](#acm20054-bib-0035){ref-type="ref"} ^,^ [^36^](#acm20054-bib-0036){ref-type="ref"} ^)^ It is also important to note that antenna effect heating depends on the angle of the conductor with respect to the applied EM wave and is maximized when the length of the conductor is parallel to the direction of the EM wave. The antenna effect for metallic implants has been studied in tissue equivalent body^(^ [^20^](#acm20054-bib-0020){ref-type="ref"} ^)^ and head^(^ [^35^](#acm20054-bib-0035){ref-type="ref"} ^,^ [^36^](#acm20054-bib-0036){ref-type="ref"} ^)^ phantoms (e.g., matched ɛ and σ). For the brain, for 3 T, the wavelength is estimated to be 25.5 cm,^(^ [^35^](#acm20054-bib-0035){ref-type="ref"} ^,^ [^36^](#acm20054-bib-0036){ref-type="ref"} ^)^ suggesting that an implant length of $\sim 12.75\,\text{cm}$ would be susceptible to antenna effect heating. At 7 T, wavelength becomes 10.6 cm leading to potential antenna effect heating for 5.3 cm implants.^(^ [^35^](#acm20054-bib-0035){ref-type="ref"} ^,^ [^36^](#acm20054-bib-0036){ref-type="ref"} ^)^ These implant dimensions give an indication of possible conductor lengths of relevance for antenna effect heating for GK patients wearing headframes. All reported MRI‐induced burns likely originate in some form from either electromagnetic induction or the antenna effect. For the case of the headframe screw heating, the results of this study show that the cause is electromagnetic induction. This finding is consistent with the dimensions of the GK headframe which, with an effective unwrapped length of $\sim 62\,\text{cm}$, is too short for antenna effect heating at 3 T (128 MHz, $\lambda/2$ approximately equal to 117 cm) or 1.5 T (64 MHz, $\lambda/2$ approximately equal to 220 cm).^(^ [^16^](#acm20054-bib-0016){ref-type="ref"} ^)^ Also, the typical $\sim 5 - 10\,\text{mm}$ length of the screws embedded in the head surface is less than the $\sim 13\,\text{cm}$ length expected for the antenna effect in the human head to occur. Thus, the cause of heating is most likely due to the induction of currents in loops formed by the headframe and the tissue of the patient\'s head, with the area of the loop oriented perpendicularly to the rapidly changing magnetic fields. Because the heat source is resistive heating, $P = i^{2}R$ (Eq. [(4)](#acm20054-math-0004){ref-type="disp-formula"}), one needs only to increase the resistance of the part of the loop outside the patient\'s head (i.e., the headframe) in order to decrease the resistive heating in the tissue. The following experimental results using melon phantoms show that heating occurs during standard MR brain scans near attached GK headframe screws, the heating mechanism is electromagnetic induction, the amount of heating depends on the material type of the headframe screws and posts, and the use of insulated headframe posts renders the induced currents and associated resistive heating negligible. II. MATERIALS AND METHODS {#acm20054-sec-0003} ========================= A. Experimental geometry, pulse sequences, and temperature measurements {#acm20054-sec-0004} ----------------------------------------------------------------------- Fresh watermelons and honeydew melons were used to simulate the human head (Fig. [1](#acm20054-fig-0001){ref-type="fig"}). The melons ranged in weight approximately from 1.5 to 3 lbs. The electrical resistance of the surface layer of both types of melon, about 1 MΩ, is comparable to the resistance of the human head measured by point contacts on the skin. Both displayed similar resistive properties, with $\left. R \right.\sim 1\, M\Omega$ within the outer shell of the melon, but decreasing to $\sim 50 - 300k\Omega$ if the inner pulp is penetrated by the meter lead (Fig. [1](#acm20054-fig-0001){ref-type="fig"}). As shown in Fig. [2](#acm20054-fig-0002){ref-type="fig"}, the melons are mounted in the GK headframe and optical thermometers are mounted with tape at various positions on the headframe, screws, and melon. All experiments were conducted using a GE 3.0 T MR scanner (Signa EXCITE, GE Healthcare, Waukesha, WI). Two pulse sequences were investigated: 1) a standard T1‐weighted GK Protocol sequence (axial T1 spin echo, with flow compensation, $\left. \text{TE} = \right.\sim 23\,\text{ms}$, $\text{TR} = 800\,\text{ms}$, $\text{NEX} = 1$, $\text{BW} = 22.73$, and a $384 \times \, 224$ matrix), and 2) in order to induce increased heating, an enhanced 13 minute fast spin echo sequence (FSCXL with flow compensation and tailored RF, $\text{TE} = 16\,\text{ms}$, $\text{TR} = 767\,\text{ms}$, $\text{NEX} = 4\,\text{MS}$, $\text{BW} = 20.83$, $\text{ETL} = 37$, and a $256 \times \, 256$ matrix). Temperature was measured using two MR‐compatible fiber optic thermometers, (Veris MR Vital Signs Monitoring System, Medrad, Inc, Warrendale, PA). The temperature is determined via a temperature sensitive phosphor located at the probe tip and energized by an LED pulse.^(^ [^37^](#acm20054-bib-0037){ref-type="ref"} ^)^ Thermal connections to the headframe screws were made by careful adhesion of the more thermally conductive *side* (as opposed to the end) of the optical thermometer to the tip of the screw (arrow, Fig. [2(c)](#acm20054-fig-0002){ref-type="fig"}. Thermal connections to the melon were made by enclosing the optical thermometers in thin plastic wrap and inserting them directly into the melon to a depth of 5 mm, preventing air cavities and ensuring good thermal contact. {#acm20054-fig-0001} {#acm20054-fig-0002} B. Antenna‐effect heating {#acm20054-sec-0005} ------------------------- To verify the absence of antenna‐effect heating, the GK headframe alone with 4 uninsulated posts and 4 tungsten tipped alumina screws (Fig. [3(a)](#acm20054-fig-0003){ref-type="fig"}) was placed in the GK standard four‐element head coil and scanned in the MRI using the current standard T1‐weighted GK protocol pulse sequence. Temperature measurements at the two anterior headframe screws were recorded every 30 seconds during the 9.5 minute pulse sequence. No heating was detected at either screw, confirming that the headframe--post--screw unit was not behaving as an antenna (Fig. [3(b)](#acm20054-fig-0003){ref-type="fig"}), triangles). Similarly, 45 and 60 mm titanium screws alone, without the headframe or posts, were embedded in the melon to depths of 5 mm and evaluated for heating with the T1‐weighted pulse sequence. Again, no heating was observed at the screw--melon interface, indicating the absence of the antenna effect for these length screws for the 0.5 mm portion of the screw within the melon. {#acm20054-fig-0003} C. Electromagnetic induction heating {#acm20054-sec-0006} ------------------------------------ To determine whether there is any heating from electromagnetic induction in a loop, a water melon was subsequently mounted on the same headframe and remeasured in the same four‐element head coil using the same pulse sequence. The screws were screwed approximately 5 mm into the surface of the watermelon and the temperature was again measured every 30 seconds at the top part of the anterior screw tips at the melon surface. A temperature increase of approximately 6°C near the screw tips was observed, confirming electromagnetic induction as the source of the heating (Fig. [3(b)](#acm20054-fig-0003){ref-type="fig"}), circles). To further characterize the temperature profile of the melon headframe composite, a watermelon was mounted on the GK headframe and with the T1 axial GK pulse sequence, the temperature was measured at the screw tip and three positions along the melon in order to determine where heating occurs. Heating was greatest along the screw surface and at two points in the melon nearest to the screw (Fig. [4](#acm20054-fig-0004){ref-type="fig"}). {#acm20054-fig-0004} D. Screw--post combinations {#acm20054-sec-0007} --------------------------- With the heating mechanism established as electromagnetic induction, the heating characteristics of different screw and post types were measured. The temperature profiles were measured for combinations of two different screw types, and two different headframe post materials. All screw and post comparison tests were done on the same melon in immediate succession to ensure the same measurement conditions. Baseline temperature was determined by the initial temperature measured just before starting the initial scan, and subsequent scans were started only after temperatures cooled to within 1°C of baseline. Baseline temperatures varied from day to day from 18°C to 23°C, with a typical value of 20°C. Titanium‐ or tungsten‐tipped alumina screws were measured in combination with either carbon or alumina posts, with both regular threads and insulated (plastic) nuts (Fig. [5](#acm20054-fig-0005){ref-type="fig"}). These combinations were chosen to represent the possible configurations available in the clinic, as well as to represent a range of safety (from "not safe" through "safe") (Table [1](#acm20054-tbl-0001){ref-type="table"}). To determine whether the manufacturer‐recommended insulating posts prevent heating, the standard alumina headframe posts were replaced with electrically insulated posts, designed by placing snap--in insulating nuts between the screw and the post in the headframe in place of the regular threaded hole (Fig. [5(a)](#acm20054-fig-0005){ref-type="fig"}), center). Experimental results confirm that this increased impedance prevents significant heating at the headframe screw--patient head interface. As shown in Table [1](#acm20054-tbl-0001){ref-type="table"}, the insulated posts are the only ones that rendered the heating negligible. ###### General heating characteristics for different screw‐post combinations. Based on temperature profile plots (see associated figures), combinations that showed anything more than negligible heating were deemed "not safe". *Screw Type* *Alumina* *Post Type Carbon* *Insulated Alumina* -------------- ------------------------------------------------------------------------------ ----------------------------------------------------------------------- ------------------------------------------------------------------------- Alumina not safe (Fig. [11](#acm20054-fig-0011){ref-type="fig"}) significant heating not safe (Fig. [12](#acm20054-fig-0012){ref-type="fig"}) some heating safe (Fig. [14](#acm20054-fig-0014){ref-type="fig"}) negligible heating Titanium not safe (Fig. [11](#acm20054-fig-0011){ref-type="fig"}) most heating not safe (Fig. [12](#acm20054-fig-0012){ref-type="fig"}) some heating safe (Fig. [14](#acm20054-fig-0014){ref-type="fig"}) negligible heating {#acm20054-fig-0005} III. RESULTS {#acm20054-sec-0008} ============ A. Electromagnetic induction responsible for MRI‐induced heating {#acm20054-sec-0009} ---------------------------------------------------------------- The initial experiment was designed to determine which of the two most likely heating mechanisms, the antenna effect or electromagnetic induction, is responsible for the heating that occurs at the GK headframe screw--melon interface. Antenna‐effect heating can be theoretically ruled out because the dimensions of the GK headframe are not large enough for this type of heating to occur. For the 128 MHz RF field of the 3 T magnet, the RF wavelength is $\sim 235\,\text{cm}$, which corresponds to $\left. \lambda/2 \right.\sim 117\,\text{cm}$. The longest posts on the GK headframe are $\sim 15\,\text{cm}$ and the overall unwrapped length is $\sim 62\,\text{cm}$, much shorter than the required $\sim 117\,\text{cm}$ necessary for antenna heating to occur in air. Antenna‐effect heating was experimentally ruled out based on temperature measurements that showed: 1) no heating for the assembled headframe suspended in air, and 2 no heating for screws of various lengths embedded at depths of 0.5 cm in the melon. This latter null result is expected based on the nominal EM wavelength produced by 3 T scans for soft tissue ($\sim 40\,\text{cm}$) and for brain tissue ($\sim 25\,\text{cm}$).^(^ [^35^](#acm20054-bib-0035){ref-type="ref"} ^,^ [^36^](#acm20054-bib-0036){ref-type="ref"} ^)^ Resonance‐induced heating is a remaining phenomenon that could possibly result in heating. We have measured negligible capacitances and inductances in the GK headframe--melon composite and there is no evidence of resonance heating occurring in this experiment. B. Induced heating location {#acm20054-sec-0010} --------------------------- Temperature measurements versus time as a function of position between the two anterior screws show that heating is greatest at the tapered part or tip of the screw along the melon--screw interface and that heating decreases with increasing distance from the screw tip (Fig. [4](#acm20054-fig-0004){ref-type="fig"}). Specifically, we found that heating occurs only near the tip of the screw and not primarily at back of the screw (Fig. [6](#acm20054-fig-0006){ref-type="fig"}). This temperature profile is consistent with the formation of a loop with most of the heating occurring at the maximum resistance spot in the loop, which is at the screw--melon interface in the conventional frame setup. {#acm20054-fig-0006} Melon resistivity was measured at different locations on the surface and at depth to determine variability that would impact temperature measurements. Using a handheld resistance meter, we measured a resistance of 1--3 MΩ for the outer flesh of the melon. Deeper into the melon, however, where the pulp is located, there is a significant drop in resistance down to $\sim 100\, k\Omega$. Experiments repeated with the thermometers placed in this deeper region show significantly decreased heating, owing to the reduced resistance (Fig. [7](#acm20054-fig-0007){ref-type="fig"}). This result shows that the region of heat is resistance dependent, again confirming that the source of the heating is the induced current loop with the same current heating more in higher resistance regions. To further characterize the heating that occurs in the melon headframe composite, the time dependence of the temperature was measured on the posterior loop, which would be expected to have less heating than the anterior loop due to the smaller area enclosed. In Figure [8](#acm20054-fig-0008){ref-type="fig"}, the time dependence of the temperature of the posterior loop (Fig. [8(a)](#acm20054-fig-0008){ref-type="fig"}), triangles) is plotted with that of the anterior loop (circles). As expected from current induced in accordance to Faraday\'s law, the induced current and its associated heating is less in the posterior loop because of the smaller cross sectional area. Other potential loops are separately measured to find the relative contributions to the screw heating of the composite system. Fig. [9](#acm20054-fig-0009){ref-type="fig"} shows the time dependence of the temperature taken near the tip of the right anterior headframe screw at the screw--melon interface during the 13 minute FSE scan for each of the indicated loops. The anterior loop shows the most heating, with the side and diagonal loops each showing less heating (Fig. [9](#acm20054-fig-0009){ref-type="fig"}). {#acm20054-fig-0007} {#acm20054-fig-0008} {#acm20054-fig-0009} C. Screw--post combinations {#acm20054-sec-0011} --------------------------- Temperature measurements for the two different screw materials, tungsten‐tipped alumina and titanium, show that there is more heating at the screw--melon interface for the titanium screws (Fig. [10](#acm20054-fig-0010){ref-type="fig"}, filled circles), indicating a higher resistance at this interface for the titanium screws compared with the alumina screws (hollow circles). Different combinations of headframe screw and post materials show different heating characteristics associated with different resistance characteristics. Several different screw--post combinations were measured. First, using the titanium screws, the time dependence of the temperature was again measured, but with carbon posts instead of alumina posts (Fig. [11](#acm20054-fig-0011){ref-type="fig"}). At the screw tips, the heating decreased for the carbon posts relative to the alumina post, but the heating increased at the screw--post interface (Fig. [12](#acm20054-fig-0012){ref-type="fig"}). This effect is also consistent with heating by electromagnetic induction. With the alumina posts, there is very little resistance at the screw--post interface, but with the carbon posts, there is significant heating, showing that there is a significant resistance at the post--screw interface. Because the size of the loop is the same, the same voltage is induced in the loop. However, now there are two locations of significant resistance, so the heat is dissipated at two locations instead of just one. Thus the heating at the screw--melon interface decreases, but increases at the screw--post interface. This occurs only for the carbon posts. Because carbon posts may offer some advantages for CT imaging, it would be possible to use these posts with minimal heating via a similar insulated nut setup as is currently used with the alumina posts. {#acm20054-fig-0010} {#acm20054-fig-0011} {#acm20054-fig-0012} Temperature measurements with the uninsulated alumina posts replaced with the insulated posts using snap‐in insulated nuts prove that the insulated posts diminish heating to a negligible level (Fig. [13](#acm20054-fig-0013){ref-type="fig"}). A close--up view shows the small amount of heating (Fig. [13(b)](#acm20054-fig-0013){ref-type="fig"}). For mixed headframe--post--screw combinations, the results for one insulated alumina post and three uninsulated alumina posts are compared with the case for one insulated alumina post and three carbon posts (Fig. [14](#acm20054-fig-0014){ref-type="fig"}). As expected, there is less heating at the screw tips when carbon posts are used, compared with alumina posts. {#acm20054-fig-0013} {#acm20054-fig-0014} In general, for the case of the GK headframe, the greatest heating occurs at the headframe screw--patient head interface, with the loop formed by the frame in contact with the skin. The results of this study show that this patient--frame loop is the only mechanism that can lead to GK patient burns that occur in the vicinity of the headframe screws. IV. DISCUSSION {#acm20054-sec-0012} ============== This set of experiments was designed to determine which physical mechanism could lead to local heating of the GK headframe for patients undergoing stereotactic MR imaging, and to test the manufacturer\'s recommended method for heating prevention using electrically insulated posts. Additional experiments were conducted to characterize typical heating profiles of the GK headframe--patient system that occurs during MRI scans, and to learn how this profile changes when headframe post and screw materials are changed. Initial experiments (Fig. [3](#acm20054-fig-0003){ref-type="fig"}) ruled out the antenna effect as a headframe heating mechanism and confirmed that the observed heating originates with the resistive dissipation of currents induced by the rapidly changing magnetic fields passing perpendicularly through the area enclosed by electrically conducting loops comprised of the headframe, headframe screws, and region of the patient\'s head between the two headframe screws, as described by Faraday\'s law. Further experiments showed that the resistive heating occurs primarily at the screw tips (Fig. [4](#acm20054-fig-0004){ref-type="fig"}, Fig. [6](#acm20054-fig-0006){ref-type="fig"}), where the electrical resistance is highest. Subsequent experiments found no antenna‐effect heating in the portion of the screw embedded in the melon with no headframe attached. The heating profile can be controlled by changing the materials of the headframe posts and screws. This was observed in the experiments comparing combinations of alumina, insulated alumina, and carbon posts, and titanium‐ and tungsten‐tipped alumina screws (Figs. [10](#acm20054-fig-0010){ref-type="fig"}--[14](#acm20054-fig-0014){ref-type="fig"}). Figures [11](#acm20054-fig-0011){ref-type="fig"} and [12](#acm20054-fig-0012){ref-type="fig"} compare the heating characteristics of the headframe composite with alumina screws and alumina posts with that of same setup, but with carbon posts in place of the alumina posts. While additional heating occurs at the carbon post--headframe screw interface (Fig. [12](#acm20054-fig-0012){ref-type="fig"}), causing the region of the post surrounding the headframe screw to heat, there is slightly less heating at the screw--patient head interface (Fig. [11](#acm20054-fig-0011){ref-type="fig"}), which again is consistent with electromagnetic induction heating. The carbon post--screw interface has a significantly higher impedance compared with the alumina post--screw interface and, thus, dissipates a larger amount of the induced current here, resulting in a reduced dissipation at the patient head--screw interface. Replacing the uninsulated posts with the insulated posts prevented all but negligible heating of the melon headframe composite (Fig. [13](#acm20054-fig-0013){ref-type="fig"}), which is consistent with Faraday\'s law. The insulating nuts greatly increase the resistance of the loop. If, for example, the resistance of the loop is doubled, then the induced voltage, which is a function only of the cross‐sectional area of the loop, is still the same, resulting in the current being halved. Resistive heating $P = i^{2}R$, so that while the resistance is doubled, the induced current squared is quartered; therefore, the heating is halved. Thus increasing the resistance in any given loop reduces the heating when the mechanism is Faraday\'s law (e.g., electromagnetic induction). The manufacturer\'s recommended method works because the heating mechanism is electromagnetic induction. It is important to note that all four posts should be insulated, because as shown in Fig. [9](#acm20054-fig-0009){ref-type="fig"}, multiple loops contribute to the heating at any given screw tip. This same antiheating technique is effective for any headframe post material, and because carbon posts may offer some advantages in CT scans, it would be possible to use these posts with minimal heating by using the same insulated nut setup that is currently used with the alumina posts. The manufacturer\'s recommended heating prevention method, however, works only to prevent heating by electromagnetic induction and would not prevent antenna‐effect heating. An entirely different method would be required to prevent heating by the antenna effect. Fortunately, the GK headframe geometry is not a problem at 1.5 T or 3 T field strengths, because the dimensions are too small to heat by this method. However, if the static field operated at 7--10 T, the implant dimensions required for the antenna effect would be decreased to $\sim 30 - 50\,\text{cm}$ in air, and as small as just a few cm in the human body, depending on the type of tissue or organ where the implant is located, with the amount of heating dependent on the implant\'s depth below the surface and its angle with respect to the applied EM wave. This effect may be a major concern for patients with small implants in the human body, such as stents or aneurysm clips, that could pose a severe burn hazard at very high field strengths. Thus, with current MRI scanners operating at fields up to 7 T, the antenna effect may become the dominant heating hazard in the near future. One of the key implications of this study is that therapists and other health professionals who image GK patients with headframes in MRI scanners and have not yet obtained insulated posts need to be aware of new safety regulations (such as the requirement to use only insulated posts for MRI scans of GK patients with headframes). It may be useful for GK health professionals to obtain additional training to understand how RF interacts with and potentially heats the human body through resistive heating in loops. It should be understood how current loops could potentially form through complex combinations of parts of a patient, a patient monitoring device, or instrument wiring via capacitive coupling, especially for higher field MRI units. The more complex heating mechanism of the antenna effect should also be understood by GK health professionals so that they are aware that thermal injuries can occur by mechanisms other than EM induction. Tables of conductivities and permittivities of different organs and tissues within the human body should be made readily available, along with dimensions of conducting implants that lead to antenna‐effect heating in various tissues in different field strength MRI scanner. V. CONCLUSIONS {#acm20054-sec-0013} ============== This study shows that heating caused by the RF field of a 3T MRI scanner due to electromagnetic induction, as described by Faraday\'s law, occurs at the GK headframe screws when melon phantoms are mounted in the headframe during stereotactic MR imaging. Titanium screws combined with the uninsulated alumina posts result in maximum heating at the screw tips. This heating can be greatly reduced to negligible levels by the use of insulating nuts (the manufacturer\'s recommended procedure) that electronically separate the metallic screws from the posts. This method would, in principle, work for any conducting headframe post material (e.g., steel and carbon fiber). There is increased risk of thermal injury at field strengths higher than 3T. The antenna effect was ruled out as the cause of headframe screw heating. However, above 3T it poses an increased risk of internal heating for GK and other patients with metallic implants having lengths suitable for standing wave formation (such as pacemaker wires), because resonant length decreases with increasing field strength and RF frequency. Thus, injury may be possible with even small‐dimension implants, such as aneurysm clips and stents. Electromagnetic induction, the principal heating mechanism identified for GK patients with attached headframes, also poses an increased risk because higher RF frequencies at field strengths above 3T may create unintended current paths through unintended capacitances. Thus, at sufficiently high magnetic fields beyond 3T, the electrically insulating nuts may no longer protect against induced heating. Their effective protection will need to be validated before use for GK patients with headframes undergoing stereotactic very high‐field (e.g., 4T, 7T, and higher) MR imaging. ACKNOWLEDGMENTS {#acm20054-sec-0014} =============== The authors would like to thank S. Shave at General Electric Healthcare, O. Eriksson at Elekta, and Lisa Wilkins and Darrell Sloan at the Department of Radiation Oncology at Wake Forest University for their time and help with this project. This work is supported by the TRADONC post‐doctoral training program at Wake Forest University School of Medicine, funded by Grant No. NCI‐32 CA113267.

Introduction {#Sec1} ============ The epidemiological effects of public disease-risk mitigation measures such as quarantines, school closures, vaccinations, trade interdictions, or travel restrictions have attracted considerable attention (see (Chowell et al. [@CR14]; Ferguson et al. [@CR21]; Chowell et al. [@CR15]; Shim and Galvani [@CR49]) for examples relating to SARS, ebola, and avian influenza). The net social benefits of such measures have also been assessed (Gupta et al. [@CR27]; Bridges et al. [@CR10]; Cauchemez et al. [@CR13]; Sadique et al. [@CR46]; Sadique et al. [@CR45]; Sander et al. [@CR47]). Less attention has been paid to the benefits of private disease-risk mitigation measures such as contact reduction, prophylaxis, private vaccination, or preferential mixing according to health status (Fenichel [@CR17]; Fenichel et al. [@CR20]; Gross et al. [@CR26]; Maharaj and Kleczkowski [@CR39]), and almost no attention has been paid to the net social benefits of private disease-risk mitigation (Reluga [@CR43]). We address this problem here. We focus on the class of infectious diseases that allow recovery with immunity (see Table [2](#Tab2){ref-type="table"}). Since we model neither memory nor learning, recovery with immunity allows us to treat each outbreak as an independent event. We also ignore births and deaths. We suppose that susceptible and infected people behave in different ways as a function of disease risk, and not as a function of the infection itself (as in the case of, for example, gonorrhea (Yorke et al. [@CR54]; Blank et al. [@CR7])). Within each health class, we assume that all individuals respond to risk in the same way and so ignore any sources of heterogeneity in the behavioral response of individuals due to, for example, age, gender, or occupation (Albarracín et al. [@CR1]; Klepac and Caswell [@CR35]; Khan et al. [@CR34]). Among possible risk mitigation strategies, we focus on the case where reactive individuals, defined as susceptible, asymptomatic, or recovered asymptomatic individuals, preferentially mix with healthy people and avoid sick people.[1](#Fn1){ref-type="fn"} How much reactive people adjust their pattern of contacts depends on the relative costs of preferential mixing and the expected cost of illness. If the disease risk is positive, where risk is the probability of illness multiplied by its cost, individuals will invest in preferential mixing up to the point at which the marginal expected costs of disease and of disease avoidance are equalized. Investment in preferential mixing will increase with the cost of disease and decrease with the cost of disease avoidance. We show that preferential mixing always reduces the size of epidemics, but increases their duration. We then compare the cost of epidemics with and without preferential mixing to measure when, and under what conditions, private disease-risk mitigation is socially beneficial. We show that the social net benefit of private disease-risk mitigation is systematically related to the characteristics both of the disease and of the society in which the disease occurs. If the benefit of avoided illness is high compared to the cost of avoidance, we find that private disease-risk mitigation always yields net benefits to society. As the relative benefits of avoided illness become smaller, however, so do the net benefits to society. Whether proportionate or preferential mixing is more costly then depends on the weight that society gives to the private cost of illness and the rate at which they discount future relative to present costs. We show the conditions under which private disease-risk mitigation results in a net loss to society and consider what this means for infectious disease management in general, for a rich/poor world in particular. A mathematical model of private disease-risk mitigation {#Sec2} ======================================================= Our modeling approach builds on existing affinity-based mixing compartment models (Hadeler [@CR29]; Hadeler and Castillo-Chavez [@CR30]; Busenberg and Castillo-Chavez [@CR11]; Morin et al. [@CR40]), where compartments represent different disease states. In most models, the probability of contact between individuals in different compartments depends only on their prevalence in the population (proportionate mixing). By contrast, we suppose that individuals mix preferentially, conditional on their own disease state and the (observable) disease states of others. As in (Fenichel et al. [@CR20]), a mixing strategy depends on the relative costs of illness and illness avoidance. The core of the model is an affinity framework for the preferential mixing structure. This framework has been shown to provide the most mathematically general solution to the problem of *who mixes with whom* (Blythe et al. [@CR8]; Levin [@CR37]). Groups may be defined by various shared attributes, including economic status, cultural or ethnic identity, geographical location, age, or disease awareness. In this paper, we define groups by their epidemiological status. The use of the affinity framework allows for three different factors to control the volume of contact between groups: (1) the size of each group, (2) the nominal activity levels of each group, and (3) the relative affinity/disaffinity between groups. We model the decision process behind changes in the affinity/disaffinity between groups, focusing on decisions made by susceptible or asymptomatically infectious (a subset of reactive) individuals. We hold the nominal level of activity (volume of contacts) constant throughout the course of the epidemic to measure more accurately the effect of changes in mixing preferences (see (Fenichel et al. [@CR20]) for a treatment that affects only the volume of contacts for the reactive class and (Morin et al. [@CR40]) for more details on varying contact volume versus contact type). To illustrate the approach, we first focus on a susceptible-infectious-recovered, *SIR*, model:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{array}{l}\frac{dS(t)}{dt}=-c\beta S(t){P}_{SI},\hfill \\ {}\frac{dI(t)}{dt}=c\beta S(t){P}_{SI}-\gamma I(t),\hfill \\ {}\frac{dR(t)}{dt}=\gamma I(t).\hfill \end{array} $$\end{document}$$ As is standard with such a model, we let *c* be the nominal contact volume of all individuals. *P* ~*SI*~ is the conditional probability that a contact made by a susceptible individual, *S*(*t*), is with an infectious individual, *I*(*t*), and *γ* is the rate at which an individual recovers and becomes immune, *R*(*t*). The affinity-based mitigation framework involves specification of a mixing matrix, *P* = (*P* ~*ij*~), that is generally taken to satisfy three mixing axioms at each moment in time *t* (Busenberg and Castillo-Chavez [@CR11]; Blythe et al. [@CR8]; Castillo-Chavez et al. [@CR12]):0 ≤ *P* ~*ij*~ ≤ 1, for all *i*, *j* ∈ {*S*, *I*, *R*},$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\displaystyle \sum_{j\in \left\{S,I,R\right\}}{P}_{ij}}=1 $$\end{document}$, for all *i* ∈ {*S*, *I*, *R*},*i*(*t*)*P* ~*ij*~ = *j*(*t*)*P* ~*ji*~, for all *i*, *j* ∈ {*S*, *I*, *R*}. The first two axioms imply that *P* is a matrix of conditional probabilities, and the third implies that it is symmetric. Susceptible individuals carry the same expected risk of encountering infection as the expected risk of infectious individuals encountering susceptible individuals. It has been shown that the unique solution to these mixing axioms is given by$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {P}_{ij}=j(t)\left[\frac{M_i{M}_j}{V}+{\phi}_{ij}\right], $$\end{document}$$where$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{array}{l}{M}_i=1-{\displaystyle \sum_{k\in \left\{S,I,R\right\}}k}(t){\phi}_{ik},\hfill \\ {}V={\displaystyle \sum_{\ell}\ell }(t){M}_{\ell },\hfill \end{array} $$\end{document}$$and *Ф* = (*ϕ* ~*ij*~) is a symmetric affinity matrix, in this case 3 × 3. The element *ϕ* ~*ij*~ may be interpreted as the effort that individuals in disease state *i* make to avoid individuals in disease state *j* (if *ϕ* ~*ij*~ \< 0) or to associate with individuals in disease state *j* (if *ϕ* ~*ij*~ \> 0). If all individuals in every disease state *i* make no effort to avoid individuals in disease state *j ϕ* ~*ij*~ = 0, we have classic proportionate mixing. The zero elements of the affinity matrix, *Ф*, reflect what we call *avoidance-neutrality*. That is, they show the individual to be neutral about a pairing event resulting from mixing behavior. By contrast, negative (positive) elements reflect the desire of an individual in one disease state to avoid (seek out) an individual in another disease state. This is a similar measure to that used in models of assortative mating (Karlin [@CR33]) and selective mixing (Hyman and Li [@CR32]) and is a form of a contact kernel (Gurarie and Ovaskainen [@CR28]). The elements of the affinity matrix describe what people want. What they actually get depends both on the preferences of others in the population and on their relative abundance. The mixing matrix *P* = (*P* ~*ij*~) for the population thus derives from the affinity matrix and represents the conditional probabilities that an individual of disease state *i* contacts someone in disease state *j*. Reactive individuals, those with incentive to avoid infection, will maximize the net present value of disease avoidance taking into account the cost of illness and illness avoidance by choosing the effort to commit to preferential mixing the elements of Ф(*t*). Formally, the decision problem for reactive individuals is to choose the level of mitigation effort, *P*(*Φ*(*t*)), to maximize the difference between the benefit of not being symptomatic, *B*, and the cost of mitigation effort, *C*(*ϕ* ~*SI*~(*t*)), give the weight they place on future wellbeing (the discount factor *ρ*) and their planning horizon, *T* $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ Ma{x}_{\phi_{SI}(t)}{\displaystyle \underset{t=0}{\overset{T}{\int }}{e}^{-\delta t}\left[B(t)\left(N(t)-I\left(P\left(\varPhi (t)\right),S(t)\right)\right)-C\left({\phi}_{SI}(t)\right)\right]dt} $$\end{document}$$ Affinity-based mixing decisions can have four different effects on *P*:Susceptible individuals seeking to reduce contact with infectious individuals can drive down the value of *ϕ* ~*SI*~ = *ϕ* ~*IS*~ directly;Recovered individuals seeking to increase non-infection-causing contacts can drive up *ϕ* ~*SR*~ = *ϕ* ~*RS*~ and lower *herd immunity* thresholds;Infectious individuals seeking to minimize contact with susceptible individuals can drive down *SI*, and possibly *RI* contacts, further reducing *ϕ* ~*SI*~ and reducing *ϕ* ~*IR*~ = *ϕ* ~*RI*~,In the limit, contact avoidance can induce an effective quarantine of infectious individuals (when *P* ~*SI*~ and *P* ~*RI*~ are very, very small, *P* ~*IS*~ and *P* ~*IR*~ are small, and *P* ~*II*~ is nearly 1). In what follows, we assume that susceptible individuals are averse to mixing with symptomatic (infectious or otherwise) individuals. In the *SIR* case, omitting all other disease-risk aversion behaviors, Ф takes the form:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \varPhi =\left(\begin{array}{ccc}\hfill 0\hfill & \hfill -a\hfill & \hfill 0\hfill \\ {}\hfill -a\hfill & \hfill 0\hfill & \hfill 0\hfill \\ {}\hfill 0\hfill & \hfill 0\hfill & \hfill 0\hfill \end{array}\right), $$\end{document}$$ with 0 representing neutrality of mixing and −*a* \< 0 representing the effort susceptible individuals make to avoid mixing with infectious individuals (Morin et al. [@CR40]). This defines$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{array}{c}\hfill {M}_S=1+ aI(t),\hfill \\ {}\hfill {M}_I=1+aS(t),\hfill \\ {}\hfill {M}_R=1,\hfill \\ {}\hfill V=1-S(t)\overline{\phi_S}-I(t)\overline{\phi_I}-R(t)\overline{\phi_R}=1+2aS(t)I(t),\hfill \end{array} $$\end{document}$$ We may then write the mixing matrix of conditional probabilities as$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ P=\left(\begin{array}{ccc}\hfill S(t)\frac{M_S^2}{V}\hfill & \hfill I(t)\left[\frac{M_S{M}_I}{V}-a\right]\hfill & \hfill R(t)\frac{M_S}{V}\hfill \\ {}\hfill S(t)\left[\frac{M_S{M}_I}{V}-a\right]\hfill & \hfill I(t)\frac{M_I^2}{V}\hfill & \hfill R(t)\frac{M_I}{V}\hfill \\ {}\hfill S(t)\frac{M_S}{V}\hfill & \hfill I(t)\frac{M_I}{V}\hfill & \hfill \frac{R(t)}{V}\hfill \end{array}\right). $$\end{document}$$ To see whether action by susceptible individuals to avoid infected individuals can be strong enough in this structure to induce isolation of infectious individuals (private quarantine), we consider the conditions under which *P* ~*SI*~ = *P* ~*IS*~ = 0. More particularly, we construct a hard upper bound for the maximum effort that may be applied to avoidance subject to relative prevalence of the epidemiological classes. Supposing that neither population is zero, *S*(*t*) *I*(*t*) ≠ 0, we consider the case$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {M}_S{M}_I-aV=0, $$\end{document}$$ which implies the convex quadratic$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ S(t)I(t){a}^2+R(t)a-1=0, $$\end{document}$$ with$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {a}^{\pm }=\frac{-R(t)\pm \sqrt{{\left(R(t)\right)}^2+4S(t)I(t)}}{2S(t)I(t)}. $$\end{document}$$ Consider the positive and negative roots, *a* ^+^ and a^−^. For a greater than the positive root, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ a\in \left(\frac{-R(t)+\sqrt{{\left(R(t)\right)}^2+4S(t)I(t)}}{2S(t),I(t)},\infty \right) $$\end{document}$, the mixing probability is less than zero, P~SI~ \< 0, and thus invalid. For $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ a\in \left(\frac{-R(t)-\sqrt{{\left(R(t)\right)}^2+4S(t)I(t)}}{2S(t),I(t)},0\right) $$\end{document}$ contacts between susceptible and infectious individuals would be desired, which violates our assumption that susceptible individuals are averse to mixing with infectious individuals. So, the effort by susceptible individuals to avoid infectious individuals is restricted to the range:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ a\in \left[0,\frac{-R(t)+\sqrt{{\left(R(t)\right)}^2+4S(t)I(t)}}{2S(t)I(t)}\right], $$\end{document}$$ with proportionate mixing at the left end point, *P* ~*SI*~(*t*) = *I*(*t*) the maximum probability of contact, and private quarantine of infectious individuals at the right end point, *P* ~*SI*~(*t*) = 0, the minimum probability of contact. Economically, the occurrence of private quarantine implies that the expected marginal cost of illness dominates the marginal cost of illness avoidance. The mixing strategy of the representative reactive individual generates infectious contact probabilities that lie anywhere between proportionate mixing and the privately driven quarantine of infectious individuals. Note that this may not be true for diseases like chicken pox or measles, where the costs of contracting the disease are much lower among children than adults. The parents of reactive children may actually seek out infection for their children to prevent the large cost of infection as an adult. In our case, however, reactive individuals assess the risk of illness and select the preferential mixing strategies that maximize the expected net benefits of those strategies where future and current costs of disease are the same. Specifically, individuals will increase effort to avoid infection up to the point where the marginal cost is offset by the marginal benefits (in terms of avoided illness) it yields. Efforts to avoid infection will be increasing in the cost of illness and decreasing in the cost of illness avoidance including any forgone benefits from contact with infectious individuals. In models without risk mitigation, disease dynamics may be completely characterized from initial conditions. With risk mitigation, the evolution of the epidemic reflects feedbacks between the cost of disease and disease avoidance on the one hand and averting behavior on the other (see (Fenichel and Horan [@CR18]; Horan et al. [@CR31]) for further discussion). The epidemiological effects of private disease-risk mitigation {#Sec3} ============================================================== We investigated several configurations of the following epidemiological classes: (S)usceptible, (E)xposed or latently infected, (A)symptomatically infectious, (I)nfectious with symptoms, and (R)ecovered and immune to the disease; not all epidemiological classes would be expected to mitigate disease risk (see Table [1](#Tab1){ref-type="table"}).Table 1Disease states and disease-risk mitigationDescriptionReactive (will mitigate risk)SSusceptibleYesELatently infected: asymptomatic and noninfectiousYesAAsymptomatically infectiousYesISymptomatically infectiousNoR~x~Recovered from disease state x and immuneR~A~ yes/R~I~ no We further considered the effect of risk mitigation by reactive individuals in four compartmental models: SIR, SEIR, and two SAIR models---a one-path and a two-path progression. Within the one-path model, susceptible individuals are first asymptomatically infectious and then progress to symptomatically infectious and then immune. In the two-path model, a susceptible individual becomes either asymptomatically or symptomatically infectious and then recovers. Individuals who recover from asymptomatic infection (R~A~) are expected to be reactive. Individuals who recover from symptomatic infection (R~I~) are not. We did not consider models with reentry to the susceptible class, e.g., SIS, SIRS, and other cyclical models. This is for two reasons. First, each of these models is capable of endemic levels of infection. This, combined with the fact that people may experience reinfection, would require individuals to form expectations (possess memory) with respect to the impact of different avoidance strategies. Second, numerical simulations of these models reveal very broad oscillations that confound comparison with single outbreak models. There are no entries (births) or removals (deaths) from the system (see Table [2](#Tab2){ref-type="table"} for example diseases listed for each model). Each model considered here is therefore a so-called single outbreak model; the population is not only kept at a fixed number, but it is also closed to the introduction of new individuals.Table 2Models studied and corresponding compartments. Red compartments denote reactive individuals who engage in disk-risk mitigation. Compartment classifications are shown in Table [1](#Tab1){ref-type="table"}. For noncommunicable diseases (such as cholera or West Nile virus), it is assumed that the number of infectious individuals is proportionate to the infectious material or the disease vector We modeled the dynamics of the epidemic types in Table [2](#Tab2){ref-type="table"} using ordinary differential equations (see Table [3](#Tab3){ref-type="table"}). This has two main implications for disease dynamics: (a) once nonzero, the state variables will never again be zero in finite time, and (b) in an infinitely small amount of time "mass" will move into each compartment as long as the transition rates are nonzero. These affect the interpretation to be given of the point at which an epidemic is "over." It is feasible that an extremely strong avoidance response early in the course of the epidemic could wipe out the infection within a population. However, within the differential equation framework, as soon as those behavioral adjustments are loosened, coupled with the fact that there is a nonzero infectious population with potentially a very large susceptible population, the infection will again spread, potentially causing additional peaks. To characterize the point at which an epidemic is over, we therefore measured the slope of a best-fit line on the last 50 time steps and then measured the variance of the data contained within that time window. If each was sufficiently close to zero (I \< 0.005) we concluded that the dynamics had stopped. The beginning of the time window where these conditions were met was then treated as the end time of the epidemic.Table 3The differential equations used for each epidemic model, other than the previously stated SIR modelSEIROPSAIRTPSAIR$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{array}{c}\hfill \frac{dS(t)}{dt}=-c\beta S(t){P}_{SI}\hfill \\ {}\hfill \frac{dE(t)}{dt}=c\beta S(t){P}_{SI}-\kappa E(t)\hfill \\ {}\hfill \frac{dl(t)}{dt}=\kappa E(t)-\gamma I(t)\hfill \\ {}\hfill \frac{dR(t)}{dt}=\gamma I(t)\hfill \end{array} $$\end{document}$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{array}{c}\kern1em \frac{dS(t)}{dt}=-cS(t)\left({\beta}_A{P}_{SA}+{\beta}_I{P}_{SI}\right)\kern1em \\ {}\kern1em \frac{dA(t)}{dt}=cS(t)\left({\beta}_A{P}_{SA}+{\beta}_I{P}_{SI}\right)-{\gamma}_AA(t)\kern1em \\ {}\kern1em \frac{dI(t)}{dt}={\gamma}_AA(t)-{\gamma}_II(t)\kern1em \\ {}\kern1em \frac{dR(t)}{dt}={\gamma}_II(t)\kern1em \end{array} $$\end{document}$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{array}{c}\hfill \frac{dS(t)}{dt}=-cS(t)\left({\beta}_A{P}_{SA}+{\beta}_I{P}_{SI}\right)\hfill \\ {}\hfill \frac{dA(t)}{dt}=cS(t)\left({\beta}_A{P}_{SA}{P}_{AA}+{\beta}_I{P}_{SI}\left(1-{p}_{II}\right)\right)-{\gamma}_AA(t)\hfill \\ {}\hfill \frac{dI(t)}{dt}=cS(t)\left({\beta}_A{P}_{SA}\left(1-{p}_{AA}\right)+{\beta}_I{P}_{SI}{p}_{II}\right)-{\gamma}_II(t)\hfill \\ {}\hfill \frac{d{R}_A(t)}{dt}={\gamma}_AA(t)\hfill \\ {}\hfill \frac{d{R}_I(T)}{dt}={\gamma}_II(t)\hfill \end{array} $$\end{document}$ The instantaneous transfer of individuals from one compartment to another also serves to induce reaction timing that may not conform to data. In recognition of this, we supposed that the differential equations represent an expectation of outcomes over a population divided between three health classes. We found that in the absence of asymptomatically infectious individuals, disease-risk mitigation has three main effects on epidemiological trajectories: (1) fewer individuals get sick, (2) the peak level of infection is lower, and (3) the epidemic lasts longer. If asymptomatically infectious individuals are present, the effects are slightly different: (1b) fewer symptomatic infections occur, (2b) the peak level of symptomatic infection is less, and (3) the epidemic lasts longer. If latently infectious individuals are present, the effects can also include secondary peaks or waves. The effects of risk mitigation in an SIR model (susceptible, infectious, recovered individuals only, with no latency or asymptomatic illness) are illustrated in Fig. [1](#Fig1){ref-type="fig"}.Fig. 1The upper panel shows trajectories for the baseline set of epidemiological parameters for an SIR model with and without mitigation (dashed and solid lines, respectively). The susceptible and recovered trajectories are removed to show the trajectory of infectious individuals in the lower panel. The final epidemic size for this example is 33 % less with mitigation than with proportionate mixing, and the peak level of infection is 70 % less. Epidemic end of I(t) = 0.0005 is marked for the two models The caveat with respect to asymptomatically infectious individuals is an important one. We found that, in two cases, contact between asymptomatically infectious and susceptible individuals may be the primary source of infection during an epidemic. The first is where asymptomatically infectious individuals actively seek to avoid symptomatic individuals. The second is where infection rates due to contact with asymptomatic individuals are greater than infection rates due to contact with symptomatic individuals either because the probability of infection or because the time during which an asymptomatic is infectious is greater than that for symptomatically infectious individuals. In mathematical terms, this is a direct consequence of the fact that $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \frac{\partial {P}_{SA}}{\partial a}>0 $$\end{document}$. Thus, while both S(t) and A(t) seek to avoid I(t), we have that P~SI~ is decreasing while P~SA~ is increasing which may result in more infections by asymptomatic than by symptomatic individuals, *β* ~*A*~ *P* ~*SA*~ \> *β* ~*I*~ *P* ~*SI*~. Since the count of A(t) and I(t) is partly controlled by their rate of exit, e.g., γ~A~(t) and γ~I~(t), it is clear that this too plays a role in the overall effect of each class on creating new infections. Sensitivity of disease dynamics to epidemiological parameters {#Sec4} ============================================================= The lack of a closed-form solution for any of the standard models studied here and thus for a(t) implies that analytical results on the interactions between mixing decisions and disease dynamics are beyond our reach. We have, however, identified a few key characteristics of these interactions through an expansive parameter sweep. For all our simulations, we numerically solved the appropriate system of epidemiological differential equations (see Table [3](#Tab3){ref-type="table"}) but assumed a daily decision process for optimal avoidance levels (i.e., we use the discrete time analog of Eq. [2](#Equ2){ref-type=""}). In the narrowest sense, we are no longer using ordinary differential equations to model the system since we have introduced a time-dependent parameter in the form of a step function, *a*(*t*) = *a*(*t*). However, the discrete nature of the decision process is consistent with the continuous time model for the evolution of the disease. We varied β~X~, the conjoined effect of transmissibility and activity level for infectious class X; γ~X~, the infectious duration for class X; κ, the duration of exposure before the onset of infectiousness; B~x~, the relative benefit gained from belonging in non-symptomatic class X; ρa^2^, the costs of avoidance mitigation; *p* ~*XX*~ = 1 − *p* ~*XY*~, the probability of an infection caused by an infectious individual of type X resulting in an individual of the same type; and T, the planning horizon. We chose a set of baseline parameters to be β = 0.12 (with 12 contacts a day, this makes the probability of infection per contact 0.01), $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \gamma =\frac{1}{14} $$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \kappa =\frac{1}{7} $$\end{document}$, *B* ~*S*~ = *B* ~*A*~ = *B* ~*E*~ = *B* ~*R*~ = 1 (with B~I~ = 0), ρ = 0.001 (making the cost function for behavior change 0.001a^2^), p~AA~ = 0.034, P~II~ = 0.965, and T = 12 days. We varied β~X~ from 0.005 to 0.012, γ~X~ and κ each from $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \frac{1}{28} $$\end{document}$ to $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \frac{1}{7} $$\end{document}$, B~X~ from 0 to 90 (with 0 representing no cost of infection) ρ from 0 to 0.1, p~AA~ and p~II~ from 0 to 1, and T from 1 to 20 (Table [4](#Tab4){ref-type="table"}).Table 4Explanation of system parameters with nominal valuesParameter nameSymbolValueUnitsProbability of infection per contact with infective of type-xβ~x~0.01--Infectious period for infective of type-x$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \frac{1}{\gamma_x} $$\end{document}$14daysLatent period$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \frac{1}{\kappa } $$\end{document}$7daysProportion of type-y infections created by a type-x infectivep~xy~p~II~ = 0.965; p~AA~ = 0.034--Contact ratec12People/dayBenefit for being type-xB~x~B~S~ = B~E~ = B~R~ = B~A~ = 1; B~I~ = 0UtilsMitigation efforta----Cost of mitigationρa^2^ρ = 0.001UtilsPlanning horizonT12Days To see the effect of disease-risk mitigation on the epidemiology of infectious diseases, we considered the impact of variation in these parameters on several metrics including the proportion of the population that avoids infection, the duration of the epidemic, and peak infection levels. The relation between these metrics and the model parameters is intuitive and does not require further discussion. However, it is important to note that β and γ each has a unique impact on the metrics. Specifically, we found that the relationship between the sensitivities is not simply proportional, as shown in Fig. [2](#Fig2){ref-type="fig"}.Fig. 2In this figure, we show the effect that varying R~0~ has on the final epidemic size and time to peak infection levels. Notice that while β and γ each have the same qualitative effect, they have different effects quantitatively. For R~0~ ≈ 1.65, we observe a switch in which parameter produces the larger magnitude metric Figure [3](#Fig3){ref-type="fig"} shows infected classes for the baseline parameters of each model with and without private disease-risk mitigation. Risk mitigation results in 32, 30, 15, and 33 % less of the population infected for SIR, SEIR, OPSAIR, and TPSAIR, respectively. Note that infections due to contact with asymptomatic individuals account for only 6 % of all infections in the two-path model, but nearly 50 % in the one-path model (in part accounted for by the reactive population of immune, and previously asymptomatically infectious, individuals in the two-path model).Fig. 3A comparison of the infection levels for each model with and without mitigation The social cost of private disease-risk mitigation {#Sec5} ================================================== Now consider what these epidemiological impacts of private risk mitigation measures imply for the social cost of infectious disease outbreaks. The social net benefit from private disease-risk mitigation is simply the difference between the aggregate cost of illness under proportionate mixing and the aggregate cost of illness plus the aggregate cost of illness avoidance under preferential mixing. For simplicity, we assume that the benefits of being well (susceptible, asymptomatic, or recovered/immune) are the same for all individuals in society, *B*(*S*) = *B*(*E*) = *B*(*A*) = *B*(*R*) = *B* \> 0, and that those benefits are lost when an individual becomes symptomatically infected, *B*(*I*) = 0. For a population of size N, we take the present value of benefits net of the costs of illness at time t to be *ρ* ^*t*^ *B*(*N* − *αI* ~*t*~), where $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \rho =\frac{1}{1+\delta } $$\end{document}$ is the social discount factor, δ is the social discount rate, and α is a measure of the difference between the private and social cost of illness. We emphasize that we do not solve the social planner's problem and hence do not estimate the magnitude of the "wedge" between optimal public and private disease-risk mitigation strategies. Rather, we consider when private disease-risk mitigation generates net public benefits and how sensitive the social net benefit of private risk mitigation is to the difference between private and social costs of illness. The discrete social net benefit of private risk mitigation is defined as $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\displaystyle \sum_{t=0}{\rho}^t}B\left[\left({N}_{\alpha }{I}_t^{pref}\right)-\left(N-\alpha {I}_t^{prop}\right)-{C}_t\right]={\displaystyle \sum_{t=0}{\rho}^t}B\left[\alpha \left({I}_t^{prop}-{I}_t^{pref}\right)-{C}_t\right] $$\end{document}$. *I* ~*t*~^*pref*^ and *I* ~*t*~^*prop*^ are infection levels corresponding to privately optimal preferential and proportionate mixing, respectively, and C~t~ denotes the cost of illness avoidance---the cost of preferential mixing. That is, the social payoff to private risk mitigation is positive only if the discounted opportunity cost of illness plus the cost of mitigation under preferential mixing is less than the discounted opportunity cost of illness under proportionate mixing. For B \> 0, the optimal private response is always to mitigate risk (even if by a small amount). At each time, the private individual uses observations on the current state of an epidemic to calculate the discounted future risk of illness associated with different levels of avoidance behavior. They then select a level of avoidance behavior so as to equate the marginal cost of avoidance and the expected marginal reduction in disease risk. To explore this, we simulated the epidemiological implications of private risk mitigation over a range of social discount rates, avoidance costs, and social weights on the cost of illness using a discrete (daily) decision process. We began by considering the relationship between the private and social costs of illness. Setting the benefits gained during symptomatic infection to zero, the private cost of illness is exactly B. The social cost of illness is then *αB* =: *B* ~*S*~. We take three cases, but focus on only one of these. Case 1: social cost of illness = private cost of illness {#Sec6} -------------------------------------------------------- In the case that α = 1, society weighs the cost of illness the same as private individuals do. There are no externalities of either proportionate or preferential mixing. Intuitively, there is complete "agreement" between the social and private payoffs to mitigation so long as the social discount rate does not deviate from the private discount rate. However, if the social discount rate is significantly lower than the private discount, rate preferential mixing may not generate social benefits beyond proportionate mixing. We illustrate this via our discussion of case 3. Case 2: social cost of illness \> private cost of illness {#Sec7} --------------------------------------------------------- In the case that α \> 1, society weighs the cost of illness higher than the individual. This is the case considered most frequently in the literature, at least implicitly. If infected individuals consider only the cost of illness to themselves, and neglect the cost of illness to others with whom they come into contact, they will underweigh the true cost of illness. There are negative externalities of proportionate mixing. In this case too, if the social discount rate substantially deviates from the private discount rate, preferential mixing may not generate social benefits beyond proportionate mixing. However, for most nonzero social discount rates, preferential mixing generates positive social net benefits over proportionate mixing. Case 3: social cost of illness \< private cost of illness {#Sec8} --------------------------------------------------------- In the case that *α* ∈ \[0, 1) society discounts the private cost of illness. The implication is that private individuals overweigh the private cost of illness relative to the cost to society. This would occur if private disease-risk mitigation that conferred a benefit on the individual imposed non-disease-related costs on others that were not taken into account by the individual. Traditional cost of illness studies sum the direct costs of illness, such as medical care, and the indirect costs of illness, such as lost production (World Health Organization [@CR52]). If private disease-risk mitigation involves similar indirect costs---frictional productivity costs to employers, say---but these are neglected, individuals will underweigh the true cost of illness avoidance. We treat this as equivalent to overweighing the cost of illness. It implies negative externalities of preferential mixing. While the three cases are symmetric, in what follows, we focus on case 3. The social net benefit of private disease-risk mitigation depends on its impact on the aggregate cost of illness and illness avoidance relative to the proportionate mixing case. This is influenced (a) by the cost of avoidance behavior, (b) by its effect on prevalence and duration of the disease, and (c) by the weight attaching to present versus future costs. In the lower panel of Fig. [1](#Fig1){ref-type="fig"}, the "obvious" trade-off is the area between the two curves in \[0,800\] and \[180, ∞\] ≈ \[180, 800\]. Due to initial conditions chosen (low values for E~0~, I~0~, and/or A~0~), the initial response to the epidemic will be small (shown in Fig. [1](#Fig1){ref-type="fig"} from time \[0,20\]). Nevertheless, if the social discount rate is extremely high (the discount factor is extremely low), even this level of avoidance could be regarded as excessive from a social perspective if the benefits associated with a reduction in incidence within \[20,180\] are effectively weighted at zero. More generally, the lower the social discount rate relative to the private discount rate, the more proportionate mixing would be expected to dominate preferential mixing. For social discount rates close to zero (discount factors close to 1), the prolongation of the epidemic, and hence of private risk mitigation efforts, means that proportionate mixing is less socially costly than preferential mixing. This range of outcomes is illustrated in Figs. [4](#Fig4){ref-type="fig"}, [5](#Fig5){ref-type="fig"}, and [6](#Fig6){ref-type="fig"}. In all cases, hot colors indicate that the net social benefits under preferential mixing are less than the net social benefits under proportionate mixing. Cold colors indicate the reverse.Fig. 4At B = 0, there is no cost of illness, and proportionate mixing dominates preferential mixing for both private individuals and society. For B \> 0 and α = 0, the social weight on illness is zero, and proportionate mixing dominates preferential mixing for society. For most B \> 0 and α \> 0, however, preferential mixing dominates proportionate mixing for both private individuals and society Fig. 5The social value of preferential mixing over proportionate mixing for four values of *α* = {.03,.05,.09,.15} as a function of both the private cost of illness and the social daily discount factor. The cost of illness is measured on the Y-axis, and the daily discount factor on the X-axis. For α ≥ 0.2, preferential mixing dominates proportionate mixing at all social discount rates. The daily discount factor used to calculate private risk mitigation behavior, 0.9986, is indicated on each panel. This corresponds to an annual discount rate of around 5 % Fig. 6The social net benefits of private disease-risk mitigation where the social cost of illness is equal to 1 irrespective of the private cost of illness. The figure reports the difference between the social cost of illness under proportionate and preferential mixing for a range of daily social discount factors and private illness costs. Cold colors indicate that the social net benefits of preferential mixing are higher than the social net benefits of proportionate mixing. Hot colors indicate the reverse. The daily discount factor used to calculate private risk mitigation behavior, 0.9986, is indicated on each panel Figure [4](#Fig4){ref-type="fig"} focuses on the cost of illness. It shows the net social benefits of preferential relative to proportionate mixing under different private costs of illness and different social weights on those costs. It shows that society would gain more from proportionate than from preferential mixing where the private cost of illness is low (B \< 5 in our example) and/or where society places a low weight on the private cost of illness (α \< 0.2 in our example). Interestingly, if this region of "disagreement" is smaller, the larger the uncertainty that individuals face. Preferential mixing is most likely to yield social benefits for diseases with OPSAIR dynamics, where reactive individuals are uncertain about the disease state of others. Preferential mixing is least likely to yield social benefits for diseases with SIR dynamics, where there is complete certainty (in our models) about disease state. Diseases with TPSAIR and SEIR dynamics lie somewhere in between. As highlighted in the previous discussion, the social discount rate can also have a large impact on the social payoff to preferential mixing. In the following two figures, we explore the implications of the social (daily) discount rate on the relative net social benefit of preferential versus proportionate mixing. We show that the lower the social discount rate, the greater the range of costs of illness over which the society will benefit more from proportionate than preferential mixing. Figure [5](#Fig5){ref-type="fig"} shows values of B and ρ for which preferential mixing dominates proportionate mixing (cold colors), given four different social weights on the private cost of illness. The higher the social weight attaching to the private cost of illness, α, the higher the discount rate at which preferential mixing dominates proportionate mixing and the lower the cost of illness at which preferential mixing dominates proportionate mixing. Increases in the private cost of illness induce private disease-risk mitigation. It is intuitive that a rise in the social weight attaching to the private cost of illness, α, will also increase the likelihood that private disease-risk mitigation will be beneficial from a social perspective. Increases in the social discount rate relative to the private discount rate increases the relative social weight attaching to present over future costs of illness and illness avoidance. It is also intuitive that this will lower the social cost of prolonging epidemics. Figure [6](#Fig6){ref-type="fig"} considers a special case---where the social weight on the private cost of illness is *α* = 1/*B*. This implies that the private cost of illness is given the same weight by society regardless of whether that cost is high or low to individuals. Individuals are assumed to respond to the cost of illness as described above, and the net payoff to society relative to proportionate mixing is judged solely in terms of the strength of the private response. In this special case, we see a different relation between the social discount rate and the likelihood that private disease-risk mitigation will be socially beneficial. As we vary the (daily) social discount factor and the private cost of illness in this case, we see that if the private cost of illness is low, proportionate mixing always dominates preferential mixing. However, as the private cost of illness rises, whether proportionate or preferential mixing dominates depends on the discount rate. If the cost of illness is extremely high, and the social discount rate is less than or equal to the private discount rate, preferential mixing dominates proportionate mixing in all cases. If the cost of illness is intermediate, however, preferential mixing dominates proportionate mixing only at either very high or very low discount rates (*ρ* ≪ 1, *ρ* ≈ 1). An interesting feature of the OPSAIR plot is that proportionate mixing is favored over preferential mixing over a range of social discount rates regardless of private cost of illness. This is because above B ≈ 70, the privately optimal mitigation strategy is chosen such that P~SI~ = 0, meaning that all infections are caused by asymptomatic individuals. Therefore, these epidemics become insensitive to further increases in the private cost of illness.[2](#Fn2){ref-type="fn"} Discussion {#Sec9} ========== Private disease-risk mitigation has been a primary driver of trends in infectious disease epidemics over the last half century. While vaccination (either privately chosen or publically mandated) is the main form of risk mitigation for many diseases, private "distancing" strategies that reduce exposure to infectious classes are also common. Private effort to avoid infection depends on the costs and benefits of that effort for the individual. Whether private risk mitigation also benefits society depends on the way it changes epidemic dynamics. We find that private disease-risk mitigation always not only reduces the number falling ill but also extends the duration of epidemics. While the social net benefits of private disease-risk mitigation are, intuitively, increasing in the social cost of illness and decreasing in the cost of illness avoidance, they are also sensitive to the timing of these costs. Whether private risk mitigation generates social benefits relative to proportionate mixing---i.e., whether the social cost of illness and illness avoidance is lower under preferential than proportionate mixing---depends both on the social weight on the private cost of illness and on the social discount rate. We find that if the private cost of illness is high relative to the social cost of illness, then society may be better off allowing a disease to run its course without incurring the cost of disease avoidance. Similarly, if the social discount rate is very low relative to the private discount rate, society may be better off avoiding the prolongation of epidemics that comes with preferential mixing. If private and social discount rates and private and social costs are not substantially different, however, private disease-risk mitigation will ordinarily generate positive social benefits over proportionate mixing. Observed data on the DALY (disability-adjusted life year) losses to disease indicate that developed countries account for only 12 % of worldwide losses due to death and disability, but for more than 90 % of health expenditure. Infectious diseases of the type modeled here are overwhelmingly a problem of the developing world, accounting for more than 34 % of DALYs in developing countries, but less than 5 % in developed countries (World Health Organization [@CR52]; Lopez et al. [@CR38]; Murray and Lopez [@CR41]; Murray et al. [@CR42]). Our findings suggest that one explanation for this lies in differences in the private response to disease risk in developed and developing countries and hence to differences in the relative private costs of illness and illness avoidance. In countries where the private cost of illness avoidance is high relative to the cost of illness (i.e., the income forgone by the sick), we would expect very little private disease-risk mitigation. In countries where the private cost of illness avoidance is low relative to the cost of illness, we would expect the opposite. The greater incidence of infectious disease in developing countries may also, however, reflect differences in the social weight attached to the private costs of illness and illness avoidance. There are various reasons why the private and social costs of illness and illness avoidance might be expected to differ. The case most often explored in the literature is the public good nature of illness avoidance. Self-protection through vaccination, for example, is an impure public good. It not only confers benefits on the individual but also confers benefits on others (Andre et al. [@CR2]; Boulier et al. [@CR9]). Since the benefits to others will typically not be part of the private vaccination decision, however, the private value of vaccination will be less than the social value (Bauch and Earn [@CR6]). In such cases, public health authorities may have an interest in increasing the level of private disease-risk mitigation (Chowell et al. [@CR16]; Sandler [@CR48]). In this paper, we consider the case where the social cost of illness is less than the private cost. This would follow if private disease-risk mitigation conferred benefits on the individual but imposed costs on society that were not taken into account by the individual. Such costs might include, for example, the frictional and productivity losses of disease-risk-related absenteeism (Rice [@CR44]). If the social cost of the private response to shorter, more intense epidemics was less than the social cost of private responses to longer, less intense epidemics, governments might have an interest in decreasing the level of private disease-risk mitigation. We have not modeled the public health authority's problem in this paper and so have not identified the socially optimal level of private risk mitigation or the levers that might be used to generate that level of risk mitigation. Our focus has been on the epidemiological implications of private disease avoidance, and the cost of that to both private individuals and society. In setting up the public health authority's problem, it is the disease dynamics that come out of this process that are relevant. There is now a growing literature on the impact of private contact and mixing decisions on disease transmission (Gersovitz and Hammer [@CR24]; Gersovitz and Hammer [@CR23]; Barrett and Hoel [@CR5]; Fenichel et al. [@CR19]; Springborn et al. [@CR50]). From a public health perspective, an understanding of the decision process improves the capacity to predict the dynamics of epidemics, but it also opens up a new set of disease management options. Options that target either the contact rate (Auld [@CR3]; Kremer [@CR36]) or the probability that contact leads to infection (Geoffard and Philipson [@CR22]) have already been assessed. Options that target preferential mixing have not. The implications of the general approach for public health policy are now being explored (Fenichel [@CR17]), but the effect of differences between social and private assessments of the appropriate weight to be given to the cost of illness now and in the future has yet to be considered. This also has implications for the way that health authorities address the management of disease risks that span communities or countries. There is a perception that the control of infectious disease is a "weakest-link" public good---that the benefits of disease control to all are limited by the capacity of the weakest link in the chain (Sandler [@CR48]; Barrett [@CR4]). One implication is that effort to build the capacity of the weakest link should be sufficient to address the problem, and capacity building is in fact written in to both the International Health Regulations (World Health Organization [@CR51]) and the Sanitary and Phytosanitary Agreement (the main multilateral agreements governing responses to infectious human, animal, and plant diseases) (World Trade Organization [@CR53]). Our findings indicate the problem may be deeper than that. It may not be that poor communities and poor countries would do more if they could, but that doing more may not be in their own interest. Interventions that best serve the public interest in rich communities may not best serve the public interest in poor communities. To prevent the wider spread of outbreaks in poor countries, countries in which the opportunity cost of illness is high may need to incentivize risk mitigation in poor countries. The West African Ebola epidemic, to take a current example, is a case in point (Gostin et al. [@CR25]). In Morin et al. [@CR40], we have shown that this preferential mixing mitigation strategy is equivalent in outcome to contact volume reduction. We acknowledge that the individual could reduce contact volume in order to reduce risk and without asymptomatic infection, it would make no difference (Morin et al. [@CR40]). However, in this study, we only consider affinity-based preferential mixing.

{#sp1 .447} {#sp2 .448} {#sp3 .449} {#sp4 .450} {#sp5 .451} {#sp6 .452} {#sp7 .453} {#sp8 .454} {#sp9 .455} {#sp10 .456} {#sp11 .456-1} {#sp12 .457} {#sp13 .458} {#f1 .456-1}

Introduction ============ The morphogenesis and functional development of the mammalian central nervous system is regulated by the interaction of genes expressed at specific times and locations during development. The roles of these genes can be better understood by studying their spatial and temporal expression patterns. Most studies of gene expression pattern analysis use invasive methods to generate their data. As a result, little spatial information is preserved during sample processing, and pattern analysis is restricted to a local area. To relate the genetic information to the brain anatomy, the Mouse Biomedical Informatics Research Network (BIRN) project has generated a multimodal atlas for adult mouse brain and created an atlas interface (Mouse BIRN Atlasing Toolkit---MBAT) which can incorporate genetic information in an anatomical context (Boline et al., [@B2], <http://www.nbirn.net/tools/index.shtm>). The ultimate goal is to use the atlas as a framework for centralizing gene expression data collected using different methods and from separate laboratories, so the cross-community, cross-modality data correlation can be standardized. Currently, this atlas interface incorporates gene expression data obtained with microarray assay to the atlas space based on anatomical concepts. It also allows registration of *in situ* data of adult mouse brain to the orthogonal planes of adult brain atlas. In this study, we aim to extend the anatomical framework to a neonatal developmental stage and use it to incorporate data of gene expression assays shown in image formats, which are usually used to compensate for the low resolution of microarray assay. To study the contribution of genes in brain development, a high-resolution anatomical framework at an early developmental stage is essential to correlate the distribution of gene products and the cell type within each structure. Although the basic anatomical architecture of a mouse brain at postnatal day 0 (P0) is similar to an adult one, the neonatal brain is not simply a smaller version of the adult brain. Due to incomplete nerve myelineation and differentiated maturation for different brain structures at P0, some of the anatomical structures at P0 cannot easily be referenced from the atlas of adult brain. Currently available high-resolution brain atlases for early developmental stages only provide a limited number of sections and structural delineations (Jacobowitz and Abbott, [@B8]; Schambra et al., [@B16]). In addition, since these neonatal atlases use paper format and individual atlas planes are not spatially in register, it is difficult to use them to integrate and present the information acquired from other sources into the atlas space. Previously, we defined a standard atlas space with stereotaxic coordinates for the neonatal (P0) C57BL/6J mouse brain using MRI brain volumes (Lee et al., [@B9]). Although this atlas represents a native space of brain volumes and provides a 3D anatomical framework, it does not provide cellular scale resolution. Here, we extend past efforts by incorporating high-resolution Nissl-stained data, which reveals cytoarchitecture of brain structures, into the previously developed P0 digital atlas. As images with detailed anatomy are co-registered to the standard space, high-resolution anatomical space can be indexed using the stereotaxic coordinates. The neonatal atlas therefore provides a region-specific framework which allows data association based on anatomical and/or spatial relations. The serviceability of the high-resolution anatomical framework of the atlas can be illustrated by incorporating gene expression data generated using invasive staining methods, such as *in situ* hybridization and immunohistochemistry staining, to the atlas space. Gene expression analyses using these methods are performed by staining thin brain slices; therefore, the results of single assays are restricted to a single plane. In order to differentiate between different gene products, one or a few genes are assayed in a single data image. It is labor intensive to perform sample preparation throughout the whole brain, and many laboratories focus their experiments on specific regions. Thus, results from single experiments usually provide only a regional picture of gene activity. These assays reveal the precise anatomical location where the gene product is distributed but reconstruction of the gene expression patterns using multiple assays compensates for the restrictions due to the staining methods, and greatly enhances the significance of single studies. This can be achieved by combining gene expression datasets in a common atlas framework. Co-displaying the data with brain anatomy also allows one to establish relationships between the differentiated distribution pattern of gene products and specific anatomical systems, potentially providing better data realization and interpretation. Individual images can be related to the 3D atlas space with a plane equation that computes the atlas brain slice corresponding to the experimental data. Since the functions of anatomical structures are the result of the interaction of several genes, associating high-resolution gene expression data with high-resolution anatomical models would provide a better insight into how gene products contribute to functional differentiation during early brain development. Materials and Methods ===================== Constructing the high-resolution P0 atlas ----------------------------------------- ### Preparation of histological volume Brain slices of 25 μm thickness were collected coronally from a C57BL/6J mouse on the first day after birth (P0). These slices were stained, aligned, and warped to a reference MRI volume using the same protocols described in MacKenzie-Graham et al., [@B13]. The registration procedures employed corrected the slice distortions introduced during sample preparation. The 3D histological volume was reconstructed using 150 registered histological sections 50 μm apart. All animals were housed and treated in accordance with the UCLA Animal Research Committee guidelines. ### Labeling anatomical structures Tissue labeling of the histological image volume was delineated on the images after 2D non-linear distortion were corrected. Labeling was done using BrainSuite2 (Shattuck et al., [@B17], <http://www.loni.ucla.edu/Software>). The delineation was based on coronal sections, aided by consultation of the orthogonal planes. Primary references were the prenatal mouse brain atlases of Jacobowitz et al., [@B8] and Schambra et al., [@B16]. Because the boundaries of most structures were left undefined in these primary references, delineations were inferred from the cytoarchitectural atlas of adult mouse brain (Paxinos and Franklin, [@B15]) based on cell arrangement, and relative position to the surrounding structure and 3D morphology. The nomenclature and abbreviation used were primarily based on Paxinos and Franklin\'s mouse brain atlas, thus remaining consistent with those used in the predefined atlas-based database (MacKenzie-Graham et al., [@B13]). If the structures in the anatomical database were hard to discriminate in the stained slices, they were labeled as their parent structures based on the hierarchical relations defined in the adult mouse brain (Paxinos and Franklin, [@B15]) and Brain Architecture Management System (Bota et al., [@B3], BAMS <http://brancusi.usc.edu/bkms>). The relationships between anatomical structures were organized hierarchically and modeled using BrainGraph (MacKenzie-Graham et al., [@B12]). ### Registering histological image volume to the standard space The reconstructed histological brain volume was re-sampled to the standard atlas space with the registration protocol described in Lee et al., [@B9]. In brief, the histological brain was first co-registered to the MRI-based atlas using a 12-parameter transformation (Woods et al., [@B19]). An anatomical label volume was reconstructed from the delineations based on the defined anatomical hierarchical relations such that it had the same 13 features defined in the low-resolution MRI-based atlas. Feature-based warping was then performed by maximizing the mutual information between the anatomical labels of the two brain volumes (Leow et al., [@B11]). Manage gene expression data using the atlas framework ----------------------------------------------------- Two XML-format documents are used to manage the gene expression data. An in-house atlas visualization interface uses these documents to display and manage data in the atlas space. The "contours" document is used to specify the areas with enriched gene expression in each image, and the "DataSet" document is used to group image data assayed for the same target (e.g., brain slices from the same batch of assay). Data management using XML format allows flexible data modeling in various atlas interfaces and will facilitate data sharing across different information systems. ### "Contours" document This document describes the spatial locations of the regions of interest (ROI, e.g., areas with positive signal in gene expression data) for each image and the expression level of these areas within that image. Each ROI is represented with a "contour" element, and the "contour" elements with the same ROI properties are organized under a "contours" element. The property of these grouped ROI (e.g., expression pattern, level) is indicated with the "ID" attribute of the "contours" element. The location of ROI in the source image is specified with pixel coordinates of the corresponding contour points and the coordinate values of each contour are retained using multiple "vertex" elements. In summary, the "contours" document represents the gene expression patterns in an image with the following schema organization:  The "Volume-Source" schema specifies the image file presenting these contours, and "Space-type" indicates whether pixel or real coordinates are used in the "vertex" schema. Note that the vertex coordinates of this document report the locations of the ROI on the experimental image. Therefore, documentation and delineation of the gene expression pattern is independent of the data-to-atlas spatial relations. Once the data-to-atlas spatial relationship is identified, the coordinates of the vertices may be transformed to the 3D atlas coordinates. ### "DataSet" document The "DataSet" document was designed for organizing the data images with a common property (e.g., assays for the same gene) and retaining the linear data-to-atlas transformations for these images. It uses multiple "Data" elements to reference the image files in the dataset, each of which indicates the absolute file path of an individual image and specifies a management plane that form the atlas. It also contains a "Signal-Contour" element that indicates the file path of the "contours" document used to represent the ROI in the image. The organization of the schemas is:  The "Data" element uses the following six elements to specify the data-to-atlas spatial relationship for the experiment image it represents. The "Atlas-Section-Center," "Atlas-Translation," and "Atlas-Rotation" elements describe the oblique plane of the atlas. The "Input-Translation" and "Input-Transform-Matrix" indicate the image translation, rotation, and scaling that linearly register the experimental image to the oblique plane of the atlas. The "Input-Section-Center" element is used to identify the orthogonal plane if the experimental data is a 3D volume. These parameters are sufficient to compute a unique management plane in the atlas to house an experimental image. These transformation schemas can also be reused to register the data to a different version (e.g., higher resolution) of atlases for the mouse brain of the same developmental stage. Reconstruct 3D gene expression patterns from multiple co-registered data ------------------------------------------------------------------------ ### Gene expression data We chose the published *in situ* hybridization data generated by Gray et al., [@B7] from the MGI database (<http://www.informatics.jax.org>) to evaluate this atlas-based management framework. This group used *in situ* hybridization to reveal the expression pattern of 1445 transcription factors in the brains of developing mice, and the data are available at both the MGI database and the Mahoney database (<http://mahoney.chip.org/mahoney>). The selected images are the gene assays at P0 that present regionally restricted gene expression patterns. The dataset for each gene expression assay contains 10--20 brain slices sectioned in the coronal orientation (rostral-caudal). After acquiring the raw images from the database, each of the brain slices was cropped and saved as individual 8-bit image file. For visualization purposes, non-brain regions were masked. The actual dimensions of the brain slices were not specified by the MGI database and were determined based on visual comparison with the atlas brain when co-displayed on the same canvas. The atlas oblique plane corresponding to each slice was determined by an experienced neuroanatomist based on visual inspection of the images and comparison of the structures on the atlas planes. This was assisted by the use of MBAT which allows for viewing an arbitrary digital section from a 3D brain volume. The selected plane is the "management plane" for all derivative data from this image. Each brain slice was further warped to its corresponding atlas section using a feature-based registration method such that the atlas anatomical labels properly mapped to the registered image. The features used for registration were the recognizable anatomical structures and the external boundaries of the image and its management atlas plane. The displacement fields used to warp the images are derived by propagating the force field used to bring the structural boundaries into register using a 2D heat diffusion model. An xml-formatted "DataSet" document was used to manage a group of registered images from any single gene assay. The file location of each of these images, as well as the rotation, translation, and scaling of the atlas used to generate the corresponding plane for each experimental slice were recorded in this document. ### Delineating gene expression image The MGI database annotated the gene expression level in large brain structures, but in most cases the expression was restricted to substructures within this area. To examine expression at a more detailed level, we only delineated the boundaries of areas that showed positive gene expression signals. The boundaries of the areas with enriched gene expression were delineated using in-house software. Blocks with the same range of intensity were outlined with single contours to differentiate the expression level. Contours of these delineations were assigned with numbers from 50 to 250 with steps of 50 to reflect the relative expression level of that region, with a higher number representing a stronger signal among the same batch of assays. If the expression pattern within an area was clustered (cellular specific), the recognizable boundaries of this area were outlined with a single contour, rather than delineating each cluster. The contours drawn on the original images were transformed using the same displacement fields used to warp the source images. The location of the contours in the 2D images was saved in a separate "contours" xml-format document. The "Signal-Contour" of the "Data" element that represents the corresponding registered image in the "DataSet" document refers to this transformed "contours" document. ### Mapping gene expression to the brain anatomy To map gene expression pattern to atlas coordinates, the vertex coordinates in the contours document are transformed to the atlas space using the registration parameters of that image. Polygons formed by these transformed vertices are modeled as 3D ROI objects and used to represent the gene expression pattern in atlas space. This allows us to examine the relation of anatomical structures to the gene expression patterns by calculating the pixels of each anatomical label within each ROI. Two ratios were computed for each anatomical structure:  Results ======= The high-resolution atlas ------------------------- The histological volume for the image base of the atlas has a resolution of 6.6 × 50 × 6.6 μm^3^ (1024 × 150 × 1024 voxels) (Figure [1](#F1){ref-type="fig"}). Plane generated by arbitrarily sectioning from the histological volume still provides high-resolution anatomy. In total, 145 delineated structures and 169 anatomical relations were defined for this histological volume (Figure [2](#F2){ref-type="fig"}). Hierarchy of these anatomical labels is modeled as a BrainGraph representation (Figure [2](#F2){ref-type="fig"}B). {#F1} {#F2} A sub-sampled version of the reconstructed histological volume (13.2 × 50 × 13.3 μm^3^) and the anatomical label volume were co-registered to the defined MRI atlas space (Figure [3](#F3){ref-type="fig"}). Global registration has been previously validated in Lee et al., [@B9]. Briefly, the anatomical labels of the MRI volumes map onto the co-registered histological volume with high accuracy in the major structures (Figure [3](#F3){ref-type="fig"}, right). The previously defined coordinate system is now applied to the high-resolution brain anatomy using the co-registered label volume. ![**Co-registration of the multimodal neonatal atlas**. Horizontal, coronal, and sagittal views of the Nissl atlas sectioned at position (−1.34, 0.77, −0.406) of the stereotaxic coordinate system defined in Lee et al., [@B9]. (Left column) The high-resolution histological volume and the anatomical delineations were co-registered to the P0 atlas space defined in Lee et al., [@B9]. (Right column) The anatomical structures from the MRI atlas brain (target for registration) are shown overlaid on the co-registered histological brain volume.](fninf-01-006-g003){#F3} The co-registered multimodal atlas as well as the original anatomical delineations are available for download at <http://www.loni.ucla.edu/Atlases/>. The visualization application SHIVA developed by LONI (<http://www.loni.ucla.edu/Software>) and MBAT developed by Mouse BIRN (Boline et al., [@B2]) can interactively describe the anatomical structures when viewing the atlas package consisting of the image volume, anatomical volume, and BrainGraph document. Comparing this atlas with the adult mouse brain (MacKenzie-Graham et al., [@B13]), two neonatal structures stood out as having significant differences from their adult counterpart: the subventricular zone (SVZ) and the cerebellum (CB). While only a small area of SVZ is left in the adult olfactory area, it is easily visible along the lateral ventricle throughout the anterior forebrain at the age of P0 (Figure [3](#F3){ref-type="fig"}, left). In contrast, the neonatal CB is unfoliated and occupies a much smaller portion in the brain, resulting in a large morphological difference from the adult. Because these areas go through such extreme changes over this period, it is logical to examine them for alterations of gene activity during brain development. This neonatal atlas offers a regional anatomical framework specific to the early developmental stage and is a valuable resource for these types of studies. Managing gene expression data in the atlas space ------------------------------------------------ The Lef1 gene assay dataset is used to illustrate our atlas-based management (Figure [4](#F4){ref-type="fig"}). The original data of the *in situ* hybridization images were acquired from the same batch of assay (assay T04972110a-Rostral from <http://www.informatics.jax.org/searches/image.cgi?10055>). Each image was registered to the corresponding atlas plane and the overlays of the registered pairs are shown in Figure [4](#F4){ref-type="fig"}. Co-displaying the management planes allows for visual realization for the spatial relationship of these data in the atlas space (Figure [5](#F5){ref-type="fig"}). {#F4} {ref-type="fig"}. The rest of the management planes are distributed between these planes and do not intersect within the brain. Left: lateral view. Right: dorsal view. The opaque blocks on the management planes are the registered brain slices of the in situ hybridization data. Color indices for the 3D objects of the anatomical structures---cyan, olfactory bulbs; red, cerebrum; yellow, diencephalons; green, midbrain-hindbrain; blue, cerebellum.](fninf-01-006-g005){#F5} Based on data-to-atlas spatial relationship specific for an image, the contours that outline the areas with enriched gene expression on this image can be transformed into the atlas space (Figure [4](#F4){ref-type="fig"}B and left of Figure [6](#F6){ref-type="fig"}A). Co-localization of the ROI objects from the same dataset reconstructs the 3D expression for a gene in the atlas space and allows for relating this pattern to multiple anatomical structures (left of Figures Figure [6](#F6){ref-type="fig"}A and [6](#F6){ref-type="fig"}B). In this example, the expression of Lef1 gene seems to be mainly distributed in the superior colliculus and the ventral-lateral part of the thalamus. {ref-type="fig"}**B** were converted into 3D objects and displayed in the atlas space together with a sagittal slice of the high-resolution atlas. Right: distribution of ROI objects derived from all contours in Figure [4](#F4){ref-type="fig"}**A** is shown on a surface model of the atlas. (***B***) Distribution of all ROI objects and two coronal sections of the atlas. Left: frontal view. Right: back view.](fninf-01-006-g006){#F6} Since the ROI objects only represent the spatial information of a gene expression pattern, this spatial information needs to associate with the anatomical concepts in order to perform further cross-modality data analysis. The ratios described in the methods were used to determine the relation between the atlas structures and the ROI objects. Together, these can give an indication of this anatomical association when dealing with an imperfect cross-modality image registration. The local anatomical ratio represents the anatomical composition of the ROI objects, and the inclusion ratio indicates a probability that an anatomical structure shows enriched gene expression. For example, in Figure [4](#F4){ref-type="fig"}B, the two types of ratios for superior colliculus suggested that it is very likely the area within contour 1 (Figure [4](#F4){ref-type="fig"}) corresponds to this anatomical structure (left in Figure [7](#F7){ref-type="fig"}A and left in Figure [7](#F7){ref-type="fig"}B). {ref-type="fig"}**B**. The anatomical trees are the hierarchical summaries of the local anatomical composition within contours in Figure [4](#F4){ref-type="fig"}**B**. The values appended to each of the structure abbreviations are the local anatomical ratios for each structure within this ROI. The highlighted structures were those with the largest local anatomical ratio within this ROI. (***B***) Structures in midbrain-hindbrain (MB-HB) (Left) and diencephalons (DI) (Right) are organized based on anatomical hierarchy. Inclusion ratios for each anatomical structure in this management plane are shown next to the structure abbreviations. The highlighted structures are those with significant ratios in both (***A***) and (***B***).](fninf-01-006-g007){#F7} In the case that more than one significant structures map onto the range of ROI, the anatomical hierarchical relations can provide the most appropriate common parental structure to associate with this ROI object. For example, the ventral nuclear group (VNG) is the most significant structure for both of the ROI objects derived from contour 2 and 3 and is consistently enclosed by them in Figure [4](#F4){ref-type="fig"}B (Figure [7](#F7){ref-type="fig"}A, right two panels). However, this structure only partially occupies either of these two ROI objects (Figure [7](#F7){ref-type="fig"}B, right panel) and is therefore not sufficient to represent them. Hence, instead of the VNG, it would be more appropriate to have these two ROI objects associate with its closer parental structure---the thalamus. The expression pattern throughout the structures can be further examined using the inclusion ratio computed from all planes in the dataset. For example, the values computed for the Lef1 gene dataset demonstrate that it is highly expressed in superior colliculus and several structures in lateral and ventral thalamus such as dorsal lateral geniculate nucleus (DLG), ventral lateral geniculate nucleus (VLG), lateral nuclear group (LNG), and VNG (Figure [8](#F8){ref-type="fig"}). Association between the ROI object and the anatomical structures relates the ROI objects to several levels of the anatomical hierarchy, thus allowing for the extension of data retrieval and comparison at multiple anatomical scales (Boline et al., [@B2]; Martone et al., [@B14]). {#F8} The digital atlas can provide a management framework that facilitates cross-dataset comparison by using multiple "DataSet" documents. To simplify visualization of multiple genes and gene levels, ROIs coming from different "DataSet" documents can be assigned different colors (Figure [9](#F9){ref-type="fig"}). Presenting these ROI in the atlas space provides a clearer view of which genes are exclusively expressed (Figure [9](#F9){ref-type="fig"}A) or co-expressed (Figure [9](#F9){ref-type="fig"}B). Even though technique difficulties prevent the delineation of several small sub-structures, for example, layers in the cerebral cortex, co-registering the gene expression patterns in the atlas may still give additional information. For instance, Figure [9](#F9){ref-type="fig"}B shows that while the ROI objects for the Mef2C and Neurod2 are both consistently associated with the cerebral cortex, the Mef2C gene is more restricted to the upper layers. {#F9} Discussions =========== In this study, we provide a high-resolution atlas of neonatal mouse brain that is registered to the space defined by the averaged shape of native brains. We used *in situ* hybridization data to demonstrate the applicability of the atlas as a framework for data normalization and reconstruction of the gene expression pattern from individual experiments. Co-registering the gene expression data in the atlas space will allow for the analysis of data in an anatomical specific framework. Significant developing structures for postnatal study ----------------------------------------------------- We observed that the SVZ and the CB of the neonate show significant differences from their adult counterparts, and they continue to differentiate postnatally. Thus, these are obvious areas to examine gene expression profile changes over the course of development. The anatomical models of these structures provided by this atlas are valuable resources for examining the developmental events in these structures. The SVZ has been known as a pool of neuronal progenitor cells during embryonic neurogenesis from which the preneural cells continuously move to the cortical plate during corticogenesis. This structure begins to decrease dramatically during postnatal development and little remains in adulthood (Gates et al., [@B5]). This structure has been identified as the harbor for neural stem cells, as cells in this region retain their capability to differentiate into neuronal cells after the brain is fully developed (Alvarez-Buylla and Garcia-Verdugo, [@B1]). To study how this developing event evolves, it is necessary to examine the co-localization of gene product, extracellular matrix, and cell type within a small region. The high-resolution digital map thus provides a region-specific framework for centralizing data from different assays. In CB, the migration of Purkinje cells has not yet completed at the time of birth, and cerebellar granule cells have just begun massive clonal expansion and differentiation. The CB at this development stage is thus unfoliated and occupies a much smaller portion in brain volume when compared with the adult. The relative ease of postnatal experimentation and well-known functional compartments make the CB a good model to study the structural pattern formation in mammalian central nervous system. Morphological differences in CB as well as in other areas result in a complicated spatial relationship between the developing and the mature brain. This issue can be handled by using consistent nomenclature and the anatomical terms for the atlases of different developmental stages. By associating with a common anatomical model, experimental data from different developmental stages can be compared with each other. The digital atlas hence offers a dynamic framework to study these postnatal developmental events across different time courses. Incorporating high-resolution gene expression data -------------------------------------------------- The BIRN project has established a microarray database and relates these data to the adult brain atlas using anatomical concepts. Microarray assays are efficient in obtaining differentiated expression profiles of all the candidate genes. However, since it is difficult to dissect an unstained structure from the brain during sample preparation, the data generated with this method is related to brain anatomy on a coarse level. Many studies use this method for systematically screening for specific genes, and then identify their exact location on a cellular scale using *in situ* or immuno-histochemistry staining. Unlike microarray data management, which can be related to anatomical concepts, management of 2D data requires a plane equation that relates the image pixels to the atlas. Several gene expression databases for mouse brain manage their image data by registering them to an orthogonal plane of the brain atlas that has defined stereotaxic coordinates (e.g., GenePaint, Carson et al., [@B4]; Allen Brain Atlas, Lein et al., [@B10]). Unlike adult mouse brain, no external reference points can be used to consistently orient the position of the developing brain. As a result, brain slices cannot be sampled with a fixed angle from an external landmark during sample collection. Since no canonical atlas plates are available to house data from oblique planes, current databases simply relate the gene expression in early development stages to brain anatomy by literally describing the expression patterns in the brain structures (e.g., MGI, Gray et al., [@B7]; GENSAT, Gong et al., [@B6]). By providing a digital neonatal atlas, data of image format now can be associated with the anatomical infrastructure. Sections collected without spatial framework can be handled in the atlas space with linear transformation. Registering the gene expression pattern to an anatomical framework provides a more comprehensive picture than purely text descriptions (Figures [6](#F6){ref-type="fig"} and [9](#F9){ref-type="fig"}). The high-resolution atlas allows the gene expression pattern in the images to be registered to a specified anatomical structure and is capable of containing the information with the anatomical models at multiple hierarchical levels (Figures [7](#F7){ref-type="fig"} and [8](#F8){ref-type="fig"}). Data from a plane thus can be transferred to an anatomical base via this structural association, and can be correlated to both image and microarray data from the database using spatial transformations and anatomical concepts. Cross-community, cross-modality data analysis can then be performed with a proper design in dataset categorization and representation for data obtained from various sources. Such information integration will potentially save a great deal of time and resources by allowing the reuse of similar data from different studies. Data analysis with reconstructed gene expression ------------------------------------------------ The data models of gene expression patterns discussed in this manuscript only reflect the patterns from their source images and therefore are planar. The anatomical information derived from reconstructing 3D gene expression (Figures [8](#F8){ref-type="fig"}) implies the areas of gene activity in each anatomical structure. The distribution of expression within an anatomical structure can be inferred by extrapolating the planar ROI objects associated with that structure between image slices. This allows for statistical correlation of the expression pattern of genes with the distribution patterns of other gene products. In addition to comparing the pattern of different genes, it is also possible to model the area with different cell type with the ROI objects and perform the same spatial analysis as the gene expression data. This in turn facilitates the examination of the interaction between gene products and their contributions to the function of the anatomical structures. Conclusion ========== The digital neonatal atlas provides a high-resolution anatomical reference for neonatal brain and also a framework for normalizing gene expression data at an early developmental stage. The Mouse BIRN project has developed tools that allow data about specific genes to be retrieved from several public databases (e.g., GeneNetwork, Allen Brain Atlas, and GENSAT) and compared using an atlas interface. Incorporating data in the atlas space allows individual researchers to use the atlas as mediator to interact with a variety of atlas-associated public resources. Data from separate data sources can be compared and correlated with each other, thus facilitating cross-modality, cross-community data analysis for individual studies. Conflict of Interest Statement ============================== This research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. This work was supported by NIH Grant U24 RR021760 to the Mouse Biomedical Informatics Research Network (BIRN, <http://www.nbirn.net>), which is funded by the National Center for Research Resources at the National Institutes of Health (NIH). List of the abbreviations of anatomical structures described in the high-resolution P0 atlas. Index[\*](#appfn1){ref-type="table-fn"} Abbreviation Full name of the structure ----------------------------------------- -------------------- -----------------------------------------------------------------------   1 Cb Cerebellum   2 AVP Anteroventral periventricular nucleus   3 Cx Cerebral cortex   4 LHA Lateral hypothalamic area   5 3V 3rd ventricle   6 lfbs Lateral forebrain bundle system   7 ZI Zona incerta   8 LDTg Laterodorsal tegmental nucleus   9 Amy Amygdala  10 RET Reticular formation  11 SPV Spinal nucleus of the trigeminal  12 DMH Dorsomedial nucleus of the hypothalamus  13 STN Subthalamic nucleus  14 VI Abducens nucleus  15 LV Lateral ventricle  16 MH Medial habenular nucleus  17 LH Lateral habenula  18 Pi Pineal gland  19 XII Hypoglossal nucleus  20 DCN Dorsal column nuclei  21 MD Mediodorsal thalamic nucleus  22 IO Inferior olivary complex  23 PRT Pretectal region  24 CN Cochlear nuclei  25 AQ Aqueduct of Sylvius  26 PG Pontine gray  27 SC Superior colliculus  28 PAG Periaqueductal gray  29 MOB Olfactory bulb  30 ns Nigrostriatal bundle  31 Olf Olfactory system  32 PCG Pontine central gray  33 RN Red nucleus  34 MSX Medial septum complex  35 IP Interpeduncular nucleus  36 VNC Vestibular nuclei  37 SI Substantia innominata  38 III Oculomotor nucleus + EW  39 AOB Accessory olfactory bulb  40 V Motor nucleus of the trigeminal  41 RA Raphe nuclei  42 PB Parabrachial nucleus  43 DLG Dorsal lateral geniculate nucleus  44 PSV Principal sensory nucleus of the trigeminal  45 4V 4th ventricle  46 VII Facial nucleus  47 IC Inferior colliculus  48 DTg Dorsal tegmental nucleus  49 AON Anterior olfactory nucleus  50 SVZ Subventricular zone  51 ic Internal capsule  52 NTS Nucleus of the solitary tract  53 MG Medial geniculate nucleus  55 AP Area postrema  56 MEV Mesencephalic nucleus of the trigeminal  57 SN Substantia nigra  58 CGB Central gray of the brain  59 NTB Nucleus of the trapezoid body  60 mfbs Medial forebrain bundle system  61 CSN Cranial & Spinal Nerves  62 mbp Middle cerebellar peduncles  63 VTA Ventral tegmental area  64 SCh Suprachiasmatic nucleus  65 cg Cingulum bundle  66 VMH Ventromedial nucleus of the hypothalamus  67 mtg Mammillotegmental tract  68 SOC Superior olivary complex  69 ml Medial lemniscus  70 CnF Cuneiform nucleus  71 Arc Arcuate hypothalamic nucleus  72 NLL Nucleus of the lateral lemniscus  73 8n Vestibulocochlear nerve  74 Pit Pituitary gland  75 mlf Medial longitudinal fasciculus  76 PH Posterior hypothalamic area  77 Pa Paraventricular hypothalamic nucleus  78 BST Bed nucleus of the stria terminalis  79 AH Anterior hypothalamic area  80 NB Nucleus of the brachium of the inferior colliculus  81 Pir + EP Piriform cortex + endopiriform nucleus  82 lo Lateral olfactory tract  83 LPO Lateral preoptic area  84 SHi Septohippocampal nucleus  85 PSC Posterior septum complex  86 LSX Lateral septum complex  87 VLG Ventral lateral geniculate nucleus  88 Su5 Supratrigeminal nucleus  89 cc Corpus callosum  90 Vn Trigeminal nerve  91 fxs Fornix system  92 st Stria terminalis  93 fi Fimbria of hippocampus  95 CC Central canal 100 CPu Caudate putamen 101 cbp Cerebellar peduncles 102 2n Optic nerve 103 PRP Nucleus prepositus 104 icp Inferior cerebellar peduncle 105 act Anterior commissure temporal limb 106 hbc Habenular commissure 107 aco Anterior commissure olfactory limb 110 sm Stria medularis 111 GP Globus pallidus 112 PR Prerubral field 113 VP Ventral pallidum 114 Tu Olfactory tubercle 115 esp Extrapyramidal fiber systems 116 VNG Ventral nuclear group 117 mt Mammillothalamic tract 118 ATN Anterior group of thalamus 119 Po Posterior thalamic nuclear group 120 fr Fasciculus retroflexus 123 PVZ Periventricular zone of the hypothalamus 125 MZ Medial zone of the hypothalamus 126 MCPO Magnocellular preoptic nucleus 127 ING Intralaminar thalamic nuclear group 128 eml External medullary lamina 129 PF Parafascicular thalamic nucleus 130 MPA Medial preoptic area 131 RtTg Reticulotegmental nucleus of the pons 132 RET(C-MO) Reticular core of medulla 133 MBO Mammillary body 134 py Pyramidal tract 135 pv Periventricular fiber system 136 CI Claustrum 137 LNG Lateral nuclear group 138 IV Trochlear nucleus 139 Rt Reticular thalamic nucleus 140 Acb Accumbens nucleus 144 IPAC Interstitial nucleus of the posterior limb of the anterior commissure 148 VTN Ventral tegmental nucleus 150 APN Anterior pretectal nucleus 153 SubG Subgeniculate nucleus 164 PRN Pontine reticular nucleus 170 MB-HB Midbrain-hindbrain 180 DI Interbrain 199 RPF Retroparafascicular nucleus 200 pc Posterior commissure 201 NPC Nucleus of the posterior commissure 202 PRC Precommissural nucleus 215 ac Anterior commissure 216 vn Vomeronasal nerve 218 IIIn Oculomotor nerve 219 Hy Hypothalamus 222 Dk Nucleus of Darkschewitsch 223 INC Interstitial nucleus of Cajal 224 MB-HB(Mot, Face) Midbrain-hindbrain, motor, face 225 MB-HB(Mot, EP) Midbrain-hindbrain, motor, extrapyramidal 228 dc Dorsal columns 229 cst Corticospinal tract 230 MB-HB(Sens-G/Visc) Midbrain-hindbrain, sensory, gustatory-visceral 231 MB-HB(Mot, Jaw) Midbrain-hindbrain, motor, jaw 232 BG Basal ganglia 233 HIP Hippocampal region 234 MB-HB(Sens-Vest) Midbrain-Hindbrain, sensory, vestibular 235 Sep Septum 236 DpMe Deep mesencephalic nucleus 237 MB-HB(Mot) Midbrain-hindbrain, motor 238 MB-HB(Sens) Midbrain-hindbrain, sensory 239 MB-HB(Sens-Aud) Midbrain-hindbrain, sensory, auditory 240 MB-HB(Sens-Vis) Midbrain-hindbrain, sensory, visual 241 Pre-Post-Cb Pre-post cerebellar nuclei 242 MB-HB(Mot, Tongue) Midbrain-hindbrain, motor, face 243 Th Thalamus 244 MB-HB(Mot, Eye) Midbrain-hindbrain, motor, eye 245 MB-HB(Sens-ss) Midbrain-hindbrain, sensory, somatosensory 246 RET(C) Reticular core 247 cpd Cerebral peduncle 249 CNS Central nervous system 250 MNG Midline thalamic nuclear group 251 scp Superior cerebellar peduncle 252 LZ Lateral zone of the hypothalamus 253 CH Cerebrum 255 Br Br ^\*^ *The intensity index of anatomical label volume for the histological-based delineations*. [^1]: Edited by: Jan G. Bjaalie, International Neuroinformatics Coordination Facility, Karolinska Institutet, Sweden; University of Oslo, Norway [^2]: Reviewed by: Trygve B. Leergaard, University of Oslo, Norway; Ilya Zaslavsky, Spatial Information Systems Laboratory, University of California San Diego, USA

All relevant data are within the paper. Introduction {#sec001} ============ The *pugilist* (*pug*) gene of *Drosophila melanogaster* encodes an enzyme with three activities in tetrahydrofolate metabolism. Null mutations of *pug* cause a subtle, transitory and recessive reduction in pteridine pigmentation in the eyes of newly-eclosed flies. The dominant *pug*^D^ mutation reduces pigment throughout the eye, with an effect that is especially noticeable in a background where only pteridines are present, e.g. *vermillion* (*v*). The eyes of such flies (*v; pug*^*D*^*/+*) exhibit an unusual ring of pigmentation around the periphery of the eye and a few scattered spots of pigment in the center of the eye, but are otherwise completely white-eyed. Ommochrome pigmentation is also affected by *pug*^*D*^, but to a lesser degree \[[@pone.0151377.ref001]\]. The protein encoded by *pug* has also been implicated in the response to Parkinson\'s disease, aging and oxidative stress, immunity \[[@pone.0151377.ref002]--[@pone.0151377.ref006]\]. The *pug*^*D*^ mutation was created at one of the junctions of an X-ray induced inversion on chromosome *3*. The mutant allele consists of three parts assembled from three different genomic locations. The upstream regulatory elements and the first approximately one-third of the coding region, including the start codon, are derived from the *pug* gene, which normally encodes the tri-functional methylenetetrahydrofolate dehydrogenase (MTHFD, located at 86C) and generates products that participate in 1-carbon transfer reactions. This activity is prominently involved in purine synthesis, and since pteridine biosynthesis is based on GTP, this provides one possible connection between *pug*^*D*^ and pigment deposition (reviewed \[[@pone.0151377.ref007]\]). The portion of the gene that remains in the *pug*^*D*^ allele should encode approximately two-thirds of the methylene tetrahydrofolate dehydrogenase and cyclohydrolase domain \[[@pone.0151377.ref008]\]. It lacks the ATP-binding site and the entire formyltetrahydrofolate synthetase domain. It seems unlikely that *pug*^*D*^ retains any of its normal enzymatic functions. The second *pug*^*D*^ coding segment is approximately one kb of highly repetitive sequence, consisting almost entirely of iterations of the satellite sequence AGAGAGA (although oriented with TCTCTCT on the sense strand in *pug*^*D*^). These almost certainly derive from centric heterochromatin where such repeats are abundant \[[@pone.0151377.ref009]\], and were captured at the inversion junction during repair of the breakpoint that generated the inversion. Transcription and translation of both coding segments is necessary for the mutant phenotype \[[@pone.0151377.ref001]\]. The translated repeats constitute about two-thirds of the pug^D^ protein. The third portion consists of *rab7* oriented in the opposite direction, and does not contribute to the *pug*^*D*^ phenotype. Because *pug*^*D*^ does not simply reproduce the phenotype of a *pug* null allele, we consider *pug*^*D*^ to be a gain-of-function, rather than loss-of-function allele. The strong phenotype produced by *pug*^*D*^ relative to *pug*-*null* alleles, even in the presence of two copies of *pug*^*+*^, also indicates that it is not merely an antimorph. For these reasons, we consider the *pug*^D^ mutation to be neomorphic. This does not exclude the possibility that it also interferes with the action of *pug*^*+*^, but the Pug^D^ protein is clearly doing more than that. The ring pattern of pigmentation produced by *pug*^*D*^ is certainly the most unusual aspect of its pug^D^ phenotype. A very limited number of mutations have been identified that produce a ring or partial ring pattern of pigmentation. The cause for such patterns has not been identified in any of these cases. In this study we explore the genesis of ring pigment patterns in *pug*^*D*^. We have observed that *pug*^*D*^ transgenes show a large degree of phenotypic variation, primarily in the thickness of the pigmented ring. The phenotypes range from those that are essentially identical to the original *pug*^*D*^, to weak alleles that are indistinguishable from wildtype \[[@pone.0151377.ref001]\]. The work we report here was carried out to determine the basis for the phenotypic variation, and to investigate models for the ring pattern of pigmentation. Materials and Methods {#sec002} ===================== Fly Stocks {#sec003} ---------- Mutations and chromosomes not described here are described by Lindsley and Zimm \[[@pone.0151377.ref010]\]. Flies carrying a *GMR-wg* construct were provided by Andrew Tomlinson \[[@pone.0151377.ref011]\]. Flies carrying the *Gla*^*1*^ mutation were provided by Konrad Basler \[[@pone.0151377.ref012]\]. Flies with the construct p*P\[w*^*+*^, *GMRP35\]* were provided by Bruce Hay \[[@pone.0151377.ref013]\]. Flies with the construct p*P\[sev-wg\]* were provided by Kenneth Cadigan \[[@pone.0151377.ref014]\]. *Glazed* and *GMR*-wg Crosses {#sec004} ----------------------------- To assess the effect of ubiquitous *wingless* expression in the eye we crossed *v; pug*^*D*^*/TM6* females to either *GMR-wg/TM7* males, to *Gla*^*1*^ *Bc bw/CyO* males \[[@pone.0151377.ref012]\] or to *+/Gla*^*1*^ *Bc Elp* males (in which the + homolog carried an unidentified deficiency). The *v* sons with the desired *wg* allele were examined for pigmentation. Eye Injection Experiments {#sec005} ------------------------- Pupae of *v; pug*^*D*^ */TM3*, *Ser* were picked at stages ranging from 0 to 2 days into pupation. They were mounted on slides with double stick tape. A glass needle similar to that used for embryo injections was employed to inject solutions into the eyes of these pupae. Injected pupae were left on slides placed on moist paper. The solutions injected were: (1) 0.85 mM guanosine in pH2.0 HCl solution (guanosine); (2) pH2.0 HCl solution (HCl); (3) 0.5X PBS solution (PBS). *pug*^*D*^ Transgenes {#sec006} --------------------- The construction of *pug*^*D*^ transgenes in the vector *P\[X97\]* has been previously described \[[@pone.0151377.ref001]\]. The construct used for most experiments reported here carried the 3.4 kb *Kpn*I---*Bam*HI *pug*^*D*^ genomic fragment. This fragment contains only the promoter for the two short transcripts identified for this gene (pug-RA, pug-RC), and contains no coding sequences from the adjacent gene CG14863 (Flybase: \[[@pone.0151377.ref015]\]). The *pug*^*S*^ gene was a derivative of the original *pug*^*D*^ clone in which the AGAGAGA repeats were spontaneously reduced to 300 bp during overnight bacterial culturing. *P* Element Transposition Screen {#sec007} -------------------------------- The *X* chromosome insertion 1A carries the *P* element construct of p*P\[X97*, *pug*^*D*^*\]* which has the *v*^*+*^ marker \[[@pone.0151377.ref001]\]. Females of the genotype *v P\[X97*, *pug*^*D*^*\]1A; ry* were crossed to *w*^*1118*^*; Sb Δ2--3 (99B)/TM6*, *Ubx* males, which carry the *Δ2--3 (99B)* insertion to provide P transposase \[[@pone.0151377.ref016]\]. Individual *v P\[X97*, *pug*^*D*^*\]1A; Sb Δ2--3 (99B)/ry* males were mated to two or three *v; ry* females. *Sb*^*+*^ male offspring that are *v*^*+*^ (with or without any pigmentation defect) were retained. From the offspring of a single male, one fertile male was kept among the *v*^*+*^ males with the same pigmentation pattern. This ensured that only males with independent *P* transposition events to autosomes were retained. Lines were established from these males. Each line was assigned with a number. All the v^+^ lines with normal pigmentation were later discarded except three: 46A, 50 and 74. All the lines were mapped by segregation from dominantly marked autosomes. No insertion on chromosome *4* or *Y* was recovered. FLP-Mediated DNA Mobilization {#sec008} ----------------------------- Mobilization experiments were done as described \[[@pone.0151377.ref017]\]. Briefly, for insertions on chromosome *3*, females that were *w*^*1118*^ *P\[ry*^*+*^, *70FLP\]3F; P\[RS3r\]/TM6* were mated to *v; P\[v*^*+*^*\]/TM6* males, with 4--5 pairs per vial. *P\[RS3r\]* is the *FRT*-bearing target site element that carries the 3' portion of the *white* gene, and P*\[v*^*+*^*\]* represents the *P\[X97\]* element carrying a derivative of the *pug*^*D*^ gene and a functional 5' portion of *white* (along with a non-functional 3' portion of *white*). Parents were transferred every 2 days. The vials were heat shocked for one hour at 37˚C immediately after each transfer. Male progeny that were *w*^*1118*^ *P\[ry*^*+*^, *70FLP\]3F; P\[RS3r\]/P\[v*^*+*^*\]* were mated to *w*^*1118*^ females with one male and 2--3 females per vial. Progeny of this cross exhibiting eye pigment, indicating that a *w*^*+*^ gene had been reconstructed by integration of the *pug-*bearing extrachromosomal circle at the target site, were retained. For donor insertions on chromosome *2*, individual *w*^*1118*^ *P\[ry*^*+*^, *ß2tFLP\]; P\[v*^*+*^*\]/+; P\[RS3r\]/+* males were mated to 2--3 *w*^*1118*^ females per male. Lines were established from pigmented progeny. For each *RS3r* target site, one of the recovered integration events was tested to confirm that it was the result of FLP-mediated mobilization to the target *FRT*. First, we simply outcrossed heterozygous females to ensure that *pug*^*D*^ and *w*^*+*^ could not be separated by recombination. We found no cases in which this occurred. A second test was based on the fact that *pug*^*D*^ and part of *w*^*+*^ are flanked by two direct *FRT*s after donor integration. When FLP is expressed in the soma, by crossing to flies carrying the heat shock inducible *70FLP* gene and heat-shocking offspring during third instar, mosaicism for both *pug*^*D*^ and *w*^*+*^ should be observed. Adult eyes were scored for *pug*^*D*^ mosaicism in a *v* background, or for *white* mosaicism in a *w* background. In previous work \[[@pone.0151377.ref017]\], only 1 of 46 events recovered by the scheme we used here was not specifically targeted, and that one event did not produce white mosaicism with FLP. In the present work, all the tested lines exhibited mosaicism for both phenotypes in response to FLP expression, supporting the conclusion that they resulted from FLP-mediated mobilization to the target *FRT*. Each mobilized *pug*^*D*^ gene was named for both the donor insertion and the target site. For example, the *pug4A-2* transgene was the result of mobilizing the *pug*^*D*^ gene in donor line 4A to the target site of *RS3r-2*. Characterization of AGAGAGA Repeat Length {#sec009} ----------------------------------------- A combination of Southern blot and PCR analyses were done to determine the genomic structure of the *pug*^*D*^ genes mobilized to *RS3r-2*. Genomic DNA was isolated as described \[[@pone.0151377.ref001]\]. Southern blot and PCR analyses were performed according to standard protocols. The probe was labeled and detected using the Genius non-radioactive kit (Roche). The exposed film was scanned, and the gel was straightened electronically in Photoshop for analysis. PCR was used to amplify regions of *pug*^D^ upstream and downstream of the AGAGAGA repeats to determine their lengths. The primers used in this work were: 1, w14178d (5'-TGTGTGTTTGGCCGAAGTAT-3'); 2, TestH3; 3, 3'pug-Bst; 4, Test1; 5: rab73'; and 6: break3' (5'-cgcgatgtgttcactttgct-3'). The sequences of the other four primers have been presented previously \[[@pone.0151377.ref001]\]. Results {#sec010} ======= The Effect of Chromosomal Position on the *Pug* Phenotype {#sec011} --------------------------------------------------------- Independent insertions of a *pug*^*D*^ transgene carried within a *P* element produce varied eye pigment phenotypes, differing primarily in the diameter of the central region that lacks pteridines. This suggests that chromosomal position effects influence the expression of *pug*^*D*^. However, *P* element transposition can lead to internal deletions of the *P* element, especially when the element contains direct repeats \[[@pone.0151377.ref018],[@pone.0151377.ref019]\]. Since the *pug*^*D*^ gene contains \~140 tandem repeats of AGAGAGA we suspected that transposing a *P* element carrying *pug*^*D*^ might generate deletions within the AGAGAGA repeats at a high frequency. To test the effects of varying chromosomal position on expression of the pug phenotype without altering the AGAGAGA repeats, we moved a *pug*^D^ transgene (called 4A) to 10 target sites along chromosome *3* using FLP-mediated DNA mobilization \[[@pone.0151377.ref017]\]. We recovered at least one integration event at each site, with multiple independent integration events recovered for most sites. Mobilization frequencies to each *RS3r* site ranged from 1.1% to 7.5%, and are reported in [Table 1](#pone.0151377.t001){ref-type="table"}. 10.1371/journal.pone.0151377.t001 ###### *RS3r* sites and mobilization frequencies. {#pone.0151377.t001g} Location[^a^](#t001fn001){ref-type="table-fn"} Integration frequency[^b^](#t001fn002){ref-type="table-fn"} ------------ ------------------------------------------------ ------------------------------------------------------------- RS3r-29 *3L*, 64A, 7.8 m.u. 2.8% (2/71) RS3r-22B *3L*, 65F, 22.8 m.u. 2.3% (2/86) RS3r-23 *3L*, 68C, 36.9 m.u. 3.8% (3/78) RS3r-2 *3L*, 75C-D, 45.0 m.u. 3.4% (3/87) RS3r-35 *3L*, 77A, 46.6 m.u. 1.1% (1/88) Centromere 80D-81F, 47 m.u. RS3r-3 *3R*, 82C, 47.1 m.u. 3.8% (3/80) RS3r-30 *3R*, 84A, 47.5 m.u. 3.8% (4/106) RS3r-41 *3R*, 85E, 49.2 m.u. 6.3% (5/79) RS3r-20B *3R*, 94A, 75.6 m.u. 4.1% (4/98) RS3r-25 *3R*, 100C, 105 m.u. 7.6% (6/79) ^*a*^ For each location a chromosome arm is indicated (except the centromere), which is followed by a cytological location and a map location in map units (m.u.). The locations of *RS3r-2* and *RS3r-3* were determined by cytology. The locations of the other eight *RS3r* sites were estimated by recombination tests based on maps in Lindsley and Zimm \[[@pone.0151377.ref010]\]. ^*b*^ The integration frequency of a *pug*^*D*^ transgene to an *RS3r* site was measured as the percentage of males that produced at least one at least one offspring with a target site insertion. Since FLP carries out reciprocal conservative recombination between *FRT*s without the participation of any host DNA exchange or repair factors, the *pug*^D^ transgene should be unchanged by the process of mobilization, and any variation in phenotype should reflect chromosomal position effects. The *pug*^*D*^ phenotype was strong at most *RS3r* sites, but did show some variation ([Fig 1](#pone.0151377.g001){ref-type="fig"}), having the strongest phenotype at *RS3r-25* and the weakest at *RS3r-35*. Within any line, there is little variation in phenotype and the pictures provided are highly representative of all individuals within that line. Five independent *pug4A-25* events and one *pug4A-35* event were chosen for examination by Southern blotting. The size of the *pug* transgene in these six mobilized lines was unchanged (not shown). In addition, independent integration events of the 4A *pug*^*D*^ donor at the same target site all showed identical pug phenotypes (not shown). Since the size of the *pug*^*D*^ transgene is the same at all sites, we conclude that chromosomal position can strongly modify the phenotypic expression of a *pug*^*D*^ transgene. {#pone.0151377.g001} The Diameter of the Central Region that Lacks Pteridines Varies with *pug*^*D*^ Copy Number {#sec012} ------------------------------------------------------------------------------------------- The original *pug*^*D*^ mutation was not viable in homozygous condition, but since it arose on a chromosome also carrying the recessive lethal *Stubble* mutation it was not known whether *pug*^*D*^ itself was viable in homozygous condition. When we recombined *Stubble* away from the *pug*^*D*^ chromosome we found that *pug*^*D*^*/pug*^*D*^ homozygotes do occasionally survive, though at a rate well below Mendelian expectation. They exhibit even less pigment than *pug*^*D*^*/+* flies, lacking pteridine pigment entirely in several places around the periphery of the eye ([Fig 2](#pone.0151377.g002){ref-type="fig"}). This suggested that quantitative variation in *pug*^*D*^ might be manifest as variation in the diameter of the pteridine-lacking central region of the eye. To test this, we combined the *pug4A-35* and *pug4A-41* insertions. Together, these two transgenes exhibited a greater reduction in pteridine pigmentation than either transgene alone, approaching the phenotype produced by the original *pug*^*D*^ mutation ([Fig 2](#pone.0151377.g002){ref-type="fig"}). Therefore, the diameter of the central zone that lacks pterdines is a quantitative reflection of *pug*^*D*^ expression. {#pone.0151377.g002} Variation in AGAGAGA Repeat Length Causes Variation in *pug*^*D*^ Phenotype {#sec013} --------------------------------------------------------------------------- This suggests that some of the variation in phenotype produced by random transgene insertions could result from variation in the AGAGAGA repeats. We previously showed that a *pug* transgene with only 300 bp of AGAGAGA repeats was ineffective at producing the *pug*^*D*^ phenotype \[[@pone.0151377.ref001]\].To test this, we made use of *P* element transposition, and its tendency to alter internal repeated segments \[[@pone.0151377.ref018],[@pone.0151377.ref019]\], to produce transgene insertions with varied phenotypes and potentially varied AGAGAGA repeat length. The *P\[X97*, *pug*^*D*^*\]*1A, marked by *v*^*+*^, was transposed from the *X* chromosome to autosomal sites with the Δ2-3(99B) transposase source \[[@pone.0151377.ref016]\]. We established lines of 23 independent insertions on chromosome *2*, and 24 on chromosome *3*, for further analyses. The majority of the *v*^*+*^ male offspring also had normal pigmentation; that is, they were pug^+^. The loss of the *pug*^*D*^ phenotype in the transposed copies could be a result of a change in the *pug*^*D*^ transgene, or it might reflect chromosomal position effects, or both. To eliminate the influence of variable position effects we used FLP to mobilize sixteen of these elements, representing a broad range of phenotypes, to the *RS3r-2* site at 75C-D. We also mobilized a *pug*^*S*^ transgene, which has only 300 bp of AGAGAGA repeats \[[@pone.0151377.ref001]\], to this site. All were examined in a *v* background to eliminate ommochrome pigmentation. When FLP was used to move independent *pug* insertions to the *RS3r-2* target site we observed a wide range of pigmentation patterns in the eye, indicating that new variants had arisen among the transposed *pug* genes ([Fig 3](#pone.0151377.g003){ref-type="fig"} and others not shown). The phenotypes range from completely normal (*pug*^*s*^*-2*) to eyes that are similar to the original *pug*^*D*^ (*pug4A-2*; [Fig 1](#pone.0151377.g001){ref-type="fig"}). We also observed that the phenotype of a *pug*^*D*^ transgene could change substantially upon mobilization to *RS3r-2*. In some cases, transgenes that were nearly *pug*^*+*^ at their original site expressed a strong *pug*^*D*^ phenotype at the *RS3r-2* site (for instance, *pug77*), or vice-versa (*pug39A*; [Fig 3](#pone.0151377.g003){ref-type="fig"}). As mentioned above, there was little variation within any particular line. {#pone.0151377.g003} Molecular analysis of the variant *pug* genes showed that AGAGAGA repeat lengths that differed substantially from the \~1.0 kb found in the original *pug*^*D*^ allele resulted in a reduced phenotype. [Fig 4](#pone.0151377.g004){ref-type="fig"} depicts the structure of the *pug*^*D*^ region after mobilization to the *RS3r* target site. We deduced the length of the repetitive stretch by first using Southern blots to measure the size of the *Eco*RI---*Bam*HI fragment. A sample of these data are shown in [Fig 5](#pone.0151377.g005){ref-type="fig"}. PCR analyses were then done to measure the sizes of the regions upstream and downstream of the repeats, using primer pairs 1,2 and 1,3 upstream and 4,6 and 5,6 downstream (locations diagramed in [Fig 4](#pone.0151377.g004){ref-type="fig"}; results not shown). With this combination we can estimate the length of the region between primers 3 and 4, which consists mostly of the AGAGAGA repeats. Any change in this region is attributed to a change in the length of the AGAGAGA repeats. (It was not possible to simply use PCR with spanning primers to determine the length of the AGAGAGA repetitive section because we were unable to perform PCR across such repeats that are longer than 300 bp.) The estimated AGAGAGA repeat length for each insertion is given in [Table 2](#pone.0151377.t002){ref-type="table"}. ![The structure of *pug*^*D*^ in the vector *P\[X97\]*.\ The 3.4 kb *Eco*RI-*Bam*HI (R, B) fragment of *pug*^*D*^ includes three parts: the 5' portion from *pug*^*+*^ (*MTHFD*); the one kb AGAGAGA repeat; and, the putative 3' UTR from the *rab7* region. The arrow indicates the direction of transcription. The solid boxes to either side represent sequences from the *white* gene of the vector. The probe used for Southern analyses is indicated. Arrow heads represent primers used for PCR. See [Materials and Methods](#sec002){ref-type="sec"} for details.](pone.0151377.g004){#pone.0151377.g004} {ref-type="fig"}. Molecular weight markers are shown in the left and right lanes, with sizes indicated in kb. The *pug* transgene line number is indicated for each lane, *e*.*g*., the first lane is *pug7-2*. The 5 kb band in all sample lanes is the endogenous *rab7* band.](pone.0151377.g005){#pone.0151377.g005} 10.1371/journal.pone.0151377.t002 ###### *P* element lines for mobilization to *RS3r-2*. {#pone.0151377.t002g} Line Chromosomal location Origin Pigmented progeny Length of the AGAGAGA repeat[^c^](#t002fn003){ref-type="table-fn"} ------------ ---------------------- --------------- ------------------------------------------ -------------------------------------------------------------------- 4A 2 Injection \+ \~1000 bp 7 2 Transposition \+ \~1000 bp 16B 2 Transposition \+ \~800 bp 27C 3 Transposition \+ \~1000 bp 30B 2 Transposition \+ \~600 bp 31 2 Transposition \+ \~1000 bp 39A 2 Transposition \+ \~650 bp 45B 3 Transposition \+ \~1000 bp 46 2 Transposition \+[^a^](#t002fn001){ref-type="table-fn"} ND 47 3 Transposition \+ \~950--1000 bp 50 3 Transposition \-[^b^](#t002fn002){ref-type="table-fn"} ND 51 3 Transposition \+ \~1400 bp 54 3 Transposition \+ \~1000 bp 70 3 Transposition \+ \~1000 bp 74 3 Transposition \-[^b^](#t002fn002){ref-type="table-fn"} ND 77 2 Transposition \+ \~1300 bp 80C 3 Transposition \+ \~1000 bp *pug*^*s*^ 2 Injection \+ \~300 bp ^*a*^ The pigmented progeny may be the result of an illegitimate integration event (see [RESULTS](#sec010){ref-type="sec"}). ^*b*^: A somatic assay did not reveal successful integration in the eye. Not characterized further. (see [RESULTS](#sec010){ref-type="sec"}). ^*c*^: The lengths of the GAGA repeat were estimated by Southern blot and PCR analyses (see [RESULTS](#sec010){ref-type="sec"}). ND, not done. The *pug4A* transgene appears to have \~1 kb of AGAGAGA repeats, the same repeat length as the original *pug*^*D*^ allele. It has a strong pug^D^ phenotype at most, though not all, sites, and serves as one standard for comparison ([Fig 1](#pone.0151377.g001){ref-type="fig"}). The other standard is the *pug*^*S*^ gene, carrying 300 bp of AGAGAGA repeats, derived by loss of a portion of the repetitive sequence during growth in bacteria \[[@pone.0151377.ref001]\]. Judged by Southern blotting, lines 4A, 7, 31, 45B, 47, 54, and 70 all have the full length repeats. However, PCR analysis with primers 4, 6 and 5,6 indicated that a non-repeat portion of *pug47* is slightly larger than the others (not shown), meaning that its repetitive portion must be slightly shorter to produce an *Eco*RI---*Bam*HI fragment of the same size. However, the difference is small, perhaps 50--100 bp, and our Southern blot is probably not capable of indicating whether there is such a small difference in the size of the *Eco*RI---*Bam*HI band. These lines, and others that have approximately the same length of repeats as *pug4A*, produced a phenotype that was indistinguishable from *pug4A* when inserted at the *RS3r-2* target site ([Table 2](#pone.0151377.t002){ref-type="table"}; [Fig 1](#pone.0151377.g001){ref-type="fig"}). In contrast, the lines with clear alterations in repeat length, both longer and shorter, have weaker phenotypes ([Fig 3](#pone.0151377.g003){ref-type="fig"}). The *pug51* and *pug77* genes both have repeat segments longer than the original *pug*^*D*^ allele, yet produce weaker phenotypes: especially *pug51*, which has the longest repeat segment of \~1400 bp. Four of the alleles, *16B*, *30B*, *39A* and *pug*^*s*^, have AGAGAGA repeats that are significantly shorter than the 1 kb found in *pug*^*D*^, and they all have a reduced phenotype. Among these four shortened alleles, *16B* has the longest repeat length (\~800 bp) and produces the strongest phenotype of that group; *pug*^*s*^ has the shortest repeat (300 bp), and produces the weakest phenotype, which is essentially pug^+^. Finally, the *pug80C* transgene appears to have a complex rearrangement, which we did not completely characterize, and it produces a strong phenotype. The correlation between repeat length and phenotype was not perfect: *pug30B* has a slightly shorter repeat than *pug39A* (600 bp *versus* 650 bp), yet, the former produces a slightly stronger phenotype (not shown). It is possible that variation in the sequence of the non-repetitive portion may account for such minor differences. Notwithstanding this particular example, the overall conclusion is that the strength of *pug*^*D*^ alleles increases as the repeat length increases from \~300 bp, reaching its strongest effect at a length of \~ 1000 bp, and then decreasing as the repeat length extends beyond \~1000 bp. The variation in phenotype was similar to that which could be produced by altering *pug*^*D*^ transgene dosage, indicating that repeat length variants effect a quantitative, not qualitative, reduction in the effect of the *pug*^*D*^ alleles. The Ring Pattern of Pigment in *pug*^*D*^ {#sec014} ----------------------------------------- We previously proposed a model in which the pattern of *pug*^*D*^ pigmentation is determined by the architecture of the compound eye (discussed further below). However, *pug*^*D*^ mosaics can be observed, as a consequence of position effect variegation \[[@pone.0151377.ref001]\], or by FLP-mediated mosaicism (results not shown). This indicates that the action of *pug*^*D*^ is cell autonomous, at least to a large degree, and suggests an alternative model: that the ring of pigment in *pug*^*D*^ could result from repression of *pug*^*D*^ expression in cells at the periphery of the eye. A prominent candidate for regulating such expression is *wingless* (*wg*). Tomlinson \[[@pone.0151377.ref011]\] summarized several features that distinguish the periphery of the eye from the center: the most interesting aspect with respect to *pug* is that there are pigment cells surrounding the eye which are not associated with photoreceptors. Tomlinson called this the Pigment Rim (PR). Cells of the PR apparently correspond to the cells that retain pteridine pigmentation in a *v; pug*^*D*^*/+* fly \[[@pone.0151377.ref001]\]. Formation of the PR results from the death of photoreceptor cells in the peripheral ommatidia, and this death is directed by *wg* expression \[[@pone.0151377.ref011],[@pone.0151377.ref014]\]. The expression pattern of *wg* is highly reminiscent of the pigment pattern seen in *v; pug*^*D*^*/+* flies, with expression in the adult head limited, for the most part, to cells of the head capsule that surround the eye (see [Fig 2D](#pone.0151377.g002){ref-type="fig"} in Tomlinson \[[@pone.0151377.ref011]\]). These cells are directly in contact with the PR cells. Thus, *wg* is a strong candidate for controlling the expression of *pug*^*D*^. To test whether *pug*^*D*^ responds to *wg* we generated flies carrying *pug*^*D*^ and *GMR-wg*, a *P* element construct that expresses *wg* throughout the eye during late development. The effect of this expression is to cause the death of photoreceptor cells throughout the eye, producing an adult eye that is composed almost entirely of pigment cells, much like the PR of a normal eye \[[@pone.0151377.ref011]\]. If *pug*^*D*^ is repressed by *wg*, we expect that the ubiquitous *wg* expression in the eye produced by *GMR-wg* should repress *pug*^*D*^ throughout the eye, allowing pteridine pigmentation throughout. This was not seen ([Fig 6](#pone.0151377.g006){ref-type="fig"}). Instead, the *pug*^*D*^ ring of pigment was still apparent. The complete absence of bristles in these eyes confirms that *wg* was expressed throughout the eye. We also combined *pug*^*D*^ with the mutation *Gla*^*1*^, a dominant allele of *wg* with expression throughout the eye, much like *GMR-wg* \[[@pone.0151377.ref012]\], and still saw the *pug*^D^ ring phenotype ([Fig 6](#pone.0151377.g006){ref-type="fig"}). Finally, a *sev-wg* construct \[[@pone.0151377.ref014]\] was tested with the same results (not shown). The expression of *wg* throughout the eye did not suppress the pug^D^ phenotype in any case. We conclude that the pug^D^ phenotype is not a response to *wg* signalling. {#pone.0151377.g006} There was some expansion of the pigmented region in some of the double mutant individuals. We suspect this is a consequence of disruption of the structure of the eye, and not a specific repression of *pug*^*D*^, because similar results can be produced by injury to the eye (see below). Though we did not count offspring, the *Gla*^*1*^ *pug*^*D*^ or *GMR-wg pug*^*D*^ double mutants were quite rare in our crosses. Many pharate adults were produced, and when dissected from their pupal cases, they were found to be the missing double mutants. These pharates appeared to be fully differentiated adults, but were extremely dehydrated, usually having completely flattened abdomens. A common characteristic of these flies was that at least one of the eyes had a large black scab-like mass on its surface. Evidence of this may be seen in the infrequent survivors, for example the fly shown at the upper left of [Fig 6](#pone.0151377.g006){ref-type="fig"}. We suspect that the cause of lethality in the double mutants is that the eye bursts from turgor pressure, allowing internal fluids to leak out and cause lethal dehydration. We have previously noted that the eyes of *pug*^*D*^ flies seem somewhat fragile \[[@pone.0151377.ref001]\]. In combination with *Gla*^*1*^, the integrity of the eye is apparently compromised to such a degree that it is often lethal. We also carried out the reciprocal test of preventing photoreceptor cell death by using *GMR-P35*, a construct that expresses the anti-apoptotic P35 protein from baculovirus throughout the eye and blocks virtually all cell death \[[@pone.0151377.ref013],[@pone.0151377.ref020]\]. The ring of pigment was still evident ([Fig 7](#pone.0151377.g007){ref-type="fig"}). In fact, the *pug*^*D*^ *GMR-P35* flies even show an increase in cells with pigment. This is likely the result of disrupting the regular ommatidial structure in these eyes (see below). ![The effect of baculovirus P35 expression on *pug*^*D*^.\ A, B: eyes from *v; pug*^*D*^*/+* flies. C, D: eyes from *v; pug*^*D*^ */P\[GMR-P35\]* flies.](pone.0151377.g007){#pone.0151377.g007} Although *wg* did not affect *pug*^*D*^, there may yet be some relationship between *wg* and pteridine pigment patterning. We noticed that in *v; Gla/+* or *v; GMR-wg/+*, there is a small region of the eye, typically in the ventral portion, that is unpigmented. It is not simply the case that this portion of the eye is devoid of pigmented cells, because in a *v/v*^*+*^*; Gla/+* or *v/v*^*+*^*; GMR-wg/+* fly this part of the eye is pigmented, though it is clearly a different color than the remainder of the eye, no doubt reflecting the loss of pteridine but not ommochrome pigmentation in this region (not shown). Eye Pigmentation Can Be Partially Restored by Direct Eye Injection {#sec015} ------------------------------------------------------------------ Previously, we proposed the model that pigment cells at the periphery of the eye are able to take up purines from neighboring non-eye tissue, providing them with a critical precursor for pigment synthesis, and that *pug*^*D*^ interferes with transport to more interior cells of the eye. To test this, we designed an experiment to deliver purines directly to cells in the center of the eye by injection into the eyes of early pupae. It has been shown that pupal eyes incubated *in vitro* can take up guanine derivatives for pigment synthesis \[[@pone.0151377.ref021]\]. We predicted that if we could supply purine precursors to the center of the eye by injection, we might be able to restore pigmentation. We previously determined that the developmental period in which *pug*^*D*^ is most effective in pigment elimination is during the first two days of pupal development \[[@pone.0151377.ref001]\]. The best time to restore pigmentation by injection is likely to be within these 2 days. We injected the eyes of *v; pug*^*D*^ */+* pupae at stages ranging from 0 to 2 days into pupation with solutions of either guanosine, PBS, or an acidified aqueous solution. For all three solutions, areas of extra pigmentation in the eye were often observed ([Fig 8](#pone.0151377.g008){ref-type="fig"}). {#pone.0151377.g008} These experiments show that the center of a *pug*^*D*^ eye still has the ability to make pteridine pigment. This is consistent with the conclusion that *pug*^*D*^ does not eliminate pigmentation by killing pigment cells. It is unlikely that a common chemical compound in all three solutions has been provided by the injection which allowed the synthesis of pteridines around the injected areas. Instead, the occurrence of injury to the eye seems to be sufficient to produce pigment in new areas of the eye. Discussion {#sec016} ========== The variation that we observed with various *pug*^*D*^ transgenes is most obviously a variation in thickness of the pteridine-pigmented ring with a complementary variation in the size of the central region that lacks pteridine pigmentation. A similar effect can be produced by varying *pug*^*D*^ gene copy number. In other words, the size of the unpigmented central zone is a quantitative manifestation of the *pug*^*D*^ phenotype. Our experiments show that chromosomal position effects are responsible for some of this variation. These results are consistent with previous findings that different chromosomal insertion sites have mostly quantitative effects on the expression of transgenes (e.g., \[[@pone.0151377.ref022],[@pone.0151377.ref023]\]). A second strong influence on the *pug*^*D*^ phenotype is the length of the AGAGAGA repeats, which also produces variation in the thickness of the pigmented ring. Phenotypic severity is not strictly correlated with the length of AGAGAGA repeats: repeated segments that are either shorter or longer than the \~1 kb found in the original *pug*^*D*^ allele can reduce the size of the central region that lacks pteridine pigmentation. Although we have shown that translation of the repeats is necessary for the pug phenotype \[[@pone.0151377.ref001]\], our results do not speak to the mechanism by which the repeat variation leads to phenotypic variation. The varied AGAGAGA repeats might affect expression of the gene, the stability of the mRNA or the protein, or the efficacy of its action. The fact that this sequence variation produces the same phenotypic effect as variance in gene copy number suggests that, whatever the molecular mechanism, the effect is essentially quantitative. It will be of great interest to determine whether the *pug*^*D*^ mode of action shares any commonalities with human diseases that arise from the translation of lengthy monotonic amino acid segments \[[@pone.0151377.ref024]\]. We previously discussed the possibility that the scattered spots of pigment seen in the center of a *v; pug*^*D*^*/+* eye are a reflection of classical position effect variegation (PEV; \[[@pone.0151377.ref025]\]) resulting from the presence of the heterochromatic AGAGAGA repeats \[[@pone.0151377.ref001]\]. However, the *pug*^*D*^ phenotype is unaffected by a suppressor of PEV (*Su(var)205*) or by a PEV-enhancing genotype (*X0* males; our unpublished results). It seems unlikely that the variegated phenotype is a result of PEV. There does seem to be a correlation between the frequency of spots in the center of the eye and the thickness of the pigment ring ([Fig 1](#pone.0151377.g001){ref-type="fig"}). Thus, the absence or appearance of spots appear to be another manifestation of the strength or weakness of the phenotype. Obtaining these results depended on our ability to alter one of two factors in isolation: either gene sequence or chromosomal position. The *pug*^*D*^ gene sequence was altered by transposase, or by the intrinsic instability of this sequence in bacteria, and the altered genes were then all placed at the identical genomic site by FLP-mediated DNA mobilization \[[@pone.0151377.ref017]\]. Conversely, chromosomal position effects were tested by using this method to move a single allele to multiple sites. Several other methods that use site-specific recombination to integrate different genes at specified target in Drosophila have been produced. The method of transgene co-placement, based on the cre-*loxP* site-specific recombination system, was developed to compare the effects of two genes at the same site and eliminate uncontrolled variability resulting from position effect \[[@pone.0151377.ref026],[@pone.0151377.ref027]\]. More recently, the integrase of bacteriophage ØC31 \[[@pone.0151377.ref028]\], the FLP recombinase \[[@pone.0151377.ref029]\] and the Cre recombinase \[[@pone.0151377.ref030]\] have been used in *Drosophila* to integrate transgenic DNA at a specific site directly upon injection. The ØC31 integrase method is efficient, stocks of flies with different chromosomal *attP* sites for integration are available, and it has therefore gained wide acceptance for transformation to a specific site \[[@pone.0151377.ref031]\]. However, we still see a unique advantage to our approach in this particular case---by generating an initial transgene inside a *P* element we were able to use transposase to generate internal sequence variants, which showed that the AGAGAGA repeat length affects the phenotype. Additionally, a very large collection of precisely localized *P\[RS3\]* insertions that can serve as target sites has been produced \[[@pone.0151377.ref032]\]. Most genes that contribute to eye color have mutant phenotypes that affect pigmentation uniformly across the eye. *pug*^*D*^ is a rare example in which the eye color defect is patterned. The fact that *pug*^*D*^ variation appears as variation in the thickness of the peripheral pigmented ring suggests the presence of a factor that originates in the periphery and spreads inward. We have considered two classes of model for what this factor might be and how it could work. One possibility is that a signaling molecule may emanate from outside the eye and act to repress the expression of *pug*^*D*^. Our experiments show that wingless is not this molecule. It is still possible that *pug*^*D*^ is negatively regulated by some peripheral signal that is unrelated to *wg* and has yet to be identified. Alternatively, *pug*^*D*^ may be responding positively to a signal that emanates from the center of the eye. However, since no particular distinctive feature of the eye center has been identified we consider this unlikely. Furthermore, our finding that pigmentation can be rescued simply by injury to the pupal eye does not support models involving such signalling. The model we favor supposes that *pug*^*D*^ creates a shortage of some metabolite required for pigment formation, but that this shortage can be overcome by uptake of some still unidentified compound. To account for the ring pattern of pigmentation, this compound would be taken up only by cells around the periphery of the eye, but it may be transported to more interior cells if it is present in excess. In *pug*^*D*^*/+* flies cells around the edge of the eye may import enough of the critical substance for their own use, accounting for the ring of pigment. If expression of *pug*^*D*^ is reduced then this substance may be transported from cell to cell further towards the center of the eye until it is no longer in surplus. This hypothesis requires that all cells in the eye are capable of transport of this compound, but only cells around the periphery are capable of taking it up from outside of the eye. One way to achieve such a limitation would be if the transport machinery were localized along the elongated surfaces of the secondary pigment cells that produce pteridine pigment, but not at the bases of those cells (represented diagrammatically in \[[@pone.0151377.ref001]\]). Localized transport of the rosy protein (xanthine dehydrogenase) has been shown, but it occurs instead at the base of pigment cells \[[@pone.0151377.ref033],[@pone.0151377.ref034]\]. We found that physical injury to the eye was sufficient to produce new regions of pigmentation. This observation is consistent with our hypothesis. Cell surfaces that would normally be in contact only with other cells of the eye may be exposed by the injury, allowing uptake of the critical compound in the new location. The increased number of internal cells with pigment in the *P35* and *wingless* expression experiments may well result from the disruption of ommatidial structure in those genotypes, allowing transport to internal cells. The identification of the transported compound might go a long way to help understand the mechanism of *pug*^*D*^ action. One possibility is that the Pug^D^ protein binds and sequesters the tetrahydrofolate substrate (MTHF) that is used as a cofactor in purine biosynthesis. The import of either folate or purines to the eye by cells around the periphery could restore pigmentation and explain the ring pattern of pigmentation produced by *pug*^*D*^. If our model is correct, one might expect to see reduced health or viability of *pug*^*D*^ flies, since mutants affecting *de novo* purine synthesis have reduced viability \[[@pone.0151377.ref035],[@pone.0151377.ref036]\]. There probably is a reduction in health of cells of the eye, since the eyes of *pug*^*D*^*/+* flies to appear to be somewhat fragile \[[@pone.0151377.ref001]\]. Both *pug*^*D*^ and the reduced dosage of *pug*^*+*^ in these heterozygotes could be contributing factors to this effect. This impairment is especially evident in the *Gla*^*1*^*/+; pug*^*D*^*/+* double mutant combination which is frequently lethal, though neither mutation by itself causes significant lethality. Cells in the eye finish their program of cell division around the time of pupariation and are likely to be particularly stressed because of the need for large quantities of purines to make pigment, which occurs after the larvae have stopped feeding \[[@pone.0151377.ref007]\]. It is conceivable that these cells have a sufficient store of purines to undergo metamorphosis, but not enough to carry out the more demanding task of producing pigment. For this they would need to synthesize additional purines, or transport purines or other precursors from outside. The authors thank Andrew Tomlinson, Konrad Basler, Bruce Hay and Kenneth Cadigan for supplying fly strains used in these experiments. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: YSR KGG. Performed the experiments: YSR MMG KGG. Analyzed the data: YSR KGG. Contributed reagents/materials/analysis tools: YSR MMG KGG. Wrote the paper: YSR KGG. [^3]: Current address: School of Life Sciences, Sun Yat-Sen University, Guangzho, China

Introduction {#s1} ============ *Diuraphis noxia* (Kurdjumov, Hemiptera: Aphididae---or Russian wheat aphid, RWA) biotypes are morphologically similar, yet display vast differences in their capacity to damage wheat cultivars upon feeding (i.e., their virulence) (Botha, [@B11]). In South Africa, the virulence of the four wild type and the mutant RWA biotypes is as follows in order from least to most virulent: SA1 \< SA2 \< SA3 \< SA4 \< SAM (Swanevelder et al., [@B92]; Jankielsohn, [@B51]). Despite the emergence of new RWA biotypes in South Africa (Tolmay et al., [@B96]; Jankielsohn, [@B50], [@B51]), and other parts of the world, including the United States of America (USA) (Haley et al., [@B42]; Burd et al., [@B19]; Randolph et al., [@B76]) and Argentina (Clua et al., [@B26]), the molecular mechanism(s) underlying the development of new biotypes is currently unknown (Shufran and Payton, [@B84]; Botha et al., [@B12]). The known genealogy of SA1 and SAM (Swanevelder et al., [@B92]), their genetic similarity (Burger and Botha, [@B20]) and their position on either end of the virulence spectrum, renders them particularly useful in the present study, to improve the understanding of the process of biotypification. The possibility of a link between RWA methylation and biotype virulence has previously been suggested (Gong et al., [@B38]; Breeds et al., [@B18]). In 2012, Gong et al. investigated the methylation of four genes encoding salivary gland proteins (putative effector genes) in RWA biotypes US1 and US2, and found these genes to be differentially methylated in the different biotypes. In the initial investigation of South African RWA methylation (Breeds et al., [@B18]), the different biotypes exhibited different banding patterns (after restriction of their DNA with methylation-sensitive enzymes), methylation levels and methylation trends, all of which support a role for methylation in biotypification. The epigenetic modification of DNA methylation involves the covalent addition of a methyl group to the 5′ position of cytosine (Glastad et al., [@B35]; Lyko and Maleszka, [@B63]). In insects, methylation occurs predominantly within genes (Walsh et al., [@B100]; Zemach et al., [@B107]; Glastad et al., [@B35]; Lyko and Maleszka, [@B63]), where to date it is reported to perform two major functions. Firstly, intragenic methylation affects alternative splicing by recruiting or interfering with different DNA binding factors (Hunt et al., [@B47]; Glastad et al., [@B34]; Yan et al., [@B106]), and secondly, it prevents the initiation of spurious transcription at cryptic binding sites within genes (Hunt et al., [@B45], [@B46],[@B47]). Three classes of DNA methyltransferase (DNMT) proteins are involved in methylation of DNA and these perform different functions, with DNMT3 and DNMT1 establishing and maintaining methylation patterns, respectively, but with a less clear function for DNMT2. This class is known to show strong conservation in sequence and is suggested to be an ancient DNA methyltransferase that changed its substrate specificity from DNA to tRNA (Sunita et al., [@B91]; Iyer et al., [@B48]; Jurkowski and Jeltsch, [@B55]; Raddatz et al., [@B75]). Insects have a variety of combinations of the *DNMT* genes, with some lineages having lost one (e.g., *Bombyx mori* and *Triboleum castaneum*) or two (e.g., *Drosophila melanogaster* and *Anopheles gambiae*) classes of *DNMTs*, and others having multiple homologs (e.g., *Apis mellifera, Nasonia vitripennis*, and *Acyrthosiphon pisum*) within a certain *DNMT* class (Kunert et al., [@B59]; Marhold et al., [@B65]; Walsh et al., [@B100]; Xiang et al., [@B105]; Glastad et al., [@B35]; Feliciello et al., [@B30]). Despite their important function in DNA methylation, knowledge of RWA *DNMTs* is still lacking. DNA methylation is removed through the process of demethylation, which can occur both passively and actively, with 5-hydroxymethylcytosine (5hmC) being a measurable intermediate of one of the active demethylation pathways (Branco et al., [@B17]; Kohli and Zhang, [@B57]). Hydroxymethylcytosine is formed through the oxidation of 5mC by ten-eleven translocation enzymes (TETs) (Tahiliani et al., [@B95]; Shen et al., [@B83]). The presence of 5hmC has only been reported in a few insects including *A. mellifera, T. castaneum, N. vitripennis* and *D. melanogaster* (Cingolani et al., [@B25]; Feliciello et al., [@B30]; Wojciechowski et al., [@B103]; Delatte et al., [@B28]; Pegoraro et al., [@B73]; Rasmussen et al., [@B77]). To determine the presence and extent of 5hmC in the RWA, an antibody specific to 5hmC was used, providing the first insight into RWA demethylation. The objective in this study was firstly to sequence and compare the epigenome of RWA biotypes SA1 and SAM, and determine the level, location (e.g., intergenic or genic, exonic or intronic), and contexts of DNA methylation (i.e., CpG, CHH, CHG) within the genomes of these RWA biotypes with differential virulence. Secondly, to quantify global methylation (5mC) and demethylation (5hmC) in the South African biotypes with shared genealogy; and thirdly, to characterize the DNA methyltransferases (*DNMTs*) and ten-eleven translocase enzyme-like (*TET*) genes and expression in these aphids, to relate these observations to the reported difference in virulence levels of the South African RWA biotypes SA1 and SAM. Materials and Methods {#s2} ===================== Aphid Rearing ------------- For whole genome bisulfite sequencing and measurement of global methylation (5mC) and hydroxymethylation (5hmC) levels, colonies of apterous parthenogenetic female aphids of South African RWA biotypes SA1 and SAM were separately established in BugDorm cages (MegaView Science Education Services Co. Ltd., Taiwan) in an insectary with the following conditions: 22.5 ± 2.5°C, 40% relative humidity, and continuous artificial lighting from high pressure sodium lamps as previously described (Breeds et al., [@B18]). In all instances triplicate colonies of each biotype were established. Stock colonies of RWA biotype SA1 were maintained on the wheat line Tugela (susceptible) and biotype SAM on the near isogenic wheat line Tugela-*Dn1* (resistant). To avoid any environmental effects due to feeding on different wheat plants, aphid biotypes were transferred to the susceptible cultivar "SST356" 1 month prior to DNA extraction for the whole genome bisulfite sequencing. In all instances, treatments were conducted using separate BugDorm cages in triplicate (*n* = 3 × 2). For *DnDNMT* expression analysis (0h), RWA biotype SA1 was maintained on the "SST 356" wheat cultivar (susceptible), while the highly virulent SAM biotype was maintained on "SST 398" (RWA resistant), and then transferred to the susceptible "SST 356" wheat cultivar 1 month prior to RNA extraction and cDNA synthesis. For the RNAseq analysis, colonies of apterous parthenogenetic female aphids of South African RWA biotypes SA1 and SAM were separately established in BugDorm cages (MegaView Science Education Services Co. Ltd., Taiwan) in an insectary with the following conditions: 22.5 ± 2.5°C, 40% relative humidity, and continuous artificial lighting from high pressure sodium lamps as previously described (Breeds et al., [@B18]). RWA biotype SA1 was maintained on the wheat line Tugela (susceptible) and biotype SAM on the near isogenic wheat line Tugela-*Dn5* (resistant). Multiple individual replicates, consisting of 50 aphids of various life stages, were collected for each biotype. Collected aphids were flash frozen in liquid nitrogen and RNA was extracted following the protocol of Qiagen\'s RNeasy RNA extraction kit performing the optional on column Qiagen DNase treatment. Extracted RNA was assessed for quality through both bleach-gel electrophoresis (Aranda et al., [@B3]) and with an Agilent 2100 Bioanalyser using the RNA Nano 6000 kit (Babu and Gassmann, [@B4]). Three RNA samples, from each biotype, representing three biological repeats, with the highest RIN values (at least above 7) were used in subsequent analysis. For the *DnDNMT* and *DnTET* host-shifts, RWA biotypes SA1 and SAM were maintained on the susceptible "SST 356" wheat cultivar, and then transferred to either near isogenic wheat lines Tugela (susceptible), or Tugela-*Dn1* (resistant), or Tugela-*Dn5* (resistant) prior to RNA extraction and cDNA synthesis. Aphids were sampled at 0, 6, and 48 h post host-shifting. In all instances, treatments were conducted using separate BugDorm cages using separate plants in triplicate (*n* = 3 × 2). For the quantitation of DNMT proteins, both biotypes were maintained on the "SST 356" wheat cultivar before protein extraction. In all instances, treatments were conducted using separate BugDorm cages in triplicate (*n* = 3 × 2). All SST cultivars were obtained from SENSAKO (Pty) Ltd., (South Africa). Identification, Cloning and Sequencing of RWA *DNMTs* and *Ten Eleven Translocation-Like (TET-Like) Genes* ---------------------------------------------------------------------------------------------------------- DNMT and TET sequences of the pea aphid (*Acyrthosiphon pisum*) were obtained through GenBank and used as BLAST (Altschul et al., [@B1]) queries against the NCBI\'s non-redundant (nr) database to obtain homologs from the Class Insecta. The obtained sequences were then aligned using MAFFT v.7.4 (Katoh and Standley, [@B56]) and through use of maximum parsimony the obtained sequences were phylogenetically rendered through use of PAUP v4.0a136. Primers were designed ([Table S1](#SM1){ref-type="supplementary-material"}) to amplify the transcripts of identified RWA *DNMTs* and *TET-like genes* using Primer3 (Rozen and Skaletsky, [@B79]). The primers were then used in a primer BLAST analysis against the RWA SAM biotype reference genome (GCA_001465515.1) to ensure they only matched genes of interest. RNA extractions and cDNA synthesis were performed for both RWA biotypes SA1 and SAM as previously described (Burger et al., [@B21]). PCR reactions for sequencing were performed using Phusion High-Fidelity DNA Polymerase (NEB) and following the manufacturer\'s protocol. PCR products were then ligated into the pTZ57R/T vector (InsTAclone PCR cloning kit, Thermo Scientific) overnight at 4°C. For PCR reactions showing non-specific amplification, gel fragments containing bands of the expected product size were excised and subjected to five freeze-thaw cycles (liquid nitrogen/60°C oven) in 20 μl of distilled water and the obtained DNA was quantified through spectrophotometry (NanoDrop 2000, Thermo). Based on these results, differing amounts of freeze-thawed DNA were used, in accordance with the kit\'s recommendations on the optimal quantity of PCR product for ligation. Transformation of DH5α competent cells (Thermo Scientific) was performed through heat shock following the manufacturers\' protocol and recombinant colonies were cultured and screened as previously performed (Burger et al., [@B21]). Plasmid minipreps (derived from at least one colony per PCR product) were submitted to the Central Analytical Facility (CAF) of Stellenbosch University for bi-directional Sanger sequencing using universal M13 forward and reverse primers ([Table S1](#SM1){ref-type="supplementary-material"}). After Sanger sequencing, raw sequences were imported into Geneious v.9.1.8 and trimmed on either end to remove poor quality or ambiguous base calls. A VecScreen BLAST (<http://www.ncbi.nlm.nih.gov/tools/vecscreen/>) was then performed using the trimmed sequences to remove any vector DNA. The sequences for both SA1 and SAM biotypes (at least one forward and one reverse per PCR product) were aligned with the respective gene from which primers were designed using Primer 3 (Sievers et al., [@B86]). Sequencing, Transcriptome Assembly, and Quality Control ------------------------------------------------------- RNA samples were sent for sequencing at Macrogen Inc., South Korea where six libraries were prepared using the TruSeq Stranded mRNA Sample Preparation Guide, Part \#15031047 Rev. E. Paired-end library construction was performed using the Illumina TruSeq stranded mRNA kit and the subsequent libraries were sequenced on the Illumina NovaSeq 6000 system to obtain 100 bp paired-end reads for three replicates of both the *Diuraphis noxia* SAM and SA1 biotypes. Raw reads obtained from the NovaSeq 6000 system were analyzed through use of FASTQC (Andrews, [@B2]) and trimmed of all poor quality reads and sequencing adaptors through use of Trimmomatic (Bolger et al., [@B9]). The trimmed reads were then used to perform a strand specific *de novo* assembly through use of the Trinity software suite (Haas et al., [@B41]). The assembled transcriptome\'s quality was assessed through mapping the reads back to the assembled transcripts using Bowtie2 (Langmead and Salzberg, [@B60]), and to assess the percentage of reads utilized to construct the transcriptome. A BUSCO v4.2 (Simão et al., [@B87]) analysis was also performed using the Insecta homolog set (accessed on 2020/02, <https://buscos.ezlab.org/datasets/prerelease/viridiplantae_odb10.tar.gz>) to establish the number of represented essential orthologs. To assess if sequencing depth was adequate to generate a high quality *de novo* assembly, successively increasing sub-samplings of total sequencing data for individual samples were performed and assembled separately. These were then compared through the use of BLASTx to the NCBI\'s nr (protein) database, and the SwissProt uniprotKB and TrEMBL databases to assess the number of full-length BLAST matches obtained for the assembled transcripts from the differently sized assemblies. Basic statistics such as the number of transcripts, transcript average length, transcript average %GC content, transcript N50 and transcript Ex90N50 were also calculated. All transcripts were analyzed through use of OmicsBox v1.2 (OmicsBox---Bioinformatics Made Easy, BioBam Bioinformatics, March 3, 2019, <https://www.biobam.com/omicsbox>) by performing BLASTx and BLASTn searches, respectively to the NCBI\'s nr and nt databases (accessed on 2019/08/21). Blast2GO (Götz et al., [@B39]) was then used to assign gene ontologies (GO) and KOG terms to all transcripts. To compare the differential gene expression between the least and most virulent biotypes, transcript abundance quantification was performed using RSEM (Li and Dewey, [@B61]) for each sample using the obtained *de novo* transcripts. Average expression and the coefficient of variation was calculated per gene for the two biotypes SA1 and SAM separately. For this purpose FPKM (fragments per kilobase of transcript per million) values were used but also estimated by RSEM. We also identified differentially expressed (DE) genes between biotypes SA1 and SAM using edgeR (Robinson et al., [@B78]) based on gene-level expected counts estimated by RSEM. Only genes with greater than two counts-per-million in at least three samples were retained for DE analysis and we considered genes DE if they had a fold-change (FC) ≥1.5 and *p* \< 0.05 after adjusting for multiple testing using the Benjamini--Hochberg (BH) procedure (Benjamini and Hochberg, [@B8]). Augustus v3.3.3 (Stanke et al., [@B90]) was utilized to predict protein coding genes from the assembled transcripts using the *Acyrthosiphon pisum* (pea pahid) training set. Through use of a Trinity provided script, the GATK v.3.8 pipeline for variant calling (Van der Auwera et al., [@B98]) was applied between the transcripts from the biotypes. Variants were accepted as true if they possessed an FS score above 30 (Phred-scaled *p*-value using Fisher\'s exact test to detect strand bias) and a QD score \<2 (Variant Confidence/Quality by Depth). Variants were also required to be present in all 3 biological replicates of one biotype and absent in all 3 biological replicates of the other biotype. Analysis of *DNMT* and *TET* Expression --------------------------------------- For *DNMT* gene expression analyses, 20 apterous aphids were collected in triplicate for each biotype (3 × *n* = 60) and their heads were removed with a liquid nitrogen-cooled scalpel by cutting carefully posterior to the prothorax ([Figure S1](#SM1){ref-type="supplementary-material"}) and RNA was extracted as previously described (Burger et al., [@B21]). cDNA synthesis was performed using the iScript™ cDNA Synthesis kit (BioRad) in accordance with the provided protocol, applying 350 to 400 ng of total RNA as template per 20 μl reaction. For the host-shift experiment, RNA was isolated from apterous aphid whole-body homogenates prepared using a micro-pestle in liquid nitrogen cooled Eppendorf tubes. Each treatment was represented by three biological replicates consisting of 30 aphids each (*n* = 90). The frozen aphids were ground with micro-pestles and RNA was extracted using RNeasy Mini Kit (Qiagen), following the manufacturers recommended protocol for insect material. cDNA synthesis was performed using SensiFAST cDNA Synthesis Kit (Bioline), with 200 ng of input RNA. Primer pairs for RT-qPCR ([Table S2](#SM1){ref-type="supplementary-material"}) were designed using Primer3 from the CDS regions of the RWA sequenced *DNMTs* and *TET* to yield products of between 100 bp and 200 bp in size. Primers were used in a primerBLAST analysis against the assembled RWA SAM biotype reference genome (GCA_001465515.1) to ensure they only matched the *DNMT* and *TET* genes from which they were designed. The relative expression of *DNMT1, DNMT2*, and *DNMT3* (in sampled aphid heads of the RWA biotypes SA1 and SAM), as well as the relative expression of *DnTET* (whole aphids of RWA biotypes SA1 and SAM that underwent host-shifts) was quantified as previously described (Burger et al., [@B21]). All samples and standards were quantified in triplicate along with a no template control as a measure of contamination. A five point, two times serial dilution of a zero-hour SA1 sample was used to generate quantification standards. The relative expression of *DnDNMT3* and *TET* were calculated using Pfaffl\'s mathematical model (Pfaffl, [@B74]) for each time point (0, 6, and 48 h). A CFX96 Real-Time System (Bio-Rad) was used to perform the real-time PCR analysis. Each reaction started with a denaturation step at 95°C for 3 min, followed 40 cycles of amplification, consisting of a denaturation step at 95°C for 10 s, an annealing step at the relevant temperature for each primer set ([Table S2](#SM1){ref-type="supplementary-material"}) for 30 s, and an extension step at 72°C for 30 s. A melt curve analysis was also performed for each reaction, to verify the absence of non-specific amplification: The incubation temperature was increased in 5 s intervals, 0.5°C at a time, from 65 to 95°C. The ribosomal genes *L27* and *L32* were used as reference genes as they have previously been shown to be constitutively expressed, respectively, in RWA and the pea aphid (Shakesby et al., [@B82]; Sinha and Smith, [@B88]). Measuring DNMT Protein Activity ------------------------------- For the extraction of aphid protein, three replicates of 150 apterous aphids (*n* = 450) of biotypes SA1 and SAM were collected, flash-frozen and stored at −80°C until use. A micropestle was used to grind aphids into a fine powder, to which 100 μl phosphate buffered saline (50 mM NaH~2~PO~4~, 50 mM Na~2~HPO~4~ and 150 mM NaCl, pH 7.5), 10 μl phenylmethylsulphonyl fluoride (1 mM) and 10 μl dithiothreitol (1 mM) were added. Homogenized mixtures were centrifuged at 15 000 rpm (4°C) for 10 min to pellet the cell debris and the resulting supernatant was transferred to a clean Eppendorf tube. Protein concentrations were quantified using the Bradford protein assay (Bradford, [@B16]) with Bovine Serum Albumin as standard (BioRad, USA), and the Glomax®-Multi Detection plate reader (Promega, USA) as described by Rylatt and Parish ([@B80]). DNA methyltransferase protein activity was quantified following the guidelines provided with Abcam\'s colourimetric DNMT Activity Quantification kit (Abcam, UK), and using the maximum recommended amount of nuclear extract, 5 μl (ranging from 7.69 to 10.96 μg, standardized using the formula below) of each of the three biological replicates per biotype (*n* = 3). DNA methyltransferase activity in OD/h/μg (optical density/hour/microgram) was calculated using the formula below. Protein activity  = ( Sample OD  −  Blank OD ) \[ Protein amount  ( ug )  x hour \]  x 1000 An ANOVA was performed to test for significant differences between the sample means, with the level of significance set at *p* ≤ 0.05. Quantifying Levels of Global Methylation (5mC) and Hydroxymethylation (5hmC) ---------------------------------------------------------------------------- Global levels of methylation were determined utilizing a colourimetric Methylated DNA Quantification kit (Abcam) using 150 ng DNA of the three biological repeats per biotype (*n* = 3). A slight modification of the protocol was followed in the "methylation capture" section, whereby incubation of DNA and diluted capture antibody was performed for 15 h at room temperature in the dark to allow for optimal antibody binding, as opposed to 1 h at room temperature. The final plate incubation, after addition of the developer solution, was carried out for the maximum recommended time of 10 min. Absorbance at 450 nm was read in triplicate (*n* = 9) within five min of adding the stop solution, using the Glomax®-Multi Detection System. Relative methylation levels were calculated for each sample using the following formula: R e l a t i v e 5 m C \% = ( S a m p l e O D \- N e g a t i v e c o n t r o l O D ) / S ( P o s i t i v e c o n t r o l O D \- N e g a t i v e c o n t r o l O D ) x 2 / P x 100 where 5mC is 5-methylcytosine, OD is optical density, S is the amount of sample DNA in ng and P is the amount of positive control in ng. An ANOVA was performed to test for significant differences between the sample means, with the level of significance set at *p* ≤ 0.05. Global hydroxymethylation levels were quantified using a colourimetric Hydroxymethylated DNA Quantification kit (Abcam), in accordance with the provided protocol. Freshly extracted DNA sample from each biotype were loaded in triplicate (*n* = 3), and standardized using the formula below (refer to S, the amount of sample DNA). The final plate incubation was carried out for 10 min, where after absorbance at 450 nm was read using the Glomax®-Multi Detection System. Relative hydroxymethylation levels were calculated for each sample using the following formula: R e l a t i v e 5 h m C \% = ( S a m p l e O D \- N e g a t i v e c o n t r o l I I O D ) / S ( P o s i t i v e c o n t r o l O D \- N e g a t i v e c o n t r o l I I O D ) x 5 / P x 100 where 5hmC is 5-hydroxymethylcytosine, OD is optical density, S is the amount of sample DNA in ng and P is the amount of positive control in ng. An ANOVA was performed to test for significant differences between the sample means, with the level of significance set at *p* ≤ 0.05. Statistical Analysis -------------------- Microsoft Excel (2010)/XLSTAT Premium (Addinsoft Inc. USA) were used for the statistical analysis, and SigmaPlot (2001) was used to plot graphs showing the average readings and standard deviation. An ANOVA was performed to test for significant differences between the sample means, with the level of significance set at *p* ≤ 0.05. The model assumptions of ANOVA (i.e., homoscedasticity and normality of the residuals), were tested for using Levene\'s test and the Shapiro-Wilk test, respectively (significance set at *p* ≤ 0.05 for both tests). If the ANOVA null hypothesis---that the means of the treatment groups are equal---was rejected, a Fisher\'s LSD test was then performed. Whole Genome Bisulfite Sequencing (WGBS) and Analysis ----------------------------------------------------- A total of 100 apterous female aphids of South African RWA biotypes SA1 and SAM were used for DNA extraction performed as described previously (Burger and Botha, [@B20]) (GenBank ID GCA_001465515.1; BioProject [PRJNA297165](PRJNA297165)). Three independent biological repeats of each biotype were conducted of each biotype (*n* = 3). Samples consisting of 2 μg DNA, of both RWA biotypes SA1 and SAM were submitted to Macrogen Inc., South Korea for bisulfite treatment, library preparation and sequencing. DNA samples were treated with the EZ DNA Methylation Lightning kit (Zymo Research) and used to construct the sequencing library utilizing the TruSeq DNA Methylation Library Kit™ (Illumina) (*n* = 1) or Accel-NGS® Methyl-Seq (Swift Biosciences) for Illumina (*n* = 2), 5′ tags were generated through random priming, followed by selective 3′ tagging. Illumina P7 and P5 adapters were ligated through amplification to the 5′ and 3′ ends, respectively. The Illumina HiSeq X platform was used to sequence the bisulfite treated samples. The obtained sequencing data was analyzed for quality using FastQC (Andrews, [@B2]). After inspecting the adapter content, per base sequence content, and per base sequence quality, Trimmomatic (Bolger et al., [@B9]) was used to remove adapter sequences and trim the paired-end reads for quality. A sliding window over 15 bp was used to trim for a quality score of 20, along with a headcrop of 10. The Illuminaclip parameter was used to search for and remove adapter sequences from the reads. After trimming, all reads were filtered for a minimum read length of 40 bp. The Bismark software program (Krueger and Andrews, [@B58]) was used to analyze the methylation status of the trimmed and filtered sequence reads. Using the RWA SAM biotype reference genome (GenBank ID GCA_001465515.1; BioProject [PRJNA297165](PRJNA297165)), the observed over expected number of cytosine bases for each methylation context was calculated as follows: C p G O E = F C p G F C . F G C H G O E = F C A G \+ F C T G \+ F C C G 3 ( F C . F 1 \- G . F G ) C H H O E = F C A A \+ F C A T \+ F C A C \+ F C T A \+ F C T T \+ F C T G \+ F C G A \+ F C G T \+ F C G G 9 ( F C . 2 F 1 \- G ) Where *F* represents the frequency of the subscripted nucleotide/dinucleotide/trinucleotide, in the reference genome. As a reference genome is not available for biotype SA1, the calculations were only performed for biotype SAM. The R-package DSS-single (Wu et al., [@B93]) was used to calculate which genes are significantly differentially methylated (*p*-value \< 0.05) between SA1 and SAM from the WGBS data. For the analysis, only genic CpG loci, with at least a ten times coverage in both biotypes, across all three repeats were considered. This amounted to 613,730 CpG sites. A Wald test (Wald, [@B99]) was conducted for differentially methylated loci with the DMLTest function. The optional "smoothing" algorithm of this function, which uses methylation data of nearby loci to generate "pseudo replicates" is only recommended for datasets where methylation loci are dense (Feng et al., [@B31]). Due to the high AT content and low methylation levels in RWA, methylation loci are sparse and the "smoothing" algorithm was not employed. The CallDMR function was then used to identify differentially methylated regions using information from the differentially methylated loci, such as the number of CpG sites in a region and the percentage of sites in a region scored as significant. This information is used to calculate a combined score for each region, referred to as an area statistic, which can be used to sort regions based on the degree of differentiation in CpG methylation (Wu et al., [@B93]). The Blast2GO suite ([@B27a], Conesa and Götz, [@B27]) was used to search for the gene ontologies (GO) and KOG terms (Burger and Botha, [@B20]) of genes containing a differentially methylated region. Results {#s3} ======= Classes of DNA Methyltransferases in RWA ---------------------------------------- The BLASTp analysis performed using the insect DNMTs against the RWA proteins, revealed three DNMT subfamilies. Comparison of the DNMTs of RWA with other aphids (i.e., *A. pisum* and *Myzus persicae*), as well as with other distant hemipteran species (i.e., *Bemisia tabaci, Cimex lectularius, Diaphorina citri, Halyomorpha halys*, and *Nilaparvata lugen*) confirmed that as with other hemipterans, RWA have three DNMT subfamilies of genes, i.e., *DNMT1, DNMT2, DNMT3* ([Figures 1--3](#F1){ref-type="fig"}). With the DNMT1 and DNMT3 sequences, those from RWA were mostly similar to that of *M. persicae* and then *A. pisum*, than to the other hemipterans included in the study. Whereas, those from *N. lugens* (the brown planthopper) and *D. citri* (Asian citrus psyllid) the most distant from the plant aphids. In the case of DNMT2, the separation was less distinct. These observations were strongly supported by bootstrap values. {#F1} Further sequence analysis revealed that the RWA had two homologs of each of the DNMT1s (DNMT1a and DNMT1b) and DNMT3s (DNMT3a and DNMT3b) ([Figures 1A](#F1){ref-type="fig"}, [3A](#F3){ref-type="fig"}), but only a single DNMT2 ([Figure 2B](#F2){ref-type="fig"}) protein. The RWA DNMTs contained several functional motifs that were recognizable and contributing to the ascribed function, all shared between the plant aphids ([Figures 1C](#F1){ref-type="fig"}, [2C](#F2){ref-type="fig"}, [3C](#F3){ref-type="fig"}). {#F2} {#F3} To assess whether the DNMTs differ between RWA biotypes, an alignment of the DNMT sequences obtained between the SA1 and SAM biotypes was conducted which revealed 36 SNPs between the biotypes ([Figures S2](#SM1){ref-type="supplementary-material"}--[S7](#SM1){ref-type="supplementary-material"}). We used the transcriptome data to investigate the expression pattern of known methylation genes in RWA biotypes SA1 and SAM ([Figures 5](#F5){ref-type="fig"}, [6](#F6){ref-type="fig"}). We again found the full complement of DNA methyltransferases which was expressed in different transcript levels, with minimal differences in *DnDNMT1* and *DnDNMT2* between the RWA biotypes, with fewer *DnDNMT3* transcripts in the more virulent SAM (9.51 ± 6.9) than in the less virulent SA1 (31.45 ± 2.4) (*p* = 0.006). To measure whether the observed SNPs had any bearing on gene expression between less and more virulent RWA biotypes, the expression of aphid head *DNMTs* among the RWA biotypes was also investigated. Biotype SAM\'s *DNMT1* expression was higher than that measured in biotype SA1, but the difference in expression was not significant (*p*-value = 0.416 for *L27* and 0.362 for *L32*), as was the expression of *DNMT2* ([Figure 5A](#F5){ref-type="fig"}). The expression of *DNMT3* however showed the most inter-biotype variation of the three *DNMT* subfamilies, and revealed that the *DNMT3* expression levels of SAM (most virulent aphid biotype) were significantly lower than that measured in SA1 (Fisher\'s LSD test; *p*-value of ≤ 0.1). To establish whether the difference in the DNMT gene expression equates into measurable differences in total DNMT protein activity, the DNMT protein activity was determined ([Figure 5B](#F5){ref-type="fig"}). The concentration of DNMT protein activity within the biotypes ranged from 44.80 to 53.54 OD/h/μg, with biotype SAM exhibiting the lower DNMT protein activity of the biotypes. However, the DNMT protein activity levels did not differ significantly between the biotypes (*p* ≤ 0.05). Sequence Analysis and Expression of *DnTET* in RWA Biotypes ----------------------------------------------------------- To shed light on the observed difference in global demethylation but not methylation levels, the *TET (N6-methyl adenine demethylase)* genes responsible for oxidation within the methylation pathway were studied (Wojciechowski et al., [@B103]). We were able to isolate and sequence the *DnTET* ortholog from RWA. Comparison of the DnTET (N6-methyl adenine demethylase) protein sequences in RWA with other aphids (i.e., *A. pisum, M. persicae)* and with other distant hemipteran species (i.e., *B. tabaci, C. lectularius, D. citri, H. halys*, and *N. lugen*) confirmed that all these species have recognizable TET-like sequences in their genomes ([Figure 4](#F4){ref-type="fig"}). Clustering of the sequences group the aphids closer to each other than to the other hemipterans, with strong bootstrap support ([Figure 4A](#F4){ref-type="fig"}). Further analysis revealed that unlike *A. pisum* and some of the other hemipteran species, RWA has only a single form of TET ([Figure 4B](#F4){ref-type="fig"}), which contain several functional motifs including DNA N6-methyl adenine demethylase ([Figure 4C](#F4){ref-type="fig"}). {#F4} We also used the transcriptome data to investigate the expression pattern of the *TET (N6-methyl adenine demethylase)* genes responsible for oxidation within the methylation pathway in RWA biotypes SA1 and SAM ([Figure 6C](#F6){ref-type="fig"}). We found more *DnTET* transcripts in the more virulent SAM (5.81 ± 0.38) than in the less virulent SA1 (2.59 ± 2.3) (*p* = 0.014). {#F5} ![Differential gene expression between RWA biotypes SA1 and SAM. **(A)** SA1 and SAM gene expression as log10 fragments per kilobase of transcript per million mapped reads (FPKM) average over three biological replicates for transcripts retained for differential expression (DE) analysis with edgeR (*n* = 64,214). Benjamini-Hochberg (BH) corrected *p* \> 0.05 and absolute fold change \[FC\] \> 1.5). **(B)** Comparison of number of transcripts expressed in SA1 and SAM. **(C)** Significant differences were observed in the expression of *DnDNMT3* and *DnTET* in SA1 and SAM; edgeR; BH corrected *p* \> 0.05 and absolute fold change \[FC\] \> 1.5.](fgene-11-00452-g0006){#F6} Expression of DNA Methylation Genes During Feeding Studies ---------------------------------------------------------- To assess further whether the sequenced *DnDNMT* and *DnTET* genes expressed in RWA biotypes SA1 and SAM, differ when challenged with different host plants, the RWA biotypes were reared on susceptible host plants (*n* = 3) ([Table 1](#T1){ref-type="table"}) and then moved to two resistant wheat cultivars (Tugela-*Dn1* and Tugela-*Dn5*). The lines were selected based on the fact that they are near isogenic and differ only with regard to the resistance gene present (i.e., *Dn1* expressing antibiosis---harmful to the aphid and *Dn5* expressing antibiosis and antixenosis---non-palatable) ([Figure 7](#F7){ref-type="fig"}). The measured *DnDNMT* expression increased significantly in both aphid biotypes when challenged with a new feeding environment (i.e., *Dn5* expressing both antibiotic and antixenotic) (*p* ≤ 0.05). The relative expression almost doubled in the less virulent biotype SA1, but even more in the more virulent biotype SAM (± six-fold *Lr27*; ±11-fold *Lr32*) within 6 h after host-shifting. ###### Relative expression of *DNMT3* and *TET* in RWA biotypes SA1 and SAM when feeding on a susceptible cultivar SST. **Relative gene expression** **Aphid biotype** ------------------------------ ------------------- --------------- --------------- --------------- DNMT 0.852 ± 0.126 0.734 ± 0.084 1.308 ± 0.531 0.749 ± 0.221 TET 0.758 ± 0.114 0.652 ± 0.072 1.222 ± 0.304 0.721 ± 0.223 *Lr27 and Lr32 was used as reference genes*. {#F7} To assess whether the sequenced *DnTET* gene was also differentially expressed in RWA, biotypes SA1 and SAM during feeding after host-shifting, the expression of *DnTET* was also measured ([Figure 8](#F8){ref-type="fig"}). The measured *DnTET* expression increased slightly but not significantly after 6 h in the less virulent SA1 biotypes when challenged with the *Dn5* wheat line (antibiotic and antixenotic) (*p* \> 0.05), but significantly in SA1 after 48 h when challenged with the antibiotic wheat line Tugela-*Dn1* (*Lr27*) (*p* ≤ 0.05). In contrast, the *DnTET* expression of virulent biotype SAM remained the same when challenged with the antibiotic wheat line Tugela-*Dn1*, but increased significantly within the first 6 h after host-shifting from the susceptible Tugela to the wheat line containing the *Dn5* resistance gene (*p* ≤ 0.05). {#F8} Global Methylation and Hydroxymethylation Quantification -------------------------------------------------------- To quantify the global levels of methylation (5mC) and dehydroxymenthylation (5hmC), antibodies specific to these (i.e., 5mC and 5hmC) were used. The use of the 5mC antibody revealed similar levels of global methylation between the less virulent SA1 and more virulent SAM biotype, with the measured levels, ranging between 0.14 and 0.16% ([Figure 9](#F9){ref-type="fig"}). The hydroxymethylation levels, however differed significantly ranging from 0.12 to 0.46% ([Figure 9](#F9){ref-type="fig"}), with biotype SA1 displaying the lowest, and biotype SAM displaying the highest 5hmC levels, respectively (*p* ≤ 0.05). {#F9} Whole Genome Bisulfite Sequencing --------------------------------- The whole genome bisulfite sequencing produced a total of 6,846,597,083 raw reads for SA1 and 7,397,965,699 raw reads for SAM, respectively, of which a total of 70,861,462 bases (SA1) and 74,073,939 bases (SAM) were methylated, which represents 1.126 ± 0.321% (SA1) and 1.105 ± 0.295% (SAM) methylation in the genome ([Tables S3](#SM1){ref-type="supplementary-material"}, [S4](#SM1){ref-type="supplementary-material"}). The sequence reads were analyzed for contexts of DNA methylation within the genome ([Table S5](#SM1){ref-type="supplementary-material"}) and the results revealed that RWA has methylation in all contexts (CpG, CHG, and CHH), with the majority of methylation within the CpG context (±5.19%), while the other contexts show much lower levels of methylation (CHG---±0.27%; CHH---±0.34%). The reads were then subjected to quality analysis, aligned and mapped to the RWA biotype SAM reference genome (GenBank ID GCA_001465515.1; BioProject [PRJNA297165](PRJNA297165)). Of the methylated reads, most of the methylation was located in genic regions (±1.58%), but intergenic methylation was also present (±0.808%) ([Table S6](#SM1){ref-type="supplementary-material"}). Methylation was evenly distributed between both strands, with the top strand only containing 0.02% more methylated calls than the bottom strand ([Table S7](#SM1){ref-type="supplementary-material"}). Within genes exonic regions were found to be overall more methylated (±0.56%) than the intronic regions ([Table S8](#SM1){ref-type="supplementary-material"}), and the most represented context of methylation (i.e., CpG, CHG, or CHH) was in the CpG context, followed by the CHG context, with the least in the CHH context ([Table S9](#SM1){ref-type="supplementary-material"}). Using the SAM biotype reference genome, the observed over expected number of cytosine bases were also calculated ([Figure 10](#F10){ref-type="fig"}). This is commonly seen for the CpG context, denoted as *CpG*~*O*/*E*~ (Hunt et al., [@B45]), with the data on all three contexts of cytosine methylation available from the Bismark pipeline, the other, often overlooked *CHG*~*O*/*E*~ and *CHH*~*O*/*E*~ ratios were also included in this study. After calculating these ratios, it was revealed that the observed CpG context, unlike with the other contexts was lower than expected ([Figure 10](#F10){ref-type="fig"}). This is particularly so for the observed CpG context in exonic regions ([Figure 10A](#F10){ref-type="fig"}). {#F10} After analysis, we identified 40 differentially methylated genes (DMEs) when we compared the genes that were differentially methylated between the less and more virulent biotypes ([Figure 11](#F11){ref-type="fig"}; [Figures S8A1-3, B1-9, C1-13](#SM1){ref-type="supplementary-material"}). Even though based on broad functional categories, the differences between the least virulent SA1 and most virulent SAM seems minimal ([Figure 11](#F11){ref-type="fig"}), further analyses of their involvement into biochemical pathways revealed that these DMEs had distinctly different predicted functions (even when involved within the same pathway, [Figures S8A1-3](#SM1){ref-type="supplementary-material"}). Interesting examples include the selective methylation of genes in more virulent SAMs, but not SA1, which include include DMEs involved in the irinotecan metabolism ([Figure S8A3](#SM1){ref-type="supplementary-material"}); metabolism of cytotoxicity by cytochrome P450 ([Figures S8C10, C11](#SM1){ref-type="supplementary-material"}); steroid hormone biosynthesis ([Figure S8C2](#SM1){ref-type="supplementary-material"}) and wax biosynthesis ([Figure S8C](#SM1){ref-type="supplementary-material"}). {#F11} Discussion {#s4} ========== Integrated pest management programs against RWA depend heavily on the breeding of wheat cultivars that provide resistance (Tolmay et al., [@B97]; Smith and Clement, [@B89]; Botha, [@B11]; Sinha and Smith, [@B88]). The effectiveness of these cultivars, however, is often short-lived as aphids overcome the resistance they impart (Botha et al., [@B13], [@B14]; Tagu et al., [@B94]; Sinha and Smith, [@B88]). Understanding how new aphid biotypes develop, as well as the mechanisms they employ to exert their virulence enabling them to breakdown plant resistance, are of utmost importance if resistant cultivars are to be used to their full potential (Botha et al., [@B12]). The availability of the highly virulent mutant RWA biotype (SAM) (Swanevelder et al., [@B92]), alongside South Africa\'s naturally occurring biotypes (SA1, SA2, SA3, and SA4) (Walters et al., [@B101]; Tolmay et al., [@B96]; Jankielsohn, [@B50], [@B51]) presents a unique opportunity for the study of biotypification. Despite having developed from SA1, which only renders *dn3*-containing cultivars susceptible (Jankielsohn, [@B50]), SAM has the remarkable ability to overcome the resistance of all the *Dn* genes that have been introduced and/or documented (Botha, [@B11]; Botha et al., [@B12]). SAM thus serves as a model to resolve aphid biotypification. In the present study, we investigated the DNMT protein family, as they catalyze the covalent addition of a methyl group to the 5′ position of cytosine in the methylation pathway. *DNMT2* seemed the most conserved with only a single form of the protein responsible for stabilizing tRNA and the regulation of protein synthesis in response to environmental cues (Becker and Weigel, [@B6]). In contrast, *DNMT1* and *DNMT3* with two forms each, are responsible for maintaining and establishing methylation patterns, respectively (Goll and Bestor, [@B36]; Goll et al., [@B37]; Jeltsch et al., [@B53]). Owing to the important role of these proteins in changing methylation patterns, it is not surprising that variations in these genes occur. RWA is also not the only insect showing multiple homologs within a specific DNMT class. Some insect lineages were shown to lack one (e.g., *B. mori* and *T. castaneum*) or two (e.g., *D. melanogaster* and *A.gambiae*) classes of *DNMTs*, while others have multiple homologs (e.g., *A. mellifera, N. vitripennis* and *A. pisum*) within a certain *DNMT* class (Kunert et al., [@B59]; Marhold et al., [@B65]; Walsh et al., [@B100]; Xiang et al., [@B105]; Glastad et al., [@B35]; Feliciello et al., [@B30]). The limited available literature on aphid *DNMTs* prompted an investigation into the baseline *DNMT* expression (i.e., expression of aphids not challenged with resistance) of South African RWA. It is widely assumed that the insect DNMTs have the same functions as their mammalian orthologs (Wang et al., [@B102]; Glastad et al., [@B34]). *DNA methyltransferase 3* (*DNMT3)* was the only gene of which the expression was significantly different between the two RWA biotypes. *DNMT3* has long been known as a *de novo* methyltransferase (Okano et al., [@B70]; Goll and Bestor, [@B36]), which establishes new methylation patterns by methylating previously unmethylated sites (Kunert et al., [@B59]; Schaefer and Lyko, [@B81]). The *DNMT3* expression of the virulent biotype used in this study, SAM, is down-regulated in comparison to the less virulent biotype, SA1, and this decrease in expression could therefore be advantageous from a virulence perspective. A role for *DNMT3A* in the facilitation of transcription has also been identified, with *DNMT3A*-dependent methylation of gene bodies promoting transcription by antagonizing polycomb repression (Wu et al., [@B104]). Although the aphid effector genes are yet to be identified (Botha et al., [@B13], [@B15]), it is possible that they contain *DNMT3A* binding sites within their gene bodies, and that their transcription could be facilitated by *DNMT3A* binding and subsequent methylation. In the current study, SA1\'s *DNMT3A* expression, and therefore DNMT3A protein production, is up-regulated in comparison to the more virulent biotype. The fact that SA1 has higher *DNMT3A* expression (and perhaps greater effector protein production) under unchallenged conditions, may provide some insight into why SA1 is the least virulent biotype. Therefore, quantifying the *DNMT3A* expression of aphids challenged by resistance may yield valuable information on *DNMT3A*\'s possible involvement in effector transcription. Other functions of *DNMT3* include its role in the removal of 5mC and 5hmC (Chen et al., [@B23], [@B24]) and a proposed involvement in the maintenance of methylation, by being able to "methylate sites missed by DNMT1 activity" (Jones and Liang, [@B54]). However, as the *DNMT3*-mediated removal of 5mC and 5hmC is dependent on certain redox conditions (Chen et al., [@B23], [@B24]), and has only been shown to occur *in vitro* (Chen et al., [@B23], [@B24]), it is difficult to draw conclusions regarding the *DNMT3* expression and its potential demethylating and dehydroxymethylating activities in RWA. The DNA methyltransferase 3 protein is assumed to help maintain methylation in densely methylated areas of mammalian genomes (Jones and Liang, [@B54]). The hydroxylation of methylated cytosines by TET enzymes, resulting in the formation of 5hmC, is one of various active demethylation mechanisms (Tahiliani et al., [@B95]; Branco et al., [@B17]). The initial functional characterization of TETs was performed in mammals, which have three TET enzymes, namely TET1, TET2, and TET3 (Iyer et al., [@B49]; Tahiliani et al., [@B95]). In contrast to this, invertebrates possess only a single TET ortholog (Pastor et al., [@B72]; Wojciechowski et al., [@B103]), which has been identified in insects containing hydroxymethylation, including *A. mellifera* (Cingolani et al., [@B25]; Wojciechowski et al., [@B103]), *T. castaneum* (Feliciello et al., [@B30])*, N. vitripennis* (Pegoraro et al., [@B73]) and *D. melanogaster* (Dunwell et al., [@B29]). In 2014, Wojciechowski et al. functionally characterized the *A. mellifera* TET ortholog, AmTET, and concluded that, like the mammalian TETs, AmTET is capable of hydroxylating 5mC to form 5hmC. This provided the first evidence that TETs play a similar role in insects, as they do in mammals. The presence of measurable amounts of 5hmC in the RWA biotypes tested, suggests that at least one active demethylation pathway (i.e., hydroxylation of 5mC by TET) is present in RWA, as confirmed in the current study when we sequenced the *DnTET* ortholog. We then studied the expression of this gene in the RWA biotypes after challenging the aphids through differential feeding, as this was previously shown to be perceived as stressful (Burger et al., [@B21]). Interestingly, when SA1 feeds on wheat with an antibiotic mode of resistance (e.g., Tugela-*Dn1*), an oxidative burst (elevated H~2~O~2~) occurs at the feeding sites (Botha et al., [@B15]; Burger et al., [@B21]), and the expression of *DnTET* more than doubles. Whereas, when SA1 feed on wheat expressing both antibiosis and antixenosis (e.g., Tugela-*Dn5*), not only is the aphid challenged by the elevated H~2~O~2~ but also by volatile substances that make the wheat unpalatable (Botha et al., [@B15]), resulting in the tripling of *DnTET* expression. A similar trend is not observed with the expression of *DnTET* in SAM. Biotype SAM feeding, however, is not associated with an oxidative burst or increased peroxidase activity levels, because SAM "avoids" detection by wheat hosts (Botha et al., [@B12]). In the present study, we also studied the epigenome of RWA biotypes SA1 (least virulent SA biotype) and SAM (most virulent SA biotype). The whole-genome bisulfite sequencing indicated that the genomes of these biotypes were globally more methylated (i.e., 1.126 ± 0.321% for SA1; 1.105 ± 0.295% for SAM) than previously reported for insect genomes. For example, the global methylation levels of *A*. *mellifera* (Lyko et al., [@B62]), *B*. *mori* (Xiang et al., [@B105]), the ants *Camponotus floridanus* and *Harpegnathos saltator* (Bonasio et al., [@B10]) and *N. vitripennis* (Beeler et al., [@B7]) are all between 0.1 and 0.2%. However, when quantified using the antibody-based methods, the global methylation levels (0.14--0.16%) are in line with other reports of insect methylation. Panikar et al. ([@B71]) investigated adult *D*. *melanogaster* methylation, also through an antibody-based method, found the adult *D*. *melanogaster* genome to be \~0.5% methylated. Russian wheat aphids thus have low, but detectable levels of methylation which are \~0.2 to 0.4-fold of that of the model organism *D. melanogaster*, as measured using the same technique, which allows a more direct comparison. Although other authors have reported lower levels of adult *D. melanogaster* methylation using bisulfite sequencing (0% -- Lyko et al., [@B64]), liquid chromatography tandem mass spectrometry (0.034% -- Capuano et al., [@B22]) and thin layer chromatography (0.05--0.1% -- Gowher et al., [@B40]). The sequence reads were analyzed for contexts of DNA methylation within the genome and the results revealed that RWA has methylation in all contexts (CpG, CHG and CHH), with the majority of methylation within the CpG context (±5.19%), but still notable methylation in the other contexts (CHG---± 0.27%; CHH---± 0.34%), with most of the methylation located in the genic regions. A similar finding was recently reported by Mathers et al. ([@B66]) in the green peach aphid, *Myzus persicae*. The authors found that exons are highly enriched for methylated CpGs, particularly at the 3′ end of genes. Their findings also alludes to sex-biased differential methylation of genes involved in aphid sexual differentiation. The model organisms, mice (*Mus musculus*) and zebra fish (*Danio rerio*), showed very low levels of CHH and CHG methylation (1% and lower) when compared to the 74.2% and 80.3% CpG methylation observed, respectively. Low levels of CHG (0.26%) and CHH (0.17%) were also reported for the honeybee (*Apis mellifera*) (Feng et al., [@B32]). In another insect example, the over expression of DNMT2-like protein in fruit flies (*Drosophila melanogaster*) resulted in observable CpT and CpA methylation (Kunert et al., [@B59]). We also wanted to assess whether the DNA strands (top vs. bottom) were methylated equally and found that the top strands were slightly more methylated (0.06%) than the bottom ones. However, interestingly the difference between the methylation of the top and bottom strands were significantly more in the more virulent SAM (0.09%), than SA1 (0.02%). This was not an unexpected finding, as biotype SAM was previously found to exhibit the highest level of hemimethylation (at the external cytosine) when its methylation was investigated using the Methylation-Sensitive Amplification Polymorphism (MSAP) technique (Breeds et al., [@B18]). Hemimethylated DNA arises during DNA replication, as the newly synthesized daughter strand contains unmodified cytosines (Jeltsch, [@B52]; Goll and Bestor, [@B36]). When the levels of 5-hmC were measured using the antibody based method, earlier observations (Breeds et al., [@B18]) were confirmed, as the levels of 5hmC differed significantly between the aphid biotypes, with SAM much higher than that measured in the less virulent SA1 (*p* \< 0.05; [Figure 9](#F9){ref-type="fig"}). Analysis of the 40 differentially methylated genes (DMEs) revealed that the less and more virulent biotypes had distinctly different DMEs. As previously indicated, DMEs in the more virulent biotype SAM include DMEs involved in the irinotecan metabolism where is it seemingly regulates the conversion between SN-38 and SN38G which regulates secretion. SN-38 produced in the body by carboxylesterase is the active metabolite of irinotecan (Fujita et al., [@B33]), with its mechanism of action thought to be its interaction with the cleavable complex of DNA and a nuclear protein topoisomerase I. This then results in a blockade of DNA replication (Hsiang and Liu, [@B44]; Hertzberg et al., [@B43]) which causes double-strand DNA breakage and cell death. This process may be linked to the metabolism of xenobiotics by cytochrome P450 and ascorbate and aldarate metabolism that may enable SAM to counter its feeding environment better than its parent, the less virulent SA1. In locusts, it was demonstrated that an ascorbate-recycling system in the midgut lumen can act as an effective antioxidant defense in caterpillars that feed on prooxidant-rich foods, emphasizing the importance of a defensive strategy in herbivorous insects based on the maintenance of conditions in the gut lumen that reduce or eliminate the potential prooxidant behavior of ingested phenols (Barbehenn et al., [@B5]). Interestingly, more virulent SAM also selectively methylates genes associated with steroid hormone biosynthesis and wax biosynthesis, both pathways producing chemicals that affect the physiological processes associated with insect development and insect survival (Niwa and Niwa, [@B69]). The less virulent SA1 have DMEs associated with the glutathione biosynthesis which may signal stress responses. Collectively, the results suggest that RWA biotype SAM has a greater capacity to actively methylate/demethylate its DNA than its parent SA1. Thus, it can be concluded with fair confidence that SAM undergoes more methylation/demethylation than its parent biotype SA1. However, the question that remains is whether this ability to actively methylate/demethylate occurs at specific sets of genes depending on the environmental cue/stress RWA is faced with, as opposed to occurring globally (although global, genome-wide demethylation was measured). As gene bodies are the predominant sites of methylation in insects (Zemach et al., [@B107]; Glastad et al., [@B35]; Lyko and Maleszka, [@B63]), it is likely that it is in these regions that methylation will be removed. Also, removal of intragenic methylation of certain genes may alter the transcripts that are produced, by exposing cryptic binding sites or intragenic promoters (Maunakea et al., [@B68]; Hunt et al., [@B46]) and/or affect the splice variants that are produced, through methylation\'s involvement in alternative splicing (Lyko and Maleszka, [@B63]; Shukla et al., [@B85]; Bonasio et al., [@B10]; Maunakea et al., [@B67]; Glastad et al., [@B34]; Yan et al., [@B106]). As demethylation can occur in a matter of hours (Glastad et al., [@B35]), the greater capability of SAM to demethylate its genome, may provide SAM with more flexibility to adapt to changing environments, and therefore may underlie SAM\'s ability to overcome plant resistance. However, this is an aspect that requires further investigation in future. Data Availability Statement {#s5} =========================== The whole epigenome sequencing data from SA1 (accession [SRX4643785](SRX4643785)) and SAM (accession [SRX4643786](SRX4643786)) were deposited to the National Center for Biotechnology Information (NCBI) (GEO SUBMISSION GSE119504) (<https://www.ncbi.nlm.nih.gov/gds/?term=diuraphis%20noxia>) with Bioproject number PRJNA489432. Author Contributions {#s6} ==================== A-MB conceived and designed the experiments as well as drafted the manuscript. Aphid rearing and sampling was performed by KB and NB. DNMT expression and TET expression was performed by KB and JT respectively, while DNMT and TET sequencing was performed by KB and HS, respectively. Alignments and phylogentic analyses of DNMT and TET sequences were performed by NB while 5mC and 5hmC analyses were performed by KB, NB, and A-MB. Whole genome bisulfite sequencing and its analysis was performed by PP and NB. All authors edited and reviewed the final manuscript. Conflict of Interest {#s7} ==================== 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. Supplementary Material {#s8} ====================== The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fgene.2020.00452/full#supplementary-material> ###### Click here for additional data file. [^1]: Edited by: Wei Guo, Institute of Zoology (CAS), China [^2]: Reviewed by: Bing Chen, Hebei University, China; Owain Rhys Edwards, CSIRO Land and Water, Australia [^3]: This article was submitted to Epigenomics and Epigenetics, a section of the journal Frontiers in Genetics

The sequence data for L. infantum 263 AMB1000.1 is available at the EMBL European Nucleotide Archive (<http://www.ebi.ac.uk/ena>) under study accession ERP001815. Introduction {#sec001} ============ Protozoan parasites belonging to the *Leishmania* genus cause several vector-borne diseases collectively referred as leishmaniases. Currently, *Leishmania* species threaten *ca*. 350 million people in 88 countries worldwide \[[@pntd.0005171.ref001]\]. Control measures primarily rely on prevention and chemotherapy (reviewed in \[[@pntd.0005171.ref002]\]). The old-fashioned and toxic antimonial derivatives top the short list of registered compounds against *Leishmania* spp. In addition to their toxicity, pentavalent antimonials require long treatment schedules and are associated with resistance \[[@pntd.0005171.ref001], [@pntd.0005171.ref003]\]. Amphotericin B (AmB) liposomal formulations were introduced for the treatment of visceral leishmaniasis in antimonial-non-responsive regions of Bihar (India) \[[@pntd.0005171.ref004]\]. Clinical resistance to AmB is rare \[[@pntd.0005171.ref005]\] but a recent study in India has reported a *L*. *donovani* field strain resistant to AmB \[[@pntd.0005171.ref006]\]. Another leishmanicidal drug introduced in the early 21^st^ century is the alkyl-phospholipid analogue miltefosine (MF). It was the first effective oral drug showing high cure rates in the treatment of several forms of leishmaniasis. However, since its registration in 2002, it has had increasing relapse rates and the emergence of drug resistance strains \[[@pntd.0005171.ref007], [@pntd.0005171.ref008]\]. None of these drugs have a well-defined mode of action against *Leishmania* spp. and primary protein drug targets are unlikely \[[@pntd.0005171.ref009]\]. AmB seems to generate channel-like pores spanning the lipid bilayer by binding preferentially to ergosterol within the membranes, hence leading to cells death \[[@pntd.0005171.ref010], [@pntd.0005171.ref011]\]. Several reports suggest that MF is able to target lipid metabolism, in addition to glycosylphosphatidylinositol (GPI) anchor biosynthesis and signal transduction \[[@pntd.0005171.ref012]\]. MF-treated parasites show an increase in phosphatidylethanolamine (PE) and *lyso*-phosphatidylcholine (PC) content in their membrane \[[@pntd.0005171.ref013]\]. Both MF and AmB affect lipids in cellular membranes \[[@pntd.0005171.ref011], [@pntd.0005171.ref014]\], and resistance mechanisms seem to involve changes in lipids. AmB-resistance in *Leishmania* mainly implies changes in cell membrane fluidity (reviewed in \[[@pntd.0005171.ref015]\]). The sterol content of *L*. *donovani* AmB-resistant promastigotes analysed by gas chromatography coupled to mass spectrometry (GC-MS) revealed an enrichment in cholesta-5,7,24-trien-3*β*-ol \[[@pntd.0005171.ref011]\], which suggests a more fluid cellular surface. On the other hand, resistance to MF primarily implies a transport defect with inactivation of the P-type ATPase miltefosine transporter (MT), or of its regulatory subunit LdRos3, causing a decrease in the uptake of *lyso*-phospholipids \[[@pntd.0005171.ref016]--[@pntd.0005171.ref018]\]. A recent study has reported changes to the metabolism of lipids in *L*. *infantum* MF-resistant parasites \[[@pntd.0005171.ref019]\], further supporting that MF influences fatty acid and/or sterol metabolism \[[@pntd.0005171.ref020]\]. We report here that the MT is mutated in both MF and AmB resistant mutants. The mutations are associated with cross-resistance and correlate with major changes in membrane lipid composition. These modifications in lipid composition were analysed through a range of lipidomic approaches and we show that different mutations in MT trigger changes in lipid compositions leading to both MF and AmB resistance. These findings are of potential clinical relevance as the sequential treatment of liposomal AmB followed by a short 7-days administration of MF has been used against visceral leishmaniasis in India \[[@pntd.0005171.ref021],[@pntd.0005171.ref022]\]. Material and Methods {#sec002} ==================== *Leishmania* cultures {#sec003} --------------------- The *Leishmania infantum* (MHOM/MA/67/ITMAP-263) wild-type strain (Ldi263 wt) and the *in vitro* generated resistant mutants AmB1000.1 and MF200.5 \[[@pntd.0005171.ref023], [@pntd.0005171.ref024]\], which are respectively resistant to 1000 nM of AmB and 200 μM MF, were grown in SDM-79 medium at 25°C supplemented with 10% fetal bovine serum, 5 μg/mL of haemin at pH 7.0 with either 200 μM of MF (Miltefosine, Cayman Chem.) or 1 μM AmB (Amphotericin B solution, Sigma) for the mutant strains, and 40 μg/mL G418 (Geneticin, Gibco-BRL) for the episomal overexpressors. EC~50~ values were calculated based on dose-response curves analysed by non-linear regression with GraphPad Prism 5.01 software. An average of at least three independent biological replicates was performed for each determination. Statistical significance between the mock-transfected wild-type and the tested strains was evaluated by unpaired two-tailed t test. Whole genome sequencing for AmB1000.1 strain {#sec004} -------------------------------------------- Genomic DNA was prepared from a mid-log phase clonal culture of *L*. *infantum* 263 AmB1000.1. A paired-ends sequencing library was prepared with the Nextera DNA sample prep kit and sequenced on an Illumina MiSeq platform with 250-nucleotide paired-ends reads. An average genome coverage of over 50-fold was obtained for the mutant. This approach allowed us to identify point mutations when comparing with the reference genome sequence of *L*. *infantum JPCM5* (TriTrypDB version 8.0) \[[@pntd.0005171.ref025]\] and *L*. *infantum* 263 wt \[[@pntd.0005171.ref026]\]. Sequence reads were aligned to the *L*. *infantum* JPCM5 genome using the software bwa-mem \[[@pntd.0005171.ref027]\]. The maximum number of mismatches was 4, the seed length was 32 and 2 mismatches were allowed within the seed. Read duplicates were marked using Picard (<http://broadinstitute.github.io/picard>) and we applied GATK for indel realignment and snp and indel discovery \[[@pntd.0005171.ref028], [@pntd.0005171.ref029]\] in *L*. *infantum* 263 AmB1000.1. PCR amplification and conventional DNA sequencing verified all putative point mutations detected by whole genome sequencing. Copy numbers variations were derived from read depth coverage by comparing the coverage of uniquely mapped reads between *L*. *infantum* 263 AMB1000.1 and *L*. *infantum* 263 wt along small non-overlapping genomic windows (5 kb) for the 36 chromosomes (normalized to the total number of uniquely-mapped reads for each strain) \[[@pntd.0005171.ref030]\]. Several python and bash scripts were created to further analyze the data. The sequence data for *L*. *infantum* 263 AmB1000.1 is available at the EMBL European Nucleotide Archive (<http://www.ebi.ac.uk/ena>) under study accession ERP001815 and sample accession ERS176091. DNA constructs and transfection {#sec005} ------------------------------- The LinJ.13.1590 and LinJ.16.1240 genes of *L*. *infantum* (LinJ10_V5.0) were amplified from genomic DNA using compatible primer pairs and PCR fragments were ligated into pGEM T-easy (Promega, Madison, WI, USA) for confirming the quality of the insert by standard sequencing, and then cloned in the *Leishmania* expression vector pSP72αNeoα \[[@pntd.0005171.ref031]\], which contains the gene neomycin phosphotransferase (NEO) as selectable marker in *Leishmania*. A total of 20 μg of plasmid DNA for episomal expression, either the empty vector (mock) or carrying the genes of interest, were transfected into *Leishmania* promastigotes by electroporation as described previously \[[@pntd.0005171.ref031]\]. Selection was achieved in the presence of 40 μg/mL G418. Miltefosine uptake {#sec006} ------------------ Miltefosine uptake was performed as described previously \[[@pntd.0005171.ref019]\]. Briefly, *Leishmania* parasites were incubated in the presence of MT-11C-BODIPY \[[@pntd.0005171.ref032]\] for 1 h. Then fluorescence emission was recorded and used to calculate the moles of internalized MF analogue. An average of three independent biological replicates run in triplicate was performed. Statistical significance between the mock-transfected wild-type and the tested strains was evaluated by unpaired two-tailed t test. Macrophage infections {#sec007} --------------------- As previously published \[[@pntd.0005171.ref009]\] macrophage infections and drug susceptibility assays were performed as following: PMA-differentiated THP-1 macrophages were infected with stationary-phase parasites at a ratio of 18:1, for 2 h at 37°C in a 5% CO~2~ atmosphere. Cells were maintained in drug-free medium for 48 h after which infected cells were either left untreated or treated for 96 h at 37°C. The number of infecting amastigotes per 100 macrophages was determined by examination of 100 macrophages per assay in two independent experiments run in triplicate, which allowed for calculating the parasitic index (P~Idx~) as the percentage of infected cells multiplied by the mean number of parasites per cell. Statistical significance between the mock-transfected wild-type and the tested strains was evaluated by unpaired two-tailed t test. Lipid extraction {#sec008} ---------------- Total lipids were extracted using a modified Bligh and Dyer method. Briefly, cells were washed with PBS, suspended in 100 μL PBS and transferred to a glass tube, 375 μL of 1:2 (v/v) CHCl~3~: MeOH added and vortexed. The samples were agitated vigorously for a further 10--15 min. The samples were now made biphasic by the addition of 125 μL of CHCl~3~, vortex and then 125 μL of H~2~O and vortexed again and centrifuged at 1000 g at RT for 5 min. The lower phase was transferred to a new glass vial and dried under nitrogen and stored at 4°C. Electrospray-mass spectrometry analysis {#sec009} --------------------------------------- Total lipid extracts were dissolved in 15 μL of CHCl~3~: MeOH (1:2) and 15 μL of acetonitrile: iso-propanol: water (6:7:2) and analysed with a AB Sciex 4000 QTrap, a triple quadrupole mass spectrometer equipped with a nanoelectrospray source. Samples were delivered using a Nanomate interface in direct infusion mode (\~125 nL/min). The lipid extracts were analysed in both positive and negative ion modes using a capillary voltage of 1.25 kV. MS/MS scanning (daughter, precursor and neutral loss scans) were performed using nitrogen as the collision gas with collision energies between 35--90 V. Each spectrum encompasses at least 50 repetitive scans. Tandem mass spectra (MS/MS) were obtained with collision energies as previously described \[[@pntd.0005171.ref033]\], phosphatidylinositol (PI)/ inositol-phosphoceramide (IPC) in negative ion mode, parent-ion scanning of m/z 241 (PI 32:0 internal standard); 35-65V, PE in negative ion mode, parent-ion scanning of m/z 196 (PE 28:0 internal standard); 20-35V, C19Δ parent ion mode scanning m/z 295. MS/MS daughter ion scanning was performed with collision energies between 35-90V. Assignment of phospholipid species is based upon a combination of survey, daughter, precursor and neutral loss scans, as well as previous assignments \[[@pntd.0005171.ref033], [@pntd.0005171.ref034]\]. The identity of phospholipid peaks was verified using the LIPID MAPS: Nature Lipidomics Gateway ([www.lipidmaps.org](http://www.lipidmaps.org)). Accurate mass spectra were also acquired (± 2 ppm) on an Orbitrap MS to assist definitive assignment of C19Δ fatty acid containing PE species. Inositol analysis {#sec010} ----------------- For inositol analysis, a fixed number of cells from different strains were collected and lipids extracted as above. An internal standard of D~6~-*myo*-inositol was added to samples prior to hydrolysis by a strong acid (6M HCl, o/n at 110°C), derivatised with trimethylsilyl ethers and analysed by GC-MS, as published elsewhere \[[@pntd.0005171.ref035]\]. *Myo*-inositol was quantified and the mean and standard deviations of three separate analyses were determined for IPC and PI inositol quantification, lipid samples underwent base hydrolysis 500 μL of concentrated ammonia and 50% propan-1-ol (1:1), followed by incubation for at least 5 h at 50°C. Upon drying under nitrogen and removal of traces of ammonia with 2 rounds of H~2~0/MeOH evaporation, the modified Bligh and Dyer method as described above was conducted to separate the IPC in the organic phase and inositol-phospho-glycerol derived from the PI, which had been deacylated into the aqueous phase. These two phases were dried down and processed for inositol content as described above. Identification and quantification of fatty acids {#sec011} ------------------------------------------------ Full characterisation and quantification of the fatty acids by conversion to the corresponding fatty acid methyl esters (FAME) followed by GC-MS analysis was performed as previously described \[[@pntd.0005171.ref036]\]. Briefly, mid-log cell-lines were spun down and triplicate aliquots equivalent to 10^8^ cells were transferred to 2 mL glass vessels and spiked with an internal standard fatty acid C17:0 (20 μL 1 mM) and dried under nitrogen. Fatty acids were released by base hydrolysis using 500 μL of concentrated ammonia and 50% propan-1-ol (1:1), followed by incubation for at least 5 h at 50°C. After cooling, the samples were evaporated to dryness with nitrogen and dried twice more from 200 μL of H~2~O/MeOH (1:1) to remove all traces of ammonia. The protonated fatty acids were extracted by partitioning between 500 μL of 20 mM HCl and 500 μL of ether, the aqueous phase is re-extracted with fresh ether (500 μL) and the combined ether phases were dried under nitrogen in a glass tube. The fatty acids were converted to FAME by adding diazomethane (3 x 20 μL aliquots) to the dried residue on ice. After 30 min the samples were allowed to warm to RT and left to evaporate to dryness in a fume hood. The FAME products were dissolved in 10--20 μL dichloromethane and analysed by GC-MS by injection of 1--2 μL on a Agilent Technologies (GC-6890N, MS detector-5973) with a ZB-5 column (30 M x 25 mm x 25mm, Phenomenex), with a temperature program at 70°C for 10 min followed by a rising gradient to 220°C at 5°C /min and held at 220°C for a further 15 min. Mass spectra were acquired from 50--500 amu. The identity of FAMEs was carried out by comparison of the retention time and fragmentation pattern with a bacterial FAME standard that contains both C17Δ and C19Δ (Supelco). Sterol analysis {#sec012} --------------- Lipid extractions of triplicate aliquots equivalent to 10^8^ cells were transferred to 2 mL glass vessels and dried down. The lipid extracts were dissolved in 20 μL dichloromethane and analysed by GC-MS by injection of 1 μL on a Agilent Technologies (GC-6890N, MS detector-5973), injector at 270°C with a ZB-50 column (15 mm x 32 mm id x 0.5 mm thickness, Phenomenex), injector at 270°C with a temperature program at 100°C for 1 min followed by a gradient to 200°C at 8°C /min and held at 200°C for a further 2 min followed by a second gradient to 300°C at 3°C /min and held for a further 15 min. Mass spectra were acquired from 50--550 atomic mass units. The identity of sterols was carried out by comparison of the retention time and fragmentation pattern with a range of standards purchased from Sigma and Materya. Results {#sec013} ======= Whole genome sequencing of *L*. *infantum* aAmB1000.1 reveals a point mutation in the miltefosine transporter MT {#sec014} ---------------------------------------------------------------------------------------------------------------- The *in-vitro* selected resistant mutant *L*. *infantum* AmB1000.1 was previously characterized by means of a large-scale proteomic study \[[@pntd.0005171.ref023]\], and here its genome was sequenced using paired-ends Illumina sequencing and compared to the one of its parent line. An average of 50-fold genome coverage was obtained for both the wild-type and the mutant. Read depth coverage analysis did not identify specific gene amplification or deletion in the mutant ([S1 Dataset](#pntd.0005171.s012){ref-type="supplementary-material"}), although aneuploidy was observed for 6 chromosomes ([S1 Table](#pntd.0005171.s009){ref-type="supplementary-material"}), a phenomenon often observed in drug resistant mutants \[[@pntd.0005171.ref018], [@pntd.0005171.ref037]--[@pntd.0005171.ref039]\]. A search for point mutations revealed 18 homozygous single nucleotide polymorphisms (SNPs) in the AmB1000.1 mutant ([S2 Table](#pntd.0005171.s010){ref-type="supplementary-material"}), 3 of which occurred within coding sequences (CDS) and caused an amino acid change ([S2 Dataset](#pntd.0005171.s013){ref-type="supplementary-material"}). These occurred within the MT ORF (LinJ.13.1590), in gene LinJ.16.1240 coding for a hypothetical transmembrane protein and in LinJ.35.0520, a large proteophosphoglycan protein made up of a short 88-times repeated sequence which is often found mutated in our various sequencing screens. The latter gene was not studied further. The SNPs in MT and LinJ.16.1240 were confirmed by sequencing PCR fragments derived from AmB1000.1. An additional 470 heterozygous SNPs were also detected in the genome of AmB1000.1, 85 of them being in CDS and non-synonymous ([S2 Table](#pntd.0005171.s010){ref-type="supplementary-material"}). Interestingly none of these occurred within the ORF coding for the regulatory subunit Ros3 that is necessary for the expression of a functional MT translocation machinery \[[@pntd.0005171.ref017]\] ([S3 Dataset](#pntd.0005171.s014){ref-type="supplementary-material"}). Mutations of the MT gene in AmB1000.1 modulate AmB resistance and is associated with cross-resistance to MF {#sec015} ----------------------------------------------------------------------------------------------------------- We next tested whether mutations in the MT or the LinJ.16.1240 gene detected in AmB1000.1 directly contributed to AmB resistance. We also included the previously characterized MF resistant mutant *L*. *infantum* MF200.5 \[[@pntd.0005171.ref024]\] with a known mutation in MT. Targeted sequencing of the MT in *L*. *infantum* MF200.5 confirmed the previously described G565R mutation, but also revealed two new mutations located at the very beginning of the gene and within a conserved domain ([S1 Fig](#pntd.0005171.s001){ref-type="supplementary-material"}). The emergence of these new mutations is probably due to continuous culturing of the *L*. *infantum* MF200.5 mutant in the presence of high MF concentrations. The mutants AmB1000.1 and MF200.5 were highly resistant to AmB ([Fig 1A and 1C](#pntd.0005171.g001){ref-type="fig"}) and MF respectively ([Fig 1B and 1C](#pntd.0005171.g001){ref-type="fig"}). Remarkably, both mutants also showed MF/AmB cross-resistance, with mutant AmB1000.1 being 3.7 fold less sensitive to MF than wild-type parasites ([Fig 1B and 1C](#pntd.0005171.g001){ref-type="fig"}) and MF200.5 being 2.7-fold cross-resistant to AmB ([Fig 1A and 1C](#pntd.0005171.g001){ref-type="fig"}). {#pntd.0005171.g001} Transfection of the wild-type MT gene as part of an episomal vector in the *L*. *infantum* AmB1000.1 and MF200.5 lines abolished, as expected and previously described \[[@pntd.0005171.ref017], [@pntd.0005171.ref038]\], MF resistance in MF200.5 and interestingly in AmB1000.1 compared to mock-transfected parasites ([Fig 1C](#pntd.0005171.g001){ref-type="fig"}). Surprisingly, transfection of the MT gene also reduced the AmB resistance to wild-type levels in the case of MF200.5 and partially reverted resistance to AmB by 4-fold for AmB1000.1 ([Fig 1C](#pntd.0005171.g001){ref-type="fig"}). Overexpression of the wild-type copy of gene LinJ.16.1240 in *L*. *infantum* AmB1000.1 had no impact on the resistance phenotype against either AmB or MF ([S2 Fig](#pntd.0005171.s002){ref-type="supplementary-material"}). This new role of MT in AmB resistance is not limited to *L*. *infantum* since selection for AmB resistance in *L*. *major* Friedlin similarly selected for a MT frameshift mutation in mutant *L*. *major* AmB1080.3 ([S1 Fig](#pntd.0005171.s001){ref-type="supplementary-material"}). The mutant displayed a 2-fold cross-resistance to MF ([S3A Fig](#pntd.0005171.s003){ref-type="supplementary-material"}) and transfection of a wild-type copy of the MT reduced not only resistance to MF, but also to AmB (3-fold) ([S3B Fig](#pntd.0005171.s003){ref-type="supplementary-material"}). We also tested the survival of AmB1000.1 and MF200.5 in PMA-differentiated THP-1 macrophages and the role of MT in resistance in intracellular parasites. The P~Idx~, represented as the percentage of infected cells multiplied by the mean number of parasites per cell, of *L*. *infantum* wild-type parasites was 540 with a mean number of 6.5 amastigotes per macrophage, while the P~Idx~ for AmB1000.1 and MF200.5 were lowered by half including the number of amastigotes inside the infected macrophages ([Fig 2A](#pntd.0005171.g002){ref-type="fig"}). The impaired infectivity was MT-related since there was a partial rescue of the phenotype in the AmB1000.1 MT add-back that reached a P~Idx~ of 420 with 5.5 parasites per macrophage ([Fig 2A](#pntd.0005171.g002){ref-type="fig"}). The AmB1000.1 intracellular amastigotes were also resistant to AmB ([Fig 2B](#pntd.0005171.g002){ref-type="fig"}) and maintained their cross-resistance to MF ([Fig 2C](#pntd.0005171.g002){ref-type="fig"}) when compared to wild-type amastigotes. Similar results were found for mutant MF200.5 that maintained their cross-resistance to AmB inside macrophages ([Fig 2B and 2C](#pntd.0005171.g002){ref-type="fig"}). Introducing a wild-type copy of the MT in AmB1000.1 led to a reduction in resistance levels to both antileishmanial agents ([Fig 2B and 2C](#pntd.0005171.g002){ref-type="fig"}). {#pntd.0005171.g002} The role of mutations in MT in miltefosine transport {#sec016} ---------------------------------------------------- Mutations in MT are often associated with defects in MF transport \[[@pntd.0005171.ref016], [@pntd.0005171.ref017], [@pntd.0005171.ref040]\]. Thus, mutations in MT in AmB1000.1 prompted us to probe the ability of the mutant to take up MF. We monitored the uptake of MT-11C-BODIPY, a fluorescent analogue of MF with *in vitro* leishmanicidal activity comparable to that of the original alkyl-phosphocholine \[[@pntd.0005171.ref032]\]. The transport of MT-11C-BODIPY was greatly impaired in the MF200.5 mutant in comparison to the wild-type strain ([Fig 3](#pntd.0005171.g003){ref-type="fig"}). In the mutant AmB1000.1 the decrease in accumulation of the fluorescent molecule was minimal but statistically significant ([Fig 3](#pntd.0005171.g003){ref-type="fig"}). The introduction of an episomal copy of wild-type MT restored MF accumulation in the MF200.5 background whilst AmB1000.1 experienced a 2.6-fold increase in the amount of MF accumulated ([Fig 3](#pntd.0005171.g003){ref-type="fig"}). This is in line with the MF re-sensitization observed for both mutants upon transfection of the rescue plasmid ([Fig 1C](#pntd.0005171.g001){ref-type="fig"}). {#pntd.0005171.g003} Changes in lipid-species content in amphotericin B and miltefosine resistant mutants harbouring mutations in the MT {#sec017} ------------------------------------------------------------------------------------------------------------------- Mutations in the MT can lead to cross-resistance to MF and AmB but it would appear that different mutations have different outcomes on MF transport ([Fig 3](#pntd.0005171.g003){ref-type="fig"}). As MT translocates *lyso*-phospholipids, we hypothesized that changes in membrane lipid content triggered by mutations in the MT could be responsible for AmB/MF cross-resistance. The lipid/fatty acid composition of the mutants AmB1000.1 and MF200.5 were thus assessed and compared to those of wild-type and add-back cells. Phospholipids (PLs) content in *L*. *infantum* 263 wild-type, MF200.5, AmB1000.1, and AmB1000.1 MT add-back parasites was analysed by electrospray ionisation-mass spectrometry (ES-MS). Negative ion ES-MS survey scans (600--1000 m/z) of total lipid extracts revealed significant changes in mutants AmB1000.1 ([Table 1](#pntd.0005171.t001){ref-type="table"} and [S4C Fig](#pntd.0005171.s004){ref-type="supplementary-material"}, upper panel) and MF200.5 ([Table 1](#pntd.0005171.t001){ref-type="table"} and [S4B Fig](#pntd.0005171.s004){ref-type="supplementary-material"}, upper panel) compared to wild-type parasites ([S4A Fig](#pntd.0005171.s004){ref-type="supplementary-material"}, upper panel). We observed that several IPC lipid species from 34:0 to 38:0 were increased in both mutants ([Table 1](#pntd.0005171.t001){ref-type="table"}, [S4A--S4C Fig](#pntd.0005171.s004){ref-type="supplementary-material"} upper panel). Transfection of the MT gene in AmB1000.1 brought back these IPC species closer to wild-type levels ([Table 1](#pntd.0005171.t001){ref-type="table"}). Two of the major PI species at 836 and 850 m/z were decreased in both mutants, probably as a result of increased IPC formation which requires PI. However, there was also a corresponding increase in the PI 42:8 species (934 m/z) ([Table 1](#pntd.0005171.t001){ref-type="table"}). Transfection of the MT gene in AmB1000.1 brought back the PI species closer to wild-type levels ([Table 1](#pntd.0005171.t001){ref-type="table"}). The PE species were relatively unchanged in MF200.5 and AmB1000.1 with the notable exceptions of a16:1/19Δ and 16:1/19Δ that were increased in both mutants ([Table 1](#pntd.0005171.t001){ref-type="table"}, [S4 Fig](#pntd.0005171.s004){ref-type="supplementary-material"} upper panels). Again, the transfection of wild-type MT in AmB1000.1 brought back those lipid species closer to wild-type levels ([Table 1](#pntd.0005171.t001){ref-type="table"}, [S4 Fig](#pntd.0005171.s004){ref-type="supplementary-material"} upper panels). 10.1371/journal.pntd.0005171.t001 ###### Comparison of the lipid PI, IPC and PE species between WT and MF200.5, AmB1000.1 or AmB1000.1+MT. These comparisons are based upon parent-ion scanning of m/z 241 for PI and IPC (PI 32:0 was used as an internal standard); 35-65V and parent-ion scanning of m/z 196 for PE (PE 28:0 was used as an internal standard). When required parent-ion scanning of m/z 295 for C19Δand accurate mass assisted distinctions between PE species with the same nominal mass ([S6 Fig](#pntd.0005171.s006){ref-type="supplementary-material"} and [S3 Table](#pntd.0005171.s011){ref-type="supplementary-material"}). The symbols show relative increase or decrease relative to wild-type cells "+++, ++, +, =, -, \--, \-\--". {#pntd.0005171.t001g} Observed mass (m/z) Lipid species MF200.5[^a^](#t001fn001){ref-type="table-fn"} AmB1000.1[^a^](#t001fn001){ref-type="table-fn"} AmB1000.1+MT[^a^](#t001fn001){ref-type="table-fn"} --------------------- ------------------------ ----------------------------------------------- ------------------------------------------------- ---------------------------------------------------- **751** IPC 32:1 = = = **753** IPC 32:0 = = = **779** IPC 34:1 ++ ++ \+ **781** IPC 34:0 +++ ++ \+ **807** IPC 36:1 +++ ++ \+ **809** IPC 36:0 +++ ++ \+ **833** IPC 38:1 = \+ = **835** IPC 38:0 \+ ++ = **861** IPC 38:1 = \+ = **836** PI 34:1 \- \-- \-- **838** PI 34:0 \-- \- \-\-- **848** PI a-36:2 \- = = **850** PI a-36:1 \-- ++ \-- **852** PI a-36:0 \- = \- **860** PI 36:3 = = ++ **862** PI 36:2 = \+ \+ **864** PI 36:1 = \+ = **866** PI 36:0 = = = **892** PI 38:1 \-- \- \+ **894** PI 38:0 \-- \- = **934** PI 42:8 +++ +++ = **690** PE 32:0 = \-- \-- **698** PE a-34:3 = \- \- **700** PE a-34:2 = \- \- **712** PE 34:3 \+ = = **714** PE 34:2/ PE a-16:1/19Δ =/+ =/= =/= **716** PE 34:1/ PE a-16:0/19Δ =/+ =/+ =/= **726** PE a-36:3 \+ = = **728** PE a-36:2/16:1/19Δ =/++ =/++ =/- **742** PE 36:2/ a-18:1/19Δ =/= =/+ =/- **744** PE 36:1/ a-18:0/19Δ =/= =/+ =/= **766** PE 38:4 = = = **768** PE 38:3 = = = **770** PE 38:2 = = = **772** PE 38:1 \+ = = ^a^Sometimes in Table 1 the lipid species share an observed mass (*e*.*g*. PE 34:1 and PE a-16:0/19Δ were detected at 716 (m/z)). While the presence of the first species may remain unaltered with respect to the wild-type, the second one may increase or decrease. The nomenclature =/+, =/++ and =/- was used in these cases. The positive ion ES-MS survey scans (600--1000 *m*/*z*) of total lipid extracts derived from both mutants indicated slight variation in individual PC species. We also observed increased presence of ceramide species at 659 and 685 m/z for mutants MF200.5 ([S4B Fig](#pntd.0005171.s004){ref-type="supplementary-material"} lower panel) and AmB1000.1 ([S4C Fig](#pntd.0005171.s004){ref-type="supplementary-material"} lower panel). Introducing a wild-type copy of MT in AmB1000.1 reversed to some extent the observed differences ([S4D Fig](#pntd.0005171.s004){ref-type="supplementary-material"} lower panel). The *lyso*-phospholipid content as determined by positive and negative ion survey scans (120--600 m/z) showed no significant differences between the WT, the resistant strains and the MT add back. *Leishmania* PE species often contain the cyclopropyl fatty acid C19Δ \[[@pntd.0005171.ref041], [@pntd.0005171.ref042]\]. Several PE species were identified and confirmed by accurate mass spectrometry ([S5 Fig](#pntd.0005171.s005){ref-type="supplementary-material"}, [S3 Table](#pntd.0005171.s011){ref-type="supplementary-material"}), by parent ion ES-MS-MS of 295 m/z corresponding to the C19Δ fragment ([S6A Fig](#pntd.0005171.s006){ref-type="supplementary-material"}) and by daughter ion fragmentation by ES-MS-MS ([S6B--S6G Fig](#pntd.0005171.s006){ref-type="supplementary-material"}). The total fatty acid content, including C19Δ species, was determined in wild-type and resistant parasites. Derivatization of the total fatty acid content into their FAME enabled their quantification by GC-MS. Total ion chromatograms were obtained from mid log phase parasites for each strain and an example of the fatty acid distribution profile for wild-type *L*. *infantum* strain can be found in [S7A and S7B Fig](#pntd.0005171.s007){ref-type="supplementary-material"}. All FAME species were identified and their relative percentages calculated ([Table 2](#pntd.0005171.t002){ref-type="table"}). Most FAMEs were similar across all strains, including the relative ratio of saturated and unsaturated fatty acids. However, both AmB1000.1 and MF200.5 mutants showed a \~4-fold increase in C19Δ. The second fatty-acid species altered significantly in both mutants corresponded to fatty acid C24:0, which was increased by 1.6-fold in AmB1000.1 and MF200.5 ([Table 2](#pntd.0005171.t002){ref-type="table"}). Episomal transfection of a wild-type copy of MT in AmB1000.1 restored both C19Δ and 24:0 fatty acids to near wild-type levels ([Table 2](#pntd.0005171.t002){ref-type="table"}). 10.1371/journal.pntd.0005171.t002 ###### Total fatty acid content quantification (relative %). GC-MS was used to determine the fatty acid content of the different *L*. *infantum* strains, in comparison with wild-type parasites. [S7 Fig](#pntd.0005171.s007){ref-type="supplementary-material"} includes an example for total ion chromatogram of derivatised fatty acids from lipid extracts of *L*. *infantum* 263 wild-type. Data are the mean of three independent experiments. Statistical significance between the mock-transfected wild-type and the rest of the strains was evaluated by unpaired two-tailed t test (\*\*\**p ≤ 0*.*001*). {#pntd.0005171.t002g} Fatty acid Retention time (min) Relative quantification (%) ------------------- ---------------------- ----------------------------- ----------- ------ ----------- 14:0 31.2 2.1 2.3 2.6 2.2 16:1 35.3 0.2 0.1 0.1 0.1 16:0 35.6 2.8 2.9 2.9 2.8 18:3 n = 3 38.8 1.1 0.9 0.7 0.9 18:3 n = 6 38.9 1.1 1.0 0.9 0.9 18:2 39.1 28.9 27.9 28.1 28.4 18:1 39.2 24.6 20.3 24.7 19.6 18:0 39.7 19.8 19.0 18.7 19.2 C19Δ 40.8 0.7 4.1\*\*\* 0.8 3.8\*\*\* 20:4 n = 3 42.1 2.9 2.5 2.7 2.7 20:4 n = 6 42.2 2.1 2.1 2.0 2.0 20:3 42.4 0.5 0.2 0.3 0.5 20:2 42.6 3.8 4.0 3.9 4.2 20:1 42.8 0.6 0.4 0.5 0.5 20:0 42.9 0.2 0.1 0.1 0.1 22:6 n-3 44.1 1.8 0.9 1.4 1.7 22:5 n-3 44.2 2.0 1.8 1.7 1.9 22:2 44.7 1.6 1.2 1.6 1.4 24:5 n-6 46.1 1.8 1.9 1.7 1.8 24:4 n-6 46.4 2.7 2.8 2.9 3.5 24:0 48.9 1.7 2.8\*\*\* 1.6 2.7\*\*\* **% Saturated** 26.6 27.1 25.7 27.0 **% Unsaturated** 73.4 72.9 74.3 73.0 Sphingolipids (SL) are major component of *Leishmania* membranes \[[@pntd.0005171.ref041]\] and *Leishmania* do not process the *de novo* biosynthetic pathways for neither sphingomyelin, nor complex glyco-SLs \[[@pntd.0005171.ref041]\]. The overall relative abundance of IPCs and PIs, as well as the total amount of lipid-inositol, were thus determined in mutants AmB1000.1 and MF200.5 and compared to those of wild-type parasites ([Table 3](#pntd.0005171.t003){ref-type="table"}). *L*. *infantum* 263 wild-type presented a relative distribution of 39% IPC and 61% PI of their total lipid containing inositol species. AmB1000.1 and MF200.5 mutants showed a significant increase in IPC formation, with percentages of 51%-53% for IPC and 49%-47% for PI, respectively. Interestingly, introduction of the wild-type copy of the MT in the AmB1000.1 mutant resulted in a shift back of the IPC-PI percentages to 45% -55%, values closer to wild-type parasites. The AmB and MF mutants also showed a significant increase in total inositol containing lipids, highlighting the significant increase in IPC content within these cells. The total inositol-lipid content was returned to wild-type like levels in the add-back line ([Table 3](#pntd.0005171.t003){ref-type="table"}). 10.1371/journal.pntd.0005171.t003 ###### IPC/PI ratio and total inositol quantification relative to the wild-type strain. Data are the mean ± s.d. of two independent experiments run in triplicate. Statistical significance of INO between the mock-transfected wild-type and the rest of the strains was evaluated by unpaired two-tailed t test (\**p ≤ 0*.*05*, \*\**p ≤ 0*.*01*, \*\*\**p ≤ 0*.*001*). {#pntd.0005171.t003g} Abundance relative to total INO (%) Total INO relative to WT (%) Abundance relative to WT (%) ------------------ ------------------ ------------------------------------- ------------------------------ ------------------------------ **Ldi263 wt** IPC 39.0 ± 1.4 \-\-- \-\-- PI 61.0 ± 1.4 \-\-- **AmB1000.1** IPC 51.4 ± 1.0\*\*\* 127.3 ± 6.0\*\* 167.5 ± 2.6\*\*\* PI 48.6 ± 1.0\*\*\* 102.5 ± 2.6 **AmB1000.1+MT** IPC 44.9 ± 1.1\*\*\* 106.2 ± 1.7\* 121.1 ± 1.7\*\* PI 55.1 ± 1.1\*\*\* 96.0 ± 1.7 **MF200.5** IPC 52.8 ± 1.1\*\*\* 116.1 ± 4.6\* 156.5 ± 2.4\*\* PI 47.2 ± 1.1\*\*\* 90.5 ± 2.4\* To complete our analysis of lipids, and because of the suspected mode of action on AmB, we also measured sterols by GS-MS in AmB1000.1, AmB1000.1+MT, MF200.5 and wild-type cells ([S8 Fig](#pntd.0005171.s008){ref-type="supplementary-material"}). Levels of sterols in wild-type cells ([Table 4](#pntd.0005171.t004){ref-type="table"}) were similar to what observed in other species \[[@pntd.0005171.ref043], [@pntd.0005171.ref044]\]. The level of ergosterol and 5-dehydroepisterol were greatly decreased in AmB1000.1 while 4-methyl-8,24-cholestadienol was strongly increased ([S8 Fig](#pntd.0005171.s008){ref-type="supplementary-material"}; [Table 4](#pntd.0005171.t004){ref-type="table"}). Surprisingly ergosterol was markedly decreased in MF200.5 as well, but episterol was now the dominant sterol ([S8 Fig](#pntd.0005171.s008){ref-type="supplementary-material"}; [Table 4](#pntd.0005171.t004){ref-type="table"}). In contrast to other lipid species (Tables [1](#pntd.0005171.t001){ref-type="table"}, [2](#pntd.0005171.t002){ref-type="table"} and [3](#pntd.0005171.t003){ref-type="table"}) transfection of wild-type MT in AmB1000.1 did not modify the distribution of sterols ([S8 Fig](#pntd.0005171.s008){ref-type="supplementary-material"}, [Table 4](#pntd.0005171.t004){ref-type="table"}). 10.1371/journal.pntd.0005171.t004 ###### GC-MS analysis of sterols in Ldi263, MF200.5, AmB1001.1 and AmB1000.1+MT. Data are the mean ± s.d. of three independent experiments. Relative percentages based upon peak areas. TIC of chromatogram 39.50--43.50 min shown in [S8 Fig](#pntd.0005171.s008){ref-type="supplementary-material"}. ND-not detected. {#pntd.0005171.t004g} Label Molecular Ion (*m/z*) Annotation Relative Percentages ------- ----------------------- ------------------------------ ---------------------- ---------- ---------- ---------- **1** 386 Cholesterol 11.0±1.8 10.7±1.6 5.3±0.7 5.2±0.4 **2** 396 5-dehydroepisterol 68.8±4.2 19.6±2.0 1.3±0.2 1.3±0.1 **3** 396\. Ergosterol 7.8±0.6 trace trace trace **4** 382 Cholesta-5,7,24-trienol 12.2±0.8 0.7±0.2 13.9±1.2 12.8±1.5 **5** 366 Episterol ND 68.2±3.9 ND ND **6** 412 14-methyl-fecosterol ND 0.8±0.1 6.6±0.2 7.0±0.2 **7** 384 Zymosterol ND ND 1.2±0.1 1.1±0.1 **8** 384 cholesta-7,24-dienol ND ND 7.6±0.3 6.7±0.5 **9** 398 4-methyl-8,24-cholestadienol ND ND 73.1±6.0 65.9±4.7 Discussion {#sec018} ========== Current clinical policies against visceral leishmaniasis in the endemic region of Bihar in India support the use of sequential treatments relying on administration of liposomal AmB followed by a short 7-days administration of MF \[[@pntd.0005171.ref021], [@pntd.0005171.ref022]\]. Resistance is not a current threat for AmB clinical use \[[@pntd.0005171.ref005]\] although *L*. *donovani* field strains unresponsive to AmB have been reported \[[@pntd.0005171.ref006]\]. Drug combination treatments involve shorter dosing schedules, which increases compliance and are less prone to select drug-resistant parasites compared to classical single-drug therapies \[[@pntd.0005171.ref008], [@pntd.0005171.ref045]\]. However, the existence of shared resistance mechanisms between two of the main leishmanicidal agents could lead to treatment failure and emergence of new refractory parasitic populations. Indeed, it has been recently reported that *L*. *donovani* can become resistant to drug combinations, including the combination of AmB/MF, and that the multi-resistant phenotypes are maintained in amastigotes \[[@pntd.0005171.ref046]\]. Similarly, the characterization of several *L*. *donovani* field isolates revealed that susceptibility profiles against MF and AmB were positively correlated, thus identifying a risk for cross-resistance \[[@pntd.0005171.ref046]\]. In this study we demonstrate that cells selected for AmB can be cross-resistant to MF and the reverse is also true. We also provide evidence for one pathway of cross-resistance through lipid content modifications, which is seemingly linked to mutations in the MT. This cross-resistance should lead to careful considerations when sequential treatments are considered in endemic regions \[[@pntd.0005171.ref021], [@pntd.0005171.ref022]\] especially because resistance is observed also for intracellular parasites ([Fig 2](#pntd.0005171.g002){ref-type="fig"}). The mutations in the MT gene of *L*. *infantum* MF200.5 and *L*. *major* AmB1080.3 lines are predicted to be in conserved domains of the MT ([S1 Fig](#pntd.0005171.s001){ref-type="supplementary-material"}), whilst the G433S substitution in the MT of *L*. *infantum* AmB1000.1 is located nearby the DKTGTLT motif of the ATPase phosphorylation domain \[[@pntd.0005171.ref047]\]. The lack of a structure for MT renders it difficult to predict the impact of this mutation on the function of the transporter, but lipid composition is altered and reintroducing an episomal copy of MT revert in part the mutation phenotype (Tables [1](#pntd.0005171.t001){ref-type="table"}, [2](#pntd.0005171.t002){ref-type="table"} and [3](#pntd.0005171.t003){ref-type="table"}). Reintroduction of a wild-type copy of MT also reverted resistance to both MF and AmB ([Fig 1](#pntd.0005171.g001){ref-type="fig"}), and intracellular survival ([Fig 2](#pntd.0005171.g002){ref-type="fig"}). Links between mutations in P-type ATPase and the import of different phospholipid species has been reported in mammalian cells \[[@pntd.0005171.ref048]--[@pntd.0005171.ref050]\]. Mutation in MT in MF200.5 is correlated with a decrease in MF uptake ([Fig 3](#pntd.0005171.g003){ref-type="fig"}) and this likely contributes to MF resistance. In contrast, the uptake of MF is only minimally changed in AmB1000.1, suggesting that the mutation in MT may lead to MF cross-resistance by another mechanism. One possibility is that the changes in lipid composition in AmB1000.1 allow for a higher accumulation of MF within its membrane. Indeed, the AmB1000.1 mutant expressing wild-type MT through episomes had a susceptibility to MF identical to that of wild-type parasites ([Fig 1](#pntd.0005171.g001){ref-type="fig"}) despite accumulating twice as much MF ([Fig 3](#pntd.0005171.g003){ref-type="fig"}). Our accumulation experiments cannot distinguish between genuine intracellular uptake from accumulation of the drug at the level of the plasma membrane. It is thus possible that lipid-related compensatory mechanisms developed by AmB1000.1 may lead to less MF intracellular translocation. While the regulatory subunit Ros3 \[[@pntd.0005171.ref007]\] is not mutated in AmB1000.1, we cannot exclude an impaired expression and its contribution, if any, to AmB-MF cross-resistance remains to be clarified. Overexpressing a wild-type version of the MT in AmB1000.1 did not totally reverted its resistance to AmB and other mechanisms, such as the upregulation of proteins implicated in protection against drug-induced oxidants \[[@pntd.0005171.ref023]\], should complement the protective effect achieved through changes in cellular-membrane lipids. For example, the level of sterols is altered in AmB1000.1 and this is independent of the MT ([Table 4](#pntd.0005171.t004){ref-type="table"}). This change in sterol may contribute to resistance but further work would be required to isolate the genes involved in those sterol changes. It is salient to point out that AmB1000.1 shows aneuploidy for 6 chromosomes ([S1 Table](#pntd.0005171.s009){ref-type="supplementary-material"}) and that many SNPs were detected in its genome, 85 of which were in coding sequences ([S2 Table](#pntd.0005171.s010){ref-type="supplementary-material"}, [S2](#pntd.0005171.s013){ref-type="supplementary-material"} and [S3](#pntd.0005171.s014){ref-type="supplementary-material"} Datasets), and some may have a role in AmB resistance. Besides resistance, MT-mediated lipid changes may also impact on parasite-macrophage interactions. As observed here for *L*. *infantum* AmB1000.1 and MF200.5, AmB-resistant \[[@pntd.0005171.ref011], [@pntd.0005171.ref051]\] and MF-resistant cells \[[@pntd.0005171.ref052]\] had previously been reported to be less infective, suggesting that parasites with mutations in the MT may be selected against in the absence of drug pressure. This is not always the case however, as MF-resistant *L*. *major* \[[@pntd.0005171.ref052]\] or *L*. *amazonensis* \[[@pntd.0005171.ref053]\] with mutations in the MT did not show reduced infectivity, and the potential for resistance to multiple drugs by a single point mutation remains real. Several studies have pointed to modifications in lipid metabolism as a major factor for both AmB \[[@pntd.0005171.ref011], [@pntd.0005171.ref054]\] and MF resistance in *Leishmania* \[[@pntd.0005171.ref013], [@pntd.0005171.ref014], [@pntd.0005171.ref019], [@pntd.0005171.ref020], [@pntd.0005171.ref040]\]. Among the lipid species whose abundances were similarly altered in the AmB1000.1 and MF200.5 mutants is the noteworthy increase of IPC, PI and 19Δ-containing PE species. Interestingly, preliminary lipid quantification experiments with the *L*. *major* AmB1080.3 mutant revealed variations in both negative and positive ion ES-MS survey scans ([S3C Fig](#pntd.0005171.s003){ref-type="supplementary-material"}) similar to those found for mutant *L*. *infantum* AmB1000.1 ([S4C Fig](#pntd.0005171.s004){ref-type="supplementary-material"}). A recent study found that *L*. *donovani* parasites exposed to MF exhibit increased levels of three PE species \[[@pntd.0005171.ref055]\], which support our findings regarding the increased levels of 19Δ-containing PE for both mutants. Interestingly, disruption of the LdMT-LdRos3 complex in *L*. *donovani* was shown to affect the asymmetry of membrane lipids and resulted in an increased exposure of PE to the exoplasmic leaflet of the plasma membrane \[[@pntd.0005171.ref056]\]. However, in contrast to our resistant selected lines, these disrupted lines did not show significant differences in the total amount of PE and PI \[[@pntd.0005171.ref056]\]. Analysis of FAMEs also confirmed the significant increase of C19Δ for both AmB1000.1 and MF200.5 mutants ([Table 2](#pntd.0005171.t002){ref-type="table"}). It has been reported that C19Δ is not abundant in wild-type *L*. *infantum* \[[@pntd.0005171.ref036]\], and since its formation requires a high energy cost \[[@pntd.0005171.ref057]\] it should provide important survival gain to the parasite. Interestingly, studies in *E*. *coli* have shown how cyclopropane fatty acids (CFAS)-mediated membrane modifications protect against environmental stresses (temperature, pH, salt concentration, etc.) \[[@pntd.0005171.ref058]. [@pntd.0005171.ref059]\]. However, the impact of CFAS on the fluidity of the lipid bilayer is unclear, some studies supporting that defect in CFAS would result in a decrease of membrane fluidity \[[@pntd.0005171.ref060]\] but others suggesting an enhanced fluidity \[[@pntd.0005171.ref061]\]. In addition to C19Δ, lignoceric acid 24:0 was also increased in both AmB1000.1 and MF200.5 mutants. Lignoceric acid was previously detected in MF-resistant *L*.*-donovani* promastigotes but not in wild-type cells \[[@pntd.0005171.ref020]\]. The content of SLs was also found to be modified in resistant parasites, with IPC being greatly increased in both mutants ([Table 3](#pntd.0005171.t003){ref-type="table"}). This correlates well with the enhanced levels of d16:0/18:0-IPC identified in MF-treated *L*. *donovani* \[[@pntd.0005171.ref055]\]. Complex SLs like IPC are found in the outer leaflet of plasma membranes together with sterols and contribute to the overall plasma-membrane fluidity \[[@pntd.0005171.ref062]\]. Moreover, changes in IPC-PI levels, most probably due to PI being the precursor for IPC formation, may also act as a compensatory effect due to the lack of optimal fatty acid containing diacyl-phospholipids secondary to defective MT activity. This in turn may also impact upon the PI species available for GPI anchor biosynthesis and usage. While MT point mutations play a direct role in MF-resistant strains, it seems they can also induce lipid-content modifications in AmB resistant cells that favour cross-resistance to MF, most likely by increasing the ability of the parasite to accumulate MF in its membrane rather than hindering its uptake. This would suggest that resistance to AmB but also cross-resistance to MF is much likely related to changes in lipid composition and possibly in the intracellular routing of MF. Different mutations in MT have thus diverse functional consequences on the transport of MF ([Fig 3](#pntd.0005171.g003){ref-type="fig"}). These results reinforce and expand the knowledge on the complex nature of AmB resistance in *Leishmania* parasites \[[@pntd.0005171.ref023], [@pntd.0005171.ref046], [@pntd.0005171.ref063]\], by including a new and unexpected character that also triggers MF cross-resistance. This could have significant impact on the control of this neglected disease. Supporting Information {#sec019} ====================== ###### Graphical representation of the different mutations identified in MT for three Leishmania drug resistant strains. The diagram includes the different conserved protein domains identified for the MT (GenBank: AAQ82704.1). (TIF) ###### Click here for additional data file. ###### Effect of the transfection of an episomal copy of the wild-type gene LinJ.16.1240 in the AmB1000.1 mutant background. \(A\) EC~50~ determination curves in the presence of amphotericin B for Ldi263 wt (62.01 ± 5.00 nM), AmB1000.1 (1.97 ± 0.04 μM) and AmB1000.1+LinJ.16.1240 (1.89 ± 0.11 μM) cell lines over 72 h. (B) EC~50~ determination curves in the presence of miltefosine for Ldi263 wt (8.34 ± 0.40 μM), AmB1000.1 (31.50 ± 2.80 μM) and AmB1000.1+LinJ.16.1240 (29.89 ± 3.03 μM) cell lines over 72 h. An average of at least three independent biological replicates is shown, with error bars depicting the standard error of the mean. EC~50~ values were determined by means of Graphpad Prism5 using non-linear regression analysis. (TIF) ###### Click here for additional data file. ###### Preliminary study for the characterization of mutant L. major Friedlin AmB1080.3. \(A\) EC~50~ determination curves in the presence of miltefosine for LmF wt+mock (9.05 ± 0.89 μM), AmB1080.3+mock (54.84 ± 5.30 μM) and AmB1080.3+MT (4.43 ± 0.62 μM) cell lines over 72 h. (B) EC~50~ determination curves in the presence of amphotericin B for LmF wt+mock (0.16 μM ± 0.01), AmB1080.3+mock (1.50 ± 0.07 μM) and AmB1080.3+MT (0.60 ± 0.02 μM) cell lines over 72 h. An average of at least three independent biological replicates is shown, with error bars depicting the standard error of the mean. EC~50~ values were determined by means of Graphpad Prism5 using non-linear regression analysis. (C) Negative (upper panel) and Positive (lower panel) ion ES-MS survey scans (600--1000 m/z) of total lipid extracts from mutant L. major Friedlin AmB1080. PEs: phosphatidylethanolamines, IPCs: inositol-phosphoceramides and PIs: phosphatidylinositols. An example of one of three independent biological replicates (which showed similar profiles) is shown. (TIF) ###### Click here for additional data file. ###### Identification of abundance changes to the membrane lipids in AmB- and MF-resistant parasites. Negative ion ES-MS survey scans (600--1000 m/z) of total lipid extracts from Ldi263 wt (A upper panel), MF200.5 (B upper panel), AmB1000.1 (C upper panel) and AmB1000.1+MT (D upper panel). Positive ion ES-MS survey scans (600--1000 m/z) of total lipid extracts Ldi263 wt (A lower panel), MF200.5 (B lower panel), AmB1000.1 (C lower panel) and AmB1000.1+MT (D lower panel). PEs: phosphatidylethanolamines, IPCs: inositolphosphoceramides and PIs: phosphatidylinositols. The different species identified in the ES-MS surveys are detailed in the Supplementary S4 Table. An example of one representative of three independent biological replicates is shown. (TIF) ###### Click here for additional data file. ###### High resolution orbitrap mass spectrometry survey scans (600--1000 m/z) of total lipid extracts from Ldi263 wt; negative ion mode (A), positive ion mode (B). (PDF) ###### Click here for additional data file. ###### Characterisation of the PE species containing C19Δ in total lipid extracts from Ldi263 wt. \(A\) ES-MS/MS parents of 295 m/z (C19Δ acyl fragment) in negative mode (600--1000 m/z). (B-G) are daughter ion ESI-MS/MS spectra of ions identified in S6A Fig: 716, 730, 744, 774, 788, 802 m/z respectively. (PDF) ###### Click here for additional data file. ###### Example of total ion chromatogram of derivatised fatty acids from lipid extracts of L. *infantum* 263 wild-type. \(A\) Chromatogram including all the FAMES species with retention times spanning from 26.00 to 50.00 min for mid log phase parasites for each strain detailed in [Table 1](#pntd.0005171.t001){ref-type="table"}. (B) Magnification of the chromatogram for the identification of the spectral peak with a retention time of ∼40 min corresponding to C19Δ. The identity of C19Δ FAME was confirmed by retention time and spectral comparison with bacterial FAME standards, which includes C19Δ. (PDF) ###### Click here for additional data file. ###### GC-MS Sterol Analysis of *Leishmania* strains. TIC of chromatogram 39.50--43.50 min for (A) Ldi263, (B) MF200.5, (C) AmB1000.1, (D) AmB1000.1+MT. Numbered peaks refer to [Table 4](#pntd.0005171.t004){ref-type="table"} for identification. (TIF) ###### Click here for additional data file. ###### Estimated ploidy for Ldi263AmB1000.1 and its WT parent (Ldi263WT). (PDF) ###### Click here for additional data file. ###### SNPs deduced from the sequencing of LdiAmB1000.1 and absent from the sequencing of its Ldi263WT parent. (PDF) ###### Click here for additional data file. ###### Mass spectrometric analysis of phosphatidylethanolamine species containing cyclopropyl fatty acid in *Leishmania infantum*. (PDF) ###### Click here for additional data file. ###### InDels. (PDF) ###### Click here for additional data file. ###### AmB_SNPs-InCDS-NoSyn-Homozygous. (PDF) ###### Click here for additional data file. ###### AmB_SNPs-InCDS-NoSyn-Heterozygous. (PDF) ###### Click here for additional data file. [^1]: The authors have declared that no competing interests exist. [^2]: **Conceptualization:** CFP PL TKS MO.**Formal analysis:** CFP PL TKS MO.**Funding acquisition:** TKS MO.**Investigation:** CFP IMV MCB MR TKS GR.**Methodology:** CFP PL TKS MO.**Resources** LR.**Supervision:** TKS MO.**Visualization:** CFP MO.**Writing -- original draft:** CFP PL TKS MO.**Writing -- review & editing:** LR TKS MO. [^3]: Current address: University of Glasgow, Glasgow, United Kingdom

{#sp1 .466}

Transplant patients may have preexisting diabetes mellitus or develop transient postoperative hyperglycemia or permanent new-onset diabetes after transplantation (NODAT). NODAT increases mortality rate due to allograft rejection and also leads to increased cardiovascular disease.\[[@ref1]\] NODAT excludes transient posttransplant hyperglycemia and pretransplant undiagnosed diabetes, which are often unrecognized due to lack of effective pretransplant screening.\[[@ref2]\] Posttransplant diabetes mellitus, that is, presence of diabetes in posttransplant setting irrespective of timing of diabetes, may be preferred nomenclature.\[[@ref3]\] NODAT risk is highest in the first 6 months post transplant and increases progressively over time.\[[@ref1]\] Perioperative stress, infection, high calcineurin inhibitors\' (CNI) exposure, and glucocorticoid induction can cause transient post-transplantation hyperglycemia (TPH) in 90% of kidney allograft recipients, so diagnosis of NODAT should be delayed until the patient is on stable maintenance doses of immunosuppressants, with stable kidney graft function and in the absence of acute infections.\[[@ref4][@ref5]\] TPH is also a risk factor for NODAT.\[[@ref1]\] The prevalence of NODAT is variable and is reported to be between 2% and 50%, probably owing to inconsistent definitions used for diagnosing NODAT.\[[@ref1]\] Though fasting glucose has a low sensitivity for diagnosing NODAT and oral glucose tolerance test (OGTT) is considered the gold standard test for diagnosis of NODAT screening patients using fasting glucose or afternoon glucose monitoring (induced by morning steroid) can identify high-risk patients requiring OGTTs.\[[@ref6]\] HbA1c greater than 6.5% is unlikely to be false-positive but may not exclude NODAT in the initial 3 months post transplantation due to anemia and rapid diabetes onset, so a lower HbA1c cut-off may be more sensitive. No glycemic indicator post transplantation is co-related with long-term outcomes. There are many nonmodifiable and modifiable risk factors for NODAT, both conventional and novel. Pretransplant risk assessment can help individualize therapy and reduce NODAT risk. Nonmodifiable risk factors such as age, sex, and HLA type help in identifying patients at risk, and interventions on modifiable risk factors such as obesity, metabolic syndrome (MS), and HCV/CMV infection may prevent complications and improve outcomes. General risk factors include family history of diabetes mellitus, ethnicity, genetic polymorphisms, increasing age, obesity, MS, and prediabetes.\[[@ref1]\] Transplant-specific risk factors include donor sex, type of underlying renal disease, graft dysfunction, biopsy-proven rejection, specific antirejection agents, cumulative steroid use, tacrolimus level, human leukocyte antigen and ABO mismatch, TPH, and hypomagnesemia.\[[@ref1]\] Pretransplantation insulin resistance contributing to MS is a risk factor for NODAT, but increasingly pancreatic beta cell dysfunction, supported by several genetic (KNNJ11, TCF7L2) polymorphism studies, has gained prominence.\[[@ref1]\] The recent Indian study by Choudhury *et al*. found 17% prevalence of NODAT with beta-cell secretory defect, high waist circumference, and trough tacrolimus level as important factors for predicting NODAT.\[[@ref7]\] Antirejection therapies such as glucocorticoids and sirolimus lead to increase in insulin resistance, while steroids increase hepatic glucose output and decrease insulin secretion in high doses. Low-dose corticosteroid is preferred but not steroid withdrawal as it increases acute rejection requiring pulse steroid which counterbalance metabolic beneficial effect.\[[@ref1]\] Split corticosteroid dosing may also reduce glycemic variability and peak hyperglycemia. CNI causes dose-dependent reduction in insulin secretion due to vacuolization and degranulation of islet cells and insulin gene transcription defect. Tacrolimus has five times higher risk of diabetes than cyclosporine. Immunosuppressive regimen therapies providing best patient and graft survival should be used irrespective of their NODAT risks as transplant rejection outweighs risks of NODAT.\[[@ref5]\] Admission evaluation includes medical history, family and past history of dysglycemia, MS. Patients discharged without hyperglycemia should have fasting plasma glucose testing at least weekly during the first month, then every 3 months for 1 year, and annually thereafter.\[[@ref2]\] Lifestyle modification (LSM) may reduce risk of NODAT.\[[@ref2]\] Stepwise approach with LSM followed by oral antihyperglycemic agents (AHA) and then insulin is appropriate for management of late (\>6 months) NODAT, but in TPH and early NODAT the reverse regimen is preferred.\[[@ref2]\] Dose adjustments or cessation of oral AHA agents in the context of renal allograft dysfunction should be individualized. Insulin is the only safe and effective agent during high glucocorticoid doses and acute illness early post transplant. Basal insulin therapy in early posttransplant hyperglycemia (\<3 weeks) reduced risk of NODAT within the first year post transplantation by 73%.\[[@ref8]\] Insulin therapy in early postoperative phase may prevent NODAT through β-cell protection.\[[@ref8]\] In summary, NODAT has a sudden onset and rapid development of complications such as graft failure, infection, and cardiovascular disease compared with a garden variety of type 2 diabetes mellitus. NODAT have unique causal factors, some of which may be modifiable. But there are diagnostic difficulties and treatment targets and agents are not well-defined. Early and aggresive management of hyperglycemia post transplant may prevent β-cell apoptosis and possibly prevent NODAT and its complications. NODAT confers an adverse prognostic outcome for transplant survival and mortality, so pretransplant risk stratification and posttransplant NODAT screening is imperative.

1. Introduction =============== Importance of C-allocation in biotechnology and ecology Recently, microalgae have attracted increased attention in the field of biomass production and biosynthesis of feed stock components or high valuable products like polyunsaturated fatty acids, antioxidants, pigments like carotenoids which can be used for food colorants, *etc*. The most challenging approach to increase the productivity is not only to develop photo-bioreactors which offer optimal conditions for algae growth, but also to increase the yield of a special product \[[@B1-metabolites-04-00453]\]. Since an increasing number of algal genomes are now available and genetic transformation protocols have been published to apply system biology for the green alga *Chlamydomonas reinhardtii* \[[@B2-metabolites-04-00453]\], but also for *Ostreococcus tauri* \[[@B3-metabolites-04-00453]\], the diatom *Phaeodactylum tricornutum* \[[@B4-metabolites-04-00453]\] and for *Nannochloropsis sp.* \[[@B5-metabolites-04-00453]\], there is increasing need to develop quantitative methods to measure the carbon flow from photosynthetic primary products into the component of interest. Typically, metabolic engineering is used to channel the carbon from sugar to lipid, starch, protein or secondary products like pigments. For this purpose the cells are genetically transformed either by inhibition of competitive pathways or by stimulation of the pathways that lead to the product. After having checked integration, expression and translation of the new gene, the simplest test to look for the metabolic effect of the transformation is to measure the content of the compound of interest during growth. However, genetic transformations of microalgae can also alter the overall activity of the cells leading for example to lower photosynthetic performance. In that case it is difficult to decide if the concentration of the product is influenced due to inefficient metabolic engineering or due to the impaired photosynthetic activity. To understand changes in metabolic fluxes in the cellular network of carbon allocation the method of choice is metabolic profiling using carbon isotopes in pulse chase experiments. However, this methodology is time consuming, costly and can be improved by well-defined sampling, when the time points of the metabolic switch from one pathway to another can be identified (see below). Since a metabolic profile is a snap shot of the metabolic state at a given moment, the changes in the carbon allocation network can be detected only by comparing snap shots from different time points. However, cells grown in a light dark cycle or in batch culture change the metabolic profile also in response to the environmental conditions, internal clock oscillations or reorganize the metabolic network during a transition period to turn back to the initial state. This problem can be solved if the time scale of metabolic dynamics can be measured by fast and easy to handle methods which can be used to find a well-defined sampling. Here we present a new method of choice involving the application of Fourier Transform Infrared (FTIR) spectroscopy, which delivers information about the relative size of the cellular macromolecular pools that might change in response to metabolic engineering (e.g., protein, carbohydrate and lipid). A second problem of metabolic engineering is to measure cell internal carbon flux rates. A metabolic profile delivers at best relative concentrations of a set of metabolites but it does not tell the kinetics of changes in the pool size of the components identified. This is especially important for those metabolites which occur in different compartments in different activated states, e.g., phosphorylated *versus* non-phosphorylated or oxidized *versus* reduced form. Here again FTIR spectroscopy of intact cells can be used to measure flux rates between and into the macromolecular pools, but with much less molecular resolution than metabolic profiling. The paper gives an example for this approach. 2. Methodology of FTIR Spectroscopy and Statistics in Spectra Interpretation ============================================================================ Traditional analysis of C-allocation or protein:lipid:carbohydrate ratios by means of biochemical methods require relative huge amount of biomass (in the mg range) and due to their time consuming procedure they are not suitable for high throughput measurements. However, analyzing the macromolecular composition that consider the whole pool of cellular compounds need new approaches to become faster and decrease the amount of sample necessary to measure. UV-VIS spectroscopy itself has been established and accepted for quantifying a wide range of cellular compounds (e.g., pigments in algae cells). Far less attention has been given to FTIR spectroscopic methods, although a single spectra can be obtained in the range from less than 500.000 cells down to single cells compared to several mg necessary for standard biochemical methods. FTIR spectroscopy was used for the first time on phototrophic organisms by Kansiz *et al.* \[[@B6-metabolites-04-00453]\] to discriminate different cyanobacterial strains but not to quantify the biomass composition. However, this work initiated a variety of different methods and applications using FTIR spectroscopy to analyze cell biochemical features in phototrophic prokaryotes and algae. Within this review we will focus on the determination of biochemical cell composition and quantification of cell constituents (e.g., lipid content). Some key works in the last decade showed the applicability of IR spectroscopic methods for quantifying cellular protein, lipid and carbohydrate contents in phototropic prokaryotes and algae. The methods in general are based on the absorption of IR light by specific chemical bonds reviewed in Movasaghi *et al.* \[[@B7-metabolites-04-00453]\]. By absorbing IR radiation changes in dipole moment of chemical bonds are required. This leads to the generation of specific vibrational features e.g., bending and stretching at different wavenumbers due to the different frequencies (energy value) of IR radiation \[[@B8-metabolites-04-00453]\]. Therefore, a single macromolecule has a very characteristic absorption spectrum in the far red region of the light spectra composed of different absorption bands like amide I, amide II and amide III band for protein. These absorbance features of single macromolecules result in a complex cell spectrum ([Figure 1](#metabolites-04-00453-f001){ref-type="fig"}) which is composed of all IR active biochemical cell components. Since vibrational bands of macromolecules are strongly overlapping each other, the interpretation of cell spectra is one of the crucial challenges in IR spectroscopy. Several methods have been developed so far, ranging from the use of peak ratios \[[@B9-metabolites-04-00453]\] over the determination of peak integrals \[[@B10-metabolites-04-00453],[@B11-metabolites-04-00453],[@B12-metabolites-04-00453]\] up to a complex mathematical fitting procedure of spectra using single reference spectra for a multiple linear regression analysis \[[@B13-metabolites-04-00453]\]. Different extinction coefficients for each chemical bond makes the interpretation even more complicate \[[@B8-metabolites-04-00453]\]. Therefore, it has to be distinguished between semi-quantitative changes of biochemical components on the one hand and absolute quantitative determination on the other hand. For semi-quantitative determination of changes in C-allocation like altered ratios of protein per lipid it may be sufficient to use absorbance peak ratios \[[@B14-metabolites-04-00453],[@B15-metabolites-04-00453]\]. Furthermore, ratios of band integrals can be used to monitor for example the relative lipid content \[[@B12-metabolites-04-00453]\]. These methods can be adequate for most of the applications e.g., monitoring cultures for responses to nutrient limitations (see below), but cannot provide absolute values. A semiquantitative approach for the determination of the macromolecular pools in species of different algae strains has been introduced by \[[@B16-metabolites-04-00453]\]. Within this study the cellular protein contents were measured by biochemical methods to correlate the ratio between the FTIR absorbance of the pool of interest and that of proteins \[[@B16-metabolites-04-00453]\]. Other applications that need more precise information on absolute quantitative amounts of protein, lipid and carbohydrate require more complex examinations. Here, two different main methodical approaches are established. The use of band integrals calibrated by external single standard substances for quantitative analysis of biomass composition has been performed by Pistorius *et al.* \[[@B10-metabolites-04-00453]\]. Here the absorption bands for lipid, protein and carbohydrate were baseline corrected and subsequently integrated ([Figure 1](#metabolites-04-00453-f001){ref-type="fig"}a). A calibration of these integrals has been performed using single substance spectra of phosphatidyl choline for lipid, creatine phosphokinase for protein and malt extract from starch hydrolysate for carbohydrate. ![Comparison of two different methods for spectra interpretation. (**a**) Peak integral quantification of macromolecule contents \[[@B10-metabolites-04-00453]\]. Peak integrals have been marked (green: lipid; blue: protein; orange: carbohydrate); (**b**) Spectra reconstruction by reference spectra (green: lipid; blue: protein; orange: carbohydrate) of the same cell spectra according to \[[@B13-metabolites-04-00453]\]. Vertical lines indicate main peaks of the cell spectra.](metabolites-04-00453-g001){#metabolites-04-00453-f001} Since Pistorius *et al.* \[[@B10-metabolites-04-00453]\] measured in transmission mode, similar procedures can also be applied to attenuated total reflectance (ATR) spectroscopy \[[@B11-metabolites-04-00453],[@B17-metabolites-04-00453]\]. Mayers *et al.* \[[@B11-metabolites-04-00453]\] characterized 30 samples by biochemical methods and subsequently related the band integrals with the percentage of compound per dry weight. A least squares polynomial regression analysis has then be performed to calibrate the peak area against the macromolecular contents for a quantitative determination of cellular macromolecules. Laurens and Wolfrum \[[@B17-metabolites-04-00453]\] also used a Partial Least Square (PLS) regression analysis but for the determination of different lipids. These experiments were based on exogenously added lipid and showed an accurate prediction of added lipid levels for triglycerides and phospholipid. This work shows the application of mathematical methods to separate between different lipids, but the quantification of this lipid biomass will still need a true calibration. However, both these methods do not take into account the overlap of vibrational bands from the different macromolecular pools. Both preferred different spectral regions for the calibration of the lipid content (2984--2780 cm^−1^ methyl and methylene groups or stretching of ester bonds at 1740 cm^−1^), which can be explained only by different cellular biochemical ratios. However, the overlap of other constituents is species-specific e.g., reference spectra of carbohydrate show a distinct peak at 2927 cm^−1^ which is in the range of the methyl bands used for lipid quantification. Absolute amounts of macromolecules will then be overestimated to some extent. This overlap of absorption bands holds true for at least all of the analyzed macromolecules ([Figure 1](#metabolites-04-00453-f001){ref-type="fig"}b). Alternatively, a multiple linear regression protocol has been performed for fitting single reference spectra into the original cell spectra to calculate the amount of each macromolecule \[[@B13-metabolites-04-00453]\] minimizing this overestimation ([Figure 1](#metabolites-04-00453-f001){ref-type="fig"}b). This spectra reconstruction also needs the calibration by reference substances. Wagner *et al.* \[[@B13-metabolites-04-00453]\] used triglycerol for lipid, bovine serum albumin for protein and glucose for carbohydrate as reference. Each of the single reference substance has been calibrated by correlating the absorption intensity *versus* the substance concentration. By a multiple linear regression algorithm, all reference spectra are then fitted into the cell spectra at once. The remaining error should then be as small as possible and is normally expressed as RMSE (root mean square error). The sum of the fitted reference spectra lead to a spectrum similar to the cell one. Furthermore, the residuals (*i.e.*, the differences between the original cell spectrum and the fitted one) can be used to identify absorption bands that are due to other IR active substances present in high amounts or characterized by high extinction coefficient. Since the reference spectra have been calibrated, the absorption of the fitted reference spectra can be used to quantify the contributing amount of the respective molecule into the cells. However, both methods have shown the applicability of IR spectroscopy with some restrictions. The validity of Lambert Beers law has been evaluated by Wagner *et al.* \[[@B13-metabolites-04-00453]\] showing that sample thickness is a crucial parameter. Also attenuation of the infrared radiation across the sample pellet can influence an absolute IR spectroscopic quantification \[[@B16-metabolites-04-00453]\], so that the amount of probe material used has to be in a narrow range and handled with care \[[@B13-metabolites-04-00453]\]. Furthermore, the method is currently restricted to cells that do not have any silica shells, since silica absorption bands overlap with the absorption bands used for carbohydrate determination \[[@B13-metabolites-04-00453]\]. Nevertheless, if the C:N ratio of the biomass is known, also FTIR spectra of Si-containing diatoms can be analyzed \[[@B18-metabolites-04-00453],[@B19-metabolites-04-00453]\]. Since FTIR-spectroscopy is still a newly developed method, several authors attempted to validate the measurements by independent techniques. Liu *et al.* \[[@B20-metabolites-04-00453]\] extracted the cellular lipids by methanol chloroform to confirm the band assignment of the lipids. Furthermore, they could show from the isolated lipid the amount of carbohydrate that is bound to the lipid acids. Since the absorption of the lipid band at 2926 cm^−1^ is much higher than the carbonyl absorption at 1740 cm^−1^ the authors suggest to use the band around 2926 cm^−1^ for quantifying the lipid content. They further used a normalization procedure by scaling the spectra to the amid I as well as amid II bands and could document that the normalization to the amid II band is more suitable than that performed on amid I. Meng *et al.* \[[@B21-metabolites-04-00453]\] validated their FTIR analysis by traditional biochemical methods, but explained only approx. 50%--70% of the dry weight. With the assumption that more than 90% of the biomass is composed of lipid, protein and carbohydrate, FTIR methods are still questionable with respect to absolute quantifications. However, the macromolecular contents measured via IR spectroscopy expressed as band area ratios correlated to the biochemically quantified macromolecular ratios. In some cases it could be interesting to analyze the biochemical content and distribution at the single cell level. Powerful IR radiation gained from a synchrotron light source can open this possibility, thus one can obtain FTIR spectra with a spatial resolution of 5 µm to 10 µm \[[@B22-metabolites-04-00453],[@B23-metabolites-04-00453]\]. However, new IR imaging techniques reach almost 0.54 µm of spatial resolution \[[@B24-metabolites-04-00453]\] but absolute quantitative measurements have not been published so far. 3. Application in Biotechnology =============================== FTIR analysis can be a helpful tool in biotechnology especially for development of lipid producing phototrophic organisms by changing culture conditions or genetically engineering. Already existing bioreactors need a dense monitoring program to identify (i) contaminations and (ii) to assess the yield of the compound of interest. The optimization of lipid production for biofuels for example needs fast and high throughput methods to monitor cellular macromolecular concentrations and lipid yields. Lipid quantification by means of FTIR spectroscopy has been shown to be a reliable method compared to the gravimetric determination \[[@B25-metabolites-04-00453],[@B26-metabolites-04-00453]\], which needs milligrams of biomass that are hard to obtain in high throughput screening. For this reason several authors used FTIR spectroscopy to evaluate the lipid productivity and cell growth with respect to heterotrophic growth, different nutrient limitations and abiotic factors like light and temperature \[[@B16-metabolites-04-00453],[@B20-metabolites-04-00453],[@B26-metabolites-04-00453],[@B27-metabolites-04-00453],[@B28-metabolites-04-00453],[@B29-metabolites-04-00453]\]. Under nitrogen limitation green algae cells change their cellular lipid and carbohydrate contents \[[@B20-metabolites-04-00453],[@B30-metabolites-04-00453]\]. Using FTIR spectroscopy Liu *et al.* \[[@B20-metabolites-04-00453]\] could discriminate cells with higher lipid contents by a principle component analysis (PCA) analysis. This method can further be used to describe the trade-off between lipid content, growth rate and cell density as a result of the nutrient starvation. In *Chlamydomonas vulgaris* cultures it was shown that FTIR spectroscopy can be used to follow the overall carbon allocation pattern during different growth phases and under different nitrogen and CO~2~ availability \[[@B27-metabolites-04-00453]\]. The authors could demonstrate that high growth rate and high lipid accumulation can be obtained under different growth conditions. From these data they could calculate the lipid productivity per dry weight, from which the best time of harvesting can be obtained \[[@B27-metabolites-04-00453]\]. In *Nannochloropsis sp.* the response to organic solvents for lipid "milking" has been investigated \[[@B31-metabolites-04-00453]\]. Some solvents increased the total lipid content but FTIR spectroscopy has shown that the composition itself does not change. In conclusion, cell homeostasis can be monitored during an experiment by FTIR spectroscopy, even if the production rate of a specific compound is increased and excreted by the cell into the medium \[[@B31-metabolites-04-00453]\]. These examples showed the variety of applications of FTIR spectroscopy to monitor the effects of the culture conditions on the macromolecular response of the cells. Hence, FTIR it is now established to compare several species to optimize the lipid yield Jiang *et al.* \[[@B32-metabolites-04-00453]\], or to identify carbohydrate accumulating strains that are also potential bio-ethanol producers \[[@B33-metabolites-04-00453]\]. In conclusion FTIR spectroscopy is a fast and simple method for evaluating and monitoring culture conditions not only in bioreactors by providing information on the macromolecular content of cells \[[@B28-metabolites-04-00453],[@B34-metabolites-04-00453]\]. Besides these monitoring of culture condition dependent accumulation of lipid and carbohydrate, FTIR spectroscopy can be applied for screening of new developed mutants with altered biochemistry. An example is the analysis of the biochemical pathway of carbon assimilation via C4 metabolism in diatoms. It has been hypothesized that mutants of a silenced PPDK would have a higher lipid accumulation. This hypothesis could be confirmed by chemometric analysis of IR spectra \[[@B35-metabolites-04-00453]\]. Finally, not only lipid accumulation, but the whole cellular macromolecule ratios can be useful for biotechnology. Mutants of cyanobacteria are potentially useful for hydrogen production. Therefore, different Synechocystis strains have been modified in antenna size, leading to altered phycobillisomes and increased electron transport rates. These patterns can then be monitored by FTIR to document the metabolic changes in mutants compared to wildtype cells \[[@B36-metabolites-04-00453]\]. 4. Physiology and C-Dynamics ============================ FTIR spectroscopy can be used for the optimization of phototrophic organism performance in biotechnology. Thereby it shows the steady state measurements of the macromolecular pools. However, this method can also be used to evaluate physiological responses within a variety of applications, ranging from whole cell physiology \[[@B37-metabolites-04-00453],[@B38-metabolites-04-00453]\] to responses on a very small molecular scale, e.g., the response to down regulation of single genes \[[@B35-metabolites-04-00453]\]. In addition measurements can be performed on different time scales ranging from days down to several minutes. Since the FTIR spectroscopy reflects all cellular macromolecules, a measure of stored energy or an energy balance of the biomass can be determined \[[@B37-metabolites-04-00453]\]. In more detail, measurements of energy balance are generally performed via gas exchange or electron transport rates based on fluorescence measurements. However, for detailed information of the cellular energy distribution, the electron allocation into different carbon pools has to be considered. There are two main physiological strategies concerning carbon allocation, especially under changing growth conditions. Either cells stay in a very stable homeostatic macromolecular composition \[[@B39-metabolites-04-00453]\] or a significant change of macromolecule ratios can be measured \[[@B19-metabolites-04-00453]\]. Nevertheless, for almost all physiological experiments FTIR spectroscopy can deliver reliable information about the cellular carbon distribution and energy content. This has been shown by a number of publications ranging from acclimation to N, Si and Fe deficiencies \[[@B19-metabolites-04-00453],[@B22-metabolites-04-00453],[@B40-metabolites-04-00453]\] or cold temperature acclimation \[[@B41-metabolites-04-00453]\] up to different CO~2~ levels \[[@B27-metabolites-04-00453]\]. As pointed out before, monitoring of such physiological responses can be done with a timescale of several minutes. In *Chlamydomonas reinhardtii* grown in a 13 h/11 h light/dark cycle the carbohydrate to protein ratio rises up to 2 fold in the first four hours of the light phase showing that the newly formed photosynthates are directed to the carbohydrate pool, whereas later on this ratio remains constant indicating that carbon is funneled equally to carbohydrate and protein. During the dark period the carbon stored in the carbohydrate pool is reallocated to protein synthesis \[[@B13-metabolites-04-00453]\]. This shows that carbon allocation is highly regulated and a time series of FTIR measurements as biochemical snap shots can resolve this dynamics to define the best time point of harvest for special application. Based on the knowledge of these dynamics the more time consuming and costly methods with higher resolution for the detection of molecular compounds (such as mass spectroscopy based metabolomics) can be applied more efficiently. The following example shows how FTIR spectroscopy can be used to identify the time point of metabolic switches in cells. In this experiment the diatom *Phaeodactylum tricornutum* was grown at two different light intensities in monochromatic blue light (BL, 470 nm) and red light (RL, 650 nm), to analyze the function of the blue-light receptors in diatoms \[[@B42-metabolites-04-00453]\]. The design of this experiment allowed the authors to obtain equal photosynthetic rates per cell by adjusting the incident RL and BL. Thus, changes in the macromolecular composition only related to light color modulated changes in the carbon allocation pattern. [Figure 2](#metabolites-04-00453-f002){ref-type="fig"} shows the relative cellular protein and carbohydrate contents. Under low BL light carbon is distributed equally into protein and carbohydrate during the whole light period and only at the end of the light phase the carbon flows to a higher proportion to the carbohydrate ([Figure 2](#metabolites-04-00453-f002){ref-type="fig"}a). In low RL this carbon distribution pattern is observed only for seven hours, after that the carbon is preferentially diverted to carbohydrate ([Figure 2](#metabolites-04-00453-f002){ref-type="fig"}c). Under high light, when the carbon flux produced from photosynthesis is higher, the new carbon is preferentially directed to carbohydrate after 3--5 h ([Figure 2](#metabolites-04-00453-f002){ref-type="fig"}b,d). This shows that the carbon allocation pattern depends on both the flux and the photoreceptor controlled activities of enzymes directing the carbon to different pathways. The dynamics show that well-defined sampling for more detailed analysis by mass spectroscopy based metabolomics can be efficiently assisted by FTIR-spectroscopy by identifying the transition phase when the carbon flux direction is changing. If the overall carbon flux can be measured, as it is the case of photosynthetic active cells either by means of Chla *in vivo* fluorescence or CO~2~ uptake, the absolute flux into the different carbon pools can be quantified. Together with the semi-quantitative information of key metabolites obtained from mass spectroscopy analysis the dynamics in flux control in the metabolic network can be displayed in quantitative terms. This shows that traditional gas flux measurements, FTIR spectroscopy and mass spectroscopy based metabolomics can synergistically act to display a metabolic network picture of life. ![Carbohydrate (orange) and protein (blue) levels of *Phaeodactylum tricornutum* grown at blue light (465 nm) of (**a**) 10 µmol absorbed photons m^−2^ s^−1^ (Q~Phar~) and (**b**) 30 µmol absorbed photons m^−2^ s^−1^ or at red light (660 nm) of (**c**) 10 µmol absorbed photons m^−2^ s^−1^ and (**d**) 30 µmol absorbed photons m^−2^ s^−1^ in a 14 h light/10 h dark cycle. The time axes correspond to the time in the light phase. FTIR spectra were detected every 2 h and used for the calculation of the macromolecular composition (according to Wagner *et al.* 2010). The carbohydrate and protein levels were normalized to the values at 3 h into the light phase (t~3~) (modified according to \[[@B42-metabolites-04-00453]\].](metabolites-04-00453-g002){#metabolites-04-00453-f002} 5. Conclusions and Perspectives =============================== FTIR spectroscopy delivers a fingerprint of the macromolecular composition of cells. Recent advances in chemometrics allow the quantification not only of major cellular pools like lipid, carbohydrate and protein but also minor components of interest. However, the resolution of complete metabolite pattern with quantitative or semi-quantitative estimates is not possible and is not in perspective. Compared to the high resolution of metabolome analysis, FTIR spectroscopy allows fast analysis with minimal amounts of material with the additional advantage of high throughput. This opens the perspective of time-resolved analysis which allows us not only to measure pattern but also fluxes. This potential is useful for biotechnological process control and to assist a well-defined sampling for mass spectroscopy based metabolomics. We thank the Deutsche Forschungsgemeinschaft (DFG) for financial support (Grant Number Wi 764/10, Wi 764/14 and Wi 764/19). Christian Wilhelm prepared the intoduction and the last section of the manuscript. Heiko Wagner and Andrea Fanesi prepared the section two to five of the manuscript. Anne Jungandreas performed the measurements and analysis of the blue red acclimation experiments. Heiko Wagner and Christian Wilhelm wrote the article. All authors read and approved the final manuscript. The authors declare no conflict of interest.

Background {#Sec1} ========== Head and neck cancer (HNC) is the sixth most common cancer worldwide \[[@CR1]\]. Currently, there are more than 62,000 people living with head and neck cancer in the UK \[[@CR2]\]. In Scotland, where incidence rates are significantly higher than the rest of the UK, a 37 % rise is predicted in the next 10 years \[[@CR3]\]. Patients are surviving longer and, due to human papillomavirus, are younger at diagnosis. The functional, psychological and social consequences of HNC cancer and its treatment can be severe and chronic. Treatment includes a combination of surgery, radiotherapy and/or chemotherapy, and the side effects of these can interfere with some of the most fundamental aspects of daily living, including eating, drinking, communication and appearance. Improved survival has been achieved at the expense of increased morbidity, especially in relation to swallowing problems or 'dysphagia' \[[@CR4]\]. Dysphagia affects up to two thirds of patients undergoing combined chemoradiotherapy (CRT) \[[@CR5]\]. Patients report major deterioration in their swallowing, and little evidence of recovery over time, with younger survivors reporting the most severe problems \[[@CR6]\]. Early side effects of CRT[1](#Fn1){ref-type="fn"} include pain, mucositis (inflammation of the mucous membranes) and xerostomia (dry mouth). Late effects include fibrosis (scarring) of the soft tissues, affecting the safety and efficiency of swallowing \[[@CR7], [@CR8]\]. Aspiration rates (food or fluid entering the airway) are high (≥60 %), leading to frequent hospitalisation for chest infection, pneumonia and even death \[[@CR5], [@CR8]\]. The effects of CRT contribute to significant weight loss, and 50--70 % of patients require a feeding tube during or after treatment \[[@CR9]\]. Tube dependency reduces the need to swallow, so increasing the likelihood of fibrosis of the muscles involved in swallowing, and in the long-term, 'disuse atrophy', sometimes leading to complete inability to swallow. Swallowing difficulties have a significant negative emotional and physical impact on social eating, return to work and everyday quality of life \[[@CR7]\]. Recent reviews \[[@CR10], [@CR11]\] suggest that prophylactic swallowing exercises may improve a range of short- and long-term outcomes, as they increase the blood flow to muscles, reducing or preventing fibrosis, and maintaining the range and speed of swallowing movements. However, not all trials have had positive results \[[@CR12]--[@CR14]\], and a number of questions remain, including the optimal timing, selection and duration of exercises, the achievement of intervention fidelity and, importantly, the support necessary to improve adherence \[[@CR10]\]. Only 13--14 % of participants practise swallowing exercises as recommended \[[@CR15], [@CR16]\]. An unpublished study found that patients with high levels of pain during treatment struggled with performing regular exercises and that it was difficult to focus on preventative exercises in the context of facing a potentially life-limiting disease. Furthermore, the intervention was not seen as integral to their overall care \[[@CR17]\]. There is, however, evidence that those who maintain their exercise schedule achieve improved swallowing outcomes \[[@CR18]\] and are significantly less likely to need a feeding tube \[[@CR19]\]---an important consideration given that feeding tubes are associated with significantly poorer quality of life \[[@CR20]\]. The multi-dimensional nature and impact of swallowing difficulties and the importance of psychological and behavioural factors on adherence to swallowing exercises has not been adequately studied \[[@CR21], [@CR22]\]. The need for further research, particularly prospective randomised trials to investigate the benefit of an intervention which includes pre-, peri- and post-treatment swallowing exercises has been highlighted repeatedly \[[@CR10], [@CR23], [@CR24]\]. This study aims to develop and test the feasibility of a Swallowing intervention Package (SiP) designed in partnership with patients, speech and language therapists (SLTs) and other members of the head and neck multi-disciplinary team (MDT), for patients undergoing primary or adjuvant CRT or radiotherapy for head and neck cancer. A development phase (phase 1), informed by focus groups with patients and consensus workshops with professionals, produced a swallowing exercise schedule, staff manual and SiP package for patients, including written materials, videos, reminder materials and an electronic e-SiP 'app' to support adherence. This protocol, based on version 5 (24.2.16) describes the feasibility study (phase 2) for this complex intervention, which aims to answer four research questions in preparation for a future multi-centre trial:What is the feasibility of delivery and potential impact of the SiP for patients and head and neck cancer teams?What are the barriers and facilitators to adherence and fidelity to the SiP?Does an e-support system (e-SiP) have potential to support patients to perform their exercises and for collecting patient-reported data through video diaries?Are the study processes and outcome measures feasible and acceptable to patients and staff? Methods/design {#Sec2} ============== This feasibility study uses quantitative and qualitative research methods, within a quasi-experimental parallel group design, to assess whether patients will tolerate and adhere to the SiP intervention, which aspects of the intervention can be implemented and which cannot, whether treatment fidelity can be achieved across different contexts, whether study processes and outcome measures will be feasible and acceptable and to what extent the intervention is likely to have an impact on swallowing dysfunction and quality of life. Forty patients will be recruited to the SiP intervention group, from three hospital outpatient settings in Scotland, and a maximum of 30 patients will be recruited to a 'usual care' group, from sites in Scotland and the North East of England. This is a pragmatic design to allow for different service delivery models in different geographical areas. There is a trade-off between precision (recruitment parameter of numbers retained by study endpoint) and number of patients recruited to the study. A sample size of 59 is thought to be sufficient to identify any potential problems in feasibility which have a 5 % probability of occurrence at least once, with 95 % confidence \[[@CR25]\]. Several new sites have expressed an interest in participating and are being adopted in order to enhance recruitment. To identify the variety of practical issues and problems faced by a 'real world' and inclusive 'typical' cohort of patients with head and neck cancer and the applicability of the intervention across this cohort, we will include people with/without prophylactic feeding tubes, from different demographic, diagnostic (e.g. oral, pharyngeal, laryngeal) and treatment (CRT and radiotherapy alone) groups. Inclusion criteria: Patients diagnosed with an index primary mucosal squamous cell carcinoma of oral cavity, nasopharynx, pharynx or larynx or unknown primary, who are planned for chemoradiotherapy or radiotherapy ≥30 fractions. This may include patients who have undergone surgery before their chemoradiation or radiotherapy. Patients must be able to give informed consent. Exclusion criteria: Patients with synchronous or metachronous head and neck cancer primaries, those undergoing palliative treatment, non­English speakers, previously diagnosed dysphagia unrelated to HNC and those who have undergone total glossectomy or total laryngectomy. Screening, recruitment and consent {#Sec3} ---------------------------------- All potentially eligible patients will be identified from the head and neck cancer MDT meetings. Eligible patients will be seen at routine appointments for treatment planning and invited to participate in the study by a member of the clinical team. The e-SiP app is being tested in Tayside, where patients eligible for the intervention are also eligible for the app version. The delegated individual will explain the study to the patient, give them the Patient Information Sheet (PIS) and answer any questions. Consent will be sought a minimum of 24 h later by a research nurse. The PIS for the intervention group explains that a Swallowing intervention Package of exercises and support has been developed and that we are now asking people to try this out and report their experiences, with the hope that we will be able to use these findings to underpin a larger trial. The PIS for the control group explains that patients will have no change to their usual care and will continue to have the swallowing support normally provided by their treatment centre. Both intervention and control patients are informed that taking part in the study will not affect any usual treatment that they may receive and that taking part may not benefit them personally. At the time of consent the participant will also be asked to fill in the baseline study questionnaires, and a standard letter will be sent to their GP specifying the recruiting hospital and whether or not the patient is in the intervention or 'usual care' group. All patients will also be asked whether they and/or a close family member or friend would be willing to be approached for a subsequent qualitative interview with the research fellow, between 6 weeks and 3 months after treatment finishes, to explore their experiences. In order to assess the representativeness of the sample, a screening log will record the number of patients who are (a) ineligible, (b) refuse to participate in the intervention or control cohorts and (c) decline specific aspects of the study (e.g. clinical assessments, questionnaires), as well as their reasons, where possible. The rights of patients to refuse to participate or withdraw without giving reasons will be respected. If patients do volunteer a reason for refusal, or for wishing not to use the e-SiP app versions, then these will be logged. Clinical and socioeconomic characteristics {#Sec4} ------------------------------------------ At the point of recruitment, all participants are allocated a unique ID number. At baseline, the following data will be extracted from NHS patient records and recorded by research nurses/clinical teams in the study case report form: Date of birth, sex, ethnicity, Index of Multiple Deprivation derived from patient's residential postcode, education level, work status, social support, tumour type, anatomic site, TNM stage, size and position of primary tumour, smoking (current/ex-smoker and pack years), use of alcohol, body mass index, weight, surgery, planned CRT regime including chemotherapy regime, type of radiotherapy (e.g. intensity-modulated radiation therapy (IMRT) or conventional), number of fractions, radiation field, feeding tube and feeding regime (whether in situ or planned). At the end of treatment, CRT details will be checked and amended as necessary to record the treatment actually delivered, including overall radiation dose to primary clinical target volume and dose to the constrictor muscles. Additional data will be recorded on: Method of feeding throughout treatment, any admissions to hospital, analgesic regime, treatment complications and weight loss will be recorded on the post-treatment case report form and updated as appropriate. Any adverse events which are not expected as a relatively common side effect of CRT (e.g. aspiration pneumonia, death) will also be recorded on this form. All case report forms are pseudonymised by their unique ID number, before being transferred to the Research Team for further analysis. Relevant staff within the project team have undergone Good Clinical Practice (GCP) training. Outcome measures {#Sec5} ---------------- A combination of patient-reported outcome measures (PROMs), adherence measures and clinical swallowing assessments will be used prior to intervention (baseline), at the end of treatment, 3 and 6 months post-treatment (see Fig. [1](#Fig1){ref-type="fig"}). Data will provide key estimates of the variability for primary and secondary outcomes to inform the best primary outcome measure and sample size estimations for a future trial. Where possible, data are being collected during or after routine clinic appointments in order to enhance retention and complete follow-up.Fig. 1Flow diagram of data collection points for SiP study Patient-reported outcome measures {#Sec6} --------------------------------- Swallowing function will be measured by the MD Anderson Dysphagia Inventory (MDADI) \[[@CR26]\] (likely primary outcome measure). Quality of life will be measured by the European Organisation for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire for cancer patients (QLQ-C30) with the supplementary Head and Neck Module (HN35) \[[@CR27], [@CR28]\] and the EuroQol five dimensions questionnaire (EQ-5D) \[[@CR29]\]. Adherence measures {#Sec7} ------------------ We have developed a 'Rehabilitation Intervention Beliefs' (RIB) questionnaire, based on Cooper et al.'s \[[@CR30]\] work on exploring patients' beliefs about cardiac rehab, to assess beliefs that may affect adherence to the SIP, including perceived necessity, concerns, practical barriers and perceived suitability for the intervention. The RIB and Brief Illness Perceptions (Brief IPQ) \[[@CR31]\] questionnaires will assess motivation towards the SIP as well as illness cognitions and emotional representations. At the 3-month data collection point, patients will be asked to complete the post-intervention RIB questionnaire, which includes three additional questions about the patient's overall experience of the SiP. Structured diary record cards (Fig. [2](#Fig2){ref-type="fig"}) will enable recording of daily exercise patterns and frequency, factors which interfered with or promoted adherence, e.g. pain, and any comments about undertaking the exercises. Furthermore, in a small sub-sample of patients, we will pilot the use of technology (e-SIP app) for reporting on adherence; iPads will be provided for the recording of video diaries, enabling real-time data on the performance of swallowing exercises and experiences to be collected.Fig. 2Diary record cards Clinical assessments of swallowing ability {#Sec8} ------------------------------------------ 100 ml Water Swallow Test (WST) \[[@CR32]\]. The WST records the duration and number of swallows required to swallow 100 ml of water. Quick, inexpensive, with good reliability and validity, excellent patient and clinician acceptability, it is sensitive to measure changes over time and follows the expected pattern in differentiating between different cancer treatment regimes. Unlike instrumental tests, the WST does not provide any information on swallow pathophysiology. It has therefore been selected to provide interim information about changes to swallowing function, as an adjunct to the more expensive Fibreoptic Endoscopic Evaluation of Swallowing (FEES) assessment.Performance Status Scale for HNC (PSS-HN) \[[@CR33]\]. This three-item scale evaluates dietary texture restrictions and social eating. It has excellent discriminatory properties, good inter-rater reliability and sensitivity to differences in performance and change over time.FEES. This instrumental test allows quantification of endpoints of swallowing pathophysiology, swallowing efficiency and airway invasion (laryngeal penetration and aspiration) \[[@CR34]\]. The alternative to FEES---videofluoroscopy (VF) \[[@CR35]\]---exposes patients to radiation, is more expensive, less practical to implement and is associated with greater loss to follow up. Blinded FEES assessments will be analysed by the SLT consultants using reliability testing. This test is used routinely in clinical practice but is not carried out for all head and neck cancer patients. An information leaflet is always provided to patients and informed consent is taken before the FEES is carried out.Functional Oral Intake Scale (FOIS) \[[@CR36]\]. This scale is scored according to a level that best describes the patient's oral intake. There are seven levels from level 1: nothing by mouth to level 7: total oral diet with no restrictions. Intervention (SiP) {#Sec9} ------------------ The SLT based in the relevant centre will teach the exercise programme and provide supporting materials, including diary record sheets, a study water bottle, a SiP patient manual[2](#Fn2){ref-type="fn"} (Fig. [3](#Fig3){ref-type="fig"}), swallowing exercise videos and an iPad containing the e-SiP (which will be piloted in a small number of patients) (Figs. [4](#Fig4){ref-type="fig"} and [5](#Fig5){ref-type="fig"}). SLTs reviewed the current literature on prophylactic exercises alongside the evidence base for specific swallowing exercises linked to the pathophysiology of post-irradiated dysphagia and achieved consensus over the exercise regime for SiP (see Table [1](#Tab1){ref-type="table"}). A staff manual was developed based on the Behaviour Change Taxonomy (BCT) \[[@CR37]\]. The original 26 point taxonomy was reduced to a 20-point taxonomy using staff consensus. During the consensus workshop conducted as part of phase 1, a multi-disciplinary group of staff reviewed the BCT and devised the manual based on how this could be applied in the SiP intervention (for example, in setting goals with patients), both for initial delivery and ongoing patient support. An example page from this manual is shown in Fig. [6](#Fig6){ref-type="fig"}. The aim of the manual is to develop maximum intervention fidelity across the different delivery sites \[[@CR38]\]. A small number of interventions will be recorded or observed to check the inclusion of particular behaviours displayed by patients or staff during intervention delivery.Fig. 3Picture of patient manuals and water bottle Fig. 4Screenshot of e-SiP Fig. 5Screenshot of e-SiP Table 1Swallowing exercisesTarget number for each dayExerciseSession 1Session 2Session 3Mendelsohn101010Masako101010Effortful swallow101010Shaker3× 1-min hold3× 1-min hold3× 1-min hold30 head lifts30 head lifts30 head liftsJaw exercises (if instructed, once patient is no longer able to insert three fingers between top and bottom teeth)5× 30-s straight opening5× 30-s straight opening5× 30-s straight opening5× 1-s jaw swing to each side5× 1-s jaw swing to each side5× 1-s jaw swing to each sideSLTs make the decision whether to reduce/remove an exercise depending on the individual patient Fig. 6Taxonomy of behaviours for SiP This initial consultation will take place before the start of CRT. At the end of the consultation, the SLT will complete an intervention receipt questionnaire, detailing the main components of the intervention delivered and an assessment of the patient's understanding and motivation towards the intervention. e-SiP delivery {#Sec10} -------------- Patients in Tayside who have expressed willingness to trial the e-SiP app version will be provided with an iPad during their initial intervention delivery appointment. They will also be asked to sign an additional consent form which sets out their responsibilities in taking care of the iPad and the boundaries for appropriate use. Two user guides are included with the iPads; the patient user guides details the use of the e-SiP app. In addition, the staff user guide explains the additional log-in functionality that staff may access to change features, e.g. the number of exercise repetitions, within the e-SiP app for that particular patient. All staff delivering the iPad have received training on encouraging the use of the e-SiP, e.g. recording video diaries, over and above the basic staff manual for the SiP delivery. On-treatment review {#Sec11} ------------------- During CRT, either the SLT or clinical nurse specialist (CNS) will review each patient regularly, according to usual on-treatment review practice in each centre. Symptoms will be managed according to local treatment and supportive care protocols. Additional on-treatment review opportunities with nurses, therapy radiographers and dietitians will be used to provide feedback on effort and progress, encourage maintenance of the intervention and monitor adherence to exercises. A weekly assessment log based on the BCT will enable practitioners to document the content of interactions about swallowing that they have had with the patient during treatment, allowing resource use, intervention fidelity and costs to be estimated. In addition, a purposive sample of consultations (pre-, during and post-treatment) between SLTs and patients in each centre will be observed or recorded, using an observation checklist \[[@CR39]\]. Patients will be encouraged to bring a family member or friend with them for intervention visits, in order that they can facilitate and support aspects of the SiP at home. Data monitoring committee {#Sec12} ------------------------- Although considered, it was deemed that the project did not justify the role of a separate Data Monitoring Committee (DMC). The intervention consists of clinically validated swallowing exercises and assessments that are used already in clinical practice, and there are no anticipated risks to participants over and above those related to the patient's cancer treatment. The role of the DMC has therefore been subsumed into the function of the Study Steering Committee, who have overseen the monitoring of adverse events. Within this, MW, ND, N H-W and EK, have formed the Data Management Team to discuss day-to-day queries concerning data entry and analysis. Qualitative evaluation {#Sec13} ---------------------- Ten to 15 patient participants (and 5 carers) will be invited for an in-depth interview between 6 weeks and 3 months after treatment finishes. Some of these interviews may be joint patient/carer dyads. Purposive sampling will be used to maximise the potential for including participants from different socio-demographic backgrounds, with a range of swallowing difficulties, who adhered to the exercises to varying degrees and used different adherence support, e.g. paper based or e-SiP. Both those who participated in the intervention and a small number from the 'usual care' group will be included. The main research fellow for the study will (a) conduct face-to-face or telephone interviews with patient and carer participants, to explore experiences, challenges and adherence to SiP and to identify strategies used by patients (and carers) to overcome any challenges, including which support measures were most and least useful and (b) undertake face-to-face or telephone interviews with 15 key members of staff (SLTs, CNS', clinic nurses, radiographers, dietitians) from feasibility study sites to elicit views of the SiP and e-SiP, problems with implementation and perceptions of patient outcomes \[[@CR40]\]. These data will provide important insights into the intervention refinements necessary for a future RCT. Data analysis {#Sec14} ------------- All data will be entered from pseudonymised case report forms (CRFs) onto a secure password protected database based at the study coordinating centre. Quantitative data will be entered onto a SPSS v21 database using a coding manual. Qualitative data will be transcribed verbatim and entered into N-VIVO for analysis. Data will be handled, computerised and stored in accordance with the Data Protection Act 1998. All study data will be retained in accordance with Research Governance and local policy. Qualitative data will be analysed using framework analysis and quantitative outcome data using SPSS v21 or Stata v14. Descriptive statistics will be used to summarise demographic and clinical data, recruitment rates and proportions of those participating, dropping out and providing usable outcome measures. These will be used to populate a CONSORT study flow chart. Missing data problems will be explored to ascertain if missing at random or not. Patient-reported and clinical swallowing outcomes will be estimated between treatment-as-usual and intervention groups as mean or median scores with respective standard deviations or interquartile ranges. Wherever possible, outcome descriptions, summaries and comparisons will be presented together with estimates of precision in accordance with stipulated CONSORT guidelines ([www.consort-statement.org](http://www.consort-statement.org)). As this is a feasibility study, the data will not be pooled with any future main trial data. This study is not powered to detect significant differences between groups, and therefore, no formal tests of statistical significance will be undertaken. Statistical models utilising the repeated measures, with baseline measures used as covariates in statistical models will be used to ascertain likely key independent predictors explaining variability of the range of possible outcome measures in this feasibility study. We do not expect to achieve significant differences from baseline to follow-up measure(s) between or within groups as this study is inadequately powered and is quasi-experimental; however, we expect to estimate relative effect sizes (and 95 % CI) for each potential main outcome measure, correlation between measurements, and intra-cluster (site) correlation coefficient for use in sample size simulations for a future randomised trial; this study aims to ascertain if the data collection is sufficient to yield usable data in hypothesis testing in a definitive multi-centre trial and to assess whether the direction of effect is in favour of the intervention. Preliminary psychometric validation of the RIB will also be undertaken, including Cronbach's alpha (for internal consistency), correlation analyses for test-retest reliability and construct validity. The EQ-5D will be used to calculate quality-adjusted life years (QALY) gains and used along with an assessment of resource implications to inform a full economic evaluation in a future definitive trial. With respect to the analysis of the e-SiP, the e-SiP itself is not considered part of the intervention (it does not represent any change in actual exercise protocol), but is a means of testing different ways of data collection, improving any missing data problem and supporting adherence. Although we intend to incorporate the e-SiP data into our analysis, our sample of patients using the e-SiP will be very small due to the available number of iPads. However, we will conduct a sensitivity analysis whereby we analyse the data with and without the e-SiP sub-sample to ascertain if there are any changes to effect size and variance. We will assess e-SiP use to see which aspects of the app patients use most and to conduct a comparison of entries between the paper and e-SiP diaries. Discussion {#Sec15} ========== The potential for swallowing exercises to improve short- and long-term outcomes in patients with head and neck cancer is increasingly recognised, but the evidence base remains limited. The support of specialist SLTs and other members of the MDT is known to be important, but the extent to which patients routinely have access to SLTs before and during treatment is extremely variable \[[@CR41]\]. This is due in part to our limited knowledge about the best ways to deliver swallowing interventions and support people who have or are at risk of developing swallowing difficulties as a result of treatment for HNC. Patients undergoing CRT experience significant fatigue and other side effects, which are likely to challenge their ability to perform swallowing exercises, but the experiences of patients and the factors influencing adherence to exercises on a day-to-day basis are largely unknown. In this study, data collected relating to patient and carer attitudes, beliefs and experiences during and after CRT will provide important insights into potential barriers and facilitators to exercise adherence in a range of real-life contexts. These will inform future studies. This collaborative study has taken a unique approach to the development of a patient-centred and evidence-based swallowing intervention, which aims to encourage and support adherence to daily swallowing exercises during and after CRT, whilst being practical for SLTs and MDTs to implement for a wide range of patients. The introduction of the e-SiP provides an exploration of the use of technology in delivering this intervention, which might be especially pertinent given the increasing numbers of younger HNC patients. We have worked in partnership with SLTs, patients and carers to develop this complex intervention. The study provides an opportunity to examine the feasibility of delivering and participating in a supported swallowing intervention across several different NHS sites and will provide the evidence needed to refine intervention and study processes for future research. We will collect a broad range of qualitative and quantitative data, including outcome measures, clinical assessments and interview. This will result in a large amount of diverse data which will inform a definitive multi-centre trial, based on sound theoretical and empirical foundations, feasible for patients and staff and implementable on a wide scale. Dissemination plans {#Sec16} ------------------- We will continue to engage and share findings with a wide range of HNC specialists and SLTs across the UK and Europe. Our study is on the National Cancer Research Institute (NCRI) portfolios for Psychosocial Oncology and Survivorship and Head and Neck Cancers. We will discuss our findings and their implications with the NCRI Clinical Studies Groups, Macmillan and HNC charities and with staff and patients from participating NHS Boards. We will present our work at National and International meetings and will submit findings for publication in clinical peer-reviewed journals. Trial status {#Sec17} ============ The proposal for this study was peer reviewed and then funded by the Chief Scientist Office in July 2014 (ref: CZH/4/1052). Progress reports are submitted to the funder on a regular basis, but no further input to the design, analysis, interpretation or writing up of the data is given by the funder. iPads have been loaned by Throat Cancer Foundation. Phase 1 was approved by NRES Committee North East---Newcastle and North Tyneside 2 in October 2014 (REC:14/NE/1168; NRS R&D: NRS14/ON593). Phase 2 was approved by the East of Scotland Research Ethics Service (EoSRES) in October 2015 (REC: 15/ES/0106; NRS R&D: NRS15/ON703), and participants are being recruited at time of submission. All protocol amendments are communicated to relevant R&D, ethics and research teams. Sponsorship, indemnity and insurance are provided by the University of Stirling. Target completion date for phase 2 is February 2017. We are happy to share the protocol and supporting documents with other researchers. Abbreviations {#Sec18} ============= CNS, clinical nurse specialist; CRF, case report form; CRT, chemoradiotherapy; EORTC HN35, European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Head and Neck; EORTC QLQ30, European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire 30; FEES, Fibreoptic Endoscopic Evaluation of Swallowing; FOIS, Functional Oral Intake Scale; HNC, head and neck cancer; IMRT, intensity-modulated radiation therapy; IPQ, Illness Perceptions Questionnaire; MDADI, MD Anderson Dysphagia Inventory; NCRI, National Cancer Research Institute; PROMs, patient-reported outcome measures; PSS-HN, Performance Status Scale for Head and Neck Cancer Patients; RIB, Rehabilitation Intervention Beliefs; RT, radiotherapy; SiP, Swallowing intervention Package; SLT, speech and language therapist; VF, videofluoroscopy; WST, Water Swallow Test For the purposes of this protocol 'CRT' includes patients undergoing chemoradiotherapy or radiotherapy, either as primary treatment or after surgery. This includes information and advice on normal swallowing, importance of maintaining swallow, tips for eating and drinking, swallowing exercises, diary record cards, strategies for managing distress and anxiety and resources for further information and support We would like to acknowledge and thank our patient advisors---Alan Clarke, Ian McLean and Jamie Rae---and all clinical research nurses, data managers, speech and language therapists, clinical nurse specialists, therapy radiographers, dietitians and medical staff who are supporting the study. Authors' contributions {#FPar1} ====================== FMcA conceived the study, MW led the study and EK provided the day-to-day coordination. ND led the statistical analysis plan. All authors contributed to study design. The manuscript was drafted by MW, EK and KT. All authors commented on and approved the final manuscript. Competing interests {#FPar2} =================== The authors declare that they have no competing interests.

Introduction {#Sec1} ============ The World Health Organization (WHO) announced on Wednesday, March 11, 2020, the new coronavirus "COVID-19" disease as "a global epidemic" (WHO [@CR37]). In response to that, governments around the world have been taking a range of actions and measures, including the closing of schools, worship places, and workplaces, postponing and canceling huge public events, restricting public transport, the lockdown of movement of people, and ceasing domestic and international flights (Bherwani et al. [@CR9]; Bashir et al. [@CR6]). According to many studies, the COVID-19 is believed to transmit through airborne bioaerosol droplets. Nevertheless, different parameters such as the extent of urban air pollution and weather conditions might have a significant impact on the elevated rates of COVID-19 cases (Farhan Bashir et al. [@CR13]; Fareed et al. [@CR12]). After the discovery of the first case of COVID-19 in the KSA on March 2, 2020, a series of actions were taken in response to the COVID-19 pandemic. The first three significant actions taken to control the spread of the pandemic were as follows: (i) suspension of Umrah pilgrimage (March 4), (ii) suspension of all schools and universities (March 8), and (iii) suspension of all international flights (March 9). With the spread of COVID-19 and the absence of vaccine and medication globally, the Saudi government imposed nationwide partial lockdown (from 7 PM to 6 AM) on March 23 followed by the restriction on movement between provinces on March 25. After that, the full lockdown was imposed nationwide on April 6. On May 28, the lockdown was partially lifted in all cities except Mecca, the movement between regions was eased, and shopping malls were opened. Following this, prayers were allowed in mosques except for the Great Mosque of Mecca, and the restrictions on domestic flights, restaurants and cafes, and parks were eased on May 31. The latest action was on June 21, where the partial lockdown was lifted on all regions. However, international flights remain suspended except repatriation flights for residents. Lastly, the Pilgrimage (Haj) was allowed to domestic residents and Saudi nationals only with limited numbers (MoH [@CR21]). The COVID-19 lockdown measures affected many aspects of human activities including vehicle use, public transportation, and industrial processes (Pata [@CR26]; Gautam [@CR14]; Bashir et al. [@CR7]; Shehzad et al. [@CR29]). In addition, several researchers around the world reported that there is a mitigation of air pollution during COVID-19 lockdown compared with before period. For instance, Berman and Ebisu ([@CR8]) assessed the air quality during the COVID-19 pandemic using air quality data, including PM~2.5~ and NO~2,~ monitored in the USA between 2017 and 2020. The study found that statistically significant PM~2.5~ and NO~2~ reductions were noticed as 11% and 26%, respectively. Shakoor et al. ([@CR27]) investigated the changes in levels of air pollutants before and after lockdown periods in USA and China. The results showed that CO, NO~2~, and PM~2.5~ concentration reduced by 19%, 37%, and 1.1%, respectively in USA, while CO, NO~2~, SO~2~, PM~2.5~, and PM~10~ concentration decreased by 27%, 39%, 18%, 18%, and 38%, respectively, in China. Agarwal et al. ([@CR1]) found that lockdown measures in India and China resulted in improvements for the air quality indexes of PM~2.5~ and NO~2~ by 65% and 66% in India and 45% and 37% in China, respectively. Singh and Chauhan ([@CR31]) analyzed air quality changes before and after COVID-19 lockdown measures over India using ground-level and satellite observations. Their results demonstrated that the PM~2.5~ concentrations dropped by to 35%%, compared with pre-lockdown phase. In another study, Sharma et al. ([@CR28]) studied the changes in concentrations of air pollutants before and after lockdown periods in seven cities of India. They reported that PM~2.5~, PM~10~, NO~2~, and SO~2~ reduced by 23--58%, 24--44%, 30--64%, and 3.5--70%, respectively. Zambrano-Monserrate and Ruano ([@CR39]) reported that NO~2~ and PM~2.5~ levels in Ecuador importantly reduced just after the implementation of lockdown measures. At the regional scale, researchers studied the effect of climatic parameters on the spread of COVID-19. Their result showed that transportation and population numbers have a forward association with the infection rates (Barbieri et al. [@CR5]; Ahmadi et al. [@CR2]). This study focuses on investigating the possible effects of the lockdown due to the COVID-19 pandemic on the air quality using meteorological and air quality datasets in the Eastern Province of the KSA. Methodology {#Sec2} =========== Description of the study area {#Sec3} ----------------------------- The Eastern Province is the easternmost of the thirteen provinces of the KSA and located between latitudes of 29.16° N and 19.11° N and longitudes of 44.65° E and 55.66° E. The Eastern Province is the third most populated province with a total population of 4.9 million and the largest province (by area) of the KSA. The weather is hot in the summer season and mild for the rest of the seasons of the year. The long-term annual average values of the temperature and relative humidity are 27 °C and 41%, respectively. Wind systems from northern directions dominate over the area with an average ground-level wind speed of 4.2 m/s. The region has an arid precipitation regime and receives a total of 100 mm rainfall in a year between November and April (Anil et al. [@CR4]). The Eastern Province has been facing severe local and long-range dust storms because of the pressure gradients. During the summer season, hot, dry, and low-level northwesterly winds called "Shamal winds" blowing at a minimum speed of 10 m/s lift dust and sand to the local and remote regions (Karaca et al. [@CR16]; Anil et al. [@CR3], [@CR4]). Remarkable amounts of PM~10~ have been inhaled during episodic periods, resulting in a tremendous rise in the number of hospital admissions regarding respiratory issues (Tsiouri et al. [@CR34]). Tawabini et al. ([@CR32]) recently reported that the daily average PM~10~ concentrations of Dhahran, Khobar, and Dammam districts of the Eastern Province were 177, 380, and 126 μg/m^−3^, respectively, which were quite above than the WHO's daily PM~10~ guideline value of 50 μg/m^−3^ (WHO [@CR36]). Air quality data and study period {#Sec4} --------------------------------- The air quality dataset discussed in this research was obtained from the database of the General Authority of Meteorology and Environment Protection (GAMEP) including CO, SO~2~, NO~2~, O~3~, and PM~10~ in seven different locations of Eastern Province, which are Jubail (Station no. 1), Qatif (Station no. 2), Dammam (Stations nos. 3--5), and Al Ahsa (Stations nos. 7--8) cities. Furthermore, hourly concentrations of air pollutants have been measured and recorded at the fully automated mobile air quality monitoring station (Station no. 6) located inside the Dammam south campus of Imam Abdulrahman Bin Faisal University (IAU). The geographic locations of the air quality monitoring stations (AQMS) are given in Fig. [1](#Fig1){ref-type="fig"}. To reveal the impact of the lockdown due to the COVID-19 on the air quality, the obtained air quality datasets were divided into three periods: (i) pre-lockdown (September 15, 2019--March 22, 2020), (ii) during-lockdown (March 23, 2020--June 20, 2020), and (iii) post-lockdown (June 21, 2020--July 18, 2020).Fig. 1Locations of air quality monitoring stations in the Eastern Province Meteorology data {#Sec5} ---------------- The meteorology data of the region during the study period was recorded by using a Davis® Vantage Pro 2+ wireless meteorology station, located at the roof of the college of engineering building of IAU (Station no. 6). The wind rose plots for the study region during the whole study, pre-lockdown, lockdown, and post-lockdown periods are illustrated in Fig. [2](#Fig2){ref-type="fig"}. The northwesterly winds governed the wind regime during the pre-lockdown, lockdown, and post-lockdown periods with the frequencies of 48.8%, 50.5%, and 58.6%, respectively. The average wind speeds of pre-lockdown, lockdown, and post-lockdown periods were calculated as 3.72 (± 2.35), 4.51 (± 2.77), and 3.92 (± 2.66) m/s, respectively.Fig. 2Wind rose plots for the study region during: **a** the whole study, **b** pre-lockdown, **c** lockdown, and **d** post-lockdown periods Statistical analysis and data visualization {#Sec6} ------------------------------------------- Descriptive statistics and outlier analysis methods were performed on the obtained air quality data sets. Box and whisker plots were depicted using the calculated mean, median, first quartile, third quartile, the range within 1.5 times of interquartile range (IQR), and outliers of each pollutant observed at each station for pre-lockdown, lockdown, and post-lockdown periods. Bivariate polar plots of concentrations were generated to determine the effect of wind velocity and wind direction data couple on the concentration of each pollutant during pre-lockdown, lockdown, and post-lockdown periods (Habeebullah et al. [@CR15]; Nadzir et al. [@CR22]; Omokungbe et al. [@CR24]). Results and discussions {#Sec7} ======================= Concentration variations of air pollutants before, during, and after lockdown periods {#Sec8} ------------------------------------------------------------------------------------- Descriptive statistics of each pollutant such as mean, median, first quartile, third quartile, the range within 1.5 times of interquartile range (IQR), and outliers are represented in box and whisker plots for pre-, during-, and post-lockdown periods (Fig. [3](#Fig3){ref-type="fig"}). During the whole study period between September 15, 2019, and July 18, 2020, the mean and median CO concentrations ranged between 0.16 and 0.14 ppm (IAU-Urban station), respectively, and between 2.34 and 2.04 ppm (Dammam-Coastal), respectively, and were quite lower than the maximum daily 8-h mean value of European Union (Directive 2008/50/EC) (EU [@CR11]). The observed CO concentrations indicated distinct spatial and temporal variation patterns before, during, and after lockdown periods. The highest CO concentration decrease during the lockdown was found for Qatif (↓0.96 ppm), followed by Rakah (↓0.33 ppm), Al Ahsa-Industrial (↓0.24 ppm), and IAU (↓0.12 ppm) sites as compared with the pre-lockdown period. On the other hand, the mean CO concentration at the Jubail site gradually increased by 0.50 ppm during the lockdown and 0.93 ppm post-lockdown phases. In addition to this, mean CO concentrations at Rakah site before and after lockdown periods increased from 1.65 to 2.48 ppm. The fluctuations of mean CO concentrations at Al Ahsa-Traffic, Dammam-Coastal, and Dammam-Industrial stations were not statistically significant within pre-, during-, and post-lockdown periods.Fig. 3Box and whisker plots of daily average CO, SO~2~, NO~2~, O~3~, and PM~10~ concentrations within the periods of pre-lockdown, during lockdown, and post-lockdown As observed in CO concentrations, temporal and spatial changes of SO~2~ concentrations were quite different. Almost all sites exhibited mild outlier values before, during, and after lockdown periods with an average of mild to extreme outliers of three points. Compared with the pre-lockdown period, the IAU site indicated the highest SO~2~ concentration reduction (↓1.38 ppb) during the lockdown, followed by Qatif (↓0.64 ppb), Jubail (↓0.41 ppb), Dammam-Industrial (↓0.21 ppb), and Al Ahsa-Industrial (↓0.16 ppb) sites. On the contrary, in comparison with the pre-lockdown period, mean SO~2~ concentrations gradually increased by 2.58 ppb (mean: 0.88 ppb) and 13.5 ppb (8.18 ppb) at Rakah and by 0.88 ppb (mean: 0.74 ppb) and 1.25 ppb (1.10 ppb) at Dammam-Coastal sites during the lockdown and post-lockdown phases, respectively. The highest SO~2~ concentration increase during the lockdown was noticed for the Al Ahsa-Traffic site (↑3.39 ppb) as compared with the pre-lockdown period. As was observed for Rakah and Dammam-Coastal sites, the mean SO~2~ concentrations within the post-lockdown phase at Al Ahsa-Traffic, Qatif, and Al Ahsa-Industrial sites increased by 2.55, 1.69, and 0.62 ppb, respectively, with respect to the pre-lockdown period. It is noteworthy to mention that only the mean SO~2~ concentration at the Rakah site during the post-lockdown period (41.9 μg/m^−3^) was higher than the WHO's daily SO~2~ guideline value of 20 μg/m^−3^ (WHO [@CR36]). The increased concentrations noticed for CO and SO~2~ pollutants at few locations could be linked to the large spectrum of industrial activities generating variable emission rates since most of the essential facilities were not shut down during the lockdown period. The impact of the lockdown period was most obvious on NO~2~ concentrations over the Eastern Province. The mean NO~2~ concentrations during pre-lockdown period were in the following order: 81.7 ± 82.8 ppb (Jubail) \> 30.1 ± 20.6 ppb (Rakah) \> 21.5 ± 5.43 ppb (IAU) \> 8.85 ± 9.30 ppb (Al Ahsa-Traffic) \> 5.49 ± 3.52 ppb (Dammam-Industrial) \> 4.59 ± 2.64 ppb (Al Ahsa-Industrial) \> 3.84 ± 0.68 ppb (Dammam-Coastal) \> 3.29 ± 1.03 ppb (Qatif). At all monitoring sites, mean NO~2~ concentrations considerably decreased during the lockdown and post-lockdown phases in reference to the pre-lockdown phase. The magnitude of decrease in mean NO~2~ concentrations was as high as about 62.5 (median: 69.3), 25.1 (25.7), and 10.6 (11.3) ppb during the lockdown and 66.4 (median: 60.5), 18.5 (24.5), and 2.99 (4.10) pbb at the post-lockdown period for Jubail, Rakah, and IAU stations, respectively. As a result of this reduction trend because of the lockdown measures, the mean NO~2~ concentrations in the pre-lockdown period at Jubail (155 μg/m^−3^), Rakah (57.2 μg/m^−3^), and IAU (40.8 μg/m^−3^) sites, being higher than WHO's NO~2~ guideline value of 40 μg/m^−3^, reduced to 27.8, 20.9, and 33.4 μg/m^−3^ even after the lockdown period. The other possible explanations for the NO~2~ concentration decline rather than the significantly reduced transportation activities during the lockdown period might be attributed to the decrease in NO emission rate and/or the photolysis reaction of NO~2~. The high combustion processes, including automobile engines, generally produce about \> 80% of NO and \< 20% of NO~2~. Therefore, the photochemical reaction of NO~2~ to produce O~3~ and quite limited NO emissions to form NO~2~ may be the reasons for the observed declines in NO~2~ concentrations due to the lockdown. Compared with the pre-lockdown phase, the mean O~3~ concentration decreased considerably at the Dammam-Coastal site by ↓25.1 ppb and ↓20.1 ppb during and after lockdown phases, respectively. The mean O~3~ concentration at Jubail site decreased from 46.5 ppb (median: 14.3 ppb) to 36.4 ppb (11.3 ppb) due to lockdown and after the lockdown increased to 150 ppb (282 μg/m^−3^), which was quite above the WHO's 8-h daily maximum guideline value of 40 μg/m^−3^. In the other five monitoring sites, the mean O~3~ concentrations during and after lockdown periods showed a clear upward trend at varying levels. This result is corroborated with recently published studies focused on the effect of lockdown measures to combat COVID-19 on the air quality of Delhi (India) (Mahato et al. [@CR20]), Almaty (Kazakhstan) (Kerimray et al. [@CR17]), Barcelona (Spain) (Tobías et al. [@CR33]), São Paulo (Brazil) (Nakada and Urban [@CR23]), Wuhan (China) (Lian et al. [@CR19]), and Hangzhou (China) (Wang et al. [@CR35]). In these cited studies, the increasing trend of O~3~ against imposed lockdown measures was linked to (i) sharp drop of NO~x~ emissions in lockdown period decreasing the possibility of NO + O~3~ titration reaction, (ii) increased VOC/NO~x~ ratio because of more reduced NOx emission enhancing O~3~ production, and (iii) stronger solar radiation compared with pre-lockdown period favoring O~3~ formation and accumulation. These three explanations on why O~3~ concentrations increased against lockdown measures are also valid for the Eastern Province where: (i) NO~2~ concentrations remarkably decreased at all sites, (ii) fugitive VOC emissions from continuously operating petrochemical/chemical manufacturing facilities and oil refineries, and (iii) increasing solar radiation throughout the season (pre-lockdown average: 7982 W/m^2^ day; lockdown: 12,000 W/m^2^ day; and post-lockdown: 13,681 W/m^2^ day). The PM~10~ pollutant demonstrated the highest number of outliers among other pollutants during the whole study period. The number of PM~10~ outliers (\> 1.5IQR) during lockdown is the highest for Jubail, Qatif, and Al Ahsa-Traffic sites. In addition, the mean values were significantly above the median values for most of the PM~10~ sampling locations during this study. The effect of lockdown on PM~10~ pollution was most apparent on Qatif, Dammam-Industrial, and Al Ahsa-Traffic sites where the mean concentrations reduced from 80.3 (median: 82.2) to 24.4 (15.5) μg/m^−3^, from 63.0 (54.1) to 49.8 (41.9) μg/m^−3^, and from 57.4 (50.1) to 24.8 (13.7) μg/m^−3^, respectively, and became below the WHO's PM~10~ guideline value of 50 μg/m^−3^. The mean PM~10~ concentrations at Jubail, Dammam-Coastal, and Rakah sites did not indicate a decreasing trend during the lockdown, implying that these sites might be under the effect of a wide range of PM~10~ emission sources. Source apportionment of air pollutants by bivariate polar plots {#Sec9} --------------------------------------------------------------- Bivariate polar plots depicted in Fig. [4](#Fig4){ref-type="fig"} illustrate the joint effect of wind velocity and wind direction on the levels of air pollutants monitored at Rakah station before, during, and after lockdown periods. The radius, angle, and colored contour on the polar graphs represent wind velocity (0--9 m/s), wind direction (0--360 in degrees), and concentration of the pollutant, respectively. The Rakah station was selected based on the extent of air pollution level relative to other stations. The main local air pollution sources in the vicinity of Rakah urban monitoring station are nearby traffic with moderate volume and extensive excavation/construction activities for the new residential area at southwest, King Abdulaziz Seaport in the north (4 km), and Industrial Area I at the west (4 km). Concentrations of CO and SO~2~ indicated similar distribution patterns on bivariate polar plots. Northerly winds with moderate speed and southerly winds with low speed usually brought higher CO and SO~2~ concentrations before and during the lockdown. After the lockdown, both CO and SO~2~ concentrations remarkably increased, and high concentrations corresponding to the winds blowing from all directions with low speeds (\< 4 m/s) point out local sources. The relatively calm and low-speed wind conditions compared with before and during lockdown might be the reason to carry plumes of a large number of vehicles and machinery operating at the nearby construction site using diesel fuel with high sulfur content. The NO~2~ concentrations showed the maximum decrease during the lockdown phase among other pollutants since the main source for NO~2~ emissions is the on-road traffic, which was drastically reduced. The winds blowing from southwestern and northeastern sectors at varying speeds carried high NO~2~ levels to the monitoring site before and after lockdown phases. The concentration contour distribution of O~3~ demonstrated an opposite pattern to that of NO~2~ within the pre-lockdown period due to the scavenging effect of NO~2~ on O~3~ (NO + O~3~ titration reaction). The mean O~3~ concentration during lockdown was almost the same as pre-lockdown and decreased by 7.21 ppb after the lockdown. This variation trend could be ascribed to the balanced VOC/NO~x~ ratio because of the VOC-limited characteristics of the Rakah area, which did not result in increasing O~3~ concentrations during and after lockdown periods as observed for other five sites in the Eastern Province. PM~10~ concentration distribution patterns on bivariate polar plots before and during lockdown were almost similar. The northwesterly and southerly winds with variable speeds carried high PM~10~ emissions to the monitoring site. Within the whole study period, the highest PM~10~ concentrations were observed when the wind speed was above 3 m/s, pointing the enhanced resuspension of dust and sand particles from open lands nearby and the possibility of regional/long-range transport. Interestingly, the mean PM~10~ concentration during the post-lockdown period reduced by 19.2 μg/m^−3^ as compared with pre-lockdown, which could be explained by changes in source emission rates, meteorological conditions, and atmospheric stability at Rakah area.Fig. 4Bivariate polar plots illustrating the joint effect of wind velocity and wind direction on observed pollutants at Rakah station Percent changes of air pollution levels during and after lockdown {#Sec10} ----------------------------------------------------------------- The percent changes in the concentrations of criteria air pollutants monitored at each station during and after the lockdown in reference to before lockdown are depicted in Fig. [5](#Fig5){ref-type="fig"}. The air quality improvements at Qatif-Industrial, Dammam-Industrial, IAU-Urban, and Al Ahsa-Industrial sites were more evident as compared with the other three sites, and the response level of each site was quite different against same lockdown measures. The most pronounced decrease was found for NO~2~ since NO~2~ concentrations reduced at all sites during the lockdown and after lockdown as well. Within the lockdown phase, the highest NO~2~ reduction rate was noticed at Dammam-Industrial site (86%), and followed by Rakah (84%), Jubail (77%), and Al Ahsa-Industrial (67%) while the reduction rates were less than 50% at other sites: IAU (49%) Al Ahsa-Traffic (34%), Qatif (26%), and Dammam-Coastal (12%). After the lockdown, NO~2~ concentrations also dropped at all sites within the range between 14% (IAU) and 81% (Jubail).Fig. 5Percent changes in air pollution levels during COVID-19 pandemic compared with pre-lockdown period In response to decreasing NO~2~ emission during lockdown period, changes in CO concentrations demonstrated decreasing trends most obviously at IAU (55%) and Qatif (42%) urban sites and then followed by Al Ahsa-Industrial (22%), Rakah-Urban (20%), and Dammam-Industrial (5.8%) sites. During the post-lockdown phase, Qatif and IAU urban stations and Al Ahsa-Industrial site indicated decreasing CO rates of 69%, 22%, and 27%, respectively. These results reveal that the CO and NO~2~ were commonly emitted from transportation activities, which were limited significantly during the lockdown and partially within post-lockdown phases, especially at urban locations. On the other hand, the CO level at Jubail industrial site gradually increased by 73% and 137% during and after lockdown, respectively. These remarkable CO increase rates could be attributed to the huge and wide range of industrial activities with variable production capacities that did not respond as expected against imposed lockdown measures at the Jubail industrial city. The ground level O~3~ is classified as a secondary pollutant, and its complex formation mechanism is governed by NOx and VOC emissions, VOC/NO~X~ ratio in the atmosphere, the strength of solar radiation, and atmospheric stability (Xu et al. [@CR38]). In addition to this, the O~3~ concentration at a receptor point could be under the effect of long-range transport mechanism as the atmospheric lifetime of O~3~ is about 22 days (Li et al. [@CR18]; Sicard et al. [@CR30]). Contrary to the significant reduction rates in NO~2~ during and after the lockdown, increasing O~3~ rates occurred at most of the sites, ranging between 6.3 and 45% within lockdown and between 18 and 263% after lockdown. This increasing pattern might be linked to the substantially dropped NO~x~ emissions (reduction in titration impact on O~3~), increased VOC/NO~x~ ratios, and stronger solar radiation as compared with before lockdown period. The highest increase rates after lifting lockdown measures were monitored at Jubail (263%) and Al Ahsa (209%) industrial sites where decreases in VOC emissions from a large number of petrochemical factories and large oil plants/refineries were not as large as NO~x~ emissions, most probably leading to an enhanced formation of O~3~ under stronger solar radiation and more stable atmospheric conditions. The decrease rate of SO~2~ during the lockdown period was in the following order of 30% (IAU) \> 25% (Qatif) \> 13% (Al Ahsa-Industrial) \> 10% (Dammam-Industrial) \> 8.7% (Jubail). After lifting the lockdown, Jubail and IAU sites showed 45% and 18% reduction rates, respectively, while other sites indicated increasing SO~2~ ratios varied between 50 and 427%. The Rakah station showed the most significant SO~2~ increase rates of 82% and 427% during the lockdown and after lockdown, respectively. The highest SO~2~ increase among other stations after lockdown may be ascribed to new excavation and dredging activities in the nearby residential area of the Rakah site. Even though the Eastern Province is under the effect of a wide range of PM~10~ emissions, the decreasing PM~10~ trend noticed in Qatif (70%), followed by Al Ahsa-Trafic (57%), AlAhsa-Industrial (29%), and Dammam-Industrial (21%) stations is an evidence for the effect of lockdown measures. In contrary to that, the Jubail station showed the most significant increases of 196% and 378% during and after lockdown, respectively. This remarkable increase compared with other sites may be attributed to local dust storm conditions since the Jubail station is approximately 180 km and 60 km far north from Al Ahsa and Qatif/Dammam monitoring stations, respectively. The observed reduction trends in NO~2~ (all sites), CO (Qatif-Urban, IAU-Urban, and Al Ahsa-Industrial), SO~2~ (Jubail-Industrial and IAU-Urban), and PM~10~ (Al Ahsa-Trafic and Rakah-Urban) pollutants even after releasing the lockdown measures could possibly be attributed to the multiple factors: (i) preference of people staying home due to the fear of the pandemic, (ii) allowing employees work from home by most of the companies, (iii) partial or full halt of production in some factories because of shattered supply chains, and (iv) different meteorological conditions and atmospheric stabilities compared with post-lockdown period. Comparison with other regions {#Sec11} ----------------------------- The variations of pollutant concentrations due to COVID-19 lockdown in different countries around the world are shown in Table [1](#Tab1){ref-type="table"}. The median values of percent changes in each pollutant monitored at the Eastern Province demonstrated in Table [1](#Tab1){ref-type="table"} are in a good agreement with those reported for the compared sites. The CO concentrations showed decreasing rates for all monitoring stations around the world. The major change in CO was observed for Almaty (Kazakhstan) from March 19 to April 14, with a percentage reduction of 49%. The same trend was observed in this study, where the median reduction value of CO was 13% during March 23 and June 21. The same similar trend was observed for NO~2~. All monitoring stations showed significant decreases during lockdown compared with the pre-lockdown period. The highest NO~2~ reduction was reported to be 96% for Salé (Morocco) from March 21 to April 2. On the other hand, O~3~ showed an increasing concentration trend in all compared sites around the world, including our study, except for the Anqing, Hefei, and Suzhou areas of China, which indicated a reduction of 9.6%. The changes in SO~2~ concentrations fluctuated between − 53% (China) and 8% (São Paulo, Brazil), while PM~10~ concentration changes varied between − 75% (Salé, Morocco) and 19% (Rome, Italy). Besides, SO~2~ and PM~10~ pollutants demonstrated decreasing trends at 75% of the locations compared in Table [1](#Tab1){ref-type="table"}.Table 1Variations of pollutant concentration due to COVID-19 control measures in different locationsLocationStudied lockdown periodChange of pollutant concentration due to COVID-19 control measures (%)ReferenceCOSO~2~NO~2~O~3~PM~10~Tehran (Iran)Mar. 21--Apr. 21− 41− 28− 33103− 30(Broomandi et al. [@CR10])Delhi (India)Mar. 24--Apr. 11− 30− 18− 530.8− 52(Mahato et al. [@CR20])Almaty (Kazakhstan)Mar. 19--Apr. 14− 497− 3515n.a.(Kerimray et al. [@CR17])Salé (Morocco)Mar. 21--Apr. 2n.a.− 49− 96n.a.− 75(Otmani et al. [@CR25])Barcelona (Spain)Mar. 14--Mar. 30n.a.1.8− 5158− 31(Tobías et al. [@CR33])São Paulo (Brazil)Mar. 24--Apr. 20− 308− 22117.7(Nakada and Urban [@CR23])Nice (France)Mar.--Apr.n.a.n.a.− 6324− 7.8(Sicard et al. [@CR30])Rome (Italy)n.a.n.a.− 461419Turin (Italy)n.a.n.a.− 30279.9Valencia (Spain)n.a.n.a.− 702.415Wuhan (China)n.a.n.a.− 5736− 32Wuhan (China)Jan. 23--Feb. 23− 23− 3.9− 53117− 40(Lian et al. [@CR19])Hangzhou (China)Jan. 24--Feb. 15− 22− 29− 58145− 48(Wang et al. [@CR35])Yangtze River Delta Region(China)Jan. 24--Feb. 25− 39− 26− 478.3− 61(Li et al. [@CR18])Feb. 26--Mar. 31− 25− 15− 292.3− 36Anqing, Hefei, and Suzhou area(China)Feb. 1--Feb. 29− 36− 53− 533.6− 49(Xu et al. [@CR38])Mar. 1--Mar. 31− 24− 41− 27− 9.6− 25Eastern Region (Saudi Arabia)Mar. 23--Jun. 20− 13 (± 36)− 9.2 (± 54)− 58 (± 26)17 (± 38)− 21 (± 83)This study Conclusions {#Sec12} =========== In this work, the possible effects of the lockdown due to the COVID-19 pandemic on the air quality were investigated in an arid region using meteorological and air quality datasets obtained from eight monitoring stations covering the Eastern Province of the KSA within the period between September 15, 2019, and July 18, 2020. The mean concentrations of NO~2~ and PM~10~ pollutants at several monitoring stations reduced to such levels lower than the guideline values of WHO by virtue of the imposed lockdown, which might reduce the impact of COVID-19, especially on elderly and sensitive groups. The NO~2~ was found to be the marker pollutant responding best to the lockdown measures since the concentration reductions of NO~2~ during- and post-lockdown periods at all sites ranged between 12--86% and 14--81%, respectively. In parallel to decreases in the NO~2~ concentrations, drops in CO concentrations fluctuated between 5.8 and 55% considering most of the sites, suggesting that CO and NO~2~ emissions were mainly from transportation activities, which were substantially restricted during the lockdown phase. Interestingly, O~3~ concentrations increased at the majority of the monitoring stations, ranging between 6.3 and 45% within lockdown and between 18 and 263% after lockdown. The results also revealed that controlling the formation of ground-level O~3~ is a quite challenging task even by reducing the emissions of primary air pollutants significantly. The Eastern Province is a low-SO~2~-emitting region due to the stringent air pollution control regulations, and the monitored SO~2~ emissions have been quite below the national standards and WHO's guidelines. For this reason, the reductions in SO~2~ concentrations were not distinct and varied between 8.7 and 30% during the lockdown period. The effect of lockdown measures was also evident in PM~10~ concentrations, which reduced within the range between 21 and 70% at most of the sites even though a wide range of emission sources dominates the PM~10~ budget over the Eastern Province's atmosphere. Besides the damaging effects of COVID-19 pandemic, the imposed lockdown countermeasures have been a unique opportunity to investigate the effects of reduced emission rates of pollutants from various sources on the air quality, to evaluate the existing and future air quality control regulations, and to build short- and long-term alternative action plans to enhance the air quality. **Publisher's note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. The authors acknowledge the funding provided by the Deanship of Scientific Research (DSR) at Imam Abdulrahman Bin Faisal University (IAU) in Saudi Arabia through research project No. 2018-044-Eng. The authors declare that they have no conflict of interest.

Background {#Sec1} ========== Breast cancer (BRCA) is among the most common cancers diagnosed in women in Europe where it also is the third cause of cancer death after lung and colorectal cancer \[[@CR1]\]. Approximately 75% of breast tumours is characterized by the expression of estrogen receptor alpha (ERα), encoded by the estrogen receptor 1 (*ESR1*) gene. These tumours require estrogen signals for continued growth and, consequently, patients generally receive endocrine treatment to inhibit ER signalling \[[@CR2]\]. Endocrine treatment comprises selective estrogen receptor modulators including tamoxifen, selective estrogen receptor down-regulators including fulvestrant, and AIs (e.g., anastrozole, letrozole and exemestane) that inhibit the production of estrogen from androgen. Unfortunately, resistance to endocrine therapy (ET) develops in approximately 30% of ER+ BRCA patients resulting in recurrence of the tumour \[[@CR3]\]. Despite many efforts the precise mechanisms leading to acquired treatment resistance remain mostly unknown and, therefore, therapies to prevent or revert resistance are currently lacking. Therefore, the identification of biomarkers, including epigenetic markers, that can predict endocrine resistance are considered of great value for patient stratification prior to ET \[[@CR4]\]. In general, BRCA development, progression, and (endocrine) drug resistance result from the cumulative burden of genetic and epigenetic changes. Moreover, post-transcriptional and post-translational modifications are likely to contribute as well \[[@CR5]--[@CR7]\]. The association of epigenetic changes with tumour characteristics, subtypes, prognosis, and treatment outcome is only partially characterized \[[@CR8]\]. Epigenetic changes have been shown to drive resistance acquisition (RA) through their effect on gene expression and/or chromosomal stability \[[@CR9]\]. For example, using RNA-seq and ChIP-seq analysis of the acetylation of lysine 27 on histone 3 (H3K27ac), an established active enhancer marker, revealed that epigenetic activation of the cholesterol biosynthesis pathway causes activation of ERα resulting in resistance \[[@CR10]\]. DNA methylation is also perturbed during BRCA development and may largely affect gene expression \[[@CR4], [@CR11]\]. Since DNA methylation has also been shown to be altered in endocrine resistant tumours \[[@CR12]\], the identification of methylation markers for disease diagnosis, prognosis, and treatment outcome is receiving increased attention. Moreover, BRCA treatment might benefit from the regulation of methylation activity by using DNA methyltransferase inhibitors \[[@CR4]\]. Treatment with the DNA methyltransferase inhibitor 5-aza-2′ deoxycytidine caused a significant reduction in promoter methylation and a concurrent increase in expression of the gene *ZNF350* that encodes a DNA damage response protein, and of *MAGED1* which is a tumour antigen and putative regulator of P53, suggesting that a methylation-targeted therapy might be beneficial \[[@CR13]\]. However, current inhibitors have weak stability, lack specificity for cancer cells and are inactivated by cytidine deaminase thus limiting their use in the treatment of BRCA \[[@CR14]\]. Several studies investigated DNA methylation in relation to disease outcome and therapy resistance. Lin et al. observed significant differences in DNA methylation profiles between tamoxifen sensitive and tamoxifen resistant cell lines \[[@CR15]\]. There, a large number of genes, several of which have been previously implicated in BRCA pathogenesis, were shown to have increased DNA methylation of their promoter CpG islands in the resistant cell lines. Similarly, Williams et al. observed a large number of hypermethylated genes in a tamoxifen-resistant cell line \[[@CR13]\]. In a meta-analysis of two human BRCA gene expression datasets, 144 genes for which methylation levels had been linked to BRCA survival were shortlisted as putative epigenetic biomarkers of survival. Kaplan-Meier survival analysis on the expression of these genes further reduced this list to 48 genes, and a subsequent correlation analysis of gene expression and DNA methylation provided evidence for the potential association of DNA methylation with survival in different BRCA subtypes including ER+/HER2- \[[@CR16]\]. Another study compared ductal carcinoma in situ to invasive BRCA and suggested that methylation changes indicate an early event in the progression of cancer and, therefore, might be of relevance for clinical decision making \[[@CR17]\]. In contrast to studies that showed the impact of promoter methylation, it has also been demonstrated that endocrine response in cell lines is mainly modulated by methylation of estrogen-responsive enhancers \[[@CR18]\]. There, increased *ESR1*-responsive enhancer methylation in primary tumours was found to be associated with endocrine resistance and disease relapse in ER+ (luminal A) human BRCA, suggesting that methylation levels can be used to identify patients that positively respond to ET. Note that, although limited ER-responsive enhancer methylation may already be present in the primary tumour, the analysis of methylation profiles of matched relapse samples showed that enhancer DNA methylation increased during treatment. Therefore, a combination of pre-existing and acquired differences in enhancer DNA methylation could be associated with the development of ET resistance. Current evidence on the association of DNA methylation and endocrine resistance is largely based on cell line models. The largest BRCA patient cohort to study genome-wide DNA methylation profiles and their association with resistance to ET is provided by The Cancer Genome Atlas (TCGA) \[[@CR19]\]. However, to the best of our knowledge, these data have hardly been used to find candidate methylation sites associated with endocrine resistance. One exception is a recent study by Zhang et al., who used the TCGA BRCA cohort to identify regions that were differentially methylated between patients resistant and sensitive to ET \[[@CR20]\]. However, their analysis was based on only a small subset of 32 samples selected based on either short-term (less than 30 months; resistant) or long-term (more than 100 months; sensitive) survival. In the current work we investigated if DNA methylation profiles of primary ER+/HER2- tumours provide information to predict endocrine resistance. We selected methylation profiles provided by TCGA from patients treated with tamoxifen or AIs, and assumed that patient survival is a proxy for absence of therapy resistance. To identify specific DNA methylation markers, we tested the association with survival using a Cox proportional hazards model. We were able to identify DNA methylation markers associated with patient outcome in a cohort of 552 ER+/HER2- patients, a sub-cohort of 172 patients treated with TAM, and a sub-cohort of 210 patients treated with AIs. We validated these markers and associated gene sets using DNA methylation profiles generated in a time course experiment using the T47D cell line treated with tamoxifen or deprived from estrogen. Methods {#Sec2} ======= Data {#Sec3} ---- We used clinical, biospecimen, gene expression (RNAseq V2) and DNA methylation (Illumina Human Methylation 450 K) data of 1098 patients with breast invasive carcinoma from TCGA ([cancergenome.nih.gov](http://cancergenome.nih.gov)). Samples represented in TCGA were all collected prior to adjuvant therapy \[[@CR21]\]. TCGA also recorded patient follow-up information describing clinical events such as type of treatment, the number of days from the date of initial pathological diagnosis to a new tumour event, death, and date of last contact. Since clinical and biospecimen data are scattered over multiple files in the TCGA repository, we first merged all information in a single table with one row per patient using the patient identifiers provided in the clinical and biospecimen data. Subsequently, we corrected drug names for tamoxifen and AIs (anastrozole, exemestane and letrozole) for spelling variants and mapped synonyms to their generic drug names (Additional File [1](#MOESM1){ref-type="media"}). Patient cohorts {#Sec4} --------------- For all patients with DNA methylation data available we selected data from primary tumours (indicated with "01" in the patient barcode) of female ER+/HER2- BRCA patients (Fig. [1](#Fig1){ref-type="fig"}). The molecular subtype was determined using TCGA gene expression data for these samples (see below). The ER+/HER2- cohort was further subdivided according to the endocrine treatment (AI or tamoxifen) that patients received during follow-up. Patients who received both drugs were included in both sub-cohorts. Consequently, we considered three patient cohorts, i.e., ER+/HER2-, AI, and TAM. Fig. 1Study flow chart and cohort definition. This figure shows the steps taken to define each of the three cohorts. First the molecular subtype was determined using TCGA BRCA gene expression data and ER+/HER2- patient samples were selected. Next, patients without follow-up data and patients for whom no methylation profiles were measured were removed. Finally, male patients were removed leading to the study cohort of ER+/HER2- patients. Patients who received tamoxifen form the TAM sub-cohort and patients who received AI form the AI sub-cohort. Dashed arrows indicate filter steps. ‡42 patients received both tamoxifen and AI and are included in both the TAM and AI sub-cohort. No missing data for TAM and AI cohorts. AI, aromatase inhibitor; BRCA, breast invasive carcinoma; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; TAM, tamoxifen; TCGA, The Cancer Genome Atlas Subtype determination {#Sec5} --------------------- Information for BRCA subtyping by immunohistochemistry of ER or HER2 is missing for 192 out of 1098 patients. Therefore, we used TCGA BRCA RNAseq V2 gene expression data to determine molecular subtypes (Additional File [2](#MOESM2){ref-type="media"}). To this end, gene expression data from primary tumours were retrieved from the Genomic Data Commons legacy archive using the R package *TCGAbiolinks* \[[@CR22]\]. RSEM estimated abundances were normalised using the upper quartile method from the R package *edgeR* \[[@CR23]\] and subsequently log2-transformed with an offset of one. BRCA subtypes ER−/HER2-, HER2+, and the lowly proliferative ER+/HER2- (luminal A) and highly proliferative ER+/HER2- (luminal B) subtypes were determined using the SCMOD2 model from the R package *genefu* \[[@CR24]\]. DNA methylation data and pre-processing {#Sec6} --------------------------------------- Illumina Human Methylation 450 K raw data (IDAT files) for the patients in the cohorts defined above were retrieved from TCGA. Pre-processing was performed using the R package *minfi* \[[@CR25]\]. Data were normalized using functional normalization with dye bias correction using a reference array \[[@CR26]\]. Detection *p*-values were calculated for each methylation probe and 82,150 probes with an unreliable signal (*p* \> 0.01) in one or more samples were removed. Probes corresponding to loci that contain a SNP in the CpG site or in the single-base extension site were removed. We also removed probes that have been shown to cross-hybridize to multiple genomic positions \[[@CR27]\]. Finally, M-values were calculated and probes with low variation across samples (standard deviation of M-values ≤ 0.4) were removed. The final data set comprised 320,504 CpG loci. Probes were annotated to genes and enhancer regions using the R package *IlluminaHumanMethylation450kanno.ilmn12.hg19*. Survival analysis {#Sec7} ----------------- ### Clinical variables {#Sec8} Based on literature \[[@CR28]--[@CR30]\] we selected menopause status (pre/post, after merging pre- and peri-menopausal; values '\[Unknown\]' and 'Indeterminate' were considered missing), AI treatment (yes, no), tamoxifen treatment (yes, no), tumour stage (I-IV, after merging subcategories; stage X was considered missing), and age at diagnosis as candidate variables predictive of survival. We tested association with survival using the Cox proportional hazards model (R package *survival*). We defined an event as the first occurrence of a new tumour event or death. For patients without an event we used the latest contact date as provided by the clinical data (right censoring). To account for missing values for the variables menopause status and stage in the ER+/HER2- cohort we used multiple imputation (R package *mice*) to generate 50 datasets and perform survival analysis on each dataset separately \[[@CR31]\]. The input data used for multiple imputation is available in Additional File [3](#MOESM3){ref-type="media"}. Rubin's rule was applied to combine individual estimates and standard errors (SEs) of the model coefficients from each of the imputed datasets into an overall estimate and SE resulting in a single *p*-value for each variable. Clinical variables with a *p*-value \< 0.10 in a univariable survival model were selected for inclusion in the multivariable survival model. Variables in the final multivariable model were determined using backward selection by iteratively removing variables with the highest *p*-value until all variables had a *p*-value \< 0.05. ### Single-locus survival analysis {#Sec9} Next we performed survival analysis to identify single methylation loci associated with patient survival using the methylation M-values in a Cox proportional hazards model. The models for each locus were adjusted for significant clinical variables from the multivariable model. To account for missing values for clinical variables, multiple imputation was used as described above. Resulting *p*-values were corrected for multiple testing using the Benjamini-Hochberg false discovery rate (FDR). Adjusted *p*-values \< 0.05 were considered statistically significant. Subsequently, single-locus survival models were also adjusted for ER+/HER2- subtypes (luminal A/luminal B) in addition to the clinical variables selected above. Kaplan-Meier curves for individual loci were determined by calculating the median of the methylation levels over all patients in a cohort and then assigning a patient to a low (methylation level \< median) or a high (methylation level ≥ median) group. ### Multi-locus survival analysis {#Sec10} We used the Cox proportional hazards model with elastic net regularization (function cv.glmnet, R package *glmnet*) \[[@CR32]\] to identify a signature of multiple methylation loci associated with survival. We followed a two-stage approach. First, the CpG signature was determined without including clinical variables using Cox regression with elastic net penalty. Secondly, from the resulting model the risk score (see below) was calculated and used in a new model that includes the clinical variables selected above in order to establish whether the methylation signature provided additional information compared to merely using clinical variables. Optimal values, minimizing the partial likelihood deviance, for the elastic net mixing parameter (α) and tuning parameter (λ) were determined by stratified (for event status) 10-fold cross-validation using a grid search varying α from 0 to 1 in steps of 0.1 and using 100 values for λ that were automatically generated for each α. We constructed one model for each of the three cohorts (ER+/HER2-, AI, TAM). Subsequently, for each cohort we used the identified signature to calculate a risk score for each patient: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{risk}\ \mathrm{score}={\sum}_i{c}_i\ast {M}_i$$\end{document}$$where for CpG locus *i, c*~*i*~ denotes the corresponding coefficient in the Cox model and *M*~*i*~ the methylation M-value. Next, multivariable Cox proportional hazards regression was performed using the risk score as a variable and adjusting for significant clinical variables from the multivariable model. Missing values for the clinical variables were imputed as described above. Finally, the risk-score-based models were also adjusted for ER+/HER2- subtypes (luminal A/luminal B) in addition to the selected clinical variables. Kaplan-Meier curves were determined for two groups of patients by calculating the median of the risk scores over all patients in a cohort and then assigning a patient to a good (risk score \< median) or a bad prognosis group (risk score ≥ median). ### Stability of multi-locus signatures {#Sec11} To assess the stability of the multi-locus signatures 30 regularized Cox models were fitted using a stratified (for event status) selection of 90% of the samples for each cohort. We counted the number of times each CpG locus was included in the 30 signatures and then selected those CpGs that occurred in at least 6 or at least 21 signatures. We refer to the resulting signatures as stability signatures. Fisher's exact test was used to determine the significance of the overlap between the original multi-locus signature and the stability signatures. Correlation between DNA methylation and gene expression {#Sec12} ------------------------------------------------------- CpGs in single-locus and multi-locus signatures were annotated to their nearest gene(s) (package *IlluminaHumanMethylation450kanno.ilmn12.hg19*). For each signature Pearson correlation coefficients (and corresponding *p*-values) were calculated between the methylation and gene expression profiles of each CpG-gene pair. Resulting *p*-values were corrected for multiple testing in each signature using the Benjamini-Hochberg FDR. Methylation profiling of resistance acquisition in an ER+ breast cancer cell line {#Sec13} --------------------------------------------------------------------------------- T47D cells were either treated with 100 nM 4-hydroxytamoxifen (TMX), long-term estrogen deprived (LTED; modelling AI treatment \[[@CR33]\]) or not treated (wild type (WT)) in two biological replicates cultured for 7 and 5 months, respectively. DNA was extracted after 0, 1, 2, 5 and 7 (only one replicate) months. Methylation profiling was performed using the Illumina MethylationEPIC BeadChip platform at the Genomic and Proteomic Core Facility (DKFZ, Germany). For each sample two technical replicates were measured. Pre-processing was performed as described above, except that a single sample approach was used for dye bias correction. The 8682 probes with an unreliable signal (detection *p*-value \> 0.01) in one or more samples were removed. Probes that cross-hybridized to multiple genomic positions as listed by Pidsley et al. \[[@CR34]\] were removed. No filtering based on M-values was performed. The final data set contains 786,872 CpG loci. Using the resulting M-values CpG-wise linear models were fitted with coefficients for each treatment (TMX, LTED, WT) and time point combination. In addition, we included a coefficient to correct for systematic differences between the two biological replicates (R package *limma*). For both LTED and TMX treated cells, contrasts were made between each individual time point *t* and the WT cell line at baseline, that is, LTED~*t*~ -- WT~0~ and TMX~*t*~ -- WT~0~, respectively. The comparison of the average of TMX and LTED treated cells versus WT baseline was estimated via the contrast (LTED~*t*~ + TMX~*t*~)/2 -- WT~0~. Differential methylation was assessed using empirical Bayes moderated statistics while also including the consensus correlation within pairs of technical replicates in the linear model fit (function duplicateCorrelation, *limma* package). The resulting signatures are referred to as the LTED, TMX and TMX/LTED signatures. Enrichment analysis {#Sec14} ------------------- We performed generalized gene set testing with a hypergeometric test using the gsameth function (R package *missMethyl*) to test if significant CpG sites are enriched in selected pathways \[[@CR35]\]. For the single-locus survival analysis, signatures were defined as those CpGs with *p*-value \< 0.006 (TAM, AI) and *p*-value \< 0.002 (ER+/HER2-) corresponding to signatures of \~ 2500 CpGs. For the T47D RA experiment signatures were defined as the top 10,000 CpGs ranked on *p*-value as determined using a moderated F-test (*limma* package), which tests whether a CpG is differentially methylated at any time point versus WT, for the three sets of contrasts (TMX, LTED, TMX/LTED) described above. We used a combination of Hallmark gene sets (collection H) and a subset of 16 curated gene sets (collection C2; gene set name contained either "tamoxifen" or "endocrine_therapy") from the Molecular Signatures Database (MSigDB) v7.0 (Entrez Gene ID version) \[[@CR36]\]. Resulting *p*-values were corrected for multiple testing using the Benjamini-Hochberg FDR. We also tested whether the methylation loci identified from the TCGA BRCA single-locus and multi-locus signatures (based on Illumina 450 K arrays) and represented on the Illumina EPIC array were enriched in the T47D RA experiment using ROAST rotation-based gene set tests (*limma* package) \[[@CR37]\]. Enrichment of TAM and AI survival signatures was assessed using the comparisons of respectively TMX and LTED treated cells to WT baseline described above. Enrichment of the ER+/HER2- survival signature was assessed using the comparison of the average of TMX and LTED treated cells versus WT baseline described above. ROAST *p*-values were calculated~,~ for two alternative hypotheses denoted as 'up' and 'down' using 9999 rotations. In the ROAST analyses directional contribution weights of 1 or − 1 were used depending on whether a CpG of the signature under consideration had a positive (corresponding to increased risk of an event) or negative (corresponding to decreased risk of an event) coefficient in the corresponding Cox model. In this case, the alternative hypothesis 'up' corresponds to methylation levels changing in the same direction in the TCGA BRCA survival signature and in the T47D RA experiment, whereas the alternative hypothesis 'down' corresponds to a change in the opposite direction (Fig. [2](#Fig2){ref-type="fig"}). The two-sided directional *p*-value is reported. Fig. 2Validation of survival signatures in T47D resistance acquisition experiment. **a** Kaplan-Meier plots for two selected CpGs significantly associated with survival in the ER+/HER2- cohort. Patients were stratified based on the methylation levels of a risk decreasing locus CpG↓ (left; higher methylation is associated with longer survival) and a risk increasing locus CpG↑ (right; higher methylation is associated with shorter survival). H, methylation levels above median; L, methylation levels below median. Shaded areas in the Kaplan-Meier plot denote the 95% CI in the H and L strata. *P*-values are based on a log-rank test. **b** Example of a barcode enrichment plot for a TCGA BRCA survival signature in the cell line comparison of treated (LTED or TMX) samples at time point *t* versus WT baseline. All methylation loci are ranked from left to right by increasing log-fold change in the cell line comparison under consideration and represented by a shaded bar. Loci within the survival signature are represented by vertical bars. Red bars correspond to risk increasing loci (for example, CpG↑ indicated with a solid bar), blue bars correspond to risk decreasing loci (for example, CpG↓ indicated with a solid bar). In this example, the risk increasing loci tend to be hypermethylated (log-fold change \> 0) in the treated cell line and the risk decreasing loci tend to be hypomethylated (log-fold change \< 0). That is, most loci change in the same direction in the survival signature and the T47D RA experiment. **c** When using directional weights of 1 and − 1 for risk increasing and risk decreasing loci respectively, the blue bars are mirrored across the black dashed line at a log-fold-change of 0. In this case for a ROAST gene set test, the alternative hypothesis 'up' corresponds to methylation levels changing in the same direction whereas the alternative hypothesis 'down' corresponds to a change in the opposite direction Quantitative real-time PCR {#Sec15} -------------------------- Total RNA was isolated from WT and T47D cells treated with tamoxifen or deprived from estrogen with RNeasy Mini kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions and treated with DNase Max Kit (Qiagen). cDNA was synthesized with the Revert Aid H Minus First Strand cDNA Synthesis Kit (Fermentas, Waltham, MA, USA). Quantitative real-time PCR (qRT-PCR) reactions for target genes were performed with the Applied Biosystems QuantStudio™ 3 Real-Time PCR System, using probes from the Universal Probe Library, UPL (Roche Diagnostics, Mannheim, Germany). The data were analyzed using the SDS software with the ΔΔCt method. The Ct values were normalized to the housekeeping gene *ACTB*. Results {#Sec16} ======= Clinical variables are associated with survival in ER+/HER2- cohort {#Sec17} ------------------------------------------------------------------- For the TCGA BRCA ER+/HER2- cohort (*N* = 552, Fig. [1](#Fig1){ref-type="fig"}) we assessed whether the clinical variables menopause status, AI treatment, tamoxifen treatment, tumour stage and age at diagnosis were associated with survival, with an event defined as first occurrence of a new tumour event or death. In a univariable Cox proportional hazards model tumour stage (HR 1.92, 95% CI 1.43--2.59; *p* = 1.63E-05) and age at diagnosis (HR 1.03, 95% CI 1.01--1.05; *p* = 2.40E-04) are significantly associated with survival (Table [1](#Tab1){ref-type="table"}). This is in agreement with previous findings that a more advanced tumour stage and increased age are associated with poorer outcome \[[@CR38]\]. Tamoxifen treatment, AI treatment and menopause status are not significantly associated with survival in our cohort. When we included the clinical variables in a multivariable Cox proportional hazards model, tumour stage, age and AI treatment were selected for inclusion in the final multivariable model using backward selection (Table [2](#Tab2){ref-type="table"}). Table 1Univariable Cox proportional hazards modelHR95% CI*P*-valueStage (per stage increment)1.921.43--2.591.63E-05Age (per 1-yr increment)1.031.01--1.052.40E-04AI treatment (vs. no AI treatment)0.680.45--1.050.0812Post-menopausal (vs. pre-menopausal)1.520.94--2.450.0913Tamoxifen treatment (vs. no tamoxifen treatment)0.670.42--1.070.0921Univariable Cox proportional hazards model for clinical variables (ER+/HER2- cohort). *HR* hazard ratio, *CI* confidence interval, *AI* aromatase inhibitorTable 2Multivariable Cox proportional hazards modelHR95% CI*P*-valueStage (per stage increment)2.151.61--2.893.05E-07Age (per 1-yr increment)1.041.02--1.052.48E-06AI treatment (vs. no AI treatment)0.610.40--0.940.026Multivariable Cox proportional hazards model for clinical variables (ER+/HER2- cohort). *HR* hazard ratio, *CI* confidence interval, *AI* aromatase inhibitor Single methylation loci associated with survival {#Sec18} ------------------------------------------------ To identify individual methylation loci associated with survival we fitted 320,504 Cox proportional hazard models using the M-value of each CpG while adjusting for the clinical variables selected in the multivariable model above (tumour stage, age and AI treatment (ER+/HER2- cohort only)). This resulted in 134, 5 and 1 CpGs for which DNA methylation is significantly (adjusted *p*-value \< 0.05) associated with survival in the ER+/HER2-, TAM, and AI cohort respectively (Additional File [4](#MOESM4){ref-type="media"}). The Kaplan-Meier curves show a significant difference in survival between the two groups stratified on median methylation level for nearly all selected loci (Additional File [5](#MOESM5){ref-type="media"}). Interestingly, apart from three CpGs in the ER+/HER2- signature, for all of the CpGs increased methylation is associated with decreased risk of an event. Additional File [6](#MOESM6){ref-type="media"} shows the overlap of the signatures for the three cohorts. Three out of five methylation loci from the TAM signature are also found in the ER+/HER2- signature and, consequently, the other two loci are specific for tamoxifen treated patients. Since all patients in the TAM cohort are also included in the ER+/HER2- cohort, overlap between the signatures is expected. TAM and AI signatures do not share methylation loci. ER+/HER2- and TAM signatures are enriched for enhancer CpGs (ER+/HER2-: 36%, *p* = 0.0005; TAM: 80%, *p* = 0.0113; Fisher's exact test, one-sided). For all selected patients we had paired DNA methylation and gene expression data (Fig. [1](#Fig1){ref-type="fig"}). We therefore calculated the Pearson correlation coefficient between the methylation profile of each locus in any of the three signatures and gene expression of the gene(s) closest to that locus (Additional File [4](#MOESM4){ref-type="media"}). DNA methylation is significantly (adjusted *p*-value \< 0.05) (anti-)correlated with gene expression for 52 (of 136), 3 (of 5) and 2 (of 2) CpG-gene pairs in the ER+/HER2-, TAM and AI signature respectively. To gain insight in the main biological processes involved in differences in survival, we performed gene set enrichment analyses on genes linked to CpG loci associated with survival. All three signatures are significantly enriched (FDR \< 0.1) in gene sets associated with ET or endocrine resistance, genes activated when upregulating the PI3K/AKT/mTOR pathway and genes upregulated in response to TGFB1, which have both been implicated in endocrine resistance \[[@CR39], [@CR40]\] (Additional File [12](#MOESM12){ref-type="media"}). Multi-locus methylation signature associated with survival {#Sec19} ---------------------------------------------------------- Next we performed a multivariable analysis with elastic net penalty to find combinations of methylation loci associated with survival in a Cox proportional hazards model. This resulted in 203, 36 and 178 CpGs that are included in the survival signatures of the ER+/HER2-, TAM, and AI cohort respectively (Additional File [7](#MOESM7){ref-type="media"}). The ER+/HER2- and AI signatures are enriched for enhancer loci (ER+/HER2-: 36%, *p* = 1.79E-05; AI: 29%, *p* = 0.044; Fisher's exact test, one-sided), whereas the TAM signature is not significantly enriched for enhancer loci (TAM: 36%, *p* = 0.051; Fisher's exact test, one-sided). The risk score calculated from the multi-locus signature and adjusted for tumour stage, age and AI treatment (ER+/HER2- cohort only) is significantly associated with survival (*p* \< 10E-12) for all three cohorts (Additional File [8](#MOESM8){ref-type="media"}) indicating that DNA methylation is an independent factor in predicting survival. The risk scores calculated from the multi-locus signatures stratify the patients in two groups for each cohort (Fig. [3](#Fig3){ref-type="fig"}a). Fig. 3Multi-locus survival analysis. **a** Kaplan-Meier plots of the patients stratified based on the risk scores of the multi-locus signature in ER+/HER2, TAM and AI cohorts. H, risk score above median; L, risk score below median. Shaded areas denote the 95% CI in the H and L strata. *P*-values are based on a log-rank test. **b** Venn diagram denoting the number of methylation loci in the multi-locus signatures for the ER+/HER2-, TAM, and AI cohorts' There is no overlap between the signatures of TAM and AI cohorts. However, the ER+/HER2- signature partly overlaps with the TAM and AI signatures (Fig. [3](#Fig3){ref-type="fig"}b). The coefficients in the Cox models corresponding to the overlapping loci have an identical sign in both cohorts. The multi-locus signatures include a large number of methylation loci that were also identified in the corresponding single-locus survival analysis. Out of 203 methylation loci in the ER+/HER2- multi-locus signature 34 were also found in the single-locus signature (Additional File [9](#MOESM9){ref-type="media"}). Moreover, all methylation loci in the TAM and AI single-locus signatures, five and one respectively, are part of the corresponding multi-locus signature. We assessed the stability of the multi-locus signatures using a 10% leave-out test. The stability signature is enriched in the original multi-locus signature for each corresponding cohort (*p* \< 0.05; Additional File [10](#MOESM10){ref-type="media"}). We calculated the Pearson correlation coefficient between the methylation profile of each locus in any of the three multi-locus signatures and gene expression of the gene(s) closest to that locus (Additional File [7](#MOESM7){ref-type="media"}). DNA methylation is significantly (adjusted *p*-value \< 0.05) (anti-)correlated with gene expression for 109 (of 235), 17 (of 37) and 57 (of 181) CpG-gene pairs in the ER+/HER2-, TAM and AI signature respectively. Profiling of resistance development in T47D cells {#Sec20} ------------------------------------------------- To investigate the possible association between DNA methylation of the loci identified in the survival analyses and ET resistance in more detail, we performed a time course experiment using the T47D ER+ BRCA cell line treated with tamoxifen or long-term estrogen deprived. qRT-PCR analysis showed that endocrine resistance associated genes *HDAC9* \[[@CR41]\] and *CD36* \[[@CR42]\] are indeed significantly increased in the treated cells compared to WT (Additional File [11](#MOESM11){ref-type="media"}). Also known tamoxifen induced genes *KRT4* and *FGF12* \[[@CR43]\] show a significant upregulation in the treated cells. Next, we generated genome-wide DNA methylation profiles for both treatments on five different time points (0 (=WT), 1, 2, 5, and 7 months). We identified three signatures, corresponding to CpGs that were differentially methylated over time in TMX treated cells, LTED cells, and in the comparison of the average of TMX and LTED cells versus WT. These signatures consist of thousands of loci that are significantly differentially methylated over time versus WT. To gain insight in the main biological processes involved in RA, we performed gene set enrichment analyses on genes associated with differentially methylated loci. All three signatures are significantly enriched (FDR \< 0.1) in gene sets associated with ET or endocrine resistance, gene sets related to metastasis such as the epithelial-mesenchymal transition, gene sets corresponding to signaling pathways implicated in endocrine resistance such as hedgehog signaling \[[@CR44]\], and a gene set defining early response to estrogen (Additional File [12](#MOESM12){ref-type="media"}). Validation of survival signatures in T47D resistance acquisition experiment {#Sec21} --------------------------------------------------------------------------- We then investigated the concordance between the CpG loci in the survival signatures and in the RA signatures (Fig. [2](#Fig2){ref-type="fig"}). The multi-locus survival signatures for ER+/HER2- and AI are significantly enriched in the comparison of the last time point (7 months) versus WT baseline in the T47D RA experiment (ER+/HER2-: *p* = 0.0017, AI: *p* = 0.0222; direction: 'up'; Table [3](#Tab3){ref-type="table"}). The signatures are not enriched at earlier time points. However, the proportion of CpGs contributing to enrichment in the same direction ('up') increases over time until it becomes significant for the last time point. The single-locus survival signature for TAM is also significantly enriched at the 7-month time point in the T47D RA experiment (*p* = 0.0032), but not for ER+/HER2- despite an increasing trend in the proportion of CpGs contributing to enrichment in the same direction ('up') over time (Table [4](#Tab4){ref-type="table"}). The single-locus AI signature consists of only one CpG and an enrichment analysis is therefore not possible. However, for this locus the change in methylation level when comparing LTED treated cells with WT baseline is not concordant with the log-hazard ratio for that locus (data not shown). Table 3ROAST test results for the multi-locus signaturesTime pointDirection*P*-valueProp. (down)Prop. (up)**ER+/HER2- (193 CpG sites)** 1Up0.39480.060.07 2Up0.67440.240.24 5Up0.31370.260.28 7Up**0.0017**0.230.35**TAM (32 CpG sites)** 1Down0.09950.060.03 2Down0.17360.160.06 5Down0.00370.280.06 7Down0.00280.220.13**AI (164 CpG sites)** 1Up0.10040.050.13 2Up0.54340.150.22 5Down0.20880.240.23 7Up**0.0222**0.150.26Direction indicates the direction of change. Methylation loci were weighted by their direction of change in the survival signature. 'Up' therefore corresponds to changes in the same direction in the survival signature and in the T47D RA experiment. That is, if a locus is risk in/decreasing in the survival signature than it is hyper/hypomethylated in the cell line signature for the indicated time point as compared to WT baseline. 'Down' corresponds to changes in the opposite direction. Prop., proportion of loci in the signature contributing to the estimated *p*-value and direction. Significant *p*-values (\< 0.05) for concordant changes ('Up') are indicated in boldTable 4ROAST test results for the single-locus signaturesTime pointER+/HER2- (128 CpG sites)TAM (5 CpG sites)Direction*P*-valueProp. (down)Prop. (up)Direction*P*-valueProp. (down)Prop. (up)1Down0.69200.110.07Down0.7699002Up0.75710.160.20Up0.8784005Down0.03650.300.25Up0.10130.20.47Up0.34550.250.31Up**0.0032**00.6Direction indicates the direction of change. Methylation loci were weighted by their direction of change in the survival signature. 'Up' therefore corresponds to changes in the same direction in the survival signature and in the T47D RA experiment. That is, if a locus is risk in/decreasing in the survival signature than it is hyper/hypomethylated in the cell line signature for the indicated time point as compared to WT baseline. 'Down' corresponds to changes in the opposite direction. Prop., proportion of loci in the signature contributing to the estimated p-value and direction. Significant p-values (\< 0.05) for concordant changes ('Up') are indicated in bold Also in terms of gene sets there is overlap between the sets enriched in both the single-locus survival signatures and RA signatures. Interestingly, one of the most significant gene sets in all six signatures consists of genes down-regulated in response to ultraviolet (UV) radiation. Many genes in this gene set are related to cell motility. Indeed, upon UV stress cells down regulate non-essential processes such as invasion and motility, whereas these processes are upregulated in resistant cells that become more invasive. Four gene sets are significantly enriched (FDR \< 0.1) in the two ER+/HER2- signatures.. These include sets associated with endocrine resistance and a gene set defining early response to estrogen (Fig. [4](#Fig4){ref-type="fig"}a, Additional File [12A](#MOESM12){ref-type="media"}). Six gene sets, several of them related to endocrine resistance, are significantly enriched in the two tamoxifen signatures (Fig. [4](#Fig4){ref-type="fig"}b, Additional File [12B](#MOESM12){ref-type="media"}). Two gene sets are significantly enriched in both the AI and LTED signature (Fig. [4](#Fig4){ref-type="fig"}c, Additional File [12C](#MOESM12){ref-type="media"}). Fig. 4Gene sets enriched in single-locus survival and resistance acquisition signatures. Gene set enrichment analysis of single-locus survival (*x*-axis) and RA signatures (*y*-axis). **a** T47D TMX/LTED signature versus ER+/HER2− single-locus signature. **b** T47D TMX signature versus TAM single-locus signature. **c** T47D LTED signature versus AI single-locus signature. Each diamond represents either a Hallmark (H) gene set or a curated gene set (C: Creighton, M: Massarweh) related to tamoxifen treatment or ET from the Molecular Signatures Database. Gene sets significantly enriched (FDR \< 0.1, that is -log10(FDR) \> 1, indicated by the red dashed lines) in both signatures are labelled with their name. See Additional File [12](#MOESM12){ref-type="media"} for a version of this figure in which more gene sets are labelled Discussion {#Sec22} ========== We investigated whether TCGA DNA methylation profiles measured in primary ER+/HER2- tumours can be used to predict development of resistance to ET in two sub-cohorts of patients treated with tamoxifen or AI. Using a single-locus Cox proportional hazard model we were able to identify 134, 5 and 1 CpGs for which DNA methylation is significantly associated with survival in the ER+/HER2-, TAM and AI cohorts respectively, while the corresponding multi-locus signatures consisted of 203, 36 and 178 CpGs. The multi-locus signatures showed a large overlap of 25, 100, and 100% with the ER+/HER2-, TAM and AI single-locus signatures respectively. The risk scores of the multi-locus signatures were significantly associated with survival. Moreover, we found that the ER+/HER2- and TAM single-locus and ER+/HER2- and AI multi-locus signatures were significantly enriched for CpGs in enhancer regions suggesting a functional effect (on gene expression) \[[@CR18]\]. For both the single-locus signatures (Additional File [6](#MOESM6){ref-type="media"}) and the multi-locus signatures (Fig. [3](#Fig3){ref-type="fig"}b) we observed no overlap of loci associated with survival between the AI and TAM cohorts. This could be indicative of a difference in development of resistance against tamoxifen or AI. This is in line with earlier observations in endocrine-resistant cells compared with wild type MCF7 cells, which also showed limited overlap in their response to tamoxifen and estrogen deprivation in terms of their gene expression \[[@CR10]\] and DNA methylation profiles \[[@CR18]\]. In our analyses we adjusted for clinical variables associated with survival (tumour stage, age and AI treatment (ER+/HER2- cohort only)) in order to estimate the independent effect of methylation on survival. It has been shown that methylation profiles can discriminate between the ER+/HER2- subtypes luminal A and B \[[@CR45]\]. Moreover, patients with a luminal B tumour have worse prognosis compared to patients with a luminal A tumour \[[@CR46]\], which is also the case in our ER+/HER2- cohort (HR 2.04, 95%CI 1.11--3.74, *p* = 0.020). We, therefore, also performed survival analyses adjusted for luminal status in addition to the clinical variables mentioned earlier. The single-locus signatures with correction for luminal status showed a considerable overlap of 85, 40, and 100% with the original (that is, without correction for luminal status) ER+/HER2-, TAM and AI single-locus signatures respectively (Additional File [13](#MOESM13){ref-type="media"}). Notably, all CpGs included in the original single-locus signatures still have an FDR \< 0.15 after correction for luminal status. The risk scores of the original multi-locus signatures were also significantly associated with survival after correction for luminal status (Additional File [13](#MOESM13){ref-type="media"}). In summary, the methylation signatures we identified are associated with survival independently of luminal status. We note that although the methylation profiles provided by TCGA are measured in untreated primary tumour samples, treatment regimens after initial diagnosis are heterogeneous. Some patients received adjuvant chemotherapy and/or radiotherapy next to ET and 42 patients in the TAM and AI cohorts received both types of endocrine treatments. Moreover, the duration of (endocrine) treatment varied among patients. Furthermore, treatment information may not be complete \[[@CR21]\]. These aspects were not taken into account in our analyses and might have biased the results. We also acknowledge that this study is limited by the relatively modest number of events (i.e., new tumour event, death) for the different cohorts (ER+/HER2-: 97 events in 552 patients; TAM: 24 events in 172 patients; AI: 32 events in 210 patients) due to the relatively short follow-up time. This affects statistical power to identify methylation loci associated with survival. In this study we assumed that the methylation events in the primary tumour, rather than acquired methylation during tumour progression, are associated with patient survival as a proxy for development of therapy resistance. To validate our results we aimed to use methylation profiles from the International Cancer Genome Consortium (ICGC). However, the number of patients in the ICGC BRCA cohort with reliable information on endocrine treatment was too small to make such a comparison meaningful. Instead, we used DNA methylation measurements obtained from T47D cells as a model system for RA in ER+ luminal A BRCA. We showed that our multi-locus signatures for the ER+/HER2- and AI cohorts were conserved among the loci that are differentially methylated in endocrine-resistant T47D cells. Similarly, our single-locus signature for the TAM cohort was also significantly enriched in the T47D experiment. At the gene set level, several sets related to ET and endocrine resistance were significantly enriched in both the survival and RA signatures. Although this is not a final validation of our results, it strongly suggests that the loci we identified from primary tumours, that is prior to any endocrine treatment, are also associated with endocrine resistance. CpGs with concordant significant changes in the survival and RA signatures and with significant (anti-)correlation between paired DNA methylation and gene expression profiles in TCGA BRCA are promising candidates for further investigation and are listed in Additional File [14](#MOESM14){ref-type="media"}. Most genes associated with these CpG sites have been implicated in survival and resistance related processes in BRCA. In particular, high levels of *TSC2* and *PXN* are associated with decreased metastasis-free survival \[[@CR47], [@CR48]\]. This is in agreement with our findings that lower methylation of the corresponding CpG loci is associated with decreased survival and that their DNA methylation profile is negatively correlated with gene expression (Fig. [5](#Fig5){ref-type="fig"}a-b, Additional File [15A-B](#MOESM15){ref-type="media"}). Interestingly, in the T47D RA experiment these loci are also significantly hypomethylated in resistant cells compared with WT (Fig. [5](#Fig5){ref-type="fig"}c, Additional File [15C](#MOESM15){ref-type="media"}). In the ER+/HER2- single-locus signature, the cg07145834 locus in the 5'UTR of *ZHX2* was selected. Low levels of *ZHX2* are associated with better overall survival \[[@CR49]\], in agreement with the findings from this study that higher methylation of the corresponding CpG locus is associated with increased survival, its DNA methylation profile is positively correlated with *ZHX2* gene expression, and the CpG locus is hypomethylated in resistant cells compared with WT T47D cells (Additional File [15D-F](#MOESM15){ref-type="media"}). Fig. 5Association of methylation levels of CpG site cg02198582 with survival and resistance acquisition and its correlation with *TSC2* expression levels. **a** Kaplan-Meier plot for CpG site cg02198582 located in the gene body of *TSC2* and significantly associated with survival in the AI cohort. Patients were stratified based on methylation levels. H, methylation levels above median; L, methylation levels below median. Shaded areas in the Kaplan-Meier plot denote the 95% CI in the H and L strata. P-value is based on a log-rank test. **b** Correlation between paired DNA methylation and gene expression profiles (cg02198582, *TSC2*). Each circle corresponds to a patient sample in the AI cohort. The Pearson correlation coefficient is indicated, together with the corresponding regression line and its 95% CI. **c** Log2-fold change of the methylation M-values of cg02198582 inT47D LTED versus WT cells Stone et al. \[[@CR18]\] recently demonstrated in a small cohort of patients who received endocrine treatment for at least five years that methylation levels in selected ESR1-enhancer loci were significantly increased in primary tumours of patients who relapsed within six years as compared to patients with 14-year relapse free survival. Moreover, these differences were even more pronounced in matched local relapse samples. DNA methylation data measured in a large number of pre- and post-treatment samples obtained from patients who received ET that either relapsed due to endocrine resistance or remained relapse-free will enable validation of the signatures identified in this and other studies. Moreover, such a cohort enables comparison of methylation levels in paired primary and local relapse samples providing the opportunity to identify epigenetic drivers of endocrine resistance \[[@CR50]\]. Conclusions {#Sec23} =========== In this study we identified individual and multivariable DNA methylation markers associated with survival and resistance in a large cohort of 552 ER+/HER2- BRCA patients from The Cancer Genome Atlas. Survival signatures were validated using time series DNA methylation profiles of T47D cells during development of resistance to endocrine therapy. A number of promising targets with concordant significant changes in survival and RA signatures were identified. These include CpG sites associated with *TSC2, PXN* and *ZHX2* that have all been implicated in survival related processes in BRCA. Our results suggest that methylation signatures associated with the development of endocrine resistance can also be identified in primary breast tumours prior to any endocrine treatment. Supplementary information ========================= {#Sec24} ###### **Additional file 1.** Mapping to generic drug names. Overview of synonyms and spelling variants for drug names used in TCGA BRCA and their mapping to a generic drug name used in our study. ###### **Additional file 2.** Molecular subtypes. Overview of the molecular subtype frequency as determined by immunohistochemistry of ER and HER2 and as predicted by the SCMOD2 model (R package *genefu*) using TGCA BRCA primary tumour gene expression data. Subtypes are listed for the 1095 patients for whom gene expression data is available (Fig. [1](#Fig1){ref-type="fig"}). ###### **Additional file 3.** Sample annotation. Sample annotation for the 552 patients in the ER+/HER2- cohort. The first sheet provides a short definition of the variables included in the second sheet. ###### **Additional file 4.** Single-locus survival analysis. Results of single-locus survival analysis on ER+/HER2-, TAM and AI cohorts. For each CpG the results of the correlation analysis and of the differential methylation analysis of month 7 versus WT in the T47D RA experiment are also included. ###### **Additional file 5**Single-locus Kaplan-Meier plots. Kaplan-Meier plots for each CpG site from the single-locus signatures. Patients were stratified based on the methylation levels of the indicated locus in ER+/HER2, TAM and AI cohorts. H, methylation level above median; L, methylation level below median. Shaded areas denote the 95% CI in the H and L strata. *P*-values are based on a log-rank test. ###### **Additional file 6.** Single-locus Venn diagram. Venn diagram of the single-locus signatures in the ER+/HER2-, TAM and AI cohorts. ###### **Additional file 7.** Multi-locus survival analysis. Results of multi-locus survival analysis on ER+/HER2-, TAM and AI cohorts. For each CpG the results of the correlation analysis and of the differential methylation analysis of month 7 versus WT in the T47D RA experiment are also included. ###### **Additional file 8.** Survival analysis using risk score. Results of survival analysis of the multi-locus signature using the risk score corrected for selected clinical variables in ER+/HER2-, TAM and AI cohorts. ###### **Additional file 9.** Overlap between single-locus and multi-locus signatures. Venn diagrams of the overlap between single-locus and multi-locus signatures in the three cohorts ER+/HER2-, TAM and AI. (PPTX 41 kb) ###### **Additional file 10.** Stability of multi-locus signatures. Results of Fisher's exact test to determine the significance of the overlap between the original multi-locus signature and the stability signature. ###### **Additional file 11.** qRT-PCR results. Primer sequences and gene expression levels of CD36, FGF12, HDAC9, and KRT4 determined by qRT-PCR after treatment with tamoxifen or long-term estrogen deprivation relative to their expression in untreated T47D cells. ###### **Additional file 12.** Gene set enrichment analysis. Gene set enrichment analysis of single-locus survival (*x*-axis) and RA signatures (*y*-axis). (A) T47D TMX/LTED signature versus ER+/HER2− single-locus signature. (B) T47D TMX signatuare versus TAM single-locus signature. (C) T47D LTED signature versus AI single-locus signature. Each diamond represents either a Hallmark gene set or a curated gene set related to tamoxifen treatment or ET from the Molecular Signatures Database. Gene sets significantly enriched (FDR \< 0.1, that is -log10(FDR) \> 1, indicated by the red dashed lines) in at least one of the two signatures are labelled with their name. Purple: gene sets that are significantly enriched in all three survival signatures. Red: gene sets that are significantly enriched in all three RA signatures. Blue: gene sets that are significantly enriched in all six signatures. ###### **Additional file 13.**Survival analyses including luminal status. Reanalysis when also including luminal status in the (*i*) multivariable survival analysis, (*ii*) single-locus survival analysis, and (iii) risk score for the multi-locus signature. ###### **Additional file 14.**CpGs with concordant significant changes in the survival and resistance acquisition signatures and with significant correlation between paired DNA methylation and gene expression profiles. CpGs in single-locus (Additional File [4](#MOESM4){ref-type="media"}) and multi-locus (Additional File [7](#MOESM7){ref-type="media"}) survival signatures were selected according to three additional criteria: (*i*) CpG DNA methylation is significantly (adjusted *p*-value \< 0.05) (anti-)correlated with expression of the nearby gene(s), (*ii*) CpG is also significantly differentially methylated (adjusted p-value \< 0.05) in the corresponding RA signature at month 7 versus WT, (*iii*) CpG changes concordantly in survival and corresponding RA signature, that is, risk increasing loci are hypermethylated and risk decreasing loci are hypomethylated. ###### **Additional file 15.**Association of methylation levels of selected CpG sites with survival and resistance acquisition and their correlation with expression levels of the associated genes. (A,D) Kaplan-Meier plot for CpG site cg14094027 located in the gene body of *PXN* (A) and CpG site cg07145834 located in the 5'UTR of *ZHX2* (D), both significantly associated with survival in the ER+/HER2- cohort. Patients were stratified based on methylation levels. H, methylation levels above median; L, methylation levels below median. Shaded areas in the Kaplan-Meier plot denote the 95% CI in the H and L strata. *P*-values are based on a log-rank test. (B,E) Correlation between paired DNA methylation and gene expression profiles (B: cg14094027, PXN; E: cg07145834, *ZHX2*). Each circle corresponds to a patient sample in the ER+/HER2- cohort. The Pearson correlation coefficient is indicated, together with the corresponding regression line and its 95% CI. (C,F) Log2-fold change of the methylation M-values of cg14094027 (C) and cg07145834 (F) in the comparison of T47D TMX/LTED versus WT. AI : Aromatase inhibitor BRCA : Breast cancer CI : Confidence interval ESR1 : Estrogen receptor 1 ER : Estrogen receptor ET : Endocrine therapy FDR : False discovery rate H3K27ac : Acetylation of lysine 27 on histone 3 HR : Hazard ratio LTED : Long-term estrogen deprived/deprivation RA : Resistance acquisition SE : Standard error TAM : Tamoxifen (patient data) TCGA : The Cancer Genome Atlas TMX : Tamoxifen (cell line experiment) **Publisher's Note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Maryam Soleimani Dodaran and Simone Borgoni: these authors contributed equally to this work Perry D. Moerland and Antoine H.C. van Kampen: these authors jointly supervised this work Supplementary information ========================= **Supplementary information** accompanies this paper at 10.1186/s12885-020-07100-z. We acknowledge Dr. Alex Michie and Dr. Age K. Smilde for their helpful suggestions while carrying out this research. The results shown here are in part based upon data generated by the TCGA Research Network: <https://www.cancer.gov/tcga>. MS performed the data analysis, interpreted the results, and drafted the manuscript. SB drafted parts of the manuscript, performed the qRT-PCR experiment and, together with ES, performed the cell line experiment under supervision of SW. SB, ES, PJV and SW critically reviewed the manuscript. MS, PJV, SW, PDM and AHCvK interpreted the results. PDM and AHCvK conceived the study, supervised the project and helped draft the manuscript. All authors read and approved the final manuscript. This work was supported by EpiPredict which received funding from the European Union's Horizon 2020 research and innovation programme under Marie Skłodowska-Curie grant agreement No 642691. The DNA methylation dataset supporting the conclusions of this article is available in the Genomics Data Commons Legacy Archive (<https://portal.gdc.cancer.gov/legacy-archive>) repository. The other datasets supporting the conclusions of this article are included within the article and its additional files. Not applicable. Not applicable. The authors declare that they have no competing interests.

Introduction ============ The rising utilization and success of *in vitro* fertilization (IVF) can be attributed to the optimization of several clinical and laboratory protocols \[[@B1]\]. One such example is the optimization of controlled ovarian stimulation (COS) by incorporating gonadotropin-releasing hormone antagonist (GnRH-ant)-based COS protocols \[[@B2][@B3]\]. These COS protocols have several advantages over traditional long GnRH-agonist protocols, including lower utilization of gonadotropins \[[@B2][@B3]\], a lower risk of ovarian hyperstimulation syndrome \[[@B2][@B3]\], and a lower rate of ovarian cyst formation \[[@B4]\]. COS with GnRH-ant based protocols can be preceded by pretreatment with oral contraceptive pills (OCPs) or estrogen, with the goal of synchronizing follicular growth \[[@B5][@B6]\]. Most of these findings have been reported in studies comparing OCPs \[[@B7][@B8][@B9][@B10][@B11][@B12]\] or estrogen \[[@B5][@B13][@B14][@B15][@B16]\] to no pretreatment. However, scarce data are available comparing transdermal pretreatment modalities to standard pretreatment modalities in terms of COS, embryological, or pregnancy outcomes. Thus, the primary objective of this study was to compare the impact of pretreatment with transdermal estradiol (E~2~) to OCPs on COS response in normal responders undergoing fresh IVF-embryo transfer (ET) cycles with GnRH-ant based protocols. Methods ======= 1. Inclusion and exclusion criteria ----------------------------------- All patients undergoing fresh IVF-ET cycles between January 2008 and June 2013 at the Round O. Perelman and Claudia Cohen Center for Reproducrive Medicine were analyzed for inclusion. Patients qualified as normal responders \[[@B17]\] when they met the following criteria: (1) age \<40 years, (2) cycle day (CD) 2/3 follicle stimulating hormone (FSH) level \<12 mIU/mL, (3) CD 2/3 E~2~ level \<75 pg/mL, and (4) anti-Müllerian hormone (AMH) level \>1 ng/mL. Patients not meeting the aforementioned criteria as well as those with known history of polycystic ovarian syndrome, or poor response or cancellation in a prior IVF-ET cycle were excluded from the analysis. The institutional review board at Weill Cornell Medical College approved our retrospective study protocol. 2. Clinical and laboratory protocols ------------------------------------ Previously described protocols for COS, ovulatory trigger, oocyte retrieval, and ET were utilized \[[@B18][@B19]\]. Patients were assigned to pretreatment with E~2~ or OCPs based on physician preference. E~2~ was administered via 0.1-mg E~2~ patches (Vivelle-Dot estradiol transdermal system, Novartis Pharmaceuticals Co., East Hanover, NJ, USA). The active E~2~ component in these patches is estra-1,3,5 (10)-triene-3,17\<β-diol. Patients applied 0.1-mg E~2~ patches 10 days after detection of the luteinizing hormone (LH) surge of the preceding menstrual cycle, and changed them every other day until the onset of menses \[[@B18]\]. COS with gonadotropins began on CD 2 of their menstrual cycle. Patient without menses after the application of the fourth E~2~ patch were underwent an assessment of their serum FSH, LH, E~2~, and β-human chorionic gonadotropin (β-hCG) levels as well as transvaginal pelvic ultrasonography, and if the findings were normal, proceeded with COS \[[@B8][@B12]\]. Patients undergoing pretreatment with OCPs began the pill (Ortho-Novum, Ortho-McNeil-Janssen Pharmaceuticals, Titusville, NJ, USA) on CD 1 of their preceding menstrual cycle. Following treatment with OCPs for 10 to 14 days, COS ensued approximately 2 to 3 days after discontinuation of the last pill. The gonadotropin doses for COS were based on the patient\'s age, body mass index (BMI, kg/m^2^), antral follicle count, and serum AMH level. COS was initiated with gonadotropins (Gonal-F, EMD Serono, Rockland, MA, USA or Follistim, Merck, Kenilworth, NJ, USA; and Menopur, Ferring Pharmaceuticals, Parsippany, NJ, USA). Ovulation was suppressed with once-daily 0.25-mg Ganirelix acetate (Merck) injections, which were started when the lead follicle was \>13 mm or the E~2~ level was 300 pg/mL \[[@B18][@B19]\]. The hCG trigger (Pregnyl, Merck or Novarel, Ferring Pharmaceuticals) was given when the two lead follicles attained a mean diameter of \>17 mm. Oocyte retrieval was performed approximately 34 to 35 hours after the hCG trigger under conscious sedation. Intramuscular progesterone (50 mg daily) was started the day after retrieval. Fertilization of oocytes was achieved with conventional *in vitro* insemination or intracytoplasmic sperm injection \[[@B20]\]. All embryos were incubated in in-house culture media \[[@B18]\]. All ETs were performed with Wallace catheters (Smiths Medical, Norwell, MA, USA). 3. Outcome variables -------------------- Baseline demographics recorded for patients included age (years), gravidity, BMI (kg/m^2^), infertility diagnosis, CD 2/3 FSH level (mIU/mL), AMH level (ng/mL), and number of previous IVF attempts. COS parameters included total days of ovarian stimulation, total days of GnRH-ant administration, total dosage of gonadotropins (IU), E~2~ level (pg/mL) on the day of trigger, peak endometrial thickness (mm), total number of oocytes retrieved, mature oocytes retrieved, and fertilization rate (%). The number of cycles canceled as well as the number of surplus embryos cryopreserved at the blastocyst stage were also noted. Any pregnancy with positive hCG but without a gestational sac was considered a biochemical pregnancy. Clinical pregnancy was defined as the number of intrauterine gestations with fetal cardiac activity per IVF-ET cycle. Pregnancy loss after visualizing an intrauterine gestation was considered a spontaneous miscarriage. A live birth was any birth after 24 weeks of gestation. 4. Statistical analysis ----------------------- Categorical variables were expressed as number of cases and percentage of occurrence and assessed using the chi-square test with the Mantel-Haenszel correction. Non-parametric variables were expressed as median (interquartile range) and were tested with the Wilcoxon rank-sum test. All continuous variables were checked for normality using the Shapiro-Wilk test and expressed as mean±standard deviation. The independent *t*-test was utilized for statistical comparisons of continuous variables. Statistical significance was set at *p*\<0.05. Based on the study of Hauzman et al. \[[@B14]\] that showed a gonadotropin dosage difference of 65 IU between the E~2~ (1,692±488 IU) and OCP (1,627±565 IU) groups, a sample of size of 1,036 patients per group was estimated, assuming an α-error of 5% and a power of 80%. Results ======= A total of 2,092 patients met the inclusion criteria: 1,057 patients in the E~2~ group and 1,035 patients in the OCP group. As shown in [Table 1](#T1){ref-type="table"}, the demographics and baseline characteristics were similar across both groups. Most patients had roughly two unsuccessful IVF-ET cycles elsewhere prior to pursuing treatment at our center. [Table 2](#T2){ref-type="table"} summarizes the COS outcomes of the study cohort. Patients in the OCP group had a longer duration of COS (10.7±1.63 days) than the transdermal E~2~ group (9.92±1.94 days). Furthermore, patients in the OCP group required higher cumulative doses of gonadotropins (2,657.3±1,187.9 IU) than the E~2~ group (2,550.1±1,270.2 IU, *p*=0.002). Overall, no difference was noted in the total days of GnRH-ant administration, E~2~ level on the day of trigger, peak endometrial thickness, number of total or mature oocytes retrieved, and the fertilization rate. No difference in IVF-ET cycle cancelation was observed in the E~2~ and OCP pretreatment groups. [Table 3](#T3){ref-type="table"} presents the pregnancy outcomes of the study cohort. No significant difference was found in the mean age of patients or the number of embryos transferred. The number of surplus embryos cryopreserved at the blastocyst stage was also comparable among the pretreatment groups. Overall, no statistically significant differences were found in the rates of biochemical pregnancy, clinical pregnancy, spontaneous miscarriage, or live birth when comparing both groups. These findings remained unchanged even after adjusting for duration of COS, gonadotropin dose, and E~2~ level on the day of the trigger. Discussion ========== Hormonal pretreatment modalities are used to suppress a patient\'s endogenous gonadotropin secretion, thereby promoting the coordinated growth of early antral follicles in response to exogenous gonadotropins \[[@B5][@B6]\]. The resulting synchronization of follicles has shown to increase oocyte and embryo yield and therefore, the overall chances of pregnancy \[[@B6][@B7]\]. OCPs are perhaps the most frequently used pretreatment modality and their impact on IVF-ET cycles have been well studied in normal responders \[[@B7][@B8][@B9][@B10][@B11]\], poor responders \[[@B12][@B21]\] and hyper-responders \[[@B22]\]. Conflicting data have been reported regarding the impact of OCP pretreatment on IVF-ET outcomes in normal responders, with some studies suggesting lower oocyte yield, clinical pregnancy, and live birth rates \[[@B10][@B11]\] than observed in patients receiving no pretreatment, while other studies have not replicated these findings \[[@B7],[@B14]\]. The variation in results may be attributed to the use of different OCPs with varying durations \[[@B7][@B14]\]. The large majority of studies do, however, emphasize that OCP pretreatment was associated with a longer duration of COS and higher gonadotropin utilization \[[@B9][@B10][@B11]\]. The use of E~2~ for pretreatment in normal responders was proposed as an alternative to OCPs given its shorter duration \[[@B5][@B14]\]. Administration of E~2~, which is generally 17-β-E~2~, during the luteal phase of the preceding cycle exerts negative feedback on FSH during the late luteal-early follicular phase transition, thereby suppressing follicular growth until the administration of exogenous gonadotropins \[[@B14]\]. In a randomized controlled trial of 100 patients comparing oral E~2~ to OCPs for pretreatment, Hauzman et al. \[[@B14]\] found no significant differences in stimulation, embryological, implantation, clinical pregnancy, spontaneous miscarriage, or live birth outcomes in fresh IVF-ET cycles. The same study also suggested that oral E~2~ could be used as an alternative to OCPs for scheduling or timing fresh IVF-ET cycles. It must be noted that the E~2~ utilized in all the aforementioned studies was E~2~ valerate (i.e., \[17β\]-3-hydroxyestra-1,3,5\[10\]-trien-17-yl valerate, which is the 17-pentanoyl ester of 17-β-E~2~). Our center has previously described the use of another form of E~2~ (17-β-E~2~, i.e., estra-1,3,5 \[10\]-triene-3,17β-diol) for pretreatment in poor responder patients \[[@B23]\]. In order to broaden the use of this regimen, we retrospectively evaluated its utility for the pretreatment of normal responder patients. In addition to its large sample size, the current study uniquely compares transdermal E~2~ to OCPs. The outcomes of the OCP group are consistent with previously published findings. However, unlike the study of Hauzman et al. \[[@B14]\], which showed no difference in stimulation outcomes between the E~2~ valerate and OCP groups, our findings revealed a shorter duration of COS and lower utilization of gonadotropins in the transdermal E~2~ group than in the OCP group. These contrasting results may be due to the difference in sample sizes and the E~2~ and OCP products utilized in the two studies. For example, patients using estrane-derived and gonane-derived OCPs, with higher androgenic properties, prior to COS have been shown to have lower oocyte yield than those using anti-androgenic OCPs or those not using OCPs \[[@B24]\]. Despite its strengths, the current study is limited by two main shortcomings. First, the sample size calculation was based on the gonadotropin dosage difference reported by Hauzman et al. \[[@B14]\]. If clinical pregnancy or live birth was considered as the primary outcome of interest, then post-hoc calculations based on the same reference would suggest a sample size of at least 9,804 patients per pretreatment group. Thus, the current study is underpowered to detect a difference in clinical pregnancy or live birth rates. Second, the assignment of patients to transdermal E~2~ or OCP pretreatment was based on physician preference, thus introducing some selection bias to the study methodology. In conclusion, our findings suggest that compared to OCPs, pretreatment with transdermal E~2~ was associated with shorter duration of COS and lower gonadotropin utilization in normal responder patients undergoing IVF-ET with GnRH-ant based protocols. The overall yield of oocytes and embryos, as well as the clinical pregnancy and live birth rates, remain unaffected by E~2~ or OCP pretreatment. As pretreatment with E~2~ begins in the preceding luteal phase, the overall length of pretreatment may also be shorter than for OCPs. While pretreatment with OCPs and oral E~2~ has previously been described for prospective scheduling of IVF-ET cycles, the current study presents reasonable data to suggest that OCP or E~2~ pretreatment may not be required in normal responders because they do not improve the overall pregnancy outcomes of fresh IVF-ET cycles. However, these suppositions need prospective validation in studies with adequate sample sizes. This article was presented at the 2015 annual meeting, American Society of Reproductive Medicine, Baltimore, MD, October 17-21, 2015. **Conflict of interest:** No potential conflict of interest relevant to this article was reported. ###### Baseline characteristics of the study cohort (n=2,092)  Values are presented as mean±standard deviation or number (%). E~2~, estradiol; OCP, oral contraceptive pill; NS, not significant; BMI, body mass index; FSH, follicle-stimulating hormone; AMH, anti-Müllerian hormone; IVF, *in vitro* fertilization. ###### Ovarian stimulation characteristics of the study cohort (n=2,092)  Values are presented as mean±standard deviation or number (%) unless otherwise indicated. E~2~, estradiol; OCP, oral contraceptive pill; NS, not significant. ###### Cycle outcomes of the study cohort (n=2,092)  Values are presented as mean±standard deviation or number (%). E~2~, estradiol; OCP, oral contraceptive pill; NS, not significant.

I am writing regarding the systematic review about clinical validity, understandability, and actionability of online cardiovascular disease (CVD) risk calculators recently published by Dr Bonner and colleagues \[[@ref1]\]. Although Dr Bonner and colleagues used a comprehensive two-step research strategy to identify Web addresses that contained a CVD risk calculator, which led to the identification of 67 Web pages, a very important CVD risk model, the Systematic COronary Risk Evaluation (SCORE) risk assessment model \[[@ref2]\], was ignored. Developed by the European Society of Cardiology, this model was derived from 12 European cohort studies (250,000 patients data collected and 3 million person-years of observation) and is based on classical risk factors such as gender, age, total cholesterol, systolic blood pressure, and smoking status. The SCORE risk assessment model should have been included because it satisfies Dr Bonner's inclusion criteria as it predicts the risk of developing a CVD event and an electronic interactive version of this model is freely available on the European Society of Cardiology\'s Web page \[[@ref3]\]. Furthermore, there are other risk assessment models locally developed in countries such as China, India, and Korea that are not taken into account in this study. Thus, it would have been useful if the authors had added to their research strategy a literature search of review documents focusing on cardiovascular risk assessment as was carried out by Zhao and colleagues \[[@ref4]\]. **Editorial note**: Authors were invited to respond but declined. They agree that additional calculators exist that could have been included if a different method was used. CVD : cardiovascular disease SCORE : Systematic COronary Risk Evaluation

Introduction ============ The incidence and mortality rates of colorectal cancer (CRC) vary around the world ([@b1-or-43-04-1187],[@b2-or-43-04-1187]). Early stage CRC commonly shows limited clinical signs, and patients are often diagnosed at the metastatic stage, rendering therapy difficult ([@b3-or-43-04-1187]). Therefore, early diagnosis and treatment of CRC remains challenging for patients and surgeons. In 2016, CRC ranked fourth and second among the most frequently diagnosed and the deadliest malignancies, respectively, in the USA ([@b4-or-43-04-1187]--[@b7-or-43-04-1187]). According to the American Cancer Society, there were approximately 13,450 newly diagnosed cancer patients in the US in 2016, 30% of whom presented with CRC ([@b8-or-43-04-1187]). In 2015, 376,000 new CRC cases were diagnosed in China, with 191,000 succumbing to the malignancy ([@b9-or-43-04-1187]). National polyp screening programs constitute an early diagnosis tool, which can markedly improve CRC prognosis ([@b10-or-43-04-1187]--[@b14-or-43-04-1187]). Early diagnosis and treatment of CRC is becoming increasingly important to surgeons. Surgery remains the principal therapeutic option for loco-regional CRC. At present, robotic and laparoscopic surgeries are performed for CRC, with improve patient outcome in comparison with traditional surgical techniques ([@b15-or-43-04-1187]--[@b17-or-43-04-1187]). However, patient prognosis is not significantly enhanced. In order to improve the prognosis of CRC patients, surgeons and pathologists have made unremitting efforts to examine the prognostic values of various tumor markers. The American Joint Committee on Cancer (AJCC) tumor-node-metastasis (TNM) staging system provides a universal modality and guides clinical treatment ([@b18-or-43-04-1187]--[@b21-or-43-04-1187]). Based on the original version, AJCC-8 (American Joint Committee on Cancer 8 edition) provides improved guidance for the individualized treatment of CRC patients and more effective treatment of patients with IVC peritoneal metastases. Tumor invasion and metastasis of CRC result from well-coordinated events involving many intracellular and extracellular factors ([@b22-or-43-04-1187]--[@b24-or-43-04-1187]). While many factors affect prognosis in CRC ([@b25-or-43-04-1187],[@b26-or-43-04-1187]), Ki67 is broadly employed in pathological analyses to evaluate cell proliferation in various malignancies ([@b27-or-43-04-1187]--[@b30-or-43-04-1187]). Although Ki67 is expressed in benign tumors, its levels are very low; however, it is found at high levels in multiple malignant lesions, and tightly associated with distant metastasis, resulting in poor patient prognosis. The prognostic value of Ki67 has been assessed in various types of cancers, particularly brain, neuroendocrine, and lymphoid tissue malignancies, and its levels are commonly utilized to grade tumors ([@b31-or-43-04-1187]). Nevertheless, its prognostic and predictive roles remain debatable mostly as standard quantification techniques for Ki67 are in existence ([@b32-or-43-04-1187]). Ki67 expression is usually examined as a percentage, which is closely related to the pathologist\'s clinical experience. Hashimoto *et al* assessed the rate and clinical significance of fascin expression in association with CRC progression and cancer cell proliferation based on Ki67 ([@b33-or-43-04-1187]). Most often, Ki67 is assessed visually by pathologists although no consensus is available concerning the specific regions to score ([@b34-or-43-04-1187]). Meanwhile, whether automated techniques could yield suitable accuracy and prognostic power for Ki67 is not known. Indeed, head-to-head comparisons between scores from automated and pathologist-based techniques in terms of prognostic value have been rarely reported, and discrepant findings in breast cancer have been obtained ([@b35-or-43-04-1187]--[@b37-or-43-04-1187]). Previous reports ([@b38-or-43-04-1187],[@b39-or-43-04-1187]) discussed the role of Ki67 expression in lung and breast cancers, examining ways to define the cutoff of Ki67 expression. Similar questions remain for CRC. How to grade Ki67 expression remains therefore an open question. A 20% cutoff has been reported ([@b38-or-43-04-1187]). Nonetheless, a previous meta-analysis assessing various cut-off levels of Ki67 in regards to prognosis suggested a visual cut-off \>25% to provide a higher discriminatory power in mortality risk compared with the remaining cut-off points evaluated ([@b39-or-43-04-1187]). Signal intensity scores were 0 (negative), 1 (weak), 2 (moderate) and 3 (strong); positivity extent was scored as 0 (\<5%), 1 (5--25%), 2 (\>25--50%, 3 (\<50--75%) and 4 (\>75%). Both sub-scores were multiplied to yield the final score, which was considered to be positive if \>5 ([@b40-or-43-04-1187]). Can a suitable cutoff increase the prognostic value of Ki67 expression in colorectal cancer? This is the starting point of the present research; as not many studies have been reported. Some scholars hold opposite views on the relationship between Ki67 expression and prognosis in CRC, suggesting that high Ki67 expression instead reflects better prognosis ([@b41-or-43-04-1187]). Other studies have reported that mean Ki67 expression is higher in p53-positive cases, and Ki67 and p53 are not correlated to clinical and pathological parameters ([@b42-or-43-04-1187]). Whether Ki67 expression is related to clinicopathological indicators and prognosis remains controversial. Meanwhile, the cutoffs vary, and the outcomes are rather controversial among previous studies. Therefore, we analyzed the associations of Ki67 expression with clinicopathological parameters and the prognosis of CRC patients in this study. Here we divided cases into four grades based on 25% intervals of Ki67 immunohistochemical signals. Associations of Ki67 expression levels with clinicopathological factors and CRC prognosis were analyzed. Prognosis in CRC was also analyzed based on Ki67 expression according to 5-year disease-free survival (DFS) and overall survival (OS) in and out of the AJCC-8 stratification. Patients and methods ==================== ### Patients In total, 2,080 CRC cases were enrolled at Huzhou Central Hospital between January 2006 and December 2012. A total of 400 cases did not undergo surgery, 400 succumbed to non-CRC causes, and 190 were lost to follow-up and thus were excluded from the present study. Therefore, 1,090 cases (stage 0 to stage IV) were involved in the final analysis. Inclusion criteria were CRC diagnosis by colonoscopy, computed tomography and pathology; no pre-surgical adjuvant therapy, radical surgery and normal lymph nodes harvested; other organ metastases found before or during surgery, and combined resection to achieve R0 resection; complete postoperative clinical and pathological data; postoperative routine immunohistochemical and pathological analyses; post-surgical chemotherapy based on the National Comprehensive Cancer Network (NCCN) guidelines; adenocarcinoma by pathological diagnosis; complete follow-up data, including recurrence and metastasis at follow-up. Exclusion criteria included severe heart, brain, liver or lung disease which may influence tolerance to surgery; non-CRC parameters causing death, interstitial or neuronal tumor, lymphoma, melanoma and other non-adenocarcinomas concomitant with CRC ([Fig. 1](#f1-or-43-04-1187){ref-type="fig"}). ### Follow-up Routine follow-up was carried out in the outpatient clinic two weeks post-operation, at 3- and 6-month intervals for the first and second years, respectively, and yearly for the remaining 3 years. Phone calls and mail were also used for follow-up. During the follow-up period, the patient statuses included i) death, censored and ii) death and recurrence and censored. ### Ethics statement The current trial followed the 2008 Declaration of Helsinki, and had approval from the Ethics Committee of Huzhou Central Hospital (Huzhou, Zhejiang, China). All patients provided signed informed consent for the use of their tissue samples for Ki67 immunohistochemistry immunoassay and medical records for research. ### Detection of tissue Ki67 Immunohistochemistry was performed by the Envision two-step method \[cat. no. ZM-0166 (Beijing Zhongshang Jinqiao Co.); K5007 (Dako)\]. The primary antibody was raised against Ki67 (cat. no. ZM-0166, 1:200 dilution) and K5007 (Dako; no dilution) was used as the secondary antibody. The steps included: i) Dewaxing with hot water; ii) antigen repair under high pressure citric with acid at pH 6.0; iii) hydrogen peroxide blocking of endogenous peroxidase; iv) primary antibody incubation at 37°C for 30 min; v) secondary antibody incubation at 37°C for 15 min; vi) DAB staining at 22°C for 5 min; vii) dehydration and mounting. We compared conventional hematoxylin and eosin (H&E) staining with Ki67 DBA immunostaining and defined + as \>0 and ≤25%; ++ as \>25 and ≤50%; +++ as \>50 and ≤75%; and ++++ as \>75% ([Fig. 2](#f2-or-43-04-1187){ref-type="fig"}). ### Surgical methods According to lesion location and the principle of malignant tumor resection, the following resection methods were used: Right hemicolectomy (RHC); left hemicolectomy specimen (LHC), with the excised portion including the left colon and the descending colon; Hartmann (HO), with 5 cm tissue distal to the tumor removed and the distal colorectal segment closed, the proximal end removed (10 cm), and the proximal end taken for fistula; anterior rectal resection specimen (AR), with the extent of resection involving the sigmoid colon and part of the rectum; anterior rectal perineal resection specimen (APR), with the scope of resection involving the sigmoid colon, whole rectum and anal canal and perineum. ### Statistical analysis SPSS 21.0 (IBM Corp.) was employed for data analysis. Clinicopathological measurement data among groups with different Ki67 expression patterns were assessed by one-way analysis of variance (ANOVA); count data were analyzed by Crosstabs and the Pearson\'s χ^2^ test. Bivariate correlation analysis was performed to assess clinical and pathological indicators with significant significance in Ki67 expression. Five-year DFS and OS were analyzed by the Kaplan-Meier method and the Breslow test. Survival rates were equally assessed by multivariable Cox\'s regression according to various clinical, pathological, and biochemical parameters, which were analyzed in our previous studies ([@b21-or-43-04-1187],[@b26-or-43-04-1187]). Patient statuses were divided into two: i) Only death was considered an event, and other parameters were censored for OS analysis; ii) death and recurrence were considered events, and other parameters were censored for DFS analysis. Weighted analysis and the non-parametric Chi-square test were applied to compare DFS and OS under different Ki67 levels. Results ======= ### General data A total of 1,090 patients of the 2,080 enrolled CRC cases were evaluated (52.4%), including 550 men (50.5%) and 540 women (49.5%). According to Ki67 expression (+, ++, +++ and ++++), the entire patient population consisted of 61 (11.1%), 141 (25.6%), 202 (36.7%) and 146 (26.5%) male patients, and 70 (13%), 144 (26.7%), 195 (36.1%) and 131 (24.3%) females, respectively. The mean age was 62.26 years (range, 17--89). ### Clinicopathological properties of the various groups based on Ki67 expression According to Ki67 expression (from low to high), sex, age, American Society of Anesthesiologists (ASA) stage, location, surgical method, operation time, invasive depth, tumor differentiation, tumor size, AJCC-8 stage, the number of lymph nodes harvested, the number of positive lymph nodes, complications, and chemotherapy status were assessed. By single factor ANONA and F test, there were no significant differences in age, operation time and the number of lymph nodes harvested. However, there were significant differences noted in regards to tumor size and the number of positive lymph nodes. Regarding measurement variations, from Ki67+ to Ki67++++, mean and standard deviations were as follows: Age, 61.64±15.80, 62.79±14.33, 62.21±14.28 and 62.07±14.31 (P=0.880); operation time (min), 153.6±34.8, 151.2±38.6, 155.0±33.2 and 154.1±33.2 (P=0.568); number of lymph nodes harvested, 14.04±1.9, 14.17±1.8, 14.25±1.8 and 14.19±1.9 (P=0.727); tumor size (cm), 3.58±1.1, 3.38±1.2, 3.74±0.9 and 3.79±0.9 (P\<0.001); number of positive lymph nodes, 0.25±0.9, 0.61±1.4, 2.45±2.2 and 2.86±2.5 (P\<0.001). From Ki67(+) to Ki67(++++) significant differences were found in count variables such as invasive depth, tumor differentiation, AJCC-8 stage and chemotherapy status (P\<0.001; P\<0.001; P=0.003 and P=0.005, respectively). However, no statistical differences were found in sex, ASA stage, location, surgical method, and complications (P=0.684, P=0.860, P=0.439, P=0.768 and P=0.587, respectively). Details are shown in [Table I](#tI-or-43-04-1187){ref-type="table"}. ### Associations of Ki67 with clinicopathological features showing significant differences We further analyzed the associations of clinicopathological indices which showed significant differences based on Ki67 expression. Spearman rho coefficients of invasive depth, tumor differentiation, tumor size, AJCC-8 stage, the number of positive lymph nodes and chemotherapy status were 0.170 (95% CI 0.113--0.225, P\<0.001), 0.456 (95% CI 0.411--0.500, P\<0.001), 0.122 (95% CI −0.063--0.181, P\<0.001), 0.195 (95% CI 0.138--0.254, P\<0.001), 0.514 (95% CI 0.468--0.558) and −0.201 (95% CI −0.253--0.148, P\<0.001), respectively, as summarized in [Table II](#tII-or-43-04-1187){ref-type="table"}. ### Five-year DFS and OS by Ki67 expression in AJCC-8 stratification The Kaplan-Meier method and Breslow test were applied to assess DFS and OS based on Ki67 expression in the AJCC-8 stratification. There were significant differences in DFS among the various Ki67 expression groups from AJCC-8=I to AJCC-8=III (all P\<0.001), but no statistical significance at AJCC-8=IV (P=0.334; [Fig. 3A-D](#f3-or-43-04-1187){ref-type="fig"}). There were significant differences in OS among the different Ki67 expression groups from AJCC-8=I to AJCC-8=III (all P\<0.001), but no statistical significance at AJCC-8=IV (P=0.334; [Fig. 4A-D](#f4-or-43-04-1187){ref-type="fig"}). DFS and OS survival rates were assessed based on Ki67 expression and AJCC-8 stratification. [Table III](#tIII-or-43-04-1187){ref-type="table"} documents the DFS and OS survival rates at different Ki67 levels in the AJCC-8 stratification. There were no patients with Ki67+++ or Ki67++++ at stage 0, not only in DFS but also in OS. [Table IV](#tIV-or-43-04-1187){ref-type="table"} shows a comparison of DFS and OS at different Ki67 expression levels. There were no significant differences among all columns (χ^2^=0.202, P=0.653; χ^2^=0.098, P=0.755; χ^2^=0.136, P=0.713 and χ^2^=0.211 P=0.646 respectively). DFS and OS showed statistical differences only between Ki67 expression groups (Breslow=164.66, P\<0.001; Brelow=166.79, P\<0.001), indicating elevated Ki67 expression was associated with poorer prognosis. Details are shown in [Table IV](#tIV-or-43-04-1187){ref-type="table"} and [Fig. 5A and B](#f5-or-43-04-1187){ref-type="fig"}. Crosstabs analysis showed that there were no significant differences between DFS and OS at different Ki67 expression levels (χ^2^=0.041, P=0.098; [Fig. 5C](#f5-or-43-04-1187){ref-type="fig"}). ### Multivariable analysis of CRC prognostic factors To identify independent predictive factors of CRC prognosis, Cox proportional hazard model analysis was performed. Sex, invasive depth, lymph node metastasis, tumor differentiation, AJCC-8 stage, chemotherapy status and Ki67 were included in the model. Regarding sex and chemotherapy status, hazard ratios (HRs) (95% CIs) for Female/Male and No/Yes were 0.976 (0.796--1.198) and 0.986 (0.522--1.862), respectively, which were not significant (P=0.819 and P=0.964, respectively). HRs (95% CIs) for invasive depth at T2/Tis and T1, T3/Tis and T1, T4/Tis and T1 were 1.5336 (0.855--2.748), 1.845 (1.034--3.290) and 1.331 (0.746--2.376), respectively, showing statistically significant differences (P=0.03). HRs (95% CIs) for lymph node metastasis at N1/N0 and N2/N0 were 0.909 (0.616--1.342) and 1.690 (1.168--2.446) (P\<0.001). Regarding differentiation (moderate/well, poor or no/well), AJCC-8 stage (I/0, II/0, III/0 and IV/0), Ki67 expression (++/+, +++/+ and ++++/+) HRs (95% CIs) were, respectively, 1.677 (1.036--2.715), 6.443 (3.883--10.756) and 134.375 (0--1.173E+29); 2,015.297 (0--1.736E+30), 1,098.443 (0--9.461E+30) and 30582.466 (0--2.637E+31); and 2.59 (1.327--50055), 4.732 (2.275--9.843) and 6.762 (3.226--14.174), showing significant differences (all P\<0.001; [Table V](#tV-or-43-04-1187){ref-type="table"}). Discussion ========== In the present study, we initially considered whether to include stage 0. Some Ki67 data were not available in this period, which inevitably resulted in that survival classified by Ki67 was not calculated. According to AJCC-8 (American Joint Committee on Cancer 5th edition) and later versions, colorectal cancer CRC) stage 0 refers to TisN0M0. This refers to carcinoma *in situ*, which is localized within the epithelium or infiltrates the lamina propria. Although it is not invasive cancer, it is invasive. We believe that stage 0 tumors are within the mucosa, but not invasive malignant tumors because invasive malignancies usually refer to advanced cancer. Therefore, we included stage 0 in this study. With the increasing attention paid to early CRC detection, this study found that stages 0-II only accounted for 34.12% of all cases, and most cases were stage III in the same period, accounting for 60%. Therefore, the diagnosis and treatment of CRC still requires further investigation to improve prognosis. As published in previous reports ([@b10-or-43-04-1187]--[@b14-or-43-04-1187]), national polyp screening programs could improve CRC prognosis drastically. Research and application of many tumor markers improve the early detection rate of tumors as well as patient prognosis ([@b43-or-43-04-1187]--[@b45-or-43-04-1187]). Ki67 expression is a tumor marker that has been used for a long time in clinical practice, but its classification criteria and relationship with prognosis remain controversial ([@b39-or-43-04-1187],[@b46-or-43-04-1187],[@b47-or-43-04-1187]). What role does Ki67 expression play in the prognosis of colorectal cancer? Melling *et al* ([@b41-or-43-04-1187]) considered that high Ki67 has a good prognostic value for CRC, contrasting with Luo *et al* ([@b40-or-43-04-1187]). Their results showed that high Ki67 expression is associated with low tumor stage and nodal status, but not with tumor grade, histological tumor type or tumor localization, representing an independent predictor of favorable survival; these findings strongly argue for a clinical utility of Ki67 immunostaining as an independent prognostic biomarker in CRC. This study showed that high Ki67 expression was associated not only with tumor stage (AJCC-8), tumor size and nodal status, but also with tumor differentiation, tumor invasive depth, and chemotherapy status, which were not discussed in Meling *et al* ([@b41-or-43-04-1187]). We used 25% as a cutoff for Ki67 expression and different study methods such as in and out AJCC-8 stratification, which may explain the discrepancy. The above findings indicate that it is ideal to use 25% as a cutoff for Ki67 expression. The nuclear protein Ki67 was first described in Hodgkin lymphoma-derived cells ([@b48-or-43-04-1187]). It is expressed throughout cell division, but is highly suppressed in resting cells (G0 phase) ([@b49-or-43-04-1187],[@b50-or-43-04-1187]). Ki67 staining is broadly utilized clinically as an index of cell proliferation, although its functions and dynamics are poorly understood. Miller *et al* ([@b51-or-43-04-1187]) tracked Ki67 amounts in single cells without external stimuli, and demonstrated that it accumulates only in the S, G2, and M phases, with continuous degradation in G1 and G0. Ki67 expression is commonly utilized in oncology as a proliferation indicator. Here we explored the association of Ki67 expression with CRC. Forones *et al* ([@b42-or-43-04-1187]) hypothesized that Ki67 and P53 are not correlated with clinical and pathologic parameters. This study showed differences in tumor invasive depth based on the Ki67 amounts, and elevated Ki67 expression was associated with increased invasive depth. This may be because tumor invasion and metastasis result from highly coordinated events involving many intracellular and extracellular factors ([@b18-or-43-04-1187],[@b19-or-43-04-1187]). Another reason may be that higher Ki67 amounts are associated with poorer tumor differentiation and high AJCC grade, as well as elevated positive lymph node rate, corroborating this study. We assessed the associations of clinicopathological features with Ki67 expression, and the results showed strong correlations, which require confirmation by molecular and genetic studies. We also showed that the surgical method was not related to Ki67 expression, demonstrating that cancer location in CRC is not associated with Ki67 expression. In this study, all tumor cases were adenocarcinomas. The associations of other types of cancer, such as melanoma, carcinoid, malignant stoma tumor and neurofibroma, with Ki67 expression, were not covered in this work. This is a flaw in the present study. We expect relevant studies to be performed. One of the highlights of this work is that DFS and OS were analyzed based on Ki67 expression in and out of the AJCC-8 stratification. Higher Ki67 expression levels reflected poorer DFS and OS out of the AJCC-8 stratification with statistical significance. Feng *et al* ([@b52-or-43-04-1187]) reported that elevated Ki67 expression is associated with poorer prognosis in breast cancer. Shin *et al* ([@b53-or-43-04-1187]) pointed out that high Ki67 reflects poor prognosis in CRC. This suggests that the gene encoding Ki67 has a similar function in tumors. However, it is puzzling that analysis based on the AJCC-8 stratification revealed comparable DFS and OS for different Ki67 levels in stage IV cases. In general, Ki-67 is closely related to RNA transcription, and shows high expression levels during cell division and proliferation, reflecting the activity of cell division and increasing the risk of tumor invasion and metastasis, which worsen patient prognosis ([@b54-or-43-04-1187]). We believe that patients with stage IV disease have poorer prognosis, and adverse factors other than Ki67 expression may play additional roles, such as surgery, chemotherapy, patient physical status, and genetic factors ([@b55-or-43-04-1187]--[@b60-or-43-04-1187]). In the present study, cases with stage IV disease were limited in number, which could have caused a bias. We look forward to undergoing future research to assess patients with stage IV CRC and Ki67 expression. Finally, the present study demonstrated that Ki67 expression is an independent risk factor for poor prognosis in CRC by multivariate analysis and Cox regression. Mucinous vs. non-mucinous lesions have different molecular mechanisms ([@b61-or-43-04-1187],[@b62-or-43-04-1187]), but this issue was not investigated in this study. We used the Broder classification of clinicopathological features to analyze Ki67 expression. Some limitations may exist, including no analysis of the marker P53. In conclusion, a cutoff of 25% is a good classification tool. In such classification, high Ki67 amounts are closely associated with poor prognosis in CRC and independently predict prognosis in the AJCC-8 stratification. We are grateful to Professors Liqing Li, Qiang Yan and Zhihong Ma for critically revising this manuscript, as well as to Professor Jingliang Ping and Wei Xu for carrying out the pathological analyses. Trial registration: Chinese Clinical Trial Registry ChiCTR20190404 Registered 4 April 2019 Retrospectively registered [http//:www.chictrorgcnindexaspx](http//:www.chictrorgcnindexaspx). Funding ======= No current external funding sources sponsored this study, which was funded by Project 2018C37090 to GT. Availability of data and materials ================================== The corresponding author will provide all data upon reasonable request. Authors\' contributions ======================= GT, GZ, JL and ZZ conceived and designed the study. GT and YC performed the procedures of data collection and statistical analysis and wrote the initial manuscript. PN and XX were involved in the conception of the study and edited the manuscript. All authors read and approved the final manuscript. Ethics approval and consent to participate ========================================== The current trial followed the 2008 Declaration of Helsinki, and had approval from the Ethics Committee of Huzhou Central Hospital (Huzhou, Zhejiang, China). All patients provided signed informed consent for the use of their tissue samples for Ki67 immunohistochemistry immunoassay and medical records for research. Patient consent for publication =============================== Not applicable. Competing interests =================== All authors declare no competing interests regarding the present study. {#f1-or-43-04-1187} {#f2-or-43-04-1187} {#f3-or-43-04-1187} {#f4-or-43-04-1187} {#f5-or-43-04-1187} ###### Association of the clinicopathological features and Ki67 expression in all involved CRC cases. N Ki67+ Ki67++ Ki67+++ Ki67++++ P-value ------------------------------ ------- ------------- ------------- ------------- ------------- -------------------------------------------------------- Sex, n (%) 0.684   Male 550 61 (46.6) 141 (49.5) 202 (50.9) 146 (52.7)   Female 540 70 (53.4) 144 (50.5) 195 (49.1) 131 (47.3) Mean age (years) 1,090 61.64±15.80 62.79±14.33 62.21±14.28 62.07±14.31 0.880 ASA stage, n (%) 0.860   I 797 94 (71.8) 212 (74.4) 293 (73.8) 198 (71.3)   II 264 33 (25.2) 64 (22.5) 93 (23.4) 74 (26.7)   III 29 4 (3.1) 9 (3.2) 11 (2.8) 5 (1.8) Location, n (%) 0.439   Ileocecum 73 11 (8.4) 20 (7.0) 24 (6.0) 18 (6.5)   Right colon 95 4 (3.1) 31 (10.9) 41 (10.3) 19 (6.9)   Transverse colon 174 25 (19.1) 42 (14.7) 60 (15.1) 47 (17.0)   Left colon 206 24 (18.3) 57 (20.0) 75 (18.9) 50 (18.1)   Sigmoid colon 108 16 (12.2) 20 (7.0) 40 (10.1) 32 (11.6)   Rectum 434 51 (38.9) 115 (40.4) 157 (39.5) 111 (40.1) Surgical method, n (%) 0.768   RHC 207 21 (16.0) 58 (20.4) 80 (20.2) 48 (17.3)   LHC 431 55 (42.0) 108 (37.9) 155 (39.0) 113 (40.8)   HO 24 3 (2.3) 7 (2.5) 6 (1.5) 8 (2.9)   AR 327 43 (32.8) 91 (31.9) 112 (28.2) 81 (29.2)   APR 101 9 (6.9) 21 (7.4) 44 (11.1) 27 (9.7) Operation time (min) 1,090 153.6±34.8 151.2±38.6 155.0±33.2 154.1±33.2 0.568 Invasive depth, n (%) \<0.001^[a](#tfn1-or-43-04-1187){ref-type="table-fn"}^   Tis and T1 127 24 (18.3) 53 (18.6) 31 (7.8) 19 (6.9)   T2 210 5 (3.8) 66 (23.2) 112 (28.2) 27 (9.7)   T3 421 78 (59.5) 92 (32.3) 132 (33.2) 119 (43.0)   T4 332 24 (18.3) 74 (6.0) 122 (30.7) 112 (40.4) Differentiation, n (%) \<0.001^[a](#tfn1-or-43-04-1187){ref-type="table-fn"}^   Well 194 58 (44.3) 97 (34.0) 30 (7.6) 9 (3.2)   Moderate 696 69 (52.7) 177 (62.1) 282 (71.0) 168 (60.6)   Poor or undifferentiation 200 4 (3.1) 11 (3.9) 85 (21.4) 100 (36.1) Tumor size (cm) 1,090 3.58±1.1 3.38±1.2 3.74±0.9 3.79±0.9 \<0.001^[a](#tfn1-or-43-04-1187){ref-type="table-fn"}^ AJCC-8, n (%) 0.003^[a](#tfn1-or-43-04-1187){ref-type="table-fn"}^   0 16 5 (3.8) 11 (3.9) 0 (0) 0 (0)   I 131 17 (13.0) 85 (29.8) 28 (7.1) 1 (0.4)   II 225 93 (71.0) 112 (39.3) 12 (3.0) 8 (2.9)   III 663 14 (10.7) 75 (26.3) 323 (81.4) 25 (90.6)   IV 55 2 (1.5) 2 (0.7) 34 (8.6) 17 (6.1) No. of lymph nodes harvested 1,090 14.04±1.9 14.17±1.8 14.25±1.8 14.19±1.9 0.727 No of positive lymph nodes 1,090 0.25±0.9 0.61±1.4 2.45±2.2 2.86±2.5 \<0.001^[a](#tfn1-or-43-04-1187){ref-type="table-fn"}^ Complications 0.587   Yes 104 13 (9.9) 28 (9.8) 32 (8.1) 31 (11.2)   No 986 118 (90.1) 257 (90.2) 365 (91.9) 246 (88.8) Chemotherapy, n (%) 0.005^[a](#tfn1-or-43-04-1187){ref-type="table-fn"}^   Yes 895 107 (81.7) 189 (66.3) 343 (86.4) 256 (92.4)   No 195 24 (18.3) 96 (33.7) 54 (13.6) 21 (7.6) P\<0.05, statistical difference. Data for age, operation time, tumor size, no. of lymph harvested and no. of positive lymph nodes are expressed as mean ± standard deviation. CRC, colorectal cancer; AJCC-8, American Joint Committee on Cancer 8 edition; ASS, American Society of Anesthesiologists. Surgical method: RHC, right hemicolectomy; LHC, left hemicolectomy; HO, Hartmann; AR, anterior rectal resection APR, anterior rectal perineal resection. ###### Significant correlations between Ki67 and clinicopathological features of the CRC cases. Ki67 Spearman rho 95% CI (lower-upper) P-value ---------------------------- -------------- ---------------------- -------------------------------------------------------- Invasive depth 0.170 0.113--0.225 \<0.001^[a](#tfn3-or-43-04-1187){ref-type="table-fn"}^ Tumor differentiation 0.456 0.411--0.500 \<0.001^[a](#tfn3-or-43-04-1187){ref-type="table-fn"}^ Tumor size 0.122 −0.063--0.181 \<0.001^[a](#tfn3-or-43-04-1187){ref-type="table-fn"}^ AJCC-8 stage 0.195 0.138--0.254 \<0.001^[a](#tfn3-or-43-04-1187){ref-type="table-fn"}^ No of positive lymph nodes 0.514 0.468--0.558 \<0.001^[a](#tfn3-or-43-04-1187){ref-type="table-fn"}^ Chemotherapy status −0.201 −0.253--0.148 \<0.001^[a](#tfn3-or-43-04-1187){ref-type="table-fn"}^ Correlation is significant at 0.05. CRC, colorectal cancer; CI, confidence interval AJCC-8, American Joint Committee on Cancer 8 edition. ###### Analysis of the 5-year DFS and OS by Ki67 expression according to AJCC-8 stage. N \+ ++ +++ ++++ ----- ----- ----- ------ ------ ------ ------ DFS 0 16 100 100 -- -- I 131 88.2 94.1 96.4 0 II 225 90.3 72.3 25.0 0 III 663 64.3 73.3 61.0 39.8 IV 55 0 0 0 0 OS 0 16 100 100 -- -- I 131 94.1 100 100 0 II 225 96.8 75.9 25.0 0 III 663 64.3 77.3 64.4 44.2 IV 55 0 0 5.9 0 -, no case of Ki67 expression. AJCC-8, American Joint Committee on Cancer 8 edition; DFS, disease-free survival; OS, overall survival. Data expressed as a percentage (%). ###### Comparison of the 5-year DFS and OS by Ki67 expression (%). N \+ ++ +++ ++++ Breslow P-value --------- ------- ------- ------- ------- ------- --------- --------- DFS 1,090 86.3 79.6 57.2 36.1 164.66 \<0.001 OS 1,090 91.6 83.9 60.7 40.1 166.79 \<0.001 χ^2^ 0.202 0.098 0.136 0.211 P-value 0.653 0.755 0.713 0.646 DFS, disease-free survival; OS, overall survival. Data are expressed as a percentage (%). ###### Multivariate analysis of prognosis for CRC using OS. Factors HR (95% CI) P-value ------------------------------------ --------------------------- -------------------------------------------------------- Sex 0.819   Female/Male 0.976 (0.796--1.198) Invasive depth 0.030^[a](#tfn5-or-43-04-1187){ref-type="table-fn"}^   T2/Tis andT1 1.5336 (0.855--2.748)   T3/Tis and T1 1.845 (1.034--3.290)   T4/Tis and T1 1.331 (0.746--2.376) Lymph node metastasis \<0.001^[a](#tfn5-or-43-04-1187){ref-type="table-fn"}^   N1/N0 0.909 (0.616--1.342)   N2/N0 1.690 (1.168--2.446) Differentiation \<0.001^[a](#tfn5-or-43-04-1187){ref-type="table-fn"}^   Moderate/well 1.677 (1.036--2.715)   Poorly or undifferentiation/well 6.443 (3.883--10.756) AJCC-8 stage \<0.001^[a](#tfn5-or-43-04-1187){ref-type="table-fn"}^   I/0 134.375 (0--1.173E+29)   II/0 2,015.297 (0--1.736E+30)   III/0 1,098.443 (0--9.461E+30)   IV/0 3,0582.466 (0--2.637E+31) Chemotherapy status 0.964   No/Yes 0.986 (0.522--1.862) Ki67 \<0.001^[a](#tfn5-or-43-04-1187){ref-type="table-fn"}^   ++/+ 2.59 (1.327--50055)   +++/+ 4.732 (2.275--9.843)   ++++/+ 6.762 (3.226--14.174) P\<0.05 Statistical significant. CRC, colorectal cancer; OS, overall survival; HR, hazard ratio; CI, confidence interval; AJCC-8, American Joint Committee on Cancer 8 edition.

Introduction ============ The *Salmonella*e are major human pathogens and represent a significant global public health issue causing morbidity and mortality resulting in a high social and economic burden worldwide ([@ref-19]). The genus consists of two species; *Salmonella enterica* and *S. bongori*. There are six subspecies of *S. enterica* differentiated by biochemical variations, namely subspecies *enterica* (I), *salamae* (II), *arizonae* (IIIa), *diarizonae* (IIIb), *houtenae* (IV) and *indica* (VI) ([@ref-30]). Subspecies I, *S. enterica* subsp. *enterica* cause 99% of human and animal infections. The two main pathologies associated with *S. enterica* are gastroenteritis and typhoidal disease. The typhoidal *Salmonella*e include *S*. Typhi and *S*. Paratyphi A, B and C. They are host restricted, monophyletic, rarely undergo recombination events and exhibit convergent evolution driven by genome degradation ([@ref-31]). The majority of gastroenteritis in the UK is caused by the host generalist serovars, such as *S*. Typhimurium and *S*. Enteritidis, and host adapted serovars that are adapted to a specific animal reservoir but can infect man and include *S*. Dublin, *S*. Gallinarum *S*. Choleraesuis, and *S*. Bovismorbificans ([@ref-17]). Approximately 8,000 isolates are referred to the *Salmonella* Reference Service (SRS) at Public Health England (PHE) each year from local and regional hospital laboratories. In April 2015, PHE implemented whole genome sequencing (WGS) as the routine typing tool for public health surveillance of *Salmonella* infections. Prior to April 2015, presumptive *Salmonella* isolates referred to SRS were speciated and sub-speciated using PCR ([@ref-13]; [@ref-12]) and grouped into serovars as described in the White-Kauffman-Le Minor scheme ([@ref-9]; [@ref-10]; [@ref-15]). This methodology is based on reactions of rabbit antisera to the lipopolysaccharide (O antigen encoded by *rfb* genes) and flagellar antigens (phases 1 and 2 of H antigen encoded by *fli*C and *flj*B). The scheme utilises this phenotypic variation, expressed as an antigenic formulae, to divide *Salmonella* into more than 2,600 serovars. Epidemiological investigations of *Salmonella* infecting humans and animals have relied on serotyping for over 70 years; national and international governmental agencies base guidelines and regulations on the serotyping method and the use of this nomenclature is a globally recognised form of communication ([@ref-26]; [@ref-7]). Furthermore, serovars have often been shown to correlate with host range and disease sequelae ([@ref-8]; [@ref-31]; [@ref-17]). There are, however, a number of issues with the serotyping approach; specifically, the expense and expertise required to produce the antisera and, furthermore, serotyping does not reflect the genetic relatedness between serovars, nor does it provide an evolutionary perspective. Alternative molecular serotyping methods have been described previously including Pulsed-field gel electrophoresis, ribotyping, repetitive extragenic palindromic sequence-based PCR (rep-PCR) and combined PCR- and sequencing-based approach that directly targets O- and H-antigen-encoding genes ([@ref-22]; [@ref-24]). In 2012, [@ref-1] proposed a sequenced based approach, multilocus sequence typing (MLST), based on the sequences of multiple house-keeping genes. Isolates that possess identical alleles for the seven gene fragments analysed are assigned a common sequence type (ST) and related STs from clonal complexes are termed e-Burst Groups (eBGs). They showed that ST and eBGs strongly correlated with serovar and so utilising this approach would facilitate backward compatibility with historical data, minimise disruption for reference laboratory service users and facilitate data exchange with other colleagues in the field. Advances in whole genome sequencing (WGS) methodologies have resulted in the ability to perform high throughput sequencing of bacterial genomes at low cost making WGS an economically viable alternative to traditional typing methods for public health surveillance and outbreak detection ([@ref-16]). Whilst WGS provides the opportunity to resolve bacterial strains to the single nucleotide resolution needed for identifying cases linked to a common source of infection ([@ref-4]), grouping isolates into higher taxonomical clones (e.g., those defined by serotyping) is an important step. The decision to adopt WGS as a routine typing method at PHE provided the opportunity to review our approach to typing *Salmonella* and to implement the MLST approach in parallel with WGS. The aim of this study was to evaluate MLST, as derived from WGS data, as a replacement for conventional serotyping of *Salmonella* for routine public health surveillance and to provide insight into the genetic population structure of all *Salmonella* species in England and Wales during a 12 month period. Methods ======= Bacterial strains ----------------- All isolates (*n* = 7,465) of *Salmonella* from human cases of gastrointestinal disease submitted to SRS from local and regional hospital laboratories in England & Wales between 1st April 2014 and 31st March 2015 were sequenced in parallel with phenotypic serotyping ([Table S1](#supp-1){ref-type="supplementary-material"}). Of these, 7,338 were identified as subspecies I and included 263 different serovars. The ten most common serovars in this dataset were Enteriditis (2,310), Typhimurium (1,407), Infantis (184), Typhi (184), Newport (173), Virchow (162), Kentucky (160), Stanley (146), Paratyphi A (135) and Java (99). One hundred and twenty seven isolates were identified as subspecies II--IV (*S. enterica* subspecies *salamae n* = 28; *S. enterica* subspecies *arizonae n* = 25; *S. enterica* subspecies *diarizonae n* = 49; *S. enterica* subspecies *houtenae n* = 20) and there was one isolate of *S. bongori*. No isolates belonging to subspecies VI (*S. enterica* subspecies *indica*) were submitted to SRS during the study period. DNA extraction for WGS ---------------------- DNA extraction of *Salmonella* isolates was carried out using a modified protocol of the Qiasymphony DSP DNA midi kit (Qiagen). In brief, 0.7 ml of overnight *Salmonella* culture in a 96 deep well plate was harvested. Bacterial cells were pre-lysed in 220 µl of ATL buffer (Qiagen) and 20 µl Proteinase K (Qiagen), and incubated shaking for 30 min at 56 °C. Four µl of RNase at 100 mg/ml (Qiagen) was added to the lysed cells and re-incubated for a further 15 min at 37 °C. This step increases the purity of the DNA for further downstream sequencing. Extraction of DNA from the treated cells was performed on the Qiasymphony SP platform (Qiagen) and eluted in 100 µl of water. DNA concentration using the GloMax system (Promega) was determined for the following sequencing steps. DNA sequencing -------------- Extracted DNA was then processed using the NexteraXT sample preparation method and sequenced with a standard 2x101 base protocol on a HiSeq 2500 Instrument in fast mode (Illumina, San Diego, CA, USA). Bioinformatics workflow ----------------------- FASTQ reads were quality trimmed using Trimomatic ([@ref-3]) with bases removed from the trailing end that fell below a PHRED score of 30. If the read length post trimming was less than 50 bp the read and its pair were discarded. The PHE KmerID pipeline (<https://github.com/phe-bioinformatics/kmerid>) was used to compare the sequenced reads with 1,769 published genomes to identify the bacterial species (and *Salmonella* subspecies) and to detect cultures submitted by the local and regional hospital laboratories that contained more than one bacterial species (mixed cultures). KmerID determines a similarity index between the FASTQ reads and each of the 1,769 published reference genomes by calculating the percentage of 18-mers in the reference that are also present in the FASTQs. Only 18-mers that occur at least twice in the FASTQ are considered present. Mixed cultures are detected by comparing the list of similarities between the sample and the references with the similarities of the references to each other, and filtering this comparison for inconsistencies. ST assignment was performed using the Metric Orientated Sequence Typer (MOST), a modified version of SRST ([@ref-14]), available from <https://github.com/phe-bioinformatics/MOST>. The primary difference between SRST and MOST is in the metrics provided around the result, while SRST gives a single score, MOST provides a larger array of metrics to give users more details on the read level associated with their result. Preliminary analysis was undertaken using the MLST database described in [@ref-1]. It takes approximately 10--15 min to run MOST using a single core on the PHE infrastructure which consists of Intel Xeon CPU E5-2680 0@ 2.70GHz, 16 cores sharing 125 Gb Memory. For isolates that had novel STs, or a ST but no associated serovar in the Achtman MLST database, the serovar was determined by phenotypic serotyping at PHE. STs and corresponding serovars of isolates serotyped and sequenced during this study were added to a modified version of the Achtman MLST database, held and curated at PHE. These novel STs were assigned a preliminary ST (PST) and an inferred serovar was determined. The PHE MLST database currently holds 7,000 strains and 1,200 serovars and is up-dated every three months. For some STs that contained two serotypes, whole genome SNP phylogenetic analysis was carried out by mapping the strains of interest against a reference genome from within the same sequence type (for ST909 H145100685 was used; for ST49, H143720759 was used), using BWA mem ([@ref-18]). SNPs were called using GATK2 ([@ref-5]) in unified genotyper mode. Core genome positions that had a high quality SNP (\>90% consensus, minimum depth 10×, GQ ≥ 30, MQ ≥ 30) in at least one strain were extracted and RAxML v8.1.17 phylogenies determined with the gamma model of rate heterogeneity and 100 bootstraps undertaken. Results ======= [@ref-1] described the population structure of *Salmonella enterica* as monophyletic lineages of STs that have evolved from a single founder node and termed these discrete clusters eBGs. The population structure of all the *Salmonella* species submitted to PHE between April 2014 and March 2015 is illustrated by the minimum spanning tree in [Fig. 1](#fig-1){ref-type="fig"}. {ref-type="supplementary-material"} for details).](peerj-04-1752-g001){#fig-1} *Salmonella* subspecies I ------------------------- The ST and corresponding serovar designation obtained from the MLST database were used to compare the WGS derived ST to the phenotypic serotype for 6,887 (94%) of 7,338 isolates of subspecies 1, and of these, 6,616 (96%) had the same result by both methods ([Table S1](#supp-1){ref-type="supplementary-material"}). It was not possible to compare phenotypic serotyping with MLST-based serotyping for 451 (6%) subspecies I isolates because either the phenotypic serotype could not be determined due to an incomplete antigenic structure (*S*. Unnamed) (*n* = 423) or the serovar could not be determined because the ST did not have a designated serotype in the MLST database (*n* = 70). Forty-two isolates were both *S*. Unnamed and had no MLST designated serotype. For the 423 (5.8%) subspecies I isolates reported as *S*. Unnamed, 318 (90%) were designated a serotype from the WGS derived MLST data. The most common serovars typed in this way included *S*. Typhimurium (118), *S*. Virchow (30), S. Stanley (17), *S*. Enteritidis (16), *S*. Infantis (14) and *S*. Thompson (13). Of the 7,338 strains tested, 70 (1%) had no serotype designation in the MLST database, of which 28 (40%) were serotyped phenotypically ([Table S1](#supp-1){ref-type="supplementary-material"}). Subspecies I novel sequence types --------------------------------- Novel sequence types that were not present in the MLST database were identified in 8.6% (*n* = 654) of the strains ([Table S1](#supp-1){ref-type="supplementary-material"}). These 654 isolates belonged to a total of 326 novel STs, designated PST; the modal number of isolates identified per PST was one ([Fig. 2A](#fig-2){ref-type="fig"}). There was no difference in the distribution of number of isolates per PST depending on whether the PST had a known serovar or belonged to an unnamed or ambiguous serotype. The rate at which PSTs were received throughout the year was plotted and revealed a linear relationship (*R*^2^ = 0.98, *y* = 1.04∗*x*, where *x* = number of days since April 1st 2014) ([Fig. 2B](#fig-2){ref-type="fig"}). {#fig-2} The serovars with the highest number of new PSTs were *S*. Typhimurium (*n* = 9), *S*. Stanley (*n* = 9), *S*. Enteritidis (*n* = 9) and *S*. Newport (*n* = 8), although the majority of these PSTs were single locus variants (SLVs) of established STs, belonging to these serovars (*S*. Typhimurium 8/9, *S*. Stanley 7/9 and *S*. Newport 7/8). There were also serovars for which a large number of PSTs were identified that were not SLVs of established STs (*S*. Agama 5/5, *S*. Agbeni 5/5, *S*. Saint-Paul 5/5, *S*. Enteritidis 4/9) which may represent new eBGs that share these serotypes. Subspecies I mismatches ----------------------- Four percent (*n* = 271) of the isolates tested exhibited a mismatch between the WGS MLST derived serovar and the phenotypic serotyping results. Of the 271 mismatches, 119 were due to a process error in the laboratory either in the phenotypic serotyping or the DNA extraction part of the WGS pipeline. With respect to the phenotypic serotyping, common errors included mislabelling samples and misinterpreting or incorrectly transcribing the antigenic structure, especially when the antigenic structures were similar. For example, *S*. Agona (I 4, 12:f, g, s:-) and *S*. Derby (I 4, 12:f, g:-). DNA extraction errors were associated with mislabelled samples. Twenty-six mismatches were potentially caused by the predicted serotype designation in the Achtman MLST database being incorrect which may be attributed to single entries that had been misidentified at the laboratory from which the MLST data was submitted. For example, in the original database ST1499 is represented by one entry identified by the submitter as *S*. Litchfield. Subsequently, phenotypic serotyping at PHE identified this ST as *S*. Bovis-morbificans in more than five isolates. ST1499 belongs to eBG34 which comprises two other STs both associated with *S*. Bovis-morbificans, indicating that the original entry in the MLST database is likely to be incorrect. The most common reason for mismatches occurring between the WGS MLST derived serotype and the phenotypic serotype (*n* = 126) occurred when two different serovars belonged to the same eBG and the same ST (see [Table 1](#table-1){ref-type="table"} and discussed in more detail below). 10.7717/peerj.1752/table-1 ###### Examples where two serovars belonged to the same eBG and the same ST.  Serotype Antigenic structure ST ------------- ---------------------- ------- Bareilly I 6,7: y: 1,2 909 Richmond I 6,7: y: 1,5 Saintpaul I 4,5,12: e,h: 1,2 49 Haifa I 4,5,12: z,10: 1,2 Sandiego I 4,12: I,v: e,n,z15 20 Brandenburg I 4,12: e,h: e,n,z15 Uganda I 3, 10: I,z13: 1,5 684 Sinstorf I 3, 10: I,v: 1,5 Agona I 4,12: f,g,s:- 13 Essen I 4,12: f,g,m:- Napoli I 1,9,12: I,z13: enx P3141 Zaiman I 1,9,12: I,v: enx For 157 of the 271 mismatches, assembly based ST calls were available from Enterobase (<http://enterobase.warwick.ac.uk/>). We compared the STs called by our mapping pipeline and the assembly based Enterobase pipeline and the same ST was called in 100% of samples. Serovars Enteritidis and Dublin ------------------------------- Of the 2,308 isolates of *S*. Enteritidis identified by both phenotypic serotyping and WGS MLST derived serotyping, 2,296 belonged to eBG4, including 2,200 ST11 and 76 ST183 ([Fig. 1](#fig-1){ref-type="fig"}). There were five additional SLVs of ST11, four of which were novel types. *S*. Gallinarum and *S*. Pulloram can be difficult to distinguish from *S*. Enteritidis ([@ref-29]) but neither of these serovars were identified in this study. Serologically, *S*. Dublin (\[1\],9,12:g,p:-) has a similar antigenic structure to *S*. Enteritidis (\[1\],9,12:g,m:-), and in [@ref-1], eBG32 (ST74) contained both *S*. Enteritidis and *S*. Dublin. However, in this study both isolates belonging to ST74 eBG32 typed as *S*. Enteriditis. Of the 2,308 isolates, 26 belonged to nine new PSTs. The most common was P3147, a previously undescribed SLV of ST11, comprising 16 cases including 10 known to have travelled to Malaysia or Singapore. Serovar Typhimurium ------------------- In this study, eBG1 contained 1,392 isolates of *S*. Typhimurium and monophasic *S*. Typhimurium (rough and non-motile variants) ([@ref-11]). The monophasic variants also belong to eBG138 (primarily ST 36) and eBG243. In contrast to eBG1 described in [@ref-1], which was represented by a large central ST19 node with at least 27 SLV STs comprising much smaller numbers of strains, eBG1 in the PHE dataset shows a predominance of both ST19 and ST34 and less allelic variation. Only nine SLVs to ST19 were identified including three undesignated STs ([Fig. 1](#fig-1){ref-type="fig"}). Serovars Java/Paratyphi B data ------------------------------ Despite the different disease outcomes associated with *S*. Paratyphi B (most commonly associated with invasive disease and paratyphoid fever) and *S*. Java (most commonly associated with gastroenteritis) it is not possible to differentiate the two serotypes by serotyping alone. *S*. Java and *S*. Paratyphi B are therefore differentiated in the laboratory by their ability to ferment dextrorotatory tartrate (*S*. Java*d*Ta + and *S*. Paratyphi B*d*Ta−) ([@ref-20]). The 99 isolates identified by both phenotypic serotyping and WGS MLST derived serotyping as *S*. Java, belonged to a diverse range of eBGs, STs and PSTs ([Table 2](#table-2){ref-type="table"} and [Fig. 1](#fig-1){ref-type="fig"}). Two of these 99 isolates belonged to ST86 and the predicted serotype from the MLST database was *S*. Paratyphi B. One of these isolates was from a blood culture (associated with invasive disease) and, therefore, likely to have been misidentified phenotypically. All 12 isolates identified as *S*. Paratyphi B phenotypically, were identified as *S*. Paratyphi B ST86 by WGS MLST. 10.7717/peerj.1752/table-2 ###### Differences in ST between S. Java and S. Paratyphi B. *S*. Java isolates in this study belonged to a diverse range of eBGs and STs associated with *S*. Java whereas *S*. Paratyphi B belonged to ST86 only.  eBG5 eBG 9 eBG59 eBG32 eBG95 ------------- ------ ------- ------- ------- ------- ---- --- --- --- --- Java 45 7 4 3 2 18 6 5 2 1 Paratyphi B 0 0 0 0 12 0 0 0 0 0 Subspecies II--IV and *S. bongori* ---------------------------------- Isolates from subspecies II, III, IV, VI and *S. bongori* were not well represented in Achtman's MLST database and thus the majority of isolates from these sub-species sequenced in this study did not belong to a previously designated eBG or ST. The population structure of the 127 non-subspecies I isolates differs markedly from that of *Salmonella enterica* (subspecies 1) ([Fig. 1](#fig-1){ref-type="fig"}) and shows some similarity to the population structure of lineage 3 in being a connected network of STs. Sixteen of the 28 isolates belonging to subspecies II were previously designated *S*. Unnamed and the 28 strains belonged to 20 different STs. There were 25 isolates classed as subspecies IIIa (belonging to 10 different STs) and 49 in subspecies IIIb (belonging to 27 different STs). Of the 20 isolates identified as subspecies IV, 10 were designated *S*. Wassenaar (P3029) by phenotypic serotyping and the 20 isolates belonged to five different STs. All isolates of subspecies II--IV and *S. bongori* were correctly speciated using the k-mer ID approach. Population structure -------------------- As highlighted by [@ref-1], the majority of isolates in the dataset belong to eBGs that have a one-to-one relationship with a specific serovar including *S*. Typhi, *S*. Paratyphi A and *S*. Heidelberg. In this study, of the serovars comprising more than 25 isolates, there were 17 serovar specific eBGs and 10 examples of a single serovar being associated with multiple eBGs ([Fig. 1](#fig-1){ref-type="fig"}). There were at least six examples of more than one serovar belonging to the same eBG but different STs, for example *S*. Hadar (ST33) and *S*. Kottbus (ST582) both belong to eBG22 and *S*. Bredeney (ST306) and *S*. Schwarzengrund (ST96) both belong to eBG33 ([Table S1](#supp-1){ref-type="supplementary-material"}). There were seven examples where two serovars belonged to the same eBG and the same ST ([Table 1](#table-1){ref-type="table"}). In all of these examples, the antigenic structures of the two serovars were similar with only one antigen differentiating the two serovars. Further analysis was carried out on two examples to determine whether this difference in antigenic structure represented a true difference in strain relatedness or a random change that is not reflected in phylogeny (for example, the insertion of phage encoded antigen). The analysis showed that the change in antigenic structure in *S*. Richmond (I 6,7: y: 1,2) and *S*. Bareilly (I 6, 7: y: 1,5), both ST 909, and in *S*. Saintpaul (I 4,5,12: e,h: 1,2) and *S*. Haifa (I 4,5,12: z,10: 1,2), both ST49, reflected a true phylogenetic difference ([Figs. 3A](#fig-3){ref-type="fig"} and [3B](#fig-3){ref-type="fig"}). {#fig-3} The same higher strata population structure referred to as lineage 3 for *S. enterica* subspecies I, as described by [@ref-1], was observed in this dataset ([Fig. 4](#fig-4){ref-type="fig"}). Genomes of these *Salmonella*e are in constant flux and homologous recombination among unrelated eBGs is frequent ([@ref-1]; [@ref-6]). Serovars in this lineage mainly consists of multiple eBGs and are polyphyletic by nature. [@ref-1] suggested that the population structure of lineage 3 does not comprise of independent startbursts, as observed with other serovars of subspecies I, but rather a connected network ([Fig. 4](#fig-4){ref-type="fig"}). The five most common examples of this in the current study, were *S*. Oranienburg, *S*. Montevideo, *S*. Chester, *S*. Poona and *S*. Bredeney ([Fig. 4](#fig-4){ref-type="fig"} and [Table S1](#supp-1){ref-type="supplementary-material"}). These five serovars are not represented in the top 10 serovars submitted to SRS during this surveillance period. {#fig-4} K-mer identification -------------------- There were 249 cultures submitted to SRS by the local hospital and regional laboratories for *Salmonella* typing that were a mix of *Salmonella* and non-*Salmonella* species. These were identified by the k-mer identification step and included 138 *Escherichia coli*, 40 *Morganella morganii*, 11 *Citrobacter species* and four *Escherichia albertii*. Discussion ========== In their seminal 2012 paper Achtman and colleagues ([@ref-1]) argued convincingly for replacing serotyping with a MLST approach based on genetic population groupings for typing *S. enterica*. The key aspects of this approach that led PHE to adopt this strategy were (i) the robustness of the population structure as defined by the natural eBG clusters (ii) the fact that eBG designation provides an accurate representation of strain relatedness and (iii) that this approach lends itself to automation. At the same time, it was necessary for PHE to maintain serovar nomenclature in order to facilitate data exchange with other colleagues in the field and maintain backward compatibility with historical data. It was suggested that by using the MLST approach to infer serovar, and by reporting both inferred serovar and ST, it would be possible to utilise the advantages of both methods and implement a state-of-the-art typing system while keeping disruption for reference laboratory service users to a minimum. The PHE dataset of 6,887 subspecies I isolates that were serotyped using both traditional phenotypic methods and a derived serotype based on MLST data extracted from the genome during a 12-month time frame, provided further evidence of the robustness of the ST/eBG approach to typing. The 96% concordance between the two techniques in a reference laboratory setting is evidence of the validity and suitability of this approach. There were 451 isolates that had to be excluded from the comparison because both types of data (phenotypic and genotypic) were not available. Of these, for 94% of the isolates, it was the phenotypic serotype that could not be determined indicating that WGS MLST derived serotyping is more robust. The PHE dataset included single serovars associated with multiple eBG, for example *S*. Typhimurium and *S*. Newport ([@ref-23]; [@ref-1]) and multiple serovars belonging to the same eBG but with different STs, for example *S*. Java (ST43) and *S*. Paratyphi (ST86) both belong to eBG5 ([@ref-1]). In both these scenarios, the correct serovar was determined from the MLST WGS data and the combination of serovar and ST/eBG provided insight into the true phylogenetic relationship between isolates. This data clearly supports Achtman and colleagues argument that eBG and ST designation provides a more accurate representation of strain relatedness than the traditional serovar designation. The phenomenon of multiple serovars belonging to the same ST (for example *S*. Richmond/*S*. Bareilly and *S*. Haifa/*S*. Saintpaul) was a rare but important example of serotyping providing a higher level of strain discrimination within a ST. These strains could be differentiated *in silico* using a tool to infer serovar from the genes that determine antigenic struture, such as seqsero ([@ref-32]). Despite the implementation of WGS, a limited phenotypic serotyping facility continues to be maintained at PHE in order to serotype isolates that cannot be matched to a serovar; either because the ST in the MLST database has no serovar designation or the ST is a novel type. Additionally, it ensures that we maintain the ability to perform the standard reference method for serotyping *Salmonella*. The PHE MLST database is regularly up-dated to include STs recently matched to a serotype by linking the ST to PHE phenotypic serotyping data and novel PSTs. This approach was adopted because at the time of analysis, the Achtman MLST database was not accepting submissions generated by WGS. There was no decrease in the rate at which PSTs were observed during the 12 month study period and the majority of PSTs were only sampled once in that time frame. Many PSTs were SLVs of known STs, indicating that we have not yet sampled the full diversity of known eBGs. New PSTs, not part of any previously identified eBG, were also observed and further diversity was found within *S. enterica* subspecies II--IV and the lineage 3 population. This suggests that there is a large amount of previously unidentified diversity within the species *Salmonellae* associated with both domestically acquired and travel related gastrointestinal disease in human cases resident in England and Wales. Isolates exhibiting monophasic properties that could not be fully serotyped phenotypically because they had an incomplete antigenic structure were matched to a ST derived serotype. The monophasic variants in this study mainly belonged to eBG1, eBG138 and eBG243 and previous studies have also shown that monophasic variants of *S*. Typhimurium have emerged as a result of multiple independent genetic events ([@ref-25]; [@ref-27]; [@ref-28]). Strains with monophasic properties are reportable to European Centre for Disease Prevention and Control (ECDC) but cannot be determined using the ST approach. Alternative strategies for determining monophasic characteristics by PCR are available ([@ref-21]) and methods for extracting this information from the genome sequencing data have been developed at PHE (P Ashton & A Lewis, pers. comm., 2015). In contrast to *S*. Typhurmurium, where ST could not be used to determine monophasic characteristics, in this study ST was able to differentiate the complex relationship between *S*. Java (Hazard Group (HG) 2 organism) and *S*. Paratyphi B (HG3) with the latter belonging to either ST42 or ST86. If this ST designation proves to be robust, MLST will facilitate the diagnosis of invasive disease and life threatening paratyphoid fever. The MLST derived serovar correlated well with the traditional serovar designation and demonstrated many advantages over traditional phenotypic serotyping. Monophasic strains with incomplete antigenic structures were accurately assigned to serotypes. Phenotypic serotyping errors, such as misinterpreting or incorrectly transcribing the antigenic structure, were avoided. Novel types were identified, confirmed and given a PST designation. Finally, this approach lends itself to automation and rapid, high-throughput processing. Two main issues arose during the evaluation of the MLST approach: (i) a number of STs did not have a serovar designation in the MLST database (including subspecies II--IV) and (ii) the unexpectedly large number of novel STs identified. Traditional phenotypic serotyping was required to type these isolates and the MLST database was modified and up-dated to incorporate the new data. Clearly, as we move forward the PHE MLST database will be constantly evolving and this data will be shared with colleagues in the field via existing MLST databases and their WGS compliant successors e.g., EnteroBase & BIGSdb. While it is difficult to draw conclusions based on our small sample size, MLST may not currently be an appropriate tool for the classification of *Salmonella* sub-species II--IV, due to the lack of a discrete population structure of EBGs. However, non-subspecies I isolates which are mainly adapted to cold blooded animals and/or reptiles contributed to less than 1.7% of the workload during the time frame of the study. Although MLST approach is generally more discriminatory than serotyping, it does not always provide the fine resolution required for public health surveillance. Further analysis based on single nucleotide polymorphisms in the core genome compared to a type strain representing the most common eBGs is performed for outbreak detection and investigation ([@ref-2]). In conclusion, serotyping inferred from 7-gene MLST results derived from WGS data is an accurate, robust, reliable, high throughput typing method that is well suited to routine public health surveillance of *Salmonella*. This approach supports the maintenance of traditional serovar nomenclature and provides further insight on the true evolutionary relationship between isolates, as well as a framework for fine level typing within eBGs for surveillance, outbreak detection and source attribution. Supplemental Information ======================== 10.7717/peerj.1752/supp-1 ###### Supplementary Table ###### Click here for additional data file. We would like to thank all the members of the *Salmonella Whole Genome Sequencing Implementation Group* including Steve Connell, Anna Lewis, Andy Levy, Clare Maguire, Clare Wend-Hansen, Martin Day, James Rogers, Siham Ibrahim, Arlene Barcenilla, Vineet Patel, Kiran Jayan, Anthony Underwood, Catherine Arnold and Ian Harrison. Additional Information and Declarations ======================================= The authors declare there are no competing interests [Philip M. Ashton](#author-1){ref-type="contrib"} and [Satheesh Nair](#author-2){ref-type="contrib"} conceived and designed the experiments, performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, wrote the paper, prepared figures and/or tables, reviewed drafts of the paper. [Tansy M. Peters](#author-3){ref-type="contrib"} conceived and designed the experiments, performed the experiments, reviewed drafts of the paper. [Janet A. Bale](#author-4){ref-type="contrib"} conceived and designed the experiments, performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, reviewed drafts of the paper. [David G. Powell](#author-5){ref-type="contrib"} analyzed the data, contributed reagents/materials/analysis tools, reviewed drafts of the paper. [Anaïs Painset](#author-6){ref-type="contrib"} analyzed the data, contributed reagents/materials/analysis tools, prepared figures and/or tables, reviewed drafts of the paper. [Rediat Tewolde](#author-7){ref-type="contrib"} conceived and designed the experiments, performed the experiments, contributed reagents/materials/analysis tools, reviewed drafts of the paper. [Ulf Schaefer](#author-8){ref-type="contrib"} contributed reagents/materials/analysis tools, reviewed drafts of the paper. [Claire Jenkins](#author-9){ref-type="contrib"} analyzed the data, wrote the paper, prepared figures and/or tables, reviewed drafts of the paper. [Timothy J. Dallman](#author-10){ref-type="contrib"} conceived and designed the experiments, analyzed the data, contributed reagents/materials/analysis tools, wrote the paper, reviewed drafts of the paper. [Elizabeth M. de Pinna](#author-11){ref-type="contrib"} and [Kathie A. Grant](#author-12){ref-type="contrib"} conceived and designed the experiments, reviewed drafts of the paper. The following information was supplied regarding the deposition of DNA sequences: All data from the Salmonella surveillance project are deposited in the BioProject of the SRA: [PRJNA248792](http://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA248792). Per sample accessions are available in [Table S1](#supp-1){ref-type="supplementary-material"}. The following information was supplied regarding data availability: MOST code for MLST and KmerID code are available here: <https://github.com/phe-bioinformatics>.

Introduction ============ Metalloporphyrin π-cation radicals are of immense interest due to their occurrence as intermediates in the catalytic cycle of many heme containing enzymes, *e.g.*, peroxidases,[@cit1] catalases,[@cit2] cytochrome P450 [@cit1a],[@cit3] *etc.*, and their photosynthetic reaction center.[@cit4] In spite of the common active intermediates, the reactivity differs from enzyme to enzyme. The extent of coupling between metal center and porphyrin π-cation radical may be linked to the various activities in the heme enzymes, and this has therefore led to the study of metalloporphyrin π-cation radicals and mixed valence π-cation radicals (where one electron is removed per two porphyrin rings).[@cit5],[@cit6] Numerous diheme enzymes such as MauG[@cit7],[@cit8] and bacterial diheme cytochrome c peroxidases (bCcP)[@cit9] serve as active catalysts in various important chemical transformations in biology. For instance, bCcP mediates peroxidase activity, whereby it transfers the oxidizing equivalents from H~2~O~2~ to cytochrome c or other small redox proteins.[@cit9] MauG ([Fig. 1](#fig1){ref-type="fig"}) is a terminal enzyme involved in the biosynthesis of the catalytic tryptophan tryptophylquinone (TTQ) cofactor of methylamine dehydrogenase (MADH). Although the two heme units are physically separated in both enzymes, they share electrons efficiently behaving as a single diheme unit rather than as independent heme centers. A tryptophan residue is, however, positioned in between two heme centers and has been proposed to act as a bridge in order to promote the electronic communication between the heme centers.[@cit7]--[@cit9] ).[@cit7a]](c5sc03120f-f1){#fig1} High-valent Fe([iv]{.smallcaps})0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000O intermediates are frequently invoked in the catalytic cycles of Fe-dependent oxidizing enzymes.[@cit1]--[@cit3] In heme-dependent enzymes, the two-electron oxidized intermediate (compound I) consists of an Fe([iv]{.smallcaps}) species (*S* = 1) coupled to an organic radical (*S* = 1/2) that is located on the porphyrin ring/axial ligand or a nearby amino acid residue (compound ES). In MauG, however, two oxidizing equivalents derived from H~2~O~2~ are distributed within the diheme system as two positive charges, giving rise to a bis-Fe([iv]{.smallcaps}) redox state ([Scheme 1](#sch1){ref-type="fig"}) in which one heme is present as Fe([iv]{.smallcaps})0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000O and the other as Fe([iv]{.smallcaps}) with axial histidine and tyrosine ligation.[@cit8] As the two hemes are physically separated by 14.5 Å, a hole-hopping mechanism was proposed wherein the tryptophan residue reversibly oxidized and reduced in order to boost the effective electronic coupling element and magnify the rate of electron transfer between the heme centers in the bis-Fe([iv]{.smallcaps}) MauG.[@cit8] ![Conversion of the hemes of MauG from diferric to bis-Fe([iv]{.smallcaps}) redox state.[@cit8]](c5sc03120f-s1){#sch1} The electronic communication between the porphyrin moieties in the ground and/or excited state can be facilitated by using covalently connected linkers in conjugation with porphyrin rings. In this context, a large number of porphyrin dimers connected *via* conjugated linkers, *viz.* alkene, alkyne, imino, and azo bridges, have been studied due to their unique optical properties.[@cit10] Our recent efforts have been directed towards exploring the role of intermacrocyclic interactions in modulating various properties, *viz.* redox potential, spin state *etc.*[@cit11] However, interactions between a pair of porphyrin π-cation radicals connected through a linker remain unexplored. In the present study, we have made synthetic analogs of the diheme centers of MauG in which two porphyrin rings have been covalently connected through a conjugated but rigid ethylene bridge which would separate the two macrocycles to the furthest extent. Step-wise oxidations up to two oxidizing equivalents have been performed using chemical oxidants. Unlike the bis-Fe([iv]{.smallcaps}) state as obtained in MauG, the two electron oxidized complex stabilizes two ferric hemes, each coupled with a porphyrin cation radical, a scenario resembling the binuclear dication diradical complex. This work describes an alternative mechanism to store two oxidizing equivalents above the ferric state in dihemes. Spectroscopic investigations have unraveled strong electronic communications between two porphyrin π-cation radicals *via* the ethylene bridge which eventually alter the nature of such bridge. The extensive π-conjugation also allows antiferromagnetic coupling between iron([iii]{.smallcaps}) centers and porphyrin radical spins of both rings. DFT calculations have been employed which further support our spin-coupling model as obtained from the magnetic measurements and also provide insight into the preferential stabilization of various geometrical conformers under certain conditions. In earlier work on monomeric metalloporphyrin π-cation radicals, spin coupling between the metal ion and the oxidized porphyrin ring in a number of derivatives that differed in metal ion, axial ligation, and/or porphyrin ligand were investigated.[@cit5],[@cit6] The inter-ring coupling was found to be closely related only to the degree of the ring overlap. In the present investigation, however, the inter-ring coupling between two monomeric Fe([iii]{.smallcaps})porphyrin π-cation radicals has been demonstrated to occur only through the bridge although they are widely separated in space. Result and discussion ===================== Dichlorodiiron([iii]{.smallcaps}) ethylene bridged octaethyl porphyrin dimers have been synthesized both in *cis* (*cis*-**1**) and *trans* (*trans*-**1**) isomers using the procedures reported earlier.[@cit11j],[@cit12] Oxidation of these complexes has been performed in a step-wise manner using chemical oxidants and monitored using UV-vis-NIR spectroscopy. [Scheme 2](#sch2){ref-type="fig"} shows the synthetic outline and list of diheme dication diradical intermediates reported here along with the abbreviations used. {#sch2} [Fig. 2](#fig2){ref-type="fig"} shows the UV-visible spectral changes upon step-wise oxidations of *cis*-**1** using Fe(ClO~4~)~3~ as an oxidant. Gradual addition of a CH~3~CN solution of Fe(ClO~4~)~3~ as an oxidizing agent in up to one equivalent to a dichloromethane solution of *cis*-**1** leads to a sharp decrease in the Soret band intensity at 390 nm along with the appearance of a low energy band at 989 nm, which is characteristic of the intra-valence charge transfer obtained due to the formation of a mixed-valence π-cation radical dimer.[@cit6],[@cit13] The absorbance of the 989 nm band has showed a linear dependence on the concentration, ruling out the possibility of intermolecular dimerization as its origin.[@cit6] On further addition of the Fe(ClO~4~)~3~ solution in up to two equivalents, the Soret band at 390 nm decreases gradually again along with the appearance of two new bands at 326 and 477 nm, while the intensity of the absorbance at 989 nm decreases with the appearance of a new broad band near 1150 nm. The 2e-oxidized complex thus obtained has been isolated in the solid state as a dication diradical species (*trans*-**2**) in good yields and has been structurally characterized. Further addition of the oxidant showed no observable change in the absorption spectra. {#fig2} Similar spectral changes have also been obtained during the oxidation of *cis*-**1** when FeCl~3~ was used as an oxidant ([Fig. 3](#fig3){ref-type="fig"}). Upon one-electron oxidation, a sharp decrease in the Soret band intensity at 390 nm has been observed along with the appearance of a low energy band at 989 nm due to the formation of a mixed-valence π-cation radical dimer. Upon addition of the FeCl~3~ solution in up to four equivalents, the Soret band at 390 nm increased gradually and underwent a blue shift to 365 nm along with the appearance of two new bands at 317 and 488 nm in the UV-visible regions. In the NIR region, the intensity of the absorbance at 989 nm decreases while a broad band near 1210 nm appears. The two-electron oxidized complex thus obtained has also been isolated as a dication diradical species (*trans*-**3**) in the solid state and has been structurally characterized. The absorption spectrum of *trans*-**3** displays a Soret band at 365 nm, two new bands at 317 and 488 nm, and a broad NIR band near 1210 nm. *Trans*-**1** also produces similar spectral changes upon oxidation with Fe(ClO~4~)~3~/FeCl~3~ leading to the formation of *trans*-**2**/**3**, respectively (Fig. S1 and S2[†](#fn1){ref-type="fn"}). Gradual addition of an acetonitrile solution of Fe(ClO~4~)~3~ to a CH~2~Cl~2~ solution of *trans*-**1** at --40 °C (Fig. S3[†](#fn1){ref-type="fn"}) also showed similar spectral changes as observed at room temperature. {#fig3} [Fig. 4](#fig4){ref-type="fig"} shows the UV-visible spectral change upon gradual addition of a CH~3~CN solution of Fe(ClO~4~)~3~ into a CH~2~Cl~2~ solution of 1,2-bis\[(chloro){5-(2,3,7,8,12,13,17,18-octaethylporphyrinato)}iron([iii]{.smallcaps})\]ethane, **4** ([Chart 1](#cht1){ref-type="fig"}), which is a highly flexible ethane-bridged analog of **1**. Upon 1e-oxidation, the intensity of the Soret band decreases slightly along with a small blue-shift while a low-intense NIR band at 1100 nm is observed which, however, upon 2e-oxidation disappears completely. {#fig4} ![1,2-Bis\[(chloro){5-(2,3,7,8,12,13,17,18-octaethylporphyrinato)}iron([iii]{.smallcaps})\]ethane, **4**.](c5sc03120f-c1){#cht1} It is interesting to compare the UV-vis-NIR spectra of 2e-oxidized complexes reported in [Fig. 5](#fig5){ref-type="fig"}. The absorption bands in the 480--500 nm, 600--900 and 1100--1300 nm regions, which are observed in both *trans*-**2** and *trans*-**3**, are absent completely in the 2e-oxidized product of **4** and are, therefore, attributed to the extensive conjugation between two porphyrin π-cation radicals (*vide infra*) in the former complexes. The intensity of the Soret band has also been drastically reduced in *trans*-**2** and *trans*-**3** (as compared to the 2e-oxidized product of **4**) which is suggestive of reduced aromaticity of the porphyrin rings in the complexes because of conjugation (*vide infra*).[@cit10c] {#fig5} The broad NIR band that is observed at 1150 and 1210 nm for *trans*-**2** and *trans*-**3** can be attributed to the charge resonance (CR) stabilization of the spins and charges in the binuclear dication diradical complex. The CR stabilization energy (Δ*E*~CR~) originates from exchange interactions between the molecular orbitals of each monomer and subsequent delocalization of the spin and charge over a greater number of atoms.[@cit6b],[@cit8],[@cit13] During the 1e-oxidation process (*vide supra*), a broad and slightly more intense NIR band at ∼1000 nm was obtained which is also attributed to charge resonance phenomena of the mixed-valence π-cation radical dimer.[@cit6],[@cit8] Interestingly, CR-transition phenomena are observed here in both 1e and 2e-oxidized complexes from the same system while the NIR bands are also clearly distinguishable by their relative position and intensity. Despite the inherent energy penalty caused by the electrostatic repulsion in the dication complex, the CR stabilization energy of the dication diradical species is larger than that of the mixed-valence π-cation radical dimer of the same system, which results in a blue-shift of the CR band.[@cit6b],[@cit8],[@cit13] In contrast, a red-shift of CR bands is observed when moving our attention from 1e to 2e-oxidation, and this is due to multiple factors related to charge delocalization through bridging ligand, intramolecular coupling between iron--radical and radical--radical spins (*vide infra*) *etc.* Through the bridging ethylene group, both porphyrin macrocycles exhibit substantial conjugation which eventually alters the nature of the bridge (ethylene to *exo*-methylene connectivity) resulting the stabilization of two unusual *trans* conformers (*U* and *P*′ type, *vide infra*). Moreover, the NIR bands in the absorption spectra of *trans*-**2** and *trans*-**3** are found to be linearly correlated with the concentration ([Fig. 6](#fig6){ref-type="fig"} and S4[†](#fn1){ref-type="fn"}). Such Beer\'s law concentration dependence further supports the intramolecular origin of the CR bands.[@cit5],[@cit6] The MauG-catalyzed reaction proceeds through a bis-Fe([iv]{.smallcaps}) intermediate which has been shown to exhibit CR phenomena. The bis-Fe([iv]{.smallcaps}) state is an electronic equivalent of two Fe([iii]{.smallcaps})porphyrin π-cation radicals with each radical spin coupled to the Fe([iii]{.smallcaps}) atom. Although there is large separation between two heme centers, such spin/charge resonance stabilization is facilitated by the presence of a Trp93 residue which bridges between heme centers.[@cit8] {#fig6} The formation of porphyrin π-cation radicals is often signalled by strong IR bands.[@cit5],[@cit6] Fig. S5[†](#fn1){ref-type="fn"} displays selected portions of the IR spectra of *trans*-**2** and *trans*-**3** along with their unoxidized analogs, *cis*-**1** and *trans*-**1**. The oxidized species, *trans*-**2** and *trans*-**3**, show π-cation radical IR marker bands at 1546, 1598 and 1544, 1595 cm^--1^, which are absent in the unoxidized complexes. The ∼1550 cm^--1^ band is due to the asymmetric C~α~--C~meso~ stretching mode while the ∼1600 cm^--1^ band is due to the C~β~--C~β~ stretch. Similar IR marker bands have also been reported earlier in octaethylporphyrinate and β-alkyl substituted porphyrinate π-cation radical derivatives.[@cit5],[@cit6] Thus, the IR spectral data further confirms *trans*-**2** and *trans*-**3** as porphyrin π-cation radicals. The molar conductances at 295 K of *trans*-**2** (41 Ω^--1^ cm^2^ mol^--1^) and *trans*-**3** (42 Ω^--1^ cm^2^ mol^--1^) in dichloromethane are similar to the values observed for 1 : 2 electrolytes in solution.[@cit14] Crystallographic characterizations ---------------------------------- Dark needle crystals of *trans*-**2** and *trans*-**3** were grown *via* slow diffusion of hexane into chloroform and dichloromethane solutions of the complexes in air at room temperature. Perspective views of the complexes along with the molecular packing are depicted in [Fig. 7](#fig7){ref-type="fig"} and [8](#fig8){ref-type="fig"}. [Table 1](#tab1){ref-type="table"} lists the crystallographic data and data collection parameters of the complexes reported here. Both molecules crystallize in the monoclinic crystal system, however, *trans*-**2** does so with a *P*2/*c* space group while *trans*-**3** crystallizes with a *P*2~1~/*n* space group. Iron centers are in six-coordinate geometry with water and perchlorate axial ligands for *trans*-**2** and five-coordinate geometry with chloride ion as axial ligand in *trans*-**3**. The porphyrin cores of the complexes showed considerable doming. Table S1[†](#fn1){ref-type="fn"} shows the selected bond distances and angles while [Table 2](#tab2){ref-type="table"} compares the structure and geometrical parameters of *trans*-**2** and *trans*-**3** along with their unoxidized complex, *trans*-**1**. {#fig7} {#fig8} ###### Crystallographic data and data collection parameters *trans*-**2** *trans*-**3** --------------------------------------------------------------------- -------------------------------- ---------------------------- Formula C~78~H~96~Cl~16~Fe~2~N~8~O~18~ C~76~H~92~Cl~14~Fe~4~N~8~ *T* (K) 100(2) 100(2) Formula weight 2112.52 1837.27 Crystal system Monoclinic Monoclinic Space group *P*2/*c* *P*2~1~/*n* *a*, Å 31.467(5) 15.165(5) *b*, Å 9.589(5) 18.236(5) *c*, Å 30.770(5) 15.623(5) *α*, deg 90 90 *β*, deg 91.556(5) 103.044(5) *γ*, deg 90 90 *V*, Å^3^ 9281(5) 4209(2) *Z* 4 2 *d* ~calcd~, g cm^--3^ 1.512 1.450 *μ*, mm^--1^ 0.842 1.166 *F*(000) 4352 1892 Crystal size 0.26 × 0.20 × 0.16 mm^3^ 0.22 × 0.16 × 0.12 mm^3^ No. of unique data 17277 7814 Completeness to theta = 25.00° 99.9% 99.9% No. of parameters refined 1166 496 GOF on *F*^2^ 1.031 1.032 *R* ~1~ [^*a*^](#tab1fna){ref-type="table-fn"} \[*I* \> 2*σ*(*I*)\] 0.0679 0.0489 *R* ~1~ [^*a*^](#tab1fna){ref-type="table-fn"} (all data) 0.1108 0.0680 w*R*~2~[^*b*^](#tab1fnb){ref-type="table-fn"} (all data) 0.1791 0.1332 Largest diff. peak and hole 1.324 and --1.101 e Å^--3^ 0.910 and --0.745 e Å^--3^ ^*a*^ {#ugt1} ^*b*^ {#ugt2} ###### Selected structural parameters *trans*-**1**[^*a*^](#tab2fna){ref-type="table-fn"} *trans*-**2** *trans*-**3** ---------------------------------------------------- ----------------------------------------------------- --------------- --------------- ------------ Fe--N~p~ (Å)[^*b*^](#tab2fnb){ref-type="table-fn"} 2.055(6) 2.027(4) 2.034(4) 2.045(3) Fe--O(H~2~) (Å) --- 2.093(3) 2.095(3) --- Fe--O(ClO~3~) (Å) --- 2.187(3) 2.178(3) --- Fe--Cl (Å) 2.241(2) --- --- 2.2249(11) ΔFe24 (Å)[^*c*^](#tab2fnc){ref-type="table-fn"} 0.48 0.10 0.04 0.50 Δ~24~ (Å)[^*d*^](#tab2fnd){ref-type="table-fn"} 0.21 0.20 0.16 0.26 Fe···Fe (Å) 9.84 8.33 8.31 9.32 *Θ* (°)[^*e*^](#tab2fne){ref-type="table-fn"} 65.79 56.59 57.13 55.82 C20--C37 (Å) 1.478(9) 1.374(6) 1.371(6) 1.364(4) C37--C37A (Å) 1.312(13) 1.435(9) 1.433(9) 1.432(7) ^*a*^Taken from [@cit11h]. ^*b*^Average value. ^*c*^Displacement of iron from the least-squares plane of the C~20~N~4~ porphyrinato core. ^*d*^Average displacement of atoms from the least-squares plane of C~20~N~4~ porphyrinato core. ^*e*^Inter-planar angle between the least-squares plane of the C~20~N~4~ porphyrinato core and the *C*~4~ plane of the bridging ethylene group (see [Fig. 9A](#fig9){ref-type="fig"}). The average Fe--N~p~ distances in the two types of complexes are very different: 2.027(4) and 2.034(4) Å for cores I and II in *trans*-**2** and 2.045(3) Å in *trans*-**3** which are characteristic of admixed high-spin and high-spin states of iron.[@cit11],[@cit15],[@cit16] The Fe--Cl bond distance of 2.2249(11) Å in *trans*-**3** is shorter than the 2.241(2) Å observed in the unoxidized complex, *trans*-**1**. The slight contraction of the Fe--Cl bond distance is consistent with the increased positive charge in the oxidized complex. The displacement of iron([iii]{.smallcaps}) from the mean plane of the C~20~N~4~ porphyrinato core (Fe···Ct~p~) in *trans*-**3** is 0.50 Å which is slightly larger than the 0.48 Å observed in *trans*-**1**.[@cit11h] However, the Fe···Ct~p~ distances in the six-coordinate complex, *trans*-**2**, are 0.10 and 0.04 Å for cores I and II, respectively, and iron is displayed in the direction of the axial H~2~O ligand. Although the ligand field strength of ClO~4~^--^ is greater than for H~2~O according to the magnetochemical series,[@cit17] this is outweighed by the strong H-bonding interactions between the ligated H~2~O and ClO~4~^--^ counterions which hold the structure in place. The most striking feature of *trans*-**2** is the unusual bowl shaped conformation (*U*-form) of the bisporphyrin unit which is stabilized by four H-bonding interactions \[O···O: 2.761(5), 2.931(5), 2.926(5) and 2.767(5) Å\] between the coordinated H~2~O molecules and the ClO~4~^--^ counterions. The inter-planar angles (*Θ*) between the plane of the ethylene bridge and that of the porphyrin rings are 56.59° and 57.13° in *trans*-**2** ([Table 2](#tab2){ref-type="table"}) as compared to 65.79° obtained for *trans*-**1**. Although steric interactions with the β-ethyl substituents in *trans*-**1** tend to prevent a coplanar arrangement of the porphyrin rings and the ethylene bridge, the extensive conjugation in the dication diradical species (*trans*-**2**) imposes a better coplanarity. The extended conjugation between two porphyrin units through the bridging ethylene moiety is manifested in the alteration of the C20--C37 and C37--C37A bond distances. The C20--C37 bond lengths of 1.374(6) (molecule-I) and 1.371(6) Å (molecule-II) in *trans*-**2** are close to the C0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000C distance, while the C37--C37A bond lengths of 1.435(9) (molecule-I) and 1.433(9) Å (molecule-II) become close to that of a C--C single bond. This results in a situation where the ethylene bridge (C--CH0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000CH--C) transforms into an *exo*-methylene (C0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000CH--CH0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000C) connectivity between two porphyrin rings ([Fig. 9B](#fig9){ref-type="fig"}). {#fig9} The X-ray structure of *trans*-**3** is similar to that of *trans*-**1** in terms of relative orientation of the two porphyrin rings; however, they differ remarkably in several structural and geometrical parameters. The C20--C37 and C37--C37A bond distances of 1.364(4) and 1.432(7) Å and the inter-planar angle (*Θ*) of 55.82° in *trans*-**3** are similar to those of *trans*-**2**, which is indicative of strong intramolecular electronic conjugation between the two rings through the ethylene bridge. The alteration of the C20--C37 and C37--C37A bond distances leads to a single bond connecting two *exo*-methylene groups at C20 of the two porphyrin rings. [Scheme 3](#sch3){ref-type="fig"} describes the three possible conformations for the *trans* isomer. In one conformation, both porphyrins are completely orthogonal to the alkene (normal *P*-type), while in other two the macrocycles are almost in plane with the bridged ethylene moiety and the rings are either cofacial (*U*-type) or anti (*P*′-type) with respect to each other.[@cit10f],[@cit10g] The 2e-oxidized complexes, *trans*-**2** and *trans*-**3**, do not stabilize in the normal *P*-type conformation but in the unusual *U* and *P*′-conformations, which facilitates extended conjugation between the two radical cores. This is the first structural evidence of the *U* and *P*′ forms in β-alkyl substituted bisporphyrins. Although the *P*′-form has been previously observed in a few β-unsubstituted bisporphyrins,[@cit10c]--[@cit10e] no such conformers could be stabilized in β-alkyl substituted bisporphyrins due to steric interactions from the extreme closeness of the alkyl groups. ![Different conformations of *trans* ethylene-bridged Fe([iii]{.smallcaps})bisporphyrin framework.](c5sc03120f-s3){#sch3} Mössbauer --------- To gain further insight into the electronic structure, we studied both *trans*-**2** and *trans*-**3** by Mössbauer spectroscopy. [Fig. 10](#fig10){ref-type="fig"} shows the solid state Mössbauer spectra of the oxidized complexes at 295 K. The Mössbauer spectra provide compelling evidence that oxidation has occurred at the porphyrin ring only. *trans*-**2** shows the presence of two quadrupole-split doublets \[*δ* (Δ*E*~q~): 0.39 (2.0) and 0.33 (1.7) mm s^--1^\], suggesting the presence of two admixed high-spin Fe([iii]{.smallcaps}) centers[@cit11],[@cit15],[@cit16] in the molecule as also observed in the X-ray structure (*vide supra*). In contrast, *trans*-**3** displays two quadrupole-split doublets, one of which has parameters comparable to the tetrachloroferrate anion[@cit18] \[*δ* = 0.19 mm s^--1^, Δ*E*~q~ = 0.23 mm s^--1^\] while the other represents a high-spin iron([iii]{.smallcaps}) centre \[*δ* = 0.29 mm s^--1^, Δ*E*~q~ = 1.1 mm s^--1^\]. {#fig10} Magnetic measurements --------------------- The magnetic susceptibility of *trans*-**2** and *trans*-**3** has been measured between 5 to 300 K using two different applied magnetic fields of 0.1 and 1 T which provide very similar results. In an attempt to obtain a quantitative description of the spin--spin interaction that led to the observed magnetic moment, the magnetic susceptibility data were fitted using the software PHI.[@cit19] Acceptable fits ([Fig. 11](#fig11){ref-type="fig"}) have been obtained using a model which included three pairs of interactions: (a) intramolecular coupling between the iron and π-cation radical spin (*J*~Fe--r~), (b) intramolecular coupling between two iron spins (*J*~Fe--Fe~), and (c) intramolecular coupling between the two radical spins of the bisporphyrin (*J*~r--r~). Each iron([iii]{.smallcaps}) center was treated as high spin (*S* = 5/2) with a *g* value set at 2.0 and the presence of small residual mononuclear iron([iii]{.smallcaps}) impurities was also taken into consideration. The parameters obtained from the fits are: *J*~Fe--r~ = --124.36 cm^--1^, *J*~Fe--Fe~ = --0.05 cm^--1^, *J*~r--r~ = --16.67 cm^--1^, *D* = 4.0 cm^--1^ for *trans*-**2**, *J*~Fe--r~ = --149.88 cm^--1^, *J*~Fe--Fe~ = --0.13 cm^--1^, *J*~r--r~ = --8.41 cm^--1^, and *D* = 6.0 cm^--1^ for *trans*-**3**. The ruffling of the porphyrin ring allows strong antiferromagnetic coupling between the porphyrin π-cation radical and the iron center, which is otherwise forbidden with planar macrocycles.[@cit5d] The extensive π-conjugation between two porphyrin units of the bisporphyrin, through the bridging ethylene group, allows antiferromagnetic coupling between the two radical spins of the two porphyrin units although the cores are well separated and not cofacial enough to have any through space interactions. {#fig11} ^1^H NMR -------- [Fig. 12](#fig12){ref-type="fig"} compares the ^1^H NMR spectra between *cis*-**1** and *trans*-**2**. For *trans*-**3** the signals are too broad to be detected. The ^1^H NMR spectra of *cis*-**1** shows the presence of eight methylene proton resonances between 34 to 55 ppm, two upfield shifted *meso* signals at --68.9 and --46.2 ppm in 2 : 1 intensity ratio, and a highly downfield shifted bridging signal at 110 ppm. However, sixteen methylene and four *meso* signals were obtained upon two electron oxidation of *cis*-**1** to form *trans*-**2**, which confirms the presence of two inequivalent iron([iii]{.smallcaps}) centers as also evident from the X-ray structure and Mössbauer spectra of the complex in the solid state. The most conspicuous features in the ^1^H NMR spectra of *trans*-**2** are the wide range of methylene resonances (39.5 to 67.1 ppm) and highly downfield shifted *meso* proton signals (at 26.5, 32.4, 37.9 and 41.8 ppm) which are attributed to the presence of porphyrin π-cation radicals and also to the increase in coordination number (from five to six).[@cit20] Mulliken spin densities have also been calculated and reveal a negative spin density on the *meso* and bridging carbon atoms (Fig. S6[†](#fn1){ref-type="fn"}), hence the downfield shifts of the corresponding proton signals which also indicates the presence of the antiferromagnetic coupling of the iron spin with the *S* = 1/2 porphyrin radical as observed in the variable temperature magnetic measurements.[@cit20a],[@cit20b] Moreover, bridging protons are found at 78.6 and 88.3 ppm which are extremely broad due to the extensive delocalization of the radical through the bridging group as also evident in the alteration of the C20--C37 and C37--C37A bond lengths (*vide supra*). The temperature dependence of the proton signals follows the Curie law (Fig. S7[†](#fn1){ref-type="fn"}) which suggests the presence of a single spin state throughout the temperature range. {#fig12} EPR --- X-band EPR spectra have been recorded at 77 K for *trans*-**2** and *trans*-**3** both in solid and solution phases and [Fig. 13](#fig13){ref-type="fig"} displays the experimental and simulated spectra while Table S2[†](#fn1){ref-type="fn"} lists the simulation parameters. Simulation of the spectra of *trans*-**2** yields *g*~⊥~ = 5.470 and *g*~∥~ = 1.945 which is consistent with admixed high spin iron([iii]{.smallcaps}) as also observed from the Mössbauer data and X-ray structure of the complex in the solid state. Trace B shows the EPR spectra of *trans*-**3** which is composed of two distinct features: a single isotropic line (*g* = 2.016) characteristic of an unresolved ^6^S~1~ state indicating spin *S* = 1/2 and a +1/2 ↔ --1/2 transition, consistent with a \[FeCl~4~\]^--^ anion,[@cit18b] and a second minor component with *g*~⊥~ = 5.750 and *g*~∥~ = 1.942, indicative of a high-spin iron([iii]{.smallcaps}). Moreover, the EPR signals of the 2e-oxidised complexes reported here are weak due to relatively weak intramolecular coupling between iron--radical (*J*~Fe--r~) and radical--radical (*J*~r--r~) spins (*vide supra*), although stronger coupling would result in no observable signals. {#fig13} MauG is an enzyme unique in several respects. In heme-dependent enzymes, the two-electron oxidized intermediate (compound I) consists of an Fe([iv]{.smallcaps}) species coupled to an organic radical that is located on the porphyrin ring/axial ligand or a nearby amino acid residue (compound ES).[@cit1]--[@cit3] However, the two oxidizing equivalents derived from H~2~O~2~ are distributed within the diheme system as two positive charges, giving rise to the bis-Fe([iv]{.smallcaps}) redox state in MauG.[@cit8] This is electronically equivalent to two ferric hemes, each coupled with a porphyrin cation radical, a scenario resembling the binuclear dication diradical complex as observed in the 2e-oxidized complexes of *trans*-**2** and *trans*-**3** reported here. Although the spins and charges are delocalized throughout the diheme system and are stabilized *via* a series of possible charge resonance structures (responsible for the observed NIR band), the most stable form is a binuclear dication diradical complex which is found spectroscopically as the most stable conformer both in the solid state and in solution. Molecular modeling ------------------ A series of DFT calculations have been carried out to get more insights into the electronic structure of the oxidized species. *Trans*-**2** has been optimized in three possible spin multiplicities: nonet (9) state (considering antiferromagnetic interactions between the π-cation radical and the iron([iii]{.smallcaps}) spin), undectet (11) state (considering antiferromagnetic coupling between the two π-cation radical centers) and tridectet (13) state (considering ferromagnetic interactions between the cation radical and iron([iii]{.smallcaps}) spin). However, the interactions between the two iron([iii]{.smallcaps}) centers of *trans*-**2** in all these states have been considered to be ferromagnetic in nature. Since we are dealing with a large computational model, we have chosen to stick with a pure high-spin state (*S* = 5/2) for the iron([iii]{.smallcaps}) centers, although experimental data suggest a minor contribution from the intermediate spin (*S* = 3/2) in *trans*-**2**. Fig. S8 and S9[†](#fn1){ref-type="fn"} show the optimized geometries of *trans*-**2** with varying spin states in unrestricted B3LYP[@cit21] and B97D[@cit22] (includes dispersion correction) functional, while [Fig. 14](#fig14){ref-type="fig"} schematically represents their relative energies. Clearly, the nonet state is energetically more stable than the undectet state, and the tridectet state is the least stable. This is also in good agreement with the experimental results. {#fig14} In an attempt to find the probable reason behind such stabilization of the *U* form over the others, we have further optimized the molecule in different conformations, namely the *P*′ and *P* forms, both in the presence and absence of counterions. Fig. S10--S15[†](#fn1){ref-type="fn"} represent the optimized geometries while [Fig. 15](#fig15){ref-type="fig"} shows the relative energies of different conformations both in the presence and absence of counterions. For *trans*-**2**, the *U* form is more stable than the *P*′ and *P* forms by 16.34 kcal mol^--1^ and 30.27 kcal mol^--1^, respectively. However, in the absence of counterions, the *P*′ form becomes the most stable one, being 5.63 kcal mol^--1^ lower in energy than the *U*-form, while the *P*-form is similar (with a difference of --0.6 kcal mol^--1^) in energy to the *U*-form. This proves that H-bonding interactions between the axial H~2~O and the ClO~4~^--^ counterion are responsible for the unusual stabilization of the *U* conformer which is otherwise strained. This is also evident in the stabilization of the *P*′ conformer for *trans*-**3** where there is no H-bonding interaction with the counterion (FeCl~4~^--^). Both the HOMO and the LUMO ([Fig. 16](#fig16){ref-type="fig"} and S16[†](#fn1){ref-type="fn"}) of *trans*-**2** were found to have significant coefficients on the bridging ethylene group for the *U* and *P*′ forms, suggesting a substantial conjugation through the bridge, while the *P* form shows no such conjugation which is also in accord with the experiment. Although metallo-OEP radicals are known to have an a~1u~ type HOMO, *trans*-**2** is best described as a mixture of both a~1u~ and a~2u~ type orbitals ([Fig. 16](#fig16){ref-type="fig"}).[@cit23] While the positions of infrared marker bands of porphyrin π-cation radicals are characteristics of an a~1u~-type symmetry, the presence of a~2u~ type is clearly evident in the large shifts of the *meso* proton signals in the ^1^H NMR (*vide supra*). Thus, computational studies clearly support the experimental observations. {#fig15} {#fig16} Summary ======= We have reported here the structural and spectroscopic evidence for two interacting Fe([iii]{.smallcaps}) porphyrin π-cation radicals that are covalently connected through an ethylene bridge and thus are too widely separated to have any through space interaction. 2e-oxidation of the diiron([iii]{.smallcaps}) porphyrin dimer leads to the conversion of the *cis* isomer to *trans* along with the stabilization of two unusual conformations (*U* and *P*′-type) which have widely different and distinct spectroscopic and geometric features and are different from the normal *trans* conformer (*P*-type). Through the bridging ethylene group, both porphyrin macrocycles exhibit notable conjugation resulting in effective coupling between iron and porphyrin radical spins of both rings. In fact, oxidation leads to a change in identity of the bridge which further highlights the key role played by the bridge in communicating the electronic properties between two rings. The conformers can be considered as a real supramolecule rather than two interacting discrete Fe([iii]{.smallcaps}) porphyrin π-cation radicals. DFT calculations further support the spin coupling model obtained from magnetic studies suggesting the nonet state of *trans*-**2** as the energetically favorable one. It also revealed that the extended π-conjugation and H-bonding interactions are the major factors in controlling the stability of the unusual conformers in the oxidized complexes. Supplementary Material ====================== Supplementary information ###### Click here for additional data file. Crystal structure data ###### Click here for additional data file. This work is dedicated to Prof. Samaresh Mitra on the occasion of his 75th birthday. We thank IIT Kanpur for providing all the facilities and support. The Council of Scientific and Industrial Research (CSIR), New Delhi, and Science and Engineering Research Board (SERB), India are gratefully acknowledged for financial support. D. S. and S. D. thank University Grant Commission (UGC), India and A. K. and S. B. thank the Council of Scientific and Industrial Research (CSIR), New Delhi for their fellowship. [^1]: †Electronic supplementary information (ESI) available: Text, figures and tables depicting synthetic procedures, detailed experimental procedures, characterisation data including UV-vis-NIR, FT-IR, DFT. CCDC [1052187](1052187) and [1052188](1052188). For ESI and crystallographic data in CIF or other electronic format see DOI: [10.1039/c5sc03120f](10.1039/c5sc03120f)

J. Liu, D. Reta, J. A. Cleghorn, Y. X. Yeoh, F. Ortu, C. A. P. Goodwin, N. F. Chilton, D. P. Mills, *Chem. Eur. J.* **2019**, *25*, 7749. Introduction {#chem201901167-sec-0001} ============ Since the discovery of ferrocene \[Fe(Cp)~2~\] (Cp=cyclopentadienyl, C~5~H~5~) in the middle of the 20th century,[1](#chem201901167-bib-0001){ref-type="ref"} there has been an explosion of research in organometallic chemistry, leading to applications in some unexpected areas such as optical and redox devices, batteries, sensing, catalysis and medicine.[2](#chem201901167-bib-0002){ref-type="ref"} Following the landmark discovery of ferrocene, derivatized metallocenes \[M(Cp^R^)~2~\] (Cp^R^=substituted Cp) have been synthesised across the s‐, p‐, d‐ and f‐block elements.[3](#chem201901167-bib-0003){ref-type="ref"} The first examples of lanthanide (Ln) Cp complexes were synthesised in 1954,[4](#chem201901167-bib-0004){ref-type="ref"} and this area has grown to the extent that Ln organometallic chemistry is now dominated by Cp^R^ ligands.[5](#chem201901167-bib-0005){ref-type="ref"} Although isolated d‐block metallocenium cations have been known since ferrocenium, \[Fe(Cp)~2~\]^+^, was isolated in 1952[6](#chem201901167-bib-0006){ref-type="ref"} and Ln metallocenium cations, \[Ln(Cp)~2~\]^+^, were first predicted in 1956 by Birmingham and Wilkinson,[7](#chem201901167-bib-0007){ref-type="ref"} it was only recently with the report of the dysprosocenium complex \[Dy(Cp^ttt^)~2~\]\[B(C~6~F~5~)~4~\] (Cp^ttt^=C~5~H~2~ *t*Bu~3~‐1,2,4, **1**‐**Dy**),[8](#chem201901167-bib-0008){ref-type="ref"} and heavy Ln metallocenium homologues \[Ln(Cp^ttt^)~2~\]\[B(C~6~F~5~)~4~\] (**1**‐**Ln**, Ln=Y, Gd, Tb, Ho, Er, Tm, Yb, Lu),[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} that structurally authenticated Ln metallocenium complexes with no equatorial interactions were achieved. The paucity of isolated Ln metallocenium cations prior to 2017 can be attributed to the propensity for large electropositive Ln^III^ cations to maximise their coordination numbers, coupled with their preference for hard donor ligands not being well‐satisfied by soft Cp^R^ coordination.[5](#chem201901167-bib-0005){ref-type="ref"} Of most relevance here, \[Sc(Cp\*)~2~{(C~6~F~5~‐κ^2^‐*F*)B(C~6~F~5~)~3~}\] (Cp\*=C~5~Me~5~)[11](#chem201901167-bib-0011){ref-type="ref"} and \[Ln(Cp\*)~2~{(μ‐C~6~F~5~‐κ^1^‐*F*)~2~B(C~6~F~5~)~2~}\]~2~ (Ln=Pr, Nd)[12](#chem201901167-bib-0012){ref-type="ref"} exhibit equatorial interactions with F atoms of the weakly coordinating anion in the solid state. Complex **1**‐**Dy** is a single‐molecule magnet (SMM) that exhibits magnetic hysteresis at 60 K;[8](#chem201901167-bib-0008){ref-type="ref"} this was a leap of 46 K over the previous record temperature for molecular magnetic hysteresis (at a comparable sweep rate) of 14 K set by Evans and Long in 2011.[13](#chem201901167-bib-0013){ref-type="ref"} Since the publication of **1**‐**Dy** peralkylated \[Dy(Cp^R^)~2~\]^+^ cations have been shown to exhibit hysteresis up to 72 K[14](#chem201901167-bib-0014){ref-type="ref"} and 80 K.[15](#chem201901167-bib-0015){ref-type="ref"} The high hysteresis temperature of **1**‐**Dy** and derivatives cannot solely be attributed to their strong axial ligand fields stabilising the most magnetic *m~J~*=±15/2 states,[16](#chem201901167-bib-0016){ref-type="ref"} as the previous record effective barrier to the reversal of magnetization was seen for \[Dy(*t*BuO)~2~(pyridine)~5~\] (1261 cm^−1^) but this complex does not exhibit hysteresis above 4 K.[17](#chem201901167-bib-0017){ref-type="ref"} State‐of‐the‐art calculations of the spin--phonon coupling in **1**‐**Dy** show that magnetic relaxation in the Orbach (over‐barrier) regime occurs due to localised molecular vibrations, and suggests that the high magnetic hysteresis temperature arises from the constrained metal--ligand vibrational modes intrinsic to the *bis*‐η^5^‐Cp^ttt^ coordination geometry.[8](#chem201901167-bib-0008){ref-type="ref"} We have subsequently shown that isolated \[Ln(Cp^ttt^)~2~\]^+^ cations exhibit anomalously low Raman exponents, distinct from analogues with equatorial ligands, suggesting the unique vibrational modes of Cp^ttt^ are important across a wide temperature range.[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} We previously reported that the bulky *bis*‐Cp^ttt^ framework in combination with the weakly coordinating anion \[B(C~6~F~5~)~4~\]^−^ provide isolated Ln metallocenium cations for smaller, heavier Ln.[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} These studies showed that \[H(SiEt~3~)~2~\]\[B(C~6~F~5~)~4~\][18](#chem201901167-bib-0018){ref-type="ref"} is effective at abstracting either fluoride or chloride from \[Ln(Cp^ttt^)~2~(X)\] precursors to yield **1**‐**Ln**. Herein, we extend these studies to the lighter, larger Ln to define the characteristic features of analogous \[Ln(Cp^ttt^)~2~\]^+^ cations across the Ln series. We find that for the largest members of the **1**‐**Ln** series one *meta*‐F of the \[B(C~6~F~5~)~4~\]^−^ anion coordinates to the exposed equatorial site in the solid state, giving \[Ln(Cp^ttt^)~2~{(C~6~F~5~‐κ^1^‐F)B(C~6~F~5~)~3~}\] (**1**‐**Ln**; Ln=La, Ce, Pr, Nd). However, smaller Sm^III^ yields an isolated cation, \[Sm(Cp^ttt^)~2~\]\[B(C~6~F~5~)~4~\] (**1**‐**Sm**), which is structurally analogous to the heavier members of the series. We were unable to complete the series with the Eu analogue **1‐Eu** as the stability associated with the +2 oxidation state[5](#chem201901167-bib-0005){ref-type="ref"} precluded the synthesis of \[Eu(Cp^ttt^)~2~(X)\] (X=F, Cl) precursors. Results and Discussion {#chem201901167-sec-0002} ====================== According to previously published synthetic methods for the synthesis of **1‐Ln** (Ln=Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu, Y),[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} \[Ln(Cp^ttt^)~2~{(C~6~F~5~‐κ^1^‐*F*)B(C~6~F~5~)~3~}\] (**1‐Ln**; Ln=La, Ce, Pr, Nd) and \[Sm(Cp^ttt^)~2~\]\[B(C~6~F~5~)~4~\] (**1‐Sm**) were all synthesised by the reactions of \[Ln(Cp^ttt^)~2~(Cl)\] (**2‐Ln**; Ln=La, Ce, Pr, Nd, Sm) with \[H(SiEt~3~)~2~\]\[B(C~6~F~5~)~4~\][18](#chem201901167-bib-0018){ref-type="ref"} in benzene (Scheme [1](#chem201901167-fig-5001){ref-type="fig"}). \[H(SiEt~3~)~2~\]\[B(C~6~F~5~)~4~\] was selected as a reagent for Ln metallocenium cation formation as: (i) it is soluble in benzene, hence strongly coordinating O‐donor solvents can be avoided; (ii) there is a significant enthalpic effect due to formation of a strong Si−Cl bond, and (iii) there is an additional entropic effect arising from two reactant species giving three product species. Complexes **1‐Ln** (Ln=Ce, Pr, Nd) were crystallised from dichloromethane (DCM) layered with hexane and became more temperature‐sensitive as soon as DCM was added; thus, these complexes were stored at −25 °C and measurements were performed below −25 °C. Despite repeated efforts we were unable to recrystallise **1‐Ln** (Ln=La, Sm) from DCM/hexane; we later found that these complexes dissolved in hot toluene and could be recrystallised and stored at room temperature. {#chem201901167-fig-5001} The precursors **2‐Ln** were synthesized from LnCl~3~ and two equivalents of KCp^ttt^ by modification of the synthesis of **2‐Pr** and **2‐Nd** from the parent LnCl~3~ and NaCp^ttt^ (Scheme [1](#chem201901167-fig-5001){ref-type="fig"}).[19](#chem201901167-bib-0019){ref-type="ref"} This strategy does not work for Eu as it has previously been shown that the reaction of EuCl~3~ with NaCp^ttt^ gives \[Eu(Cp^ttt^)~2~\], with elimination of half an equivalent of (Cp^ttt^)~2~ via reductive coupling.[20](#chem201901167-bib-0020){ref-type="ref"} We attempted the synthesis of an alternative precursor \[Eu(Cp^ttt^)~2~(F)\] by the oxidation of \[Eu(Cp^ttt^)~2~\][20](#chem201901167-bib-0020){ref-type="ref"} with \[Fe(Cp)~2~\]\[PF~6~\][21](#chem201901167-bib-0021){ref-type="ref"} in toluene, envisaging that this would eliminate gaseous PF~5~ and \[Fe(Cp)~2~\]. However, this did not occur and instead we obtained \[{Eu(Cp^ttt^)(μ‐PF~6~‐κ^4^‐*F*)(THF)}~2~\] (**3**) in 60 % yield, along with several crystals of (Cp^ttt^)~2~ (**4**), presumably due to the contamination of the Eu^II^ starting material with THF solvates, such as \[Eu(Cp^ttt^)~2~(THF)\] or \[Eu(Cp^ttt^)(I)(THF)~*x*~\]~*n*~, and subsequent ligand scrambling. Therefore, Eu is the only Ln apart from radioactive Pm that we were unable to access via these anion abstraction procedures. Crystalline yields for **2‐Ln** are similar (46--52 %) except for **2‐La**, which was isolated in 31 % yield. For **1‐Ln**, the crystalline yields were dependent upon the solvent system; yields were significantly higher when crystallised from toluene (67 % for **1‐La** and 69 % for **1‐Sm**) compared to those crystallised from DCM layered with hexane (21--59 %). Elemental analysis results obtained for **1‐Ln** and **2‐Ln** were generally in excellent agreement with expected values, although low carbon values, likely due to carbide formation from incomplete combustion, were consistently obtained in some cases. In spite of this, the bulk purity of **1‐Ln** was indicated by the consistency of other analytical data with their formulations (see below). Complex **1‐Sm** contains the largest Ln^III^ ion in a Ln metallocenium cation with no equatorial interactions to be isolated to date. The most appropriate explanation for the structural change for the largest Ln^III^ cations is a simple argument based on ionic radii, with the two Cp^ttt^ rings offering a sufficient amount of steric protection around the smaller Sm^III^ ion (CN=6, IR=0.958 Å; CN: effective coordination number; IR: ionic radius)[21](#chem201901167-bib-0021){ref-type="ref"} to saturate the coordination sphere and prevent coordination of the counter‐ion, whereas the early Ln^III^ ions are large enough to accommodate a further interaction (e.g. Nd^III^: CN=6, IR=0.983 Å).[22](#chem201901167-bib-0022){ref-type="ref"} This hypothesis can be tested in future by steric and electronic variations of the Cp^R^ ligands and the weakly coordinating anion. ^1^H NMR spectra were recorded from −350 to +350 ppm for **1‐Ln** and **2‐Ln** (Table [1](#chem201901167-tbl-0001){ref-type="table"}). Significant broadening of signals was seen for all paramagnetic species: the *t*Bu group protons were observed in all cases apart from **1‐Pr**, however, the Cp^ttt^ ring protons were only seen in **1‐La**, **2‐La**, **1‐Sm** and **2‐Sm**. Three signals were observed in the ^1^H NMR spectra of diamagnetic **1‐La** and **2‐La**, and paramagnetic **2‐Sm**, in a ratio of 18:36:4, corresponding to two sets of *t*Bu groups and the Cp^ttt^ ring protons, respectively. The paramagnetism of most **1‐Ln** precluded assignment of their ^13^C{^1^H} NMR spectra. However, for diamagnetic **1‐La** and **2‐La** these could be interpreted: for **1‐La** the expected three Cp^ttt^ ring carbon signals were identified at 135.96, 147.78 and 149.70 ppm, whereas six resonances for the *t*Bu groups were seen between 29.86 and 32.82 ppm. These are assigned as three quaternary and three methyl signals; multiple signals in this region were previously observed for **1‐Lu**.[9](#chem201901167-bib-0009){ref-type="ref"} The \[B(C~6~F~5~)~4~\]^−^ anions in **1‐Ln** were observed by ^11^B{^1^H} and ^19^F{^1^H} NMR spectroscopy (see Supporting Information). In the case of **1‐La**, we could also employ ^13^C{^19^F} NMR spectroscopy to characterise the \[B(C~6~F~5~)~4~\]^−^ anion: the expected four aromatic resonances were located at 124.26 (*ipso*‐), 136.91 (*m*‐), 138.79 (*p*‐) and 148.71 ppm (*o*‐); the signal for the *ipso*‐carbon is a quartet from coupling with ^11^B (^1^ *J* ~BC~=51.8 Hz). Signals in the ^11^B{^1^H} NMR spectra of **1‐Ln** were observed between −20 and −16 ppm, and three peaks were observed in the ^19^F{^1^H} NMR spectra, corresponding to *meta*‐, *para*‐ and *ortho*‐F. In the ^19^F{^1^H} NMR spectrum of **1‐Sm**, the *para*‐F signal is a triplet from coupling to two adjacent *meta*‐F atoms (*J* ~FF~=20.4 Hz), but this is a singlet for **1‐La**, presumably due to broadening by quadrupolar ^139^La (*I*=7/2). The ^11^B and ^19^F NMR spectra of \[Sc(Cp\*)~2~{(C~6~F~5~‐κ^2^‐*F*)B(C~6~F~5~)~3~}\][11](#chem201901167-bib-0011){ref-type="ref"} and \[Ln(Cp\*)~2~{(μ‐C~6~F~5~‐κ^1^‐*F*)~2~B(C~6~F~5~)~2~}\]~2~ (Ln=Pr, Nd)[12](#chem201901167-bib-0012){ref-type="ref"} have not been reported to the best of our knowledge, but all NMR spectra reported herein are similar to those previously discussed for the heavy **1‐Ln** (Ln=Gd--Lu).[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} Together, these data indicate that the \[B(C~6~F~5~)~4~\]^−^ anions in light **1‐Ln** are only weakly bound in solution, and dissociate to give isolated \[Ln(Cp^ttt^)~2~\]^+^ cations, as evidenced by the symmetry of ^19^F NMR spectra.[23](#chem201901167-bib-0023){ref-type="ref"} To probe this further, we performed VT ^19^F NMR studies of **1‐La** in C~7~D~8~, but the spectra were unchanged down to 218 K. Finally, the paramagnetism of **3** precluded assignment of its ^1^H, ^13^C, ^19^F and ^31^P NMR spectra. ###### ^1^H NMR spectra assignments of **1‐Ln** in \[D~2~\]DCM and **2‐Ln** in \[D~6~\]benzene. ----------------------------------------------------------------------------------------- Complex *δ* ^1^H \[ppm\] {C~5~H~2~C(C*H* ~3~)~3~}^−^\ *δ* ^1^H \[ppm\]\ (FWHM/Hz) {C~5~ *H* ~2~C(CH~3~)~3~}^−^ ---------- ----------------------------------------------- ------------------------------ **1‐La** 1.28, 18 H; 1.46, 36 H 6.26, 4 H **1‐Ce** −13.26 (450), 18 H;\ -- −7.93 (920), 36 H **1‐Pr** -- -- **1‐Nd** −17.88 (310), 18 H;\ -- −11.94 (700), 36 H **1‐Sm** −1.37, 36 H 19.69, 2 H **2‐La** 1.25, 18 H; 1.52, 36 H -- **2‐Ce** −13.06, 18 H; −2.53, 36 H -- **2‐Pr** −36.08, 18 H; −7.74 (750), 36 H -- **2‐Nd** −18.95, 18 H; −5.58, 36 H -- **2‐Sm** −6.01, 18 H; 0.55, 36 H 19.80, 4 H ----------------------------------------------------------------------------------------- Wiley‐VCH Verlag GmbH & Co. KGaA The solid‐state structures of **1‐Ln**, **2‐Ln**, **3** and **4** were determined by single crystal XRD (**1‐Nd** is depicted in Figure [1](#chem201901167-fig-0001){ref-type="fig"}; **1‐Sm** in Figure [2](#chem201901167-fig-0002){ref-type="fig"} and **2‐Nd** in Figure [3](#chem201901167-fig-0003){ref-type="fig"}; see the Supporting Information for other structures. Selected bond distances and angles of **1‐Ln** are compiled in Table [2](#chem201901167-tbl-0002){ref-type="table"}). The \[Sm(Cp^ttt^)~2~\]^+^ cation of **1‐Sm** exhibits identical structural features with **1‐Dy**,[8](#chem201901167-bib-0008){ref-type="ref"} whereas, the early Ln (Ln=La, Ce, Pr, Nd) all exhibit interactions between the Ln centre and a *meta*‐fluorine of one of the C~6~F~5~ rings of the \[B(C~6~F~5~)~4~\]^−^ anion. In all cases, the \[Ln(Cp^ttt^)~2~\]^+^ fragments exhibit bent geometries and the Cp~centroid~⋅⋅⋅Ln⋅⋅⋅Cp~centroid~ angles for **1‐Ln** are closer to linearity than the corresponding precursors **2‐Ln**. In the cases of **1‐Sm** and the late Ln metallocenium cations in **1‐Ln** (Ln=Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu),[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} the Cp~centroid~⋅⋅⋅Ln⋅⋅⋅Cp~centroid~ angles are significantly more linear than the \[Ln(Cp^ttt^)~2~\]^+^ fragments in early **1‐Ln** (Ln=La, Ce, Pr, Nd) which have Ln⋅⋅⋅F interactions \[range Cp~centroid~⋅⋅⋅Ln⋅⋅⋅Cp~centroid~: early Ln 148.10(13)° to 149.48(11)°; late Ln 150.2(2)° to 155.11(6)°; **1‐Sm**: 153.23(6)°\]. Two nearly eclipsed η^5^‐Cp^ttt^ rings coordinate to the Ln^III^ centres in all **1‐Ln**.[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} The mean Ln⋅⋅⋅Cp~centroid~ distances decrease regularly across the Ln series \[mean La⋅⋅⋅Cp~centroid~=2.503(3) Å; Lu⋅⋅⋅Cp~centroid~=2.246(4) Å\].[9](#chem201901167-bib-0009){ref-type="ref"} In the complex with isolated cations **1‐Sm**, two equatorial electrostatic Ln⋅⋅⋅C interactions with carbon atoms from a *t*Bu group were observed \[C(7) 2.966(4) Å, C(24) 2.998(4) Å\], with a C⋅⋅⋅Ln⋅⋅⋅C angle of 147.56(11)°; this feature has been noted for the heavier Lns previously.[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} In the case of **1‐Ln** (Ln=La, Ce, Pr, Nd), Ln⋅⋅⋅F distances were observed between 2.679(9) Å (**1‐La**) to 2.632(4) Å (**1‐Nd**); similar contacts between LnCp^R^ ~2~ fragments and \[B(C~6~F~5~)~4~\]^−^ anions have previously been seen for \[Sc(Cp\*)~2~{(C~6~F~5~‐κ^2^‐*F*)B(C~6~F~5~)~3~}\][11](#chem201901167-bib-0011){ref-type="ref"} and \[Ln(Cp\*)~2~{(μ‐C~6~F~5~‐κ^1^‐*F*)~2~B(C~6~F~5~)~2~}\]~2~ (Ln=Pr, Nd).[12](#chem201901167-bib-0012){ref-type="ref"} In the Pr and Nd examples, the two \[Ln(Cp\*)~2~\]^+^ fragments are bridged by two \[B(C~6~F~5~)~4~\]^−^ anions to give a total of four *meta*‐F⋅⋅⋅Ln interactions, but in the Sc example a single C~6~F~5~ ring binds to one Sc^III^ centre via both a *meta*‐ and a *para*‐F atom. The shortest Ln⋅⋅⋅F distance in **1‐Sm** is 5.693(2) Å, which indicates that Ln⋅⋅⋅F interactions are only observed for the **1‐Ln** series when the Ln^III^ ionic radius is larger than a critical value, when the *bis*‐Cp^ttt^ framework no longer provides sufficient coverage to prevent equatorial interactions with \[B(C~6~F~5~)~4~\]^−^ anions. ![Molecular structure of \[Nd(Cp^ttt^)~2~{(C~6~F~5~‐κ^1^‐*F*)B(C~6~F~5~)~3~}\] (**1‐Nd**). Displacement ellipsoids set at 30 % probability level and hydrogen atoms are omitted for clarity.](CHEM-25-7749-g001){#chem201901167-fig-0001} ![Molecular structure of \[Sm(Cp^ttt^)~2~\]\[B(C~6~F~5~)~4~\] (**1‐Sm(C~7~H~8~)~1.5~**). Displacement ellipsoids set at 30 % probability level and hydrogen atoms and lattice solvent are omitted for clarity.](CHEM-25-7749-g002){#chem201901167-fig-0002} ![Molecular structure of \[Nd(Cp^ttt^)~2~(Cl)\] (**2‐Nd**). Displacement ellipsoids set at 30 % probability level and hydrogen atoms are omitted for clarity.](CHEM-25-7749-g003){#chem201901167-fig-0003} ###### Selected bond distances and angles of \[Ln(Cp^ttt^)~2~\]^+^ fragments in **1‐Ln**. Complex Ln⋅⋅⋅Cp~cent~ \[Å\] Cp~cent~⋅⋅⋅Ln⋅⋅⋅Cp~cent~ \[°\] Ln⋅⋅⋅F \[Å\] -------------------------------------------------------- --------------------- -------------------------------- -------------- **1‐La** 2.507(2), 2.499(2) 148.13(2) 2.679(9) **1‐Ce** 2.507(4), 2.478(5) 149.1(2) 2.677(5) **1‐Pr** 2.479(3), 2.459(4) 149.48(11) 2.664(3) **1‐Nd** 2.445(5), 2.464(4) 148.10(13) 2.632(4) **1‐Sm** 2.384(2), 2.392(2) 153.23(6) -- **1‐Gd** [9](#chem201901167-bib-0009){ref-type="ref"} 2.364(5), 2.345(5) 150.2(2) -- **1‐Tb** [10](#chem201901167-bib-0010){ref-type="ref"} 2.325(4), 2.339(5) 152.2(2) -- **1‐Dy** [8](#chem201901167-bib-0008){ref-type="ref"} 2.318(2), 2.314(2) 152.56(7) -- Wiley‐VCH Verlag GmbH & Co. KGaA The ATR‐IR spectra of **1‐Ln** and **2‐Ln** were obtained as microcrystalline powders (spectra of **1‐Ln** compiled in Figure [4](#chem201901167-fig-0004){ref-type="fig"}; see Supporting Information for full ATR‐IR spectra). As expected, most spectra show broadly similar features, despite the structural differences between **1‐Sm** and other light **1‐Ln**. However, **1‐Pr** is anomalous in exhibiting three additional strong absorptions at 798, 1086 and 1261 cm^−1^; these differences could not be assigned. {#chem201901167-fig-0004} The electronic spectra of **1‐Ln** and **2‐Ln** were obtained at room temperature as 1 m[m]{.smallcaps} solutions in DCM (spectra of **1‐Ln** compiled in Figure [5](#chem201901167-fig-0005){ref-type="fig"}; see Supporting Information for full UV/Vis/NIR spectra). Due to their Laporte‐forbidden nature, f--f transitions are relatively weak,[5](#chem201901167-bib-0005){ref-type="ref"} so charge transfer (CT) bands tailing in from the UV region dominate the spectrum. Dilute solutions of **1‐Nd** and **1‐Sm** are pale green and pale orange, respectively, and some f--f transitions could be clearly observed: **1‐Nd** shows absorptions at $\widetilde{\nu}$ ~max~=16600 cm^−1^ (*ϵ*=46 mol^−1^ dm^3^ cm^−1^), owing to the ^4^I~9/2~→^4^G~5/2~ transitions;[24](#chem201901167-bib-0024){ref-type="ref"} and **1‐Sm** shows absorptions at $\widetilde{\nu}$ ~max~=11 400 cm^−1^ (*ϵ*=20 mol^−1^ dm^3^ cm^−1^) and 8400 cm^−1^ (*ϵ*=50 mol^−1^ dm^−3^ cm^−1^), which likely arise from ^6^H~5/2~→^6^F~11/2~, ^6^F~7/2~ transitions, respectively.[24](#chem201901167-bib-0024){ref-type="ref"} Similar absorptions can be seen in the spectra of the precursors **2‐Nd** and **2‐Sm**. ![UV/Vis/NIR spectra of **1‐Ln** (ca. 1 m[m]{.smallcaps} in CH~2~Cl~2~) from 6000 to 21 000 cm^−1^ at room temperature.](CHEM-25-7749-g005){#chem201901167-fig-0005} Complete active space self‐consistent field spin‐orbit (CASSCF‐SO) calculations were performed on unoptimized XRD structures of **1‐Ln** and **2‐Ln** using MOLCAS 8.0[25](#chem201901167-bib-0025){ref-type="ref"} to determine their electronic structures (see Supporting Information for full details). We split our discussion between the Kramers (odd number of unpaired electrons) and non‐Kramers (even number of unpaired electrons) ions for clarity. Kramers ions (Ce, Nd and Sm) {#chem201901167-sec-0003} ---------------------------- In the case of Kramers ions with an odd number of unpaired electrons, there will always be two‐fold electronic degeneracy in zero magnetic field, irrespective of a low‐symmetry CF. Each of these Kramers doublets is a linear combination of *m~J~* projections of the total angular momentum *J*, but can also be defined as an effective *S*=1/2 state. With this definition, the *S*=1/2 states have an effective *g*‐matrix describing the magnetic anisotropy owing to the varying *m~J~* composition of each doublet. The 3×3 *g*‐matrix can always be diagonalised to yield three principal *g*‐values and their corresponding orientations in the molecular frame; here we define the principal *g*‐values such that *g* ~1~\<*g* ~2~\<*g* ~3~. While the choice of quantisation axis for the CF is irrelevant for any physical observable, it defines how each state is decomposed into linear combinations of *m~J~* projections (*m~J~* is the projection of the total angular momentum *J* along the quantisation axis). In the case of pure *m~J~* states (*m~J~*\>1/2), *g* ~1~=*g* ~2~=0 and *g* ~3~=2*g~J~m~J~* where *g~J~* is the Landé *g*‐factor of the *J* multiplet of the Ln^III^ ion. In the general case of mixed *m~J~* states it is useful to define \<*J* ~z~\>, the expectation value of the *J* ~z~ operator which measures the projection of *J* on the quantisation axis, to understand the magnetic nature of the states; in the limiting case of a pure *m~J~* state, \<*J* ~z~\>=*m~J~*. For **1‐Ce** (Table [3](#chem201901167-tbl-0003){ref-type="table"}) we observe that the ground Kramers doublet is highly axial (*g* ~1~≈*g* ~2~\<0.1, *g* ~3~=4.16), and is thus well‐described as *m~J~*=±5/2 when the CF is quantised along the *g* ~3~ direction (Tables [3](#chem201901167-tbl-0003){ref-type="table"} and S17). For this choice of quantisation axis, the two excited Kramers doublets are dominated by *m~J~*=±3/2 and ±1/2, respectively. Here, the strong *pseudo*‐linear CF of the *bis*‐Cp^ttt^ ligand set stabilises the oblate spheroid‐shaped electron density of *m~J~*=±5/2 and destabilises the prolate spheroid‐shaped *m~J~*=±1/2 state to give a typical double‐well potential of easy‐axis magnetic anisotropy (Figure [6](#chem201901167-fig-0006){ref-type="fig"}), as expected from simple electrostatic principles.[16](#chem201901167-bib-0016){ref-type="ref"} This result shows that despite the counter ion binding to the metal ion, its influence on the electronic structure of the Ce^III^ ion is much weaker compared to the *bis*‐Cp^ttt^ CF; this can be verified by repeating the calculations without the \[B(C~6~F~5~)~4~\]^−^ anion (Table S18), showing only a marginal increase in axiality. For **2‐Ce** however, while the ground doublet is also *m~J~*=±5/2 with the CF defined along the Cp^ttt^--Cp^ttt^ direction (Table S19, Figure S131), the first excited state is almost isotropic and the most excited state also resembles the *m~J~*=±5/2 state when the CF is quantised along the "*pseudo*‐C~3~" direction perpendicular to both the Cp^ttt^--Cp^ttt^ direction and the Ce--Cl vector (Figure S132); this demonstrates that the Cl^−^ anion is competing with the Cp^ttt^ ligands and has a significant influence on the CF. ###### Low‐lying electronic structure of **1‐Ce** calculated with CASSCF‐SO in zero‐field. Wavefunction decomposition quantised along the *g* ~3~ direction of the ground doublet. Energy \[cm^−1^\] *g* ~1~ *g* ~2~ *g* ~3~ Angle \[°\] Wavefunction ------------------- --------- --------- --------- ------------- ---------------------------------------- 0.0 0.05 0.10 4.16 -- 100 %$\left| {\pm 5/2} \right.\rangle$ 786.6 0.48 0.52 2.21 2.74 93 %$\left| {\pm 3/2} \right.\rangle$ 1577.1 1.82 3.00 0.84 2.33 94 %$\left| {\pm 1/2} \right.\rangle$ Wiley‐VCH Verlag GmbH & Co. KGaA {#chem201901167-fig-0006} The electronic structure of **1‐Sm** (Table [4](#chem201901167-tbl-0004){ref-type="table"}), like **1‐Ce**, can be explained with simple electrostatic principles.[16](#chem201901167-bib-0016){ref-type="ref"} The CF is still dominated by the near‐axial potential of the Cp^ttt^ ligands, but as the electron density distributions of the *m~J~* states of Sm^III^ are in the opposite sense to Ce^III^ (i.e., *m~J~*=±5/2 has a prolate spheroidal electron density and *m~J~*=±1/2 is oblate spheroidal), this makes **1‐Sm** the antithesis of an SMM with the *m~J~*=±1/2, ±3/2 and ±5/2 doublets in ascending energy (Figure [6](#chem201901167-fig-0006){ref-type="fig"}; Tables [4](#chem201901167-tbl-0004){ref-type="table"} and S28). For **2‐Sm**, the *g* ~3~ direction of the ground Kramers doublet lies along the "*pseudo*‐C~3~" direction (Figure S140), yet when the CF is quantised along this direction there is strong mixing between *m~J~*=±1/2 and ±3/2 owing to the competition between the Cp^ttt^ and Cl^−^ ligands (Table S29). For both **1‐Sm** and **2‐Sm** the *g*‐values are much reduced from those expected for an isolated ^6^H~5/2~ multiplet; this is due to considerable *J*‐mixing with the low‐lying ^6^H~7/2~ and ^6^H~9/2~ multiplets, rendering a CF model in the basis of a pure *J*=5/2 state an inaccurate representation of the wavefunction. ###### Low‐lying electronic structure of **1‐Sm** calculated with CASSCF‐SO in zero‐field. Wavefunction decomposition quantised along the *g* ~3~ direction of the ground doublet. Energy \[cm^−1^\] *g* ~1~ *g* ~2~ *g* ~3~ Angle \[°\] Wavefunction ------------------- --------- --------- --------- ------------- ---------------------------------------- 0.0 0.18 0.38 1.08 -- 100 %$\left| {\pm 1/2} \right.\rangle$ 308.69 0.12 0.13 2.57 2.26 97 %$\left| {\pm 3/2} \right.\rangle$ 905.10 0.06 0.07 4.04 5.11 99 %$\left| {\pm 5/2} \right.\rangle$ Wiley‐VCH Verlag GmbH & Co. KGaA For Nd^III^ the situation is less clear than for Ce^III^ and Sm^III^ because the magnitude of the quadrupolar terms in the expansion of the *m~J~* electron densities are smaller than for the other Ln^III^ ions,[26](#chem201901167-bib-0026){ref-type="ref"} and therefore simple electrostatics do not dominate the electronic structure and mixing of *m~J~* states by the low symmetry CF is more severe. For **1‐Nd** the two lowest lying doublets are 86 % *m~J~*=±9/2 and 86 % *m~J~*=±7/2, respectively, when the CF is quantised along the Cp^ttt^‐Cp^ttt^ direction (Table S25), however the subsequent excited states are quite mixed and the highest‐energy doublet is 86 % *m~J~*=±9/2 when the CF is quantised along the "*pseudo*‐C~3~" direction (based on the nearest F‐atom; Figure S136). Removal of the counterion reduces the axiality of the most excited state (Table S26), however this does not change the overall picture of the electronic structure. This suggests that it is the bent arrangement of the Cp^ttt^ rings, and not the influence of the \[B(C~6~F~5~)~4~\]^−^ counterion, that forces the magnetic anisotropy of the most excited state to be perpendicular to that of the ground state, further highlighting the minimal electronic influence of the bulky anion. The electronic structure of **2‐Nd** is generally similar to that of **1‐Nd**, however the mixing is more severe (Table S27) and is not discussed further. Non‐Kramers ions (Pr) {#chem201901167-sec-0004} --------------------- In the case of non‐Kramers ions, the action of a low‐symmetry CF can completely remove the degeneracy of the electronic states, although it does not have to do so (e.g. in high symmetry environments). In the case of a near‐axial CF, there will be a number of *pseudo*‐doublets that are well described as pure *m~J~* functions and a singlet state corresponding to *m~J~*=0; the departure from axiality determines the degree to which the *pseudo*‐doublets are split into singlets. For *pseudo*‐doublets only a single principal *g*‐value can be defined and the other two are zero (note that this does not necessarily indicate a pure *m~J~* state); thus, \<*J* ~z~\> is a useful indicator for non‐Kramers ions. The situation for **1‐Pr** (Table [5](#chem201901167-tbl-0005){ref-type="table"}) is similar to that of **1‐Ce**, where the ground *pseudo*‐doublet is well defined as *m~J~*=±4 when the CF is quantised along the Cp^ttt^‐Cp^ttt^ direction (Tables [5](#chem201901167-tbl-0005){ref-type="table"} and S20), however the low‐symmetry CF leads to mixing between the opposing *m~J~* projections and small splittings in the *pseudo*‐doublets at low energies and substantial splittings at the more energetic end of the spectrum. These mixings can be partially lifted by applying a small magnetic field along the *g* ~3~ direction of the ground *pseudo*‐doublet (Table S21). Removal of the counterion in the CASSCF‐SO calculations for **1‐Pr** results in larger energy gaps between the CF states and reduces the splitting within the *pseudo*‐doublets at higher energies, but generally does not have a large effect on the electronic structure (Table S22). For **2‐Pr**, with an extra competitive element in the CF (the Cl^−^ ion), the resulting electronic structure is nearly all singlets (Table S23), and application of a small magnetic field is inconsequential (Table S24). ###### Low‐lying electronic structure of **1‐Pr** calculated with CASSCF‐SO in zero‐field. Wavefunction decomposition quantised along the *g* ~3~ direction of the ground doublet. Energy \[cm^−1^\] *g* ~3~ Angle \[°\] Wavefunction ------------------- ---------------------------------------------------------------------- ------------- ---------------------------------------------------------------------- 0.00 6.18 -- 50 %$\left| {- 4} \right.\rangle$ +50 %$\left| {+ 4} \right.\rangle$ 0.31 50 %$\left| {- 4} \right.\rangle$ +50 %$\left| {+ 4} \right.\rangle$ 852.05 4.76 10.3 49 %$\left| {- 3} \right.\rangle$ +49 %$\left| {+ 3} \right.\rangle$ 863.29 48 %$\left| {- 3} \right.\rangle$ +48 %$\left| {+ 3} \right.\rangle$ 1050.72 3.28 20.8 48 %$\left| {- 2} \right.\rangle$ +48 %$\left| {+ 2} \right.\rangle$ 1054.69 47 %$\left| {- 2} \right.\rangle$ +47 %$\left| {+ 2} \right.\rangle$ 1210.68 -- -- 49 %$\left| {- 1} \right.\rangle$ +49 %$\left| {+ 1} \right.\rangle$ 1377.28 -- -- 49 %$\left| {- 1} \right.\rangle$ +49 %$\left| {+ 1} \right.\rangle$ 1403.77 -- -- 97 %$\left| 0 \right.\rangle$ Wiley‐VCH Verlag GmbH & Co. KGaA To experimentally probe the nature of the ground states, we have performed cryogenic electron paramagnetic resonance (EPR) spectroscopy on the Kramers ion analogues (Ce^III^, Nd^III^ and Sm^III^) of **1‐Ln** and **2‐Ln**. We note generally that the intensities of the signals are very weak, however that there is excellent agreement between experiment and theory (Table [6](#chem201901167-tbl-0006){ref-type="table"}; Figures S126--S129). We do not observe a signal for **1‐Nd** or **2‐Sm**: the former is predicted to be EPR silent from CASSCF‐SO (the ground doublet is dominated by *m~J~*=±9/2 and has very small *g* ~1~ and *g* ~2~, thus there is a vanishing component of Δ*m~J~*=±1 between the two states, which is the EPR selection rule), however the latter is apparently too weak or is broadened beyond detection. ###### Calculated and measured *g*‐values for the Kramers analogues of **1‐Ln** and **2‐Ln**. Complex CASSCF‐SO Experiment ---------- ------------- ------------ **1‐Ce** 4.16 4.22 **2‐Ce** 4.09 4.19 **1‐Nd** 6.32^\[a\]^ Silent **2‐Nd** 6.00 6.10 **1‐Sm** 1.08 ≈1.5 **2‐Sm** 0.71 Silent \[a\] **1‐Nd** is predicted to be EPR silent. Wiley‐VCH Verlag GmbH & Co. KGaA The χ~M~T products for **1‐Ln** and **2‐Ln** show a gradual decrease with temperature for all samples (Figure S109), however the data for **2‐Pr** decrease rapidly at lower temperatures, and those for **1‐Pr** decrease slowly below 20 K. For **2‐Pr** this is due to a singlet ground state, which is confirmed by magnetisation measurements (Figure S111) and CASSCF‐SO calculations (Table S23), but for **1‐Pr** the experimental decrease is much more substantial than predicted by CASSCF‐SO, suggesting that the calculations have underestimated the splitting within the ground pseudo‐doublet on the order of a few cm^−1^ (Table S20). Generally, these magnetic data are in good agreement with CASSCF‐SO‐calculated electronic structures, however we note that due to the small magnetic moments of the light Ln^III^ ions (where the SO coupling gives *J*=\|*L*−*S*\| ground multiplets) the experimental data are sensitive to small errors in sample masses and diamagnetic corrections. The data for **1‐Sm** and **2‐Sm** are the most susceptible to this, as they feature not only the lowest magnetic moments, but also because they are highly sensitive to *J* mixing (see above) such that the calculations only reveal an approximate electronic structure. We were curious if the anomalous Raman exponents we observed previously for the heavy \[Ln(Cp^ttt^)~2~\]^+^ series[8](#chem201901167-bib-0008){ref-type="ref"}, [9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} could also be observed for these light Ln^III^ analogues. Therefore, we investigated the Kramers ions with AC susceptibility studies to examine their magnetisation dynamics (Figures [7](#chem201901167-fig-0007){ref-type="fig"} and S114--S125). Only in the case of **2‐Ce** was it possible to observe slow magnetic relaxation in zero DC field, and therefore we employed a 0.1 T DC field for all measurements; however, no slow relaxation could be detected for **1‐Sm** and **2‐Sm**. Only for **2‐Nd** is relaxation via an Orbach mechanism plausible (Figure S125), giving an effective barrier of *U* ~eff~=51.2 cm^−1^ with *τ* ~0~=9.64×10^−8^ s; CASSCF‐SO predicts a first excited state at around 72 cm^−1^, so this is not unreasonable. In all other cases a Raman mechanism is the best model for the data, and indeed is also a possible explanation for **2‐Nd**. For both pairs of **1‐Ce** vs. **2‐Ce** and **1‐Nd** vs. **2‐Nd**, the Raman exponent is smaller for **1‐Ln** than for **2‐Ln** where the Cl^−^ is bound, similar to the trend we have observed previously for the heavy Ln metallocenium cations (Table [7](#chem201901167-tbl-0007){ref-type="table"}),[9](#chem201901167-bib-0009){ref-type="ref"}, [10](#chem201901167-bib-0010){ref-type="ref"} which supports our previous conclusion that the absence of monodentate ligands in \[Ln(Cp^ttt^)~2~\]^+^ is responsible for lowered Raman exponents. However, the exponents for **1‐Ce** and **1‐Nd** are considerably larger than for **1‐Tb**, **1‐Dy** and **1‐Ho**, which suggests that there could be an influence of the proximate \[B(C~6~F~5~)~4~\]^−^ anion on the relaxation dynamics of these species. {#chem201901167-fig-0007} ###### Raman relaxation parameters from AC magnetometry. Complex *H* \[T\] *C* \[s^−1^ K^−*n*^\] *n* ----------------------------------------------------------------------------- ----------- ----------------------- ------- **1‐Ce** 0.1 0.0308 5.37 **2‐Ce** 0.1 0.00475 6.48 **1‐Nd** 0.1 0.00117 6.29 **2‐Nd** 0.1 0.000300 8.74 **1‐Tb** [10](#chem201901167-bib-0010){ref-type="ref"} 0 24 1.2 **\[Tb(Cp^ttt^)~2~(BH~4~)\]** [10](#chem201901167-bib-0010){ref-type="ref"} 0.1 0.26 4.6 **1‐Dy** [8](#chem201901167-bib-0008){ref-type="ref"} 0 0.000001664 2.151 **2‐Dy** [9](#chem201901167-bib-0009){ref-type="ref"} 0.1 0.0023 5.3 **1‐Ho** [9](#chem201901167-bib-0009){ref-type="ref"} 0.1 3.4 2.9 **2‐Sm** [9](#chem201901167-bib-0009){ref-type="ref"} 0.1 0.015 8.5 **2‐Yb** [9](#chem201901167-bib-0009){ref-type="ref"} 0.1 0.017 7.1 Wiley‐VCH Verlag GmbH & Co. KGaA Conclusions {#chem201901167-sec-0005} =========== We have completed a series of analogous Ln metallocenium cations \[Ln(Cp^ttt^)~2~\]^+^, with the exception of Eu and Pm. In the case of the early Ln (Ln=La‐Nd) the \[B(C~6~F~5~)~4~\]^−^ counter‐ions weakly coordinate to the metal centre through a *meta*‐fluorine atom in the solid state, as the metals are large enough to incorporate this additional interaction; these equatorial interactions are absent in the solid state for the later Ln (Ln=Sm, Gd--Lu) analogues which contain smaller Ln centres. However, in the solution phase these interactions could not be detected by VT ^19^F NMR spectroscopy, indicating that isolated Ln metallocenium cations are present in fluid solution.[23](#chem201901167-bib-0023){ref-type="ref"} Analysis of the electronic structures of \[Ln(Cp^ttt^)~2~\]^+^ for the early Ln suggests that the weak equatorial *meta*‐fluorine interaction has little effect on the axiality of these systems. Measurement of the relaxation dynamics shows a consistent picture across the \[Ln(Cp^ttt^)~2~\]^+^ series, where the absence of monodentate ligands leads to lower Raman exponents than when they are present, indicating that this effect is a hallmark of all Ln metallocenium cations. We cannot rule out the influence of \[B(C~6~F~5~)~4~\]^−^ on the relaxation dynamics of **1‐Ce** and **1‐Nd**, and suggest that the effect of weak equatorial donors on magnetic relaxation mechanisms may be explored in the future by using different Cp^R^ ligands and a range of weakly coordinating anions. Experimental Section {#chem201901167-sec-0006} ==================== **Materials and methods**. All manipulations were carried out using standard Schlenk line and glove box techniques under dry argon. Solvents were passed through columns containing alumina or were dried by refluxing over K or CaH~2~ (DCM), and were stored over K mirrors or 4 Å molecular sieves (THF, DCM) and degassed before use. For NMR spectroscopy, \[D~6~\]benzene and \[D~8~\]toluene were dried by refluxing over K, and \[D~2~\]DCM was dried by refluxing over CaH~2~. NMR solvents were degassed by three freeze‐pump‐thaw cycles, and vacuum‐transferred before use. Anhydrous LnCl~3~ were purchased from Alfa Aesar and were used as received. KCp^ttt^,[27](#chem201901167-bib-0027){ref-type="ref"} \[H(SiEt~3~)~2~\]\[B(C~6~F~5~)~4~\][18](#chem201901167-bib-0018){ref-type="ref"} and \[Eu(Cp^ttt^)~2~\][20](#chem201901167-bib-0020){ref-type="ref"} were prepared according to literature methods, whilst **2‐Ln** (Ln=Pr, Nd) were made by a modification of published procedures.[19](#chem201901167-bib-0019){ref-type="ref"} ^1^H (400 and 500 MHz), ^13^C{^1^H} (100 and 125 MHz), ^13^C{^19^F} (125 MHz),^11^B{^1^H} (128 and 160 MHz), and ^19^F{^1^H} (376 MHz) NMR spectra were obtained on Avance III 400 or 500 MHz spectrometers at 298 K. UV/Vis/NIR spectroscopy was performed on samples in Youngs tap‐appended 10 mm path length quartz cuvettes on an Agilent Technologies Cary Series UV/Vis/NIR spectrophotometer at 175--3300 nm. ATR‐Fourier transform infrared (ATR‐FTIR) spectra were recorded as microcrystalline powders using a Bruker Tensor 27 spectrometer. Elemental analyses were performed by Mrs Anne Davies and Mr Martin Jennings at The University of Manchester School of Chemistry Microanalysis Service, Manchester, UK. General synthetic procedures for **1‐Ln** and **2‐Ln** are given below; full details are in the Supporting Information. **\[Ln(Cp^ttt^)~2~{(C~6~F~5~‐κ^1^‐*F*)B(C~6~F~5~)~3~}\] (La=La‐Nd), \[Sm(Cp^ttt^)~2~\] \[B(C~6~F~5~)~4~\] (1‐Ln)**: Benzene (15 mL) was added to a mixture of \[H(SiEt~3~)~2~\]\[B(C~6~F~5~)~4~\] (0.5--0.8 mmol) and **2‐Ln** (0.5--0.8 mmol) at room temperature. The mixture was stirred for 16 hours and a precipitate formed. The volatiles were removed in vacuo to give a powder, which was washed with hexane (10--15 mL) and benzene (10--15 mL). In some cases, the crude material was dissolved in DCM (typically 1.4--3 mL) at −78 °C, and layered with 1--1.5 equiv of hexane. Storage at −25 °C overnight gave crystals of **1‐Ln** (Ln=Ce, Pr, Nd). In other cases, the crude material was dissolved in hot toluene (5 mL). Storage at room temperature gave **1‐Ln** (Ln=La, Sm). **1‐La**: Colourless crystals (0.432 g, 67 %). Anal. Calcd (%) for C~58~H~58~BF~20~La: C, 54.22; H, 4.55; Found: C, 52.22; H, 4.15. ^1^H NMR (\[D~2~\]DCM, 400 MHz, 298 K): *δ*=1.38 (s, 18 H, Cp‐C(C*H* ~3~)~3~), 1.46 (s, 36 H, Cp‐C(C*H* ~3~)~3~), 6.26 ppm (s, 4 H, Cp‐C*H*). ^11^B{^1^H} NMR (\[D~2~\]DCM, 128 MHz, 298 K): *δ*=−16.63 ppm (s). ^13^C{^1^H} NMR (\[D~2~\]DCM, 125 MHz, 298 K): *δ*=29.86 and 30.00 (C(*C*H~3~)~3~), 30.53 (C(*C*H~3~)~3~), 31.03 and 31.18 (*C*(CH~3~)~3~), 32.82 (*C*(CH~3~)~3~), 135.96 (Cp‐*C*H), 147.78 (Cp‐*C*), 149.70 ppm (Cp‐*C*). ^13^C{^19^F} NMR (\[D~2~\]DCM, 125 MHz, 298 K): *δ*=124.26 (q, B‐C~*ipso*~, ^1^ *J* ~BC~=51.8 Hz), 136.91 (s, *m*‐*C*F), 138.79 (s, *p*‐CF), 148.71 ppm (s, *o*‐*C*F). ^19^F{^1^H} NMR (\[D~2~\]DCM, 376 MHz, 298 K): *δ*=−166.98 (s, *m*‐*F*), −162.89 (s, *p*‐*F*), −132.67 ppm (s, *o*‐*F*). The low solubility of **1‐La** in \[D~6~\]benzene precluded assignment of ^1^H and ^13^C{^1^H} NMR spectra in this solvent. ^11^B{^1^H} NMR (\[D~6~\]benzene, 160 MHz, 298 K): *δ*=−16.00 ppm (s). ^19^F{^1^H} NMR (\[D~6~\]benzene, 376 MHz, 298 K): *δ*=−165.32 (s, *m*‐*F*), −160.90 (s, *p*‐*F*), −131.66 ppm (s, *o*‐*F*). FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2967 (br, m), 2874 (w), 2819 (w), 1643 (m), 1513 (s), 1459 (s), 1365 (m), 1273 (m), 1240 (m), 1087 (s), 977 (s), 924 (w), 828 (s), 773 (s), 756 (s), 683 (s), 660 (s), 609 (m), 574 (m), 470 (w), 440 cm^−1^ (w). **1‐Ce**: Yellow crystals (0.137 g, 21 %). Anal. Calcd (%) for C~58~H~58~BF~20~Ce⋅1.5CH~2~Cl~2~: C, 50.56; H, 4.35; Found: C, 50.90; H, 4.18. *μ* ~eff~ (Evans method, 298 K, \[D~2~\]DCM): 2.12 μ~B~. ^1^H NMR (\[D~2~\]DCM, 500 MHz, 298 K): *δ*=−13.26 (br, 18 H, *ν* ~1/2~∼450 Hz, Cp‐C(C*H* ~3~)~3~), −7.93 ppm (br, 36 H, *ν* ~1/2~∼920 Hz, Cp‐C(C*H* ~3~)~3~); no other signals observed. ^11^B{^1^H} NMR (\[D~2~\]DCM, 160 MHz, 298 K): *δ*=−18.02 ppm (s). The paramagnetism of **1‐Ce** precluded assignment of its ^13^C{^1^H} NMR spectrum. ^19^F{^1^H} NMR (\[D~2~\]DCM, 376 MHz, 298 K): *δ*=−170.26 (s, *m*‐*F*), −164.33 (s, *p*‐*F*), −134.61 ppm (s, *o*‐*F*). FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2963 (br, m), 2871 (w), 2821 (w), 1643 (m), 1512 (s), 1459 (s), 1365 (m), 1273 (w), 1241 (w), 1087 (s), 976 (s), 924 (w), 867 (w), 830 (m), 773 (s), 756 (s), 683 (s), 660 (s), 609 (m), 574 (m), 441 cm^−1^ (w). **1‐Pr**: Yellow crystals (0.255 g, 40 %). Anal. Calcd (%) for C~58~H~58~BF~20~Pr⋅2 CH~2~Cl~2~: C, 49.47; H, 4.29; Found: C, 47.68; H, 4.27. *μ* ~eff~ (Evans method, 298 K, \[D~2~\]DCM): 2.70 μ~B~. The paramagnetism of **1‐Pr** precluded assignment of its ^1^H and ^13^C{^1^H} NMR spectra. ^11^B{^1^H} NMR (\[D~2~\]DCM, 128 MHz, 298 K): *δ*=−19.42 ppm (s). ^19^F{^1^H} NMR (\[D~2~\]DCM, 376 MHz, 298 K): *δ*=−172.86 (s, *m*‐*F*), −166.16 (s, *p*‐*F*), −136.51 ppm (s, *o*‐*F*). FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2963 (m), 2909 (w), 2871 (w), 1643 (m), 1460 (s), 1365 (m), 1261 (s), 1242 (w), 1086 (s), 1021 (m), 977 (s), 798 (s), 774 (m), 755 (m), 683 (s), 660 (s), 609 (w), 573 (m), 473 (w), 441 cm^−1^ (w). **1‐Nd**: Green crystals (0.381 g, 59 %). Anal. Calcd (%) for C~58~H~58~BF~20~Nd⋅2 CH~2~Cl~2~: C, 49.36; H, 4.28; Found: C, 49.51; H, 4.17. *μ* ~eff~ (Evans method, 298 K, \[D~2~\]DCM): 3.42 μ~B~. ^1^H NMR (\[D~2~\]DCM, 400 MHz, 298 K): *δ*=−17.88 (br, 18 H, *ν* ~1/2~∼310 Hz, Cp‐C(C*H* ~3~)~3~), −11.94 ppm (br, 36 H, *ν* ~1/2~∼700 Hz, Cp‐C(C*H* ~3~)~3~); no other signals observed. ^11^B{^1^H} NMR (\[D~2~\]DCM, 128 MHz, 298 K): *δ*=−18.55 ppm (s). The paramagnetism of **1‐Nd** precluded assignment of its ^13^C{^1^H} NMR spectrum. ^19^F{^1^H} NMR (\[D~2~\]DCM, 376 MHz, 298 K): *δ*=−170.31 (s, *m*‐*F*), −164.92 (s, *p*‐*F*), −135.17 ppm (s, *o*‐*F*). FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2967 (br, m), 2875 (w), 2820 (w), 1643 (m), 1512 (s), 1460 (s), 1394 (w), 1365 (m), 1254 (w), 1241 (w), 1087 (s), 977 (s), 955 (w), 924 (w), 829 (m), 773 (s), 756 (s), 683 (s), 660 (s), 609 (w), 574 (m), 477 (w), 440 cm^−1^ (w). **1‐Sm**: Red crystals (0.790 g, 69 %). Anal. Calcd (%) for C~58~H~58~BF~20~Sm: C, 53.74; H, 4.51; Found: C, 53.88; H, 4.49. *μ* ~eff~ (Evans method, 298 K, \[D~2~\]DCM): 1.90 μ~B~. The paramagnetism of **1‐Sm** precluded assignment of its ^13^C{^1^H} NMR spectrum. ^1^H NMR (\[D~2~\]DCM, 400 MHz, 298 K): *δ*=−1.37 (s, 36 H, Cp‐C(C*H* ~3~)~3~), 19.69 ppm (s, 2 H, Cp‐C*H*); no other signals observed. ^11^B{^1^H} NMR (\[D~2~\]DCM, 128 MHz, 298 K): *δ*=−16.76 ppm (s). ^19^F{^1^H} NMR (\[D~2~\]DCM, 376 MHz, 298 K): *δ*=−167.89 (s, *m*‐*F*), −163.76 (t, *J* ~FF~=20.4 Hz, *p*‐*F*), −133.13 ppm (s, *o*‐*F*). ^1^H NMR (\[D~6~\]benzene, 400 MHz, 298 K): *δ*=−2.58 (s, 18 H, Cp‐C(C*H* ~3~)~3~), −0.90 (s, 36 H, Cp‐C(C*H* ~3~)~3~), 18.80 ppm (s, 4 H, Cp‐C*H*). ^11^B{^1^H} NMR (\[D~6~\]benzene, 128 MHz, 298 K): *δ*=−16.32 ppm (s). ^19^F{^1^H} NMR (\[D~6~\]benzene, 376 MHz, 298 K): *δ*=−168.96 (s, *m*‐*F*), −162.02 (s, *p*‐*F*), −131.93 ppm (s, *o*‐*F*). FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2964 (br, m), 2872 (w), 2796 (w), 1642 (m), 1512 (s), 1459 (s), 1367 (m), 1276 (m), 1239 (m), 1083 (s), 977 (s), 843 (w), 806 (w), 774 (m), 756 (m), 735 (m), 683 (m), 661 (s), 611 (w), 574 (w), 466 (w), 449 cm^−1^ (w). **\[Ln(Cp^ttt^)~2~(Cl)\] (Ln=La, Ce, Pr, Nd, Sm) (2‐Ln)**: THF (30 mL) was added to a Teflon tap‐appended ampoule containing a pre‐cooled (−78 °C) mixture of LnCl~3~ (2 mmol) and KCp^ttt^ (4 mmol). The reaction mixture was allowed to warm to room temperature and heated in an oil bath at 80 °C for 16 hours. The solvent was removed in vacuo and toluene (30 mL) was added. The reaction mixture was heated in an oil bath at 120 °C for 16 hours. The resultant suspension was allowed to settle for 3 hours and filtered. The solution was concentrated to 2 mL and stored at 8 °C to afford crystals of **2‐Ln**. **2‐La**: Colourless crystals (0.395 g, 31 %). Anal. Calcd (%) for C~34~H~58~LaCl: C, 63.69; H, 9.12; Found: C, 63.57; H, 9.30. ^1^H NMR (\[D~6~\]benzene, 400 MHz, 298 K): *δ*=1.25 (s, 18 H, C(C*H* ~3~)~3~), 1.52 (s, 36 H, C(C*H* ~3~)~3~), 6.51 ppm (s, 4 H, Cp‐*H*). ^13^C{^1^H} NMR (\[D~6~\]benzene, 100 MHz, 298 K): *δ*=31.19 (C(*C*H~3~)~3~), 32.86 (*C*(CH~3~)~3~), 34.50 (*C*(CH~3~)~3~), 34.68 (C(*C*H~3~)~3~), 114.78 (Cp‐*C*H), 138.17 (Cp‐*C*), 139.58 ppm (Cp‐*C*). FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2956 (s), 2904 (w), 2869 (w), 1461 (m), 1389 (m), 1361 (s), 1260 (s), 1241 (m), 1091 (br, w), 1016 (s), 866 (w), 797 (s), 678 (s), 590 (w), 566 (w), 551 (w), 436 cm^−1^ (w). **2‐Ce**: Orange crystals (0.651 g, 51 %). Anal. Calcd (%) for C~34~H~58~CeCl: C, 63.57; H, 9.10. Found: C, 63.60; H, 9.22. *μ* ~eff~ (Evans method, 298 K, \[D~6~\]benzene): 2.34 μ~B~. The paramagnetism of **2‐Ce** precluded assignment of its ^13^C{^1^H} NMR spectrum. ^1^H NMR (\[D~6~\]benzene, 400 MHz, 298 K): *δ*=−13.06 (s, 18 H, C(C*H* ~3~)~3~), −2.53 ppm (s, 36 H, C(C*H* ~3~)~3~); no other signals observed. FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2954 (s), 2904 (m), 2868 (w), 1460 (s), 1389 (s), 1358 (s), 1241 (s), 1165 (m), 1001 (s), 958 (m), 816 (s), 774 (s), 678 (s), 566 (w), 436 cm^−1^ (s). **2‐Pr**: Pale green crystals (0.672 g, 52 %). Anal. Calcd (%) for C~34~H~58~PrCl: C, 63.49; H, 9.09; Found: C, 63.37; H, 9.22. *μ* ~eff~ (Evans method, 298 K, \[D~6~\]benzene): 3.35 μ~B~. The paramagnetism of **2‐Pr** precluded assignment of its ^13^C{^1^H} NMR spectrum. ^1^H NMR (\[D~6~\]benzene, 400 MHz, 298 K): *δ*=−36.08 (s, 18 H, C(C*H* ~3~)~3~), −7.74 ppm (br, 36 H, *ν* ~1/2~∼750 Hz, C(C*H* ~3~)~3~); no other signals observed. FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2955 (s), 2905 (m), 2869 (w), 1460 (s), 1389 (s), 1359 (s), 1241 (s), 1166 (m), 1001 (s), 959 (m), 832 (m), 818 (s), 775 (s), 679 (s), 567 (m), 437 cm^−1^ (s). **2‐Nd**: Blue crystals (0.593 g, 46 %). Anal. Calcd (%) for C~34~H~58~NdCl: C, 63.16; H, 9.04; Found: C, 61.22; H, 9.02. The paramagnetism of **2‐Nd** precluded assignment of its ^13^C{^1^H} NMR spectrum. *μ* ~eff~ (Evans method, 298 K, \[D~6~\]benzene): 3.55 μ~B~. ^1^H NMR (\[D~6~\]benzene, 400 MHz, 298 K): *δ*=−18.95 (s, 18 H, C(C*H* ~3~)~3~), −5.58 ppm (s, 36 H, C(C*H* ~3~)~3~); no other signals observed FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2955 (s), 2907 (m), 2870 (w), 1460 (s), 1389 (s), 1358 (s), 1241 (s), 1166 (m), 1001 (s), 833 (s), 820 (s), 775 (s), 679 (s), 439 cm^−1^ (s). **2‐Sm**: Yellow crystals (0.656 g, 50 %). Anal. Calcd (%) for C~34~H~58~SmCl: C, 62.57; H, 8.96; Found: C, 60.93; H, 9.19. *μ* ~eff~ (Evans method, 298 K, \[D~6~\]benzene): 1.69 μ~B~. The paramagnetism of **2‐Sm** precluded assignment of its ^13^C{^1^H} NMR spectrum. ^1^H NMR (\[D~6~\]benzene, 400 MHz, 298 K): *δ*=−6.01 (s, 18 H, C(C*H* ~3~)~3~), 0.55 (s, 36 H, C(C*H* ~3~)~3~), 19.80 ppm (s, 4 H, Cp‐*H*). FTIR (ATR, microcrystalline): $\widetilde{\nu}$ =2956 (s), 2905 (m), 2870 (w), 1460 (s), 1389 (s), 1356 (s), 1241 (s), 1221 (w), 1166 (s), 1000 (s), 959 (m), 836 (w), 824 (s), 776 (s), 679 (s), 591 (w), 438 cm^−1^ (s). Conflict of interest {#chem201901167-sec-0008} ==================== The authors declare no conflict of interest. Supporting information ====================== As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. ###### Supplementary ###### Click here for additional data file. We thank the China Scholarship Council (studentship for J.L.) the Engineering and Physical Sciences Research Council (Doctoral Prize Fellowship to C.A.P.G. and EP/P002560/1 for F.O. and D.R.), the Ramsay Memorial Fellowships Trust (fellowship to N.F.C.) and The University of Manchester. We thank the EPSRC UK National Electron Paramagnetic Resonance Service for access to the EPR facility and the SQUID magnetometer, and the University of Manchester for access to the Computational Shared Facility. Research data files supporting this publication are available from Mendeley Data at DOI: <https://doi.org/10.17632/hgnvnjhz2v.1>.

Introduction {#s1} ============ Friedreich\'s ataxia (FRDA) is an autosomal recessive progressive hereditary neurodegenerative disorder caused by a GAA repeat expansion in the first intron of the *FXN* gene on chromosome 9 ([@B1]). The prevalence in the Caucasian populations is 2--5:100.000 ([@B2]). FRDA is characterized by early onset and progressive deterioration of the motor and sensory functions, scoliosis, cardiomyopathy, and eventually nystagmus ([@B3], [@B4]). Age at onset (AAO), clinical progression and severity are not uniform across patients, but variably correlate with the short-allele expansion size ([@B5]). Magnetic Resonance Imaging (MRI) studies have provided several insights over the damage in cerebellar, cerebral, and spinal cord areas involved in FRDA. Cerebral, cerebellar, and spinal cord involvement in FRDA has been documented and established with different MRI-based techniques. Volumetric MRI studies have shown widespread involvement of white (WM) and gray matter (GM). Atrophy has been documented in the infratentorial compartment at the level of dentate nuclei (DN) ([@B6]), peridentate WM ([@B7]), posterior cerebellar lobules, vermian cortex, as well as in the dorsal medulla ([@B7]--[@B10]) and in supratentorial GM areas ([@B6], [@B8], [@B9]). Diffusion Tensor Imaging (DTI) studies further characterized structural changes in WM revealing alterations in the cerebellar peduncles ([@B6], [@B11]--[@B16]) and in the cerebellum ([@B11], [@B13]). Alterations in the corticospinal tracts (CST) were observed at the level of subcortical pre-central WM, posterior limb of internal capsule (PLIC) ([@B11]) and in the brainstem ([@B6], [@B11], [@B13], [@B15]). Furthermore, other DTI based studies have reported alterations of the posterior thalamic radiations ([@B14], [@B15], [@B17]), optic radiations ([@B18]), and the long associative tracts ([@B11], [@B14], [@B15]). Few functional MRI (fMRI) studies have been performed in FRDA, including either motor ([@B19]--[@B23]) or non-motor tasks ([@B15], [@B24], [@B25]). These studies reported overall significant differences of activation in the posterior cerebellar lobules ([@B20], [@B22]) and in the cortical motor and sensory areas ([@B19]--[@B22]). The rationale behind the present work is the existence of a pattern of functional and structural alterations characterizing WM and GM in FRDA which correlates with specific clinical measures known as ataxia scales ([@B26]--[@B28]) routinely used to assess disease severity. Once identified, this pattern could be directly used for the implementation of longitudinal studies possibly overcoming some of the sensibility limitations recently demonstrated for the clinical scales ([@B29], [@B30]). For this reason, we designed a cross-sectional study of FRDA from a neuroimaging (Voxel-Based Morphometry, DTI, fMRI) and a clinical prospective in order to provide a composite overview of the CNS damage in FRDA versus healthy controls (HCs). In addition, we investigated the correlation between clinical functional scales and neuroimaging metrics. Materials and methods {#s2} ===================== Participants ------------ We recruited a cohort of 21 patients with a molecularly confirmed diagnosis of FRDA at the research centers "Eugenio Medea" in Conegliano/Pieve di Soligo (TV, Italy) and Bosisio Parini (LC, Italy). The recruited patients were older than 12 years and had an early onset FRDA (under 25 years old). All participants, but three, were native Italians, mostly originating from Central and North Italy. Non-Italian patients came from Albania (*n* = 2) and Germany (*n* = 1). A group of 18 age and sex matched HCs was recruited for inter-group comparison following a detailed anamnestic interview and cognitive assessment. The demographic data of patients and controls are presented in Table [1](#T1){ref-type="table"}. ###### Demographic and clinical data. **FRDA (*n* = 21)** **HC (*n* = 18)** ---------------- ----------------------------- ------------------- Gender F (%) 16 (76.19) 11 (61.11) Hand dominance 19R, 2L 18R AAV (y) 26.95 ± 10.35 (12--50) 27.05 ±9 (16--46) GAAsr 671.24 ± 210.5 (170--946) -- GAAlr 812.6 ± 225.04 (350--1230) -- AAO (y) 10.62 ± 4.58 (4--20) -- DD (y) 16.33 ± 8.82 (3--32) -- SARA 21.38 ± 7.76 (8--32) -- ICARS 52.95 18. 53 (22--84) -- FARS -ne 62.25 ± 19.37 (31.33--92.5) -- *FRDA, Friedreich\'s Ataxia, HC, healthy control GAAsr, GAA short repeat; GAAlr, GAA long repeat; SARA, Scale for the Assessment and Rating of Ataxia (0--40); ICARS, International Cooperative Ataxia Rating Scale (0--100); FARS-ne, Friedreich Ataxia Rating Scale -neurological examination (0--117)*. Ethic committee approval and patients consent --------------------------------------------- The study has been reviewed and approved by the Ethic Committee IRCCS E. Medea---Associazione La Nostra Famiglia---Bosisio Parini (LC) (Prot. No 051/11-CE) and all participants gave their written informed consent in accordance with the Declaration of Helsinki. Clinical measurement tools -------------------------- All patients underwent a complete clinical and neurological assessment. The Scale for the Assessment and Rating of Ataxias scale (SARA) ([@B28]), International Cooperative Ataxia Rating Scale (ICARS) ([@B26]), and the neurological section of the Friedreich Ataxia Rating Scale (FARS) ([@B27]) were implemented. Patients and HCs underwent a cognitive assessment specific for 2 age groups: 12--16 and 16--50 years old. The cognitive functions of the subjects aged 12--16 years were assessed by using the Wechsler Intelligence Scale for Children III (WISC-III) ([@B31]). The group of adults and older adolescents in both FRDA and HCs were assessed by using the Wechsler Adult Intelligence Scale Revised (WAIS- R) ([@B32]). Neuroimaging protocol --------------------- FRDA and HCs underwent an MRI session with a 3T Philips Achieva Scanner (Philips Medical System, The Netherlands), equipped with a digital 32-channel head coil. The acquisition protocol included a T1-weighted (T1w) high resolution sequence (TE/TR = 3.5/8 ms, flip angle 8°, SENSE factor 2, voxel-size 1 × 1 × 1 mm^3^ matrix size 256 × 256 × 160), a multi-shell diffusion MRI acquisition (15 directions at b = 300 s/mm^2^, 53 directions at b = 1,100 s/mm^2^, 8 volumes at b = 0 s/mm^2^, TE/TR = 100/8,800 ms, SENSE factor 2, SPIR fat suppression, voxel size 2.2 × 2.2 × 2.2 mm^3^, matrix size 112 × 112 × 80), a T2-weighted (T2w) fat suppressed scan (for DTI processing purposes, TE/TR = 100/4,700 ms, SENSE factor 2, voxel-size 1.5 × 1.5 × 1.5 mm^3^, matrix size 160 × 146 × 110), and a fMRI sequence (FOV = 240 × 240 mm^2^, 40 slices interleaved without gap, voxel size 2.5 × 2.5 × 3.5 mm^3^, TE/TR = 20/2,000 ms, flip angle 85°, 178 time points). fMRI motor task --------------- The fMRI protocol included a standard block design finger tapping task involving both hands. Subjects were asked to press the buttons of an MRI-compatible response-device using all the fingers in sequence from the thumb to the little finger, always starting from the thumb. Blocks lasted 20 s for each hand with 16 s inter-stimulus interval. The fMRI task was paced according to the screen commands that were provided with a regular pattern. A drawing of the right or left hand with a caption ("right hand" or "left hand") was projected on MR compatible goggles worn by the patients during the stimulus. A fixation point was projected in the inter-stimulus interval. Subjects were trained before the scan to familiarize with the projected instructions and with the hand device and to ensure comprehension of the task. Gray matter analysis -------------------- Voxel-Based Morphometry (VBM) pipeline ([@B33]) was applied to T1w images to detect morphological differences in the volumes between patients compared to HCs. Data was pre-processed with the N4 tool of ANTs ([@B34], [@B35]) to remove intensity field inhomogeneity, then the FSL anatomical pipeline was applied ([@B36]--[@B38]) to remove the skull and perform tissue segmentation, finally obtaining the Partial Volume Estimate (PVE) map of GM. A population template of GM was built among all subjects (including patients) with ANTs ([@B34], [@B39], [@B40]), then the GM PVE of each subject was non-linearly moved to template space and multiplied by the determinant of the transformation Jacobian. Voxel-wise statistics were computed with a general linear model (GLM) using FSL ([@B41]), with age, sex, and intracranial volume (ICV) in native space as covariates. The critical value for null hypothesis rejection was set at 0.05 corrected for multiple comparisons and employing the TFCE technique ([@B42]). White matter analysis --------------------- All the individual DTI data werepre-processed with Tortoise ([@B43]), taking advantage of the T2w acquisition to correct for motion and eddy current artifacts. With the same software the mono-exponential non-linear model was fitted on the data including all shells ([@B44], [@B45]) to compute the diffusion tensor and DTI derived measures to associated quantitative metrics, as fractional anisotropy (FA), axial diffusivity (AD), mean diffusivity (MD), and radial diffusivity (RD). Two study specific templates were built with DTI-TK ([@B46], [@B47]), one for HCs and one for patients. DTI-TK takes advantage of all the tensor elements to perform the registrations, delivering more accurate spatial alignments than tools based on intensity registration, especially in regions with complex fiber architecture such as the brainstem. The template of FRDA was non-linearly registered to the template of HCs, therefore then the transformations were concatenated to move the quantitative maps of FRDA subjects to the template space of HCs with a single interpolation. Statistical analysis were performed both at region of interest (ROI) and at voxel level. Voxel-wise statistics were performed for each DTI derived map with a general linear model (GLM) using age and sex as covariates, with the same approach previously described for VBM. To perform ROI level statistics, we moved the two JHU DTI-based WM atlas ([@B48]) to the common space. For each of the 43 regions defined in the atlas (see list in Table [S1](#SM1){ref-type="supplementary-material"}), we tested FA, MD, AD, and RD with a GLM comparing HCs and FRDA patients while accounting for age and sex as covariates. For the ROI level analysis, the critical threshold was set to 0.05 corrected with the Bonferroni method for multiple comparisons. Functional MRI analysis ----------------------- Data were pre-processed using SPM12 ([@B49]) and ANTs. Functional volumes were realigned using a two-step realignment process (to the first volume of the sequence and then to the mean volume), then the mean volume was co-registered to the T2w anatomical volume to correct for EPI distortions ([@B50]). The T2w volume was rigidly aligned to the corresponding T1w anatomical volume after normalization to the MNI152 space. Finally, all the transformations were combined together to avoid further interpolation errors. The combined transformation was applied to normalize the functional sequence on the MNI space, with a final voxel size of 2 × 2 × 2 mm^3^. Before the single-subject analysis, a Gaussian spatial filter, with FWHM equal to 6 mm, was applied to the functional data to increase the Signal-to-Noise-Ratio (SNR) and to deal with the residual anatomical differences between subjects. Additionally, a high-pass temporal filter with cut-off frequency of 128 s was used to correct for slow signal drifts. The GLM approach was adopted for the single-subject analysis: the design matrix included one regressor for each condition of interest: rest, right (R) hand movement, left (L) hand movement and one regressor of no interest for each as nuance factors (movement parameters of the realignment process and outlier volumes). The ARtifact Detection Tool (ART) was employed to create regressors and detect outliers ([@B51]). A volume was defined as an outlier when its displacement from the reference volume was \>2 mm or when its global mean Z-score was \>5. After the estimation of the model was completed, we created a map for each effect we wanted to test. Contrast maps between the R and the L hand movement task (R\>L and L\>R) were computed to exclude confounding factors due to the non-motor component of the task (attention, visual stimulation). These maps were subsequently used as inputs for the following GLM group analysis. A one-sample *t-*test was performed among the two groups individually to verify which anatomical regions were involved in the task, then a two-sample *t-*test was performed to verify differences between groups. All the statistics were performed with SPM12 using a threshold *p* \< 0.05 corrected for the False Discovery Rate (FDR) at voxel level. Following the same rationale of statistical analysis performed on the DTI data, in the fMRI dataset we performed a ROI-based GLM correlation analysis of beta values, including the clinical variables as predictors. However, differently from the DTI dataset, the fMRI dataset included only 2 male subjects. For this reason, the sex of patients was not included as confounding factor. For the correlation analysis we used clusters of significant activations generated by R hand and L hand movement (contrasts "R Hand \> L Hand" and "L Hand \> R Hand) in the HC group. This approach guaranteed the independence of the ROI-selection step from the data used in the GLM correlation analysis. We found eight ROIs related to the movement of the R hand only, and seven ROIs related to the movement of the L hand only. Significance was set using a *p* \< 0.05 corrected for the number of variables tested. Statistical analysis -------------------- Computed descriptive statistics mainly included mean values, modes and standard deviations of the variables. Inferential statistics on the clinical variables were performed with SPSS v24. Further statistical investigations were performed within the group of FRDA patients to assess the relation between diffusion metrics FA and MD and clinical covariates scoring disease severity. The considered covariates were the functional scales SARA, ICARS, and FARS, the short repeat triplet GAAsr and the age at onset (AAO). Thus, a battery of multi-variate GLMs was fit for each ROI, following a design scheme with intercept, age, sex, and one additional covariate at time. The significance *p*-value used to consider a covariate significant was set to 0.05 divided by the total number of considered covariates. Results {#s3} ======= Clinical results ---------------- Twenty-one patients with a molecularly defined diagnosis of FRDA were examined. The mean age at the time of the clinical assessment (AAV) was 26.95 ± 10.35 years (range 12--50, mode 26 years) (Table [1](#T1){ref-type="table"}). The mean disease duration (DD) was 16.33 ± 8.82 years (range 3--32). The gender ration was F/M: 16/5. Patients declared an AAO of about 10.62 ± 4.58 (range 4--20), with a bimodal distribution (10 and 11 years). Twenty of the patients\' cohort were homozygous for the GAA repeat expansion, and one presented a 170 GAA repeat expansion on one allele and a nonsense point mutation on the other ([@B52]). The mean GAA repeat expansion in the short-allele (GAAsr) was 671.24 ± 210.5 (range 170--946), while the long allele (GAAlr) counted for 812.6 ± 225.04 (range 350--1,230). The Spearman test correlations showed that GAAsr correlated negatively with the AAO (*R*^2^ = −0.692, *p* = 0.001) and positively with clinical severity scores SARA (*R*^2^ = 0.714, *p* = 0.000^\*^) and ICARS (*R*^2^ = 0.690, *p* = 0.001). The DD correlated positively to all the clinical scales SARA (*R*^2^ = 0.529, *p* = 0.014), ICARS (*R*^2^ = 0.582, *p* = 0.006), and FARS-ne (*R*^2^ = 0.767, *p* = 0.016). The clinical features of the patients at onset, and the involvement of the CNS, sensory and other systems derived from the neurological examination, the ataxia scales and other complementary assessments are shown in Table [S2](#SM2){ref-type="supplementary-material"}. Neuroimaging findings --------------------- ### Gray matter analysis VBM showed a significant bilateral reduction of GM volume in lobule V, VI, VIII (L\>R) and in the crus of cerebellum in FRDA patients compared to HCs. A significant volume reduction was observed also in the posterior lobe of the vermis, in both the flocculi and in the L tonsil (Figure [1](#F1){ref-type="fig"}). No differences between patients and controls were found in the supratentorial GM. {#F1} ### White matter analysis Voxel-wise analysis of DTI derived maps showed a diffuse reduction of FA values in both infratentorial and supratentorial WM of FRDA patients (Figure [2](#F2){ref-type="fig"}). In particular FA reduction involved deep cerebellar WM, inferior (ICP), middle (MCP), and superior (SCP) cerebellar peduncles (including the decussation in the brainstem), CST both in the brainstem, in the PLIC and in the subcortical WM close to motor cortex, posterior thalamic radiations and optic radiation, and corpus callosum (CC). Optic tracts were also partially involved. The same areas showed increased MD, AD, and RD values in FRDA patients, with a more diffuse pattern for RD (Figure [3](#F3){ref-type="fig"}). {#F2} {#F3} The ROI-based analysis showed similar results. Significant FA reductions in FRDA patients were found in ROIs representing CSTs, forceps major, CC, posterior thalamic radiations, SCP, MCP, and ICP. The same areas showed increased MD and RD values, while AD was increased only in MCP and SCP (Table [2](#T2){ref-type="table"}). ###### DTI metrics distribution in the White Matter in FRDA and HCs and the intergroup variation. **TRACTS** **Mean FA \[μm2/s\]** **Mean MD \[μm2/s\]** **Mean AD \[μm2/s\]** **Mean RD \[μm2/s\]** -------------------------------- ----------------------- ----------------------- ----------------------- ----------------------- ------- ------ ------- ------- ------ ------- ------- ------ SCP L 0.537 0.349 −34.9 3.065 4.400 43.6 1.667 1.958 17.4 0.699 1.221 74.7 SCP R 0.550 0.369 −33.0 2.984 4.099 37.4 1.646 1.865 13.3 0.669 1.117 66.9 ICP R 0.501 0.390 −22.2 2.164 2.607 20.5 . . . 0.505 0.682 35.0 ICP L 0.497 0.395 −20.5 2.190 2.605 19.0 . . . 0.514 0.679 32.2 Forceps major 0.506 0.447 −11.8 2.496 2.786 11.6 . . . 0.570 0.682 19.6 Body of CC 0.627 0.574 −8.6 . . . . . . 0.447 0.552 23.6 Posterior thalamic radiation L 0.560 0.515 −8.0 2.209 2.407 9.0 . . . 0.480 0.557 16.1 CST R 0.510 0.467 −8.5 2.187 2.356 7.7 . . . 0.509 0.575 13.0 Posterior thalamic radiation R 0.561 0.517 −7.8 2.165 2.363 9.1 . . . 0.469 0.543 15.8 MCP and pontine crossing tract 0.490 0.453 −7.5 2.066 2.289 10.8 1.087 1.157 6.5 0.490 0.566 15.6 CST L 0.516 0.478 −7.4 . . . . . . 0.496 0.551 11.2 Splenium of CC 0.654 0.614 −6.1 2.453 2.703 10.2 . . . 0.465 0.553 18.9 *GLM p \< = 0.05 Bonferroni on FA. FA, Fractional anisotropy; MD: medial diffusivity; RD, radial diffusivity; HCs, healthy controls, FRDA, Friedreich\'s ataxia patients; SCP, superior cerebellar peduncle; L, left; R, right; ICP, inferior cerebellar peduncle; MCP medial cerebellar peduncle; CC, corpus callosum; CST, cortical spinal tract*. No differences emerged at the level of association tracts like superior or inferior longitudinal fasciculi, uncinate fasciculi, or inferior fronto-occipital fasciculi. The relative change of DTI-derived measures varied among significant tracts. The SCPs showed the most important alterations, with more than 30% FA reduction and more than 35, 13, and 65% increments in MD, AD, and RD values respectively (Table [2](#T2){ref-type="table"}). The ICPs were the second most-affected tracts with a FA reduction around 20% and an increment of MD and RD of more than 18 and 32%, respectively. Other significant tracts showed less severe modifications of DTI measurements. WM and clinical data correlation -------------------------------- Table [3](#T3){ref-type="table"} shows significant correlations between DTI-indexes (FA and MD) and clinical variables. Negative and positive correlations were found for FA and MD values. The AAO correlated with FA and MD in both supratentorial (thalamic radiations, CC, forceps, fornix, internal capsule) and infratentorial (ICP) tracts. Clinical scales and the number of GAA repetitions were negatively correlated with FA and positively correlated with MD in SCP, forceps, and fornix. ###### Significant coefficients (β) of the GLM regression of DTI metrics with clinical covariates. **Description** **AAO(β)** **GAAsr(β)** **ICARS(β)** **SARA(β)** **FARS-ne (β)** --------------------------------------------------------------- ------------ -------------- -------------- ------------- ----------------- **FA** Anterior thalamic radiation L 0.0035 -- -- -- -- Anterior thalamic radiation R 0.0032 -- -- -- -- Posterior thalamic ratiation and retrolenticular part of IC R 0.0050 -- -- -- -- Posterior thalamic ratiation and retrolenticular part of IC L 0.0060 -- -- -- −0.0010 Forceps major 0.0059 −0.0001 −0.0013 −0.0029 -- Inferior fronto--occipital fasciculus R 0.0039 -- -- -- -- Body of CC 0.0071 -- -- -- -- Anterior limb of IC L 0.0049 -- -- -- -- Fornix (column, body, cres with stria terminalis) 0.0051 -- -- −0.0028 -- SCP R -- −0.0001 −0.0016 −0.0040 −0.0016 SCP L -- −0.0001 −0.0012 −0.0032 −0.0012 **MD** Anterior thalamic radiation L −0.0283 -- -- -- -- Anterior thalamic radiation R −0.0332 -- -- -- -- Posterior thalamic ratiation and retrolenticular part of IC L −0.0347 -- -- -- -- CST R −0.0280 -- -- -- -- Forceps major −0.0360 -- -- 0.0181 -- Inferior fronto--occipital fasciculus R −0.0200 -- -- -- -- Superior longitudinal fasciculus L −0.0229 -- -- -- -- Superior longitudinal fasciculus R −0.0200 -- -- -- -- Splenium of CC −0.0322 -- -- -- -- Fornix (column, body, cres with stria terminalis) -- -- -- 0.0220 0.0095 ICP R −0.0322 -- -- -- -- SCP R -- -- 0.0137 0.0334 0.0132 SCP L -- -- 0.0120 0.0313 0.0119 *p \< 0.05/nr of covariates. FA, Fractional anisotropy; MD, medial diffusivity; FRDA, Friedreich\'s ataxia patients; SCP, superior cerebellar peduncle; L, left; R, right; ICP, inferior cerebellar peduncle; CST, cortical spinal tract; AAO, age at onset; GAAsr, GAA short-allele; SARA, Scale for the Assessment and Rating of Ataxia (0--40); ICARS, International Cooperative Ataxia Rating Scale (0--100); FARS-ne, Friedreich Ataxia Rating Scale -neurological examination (0--117)*. fMRI analysis ------------- Due to excessive motion artifacts and collaboration problems during the acquisition, 7 FRDA and 3 HCs were excluded from the following fMRI analysis. The FRDA group (*n* = 14) mean age was 27.6 ± 11.1 years (range 12.1--50.5) and gender ratio 12F/2M. The HCs (*n* = 15) group had a mean age 27.9 ± 9.8 years (range 15.9--45.7) and gender ratio 10F/5M. All the FRDA and HCs were R handed. From intragroup analysis, in both FRDA and HCs, the multi-finger tapping task activated similar brain regions. In particular, while moving the R fingers (contrast R\>L), significant activations were found in the L M1, L insula, and R superior cerebellar hemisphere (lobules V, VII, VIII). During the movement of L fingers (contrast L\>R), activations involved R M1 cortex, R insula, and L superior cerebellar hemisphere (lobules V, VI, VIII) (Figure [4](#F4){ref-type="fig"}). {#F4} From intergroup analysis, significant differences emerged only during the movement of the non-dominant hand (left-hand), with HCs showing a stronger activation (cluster size k = 42 voxels, *t* \> 5.174, *p* \< 0.05 FDR corrected) in the L superior cerebellar hemisphere (Figure [5](#F5){ref-type="fig"}). No significant differences were found during the movement of the dominant hand. {#F5} fMRI and clinical data correlation ---------------------------------- Table [4](#T4){ref-type="table"} shows the significant correlations between fMRI activation magnitude and the clinical parameters in the FRDA group. Significant positive correlations (*p* \< 0.05/nr of variables) were found between AAO and activation in the cerebellar anterior lobe, insula, motor cortex, and temporal lobe. Negative correlations (*p* \< 0.05) that did not survive to multiple comparisons correction, were found between disease severity measures (ICARS, SARA; FARS-ne), GAAsr and fMRI activation in the cerebellum, insula, and temporal lobes. Task-related cerebral activation generated more correlation during the R-hand task (dominant hand). ###### Correlations between fMRI activation and clinical covariates. **AAO (β)** **GAAsr (β)** **ICARS (β)** **SARA (β)** **FARSne (β)** **DD (β)** ------------------------------------------------- ----------------------------------------- --------------- --------------- -------------- ---------------- ------------ **RIGHT-HAND TASK** Motor Cortex L 1.0139[^\*^](#TN1){ref-type="table-fn"} -- -- -- -- -- Middle temporal lobe 0.8804[^\*^](#TN1){ref-type="table-fn"} −0.622 −0.460 −0.465 −0.516 -- Insula L 1.0634[^\*^](#TN1){ref-type="table-fn"} −0.647 −0.671 −0.599 −0.620 -- Posterior Cingulum R 0.858 -- -- -- -- -- Cuneus R 0.904 -- -- -- -- -- Cerebellar anterior lobe (lobules V--VI) R 0.9486[^\*^](#TN1){ref-type="table-fn"} -- −0.602 -- -- -- **LEFT-HAND TASK** SMA R -- -- -- -- -- −0.540 Cerebellar anterior lobe (lobules V--VI) R 0.9598[^\*^](#TN1){ref-type="table-fn"} −0.739 −0.648 −0.625 −0.587 -- Cerebellar posterior lobe (lobules VII--VIII) L 0.744 −0.758 −0.605 −0.647 −0.586 -- p \< 0.05; *p \< 0.05/nr of covariates. SMA, supplementary motor area; L, left; R, right; AAO, age at onset; DD, disease duration; GAAsr, GAA short-allele; SARA, Scale for the Assessment and Rating of Ataxia (0--40); ICARS, International Cooperative Ataxia Rating Scale (0--100); FARS-ne, Friedreich Ataxia Rating Scale -neurological examination (0--117)*. Cognitive assessment results ---------------------------- Nineteen patients underwent the cognitive assessment (Table [S3](#SM3){ref-type="supplementary-material"}). The protocol was not administered to two patients, as they were non-Italian speakers. The mean level of education was 12.84 years (SD ± 2.87). The intelligence quotient (IQ) level was normal in 13 FRDA (68.41%), borderline in 4 FRDA and in the intellectual disability range in 2 subjects (10.52%). In 4 out of 6 FRDA patients with reduced IQ, the verbal performance was more strongly impaired. In 10 cases, FRDA patients showed a disharmonic profile, with better verbal component. The IQ level components fell into the normal range in all the HCs. Considering the IQ level as index of cognitive functioning and using an intellectual disability cut off of total IQ \< 70; we found only 2 patients with cognitive deficits. Due to the unbalanced samples of patients with vs. without cognitive deficits, we did not perform group comparison of DTI and fMRI indexes. We tested the occurrence of any correlation between the IQ values and the DTI and fMRI metrics, but did not find any significant correlation. Discussion and conclusions {#s4} ========================== Our study demonstrates the extent of CNS brain damage in FRDA by using a composite protocol of clinical and multimodal neuroimaging tools as VBM, DTI and fMRI in a cross-sectional study. Our findings support the need for future longitudinal studies and highlights the possibility that MRI studies could provide valuable paraclinical biomarkers in FRDA. Voxel-based morphometry in the gray matter ------------------------------------------ By using the VBM technique we demonstrated infratentorial GM volume reduction, particularly in the lobules V, VI, VIII, crus of cerebellum, posterior lobe of vermis, flocculi bilaterally, and L tonsils. These results are consistent with previous studies ([@B6], [@B8], [@B9], [@B15], [@B53]) (Table [S4](#SM4){ref-type="supplementary-material"}). We did not find any supratentorial GM volume reduction or DN alteration. This is in line with reports of no cerebral atrophy in FRDA ([@B11], [@B53]), although opposite results have also been reported ([@B6], [@B8], [@B15], [@B54]). Our findings support previous investigations reporting involvement of the anterior vermis and of the anterior cerebellar lobes ([@B6], [@B11]), otherwise known as the primary sensorimotor regions of the cerebellum. Yet, we describe for the first time GM loss in FRDA in the lobule VIII, also known as the secondary sensorimotor cerebellum ([@B55]), in the flocculi bilaterally and in the left tonsil. This morphometric distribution of GM changes implies the involvement of both the cerebellar primary and secondary body maps in FRDA. In addition, the GM reduction in the flocculi, in the lobules V--VI and in the dorsal vermis may provide the structural explanation for the eye movements impairments (Table [S2](#SM2){ref-type="supplementary-material"}) ([@B56]), as it was demonstrated in other neurodegenerative ataxias such as SCA6 ([@B57]) where the floccular atrophy was linked to gaze holding and pursuit impairment. Microstructural changes in the white matter ------------------------------------------- Voxel-wise and ROI-based analysis of DTI metrics reported diffuse microstructural changes in infra- and supratentorial WM. These findings are in line with previous DTI reports of alterations in the cerebellar peduncles ([@B6], [@B11]--[@B16]) (Table [S4](#SM4){ref-type="supplementary-material"}) and of the supratentorial involvement in areas such as CC, forceps major, posterior thalamic radiations, optic tracts, and optic radiations ([@B6], [@B14], [@B15], [@B18], [@B58], [@B59]). In particular, we observed significantly different FA and MD values between FRDA and HCs in the CST, as consistently reported in the literature ([@B6], [@B11], [@B13], [@B15]). The intergroup variation of the four DTI metrics pointed toward a primary involvement of the SCPs and, to a less extent, of the ICPs. The infra- and supratentorial FA and MD values correlated with AAO in FRDA. In addition, the disease severity scored with SARA and FARS showed a strong correlation with FA and MD values in SCP and forceps major. These findings support previous correlating DTI metrics and disease severity scores in areas as the cerebellar peduncles and the CC ([@B6], [@B7], [@B13], [@B16]). Although other studies denied such correlations, we presume that this difference was due to the different techniques employed and study samples ([@B12], [@B14], [@B59], [@B60]). We did not observe any structural alteration in the associative tracts. Others have reported FA reduction in the inferior fronto-occipital fasciculus ([@B11], [@B14], [@B15]) and in the inferior longitudinal fasciculus ([@B11], [@B15]). However, the FRDA intragroup FA and MD values correlated with the AAO in the inferior fronto-occipital and superior longitudinal fasciculi, suggesting sensitivity of this measure to damage over time. The relevance of the consistent variation of the different DTI-indexes, in particular localized in the cerebellar peduncles and forceps major is strongly supported by the correlation with the different disease severity measures and with AAO. Motor function exploration -------------------------- We observed similar areas of activation in both groups from the intragroup analysis of fMRI data with bi-manual finger tapping task. During the R-hand motor task (dominant hand in all the subjects) we observed significant activation in L M1, L insula, and the R superior cerebellar hemisphere (lobules V, VI, VIII). From the L-hand motor task, we observed significant activations of the R M1, R insula, and L superior cerebellar hemisphere (lobules V, VII, VIII). The areas activated in the cerebellum with R and L hand during the motor task were comparable to the areas found in other studies ([@B20]--[@B23]) (Table [S4](#SM4){ref-type="supplementary-material"}). Our intergroup analysis showed no significant differences during the dominant-hand task, differently from other studies that have reported either reduction of activation in M1, thalamus, DN, and lobule V ([@B21], [@B22]) or increase of activation in parietal cortex, striatum, supplementary motor area, and lobule VII ([@B20]--[@B22]) Interestingly, we found a significant difference during the non-dominant hand motor task with a stronger activation in the L superior cerebellar hemisphere in HCs when compared to FRDA. This difference was not previously described. One possible explanation is that while the dominant-hand is controlled by the dominant hemisphere, the non-dominant hand is controlled by both hemispheres ([@B61], [@B62]). The intergroup differences in our study might be explained by probable disruption in FRDA of the cerebellar circuitry involved in the primary somatosensory map in the cerebellum. In our study, we propose a bimanual device-mediated finger tapping task, while other studies have performed only dominant-hand motor tasks ([@B19]--[@B22]). The correlation analysis demonstrates that the cerebral and cerebellar functionality can be related to clinical and genetic factors. Predominantly, the activation of the anterior cerebellum is directly correlated to AAO, GAAsr, and severity measures as reported elsewhere ([@B4], [@B23]). Although the FRDA group had a lower IQ than HCs, the mean IQ level of both groups was within normal range, similarly to the results reported in Mantovan et al. ([@B19]). The lack of correlation between results of the neuropsychological assessment and either fMRI, VBM, or DTI data is not surprising since the finger tapping task was not constructed to explore cognitive relevant performances and the areas found impaired with VBM and DTI techniques are not primarily engaged in cognitive functions. Our findings derive from a 3 Tesla scanner. This magnet strength was widely used in other studies ([@B6], [@B8], [@B9], [@B15], [@B17], [@B20], [@B24], [@B25]) (Table [S4](#SM4){ref-type="supplementary-material"}) and perhaps is sufficient to detect the CNS structural and functional damage in FRDA. Only two other works have tried a higher magnet field of 7 Tesla ([@B22], [@B54]) but their findings do not provide more insight. Limitations in our study are the relatively small sample size, which impacts on the statistical strength of the study, the heterogeneity in disease severity of the recruited subjects and the lack of a longitudinal appraisal of both clinical and neuroimaging assessments. Few studies have recruited more patients, but the sample size was generally not remarkably different. The difficulty in recruitment of larger cohorts could be overcome by sharing common protocols in multiple centers and pooling the results. A step in this direction is already taking place with the ENIGMA network (*[http://enigma.ini.usc.edu/](http://http://enigma.ini.usc.edu/)*) which promotes neuroimaging data sharing for rare genetically determined diseases. There have been few longitudinal studies in FRDA. Two longitudinal studies reported DTI changes as FA reduction in CC ([@B6]) and AD changes in CC splenium and WM deep subcortical parietal lobes after a follow-up period ([@B59]) with no macrostructural changes or atrophy. Two more longitudinal interventional studies have tried to investigate the effect of rhuEPO treatment in FRDA and showed increase in the GM volume in pulvinar and posterior parietal cortex ([@B63]) and WM bilateral increase in FA and increase in AD in cerebral hemispheres bilaterally ([@B60]). However, more longitudinal studies are needed in order to validate the cross-sectional study derived findings. This would lead to a better definition of the significance and of the temporal dynamics of the observed alterations. In such contexts, the neuroimaging study could provide the sensitive and objective biomarkers very much needed for the rapid and efficient clinical trial design and results evaluation. In summary, we demonstrated infratentorial GM volume reduction suggesting an alternative sensorimotor cerebellar map involvement in FRDA with no supratentorial GM involvement. This finding shows the extension of cerebellar damage in FRDA. The microstructural WM findings were consistent with the known areas of CNS damage and correlated with the clinical measures. The established correlation between DTI metrics and clinical variables as the AAO or the disease severity scores, should now be validated in larger cohorts and in longitudinal studies, which then could support the use of these measures as disease biomarkers in future studies. The non-dominant hand motor task differences between the FRDA and HCs gives hints on cerebral-cerebellar circuitry disruption in both hemispheres. In conclusion, our multimodal imaging study provided convergent results, with a strong involvement of the cerebellar cortex, cerebellar WM tracts, in particular SCPs and ICPs and a strong functional involvement of the anterior lobe of the cerebellum during the non-dominant hand motor task. These findings bring a new dimensional role of the cortical circuitry involved in FRDA. Authors note {#s5} ============ Paper previously presented by MV in part as a scientific poster at the 1st Congress of the European Academy of Neurology, June 20--23, 2015, Berlin, Germany. The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher. Author contributions {#s6} ==================== MV provided the design of the study, supervision, collected and analyzed the data, and prepared the manuscript. FA provided the design of the study and analyzed the data. AN and AD analyzed the data. SP provided the data and analyzed the data. EP, GP, MD, EB, ER, MF, and PC provided the data. AM provided the design of the study, supervision, and financial support. All authors revised the manuscript for important intellectual content, provided approval for publication and agree to be accountable for all aspects of the work. 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. We are very grateful to patients who participated in this study and their families. We thank the MRI technicians Giancarlo Lembo, Luca Brenz, and the nurse Lucia for technical assistance during the acquisition of the MRI scans. This work was supported by the Italian National Institutes of Health \[Grant Ricerca Corrente 2011\] and by Ogni Giorno per Emma ONLUS Association. Supplementary material {#s7} ====================== The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fneur.2018.00747/full#supplementary-material> ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. [^1]: Edited by: Kathrin Reetz, RWTH Aachen Universität, Germany [^2]: Reviewed by: Laura Canafoglia, Istituto Neurologico Carlo Besta (IRCCS), Italy; Jordi A. Matias-Guiu, Hospital Clínico San Carlos, Spain [^3]: This article was submitted to Applied Neuroimaging, a section of the journal Frontiers in Neurology [^4]: †Present Address: Alberto De Luca, Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands

Introduction {#S1} ============ In human and nonhuman primates, the face plays an important role in social communication ([@R10]; [@R49]; [@R64]). In some species, such as macaques (*Macaca mulatta*, *M. fuscata*), and mandrills (*Mandrillus sphinx*), facial morphology may signal fertility ([@R21]; [@R53]; [@R56]). The face may also provide cues to health in humans ([@R31]; [@R36]; [@R61]) and rhesus macaques ([@R42]). In mandrills and drills (*Mandrillus leucophaeus*), males with stronger facial color saturation tend to be higher ranking ([@R43]; [@R57]). In rhesus macaques, male facial coloration may indicate mate quality ([@R22]; [@R63]) and is linked to competition for mating opportunities ([@R51]), but is not related to dominance status ([@R32]). In addition to coloration, other aspects of facial morphology may play a role in signaling social status. One such measure is the facial width-to-height ratio (fWHR), which measures bizygomatic width---the distance between the zygomatic arches---divided by superior facial height (mid-face height, i.e., nasion-prosthion) (see [Fig. 1](#F1){ref-type="fig"}). Sexual dimorphism in fWHR was inversely related to dimorphism in canine size across 14 primate species ([@R71]), suggesting that weak dimorphism in canine size is not due to low sexual selection, but instead to selection for sexually dimorphic face width. The authors theorized that sexual dimorphism in human fWHR might be driven by female mate choice for larger cheek bones ([@R72]), a feature indicative of facial attractiveness ([@R18]; [@R19]), although other facial features may be stronger attractiveness indicators ([@R45]). Mate choice might not be the only driver of sexual dimorphism; fWHR may be a signal in male-male competition ([@R12]), indicating differences in dominant and aggressive tendencies that could be linked to facial morphology through underlying differences in testosterone levels ([@R39]; [@R47]). Since these reports, additional studies have established that fWHR is not sexually dimorphic in humans ([@R37]; [@R38]; [@R48]); however, fWHR has been linked to aggressiveness and fighting ability in human males ([@R4]; [@R30]; [@R62]; [@R76]). Wider male faces are also perceived as being more aggressive ([@R2]; [@R39]; [@R44]; [@R58]), suggesting that relative face width is a social cue to aggressive and assertive behavior. While a meta-analysis of 19 studies (*N* \> 4000) found support for the link between fWHR and aggression ([@R30]), a recent study in a large sample of humans (*N* \> 137,000) found little evidence for any link between fWHR and self-reported behaviors such as impulsiveness ([@R35]). Moreover, researchers have conducted most human studies of fWHR in western, educated, industrialized, rich, and democratic populations ([@R33]), which are largely socially monogamous and not representative of the diversity among human mating and societal systems. Given these limitations and the debates within the current literature, returning to the roots of this literature and considering the role of fWHR in nonhuman primates could improve our understanding of the potential role of human facial morphology in mate choice and mate competition. Doing so provides two specific benefits. First, examining fWHR links to dominance in other primate species could allow a better appreciation of the social selection pressures that led to this association, particularly regarding sex differences. Examining links to dominance behavior among species with varying levels of fWHR dimorphism could provide a stronger ecological basis for understanding the fWHR-dominance relationship found in humans. Second, as human studies often rely on self-reported measures or proxies of dominance ([@R35]), by examining nonhuman primates we can assess not only trait ratings of assertive behavior but also specifically rank, which could be more biologically relevant for understanding behavioral links to fWHR. We also examined a second facial metric, the facial lower-height-full-height ratio (fLHFH), which is introduced after fWHR. Given the inverse relationship between canine height sexual dimorphism and fWHR dimorphism ([@R71]), brown capuchins (*Sapajus apella*) proved an ideal candidate for exploring fWHR links to dominance, given their low sexual dimorphism in canine size, similar to humans, but apparent dimorphism in fWHR ([@R40]; [@R71]). Findings in brown capuchins revealed that in adults of both sexes, fWHR was positively related to ratings of Assertiveness, and furthermore, fWHR was higher among alpha individuals ([@R40]). Whether the dimensions of the capuchin face are a social cue of assertiveness is still debated ([@R74]). However, as has been suggested in humans, a higher fWHR could prove advantageous in male-male competition, possibly through links to stronger bite force or more robust skull structure ([@R40]). Research in the *Macaca* genus supports the theory that fWHR is a cue to fighting ability ([@R9]). Across 11 macaque species, those with despotic dominance styles, such as in rhesus macaques, had higher fWHR in both sexes compared to more socially tolerant species, such as Tonkean macaques (*Macaca tonkeana*). These findings suggest that face width could be a signal of aggressive tendencies, particularly in females, that reduces the need for conflict within species for which escalated conflict could have serious consequences ([@R9]). This result fits with findings that rhesus macaques can differentiate human faces of varying fWHR, looking longer at faces with lower fWHR ([@R17]). In both humans ([@R28]; [@R47]; [@R69]) and capuchins ([@R11]) it has been suggested that the relationship between fWHR and aggressive behavior is driven by low social status, as it is significant only among low-status individuals. In brown capuchins, for example, although higher ranking individuals are typically higher in Assertiveness, correlations between fWHR and Assertiveness are significant only in non-alpha individuals ([@R11]). Given that group members are typically aware of which is the highest ranking member of their group, there may be no need for high-ranking individuals to physically advertise dominance, as social knowledge obviates the need for this. This hypothesis proposes that low-status individuals are not necessarily low in the Assertiveness personality dimension. Assertiveness is a construct of multiple assessments of behavior, and tends to capture, among other descriptors, how independent, submissive, bullying, and manipulative ([@R68]) an individual is. While the current literature on facial morphology provides insights into the social role of physical features, to date, investigations of social behavior and physical features have focused on only a few species. In the current study, we expand this line of research to focus on an Old World monkey species. Given the links between fWHR and female social tolerance across the *Macaca* genus ([@R9]), we aimed to explore whether this ratio is linked specifically to dominance behavior in a despotic macaque species ([@R60]), rhesus macaques. In contrast to brown capuchins, which have low canine dimorphism and higher fWHR dimorphism ([@R40]; [@R71]), rhesus macaques exhibit both medium fWHR and canine height dimorphism, as well as being relatively more despotic with high levels of intragroup aggression ([@R59]), at least among females, factors that make them a useful comparison species for exploring links between fWHR and dominant/aggressive behaviors. This study should therefore help to build a bigger picture, across the primate lineage, of what factors might drive fWHR as a cue for dominance. In two samples of captive rhesus macaques we studied the relationships between fWHR and two different measures of dominance: 1) hierarchical dominance status measured using normalized David's scores and 2) a rater-derived personality dimension, Assertiveness, which assesses overall tendencies toward assertive and aggressive behavior, rather than dominance status. If rhesus macaques parallel brown capuchins, fWHR might be positively related to ratings of Assertiveness, similar to links to despotism ([@R9]), with higher fWHRs found among individuals with higher social rank. Moreover, if there is a relationship between fWHR and Assertiveness, and the association is driven by low social status, then we could find a significant association between fWHR and dominance status only among individuals with lower social status, but not higher social status, as measured by normalized David's scores ([@R40]). Following earlier work on personality and facial morphology ([@R73]), we also simultaneously examined fWHR associations with ratings on five other personality dimensions, labeled Confidence, Openness, Friendliness, Activity, and Anxiety ([@R68]), a conservative approach that allowed us to control for covariance between dimensions. Brown capuchin males have higher fWHR than females, although this is particularly driven by mature, alpha males that have even higher fWHRs ([@R40]). Similarly, the link between fWHR and behavior is found predominantly among human males ([@R50]; [@R71]), so it would be consistent for fWHR associations in rhesus macaques to be driven by males. If fWHR--personality associations are driven by intrasexual selection, then given low male--male competition in rhesus, we might not find similar male-driven effects. Moreover, rhesus macaques of both sexes demonstrate significant facial skeletal growth past the point of reproductive maturity ([@R65]), so if macaques parallel capuchins, we might also see differences in the associations between skeletally mature and immature macaques (over and under age 8 yr., respectively). Although not as widely studied as fWHR, fLHFH has previously been used as part of a masculinity index in human males to study facial attractiveness ([@R50]). Unlike fWHR, in humans fLHFH is sexually dimorphic, with males having higher fLHFH ratios than females ([@R50]), which likely captures a longer, lower face in males ([@R55]). Human jaw size is sexually dimorphic, with more prominent jaw bones being rated as more attractive in male faces ([@R45]), suggesting that preference for males with larger jaws may have driven sex differences in face height. Contrastingly, fLHFH was not found to be sexually dimorphic in brown capuchin monkeys ([@R73]), which calls into question what selection pressures could drive species differences in mate preference for jaw size. Whether, and why, this metric is sexually dimorphic in other primate species requires exploration. Also unlike fWHR, fLHFH has been associated with higher ratings of Neuroticism and lower ratings of Attentiveness in brown capuchins of both sexes, traits strongly related to vigilance behavior ([@R46]; [@R73]). These traits may be operationalized as one form of social status, prosocial competence ([@R41]; [@R73]), which manifests as policing behavior ([@R25]), suggesting that facial height may also play a role in social cues. The mechanisms that underlie links between fLHFH and vigilance or policing behavior, however, are poorly understood, especially given that in capuchins, this potential cue to status does not differ by sex. It is possible that this morphological feature shares underlying variance with hormones that also drive social attentiveness. One such candidate could be testosterone, which has been implicated in vigilance to social threat ([@R24]). The role of fLHFH in social behavior thus warrants further investigation, to examine whether similar effects occur in other species. As with fWHR, if rhesus macaques parallel brown capuchins ([@R73]), we might find that ratings of lower Confidence or higher Anxiety would be most likely to be associated with higher fLHFH ratios, but we might not find a relationship between face height and David's scores. We additionally aimed to examine the relationship between 1) fLHFH and ratings on the personality dimensions Confidence, Openness, Assertiveness, Friendliness, Activity, and Anxiety. We also examined the relationship between 2) fLHFH and normalized David's scores. Because we had access to a large number of independent variables, particularly personality dimensions, and wished to carry out multiple analyses within subgroups, our analyses for both fWHR and fLHFH were primarily exploratory. Methods {#S2} ======= Samples {#S3} ------- We studied two samples of rhesus macaques: 65 (34 male) housed in 5 social groups at the ONPRC and 44 (13 male) rhesus macaques housed in 3 social groups at the CNPRC. The mean age of the ONPRC sample was 5.1 yr. (SD = 2.7, range: 2--11.6 yr) and the mean age of the CNPRC sample was 8.1 yr. (SD = 4.9, range: 1--21.1 yr). Personality Ratings {#S4} ------------------- We used two versions of the Hominoid Personality Questionnaire (HPQ) ([@R67]) to assess personality (see <http://extras.springer.com/2011/978-1-4614-0175-9/weiss_monkey_personality.pdf>): a 12-item version and the full 54-item version. Questionnaire-based approaches to personality assessment have previously demonstrated interrater and test--retest reliability ([@R27]) as well as behavioral validity ([@R23]; [@R46]). The HPQ has been used to assess personality across multiple species of apes, Old World, and New World monkeys ([@R66]). Each HPQ item consists of an adjective followed by one to three sentences describing that adjective in the context of monkey behavior. For example, "FEARFUL: Subject reacts excessively to real or imagined threats by displaying behaviors such as screaming, grimacing, running away, or other signs of anxiety or distress." Individuals familiar with the studied animals rate each monkey on each item using a 7-point Likert scale, where 1 indicates "Displays either total absence or negligible amounts of the trait" and 7 indicates "Displays extremely large amounts of the trait." We instructed raters not to discuss their ratings with each other. Five staff members and one researcher (LMR) rated the ONRPC macaques on the 12-item version of the HPQ. The 12-item HPQ (hereafter, short form) covered five of the six rhesus macaque personality dimensions: Confidence, Anxiety, Openness, Assertiveness (referred to as Dominance in [@R68]), and Friendliness. One CNPRC staff member and (LMR) rated the CNRPC macaques on the 54-item HPQ, which covered the same dimensions as the 12-item HPQ with the addition of the Activity dimension ([@R68]). We calculated unit-weighted component scores ([@R29]) using the published six component structure ([@R68]). Because the short form was a subset of the full HPQ, we calculated both full- and short-form component scores in the CNPRC sample, and we built full HPQ and short-form models separately where appropriate. Dominance Status {#S5} ---------------- LMR observed 41 of the studied ONPRC macaques and all 45 CNPRC macaques using focal observations ([@R3]) as part of research on rhesus macaque health and welfare (see Robinson et al. 2019 for full details). Of the eight groups, six groups were observed using focal animal observation (*N* = 85; see [Electronic Supplementary Material \[ESM\] Table SI](#SD1){ref-type="supplementary-material"} for individual group characteristics) resulting in some missing data (e.g., normalized David's scores) for the 24 macaques housed in the remaining 2 groups. A complete description of the observed ONPRC and CNPRC samples is available in Robinson et al. (2019). We observed the macaques for multiple behaviors including supplant behaviors, which we defined as: focal macaque is touched by a conspecific and the focal macaque moves and conspecific may or may not take the focal macaque's spot, or focal macaques moves in response to a conspecific's touch. For the full ethogram see [Supplementary Table \#1](#SD1){ref-type="supplementary-material"} in Robinson et al. (2019). Across the two facilities, we observed each macaque for a mean of 224.09 ± SD 57.22 minutes. We then used these data to create directional supplant matrices for each rhesus macaque group, which we then used to calculate each macaque's normalized David's score ([@R20]). We pooled both sexes into the same dominance hierarchies, and we investigated sex differences via interactions in regression models to preserve sample size, rather than splitting males and females. Age Categories {#S6} -------------- Significant physical changes in body size ([@R7]) and facial morphology ([@R13]) occur as rhesus macaques sexually mature and become adult. Although sexual maturity occurs at ages 3 yr. and 4 yr. in females and males respectively, skeletal maturity does not occur until age 8 in both sexes ([@R13]; [@R65]). We divided our sample into younger (81 individuals \<8 yr) and older (28 individuals \>8 yr) subsamples to determine if the associations we investigated were unique to one age group but not the other ([@R65]). The younger age group included all individuals who may still experience developmental changes in the skeletal structure of their faces, whereas the older group contained only mature individuals who are no longer growing. In some analyses, we examined these subgroups in addition to the entire sample that encompasses both age groups. Facial Measurements {#S7} ------------------- We photographed macaques' faces after we completed the focal observations for each project, either while the macaques were in their social groups or during regularly scheduled sedation as part of health checks using the same photography method each time. We took multiple photographs of neutral expressions for most individuals (mean = 2.38 ± SD 1.39), up to a maximum of 7; 20 individuals were represented by only one photograph. We did not use photos if key facial features were not visible, or if the individual's face was not parallel and ventrally head-on to the camera; we measured 238 usable photographs. We defined facial metrics with a total of seven points ([Fig. 1](#F1){ref-type="fig"}) and we computed fWHR as the ratio of bizygomatic-width (maximum horizontal distance from the left to the right facial boundary) to upper face height (vertical distance from the midpoint of the upper lip to the highest point of the eyelids) ([@R40]; [@R73]). We calculated fLHFH as the distance between the highest point of the eyelids and the lowest point of the chin divided by the length of the whole face, i.e., from brow to chin. We reassessed 24 images at random (from 21 different monkeys), i.e., 10% of the total number of face photographs, and compared them to the original assessments to evaluate the reliability of our measurements. Study Design and Statistical Analyses {#S8} ------------------------------------- With these demographic, personality, and facial morphology data we carried out a series of exploratory regression analyses. These analyses were exploratory because 1) that was an analytical choice we made before carrying out any analyses: we would investigate many variables and interactions of potential interest, and 2) it was a necessity: macaque personality structure is not perfectly analogous to capuchin or human personality structure ([@R68]). Direct predictions cannot be made, particularly when on the human side of the field, the existence of associations between fWHR and personality is contested ([@R30]; [@R35]). Our aim was thus to carry out exploratory analyses and generate rhesus macaque specific hypotheses, which can then be tested in future work. Although this study was exploratory, we nevertheless planned the analytic approach described in the text that follows in advance, although we did not formally preregister our analyses. We assessed reliability of facial measurements using three metrics. Cronbach's alpha (*α*) is a test of internal consistency, that is, if two assessments are measuring the same thing. In this case the assessments are the originals and retests of a given face photograph by the same rater. Though widely used, *α* overestimates reliability, so we supplemented it with the Pearson product-moment correlation coefficient (*r*) and Guttman's lambda 6 reliability ([@R52]). fWHR and fLHFH were calculated for both the original and retest measurements, then test--retest reliability was compared using these statistics. We also assessed the reliability of the individual point placements on macaques' faces. We primarily assessed the relationships between variables using linear mixed models ([@R5]), and the assumptions made by these models were checked for violations ([@R8]). Where possible, we estimated bootstrapped 95% confidence intervals, but when this was not possible, we computed profile or Wald intervals. We conducted all analyses in the R programming language (version 3.4.2). Our exploratory approach followed a model building procedure that minimized individual steps; we formulated steps in terms of successive sets of variables explaining variance over and above the preceding sets. Variables that did not improve model fit were not retained to avoid collider bias ([@R15]). First, we examined age, age^2^, and age^3^ associations with our outcome variables, fWHR and fLHFH. Age^2^ and age^3^ were included to assess nonlinear age changes. Second, we added sex and a sex × age interaction. Third, we added normalized David's scores and its interaction with sex. Fourth, we added personality variables, in separate models for the full and short-form HPQ, as using only the short-form HPQ allowed us to enlarge the sample. Finally, for fWHR, at step 4 we also examined a David's score × Anxiety interaction. At each step, we retained variables if they were significant or if overall model fit was improved. After step two analyses of age associations across the whole sample, we carried out subsequent analyses simultaneously in the full sample, younger subsample, and older subsamples using the same variables; if we retained a variable in one sample it was retained in all for ease of comparability. fWHR and fLHFH were analyzed in separate models but following the same process. ### Data Availability {#S9} Data are available in the [online supplemental files](#SD1){ref-type="supplementary-material"} at the *International Journal of Primatology* website. Ethical Note {#S10} ------------ This project was noninvasive and purely observational and complied with the US Animal Welfare Act. The Institutional Animal Care and Use Committees at the Oregon National Primate Research Center (ONPRC) and California National Primate Research Center (CNPRC) approved these studies. Ethical approval was given by the ONPRC, CNRPC, and University of Edinburgh's Biological Services Unit, AWERB OS2-14 and A3433-01. The authors declare that they have no conflict of interest. Results {#S11} ======= Reliability of Facial Measurements {#S12} ---------------------------------- All statistics indicated that our ratings of faces were reliable from one individual to the next ([Table I](#T1){ref-type="table"}). We also found only small deviations in placement of points A through G from one assessment to the next ([ESM Table SII](#SD1){ref-type="supplementary-material"}). We describe the complete details of our facial reliability analyses in the [online supplement](#SD1){ref-type="supplementary-material"}, along with the correlation matrix for all variables of interest ([ESM Table SIII](#SD1){ref-type="supplementary-material"}). Facial Width/Height Ratio {#S13} ------------------------- ### Age, Sex, and Dominance Status {#S14} The best fit model contained age, age^2^, and age^3^ (χ^2^ = 4.24, df = 7, *P* = 0.039), although none of the variables were significant on their own in the model ([Table II](#T2){ref-type="table"}). Building on this model by including sex and an age × sex interaction revealed a significant effect of sex, but no interaction ([ESM Table SIV](#SD1){ref-type="supplementary-material"}). In our models of age and sex in the younger and older subsamples, we found no age × sex interaction ([Table II](#T2){ref-type="table"}). We found a significant effect of sex for the younger group, but not for the older group, suggesting that sex differences in face structure within the younger age group drive the overall effect. Retaining sex, we added normalized David's scores to our models. Normalized David's score was not associated with fWHR in the full sample or subsamples ([Table III](#T3){ref-type="table"}). Normalized David's scores also interacted with sex in similar models, but in these models neither David's scores nor the interaction was significant. We thus excluded normalized David's scores from additional models of fWHR. ### Personality {#S15} In a model including monkeys rated on the full HPQ, we found no personality dimensions associated with fWHR ([Table IV](#T4){ref-type="table"}). However, in the younger sample, we found Activity was negatively associated and Assertiveness positively associated with fWHR, while in the older sample, Assertiveness was negatively associated with fWHR, but Confidence was positively associated with fWHR. We built similar models based on the shortened definition of four macaque personality dimensions, which allowed us to use animals from both facilities. In these models, we again found no associations between personality and fWHR that spanned the full sample. However, when we split the sample we increased discriminability within age groups, which allowed these models to detect a positive association between fWHR and Assertiveness and a negative association with Confidence, in the younger sample. We found no associations in the older sample. The positive linear association between Assertiveness and fWHR is visible in the under-8 individuals but the fit for the smaller, older group is not apparent ([Fig. 2](#F2){ref-type="fig"}). We were also interested in further exploring the relationship between Anxiety and fWHR, as lower ranking primates tend to experience more stress under certain social conditions and may express this through higher Anxiety ([@R1]). Thus, in post hoc analyses we built models that included all age variables, sex, dominance status, Anxiety, and the interaction of dominance status and Anxiety ([Table V](#T5){ref-type="table"}). The main effects of Anxiety (using the shortened questionnaire) and dominance status were significant in the full sample, as was the interaction between dominance status and Anxiety. These effects were all significant in the younger sample as well. Though none were significant in the older sample, the directions of effects were consistent in all groups. These results indicate that higher Anxiety individuals tend to have lower fWHR, and individuals that are high in dominance status and high in Anxiety or low in dominance status and low in Anxiety have higher fWHR. We examined models that used the full HPQ subsample, but we found no significant effects. Facial Lower-Height/Full-Height Ratio {#S16} ------------------------------------- ### Age, Sex, and Dominance Status {#S17} We followed the same analytic strategy with fLHFH as with fWHR. Modeling the effects of age, age^2^, and age^3^, on fLHFH, we again found that the best fit model contained all three age variables (χ^2^ = 12.4, df = 7, *P* \< 0.001). Unlike for fWHR, the associations between the age variables and fLHFH were significant ([Table VI](#T6){ref-type="table"}). We also split the sample into younger and older monkeys at this stage; the age effects were not significant in either subsample, but were in the same direction in all variations. Adding sex and an age × sex interaction to the best fit model, we found no effect for either sex (*B* = −0.015, CI: \[−0.044, 0.016\]) or the interaction (*B* = −0.011, CI: \[−0.047, 0.024\]). This held true in the younger (sex: *B* = 0.034, CI: \[−0.092, 0.023\]; interaction: *B* = −0.048, CI: \[−0.127, 0.044\]) and older (sex: *B* =0.034, CI: \[−0.202, 0.274\]; interaction: *B* = −0.031, CI: \[−0.258, 0.192\]) subsamples, so we did not retain sex in subsequent models of fLHFH. With normalized David's scores, we followed the same process as we did with our fWHR models, and similarly, found no association between dominance status and fLHFH in the full sample (*B* = 0.002, CI: \[−0.005, 0.008\]) or either younger (*B* =0.002, CI: \[−0.005, 0.009\]) or older (*B* =0.003, CI: \[−0.016, 0.012\]) subsamples. ### Personality {#S18} We first modeled personality's influence on fLHFH using the sample with full ratings for all six personality dimensions. As with our initial models of fWHR and personality, we found no associations between the variables ([Table VII](#T7){ref-type="table"}). We also found no association between the personality dimensions and fLHFH in the older group, but in the younger group we found a positive association between Assertiveness and fLHFH, and a negative association between Confidence and fLHFH. Using the shortened questionnaire dimensions from animals at both facilities, we also found no effects of personality across the entire sample, or in either subsample ([ESM Table SVI](#SD1){ref-type="supplementary-material"}). Discussion {#S19} ========== We found facial dimensions in rhesus macaques to be related to variables of age, sex, and personality, although the nature of these relationships varied somewhat. There were no specific, consistent effects of age, age^2^, or age^3^, but their inclusion in our models did improve fit, and thus they made a meaningful contribution. In other words, although the age variables were not always significant in our models, the interdependence between these variables and the outcomes supports an age effect across the macaque lifespan (at least through the ages present in our samples), such as the differences we found between skeletally mature and immature macaques. Splitting the sample by these age groups ([@R7]; [@R13]; [@R65]) consistently made our models more interpretable, and differences we found in the associations between facial dimensions and age group suggest that changes in facial morphology may co-occur with psychological development. [Table VIII](#T8){ref-type="table"} summarizes our findings. In contrast to work across the macaque genus ([@R9]), which found no sex difference in fWHR across 11 macaque species, we found that females across the sample have wider faces, particularly in the younger sample of macaques. There were, however, no sex effects on fLHFH. We also found no interactions between age and sex, which contrasts with findings in capuchin monkeys, in which sex was not associated with either fWHR or fLHFH, but an age × sex interaction was found with both ([@R40]; [@R73]). We also found that fWHR was associated with aspects of personality, but in different ways depending on which age category the individual fell within. Assertiveness was negatively associated with fWHR in older monkeys (above age 8), but positively associated in younger monkeys (under age 8); contrastingly, Confidence was negatively associated with fWHR in young monkeys. Activity was also negatively associated with lower fWHR in the young macaques. The relationship between Assertiveness (item loadings from Dominant, Bullying) and fWHR in the young group reflects findings in capuchins, where Assertiveness (item loadings Bullying, Aggressive, Dominant) was positively associated with fWHR ([@R40]). Associations between fLHFH and personality and other variables were weaker. fLHFH was associated with Confidence and Assertiveness in the same way as fWHR, again, only in the young subsample of macaques. It has been hypothesized that the fLHFH ratio is a potential signal of status in capuchins, conferred not through aggressive or dominant behaviors but through a policing role ([@R25]), which could explain the links to vigilant behavior in brown capuchins ([@R73]). In rhesus macaques, the lack of a relationship in the older sample suggests that this association is not indicative of some form of social cue. That these findings for rhesus macaques do not reflect findings in brown capuchins suggests that the relationship between status and face height may differ between species. Our evidence only suggests that, at least in macaques, fLHFH is measuring the same underlying physical characteristics as fWHR. This is consistent with the correlation of *ρ* = 0.19 ([Table SIII](#SD1){ref-type="supplementary-material"}) that we calculated between fLHFH and fWHR. However, in brown capuchins the correlation was much lower, 0.02 ([@R73]), so to determine whether fLHFH is a distinct social signal of some kind, further investigation in other primate species are warranted. We found no relationships between dominance status (i.e., normalized David scores) and fWHR or fLHFH. Earlier theory and results from a variety of macaque species suggest that fWHR could be a cue to social rank ([@R9]). Moreover, work in capuchins found an association between alpha status and fWHR ([@R40]). Despite finding no relationship between dominance status and fWHR, it is notable that dominance status had considerable overlap with both the Confidence and Assertiveness personality dimensions ([Table SIII](#SD1){ref-type="supplementary-material"}), which were both related to fWHR. Thus, it seems that personality alone, rather than personality as a proxy for social standing, relates to facial morphology. This runs counter to the suggestion that fWHR is a visual signal for social status, or is driven by low-status individuals ([@R11]; [@R28]; [@R69]), though we were unable to directly test this hypothesis with this sample. Another possibility is that dominance status is more important as a physical cue in New World monkeys, whereas in Old World monkeys Assertiveness rather than dominance status is a stronger predictor of fWHR variance. This fits with the findings that rhesus macaques differentiate between faces based on trustworthiness ([@R17]). Rhesus macaques are "despotic" ([@R59]) and among the most intolerant of macaque species, whereas capuchins are known for being socially tolerant within their groups ([@R26]). Differing levels of social tolerance may explain the differences in personality associations between rhesus macaques and brown capuchins. It would be informative to explore to what extent social tolerance of different species varies with differences in fWHR. If the relationship between fWHR and dominance behavior varies with social status, then one would expect to find larger differences in this ratio between alpha and non-alpha individuals in despotic species over socially tolerant species. One difficulty of this approach is that one must establish a measure that can account for not only species differences but also intraspecific sex differences in social tolerance, something that we were unable to fully address in this study. The presence of both an Assertiveness and Confidence dimension in rhesus macaque personality structure complicates interpretations of these associations with facial morphology. Confidence is characterized by items like "fearful," "submissive," "cautious" -- all reverse coded. Assertiveness is characterized by "dominant," "bullying," and "irritable." Thus, these constructs are already measuring similar underlying processes, so a possible explanation for our results is that as younger monkeys mature, a shift in behavior changes the associations between individual features and both Assertiveness and Confidence. One such feature could be fWHR. We found a negative association between Anxiety and fWHR. Typified by "quitting," "anxious," "erratic," and the reverse coding of "cool," Anxiety overlaps with both human and capuchin Neuroticism ([@R16]; [@R46]). Neuroticism was associated with fLHFH in capuchins, but not with fWHR. Anxiety has a low correlation with dominance status, so as with Confidence and Assertiveness, it is unlikely that dominance status confounded this association. However, the positive interaction we found between dominance status and Anxiety indicates that more dominant individuals tend to either have higher Anxiety or higher fWHR. One possible reason for this is that without the physical social signal of fWHR to support their position, more dominant individuals experience more stress and have higher Anxiety as a result. While a growing body of literature has developed around fWHR ([@R9]; [@R30]; [@R35]; [@R40]; [@R72]) and fLHFH ([@R50]; [@R73]), associations between these physical variables and psychological measures are understudied in juvenile and subadult primates ([@R33]; [@R70]; [@R75]). Our study suggests that age may influence how facial dimensions relate to personality. None of the associations between personality and our facial dimension measures spanned both age categories. As we suggested, the same underlying processes could be involved in different associations but unearthing the physiological and developmental processes involved is beyond the scope of this study. Despite having a larger sample of macaques to work with when we analyzed the short-form data, associations between facial ratios and the shortened personality variables were often weaker. This is consistent with the finding in the field of personality psychology that fewer items generally results in poorer validity ([@R14]). We found associations only in the younger group, which fit the data: we used the shortened questionnaire at the ONPRC and the mean age of that sample was younger than that of the CNPRC. Owing to the limitations of the shortened questionnaire, we likely only had the power to detect the same effects in the younger subsample of monkeys. However, the associations we did find between the shortened dimensions and facial ratios were consistent with our overall results. The relationships among sex, social status, and dominance behaviors are complex and variable ([@R6]; [@R34]), and incorporating social style is a particular issue for studies of macaques because social style is derived from female behavior. Social style appears to be related to fWHR across sexes ([@R9]), but without a comprehensive understanding of the male behaviors that complement tolerant or despotic female behavior, it is difficult to conclude how pressures on male social position would relate to social signals like fWHR or fLHFH. Our sample of skeletally immature macaques was larger than our sample of skeletally mature macaques. Our overall results are more representative of the younger population, and further work is necessary to derive any solid conclusions about the associations among personality, social status, sex, and fWHR in older rhesus macaques. Our findings show that even from a young age, facial differences emerge that can be linked to personality differences. These results are broadly consistent with previous work in capuchin monkeys and humans. In humans, fWHR is most often associated with aggression ([@R30]), though the effect is small. Rhesus macaques are a despotic, sexually dimorphic, species, and associations between facial dimensions and dominance-related traits are likely to be stronger than the relationships one might find in humans. Thus, the multiple relationships between dominance-related traits and facial measures we found in rhesus macaques suggest that associations between personality and facial morphology may have been present deep in our phylogenetic history, beyond the common ancestor of Old World and New World monkeys. However, our results are not consistent with associations found between fLHFH and personality in capuchin monkeys, suggesting in this respect, the associations among personality, behavior, and fLHFH may have diverged since the split between Old and New World monkeys. Supplementary Material {#SM1} ====================== We thank the staff, particularly those who performed personality ratings, at the Oregon and California National Primate Research Centers for their help with this project. We thank Adriane Maier and Daniel Gottlieb for their help. We acknowledge support from NIH award P51OD011092 and P51OD011107. We also thank Oliver Schulke, Joanna Setchell, and three anonymous reviewers for their helpful feedback on an earlier version of the manuscript. Electronic supplementary material The online version of this article (<https://doi.org/10.1007/s10764-019-00104-y>) contains [supplementary material](#SD1){ref-type="supplementary-material"}, which is available to authorized users. ![Example of measuring points used for rhesus macaque morphometric calculations. Measure for facial width-to-height ratio: (A--F)/\[midpoint (C, D)--G\]. Lower-height/full-heightratio: \[midpoint (C, D)--B\]/(A--B). We observed the rhesus macaques (*Macaca mulatta*) at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014.](nihms-1049742-f0001){#F1} {#F2} ###### Reliability statistics for two facial measures Cronbach's *α* Guttman's λ, *G6* Pearson correlation, *r* -------------------------------- ---------------- ------------------- -------------------------- Facial width-to-height ratio 0.81 0.66 0.81 Lower-height/full-height ratio 0.97 0.90 0.95 Photographs were taken of rhesus macaques (*Macaca mulatto*) observed at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014 ###### Associations between facial width-to-height ratio, age, and sex, split by age group Variable Younger Older ----------- --------- ----------------- ----------- ---------------------- ------- ----------------- ------- ----------------- Age −0.88 \[−1.85, 0.20\] **−0.93** \[**−1.99, −0.05**\] −0.59 \[−4.19, 2.59\] 1.26 \[−5.90, 7.88\] Age^2^ 2.90 \[−1.90, 7.32\] 3.55 \[−0.35, 8.18\] 0.92 \[−3.19, 5.59\] −1.74 \[−10.9, 8.50\] Age^3^ −2.72 \[−8.34, 3.48\] −3.88 \[−9.72, 1.10\] −0.43 \[−2.27, 1.19\] 0.76 \[−3.59, 4.78\] Sex **−0.12** **\[−0.21, −0.04\]** 0.20 \[−0.48, 0.84\] Sex × Age −0.05 \[−0.18, 0.07\] −0.20 \[−0.81, 0.40\] We fitted two models each to the younger and older data, the first without sex and an age × sex interaction, and a second model with. Bold indicates estimates whose confidence interval did not overlap with 0. We observed rhesus macaques (*Macaca mulatta*) at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014 ###### Mixed models of facial width-to-height ratio, sex, dominance status, and age variables Variable All individuals Younger Older ------------------ ----------------- ---------------------- ----------- ---------------------- ------- ----------------- Age −0.19 \[−0.44, 0.06\] −0.84 \[−1.85, 0.01\] −0.59 \[−4.08, 2.89\] Age^2^ 0.37 \[−0.13, 0.87\] 3.31 \[−0.47, 7.67\] 1.17 \[−3.40, 5.75\] Age^3^ −0.20 \[−0.47, 0.07\] −3.88 \[−9.33, 0.94\] −0.63 \[−2.48, 1.21\] Sex **−0.07** **\[−0.14, −0.02\]** **−0.10** **\[−0.15, −0.05\]** 0.17 \[−0.17, 0.50\] Dominance status −0.01 \[−0.02; 0.01\] 0.01 \[−0.02, 0.01\] −0.03 \[−0.08, 0.02\] Bold indicates estimates whose confidence interval did not overlap with 0. We observed rhesus macaques (*Macaca mulatta*) at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014 ###### Mixed models of facial width-to-height ratio, personality, age, and sex variables Variable All individuals Younger Older --------------- ----------------- ----------------- ----------- ---------------------- ----------- ---------------------- Age −0.21 \[−0.56, 0.13\] −0.06 \[−1.61, 1.35\] −2.19 \[−5.65, 1.30\] Age^2^ 0.41 \[−0.25, 1.05\] −0.36 \[−7.33, 6.94\]\] 3.20 \[−1.25, 7.60\] Age^3^ −0.22 \[−0.56, 0.13\] 0.30 \[−9,87, 9.91\] −1.37 \[−3.07, 0.36\] Sex −0.01 \[−0.12, 0.11\] 0.11 \[−0.11, 0.34\] 0.13 \[−0.21, 0.46\] Confidence 0.03 \[−0.14, 0.19\] −0.15 \[−0.35, 0.05\] **0.37** **\[0.03, 0.72\]** Openness 0.06 \[−0.08, 0.19\] 0.07 \[−0.10, 0.27\] 0.29 \[−0.01, 0.59\] Assertiveness −0.03 \[−0.21, 0.14\] **0.23** **\[0.05, 0.46\]** **−0.44** **\[−0.77, −0.11\]** Friendliness 0.01 \[−0.07, 0.08\] 0.06 \[−0.12, 0.22\] −0.09 \[−0.27, 0.09\] Activity −0.10 \[−0.2, 0.01\] **−0.22** **\[−0.44, −0.01\]** −0.13 \[−0.28, 0.02\] Anxiety 0.05 \[−0.05, 0.16\] 0.00 \[−0.25, 0.22\] 0.10 \[0.05, 0.26\] Removing age variables from the model of younger macaques did not substantively change the results. Bold indicates estimates whose confidence interval did not overlap with 0. We observed rhesus macaques (*Macaca mulatta*) at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014 ###### Mixed models of facial width-to-height ratio and Anxiety by dominance status interactions Variable All individuals Younger Older ---------------------------- ----------------- ---------------------- ----------- ---------------------- ------- ----------------- Age −0.19 \[−0.43, 0.06\] **−1.00** **\[−1.97, −0.17\]** −0.79 \[−4.12, 2.55\] Age^2^ 0.38 \[−0.10, 0.86\] **4.14** **\[−0.50, 8.40\]** 1.52 \[−2.88, 5.92\] Age^3^ −0.22 \[−0.48, 0.04\] **−5.04** **\[−10.4, −0.40\]** −0.83 \[−2.62, 0.97\] Sex **−0.07** **\[−0.13, −0.01\]** **−0.09** **\[−0.15, −0.04\]** 0.15 \[−0.24, 0.54\] Dominance status **−0.02** **\[−0.03, 0.00\]** **−0.02** **\[−0.03, −0.00\]** −0.06 \[−0.13, 0.01\] Anxiety **−0.14** **\[−0.26, −0.02\]** **−0.13** **\[−0.23, −0.03\]** −0.32 \[−0.83, 0.17\] Dominance status × Anxiety **0.02** **\[0.00, 0.03\]** **0.02** **\[0.00, 0.03\]** 0.04 \[−0.02, 0.10\] Bold indicates estimates whose confidence interval did not overlap with 0. We observed rhesus macaques (*Macaca mulatta*) at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014 ###### Mixed models of lower-height/full-height ratio and age variables Variable All individuals Younger Older ---------- ----------------- --------------------- ------- ----------------- ------- ------------------ Age **−0.18** **\[−0.3, −0.06\]** −0.29 \[−0.81, 0.25\] −0.04 \[−1.25, 01.16\] Age^2^ **0.40** **\[0.16, 0.63\]** 0.70 \[−1.71, 2.98\] 0.28 \[−1.27, 1.83\] Age^3^ **−0.23** **\[−0.36, −0.1\]** −0.36 \[−3.27, 2.74\] −0.21 \[−0.82, 0.40\] Bold indicates estimates whose confidence interval did not overlap with 0. We observed rhesus macaques (*Macaca mulatta*) at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014 ###### Mixed models of lower-height/full-height ratio, personality, and age variables Variable All individuals Younger Older --------------- ----------------- ---------------------- ----------- ---------------------- ------- ----------------- Age −0.13 \[−0.28, 0.04\] −0.15 \[−0.72, 0.43\] −0.82 \[−2.51, 0.87\] Age^2^ **0.33** **\[0.02, 0.61\]** −1.01 \[−3.60, 1.57\] 1.24 \[−0.98, 3.27\] Age^3^ **−0.20** **\[−0.35, −0.03\]** 2.85 \[−0.56, 6.27\] −0.54 \[−1.32, 0.31\] Confidence −0.04 \[−0.11, 0.03\] **−0.14** **\[−0.23, −0.05\]** 0.05 \[−0.10, 0.19\] Openness 0.04 \[−0.02, 0.10\] −0.01 \[−0.10, 0.07\] 0.12 \[−0.03, 0.25\] Assertiveness 0.03 \[−0.05, 0.10\] **0.16** **\[0.07, 0.27\]** −0.05 \[−0.20, 0.10\] Friendliness 0.01 \[−0.03, 0.04\] 0.02 \[−0.05, 0.79\] −0.06 \[−0.13, 0.03\] Activity −0.02 \[−0.06, 0.03\] −0.04 \[−0.09, 0.02\] −0.03 \[−0.11, 0.04\] Anxiety −0.02 \[−0.07, 0.03\] −0.05 \[−0.15, 0.04\] −0.02 \[−0.04, 0.07\] Bold indicates estimates whose confidence interval did not overlap with 0. We observed rhesus macaques (*Macaca mulatta*) at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014 ###### Summary of results from rhesus macaques (*Macaca mulatta*) observed at the Oregon National Primate Research Center between March and June 2014, and at the California National Primate Research Center between January and April 2014 Facial width-to-height ratio Lower-height/full-height ratio ---------------------------- ------------------------------ -------------------------------- ---- --- Sex − . . . Sex × Age . . . . Confidence − . − . Openness . . . . Assertiveness \+ − \+ . Friendliness . . . . Activity − . . . Anxiety (−) (−) . . Dominance status (−) (−) . . Anxiety × Dominance status − − A plus sign indicates a positive relationship, a minus sign indicates a negative relationship, and a dot indicates no significant relationship, all via regression models. Associations between Anxiety and facial width-to-height ratio are in parentheses to indicate that this association was present only when the Anxiety by Dominance status interaction was included [^1]: **Author's Contributions** LMR, DMA, and VADW initially conceived the project. LMR submitted the ethical forms and collected the data. DMA, LMR, and VADW wrote and revised the manuscript. VADW developed the methodology. DMA and VADW performed the statistical analyses. KC and JPC supervised this project at their respective facilities and provided editorial support.

Dear Editor-in-Chief ==================== Most of anti-cancer drugs have mutagenic, clastogenic, and carcinogenic properties. Studies on oncology nurses and personnel handling cytostatic drugs showed that the frequencies of chromosomal aberrations, sister chromatid exchanges, and micronuclei were significantly increased in personnel handling anti-cancer drugs compared to control group ([@B1]). β-Lapachone (β-Lap) is an anti-cancer drug which exerts cytotoxic effect via its induction of ROS generation which ultimately leads to DNA damage ([@B2]). Nucleotide excision repair (NER), base excision repair (BER), and non-homologous end joining repair (NHEJ) pathways are very important in genome stability. XRCC1 (OMIM: 194360, an essential scaffolding protein for both long and short patch BER), GADD45A (OMIM: 126335, a growth arrest and DNA-damage-inducible protein) and LIG4 (OMIM: 601837) have important roles in the BER, NER, and NHEJ pathways, respectively. LIG4 efficiently joined single-strand breaks in a double-stranded polydeoxynucleotide in an ATP-dependent reaction ([@B3]). LIG4 efficiently joined single-strand breaks in a double-stranded polydeoxynucleotide in an ATP-dependent reaction. LIG4 efficiently joined single-strand breaks in a double-stranded polydeoxynucleotide in an ATP-dependent reaction. Depletion of XRCC1 dramatically sensitized cells to β-Lap ([@B4]) and Gadd45a-null mice showed genomic instability ([@B5]). β-Lap efficiency can be affected by NHEJ performance ([@B2]). Due to anti-cancer property of β-Lap, so it's occupationally exposure as a public health concern is expected. To our knowledge there is no study on the effect of β-Lap on the transcript levels of *XRCC1*, *GADD45A* and *LIG4* genes. Therefore the present study was carried out. SH-SY5Y neuroblastoma cell was cultured in DMEM/F12 enriched with 10% FBS (Gibco), penicillin (100 U/ml, Sigma) and streptomycin (100 μg/ml, Sigma). The cells were seeded at 3 × 10^5^ cells/ml and incubated at 37 °C for 24 h and then cells were treated with β-Lap. Cells were harvested after 24 h and RNA extraction was done. Quantitative real-time PCR and primers specific for the examined genes were described previously ([@B6]). 3.2 and 2.0 μm M β-Lap showed about 18% cytotoxicity and no cytotoxicity, respectively. The experiments were done in triplicates. Data were shown as means ± standard error (SE). [Figure 1](#F1){ref-type="fig"} shows the alteration of mRNA levels of *XRCC1*, *GADD45A* and *LIG4* genes in different treatments. The *XRCC1* mRNA level was significantly decreased at non-toxic concentration. The *GADD45A* mRNA levels did not alter at nontoxic concentration of β-Lap, however, it was significantly increased at toxic concentration of β-Lap, compared with the control culture. The mRNA levels of *LIG4* were significantly decreased at both toxic and non-toxic concentrations of β-Lap. The expression levels of the *XRCC1* and *LIG4* significantly decreased at nontoxic concentrations of β -Lap, cellular DNA repair system cannot repair DNA damages. {#F1} We know that nurses of oncology departments and workers handling neoplastic drugs showed higher chromosomal damage compared to control persons ([@B1]), which may interpreted by their lower DNA repair capacity due to exposure of non-toxic levels of anti-cancer drugs. Alterations in mRNA levels of DNA repair related genes seem to be a rapid, simple and sensitive method for biomonitoring of effect(s) of occupationally exposure to anti-cancer drugs. For public health programs, the early detection of alterations may permit the adoption of preventive biological controls such as hygienic improvements in the workplace or the reduction of work hours. Further experiments needs to investigate the effects of other anti-cancer drugs of expression levels of DNA repair genes at non-toxic concentrations. This work was supported by the Shiraz University, Iran (Grant number: 93GCU1M1741). **Conflict of interest** The authors declare that there is no conflict of interests.

Introduction {#Sec1} ============ *Acanthaspis cincticrus* (Stål) (Hemiptera: Reduviidae) is a predatory assassin bug, which feeds on ants and can be found in the vicinity of ant nests, waiting for the preys^[@CR1]^. This species is native to Oriental Region with one generation per year. Nymphs of this species have five instars and exhibit both natural and corpse camouflaging behaviors^[@CR1]^. They cover themselves with a range of materials found in their environment, including ants corpses and other insects, dust and soil particles, which they affix on themselves with the viscid secretions from specialized setae on the abdomen^[@CR2]^. Camouflaging in *A. cincticrus* is specific to the nymphs (Fig. [1A](#Fig1){ref-type="fig"}) and is absent in the adults (Fig. [1B](#Fig1){ref-type="fig"}). This behavior has been documented in some neuropterans^[@CR3]^. Camouflaging may benefit nymphs in two ways: by being less visible in stalking their preys and by not being obvious to potential predators. While studies on *A. cincticrus* thus far have focused on the morphology, biology and behaviors related to camouflaging in nymphs^[@CR1],[@CR2]^, the molecular mechanisms underlying development and predation are unknown.Figure 1The nymph and adult of *A. cincticrus*. (**A**) A masked nymph camouflaged with ant corpses, dust and soil particles. (**B**) An adult male. Photographs were taken by F.K. In predatory insects, arousing, paralyzing and sucking are important for predatory process^[@CR4]^. These processes are influenced or regulated by several genes or pathways involved in the identification of chemical signals, and in the regulation of neuromodulators for agonistics and the digestion of the prey^[@CR5]--[@CR7]^. Predators locate and interact with their prey by using species-specific semiochemicals, such as the detection and discrimination of odorant-binding proteins (OBPs) and chemosensory proteins (CSPs)^[@CR5]^. After locating the target, aggression is displayed by predaceous arthropods against their prey^[@CR8]^. Factors that affect aggressiveness and/or fighting success in predaceous arthropods include body size^[@CR9]^ and contestants' age^[@CR10]^. Previous studies reported that many genes, such as *cyp6Q20*, *tyramine receptor*, *octopamine receptor* and *metabotropic glutamate receptor B*, were likely to be involved in the regulation of complex behavioral phenotypes^[@CR11],[@CR12]^. *Tyramine receptor* and *octopamine receptor* serve as aggression modulating neurotransmitters that affect excessive aggressiveness and impulsiveness in the biogenic amine signaling pathway^[@CR11]^; *metabotropic glutamate receptor B* is a part of G protein coupled metabotropic receptors family, which is the major excitatory neurotransmitter acting through multiple independent receptors, involved in regulating aggressive behavior of mice^[@CR12]^. The paralyzing and digestion process of assassin bugs includes extra oral digestion (EOD)^[@CR13]^ which is a chemical pretreatment to mobilize nutrients and is common in the predation of Heteroptera^[@CR7]^. Venom and/or digestive enzymes have a biochemical role of typically stunning, killing prey and even aiding in prey digestion^[@CR7],[@CR14]^. The venomous saliva of predatory reduviid bugs is known to contain a complex mixture of proteins^[@CR15]^, peptides^[@CR16]^ and enzymes^[@CR17]^. The most abundant enzymes that are present in the salivary secretion include proteinases, phospholipase, trypsin like enzymes, esterase, serine proteases, etc^[@CR18]^. Serine proteases, found in many organisms, have attracted broad interests because they have diverse physiological functions that affect processes such as digestion, immune response, cellular differentiation and prothrombin activator^[@CR7],[@CR13]^. The presence of trypsin-like enzyme in the salivary glands of *Zelus renardii* indicates that it has evolved or retained *trypsin-like* genes for protein digestion^[@CR7]^. Although most reduviids are common predators, studies on the factors influencing the feeding process focused on digestion^[@CR7],[@CR15]--[@CR18]^. The transcriptomic and bioinformatic approaches presented in previous studies have been served as a foundation to explore the relationships between gene regulation and behavioral evolution in other species^[@CR5],[@CR11]^. In this study, we performed a comprehensive transcriptome analysis during eight life stages of *A. cincticrus*, including egg, five instar nymphs, female and male adults. We identified 13,479 significantly differentially expressed genes (SDEGs) and also identified and characterized 115 SDEGs involved in predation, indicating differences between genders and among life stages. We performed quantitative real-time PCR (qRT-PCR) analysis to determine the expression profiles of ten SDEGs involved in predation from different life stages. The results could help elucidate the role of venom-related, aggression-related and olfactory-related genes involved in the predation of reduviids. Results {#Sec2} ======= Whole-transcriptome sequencing and annotation of the predicted proteins {#Sec3} ----------------------------------------------------------------------- cDNA libraries from all life stages of *A. cincticrus* yielded about 176 million clean sequence reads with an average quality value ≥ 30. GC content of the sample was averaged at 38.03% (Table [S1](#MOESM1){ref-type="media"}). After assembling the clean reads, we obtained 164,745 transcripts and 84,055 unigenes with an average sequence length of 1,366 bp and 667 bp, and N50 with average lengths of 2,731 bp and 1,087 bp, respectively (Table [1](#Tab1){ref-type="table"}). The length of unigenes ranged from 210 bp to 29,391 bp (Fig. [2A](#Fig2){ref-type="fig"}). Of which, 12, 679 (or 15.09%) unigenes had sequence length more than 1,000 bp. The result is similar to the number of unigenes reported in the moths *Dendrolimus punctatus* (70,664)^[@CR19]^ and *Athetis lepigone* (81,356)^[@CR20]^. A proportion of unigenes was annotated based on the available protein database for *A. cincticrus*. 24,402 unigenes (or 29.03%) were annotated as coding hypothetical proteins. Homology analysis of *A. cincticrus* unigenes showed that they best matched with the species from Hemiptera, the pea aphid *Acyrthosiphon pisum* and the bean bug *Riptortus pedestris* (Fig. [2B](#Fig2){ref-type="fig"}).Table 1Summary of *Acanthaspis cincticrus* transcriptomes.Length rangeTranscriptsUnigenes200--30039,146(23.76%)33,985(40.43%)300--50031,343(19.03%)23,125(27.51%)500--100029,635(17.99%)14,266(16.97%)1000--200028,485(17.29%)7,202(8.57%)2000+36,136(21.93%)5,477(6.52%)Total number164,74584,055Total length225,069,31156,032,206N50 length2,7311,087Mean length1,366667 Figure 2The statistics of assembly and homology analyses. **(A**) Size distribution of unigenes. (**B**) Species distribution of the BLASTX against NCBI-NR database, proportions of more than 1% were shown. Functional classification and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis {#Sec4} ------------------------------------------------------------------------------------- The functions of unigenes were predicted by Gene Ontology (GO) analysis at the macro level. In total, 11,682 unigenes were classified into 58 sub-categories belonging to three GO functional categories: cellular component (CC), molecular function (MF) and biological process (BP) (Fig. [S1](#MOESM1){ref-type="media"}). The GO classification of unigenes indicated that 'cell part' (20.79%) and 'cell' (20.68%) were the dominant sub-categories in the CC category, 'catalytic activity' (42.40%) and 'binding' (38.83%) were the most dominant sub-categories in the MF category, and 'metabolic process' (24.23%) and 'cellular process' (21.81%) were the most dominant sub-categories in the BP category. We also annotated the unigenes by searching the Cluster of Orthologous Groups (COG) database to classify the functions of the predicted proteins. A total of 6,822 unigenes had a COG classification. Among the 25 COG categories (Fig. [S2](#MOESM1){ref-type="media"}), the cluster for 'general function prediction only' was the most dominant group (18.25%), followed by 'translation, ribosomal structure and biogenesis' (8.66%) and 'post-translational modification, protein turnover, chaperones' (8.58%). In order to identify the biological pathways represented in the transcriptome of *A. cincticrus*, 7,193 unigene sequences were mapped to 193 KEGG pathways (Table [S2](#MOESM1){ref-type="media"}). The most prominent metabolic pathways were protein processing in endoplasmic reticulum (4.25%), oxidative phosphorylation (3.99%) and RNA transport (3.70%). The SDEGs in different life stages {#Sec5} ---------------------------------- We evaluated the gene expression levels to measure the genetic landscape in different life stages of *A. cincticrus*. We performed nine comparisons of the transcriptomes from pairs of life stages (e.g. egg vs 1^st^ instar, 1^st^ vs 2^nd^ instar, 2^nd^ vs 3^rd^ instar, 3^rd^ vs 4^th^ instar, 4^th^ vs 5^th^ instar, 1^st^ vs 5^th^ instar, 5^th^ instar vs adult male, 5^th^ instar vs adult female, male vs female adults) (Table [S3](#MOESM1){ref-type="media"}). The results yielded 13,479 SDEGs (Table [S4](#MOESM1){ref-type="media"}), which were then hierarchically clustered in a heat map (Fig. [S4](#MOESM1){ref-type="media"}). In this heat map, the gene expression profiles differed through all life stages with each comprising a main cluster of high expression genes. In the egg stage, many crucial embryogenesis related genes also comprised maternal genes such as *kruppel*, *hunchback*, *nanos-like protein* and *serine protease snake-like*, *protein takeout-like*, *maelstrom*, *pumilio homolog 1-like* and *staufen* (Table [S5](#MOESM1){ref-type="media"}). The nymph and adult stages also had their own high expression unigenes, including *cuticular proteins*, *histone H2A*, *cytochrome P450s*, *ecdysone-induced protein* and *juvenile hormone binding proteins* in nymph stages, while *vitellogenin* and *sperm flagellar protein 1-like* in the adults. A previous study reported that these maternal genes, e.g. *snake* and *nanos*, might be involved in the embryonic development of *Oncopeltus fasciatus* ^[@CR21]^. The distribution of upregulated, downregulated and uniqueness of SDEGs in the different comparisons were shown in the Table [S6](#MOESM1){ref-type="media"}. We also analyzed the changes in gene expression between the nymph stages and displayed the distribution of SDEGs using a Venn diagram that illustrated the intersection between the expressed genes in various developmental nymph stages (Fig. [S4](#MOESM1){ref-type="media"}). In total, 38 SDEGs were shared by all nymph stages. 203, 937, 333 and 614 SDEGs were uniquely expressed in 1^st^ vs 2^nd^ instar, 2^nd^ vs 3^rd^ instar, 3^rd^ vs 4^th^ instar, and 4^th^ vs 5^th^ instar indicating that different life stages had different highly expressed unigenes that participated in specific life activities. In the comparison of egg vs 1^st^ instar, upregulated SDEGs with the high variation (10+ fold) were *serine protease*, *trypsin-like protease*, *OBPs*, *cuticular protein*, etc. The downregulated SDEGs (3+ fold) were annotated to *early cuticle protein 6*, *hunchback*, *nanos-like protein*, *serine protease snake-7*, *vitellogenin receptor*, etc. In the comparisons of 1^st^ vs 2^nd^ instar and 2^nd^ vs 3^rd^ instar, the most enriched GO terms were oxidation-reduction process (BP) and structural constituent of ribosome (MF). Lipid particle (CC) and cytoplasm (CC) were respectively enriched in 1^st^ vs 2^nd^ instar and 2^nd^ vs 3^rd^ instar (Table [S7](#MOESM1){ref-type="media"}). Oxidative phosphorylation, ribosome and lysosome were the top three enrichment pathways in the KEGG pathway analysis (Table [S8](#MOESM1){ref-type="media"}). When comparing 3^rd^ vs 4^th^ instars, most upregulated genes were enriched for GO categories including protein metabolic process (BP), intracellular (CC) and catalytic activity (MF) (Table [S7](#MOESM1){ref-type="media"}). According to KEGG enrichment analysis, the upregulated genes were enriched for pathways such as arginine and proline metabolism, wnt signaling pathway and glutathione metabolism (Table [S8](#MOESM1){ref-type="media"}). When comparing 4^th^ vs 5^th^ instar, most of the SDEGs were categorized into nucleosome assembly (BP), intracellular (CC) and catalytic activity (MF) (Table [S7](#MOESM1){ref-type="media"}). The KEGG analysis indicated that arginine and proline metabolism, glutathione metabolism and progesterone-mediated oocyte maturation were the top enriched pathways (Table [S8](#MOESM1){ref-type="media"}). We also performed a comparison of 1^st^ vs 5^th^ instar. The upregulated genes with the high expression variation (5+ fold) were *pupal cuticle protein precursor*, *defensin*, *histone*, etc. The downregulated genes with high expression variation were *ecdysone-induced protein*, *juvenile hormone binding protein-like precursor*, etc. Between the 5^th^ instar nymph vs adult male transcriptomes, the upregulated genes with the high expression variation (5+ fold) included *venom serine protease*, *odorant-binding protein*, *methylmalonyl-CoA decarboxylase*, etc. The downregulated genes included digestive and metabolic genes such as *histone H2A*, *prostaglandin reductase*, *cytochrome P450 monooxygenase CYP6*  *X* *1v1*, etc. When comparing female vs male adults, the upregulated genes (5+ fold) included *venom serine protease*, *vitellogenin receptor*, etc. The downregulated genes included *sperm flagellar protein 1-like*, *ejaculatory bulb-specific protein 3*, *heat shock protein 60*, etc. The three comparisons (egg vs 1^st^ instar, 1^st^ vs 5^th^ instar, 5^th^ instar vs adult male), including two developmental transitions (eggs to nymphs, nymphs to adults), shared two pathways for the upregulated SDEGs: wnt signaling pathway and neuroactive ligand-receptor interaction (Fig. [3A](#Fig3){ref-type="fig"}), which are important pathways in processing information^[@CR22]^. Three pathways, including arginine and proline metabolism, purine metabolism and oxidative phosphorylation (Fig. [3A](#Fig3){ref-type="fig"}), were enriched in two comparisons (egg vs 1^st^ instar, 5^th^ instar vs adult male). The oxidative phosphorylation pathway plays a critical role in the supply of energy^[@CR23]^. The enrichment analysis indicated that energy metabolism increased during development in *A. cincticrus*. In addition, other genes in this pathway encoding transport oxidoreductase and heat shock proteins (*hsp*s) were upregulated as well. Glutathione metabolism was downregulated in two comparisons (5^th^ instar vs adult male, adult male vs adult female), and arginine and proline metabolism were downregulated in two comparisons (1^st^ vs 5^th^ instar, adult male vs adult female). Inositol phosphate metabolism and phosphatidylinositol signaling system were downregulated in egg vs 1^st^ instar nymph (Fig. [3B](#Fig3){ref-type="fig"}). The changes in phosphatase in *Triatoma* saliva could interfere with anti-haemostatic response and facilitate their feeding^[@CR24]^.Figure 3Enrichment KEGG pathways for the upregulated (**A**) and downregulated (**B**) SDEGs in comparisons of different life stages. In the GO enrichment of BP, the most enrichment of upregulated SDEGs, including metabolic process, cellular process, single-organism cellular process, signal transduction, proteolysis, and oxidative phosphorylation, were shared by two comparisons (egg vs 1^st^ instar and 5^th^ instar vs adult male) (Fig. [S5A](#MOESM1){ref-type="media"}). The most enrichment of upregulated SDEGs in the CC (Fig. [S5B](#MOESM1){ref-type="media"}) included integral component of membrane, cytoplasm, membrane, lipid particle, plasma membrane and microtubule associated complex, were shared in two comparisons (egg vs 1^st^ instar and 5^th^ instar vs adult male). As for the MF (Fig. [S5C](#MOESM1){ref-type="media"}), two comparisons (egg vs 1^st^ instar and 5^th^ instar vs adult male) revealed five common GO enrichment categories of upregulated SDEGs, including catalytic activity, binding, transferase activity, nucleic acid binding and ATP binding. Four comparisons (egg vs 1^st^ instar, 1^st^ vs 5^th^ instar, 5^th^ instar vs adult male, adult male vs adult female) shared one common GO term (catalytic activity). Serine protease, involved in general digestion and protein metabolism^[@CR7],[@CR25]^, was identified in catalytic activity indicating that the organism was undergoing continuous growth. Together, the KEGG pathway and GO enrichment analyses identified a diverse group of development-related genes that undergo differential transcriptional responses during the developmental stages and predation process. These genes include maternal genes, cuticular proteins, metabolite synthases, hormone-related genes (Table [S5](#MOESM1){ref-type="media"}), venom-related genes, aggression-related genes and olfactory-related genes (Table [S8](#MOESM1){ref-type="media"}). Genes putatively involved in predation {#Sec6} -------------------------------------- By comparing the transcriptomes from different life stages, we identified 115 genes involved in predation, including 37 venom-related, 23 aggression-related and 55 chemosensory-related genes (Table [S9](#MOESM1){ref-type="media"}). The venom-related genes included *esterase*, *trypsin-like protease*, *trypsin precursor*, *lipase*, *venom serine protease 34-like*, *serine carboxypeptidase 1*, *lysosomal acid phosphatase precursor*, etc. It appeared that these genes exhibited low expression level in egg stage and relatively high expression level in nymph and adult stages. The transcript levels of *trypsin-like protease*, *trypsin precursor* and *esterase* showed the similar changing trend to that of the most venom-related genes, except *esterase* being low in 5^th^ instar (Fig. [4A--C](#Fig4){ref-type="fig"}). The majority of these genes were enriched in hydrolase activity, catalytic activity and peptidase activity.Figure 4Temporal distribution of ten SDEGs involved in predation. (**A**--**C**) Three venom-related genes. (**D**--**F**) Three aggression-related genes. (**G--J**) Four olfactory-related genes. The transcript levels of SDEGs were calculated as the log~10~ FPKM of each comparison in transcriptome analysis and were shown on the y-axis. The aggression-related genes included *tyramine receptor*, *octopamine receptor*, *cyp6Q20*, *deaf1*, *metabotropic glutamate receptor B*, *progesterone*, *androgens*, *serotonin*, etc. Apparently egg had the relatively low expression of these genes compared to other stages. As in the nymph stages, no obvious trends were observed regarding the expression of most aggression-related genes. 5^th^ instar nymph had the relatively low expression of *tyramine receptor* (Fig. [4D](#Fig4){ref-type="fig"}) and highest expression level of *octopamine receptor* (Fig. [4E](#Fig4){ref-type="fig"}). *Androgens* and *serotonin* with a significant downregulated expression in the 3^rd^ instar. Adult male had the higher expression of most aggression-related genes, e.g. *tyramine receptor*, *metabotropic glutamate receptor B* (Fig. [4F](#Fig4){ref-type="fig"}), *deaf1*, *trp*, *progesterone*, *androgens*, and *serotonin* (Table [S9](#MOESM1){ref-type="media"}), than adult female. The enrichment GO categories were signal transducer activity, cell communication and G-protein coupled receptor signaling pathway. We identified 22 *OBPs*, 17 *CSPs*, 11 *olfactory receptors* (*ORs*), four *sensory neuron membrane proteins* (*SNMPs*) and one *pheromone-binding protein* (*PBP*)-like transcripts. *OBPs* and *CSPs* exhibited relatively low expression in egg stage and high expression in nymph and adult stages (Fig. [4G,H](#Fig4){ref-type="fig"}). *PBP* exhibited relatively low expression in egg, 2^nd^ instar and 5^th^ instar, and showed high expression in the other life stages (Fig. [4I](#Fig4){ref-type="fig"}). *SNMPs* seemed to have a constant expression during all life stages (Fig. [4J](#Fig4){ref-type="fig"}). Adult male had the higher expression levels of most *ORs* than other stages. These genes were mainly enriched for odorant binding, response to stimulus, phosphorylation as well as in the olfactory receptor pathway. Trend analysis {#Sec7} -------------- To understand the expression patterns of the 13,479 SDEGs, gene data from all life stages were clustered into 50 model profiles (Fig. [S6](#MOESM1){ref-type="media"}). 4,183 genes represented 13 significant gene expression patterns, and 33 of 115 SDEGs related to predation could be clustered into five profiles with significance (*p* \< 0.05). The genes assigned to most profiles showed variation in nymph stages and demonstrated a bias between genders, e.g. *trypsin precursor*, *trypsin-like protease* and *esterase* showing a bias towards females; *metabotropic glutamate receptor B* and *ORs* was observed as a male-biased expression. qRT-PCR validation {#Sec8} ------------------ To further confirm the quality of the transcriptome, we compared the expression patterns of the seven libraries, including those from 1^st^ to 5^th^ instar nymphs, male and female adults, of ten selected SDEGs (three venom-related, three aggression-related, four olfactory-related genes) involved in predation using qRT-PCR (Fig. [5](#Fig5){ref-type="fig"}). The expression trend of seven SDEGs in qRT-PCR analysis was consistent with that detected in the transcriptome analysis earlier (Figs [4](#Fig4){ref-type="fig"} and [5](#Fig5){ref-type="fig"}). Inconsistencies were found in the three SDEGs, *tyramine receptor*, *OBP* and *SNMP*, which were different in some stages between qRT-PCR and transcriptome analysis. This difference in gene expression might be caused by the difference in the accuracy of these two assay methods. The SDEGs analyzed in this study were unigenes obtained from transcriptome that were assembled and mapped, while qRT-PCR showed a lower sensitivity than transcriptome sequencing. Nevertheless, qRT-PCR analysis confirmed the direction of change detected by transcriptome analysis, indicating that our results are reliable.Figure 5The dynamic expression patterns of ten SDEGs by using qRT-PCR. (**A**--**C**) Three venom-related genes. (**D**--**F**) Three aggression-related genes. (**G**--**J**) Four olfactory-related genes. The relative transcript level of each SDEGs was shown as the mean ± SE. Significant differences among different life stages were indicated by different letters (*p* \< 0.05, *t*-test). Discussion {#Sec9} ========== In this study, we sequenced and characterized the transcriptome from the different life stages of *A. cincticrus*, with a particular focus on SDEGs involved in the development and the predation process. The transcripts displayed a variety of interesting differential expression patterns observed across all life stages. The characteristics of gene expression patterns between life stages and predation related unigenes were discussed below. Genes involved in development {#Sec10} ----------------------------- During its life cycle, *A. cincticrus* undergoes incomplete metamorphosis. Many crucial genes and transcriptional factors, such as *nanos-like protein*, *hunchback* and *serine protease snake-like*, participate in embryonic development of model organisms. For instance, the *nanos* and *snake* gene are core components of the germplasm and are involved in the formation of anterior-posterior body axis during early embryogenesis^[@CR21]^. These genes were categorized in the GO terms of 'catalytic activity' and 'hydrolase activity', which are the essential first steps in the eggs activation cascade^[@CR26]^. In this study, the high expression of genes related to embryonic development revealed the ongoing drastic cellular differentiation process in the eggs. The nymphs should continually prey on ants to sustain the energy for molting through five nymph stages. They begin movement right after hatching from the eggs and feed on ants to store energy for development and various activities. Genes participating in muscle growth^[@CR27]^ including *histone H2A* and genes pertaining to metabolic detoxification^[@CR28]^ including *glycosidases* and *cytochrome P450s* were identified in the transcriptomes of nymphs. *Glycosidases* have been previously proposed to aid in the digestion of endosymbiont cell walls in *Rhodnius prolixus* ^[@CR29]^. *Cytochrome P450s* are highly expressed in the "oxidoreductase activity", and are known to be involved in steroid and lipid metabolism^[@CR30]^ and can participate in insecticide resistance^[@CR31]^. Insect molting is rigorously choreographed and is coordinated by fluctuations in the juvenile hormone (JH) and ecdysone. JH plays a key role in regulating development and metamorphosis^[@CR32]^, while ecdysone-regulated genes play key roles during molting and metamorphosis of *Oncopeltus fasciatus* ^[@CR21]^. We discovered that the expression of *ecdysone-induced protein* and *juvenile hormone binding protein* was high in the early nymphal stages. In general, the wings in reduviids become externally visible during the 2^nd^ instar nymph, and they grow larger at each successive instar, then eventually they are fully developed when reduviids molt into adulthood^[@CR33]^. We herein found *notch*, which is critical for wing disc development by secreting signal molecules^[@CR34],[@CR35]^, were activated and highly expressed in the 2^nd^ instar nymph of *A. cincticrus*. In the final stage of development, adults reach sexual maturity followed by the behaviors like courtship, mating and oviposition. The *hsps* in *Drosophila* ^[@CR36]^ potentially mediate the cellular response to thermal stress and indirectly influence the downregulation of sex steroid receptors, resulting in repressing the proliferation of growing follicles in proestrus, and therefore impact the organism's fecundity and longevity^[@CR37]^. Expression of *hsps* was lower in adult female than adult male. The KEGG enrichment term "oocyte meiosis" was enriched among the upregulated genes in the comparison of transcriptomes from male vs female adults, with *vitellogenin* highly expressed in the females. *Vitellogenin* encodes the vitellin in oocytes, and is expressed as an egg yolk precursor in female insects and sterile worker bees, as well as in other insects^[@CR37]^. Genes involved in predation {#Sec11} --------------------------- The predatory process of reduviids involves the following steps: arousing, locating, approaching, paralyzing, sucking, releasing and cleaning^[@CR4]^. Neuroactive ligand-receptor interaction, wnt signaling pathway, glutathione metabolism, arginine and proline metabolism are important metabolic pathways in managing information and protein digestion. These processes are coordinated by a series of genes. SDEGs involved in predation were enriched in these pathways in nymphs and adults. Because reduviids can only ingest liquid food, the salivary secretions play a critical role in paralyzing and feeding of their prey. The toxic protein content of venomous saliva in reduviids varies with respect to males and females^[@CR38]^. Records of adult female reduviids paralyzing their prey more rapidly than adult males reinstate the above fact^[@CR39]^. The *esterase* was abundant in the saliva of *A. cincticrus* may likely aid in paralysis of prey^[@CR18]^. In the salivary glands, *trypsin-like mRNA* is dominant which is involved in cleaving the proteins at lysine and arginine residues^[@CR40]^. The transcript levels of most SDEGs were low in eggs, relatively high in nymph and adult stages and exhibited a bias towards adult female. This result was consistent with the eggs being in a static state, while the nymphs feed on mass of ants for development and the females reserve energy to oviposition. The females have high expression of *trypsin precursor*, *trypsin-like protease* and *esterase*, leading to the stronger toxicity of the saliva than males and obtaining more ants due to their ability in paralyzing the prey more rapidly than the males. The activation and upregulation of these genes are critical for successful paralyzing and digestion of the prey. Success rate of aggression can be influenced by several factors, such as physical disparities (e.g. size, strength and weapon)^[@CR41],[@CR42]^, as well as physical exertion and experience in previous fights^[@CR43],[@CR44]^. For example, different size of the parasitoid wasps has a different fighting propensity and weak-fighting ones usually stay hidden to avoid the stress of fighting^[@CR45]^. Males usually are more likely to respond aggressively than females^[@CR10]^ due to the limited opportunities to mate^[@CR46]^. Monoamines, neuropeptides, and pheromones have been implicated as important neuromodulators for agonistic displaying in insects^[@CR47]^. The expression of *cyp6Q20* in olfactory sensory organs has also been reported to regulate hereditary-related and environmental-related aggression in *Drosophila melanogaste*r^[@CR48]^, suggesting a function in response to aggression-related stimuli. These results suggest that changes in *octopamine receptor* of nymph during an aggressive confrontation may be related to sufficient amounts of *octopamine* that induce the initiation, level and duration of aggressive behaviors^[@CR49]^. We found that the *metabotropic glutamate receptor B* was highly abundant in *A. cincticrus* adult male, which is implicated in the regulation of aggression in honey bee^[@CR11]^ and in the competition of reproduction in *Nasonia vitripennis* ^[@CR9]^. The *progesterone* synthesized by the adrenocortical tissue was the highest in males, and may act indirectly on the brain after local conversion to other steroids that affect aggressive behavior. Allopregnanolone may be synthesized largely in males and is a neurosteroid that has been shown to play a role in regulating mammalian aggressive behavior^[@CR50],[@CR51]^. These results of gene expression are consistent with the behavior that occurs in nature, i.e., *A. cincticrus* males exhibiting stronger aggressive behaviors against ants and other males to obtain food and win the mating chance. Similarly, *A. cincticrus* adults are generally larger in body size and inhabit in larger range of space suggesting they may exhibit stronger aggression than nymphs. Nymphs usually live underneath the stone, most often gregariously, likely consistent with their aggressive ability to avoid stress of fighting. Thus, it is not difficult to understand the camouflaging behavior in nymphs to protect themselves when facing danger or their enemies. Insect predation is known to activate a series of mechanisms in order to locate prey, accompanied by sensitive identification of odor. Signal transmission from receptors to higher central nervous system rely on certain proteins, such as olfactory receptors (ORs) and OBPs^[@CR52]^, that could discern stimulants of insects. For instance, green lacewing *Chrysopa pallens* use CpalOBP2 to detect and recognize the alarm pheromone in aphid species^[@CR5]^. Hull *et al*.^[@CR53]^ speculated that the plant bug *Lygus lineolaris* presented more *OBP*-like transcripts because of its broad host range. However, in pea aphid, there were relatively small number of *OBP*s owing to its parasitic lifestyle and specialized ecology^[@CR54]^. Hekmat-Scafe *et al*.^[@CR55]^ suggested that a moderate number of OBPs could act in a combinatorial manner with a moderate number of ORs to greatly increase the recognition power of an insect's olfactory system. *ORs* were found highly expressed in male antennae of Asian corn borer due to its specific olfactory behavior and finding the emerging female^[@CR56]^. Antennae-enriched CSPs may be involved in *C. pallens* identifying and binding volatile from pests, such as aphids, or pest-damaged plants^[@CR57]^. Interestingly, the transcript abundance of these genes varied in different life stages of *A. cincticrus*. The transcripts were abundant in nymph and adult stages, suggesting a potential role in the detection and discrimination of odors. Surprisingly, greater sex bias was seen for some transcripts between genders, and *OBP*s were differentially expressed in nymph stages. The majority of these olfactory genes are highly expressed in the olfactory organ of female antennae than that of adult male, in contrast to their usual function for adult male to detect mates at long range, may be useful for females to detect pheromones released from ants in order to locate them accurately. Male-biased expression was observed in *ORs* in antennae, suggesting that they may mediate olfactory behavior specific to males, for example finding emerging adult female. The high expression of *OBPs* in *A. cincticrus* nymphs might suggest the unique functions in chemoreception. Compared with the egg stage, the nymphs and adults are active and had the need to detect odorants for their survival and reproduction in the environment. Methods {#Sec12} ======= Collection and colony maintenance {#Sec13} --------------------------------- The colony of *A. cincticrus* used in this study was originally obtained from Yu County (39.9465°N, 114.9379°E), Hebei, China, then was maintained in a light incubator at 25 ± 2 °C, 60 ± 5% relative humidity (RH) under a 16 h: 8 h light/dark cycle. Specimens of eight life stages were collected as eggs, 1^st^ to 5^th^ instar nymphs, male and female adults. These samples were immediately frozen in liquid nitrogen and stored at −80 °C until RNA extraction. RNA isolation, cDNA library construction and Illumina sequencing {#Sec14} ---------------------------------------------------------------- Total RNA was isolated using TRIzol reagent (Invitrogen, USA) according to the manufacturer's protocols. RNA integrity was assessed using the RNA Nano 6000 Assay Kit in the Agilent Bioanalyzer 2100 system (Agilent Technologies, USA). The extracted RNA was used to establish a cDNA library for transcriptome sequencing using an Illumina Hiseq. 2500 sequencer platform at Biomarker Technologies Corporation (Beijing, China). Raw reads cleaning and quality control {#Sec15} -------------------------------------- In order to guarantee high-quality and clean reads, raw reads containing more than 5% unknown nucleotides and low-quality reads containing more than 50% of bases with a Q-value ≤ 20% were excluded from the sequence assembly. The clean reads were subjected to subsequent analyses and deposited in the NCBI SRA database under accession numbers: SRR5099968 (egg), SRR5099969-73 (1^st^ to 5^th^ instar nymphs), SRR5106081 (adult female) and SRR5106242 (adult male). Assembly and functional gene annotation {#Sec16} --------------------------------------- The clean reads were assembled using Trinity (<http://trinityrnaseq.sourceforge.net/>) with default assembly parameters. Clean reads with a certain overlap length were initially combined to form long fragments without N (named contigs). Related contigs were clustered using the TGICL software^[@CR58]^ to yield unigenes (without N) that cannot be extended on either ends, and redundancies were removed to acquire non-redundant unigenes. Subsequently, non-redundant unigenes were analyzed and annotated. Unigenes were mapped against the NCBI-NR, SwissProt and COG databases using the BLASTX algorithm with an e-value cut-off of 10^−5^ to retrieve functional annotations based on sequence similarity. High-priority databases (followed by NBCI-NR, SwissProt, and KEGG) were selected to determine the direction of the unigene sequences. According to the best alignment results, the coding sequences were generated. The ESTScan software^[@CR59]^ was used to determine the sequence direction of the unigenes that could not be aligned to any of the above databases. Functional gene annotations were collected for all unigene sequences ≥ 150 bp using Blast2GO^[@CR60]^. To determine the distribution of gene functions at the macro level, Blast2GO was used to retrieve the GO terms for each sequence by searching the NCBI-NR database. The WEGO software^[@CR61]^ was used to plot the distribution of the GO functional classification of the unigenes. Functional pathway analysis was performed using the KAAS webserver^[@CR62]^ from KEGG. Differential gene expression and trend analysis {#Sec17} ----------------------------------------------- Gene expression levels were calculated by RSEM^[@CR63]^. The read counts were adjusted by the eDEGR program package through one scaling normalized factor. The DEGSeq package was used to analyze the gene expression difference between two groups. DEGSeq provided statistical routines for determining differential expression in unigenes using a model based on the negative binomial distribution. *P*-values were adjusted using the Benjamini-Hochberg's approach for calculating the false discovery rate (FDR). Fragments per kilobase of exon per million fragments mapped (FPKM) was used to quantify the expression level of unigenes. Genes with an adjusted *p* \< 0.05, FDR ≤ 0.01 and the absolute value of log~2~ Fold change (FC) ≥ 1 were set as the thresholds to identify the SDEGs. FC is the ratio of FPKM between the two samples. GO enrichment analysis of SDEGs was carried out based on an algorithm provided by GOstat^[@CR64]^, with the complete annotation results set as the background. The *p*-value was approximated using the Chi-square test. Fisher's exact test was used when any expected value was below 5. This program was implemented as a pipeline^[@CR65]^. STEM (Short Time-Series Expression Miner) software was used to profile gene expression time series and identify significant expression tendencies^[@CR66]^ in these life stages based on FPKM values. Based on different signal density of genes under different situations, we identified a set of unique model expression tendencies. The raw expression values were converted into log~2~ ratios. Some unique profiles were further defined using a strategy for clustering short time-series gene expression data. Each model profile includes the actual or the expected number of genes. Fisher's exact test and multiple-comparison tests were used to determine whether the significant profiles had higher probability than expected^[@CR67]^. qRT-PCR validation {#Sec18} ------------------ To validate the quality of the transcriptome data and estimate the relative expression, qRT-PCR was performed using the SYBR Premix Ex Taq (Takara, Japan) according to the manufacture's protocols and a CFX96 real-time PCR detection system (Bio-Rad, USA). Primer sequences of selected genes were designed by DNAMAN and were listed in Table [S10](#MOESM1){ref-type="media"}. Total RNA was extracted as described in the RNA isolation and cDNA was synthesized by Promega Goscript Reverse Transcription Systems (Fisher Scientific, USA). All qRT-PCR reactions had three biological and three technical replicates. The average threshold cycle (Ct) was calculated from values in all six replicates per sample. The *actin* gene was chosen as an endogenous control to normalize expression between different samples, as has been used in other insects^[@CR68]^. The 2^−△△CT^ method^[@CR69]^ was used to evaluate the relative gene expression. All data were statistically analyzed by independent sample students *t*-test using SPSS 17.0. The relative expression level of each unigene was presented as mean ± standard error (SE). Electronic supplementary material ================================= {#Sec19} Dataset 1 Supplementary information **Electronic supplementary material** **Supplementary information** accompanies this paper at 10.1038/s41598-017-12978-0. **Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This work was funded by the National Basic Research Program of China (No. 2013CB127600), the National Key Technology R & D Program of the Ministry of Science and Technology (No. 2012BAD19B00) and the Special Fund for Scientific Research (No. 2012FY111100). F.K., H.L. and W.C. designed and performed the research. F.K. and Y.Z. collected samples. F.K. and Y.Z. performed the molecular work. F.K. and H.L. analyzed the data. All authors discussed results and implications. F.K., H.L., S.L., Y.Z., X.Z. and W.C. wrote the manuscript. All authors have read and approved the final manuscript. Competing Interests {#FPar1} =================== The authors declare that they have no competing interests.

Background ========== ERK mitogen-activated protein (MAP) kinases are commonly activated by 7TMRs, which leads to a wide array of cellular processes including cell proliferation and cell differentiation. In the last decade, a tremendous amount of works have been dedicated to elucidate the cell signaling mechanisms whereby 7TMRs activate ERK. To achieve ERK activation, some 7TMRs, such as the lutropin receptor \[[@B1]\], rely solely on G protein activation and to second messenger production. Besides, several reports support the view that MAP kinase activation requires receptor internalization, mediated by β-arrestins \[[@B2]\]. Originally, β-arrestins have been viewed as responsible for receptor desensitization, by uncoupling an agonist-activated receptor from its effector G proteins, and then by driving the uncoupled receptor to clathrin-coated pits \[[@B3],[@B4]\]. β-arrestin-dependent internalization of 7TMRs involves the direct interaction of the carboxy-terminal part of β-arrestins with the β2-adaptin subunit of the adaptor protein (AP)-2 complex \[[@B5]\]. Mutation of two arginines in this region abrogates both the β-arrestin/AP2 interaction and the clustering of β2-adrenergic receptor into clathrin-coated pits \[[@B6]\]. Furthermore, β-arrestins bind directly to clathrin *in vitro*\[[@B7]\]. As endocytic adaptors, β-arrestins also interact with the small GTPase ADP-ribosylation factor (ARF)-6 and its exchange factor nucleotide-binding site opener (ARNO), and with the N-ethylmaleimide-sensitive fusion protein (NSF) \[[@B8]\]. In HEK 293 cells stimulated by isoproterenol, overexpression of β-arrestin V53D, or of a β-arrestin (319--418) peptide, both impaired in their receptor-binding ability \[[@B9]\], not only reduces the β2-adrenergic receptor internalization level, but also decreases ERK activation \[[@B10]\]. Likewise, inhibition of β-arrestin 1 or 2 expression by RNA interference levels off the isoproterenol-induced ERK phosphorylation \[[@B11]\]. Besides, fission of the clathrin endocytic vesicle from the plasma membrane is in part achieved by the GTPase dynamin. Overexpression of a defective K44A dynamin mutated in its catalytic domain \[[@B12]\] impairs both receptor internalization as well as ERK stimulation transduced by the δ-opioid receptor \[[@B13]\]. In sharp contrast, some 7TMRs, such as the α2a adrenergic receptor \[[@B14]\], activate ERK without being internalized, whereas some others, such as the metabotropic glutamate mGlu1 receptor, require β-arrestins to activate ERK, but not through their endocytosis-promoting ability \[[@B15]\]. Therefore, whether 7TMR-mediated ERK activation will depend on β-arrestin-promoted internalization or not seems to be a receptor-related issue. The follicle-stimulating hormone receptor (FSH-R) is a 7TMR whose main effector is adenylate cyclase \[[@B16]\]. Once bound to its agonist, the FSH-R gets phosphorylated by G protein-coupled receptor kinases (GRKs), recruits β-arrestins \[[@B17]\] and undergoes internalization \[[@B18]-[@B21]\]. Overexpression of β-arrestin 1 or 2 or of the β-arrestin (319--418) peptide respectively reduces or increases cAMP in response to FSH, as measured by a luciferase gene reporter assay \[[@B17],[@B22]\]. The FSH-R is expressed by two cell types of the gonad, namely Sertoli cells in the testis, and granulosa cells in the ovarian follicle \[[@B23]\]. ERK MAP kinases have been shown to be activated upon FSH stimulation of primary cultures of both cell types \[[@B24]-[@B26]\], and this signaling pathway mediates the mitogenic response of Sertoli cells to the hormone \[[@B24]\]. Previously, overexpression of β-arrestin 1 or 2 \[[@B21]\] or of the β-arrestin (319--418) peptide and of the dynamin K44A \[[@B20]\] mutant had been shown to affect the FSH-R internalization. But to date, nothing is known about the role of β-arrestin-dependent internalization in ERK activation by the FSH-R. Here, we addressed this question in HEK 293 cells transiently expressing the FSH-R, by enhancing internalization with overexpressed wild-type β-arrestins or by interfering with receptor internalization with the β-arrestin (319--418) construct or by the dynamin K44A mutant. Methods ======= Materials --------- Porcine FSH (apparent molecular weight = 33,500 g/mol) was purified by Dr Jean Closset (Université de Liège, Belgium) \[[@B27]\]. Amphotericin B, penicillin, streptomycin, glutamin, phenylmethylsulfonyl (PMSF), Na~3~VO~4~, leupeptin, pepstatin and aprotinin were from Sigma Chemical Co (St. Louis, MO). Dulbecco\'s minimum essential medium (DMEM), minimum essential medium (MEM) with Earle\'s salt, foetal calf serum (FCS), non essential amino acids, trypsin-EDTA were all from Gibco-BRL Life Technologies (Gaithersburg, MD). The Transfast™ transfection reagent was from Promega Corp., Madison, WI. Plasmids -------- The pRK-FSHR/3 was a kind gift of Dr R. Sprengel (Heidelberg, Germany). The pCMV5-rat β-arrestin 1 and pCMV5-rat β-arrestin 2 were gifts of Dr R.J. Lefkowitz (Durham, NC). The pcDNA3-β-arrestin (319--418) was given by Dr J.L. Benovic (Philadelphia, PA) and the pcDNA3-dynamin-K44A plasmid was provided by Dr S.L. Schmid (La Jolla, CA). Cell culture and transfection ----------------------------- HEK 293 cells were grown in MEM supplemented with 20 μM glutamin, 100 μM non essential amino acids, 10% heat-inactived FCS, 10 U/ml penicillin and 10 μg/ml streptomycin. HEK 293 cells were grown and transfected in 75 cm^2^flasks. Cells were maintained at 37°C in a humidified atmosphere of 5% CO~2~. Fifty to eighty % confluent cells in FCS-free medium were incubated for 1 hour with Transfast reagent (800 ng per cm^2^culture) and plasmid DNA encoding the FSH-R (200 ng per cm^2^) and β-arrestin (319--418) (600 ng per cm^2^) or wild-type β-arrestins (200 ng per cm^2^) or dynamin K44A (600 ng per cm^2^), unless otherwise stated. Empty plasmid was added in every culture wells to equalize transfected plasmid concentrations. Twenty-four hours after transfection, cells were treated for 90 sec with 3 ml 0.25% trypsin and 1 mM EDTA, centrifuged 10 min at 100 *g*and seeded in 9.6 cm^2^culture plates with a dilution factor of 0.4. Seventy hours after transfection, cells were FCS-starved for 2 hours in 1 ml, and then stimulated with 1 to 10 nM pFSH. Media were collected and cells were scrapped. For further Western blot analysis, HEK 293 cells were scrapped directly in Laemmli sample buffer (Tris HCl 0. 25 M pH 6.8, 5% SDS, 50% glycerol, 50 mM β-mercaptoethanol, 0.01% bromophenol blue). Western blots ------------- HEK 293 cell lysates were preincubated for 30 min at 37°C before gel loading. Samples were resolved by SDS-PAGE, electrophoretically transferred to polyvinylidene difluoride (PVDF) membranes (NEN Life Science Products, Boston, MA) and hybridized with the antibodies mentionned in the following. Anti-p44^ERK1^/p42^ERK2^rabbit polyclonal antibody (Cell Signaling Technology Inc.) was used at 1:1,000 dilution. The anti-ERK rabbit polyclonal antibody purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), was used at 1:10,000 dilution. The anti-arrestin A1CT polyclonal antibody kindly provided by R.J. Lefkowitz (Durham, NC), was used at a 1:10,000 dilution. It recognizes the C-terminus of β-arrestins, and therefore recognizes endogenous, overexpressed wild-type as well as overexpressed (319--418) β-arrestins. Primary antibodies were incubated with membranes in TBS (20 mM Tris, 150 mM NaCl) 0.1% Tween-20 supplemented with 5% unfat milk for 18 hours at 4°C under constant agitation. Horseradish peroxydase-coupled anti-rabbit antibody (Bio-Rad Laboratories Inc., Marnes-la-Coquette, France) was used at 1:5,000 dilution, to detect antigen-antibody interactions by enhanced chemioluminescence (NEN Life Science Products). To monitor protein loading, a first Western blot was hybridized with the anti-P-ERK antibody, the membrane was stripped 30 min at 50°C in 100 mM β-mercaptoethanol, 2% SDS, and rinsed twice for 10 min at room temperature in 150 mM NaCl, 0.05% Triton X100. Then, the membrane was reprobed with anti-ERK antibody. When β-arrestins were also probed with the A1CT antibody, a second SDS-PAGE was achieved with equal quantities of proteins as in the first gel, and Western blot was carried on. The ratio of phosphorylated ERK2 to total ERK2 was quantified using the ImageMaster 1D Elite version 4 Software (Amersham Biosciences, Arlington Heights, IL) and the results were expressed as phospho-ERK/ERK. ^125^Iodo-FSH labelling (iodination) ------------------------------------ Five μl (70 pmol) pFSH were incubated for 20 minutes with 25 μg of lyophylized iodo-gen and 2 μl containing 100 μCi ^125^Iodine with specific activity of 2,670 Ci/mmol. Separation of ^125^Iodo-FSH was performed on a Sephadex G50 column and the iodination efficiency was around 65%. Cell surface and internalized ^125^Iodo-FSH binding --------------------------------------------------- Internalization assay procedures were described previously \[[@B18]\]. Briefly, cells plated in 9.6 cm^2^wells were placed in 950 μl of Waymouth\'s MB752/1 containing 1 mg/ml bovine serum albumin (BSA) and 20 mM Hepes pH 7.4 for 2 h at 37°C. Each well received around 500, 000 cpm ^125^I-FSH (160 pM final concentration) alone or added with unlabelled pFSH (160 nM final concentration) in 50 μl. Total and non-specific binding were assayed in triplicates. At the indicated timepoints, cells were placed on ice and washed three times with 1 ml of cold Hanks\' balanced salt solution containing 1 mg/ml BSA. The surface-bound hormone was then released by incubating cells in 1 ml of cold 50 mM glycine, 150 mM NaCl, pH 3.0 for 4 min. The buffer was withdrawn and counted, the cells were collected with 300 μl of 0.5 N NaOH and were also counted. Expression of results and statistical analysis ---------------------------------------------- Results were expressed as mean + S.E.M., unless otherwise indicated. Comparison of the results was based on variance analysis. A probability (P) value below 5% was considered as significant. Results ======= FSH stimulates ERK phosphorylation in HEK 293 cells --------------------------------------------------- In HEK 293 cells transiently expressing the rat FSH-R, FSH stimulated ERK phosphorylation in a dose-dependent manner (Figure [1A, 1B](#F1){ref-type="fig"}). Based on these data, the optimal dose of FSH to be used in the following was determined as 3 nM FSH, which is close to the K~D~of FSH in testicular fractions \[[@B23]\]. In a time-course experiment, FSH stimulated ERK phosphorylation as soon as 2 minutes of exposure. ERK phosphorylation peaked around 6 min, then declined slowly, since 44% of the maximum phosphorylation was still observed by 15 min of FSH stimulation (Figure [1C, 1D](#F1){ref-type="fig"}). By 60 min, there was still more than twice as much the basal level of ERK phosphorylation (data not shown). No ERK phosphorylation was detected in cells exposed to a vehicle or in control cells transfected with an empty plasmid (Figure [1D](#F1){ref-type="fig"}). {#F1} Characterization of the FSH-R internalization --------------------------------------------- In order to evaluate whether blocking internalization impacts on FSH-stimulated ERK phosphorylation, it was first necessary to characterize the parameters of FSH-R internalization in our experimental conditions. For that purpose, cells were placed in the same experimental conditions as those used to visualize ERK phosphorylation, except that pFSH was replaced by I^125^-labeled pFSH. The kinetics of FSH binding we obtained (Figure [2A](#F2){ref-type="fig"}) were in agreement with reference data obtained in testicular fractions \[[@B28]\] and in HEK 293 cells exogenously expressing the FSH-R \[[@B28]\]. The ratio of internalized ^125^I-FSH binding *versus*total binding increased rapidly up to 10 min, then more slowly over time (Figure [2B](#F2){ref-type="fig"}), so that internalization of the FSH-R in response to FSH did not exceed 50% over time. {#F2} The dominant negative β-arrestin (319--418) construct or the dynamin K44A mutant reduces the ratio of internalized ^125^Iodo-FSH binding without altering FSH-induced ERK phosphorylation ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- We next investigated whether FSH-R internalization was required to generate the FSH-induced ERK response. Transfection of 600 ng of β-arrestin (319--418) or of the GTPase-deficient dynamin K44A mutant reduced the ratio of internalized FSH-R after 8 min of ^125^Iodo-FSH exposure, a timepoint when ERK phosphorylation peaked (Figure [3A](#F3){ref-type="fig"}). Even in the presence of 2 μg of β-arrestin (319--418) or dynamin K44A, it was not possible to decrease further the internalization level, suggesting that at maximum 40% inhibition of internalization could be reached in these conditions. In time-course experiments with ^125^I-FSH exposure from 4 to 60 min, the level of inhibition of FSH-R internalization by β-arrestin (319--418) (Figure [3B](#F3){ref-type="fig"}) or by dynamin K44A (Figure [3C](#F3){ref-type="fig"}) was constant over time. Transfection of β-arrestin (319--418), visualized by immunoreaction with the A1CT anti-arrestin antibody, did not alter ERK phosphorylation levels after 4, 8 or 12 minutes of FSH exposure (Figure [4A](#F4){ref-type="fig"}). Likewise, transfection of the K44A dynamin mutant left the FSH-stimulated ERK phosphorylation level unchanged. Figure [4B](#F4){ref-type="fig"} shows that there is no difference in the quantification of phosphorylated *versus*total ERK between cells transfected with β-arrestin (319--418) or dynamin K44A and control cells. Therefore, inhibiting β-arrestin- or dynamin-mediated FSH-R internalization did not alter FSH-induced ERK phosphorylation. {#F3} {#F4} Overexpression of wild-type β-arrestin 1 or 2 increases internalized ^125^Iodo-FSH binding without affecting FSH-induced ERK phosphorylation -------------------------------------------------------------------------------------------------------------------------------------------- Symetrically, overexpression of wild-type β-arrestin 1 or β-arrestin 2 enhanced the FSH-R internalization (Figure [5A](#F5){ref-type="fig"}). In time-course experiments with ^125^I-FSH exposure from 4 to 60 min, the increase in FSH-R internalization by β-arrestin 1 or 2 was constant over time (Figure [5B](#F5){ref-type="fig"}). β-arrestin 2 appeared slightly more effective than β-arrestin 1 in enabling internalization of the FSH-R. However, expression of both arrestins had no additive effect. β-arrestin overexpression led to more than 70% of the control level of FSH-R internalization. The actual level of overexpressed β-arrestins was individually immunodetected with the A1CT antibody (Figure [5C](#F5){ref-type="fig"}). The increase in FSH-R internalization obtained after transfection of 50 or 200 ng of β-arrestin 1 or of β-arrestin 2 did not alter FSH-induced ERK phosphorylation. Both quantities of β-arrestins led to a plateau in the FSH-R internalization level (Figure [5C](#F5){ref-type="fig"}, lower part). Figure [5D](#F5){ref-type="fig"} shows that there is no difference in the quantification of phosphorylated *versus*total ERK between cells overexpressing β-arrestin 1 or β-arrestin 2 and control cells. {#F5} In conclusion, these data show that an increase in β-arrestin-mediated FSH-R internalization level had no effect on the ERK response. Discussion ========== Our results clearly show that interfering with the FSH-R internalization by β-arrestin (319--418) or by dynamin K44A did not affect the ability of FSH to enhance ERK phosphorylation. Consistently, overexpression of β-arrestins 1 and 2 increased the percentage of FSH-R internalization but did not alter the ability of FSH to increase ERK phosphorylation. Therefore, we conclude that β-arrestin-dependent internalization is not required for FSH-induced ERK phosphorylation. Our data are concordant with previous reports showing that expression of β-arrestin (319--418) or of dynamin K44A impaired the FSH-R internalization rate \[[@B18],[@B20],[@B29]\]. Here, we extended these works, but even by increasing doses of interfering mutants, and in kinetics experiments, the FSH-R internalization could not be inhibited by more than 50%, in agreement with these reports. To validate our experimental framework, we also confirmed that overexpression of β-arrestin 1 \[[@B21]\] and β-arrestin 2 \[[@B20],[@B21]\] enhanced the FSH-R internalization rate. By a similar approach, it is possible to enhance ERK phosphorylation in response to angiotensin II \[[@B30]\]. Interestingly, our results suggest a difference in the efficiency of β-arrestins 1 and 2 to promote FSH-R internalization, in contrast to previously reported data \[[@B21]\]. Non-redundant functions for β-arrestins have previously been documented for other receptors. For example, whereas internalization of the agonist-induced β2-adrenergic receptor is insensitive to β-arrestin-1 depletion by RNA interference, β-arrestin-2 depletion has a dramatically inhibitory effect \[[@B31]\]. In addition, a reduction in β-arrestin 2, but not in β-arrestin 1 expression by siRNA, inhibits the ability of Angiotensin II to increase ERK phosphorylation, in HEK 293 cells transfected with the Angiotensin II type 1A receptor \[[@B32]\]. Likewise, vasopressin type 2 receptor-mediated ERK phosphorylation is sensitive to β-arrestin 2, but not to β-arrestin 1 suppression \[[@B33]\]. In sharp contrast, β-arrestin 1 and 2 have similar effects on parathyroid hormone receptor-mediated \[[@B34]\] or on β2-adrenergic receptor-dependent ERK activation \[[@B11]\]. Therefore, whether or not each β-arrestin has a specific action seems to vary according to the receptor. Albeit the FSH-R internalization level does not seem crucial for FSH-induced ERK phosphorylation, we cannot exclude an internalization-independent role of β-arrestins to achieve this response. Importantly, in the last years, β-arrestin function has extended far beyond receptor desensitization and endocytosis, by virtue of their ability to scaffold MAP kinase signaling modules, as well as elements of the inositide-dependent pathways \[[@B35],[@B36]\]. It has been extensively reported that β-arrestins can scaffold an entire MAP kinase cascade, including a MAPKKK such as Raf-1 or ASK1, a MAPKK such as MKK4/7 or MEK1/2 and a MAPK such as ERK \[[@B37]\], JNK \[[@B38]\] and p38 \[[@B39]\]. The mechanism of ERK activation is a crucial determinant of ERK-induced physiological response. A two-step process has recently been proposed to lead to ERK activation in response to 7TMR agonists: an early G protein-dependent mechanism followed by a β-arrestin-mediated mechanism leading to ERK phosphorylation by 10 minutes of Angiotensin II exposure \[[@B40]\]. Whereas the early and transient G-protein-activated ERK translocates to the nucleus to phosphorylate its target transcriptional regulators, the late and sustained β-arrestin-activated ERK accumulates in a pool of cytoplasmic vesicles \[[@B41]\] to constrain ERK activity to phosphorylation of its extra-nuclear targets. Likewise, retention of the proteinase-activated receptor-2 in the cytoplasm by a β-arrestin-containing signaling complex retains ERK in the cytoplasm, thus preventing a mitogenic response \[[@B37]\]. Therefore, fine-tuning the mechanism of ERK activation by G protein and/or by β-arrestins is a major determinant of an agonist-induced ultimate cellular response. Our own studies on the FSH-R using β-arrestin siRNA indicate that prolonged ERK phosphorylation upon FSH stimulation requires β-arrestins (See additional file 1: Kara *et al*., submitted). Therefore, since β-arrestins are required for FSH-induced ERK phosphorylation, and since ERK phosphorylation occurs even though internalization is markedly reduced, our results raise the appealing possibility that FSH would activate ERK at the plasma membrane. β-arrestins would assemble their MAPK signaling modules at the plasma membrane, or, alternatively, they would recruit a preformed complex, similarly to the JNK3 signaling module \[[@B38]\]. This point requires further investigations. In granulosa cells, B-Raf, Rap-1 and MEK are constitutively active and lead to ERK phosphorylation. However, ERK is constitutively dephosphorylated by a phosphatase which is blocked by FSH, in a G-protein/PKA-dependent manner \[[@B26]\]. This leads to ERK translocation to the nucleus, as further substantiated by our own data obtained in Sertoli cells \[[@B24]\]. These results do not exclude that β-arrestins could intervene later in the kinetics of activation, to sequester activated ERK in the cytosol, as previously reported for other GPCRs, such as the AT1a-R \[[@B30]\] or the PAR-2 \[[@B37]\]. In conclusion, our study provide clear evidence that, in contrast to many 7TMRs, β-arrestin-mediated FSH-R internalization is not required for ERK activation by FSH. Acknowledgements ================ The authors would like to thank Dr R.J. Lefkowitz, Dr R. Sprengel, Dr J.L. Benovic, and Dr S. Schmid for kindly providing plasmids and antibodies. E.K. was funded by fellowships from the Région Centre, the Fondation pour la Recherche Médicale, and the Fonds d\'Aide à la Recherche Organon. This work was funded by the Institut National de la Recherche Agronomique and by the Centre National de la Recherche Scientifique.

Background {#Sec1} ========== The bacteria *Streptococcus pneumoniae* (pneumococcus) and *Haemophilus influenzae* type b (Hib) are leading causes of childhood pneumonia and meningitis and are major contributors to worldwide mortality in children younger than 5 years of age. In 2011, an estimated 411,000 deaths occurred worldwide due to pneumococcal pneumonia and 197,000 due to Hib pneumonia \[[@CR1]\]. In China, it is estimated that there are 261,000 cases and 11,000 deaths each year in children under 5 years of age due to pneumococcal pneumonia and meningitis \[[@CR2]\], and 19,000 childhood deaths result from Hib infection \[[@CR3]\]. Pneumococcus and Hib colonize the upper airways. Asymptomatic carriers can still transmit to other individuals, with disease resulting if the pathogens descend the airway into the lower respiratory tract \[[@CR4], [@CR5]\]. Over 90 different serotypes of pneumococcus have been found, and vaccinations need to strike a reasonable balance between the financial cost against the value of including more serotypes known to cause disease \[[@CR6]\]. In contrast to the serotypic diversity with pneumococcus, Hib is a much more predominant cause of disease than non-type b *H. influenzae* \[[@CR7]\]. The Hib conjugate vaccine has been available since 1987 in the US \[[@CR7]\] and 2000 in China \[[@CR8]\], and a pneumococcal conjugate vaccine (PCV) was licensed in the US in 2000 \[[@CR9]\] and introduced to China in 2008 \[[@CR10]\], although it was taken off the market in China in 2014 \[[@CR11]\]. Neither the Hib vaccine nor PCV are included on the publically-funded Expanded Program on Immunization (EPI) in China, but they are available at immunization clinics for a fee. The Hib vaccine costs \$13--\$17 whereas PCV is much more expensive (\$127) \[[@CR10]\]. As a result coverage of Hib vaccine is much higher than PCV (50.9 % versus 11.4 % in Shanghai); moreover, although infants as young as 6 weeks are eligible for Hib vaccine and PCV administration, vaccination is typically delayed until after 1 year of age. Coverage at 12 months is around 20 % for Hib vaccine and \<5 % for PCV and at 24 months, coverage is approximately 40 % for Hib vaccine and 5 % for PCV \[[@CR12]\]. The first PCV protected against 7 serotypes of pneumococcus (PCV-7), but more recently, 10-valent (PCV-10) and 13-valent (PCV-13) vaccines have been licensed \[[@CR6]\], and a 15-valent vaccine (PCV-15) is under development \[[@CR13]\]. The World Health Organization (WHO) recommends that countries choose a PCV depending on the distribution of serotypes within their population, along with vaccine supply and cost-effectiveness considerations \[[@CR6]\]. Two studies in China have previously evaluated the coverage between PCVs and serotypes in the population. A study of clinical isolates from 8 cities in China during 2005 and 2006 found coverage of PCV-7 was 62.6 %, PCV-10 was 64.8 %, and PCV-13 was 79.6 % \[[@CR14]\]. Similarly, a study of clinical isolates from 2006 to 2008 found coverage of 60.3, 66.7 and 87.8 % for the 3 vaccines, respectively \[[@CR15]\]. Because the distribution of pneumococcal serotypes may vary geographically, however, these findings may not be representative of potential coverage from vaccines in all areas. Few studies have been published from China in the English literature looking at Hib and pneumococcal carriage. This information is important for modeling the burden of disease and underscoring the importance of vaccination, particularly in a country with low coverage of Hib and pneumococcal vaccines but with widespread resistance to antibiotics \[[@CR16], [@CR17]\]. In this study, we calculate the prevalence of nasopharyngeal carriage of Hib and pneumococcus (including specific pneumococcal serotypes), describe the prevalence of antimicrobial resistance in children with carriage, and examine risk factors for nasopharyngeal colonization with pneumococcus or Hib. Methods {#Sec2} ======= In 2009, April and May for spring, October and November for autumn, healthy children aged 12--18 months were recruited from a total of 5 immunization clinics in Shanghai. We purposely sampled in both the spring and autumn in order to determine if there was a seasonal fluctuation in carriage (although this has not been borne out in previous studies in the UK \[[@CR18]\] or Italy \[[@CR19]\]). These clinics purposefully selected from 3 districts (Huangpu, Xuhui, and Pudong) to represent different areas of the city. For 1 or 2 days, research staff from the municipal Centers for Disease Control and Prevention (CDC) and district CDCs attended these clinics, and enrolled children between 12 and 18 months who attended the clinic on those days. Children were excluded for the following reasons: 1) antibiotic use within 15 days of enrollment; 2) congenital anomalies of the nasopharynx; 3) existence of a long-term infectious disease (such as chronic otitis media, or chronic sinusitis); 4) previous immunization for pneumococcal disease; 5) coagolopathy; or 6) body temperature over 38 °C at time of enrollment. Parents of the children were asked about their residency status (local or Shanghai residency vs non-local residency), their household monthly income, daycare attendance, household size, number of children, highest educational level attainment of father and mother, smoking in the household,, if child was breastfed, child history of serious disease, and number of doses of Hib vaccine the child had received. Research staff obtained a nasopharyngeal sample using one infant-sized calcium alginate swab (Thermo Fisher Scientific, Hampton, NH, USA) for each child. These specimens were streaked on Columbia agar and chocolate agar immediately upon collection, then transported in 12 h to the Shanghai Municipal CDC for culturing and preliminary pathogen identification. The swabs were streaked directly onto the plates. Positive isolates were stored at −70 °C, then transported to a centralized laboratory in Beijing for pathogen confirmation, pneumococcal serotyping, and antibiotic resistance testing. The site of specimen collection (nasopharynx), swab material (calcium alginate fiber), storage temperature (−70 °C), and choice of agar (5 % sheep blood) followed standard methods \[[@CR20]\]. Isolate confirmation {#Sec3} -------------------- The specimens streaked on the Columbia and chocolate agar were incubated at 37 °C in 5 % CO~2~-enriched atmosphere for 24--48 h. Identification of pneumococcal colonies was based on the following conventional microbiological methods: colony morphology, growth on Columbia agar, susceptibility to optochin, and bile solubility. Colony morphology, growth on chocolate agar, and the X + V factor were requirements for identification of *H. influenzae* colonies. For both pneumococcus and Hib, a diameter in the inhibition zone \>7 mm was recorded as a positive result. *S. pneumoniae* isolates were serotyped by the Neufeld (Quellung) reaction, by inoculating a broth with culture from the agar plate, and then mixing a selection from the broth with antiserum and determining if a reaction occurred. Antimicrobial susceptibility tests {#Sec4} ---------------------------------- Minimum inhibitory concentrations (MICs) of anti-microbial agents were determined by E-test method. The pneumococcal and Hib isolates were tested for susceptibility to azithromycin, cefuroxime, ceftriaxone, levofloxacin, and moxifloxacin. Pneumococcal isolates were additionally tested for resistance to amoxicillin/clavulanic acid, erythromycin, and penicillin. *H. influenzae* isolates were tested for susceptibility to ampicillin. Standard references were included for quality control. Statistical analysis {#Sec5} -------------------- Descriptive statistics were used to illustrate the prevalence of Hib and pneumococcal carriage, antibiotic resistance, *S. pneumoniae* serotypes, season of enrollment, and other demographic characteristics of the enrolled children. For bivariate analyses, the Pearson chi-square test was used to test the association between dichotomous demographic factors and 1) carriage of *S. pneumoniae* and 2) carriage of Hib. The Cochran-Armitage Trend Test was used to compute *P*-values for ordinal variables (age of child, age of mother at childbirth, urbanicity, household income, size of house, mother's education, father's education, feeding pattern, and Hib vaccine history). Exact tests were used if any cell counts were ≤5. Correction for multiple testing was done through the sequential Holm-Bonferroni Method. For the multivariable analysis, the explanatory variables input into the logistic regression model were either related due to (1) a priori considerations (age of child), (2) selection criteria (season, urbanicity), or (3) variables found to be significant (*P*-value \<0.05) in the bivariate analysis. Mother's education level and father's education level were not placed in the same model together because of concerns about multicollinearity. Analyses were performed in SAS version 9.3 (SAS Institute, Inc, Cary, NC, USA). *P*-value correction for multiple testing was computed in R version 3.0.3 (R Foundation for Statistical Computing, Vienna, Austria). Results {#Sec6} ======= In this study, 6 parents refused to participate, leaving a total of 614 children aged 12 to 18 months enrolled: 308 in the spring and 306 in the autumn. Table [1](#Tab1){ref-type="table"} highlights the demographic and carrier state risk factors; 338 were male, and 160 were local residents, whereas 454 were non-local children whose family had relocated to Shanghai from another province. Children with PCV already administered were excluded from participating in the study, but 80.7 % of children had received at least one dose of Hib vaccine. No child in the study attended a daycare.Table 1Prevalence of nasopharyngeal carriage of pneumococcus and *Haemophilus influenzae* type b (Hib) and demographic and medical correlates among children in Shanghai, 2009CountPneumococcal carriage (%)*P*-value^a^Hib carriage (%)*P*-value^a^Overall614102 (16.6 %)49 (8.0 %)Season of enrollment0.07801.0000 Autumn30638 (12.4 %)23 (7.5 %) Spring30864 (20.8 %)26 (8.4 %)Sex1.00001.0000 Male33860 (17.8 %)30 (8.9 %) Female27642 (15.2 %)19 (6.9 %)Age of child0.15301.0000 12 months17035 (20.6 %)14 (8.2 %) 13--14 months18435 (19.0 %)16 (8.7 %) 15--16 months15820 (12.7 %)12 (7.6 %) 17--18 months10212 (11.8 %)7 (6.9 %)Age of mother at childbirth0.39841.0000 18--24 years17935 (19.6 %)18 (10.1 %) 25--29 years25744 (17.1 %)18 (7.0 %) 30--34 years12318 (14.6 %)9 (7.3 %) 35--44 years555 (9.1 %)4 (7.3 %)Household residency0.00090.0008 Local16011 (6.9 %)2 (1.3 %) Non-local45491 (20.0 %)47 (10.4 %)Urbanicity0.15300.0077 Urban42262 (14.7 %)25 (5.9 %) Suburban12622 (17.5 %)12 (9.5 %) Rural6618 (27.3 %)12 (18.2 %)Household income0.11660.0490 \<2000 yuan (\$292)23247 (20.3 %)26 (11.2 %) 2000--4999 yuan (\$292--\$730)28647 (16.4 %)21 (7.3 %) ≥5000 yuan (\$731)968 (8.3 %)2 (2.1 %)Size of house0.21630.0008 \<18 m^2^14532 (22.1 %)17 (11.7 %) 18--29 m^2^12020 (16.7 %)19 (15.8 %) 30--69 m^2^15625 (16.0 %)8 (5.1 %) ≥70 m^2^19325 (13.0 %)5 (2.6 %)Siblings0.15300.0999 No40757 (14.0 %)24 (5.9 %) Yes20745 (21.7 %)25 (12.1 %)Mother's education0.06500.0008 Primary school or less7318 (24.7 %)9 (12.3 %) Junior high28654 (18.9 %)32 (11.2 %) High school/vocational13416 (11.9 %)8 (6.0 %) College or more12114 (11.6 %)0 (0.0 %)Father's education0.04060.0003 Primary school or less234 (17.4 %)4 (17.4 %) Junior high29965 (21.7 %)34 (11.4 %) High school/vocational13516 (11.9 %)10 (7.4 %) College or more15717 (10.8 %)1 (0.6 %)Smoking in home1.00001.0000 No28847 (16.3 %)27 (9.4 %) Yes32655 (16.9 %)22 (6.7 %)Feeding pattern (first 6 months)0.86120.3392 Exclusive breastfeeding31758 (18.3 %)35 (11.0 %) Breastfeeding and formula20732 (15.5 %)7 (3.4 %) Only formula9012 (13.3 %)7 (7.8 %)Any history of disease0.68700.8154 No12326 (21.1 %)14 (11.4 %) Yes49176 (15.5 %)35 (7.1 %)Hib vaccine history1.00000.3392 None7114 (19.7 %)9 (12.7 %) 1 dose14622 (15.1 %)14 (9.6 %) 2 doses14824 (16.2 %)12 (8.1 %) 3 doses18231 (17.0 %)10 (5.5 %) 4 doses6711 (16.4 %)4 (6.0 %)^a^From chi-square test, except for age of child, age of mother at childbirth, urbanicity, household income, size of house, mother's education, father's education, feeding pattern, and Hib vaccine history, which were from Cochran-Armitage Trend Test. Exact tests were used if cell counts were ≤5. *P*-values were corrected for multiple testing through the sequential Holm-Bonferroni Method Overall, 137 (22.3 %) of participants were carriers of one or more pathogen. More children were carriers of pneumococcus (16.6 %) than Hib (8.0 %). Household residency and father's education were the two significant risk factors for pneumococcal carriage: more non-locals (20.0 %) than locals (6.9 %) were carriers (*P* = 0.0009), and children having a father with less education were more likely to be carriers (*P* = 0.0406). For Hib carriage, in addition to household residency and father's education, urbanicity, household income, size of house, and mother's education were significant risk factors. Urban dwellers had lower carriage (5.9 %) than those in suburban (9.5 %) or rural areas (18.2 %) (*P* = 0.0077), and having greater income and a larger house were protective against Hib carriage. Hib vaccination was not associated Hib carriage status (*P* = 0.3392). The most common pneumococcal serotype was 19 F, which was in 54 children. In descending order of prevalence, the other isolated serotypes present in the sample were 19A (*n* = 8), 6A (*n* = 6), 6B (*n* = 5), 6A/6B cross-reactive (*n* = 5), 23 F (*n* = 4), 15 (*n* = 4), 14 (*n* = 2), and 1 each for 17 and 22Twelve specimens were untypeable. The prevalences we found correspond to a coverage of 68.6 % (70) for PCV-7 and PCV-10, 82.3 % (84) for PCV-13, and 83.3 % (85) for PCV-15. Table [2](#Tab2){ref-type="table"} shows the frequency of antimicrobial resistance for each pathogen. All 102 pneumococcal isolates were sensitive to levofloxacin and moxifloxacin. Highest levels of resistance were to azithromycin (51.0 %) and erythromycin (51.0 %). Most Hib isolates were sensitive to all antibiotics, except 12.2 % were resistant to ampicillin.Table 2Antibiotic Resistance among *S. pneumoniae,* and *H. Influenzae type b* isolates in 614 Shanghai children, 2009Antibiotic*S. pneumoniae* (*n* = 102)*H. influenzae* type b (*n* = 49)SensitiveIntermediaryResistantSensitiveIntermediaryResistantAmoxicillin/Clavulanic Acid96 (94.1 %)5 (4.9 %)1 (1.0 %)------Ampicillin------43 (87.8 %)06 (12.2 %)Azithromycin50 (49.0 %)052 (51.0 %)49 (100.0 %)00Cefuroxime82 (80.4 %)4 (3.9 %)16 (15.7 %)49 (100.0 %)00Ceftriaxone102 (96.1 %)3 (2.9 %)1 (1.0 %)49 (100.0 %)00Erythromycin50 (49.0 %)052 (51.0 %)------Levofloxacin102 (100.0 %)0049 (100.0 %)00Moxifloxacin102 (100.0 %)0049 (100.0 %)00Penicillin96 (94.1 %)6 (5.9 %)0------ Results of a multivariable analysis are shown in Table [3](#Tab3){ref-type="table"}. Pneumococcal carriage was significantly lower in the spring compared to autumn (odds ratio (OR) = 0.46, 95 % confidence interval (CI) = 0.27, 0.78). Locals had 0.30 times the odds of pneumococcal carriage compared to non-locals (95 % CI = 0.13, 0.69). No significant predictors of Hib carriage were found; a more parsimonious model of Hib carriage, only including the covariates from the pneumococcal carriage model, also revealed no statistically significant predictors (results not shown).Table 3Odds ratios and 95 % confidence intervals for risk factors of nasopharyngeal carriage of pneumococcus or *Haemophilus influenzae* type b (Hib) among 614 children in Shanghai, 2009Pneumococcal carriage*P*-value^a^Hib carriage*P*-value^a^Spring vs autumn0.46 (0.27, 0.78)0.00440.78 (0.37, 1.64)0.5166Age of child0.47030.9533 12 months1.46 (0.69, 3.09)0.95 (0.34, 2.65) 13--14 months1.34 (0.64, 2.84)1.08 (0.39, 2.94) 15--16 months0.94 (0.42, 2.09)1.21 (0.43, 3.40) 17--18 monthsrefrefLocal vs non-local residency0.30 (0.13, 0.69)0.00480.45 (0.09, 2.32)0.3388Urbanicity0.17730.1067 Urban0.66 (0.34, 1.25)0.43 (0.19, 0.97) Suburban1.06 (0.46, 2.44)0.63 (0.21, 1.85) RuralrefrefFather's education0.40370.4163 Primary school or less0.79 (0.21, 2.90)8.53 (0.70, 103.52) Junior high1.10 (0.52, 2.34)5.61 (0.60, 52.28) High school/vocational0.66 (0.29, 1.52)5.29 (0.58, 48.43) College or morerefrefHousehold income0.5869 \<2000 yuan2.21 (0.48, 10.18) 2000--4999 yuan2.19 (0.47, 10.18) ≥5000 yuanrefSize of house0.1010 \<18 m^2^2.06 (0.67, 6.32) 18--29 m^2^2.96 (0.99, 8.85) 30--69 m^2^1.17 (0.35, 3.88) ≥70 m^2^ref≥1 dose Hib vaccine vs no doses0.55 (0.24, 1.27)0.1607^a^ *P*-value from Wald Chi-square test of Type 3 Analysis Discussion {#Sec7} ========== In a study of a group of otherwise healthy children 12--18 months of age attending immunization clinics in Shanghai, we found a substantial number had asymptomatic nasopharyngeal colonization with either pneumococcus (16.6 %) or Hib (8.0 %). Carriage of *Haemophilus influenzae* type b in Shanghai was higher compared to concurrent studies taking place in Dongguan and Beijing \[[@CR21], [@CR22]\]. Pneumococcal carriage was lower in this study than that in Dongguan \[[@CR21]\], but higher than that in Beijing \[[@CR22]\]. The rate of pneumococcal carriage we found is slightly lower than what has been found in the United States (29 %) and Australia (26 %) prior to widespread pneumococcal vaccination \[[@CR23], [@CR24]\], and it is much lower than what has been found in other developing countries before vaccination introduction; for example almost all infants are colonized with pneumococcus in Papua New Guinea and The Gambia \[[@CR25]\]. Low carriage of these pathogens among persons in Shanghai may result from high levels of antibiotic overuse \[[@CR16]\], and Hib carriage might be lower than pneumococcal carriage because children with a history of PCV administration were excluded, whereas most children in the study had received at least one dose of Hib vaccine. There may be a substantial reduction in carriage due to Hib vaccine coverage in the population; a study in The Gambia found nasopharyngeal carriage decreased from 12 to 0.5 % after the Hib vaccine was introduced \[[@CR26]\]. It has been hypothesized that crowding, having more children in the family, and exposure to smoking are reasons why carriage is higher in developing countries than developed countries \[[@CR25]\]. We found an inverse association between household size and Hib (but not pneumococcal) carriage, and carriage was higher (but not significantly so) for children with siblings compared to single children, indicating possible effects from human crowding. However, families in this study showed similar characteristics to families throughout China---they had only child, and typically family members (like grandparents) take care of children during the day instead of parents utilizing community daycare facilities. Accordingly, risk factors (like large families and daycare use) common in other places throughout the world may have a low prevalence in China and may not be as important for describing the overall distribution of nasopharyngeal carriage in the population. Approximately two-thirds of the pneumococcal isolates identified are included in the current PCV-7 formulation, and there is increased coverage, to about four-fifths, with PCV-13. In contrast, the additional serotypes in PCV-10 and PCV-15, which build upon PCV-7 and PCV-13, respectively, do not appear to be highly prevalent in Shanghai children. This suggests that PCV-13 is an appropriately comprehensive vaccine for use in Shanghai. Antibiotic resistance tests are important to ensure that the clinical treatment is effective. We encountered high resistance to erythromycin and azithromycin among pneumococcal isolates in this study, and previous studies have even higher levels of antibiotic resistance to erythromycin and penicillin \[[@CR27]\]. Antibiotic resistance in the Hib isolates was lower than in the pneumococcal isolates, but the relatively high resistance rate for Hib (12.2 %) is concerning. Hib and pneumococcus can cause a similar spectrum of invasive disease---pneumonia, meningitis, or septicemia; and doctors in China often do not identify the causative agent of disease or assess for antimicrobial susceptibility before prescribing a course of treatment. High antibiotic resistance in the population can therefore prevent a child from receiving an effective treatment. As Hib and pneumococcal vaccination coverage increases in China, sustained laboratory investigations of antibiotic resistance will be necessary to ensure that the standard of care for invasive disease is effective. Because penicillin sensitivity was still quite high in this population, it may be the antibiotic of choice for physicians in Shanghai treating pneumococcal disease. This study has several limitations. We collected specimens from asymptomatic children, and, theoretically, the pneumococcal serotypes in healthy children may differ from those with invasive pneumococcal disease. However, previous studies in China have found some concordance between the serotype distribution in invasive and noninvasive cases \[[@CR14]\]. We also did not test an individual child for multiple carriage of pneumococcal serotypes, although previous studies have found a large proportion of children may be colonized in this manner \[[@CR28]\]. Additionally, our small sample size may limit our ability to identify important risk factors for pneumococcal and Hib colonization. Lastly, it is difficult to compare the distribution of risk factors for Hib carriage to those for pneumococcal carriage because we did not exclude participants based on Hib vaccination, and Hib vaccine coverage may be high enough in Shanghai to effectively lower Hib carriage \[[@CR26]\]. Conclusions {#Sec8} =========== We found a substantial proportion of children with asymptomatic pneumococcal and Hib colonization. In recent years, there have been for-fee vaccines available in China to protect against pneumococcal and Hib disease. Because carriage of these bacteria was higher in more disadvantaged children, indicating a higher burden of disease in these children, these vaccines may not be reaching the children who have the highest burden of disease. Moreover, the presence of high antibiotic susceptibility towards pneumococcus, and to a lesser extent towards Hib, underscores the need for preventive protection against these diseases. Public funding of pneumococcal and Hib vaccines would be one mechanism to increase uptake of these vaccines. Ethics approval and consent to participate {#Sec9} ========================================== The Shanghai CDC Ethics Committee approved this study. Informed consent was obtained from each child's parents prior to participation. Consent for publication {#Sec10} ======================= Not applicable. Availability of data and materials {#Sec11} ================================== Please contact the corresponding author for details about the data. Data access is subject to approval from the Shanghai CDC. CDC : Centers for Disease Control and Prevention EPI : Expanded Program on Immunization Hib : *Haemophilus influenzae* type b MICs : minimum inhibitory concentrations PCV : pneumococcal conjugate vaccine Pneumococcus : *Streptococcus pneumonia* WHO : the World Health Organization **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** JH contributed to the conception and design of study, and acquisition of data. XS contributed to analysis of data, interpretation of results, and critically reviewed and revised the manuscript. ZH acquired data, analyzed and cleaned the data, interpreted results, and drafted the manuscript. AW analyzed the data, contributed to interpretation of results, and drafted the manuscript. BC contributed to interpretation of results, and critical review of the manuscript. JY participated in collection and management of data. ST was responsible for experiments and data analysis. YL participated in experiments, and data analysis. MB contributed to interpretation of results ad critical review of the manuscript. ZY contributed to conception and design of study, and critical review of the manuscript. All authors read and approved the final manuscript. We appreciate the work of the research staff who contributed time and effort to this project. Funding {#FPar1} ======= Funding for this project was provided by Wyeth Pharmaceutical Co., Ltd. (acquired by Pfizer in 2009).

Introduction ============ Our group has previously demonstrated that antibodies to a 140 kDa protein detected by protein immunoprecipitation (known as anti-p140 or P140) in JDM were strongly associated with the development of calcinosis, a major cause of morbidity \[[@keu259-B1]\]. The presence of autoantibodies to a 142 kDa antigen (anti-MJ) in JDM has also been shown by others to be associated with a severe disease course, worse functional status and persistent disease activity \[[@keu259-B2]\]. It is now clear that the target of these autoantibodies is a 140 kDa protein, NXP2 (molecular weight 140 kDa). Anti-NXP2 autoantibodies are common in JDM and identifiable in 13--23% of patients \[[@keu259-B1], [@keu259-B2]\]. JDM is a heterogeneous disease and autoantibodies are potentially useful biomarkers to divide patients into homogeneous subgroups and inform on prognosis. Anti-NXP2 autoantibodies are of particular interest, given their frequency in JDM and their association with important disease features such as calcinosis, a major cause of morbidity. In addition to autoantibody status, age at disease onset has also been shown to influence the clinical phenotype and overall prognosis in JDM, although the nature of the relationship between age at disease onset and outcome has been variably reported \[[@keu259-B3], [@keu259-B4]\]. Here we analyse the clinical associations of anti-NXP2 measured by ELISA and protein immunoprecipitation within an extended and large cohort of juvenile myositis patients stratified by age of onset. We specifically assess the relationships between autoantibody status, calcinosis and disease activity. Methods ======= Patients -------- Patient serum samples and clinical data were obtained through the UK Juvenile Dermatomyositis Cohort and Biomarker Study (JDCBS). The JDCBS is a large cohort of UK patients with idiopathic inflammatory myopathies (IIMs), the majority with JDM \[[@keu259-B5]\]. Patients are recruited consecutively on presentation to paediatric rheumatology departments across the UK. Children and young people ≤16 years of age are included based on a diagnosis of definite or probable JDM or PM by Bohan and Peter criteria \[[@keu259-B6]\], as well as JDM or PM with overlap CTD features. Multicentre research ethics approval was obtained for the JDCBS and parental consent for children and consent or age-appropriate assent was obtained for all patients in accordance with the Declaration of Helsinki. The JDCBS Steering Committee approved this project and biological samples and clinical data were provided in accordance with JDCBS ethics and consent. Remission in JDM was defined as a full-strength Childhood Myositis Assessment Score (CMAS) \>48 \[[@keu259-B7]\], the absence of skin disease (no rash, no Gottron's, no oedema and no ulceration) and a Physician's Global Assessment Score (PGAS) \<1. While this definition of remission has not been validated, all are standard outcome measures in JDM. Where possible, we also utilized the recently proposed PRINTO criteria for disease inactivity in JDM, defined as at least three of the following criteria: creatinine kinase ≤150, CMAS ≥48, manual muscle test (MMT) ≥78 and PGAS ≤0.2 \[[@keu259-B8]\]. Autoantibody detection ---------------------- Immunoprecipitation of radio-labelled K562 cells was performed on all samples to determine the presence of autoantibodies. Anti-NXP2 ELISA was subsequently performed on all samples. Patients were classified as anti-NXP2 positive if a 140 kDa band was seen on immunoprecipitation and anti-NXP2 ELISA was subsequently positive. ### Immunoprecipitation Ten microlitres of sera were mixed with 2 mg of protein A--Sepharose beads (Sigma, St Louis, MO, USA) in immunoprecipitation (IPP) buffer (10 mM Tris--HCl, pH 8.0, 500 mM NaCl, 0.1% v/v Igepal) at room temperature for 30 min. Beads were washed in IPP buffer prior to the addition of 120 µl of \[^35^S\]methionine-labelled K562 cell extract in IPP buffer. Samples were mixed at 4°C for 2 h. Beads were washed in IPP buffer and Tris-buffered saline (TBS) (10 mM Tris--HCl, pH 7.4, 150 mM NaCl) before being resuspended in 50 µl of SDS sample buffer (Sigma). After heating, proteins were fractionated by 9% SDS-PAGE gels, enhanced, fixed and dried. Labelled proteins were analysed by autoradiography. ### ELISA ELISA was conducted as previously described \[[@keu259-B9]\], with some modifications. Ninety-six-well polystyrene plates were coated with rNXP2 (Origene Technologies, Rockville, MD, USA) at 4°C for 16 h. Samples were diluted to 1:200. Secondary antibodies were conjugated goat anti-human IgG/M (Sigma). Tetramethylbenzidine liquid substrate (Sigma) was then added. All samples were tested in duplicate and optical density was read using an automatic plate reader. The negative cut-off was defined as \>3 [s.d.]{.smallcaps} above the mean of 42 normal healthy (adult) serum controls. Data analysis ------------- Multivariate regression analyses were performed using generalized additive models within R \[[@keu259-B10]\], which allows for the possibility of non-linear relationships between covariates and response variables \[[@keu259-B11]\]. Within this framework, models were used to assess the significance of age at disease onset and anti-NXP2 in relation to calcinosis as well as disease outcome 2 and \>4 years post-diagnosis. Disease duration was adjusted for. Potential differences between two groups were assessed using chi-squared analysis with Yates's continuity correction. The Mann--Whitney *U* test was used to compare non-normally distributed continuous data. Results ======= Demographic data and key findings are summarized in [Table 1](#keu259-T1){ref-type="table"}. T[able]{.smallcaps} 1Demographic data and key findings of 285 patients analysedAll patientsAnti-NXP2 positiveFemale, *n*/*N* (%)206/285 (72)41/56 (73)IIM subtype^a^, *n*/*N* (%)    DM242/285 (85)55/56 (98)    PM1/278 (0.4)0    Overlap33/285 (12)1/56 (2)^b^Age at onset, average (IQR), years6.2 (4--10)6.5 (4--10)Time to diagnosis, average (IQR), months4 (2--10)4 (2--10)Length of follow-up, average (IQR), years9 (5--12)9 (5--12)Proportion with calcinosis, *n*/*N* (%)94/283 (29)24/56 (43), *P* = **0.025**Lowest ever CMAS score, average (IQR)40 (27--48)29.6 (18--43), *P* \< **0.001**[^1] Autoantibody detection ---------------------- Of the 365 patients enrolled in the JDCBS at the time of analysis, 285 (78%) had both data and serum available. This group was representative of the cohort as a whole, showing no significant differences in gender (72% *vs* 68% female), age at disease onset (median 6.2 *vs* 6.7 years) or JDM clinical type (12% *vs* 14% JDM overlap) compared with those patients not analysed. Anti-NXP2 autoantibodies were identified in 56/285 (20%) patients. Two further patients were positive by anti-NXP2 ELISA but did not have a 140 kDa band on immunoprecipitation. Further analysis was carried out on only those patients definitively positive using both techniques. Calcinosis ---------- Overall, 33% of patients developed calcinosis during the follow-up period. Age at disease onset influenced the likelihood of calcinosis allowing for the duration of disease and there was a significant decrease in calcinosis with increasing age at disease onset \[odds ratio (OR) 0.90, 95% CI 0.83, 0.97, *P* = 0.005\]. Furthermore, this relationship with age proved to be near linear, a striking result given the flexibility of the model meant there were no prior constraints on its shape ([Fig. 1](#keu259-F1){ref-type="fig"}). F[ig]{.smallcaps}. 1The effect of anti-NXP2 autoantibodies on the risk of calcinosis by age at disease onset (with 95% CI)A near-linear relationship is seen between younger age at disease onset and increased risk of calcinosis. There was a significant association with anti-NXP2 autoantibodies and calcinosis: 43% anti-NXP2-positive patients developed calcinosis *vs* 30% anti-NXP2-negative patients (OR 2.10, 95% CI 1.10, 4.01, *P* = 0.025 unadjusted for age, *P* = 0.039 adjusted). The presence of anti-NXP2 autoantibodies increased the risk of calcinosis after allowance for the possible effects of other variables, including age. As some patients were recruited an appreciable time after their initial diagnosis, information on when calcinosis was first noted was not always available. However, seven children with anti-NXP2 autoantibodies developed calcinosis after their first study visit and all of these had anti-NXP2 autoantibodies present in blood samples predating calcinosis development (taken 2 months--4 years after disease onset). The average time to develop calcinosis in this group was 5.3 years (range 1.2--13 years). Thirty-two children without anti-NXP2 autoantibodies developed calcinosis after their first study visit and the average time to calcinosis was similar at 4.1 years after disease onset (range 0--10.9 years). Muscle weakness --------------- Children with anti-NXP2 autoantibodies had a lower median lowest ever CMAS, corresponding to a greater degree of muscle weakness, compared with those without anti-NXP2 autoantibodies {29.6 \[interquartile range (IQR) 18--43\] *vs* 42 (IQR 31--48)} and this was statistically significant (*P* \< 0.001). Furthermore, 53% of children with anti-NXP2 had a lowest ever CMAS ≤30, corresponding to a moderate or severe degree of weakness \[[@keu259-B7]\], compared with just 25% of those children without anti-NXP2 (*P* \< 0.001). Disease outcome in JDM ---------------------- Disease outcome was assessed 2 years post-diagnosis and at the last clinic visit, providing this was at least 4 years post-diagnosis (mean 7.9 years). Data at 2 years post-diagnosis were not yet available in children who had been diagnosed with JDM \<2 years previously, had been recruited into the study \>2 years post-diagnosis or had not been reviewed 20--28 months post-diagnosis. Information was available for 152 children 2 years post-diagnosis (25 with anti-NXP2 autoantibodies) and 136 children \>4 years post-diagnosis (24 with anti-NXP2 autoantibodies). Using our own predefined definition of remission, children with anti-NXP2 antibodies were less likely to be in remission 2 years post-diagnosis compared with those without anti- NXP2 (8% *vs* 32%; OR 0.23, 95% CI 0.05, 1.05, *P* = 0.04). Using the recently proposed PRINTO definition of disease inactivity in JDM \[[@keu259-B8]\], fewer children with anti-NXP2 had inactive disease 2 years post-diagnosis (37.5% *vs* 48%), but this did not reach significance. Outcomes \>4 years post-diagnosis were similar in the two groups, irrespective of the definition used (our own criteria, 47% *vs* 50%; PRINTO criteria, 72% *vs* 73%). Discussion ========== This study shows that anti-NXP2 autoantibodies are the most common autoantibody identifiable in our JDM cohort and can be identified in one in five UK children with JDM. They are a potentially important clinical biomarker and can both facilitate diagnosis and provide important prognostic information, particularly with regard to calcinosis development. Younger children with JDM are believed by some specialist clinicians to have more aggressive disease, but recent studies have challenged this view \[[@keu259-B4]\]. We have clearly demonstrated that children presenting at a younger age have a higher risk of developing calcinosis, a major cause of morbidity. The presence of anti-NXP2 autoantibodies substantially increases the risk of calcinosis across all ages. Anti-NXP2 autoantibodies were associated with a greater degree of muscle weakness as determined by CMAS, suggesting more severe disease. This difference is of clinical significance and children with anti-NXP2 autoantibodies were more likely to have a moderate or severe degree of clinical weakness \[[@keu259-B7]\]. In JDM, the development of calcinosis has previously been associated with a delayed diagnosis, a chronic disease course and inadequately treated disease \[[@keu259-B12]\]. In our cohort, no difference in time from symptom onset to diagnosis was identified between the anti-NXP2-positive and anti-NXP2-negative groups. We investigated the prevalence of disease remission in JDM at two time points using two different definitions of disease inactivity. While a significant association between anti-NXP2 autoantibody status and more persistent disease activity 2 years post-diagnosis was suggested using our own definition of remission, the CIs were wide. We were unable to clearly demonstrate a significant relationship using the recently proposed PRINTO criteria, although the same trend existed. Persistent disease activity remains a possible mechanism driving calcinosis development in children with anti-NXP2 autoantibodies, but larger patient numbers and validated remission criteria are required to firmly establish the nature of any association. In summary, anti-NXP2 autoantibodies are associated with features of severe disease in JDM. The development of calcinosis in JDM is influenced by both age at disease onset and anti-NXP2 autoantibody status and both factors must be considered when attempting to predict the risk of this clinically important complication. Rheumatology key messagesEarlier age at disease onset is associated with greater risk of calcinosis in JDM.Anti-NXP2 substantially increases the risk of calcinosis across all age groups in JDM.Anti-NXP2 autoantibodies are associated with features of severe disease in JDM. Supplementary Material ====================== ###### Supplementary Data The UK Juvenile Dermatomyositis Cohort and Biomarker Study (JDCBS) has been supported by generous grants from the Wellcome Trust UK (085860), Action Medical Research UK (SP4252), the Myositis Support Group UK, Arthritis Research UK (14518) and the Henry Smith Charity. The JDCBS has been adopted onto the Comprehensive Research Network through the Medicines for Children Research Network ([www.mcrn.org.uk](http://www.mcrn.org.uk)). S.T. and the serotyping of much of the JDCBS cohort were funded by the British Medical Association (Doris Hillier grant 2012). A list of the members of the Juvenile Dermatomyositis Research Group is available as [supplementary material](http://rheumatology.oxfordjournals.org/lookup/suppl/doi:10.1093/rheumatology/keu259/-/DC1) at *Rheumatology* Online. *Disclosure statement*: The authors have declared no conflicts of interest. Supplementary data ================== [Supplementary data](http://rheumatology.oxfordjournals.org/lookup/suppl/doi:10.1093/rheumatology/keu259/-/DC1) are available at *Rheumatology* Online. [^1]: IQR: interquartile range; IIM: idiopathic inflammatory myopathy; CMAS: Childhood Myositis Assessment Score. ^a^In addition to the diagnostic categories listed, nine juvenile patients were labelled as other IIM or focal myositis. ^b^This patient had JDM--scleroderma overlap. Bold text indicates significant *P*-values. Demographic data were similar between anti-NXP2-positive and -negative groups.

1. Introduction {#sec1} =============== Secondary amyloidosis can complicate chronic inflammatory autoimmune diseases such as rheumatoid arthritis (RA) and ankylosing spondylitis. However, the clinical findings of primary amyloidosis may mimic those of primary rheumatologic disorders. We present the case of a 53-year-old woman with a history of hypertension, bilateral carpal tunnel syndrome, and irritable bowel syndrome (IBS) who presented for evaluation of fingernail changes for the previous three years. The nails had vertical ridges and had cracked to the nail base. Occasionally, she also had mild pallor and bluish discoloration of her fingers on exposure to cold, suggestive of Raynaud\'s phenomenon. She had had diarrhea alternating with constipation and mild abdominal cramping for 3 years. Colonoscopy was done a year earlier and showed only diverticulosis of sigmoid colon; she was given a diagnosis of IBS by a gastroenterologist. Other symptoms included diffuse alopecia, dry eyes, fatigue, and numbness in both forearms and fingers bilaterally. She had had unilateral carpal tunnel release in the prior year; tissue was not sent for pathologic examination. Family history was significant for her mother having myelofibrosis and acute leukemia, but there was no family history of autoimmune disease. Her physical examination was normal except for vertical ridging of dystrophic nails of the first, second, and third digits of both hands, diffuse alopecia with no scalp rash, scaling or scarring, and numbness in her first, second, and third digits of both hands with positive Phalen and Tinel test. She had no skin tightening, rashes, or telangiectasias. Complete blood count and complete metabolic panel including serum calcium, urinalysis, erythrocyte sedimentation rate, and C-reactive protein were unremarkable except for total protein of 5.4 g/dL (normal: 6.1--7.9 g/dL). She had a normal serum iron, vitamin B12, folic acid, TSH, free T4, and chest X-ray. Serologic testing revealed an ANA of 1 : 1280 in a nucleolar pattern. Antibodies to SSA, SSB, Smith, RNP, centromere, Scl-70, and dsDNA were negative; C3 and C4 were normal. She was initially diagnosed with Undifferentiated Connective Tissue Disease (UCTD) by her rheumatologist. But, given her neuropathy, serum protein electrophoresis (SPEP) was sent and it showed an increase in *α*2-globulin fraction (16.7%; normal: 8.9--14.5%) and decrease in *γ*-globulin level and fraction (0.4 and 7.0%, resp.; normal: 0.6--1.9 g/dL; 9.8--24.4%) with an albumin-to-globulin ratio of 1.4. Serum immunofixation did not show any evidence of monoclonal gammopathy. Serum-free light chain assay revealed elevated free lambda light chains of 34.22 mg/dL (normal: 0.57--2.63 mg/dL) with low free kappa light chains of \<0.29 mg/dL (normal: 0.33--1.94 mg/dL) and global decrement of IgG, IgA, and IgM. Urine protein electrophoresis and immunofixation were unremarkable. She was referred to the Hematology-Oncology consultation team for further evaluation of the light chain monoclonal gammopathy. A bone marrow biopsy (Figures [1](#fig1){ref-type="fig"} and [2](#fig2){ref-type="fig"}) showed normocellular bone marrow with 15% plasma cells and lambda light chain restriction; Congo red staining was negative. A skeletal survey, LDH, and *β*2-microglobulin were negative. Given her constellation of symptoms, amyloidosis remained a diagnostic concern. An abdominal fat pad biopsy was performed, revealing positive Congo red staining ([Figure 3](#fig3){ref-type="fig"}). Electron microscopic examination revealed clusters of filamentous material with average diameter of 8.7 nm, lined between lipid droplets, compatible with amyloid; no amyloid typing was performed. The patient was diagnosed with free lambda light chain amyloidosis. Echocardiogram, pro-BNP, and Troponin T were obtained for staging; these were within normal limits. The patient was treated with 4 cycles of bortezomib and dexamethasone. When near-complete remission was achieved with near-normalization of serum-free lambda light chains, decreasing to 3.07 mg/dL over 3 months, she subsequently underwent autologous peripheral blood stem cell transplantation after melphalan conditioning. Her posttransplantation serum-free lambda light chains decreased to 2.69 mg/dL at three months and then normalized to 2.57 mg/dL at six-month follow-up. The patient returned to her normal level of excellent function, with resolution of her initial symptoms, and returned to work one month after the transplant procedure. She had no clinical features of autoimmune disease. No posttransplant serologies were obtained. 2. Discussion {#sec2} ============= Deposits of amyloid may be distributed in many organs of the body (systemic amyloidosis) or may be restricted to a single organ (localized amyloidosis) \[[@B1], [@B2]\]. As demonstrated here, primary (AL) amyloidosis may mimic primary rheumatologic diseases such as SLE and Sjögren\'s syndrome. When this occurs, no typical pathologic findings of rheumatologic diseases are found in the biopsies of joints, salivary glands, or blood vessels except amyloid accumulation in these tissues \[[@B2], [@B6]\]. Others also have described amyloidosis in rheumatic diseases such as Sjögren\'s syndrome, scleroderma, and primary biliary cirrhosis as a secondary phenomenon \[[@B6], [@B4]\]. In our patient, dry eyes can likely be explained by the amyloid deposition in lacrimal glands, masquerading as Sjögren\'s syndrome \[[@B2]\]. AL amyloidosis associated with multiple myeloma may result in an arthropathy resembling RA \[[@B5]\]. In this setting, treating the underlying hematologic malignancy is the most important intervention as patients tend to respond better to multiple myeloma treatment than to treatment of the amyloid arthropathy. Sensitivity of selected physical and laboratory findings for detecting amyloidosis in cases of multiple myeloma (MM) associated arthropathy has been illustrated in literature ([Table 1](#tab1){ref-type="table"}) \[[@B5]\]. While our patient did not have arthropathy, she had bilateral carpal tunnel syndrome and hypogammaglobulinemia, which raised the suspicion of AL amyloidosis. The carpal tunnel symptoms of the prior year are consistent with the typical 2-year delay between the onset of symptoms and the recognition of amyloidosis \[[@B3]\]. Amyloid-associated rheumatic disease manifestations in primary amyloidosis have been associated with an anticentromere antibody pattern of ANA \[[@B3]\]. However, to the best of our knowledge, a nucleolar pattern ANA has not been previously described in this setting. Given the association of a nucleolar pattern ANA and scleroderma, our patient warrants follow-up for the progression to the development of scleroderma, though this seems unlikely in the setting of a successful stem cell transplantation and resolution of symptoms. Alopecia or nail dystrophy can rarely present as a sign of occult amyloidosis \[[@B7], [@B8]\]. Unusual manifestations of primary cutaneous amyloidosis in association with Raynaud\'s phenomenon and livedo reticularis have been reported \[[@B1]\], but, in this case, Raynaud\'s phenomenon preceded the cutaneous amyloidosis by ten years. Histopathology described in scalp biopsies of amyloid patients with alopecia has demonstrated that amyloid deposits were seen to be compressing pilosebaceous units, accompanied by atrophy of these structures and loss of hair from the shafts \[[@B9]\]. Nail biopsy may show amyloid deposits in papillary dermis of the matrix, nail bed, and perivascular areas, which perhaps cause vascular impairment with subsequent nail brittleness, crumbling, onycholysis, subungual thickening, striation, or anonychia \[[@B10]\]. Initially, the bone marrow biopsy in our patient was performed for the purpose of diagnosing plasma cell dyscrasia. The biopsy revealed lambda light chain restriction on*in situ* hybridization but negative Congo red staining. The diagnostic yield of bone marrow biopsy in primary amyloidosis is estimated at 63% \[[@B11], [@B12]\]. We interpreted that the false-negative results could be due to variable and sometimes pale staining of amyloid fibrils, inadequate material, or improper use of polarizing instruments. In addition, the correct interpretation of Congo red staining requires experience and is somewhat subjective \[[@B12]\]. Thioflavin staining is another option to confirm the validity of negative Congo red staining because it can detect smaller deposits, although it is less specific \[[@B12]\]. Abdominal fat pad aspiration has replaced most of the previously used tissue biopsies because it is easy to perform and safe and involves minimal patient discomfort and morbidity. Studies have shown that, in about 11 percent of cases, fat pad biopsy can be positive in the setting of negative bone marrow staining \[[@B3]\]. The sensitivity and specificity for abdominal fat pad aspiration are 55--85% and 75--100%, respectively \[[@B11], [@B12]\]. Labial salivary gland (LSG) biopsy is a good diagnostic modality to consider for diagnosing amyloidosis mimicking Sjögren\'s syndrome because it shares the advantages of abdominal fat pad aspiration with similar diagnostic yield and offers the ability to rule out Sjögren\'s syndrome simultaneously. The diagnostic yield of LSG is roughly 58--86%, with specificity of 100% \[[@B13], [@B14]\]. Biopsy of another involved organ such as kidney, liver, colon, or heart can be utilized as well due to the high sensitivity (87--98%), if Congo red staining on the bone marrow and abdominal fat pad biopsy is negative, but there is high clinical suspicion of amyloidosis \[[@B11], [@B12]\]. To summarize, our patient presented with dystrophic nail changes, carpal tunnel syndrome, Raynaud\'s phenomenon, and high titer positive nucleolar pattern ANA and was diagnosed with lambda light chain systemic amyloidosis by abdominal fat pad biopsy. Fortunately, in our case, the diagnosis was made before the development of cardiomyopathy, a complication significantly reducing life expectation \[[@B15]\]. Based on the Revised Mayo Clinic Staging for amyloidosis \[[@B15]\], our patient was staged as Stage II, which has median survival of 97 months in stem cell transplant eligible patients versus 19 months in ineligible patients. Thus we treat amyloidosis early on, since prognosis in untreated AL amyloidosis is dismal. We suggest further monitoring for her positive nucleolar pattern ANA due to the possible progression to scleroderma or other rheumatic diseases, though this would be a remote possibility following her bone marrow transplantation. Furthermore, due to the relatively low diagnostic yield of bone marrow biopsy, we employed an alternate diagnostic modality with little morbidity (abdominal fat pad aspiration) to establish this patient\'s diagnosis. Our case also highlights the many similarities in the clinical features between systemic light chain amyloidosis and primary rheumatological disorders. It also underlines the importance of the prompt recognition of "red flags" of amyloidosis which are crucial to triggering appropriate diagnostic procedures, since early diagnosis prior to the development of advanced cardiomyopathy is a key to improving outcomes in this disease. Competing Interests =================== All authors declare that there are no competing interests. {#fig1} {#fig2} {#fig3} ###### Sensitivity of selected physical and laboratory findings for detecting amyloidosis in the setting of multiple myeloma \[[@B5]\]. Finding Sensitivity (%) ------------------------------------------------------------------------ ----------------- \(1\) Macroglossia, carpal tunnel syndrome, or shoulder pads (any one) 47 \(2\) Hypercalcemia 26 \(3\) Hypogammaglobulinemia 22 \(4\) Renal failure 23 \(5\) Juxta-articular bone lesions 25 \(6\) Bence Jones proteinemia 56 (1), (2), (3), or (4) 67 (1), (2), (3), (4), or (5) 74 (1), (2), (3), (4), or (6) 86 (1), (2), (3), (4), (5), or (6) 91 [^1]: Academic Editor: Mamede de Carvalho

All relevant data are within the paper. Introduction {#sec005} ============ Preparedness is a state of readiness to respond to crises \[[@pone.0156536.ref001]\], with one aim being to reduce negative outcomes. When preventive work is effective, it is invisible. Preparedness for disaster often includes a written emergency plan. The plan represents a structured definition of working roles and how to act in disaster situations. Such plans aim to clarify the working roles. Working during a disaster is a dynamic and ongoing process, and a written emergency plan, even though helpful, often does not cover the unfamiliar and unexpected demands that may occur in major disasters. During a disaster, it is important to obtain a balance between a standardization of behavior using plans as well as an ability to improvise and exercise professional judgment \[[@pone.0156536.ref002]\]. This is particularly critical when rescue workers are in situations of compromised security and uncertainty. Weisæth and Kjeserud \[[@pone.0156536.ref003]\] state that emergency planning and management of disasters cannot only be learned, but must be experienced in order to attain the knowledge one needs. They infer that very few people without training will be able to spontaneously manage a disaster or serious emergency, and conclude that readiness without training is of sparse value. Preparedness in rescue workers are obtained through an emergency plan, but even more through every-day experience with physically traumatized people either at the site of the accident or in hospital, through disaster drills or during simulation training with a team \[[@pone.0156536.ref004], [@pone.0156536.ref005]\]. Successful training will most likely have an effect on a better understanding of the tasks and roles during the operation, which may lead to less psychological distress during and after the operations, and even help mastering of future disaster operations. However, studies on the effects of training on rescue operations after terror attacks are limited. Some papers present an evaluation after a disaster drill, followed by suggestions for future events \[[@pone.0156536.ref006]\]. In a review studying the effects of disaster training, a lack of evidence between training interventions and improvement of knowledge and skills during a disaster response was found \[[@pone.0156536.ref007]\]. Few studies explore emergency preparedness per se and most of the work remains unreported in the literature. Some information and preliminary reports have been presented at conferences \[[@pone.0156536.ref008]\]. The predictors of post-traumatic stress symptoms in rescue workers after a disaster have been studied, but the results are conflicting regarding the relationship with prior experience, preparation and training as resilient factors \[[@pone.0156536.ref009]\]. On July 22, 2011, Norway was struck by two terror attacks. A car bomb was detonated in the Oslo Government district. Eight people were killed, many were wounded and governmental buildings were severely damaged. Approximately two hours after the first attack, the terrorist initiated shooting on a political youth camp on Utøya Island, where 69 were killed and many physically injured \[[@pone.0156536.ref010]--[@pone.0156536.ref012]\]. More than 500 people participated in the youth camp and were psychologically traumatized by the shooting \[[@pone.0156536.ref013]\]. The terror attacks of July 22, presented an unusual requirement for those who took part in the rescue operations and subsequent treatment of the wounded. Because so many were killed, wounded or emotionally traumatized, the pressure on rescue workers was demanding and extraordinary compared to their normal daily routines. In addition to facing severe human and material losses, the rescue workers also had to be prepared for the possibility of more terror attacks. An ongoing disaster presents the possibility for real-life training. In this study we define rescue workers as personnel with professional education/training, affiliated either in the police force, fire department or in the emergency departments of a hospital. They are familiar with an emergency preparedness plan. To the best of our knowledge, there is no systematic examination of preparedness of rescue workers and their response capabilities during large-scale disaster operations. On the basis of what is outlined above, we had two aims for the present study: To examine 1. The level of preparedness, exposure and role clarity, and; 2. The relationship between demographic, preparedness and exposure and Role clarity during the rescue operations and;Achieved mastering for future disaster operations. Material and Methods {#sec006} ==================== Study design and setting {#sec007} ------------------------ This cross-sectional study investigated professional personnel involved in rescue operations after the terror attacks in Norway. The present study is part of a larger project examining the challenges that the professional personnel met during these operations \[[@pone.0156536.ref014]\]. A questionnaire, which included background variables, contributions and exposure during the rescue operations, and how the events affected them, was sent to the rescue workers. The questionnaire was distributed approximately eight to 11 months after the terror attacks on July 22, 2011 (mean = 10 months) to personnel who worked with the terror victims or their relatives from July 22 to August 5, 2011, with a reminder sent one month after the first request. Those surveyed were kept anonymous. The questionnaire was distributed with an information letter while the return of the questionnaire was assumed to imply informed consent. Subjects {#sec008} -------- One could assume that personnel affiliated to an organization with a written plan were prepared for disaster, therefore three organizations with a consistent emergency preparedness plan were selected from the study population: healthcare professionals working in hospitals (n = 859), police officers (n = 252) and firefighters (n = 102). Police officers from three counties, firefighters from 10 independent units in four counties and healthcare providers from three hospitals were included. Among health care workers, 16% of the emergency medical service personnel worked close to the scene of terror. The corresponding numbers were 37% and 97% for police officers and the firefighters respectively. The other worked at their regular working place. The leader from each unit was responsible for the distribution and collection of the questionnaires. The respondent placed the completed questionnaire in an envelope. The envelope could be sealed, and were dropped in a sealed box. In the current paper, general and psychosocial healthcare personnel working within a hospital organization were merged into one group (n = 859). Personnel working in municipal emergency services (n = 72) and a center for next-of-kin outside a hospital (n = 95) were excluded, as they were unlikely to be working in an organization with a consistent emergency plan. Personnel working in a rehabilitation ward in one of the hospitals (n = 28) were excluded because their work started after August 5, and eight respondents with incomplete questionnaires were also excluded. The response rates were 54% among healthcare providers, 51% among police officers and 82% among firefighters. Assessments {#sec009} ----------- Most of the items included in the questionnaire were developed by The Norwegian Centre for Violence and Stress Studies, and used in a cross-sectional study of Norwegian personnel mobilized during the 2004 Indian Ocean tsunami. A replica of some assessments from that study made it possible to compare data from two samples of Norwegian rescue workers \[[@pone.0156536.ref009]\]. ### Sociodemographic characteristics {#sec010} Age, sex, duration of work at their current organization and working place during the rescue operation measured sociodemographic characteristics. These were classified as shown in [Table 1](#pone.0156536.t001){ref-type="table"}. 10.1371/journal.pone.0156536.t001 ###### Background characteristics. {#pone.0156536.t001g} Healthcare providers Police officers Fire-fighters *P value* ------------------------------------------------------------------------------------------------------ ---------------------- ----------------- --------------- ----------------------------------------------- **Gender**, male, n (%) 287 (33.6) 169 (67.9) 100 (99.0) \<.001[\*\*](#t001fn003){ref-type="table-fn"} \< 30 years 144 (16.8) 29 (11.6) 11 (10.8) 30--49 years 524 (61.3) 172 (69.1) 68 (66.7) \>50 years 187 (21.9) 48 (19.3) 23 (22.5) \< 1 year 44 (5.2) 19 (7.6) 2 (2.0) 1--5 years 242 (28.4) 74 (29.6) 18 (17.6) .015[\*](#t001fn002){ref-type="table-fn"} \>5 years 567 (66.5) 156 (62.7) 82 (80.4) Work experience in similar tasks 541 (63.9) 163 (64.7) 70 (68.6) ns Training based on simulation 585 (68.8) 176 (69.8) 74 (72.5) ns Disaster drill 588 (69.2) 173 (68.9) 68 (66.7) ns Previous event with \>5 fatalities 223 (26.1) 66 (26.3) 31 (30.4) ns Sites of terror 138 (16.2) 92 (36.8) 99 (97.1) Other (in hospital/out-patient emergency service. Centre for next of kin, patrolling, investigation) 713 (83.8) 157 (63.1) 3 (2.9) \<.001[\*\*](#t001fn003){ref-type="table-fn"} Note: \* p = \<.05, \*\* p = \<.001. Chi-square ### Preparedness/training {#sec011} Four items measured preparedness/training (see [Table 1](#pone.0156536.t001){ref-type="table"}), and most items had three response alternatives: *1) yes---once*, *2) yes---several times*, *and 3) no*, *never*. The response alternatives were dichotomized into no and yes. The item, event with \> 5 deceased, had two response alternatives, no and yes. ### Perceived threat and inner psychological responses {#sec012} Four self-designed items were used to assess possible perceived threats after the terror attacks. We questioned whether subjects experienced a (1) fear of explosion/shooting, (2) fear of being injured, or (3) other risks/uncertainty. The response alternatives for all items were: 0 = *no*, *not experienced*; 1 = *yes*, *but not stressful*; 2 = *yes*, *moderately stressful*; and 3 = *yes*, *very stressful* ([Table 2](#pone.0156536.t002){ref-type="table"}). The items were summarized and named as perceived threats. Two questions assessed perceived psychological responses: (1) Did you feel overwhelmed, and (2) Did you feel a lack of control? These items were scored on a five-point Likert scale: 1 = *not at all* to 5 = *to a very high degree*. 10.1371/journal.pone.0156536.t002 ###### Exposure and psychological responses. {#pone.0156536.t002g} Healthcare providers Police officers Fire-fighters *p value* -------------------------------------------- ---------------------- ----------------- ---------------- ---------------------------------------------------- **Witness experience** sum all items: yes 694 (80.3) 169 (67.1) 94 (92.2) \< **.001**[\*\*](#t002fn002){ref-type="table-fn"} Cronbach's Alpha = 0.70 **Perceived threats** (scale 0--3) .5 (.5-.6) .8 (.7-.9) .9 (.7--1.1) **\< .001**[\*\*](#t002fn002){ref-type="table-fn"} Cronbach's Alpha = 0.87 **Feeling overwhelmed** (scale 1--5) 2.4 (2.4--2.5) 2.6 (2.5--2.8) 2.5 (2.3--2.7) ns **Feeling a lack of control** (scale 1--5) 2.1 (2.0--2.1) 2.5 (2.3--2.6) 2.1 (1.1--1.4) \< **.001**[\*\*](#t002fn002){ref-type="table-fn"} **Dissociation** 1.7 (1.7--1.8) 1.7 (1.6--1.8) 1.7 (1.6--1.8) ns Cronbach's Alpha = 0.78 **Arousal** 2.9 (2.9--3.0) 3.2 (3.0--3.3) 2.6 (2.4--2.8) \< **.001**[\*\*](#t002fn002){ref-type="table-fn"} Cronbach's Alpha = 0.76 (PCL) (scale 17--85) Median (range) 19.0 (17--64) 19.0 (17--69) 19.0 (17--64) PCL 35--50 18 (2.1) 7 (2.8) 1 (1.0) PCL \> 50 3 (.4) 1 (.4) 2 (2.0) Cronbach's Alpha = 0.91 *Note*: Mean (95% CI) or n (%). Scale 1--5: 1 = not at all, 5 = to a very high extent for all items except threat and PCL, \*\* p = \<.001. Chi-square or ANOVA ### Exposure {#sec013} Seven items measured exposure: witnessing disaster victims (1) searching for next- of-kin, (2) in despair at the campsite, (3) with major physical injuries, (4) seeing dead bodies, (5) having physical contact with dead bodies, (6) seeing torn body parts, and (7) strong smells or other sensory perceptions. All items were dichotomized into no and yes, with the sum of all items reflecting the number of positively scored items (0--7). ### Dissociation and arousal {#sec014} Eight items assessed dissociation and arousal during the rescue operations: a feeling of (1) "numbness", (2) not being aware of the surroundings, (3) what you experienced was not real, (4) not being yourself, (5) not remembering what happened or only parts of it, (6) sharpened attention, (7) a reduced need for sleep and/or rest, and (8) positive activation (more energy or an intense sense of coping). A five-point Likert scale was used: 1 = *not at all* to 5 = *to a very high degree*. A factor analysis revealed two constructs: dissociation (items 1--5) and arousal (items 6--8). ### PTSD symptoms {#sec015} The PTSD Checklist (PCL-S) \[[@pone.0156536.ref015]\] was used to screen for posttraumatic stress symptoms (PTSD) and is a widely used and validated self-reported measure of PTSD \[[@pone.0156536.ref016]\]. Each item was scored on a five-point Likert scale: 1 = *not at all*, to 5 = *very often*, in which the overall scores ranged from 17 to 85. A cut off score of 31--38 has been reported to identify most cases with posttraumatic stress symptoms of clinical significance \[[@pone.0156536.ref017], [@pone.0156536.ref018]\] and a cut-off score of 50 on the PCL scale has been recommended for possible PTSD (15). That is the reason for the selection, but lower scores have also identified persons with possible PTSD (17). ### Work-related variables {#sec016} The items: "Working with adequate resources to perform the rescue work satisfactorily" and "Know your professional role/working tasks" (Role clarity) were measured on a five-point Likert scale: 1 = *not at all* to 5 = *to a very high degree* \[median split score: \< 4 *(to a low degree)* and ≥ 4 *(to a high degree)*\]. The latter item was used as an outcome in the regression analysis. A Likert scale was used on the items, "Comprehend people's reactions better" and "More prepared to master similar situations after the rescue work". The latter was dichotomized and used as an outcome in the regression analysis \[median split score: \<3 (*to a low degree)* and ≥ 3 (*to a high degree*)\]. ### Evaluation of the rescue work {#sec017} Seven items were used to evaluate the rescue work: To what degree did you find the rescue work a) successful, b) meaningful, c) I/we fell short, d) I/we were hindered, e) competition was a problem, f) I received sufficient advice and support, and g) I stretched myself too far because a high effort was expected. The items were measured on a five-point Likert scale: 1 = *not at all* to 5 = *to a very high degree*, with a factor analysis identifying two factors: successful (items a, b, f) and obstructed (c, d, e, g). Ethics {#sec018} ------ Prior to the study, the Data Protection Officer at Oslo University Hospital was contacted to clarify further approval from the regional Ethics Committee. The Data Protection Officer informed that further approval from The Regional Ethics Committee was not necessary due to anonymously collected data. The data was stored on the research server at the hospital. Data analyses {#sec019} ------------- The data was presented as means with 95% confidence intervals, which means the alpha was set at 5% (0.05), or percentages, and in general there were few missing data (1--2%). Where appropriate, the variables were dichotomized. Chi-squared and Kruskal--Wallis tests were used to compare proportions; whereas independent samples *t* tests were used to compare means. A logistic regression analysis (Forward Wald) identified the predictors of role clarity and being more prepared to master future disaster operations, and SPSS statistical software (version 21; SPSS, Chicago, IL) was used. Results {#sec020} ======= Previous training and work experience {#sec021} ------------------------------------- Approximately two-thirds of all respondents reported previous training in either work experience from similar tasks, simulations or disaster drills ([Table 1](#pone.0156536.t001){ref-type="table"}). In addition, more than one-fourth reported to have participated in a rescue event involving more than five fatalities, and there were no significant differences between the groups in any of these items. A total of 99 (97%) of the firefighters worked at the sites of terror, while approximately half of the police officers had other duties such as office work, patrolling and investigation. Most of the healthcare providers 713 (84%) worked in a hospital. Perceived threat and psychological responses {#sec022} -------------------------------------------- The level of perceived threat (fear of shooting, fear of getting hurt or other danger) was generally low ([Table 2](#pone.0156536.t002){ref-type="table"}). Healthcare providers perceived less of a threat during the rescue work compared to police officers and firefighters (*p* \< .001) and the police officers reported significantly more lack of control (*p* \< .001). More firefighters and healthcare providers witnessed victims compared to the police officers (p \< .001). Police officers perceived the highest level of arousal, and healthcare providers perceived more arousal than firefighters (*p* \< .001). Few responders in either group had post-traumatic stress symptoms at clinical sub-threshold (1.0--2.8%) or at a PTSD level (0.4--2.0%) (ns). Work-related variables {#sec023} ---------------------- [Table 3](#pone.0156536.t003){ref-type="table"} shows work-related variables. The great majority knew their professional role (role clarity) (healthcare providers *M* = 4.1 vs. police officers: *M* = 3.9 vs. firefighters: *M* = 4.2, *p* \< .001). The police officers reported less adequate resources (*M* = 2.8 vs. firefighters: *M* = 3.8 and healthcare providers *M* = 4.1; *p* \< .001), as well as more often perceiving the rescue work to be obstructed as in falling short, being hindered, experiencing competition or stretching themselves too far (*M* = 2.4 vs. firefighters: *M* = 1.9 vs. healthcare providers *M* = 1.9, *p* \< .001). All groups reported the rescue work to be successful, but the police officers significantly less so (*M* = 3.7 vs. firefighters: *M* = 4.2 vs. healthcare providers: *M* = 4.0, *p* \< .001). All groups reported being more prepared to master similar situations in the future (*M* = 3.5--3.6), and with no significant differences between groups. 10.1371/journal.pone.0156536.t003 ###### Working role, resources and evaluation of the rescue work. {#pone.0156536.t003g} Healthcare providers Police officers Fire-fighters *p value* -------------------------------------------------------------------------------------------------------------- ---------------------- ----------------- ---------------- ---------------------------------------------------- **Role clarity** 4.1 (4.0--4.1) 3.9 (3.8--4.0) 4.2 (4.0--4.3) **\< .001**[\*\*](#t003fn002){ref-type="table-fn"} **Working with adequate resources** 4.1 (4.0--4.2) 2.8 (2.6--2.9) 3.8 (3.6--3.9) **\< .001**[\*\*](#t003fn002){ref-type="table-fn"} **The rescue work was obstructed** (Scale 1--5): (fell short, hindered, competition, stretch myself too far) 1.9 (1.8--1.9) 2.4 (2.3--2.5) 1.9 (1.8--2.0) **\< .001**[\*\*](#t003fn002){ref-type="table-fn"} **The rescue work was successful** (Scale 1--5): (Successful, meaningful, sufficient advice and support) 4.0 (4.0--4.1) 3.7 (3.6--3.7) 4.2 (4.1--4.3) **\< .001**[\*\*](#t003fn002){ref-type="table-fn"} **More prepared to master similar situations in the future** 3.6 (3.6--3.7) 3.6 (3.5--3.7) 3.5 (3.3--3.6) ns *Note*: Mean (95% CI) or n (%). Scale: 1 = not at all, 5 = to a very high extent. \*\* p = \<.001. ANOVA Predictors of role clarity {#sec024} -------------------------- Eighteen variables were univariately significantly associated with role clarity during the rescue work ([Table 4](#pone.0156536.t004){ref-type="table"}). In the multivariable analysis, being female (OR 1.4), more years of work experience in the current organization (OR 2.3--3.5), experiencing training based on simulation (OR 1.6), having previously experienced an event with \> 5 fatalities (OR 1.6), not feeling a lack of control (OR .8), not being obstructed (OR .4) and perceiving the rescue work as a success (OR 2.3) were all associated with role clarity. There were no significant differences across professional groups. 10.1371/journal.pone.0156536.t004 ###### Predictors of role clarity during the rescue work. {#pone.0156536.t004g} Variables Univariable (unadjusted) Multivariable (adjusted) ------------------------------------------------------- -------------------------- -------------------------- ------------------------------------------------ ----- ---------- --------------------------------------------------- **Gender** (female vs. men) 1.3 1.0--1.8 .036[\*](#t004fn002){ref-type="table-fn"} 1.4 1.0--2.0 **.041**[\*](#t004fn002){ref-type="table-fn"} 30--49 years (reference group)1.5 1.0--2.2 .026[\*](#t004fn002){ref-type="table-fn"} \>50 years 1.7 1.1--2.6 .024[\*](#t004fn002){ref-type="table-fn"} \< 1 year (reference group) 1--5 years 2.0 1.2--3.5 .011[\*](#t004fn002){ref-type="table-fn"} 2.3 1.3--4.3 **.006**[\*](#t004fn002){ref-type="table-fn"} \> 5 years 3.6 2.1--6.0 \<.001[\*\*](#t004fn003){ref-type="table-fn"} 3.5 1.9--6.2 **\<.001**[\*\*](#t004fn003){ref-type="table-fn"} **Rescue worker group**: Police officers (reference group) Healthcare providers 1.8 1.3--2.5 \<.001[\*\*](#t004fn003){ref-type="table-fn"} Fire-fighters 2.6 1.4--4.7 .002[\*](#t004fn002){ref-type="table-fn"} Previous work experience 1.4 1.1--1.9 .015[\*](#t004fn002){ref-type="table-fn"} Training based on simulation 1.6 1.2--2.1 .001[\*](#t004fn002){ref-type="table-fn"} 1.6 1.1--2.2 **.007**[\*](#t004fn002){ref-type="table-fn"} Previous disaster drill 1.5 1.1--2.0 .004[\*](#t004fn002){ref-type="table-fn"} Previous event with \> 5 fatalities 1.7 1.2--2.4 .002[\*](#t004fn002){ref-type="table-fn"} 1.6 1.1--2.5 **.019**[\*](#t004fn002){ref-type="table-fn"} Working with adequate resources (high vs. low) 3.1 2.3--4.1 \<.001[\*\*](#t004fn003){ref-type="table-fn"} Perceived threat .8 .7--.9 .004[\*](#t004fn002){ref-type="table-fn"} Witnessing ns Present at the site of terror (yes vs. no) ns Arousal ns Dissociation .7 .5--.9 .003[\*](#t004fn002){ref-type="table-fn"} Felt overwhelmed (scale 1--5) .7 .6--.8 \< .001[\*\*](#t004fn003){ref-type="table-fn"} Felt a lack of control (scale 1--5) .6 .5--.7 \< .001[\*\*](#t004fn003){ref-type="table-fn"} .8 .7--.9 **.003**[\*](#t004fn002){ref-type="table-fn"} **Satisfaction** Perceived success in the rescue work (scale 1--5) 2.9 2.3--3.7 \< .001[\*\*](#t004fn003){ref-type="table-fn"} 2.3 1.7--3.0 **\<.001**[\*\*](#t004fn003){ref-type="table-fn"} Perceived obstruction in the rescue work (scale 1--5) .3 .3--.4 \< .001[\*\*](#t004fn003){ref-type="table-fn"} .4 .3--.5 **\<.001**[\*\*](#t004fn003){ref-type="table-fn"} *Note*: Logistic regression analysis, forwald wald. \* p \> .05, \*\* p \> .001. The dependent variable is dicotomized: 0 = 1--3: 1 = 4--5 on the 1--5 scale Predictors of being more prepared to master future operations {#sec025} ------------------------------------------------------------- Four variables were univariately significantly associated with being more prepared to master similar situations in the future. In the multivariable analysis, being aroused (OR 2.0) and perceiving the rescue work as successful (OR 1.5) were both independent predictors. Discussion {#sec026} ========== Most respondents were experienced, had been working in their current organization for at least five years and had participated in disaster training and/or drills. The prevalence of possible PTSD was very low. A high level of role clarity was significantly associated with work-related variables such as time working at their current organization, disaster training and/or prior experience from mass casualty events with more than five fatalities. Role clarity was also associated with a feeling of control, that the rescue work was perceived as a success, and with female respondents. When perceived mastering for future disasters was examined, there was an association between higher levels of arousal and the rescue work being perceived as a success. Moreover, there were no statistically significant differences between professional groups when predictors were taken into account. Role clarity {#sec027} ------------ Two key areas were associated with a higher level of role clarity: 1) preparedness, and 2) the perception of control, and rescue work success. ### 1) Preparedness {#sec028} We wanted to study personnel working in organizations with an emergency preparedness plan. A written plan is not sufficient to handle a disaster, and the local organizations are responsible for training and drills to prepare for emergencies. Each department within an organization must implement the plan according to its specific functions and needs. Every large-scale event is different \[[@pone.0156536.ref019]\] and improvisation is necessary in order to meet specific demands. Being prepared strengthens the ability one has to cope during a chaotic situation ^8^, and everyday training and hands-on practice give an opportunity to develop skills and to be more confident with a given role. Based on preparation and training, when a disaster strikes, the personnel can perform everyday work with an adaptation to the particular large-scale event at hand \[[@pone.0156536.ref020]\]. In the present study, the respondents were experienced, and most had been trained in emergency preparedness. These factors were associated with higher role clarity, which may support Weisæth and Kjeserud \[[@pone.0156536.ref003]\] who emphasize that readiness without training is of sparse value. In addition, the terror sites were situated in areas with large populations, hence increasing the odds of experienced personnel. In Scandinavia, a number of hospitals receiving critically injured patients are offered systematic and regular trauma training \[[@pone.0156536.ref005]\]. Multidiciplinary systematic large-scale disaster drills are part of the training for rescue personnel \[[@pone.0156536.ref021]\]. This may reflect the significant relationship between preparedness and a higher degree of role clarity during the terror events, and thereby highlight the importance of regular systematic training. This is in accordance with studies of systematic team-based training in trauma care that showed beneficial in-hospital effects \[[@pone.0156536.ref004]\], well-functioning cooperation \[[@pone.0156536.ref022]\] and increasing role clarity after systematic multidisciplinary team training \[[@pone.0156536.ref023]\]. On the other hand, a systematic review found insufficient evidence for the effectiveness of training interventions to help improve knowledge and skills during an exercise \[[@pone.0156536.ref007]\]. That review included five papers, which studied out-of-hospital healthcare providers. The studies were from heterogenic samples working in different organizations, which may explain the different outcome relative to our study. ### 2) Perception of control and success {#sec029} A greater sense of control was also associated with a higher level of role clarity. It is therefore likely to assume that there is a connection between a sense of control in performing everyday routines (preparedness) and being confident in performing similar tasks during a disaster. The police officers reported less control than the other groups. Some of those who worked at the sites of terror faced extremely difficult tasks, such as risk of collapse of buildings, risk of other terror attacks, an overloaded rescue boat, long response time for the police helicopter and guidelines to hold the other rescuers back because of the unsecured situation with the terrorist A feeling of success also had a significant association with role clarity, which can be attributed to practiced routines, and a corresponding greater sense of control. Helping victims during the senseless terror attacks was a positive experience for many professionals. Female gender was also associated with role clarity when we analyzed the groups together. Almost all firefighters were men, so we performed a regression analysis for each group separately. Female gender was not a significant predictor in any of the groups, but became significant when all groups were analyzed together. We do not have any data to explain this, but more women worked as healthcare providers relative to the other groups. Despite a demanding situation and unusual tasks, role clarity may be related to working in familiar surroundings within the hospital with well-known equipment and sufficient recourses. Better prepared to master similar situations in the future {#sec030} ---------------------------------------------------------- Arousal and a feeling of success were associated with a sense of being better prepared to master similar situations in the future. The degree of arousal was moderate and may have contributed to feelings of mastery, control, and thus to a sense of a successful operation. Under demanding conditions, a low level of arousal can be a sign that the work was perceived as being routine. In this special situation, there were many unknown events, but the rescue personnel had a high level of knowledge and training, which may have influenced their level of arousal. One can wonder if a high level of arousal can contribute to a feeling of loss of control, and hence a negative experience that could subsequently reduce the learning effect. A positive experience can confirm that one has handled a situation properly, something that will increase confidence in managing similar situations. One study supports this conclusion \[[@pone.0156536.ref024]\], finding that disaster training increases one's ability to cope with complex situations. Post-traumatic stress symptoms {#sec031} ------------------------------ Personnel in all groups were exposed to highly traumatic experiences, though the prevalence of possible PTSD (PCL-S \>50) was low (0.6%), and with no significant differences between the groups. This is even lower than the findings in policemen after the 2004 Madrid train bombings (3.9%) \[[@pone.0156536.ref025]\]. A large number of the rescue personnel in the present study witnessed people in despair and injured or deceased victims, and even felt threatened, but few reported this as very/extremely stressful. This may be the result of the extensive training and experience among the rescue workers \[[@pone.0156536.ref026]\]. Among deployed personnel after the 2004 Indian Ocean tsunami \[[@pone.0156536.ref009]\] preparedness, including training in similar tasks and exposure to mass casualty events, was also related to a lower degree of PTSD. Most of the rescue personnel were offered debriefing and/or peer-support after the duty. This may have partly influenced the low rates of possible PTSD. We find it however, more likely that the main reason for the low rate of possible PTSD was that the staff was well prepared and had high role clarity. It is unlikely that a single debriefing should be sufficient to prevent PTSD. Strengths and Limitations {#sec032} ------------------------- The moderate to high response rate was considered to be an advantage of the present study, especially for the firefighters, while the use of a well-known questionnaire such as the PCL-S was also a strength. Lower response rate among healthcare providers and police may represent a selection bias. We have little information on non-responding personnel. The variables "preparedness" and "role clarity" had only one item, which may be considered a limitation, even though we think it should be clear to the participants what is meant by the question. More detailed questions about working tasks, responsibilities and work exposure could have yielded additional information (e.g. interview), but this would have required considerable resources that were not available. However, it is unlikely that this would have changed our main finding, namely that the personnel were experienced and trained, and that preparedness was associated with role clarity. Respondents on sick leave might not have received the questionnaires, which may have biased the results. We did not perform assessments of previous or current mental problems, coping styles or personality traits. In addition, we did not determine the marital status or educational levels of the participants, or whether they had lost someone close during the attacks. Even so, it is unlikely that this would have altered the main findings of the study. The participants completed the questionnaire 8--11 months after the event. Time elapsed from the event to inclusion may represent a recall bias. It may be considered a limitation of this study that we did not conduct a prospective study with at least two time points. The development of e.g posttraumatic stress symptoms could then been found. On the other hand, a prospective design could not be anonymous, which would probably have lowered the response rate. By determining the symptoms after almost one year, we obtained data that demonstrated the long-term effects of the event. Clinical Implications {#sec033} --------------------- The majority of the rescue workers had been trained and prepared to handle both everyday and large-scale mass casualty events. There was an association between this and role clarity, but also between role clarity and perceiving control and success. The low level of post-traumatic stress symptoms indicates that the rescue workers who participated were quite resilient, and there are reasons to believe that this also may be associated with preparedness. This strengthens the importance of preparedness, not only in relation to the importance of emergency preparedness plans, but especially for training within the organization. Conclusion {#sec034} ========== The rescue workers who participated in the terror operations in 2011 were exposed to high levels of death, injuries and destruction. Most personnel within all groups were experienced and prepared for handling traumatic events. Preparedness predicted a higher perceived role clarity together with a perception of feeling in control. Perceiving the work as successful predicted both a higher role clarity and preparedness for future incidents. Preparedness provides role clarity, and must be prioritized in training programs among rescue workers. This study is a collaboration between Oslo University Hospital (OUS), Vestre Viken Hospital Trust (VV), the University of Oslo (UiO) and The Norwegian Centre for Violence and Traumatic Stress Studies (NKVTS). We owe a debt of gratitude to professor Edvard Hauff (UiO) and professor Trond Heir (NKVTS) for their collaboration in planning the study. The authors appreciate the participation of 1,213 police officers, firefighters and healthcare providers, who worked closely to the human suffering after the terror attacks in Norway. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: LS ØE. Performed the experiments: LS MJBP AG. Analyzed the data: LA ØE. Wrote the paper: MJBP AG BRR ØE LS.

Findings {#Sec1} ======== Infections are uncommon causes of scleral inflammation \[[@CR1]\]. Diagnosis is often difficult and gets delayed as the clinical picture appears similar to the more common cause: the immune-mediated disease. Fungal infections of the sclera have been reported following surgeries for retinal detachment \[[@CR2]--[@CR5]\], pterygium \[[@CR6], [@CR7]\], cataract \[[@CR8], [@CR9]\] and as a part of systemic fungal infections \[[@CR10], [@CR11]\]. All these reports are in immunocompetent individuals with a significant inflammatory response. Acquired immune deficiency syndrome (AIDS) patients are prone to many opportunistic fungal infections, but ocular fungal infections are rare and usually do not involve the sclera \[[@CR12]\]. We report a case of fungal scleral abscess caused by *Candida albicans* in a patient with AIDS and its successful resolution following antifungal therapy. Case report {#Sec2} ----------- A 57-year-old Asian (Indian) male was first seen at our hospital in July 2015. He came to us with a history of redness in his right eye for 15 days associated with pain and watering. He had no complaints regarding his vision. He was a known diabetic for 20 years and was detected to be infected with human immunodeficiency virus (HIV) 10 years back. His CD4 count was 461 cells/mm^3^, and he was on highly active antiretroviral therapy (HAART). His blood sugar levels were moderately controlled. There was no history of trauma or any other significant history. On examination, his best corrected visual acuity (BCVA) was 6/9, N6 in both eyes. Slit lamp examination of the right eye revealed evidence of conjunctival congestion and an oval scleral ulcerative lesion around 3 mm posterior from the corneal limbus at around 6 o'clock meridian, measuring 6 mm by 4.5 mm (Fig. [1](#Fig1){ref-type="fig"}). The ulcer had dense infiltrates with a pseudomembrane. The cornea was clear and the anterior chamber was deep and quiet. Posterior segment examination showed a clear vitreous and a normal fundus. Left eye examination was within normal limits. Intraocular pressure by applanation tonometry was 14 mmHg in both the eyes. On systemic examination, there was no *Candida* or any fungal infection elsewhere in the body.Fig. 1Slit lamp photograph showing the scleral abscess of about 6 mm by 4.5 mm He was diagnosed as an infective scleral abscess of the right eye. Scraping of the lesion was done and was put on empirical treatment of oral indomethacin and topical moxifloxacin eye drops one hourly. Scraping did not reveal any fungal or bacterial elements, apart from occasional pus cells. He was advised to continue medications. However, after 3 days, culture on Sabouraud's dextrose agar showed *C. albicans* species (Fig. [2](#Fig2){ref-type="fig"}) which confirmed the clinical suspicion. He was given amphotericin B 0.25 % eyedrops one hourly, voriconazole 1 % eyedrops one hourly and ciprofloxacin 0.3 % eyedrops four times per day along with oral voriconazole 200 mg tablet twice daily. The patient was continued on the same medication, while the lesion started increasing in size.Fig. 2Sabouraud's dextrose agar (SDA) showing the isolates of *Candida albicans* Two days later, the scleral abscess showed signs of improvement. The epithelial defect and congestion persisted. Culture and sensitivity was done which revealed the fungus sensitive to amphotericin B and natamycin and resistant to voriconazole, fluconazole and itraconazole. Hence, topical natamycin was added and voriconazole was stopped. The rest of the medications were continued as earlier. The lesion started regressing and the patient was reviewed after 2 weeks. At 2 weeks follow-up, his BCVA was 6/9, N6 in both the eyes. The patient had no fresh complaints. Slit lamp examination revealed complete healing of the lesion (Fig. [3](#Fig3){ref-type="fig"}). He was advised to stop topical and oral medications and was continued on only topical lubricants. At 1 month of follow-up, there was complete resolution of the lesion. His viral load was increasing along with decrease in CD4 counts and was shifted to second line HAART. At his last visit in January 2016, his viral load was less than 150 copies/ml and was doing well on second line HAART.Fig. 3Slit lamp photograph showing completely resolved scleral lesion Discussion {#Sec3} ---------- Fungal infections of the sclera are devastating cause of infectious scleritis as they are difficult to diagnose and often diagnosed late. The reported incidence of fungal scleritis is around 11 to 38 % of the total infectious causes of scleritis \[[@CR13]--[@CR15]\]. *C. albicans* is a dimorphic commensal fungus. Candidiasis is usually seen in immunocompromised individuals like HIV-infected patients. Candidiasis has a varied presentation. *C. albicans* usually causes keratitis, chorioretinitis and endogenous endophthalmitis in HIV/AIDS patients \[[@CR16]--[@CR18]\]. Scleral infection by *C. albicans* is very rare. Ahn et al. have reported two cases of fungal scleral infection in immunocompetent individuals \[[@CR19]\]. Garelick et al. have described a case of *Cryptococcus albidus* in a patient with AIDS \[[@CR20]\]. No case of scleral abscess has been reported in any patient with HIV/AIDS. Our patient is a HIV-positive patient and has a scleral abscess caused by *C. albicans*. Hence, it should be considered as a possible diagnosis and early investigation and treatment should be done, as it can lead to devastating complication like endophthalmitis. Our case also highlights the fact that a strong degree of clinical suspicion backed by appropriate anti infective (antifungal therapy) is a must in complete resolution of the lesion. Our case demonstrates the utility of culture and sensitivity in choosing the appropriate antifungal agent, since the initial use of a broad spectrum antifungal did not yield the required result. Based on culture and sensitivity, specific drugs were used which lead to complete resolution of lesions. Conclusion {#Sec4} ---------- In conclusion, we report an uncommon presentation of a *C. albicans* scleral abscess in an AIDS patient, who was treated promptly by appropriate topical and oral antifungals. Proper scraping and culture and sensitive reporting are an essential component of diagnosis and treating such a rare case thus preventing grave consequences. Consent {#Sec5} ======= Written informed consent was obtained from the patient. AIDS : acquired immune deficiency syndrome BCVA : best corrected visual acuity HAART : highly active antiretroviral therapy HIV : human immunodeficiency virus The authors declare no financial interest or sources of support. This paper has never been submitted to any journal nor published in any form of media before. **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** HS was involved in the data analysis and drafting of the manuscript. SS helped in the patient interaction and diagnosis and was involved in revising the manuscript critically. LT did the microbiological support and drafting of the manuscript. MA helped in the patient interaction and diagnosis and was involved in revising the manuscript critically. JB helped in the patient interaction and diagnosis, was involved in revising the manuscript critically and gave final approval of the version to be published. All authors read and approved the final manuscript. We acknowledge the Departments of Uvea and Microbiology for their support.

1.. Introduction {#s0001} ================ Carbonic anhydrases (CAs, EC 4.2.1.1) are important metalloenzymes involved in the hydration of carbon dioxide, a quite simple reaction whose balance is important in many cellular and physiological processes, spanning from pH homoeostasis and respiration to biosynthetic pathways including lipogenesis, glucogenesis, and ureagenesis[@CIT0001]. Human CAs (hCAs) exist in fifteen isoforms, which possess different characteristics as catalytic activity, tissues distribution, cellular localisation (cytosol, mitochondria and cell membrane) and sensibility to inhibitors. The researches of the last years showed all the consequences of their altered activity, either in the case of excessive or deficient ones[@CIT0004]^,^[@CIT0005]. In this regard, investigations on synthetic and natural compounds have been done with the aim to discover new selective and potent inhibitors and/or activators, which could restore the normal functions of these enzymes[@CIT0006]. One of the most studied alterations has been observed in tumours, where some hCA isoforms appear overexpressed; the so-called tumour-related isoforms, hCA IX and XII which are transmembrane enzymes, participate along with cytosolic hCA II in the complex pH machinery which controls the *milieu* of hypoxic tumours, promoting drug resistance as well as proliferation, migration, invasion and metastasis of tumour cells[@CIT0014]. To reverse this condition, efforts have been made to develop inhibitors that act effectively and selectively towards these targets. The design of more molecules and scaffolds, and the obtaining of co-crystals of some of them with hCA enzymes (mostly with hCA II), prompted the discovery of four different mechanisms of inhibition for hCAs[@CIT0018]^,^[@CIT0019]. The most recently reported one was observed for the first time with 2(benzylsulfinyl)benzoic acid ([Figure 1](#F0001){ref-type="fig"}) co-crystallized with hCA II[@CIT0020]. This compound showed an atypical mechanism involving the occupancy of a pocket next to the entrance of the active site, where it established some interactions holding the His64 residue in the "*out*" conformation. This amino acid, which exists in two conformations called "*in*" and "*out*", acts as a proton-shuttle system transferring a proton from the zinc-coordinated water molecule to the environment, to reconstitute the catalytic active form. Interfering with this process (the rate-determining step of the entire catalytic cycle) is equal to block or strongly reduce the enzymatic activity, obtaining the enzyme inhibition. Our lead compound was inactive against hCA I and XII (*K*~i~ hCA I/XII \>10 µM), while showed inhibitory activity in the low micromolar range towards both hCA II and IX, with a slight preference for the former (*K*~i~ hCA II = 0.15 µM, *K*~i~ hCA IX = 1.29 µM). With this in mind, we developed a novel series of derivatives based on the lead compound scaffold, pursuing some changes on its structure in order to improve activity and selectivity ([Figure 1](#F0001){ref-type="fig"}). {#F0001} The carboxylic acid moiety (red in [Figure 1](#F0001){ref-type="fig"}) was replaced with methyl ester, amide, *N*-methyl amide, hydroxamic acid and ketone. The selected functionalities possess different hydrogen-bond acceptor/donator profiles if compared with the carboxylic acid one; furthermore, except for amide and hydroxamic acid moieties that still conserve a "deprotonatable" group, the others lack this feature whose importance has been in this way deepened. The benzyl moiety (blue in [Figure 1](#F0001){ref-type="fig"}) was challenged through the insertion of substituents or by its entire substitution with benzoylmethyl and phenylacetic moieties as well as unsaturated C~3~--C~5~ alkyl chains. The aim of these changes was to understand the importance of the hydrophobic interactions with Phe231 and Asn232 that the phenyl ring established in the hCA II-inhibitor adduct. The addition of substituents on this phenyl ring changed the π-system charge distribution along with its hydro/lipophilicity. On the contrary, moving the phenyl group away from the sulphur atom with the addition of a carbonyl group (in the benzoylmethyl or phenylacetic derivatives) shed light on the importance of this system in that position and its conformational freedom. Attempts with unsaturated alkyl chains with increasing lengths have been done in order to evaluate if the presence of groups, able to give hydrophobic interactions, could work similarly to the phenyl ring. The role of sulphur atom was also challenged (yellow, green and violet in [Figure 1](#F0001){ref-type="fig"}), evaluating how both its oxidation state and its bioisosteric replacement affected the inhibitory activity. In the first case, we performed sulphide oxidation in order to obtain sulphinyl (sulfoxides) and sulphonyl (sulfones) derivatives. This synthesis was performed without chiral auxiliaries, obtaining sulfoxides in a mixture of enantiomers (racemate) that were resolved using a chiral HPLC system[@CIT0021]^,^[@CIT0022] and individually tested; this permitted the evaluation of how chirality could affect inhibitory activity of these derivatives. On the contrary, the bioisosteric replacement of sulphur atom with oxygen and nitrogen ones or with methylene group has been done, unravelling the importance of that atom for the inhibitory activity (violet in [Figure 1](#F0001){ref-type="fig"}). 2.. Chemistry and HPLC enantioseparation {#s0002} ======================================== In order to obtain the novel compounds, we followed the synthetic strategies reported in Schemes 1--4. For the synthesis of derivatives **1--3** and **16--18** we started from 1--(2-bromophenyl)ethan-1-one and 2-nitrobenzamide, respectively ([Scheme 1](#SCH0001){ref-type="scheme"}); both of them (individually) underwent an aromatic nucleophilic substitution with benzyl mercaptan. The reactions were performed in *N,N'*dimethylformamide (DMF) at reflux, in the presence of K~2~CO~3~. The so obtained compounds (**1** and **16**) were treated in the next oxidative reaction with *meta*-chloroperbenzoic acid (mCPBA). Although many routes able to afford sulphur oxidation exist[@CIT0023], we have deliberately chosen this approach to obtain the two oxidation products, the sulfoxide and the sulphone, in the same reaction. {#SCH0001} In fact, controlling the amount of oxidant added during the reaction we obtained both the species that were easily separated owing to the different chromatographic profiles. Compounds **7--12** and **19--21** were obtained with a nucleophilic substitution reaction between methyl 2-mercaptobenzoate and (un)substituted benzyl bromides, followed by the oxidative step. The carboxylic acid derivatives were obtained through the hydrolysis of compound **7** with aqueous 2 N sodium hydroxide (NaOH) in a mixture of equal amounts of water and 1,4-dioxane (50/50, *v*/*v*) to give **4**, that was subsequently oxidised to **5** and **6** ([Scheme 1](#SCH0001){ref-type="scheme"}). Compounds **13**--**15** were synthesised performing the first step of nucleophilic substitution between methyl 2-mercaptobenzoate and benzoyl bromide at room temperature, with successive oxidation in the same conditions previously seen ([Scheme 2](#SCH0002){ref-type="scheme"}). The synthetic pathway of derivatives **22**--**38** is reported in [Scheme 3](#SCH0003){ref-type="scheme"}. Methyl 2mercaptobenzoate was reacted in the presence of bromoalkanes of increasing length, in DMF at reflux in the presence of K~2~CO~3~ to obtain intermediates **E2--E5** (**E1**, methyl 2-(methylthio)benzoate, was commercially available). Only two of the **E** intermediates bearing butyl and pentyl chain respectively (**E4** and **E5**), underwent the oxidative step obtaining the sulphinyl derivatives **22--23**. All the **E1**--**E5** intermediates were hydrolysed with aqueous 2 N sodium hydroxide in a mixture of equal amounts of water and 1,4-dioxane (50/50, *v*/*v*). After synthesis completion, the reactions were quenched with hydrochloric acid (HCl), giving the intermediates **A1--A5** ([Scheme 3](#SCH0003){ref-type="scheme"}). The activation of the carboxylic acid group for the amide synthesis has been obtained using the mixed anhydride approach. In this regard, we used ethyl chloroformate in tetrahydrofuran (THF) under nitrogen atmosphere (N~2~) in the presence of triethylamine (Et~3~N) excess. Monitoring the reaction with thin layer chromatography (TLC) we detected the disappearance of the acidic intermediate, until the completion of anhydride formation. Then, ammonium chloride (NH~4~Cl) was added. The presence of the triethylamine excess involved the "de-blocking" of NH~3~ from NH~4~Cl, giving the final products (**24**, **27**, **30**, **33**, **36**). {#SCH0002} {#SCH0003} Although chiral properties are often linked with the presence of a quaternary carbon atom binding different substituents, even the sulphur atom can have chiral behaviour under appropriate conditions. The sulfoxides have a pyramidal structure where one vertex contains the sulphur electron pair, which can be considered as the fourth substituent bound to the sulphur atom. Therefore, if the two groups bound on the sulphinyl moiety are diverse, there is a chiral centre and the presence of two enantiomers. Differing from tertiary amines, which had low pyramidal inversion energy barrier, the sulfoxides possess energy barrier that allows the existence of two stable enantiomers, which can be resolved. Some sulfoxides (**5**, **8**, **11**, **17**, **20**) obtained in of the synthetic pathway shown in [Schemes 1--4](#SCH0001 SCH0002 SCH0003 SCH0004){ref-type="scheme"}, were selected and mg-quantities of their pure enantiomeric forms easily isolated by semipreparative enantioselective HPLC on the 250 mm × 10 mm i.d. Chiralpak IC column using pure ethanol as a mobile phase[@CIT0024]. In compounds **39**--**53**, the sulphur atom characterising molecules **1**--**38** was replaced by an isosteric oxygen (**39**, **44**, **47**, **52**), nitrogen atom (**40, 42**, **43**, **45**, **48**, **50**, **51**, **53**), or alkyl chain (**46**, **41**, **49**). Compounds **39--41** and **43** (anthranilic acid) were commercially available and used as purchased. Compound **42** was obtained starting from anthranilic acid, which was treated with sodium hydride (NaH) in dry THF. After 10 min, the phenylacetyl bromide was added at room temperature under nitrogen, giving the desired amide ([Scheme 4](#SCH0004){ref-type="scheme"}). Compounds **44--46** were synthesised starting from the corresponding carboxylic acids **39**--**41**, which were treated with sulphonyl chloride (SOCl~2~) in boiling dry methanol to give the corresponding methyl esters ([Scheme 4](#SCH0004){ref-type="scheme"})[@CIT0025]. {#SCH0004} The reaction of the ester **45** with hydroxylamine hydrochloride in methanol at room temperature for 48 h gave the *N*-hydroxy benzamide **51**[@CIT0026]. Compounds **47--49** were obtained converting the carboxylic acids **39--41** to the corresponding acyl chlorides by means of oxalyl chloride in the presence of a catalytic amount of dry DMF. Acyl chlorides were subsequently treated with ammonium hydroxide (NH~4~OH), to afford the desired amides ([Scheme 4](#SCH0004){ref-type="scheme"})[@CIT0027]. The amide **50** was obtained by a different route ([Scheme 5](#SCH0005){ref-type="scheme"}). At first, it was synthesised the amide from the anthranilic acid activated by SOCl~2~ in DMF at room temperature as previously reported. The addiction of NaH and phenylacetyl chloride in dry THF at room temperature let to the final amide **50**. {#SCH0005} Finally, ketones **52** and **53** were easily obtained through the reaction of 2-hydroxy or 2-aminoacetophenone, respectively, with benzylbromide in the presence of K~2~CO~3~ in dry acetone or acetonitrile, respectively ([Scheme 5](#SCH0005){ref-type="scheme"}). 3.. Experimental protocols {#s0003} ========================== 3.1.. General {#s0004} ------------- Solvents were used as supplied without further purification. Starting materials and other chemicals were purchased by Sigma-Aldrich (Milan, Italy) and used in the syntheses and in the biological assays without further purification. All synthesised compounds have been fully characterised by analytical and spectral data. Column chromatography was carried out using Sigma-Aldrich^®^ silica gel (high purity grade, pore size 60 Å, 200--425 mesh particle size). Analytical thin-layer chromatography was carried out on Sigma-Aldrich^®^ silica gel on TLC aluminium foils with fluorescent indicator 254 nm. Visualisation was carried out under UV irradiation (254 and 365 nm). ^1^H NMR spectra were recorded on a Bruker AV400 (^1^H: 400 MHz, ^13 ^C: 101 MHz). Chemical shifts are quoted in ppm, based on appearance rather than interpretation, and are referenced to the residual non deuterated solvent peak. Missing signals must be attributed to overlapping peaks. Infrared spectra of the most representative compounds were recorded on a Bruker Tensor 27 FTIR spectrometer equipped with an attenuated total reflectance attachment with internal calibration. Absorption maxima (ν~max~) are reported in wavenumbers (4000--400 cm^−1^). Elemental analyses for C, H, and N were recorded on a Perkin-Elmer 240 B microanalyzer obtaining analytical results within ± 0.4% of the theoretical values for all compounds. All melting points were measured on a Stuart^®^ melting point apparatus SMP1 and are uncorrected (temperatures are reported in °C). Where given, systematic compound names are those generated by ChemBioDraw Ultra^®^ 12.0 following IUPAC conventions. Mass spectra on the most representative compounds were performed on a LCQ (Thermo Finnigan) ion trap mass spectrometer (San Jose, CA, USA) equipped with an electrospray ionisation (ESI) source. The capillary temperature was set at 300 °C and the spray voltage at 4.25 kV. The fluid was nebulised using nitrogen (N~2~) as both the sheath gas and the auxiliary gas. All the characterisation data for each compound were reported as Supplemental material. 3.2.. Enzyme inhibition assays {#s0005} ------------------------------ An Applied Photophysics stopped-flow instrument has been used for assaying the CA catalysed CO~2~ hydration activity[@CIT0028]. Phenol red (0.2 mM) has been used as indicator, working at the absorbance maximum of 557 nm, with 20 mM Hepes (pH 7.5, for α-CAs) as buffer and 20 mM NaClO~4~ (for maintaining constant the ionic strength), following the initial rates of the CA-catalysed CO~2~ hydration reaction for a period of 10--100 s. The CO~2~ concentrations ranged from 1.7 to 17 mM for the determination of the kinetic parameters and inhibition constants. In particular, CO~2~ was bubbled in distilled deionised water for 30 min till saturation. A CO~2~ kit (Sigma, Milan, Italy) was used to measure the concentration in serially diluted solutions from the saturated one at the same temperature. For each inhibitor at least six traces of the initial 5--10% of the reaction have been used for determining the initial velocity. The uncatalyzed rates were determined in the same manner and subtracted from the total observed rates. Stock solutions of inhibitor (1 µM) were prepared in distilled-deionised water and dilutions up to 0.1 nM were done thereafter with the assay buffer. Inhibitor and enzyme solutions were preincubated together for 15 min at room temperature prior to assay to allow for the formation of the E-I complex or for the eventual active site mediated hydrolysis of the inhibitor. The inhibition constants were obtained by non-linear least-squares methods using PRISM 3 and the Cheng-Prusoff equation[@CIT0029], and represent the average from at least three different determinations. All recombinant CA isoforms were obtained in-house as previously reported[@CIT0030]^,^[@CIT0031]. 3.3.. Molecular modelling studies {#s0006} --------------------------------- The three-dimensional structures of all ligands were prepared in their lowest energy conformation using the MOE software package (v2019.01, Chemical Computing Group, Inc, Montreal, Canada). The sulphonamide nitrogen atoms of the ligands were assigned a negative charge (R-SO~2~NH^-^) and the ligands were energy minimised (MMFF94x force field). All protein structures were obtained from the RCSB protein databank: hCA I (pdb: 3lxe, 1.90 Å), hCA II (pdb: 4e3d, 1.60 Å), hCA IX (pdb: 3iai; 2.20 Å) and hCA XII (pdb: 1jd0; 1.50 Å). The protein atoms and the active site zinc ions were retained and all other atoms were omitted. The remaining structure was protonated using the protonate 3 D functionality of MOE and subsequently, the obtained structure was energy-minimised (AMBER14:EHT)[@CIT0032]. Finally, the obtained protein models were superposed on the hCA I structure using the backbone Cα-atoms and all Zn^2+^-ions, zinc-binding histidines and the overall backbone atoms superposed well (RMSD value: 1.281 Å). Docking calculations were performed using the FlexX docking tool (v2.3.2; BioSolveIT GmbH, St. Augustin, Germany) within MOE. The binding pocket was defined as all residues within 6.5 Å of the reference ligand acetazolamide. The sulphonamide tail of the ligands was forced to adopt a similar orientation and interactions to the Zn^2+^ ion as observed for acetazolamide using a pharmacophore model. All ligands were docked fifty times and the best scoring three poses were subjected to refinement calculations[@CIT0033]. To this end, the ligand and binding pocket residues were energy minimised and rescored using GBVI/WSA force field[@CIT0034]. 4.. Carbonic anhydrase inhibition studies {#s0007} ========================================= All the tested compounds had no affinity for the common off-target hCA I isoform (*K*~i~ \>100 µM) and some of them were more active against the tumour-related isoform hCA IX if compared with the parent drug 2-(benzylsulfinyl)benzoic acid ([Table 1](#t0001){ref-type="table"}). The derivatives **1**--**3** bearing the ketone group and with benzyl moiety bound to the sulphur atom, exhibited affinity and selectivity against the tested isoforms depending on the sulphur atom oxidation state. Compound **1** exhibited better activity than the lead compound against hCA IX (*K*~i~ hCA IX = 1.1 µM) and was ineffective against the two off-targets (*K*~i~ hCA I/II \>100 µM). The oxidation to sulfoxide (**2**) impaired the selectivity "restoring" the inhibitory activity owned by the lead compound for hCA II, although in the high micromolar range, while the affinity towards hCA IX (**2**, *K*~i~ hCA IX = 2.0 µM) was slightly inferior than **1**. The sulphone **3** displayed affinity exclusively for hCA IX, although weaker than **1** and **2** (**3**, *K*~i~ hCA IX = 16.4 µM). ###### Inhibitory activity of derivatives **1**--**53** and the reference drug (acetazolamide, AAZ) against the four selected hCA isoforms by a stopped-flow CO~2~ hydrase assay[@CIT0028]. Compound Structure *K*~i~ (μM)[^a^](#TF1){ref-type="table-fn"} -------------- ----------------------------------- --------------------------------------------- ------- ------- ------- **1**  \>100 \>100 1.1 nt **2**  \>100 63.2 2.0 nt **3**  \>100 \>100 16.4 nt **4**  \>100 \>100 21.8 nt **(*R*)-5**  \>100 0.21 1.4 nt **(*S*)-5** \>100 0.093 1.2 nt **6**  \>100 \>100 2.3 nt **7**  \>100 75.2 15.0 nt **(*R*)-8**  \>100 84.9 15.3 nt **(*S*)-8** \>100 65.7 20.0 nt **9**  \>100 \>100 11.4 nt **10**  \>100 45.7 1.0 nt **(*R*)-11**  \>100 \>100 1.4 nt **(*S*)-11** \>100 \>100 18.1 nt **12**  \>100 40.1 21.1 nt **13**  \>100 \>100 \>100 nt **14**  \>100 \>100 \>100 nt **15**  \>100 \>100 \>100 nt **16**  \>10 8.22 0.046 2.66 **(*R*)-17**  \>100 52.2 2.2 nt **(*S*)-17** \>100 51.9 1.9 nt **18**  \>10 2.67 \>10 0.066 **19**  \>100 86.3 19.8 nt **(*R*)-20**  \>100 52.2 22.8 nt **(*S*)-20** \>100 38.9 24.6 nt **21**  \>100 68.2 25.8 nt **22**  \>100 \>100 \>100 nt **23**  \>100 \>100 \>100 nt **24**  \>100 \>100 6.5 \>100 **25**  \>100 \>100 34.5 \>100 **26**  \>100 \>100 14.0 \>100 **27**  \>100 \>100 16.7 \>100 **28**  \>100 \>100 40.9 \>100 **29**  \>100 \>100 22.7 \>100 **30**  \>100 90.9 \>100 \>100 **31**  \>100 \>100 38.2 \>100 **32**  \>100 \>100 2.3 \>100 **33**  \>100 \>100 2.5 \>100 **34**  \>100 \>100 36.4 \>100 **35**  \>100 \>100 13.5 \>100 **36**  \>100 \>100 2.7 \>100 **37**  \>100 \>100 27.3 \>100 **38**  \>100 \>100 8.7 \>100 **39**  \>100 \>100 \>100 \>100 **40**  \>100 \>100 \>100 \>100 **41**  \>100 \>100 \>100 \>100 **42**  \>100 \>100 \>100 \>100 **43**  \>100 \>100 \>100 \>100 **44**  \>100 \>100 \>100 \>100 **45**  \>100 \>100 \>100 \>100 **46**  \>100 \>100 \>100 \>100 **47**  \>100 \>100 \>100 \>100 **48**  \>100 \>100 \>100 \>100 **49**  \>100 \>100 \>100 \>100 **50**  \>100 \>100 \>100 \>100 **51**  \>100 \>100 \>100 \>100 **52**  \>100 \>100 \>100 \>100 **53**  \>100 \>100 \>100 \>100 **AAZ**   0.25 0.012 0.25 0.006 Mean from three different determinations (errors in the range of 5--10% of the reported values). nt: not tested. The replacement of **1** methyl group with the hydroxyl one to give the **4**, the sulphide analogue of the lead compound, had detrimental effects reducing affinity for hCA IX (*K*~i~ hCA IX = 21.8 µM) but still conserving the selectivity (*K*~i~ hCAI/II \>100 µM). The oxidation to sulfoxide ((*R*/*S*)**-5**) and sulphone (**6**) improved inhibitory activity in the low micromolar range against the tumour-related isoform. However, while **6** bearing the sulphonyl moiety was inactive towards the two off-target actings selectively against hCA IX (**6**, *K*~i~ hCA I/II \> 100 µM; *K*~i~ hCA IX = 2.3 µM), the two sulphinyl enantiomers of **5** exhibited their preference for hCA II ((*R*)-**5**, *K*~i~ hCA II = 0.21 µM; (*S*)-**5**, *K*~i~ hCA II = 0.093 µM) rather than hCA IX ((*R*)-**5**, *K*~i~ hCA IX = 1.4 µM; (*S*)-**5**, *K*~i~ hCA IX = 1.9 µM). Comparing the inhibition data against hCA II of **5** enantiomers, it was quite clear the preference of the enzyme for (*S*)-**5**, which was the eutomer. Indeed, after incubation with the racemic mixture of **5**, only the adduct between hCA II and (*S*)-**5** was observed[@CIT0020]. Compounds **7--9** were endowed with a methyl ester functional group, keeping constant the benzyl group bound to the sulphur atom (in the different oxidation states). These derivatives showed reduced inhibitory activity against hCA IX, regardless the sulphur oxidation number. Furthermore, compounds **7** and (*R*/*S*)-**8** were also weakly active against hCA II (65.7 \< *K*~i~ hCA II (µM)\<84.9). The unfavourable effect of the additional methyl group was mitigated by the insertion of a chloro atom at the *para-*position of the phenyl ring (**10**--**12**); this substitution counteracted the detrimental effects of the methyl group, probably enforcing other interactions inside the enzyme. Compound **10** was a good inhibitor of hCA IX exhibiting *K*~i~ against hCA IX of 1.0 µM, with a high micromolar residual activity against hCA II (*K*~i~ hCA IX = 45.7 µM). The oxidation to sulphinyl group ameliorated selectivity, because the two enantiomers (*R*/*S*)-**11** were ineffective against both of hCA I and II (*K*~i~ hCA I/II \>100 µM). These two optical isomers showed different affinity towards hCA IX, with the (*R*)-enantiomer that inhibited this isoform better than the (*S*)-one ((*R*)-**11**, *K*~i~ hCA IX = 1.4 µM; (*S*)-**11**, *K*~i~ hCA IX = 18.1 µM). The further oxidation to sulphone (**12**) gave a detrimental effect, reducing both activity against hCA IX (*K*~i~=21.1 µM) and selectivity (*K*~i~ hCA II = 40.1 µM). The formal insertion of a carbonyl moiety between the methylene group and the phenyl ring in order to move the latter away from the sulphur atom, led to the **13**--**15** derivatives. The presence of this benzoyl moiety was not tolerated by this scaffold, giving compounds which did not exhibit effects towards all the tested isoforms (*K*~i~\>100 µM). The consequence of the amidic moiety insertion was also evaluated (**16--18**). Compound **16**, containing a non-oxidised sulphur atom, was the best inhibitor of the series against hCA IX displaying *K*~i~ of 0.046 µM. This result was slightly impaired by the oxidation to sulphinyl derivatives ((*R*)-**17**, *K*~i~ hCA IX = 2.2 µM; (*S*)-**17**, *K*~i~ hCA IX = 1.9 µM), while the corresponding sulphone (**18**) completely lost the affinity for hCA IX. All the amidic derivatives showed affinity for hCA II, albeit the sulphide (**16**) and sulphone (**18**) inhibited this isoform better than the sulphinyl enantiomers, which exhibited *K*~i~ values around 50 µM (**16**, *K*~i~ hCA II = 8.22 µM; **18**, *K*~i~ hCA II = 2.67 µM). The presence of a methyl group on the amidic moiety was detrimental for the inhibitory activity towards hCA IX. Indeed, the derivatives **19**--**21**, endowed with monomethyl amide, were less effective against the cancer-related isoform (19.8 \< *K*~i~ hCA IX (µM) \<25.8) than the analogues containing primary amide. Furthermore, the activity against hCA II was also lowered in the high micromolar range (36.9 \< *K*~i~ hCA II (µM)\<86.3). This behaviour was very similar to that observed for molecules containing a methyl ester moiety, which were less effective compared with their carboxylic acid analogues. The effect and the importance of the phenyl moiety were also challenged by substituting it with linear alkyl chains of increasing length (**22--38**) and testing the influence against the two off-target hCA I and II, and the two tumour-related isoforms hCA IX and XII. The first attempts were done with the two compounds **22** and **23**. These derivatives, bearing methyl ester with the sulfoxide group linked to butyl and pentyl chain respectively, were inactive against all the tested isoforms, underlying the detrimental effects of this structural combination. However, better results were obtained when the amide moiety was placed instead of methyl ester one (**24--38**). The derivatives **24--38** displayed their effect exclusively against hCA IX, without any inhibitory activity against hCA I, II and XII. Only for derivative **30**, bearing propyl chain bound to the sulphur atom, we observed activity against hCA II, but with a very low affinity (*K*~i~ hCA II = 90.9 µM). Compound **32**, the sulphone analogue of sulphide **30**, was the best inhibitor against hCA IX among the compounds substituted with alkyl chains, inhibiting the cancer-related isoform with a *K*~i~ value of 2.3 µM. Other good results were found with **33** and **36**, containing a sulphur atom and bearing butyl or pentyl chain, respectively (**33**, *K*~i~ hCA IX = 2.5 µM; **36**, *K*~i~ hCA IX = 2.7 µM). Except for derivatives with propyl chain (**30--32**), the compounds containing the non-oxidised sulphur atom showed better inhibitory activity than the related sulfoxide and sulphone analogues ([Table 1](#t0001){ref-type="table"}). In order to evaluate the role and the importance of sulphur atom in this scaffold, we performed its bioisosteric replacement with methylene moiety and oxygen or nitrogen atoms (**39--53**), keeping constant the functional groups constituting the scaffold seen above (e.g. carboxylic acid, amide, etc.). Some of the obtained compounds showed structures completely comparable to the lead compound (see for example **39, 41**). As one can see from [Table 1](#t0001){ref-type="table"}, all the attempts done changing the sulphur atom were ineffective, giving compounds unable to inhibit all the tested isoforms. Furthermore, also the influence of the tested functional groups like the acidic or amidic one, were useless, enforcing and underling the importance that sulphur atom has for the activity of this scaffold. In light of the above, some considerations can be done. Among the functional groups tested (carboxylic acid, methyl ester, etc.) the best hCA IX inhibitor was provided with amidic moiety, although ketone and carboxylic acid derivatives also showed good activity. The insertion of a methyl group on the carboxyl acid or the amide impaired activity. This could be due to the altered H-bond donator/acceptor profile, or most likely to the increased steric hindrance. From all the data evaluated, it was quite clear that the oxidation of sulphur atom did not influence the activity or selectivity in a specific manner. Furthermore, in most of the efforts done, the sulphide derivative was the best inhibitor of hCA IX among the sulphide/sulfoxide/sulphone analogues endowed with the same substituent. The only exception was for carboxylic acid derivatives, whose better activity was exhibited by the sulphinyl ones. The chiral resolution of the single sulfoxide enantiomers did not give a specific trend, with few enantiomers exhibiting better activity compared with the related optical isomer. The benzyl moiety was essential for the inhibition and the presence of halogen on the ring improved activity when compared with the unsubstituted analogues. The replacement of phenyl ring with linear alkyl chains impaired the activity, while raising the selectivity. The benzoyl moiety was not tolerated by this scaffold, affording compounds ineffective against all the tested isoforms. These outcomes could be related to the loss of hydrophobic interactions due to the phenyl ring shift, clashes with amino acids owing to the ring new position, or both of them. Nevertheless, the reasons of these data might be other. In fact, the inserted "spacer" (the carbonyl group), has an own steric hindrance along with specific bond angles and a stable dipole; all these features could affect negatively the interaction with the isoform IX. Finally, the substitution of the sulphur atom with oxygen, nitrogen one, or methylene underlined the importance of the sulphur atom for the activity of these compounds, because all the obtained derivatives were inactive against all the tested isoforms. Furthermore, some selected compounds were also tested against hCA XII to preliminary provide information on this target. All the derivatives, except compounds **16** and **18**, were inactive. Sulphone **18** was a potent and selective hCA XII inhibitor with *K*~i~ 0.066 µM, whereas sulphur-based **16** was a medium-potency inhibitor. 5.. Molecular modelling studies {#s0008} =============================== 5.1.. Docking studies into the active site of hCA IX {#s0009} ---------------------------------------------------- The lowest *K*~I~ value for hCA IX was measured for compound **16** (*K*~i~ = 46 nM), which is lower compared to the reference compound acetazolamide (*K*~i~ = 250 nM). Two different binding interactions with the hCA IX active site were suggested. In the first docked pose, the ligand directly formed an interaction with the Zn^2+^-ion via its carbonyl group ([Figure 2(A))](#F0002){ref-type="fig"}. The amine group of the ligand formed hydrogen bonds with the side chain hydroxyl groups of Thr199 and Thr200. One of the ligand's phenyl group interacted with the side chain of His64 (arene-H interaction). Changing the amino group or the phenyl group or replacing the S atom by SO or SO~2~ groups made it more difficult for the ligands to adopt this pose. Several compounds (including compounds **1**, **5**, **10**, **11**, **24** and **32**) can also adopt poses in which a direct interaction with the Zn^2+^ is possible. In the second docked pose, the ligand carbonyl group formed a hydrogen bond to the zinc-bound water molecule ([Figure 2(B))](#F0002){ref-type="fig"}. In addition, the carbonyl and amine group of the ligand formed hydrogen bonds with the side chain hydroxyl group of Thr200. Hydrophobic interactions were formed with the side chains of His94 and Val121. The replacement of the amino group with a hydroxyl group or the replacement of the S atom by SO or SO~2~ groups may be tolerated according to this pose. The later may form hydrogen bonds with the side chain of Gln92. Furthermore, many compounds can adopt docked poses in which an interaction with the water molecule is possible. {#F0002} 5.2.. Docking studies into the active site of hCA XII {#s0010} ----------------------------------------------------- The lowest *K*~i~ value for hCA XII was measured for compound **18** (*K*~i~ = 66 nM) and the second-lowest *K*~i~ value was measured for compound **16** (*K*~i~ = 2.66 µM). Two docked poses have been obtained for these compounds: both formed an interaction with the Zn^2+^-ion via the ligand carbonyl groups ([Figure 3](#F0003){ref-type="fig"}). The amino group of compound **18** formed hydrogen bonds with the side chains of Thr199 and Thr200, while the amino group of compound **16** only established a hydrogen bond to the side chain of Thr199. Both ligands formed hydrophobic interactions with His94. In addition, compound **18** established a hydrophobic interaction with the side chain of Val121. No poses have been obtained for these compounds with an interaction with the zinc-bound water molecule. {#F0003} 5.3.. Docking studies into the active site of hCA I and hCA II {#s0011} -------------------------------------------------------------- Although the actives sites of the investigated hCA isozymes are very similar to each other, important differences still exist that most likely influence the binding interactions of the ligands. In hCA I, His200 is present instead of Thr200 and therefore this hydrogen bonding opportunity (see [Figures 2](#F0002){ref-type="fig"} and [3](#F0003){ref-type="fig"}) was no longer present and the ligands cannot approach the Zn^2+^-ion to form direct interactions or interactions with a zinc-bound water molecule. This may be responsible for the low activity of these compounds against hCA I. Similarly, a relatively large Phe131 is present in hCA II instead of Ala131 in hCA XII. As such, the docked pose as observed of compound **18** in the active site of hCA XII ([Figure 3(A)](#F0003){ref-type="fig"}) would not be possible due to a clash with Phe131. 6.. Conclusion {#s0012} ============== In conclusion, we have explored the chemical space of the main functional groups of the 2(benzylsulfinyl)benzoic acid, an innovative and atypical hCA inhibitor. We have designed, synthesised, and tested 53 derivatives against the most important hCA isoforms establishing robust SARs within this scaffold. Most of them were selective hCA IX inhibitors in the low micromolar/high nanomolar range. The stereochemistry of the sulfoxide group had no impact on the biological activity, whereas the isosteric replacement of the sulphur atom with oxygen, nitrogen or methylene led to a total loss of activity. Molecular modelling studies demonstrated different binding poses of the inhibitors within the active site, differently from that observed for the parent compound. Disclosure statement {#s0013} ==================== The authors declare no conflict of interest. [^1]: Supplemental data for this article can be accessed [[here]{.ul}](https://doi.org/10.1080/14756366.2019.1651315).

Study rationale {#sec1} =============== Infectious diseases, nutritional deficiencies and maternal and neonatal conditions are the dominant contributors to the overall burden of disease worldwide \[[@ref1]\]. However, the burden of non-communicable diseases (NCDs) such as diabetes, cardiovascular disease and cancer, has increased in the last decade. In 2012, an estimated 2.8 million deaths in sub-Saharan Africa (SSA) were attributable to NCDs; they are projected to be the most common cause of death in SSA by 2030 \[[@ref2], [@ref3]\]. The increasing prevalence of diabetes is a key component of the global rise in NCDs. The International Diabetes Federation currently estimates a 54% increase in global prevalence of diabetes, from 382 to 592 million by 2035 \[[@ref4]\]. SSA is projected to have a 109% increase in the prevalence of diabetes, from 19.8 million in 2013 to 41.5 million by 2035, representing the highest proportional increase of any region in the world. The rise in diabetes prevalence is thought to be a consequence of rapid demographic and epidemiological transitions and urbanisation in the region \[[@ref4], [@ref5]\]. Within SSA, the prevalence of chronic diseases such as diabetes varies markedly among populations, suggesting that these disorders arise from complex interrelations between environmental and biological factors \[[@ref6]\]. Socioeconomic and demographic differences, diagnostic approaches, variation in lifestyle factors and genetic predisposition may explain some of these differences \[[@ref7]--[@ref9]\]. However, there is a lack of large-scale population-based studies that have assessed these systematically in SSA populations. Recent population-based diabetes studies have varied in size and scope (as shown in online Supplementary Table S1); many have only assessed a subset of potential risk factors, limiting comparative analyses. Relatively few have utilised validated survey designs, such as the WHO STEPwise approach to Surveillance (STEPS), to capture a broadened set of variables, as in the DDS. There is, therefore, a need for high-quality data to more reliably assess the burden of chronic diseases and the full spectrum of their risk factors, in order to determine the individual contribution and impact of these risk factors in SSA. In 2013, an estimated 9.3% of the population in South Africa had diabetes, which is one of the highest national prevalences of diabetes in SSA \[[@ref10]\]. Studies investigating the prevalence of diabetes in South Africa have reported differing prevalence between urban and rural areas and by ethnicity \[[@ref11]--[@ref14]\]. However, the only study population-based epidemiological study of diabetes in Durban was conducted in the early 1990s and found a prevalence of 5.3% \[[@ref15]\]. We therefore established the DDS as a population-based research framework to investigate a broad range of lifestyle, medical and genetic factors and their association with diabetes -- the primary research focus of the study -- together with other chronic diseases and traits including hypertension, dyslipidaemia, human immunodeficiency virus (HIV) and hepatitis C virus (HCV) in an urban South African (black) population living in the eThekwini Municipality (also known as the city of Durban). The study provides a resource to examine a wide range of research questions in order to inform health policy and practice around the burden and aetiology of chronic diseases; this in turn will enable effective targeting of resources in South Africa and across the region. The DDS also provides a framework for future research including the possibility of interventional studies. The DDS is a collaborative initiative between the University of KwaZulu-Natal, the University of Cambridge, the University of Oxford and the Wellcome Trust Sanger Institute. Study population {#sec2} ================ The study area is the eThekwini Municipality (city of Durban) in KwaZulu-Natal, South Africa as shown in [Fig. 1](#fig01){ref-type="fig"}. \[[@ref16]\] Based on the 2011 population census for South Africa, black Africans constitute 79.6% (41 000 938) of the total population of 50.8 million \[[@ref17]\]. KwaZulu-Natal is the second most densely populated province in South Africa, accounting for around one-fifth of the total population (10 267 300), of whom 86.8% are black African, 52.3% are female and 73.7% are aged 35 years or older \[[@ref18]\]. Fig. 1.Map of KwaZulu-Natal and eThekwini, South Africa. The map shows the municipal boundaries in KwaZulu-Natal, with the eThekwini Metropolitan Municipality highlighted in red. Adapted from Naudé *et al*. \[[@ref16]\]. The DDS sampling frame was developed independently by the Research and Policy Department of the eThekwini Municipality, and is divided into nine planning unit clusters (PUCs), representing nine townships (Umlazi, Inanda, KwaMashu, Ntuzuma, Mpumalanga Complex, Cato Manor, Clermont, Lamontville and Chesterville). Geographic information system mapping techniques were used to gather accurate geographical and boundary data on the selected PUC areas. The study region is predominantly urban with a total population of approximately 1 378 750 individuals, 61% of whom live in informal dwellings. In each PUC streets are randomly selected, with the number of streets selected being proportional to the size of the PUC population. Next, within each randomly selected street, households are nominated by systematic cluster sampling, with the number of formal or informal houses selected within each street proportional to the ratio of formal to informal housing in that PUC. Further, in each street the number of selected houses is oversampled, allowing for up to 20% non-participation without compromising overall study size. Within each nominated household, all family members who are of African descent, not pregnant and aged 18 years or older are eligible for recruitment into the study. Before participant recruitment and data collection begin, PUC areas are sensitised to the study using flyers, neighbourhood-specific community activation events and reminders for local residents. These stages of community engagement and mobilisation are conducted with the support of the provincial department of health (KwaZulu-Natal), local health authorities and community leaders. Once the study rationale and procedures have been explained, all eligible individuals present at the time of mobilisation are invited to participate in the study. A non-responder is defined as an individual who has declined an invitation to participate; a non-response household is defined as a household who has been contacted on at least three occasions with all habitants absent on each occasion. Each non-response household is replaced by another house pre-selected from the same street as described above. Data and sample collection are undertaken at pre-arranged study appointments at a central community venue. All study appointments occur in the morning; this is essential as participants are required to fast overnight for a 75 g oral glucose tolerance test (OGTT). Statistical power {#sec3} ================= The DDS sample size -- based on population size, prevalence precision estimates and power for epidemiological and genetic association studies -- is 10 000 participants. Previous studies in SSA have estimated a diabetes prevalence of around 10% (which is consistent with preliminary findings from our study; data not shown). Based on this estimate, and taking into account design effects, our sample size is sufficient to estimate the population prevalence of diabetes with a 95% confidence interval (CI) of less than 1%. For logistic regression analyses that involve diabetes as an outcome, based on 1000 cases and 9000 controls, the study has more than 80% power to detect odds ratios (ORs) as low as 1.2 for exposure prevalences equal to 30%. For exposure prevalences greater than 30%, ORs less than 1.2 can be detected with the same power. Further, for exposure prevalences as low as 5% the study will have more than 80% power to detect ORs of at least 1.5. At this sample size we will also have the ability to examine genetic variants that explain 0.5 % of the variation in a relevant trait with more than 80% power using a genome-wide statistical threshold. Data collection {#sec4} =============== The DDS collects detailed questionnaire information on participant health, lifestyle and socioeconomic indices, as well as biophysical measurements, biochemical, haematological and serologic biomarkers for non-communicable and infectious diseases, and genetic information, as shown in [Table 1](#tab01){ref-type="table"}. The questionnaire used in this study is an adaptation of the standardised and validated WHO STEPS tool designed for the collection of disease risk factors \[[@ref19]\]. The questionnaire responses are collected on an electronic data capture system \[electronic questionnaire (EQ)\]. For increased accuracy and efficiency the EQ has built-in data checks, including double entry of numbers, plausibility of answers and appropriate numerical limitations, the detailed methods of which have been previously described \[[@ref20]\]. Biophysical measurements (height, weight, waist and hip circumferences and blood pressure) and samples of venous blood (baseline fasting sample, 30- and 120-min) and urine are collected in the same study appointment to maximise efficiency. Adhering to standardised protocols, trained study personnel perform all study procedures and sample collections. Study equipment is calibrated regularly. Blood and urine samples are stored in cold boxes maintained at 4--8 °C until transported to a laboratory within 6 h of collection. Full blood count, plasma glucose and HbA~1c~ analyses are undertaken on the day of sample collection, at a local private laboratory. All other samples are transported to the University of KwaZulu-Natal laboratory and processed for storage (including centrifugation, separation into serum and plasma, and DNA extraction). The processed samples are stored at −70 °C for future analyses. Extracted DNA is sent to the Wellcome Trust Sanger Institute in the UK for next-generation genomic analyses. Table 1.Planned data component domains in the DDSDomainComponentsHealth and lifestyle questionnaireDemographic informationEducation, occupation and livelihoodSocioeconomic indices (household features and ownership)Tobacco useAlcohol consumptionDietary behaviourPhysical activityMedical historyFamily medical historyBiophysical measurementsBlood pressure and heart rateHeight and weightWaist circumference and hip circumferenceNCD & infectious disease biomarkers and genetic data Glycaemic biomarkersOGTT: 0-, 30-, 120-min plasma glucose 0-, 30-, 120-min serum insulinHbA~1c~ Cardiometabolic and haematological biomarkersFull blood countTotal cholesterolLDLHDLLp(a)TriglyceridesAspartate amino transferaseAlanine amino transferaseAlkaline phosphataseGamma glutamyl transferaseBilirubinAlbuminAlbumin--creatinine ratio (urine sample)Sodium and potassium (urine sample) Infectious disease biomarkersHIVHBVHCV Genetic markers (host and pathogen)Illumina SNP chip arraysIllumina whole genome and exome sequencingViral whole genome sequencing[^2] Feedback of results to the participants starts during the study appointment. All participants receive blood pressure and body mass index results immediately, accompanied by appropriate lifestyle and health guidance from trained study personnel including advice to seek medical referral. Results from the OGTT are posted to participants, unless they choose not to receive them. Participants may also choose to receive the results of other tests such as HIV, HCV and liver function tests, by appointment at the DDS feedback office. Where participant results are abnormal, according to clinical algorithms defined in accordance with national guidelines, participants are advised to seek further assessment (investigation and management) at the appropriate local health facility as part of their standard clinical care. Ethical considerations {#sec5} ====================== The DDS has obtained full ethical approval from the Biomedical Research Ethics Committee at the University of KwaZulu-Natal (reference: BF030/12) and the UK National Research Ethics Service (reference: 14/WM/1061). Written informed consent is obtained from all study participants. The consent form covers the following activities: data collection by questionnaire and biophysical measures, sample collection and analysis including linked-anonymous HIV testing, feedback of results, consent for DNA extraction and genomic analyses, consent for use of data and samples for future research, and permission to re-approach participants for new studies. Study progress {#sec6} ============== Data collection began in November 2013. Through successful community sensitisation, engagement and mobilisation, the study has achieved a high level of participation: to date, 1300 individuals have been invited join the study, of whom 96 refused participation, with a resulting response of 92.6% (*n* = 1204) ([Table 2](#tab02){ref-type="table"}). The age and sex structure of this DDS intermediate dataset are compared with the 2011 census data for the eThekwini population in [Fig. 2](#fig02){ref-type="fig"} \[[@ref18]\]. Of the 1204 participants recruited to date, 28.3% (341) are men compared with 48.2% found in the general population census data. This low proportion of men in the DDS is consistently observed in population-based studies in South Africa. Likely explanations include high levels of unemployment in the townships sampled leading to men moving away for work (migrant labour system) \[[@ref21], [@ref22]\], and challenges in participant mobilisation as male potential participants are often away during office hours when the study mobilisation takes place \[[@ref11]--[@ref14]\]. The age distribution of the DDS intermediate dataset is broadly similar to the general population of eThekwini: 46.6% were aged 18--34 years in the DDS compared with 49.1% in the 2011 eThekwini census; 44.1% (DDS) compared with 43.6% (census) were aged 35--64 years; and 9.2% (DDS) compared with 7.3% (census) were aged 65 years or older. The employment status of DDS participants for the preceding 12 months is shown in [Fig. 3](#fig03){ref-type="fig"}: 37.2% were in full- or part-time employment, 13.9% were retired, and 10.2% were students. Thirty-eight per cent of the DDS population reported being unemployed (including both individuals able and unable to work), compared with 30.2% in the 2011 eThekwini census. Fig. 2.Population pyramid comparing the age and sex structure of the DDS intermediate dataset (*n* = 1204) with the 2011 census population for the eThekwini municipality (city of Durban). DDS data for women (red bars) and men (blue bars) are superimposed over census data for women (black and white bars, hatched) and men (black and white bars, plain). Fig. 3.Proportion of participants in the DDS intermediate dataset (*n* = 1204) by category of employment in the preceding 12 months. Table 2.Demographic characteristics of participants in the DDS intermediate dataset (n = 1204)Variable*N* (%)Sex Men341 (28.3) Women863 (71.7)Age category (years) 18--24284 (23.6) 25--34277 (23.0) 35--44189 (15.7) 45--54198 (16.4) 55--64144 (12.0) 65+111 (9.2) Do not know1 (0.1)Language isiZulu1084 (90.0) isiXhosa79 (6.6) Sesotho28 (2.3) Other13 (1.1)Level of education Nil16 (1.3) Nursery or preschool17 (1.4) Primary246 (20.4) Secondary848 (70.5) Post-secondary/vocational19 (1.6) University52 (4.3) Do not know6 (0.5) Strengths and limitations {#sec7} ========================= The DDS is a platform for population-based research encompassing the distribution, interrelation and aetiology of chronic diseases and their risk factors in an urban South African setting, providing a valuable research resource and creating an evidence base to inform local and national health policy and public health programmes. The study is designed to characterise a wide range of phenotypes -- including banked biological samples for future analyses -- and to align these with genomic data. In this way, the DDS has the scope to address a broad set of research questions and the ability to provide aetiological insights into the variation in risk factors for diabetes and other chronic diseases among adults in this setting. The study has no upper age limit for participants and is therefore well placed to investigate disease and risk factors in older populations \[[@ref23]\]. The support of community leaders and local residents, cultivated through a careful and involved dialogue with these communities, enables the recruitment of informed and confident participants who are willing to enter a survey containing sensitive questions and encourages a high proportion of responders. A limitation of the DDS is its cross-sectional design; it is therefore not possible to assess incidence in this study or the temporal association between exposures and outcomes. However, the DDS has obtained consent to follow up and re-approach participants, thus enabling the potential for longitudinal studies, prospective case ascertainment of incident disease, and interventional studies. The data collection framework of the DDS has been standardised and aligned to the data collection framework of both the APCDR and the pan-African Human Heredity and Health in Africa (H3A) diabetes study \[[@ref24]\]. This will allow future comparison with other SSA countries, including pooled analyses of equivalent data from existing studies (as shown in online Supplementary Table S1) to better characterise the epidemiology of diabetes in SSA. The study specifically recruits participants of African descent in order to better assess the association between exposures and disease in this population \[[@ref25], [@ref26]\]. By focusing on a single ethnic group, the DDS may limit its generalisability to other populations; this highlights the importance of conducting similar studies nationally and across SSA to fully understand the aetiology and determinants of chronic disease. However, the intermediate dataset presented here indicates that the DDS population is broadly representative of its target population. Future plans {#sec8} ============ The DDS continues to recruit participants towards the sample size of 10 000 participants, with sufficient statistical resolution to present accurate prevalence estimates and identify and assess epidemiological associations between risk factors and disease that are informative for policy-makers and public health programmes in the region. Given the planned sample size, the DDS will be powered to detect less prevalent exposures, and distinguish more subtle associations between risk factors and disease. The final sample size will also have power to identify subpopulations at high risk of chronic diseases, whilst acknowledging all the caveats of undertaking such subgroup analyses. Additionally, with the option to re-approach participants, the DDS has the potential as a platform for future research including studies of diabetes complications and intervention studies. A central aim of the DDS is to combine epidemiological methods with population-based genome-wide technologies. This integration has already provided new insights into the biology of chronic diseases and their risk factors in Western populations \[[@ref27]\]. However, the relevance of many recent genomic findings to populations in SSA is not known. Given the marked genomic diversity among populations in SSA, understanding the genomic basis for chronic diseases and their risk factors in populations of African descent is likely to provide additional insights into the genetic determinants of chronic diseases and potential therapeutic strategies \[[@ref8]\]. The DDS is thus well positioned to take advantage of these opportunities to conduct genomic analyses and contribute to the global genomics research arena. Data access {#sec9} =========== All data (health, lifestyle and socioeconomic questionnaire, biophysical measurements, NCD and infectious disease biomarkers and genetic) generated in the DDS are stored and curated as part of the APCDR. For data access inquiries you may contact the APCDR data access committee (<data@apcdr.org>). The study was supported by the Wellcome Trust (grant number 098051), the African Partnership for Chronic Disease Research (Medical Research Council UK partnership grant number MR/K013491/1), the National Institute for Health Research Cambridge Biomedical Research Centre (UK), the Gates Cambridge Scholarship programme (UK), Novo-Nordisk (South Africa), Sanofi-Aventis (South Africa), and MSD Pharmaceuticals (Pty) Ltd (Southern Africa). MSS, EHY and AAM conceived the idea for the study. MSS, EHY, FJP, JR, TME, BO and AAM designed the study. MSS, EHY, FJP, JR and AAM wrote the study protocol. TRH analysed the data and, with EHY, wrote the manuscript. FJP, TME, BO, MIM, MSS and AAM contributed to discussion, and reviewed and edited the manuscript. The authors would like to thank the study participants for their cooperation and express their gratitude to Nonhlanhla Nombula, the field coordinator and the field team staff. Supplementary material {#sec10} ====================== For supplementary material accompanying this paper visit http://dx.doi.org/10.1017/gheg.2015.3 ###### click here to view supplementary material No other potential conflicts of interest relevant to this article were reported. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. [^1]: These authors contributed equally to this work. [^2]: NCD, non-communicable disease; OGTT, oral glucose tolerance test; HbA~1c~, glycated haemoglobin; LDL, low-density lipoprotein; HDL, high-density lipoprotein; Lp(a), lipoprotein(a); HIV, human immunodeficiency virus; HBV, hepatitis B virus; HCV, hepatitis C virus; SNP, single-nucleotide polymorphism.

Dear Editor: Hailey-Hailey disease (HHD), also known as benign familial pemphigus, is a chronic recurrent bullous and vesicular dermatitis, presented histologically withsuprabasilar acantholysis[@B1]. Flexural sites, especially lateral side of neck, groin and axillae are affected. Patients are suffered by chronic erosions of the aforementioned sites, which is painful and commonly become infected secondarily[@B2]. Even many studied reported various medical treatments, such as topical and systemic corticosteroid, antifungals, and antibiotics, no radically curative therapy for HHD was found[@B3]. Good results for HHD have been obtained from such surgical interventions as excision, split-thickness skin graft, and dermabration[@B4]. However, there are few reports regarding the efficacy of cryotherapy for HHD. Here, we describe the successful treatment for recalcitrant HHD. A 33-year-old man clinically presented with signs and symptoms of HHD in the area of bilateral crural folds. HHD was diagnosed with biopsy. His signs and symptoms had been presented for two years, despite multiple standard treatment and anecdotal therapies: numerous topical corticosteroids, tetracycline, cyclosporin, isotretinoin, prednisone, topical pimecrolimus cream, and tacrolimus ointment. Surgical, laser, or radiation therapy had never been tried. As multiple therapies were either ineffective or associated with unacceptable side effects, cryotherapy and CO~2~ laser (UM-L30; Union Medical Co., Seoul, Korea) were initiated. The patient was started on cryotherapy (left side) and CO~2~ laser (right side) monthly for 3 months, after intense history taking and carefulphysical examination. We performed cryotherapy with two cycles using an open-spray technique at 10\~15 seconds and CO~2~ laser with (1.2 W in a continuous wave) one pass each session. Cryotherapy with the Cryosurg® spray gun (Coopersurgical, Trumbull, CT, USA) was done. The cryotherapy was perfomed perpendicular from the lesion at a 1\~2 cm distance. When the ice-ball had presented from the center to include the edge of the area with a 2 mm margin, cryotherapy was stopped. It was then allowed to melt. After each therapy, no extra dressing was done. Photographs of lesional skin were taken at baseline and week 12 ([Fig. 1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}). The area receiving cryotherapy was disease free and asymptomatic 3 months later. The region subjected to CO~2~ laser treatment showed mild to moderate response, but less than the area receiving cryotherapy. A variety of measures have been recommended for recalcitrant HHD. Cryotherapy has been advocated although freezing may induce lesions[@B5]. Necrosis, after the freezing and thawing of cells is the mechanism of cryotherapy. Re-epithelialize in treated areas occured. The HDD lesion was shown to have altered cellular connections within desmosomes and adherens junctions of the epidermis. The mechanism of cryotherapy for treatment of HHD may eliminate diseased epidermis and re-epithelialize normal epidermis. Our patient presented with recalcitrant HHD. In management, we found, for the condition of patient\'s inguinal area, cryotherapy is more effective than was CO~2~ laser. Moreover, No short-term untoward side-effects was observed, such as hypopigmentation, after the use of cryotherapy as the treatment for HHD. During at least 3 months, patient went through relatively disease-free remission. These observations represent the successful use of cryotherapy for HHD. This work was supported by the National Research Foundation of Korea (NRF), funded by the Ministry of Science, ICT & Future Planning (NRF-2014R1A1A3A040-49491), and the Hallym University Research Fund 2014 (HURF-2014-53, HURF-2014-58). {#F1} {#F2}

1. Introduction {#sec1-polymers-11-01937} =============== The use of naturally renewable biopolymers (e.g., starch, cellulose, chitosan/chitin, caseinate, and pectins) as edible and biodegradable films and coatings constitute an actual and sustainable choice to prevent the post-harvest losses of very perishable fruits, such as blueberries. The production of blueberries (family Ericaceae, genus *Vaccinium*) has been rapidly growing in the last years, reaching over 596,813 t worldwide in 2017, being led by the United States with 40% of total production and Canada with 27% \[[@B1-polymers-11-01937]\]. Highbush blueberries (*Vaccinium corymbosum*) are the most commercial blueberries. However, several other wild shrubs of the genus *Vaccinium* also produce commonly eaten blueberries, such as the European *Vaccinium myrtillus*. *Vaccinium meridionale* Swartz (Andean blueberry) is one of the species of the genus *Vaccinium* that grows in the Andean region of South America at 2300--3300 m above sea level (m.a.s.l.) \[[@B2-polymers-11-01937]\]. The fruit is a berry, 5--10 mm in diameter, dark reddish, with a sour and tart character. In Colombia, Andean blueberry is sold at the local markets, and it is eaten fresh, dried, and cooked into sauces, jellies, and jams \[[@B3-polymers-11-01937]\]. *Vaccinium meridionale* Swartz is considered to be health-promoting food due to its high content of anthocyanins (\~329 ± 28 mg/100 g, mainly cyanidin 3-glucoside), and phenolic acids (\~ 758.6 ± 62.3 mg gallic acid equivalent/100 g, mainly chlorogenic acid) \[[@B3-polymers-11-01937]\]. Several studies have shown than Andean blueberry has antioxidant, anticarcinogenic, and anti-inflammatory properties \[[@B3-polymers-11-01937],[@B4-polymers-11-01937],[@B5-polymers-11-01937]\]. Therefore, *Vaccinium meridionale* Swartz has a high potential of use as an antioxidant additive or functional ingredient in food, pharmaceutics, and cosmetic applications \[[@B6-polymers-11-01937]\]. As for most blueberries fruit, Andean blueberries are perishable fruits, and, therefore, it is necessary to develop strategies to increase their storage life \[[@B7-polymers-11-01937]\]. It has been reported that the fruit decays quickly at ambient temperature after harvest, thus affecting its commercialization \[[@B8-polymers-11-01937]\]. Moreover, blueberry fruits are susceptible to bruising from mechanical impact, water loss, and microbial attack \[[@B9-polymers-11-01937],[@B10-polymers-11-01937]\]. Coatings are defined as mixtures of film-forming materials plus solvents and other additives (e.g., plasticizers), which, when applied to a surface and cured or dried, yield a solid protective, decorative and/or functional adherent thin layer \[[@B11-polymers-11-01937]\]. Moreover, coatings can be added of antioxidants, antimicrobial agents, oxygen scavengers, or moisture absorbing, among others, to obtain active coatings \[[@B12-polymers-11-01937],[@B13-polymers-11-01937]\]. Over the past years, the use of coatings has become more and more important in the food field. It has been reported that the application of coatings allows extending the shelf life of perishables and sensitive products such as fruits and vegetables because these materials act as an external protective layer slowing the respiration rate and reducing moisture and solute migration, gas exchange, oxidative reaction rates, as well as suppress physiological disorders of fresh-cut fruits \[[@B12-polymers-11-01937],[@B14-polymers-11-01937],[@B15-polymers-11-01937]\]. Several polymers from natural origin have been used for the fabrication of food edible coatings, including starch \[[@B16-polymers-11-01937]\], alginate \[[@B17-polymers-11-01937]\], chitosan \[[@B18-polymers-11-01937]\], pectin \[[@B17-polymers-11-01937]\] and cellulose derivatives \[[@B19-polymers-11-01937]\], among others. Between them, starches are well known for their good film-forming properties and functionalities \[[@B20-polymers-11-01937]\]. Moreover, it has been stated that because the edible films and coatings based on starches are reported to be transparent, odorless, tasteless, and colorless, they would not exhibit adverse effects on the sensory quality of a product \[[@B21-polymers-11-01937],[@B22-polymers-11-01937],[@B23-polymers-11-01937]\]. Despite that starches from diverse sources have been widely used as edible coating material and there is a tremendous amount of knowledge on composition, structure, properties, and modifications \[[@B24-polymers-11-01937],[@B25-polymers-11-01937],[@B26-polymers-11-01937]\], starches from Colombian native potatoes used in this research have not been studied until now. It is commonly hypothesized that starch granules architecture varies widely between different botanical sources, and so has a significant impact on its physicochemical properties \[[@B24-polymers-11-01937],[@B25-polymers-11-01937],[@B26-polymers-11-01937]\] In the current work, starches were isolated from three different Colombian potatoes varieties (mora, pacha negra, and alcarrosa), and their physicochemical properties were investigated and compared. Edible coatings based on the different starches were applied on Andean blueberries, and the changes in fruit physicochemical parameters were monitored during 12 days of storage. As far as the authors are aware, this is the first time that starches isolated from Colombia native potatoes have been characterized in detail and applied as edible coatings on Andean blueberry. 2. Materials and Methods {#sec2-polymers-11-01937} ======================== 2.1. Materials {#sec2dot1-polymers-11-01937} -------------- Starches were isolated from three varieties of native potatoes (*Solanum tuberosum*), grown in Ventaquemada (Boyacá, Colombia) at 2630 (m.a.s.l.) ([Figure 1](#polymers-11-01937-f001){ref-type="fig"}). Andean berries (*Vaccinium meridionale* Swartz) at maturity stage 4 (100% purple) and dry content (%) of 23.58 ± 1.90 were obtained from a local supermarket (Duitama, Colombia). The berries were examined previous to its use to separate fruits with physical, mechanical, or microbial damages. All chemicals used were of analytical grade. Dimethyl Sulfoxide was purchased from Sintorgan^®^ (Buenos Aires, Argentina). Urea was purchased from YPF (Buenos Aires, Argentina). Iodine and Potassium iodide were purchased from Anedra (Buenos Aires, Argentina). Ethanol 95.5% was purchased from Merck^®^ (Darmstadt, Germany). Glycerol was purchased from J. T. Baker (Phillipsburg, NJ, USA). Sodium hydroxide was purchased from Sigma Aldrich (St. Louis, MO, USA). 2.2. Isolation and Characterization of Starches {#sec2dot2-polymers-11-01937} ----------------------------------------------- ### 2.2.1. Starches Isolation and Yield {#sec2dot2dot1-polymers-11-01937} Starches isolation was carried out from each potato cultivar according to the optimized protocol by Doporto, Dini, Mugridge, Viña, & García, 2012 \[[@B27-polymers-11-01937]\]. Briefly, potatoes were washed, sanitized (250 ppm of chlorine, 10 min), peeled and pulped with a grater. The grated potatoes were blended with water (2 L water/kg) and stored at 4 °C for 24 h. The blend was filtered using a cheesecloth, and the starch slurry was decanted at 4 °C. The supernatant was discarded, and the starch cake was recovered, dried at 40 °C for 14 h in a hot-air oven, and milled. The starch yield was calculated as the percentage of dry weight (dehydrated at 40 °C) of isolated starch related to the fresh weight of the potato tuber. ### 2.2.2. Amylose Content {#sec2dot2dot2-polymers-11-01937} Amylose content of starches was determined by the iodine binding method \[[@B28-polymers-11-01937]\]. An amount of each potato starch was weighted (*m~starch~*) and mixed with 10 mL of urea-DMSO solution (1:9 v/v). The blend was heated in an oven at 100 °C for 1 h and cooled to room temperature. Then, 0.5 mL of the prepared solution were weighed (*m~solution~*) and diluted in 5 mL of ethanol. The solution was centrifuged (5000 rpm, 30 min) and the supernatant was discarded. Urea-DMSO solution (1 mL), I~2~/IK solution (1 mL) and distilled water (to a final volume of 50 mL) were incorporated. Absorbance was measured at 635 nm, and amylose content was calculated according to the following equation:$$Amylose\ content\ \left( \% \right) = 28.414 \times \frac{A_{635}}{2 \times m_{starch} \times m_{solution}} \times 100$$ where *m~starch~* is the amount of each potato starch used in the assay, *m~solution~* is the weight of the solution, and *A*~635~ is the absorbance measured at 635 nm. ### 2.2.3. Size and Morphology of Starch Granules {#sec2dot2dot3-polymers-11-01937} The morphology of potato starch granules was examined by scanning electron microscopy using a field emission gun (FE-SEM) (SUPRA 40, Carl Zeiss NTS, Oberkochen, Germany). Starch granules were attached to aluminum stubs using a two-sided adhesive tape and sputtered with a thin layer of platinum before observation. Size distributions of granules were determined using a Laser Diffraction Particle Size Analyzer SALD-3101 (LS, Shimadzu, Kyoto, Japan). For this purpose, starches were dispersed in ethanol (96% *v/v*) and diluted until an absorbance between 0.12 and 0.20 was obtained. ### 2.2.4. Infrared Spectra {#sec2dot2dot4-polymers-11-01937} FTIR analysis was performed in a Jasco FT-IR 4100 spectrometer (Hachioji, Tokyo, Japan) equipped with attenuated total reflectance (ATR) module. Samples were placed on the ATR accessory, and they were analyzed, taking 64 scans per experiment with a resolution of 4 cm^−1^. ### 2.2.5. X-ray Diffraction Patterns {#sec2dot2dot5-polymers-11-01937} X-ray diffractograms of the starch granules were recorded at 2θ between 5° and 50° with a step size of 0.02° in a Philips-APD PW 1710 diffractometer (Eindhoven, North Brabant, The Netherlands) using Cu-Kα radiation (λ = 1.542 A). The crystalline fractions were calculated as reported Lopez-Rubio, Flanagan, Gilbert, & Gidley \[[@B29-polymers-11-01937]\]. Patterns were fitted with an amorphous background and several crystalline peaks using Gaussian functions. The fitting coefficients were refined to minimize the value of Chi-squared using the Levenberg-Marquardt algorithm. The crystalline fraction was calculated according to Equation (2):$$X_{c} = \frac{\sum^{\ }Ac_{i}}{A_{t}}$$ where $X_{c}$ corresponds to the crystalline fraction, $Ac_{i}$ to the area of each crystalline peak, and $A_{t}$ to the total area. The fitting procedure was repeated five times with different initial inputs. ### 2.2.6. Thermal Properties {#sec2dot2dot6-polymers-11-01937} The temperature and the enthalpy of gelatinization of the starch granules were determined by differential scanning calorimetry (DSC) (Mettler Toledo, Schwerzenbach, Switzerland). The starch samples (\~2.5 mg) were weighed into aluminum DSC pans, to which 7.5 mg of deionized water was added. After sealing, the pans were heated from room temperature to 90 °C with at a heating rate of 10 °C /min, under nitrogen atmosphere (30 mL/min). An empty aluminum pan was used as a reference. The onset (*T*~on~), peak (*T*~p~), conclusion (*T*~c~) temperatures, and gelatinization enthalpies (Δ*H*) were estimated using the TA Instrument universal analysis software. Simultaneous Thermogravimetric/Differential thermal analyses (TGA/DTA) were performed in a Shimadzu DTG-60 (Kyoto, Japan). Dry samples (\~5 mg) were placed in aluminum pans inside the thermogravimetric balance and then heated under a nitrogen atmosphere (20 mL/min) in a range of 30 to 450 °C at a heating rate of 10 °C/min. TGA curves were derived with respect to temperature to determine the degradation temperatures. 2.3. Preparation and Application of Starch Edible Coatings {#sec2dot3-polymers-11-01937} ---------------------------------------------------------- The preparation of edible starch coatings was carried out as reported in a previous work \[[@B30-polymers-11-01937]\]. Blends of each potato starch (2.0 g), glycerol (0.6 g) and distilled water (97.4 g) were homogenized for 40 min and then heated to 92 °C, under constant stirring, until complete starch gelatinization. The formulations were cooled to room temperature for later application on the fruits. A total mass of 6 kg of blueberries was randomly divided into four groups (control, pacha negra starch edible coating, mora starch edible coating, alcarrosa starch edible coating); each group containing 1.5 kg of Andean blueberries. The fruits were dip-coated by immersion in the coating solutions for 90 s, drained of excess coating, and air-dried at room temperature. Control samples (without coating) were also prepared by immersing of fruits in distilled water and kept under the same storage conditions than treated ones, for comparison. 2.4. Evaluation of Quality Attributes of Andean Blueberries along Storage {#sec2dot4-polymers-11-01937} ------------------------------------------------------------------------- Coated and uncoated Andean blueberries were packed in PET trays with perforated vents and stored for 12 days. Evaluations of quality attributes were performed at 1, 5, 9, and 12 days of storage at room conditions (14 °C and 80% RH). For every sampling time, three trays containing 125 g (\~250 units) of Andean blueberries each were prepared. ### 2.4.1. Respiration Rate {#sec2dot4dot1-polymers-11-01937} Respiration rate was measured as reported Hasperué, Rodoni, Guardianelli, Chaves, & Martínez, 2016 \[[@B31-polymers-11-01937]\]. Approximately 120 g of Andean blueberries were placed for 30 min at 25 °C inside hermetically sealed 2 L flasks. Then, the CO~2~ concentration was determined using an infrared analyzer (LabQuest^®^2 Model LQ2-LE, Beaverton, OR, USA), and the results were expressed as the rate of respiration (CO~2~) in mg kg^−1^ s^−1^. ### 2.4.2. Soluble Solids Content, pH and Titratable Acidity (%) {#sec2dot4dot2-polymers-11-01937} The soluble solids content was measured in the fruit juice using an Atago refractometer model PR 101 (Atago CO., Tokyo, Japan) and expressed as °Brix (AOAC 932.12). Fruit samples were crushed using a blender and filtered through filter paper to obtain the fruit juice. The pH of the fruit samples was assessed using a digital pH meter (Oakton Instruments, Vernon Hills, IL, USA) (AOAC 981.12). Titratable acidity (%) was determined by titration with 0.1 N NaOH up to pH 8.1, using 1 g of sample in 10 mL of distilled water (AOAC 942.15). The results were expressed in citric acid percentage. ### 2.4.3. Weight Loss {#sec2dot4dot3-polymers-11-01937} Weight loss of Andean blueberries during storage was determined by weighting all fruit trays at the beginning of the storage and every day of analysis. The weight loss (% *W*) was calculated with the following equation:$${\%\ }W = \ \left( \frac{m_{0} - m_{f}}{m_{0}} \right)x\ 100$$ where *m~f~* is the weight at each time and *m*~0~ the initial weight of each sample ### 2.4.4. Firmness Analysis {#sec2dot4dot4-polymers-11-01937} Firmness was determined using a digital Force Gauge PCE-FM200 (Southampton, UK) equipped with a 6 mm diameter stainless steel probe. Firmness was defined as the maximum force to disrupt the tissue at the penetration time used (5 s) \[[@B32-polymers-11-01937]\]. The results were expressed as an average of at least five measurements. 2.5. Statistically Analysis {#sec2dot5-polymers-11-01937} --------------------------- The statistical analysis was performed using the Systat Inc. software (Evanston, IL, USA). Analysis of variance (ANOVA) and Tukey pairwise comparisons were carried out using a level of 95% confidence. The experiments were performed at least in duplicate, and the data were reported as mean ± standard deviation. 3. Results and Discussion {#sec3-polymers-11-01937} ========================= 3.1. Characterization of Starches {#sec3dot1-polymers-11-01937} --------------------------------- [Table 1](#polymers-11-01937-t001){ref-type="table"} shows the characteristics of the starches isolated from the Colombian native potatoes. All three starches had similar isolation yield (\~11%). It was recently reported isolation yields (%) ranging between 7.0% and 16.5% when worked with yellow skin potatoes for extraction of starch \[[@B33-polymers-11-01937]\]. Concerning the amylose content (%), alcarrosa starches showed higher amylose content than the mora and pacha negra starches, even though these differences were not statistically significant (*p* \> 0.05) ([Table 1](#polymers-11-01937-t001){ref-type="table"}). Amylose/amylopectin ratio is a leading parameter since it determines the physicochemical properties of starches, such as gelatinization energy and crystalline structure \[[@B34-polymers-11-01937]\]. Besides, the properties of starch coatings are also affected by the amylose/amylopectin ratio. Generally, starch films with higher amylose content present a better barrier to oxygen \[[@B35-polymers-11-01937]\]. According to the literature, the amylose content of potato starch generally varies between 18% and 30% \[[@B25-polymers-11-01937],[@B34-polymers-11-01937]\], indicating that the potato varieties used in this work are of relative low amylose content. Size and morphology of starch granules depend strongly on the botanical source. Besides, these parameters affect the properties of the corresponding starch gel directly; therefore, the different isolated starches were characterized by means of their granule sizes and morphology. [Figure 2](#polymers-11-01937-f002){ref-type="fig"} presents SEM images and size distributions of the different samples, and the mean diameters of the granules are reported in [Table 1](#polymers-11-01937-t001){ref-type="table"}. All the starches showed an average granule size of \~47 µm without significant differences between them. Besides, a wide distribution of granule size ([Figure 2](#polymers-11-01937-f002){ref-type="fig"}) was observed in all systems. Starch granules resulted mostly ellipsoidal, but some small granules were circled shaped, and some large granules were irregularly shaped. Size and morphology of granules resulted similarly to those reported in the literature for potato starches from different varieties \[[@B36-polymers-11-01937],[@B37-polymers-11-01937],[@B38-polymers-11-01937]\]. [Figure 3](#polymers-11-01937-f003){ref-type="fig"}a shows the X-ray diffraction patterns of the studied samples, and the percentages of crystallinity are reported in [Table 1](#polymers-11-01937-t001){ref-type="table"}. Similar crystalline fractions were obtained for all the studied samples (\~45%), and the obtained values were in agreement with the literature (15--45%) \[[@B39-polymers-11-01937],[@B40-polymers-11-01937]\]. Starches showed a semi-crystalline structure with diffraction peaks centered at 5.7°, 14.7°, 17.1°, 19.7°, 22.1°, 24.0°, 26.6°, 30.6°, and 34.7°. These patterns correspond to a B-type structure, which is characteristic of potato starch \[[@B41-polymers-11-01937]\] and consists of double helices of amylose packed with a P61 space group in a hexagonal unit cell \[[@B42-polymers-11-01937]\]. Some differences could be observed in the obtained diffraction patterns. The intensity of the crystalline peaks centered at 5.7° and 26.6° was slightly higher in mora starch, while the intensity of the peaks at 14.7° and 19.7° was slightly higher in alcarrosa starch. Besides, alcarrosa starch showed an additional peak at 23.1°. These differences indicate that polymer packing is different, which could be attributed to different molecular weight distributions of amylose and amylopectin in each potato starch variety. FTIR constitutes a standard method to study chemical interactions in polymers and soft matter. Mainly it is useful to characterize starch granules concerning its short-range structure. FTIR spectrum of the studied systems showed the characteristic peaks of starch: O--H stretching (3300 cm^−1^), C--H symmetric and asymmetric stretching (2930 and 2890 cm^−1^), C--O stretching (1149 cm^−1^), C--C stretching (1077 cm^−1^), and C--OH bending (1100--900 cm^−1^) ([Figure 3](#polymers-11-01937-f003){ref-type="fig"}b) \[[@B43-polymers-11-01937]\]. No additional bands were observed in any sample, which suggests that the isolation procedure was effective. Besides, no significant differences were observed between the spectra of the different potatoes starches, indicating that interactions between polymer chains are similar for all systems. In particular, the bands in the region 1100--900 cm^−1^ are highly sensitive to changes in the short-range structure. Since there were no differences in the relative intensity of the bands in this region, it can be concluded that the short-range structure is similar for all the studied starch granules. [Figure 4](#polymers-11-01937-f004){ref-type="fig"} shows the DSC thermograms of the starches isolated from the different Colombian potato cultivars. All samples showed similar temperatures (*T*~on~ \~ 58 °C; *T*~p~ \~ 62.8 °C; *T*~c~ \~ 70.1 °C) and enthalpies (∆*H* \~ 12 J/g) of gelatinization. Martínez et al., 2019 reported similar thermal gelatinization properties when worked with starches isolated from native potatoes of the Andean region (Imilla blanca, Imilla negra and, Loc'ka). On the other hand, the thermal degradation of all samples occurred in two steps. The first event occurred between 50 and 150 °C and corresponds to moisture loss, while the second event occurred between 250 and 342 °C, with the maximum of the peak at 305 °C, and corresponds to thermal decomposition of the starch polymer \[[@B44-polymers-11-01937]\]. No significant differences were detected between the samples (TGA curves not shown). [Figure 5](#polymers-11-01937-f005){ref-type="fig"} presents the DTA curves of the different starch granules. In all samples, four endothermic processes were observed. The first event occurred between 30 and 160 °C and corresponded to the melting of crystallites \[[@B45-polymers-11-01937]\]. The rest of the thermal events occurred between 270 and 320 °C and were associated with the thermal degradation of starch (according to TGA results). Generally, the thermal degradation of starch reveals two endothermic peaks corresponding to the degradation of amylose (T \~ 275 °C) and amylopectin (T \> 300 °C) \[[@B46-polymers-11-01937]\]. However, in this work, the degradation of starch consisted of three endothermic peaks (insert in [Figure 5](#polymers-11-01937-f005){ref-type="fig"}). The thermal event at higher temperatures was associated with amylopectin degradation, and it was centered at 308 °C for pacha negra and alcarrosa starches, and at 305 °C for mora starch. Then, the peaks at lower temperatures were associated with the degradation of amylose fractions with different molecular weights. Tong et al. \[[@B25-polymers-11-01937]\] recently characterized different pigmented potatoes from China and reported a binomial distribution of amylopectin molecular weight. The endothermic peaks observed in this work were centered at 283 °C and 290 °C for pacha negra starch, at 279 °C and 286 °C for mora starch, and at 283 °C and 288 °C for alcarrosa starch. Both amylose and amylopectin degradation temperatures were found at lower temperatures for the samples corresponding to mora starch. 3.2. Effect of Edible Coating on the Visual Appearance of Andean Blueberry {#sec3dot2-polymers-11-01937} -------------------------------------------------------------------------- [Figure 6](#polymers-11-01937-f006){ref-type="fig"} shows images of Andean blueberries without and with edible starch coatings. It can be seen that the coatings-forming solutions form a good film on the surface of the Andean blueberries, giving the fruit a brighter, translucent, fresh-like appearance, compared to the control fruit. It has been reported that edible coatings have useful effects on the bright appearance of the fruits due to that the coating forms a smoother surface compared to the fruit skin. In this sense, a surface with less roughness would lead to a greater reflection of visible light \[[@B47-polymers-11-01937]\]. 3.3. Changes in Andean Blueberry Quality Parameters during Storage {#sec3dot3-polymers-11-01937} ------------------------------------------------------------------ [Figure 7](#polymers-11-01937-f007){ref-type="fig"}A shows the changes in the respiration rate (%) of uncoated and coated Andean blueberries along storage. In general, the respiration rate of the fruits was decreased by the application of the edible coatings based on all potato starches studied. At the end of the storage, the respiration rate of the coated Andean blueberries was \~27% lower compared to the uncoated fruits (*p* \< 0.05), regardless of the starch type used. It has been reported that polysaccharide coatings exhibit a high barrier to oxygen, significantly limiting respiration of the fruit \[[@B48-polymers-11-01937],[@B49-polymers-11-01937],[@B50-polymers-11-01937]\]. Control fruits showed a significant increase in their respiration rate after 5 days of storage (*p* \< 0.05); while in the samples with edible coatings based on mora and alcarrosa starches, significant increases were not found along the assay (*p* \< 0.05). The samples coated with pacha negra starch showed a significant descent in the respiration rate at 9 days of storage and then an increase at the end of storage. This behavior was probably due to that with the passing of the days of storage, the oxygen permeability of the pacha negra edible coatings may have been improved. However, at the end of storage, the water loss of the fruit can be slightly affected the integrity of the edible coating, and therefore their efficiency facing the gaseous exchange was decreased. For what concern the titratable acidity (%), uncoated Andean blueberries decreased their acidity (%) as from 5 days of storage, in relation to their initial value ([Figure 7](#polymers-11-01937-f007){ref-type="fig"}B). Similar behavior was observed for the samples with edible coatings based on mora and pacha negra starches ([Figure 7](#polymers-11-01937-f007){ref-type="fig"}B). A decrease in the titratable acidity (%) of Andean blueberries stored at different temperatures was reported recently \[[@B8-polymers-11-01937]\]. It has been well documented that organic acids are primary substrates involved in the respiration process of climacteric fruits; thus, a reduction in acidity is commonly expected during fruit ripening \[[@B51-polymers-11-01937],[@B52-polymers-11-01937]\]. Unlike the other starches, the edible coatings based on alcarrosa starch produced a significant increase in the titratable acidity (%) of Andean blueberries as from 5 days of storage ([Figure 7](#polymers-11-01937-f007){ref-type="fig"}B). This behavior was probably due to that the alcarrosa starch coatings could have acted delaying the utilization of organic acids as respiration substrate and providing higher rigidity to the skin of the fruit preventing the loss of the organic acids (e.g., citric acid). Similar observations have been reported by other authors \[[@B53-polymers-11-01937],[@B54-polymers-11-01937]\]. The differences in the effect of alcarrosa starch edible coating on the changes in the titratable acidy during storage can also be associated with the slight differences in the amylose content of the starches ([Table 1](#polymers-11-01937-t001){ref-type="table"}). Concerning the pH, no relevant differences were detected among the uncoated and the differently coated samples at each considered storage time ([Figure 7](#polymers-11-01937-f007){ref-type="fig"}C). On the other hand, Andean blueberries with and without edible starch coatings showed a gradual increase in their soluble solids content during storage ([Figure 7](#polymers-11-01937-f007){ref-type="fig"}D). After 5 days of the assay, statistically significant differences were found between the samples with and without starch coatings (*p* \< 0.05). At the end of the storage, the soluble solids content of the coated Andean blueberries was 14% higher compared to the uncoated fruits (*p* \< 0.05), and no significant differences were found between the different starches used. Generally, the concentration of soluble solids is associated with dehydration or weight loss. A more significant weight loss corresponds to a higher concentration of soluble solids. Moreover, it has been stated that quick respiration rates, in turn, accelerated the synthesis and use of metabolites that result in increasing the soluble solid concentration \[[@B55-polymers-11-01937]\]. However, in the present work, a non-conventional behavior was observed; the coated fruits showed lower respiration rates (%) and similar weight loss (%) than the uncoated fruits ([Figure 7](#polymers-11-01937-f007){ref-type="fig"}A and [Figure 8](#polymers-11-01937-f008){ref-type="fig"}A). Therefore, it was hypothesized that edible coatings based on all the starches studied were able to act as an effective barrier avoiding the loss of soluble solids. [Figure 8](#polymers-11-01937-f008){ref-type="fig"}A shows the changes in the water loss of uncoated and coated Andean blueberries along storage. All fruits showed a progressive increase of this water loss (%) until reach values of around 20% at the end of the storage, regardless of the presence of edible coatings. The loss of weight in fresh blueberries is mainly due to the loss of water caused by transpiration and respiration processes, which is determined by the gradient of water vapor pressure between the fruit and the surrounding air \[[@B56-polymers-11-01937]\]. The water loss was apparently higher for the uncoated samples than for the coated fruits; however, there were no significant differences between all samples (*p* \< 0.05). It has been reported that water loss depends on several factors such as fruit surface area/volume, fruit integrity, storage conditions (relative humidity and temperature), among others \[[@B53-polymers-11-01937],[@B54-polymers-11-01937],[@B55-polymers-11-01937],[@B56-polymers-11-01937],[@B57-polymers-11-01937]\]. Rincón Soledad et al. reported water loss (%) for the Andean blueberry of 37.3% after 30 days of storage at 20 °C. Moreover, they stated that water loss higher than the 10% provoked changes in the freshness of the fruit \[[@B8-polymers-11-01937]\]. The use of edible coating has been reported as an alternative to control the water loss of fruits. However, it has been stated that edible coatings based on natural polymers such as starches and chitosan did not act as an effective barrier against the weight loss \[[@B56-polymers-11-01937],[@B57-polymers-11-01937],[@B58-polymers-11-01937]\]. At the beginning of storage, the firmness of all coated Andean berries was 41% higher compared to the uncoated fruits (*p* \< 0.05). With the passing of 5 days of storage, the increase in the firmness of the coated samples was higher (\~58%), with respect to the control samples (*p* \< 0.05) ([Figure 8](#polymers-11-01937-f008){ref-type="fig"}B). At 9 day of storage, all samples showed similar firmness (*p* \< 0.05). Among all treatments, Andean berries coated with alcarrosa starches were more firm, and the fruits coated with mora starches showed lower firmness at the end of storage. Mannozzi et al. observed an increase in the firmness of blueberries immediately after coating with chitosan, and this was attributed to that the coating provides rigidity to the skin of the fruit \[[@B18-polymers-11-01937]\]. In the uncoated fruits and the samples with alcarrosa potato starch coating, their firmness increased around 75% and 58%, respectively, at the end of storage. It has been stated that water loss leads to increased firmness during post-harvest storage \[[@B7-polymers-11-01937],[@B8-polymers-11-01937],[@B9-polymers-11-01937],[@B10-polymers-11-01937],[@B11-polymers-11-01937],[@B12-polymers-11-01937],[@B13-polymers-11-01937],[@B14-polymers-11-01937],[@B15-polymers-11-01937],[@B16-polymers-11-01937],[@B17-polymers-11-01937],[@B18-polymers-11-01937],[@B19-polymers-11-01937],[@B20-polymers-11-01937],[@B21-polymers-11-01937],[@B22-polymers-11-01937],[@B23-polymers-11-01937],[@B24-polymers-11-01937],[@B25-polymers-11-01937],[@B26-polymers-11-01937],[@B27-polymers-11-01937],[@B28-polymers-11-01937],[@B29-polymers-11-01937],[@B30-polymers-11-01937],[@B31-polymers-11-01937],[@B32-polymers-11-01937],[@B33-polymers-11-01937],[@B34-polymers-11-01937],[@B35-polymers-11-01937],[@B36-polymers-11-01937],[@B37-polymers-11-01937],[@B38-polymers-11-01937],[@B39-polymers-11-01937],[@B40-polymers-11-01937],[@B41-polymers-11-01937],[@B42-polymers-11-01937],[@B43-polymers-11-01937],[@B44-polymers-11-01937],[@B45-polymers-11-01937],[@B46-polymers-11-01937],[@B47-polymers-11-01937],[@B48-polymers-11-01937],[@B49-polymers-11-01937],[@B50-polymers-11-01937],[@B51-polymers-11-01937],[@B52-polymers-11-01937],[@B53-polymers-11-01937],[@B54-polymers-11-01937],[@B55-polymers-11-01937],[@B56-polymers-11-01937],[@B57-polymers-11-01937],[@B58-polymers-11-01937],[@B59-polymers-11-01937]\]. Besides, mora and pacha negra potato starch coatings allow preserving the firmness of the blueberries with the passing of the storage days. It has been reported that the fruit firmness commonly decreased along storage due to the ethylene production, which encourages the synthesis of enzymes responsible for softening such as polygalacturonase \[[@B57-polymers-11-01937]\]. Therefore, it can be stated that mora and pacha negra potato starch coatings were able to maintain the firmness of the Andean blueberries along storage. This behavior is positive from a commercialization point of view because firmness is one of the factors that most influence the purchase intent of the consumers, who prefer fresh-like fruits. 4. Conclusions {#sec4-polymers-11-01937} ============== Colombian native potatoes constitute an important source of starches that can be potentially used in the development of several new food and non-food products, including edible coatings. The application of edible coatings based on starches isolated from Colombian native potatoes proved to be a useful and straightforward method to decrease the respiration rate and the loss of soluble solids of Andean blueberries during storage. Moreover, the coatings maintained the firmness of the fruits. However, the application of these edible coatings did not avoid the water loss of the fruits during storage, and, therefore, more studies are necessary to improve the water barrier properties of the coatings. The new edible coatings can be helpful for fruit farmers and agro-food enterprises interested in adding value to their products thought the improvement of the appearance and the increase of the shelf life. A.L.-C. would like to thank Julian Cardozo-Becerra and Roxana Puerto-Torres for assistance with the preliminary assays and the digital images of Colombia potato cultivars. Conceptualization was devised by A.L.-C. and S.G.; Methodology, Validation and Formal Analysis was carried out by C.M.-J., S.E.-A., S.G. and A.L.-C. Investigation, Resources, Data curation, Writing Original Draft Preparation and Writing---Review & Editing was performed by C.M.-J., S.E.-A., S.G. and A.L.-C.; Project administration and Funding acquisition S.G. and A.L.-C. This study was funded by Colciencias, the programa Colombia BIO, and the Gobernación de Boyacá through the Fondo de Ciencia, Tecnología e Innovación del Sistema General de Regalías, managed by the Fondo "Francisco José de Caldas" (project 61891. Conv. 794-2017). Furthermore, A.L.-C. would like to thank the Universidad Pedagógica y Tecnológica de Colombia (UPTC) for their financial support. S. G. and S. E.-A. would like to thank the Agencia Nacional de Promoción Científica y Tecnológica (ANPCyT PICT 2017-2362), to the Secretaría de Política Universitarias (SPU 2017--2018 nro 1655), and to Universidad de Buenos Aires (UBACyT 20020170100381BA) for their financial support. The authors declare no conflict of interest. {#polymers-11-01937-f001} {#polymers-11-01937-f002} {#polymers-11-01937-f003} {#polymers-11-01937-f004} {#polymers-11-01937-f005} {#polymers-11-01937-f006} {#polymers-11-01937-f007} {#polymers-11-01937-f008} polymers-11-01937-t001_Table 1 ###### Characteristics of the starches isolated from the Colombian native potatoes. Sample Extraction Yield (%) Amylose Content (%) Mean Granule Size (µm) Shape Description Crystallinity Fraction (%) -------------------- ---------------------- --------------------- ------------------------ ------------------- ---------------------------- Pacha negra starch 11.9 ± 1.5 ^a^ 17.92 ± 0.26 ^a^ 49.3 ± 1.7 ^a^ Ellipsoid 48.0 ± 1.6 ^a^ Mora starch 10.1 ± 1.3 ^a^ 18.27 ± 0.59 ^a^ 47.6 ± 1.5 ^a^ Ellipsoid 44.7 ± 2.2 ^a^ Alcarrosa starch 11.0 ± 1.2 ^a^ 19.58 ± 0.62 ^a^ 45.9 ± 1.7 ^a^ Ellipsoid 45.9 ± 3.9 ^a^ Different letters within the same column indicate statistically significant differences (*p* \< 0.05).

Introduction ============ The omohyoid muscle typically has an inferior belly originating from the superior border of the scapula, near the suprascapular notch, and occasionally, from the superior transverse scapular ligament \[[@B1]\]. This muscle then passes deep to the sternocleidomastoid where its superior belly passes almost vertically upward next to the lateral border of sternohyoid to attach to the inferior border of the body of the hyoid bone lateral to the insertion of sternohyoid \[[@B2]\]. Reported variants of the omohyoid muscle include absence of its superior belly, duplicated superior belly, coursing deep to the internal jugular vein, and existence as the variant cleido-hyoideus muscle \[[@B3]\]. As the omohyoid muscle is commonly used as a surgical landmark during neck dissections (e.g., identified as the surgical landmark for level III and IV lymph node metastases), knowledge of its variations is important to surgeons. Herein, we report an unusual variant of the omohyoid and sternohyoid muscles. Case Report =========== During the routine dissection of the neck in a fresh frozen cadaver head, fusion of the omohyoid and sternohyoid muscles into a single large sheet was found on the right side. The specimen was a Caucasian male cadaver whose age at death was 86 years old. The neck had been cut horizontally at the seventh cervical vertebra, so the inferior belly of the omohyoid muscle and attachment to the sternum of the sternohyoid muscle could not be observed. This large conjoined muscle was deep to the pretracheal fascia and ascended to attach onto the hyoid bone and thyroid cartilage ([Fig. 1A](#F1){ref-type="fig"}). Two white tendons were found at the inferolateral part of this muscle; the superior tendon represented the intermediate tendon of the omohyoid muscle and the inferior tendon and traveled anteriorly to cross the midline and attach onto the left sternohyoid muscle ([Fig. 1B](#F1){ref-type="fig"}). The muscle was innervated by branches of the cervical plexus and the ansa cervicalis was not observed. Deep to this abnormal muscle, the sternothyroid and thyrohyoid muscles were found and were normal in their morphology. The lateral edge of this fused infrahyoid muscle crossed the carotid sheath and its contents. On the left side, the infrahyoid muscles were normal, However, a slip from a left levator glandulae thyroidea muscle ascended and attached to the surface of the thyrohyoid muscle ([Fig. 2](#F2){ref-type="fig"}). As a cadaveric examination, the present study did not require approval by an ethics committee at our institutions, and the work was performed in accordance with the requirements of the Declaration of Helsinki (64th WMA General Assembly, Fortaleza, Brazil, October 2013). Discussion ========== In some studies, a doubled or duplicated superior belly of the omohyoid muscle has been reported \[[@B4]\]. In these reports, the inferior head of the superior belly of the omohyoid muscle was described as it fused to the sternohyoid muscle. Wood \[[@B5]\] reported duplication and triplication of the superior belly of this muscle and insertion onto the thyroid cartilage. The region between the sternohyoid and omohyoid has been reported to be filled with muscle slips instead of fascia \[[@B6]\]. However, no reports have shown the massive fusion between the omohyoid and sternohyoid muscles like the present case. Embryologically, Loth \[[@B7]\] considered the omohyoid muscle to be a remnant of the sternocleido-omohyoid muscle, which consists of two layers, a sternocleidohyoid portion and an omohyoid portion. Buntine \[[@B8]\] described this primitive sheet of the muscle as the episterno-cleido-hyoideus or sterno- omo-hyoid. Unfortunately, in the present case, the inferior part of the muscle could not be observed due to prior dissection. However, we speculate that this sheet might represent the sternocleido-omohyoid muscle, which had not undergone atrophy. The omohyoid muscle is used as a landmark to identify levels III and IV lymph node metastases of head and neck cancers \[[@B9]\] with the lateral border of the sternohyoid muscle defined as the anterior boundary of levels III and IV. In terms of surgical neck dissection, fusion such as reported in our case might mislead the surgeon and confuse the level classification of metastasis and thus alter postoperative therapy and prognosis by altering the anatomical landmarks used to localize lymph nodes in the region \[[@B10]\]. The omohyoid muscle is an important landmark for head and neck cancer treatment. Therefore, surgeons should be aware of its variations such as the one described in the current report. The authors wish to thank all those who donate their bodies and tissues for the advancement of education and research. {#F1} {#F2}

Background {#Sec1} ========== Acrodysostosis (ACRDYS) is a rare skeletal dysplasia with skeletal abnormalities and multi-hormonal resistance similar to PHP1A, but caused by mutations downstream of the genes involved in PHP1A. Whereas PHP1A results from mutations of the gene encoding the Gα-stimulatory subunit (*GNAS*), ACRDYS is caused by mutations involving the genes for downstream effector proteins of *GNAS* in the cyclic AMP (cAMP)/ protein kinase A (PKA) pathway. In ACRDYS1, the protein kinase A regulatory subunit 1A (*PRKAR1A*) is mutated and in ACRDYS2 the gene coding for phosphodiestarase type 4D (*PDE4D*) is mutated. *PRKAR1A* encodes the cAMP-dependent regulatory subunit of protein kinase A. The mutated subunit impairs the protein kinase A response to cAMP accounting for the hormonal resistance and skeletal abnormalities in ACRDYS1. *PDE4D* is also involved in the cAMP signaling pathway, encoding a class-IV cAMP-specific phosphodiesterase, regulating cAMP concentrations and leading to ACRDYS2 \[[@CR1]\]. The two subtypes of ACRDYS are both characterized by similar skeletal features (short stature, facial dysostosis, and brachydactyly with cone-shaped epiphyses) and variable multi-hormone resistance. Patients with ACRDYS1 tend to have less facial dysostosis, and normal intelligence with multi-hormone resistance. Patients with ACRDYS2 tend to have more characteristic facial dysostosis and intellectual disabilities but have only subtle or absent hormonal resistance \[[@CR1]\]. The G-protein coupled hormones which are commonly affected in ACRDYS1 include parathyroid hormone (PTH), thyroid-stimulating hormone (TSH), growth hormone-releasing hormone (GHRH), luteinizing hormone (LH), and follicle-stimulating hormone (FSH). Accepted standard therapy involves thyroid hormone replacement and calcitriol to normalize thyroid and calcium metabolism. Growth hormone (GH) therapy is an accepted treatment for PHP1A patients with short stature and GH deficiency \[[@CR2]\] but efficacy of GH therapy is unknown in ACRDYS1. We report a case of ACRDYS1 in an 8-year-old female with severe short stature treated empirically with GH which was complicated by LCPD. Case presentation {#Sec2} ================= An 8-year-old female was referred to pediatric endocrinology for short stature. She was born at 38 weeks and small for gestational age (SGA) - birth weight of 4 pounds 8 oz (2.04 kg), and birth length of 16 in. (40.6 cm). Her father was from Mexico and her mother was from Puerto Rico and their heights are 5′9″ (175.3 cm) and 5′5″ (165.1 cm) respectively giving her a mid-parental height of 5′4.5″ (163.8 cm). She had normal development, without any significant medical problems. Physical examination demonstrated severe short stature (height Z-score − 3.46), midface hypoplasia, and severe brachydactyly of her hands and feet. Laboratory analysis showed mildly elevated TSH with normal free thyroxine, mildly low calcium with elevated PTH and low normal IGF-1 level (Table [1](#Tab1){ref-type="table"}). Radiographs revealed severe brachymetatarsia, brachymetacarpia, and brachydactyly with cone-shaped epiphyses in the proximal phalanges (Fig. [1](#Fig1){ref-type="fig"}). Genetic testing revealed a mutation of the *PRKAR1A* gene, specifically R368X (CGA \> TGA). To address her TSH and PTH resistance, she was started on levothyroxine and calcitriol. We postulated that her severe short stature was from a combination of GHRH resistance and being born SGA. GH therapy was initiated at 0.35 mg/kg/week -- a mid-range dose for children with SGA. Three months later, her serum IGF-1 Z-score increased from − 1.2 to 2.1 at an effective GH dose of 0.33 mg/kg/week. Table 1Laboratory analysis upon initial presentation at age 8 yearsReference rangeResultsTSH0.6--5.5 μU/ml10.6Free T40.9--1.4 ng/dl1.0PTH11--74 pg/ml149Calcium8.9--10.4 mg/dl8.7Phosphorus3--6 mg/dl4.8IGF-176--424 ng/ml91 (z score − 1.2)IGF-BP31.6--6.5 mg/dl3.0Fig. 1Brachydactyly and cone shaped epiphyses seen in patient which is characteristic of ACRDYS After 7 months, her annualized growth velocity had increased from 3 cm/year to 11 cm/year and she had grown 6.5 cm (Fig. [2](#Fig2){ref-type="fig"}). However, she also reported right knee pain for 1 week. Radiographs (Fig. [3](#Fig3){ref-type="fig"}) revealed flattening of the right femoral head consistent with LCPD. She was made non-weight bearing and GH therapy was discontinued. Fig. 2Linear Growth ChartFig. 3At onset of knee pain: Right capital femoral epiphyseal height loss was observed without evidence of slippage. No significant sclerosis Six weeks later, radiographs (Fig. [4](#Fig4){ref-type="fig"}) revealed further collapse with involvement of the entire femoral head, and she was now using crutches/wheelchair. Her lesion stabilized after 8 months with conservative management. She remained off GH, had menarche at 11 years 8 months and reached skeletal maturity at 13 years old with final adult height at 4′2″ (128 cm); Z-score of − 5.5. She is ambulating independently with a leg length discrepancy \< 1 cm. She occasionally has right hip pain after walking for long periods and is being monitored closely for early hip arthritis. She has regular menses and continues on levothyroxine and calcitriol for TSH and PTH resistance respectively. Fig. 4Six weeks later: Increased flattening and sclerosis of right capital femoral epiphysis and mild increased fragmentation laterally Discussion {#Sec3} ========== Leg Calves Perthes Disease (LCPD), also called avascular necrosis of the femoral head epiphysis, is most commonly seen amongst children 4--8 years of age. LCPD is grouped together with slipped capital femoral epiphysis (SCFE) and scoliosis as possible adverse skeletal effects of the rapid growth caused by GH therapy in children with GH deficiency \[[@CR3]\]. The incidence rate for LCPD in idiopathic growth hormone deficient patients is reported to be 10 per 100,000 compared with 5--7 per 100,000 for the general population. Interestingly, the rate did not vary if the GH deficient children were on growth hormone therapy or not \[[@CR4]\]. LCPD is not known to be associated with ACRDYS, but LCPD was reported in a patient with PHP1A not on GH \[[@CR5]\]. Children with PHP1A treated with GH have not been reported to have an increased risk of LCPD. Proposed underlying risks of LCPD include glucocorticoid exposure, genetic mutations of COL2A1, coagulation abnormalities, traumatic injury to the blood supply, transient synovitis, second-hand smoke exposure, and venous congestion \[[@CR6]\]. Patients with underlying renal failure or kidney transplants who are on GH therapy are also reported to be at higher risk to develop LCPD \[[@CR4]\]. Similar to a previous case report of a pre-pubertal female with GH deficiency who developed LCPD \[[@CR7]\], we hypothesize that growth spurt due to GH treatment led to an insufficient blood supply to the epiphysis, resulting in LCPD in our patient. It is currently advised to stop GH treatment in cases of LCPD to decrease the severity of the disease and prevent contralateral disease \[[@CR7]\]. This is what we did and she did not develop contralateral disease. The 2018 consensus guidelines for PHP1A patients include a recommendation to screen all patients for GH deficiency \[[@CR8]\] and then to treat those who are deficient. Current standard starting dose of GH is 0.3 mg/kg/week divided daily and titrated utilizing the IGF-1 level and the linear growth velocity \[[@CR2]\]. For PHP1A children born small for gestational age (SGA), GH treatment can be considered, and typically higher doses are used than for GH deficiency \[[@CR2], [@CR8]\]. De Zegher et al. reported higher doses of GH at 0.46 mg/kg/week used in patients with SGA without skeletal dysplasia and GH was well tolerated \[[@CR9]\]. Our patient who was born SGA was started on higher GH dose of 0.35 mg/kg/week as compared to GH deficient PHP1A patients. It may be that this higher dose contributed to the development of LCPD. Conclusion {#Sec4} ========== GH is the standard of care for patients with PHP1A with GH deficiency and has been shown to increase final adult height significantly and there have been no reported cases as of yet of LCPD on GH. Efficacy of GH in improving final adult height in patients with ACRDYS1 is unknown. The risk of LCPD during GH treatment in patients with ACRDYS1 needs to be considered and patients should be counseled accordingly. Clinicians may also consider using a lower initial GH dose such as 0.16--0.24 mg/kg/week used for growth hormone deficiency \[[@CR10]\]. ACRDYS 1 : Acrodysostosis type 1 ACRDYS 2 : Acrodysostosis type 2 AHO : Albright hereditary osteodystrophy PHP1A : Pseudohypoparathyroidism type 1A *GNAS* : Gα-stimulatory subunit *PRKAR1A* : Protein kinase A regulatory subunit 1A *PDE4D* : Phosphodiestarase type 4D PKA : Protein kinase A GH : Growth hormone PTH : Parathyroid hormone TSH : Thyroid stimulating hormone LH : Luteinizing hormone FSH : Follicle stimulating hormone GHRH : Growth hormone releasing hormone LCPD : Legg-Calve-Perthes Disease SCFE : Slipped Capital Femoral Epiphysis IGF-1 : Insulin like growth factor-1 IGF-BP3 : Insulin like growth factor binding protein 3 **Publisher's Note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. not applicable. All authors contributed equally to the case report. All authors read and approved the final manuscript. Funding is not needed for this case report. not applicable. not applicable. consent obtained from patient. The authors declare that they have no competing interests.

Introduction {#s1} ============ In the gastrointestinal tract, interstitial cells of Cajal (ICC) function as pacemakers, neurotransmitter transducers, and mechanosensors that respond to physical and chemical signals, and thereby modulate smooth muscle contractility [@pone.0048897-Won1], [@pone.0048897-Choi1], [@pone.0048897-Ward1], [@pone.0048897-Nakayama1]. Alterations in ICC function have been linked to more than a dozen gastrointestinal diseases [@pone.0048897-Streutker1], [@pone.0048897-Mostafa1]. In the last decade, novel ICCs have also been identified in the urinary bladder in several species, including guinea pigs, rats, and humans [@pone.0048897-McCloskey1], [@pone.0048897-PiasecznaPiotrowska1], [@pone.0048897-Shafik1], [@pone.0048897-Kim1], [@pone.0048897-Okada1], [@pone.0048897-Johnston1]. Unlike ICC in gut, the function of ICC in bladder is poorly understood, but emerging data indicates that they too, are implicated in several bladder diseases. These disorders offer the chance to gain insights into ICC functioning. In megacystis-microcolon intestinal hypoperistalsis syndrome (MMIHS), a congenital lethal disease in newborns, patients are unable to void spontaneously and have a massively dilated bladder. It is thought that the lack of ICC in the MMIHS bladder is responsible for this lethal voiding dysfunction [@pone.0048897-PiasecznaPiotrowska1]. Proto-oncogene c-Kit (C-kit, tyrosine-protein kinase Kit, or CD117) is a receptor tyrosine kinase (RTK) expressed on the surface of hematopoietic as well as other cell types such as mast cells. Signaling through c-kit plays a role in cell survival, proliferation, and differentiation, and gain of function mutations in this protein are associated with multiple tumors [@pone.0048897-Papaspyrou1], [@pone.0048897-Corless1], [@pone.0048897-Arock1], [@pone.0048897-Rulina1]. In the digestive tract, c-kit is used as the "gold standard" for identification of ICC. C-kit has also been identified in urinary bladder in guinea pig, rat and human, and further functional characterization has suggested these c-kit-positive cells are like the ICC in gastrointestinal tract [@pone.0048897-Okada1], [@pone.0048897-McCloskey2], [@pone.0048897-Biers1], [@pone.0048897-Johnston2], [@pone.0048897-Vahabi1], [@pone.0048897-Wang1]. In the mouse urinary tract, there has been some confusion about the presence of c-kit in bladder ICC. Pezzone and co-authors reported c-kit positive cells in ureter, but not in bladder [@pone.0048897-Pezzone1]. Meanwhile, McCloskey *et al.* identified c-kit positive cells in both wild-type and *W/Wv* (c-kit mutant) mice [@pone.0048897-McCloskey3], while other investigators have failed to find c-kit in mouse bladder at all [@pone.0048897-Lagou1], [@pone.0048897-Koh1]. ICC are stellate-like cells with long dendrites or spikes. They have close contacts with nerve varicosities and smooth muscle cells and form gap junctions with each other, which provide a route for the diffusion of low molecular weight materials as an important intercellular signal communication pathway between these types of cells. Thus gap junctions have a crucial role in mediating the synchronized contraction of smooth muscle cells. Nemeth *et al.* reported that gap junction protein connexin 43 is present in ICC with convincing co-localization of c-kit. ICC could be seen to form a three-dimensional network in the normal colonic bowel wall and lack of connexin43 expression in the aganglionic bowel of Hirschsprung\'s disease (HD) may be partly responsible for the smooth muscle motility dysfunction in HD patients [@pone.0048897-Nemeth1]. It has remained unclear as to whether ICC in the urinary bladder also express connexin 43, or where connexin 43 is located. While dozens of studies indicate that connexin 43 plays a crucial role in modulating bladder overactivity, its precise localization is still the subject of debate and attributed mostly to bladder smooth muscle cells and/or myofibroblasts. We therefore sought to define more clearly the molecular signature of ICC and in doing so resolve some of these uncertainties [@pone.0048897-Hashitani1], [@pone.0048897-Haefliger1], [@pone.0048897-Sui1], [@pone.0048897-Neuhaus1], [@pone.0048897-Ikeda1], [@pone.0048897-Neuhaus2]. We recently identified a population of cells in mouse bladder that express ectonucleoside triphosphate diphosphohydrolase 2 (NTPDase2), an ectonucleotidase that degrades ATP/UTP to ADP/UDP and further to AMP/UMP [@pone.0048897-Yu1]. These NTPDase2-positive cells are a unique subset of cells in bladder that exhibit narrow elongated and branched cell processes and clearly wrap around smooth muscle cell bundles. They are also expressed in a sub-region of the lamina propria that is immediately adjacent to the bladder smooth muscle layer and seems to be a natural extension from between the smooth muscle spaces. Intriguingly our data indicated they were distinct from smooth muscle, fibroblasts, myofibroblasts and neurons, indicating they might be ICC [@pone.0048897-Yu1]. In this study we have used confocal immunofluorescent microscopy of multiple cellular markers to define the identity of these cells. Our data demonstrate that they are ICC by the usual accepted definitions and that ICC in mouse bladder can be clearly defined by the presence of c-Kit, NTPDase2, CD34, Ano1, connexin 43, vimentin, desmin, PDGF receptor and merlin/NF2. These data establish a molecular expression profile for ICC in mouse bladder, which can be used to assist in explorations of their functional roles in future. Materials and Methods {#s2} ===================== Materials {#s2a} --------- Unless otherwise specified, all chemicals were obtained from Sigma (St. Louis, MO) and were of reagent grade or better. Animals {#s2b} ------- Mice used in this study were C57BL/6J mice (19--21 g) from Charles River Laboratories (Wilmington, MA). Mice were euthanized by inhalation of 100% CO2. After euthanasia and thoracotomy, the bladders were rapidly excised and processed as described below. All animal studies were carried out with the approval of the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee (Protocol **\#**051-2009). Antibodies and labeled probes {#s2c} ----------------------------- Affinity-purified polyclonal sheep anti-NTPDase2 antibody (AF5797) was purchased from R&D systems (Minneapolis, MN). Affinity-purified monoclonal rat anti-CD34 \[MEC14.7\] antibody (ab8158) and monoclonal rabbit anti-TMEM16A \[SP31\] antibody (ab64085) were purchased from Abcam (Cambridge, MA). Affinity-purified polyclonal rabbit anti-c-kit antibody (A0501), affinity-purified polyclonal rabbit anti-connexin 43 antibody (C0158), affinity-purified polyclonal rabbit anti-vimentin antibody (C0390), affinity-purified polyclonal rabbit anti-desmin antibody (C0171), affinity-purified polyclonal rabbit anti-PDGFβ antibody (B7194), affinity-purified polyclonal rabbit anti-merlin antibody (A8046) were purchased from Assay Biotechnology (Sunnyvale, CA). Affinity-purified polyclonal goat anti-mast cell tryptase (G-12) antibody (sc-32474) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Affinity-purified monoclonal mouse anti-αSMA antibody is a kind gift from the laboratory of Dr. Raghu Kalluri (Beth Israel Deaconess Medical Center). Secondary donkey anti-rabbit/sheep/goat/mouse antibodies conjugated to Alexa 488/546, and Topro-3 were purchased from Invitrogen-Molecular Probes (Carlsbad, CA). Immunofluorescence analysis {#s2d} --------------------------- Excised bladders were fixed in 4% (w/v) paraformaldehyde dissolved in 100 mM sodium cacodylate (pH 7.4) buffer for 2 h at room temperature. Alternatively, tissue was fixed in 4°C methanol for 10 min. Fixed tissue was cut into small pieces with a razor blade, cryoprotected with 30% sucrose solution (w/v), frozen, sectioned (5 µm), and incubated with primary antibodies (1∶50--1∶500 dilution) overnight at 4°C as described previously [@pone.0048897-Yu1]. After washing away unbound primary antibody, the sections were incubated with a mixture of Alexa 488-conjugated secondary antibody (diluted 1∶100), Alexa 546-conjugated secondary antibody (diluted 1∶100), and Topro-3 (1∶1,000). The sections were washed with PBS, postfixed with 4% (w/v) paraformaldehyde, and mounted under coverslips with p-diaminobenzidine-containing mounting medium. All immunofluorescent localization data shown are representative images of staining performed on at least three individual bladders. Scanning laser confocal analysis {#s2e} -------------------------------- Imaging was performed on a Zeiss LSM-510 confocal microscope equipped with argon and green and red helium-neon lasers (Thornwood, NY). Images were acquired by sequential scanning with a 63× (1.4 numerical aperture) planapochromat oil objective and the appropriate filter combinations. Serial sections were captured with a 0.25 µm step size. The images (1024 & 1024 pixels) were saved as TIFF files. Serial sections were projected into one image using LSM-510 software. The contrast level of the final images was adjusted in Adobe Photoshop, and the images were imported into Adobe Illustrator CS3. Results {#s3} ======= Immunostaining of frozen bladder sections was performed for multiple proteins of interest. [Figure 1](#pone-0048897-g001){ref-type="fig"} shows the spatial relationship between ectonucleotidase NTPDase2 and bladder smooth muscle stained with antibody to α-smooth muscle actin (αSMA, red). NTPDase2-positive cells (green) occur in close physical proximity to smooth muscle cells with clear evidence of elongated cell processes. NTPDase2 is also expressed in the stromal compartment next to the detrusor [@pone.0048897-Yu1]. Arrows in the merged image show that despite lying adjacent to smooth muscle cells, the green signal remains spatially separate from αSMA. {#pone-0048897-g001} Confocal immunofluorescent laser scanning microscopy revealed that spindly NTPDase2-positive cells were dispersed between smooth muscle cell bundles and densely occupied the lamina propria next to the muscle layer [@pone.0048897-Yu1], indicating morphological characteristics resembling those of ICC. To confirm this, we stained the same cells with a c-kit antibody, and both signals showed excellent co-localization ([Fig. 2](#pone-0048897-g002){ref-type="fig"}). We only observed c-kit staining in NTPDase2-positive cells, and not in smooth muscle or urothelial cells, indicating that the NTPDase2-positive cells in mouse bladder are likely to be ICC. {#pone-0048897-g002} To extend this initial characterization we stained tissue sections with CD34. CD34 is an important adhesion molecule. It has been identified in hematopoietic cells, endothelial cells, mast cells, as well as being reported in ICC [@pone.0048897-Nunomura1], [@pone.0048897-Nielsen1], [@pone.0048897-Lorincz1], [@pone.0048897-Junquera1]. CD34 antibodies strongly stained the NTPDase2-positive cells ([Fig. 3 A--C](#pone-0048897-g003){ref-type="fig"}). CD34 also stained a few NTPDase2-negative cells in lamina propria beneath the urothelium, possibly indicating hematopoietic cells or mast cells. It has been shown that both c-kit and CD34 also stain mast cells in addition to ICC. To exclude the possibility that NTPDase2-positive cells are mast cells, we stained bladder tissue sections with tryptase. The tryptase signal co-localized with the CD34 signal beneath the urothelium in lamina propria, but not in all cells ([Fig. 3 D--F](#pone-0048897-g003){ref-type="fig"}). Importantly, c-kit positive cells wrapping around muscle bundles are typtase negative ([Fig. 3 G--I](#pone-0048897-g003){ref-type="fig"}). These data strongly suggest that NTPDase2-positive, c-kit and CD34 positive but tryptase negative cells are ICC in mouse urinary bladder tissue. {#pone-0048897-g003} Recently, anoctamin 1 (Ano1, DOG1, TMEM16A) has been recognized as a selective marker of ICC [@pone.0048897-GomezPinilla1]. Ano1 is a Ca^2+^ - activated Cl^−^ channel and its expression in ICC is fundamental for slow wave activity in gastrointestinal muscles [@pone.0048897-Hwang1]. Interestingly, we observed Ano1 signal in mouse bladder tissue between muscle bundles with a fibrous structure ([Fig. 4 A--C](#pone-0048897-g004){ref-type="fig"}), and this signal has nice co-localization with CD34 signal ([Fig. 4 D--F](#pone-0048897-g004){ref-type="fig"}). These data provide additional confirmatory evidence that these cells are urinary tract ICC. {#pone-0048897-g004} NTPDase2-positive cells also exhibited convincing colocalization with connexin 43 ([Fig. 5](#pone-0048897-g005){ref-type="fig"}) and interestingly there was no evidence for broader expression of connexin 43 in smooth muscle itself, or any other types of cells in bladder. This data supports the finding of c-kit and connexin 43 co-localization in colonic bowel ICC [@pone.0048897-Nemeth1], indicating a number of similarities between bladder and gastrointestinal tract ICC. {#pone-0048897-g005} Vimentin is the major cytoskeletal component of mesenchymal cells and is often used as a marker of mesenchymally derived cells. Because of this it is also often used as an ICC marker and has been reported in guinea pig bladder ICC and ICC in other organs [@pone.0048897-Johnston1], [@pone.0048897-Davidson1], [@pone.0048897-Popescu1], [@pone.0048897-Komuro1]. Immunostaining showed co-localization of NTPDase2-positive cells with vimentin ([Fig. 6A--C](#pone-0048897-g006){ref-type="fig"}). Vimentin antibody also labels NTPDase2-negative cells in the lamina propria (not shown in the photo) and these are likely to be myofibroblasts and/or fibroblasts. Desmin is another cytoskeletal marker closely related to muscle tissue development and differentiation. NTPDase2-positive cells also showed partial co-localization with desmin signal, and like vimentin, labeled other cells in the lamina propria that might be myofibroblasts ([Fig. 6 D--F](#pone-0048897-g006){ref-type="fig"}, lower right quadrant). PDGF is a potent mitogen for cells of mesenchymal origin. PDGF receptor plays important roles in driving mesenchymal proliferation, migration, differentiation, and in disease states such as fibrosis. Recently PDGF receptors (PDGF-R) have been reported in ICC in gastrointestinal tract [@pone.0048897-Kurahashi1], [@pone.0048897-Chan1]. Consistent with these reports, we detected strong PDGFβ-R signals in NTPDase2-positive cells. PDGFβ-R was also expressed in other cell types, like smooth muscle cells and possibly fibroblasts but generally staining was much weaker ([Fig. 6 G--I](#pone-0048897-g006){ref-type="fig"}). {#pone-0048897-g006} Gastrointestinal stromal tumor (GIST) is one of the most common mesenchymal tumors of the gastrointestinal tract. Tumorigenesis is usually associated with gain of function mutations in the c-kit gene and activation of inappropriate tyrosine kinase signaling (∼95%). Activating mutations in other receptors can also result in GIST e.g PDGF-Rα. In recent years, investigators have studied the relationship between GIST and merlin (also known as neurofibromin 2, NF2). Merlin is a membrane-cytoskeletal marker that is mainly located in adherens junctions [@pone.0048897-Lallemand1], [@pone.0048897-Rangwala1], [@pone.0048897-Flaiz1]. It is predominantly found in nerve tissue and functions as a tumor suppressor. Several recent studies suggest that merlin/NF2 could be located in c-kit positive GIST cells [@pone.0048897-Fukasawa1], [@pone.0048897-Pylkkanen1], [@pone.0048897-Taddei1]. Therefore we were interested to see whether we could detect its expression in bladder ICC. Our data indicate that this protein is expressed in NTPDase2-positive cells in mouse bladder with convincing co-localization of these two proteins ([Fig. 7](#pone-0048897-g007){ref-type="fig"}). {#pone-0048897-g007} Discussion {#s4} ========== ICC were originally discovered as a morphologically distinct cell in gastrointestinal tract [@pone.0048897-FaussonePellegrini1], [@pone.0048897-Thuneberg1]. They now have a widely accepted role as a 'pacemaker' cell responsible for generating slow waves involved in peristalsis. ICC or ICC-like cells are now being described in many other muscular organs [@pone.0048897-McCloskey1], [@pone.0048897-Popescu1], [@pone.0048897-Yuzuriha1], [@pone.0048897-Formey1], [@pone.0048897-Wang2], but their functions in these systems are much less clear. The urinary bladder has been reported to possess ICC in several species, but there is some question as to whether they are universally expressed in mammals or have similar functions to their counterparts in the GI tract [@pone.0048897-Pezzone1], [@pone.0048897-McCloskey3], [@pone.0048897-Lagou1], [@pone.0048897-McCloskey4]. In this study, we have clearly demonstrated the presence of ICC marker c-kit, in NTPDase2-positive cells of mouse bladder. NTPDase2 is not expressed by fibroblasts, myofibroblasts, neurons or smooth muscle [@pone.0048897-Yu1], therefore colocalization with c-Kit identifies it as another specific marker for ICC in bladder. The localization of NTPDase2 expression in ICC cells raises an interesting question as to whether purinergic signaling plays a role in the regulation of their function. Multiple purinergic receptors have been reported in ICC, including P2X~2/5~, P2Y~4~ and P2Y~1~ [@pone.0048897-Burnstock1], [@pone.0048897-VanNassauw1], [@pone.0048897-Chen1]. In a c-kit mutant mouse (W/Wv), purinergic contraction of bladder smooth muscle has been significantly up-regulated, indicating an intriguing relationship between c-kit signaling pathway and purinergic signaling pathways [@pone.0048897-McCloskey3]. The presence of purinergic receptors in ICC and the clear functional importance of purines in bladder pathophysiology, suggests that NTPDase2/purinergic signaling might be important in regulating ICC pacemaker activity or mechanosensory function. Mechanical stretch of cells causes release of ATP [@pone.0048897-Burnstock2], and mechanical stretch of ICC regulates pacemaker activity and smooth muscle contraction in both stomach and bladder [@pone.0048897-Won1], [@pone.0048897-Wang1]. Furthermore, the presence of connexin 43 in ICC provides an ideal pathway for ATP and other purines to communicate through connexin hemichannels or intercellular gap junctions [@pone.0048897-Lazarowski1], [@pone.0048897-Hanani1]. Our findings suggest that NTPDase 2 might play an important regulatory role in ICC purinergic signaling, possibly by temporally and spatially limiting P2-receptor exposure to nucleotides and nucleosides. There has been some controversy as to the existence of c-kit positive cells in mouse bladder and our own experience has been that many commercially available c-kit antibodies fail to immunostain frozen sections [@pone.0048897-Yu1]. This is possibly due to variants or isoforms of c-Kit expressed in different tissues and/or substantial species heterogeneity (<http://www.uniprot.org/uniprot/P05532>). In this study, we co-localized seven putative ICC related markers with NTPDase2-positive cells in mice bladder ([Table 1](#pone-0048897-t001){ref-type="table"}). In addition to c-kit we show these cells also express strong CD34, a protein that also often functions as a mast/stem cell marker, which is consistent with c-kit as a mast/stem cell growth factor receptor, indicating a possible functional relationship. 10.1371/journal.pone.0048897.t001 ###### Summary of expression profile of molecular markers in bladder ICC. {#pone-0048897-t001-1} Markers Bladder ICC Myofibroblast/fibroblast Smooth muscle Neuron Mast cell --------------- ------------- -------------------------- --------------- -------- ----------- NTPDase1 (33) − − \+ − NTPDase2(33) \+ − − − − Nt5e (33) − − \+ − FSP1 (33) − \+ − − − αSMA (33) − +/− \+ − − CGRP (33) − − − \+ − C-kit \+ − − − \+ CD34 \+ − − − \+ Tryptase − − − − \+ Ano1 \+ − − − Connexin 43 \+ − − − Vimentin \+ \+ − − Desmin \+ \+ − − PDGFRβ \+ \+ \+ Ano1 has been shown to be a specific marker for ICC [@pone.0048897-GomezPinilla1], [@pone.0048897-Hwang1]. It has also been reported to be a highly sensitive and specific marker for GIST, with 100% sensitivity compared with 95% for kit signaling [@pone.0048897-Hwang2]. Furthermore we show clear evidence for expression of gap junction protein connexin 43, three mesenchymal related markers in vimentin, desmin and PDGF-R and finally a novel neural marker that has been reported in GIST ICC cells, merlin/NF2. Combining our findings in this study with our previously published data on NTPDase2 positive cells in mouse bladder [@pone.0048897-Yu1], we summarize the cellular protein expression profile in [table 1](#pone-0048897-t001){ref-type="table"} and compared these markers with other cell types in the urinary bladder. The picture which emerges strongly indicates that the NTPDase2-positive cells are mouse bladder ICC. This expanded molecular profile may be useful for identifying ICC in mice given the technical difficulties associated with c-kit detection. Connexin 43 appears to be an important participant in disease states like bladder carcinoma and bladder overactivity, but it is almost universally considered to be located on bladder smooth muscle cells or myofibroblasts [@pone.0048897-Hashitani1], [@pone.0048897-Haefliger1], [@pone.0048897-Sui1], [@pone.0048897-Neuhaus1], [@pone.0048897-Ikeda1], [@pone.0048897-Neuhaus2], [@pone.0048897-Roosen1]. As but one example, connexin 43 was recently shown to be important in regulating diurnal changes in bladder capacity and micturition [@pone.0048897-Negoro1]. While the major findings are not in question and indeed are of great significance, the authors attributed this function to connexin 43 in bladder smooth muscle. The connexin 43 labeling showed a sparse staining pattern that is likely located at the edge of muscle bundle, exactly the location of ICC [@pone.0048897-Yu1]. Since the presence of ICC in mouse bladder has been difficult to define by c-kit staining, it is understandable that immunostaining deep within the detrusor is usually considered to be smooth muscle cells. However, we have previously shown that these cells are distinct from smooth muscle cells with multiple markers, including α-SMA, Entpd1, Nt5e, NTPDase2 ([table 1](#pone-0048897-t001){ref-type="table"}) [@pone.0048897-Yu1], and now several more in this study. In other published studies, connexin 43 found in the lamina propria, are attributed as myofibroblasts. In our experience, myofibroblasts are enriched immediately beneath the urothelial layer instead of next to smooth muscle layer and this was demonstrated by labeling myofibroblasts with α-SMA while the connexin 43 positive cells do not stain α-SMA [@pone.0048897-Yu1]. The finding of connexin 43 in ICC indicates that its function in bladder is likely due to the activity of ICC. A somewhat surprising finding was that merlin/NF2, most often considered a neuronal marker, was expressed in bladder ICC. Does this indicate that bladder ICC have nerve cell like function? Considering the pacemaker and neurotransducer functions of gastrointestinal ICC, it may be true. The role of ICC in bladder is not well defined but by analogy with GI tract may be responsible for tonic regulation of smooth muscle contractility. Since dysregulation of ICC in GI tract leads to serious pathology ranging from GIST to motility disorders, there is a strong likelihood of ICC involvement in certain syndromes of voiding dysfunction. Indeed as noted, alterations to connexin 43 expression have been linked to cancer and overactive bladder. Better ways to define the morphological signature of ICC will hopefully assist investigations into their function. In summary, we defined a unique set of molecular markers of NTPDase2 positive cells in mouse urinary bladder. These findings are in line with other reports in gastrointestinal ICC or previous observations in bladder ICC. We believe that, as in other species, these NTPDase2 positive cells are mouse ICC, and we hope that our findings will help other investigators to avoid misidentifying this unique cell type. A clear definition of molecular expression in bladder ICC will also facilitate directed investigations in mouse knockouts to understand the role of ICC in spontaneous contraction and mechanosensation. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: WY MLZ WGH. Performed the experiments: WY. Analyzed the data: WY MLZ WGH. Contributed reagents/materials/analysis tools: WY MLZ WGH. Wrote the paper: WY MLZ WGH. NA.

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 and background =========================== Fulminant hepatic failure is defined as severe liver injury resulting in impaired synthetic capacity and encephalopathy in patients with previously normal liver function \[[@REF1]\]. It is typically characterized by increases in transaminases, which serves as the first clue to the clinician that acute liver injury is occurring. Depending on how elevated the patient's transaminases are, one can then narrow the list of differential diagnoses. Elevations in transaminases greater than ten times the upper limit of normal are suggestive of ischemic, toxic or viral liver injury \[[@REF2]\]. Of these, toxic (acetaminophen) and viral are the most common etiologies worldwide \[[@REF3]\]. Hepatitis A, B, C, D, and E are the most frequently encountered forms of viral hepatitis in the medical literature. Rarely will non-hepatitis viruses cause fulminant hepatic failure. The goal of this review article is to aid clinicians in the early identification of herpes simplex virus (HSV) hepatitis in order to initiate rapid treatment and ultimately prevent mortality. Review ====== Epidemiology HSV hepatitis in adults is a rare entity that was first reported in 1969 by Flewett et al. \[[@REF4]\]. Since its discovery, it is thought to comprise a mere 1% of all cases of acute liver failure (ALF), and only 2% of viral associated ALF \[[@REF5]\]. In the limited amount of cases that have been reported in the literature, it has occurred more frequently in patients that are in an immunocompromised state. These states include pregnancy, corticosteroid use, human immunodeficiency virus (HIV), and autoimmune disease \[[@REF6]\]. There also seems to be a female predilection of the disease. In 2007, Norvell et al. compiled 137 cases of HSV hepatitis that were available in the literature at the time. Their data demonstrated a female predominance of 62%. In the same study, 76% of inflicted patients were found to be either pregnant or immunosuppressed \[[@REF7]\]. It is important note however that HSV hepatitis is non-discriminating and may occur in male immunocompetent patients as well \[[@REF8]\]. Pathogenesis There are multiple theories regarding the pathogenesis of HSV hepatitis. The first theory postulates that a large HSV inoculum at the time of initial infection overwhelms the hosts defense system, and results in visceral dissemination into the liver. The second theory suggests that visceral dissemination derives from herpetic lesions in the setting of impaired macrophages, cytotoxic T-lymphocytes, and delayed type hypersensitivity. The third theory suggests that an acute HSV infection, superimposed on a latent HSV reactivation, may be the culprit in the development of fulminant hepatitis. Finally, a fourth theory takes into consideration the heterogeneity of HSV. Studies have proven that certain strains of HSV are neurovirulent causing herpetic encephalitis. Similarly, there may be certain strains that are hepatovirulent and result in fulminant hepatic failure \[[@REF9], [@REF10]\]. Clinical manifestations Identifying HSV hepatitis can be a precarious task. Clinically its presentation is indistinguishable from other etiologies of acute hepatitis. In addition, many of the cases reported in the literature have presented with an anicteric pattern \[[@REF11]\]. To further complicate matters, less than half of the reported cases present with mucocutaneous lesions suggestive of HSV \[[@REF12]\]. Leukopenia and thrombocytopenia have also been reported to be associated with HSV hepatitis \[[@REF13]\]. Other findings that are often seen in HSV hepatitis include fever, coagulopathy (resulting in life-threatening hemorrhage), and acute renal failure \[[@REF7]\]. Given the lack of objective physical findings, and absolute laboratory results, history gathering becomes essential. Diagnosis Diagnostic tools that can aid a physician in recognizing HSV hepatitis include HSV serology, HSV DNA by polymerase chain reaction (PCR), computed tomography (CT) scan and liver biopsy. Regarding laboratory studies, HSV DNA by PCR has been shown to be more discriminating than serologic testing for diagnosing or excluding HSV as a cause of ALF \[[@REF14]\]. CT scan findings are very non-specific, but may show hepatomegaly along with diffuse hypodense 1-4 mm lesions, which represent foci of acute hepatic necrosis. Ultrasound may also be useful in excluding other diagnoses that are suspected. This is especially true in pregnant patients where fatty liver disease of pregnancy is often the cause of acute liver failure \[[@REF15]\]. Despite the utility of these diagnostic modalities, liver biopsy remains the gold standard for diagnosing HSV hepatitis. Histologically, biopsy will show focal or confluent areas of acidophilic type necrosis, with little associated inflammation. In non-necrotic areas, Cowdry type A bodies surrounded by halos, may also be found \[[@REF16]\]. Treatment Treatment of HSV hepatitis, and subsequent prognosis, is highly contingent on time. Rapidly progressive acute liver failure occurs in 74% of cases, with mortality rates reaching 90% \[[@REF17]\]. Studies have demonstrated better clinical outcomes in patients who were started on acyclovir early in their hospital course. One literature review in particular, concluded that patients who received early treatment were less likely to die, or require liver transplantation \[[@REF18]\]. It is important to note however that there have been cases of acyclovir resistant HSV hepatitis reported in the literature. Resistance of HSV to acyclovir is most often seen in immunocompromised patients and rates have been reported to range from 3.5% to 10% in this population \[[@REF19]\]. Acyclovir is a nucleoside analogue that is phosphorylated into its active form by thymidine kinase in HSV-infected cells. Acyclovir resistance results from mutations in the thymidine kinase gene that cause decreased production or complete absence of thymidine kinase \[[@REF20]\]. In cases of acyclovir resistant HSV hepatitis, Foscarnet or Cidofovir may be employed. These two agents inhibit the catalytic unit of viral DNA polymerase and do not require activation by thymidine kinase \[[@REF21]\]. Chaudhary et al. reported one case of HSV hepatitis that showed no improvement with acyclovir, but when foscarnet was added the patient's mental status improved and she was ultimately discharged home with no need for liver transplantation \[[@REF22]\]. Other important questions to consider with regards to treatment are the optimal route of administration (intravenous vs oral) and the subsequent duration of therapy. This is highlighted by a case reported by Czartoski et al. in which a patient diagnosed with HSV hepatitis initially showed improvement on intravenous acyclovir, but quickly decompensated when switched to oral valacyclovir \[[@REF23]\]. Despite completing 43 days of intravenous acyclovir prior to discontinuation, the patient ultimately expired. Once fulminant hepatic failure occurs the only option that remains is liver transplantation. In Norvell et al.'s literature review, seven patients with HSV hepatitis underwent orthotopic liver transplantation after developing fulminant hepatic failure. A mere three patients ultimately survived, expiring due to complications associated with liver transplantation \[[@REF7]\]. Despite a lack of evidence, there may be a role for plasmapheresis in HSV hepatitis. Holt et al. reported one such case in which a pregnant patient presented to their institution with acute liver failure. Given the patient\'s anemia, thrombocytopenia and transaminitis the presumptive diagnosis of HELLP (hemolysis, elevated liver enzymes, low platelet count) syndrome was made. This was followed by the rapid initiation of therapeutic plasma exchange (TPE). One day after the initiation of TPE the patient's transaminitis markedly improved (Alanine transaminase decreased by 5,131 IU/L; Aspartate transaminase decreased by 1,282 IU/L). The patient was later diagnosed with HSV hepatitis through liver biopsy and was started on acyclovir. She was ultimately discharged home with normal transaminase levels \[[@REF24]\]. Conclusions =========== In conclusion, HSV hepatitis is a rare cause of rapidly progressive acute hepatitis that oftentimes results in fulminant hepatic failure. Due to its rarity, and lack of discernible clinical features, early recognition and history gathering become indispensable. Current guidelines do not recommend empiric treatment with acyclovir, but given its high mortality rate and complications associated with liver transplantation, one should consider empiric treatment in selected patient populations where there is a high index of clinical suspicion for HSV hepatitis. These would include immunocompromised patients or promiscuous patients with new sexual partners who present with acute liver failure. In these patients empiric treatment should be strongly considered along with early testing to rule out HSV hepatitis. The authors have declared that no competing interests exist.

A carpometacarpal joint (CMCJ) of the thumb is important for the function of the thumb and in the performance regarding strong pinch and grasp. The dislocation of the CMCJ in the thumb accounts for less than 1% of all hand injuries.[@B1] Mechanical instability for the CMCJ of the thumb is an important factor, which may lead to articular degeneration of the joint and thus interfere with the normal function of the hand. Recommended treatment for this injury has ranged from closed reduction with or without percutaneous K-wire fixation to reconstruction of the ligaments. We report a case of bilateral thumb CMCJ dislocations: a unique combination of injuries. It was successfully treated with closed reduction and percutaneous K-wires fixation on one side, and an open reduction and reconstruction of the ligament in the other side. Our patient was informed that the data concerning this case would be submitted for publication. CASE REPORT =========== A 50-year-old man was taken to the emergency room as a result of a motorbike accident. At the time of impact, he was firmly grasping the handlebars with both hands. A physical examination revealed severe tenderness and dorsal prominence at the CMCJ regarding both thumbs. There were no neurovascular injuries or skin lesions. The radiographs showed dorsal dislocation of the CMCJ for both thumbs with a tiny fracture fragment in the right hand ([Fig. 1](#F1){ref-type="fig"}). His accompanying injuries were a bilateral haemothorax, nasal bone fracture, and right distal tibia fracture. Under lidocain block, a closed reduction was performed by gentle longitudinal traction. After the reduction, 3-dimensional computed tomography showed that the dislocations of both CMCJs still remained. Due to a bilateral haemothorax, surgery was performed 2 weeks after the injury. His right hand was treated with a closed reduction and percutaneous K-wires fixation under fluoroscopic guidance. However, his left hand was significantly unstable, so an open reduction with ligamentous reconstruction was performed ([Fig. 2](#F2){ref-type="fig"}). During the operation, the dorsal capsule and volar oblique ligament were ruptured, making it impossible to suture securely. Some small cartilage fragments and remnants of ligament interposed in the joint space were removed. Reconstruction of the volar oblique ligament was performed with the radial half of the flexor carpi radialis remaining in continuity at its insertion on the second metacarpal base. It was routed through a drill hole in the base of the metacarpal in the sagittal plane perpendicular to the thumb nail, using a 28-gauge stainless steel wire, passed deep to the abductor pollicis longus insertion, and then passed around the remaining flexor carpi radialis and secured over the dorsal capsule ([Fig. 3](#F3){ref-type="fig"}). Both thumbs were immobilized in a thumb spica cast for 6 weeks. Routine activities were recommended immediately upon removal of the cast. The K-wires were removed 7 weeks after surgery. At the 16-month follow-up, the patient complained of mild stiffness of the left thumb. However, there was no pain or chronic instability. DISCUSSION ========== A CMCJ of the thumb is a biconcave saddle joint: the trapezium is convex on anteroposterior views and concave on lateral views, whereas the thumb metacarpal is the opposite. This unique configuration provides a wide range of motion varying from abduction to opposition while the joint remains stable.[@B2] Furthermore, it is supported with a thickened joint capsule composed of sixteen ligaments, although stability is primarily provided by 4 main ligaments: the anterior oblique, the intermetacarpal, the dorsoradial, and the posterior oblique ligament.[@B3] Controversy concerning which ligaments are damaged in joint dislocation and which ligaments are the true key stabilizers for joint stability still exists. However, for years the volar oblique ligament has been believed to be the key stabilizer for CMCJ of the thumb. In this case, the dorsal capsule and volar oblique ligament were ruptured from the base of the first metacarpal bone. This intrinsic joint stability makes a dislocation of the thumb CMCJ quite a rare injury. Mueller[@B1] reported that the CMCJ dislocation of the thumb accounts for less than 1% of all hand injuries. Bilateral CMCJ dislocation of the thumb was first described by Khan et al.[@B4] in 2003. In their report, both joints appeared stable after closed reduction and the patient was treated with a thumb spica cast for 6 weeks. At a 15-month follow-up, the functional result was good. We believe this is the second case described regarding a bilateral CMCJ dislocation of the thumb in the English literature. Two mechanisms have been reported for a traumatic dislocation of the CMCJ of the thumb.[@B5]-[@B8] The first is a longitudinally directed force along the axis of the metacarpal with the joint in full flexion. The second is a force driven into the first web space of the thumb. This force separates the base of the first and second metacarpal joints and produces a CMCJ dislocation. It commonly occurs when motorcycle drivers grip the handlebars before an impact and the traumatic event of our case suggested that the injury was caused by the second mechanism. There is some controversy regarding which treatments are optimal for acute thumb CMCJ dislocations. The treatments have ranged from a closed reduction and immobilization to a closed or open reduction and K-wire fixation with or without the reconstruction of the capsule and ligaments.[@B2] It is believed that the most important point in the management of acute CMCJ dislocations is to assess the stability of the joint after reduction.[@B9] If the first metacarpal bone was not well seated radiographically or was loose or sloppy clinically following the reduction, surgical reduction was indicated. Simonian and Trumble[@B10] advocated thumb spica casting if stable closed reduction was obtainable, but suggested early ligament reconstruction for an unstable thumb CMCJ. They reported that early reconstruction might decrease the incidence of recurrent instability and posttraumatic arthritis. Similarly, Shah and Patel[@B8] reported that open reduction and internal fixation alone were not adequate for an unstable thumb CMCJ dislocation and that the reconstruction of the capsular as well as ligamentous structures are needed. We performed a closed reduction and percutaneous K-wires fixation on the right hand. Although the extent of ligamentous injury was not determined, we believed that the volar oblique ligament was intact since the avulsion of the first metacarpal base was seen. On the other hand, an open reduction and reconstruction of the volar oblique ligament according to Eaton et al.[@B7] were performed on the left hand due to severe instability. This had the advantage of reconstructing the joint in two planes, reconstituting the volar oblique ligament and also creating a new ligament radially in a part of the joint capsule, which was weak and membranous. At the 16-month follow-up, the patient demonstrated a normal range of motion, strength and no joint instability. Overall, he was satisfied with the outcome. Although the follow-up period is short in order to allow a comment on the development of posttraumatic arthritic changes, this positive result would encourage us to adopt this technique in future cases. No potential conflict of interest relevant to this article was reported. {#F1} {#F2} {#F3}

Familial glucocorticoid deficiency (FGD) is a rare autosomal recessive disorder in which cortisol and adrenal androgen secretions are deficient due to adrenal unresponsiveness to ACTH stimulation. The most significant phenotypic features are severe neonatal hypoglycemia, frequent childhood infection, and excessive generalized skin pigmentation related to elevated proopiomelanocortin (POMC) products ([@B1]). FGD can be caused by mutations in genes encoding the ACTH receptor \[melanocortin 2 receptor (*MC2R*)\] or its accessory protein (*MRAP*), and a similar clinical presentation has been observed in some patients with specific "mild" mutations in the steroidogenic acute regulatory protein ([@B2]--[@B4]). Most commonly, *MC2R* is mutated, and this is referred to as FGD type 1 (OMIM 202200). MSH (melanotropin) and ACTH are products of the same gene, *POMC*, and they regulate pigmentation and adrenocortical function, respectively ([@B5]). Mutations in *POMC* result in a red hair phenotype with metabolic abnormalities, including adrenal insufficiency and obesity ([@B6]). MC1R plays a central role in the regulation of skin pigmentation, is expressed in melanocytes, and binds α-MSH and ACTH with similar affinity. MC1R activation increases the ratio of black, strongly photoprotective eumelanins to reddish, poorly photoprotective pheomelanins ([@B7]). Several MC1R variants are associated with red hair/fair skin ([@B7]--[@B9]). Increased circulating ACTH acting through MC1R is believed to be the cause of the hyperpigmentation seen in primary adrenal failure. There are, however, a few reports of Addison\'s cases without hyperpigmentation ([@B10]--[@B21]). So far, no pathogenetic mechanism has been described to explain this phenomenon, although it has been noted that it is more frequently observed in fair-skinned individuals. Here we report coexistent homozygous MC2R and MC1R mutations in the same individual, causing an unusual presentation of FGD without hyperpigmentation. Patient and Methods =================== The patient presented at 4 yr of age with hypoglycemia during a respiratory tract infection after prolonged fasting. She had three further hypoglycemic attacks in the following 2 yr during infections. At the age of 6, during evaluation of her fifth hypoglycemic attack, she was diagnosed with hypocortisolemia with elevated ACTH levels. A standard ACTH stimulation test revealed a subnormal response ([Table 1](#T1){ref-type="table"}). Absence of salt wasting with normal plasma renin activity and without elevation of adrenal glucocorticoid precursors was consistent with isolated glucocorticoid deficiency ([Table 1](#T1){ref-type="table"}). ###### Biochemical and hormonal values at baseline and after a standard dose (250 μg) ACTH test Patient levels Reference ranges ---------------------------------------------------------- ------------------------------------------------------------------- ----------------------------------------------------------------------- ACTH \>1250 pg/ml (275 pmol/liter)[*^b^*](#TF1-2){ref-type="table-fn"} \<46 pg/ml (10.1 pmol/liter)[*^b^*](#TF1-2){ref-type="table-fn"} Cortisol at diagnosis 2.56 μg/dl (71.1 nmol/liter)[*^a^*](#TF1-1){ref-type="table-fn"} 5--23 μg/dl (139--639 nmol/liter)[*^a^*](#TF1-1){ref-type="table-fn"} Cortisol in standard dose ACTH stimulation test (250 μg) 4.9 μg/dl (136.1 nmol/liter)[*^b^*](#TF1-2){ref-type="table-fn"} 5--23 μg/dl (139--639 nmol/liter)[*^b^*](#TF1-2){ref-type="table-fn"} 6.6 μg/dl (183.3 nmol/liter)[*^c^*](#TF1-3){ref-type="table-fn"} \>20 μg/dl (\>550 nmol/liter)[*^c^*](#TF1-3){ref-type="table-fn"} DHEAS \<15 μg/dl (0.39 μmol/liter) 2.3--15 μg/dl (0.06--0.39 μmol/liter) Na and K 136 and 3.14 mEq/liter 134--145 and 3.5--5.3 mEq/liter Aldosterone 50 pg/ml (138.8 pmol/liter) 10--180 pg/ml (27.7--500 pmol/liter) Plasma renin activity 2.5 ng/ml · h (3.2 pmol/ml · h) 1--6.5 ng/ml · h (1.3--8.4 pmol/ml · h) The patient was found to have hypocortisolemia at the time of hypoglycemia, but ACTH levels were not measured at this time because no hyperpigmentation was present. DHEAS, Dehydroepiandrosterone sulfate. Conversion factors: DHEAS, 1 μg/dl = 38.46 μmol/liter; cortisol, 1 μg/dl = 0.036 nmol/liter; aldosterone, 1 pg/ml = 0.36 pmol/liter; ACTH, 1 pg/ml = 4.54 pmol/liter; plasma renin activity, 1 ng/ml · h = 0.77 pmol/ml · h. At the time of hypoglycemia (glucose, 25 mg/dl); before, and after ACTH stimulation test. Her parents were consanguineous, and she had two unaffected sisters. Her physical examination was normal, except that her height and weight were above the 97th centile for height and weight of a sex- and age-matched reference population. It was noted that she had no hyperpigmentation, despite very high ACTH levels. Her eye color was bluish-gray, and skin and hair pigmentation was similar to unaffected family members. The family noted that her hair color was red during infancy and darkened as she got older. At follow-up, after replacement with hydrocortisone and reduction of serum ACTH levels, her hair had reverted to a reddish color ([Fig. 1](#F1){ref-type="fig"}). {#F1} Sequencing ---------- Genomic DNA was extracted from whole blood samples after individuals gave informed consent. The sequences of the coding exons of *MC1R* and *MC2R* and their intron/exon junctions were determined by PCR and automated sequencing. Results ======= Nucleotide sequence analysis of *MC2R* revealed a homozygous mutation NC_000018.9:g.13885083C\>A, c.455C\>A (p.T152K), a loss-of function mutation previously shown to be trafficking defective ([@B22]) ([Fig. 2](#F2){ref-type="fig"}). The parents and one unaffected sister were heterozygous for the mutation, and her other unaffected sister was wild-type. Hyperpigmentation is a classical clinical finding in FGD; to find the reason for this patient\'s lack of pigmentation, we sequenced *MC1R*. A homozygous NC_000016.9:g.89986144C\>T, c.478C\>T variation (p.R160W) was detected in *MC1R* in the proband. Her parents and sisters were heterozygous for the change ([Fig. 2](#F2){ref-type="fig"}). This change, SNP ID rs1805008, has been linked to a red hair phenotype ([@B8]). {#F2} Discussion ========== Pigmentary skin changes are one of the most important clinical features and diagnostic clues in adrenal failure for clinicians. Here, we report a case of FGD without this clinical feature due to homozygous mutations in two unrelated genes, located on different chromosomes but belonging to the same family, the melanocortin receptors. The ligands of MC2R and MC1R are ACTH and α-MSH, respectively, both products of the same gene, *POMC*, which encodes a polypeptide hormone precursor that undergoes tissue-specific, posttranslational processing via cleavage by prohormone convertases. ACTH and α-MSH share the first 13 amino acids, and MC1R binds both peptides with similar affinity because of this structural similarity. Therefore, increased ACTH levels lead to hyperpigmentation in adrenal failure; in our case, skin pigmentation is absent due to a loss-of-function mutation in *MC1R*. MC1R is a control point in the regulation of pigmentation; genetic variants are detected in over 80% of individuals with red hair and/or fair skin that tan poorly, but in fewer than 20% of individuals with brown or black hair ([@B7]). The R160W change reported in our case is one of most common changes reported in red-haired individuals ([@B8]) and reduces the cAMP response to α-MSH ([@B23]--[@B25]). Our patient had red hair when she was born, which darkened as she got older; after treatment and normalization of ACTH levels, her hair gradually lightened. There are no reports of adrenal failure without pigmentation caused by *AAAS*, *DAX-1*, or *SF-1* mutation, but some nonpigmented, so-called "white Addison\'s" cases have been reported ([@B10]--[@B21]). All such cases have been recorded in Europeans with fair skin, which suggests that MC1R variants could be implicated; in many instances, diagnosis was unexpected or delayed because of the absence of pigmentation. In only one patient was a mechanism for the absence of hyperpigmentation proposed; it was attributed to generalized vitiligo, with a high degree of melanosome degradation in secondary lysosomes being reported in a skin biopsy ([@B15]). Here we report the first case of coexistent *MC2R* and *MC1R* homozygous mutations resulting in an unusual case of FGD1 without hyperpigmentation. The finding of a *MC1R* mutation may explain the absence of hyperpigmentation in this case. Additionally, MC1R variants could explain cases of white Addison\'s disease and might also handicap early phenotypic recognition of adrenal failure. For editorial see page E802 Abbreviations: FGDFamilial glucocorticoid deficiencyMC2Rmelanocortin 2 receptorPOMCproopiomelanocortin. This work has been supported by the Medical Research Council UK (New Investigator Research Grant G0801265, to L.A.M.; and Clinical Research Training Fellowship Grant G0901980, to C.H.). Disclosure Summary: The authors have nothing to disclose.

Introduction {#Sec1} ============ Previous studies suggested that surgery using polypropylene mesh could offer a better anatomical cure of pelvic organ prolapse (POP) \[[@CR1]\]. A large UK-based surgical randomised controlled trial (RCT) (PROSPECT, PROlapse Surgery: Pragmatic Evaluation and randomised Controlled Trials) \[[@CR2]\] was conducted to compare outcomes of native tissue and mesh-augmented repairs. This study showed that the outcomes for both categories of repair are similar but mesh-augmented repairs have an additional 10% risk of mesh complications. PROSPECT was a pragmatic RCT in which surgeons used the surgical techniques routinely used in their clinical practice. At the start of the study surgeons completed a questionnaire to document their surgical techniques for both native tissue and mesh/graft repairs \[[@CR3]\]. This demonstrated significant variation in the surgical technique used to perform an anterior repair. The limitations of the questionnaire study design, including the uncertainty about surgeon's terminology, gave cause for further evaluation. The current mesh pause adds additional importance to this study. This prospective qualitative study, Variation in Surgical Technique (VaST), was proposed to gain greater insight into the surgical technique variations that exist and to understand why practice variation continues to exist despite the known importance of evidence-based medicine. The objective of this study was to describe and categorise surgical techniques used to perform native tissue anterior repairs so that future studies can assess the impact of surgical technique on the outcome of surgery. Methods {#Sec2} ======= This multi-centered UK-based prospective observational study used qualitative methodology to evaluate the surgical techniques used for native tissue anterior POP repairs. A purposive sample was drawn from a cohort of surgeons who had recruited to the large surgical prolapse study, PROSPECT \[[@CR2]\]. This sample was chosen to allow a future subgroup analysis of the influence of surgical technique on outcome. An additional sample of surgeons who had chosen not to participate in PROSPECT was included to ensure techniques were representative of common practice. Recruitment concluded following saturation of themes. Data collection was performed at the individual surgeons' hospital site and the surgery was observed during routine theatre schedules. The same investigator performed all interviews and observations. Each surgeon was filmed performing a native tissue anterior repair. This was followed by a face-to-face semi-structured audio-recorded interview with the participating surgeon about their surgical technique. Additional field notes were taken. All interviews were professionally transcribed in a verbatim manner and a subset sent to the surgeons to ensure accuracy. Thematic analysis \[[@CR4]\] using all data was performed and the six phases of analysis were followed. Stages 1--3 (familiarisation with data, generating initial codes and searching for themes) involved two of the investigators. A further investigator was involved in stages 4--6 (reviewing themes, defining and naming themes and producing a report) and in independently reviewing a subset of videos. The computer software (NVIVO) was used in the analysis of data to code and develop themes. The first objective of VaST was to directly observe, describe and categorise techniques used to perform a native tissue anterior repair. Ethical approval was gained from the Sunderland Ethics Committee (REC no. 13/NE/0158, 29/05/13). Results {#Sec3} ======= Thirty surgeons were recruited to VaST; 2 surgeons were interviewed and 28 surgeons interviewed and filmed performing a native tissue anterior repair. These UK-based consultant surgeons worked in 1 of 21 centres (tertiary and district general hospitals) across England and Scotland. Table [1](#Tab1){ref-type="table"} summarises the background demographics of the surgeons and the procedures performed, both isolated anterior repairs and those with concomitant procedures.Table 1Demographics of surgeons and details of concomitant surgeryNumberType of surgeon  PROSPECT22  Non-PROSPECT8Gender of surgeon  Male20  Female10Type of consultant appointment  General gynaecologist1  Gynaecologist with special interest14  Accredited subspecialist in Urogynaecology14  Urologist1Years since consultant appointmentMean 12 years (range 3--31)Procedures observed32  Anterior repair alone12  Anterior repair + sacrospinous fixation4  Anterior repair, posterior repair + sacrospinous fixation4  Anterior and posterior repair5  Anterior repair, posterior repair and vaginal hysterectomy6  Anterior repair, posterior repair and Manchester repair1 Current techniques used to perform native tissue anterior repairs {#Sec4} ================================================================= Infiltration {#Sec5} ------------ At the start of the procedure most surgeons used infiltration in the anterior wall (*n* = 27/30). However, there was large variation in the volume used, from 3--80 ml (median 20 ml). The type of infiltration used included: local anaesthetic alone; local anaesthetic with saline; local anaesthetic with adrenaline; local anaesthetic with both adrenaline and saline; and finally adrenaline with saline. None used saline alone. The local anaesthetics used include lidocaine, bupivicaine and levobupivaciane. Surgeons were asked where they injected the infiltration. Some of the surgeons stated that they placed the infiltration within a specific place in either a superficial or deep plane."Surgeon I: It's just underneath the vaginal skin. Obviously it is very difficult when you're infiltrating to judge whether you are underneath the fascia or not but I try to be superficial so that I get a layer between the fascia and the skin.""Surgeon L: That's an interesting one and we have been arguing for years as to exactly where you are but I think that I am sub-fascial." Others were less certain of where the infiltration was placed and one surgeon described letting the infiltration, 'find the plane itself' (Surgeon AB). The surgeon's description of the depth of infiltration did not always match the investigators\' observations (Table [2](#Tab2){ref-type="table"}).Table 2Surgeons\' and investigators\' views of infiltration placementPlacement of infiltrationSurgeon\'s viewInvestigator\'s viewNone33Superficial108Deep125Uncertainty51Mixed011 Some surgeons used the presence or absence of blanching of the skin to inform whether the infiltration was in the place they wished it to be. Some took the presence of blanching to signify a superficial placement and others were unsure of what blanching signified."Surgeon Q: I\'m infiltrating it so that...the skin goes white. What layer that is, I have no idea, but essentially what I\'m trying to do, without any good evidence, is to make it go whiter.""Surgeon R: I inject local and adrenaline in the operation site underneath the fascial layer, so I don't want to see skin blanching." Incision {#Sec6} -------- Most surgeons used a longitudinal midline incision, performed with either a scalpel or scissors. One routinely used an elliptical incision (Surgeon Z) and one used a diathermy pen (Surgeon J). When considering the caudal aspect of the incision, all surgeons expressed that they avoided the area overlying the urethra. The terminology to describe this landmark varied including 'bladder neck'; '2 cm', '3 cm' or '4 cm below the urethra'; 'just below the urethra'; 'urethro-vaginal sulcus'; 'where rugosity is lost' and 'at the extent of the bulge'. When considering the cephalad extent of the incision most surgeons stated: the cervix or the vault. Other surgeons stated: 1 cm from the cervix/vault, as far as they could reach or to the extent of the prolapse. One surgeon failed to articulate an anatomical landmark and stated, 'It is related to experience' (Surgeon J). Dissection {#Sec7} ---------- The depth of the dissection through the anterior vaginal wall varied. Some performed a superficial dissection aiming to leave the vaginal muscularis, often called fascia, on the bladder and others described a deep dissection aiming to leave the vaginal muscularis on the vaginal epithelium. Figure [1](#Fig1){ref-type="fig"} shows photographic illustrations of the different depths of dissection. Surgeon F described dissecting the vaginal muscularis from both the vaginal epithelium and the underlying bladder creating 'fascial flaps' (Fig. [1](#Fig1){ref-type="fig"}c). One surgeon described dissecting to the plane 'that seems right' but was unable to specify what this plane was.Fig. 1Photographic illustrations of the levels of dissection. **a** Superficial dissection. **b** Deep dissection. **c** Fascial flap dissection The extent of lateral dissection was discussed with the surgeons. Some boney landmarks were described including the pubic arch (Surgeon P1), underneath the pubic rami (Surgeon P2, Surgeons E, F, O, S, AB) and behind the symphysis pubis. Others described muscular landmarks including the white line (arcus tendoneous fascia pelvis, ATFP) (Surgeon A), obturator internus (Surgeons J, Y) and pelvic side wall (Surgeon AA). A proportion of surgeons did not identify a landmark and explained the lateral extent of dissection as being something difficult to articulate or related to a surgeon\'s intuition."Surgeon M: I go as far as I think I need to go; that perhaps sounds rather vague and unacceptably vague but that's what I do." Fascial repair methods {#Sec8} ---------------------- Interviews and video analysis identified two levels of fascial repair, which were dependent on the depth of dissection. When a superficial dissection was performed sutures were placed in the vaginal muscularis, which was left attached to the bladder. When deep dissection was performed the sutures were also placed in the vaginal muscularis, which in these cases was left attached to the vaginal epithelium. Superficial fascial repair was the most commonly observed method. The most common method of repair, no matter the depth or structure repaired, was a form of midline suturing. The number of layers the repair included varied from one to three and the lateral extent of suture placement also varied (Fig. [2](#Fig2){ref-type="fig"}). A variety of suturing methods were observed. These methods were either interrupted or continuous and are summarised in Fig. [2](#Fig2){ref-type="fig"}. Other methods observed included a superficial dissection followed by repair of isolated defects in the vaginal muscularis repair (separate fascial defects, *n* = 2), placement of sutures in the skin when the vaginal muscularis was left attached to the bladder (n = 2), placement of sutures in the bladder wall when the vaginal muscularis was attached to the vaginal epithelium (n = 2) and repair of fascial flaps.Fig. 2Site of fascial suture placement and methods of fascial repair suturing. **a** Site of fascial placement. **b** Methods of fascial repair suturing The suture material used to repair the vaginal muscularis included polygycolic acid suture (PGA) or polydioxanone suture (PDS) and sometimes a combination of both when two or more layers of fascial repair were performed. Skin trimming {#Sec9} ------------- Skin excision was performed to some extent by nearly all surgeons. Those who did not routinely trim the vaginal skin discussed situations when vaginal skin was excised. Most surgeons who trimmed the vaginal skin stated they were careful only to remove a small amount. Some stated that the amount of skin trimmed was dependent on the amount of redundant skin or the size of the bulge/prolapse. Surgeon AA explained, 'enough so the skin lies without being under any tension'. Skin closure {#Sec10} ------------ The vaginal epithelium was closed predominantly with PGA suture but of varying calibres (0, 1, 2.0). One surgeon closed the vaginal epithelium with poliglecaprone 25 (monocryl). The main method of closure was a continuous locking (CL) (*n* = 23/28) suturing. The other methods included continuous non-locking (CNL) (*n* = 2/28), interrupted (*n* = 1/28) and mattress sutures (*n* = 1/28) and one surgeon described a specific suture method called 'the bunny stitch' (surgeon S) (*n* = 1/28). It was described as repetitions of three continuous non-locking sutures in the vaginal epithelium followed by a separate interrupted suture that included tissue from the vaginal epithelium and muscularis. Surgical terminology {#Sec11} -------------------- Most surgeons used the term 'fascia' to describe the tissue they were repairing. However, the concept of fascia was not uniform amongst surgeons. Some described fascia as being part of the vaginal skin (Surgeon X), others described it as a separate entity (Surgeon L) and some expressed uncertainty about what the tissue was or what planes they were operating in (Surgeon J)."Surgeon X: It's part of the skin; there are two layers of the fascia.""Surgeon L: Well it's a layer and you can separate it off the skin.""Surgeon J: I can get into a plane quite comfortably...I think I'm leaving most of the 'fascia' on the bladder, but would I describe it as a bladder tissue or a vaginal tissue? I don't know. Can I sit on the fence on that one?" For the purpose of the above descriptions the histological terms were used to ensure clarity of the plane being described \[[@CR5]\]. From observations and later video analysis it was concluded that when surgeons used the term 'fascia' this corresponded to the layer of the vagina which in histological terms is called the vaginal muscularis. There were cases however, where the verbal description of infiltration placement and depth of dissection did not match the investigator\'s view. Despite surgeons\' description of infiltration placement in a specific plane, it was the investigators\' view that just over a third of surgeons placed the fluid in multiple planes rather than in one distinctive plane (Table [2](#Tab2){ref-type="table"}). As outlined in the descriptions above, surgeons used varying terms to describe the caudal and cephalic aspects of the incisions, the lateral extent of dissection and the fascial and skin suture methods. The terms used to describe the repair of the vaginal muscularis included plication, buttressing and repair; in some instances these terms were used interchangeably by the same surgeon. Discussion {#Sec12} ========== Main findings {#Sec13} ------------- Despite the known importance of evidence-based medicine, the VaST study found significant variations exist between surgeons in the techniques they used to perform anterior repairs. Qualitative methods (video observations and interviews) have allowed categorisation of the entire procedure. The degree of variation seen was greater than had previously been described in the literature when simple questionnaires were used \[[@CR6]--[@CR8]\]. The combination of these variable steps results in potentially hundreds of different types of native tissue anterior repair. VaST is the first study to visually categorise surgical technique hence removing errors related to terminology that are inherent in questionnaire-based studies. It is also the first study to relate the techniques in every step of the procedure with outcome \[[@CR9]\]. When performing anterior repairs surgeons did not follow a single method described in the literature \[[@CR10]--[@CR12]\] but instead the techniques used by individuals were a mixture of multiple methods. The categorisation of surgery and development of overarching themes of technique were not possible. The themes developed reflect this and represent the variations seen in the steps of the procedure rather than reflecting the procedure as a whole. The themes of technique include depth of infiltration and dissection, fascial repair method, fascial suture placement, number of fascial repair layers, fascial suture material, fascial suture method, skin trimming, skin suture material and skin suture method. In a number of cases, there was a difference between the investigator\'s view of the techniques observed in real time and on video and the techniques described by the surgeons during interview. The lack of agreed terminology to describe these surgical techniques and anatomical landmarks is likely to be a contributing factor. In this group of surgeons the term fascia was commonly used but on further questioning it was poorly defined. In addition there are aspects of technique that surgeons had more difficulty in describing because they were more subjective, the most significant being the extent of lateral dissection. Previous questionnaire studies will not have been able to capture these tacit issues. Strengths and Limitations {#Sec14} ------------------------- A key strength of the VaST study is that qualitative methods allowed a greater understanding of the variation of surgical techniques used in an anterior repair procedure. A good sample size was gained (30), at participant 27 saturation of themes was reached and 3 further surgeons were recruited for confirmation. The demographic spread of the surgeons was likely representative of UK practice as a whole. Video observations have given a perspective that is not possible to gain from questionnaires or interviews alone. The sequence of observation of surgery followed by interview allowed immediate validation of findings with the surgeons and generated areas for discussion. Filming vaginal surgery proved to be relatively easy and it may be useful to include film material in future surgical trials and training. Surgeons were filmed operating in their own surroundings, hence geographical logistics limited the researcher to one visit per site. This limited the ability to observe variation of technique within the individual surgeon's practice; however this was discussed in the subsequent interviews with the surgeons. This triangulation of methods should have reduced the impact of this limitation. Interpretation {#Sec15} -------------- This study has developed themes of surgical technique for native tissue anterior repair. The categorisation of this procedure was not straightforward because of variations existing in all steps of the procedure, inability of surgeons to articulate aspects of their surgical technique and lack of agreement on terminology. Within the literature there are descriptions of different techniques, which are categorised under the umbrella term of 'anterior repair' \[[@CR6]--[@CR8], [@CR10]--[@CR16]\]. As with previous questionnaire studies \[[@CR3], [@CR6], [@CR8]\] this study identified that most surgeons dissect in a superficial plane and this was frequently combined with a midline repair of the vaginal muscularis. This technique was first described by Kelly in 1913 for the treatment of stress incontinence \[[@CR10]\]. However most surgeons now avoid the first 3 cm of the anterior vaginal wall, which is a significant variation of technique from that described by Kelly. This is likely due to the indication for anterior repair changing from management of urinary incontinence to a surgery for prolapse. When reviewing methodology in randomised control trials making an assessment of prolapse repairs, surgical technique variance should be considered a confounding factor on outcome. In the literature, there are reports of a standard anterior repair or midline plication being performed. However, we know from our study that in clinical practice there is considerable variation in each step of the procedure as there is nothing 'standard' about repair of the anterior compartment. This highlights the importance of studying surgery in pragmatic trials across multiple centres to ensure the external validity of the results. In the future we would suggest more detailed descriptions of surgical technique. Surgery consists of explicit and tacit techniques; explicit ones such as a suture type are easy to define and record but tacit techniques such as the extent of lateral dissection are difficult to assess and describe. Video analysis allows us to view aspects of tacit technique not possible from questionnaires or interviews alone. However, the difficulty of teaching tacit aspects of surgery did not appear to account for all variation in practice recorded in our study because there was an equal amount of variation in both the explicit and tacit steps of the procedure. There is contention within the literature as to the existence of 'fascia' \[[@CR17]\] and this could explain surgeons\' difficulty in articulating the origins of this tissue. The extent of variation in terminology was unexpected but is an important finding because until an agreement is made it will be difficult to conclude which technique is most effective. A previous cadaveric study has categorised the layers of the anterior vaginal wall in histological terms and identified three layers including mucosa (non-keratinised squamous epithelium overlying loose connective tissue), muscularis (smooth muscle, collagen and elastin) and adventitia \[[@CR5]\]. When describing surgery, the use of histological terms could improve descriptions and understanding of the techniques used. It is our interpretation that when performing a superficial dissection, this plane is between the vaginal mucosa and muscularis, a 'deep dissection' between the adventitia and bladder. Future research assessing the excised vaginal tissue from the anterior repair could confirm the histological level of dissection and more accurately define the \'fascia\' we plicate. Video footage has shown a deeper plane to be less vascular and required minimal force, with mainly blunt dissection to develop it. This 'deep dissection' technique has previously been described in the literature, being the level at which mesh/grafts are placed \[[@CR18]\]. It is likely that this technique has been extrapolated from the dissection used for insertion of graft/mesh because only PROSEPCT surgeons in this study who inserted mesh/grafts performed dissection at this depth for native tissue repairs. The themes of surgical technique generated from this study will be used to assess the influence of surgical technique on the outcome of surgery. As well as having an understanding of how the operation varies we need to consider why surgical technique varies and this will be the subject of a further research paper. Conclusion {#Sec16} ========== In the UK there is not a 'standard' native tissue anterior repair. Compared with previous questionnaire studies the use of qualitative methods has given a greater insight into the variation of surgical techniques used to perform native tissue anterior repairs. Furthermore, this study highlights the need to standardise surgical terminology. Further research is required to evaluate which anterior repair techniques are the most effective. A histological study of the excised tissue could more accurately confirm the origin of the tissue that is repaired and generically called fascia. **Presentation of work** UKCS Conference; Aberdeen, 2015 (Oral Presentation) IUGA Meeting, NICE, France 2015 (Oral poster) **Publisher's note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Surgeons who participated in VaST study. E Fairclough: Development of protocol, data collection, data analysis, manuscript writing. J Myers: Development of protocol, manuscript editing. J Segar: Manuscript editing. ARB Smith: Concept, development of protocol, data analysis, manuscript editing. F M Reid: Concept, development of protocol, data analysis, manuscript editing. Ethical approval was gained from the Sunderland Ethics Committee (REC number: 13/NE/0158, 29/05/13).

1. Introduction {#sec1} =============== Adipose tissue-derived mesenchymal stem cells (ASCs) with autologous fat improve the regenerative ability and retention of fat grafts and are increasingly being used for breast reconstruction of breast cancer patients following mastectomy \[[@B1]\]. However, increasing evidence has shown that ASCs may promote the growth and metastasis of breast cancer cells \[[@B2]--[@B5]\], and several studies have demonstrated that ASCs inhibit the growth of breast cancer \[[@B6], [@B7]\]. These contradictory observations may be due to different sources of ASCs, tumor models, and biomarkers for identifying ASCs. To enhance the safety of ASC application in breast reconstruction, it is very important to identify specific biomarkers to distinguish the breast cancer cell growth-promoting ASC subpopulation from other ASC subpopulations that do not enhance the growth and metastasis of breast cancer cells. c-Kit is a protooncogene located at chromosome 4q12, and its encoding protein is a transmembrane receptor tyrosine kinase \[[@B8], [@B9]\]. c-Kit is expressed in many cells of the tumor microenvironment, including mesenchymal, mast, and progenitor cells. In breast cancer, the c-Kit/Kit ligand (KitL) signaling pathway promotes the proliferation, survival, and metastasis of tumor cells \[[@B10]\]. Moreover, the expression level of c-Kit is closely related to triple-negative breast cancer \[[@B11]\]. Recently, it was found that c-Kit^+^ ASCs display a higher differentiation potential in comparison to c-Kit^−^ ASCs \[[@B12], [@B13]\]. These facts suggest that c-Kit may be a potential biomarker that could distinguish the breast cancer cell growth-promoting ASC subpopulation from other ASC subpopulations. The growth and metastasis of tumor cells is dependent on vessel formation in the tumor mass \[[@B14]\]. It has been shown that tumor cells recruit bone marrow-derived vascular endothelial progenitor cells (BM-EPCs) by increasing the expression of hypoxia-inducible factor-1*α* (HIF-1*α*) and vascular endothelial growth factor (VEGF), both of which play an important role in angiogenesis \[[@B15], [@B16]\]. The interaction of EPCs and tumor cells can enhance angiogenesis, which plays a crucial role in the growth and metastasis of tumor cells \[[@B17], [@B18]\]. ASCs have been demonstrated to promote angiogenesis by secreting growth factors within a variety of tumor types \[[@B19]--[@B22]\]. However, since the tumor microenvironment is very complex, whether ASCs differentiate into endothelial-like cells or recruit endothelial cells for vessel formation during tumor angiogenesis remains to be determined. To explore the role and mechanism of c-Kit^+^ ASCs in breast cancer progression, in this study, we established a coculture model of ASCs and breast cancer cells. Furthermore, we analyzed the impact of c-Kit^+^ ASCs on tumor angiogenesis using breast cancer mouse models. 2. Materials and Methods {#sec2} ======================== 2.1. Cell Culture {#sec2.1} ----------------- 4T1 breast cancer cells were purchased from the American Type Culture Collection (Manassas, VA, USA) and cultured in RIPM-1640 medium (Lonza, Walkersville, MD, USA) supplemented with 10% fetal bovine serum (FBS) (Gibco, Grand Island, NY, USA), 2 mM glutamine, 100 U/mL penicillin, and 100 *µ*g/mL streptomycin. 2.2. Animals {#sec2.2} ------------ Four-week-old female nude mice (Balb/c) were obtained from the SLAC Laboratory Animal Corporation (Shanghai, China) and were housed in a specific pathogen-free room. The Animal Committee of Harbin Medical University approved all the experimental protocols and animal handling procedures. All experimental procedures and postoperative animal care were conducted in accordance with the National Institute of Health\'s Guidelines for the Care and Use of Laboratory Animals. 2.3. Preparation of c-Kit^+^ ASCs {#sec2.3} --------------------------------- Six-week-old Balb/c female mice were sacrificed by cervical dislocation, and inguinal fat tissues were dissected. After mincing the tissues into 2-3 mm pieces, they were digested with 200 U/mL collagenase II (Sigma, St. Louis, MO, USA) for 30 min, followed by centrifugation at 1200 rpm for 10 min. The pellets were sequentially filtered through 200 mesh filters and centrifuged at 12,000 rpm for 10 min. The pelleted cells were washed twice with phosphate-buffered saline (PBS) and resuspended in Dulbecco\'s modified Eagle medium (DMEM; Lonza) containing 10% FBS. The cells were grown at 37°C in a humidified atmosphere containing 5% CO~2~, and the medium was changed daily for 2-3 days. After 3 weeks of culture, c-Kit^+^/CD90^+^ cells were isolated by magnetic bead separation and further purified by fluorescence-activated cell sorting. The adipogenic differentiation potential of ASCs was routinely induced for 2 weeks using medium supplementation (1 : 1 DMEM/Hams F-12 containing 3% fetal calf serum, 100 nM insulin, 1 *μ*M dexamethasone, and 0.25 mM 3-isobutyl-1-methylxanthine) and determined using oil red O staining, following standard protocols. 2.4. Preparation of BM-EPCs {#sec2.4} --------------------------- Bone marrow cells were isolated from the femurs of Balb/c female mice and diluted in Histopaque-1083 (Sigma) (7 : 4), immediately followed by centrifugation at 2400 rpm for 25 min at room temperature. The layer of bone marrow cells at the opaque interface was transferred to a tube containing PBS and centrifuged at 1500 rpm for 10 min at 4°C. EPCs from the mononuclear cells were isolated with CD34 and VEGFR2 magnetic bead separation (Miltenyi Biotech Inc., Auburn, CA, USA) and cultured in endothelial cell growth medium-2 (EGM-2; Lonza) at 37°C and 5% CO~2~ in a humidified incubator. Medium was changed daily for 2-3 days. 2.5. Immunofluorescence {#sec2.5} ----------------------- The adherent cells were trypsinized and plated on EZ slides (Millipore, Billerica, MA, USA) for immunofluorescence assays. The cells were labeled with monoclonal rat anti-CD90 antibody (Abcam, Cambridge, MA, USA), followed by donkey anti-rat secondary antibody conjugated with Alexa Fluor 488 (Thermo Scientific, Waltham, MA, USA). For c-Kit staining, cells were incubated with polyclonal rabbit anti-c-Kit (Abcam) antibody, followed by incubation with goat anti-rabbit secondary antibody conjugated with Alexa Fluor 647 (Abcam). After incubation with 4′,6-diamidino-2-phenylindole (Thermo Scientific) for 1 min, the cells were observed under a fluorescence microscope (Olympus, Tokyo, Japan). The percentages of c-Kit^+^ and CD90^+^ cells during each isolation were analyzed using Image J software. 2.6. Direct Coculture of ASCs with 4T1 Cells {#sec2.6} -------------------------------------------- c-Kit^+^ or c-Kit^−^ ASCs were cocultured with 4T1 cells at a ratio of 1 : 1. The cells were plated on gelatin-coated (1% in PBS) 24-well plates and cultured in DMEM at a density of 50,000/cm^2^. As controls, 4T1 cells or ASCs alone were cultured under the same conditions. The cells were incubated for 1--5 days at 37°C with humidified 5% CO~2~. 2.7. Indirect Coculture of ASCs with 4T1 Cells {#sec2.7} ---------------------------------------------- In a 24-well Transwell culture plate (Corning, NY, USA), 3 × 10^4^ 4T1 cells were plated in the bottom chamber and 3 × 10^4^ ASCs were plated in the upper chamber in DMEM. The chambers were incubated for 1--5 days in a 37°C incubator with humidified 5% CO~2~. 2.8. Tube Formation Assay {#sec2.8} ------------------------- Matrigel (BD, Franklin Lakes, NJ, USA) was added to a 96-well plate, 50 *μ*L per well. After incubation at 37°C with 5% CO~2~ for 1 h, ASCs or EPCs (10^4^) in 100 *μ*L of culture medium were added. The cells were incubated for 18 h, and the area of tube formation was recorded using imaging software (Olympus, Tokyo, Japan). 2.9. Cell Viability Assay {#sec2.9} ------------------------- Approximately 3 × 10^3^ 4T1 cells were cultured in culture supernatant from ASCs in 96-well plates. After 1--5 days, 10 *μ*L of cell counting kit-8 (CCK-8) solution was added per well, and the cells were further incubated at 37°C with 5% CO~2~ for 3 h. Then, the absorbance at 450 nm was recorded using a microplate reader. 2.10. Cell Proliferation {#sec2.10} ------------------------ DNA quantification was performed to assess cell proliferation using the Quant-iT PicoGreen dsDNA reagent and kit (Invitrogen, Carlsbad, CA, USA). After culturing for 1--5 days, the 4T1 cells in the direct coculture model were washed two times with PBS, and 200 *μ*L of 0.2% (v/v) Triton X-100/5 mM MgCl~2~ was added to the cells. Following digestion, the samples were centrifuged at 12,000 rpm for 10 min at 4°C. The supernatant (100 *μ*L) was transferred to a 96-well plate, and 100 *μ*L of PicoGreen fluorescence reagent (1 : 200 in Tris-EDTA buffer) was added. Fluorescence was measured in a microplate reader (Nanodrop 3300, Thermo Scientific), and values were calculated using known DNA standards. 2.11. RNA Extraction and Quantitative Real-Time Polymerase Chain Reaction (qPCR) {#sec2.11} -------------------------------------------------------------------------------- Total RNA was extracted using RNeasy Mini Kits (QIAGEN, Düsseldorf, Germany), according to the manufacturer\'s protocol. Total RNA (1 *μ*g) was reverse-transcribed into cDNA using a QuantiTect Reverse Transcription Kit (QIAGEN). The primer sequences used were as follows: mouse c-Kit forward, 5′-CTGACGTGCATTGATCCCGA-3′, reverse, 5′-CTCGTGAGGCCATTGCTGTT-3′; GAPDH forward, 5′-AGGTCGGTGTGAACGGATTTG-3′, reverse, 5′-GGGGTCGTTGATGGCAACA-3′. The qPCR was performed on a CFX900 thermal cycler (Bio-Rad, Hercules, CA, USA) using THUNDERBIRD SYBR PCR Mix (TOYOBO, Osaka, Japan), according to the manufacturer\'s protocol. GAPDH was used as an internal control. Amplification parameters were as follows: one cycle of 50°C for 2 min, 95°C for 3 min, followed by 40 cycles of 95°C for 10 s and 60°C for 30 s. Relative expression was determined using the comparative threshold cycle method (2^−ΔΔCt^). 2.12. Western Blot Analysis {#sec2.12} --------------------------- Total proteins were extracted using a total protein extraction kit (Vazyme Biotech, Nanjing, China), and concentration was measured using a bicinchoninic acid protein assay kit (Beyotime Biotechnology, Shanghai, China). A total of 20 *μ*g of protein was separated by 7.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred onto a polyvinylidene difluoride membrane. The membrane was incubated in blocking buffer containing 5% nonfat dry milk in Tris-buffered saline and Tween 20 (10 mM Tris-HCl, pH 8.0, 100 mM NaCl, and 0.05% Tween) for 1 h at room temperature. After incubating the membrane with primary antibody (rabbit anti-c-Kit antibody diluted at 1 : 3000) overnight at 4°C and horseradish peroxidase- (HRP-) conjugated goat anti-rabbit antibody (1 : 4000, Abcam) for 2 h at room temperature, protein-antibody complexes were visualized with an Enhanced Chemiluminescence Western Blotting Detection Kit (Beyotime Biotechnology, Shanghai, China) and analysis system (Bio-Rad). *β*-Actin was detected by the same method as a loading control. 2.13. Enzyme-Linked Immunosorbent Assay (ELISA) {#sec2.13} ----------------------------------------------- Cell culture supernatant was collected after culturing for 1, 3, and 5 days by centrifugation for 20 min at 1000 rpm. Primary tumor tissues were rinsed in ice-cold PBS to remove excess blood and weighed. The tissues were minced into small pieces and crushed with liquid nitrogen in a mortar. The resulting suspension was sonicated and centrifuged for 5 min at 5000 rpm. The supernatant was immediately subjected to cytokine and chemokine detection with a commercially available ELISA kit (Cloud-Clone Corp., Houston, TX, USA), according to the manufacturer\'s protocols. 2.14. Animal Model {#sec2.14} ------------------ Subcutaneous orthotopic injection in 4-week-old female nude mice (BALB/c) was performed under general anesthesia (1.2% Avertin, 0.1 mL/10 g). The cells, including 10^5^ 4T1 cells, 10^5^ ASCs, and 10^4^ EPCs, were resuspended in 200 *µ*L of PBS/Matrigel and injected into the mammary fat pads of female nude mice alone or EPCs + 4T1 or ASCs + 4T1/EPCs coinjections. In every group (*n* = 5), the tumor size was measured twice/week, and the tumor volume was calculated according to the following formula: tumor volume = 0.5 × (*D*~max~ × *D*~min~^2^). Three weeks after injection, the mice were sacrificed and the primary tumors were removed for further evaluation. 2.15. Immunohistochemistry {#sec2.15} -------------------------- After tumor removal, the fresh tumors were immediately embedded in optimum cutting temperature compound (Sakura, Zoeterwoude, Netherlands) and sectioned (5 mm), followed by staining with hematoxylin and eosin (H&E), CD31 antibody, and oil red O. Briefly, sections mounted on slides were dehydrated in ethanol, rinsed in PBS containing Tween 20 (PBST), and incubated with 0.3% hydrogen peroxide for 15 min. After washing with PBST, sections were blocked by incubation in 3% bovine serum albumin for 30 min, followed by overnight incubation with primary antibody (rabbit anti-CD31 diluted 1 : 50; Abcam). Slides were washed with PBST followed by a 1 h incubation with HRP-conjugated goat anti-rabbit secondary antibody (1 : 4000, Abcam), rinsed in PBST, and exposed to 3,3′-diaminobenzidine (Solarbio, Beijing, China). Then, counterstaining was performed with hematoxylin (Solarbio, Beijing, China). For H&E staining, sections were stained in hematoxylin for 3 min, washed in water, and then exposed for 5 min to eosin (Solarbio). The immunostaining results were analyzed using imaging software (Olympus, Tokyo, Japan). 2.16. Statistical Analysis {#sec2.16} -------------------------- All experiments were performed in triplicate. All data were statistically analyzed using SPSS version 13.0, and graphs were made using GraphPad Prism version 5.0 software. One-way analysis of variance and Newman-Keuls post hoc tests were used to compare variance between groups. A *p* value ≤ 0.05 was considered as a significant difference. Differences were considered highly significant when *p* ≤ 0.01. 3. Results {#sec3} ========== 3.1. Isolation and Characterization of ASCs and EPCs from Mice {#sec3.1} -------------------------------------------------------------- To investigate whether c-Kit^+^ ASCs promote the growth of breast cancer cells, we isolated ASCs from mouse inguinal adipose tissues. The isolated ASCs appeared as a spindle shape, and oil red O staining showed that adipogenic differentiation of sorted ASCs contained lipid drops inside their cytoplasms, a feature of mature adipocytes ([Figure 1(a)](#fig1){ref-type="fig"}). The cells obtained from mouse adipose tissues were mostly CD90^+^ cells and included a c-Kit^+^ subpopulation (Figures [1(b)](#fig1){ref-type="fig"}, [1(c)](#fig1){ref-type="fig"}, and [1(d)](#fig1){ref-type="fig"}). Nevertheless, there were very few cells that were positive for the endothelial progenitor cell marker CD34, and no CD45^+^ subpopulation was found by immunofluorescence staining ([Figure 1(e)](#fig1){ref-type="fig"}). These results indicated that the isolated and expanded cells including c-Kit^+^/CD90^+^ ASCs were not contaminated with endothelial or hematopoietic cells. To assess whether the interaction of ASCs and EPCs promotes breast cancer angiogenesis, we isolated BM-EPCs from mice. The isolated mononuclear cells appeared as round-shaped cells that attached on plates at days 3--5. At days 7--14, the cells demonstrated a cobblestone appearance on gelatin-coated plates, which is a characteristic of EPCs. In methylcellulose media, EPCs demonstrated cell-cluster formation consistent with the ability to form colonies ([Figure 1(f)](#fig1){ref-type="fig"}). These results indicated a successful isolation of EPCs from mouse bone marrow. 3.2. c-Kit^+^ ASCs Promote the Viability and Proliferation of Breast Cancer Cells {#sec3.2} --------------------------------------------------------------------------------- To test the effects of c-Kit^+^ ASCs on the viability and proliferation of breast cancer cells, we cocultured ASCs with 4T1 cells. With a direct coculture model, we found that the mRNA expression of c-Kit was significantly higher in the coculture group than in the c-Kit^+^ ASCs alone group (*p* \< 0.001, [Figure 2(a)](#fig2){ref-type="fig"}). Western blot analysis showed no detectable c-Kit protein in the 4T1 cells alone group, but there was a higher c-Kit protein level in the coculture group than in the c-Kit^+^ ASCs alone group ([Figure 2(b)](#fig2){ref-type="fig"}). However, we did not observe c-Kit expression in the c-Kit^−^ ASCs alone or coculture group, which suggests that 4T1 cancer cells may increase c-Kit expression in c-Kit^+^ ASCs ([Figure 2(b)](#fig2){ref-type="fig"}). To assess the effect of ASCs on 4T1 cell proliferation, we performed an indirect coculture of ASCs with 4T1 cells and determined the viability of 4T1 cells for 5 days. After cell culture for 3 days, the viability of 4T1 cells was significantly enhanced in the coculture with c-Kit^+^ ASCs, compared to the 4T1 cell culture alone group (*p* \< 0.05, [Figure 2(c)](#fig2){ref-type="fig"}). The cell proliferation assay using a Quant-iT PicoGreen kit found that the proliferation of 4T1 cells was significantly increased in the coculture with c-Kit^+^ ASCs, compared to the culture of 4T1 cells alone for 4-5 days (*p* \< 0.01, [Figure 2(d)](#fig2){ref-type="fig"}). These results suggest that c-Kit^+^ ASCs promote the viability and proliferation of breast cancer cells. 3.3. Effect of c-Kit^+^ ASCs on Primary 4T1 Tumor Growth In Vivo {#sec3.3} ---------------------------------------------------------------- ASCs affect tissue regeneration and homeostasis, and c-Kit^+^ ASCs have a higher differentiation potential compared to other subpopulations of mesenchymal stem cells in adipose tissue \[[@B12], [@B13]\]. In addition, tumor growth depends on vessel formation. Nevertheless, our Matrigel assay showed that both c-Kit^+^ and c-Kit^−^ ASCs showed no tube formation ability in comparison to EPCs ([Figure 2(e)](#fig2){ref-type="fig"}). To explore whether c-Kit^+^ ASCs promote breast cancer growth in vivo, we performed subcutaneous coinjection of 4T1 cells with ASCs in combination with EPCs into nude mice and observed the effect of c-Kit^+^ ASCs on 4T1 tumor growth. We found that injection of 4T1 cells alone formed a palpable tumor ([Figure 3(a)](#fig3){ref-type="fig"}). In the coinjection groups, the tumor growth was markedly stimulated in comparison to the injection of 4T1 cells alone (*p* \< 0.01, Figures [3(a)](#fig3){ref-type="fig"} and [3(b)](#fig3){ref-type="fig"}). Furthermore, the tumor volume was significantly increased in the c-Kit^+^ASCs + 4T1/EPCs group, compared with the c-Kit^−^ASCs + 4T1/EPCs group and the EPCs/4T1 group (2275 mm^2^ versus 2052 mm^2^ and 2275 mm^2^ versus 1705 mm^2^, respectively; *p* \< 0.05, Figures [3(a)](#fig3){ref-type="fig"} and [3(b)](#fig3){ref-type="fig"}). Although the tumor volume in the c-Kit^−^ASCs + 4T1/EPCs group was larger than that of the EPCs/4T1 group up to 21 days after injection (2052 mm^2^ versus 1705 mm^2^, Figures [3(a)](#fig3){ref-type="fig"} and [3(b)](#fig3){ref-type="fig"}), the difference was not significant (*p* \> 0.05, [Figure 3(b)](#fig3){ref-type="fig"}). In addition, 14 days after injection, the weights of the nude mice significantly declined by 0.24% of the original weight in the c-Kit^−^ASCs + 4T1/EPCs group (19.14 g versus 19.65 g, [Figure 3(c)](#fig3){ref-type="fig"}). However, there was no significant difference of the weight in each group (*p* \> 0.05, [Figure 3(c)](#fig3){ref-type="fig"}). 3.4. c-Kit^+^ ASCs Promote EPC-Mediated Tumor Angiogenesis {#sec3.4} ---------------------------------------------------------- Tumor angiogenesis is crucial to tumor growth. However, the role of ASCs in tumor angiogenesis is unclear. To explore the role for c-Kit^+^ ASCs in tumor angiogenesis, tumor grafts were analyzed by H&E staining and mouse-specific anti-CD31 staining. Oil red O staining showed lots of mature adipose cells in the tumor tissues of the ASCs + 4T1/EPCs group ([Figure 3(d)](#fig3){ref-type="fig"}); however, H&E and CD-31 immunostaining of tissue sections revealed broad vascularization in the c-Kit^+^ASCs + 4T1/EPCs group, in contrast to the other groups ([Figure 3(e)](#fig3){ref-type="fig"}). Quantification analysis revealed that the microvascular density was significantly higher in the coinjection groups containing c-Kit^+^ ASCs than in the other injection groups (81.3 ± 3.1 vessels/mm^2^ versus 65.0 ± 10.0 vessels/mm^2^ in the c-Kit^−^ASCs + 4T1/EPCs group, *p* \< 0.05; 37.3 ± 4.2 vessels/mm^2^ in the EPCs/4T1 group, *p* \< 0.001; 23.7 ± 5.1 vessels/mm^2^ in the 4T1 group, *p* \< 0.001, [Figure 3(f)](#fig3){ref-type="fig"}), suggesting that c-Kit^+^ ASCs may enhance EPC-mediated tumor angiogenesis of breast cancer. 3.5. c-Kit^+^ ASCs Stimulate the Release of Interleukin-3 (IL-3), Stromal Cell-Derived Factor-1 (SDF-1), and VEGF-A {#sec3.5} ------------------------------------------------------------------------------------------------------------------- ASCs produce several cytokines and chemokines that affect other surrounding cells, especially EPCs, thereby increasing tumor angiogenesis. To explore the molecular mechanisms by which c-Kit^+^ ASCs enhance EPC-mediated tumor angiogenesis of breast cancer, we analyzed the relevant cytokines, chemokines, and angiogenic growth factors. In direct coculture models, ELISA analyses revealed that the release of IL-3 significantly increased in the c-Kit^+^ASCs/4T1 group in comparison to the c-Kit^−^ASCs/4T1 group (1416.7 ± 728.6 pg/mL versus 633.3 ± 208.2 pg/mL, *p* \< 0.05, [Figure 4(a)](#fig4){ref-type="fig"}) and the single culture groups (483.3 ± 175.6 pg/mL in the 4T1 group, *p* \< 0.05; 443.3 ± 172.1 pg/mL in the c-Kit^−^ASCs group, *p* \< 0.05; 560.0 ± 182.5 pg/mL in the c-Kit^+^ASCs group, *p* \< 0.05; [Figure 4(a)](#fig4){ref-type="fig"}). However, there was no significant difference in the SDF-1 (*p* \> 0.05, [Figure 4(b)](#fig4){ref-type="fig"}) and VEGF-A (*p* \> 0.05, [Figure 4(c)](#fig4){ref-type="fig"}) levels among the groups. To extend our observation in vitro to mouse models in vivo, the levels of IL-3, SDF-1, and VEGF-A were assessed in tumor tissues. Interestingly, we found that IL-3 was significantly increased in the c-Kit^+^ASCs + 4T1/EPCs group, compared with the c-Kit^−^ASCs + 4T1/EPCs group (9900.0 ± 141.4 pg/mL versus 7250.0 ± 353.6 pg/mL, *p* \< 0.05, [Figure 4(d)](#fig4){ref-type="fig"}) and the other groups (5900 ± 565.7 pg/mL in the EPCs/4T1 group, *p* \< 0.05; 2500 ± 707.1 pg/mL in the 4T1 group, *p* \< 0.01, [Figure 4(d)](#fig4){ref-type="fig"}). Moreover, the release of SDF-1 significantly increased in the c-Kit^+^ASCs + 4T1/EPCs group in comparison with the c-Kit^−^ASCs + 4T1/EPCs group (15500.0 ± 707.1 pg/mL versus 8150.0 ± 495.0 pg/mL, *p* \< 0.01, [Figure 4(e)](#fig4){ref-type="fig"}) and the other groups (9050.0 ± 777.8 pg/mL in EPCs/4T1 group, *p* \< 0.01; 4400.0 ± 565.7 pg/mL in the 4T1 group, *p* \< 0.01, [Figure 4(e)](#fig4){ref-type="fig"}). In addition, the level of VEGF-A was significantly higher in the c-Kit^+^ASCs + 4T1/EPCs group than in the c-Kit^−^ASCs + 4T1/EPCs group (7650.0 ± 495.0 pg/mL versus 3250.0 ± 353.6 pg/mL, *p* \< 0.01, [Figure 4(f)](#fig4){ref-type="fig"}) and the other groups (4150.0 ± 212.1 pg/mL in EPCs/4T1 group, *p* \< 0.01; 1800.0 ± 282.8 pg/mL in the 4T1 group, *p* \< 0.01, [Figure 4(f)](#fig4){ref-type="fig"}). However, there were no significant differences between the c-Kit^−^ASCs + 4T1/EPCs group and the EPCs/4T1 group in terms of the IL-3, SDF-1, and VEGF-A levels (*p* \> 0.05, Figures [4(d)](#fig4){ref-type="fig"}--[4(f)](#fig4){ref-type="fig"}). 4. Discussion {#sec4} ============= ASCs have been proposed to promote the viability of fat grafts and the efficacy of the procedure after breast cancer surgery \[[@B23], [@B24]\]. However, emerging evidence indicates that ASCs may contribute to the growth and metastasis of breast cancer \[[@B25], [@B26]\]. Conversely, other studies have demonstrated an inhibitory role of ASCs in breast cancer \[[@B6], [@B7]\]. The contradiction warrants the selection of immunophenotypic markers of ASCs in clinical application. Our results demonstrated that c-Kit^+^ cells with characteristics of adipogenesis could be isolated from ASCs and that the c-Kit^+^ subpopulation was very rare, which is consistent with published observations that most stromal vascular fraction cells are c-Kit^−^ \[[@B27]--[@B29]\]. There are distinct expression levels of c-Kit in stem cells/progenitor cells, melanocytes, and mast cells \[[@B30]\]. The c-Kit signaling network supports the proliferation, differentiation, and survival of c-Kit-expressing cells \[[@B31], [@B32]\]. In BRCA1 mutation-associated breast cancers, c-Kit is required for the growth and survival of the tumor cells, and c-Kit activity may be downregulated to allow normal differentiation in adult tissue \[[@B33]\]. At present, to the best of our knowledge, no studies have shown the effect of c-Kit^+^ ASCs on breast cancer progression compared with the other subpopulations of ASCs. In the present study, we showed that the expression of c-Kit increased in the c-Kit^+^ASCs/4T1 coculture group compared with single cultures and that the proliferation of breast cancer cells was enhanced by c-Kit^+^ ASCs after coculture. Moreover, the release of membrane-bound KitL, as an adhesion/survival-promoting molecule for stem cells, depends on IL-3 \[[@B34]\]. IL-3 acts as a nonspecific proinflammatory cytokine and drives cell proliferation by the JAK/STAT and PI3K/AKT pathways \[[@B35]--[@B37]\]. Our results demonstrated that c-Kit^+^ ASCs stimulated IL-3 release both in vitro and in vivo. Taken together, our results suggest that c-Kit may stimulate the proliferation of 4T1 breast cancer cells by promoting IL-3 release. Tumor angiogenesis is crucial to tumor growth \[[@B38]--[@B40]\]. However, we found that the ASCs had no tube formation ability in vitro, and no palpable tumor masses formed in the ASCs injection alone group in mice. A previous study has shown the perivascular properties of ASCs in stabilizing these neovessels and the effect on EPC/endothelial cells by paracrine signaling in vivo \[[@B41]\]; accordingly, we implanted ASCs together with the EPCs that were isolated and cultured in the same conditions in order to eliminate their effect. Interestingly, we observed that c-Kit^+^ ASCs significantly promoted tumor growth in combination with EPCs, compared to the c-Kit^−^ ASCs coinjection group and the other groups, especially at 1 week after injection. Tumor angiogenesis is sustained by complicated cytokine networks among tumor cells and other cells in the tumor microenvironment \[[@B42]--[@B45]\]. We found that c-Kit^+^ ASCs had an intrinsic proangiogenic capability and stimulated IL-3 release, which could promote the proliferation of 4T1 breast cancer cells or the growth of mature adipose cells. However, the tumor microenvironment in solid tumors is hypoxic, and c-Kit has been reported to participate in the formation of the tumor vasculature via promoting HIF-1*α*-mediated VEGF expression \[[@B46], [@B47]\]. Boesiger et al. have shown that, in response to c-Kit/KitL, mouse or human mast cells rapidly release VEGF by degranulation and then sustain the release by secreting newly synthesized proteins \[[@B48]\]. Meanwhile, an increase of angiogenic factors, such as HIF-1*α*, SDF-1, and VEGF, in turn causes activation of matrix metalloproteinase-9 and thus initiates recruitment and mobilization of BM-EPCs in the peripheral circulation \[[@B49]\]. Our results indicate that active c-Kit^+^ ASCs may recruit myeloid cells by SDF-1 for angiogenesis, thereby stimulating EPCs to form vessels followed by the release of VEGF-A, which are crucial factors for the proliferation of breast cancer cells in the tumor microenvironment \[[@B50], [@B51]\]. Accumulating evidence suggests that ASCs may favor tumor progression by autocrine and paracrine signaling in different transplanted and metastatic tumor models. However, the interaction mechanism between ASCs and breast cancer in the tumor microenvironment remains to be determined. One limitation of our current study is the lack of continuous observation and cell tracking to discriminate different cell types and illuminate the specific role of cells in tumor angiogenesis in every group. Moreover, human tissue-derived c-Kit^+^ ASCs are needed to prove the function of c-Kit^+^ ASCs in breast cancer progression. In addition, multicenter clinical outcomes based on an adequate follow-up of breast cancer patients with an autologous fat graft are needed to compare c-Kit^+^ ASCs with other subpopulations of ASCs. In conclusion, c-Kit^+^ ASCs can promote the tumor angiogenesis and growth of breast cancer by recruiting EPCs via a synergistic effect of c-Kit and IL-3. Our findings suggest that identification of different phenotypes of ASCs may be required in reconstructive efforts with stem cell-enhanced fat grafting. Furthermore, c-Kit^+^ ASCs should be avoided for use in breast reconstruction. This work was supported by the Doctorial Innovation Fund of Harbin Medical University. Conflicts of Interest ===================== The authors declare that they have no conflicts of interest. {#fig1} {#fig2} {#fig3} {#fig4} [^1]: Academic Editor: Stephen H. Safe

Data are owned by each centre that contributed to the ARPEC project. The ARPEC steering committee was in charge of approval of proposed analyses for abstracts and publications, delegating this to work package leads, if appropriate. The work package lead for ARPEC point prevalence surveys was Prof. Herman Goossens who together with the Project Co-ordinator Prof. Mike Sharland may be contacted at <arpec@sgul.ac.uk>. Introduction {#sec007} ============ Antibiotics are among the most commonly used medications for hospitalized children \[[@pone.0199878.ref001]\]. On any day, 30% to 60% of children admitted to hospital around the world will receive at least one antibiotic, with many being prescribed multiple systemic antimicrobials \[[@pone.0199878.ref002],[@pone.0199878.ref003]\]. Antimicrobial stewardship interventions can improve antibiotic use in this vulnerable population and are usually implemented at a high level of aggregation, for example at hospital level \[[@pone.0199878.ref004],[@pone.0199878.ref005]\]. It is often desirable to compare the use of antibiotics, especially of last-resort agents, between hospitals or regions to identify outliers and therefore areas for intervention. However, merely comparing the overall volume of use or crude proportions for antibiotics of interest is unlikely to be useful because prescription patterns vary markedly, and this is partially due to differences in patient case-mix \[[@pone.0199878.ref006]--[@pone.0199878.ref011]\]. In many areas of infection control, regression models are used to adjust metrics for differences in patient case-mix \[[@pone.0199878.ref012]--[@pone.0199878.ref014]\]. However, these risk-adjustment models can easily become complex, may be based on specific data that are not widely available and/or comparable, and can require the adoption of extensive, costly data collection processes. Another method is to apply a stratification system and examine exposure within groups of similar patients. An example of this method from another area of medical practice is the Robson classification, which stratifies pregnant women according to simple and widely available clinical characteristics that influence their a priori risk of having a Caesarean delivery \[[@pone.0199878.ref015]--[@pone.0199878.ref017]\]. We examined whether a risk-adjustment model could be developed from readily available variables that would facilitate the fair comparison of statistics from point prevalence surveys (PPS) on the prescribing of antibiotics to children with sepsis/bloodstream infections. We focused on three "pediatric conserve antibiotics" (pCAs) for severe Gram-negative and Gram-positive neonatal and pediatric infections. These antibiotics are part of the newly defined World Health Organization Watch group of antibiotics. This group has been identified to have a higher resistance potential, and should only be used for specific indications or in infections caused by bacteria suspected or proven to be resistant to less broad-spectrum options \[[@pone.0199878.ref018]\]. We evaluated whether available variables enabled the creation of: (i) a risk-adjustment model to fairly compare the prevalence rates across world regions, and (ii) a simple stratification system that identified patient groups who would be expected to have similar exposures due to their characteristics. Materials and methods {#sec008} ===================== Data collection {#sec009} --------------- The study used data collected as part of the Antibiotic Resistance and Prescribing in European Children (ARPEC) project global PPS \[[@pone.0199878.ref003]\]. PPS are simple, standardized tools used widely internationally to collect data on antimicrobial use to facilitate monitoring within centers and countries \[[@pone.0199878.ref019]\]. Participating centers were asked to conduct a one-day cross-sectional survey of antimicrobial prescriptions for inpatients on neonatal and pediatric wards during three periods in 2011/2012 \[[@pone.0199878.ref002],[@pone.0199878.ref003]\]. During each PPS all neonatal and pediatric wards in participating institutions had to be surveyed once within the defined auditing period. All patients present in the wards at 8:00 am, and at least since midnight on the day of the survey, were recorded. For each patient treated with at least one antimicrobial, detailed data on the prescription as well as about the patient were collected according to a standardized protocol. The ARPEC PPS were conducted in 226 participating centers located in 41 countries, which were grouped into continental regions (Africa, Asia, Australia, Europe--East, Europe--North, Europe--South, Europe-West, Latin/South America and North America) according to the UN geoscheme classification \[[@pone.0199878.ref002],[@pone.0199878.ref003]\]. The PPS methodology and data collection approaches have previously been described in detail \[[@pone.0199878.ref002],[@pone.0199878.ref003]\]. During data collection no unique identifiers, such as hospital numbers or dates of birth, were recorded. As the PPS was therefore a completely anonymized audit of antimicrobial prescribing to inpatient neonates and children, formal ethical review was not a requirement. Individual participating centres were asked to ascertain any local requirements for ethical review. By entering data, centres confirmed that they had taken the required steps according to their local and national regulatory and legal requirements. Study population and definition of patient and treatment characteristics {#sec010} ------------------------------------------------------------------------ The study used the records of surveyed patients who were prescribed systemic antibiotics (J01) \[[@pone.0199878.ref020]\] for the most common indication of suspected or definitive sepsis/bloodstream infection \[[@pone.0199878.ref003]\], excluding febrile neutropenia and catheter-related bloodstream infection. A single key infection syndrome was selected as different factors may drive prescribing of antibiotics depending on the type of infection being treated. Relevant prescriptions were identified from the PPS information on "reason for prescription". In terms of antibiotic use, we focused on carbapenems (J01DH), glycopeptides (J01XA) and linezolid (J01XX08). Prescribing of these antibiotics may reflect actual or feared infection caused by resistant organisms, such as extended-spectrum beta-lactamase producing Gram-negative bacteria or methicillin-resistant *Staphylococcus aureus*. The World Health Organization confirms these antibiotics, among others, as key targets for national antibiotic stewardship \[[@pone.0199878.ref018]\]. Our study is limited to the indicated groups and follows the same approach as a recent study evaluating the impact of antimicrobial stewardship on antibiotic prescribing in US children's hospital \[[@pone.0199878.ref021]\]. Exposure to pCAs was defined at the patient level, with a patient classified as exposed if one or more of the antibiotics listed above was prescribed. At the patient level, the ARPEC dataset included information on a patient's age, whether they had any chronic conditions, and the type of ward the patient was on. Data were also collected on the type of prescription (empiric or targeted). Neither the microbiological results for individual patients nor hospital antibiograms were available. Finally, timing of prescription was available as having been issued \>48 hours after hospitalization (hospital-acquired) or ≤48 hours after hospitalization (community-acquired). Any prescription for sepsis/bloodstream infection in the first three days of life was considered neonatal early onset sepsis. Wards were classified as either a neonatal intensive care unit (NICU, all care levels), pediatric intensive care unit (PICU) or other pediatric wards. Patients with any recorded underlying disease from a predefined list including surgical malformations, chronic neurological, gastrointestinal, endocrine, lung and renal disease as well as congenital heart disease, oncologic/hematologic diseases, genetic or metabolic disorders, rheumatological or autoimmune disease and chronic infections were labeled as having underlying disease ([S1 File](#pone.0199878.s001){ref-type="supplementary-material"}). Patients receiving any targeted prescriptions for a sepsis/bloodstream infection (according to the ARPEC protocol based on pathogen identification and/or antimicrobial susceptibility testing) were defined as receiving targeted treatment, even when additional prescriptions were empiric. All other patients were labeled as receiving empiric treatment. Statistical analysis {#sec011} -------------------- Logistic regression was used to assess the association between pCA exposure and the individual patient and treatment characteristics. Age was dichotomized into neonates aged 3 days or younger versus infants aged 4 days or older and children (reflecting clinical differences between early-onset and late-onset sepsis among neonates). We then developed a risk model using multivariable logistic regression. The model was developed by sequentially adding each available patient variable, starting with the variable that had the strongest univariate association and ending with the weakest. A Wald test was used to assess the contribution of an added variable to the model and a p value of 0.05 was used as the threshold for inclusion. Following this, interactions between included variables were explored. The performance of the model was assessed in terms of its calibration and discrimination. Calibration describes the level of agreement between the predicted and observed risks, and was evaluated using the Hosmer-Lemeshow test. Discrimination indicates the ability of a model to distinguish patients with a lower and higher risk of pCA prescription. We evaluated this by using the c-statistic (equivalent to the Area under the ROC curve). The regression model was used to calculate risk-adjusted regional pCA exposure rates. These were derived using indirect standardization, which involved multiplying the ratio of observed/expected exposure rates by the mean exposure rate in the whole cohort \[[@pone.0199878.ref009]\]. Approximate 95% confidence intervals were derived for proportions and indirectly standardized rates using the Wilson Score and Byar's Method, respectively. As a sensitivity analysis, we repeated the above process using a multilevel logistic model, which incorporated a random-intercept term for the subregions as well as the explanatory variables. This accounted for any lack of independence in the data due to patients being clustered within subregions. Finally, a small number of mutually exclusive and comprehensive patient subgroups were defined on the basis that they described clinical situations in which we would expect a patient's chance of receiving pCA to be similar given the seriousness of the situation and the effectiveness of current antiobiotics. We used the same variables that were considered in the risk-model development process because they represented information that is easy to collect and can be standardized. Inspection of the variables identified six patient groups that were expected to be associated with different levels of exposure to pCA: 1. Neonatal early onset sepsis (infants ≤3 days of age): High reported coverage provided by narrow-spectrum regimens \[[@pone.0199878.ref022]\]. 2. Community-acquired sepsis in otherwise healthy infants \>3 days of age and children: Lower levels of colonization and infection by multidrug-resistant pathogens \[[@pone.0199878.ref023]\]. 3. Community-acquired sepsis in infants \>3 days of age and children with underlying disease: Colonization by multidrug-resistant pathogens possible with reported epidemiology similar to hospital-acquired bloodstream infection \[[@pone.0199878.ref023]\]. 4. Empiric treatment of hospital-acquired sepsis in infants and children of any age outside of PICU: Colonization by multidrug resistant pathogens possible, but colonization pressure less than in intensive care \[[@pone.0199878.ref024]\]. 5. Targeted treatment of hospital-acquired sepsis in infants and children of any age outside of PICU: May include patients having been discharged from intensive care to complete treatment after stabilization, therefore likely to partially reflect intensive care epidemiology \[[@pone.0199878.ref025]\]. 6. Hospital-acquired sepsis in infants and children of any age in PICU: Colonization by multidrug-resistant bacteria expected with high colonization pressure in intensive care \[[@pone.0199878.ref025]\]. We examined the ability of these subgroups to reduce the heterogeneity within the patient population using the measures of discrimination and calibration described above. All statistical analyses were carried out using Stata/IC 13.1®, Statacorp, USA. Results {#sec012} ======= Description of cohort {#sec013} --------------------- The complete global ARPEC PPS cohort contained data on 11899 prescriptions on 6499 patients. Among these, there were 2668 prescriptions for sepsis, but limiting the cohort to patients with complete records led to the exclusion of a further 415 prescription records ([Fig 1](#pone.0199878.g001){ref-type="fig"}). The final dataset contained 2253 systemic antibiotic prescriptions for 1281 infants and children, representing 19% of a total of 11899 recorded prescriptions. {#pone.0199878.g001} Overall pCA exposure {#sec014} -------------------- Of the 1281 included patients, 445 patients (34.7%; 159 children ≤30 days of age of which two were ≤3 days of age, 286 children \>30 days of age) were exposed to pCAs. In total, 18.4% (235/1281) were receiving carbapenems, 25.4% (325/1281) glycopeptides and 1.2% (16/1281) linezolid. For each of the patient and treatment characteristics, the proportion of exposed patients varied across the levels of each variable by at least 10%, as shown in [Table 1](#pone.0199878.t001){ref-type="table"}. 10.1371/journal.pone.0199878.t001 ###### Association of key patient characteristics with exposure to pCA antibiotics (group comparisons using Χ^2^ testing). {#pone.0199878.t001g} Total patients with sepsis/BSI \% in group pCA-exposed \% exposed p-value --------------------------------- -------------------------------- ------------- ------------- ------------ ---------- **Age** Neonate ≤3 days of age 123 9.6 2 1.6 p\<0.001 Infant or child \>3 days of age 1158 90.4 443 38.3 **Ward** Pediatric ward 466 36.4 117 25.1 p\<0.001 Neonatal intensive care 635 49.6 219 34.5 Pediatric Intensive care 180 14.1 109 60.6 **Underlying disease** Absent 311 24.3 32 10.3 p\<0.001 Present 970 75.7 413 42.6 **Acquisition of infection** Community 649 50.7 78 12.0 p\<0.001 Hospital 632 49.3 367 58.0 **Type of treatment** Empiric 980 76.5 285 29.1 p\<0.001 Targeted 301 23.5 160 53.2 **Total** 1281 445 34.7 Multivariable logistic regression model for exposure to pCA {#sec015} ----------------------------------------------------------- Each individual patient and treatment characteristic was found to be associated with antibiotic use, and improved the performance of the multivariable logistic model when added ([Table 1](#pone.0199878.t001){ref-type="table"}). There was evidence of an interaction between the variables "ward" and "acquisition of infection" as well as between the variables "underlying disease" and "type of treatment". These separate variables were replaced by variables that captured the combination of categories. [Table 2](#pone.0199878.t002){ref-type="table"} shows the results of the model that takes into account these interactions. Overall, the following were associated with increased odds of pCA exposure: (1) presence of any underlying disease, (2) treatment in PICU; (3) receiving targeted treatment; (4) treatment for hospital-acquired infection. Being ≤3 days old was associated with lower odds of pCA exposure. 10.1371/journal.pone.0199878.t002 ###### Logistic regression results showing adjusted odds ratios for exposure to pediatric reserve antibiotics (pCAs) with 95% confidence intervals. {#pone.0199878.t002g} Group Adjusted OR 95%CI ----------------------------------------------------------------------- ------------- ----------- **Patients according to ward type and acquisition of infection** Non-ICU / community-acquired Ref \- Non-ICU / hospital-acquired 5.0 3.0--8.3 NICU / community-acquired 0.6 0.3--1.1 NICU / hospital-acquired 5.7 3.7--8.8 PICU / community-acquired 4.2 2.2--8.1 PICU / hospital-acquired 12.7 7.3--22.2 **Patients according to underlying disease and type of prescription** No underlying disease / empiric Ref \- No underlying disease / targeted 4.3 1.8--10.0 Underlying disease / empiric 3.8 2.2--6.7 Underlying disease / targeted 7.1 3.9--13.0 **Patients according to age** Neonate ≤3 days of age Ref \- Neonate \>3 days of age, infant or child 16.9 4.0--70.9 This final model demonstrated strong discrimination, with a c-statistic of 0.83. There was also evidence of good calibration (Hosmer-Lemeshow test, p = 0.38, see [Fig 2](#pone.0199878.g002){ref-type="fig"} for calibration plot). {#pone.0199878.g002} Multilevel random-intercept logistic model for exposure to pCA {#sec016} -------------------------------------------------------------- The analysis using the multilevel model gave similar results to the main analysis. We found only modest variation in the random-intercepts of the subregions (variance = 0.21; SE(var) = 0.14) and the coefficients of the explanatory variables were similar to those estimated in the standard model. In addition, the Pearson correlation coefficient between the predicted risks for individuals from the two models was 0.97, with the predictions from the multilevel model producing to almost identical calibration and discrimination figures. Adjusted regional patterns of pCA exposure {#sec017} ------------------------------------------ [Fig 3](#pone.0199878.g003){ref-type="fig"} demonstrates the impact of using the risk model to adjust for differences in patient characteristics on regional pCA exposure levels. Crude regional exposure rates ranged from 10.3% (Africa) to 67.4% (Latin America). After adjustment, there was substantially less variation between the regions, with the adjusted regional exposure rates ranging from 17.1% (Africa) to 42.8% (Latin America). The 95% confidence intervals around adjusted pCA exposure rates indicate that, with the exception of Africa, regional estimates may not differ from the overall cohort mean pCA exposure level once key characteristics have been taken into account. {#pone.0199878.g003} pCA exposure in predefined groups {#sec018} --------------------------------- [Table 3](#pone.0199878.t003){ref-type="table"} shows the characteristics of the six patient groups that were derived from clinical reasoning. 10.1371/journal.pone.0199878.t003 ###### Overall proportions of treated patients within predefined group and the expected rate of exposure to pediatric reserve antibiotics (pCAs). {#pone.0199878.t003g} Patient group Total patients (n) \% in group Exposed to pCAs (n) \% Exposed 95% CI ------- --------------------------------------------------------------- -------------------- ------------- --------------------- ------------ -------------- **1** Neonatal early onset sepsis 123 10% 2 1.6% 0.2 to 5.8 **2** CA sepsis/BSI in otherwise healthy infants and children 251 20% 17 6.8% 4.0 to 10.6 **3** CA sepsis/BSI in infants and children with underlying disease 295 23% 60 20.3% 15.9 to 25.3 **4** Empiric treatment of HA sepsis/BSI outside of PICU 327 25% 162 49.5% 44.0 to 55.1 **5** Targeted treatment of HA sepsis/BSI outside of PICU 173 13% 120 69.4% 61.9 to 76.1 **6** HA sepsis/BSI on PICU 112 9% 84 75.0% 65.9 to 82.7 1281 445 34.7% CA: community-acquired, HA: hospital-acquired, BSI: bloodstream infection, PICU: pediatric intensive care unit. [Table 4](#pone.0199878.t004){ref-type="table"} shows the distribution of patient groups by region. 10.1371/journal.pone.0199878.t004 ###### Distribution of included patients for 6 predefined groups by region. {#pone.0199878.t004g} Patient group Subregion ----------------------- --------------------------------------------------------------- ----------- ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ ------ **1** Neonatal early onset sepsis 17 10.3 12 6.1 10 16.7 43 13.0 18 7.1 18 10.7 0 0 5 11.6 **2** CA sepsis/BSI in otherwise healthy infants and children 38 23.0 24 12.2 14 23.3 64 19.4 70 27.6 28 16.6 1 2.2 4 9.3 **3** CA sepsis/BSI in infants and children with underlying disease 74 44.9 54 27.4 12 20.0 55 16.7 40 15.8 44 26.0 9 19.6 10 23.3 **4** Empiric treatment of HA sepsis/BSI outside of PICU 24 14.6 54 27.4 15 25.0 102 30.9 65 25.6 30 17.8 14 30.4 16 37.2 **5** Targeted treatment of HA sepsis/BSI outside of PICU 11 6.7 27 13.7 4 6.7 47 14.2 35 13.8 30 17.8 12 26.1 5 11.6 **6** HA sepsis/BSI on PICU 1 0.6 26 13.2 5 8.3 19 5.8 26 10.2 19 11.2 10 21.7 3 7.0 Carbapenem exposure 17 10.3 57 28.9 5 8.3 44 13.3 53 20.9 33 19.5 20 43.5 3 7.0 Glycopeptide exposure 12 7.3 59 30.0 12 20.0 74 22.4 75 29.5 56 33.1 19 41.3 17 39.5 Total n 165 197 60 330 254 169 46 43 CA: community-acquired, HA: hospital-acquired, BSI: bloodstream infection, PICU: pediatric intensive care unit. The proportions refer to contributions of each group for the region in question. Overall, nearly 50% of children fell into groups 4 to 6, as they were being treated for hospital-acquired sepsis/bloodstream infection. In terms of the pCA exposure rates, levels were lowest in neonates treated for early onset sepsis (1.6%) and highest in patients with hospital-acquired sepsis on PICU (75.0%). We assessed the performance of this simple classification by using a logistic regression model that included only these six predefined groups. The model had a similar level of performance as the full risk model, with good levels of discrimination (c-statistic = 0.81) and calibration (Hosmer-Lemeshow test, p = 0.813; [Fig 4](#pone.0199878.g004){ref-type="fig"}). The regional distribution of patients may explain very high crude pCA exposure levels in Latin America: Nearly 80% of patients in this region fell into groups 4 to 6 compared with maximally 50--60% in other regions. These patients would be expected to have higher pCA exposure rates than patients in groups 1 to 3. {#pone.0199878.g004} Discussion {#sec019} ========== The data from global point prevalence surveys of inpatient neonatal and pediatric systemic antibiotic prescriptions for sepsis/bloodstream infection revealed large differences in the crude pCA prevalence rates across the regions. But, the interpretation of these differences is hampered by the considerable systematic differences between the regions in the patterns of disease, antimicrobial resistance and population structure. In this study, we demonstrated that having data on a few easily collected variables related to patient and treatment characteristics, it is possible to develop a risk adjustment model to produce adjusted pCA exposure rates, thereby allowing a fairer comparison of regions. In addition, the variables could be combined into a simple patient classification that differentiated various clinical situations in which the expected pCA exposure rates would be expected to differ. While a risk-adjustment approach based on logistic regression is preferable for making comparisons against a standard population, the classification facilitates benchmarking by creating relatively homogenous groups of patients who would be expected to have similar exposure to pCA due to their clinical circumstances. When evaluated, both approaches performed well at discriminating between children in terms of their likely exposure to pCA. Overall, the average pCA exposure rate was high at 35%. Exposure rates to pCA were higher among older children, those on PICU, children with underlying disease, and receiving targeted treatment for hospital-acquired sepsis/bloodstream infection. That targeted treatment was strongly associated with higher pCA exposure may reflect a high rate of resistant bacteria identified in those children with culture-confirmed sepsis/bloodstream infection. Given the reported high rates of antimicrobial resistance in key pathogens globally, but especially in low- and middle-income countries \[[@pone.0199878.ref026]--[@pone.0199878.ref028]\], this is a worrying sign of the prevalence of multidrug-resistant infections, especially among hospital-acquired infections, in this population. Regional crude prevalence rates varied considerably, the lowest and highest differing by a factor of 6.5. After adjustment, the prevalence rates varied by a factor of 2.5, demonstrating that a large proportion of variation arose from differences in the distribution of the measured patient and prescription characteristics. Previous analyses of case-mix adjustment in benchmarking of inpatient antibiotic prescribing have used variables that require detailed knowledge about each patient \[[@pone.0199878.ref007],[@pone.0199878.ref009]\] or detailed hospital-level data \[[@pone.0199878.ref008],[@pone.0199878.ref010],[@pone.0199878.ref011]\]. While models based on these variables may have demonstrated even better discrimination and calibration in this dataset, the value of our study is to demonstrate that pCA prescribing rates from prevalence surveys can be adequately risk-adjusted using easily collected variables. The effectiveness of this approach needs to be replicated in other datasets, and the benefit of including other factors also needs evaluation. Nonetheless, the results underline the importance of focusing on the complete and accurate measurement of important patient-level variables and treatment characteristics during data collection to enable optimal utilization of PPS data. The predefined patient groups based on clinical reasoning proved to have a similar level of performance to the full logistic regression model. In practice, the application of a logistic regression model to inform quality improvement at a hospital level could be challenging because it produces a single composite statistic that describes overall performance. In contrast, a classification-based approach makes it possible to monitor the prevalence of pCA in distinct types of patients, for which the action required to tackle above average rates is likely to be different. This has been found to be a key issue in the development and use of classification systems in other circumstances \[[@pone.0199878.ref015],[@pone.0199878.ref016]\]. The clinical logic underpinning the classification gives it a face-validity that suggests it could be applicable in other situations. But, we recommend that, before it is adopted for use in other infection syndromes and healthcare settings (e.g. adult care), its performance is evaluated further using data collected in that setting. Our analysis has a number of limitations. First, despite this being as far as we are aware the largest neonatal and pediatric antibiotic prescribing PPS database globally, some regions contributed only a small number of patients. With a larger sample size, we would have been able to better estimate true differences in regional pCA exposure rates. Sample size limitations will also impact the application of our approach at hospital-level. Assuming that prescribing practices remain relatively stable, the pooling of data from several PPS may be one approach to overcome small sample sizes. Second, we only included prescriptions that were recorded as being for sepsis/bloodstream infection. Patients in our cohort may have received additional antibiotics for another indication (e.g. lower respiratory tract infection), which we did not include in our evaluation. Whether the same risk factors are associated with pCA exposure in patients treated for other infections needs to be tested. Third, data on the causative organism in targeted treatment were not recorded. We therefore rely on local contributors having correctly identified the recorded treatment as the most suitable narrow-spectrum antibiotic option for the target pathogen. In the future, pCA exposure rates should be interpreted together with information on actual resistance at patient or aggregate levels \[[@pone.0199878.ref029],[@pone.0199878.ref030]\] to gauge whether pCA exposure levels are high in response to high antibiotic resistance rates or are mainly driven by prescriber behavior. Fourth, PPS data provides no information on duration of pCA exposure, which may have an important impact on the volume of pCAs used in a specific setting. Fifth, our analysis would need to be repeated analyzing data from a variety of hospitals. ARPEC PPS participant centers were predominantly tertiary and/or university hospitals, and the relevance of our findings for benchmarking involving smaller secondary hospitals would have to be confirmed. Finally, the cluster sizes of the participating centers were too small to support a multilevel model with centre as the cluster. Instead, we fitted a random-intercept logistic model with subregions to take account into account the hierarchical structure of the data. This did not change the conclusions about each variable and there was excellent agreement between the predictions from the two models. We therefore chose to present the results from the simpler standard logistic model. In addition to conventional case-mix adjustment approaches, predefined patient groups, such as those described in our analysis, enable the generation of aspirational targets for aggregate pCA exposure rates, either in local, regional or national settings. These targets could be based on current average levels of exposures or be based on expert consensus about desirable practice. This would allow the comparison of (i) overall standardized exposure rates; (ii) variations in distribution of patient strata; (iii) variations in exposure rates for specific patient groups. The advantages of this approach is that evaluations of pCA exposure would take into account key characteristics of the patients and infection episodes that are highly likely to influence pCA prescribing decisions and as such reflect justified use of these antibiotics. This may enable identification of specific target areas for intervention, while taking into account that what is appropriate may differ between facilities and/or regions. For such comparisons and target setting treatment and patient characteristics need to be captured, as described in this manuscript and as is standard during point prevalence surveys. Given the good performance of the logistic regression classification model in our analysis, the level of detail for the variables included in the ARPEC PPS may be sufficient for evaluations of childhood antibiotic use. However, additional or different variables are likely to be useful for similar analyses in other patient groups. Case-mix adjustment, preferably using a few easy-to-collect patient and prescription characteristics, is key to accurately and fairly comparing prescribing patterns between health care providers, regional health care administrations and countries. Furthermore, assessing antibiotic exposure rates in a clinically relevant manner within homogenous and easily identifiable patient groups can be a rich source of information about key areas for intervention to improve antibiotic prescribing. Quality of antibiotic prescribing could then be assessed in such patient groups using validated indicators. In this way, interventions that will achieve a safe and reasonable reduction in the use of critically important antibiotics at aggregate level can be defined and evaluated. Supporting information {#sec020} ====================== ###### Excerpt of data collection instructions for coding of underlying disease for ARPEC PPS. (PDF) ###### Click here for additional data file. We thank all partners and collaborators of the ARPEC PPS work package for their data contributions. [^1]: **Competing Interests:**JAB's husband is senior corporate counsel at Novartis International AG, Basel, Switzerland, and holds Novartis stock and stock options. MS chairs the UK Department of Health Expert Advisory Committee on Antimicrobial Resistance and Healthcare Associated Infection (ARHAI). All other authors declare no competing interest. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Introduction {#S0001} ============ Conjunctivitis is a collective term for a diverse group of diseases that are characterized by inflammation of the conjunctiva.[@CIT0001] The most common cause of infectious conjunctivitis is viral infection (\~80% of cases), followed by bacterial.[@CIT0001],[@CIT0002] The noninfectious forms are allergic, mechanical/irritative/toxic, immune-mediated, and neoplastic.[@CIT0001] Of noninfectious conjunctivitis, the allergic form is the most common, affecting approximately 40% of the US population.[@CIT0002] Conjunctivitis can also be classified as acute, chronic, or recurrent,[@CIT0001],[@CIT0002] according to the mode of onset and severity of the clinical response.[@CIT0002] Most cases of viral and bacterial conjunctivitis are considered acute and, less frequently, as hyperacute. The mechanical/irritative/toxic, immune-mediated, and neoplastic forms are typically associated with chronic inflammation,[@CIT0001] and the toxic form can have an acute onset. Allergic conjunctivitis can be chronic, with possible onset in childhood, and present with acute exacerbations related to seasonal factors or contact lens use, for example.[@CIT0001] Conjunctivitis is a highly common presentation in the primary care setting. It is responsible for \~2% of all medical consultations in the United States[@CIT0003] and countless self-referrals to community pharmacies.[@CIT0004] It is difficult to accurately capture the true epidemiology of conjunctivitis because of the high numbers of patients who may self-medicate or delay or decline to seek medical care, and because of inaccurate diagnosis in the primary medical care setting. Approximately 70% of all patients with acute red eye present to primary care and urgent care,[@CIT0005] posing a large economic and social burden.[@CIT0006]--[@CIT0008] The pharmacoeconomic impact of infectious conjunctivitis is noteworthy and includes the cost of repeat diagnoses, referrals, insurance copays, patient medications, and lost productivity associated with absence from work or school. In the United States, acute conjunctivitis affects an estimated 6 million people per year,[@CIT0002] and bacterial conjunctivitis treatment alone costs the United States \$377--857 million annually. To the best of our knowledge, no health-related quality-of-life studies of patients with acute infectious conjunctivitis have been published. The purpose of this narrative review is to provide practical guidance on the differentiation of acute infectious conjunctivitis in the primary care setting and to highlight disconnects between diagnosis and treatment. Challenges of Acute Infectious Conjunctivitis Management and Diagnosis {#S0001-S2001} ---------------------------------------------------------------------- Eye care professionals tend to evaluate acute infectious conjunctivitis cases at secondary care visits. Patients with infectious conjunctivitis are suboptimally managed in primary care due to an underappreciation of the prevalence of viral conjunctivitis, as well as frequent antibiotic prescriptions (\~80%) for infectious conjunctivitis,[@CIT0009] despite guidelines that discourage antibiotic use for minor self-limiting illnesses.[@CIT0010] Antibiotic prescription is a measure of presumed diagnosis of bacterial conjunctivitis, but the rate of clinical accuracy in diagnosing viral conjunctivitis may be \<50%.[@CIT0009],[@CIT0011] Observational evidence indicates that laboratory confirmation isolates bacterial pathogens in just 30--50% of suspected cases.[@CIT0009],[@CIT0011] Misdiagnoses expose patients to unnecessary and ineffective antibiotic treatment. In the Antibiotic Resistance Monitoring in Ocular Microorganisms (ARMOR) surveillance study of 3237 ocular isolates collected from patients with bacterial eye infections (from 72 US centers from 2009 to 2013), nearly one-half had methicillin-resistant staphylococci.[@CIT0012] Although the contribution of topical antibiotics to antibiotic resistance is not easily confirmed,[@CIT0011] current resistance trends should be considered before treating common eye infections with antibiotics.[@CIT0013] Furthermore, prescribing antibiotics in cases of unconfirmed bacterial infection is not compliant with Antibiotic Stewardship initiatives, which are part of primary care and pediatric practice protocols.[@CIT0014] In addition to the risk of antibiotic resistance, misdiagnosis can result in recurrent cases of infectious conjunctivitis, particularly among pediatric patients, or in cases that progress to serious ocular and extra-ocular complications. There are several challenges to accurately diagnosing acute infectious conjunctivitis. First, clinical ambiguity exists between the acute viral and bacterial as well as the allergic forms, which can confound diagnosis.[@CIT0011] Discrimination between viral or bacterial origins of infectious conjunctivitis based on historical, nonspecific, clinical signs and symptoms, such as type of discharge, is difficult and not supported by evidence-based diagnostic criteria.[@CIT0011],[@CIT0015] Routine bacteriologic examinations are neither typically performed nor practical in the management of clinically suspected acute bacterial conjunctivitis,[@CIT0015] except in neonates.[@CIT0001] Culturing should also be considered for immunocompromised patients or for patients with hyperacute cases of infectious conjunctivitis. The misdiagnosis of infectious conjunctivitis is compounded by the view that all cases should be treated with topical ophthalmic antibiotics, even though most are self-limiting.[@CIT0016],[@CIT0017] Social factors, such as school policies or pressure to reduce absence from work, may also influence general practitioners to prescribe topical antibiotics.[@CIT0011] Lastly, the level at which patients enter the health care system can impact the accuracy of the diagnosis. Guidelines and criteria for diagnosing and treating acute infectious conjunctivitis based on natural history and etiology are available.[@CIT0001] However, greater awareness of and adherence to differential diagnosis should be kept in mind to improve outcomes in patients presenting with suspected acute infectious conjunctivitis.[@CIT0017] Epidemiology and Clinical Presentation {#S0001-S2002} -------------------------------------- Discriminating features of viral, bacterial, and allergic acute conjunctivitis are summarized in [Table 1](#T0001){ref-type="table"}.[@CIT0001],[@CIT0002],[@CIT0018]Table 1Discriminating Clinical Features of Suspected Acute ConjunctivitisTypeConjunctivalItchingDischargeLymphadenopathyAssociated Fever and Sore ThroatViralFollicularMinimalWateryCommon (\~50%)Common Adenoviral HSV VZVBacterialPapillaryMinimalPurulentUncommonOccasionally NongonococcalMucopurulent GonococcalHyperpurulentChlamydialFollicularMinimalMucopurulentCommonNoAllergicPapillary with chemosisSevereWateryNoneNoMucoid[^1][^2] ### Viral Conjunctivitis {#S0001-S2002-S3001} Adenoviruses are the cause of most (65--90%) cases of viral conjunctivitis, while the herpes simplex virus (HSV) is the cause of 1.3--4.8% of all cases of acute conjunctivitis.[@CIT0002] Other viruses that are associated with conjunctivitis include varicella (herpes) zoster virus (VZV) and *Molluscum contagiosum*.[@CIT0001] Adenoviruses are nonenveloped viruses that are relatively resistant to disinfection.[@CIT0019] Adenoviral conjunctivitis is highly contagious, due in part to the capacity of the virus to survive in a desiccated state at room temperature for several weeks.[@CIT0020] Transmission risk is 10--50%, and the infection may spread via personal contact or indirectly through shared items.[@CIT0002] Adenoviral conjunctivitis is self-limiting, with improvement of signs and symptoms within 5--14 days.[@CIT0001] Often, the condition presents as unilateral in early phases and eventually spreads to both eyes. Adenovirus serotypes are associated with different types of ocular infection,[@CIT0021] including pharyngoconjunctival fever and epidemic keratoconjunctivitis (EKC; patient photos shown in [Figure 1](#F0001){ref-type="fig"}). Pharyngoconjunctivitis typically manifests bilaterally with fever and enlargement of the periauricular lymph nodes.[@CIT0002] EKC is more severe because of the adverse effect it can have on visual acuity; it is characterized by watery discharge, swelling, and redness, and involvement of lymph nodes on both sides of the neck.[@CIT0002],[@CIT0021] EKC is highly contagious, and asymptomatic patients who have contracted the disease may unknowingly spread the virus.[@CIT0021]Figure 1Case photographs of adenoviral conjunctivitis. (**A**) gross examination of acute adenoviral conjunctivitis; (**B**) bilateral AdenoPlus-positive EKC; (**C**) high magnification of the conjunctiva of positive EKC. Photos provided by Scott Hauswirth and Elizabeth Yeu. The patients have given permission for publication of these photos.**Abbreviation:** EKC, epidemic keratoconjunctivitis. A complication of EKC is the formation of pseudomembranes on the palpebral conjunctiva in the early and later stages of the condition.[@CIT0021],[@CIT0022] Manual removal of these membranes causes minor bleeding but no damage to the underlying epithelium, and may prevent symblepharon formation in cases with severe inflammation.[@CIT0022] Multifocal subepithelial infiltrates may develop in the cornea within 7--10 days after the onset of clinical signs of infection.[@CIT0023] These can persist for weeks or even years in some cases.[@CIT0022] If left untreated, subepithelial infiltrates can cause scarring in the cornea, leading to irregular astigmatism and loss of visual acuity.[@CIT0024] Topical corticosteroids or steroid-sparing agents such as cyclosporine eye drops can be used to reduce the risk of scarring;[@CIT0022] however, topical corticosteroids should be used with caution, especially in later stages of recovery from adenoviral infection.[@CIT0021],[@CIT0025] It should be noted that subepithelial infiltrates may also be a feature of chlamydial conjunctivitis.[@CIT0001] As no medications are currently approved for adenoviral conjunctivitis, treatment is mainly supportive (cold compresses, chilled artificial tears)[@CIT0001] with instruction on infection control, as viral shedding continues after onset.[@CIT0001],[@CIT0025] There are few virucidal agents in preclinical development. Povidone-iodine (PVP-I) was under investigation as a potential treatment because of its broad-spectrum antimicrobial activity.[@CIT0001] An ophthalmic suspension composed of PVP-I 0.6% and the corticosteroid dexamethasone 0.1% (PVP-I 0.6%/DEX 0.1%) has been evaluated in a Phase 2 trial.[@CIT0026] Results from this study showed that PVP-I 0.6%/DEX 0.1% treatment of patients with acute adenoviral conjunctivitis significantly improved clinical resolution and eradication of adenovirus in the study eye.[@CIT0026] Another compound, N-chlorotaurine, has broad-spectrum microbicidal activity due to its nonspecific reaction mechanism, and was investigated for the treatment of viral conjunctivitis.[@CIT0027] A 1% solution of N-chlorotaurine was well tolerated in patients with EKC and shortened the course of disease compared with gentamicin eye drops in a randomized phase 2b study.[@CIT0027] Topical ganciclovir 0.15% ophthalmic gel is approved in the United States for acute herpetic keratitis only.[@CIT0001] Prior HSV infection is a predisposing factor for HSV conjunctivitis, and reactivation may be triggered by stress, trauma, ultraviolet light exposure, or other viral illnesses.[@CIT0001] HSV conjunctivitis is usually unilateral and characterized by bulbar conjunctival injection, thin and watery discharge, and mild follicular reaction of conjunctiva.[@CIT0001],[@CIT0002] Distinctive signs are vesicular eyelid lesions and pleomorphic or dendritic epithelial keratitis on the cornea or conjunctiva.[@CIT0001] As these signs on the cornea or conjunctiva are often present in the absence of eyelid or dermal lesions, they are best seen by instillation of vital dyes such as sodium fluorescein, lissamine green, or rose Bengal to assess the ocular surface. HSV conjunctivitis usually subsides within 4--7 days unless complications occur. Topical and/or oral antivirals are recommended to shorten the duration of illness.[@CIT0001],[@CIT0002] Because topical corticosteroids can potentiate HSV,[@CIT0001],[@CIT0002] they should be avoided,[@CIT0001] particularly in the presence of corneal or conjunctival epithelial disease.[@CIT0001],[@CIT0002] HSV can also cause EKC, which is difficult to distinguish clinically from adenoviral conjunctivitis.[@CIT0028] Active chicken pox and prior exposure to chicken pox or shingles are predisposing factors for VZV conjunctivitis.[@CIT0001] This condition is commonly unilateral in presentation. Distinctive signs of VZV conjunctivitis include inflammation of the eyelids, conjunctiva, cornea, or uvea.[@CIT0002] Uveitis may present with increased intraocular pressure, which in some cases may rise to dangerous levels.[@CIT0029] VZV conjunctivitis usually abates in a few days.[@CIT0001] A combination of oral antivirals and topical corticosteroids can be used with caution to manage VZV conjunctivitis,[@CIT0001],[@CIT0030] as topical antivirals alone are not effective in treating VZV conjunctivitis.[@CIT0001] ### Bacterial Conjunctivitis {#S0001-S2002-S3002} Approximately 32% of acute infectious conjunctivitis cases are caused by bacteria, according to an evaluation of the prevalence of positive bacterial culture in adult patients who presented to their general practitioner with a red eye and either (muco)purulent discharge or glued eyelids.[@CIT0009] Bacterial conjunctivitis is typically unilateral but can sometimes affect both eyes.[@CIT0001] The staphylococcal species are the most common pathogens for bacterial conjunctivitis in adults, followed by *Streptococcus pneumoniae* and *Haemophilus influenzae*.[@CIT0002] In children, the bacterial disease is responsible for 50--75% of conjunctivitis cases, often caused by *S. pneumoniae, H. influenzae*, or *Moraxella catarrhalis*. The hyperacute form of bacterial conjunctivitis is caused by either *Neisseria gonorrhoeae* or *Chlamydia trachomatis*. Acute bacterial conjunctivitis is characterized by a rapid onset of conjunctival redness and mucopurulent discharge.[@CIT0001],[@CIT0002] Symptoms may also include burning and itching, though these are less common.[@CIT0031] Acute bacterial conjunctivitis is self-limiting, resolving within 7--10 days.[@CIT0002] Use of broad-spectrum topical antibiotics should be delayed,[@CIT0001],[@CIT0002],[@CIT0010] and no evidence demonstrates the superiority of any topical antibiotic agent.[@CIT0001] Methicillin-resistant *Staphylococcus aureus* (MRSA) conjunctivitis is the cause of an estimated 3--64% of ocular staphylococcal infections[@CIT0002] and requires management by an ophthalmologist with antibiotics that are active against MRSA, such as besifloxacin ophthalmic suspension.[@CIT0032] The rapid onset and progression of hyperacute bacterial conjunctivitis (often caused by *N. gonorrhoeae*) warrants immediate ophthalmic management to prevent corneal involvement and potentially perforation.[@CIT0001] Symptoms include purulent discharge, diminished visual acuity, eye tenderness, and swollen lymph nodes. The increasing incidence of gonorrhea infection was recognized in 2015 as an urgent public health threat in the United States.[@CIT0033] From 2013 to 2017 the rate of reported gonorrhea cases increased by 67% (555,608 cases in 2017),[@CIT0034],[@CIT0035] with an 18.6% increase between 2016 and 2017.[@CIT0034] Gonococcal ocular infection can present in neonates after vaginal delivery by infected mothers.[@CIT0001] Treatment includes intramuscular ceftriaxone for adults, children, and neonates, and oral azithromycin or doxycycline for adults.[@CIT0002] The Centers for Disease Control and Prevention treatment guidelines changed in 2015 to support the use of dual therapy with ceftriaxone and azithromycin due to the increased resistance of *N. gonorrhoeae* to the antimicrobials previously used for treatment.[@CIT0036] Bacterial conjunctivitis lasting more than 4 weeks is considered chronic.[@CIT0002] Chlamydia is the cause of 1.8--5.6% of all acute conjunctivitis cases.[@CIT0002] Chlamydial conjunctivitis primarily has a unilateral presentation with concurrent genital infection[@CIT0001] and presents with conjunctival redness, lymphoid follicle formation, or mild purulent discharge.[@CIT0001],[@CIT0002] A distinctive sign of this infection in adults is bulbar conjunctival follicles, which are absent in neonates/infants.[@CIT0001] Chlamydial conjunctivitis is typically spread through sexual contact.[@CIT0001] Chlamydia is the most common sexually transmitted disease in the United States, with 1,708,569 cases reported in 2017.[@CIT0037] From 2013 to 2017, the number of reported cases of chlamydia in the United States increased by 22%.[@CIT0038] In neonates, the eyes can be infected after vaginal delivery by infected mothers.[@CIT0001] Treatment includes oral azithromycin or oral doxycycline in adults, and erythromycin in neonates. ### Allergic Conjunctivitis {#S0001-S2002-S3003} Most cases of allergic conjunctivitis are attributable to seasonal allergies and present with bilateral symptoms.[@CIT0001] Mild to severe itching and redness are the key symptoms of allergic conjunctivitis,[@CIT0018] which can present with a watery or mucoid discharge.[@CIT0001] Treatment includes supportive measures (cold compress, artificial tears), topical or oral antihistamines, and mast cell inhibitors. A short course of topical corticosteroids can be used in select cases in which corneal involvement and herpetic infection have been ruled out; both conditions could worsen with steroid use.[@CIT0001],[@CIT0002] Differential Diagnosis {#S0001-S2003} ---------------------- To satisfy the American Academy of Ophthalmology Preferred Practice Pattern guidelines that treatment of conjunctivitis is directed at the root cause,[@CIT0001] physicians should manage patients by taking a systematic approach ([Figure 2](#F0002){ref-type="fig"}), considering the epidemiologic spectrum of infectious conjunctivitis, patient history, clinical exam, and---if needed---diagnostic tests.Figure 2Differential diagnosis algorithm for suspected acute conjunctivitis. Reproduced with permission from *JAMA*. 2013;310(16):1721--1729.  Copyright©2013 American Medical Association. All rights reserved.[@CIT0002] ### Clinical Examination {#S0001-S2003-S3001} Some initial diagnosis is necessary to exclude any potentially serious cause of acute red eye. Historical features for evaluation include pain or photophobia, which can be signs of bacterial keratitis, anterior uveitis, or acute angle-closure glaucoma.[@CIT0039] Associated symptoms such as upper respiratory tract infection, or known exposure to others who have presented with this, suggest a viral form of conjunctivitis.[@CIT0039] Patient history of ocular trauma should be considered as a potential alternative cause of inflammation. A corneal abrasion resulting from a foreign body or exposure to the sun or ultraviolet light can present as red eye.[@CIT0039] Red, painful and watery eyes due to corneal or conjunctival foreign bodies can masquerade as acute and chronic forms of conjunctivitis. Presence of a foreign body is elucidated via history, and by careful biomicroscopic exam. In those who wear contact lenses, the risk of corneal involvement and bacterial keratitis is high; therefore, these individuals should be referred to an eye care provider.[@CIT0002],[@CIT0039] Some signs and symptoms of acute infectious conjunctivitis are mimicked by dry eye disease such as hyperemia,[@CIT0032] grittiness and stinging.[@CIT0040] Adhering to the triage questions and risk factor analysis recommended by the Tear Film & Ocular Surface Society Dry Eye Workshop II (TFOS DEWS II)[@CIT0040] aids the exclusion of dry eye disease, although it may coexist with acute infectious conjunctivitis owing to the high prevalence of both conditions. Of the immune-mediated forms of conjunctivitis (eg, ocular mucous membrane pemphigoid, graft-versus-host disease, Stevens-Johnson syndrome), generally early and milder manifestations of immune-mediated conjunctivopathy may share symptomatology with acute infectious conjunctivitis because of the presence of nonspecific complaints (eg, redness, tearing, clear discharge) and conjunctival injection.[@CIT0001] However, these types of conjunctivitis often differ from viral conjunctivitis in their time course; they are generally protracted and more easily confused with a chronic process rather than an acute episode. Medication history will also be informative. For example, anticoagulants are associated with subconjunctival hemorrhage, and topiramate is associated with angle-closure glaucoma. Patients requiring steroids may be at risk of more severe disease or other pathology, and referral to an eye care provider is recommended.[@CIT0002] The clinical examination should also include an assessment of the lymph nodes (swollen submandibular and preauricular lymph nodes). Lymphadenopathy, such as an enlarged, tender preauricular lymph node, is more common in viral versus bacterial conjunctivitis.[@CIT0031] The physical eye examination should inspect for visual acuity, discharge type, corneal opacity, abnormal pupil size or shape, eyelid swelling, and orbit protrusion or asymmetry.[@CIT0001],[@CIT0002],[@CIT0031] Referral to an ophthalmologist or eye care professional should be performed immediately for cases that involve visual loss, moderate to severe pain, severe purulent discharge, or any corneal involvement.[@CIT0001],[@CIT0002] Useful information can be obtained with a simple penlight in the absence of a slit lamp, especially as it pertains to measurement of the anterior chamber depth and the ruling out of angle closure.[@CIT0002] Viral conjunctivitis can vary in presentation, but tearing, clear to white discharge, redness, and bilaterality are very common. Viral conjunctivitis usually starts with one eye, with the second eye to follow within a few days.[@CIT0015] Bacterial conjunctivitis is strongly suggested by glued eyelids, lack of itching, and lack of history of conjunctivitis.[@CIT0009] A diagnostic index for bacterial conjunctivitis by van Weert et al suggests that younger adult patients with conjunctivitis who present without glued eyelids at awakening are less likely to have conjunctivitis with a bacterial origin; this risk increases with age and presentation of two glued eyes.[@CIT0011] A slit lamp provides a more detailed appearance of the conjunctival epithelium that enables differentiation between various types of conjunctival inflammation (follicular, mainly infectious; papillary, mainly allergic; pseudomembranous, severe EKC; eosinophilic, allergic). Fluorescein staining in conjunction with slit lamp examination, highlights patterns of corneal fluorescein uptake and exposes anterior chamber inflammation.[@CIT0039] HSV infection typically appears as branching lesions with intensely stained end bulbs, while highly branched lesions without end bulbs are typical of VZV infection.[@CIT0039] ### Diagnostic Tests {#S0001-S2003-S3002} Viral diagnostic tests are not used routinely in a primary care setting, but access to these diagnostics has increased and may provide some guidance in obtaining a more accurate diagnosis. Polymerase chain reaction (PCR) for measurement of adenoviral DNA from tear fluid has become more common, but use of the technology is still hindered by costs of equipment and training.[@CIT0028] Viral cell culture with confirmatory immunofluorescence assay is time consuming and rarely performed.[@CIT0028] The availability of an in-office rapid test, such as AdenoPlus (Quidel, San Diego, CA, USA), to detect adenoviruses is a promising approach to obtain a timely diagnosis and reduce unnecessary use of ophthalmic antibiotics.[@CIT0006],[@CIT0028],[@CIT0041],[@CIT0042] In addition to the health benefits, use of a rapid and accurate point-of-care test for adenovirus detection could reduce societal costs associated with acute conjunctivitis.[@CIT0006] AdenoPlus specifically detects adenoviral serotypes through antigen-based microfiltration sampling.[@CIT0028],[@CIT0042] Observed sensitivity of the test ranges from 39.5 to 93%.[@CIT0028],[@CIT0041],[@CIT0042] Sambursky et al noted that testing conditions may cause an artificial inflation of the sensitivity,[@CIT0028] and Holtz et al suggested verifying negative results by real-time PCR.[@CIT0042] Also, clinical judgement may supersede the result of this test vis-à-vis treatment, in part because no available ophthalmic agent is indicated for the treatment of viral conjunctivitis. Conjunctival cultures should be obtained for all diagnoses of suspected infectious neonatal conjunctivitis.[@CIT0001] Bacterial cultures are recommended for patients of all ages if purulent infectious conjunctivitis recurs, is severe, or becomes chronic.[@CIT0001] In places where culture plating is less common, ESwab (Copan Diagnostics, Murrieta, CA, USA) offers a convenient method of sample preparation for bacterial testing.[@CIT0043] ESwab is a liquid-based multipurpose collection and transport system that is suitable for automation, Gram stains, and traditional culture.[@CIT0043] Also, the ESwab system maintains testing viability of bacteria at room temperature and for ≤48 hrs.[@CIT0043] Suspected cases of adult and neonatal chlamydial conjunctivitis can be confirmed by diagnostic testing using an immunofluorescent antibody test and enzyme-linked immunosorbent assay.[@CIT0001] Testing ocular samples by PCR is not currently approved by the US Food and Drug Administration. Gram or Giemsa staining of conjunctival cytology smears is recommended for diagnosing infectious neonatal conjunctivitis or gonococcal conjunctivitis in any age group.[@CIT0001],[@CIT0044] In cases of conjunctivitis unresponsive to therapy, conjunctival biopsy in conjunction with immunofluorescent staining can be considered.[@CIT0001] Conclusions {#S0002} =========== Diagnosis of acute infectious conjunctivitis is complicated because of the signs and symptoms that overlap with other inflammatory conjunctival conditions; therefore, likelihood for misdiagnosis remains high. Misdiagnosis of viral conjunctivitis as bacterial conjunctivitis may result in inappropriate use of antibiotics, which in turn contributes to increased antibiotic resistance. Additionally, misdiagnosis may lead to impaired quality of life and lost productivity due to poor outcomes. Adherence to the American Academy of Ophthalmology evidence-based treatment guidelines and greater educational efforts directed at both patients and physicians could improve the management of acute infectious conjunctivitis. Prescription of antibiotics should be delayed until a bacterial etiology is confirmed, and educational pamphlets should be distributed to general and emergency care practitioners. Development of novel agents with broad-spectrum antimicrobial activity may offer potential new approaches without the drawback of resistance.[@CIT0026],[@CIT0027],[@CIT0045] The authors thank Dr. Terry O'Brien for his helpful discussion and comments at the initial stage of this manuscript. The authors also thank Daniella Babu, PhD, of Excel Medical Affairs, who provided medical writing assistance, funded by Shire, a Takeda company. Funding {#S0003} ======= Medical writing support was funded by Shire (a Takeda company). Shire (a Takeda company), reviewed the manuscript for medical accuracy only. Disclosure {#S0004} ========== EY has served as a consultant for Allergan, Bausch & Lomb, JJV, Novartis, Shire (a Takeda company), TearLab, and TissueTech, during the conduct of the study. EY has served as a consultant for Alcon, Aurea Medical, Avedro, Avellino, BioTissue, Beaver Visitec, BlephEx, Bruder, CorneaGen, Dompe, EyePoint Pharmaceuticals, iOptics, Glaukos, Guidepoint, LENSAR, Kala Pharmaceuticals, Merck, Mynosys, Ocular Science, Ocular Therapeutix, Ocusoft, Omeros, Oyster Point Pharmaceuticals, Science Based Health, Sight Sciences, Sun, Surface, TopCon, TearLab Corporation, and Zeiss, outside the submitted work. SH has served as an advisor or investigator for Allergan, Alcon, Avedro, BioTissue/TissueTech, Dompe, Eyevance, EyePoint Medical, Johnson & Johnson, Kala, NuSight Medical, Ocular Therapeutix, Quidel, Shire (a Takeda company), Science Based Health, Sight Sciences, Sun, Tear Solutions, Tear Restore, and TissueTech. The authors report no other conflicts of interest in this work. [^1]: **Note:** Data from these studies.[@CIT0001],[@CIT0002],[@CIT0018] [^2]: **Abbreviations:** HSV, herpes simplex virus; VZV, varicella (herpes) zoster virus.

INTRODUCTION ============ Nowadays, the internet is an essential part of everyday life, as it enables the fluidity necessary for the circulation of information and global communication.[@B1] ^,^ [@B2] Although there are benefits to this, a new pathology called internet addiction has occurred.[@B1] ^,^ [@B3] ^,^ [@B4] As such, a fine line separates essential access to work, the teaching-learning process, social communication, and information searching with the compulsive and pathological use of the internet. Therefore, internet dependency is a new and growing problem in public health, requiring greater understanding by the scientific community in order to propose community interventions. Several factors are associated with internet addiction, such as sociodemographic aspects related to internet use and psychosocial habits[@B5] ^,^ [@B6] ^,^ [@B7]. Some studies have reported that types of leisure activities done in one's free time, in addition to the multiple activities performed online, are predisposing factors for internet dependence.[@B8] ^,^ [@B9] Specifically, systematic reviews have indicated several adverse effects of internet addiction on mental health, such as attention deficit hyperactivity disorder, depression, hostility and low self-esteem.[@B1] ^,^ [@B5] ^,^ [@B6] ^.^ [@B7] Data from the Brazilian National Household Sample Survey indicate that, in Brazil, 49.4% of people are connected to the internet, and this is mostly concentrated in the 15-17-year-old age group.[@B10] Thus, the aim of this study was to investigate the prevalence of internet addiction and associated factors in high school students in the municipality of Rio Branco, in the Brazilian state of Acre. METHOD ====== This is a population-based, cross-sectional study with high school students in the municipality of Rio Branco, conducted in the months of May and June 2015. The study population consisted of 20,476 students enrolled in 37 high schools in the urban area of Rio Branco. To determine sample size, an expected prevalence for internet dependency of 12% was considered, as observed in high school students in China.[@B8] Precision was set for a sampling error of 0.03 and a level of confidence of 95% was set. With these criteria, the study required a sample of 1,182 students. To protect the investigation from the effects of non-response, 18% was added to the sample size, taking into account operational restrictions, such as the number of researchers and the time available for data collection. The final sample size was estimated at 1,391 students. The selection of students took place in three stages. First, five schools were selected by simple random sampling. Three schools were public and two were private. This number of schools was based on operational restrictions for data collection. In the second stage, classrooms from each school were chosen by allocation proportional to size, and students were chosen by classroom in the third stage. As inclusion criteria, students aged between 14 and 18 years and 11 months, who were regularly enrolled in a high school institution in the city of Rio Branco were considered. The exclusion criteria were pathologies that would make it impossible for students to participate in data collection without assistance, given the importance of privacy in answering the questions. The study was approved by the Research Ethics Committee of the Universidade Federal do Acre under protocol number 39594914.8.0000.5010. Before beginning data collection, the parents and adolescents signed a free and informed consent form. A structured and self-answered questionnaire was applied to students with questions regarding demographic (gender and age), social and family (marital status, being an only child, number of true friends and number of teenagers at home), leisure activities (physical activity; going to the movies and theater, going out to dance at nightclubs and shows; and reading newspapers, magazines or books), parental control and computer usage time. To obtain information on physical activity, the self-administered physical activity checklist was applied, and then validated by Farias Junior et al.[@B11] Using this instrument, the types of physical activity and the number of minutes engaging in physical activity per week were verified. This variable was categorized into: equal to or greater than 300 minutes, between 150 and 299 minutes, from 1 and 149 minutes, and does not practice physical activity. According to the protocol suggested by Petroski,[@B12] body weight was measured using a digital portable scale with a maximum capacity of 150 kg and a sensitivity of 50 g, while height was measured using a portable stadiometer with a maximum extension of 2 subdivided into 0.1 cm. The variable body mass index (BMI) was used as recommended by the World Health Organization,[@B13] with excess weight (sum of overweight and obese individuals) determined using BMI values for age equal to or greater than 1 Z score. The test for internet addiction was verified by the internet addiction test (IAT),[@B14] which was validated and adapted into the Portuguese language.[@B2] ^,^ [@B15] The test consists of 20 questions on a Likert scale ranging from 1 (rarely) to 5 (always), for the individuals to fill out themselves. Individuals with a score equal to or greater than 70 points were considered to be internet dependent. Statistical analysis of the data was performed using the Stata program, version 10 (Stata Corp. College Station, TX, United States). The associated factors were identified in two stages. First, the independent variables that showed associations with internet dependence with p≤0.20 by the Wald test in simple logistic regression were selected to compose the final multiple model. Subsequently, using multiple logistic regression, using the step-by-step procedure with retrograde elimination, variables with p≤0.05 were selected to compose the final multiple model. The variables with a -value between 0.05 and 0.10 remained in the model as adjustment variables. RESULTS ======= Of the 1,391 adolescents contacted, there were four losses due to refusal to participate in the study. Thus, 1,387 adolescents were considered in the analysis, of which 76.8% were from public schools and 23.2% were from private schools. Of these, 53.1% were female, 44.5% were between 14 and 15 years old, and 55.5% were between 16 and 18 years old. The overall prevalence of internet addiction was 10.6%. Male students showed a lower prevalence (6.9%) of internet dependence in contrast to female students (13.9%). However, no statistical difference was identified between the ages for internet dependence. The variables sex, true friends, use of a computer in the bedroom, times using a computer in the middle of the week and on the weekend, going to the movies or theater, going out to dance in nightclubs or *shows*, reading the newspaper, magazines or books, and practicing physical activity were candidates for the final multiple model ([Tables 1](#t1){ref-type="table"} and [2](#t2){ref-type="table"}). Table 1Prevalence (%) and Odds Ratio of internet dependence according to demographic, social and family variables and body mass index in high school adolescents, Rio Branco, Acre, Brazil.n\*%ORp-valueSexMale6506.91Female73713.92.18\<0.001Age14-15 years61611.2116-18 years77110.20.900.567Only childNo1,19910.51Yes16911.21.070.772True friends6 or more friends4118.511 to 5 friends78111.91.450.073No Friends7712.91.600.217Another teenager at homeYes84410.01No51511.41.150.421BMIEutrophic1,04610.51Overweight30411.51.100.621[^2][^3] Table 2Prevalence (%) and Odds Ratio of internet dependence according to variables of leisure activities, parental control and time spent using computers in high school adolescents, Rio Branco, Acre, Brazil.n\*%ORp-valueUsing a computer in their roomNo1,0419.11Yes33115.71.850.001Parental control over computer useYes5229.91No85211.01.120.532Computer time during the weekUp to two hours a day1,3299.81More than two hours a day5829.33.79\<0.001Computer time during the weekendUp to two hours a day1,0778.41More than two hours a day31018.32.44\<0.001Going to the movies or theaterRarely or never54911.11Monthly6198.50.740.144Weekly20615.01.410.142Going out to dance at nightclubs or showsRarely or never9748.21Monthly24712.91.660.022Weekly13822.43.23\<0.001Reading newspapers, magazines or booksRarely or never41912.61Monthly31411.40.890.627Weekly6418.70.660.041Physical activity\> 300 minutes5917.91150 to 299 minutes1888.51.070.8071 to 149 minutes19710.61.380.243Does not practice physical activity32317.02.37\<0.001[^4][^5] In the analysis of factors associated with internet dependence ([Table 3](#t3){ref-type="table"}), girls were 1.84 times more likely to be dependent on the internet than boys. Adolescents who spent more than two hours on the computer on weekdays and on the weekend showed associations of 2.39 and 2.08, respectively, with regard to internet dependence. Adolescents who did not practice physical activity were 2.27 times more likely to be dependent on the internet when compared to those who practiced more than 300 minutes per week, while adolescents who went out weekly to dance at nightclubs or shows were 3.32 times more likely to be dependent on the Internet than those who rarely or never frequented such places. The habit of reading weekly was presented as an adjustment variable in the final model of internet dependence. Table 3Multiple model of internet addiction in high school adolescents, Rio Branco, Acre, Brazil, described in Odds Ratio and 95% confidence interval.OR95%CIp-valueSexMale1Female1.841.21 - 2.800.004Computer time during the weekUp to two hours a day1More than two hours a day2.391.18 - 4.850.015Computer time during the weekendUp to two hours a day1More than two hours a day2.081.36 - 3.180.001Physical activity\> 300 minutes1150 to 299 minutes1.210.64 - 2.300.5431 to 149 minutes1.240.68 - 2.280.469Does not practice physical activity2.271.42 - 3.640.001Reading newspapers, magazines or booksRarely or never1Monthly1.080.66 - 1.770.744Weekly0.690.44 - 1.070.099Going out to dance in nightclubs or *shows*Rarely or never1Monthly1.741.08 - 2.790.021Weekly3.321.99 - 5.56\<0.001[^6] DISCUSSION ========== In this cross-sectional population-based study in the Brazilian Amazon region with high school teenagers, a prevalence of 10.6% internet dependence was identified. Girls showed a higher prevalence of dependence in contrast to boys. The factors associated with internet dependency were use of a computer in free time for more than two hours a day, physical inactivity, and going out to dance in nightclubs and shows. The high prevalence of internet addiction observed here may have been stimulated by the digital inclusion policy, which made free internet access available in various public places in the city of Rio Branco, including schools. Kuss et al.[@B16] suggest that unrestricted accessibility to the internet may contribute to internet dependence. Investigations on internet addiction have shown different prevalence levels in several countries. This variability in prevalence may be due to the numerous measurement instruments and cutoff points used to demarcate levels of dependence. In addition, the inequality in access to the internet in various locations, due to cultural diversity, may have contributed to this variation in the prevalence of internet dependence.[@B17] Therefore, caution should be used when comparing the prevalence in studies with different methods. However, considering only research that used a cutoff point equal to or greater than 70 points in the IAT, the prevalence of internet dependence verified in the present study was higher in relation to that of students in South Korea (2.3%)[@B18] and Lebanon (4.2%).[@B19] In modeling the associated factors, the female gender presented a greater association with internet dependence. Controversially, other studies show that boys are more dependent than girls.[@B20] ^,^ [@B21] ^,^ [@B22] However, the level of internet use among females has increased in recent years.[@B23] In the present study, the time spent on a computer both during the week and over the weekend showed an association with internet dependence. Analogous information was verified in the city of Shanghai, with respect to the hours spent online on weekdays and weekends.[@B8] According to studies in China and Turkey, adolescents who spent more than two hours a day on a computer were more likely to be dependent on the internet compared to those who spent less time.[@B9] ^,^ [@B24] Thus, spending more than two hours a day on a computer with Internet access is considered to be sedentary behavior and has the potential to have a negative impact on health.[@B25] Therefore, controlling adolescents' time spent on computers is an effective method for preventing internet addiction. Engaging in physical activity has been recommended as an important public health strategy to improve the health of children and adolescents. Nowadays, there is a decrease in physical activity and leisure time caused by computerization.[@B26] ^,^ [@B27] In conjunction with the results demonstrated in this study, research has identified an association between the lowest levels of physical activity and internet addiction in adolescents.[@B28] ^,^ [@B29] ^,^ [@B30] Also Babey et al.[@B28] state that teenagers are likely to replace physical activity in their spare time with time on the computer. The teenagers who go out to dance in nightclubs or shows showed greater dependence on the internet. This result corroborates the argument of Zhou et al.,[@B9] who inferred that internet usage contributes to maintaining or increasing social commitments. Usually, teenagers attend shows and nightclubs as part of a social time. Thus, it is believed that they experience real time in connection with the virtual space of the internet, using it for their personal promotion in search of self-affirmation. As an adjustment variable in the modeling, the weekly habit of reading a newspaper, magazine or book is an indicator of protection against internet dependence. In this sense, Sasmaz et al.[@B24] affirmed that the construction of a healthy living environment through the control of computer use and internet access, and the promotion of book reading, favor the prevention and treatment of internet addiction in adolescents. There are some limitations in this study that should be highlighted. First, the cross-sectional design used in the present study made it impossible to identify the causality of factors associated with internet dependence. Second, the data were obtained from a self-administered questionnaire by the adolescents, with no complementary information from parents or other informants. In conclusion, the present study showed a prevalence of 10.6% for internet addiction in a sample of high school adolescents in Acre. Female adolescents were more susceptible to internet addiction when compared to male adolescents. Time spent on the computer and physical inactivity were major factors for internet dependence. From the leisure time activities, the habit of reading proved to be indicative of being protective for internet dependence. On the other hand, attending shows and nightclubs was positively associated with internet dependence. In view of this evidence, for the prevention of internet addiction in schoolchildren, implementing strategies aimed at reducing internet time and computer use at school and in a social-family context are recommended. Concomitantly, encouraging the participation in physical activities and in reading newspapers, magazines and books is suggested. The research project was funded by the Research Support Foundation of the State of Acre (Fundação de Amparo à Pesquisa do Estado do Acre - FAPAC) (015/2013). [^1]: The authors declare no conflict of interests. [^2]: \*The total n does not add up to 1,387 for all variables because some questions did not have answers. [^3]: OR: Odds Ratio; BMI: body mass index. [^4]: \*The total n does not add up to 1,387 for all variables because some questions did not have answers. [^5]: OR: Odds Ratio. [^6]: OR: Odds Ratio; 95%CI: 95% confidence interval.

1. Introduction {#sec1-ijerph-14-00012} =============== Worldwide, *Legionella* is an opportunistic pathogen of public health concern \[[@B1-ijerph-14-00012],[@B2-ijerph-14-00012]\]. It is the causative agent of Legionellosis which includes Legionnaires' disease, an atypical pneumonia, and Pontiac fever, an acute febrile illness \[[@B3-ijerph-14-00012],[@B4-ijerph-14-00012]\]. As such, it is responsible for both nosocomial and community-acquired infections \[[@B5-ijerph-14-00012]\]. One of the primary sources of Legionellosis has been identified as potable water systems \[[@B6-ijerph-14-00012],[@B7-ijerph-14-00012],[@B8-ijerph-14-00012]\]. In the U.S. from 2009--2010, 58% of all drinking water-related outbreaks were caused by *Legionella* \[[@B9-ijerph-14-00012]\]. Across Europe, in 2010, there were 6305 notified cases of Legionnaires' disease reported to the European Legionnaires' Disease Surveillance Network (ELDSNet) \[[@B1-ijerph-14-00012]\]. In the U.S. there has been a significant increase in the incidence of Legionellosis from 0.39 cases per 100,000 in 2000 to 1.36 per 100,000 in 2011 (which equates to a total of 4202 cases) \[[@B10-ijerph-14-00012]\], although estimates state that that this could actually be as high as 50,000 cases, with many going undiagnosed \[[@B8-ijerph-14-00012]\]. It has also been reported that the annual health care cost of *Legionella* infection in the U.S. is over \$430 million \[[@B11-ijerph-14-00012]\]. Opportunistic pathogens are likely to become increasingly significant to public health as a consequence of the global aging population \[[@B12-ijerph-14-00012]\]. It is estimated that in the next five years, there will be more people over the age of 65 than there will be under the age of five \[[@B13-ijerph-14-00012]\]. This will mean that a greater percentage of the population will be considered high risk \[[@B14-ijerph-14-00012]\]. Consequently, opportunistic pathogens linked to potable water distribution systems have been identified as an emerging waterborne disease problem of public health significance \[[@B15-ijerph-14-00012]\]. This identifies the need to discuss and evaluate the guidelines for the control of *Legionella* in water distribution systems \[[@B8-ijerph-14-00012]\]. Historically, an issue that has been subject to some debate is the role of routine testing of water systems for the monitoring of *Legionella* \[[@B14-ijerph-14-00012]\]. The lack of correlation between test results and human health risk has been previously acknowledged \[[@B7-ijerph-14-00012],[@B16-ijerph-14-00012]\]. Additionally, it has been identified that there is the potential for overreliance on test results at the detriment of risk management strategies \[[@B14-ijerph-14-00012]\]. This commentary will explore current literature to present the argument that routine testing of water systems for *Legionella* should be removed from all guidelines. Instead, the presence of the pathogen should be assumed and appropriate control strategies should be identified and managed accordingly. 2. The Presence of *Legionella* in Potable Water Sources {#sec2-ijerph-14-00012} ======================================================== There have been numerous studies which have linked potable water to outbreaks of Legionellosis \[[@B6-ijerph-14-00012],[@B14-ijerph-14-00012],[@B17-ijerph-14-00012],[@B18-ijerph-14-00012],[@B19-ijerph-14-00012],[@B20-ijerph-14-00012],[@B21-ijerph-14-00012],[@B22-ijerph-14-00012],[@B23-ijerph-14-00012]\]. However, there have also been numerous studies which have detected the pathogen from potable water not associated with a specific outbreak; this includes studies from the U.S. \[[@B24-ijerph-14-00012],[@B25-ijerph-14-00012]\], The Netherlands \[[@B26-ijerph-14-00012],[@B27-ijerph-14-00012],[@B28-ijerph-14-00012]\], Germany \[[@B29-ijerph-14-00012],[@B30-ijerph-14-00012]\], Sweden \[[@B31-ijerph-14-00012]\], Israel \[[@B32-ijerph-14-00012]\], Australia \[[@B33-ijerph-14-00012]\], Latvia \[[@B34-ijerph-14-00012]\], Italy \[[@B35-ijerph-14-00012]\], Spain \[[@B36-ijerph-14-00012]\], UK \[[@B37-ijerph-14-00012]\], Croatia \[[@B38-ijerph-14-00012]\], France \[[@B39-ijerph-14-00012]\], Iran \[[@B40-ijerph-14-00012]\] and China \[[@B41-ijerph-14-00012]\]. A recent study by Donohue \[[@B42-ijerph-14-00012]\] collected 272 water samples from public and private cold water taps across the U.S. and found that 47% were positive for *L. pneumophila* Sg1 using qPCR. Another study in Australia used qPCR to consistently detect *Legionella* from two different potable water distribution pipelines sampled four times over the year \[[@B33-ijerph-14-00012]\]. The evidence presented in these studies suggests that *Legionella* is ubiquitous in potable water distribution systems \[[@B28-ijerph-14-00012],[@B43-ijerph-14-00012]\]. 3. Interpreting *Legionella* Test Results {#sec3-ijerph-14-00012} ========================================= Currently, there are numerous *Legionella* qMRA models but there is no consensus with regards to the concentration that will cause Legionellosis \[[@B44-ijerph-14-00012],[@B45-ijerph-14-00012],[@B46-ijerph-14-00012],[@B47-ijerph-14-00012],[@B48-ijerph-14-00012]\]. Infectious doses are based on animal models or back-calculated from exposure estimates during outbreaks \[[@B44-ijerph-14-00012],[@B49-ijerph-14-00012],[@B50-ijerph-14-00012],[@B51-ijerph-14-00012]\]. Attempts to calculate infectious doses are further complicated by variations in *Legionella* virulence based on strain type, life cycle and environmental conditions \[[@B52-ijerph-14-00012],[@B53-ijerph-14-00012],[@B54-ijerph-14-00012],[@B55-ijerph-14-00012]\], as well as the disparity of illness as a consequence of exposure to *Legionella* \[[@B45-ijerph-14-00012]\]. This includes the differences between infectious mechanisms and outcomes of Legionnaires' disease compared to Pontiac fever \[[@B14-ijerph-14-00012],[@B56-ijerph-14-00012],[@B57-ijerph-14-00012]\]. Also, there is potential for exposure to *Legionella* to cause no illness but instead an asymptomatic increase in *Legionella* antibodies \[[@B58-ijerph-14-00012]\]. 4. Limitations with the Standard Method for *Legionella* Detection {#sec4-ijerph-14-00012} ================================================================== The International Standard ISO 11731 describes the standard culture method for the isolation and enumeration of Legionella from environmental samples \[[@B59-ijerph-14-00012],[@B60-ijerph-14-00012]\]. However, this standard culture method is time-consuming and fraught with limitations \[[@B61-ijerph-14-00012],[@B62-ijerph-14-00012],[@B63-ijerph-14-00012],[@B64-ijerph-14-00012]\]. The main limitation is that the culture cannot detect viable but non-culturable (VBNC) *Legionella* which has been shown to be induced by numerous factors commonly found in potable water systems \[[@B45-ijerph-14-00012],[@B56-ijerph-14-00012]\]. This includes the presence of disinfection chemicals, low nutrients, high temperatures and low oxygen \[[@B65-ijerph-14-00012],[@B66-ijerph-14-00012],[@B67-ijerph-14-00012],[@B68-ijerph-14-00012]\]. Additionally, a study by Borges et al. \[[@B63-ijerph-14-00012]\] utilized the standard detection method and identified the occurrence of false positives as a consequence of the misidentification of *Chitinophagaceae*. However, despite the occurrence of false positives, Borges et al. \[[@B63-ijerph-14-00012]\] concluded that, overall, the culture method underestimated *Legionella* populations. This is supported by a desktop study that collated all published data testing environmental samples for *Legionella* concurrently using culture and qPCR and found that 72% of samples were positive using PCR compared to 34% using culture \[[@B62-ijerph-14-00012]\], although it is important to note that qPCR overestimates as it detects both live and killed *Legionella* and the actual numbers are likely somewhere in between \[[@B62-ijerph-14-00012],[@B69-ijerph-14-00012]\]. The inaccuracy with the standard detection method was also demonstrated by a pilot study conducted in the U.S. for the Environmental *Legionella* Isolation Techniques Evaluation (ELITE) Program. This study sent environmental samples to 20 different U.S. laboratories for the detection of *Legionella* and compared results to those conducted at the Centers for Disease Control and Prevention (CDC) reference laboratory. Of these, 37% of variable samples (containing either low concentrations of *Legionella* in pure culture or a mix culture containing a high ratio of heterotrophs to *Legionella*) were identified incorrectly as negative compared to the CDC reference laboratory results. Additionally, for samples identified as positive the mean concentration was underestimated by 1.25 logs CFU/mL. This study concluded that: "The large enumeration error observed suggests that the need for remediation of a water system should not be determined solely by the concentration of *Legionella* observed in a sample since that value is likely to underestimate the true level of contamination" \[[@B70-ijerph-14-00012]\]. To further confound these limitations is the lack of standardized protocols detailing sampling frequency and the selection of sampling sites \[[@B14-ijerph-14-00012]\]. 5. Guidelines {#sec5-ijerph-14-00012} ============= The U.S. Environmental Protection Agency (EPA) first acknowledged that potable water distribution systems presented a major source of *Legionella* in 1985 \[[@B71-ijerph-14-00012]\]. More recently in Australia, the EnHealth Guidelines for *Legionella* control in the operation and maintenance of water distribution systems in health and aged care facilities state that "health and aged care infrastructure managers need to be aware that even a well-managed water supply from a water service provider does not guarantee Legionella-free water" \[[@B72-ijerph-14-00012]\]. In the UK, the Department of Health's Health Technical Memorandum 04-01: Safe water in healthcare premises states, "There is a strong likelihood of low concentrations of Legionella existing in all open water systems including those of building services" \[[@B73-ijerph-14-00012]\]. In recognition of the ubiquitous nature of *Legionella,* current recommendations from the U.S. Centers for Disease Control (CDC) do not include routine testing for *Legionella* as a monitoring strategy or to indicate the need for decontamination \[[@B8-ijerph-14-00012]\]. However, they do state that environmental testing for *Legionella* is useful to validate the effectiveness of control measures utilized within the risk management strategy \[[@B74-ijerph-14-00012]\]. The problem with this arises with the presence of VBNC *Legionella* which may give misleading results regarding the success of a control measure \[[@B62-ijerph-14-00012]\]. For example, the U.S. EPA \[[@B64-ijerph-14-00012]\] cites numerous studies which demonstrate the effectiveness of monochloramine disinfection against *Legionella*. However, it is challenging to interpret these results given that recent studies demonstrate that monochloramine actually induces VBNC *Legionella* \[[@B65-ijerph-14-00012],[@B66-ijerph-14-00012]\]. This issue was identified by Turetgen \[[@B65-ijerph-14-00012]\], who stated, "These VBNC cells are undetectable by standard enumeration methods \[[@B29-ijerph-14-00012]\]; this could lead to an underestimation of the real number of Legionellae present in the sample and an overestimation of the efficacy of disinfectants." The European Working Group for *Legionella* Infections (EWGLI)Technical Guidelines 2011 state that monitoring for *Legionella* should be carried out monthly for the first 12 months and then quarterly for hot water systems. In systems where control levels of treatment are not being achieved, sampling should be conducted weekly until the system is under control. If samples return a positive *Legionella* count over 1000 CFU/L, then an immediate review of control measures should be conducted and a risk assessment needs to be carried out to identify any remedial actions \[[@B75-ijerph-14-00012]\]. The issue with this message is that there is the potential for it to be misinterpreted by managers, who may think that this suggests a review of control measures is only required when *Legionella* is detected. Additionally, the U.S. Occupational Safety and Health administration provides guidelines for remedial action based on quantitative *Legionella* counts (10 CFU/mL requires prompt cleaning and biocide treatment, but 100 CFU/mL also requires additional steps to prevent employee exposure) \[[@B76-ijerph-14-00012]\]. This is quite misleading given that there is limited evidence to demonstrate that an increase in the concentration results in an increased risk to public health, given that the infectious dose is influenced by numerous variables and the lack of reliability or reproducibility of testing methods \[[@B45-ijerph-14-00012],[@B70-ijerph-14-00012]\]. A study from 1987 actually demonstrated the high likelihood of *Legionella* presence in cooling towers and potable water and suggested that given this, there should be an epidemiological association with an environmental source established before intervention methods are taken \[[@B77-ijerph-14-00012]\]. 6. Control Mechanisms {#sec6-ijerph-14-00012} ===================== The abolishment of routine testing for *Legionella* combined with a change in attitude to consider potable water as a constant source of *Legionella* would place an emphasis on maintaining control mechanisms. This includes maintaining temperature control and disinfection residuals, as well as preventing stagnant water or warm water causing significant biofilm formation (through dead-legs or heat exchange due to pipe proximity or limited pipe insulation) \[[@B64-ijerph-14-00012],[@B72-ijerph-14-00012],[@B74-ijerph-14-00012]\]. Additionally, the assumption that the water may be constantly contaminated with *Legionella* identifies the importance for additional control mechanisms to prevent exposure to high-risk patients (i.e., transplant patients or intensive care patients) such as the use of point-of-use filters \[[@B72-ijerph-14-00012],[@B78-ijerph-14-00012]\]. 7. Conclusions {#sec7-ijerph-14-00012} ============== Currently, the U.S. CDC recommendations do not include routine environmental testing for *Legionella.* Given the uncertainties associated with the relevance of environmental monitoring to human health risk and the limitations with the sampling/testing methods, routine environmental sampling should be removed from all guidelines. The high chance of false negative results could potentially cause managers to underestimate the risks or to overestimate a control mechanism. Given that numerous studies have demonstrated the ubiquitous nature of *Legionella* in potable water, an alternative is to assume the pathogen's presence and focus on the management of appropriate control measures and protecting high-risk populations. The abolishment of routine monitoring will prevent managers from overreliance on test results and instead focus on risk management strategies. The author declares no conflict of interest.

Correction to: Protein Cell 10.1007/s13238-019-00668-8 {#Sec1} ====================================================== In the original publication the bands in Fig. 1J and Fig. 2B were not visible. The correct versions of Fig. [1](#Fig1){ref-type="fig"}J and Fig. [2](#Fig2){ref-type="fig"}B are provided in this correction.Figure 1EGFR activation promotes TLR4 phosphorylation and cell surface expression of TLR4 in response to LPS. (A and B) BMDM were treated with LPS (1 μg/mL) for 6, 12, or 24 h in the presence or absence of pretreatment of PD or TAPI-1. (A) Flow cytometry analysis of cell surface TLR4 intensity in BMDM. (B) Flow cytometry analysis of cell surface TLR4 intensity in BMDM. (C and D) WT (C57BL/6) mice were treated with LPS (10 mg/kg, i.p.). In some groups, mice were pretreated with erlotinib (100 mg/kg, gavage administration) at 30 min prior to LPS i.p. Peritoneal lavage fluids were collected at 24 h after LPS treatment and peritoneal macrophages were identified with F4/80. TLR4 intensity on the surface of peritoneal macrophage was analyzed by flow cytometry. (E and F) BMDM isolated from WT and *EGFR*^−/−^ mice were treated with LPS (1 μg/mL) *in vitro* for 1 h followed by flow cytometry analysis of cell surface TLR4 intensity. (G and H) WT (C57BL/6) and *EGFR*^−/−^ mice were treated with LPS (10 mg/kg, i.p.) for 24 h. Peritoneal lavage fluids were collected, and peritoneal macrophages were identified with F4/80. TLR4 intensity on the surface of peritoneal macrophage was analyzed by flow cytometry. (I) Western blot analysis of phosphor-TLR4 in BMDM treated with LPS (1 μg/mL) for 30 min with or without PD168393 (PD, 10 μmol/L) pretreatment for 30 min. (J) Western blot analysis of phosphor-TLR4 in *EGFR*^−/−^ BMDM treated with LPS (1 μg/mL) for 30 min. (K--N) HEK293 cells were transfected with *TLR4*, *MD2*, *CD14*, *EGFR*, or *TLR4* mutant for 48 h, with treatment of LPS (1 μg/mL) for 30 min or 24 h. (K) Diagram of the TLR4 phosphorylation site mutated plasmid. (L) Western blot analysis of the phosphor-TLR4 and phosphor-EGFR in transfected HEK293 treated with LPS for 30 min. (M and N) Flow cytometry analysis of cell surface TLR4 intensity in transfected HEK293 treated with LPS for 24 h. (O and P) BMDM were treated with LPS (1 μg/mL) for 30 min with or without PD168393 (PD) pretreatment for 30 min. (O) Immune-staining of TLR4 and EGFR in BMDM. (P) Co-immunoprecipitation of TLR4 with EGFR in BMDM. (Q) Immune-staining of TLR4 and GM130 in BMDM treated with LPS (1 μg/mL) for 24 h with or without PD168393 pretreatment for 30 min. (R) Immune-staining of TLR4 and GM130 in *EGFR*^−/−^ BMDM treated with LPS (1 μg/mL) for 24 h. All images and flow cytometric plots are the representatives from at least 4 experiments. The graphs depict mean ± SD of four to six experiments or mice. \**P* \< 0.05 as compared with control group; †*P* \< 0.05 as compared with the time-matched LPS alone groupFigure 2Rab5a-mediated concurrent internalization of TLR4 and EGFR results increased cell surface expression of the receptors. (A and B) BMDM were treated with LPS for 6, 12, or 24 h in the presence or absence of pretreatment of PD168393 (PD) for 30 min. (A) Real time PCR analysis of Rab5a expression. (B) Western blot analysis of Rab5a expression. (C and D) BMDM transfected with si-NC and si-Rab5a for 48 h were treated with LPS (1 μg/mL) for 24 h. Flow cytometry analysis of cell surface TLR4. (E--H) BMDM were treated with LPS (1 µg/mL) for 1 h or 24 h, with or without clathrin inhibitor chlorpromazine (CPZ 12.5 μmol/L) or PD168393 (PD 10 μmol/L) pretreatment for 30 min. Flow cytometry analysis of cell surface TLR4 at 24 h or after LPS. (I and J) BMDM cells transfected with si-NC or si-Rab5a for 48 h followed by LPS treatment (1 μg/mL) for 1 h. Flow cytometry analysis of cell surface TLR4. (K and L) WT and *Rab5a*^−/−^ BMDM were treated with LPS (1 μg/mL) for 1 h or 24 h. Flow cytometry analysis of cell surface TLR4 at 1 h or 24 h after LPS. (M--O) BMDM were treated with LPS (1 μg/mL) for 1 h with or without PD168393 (PD) pretreatment for 30 min. (M) Immune-staining of TLR4 and EEA1 in BMDM. (N) Immune-staining of TLR4 with Rab5a. (O) Co-immunoprecipitation between TLR4 and Rab5a in BMDM. All flow cytometric plots are the representative from at least 4 experiments. The graphs depict mean ± SD of four to six experiments or mice. \**P* \< 0.05 as compared with control group; †*P* \< 0.05 as compared with the time-matched LPS alone group

Background {#Sec1} ========== Atypical antipsychotics (AAPs) are an effective treatment for many types of mental illnesses. According to treatment guidelines, antipsychotics are recommended for schizophrenia treatment \[[@CR1]--[@CR3]\], for the treatment of bipolar disorders \[[@CR4], [@CR5]\], and in some cases, as adjunct therapy for major depressive disorder (MDD) \[[@CR6], [@CR7]\]. Although effective, AAPs are often associated with treatment-emergent adverse events (TEAEs), which can be highly burdensome and can affect quality of life and medication adherence \[[@CR8]--[@CR10]\]. Accordingly, treatment guidelines recommend that physicians modify treatment regimens based on patients' response and ability to tolerate side effects \[[@CR1]--[@CR7]\]. This report focused on schizophrenia and MDD as two groups that could be anticipated to experience the effects of the medication very differently, as it was important to study the scope of how patients experience TEAEs. In recent decades, AAPs have been introduced for the treatment of schizophrenia and MDD, with both improved treatment efficacy and reduced neurological side effect burden compared with older, first generation antipsychotics \[[@CR11], [@CR12]\]. Schizophrenia affected an estimated 1.1% of adults (2.6 million) in the United States in 2013, and onset in early adulthood is common, often leading to chronic lifelong disability \[[@CR13], [@CR14]\]. MDD is even more prevalent in the United States, with an estimated 6.7% (15.7 million) adults having experienced an MDD episode in 2013, and more than 10 million who received treatment for depression \[[@CR15]\]. However, despite the improved efficacy and tolerability profiles of AAPs, the risk of TEAEs still associated with these agents often includes weight gain and metabolic syndrome, extrapyramidal symptoms (EPS), sexual dysfunction, and sedation and somnolence, depending on the specific agent \[[@CR9], [@CR16], [@CR17]\]. A multiple treatment meta-analysis of schizophrenia trials has shown that although antipsychotics had small but robust differences in efficacy, they differed substantially in side effects \[[@CR12]\]. Perspectives on the importance of TEAEs differ across patients and between patients and physicians. That is, how patients are affected by TEAEs is specific to each individual and may also be seen differently by physicians. These differences in perspectives and preferences must be taken into account within the therapeutic alliance, as they may impact treatment decisions when considering the overall benefit-risk profile. In fact, increasing importance is being placed on bringing the patients' perspectives to the evaluation of the overall benefit-risk profile for treatment \[[@CR18]\]. In a study reviewing adverse events of antipsychotics as outcome measures, it was concluded that a patient's subjective experience of medication should be given more consideration \[[@CR19]\]. Although TEAEs are an important consideration for treatment, they are hard to quantify. The patient's perspective may assist in this and has been used in other disease fields. For example, in rheumatology, a tolerability index that has been used in clinical trials incorporates a patient-based method of assessing TEAEs \[[@CR20]\]. In cancer clinical trials, a recommended core set of patient-reported symptoms for measuring side effects has been established to promote consistent assessment of treatment-related symptoms \[[@CR21]\]. There are a number of neuroleptic side-effect assessment scales available \[[@CR22], [@CR23]\], and among the most complete are the 48-item Udvalg for Kliniske Undersogelser (UKU) rating scale \[[@CR24]\] and the Liverpool University Neuroleptic Side-Effect Rating Scale (LUNSERS) \[[@CR25]\]. However, these scales do not fully account for the patient's subjective experience or preference. Although considered important, the differences in patient and physician perspective in schizophrenia and MDD appear to be lacking in the literature. The current study investigates TEAEs of AAPs from both patient and physician perspectives. The goal of the study was to provide better understanding of the occurrence and burden of TEAEs associated with AAP medications, as reported by patients and physicians. The initial feedback gained in this study will be used for a future project to generate an algorithm for a tolerability index score that will quantify the burden of AAP TEAEs and fully accommodate patient preference through a discrete choice experiment. The development of this index measure to assess the burden of TEAEs could help facilitate the prescribing of AAP medications to individuals with MDD and schizophrenia. A first step to establishing this potential tolerability index algorithm would be to evaluate which TEAEs are most bothersome to patients. Although the long-term goal is to allow the approach towards assessing TEAEs to be transnational, the patient selection was based in the US, partly due to the push by the FDA to include patients' perspectives in the overall benefit-risk profile for treatment \[[@CR18]\]. Methods {#Sec2} ======= Study design {#Sec3} ------------ To gain insight into the occurrence and burden of TEAEs associated with AAPs, focus groups with patients with MDD and interviews with patients with schizophrenia were conducted at two qualitative research facilities: one in Raleigh, North Carolina, USA, and one in St. Louis, Missouri, USA. Although focus groups are more efficient in terms of the time required for data collection and appropriate for data collection in the MDD patient population, previous studies have demonstrated that individual interviews are more successful in obtaining adequate feedback from individuals with schizophrenia, as these patients are generally more comfortable discussing their symptoms and experiences on an individual basis \[[@CR26]\]. A physician focus group was also conducted with psychiatrists at the North Carolina location to obtain their perspectives on the occurrence and importance of TEAEs in these patient populations. This study fully adhered to COREQ guidelines and methodology. Medical recruiters at each facility screened all participants following a study review by an institutional review board (RTI IRB Approval 2/18/15; \# 13733); informed, written consent was obtained prior to initiation of the study. Patients and physicians were provided an honorarium in appreciation for their time. Physician and patient recruitment {#Sec4} --------------------------------- Patients were selected from the individual site databases of general community residents who had previously agreed to be contacted for potential research opportunities. Patients were identified and screened, based on their own reports, to meet the inclusion criteria of being an English-speaking adult with a clinician-administered diagnosis of MDD or schizophrenia, taking one or more AAPs within the past year, and reporting one or more TEAEs associated with an AAP. The AAPs included were: aripiprazole, asenapine, clozapine, iloperidone, lurasidone, olanzapine, paliperidone, quetiapine, risperidone, or ziprasidone. Physicians interested in participating in the study were identified from the North Carolina site database. Each physician was selected by meeting the criteria of being a practicing psychiatrist providing direct care of adult patients with MDD and/or schizophrenia and regular treatment of these patients with AAP medication (e.g., 10% or more of patients with MDD requiring adjunctive therapy). Procedures {#Sec5} ---------- In this study, a TEAE was defined as "any untoward or undesirable medical occurrence in a patient that was linked in time with the use of a pharmaceutical/medicinal product and that may or may not be considered to be related to that product." Adverse drug events were not actively solicited, ascertained, or evaluated in the study; however, because this project was conducted by Lundbeck, if a potential TEAE associated with a Lundbeck product became evident through the conduct of this qualitative research, a TEAE report was submitted to Lundbeck US Pharmacovigilance. Semi-structured interview guides were utilized to provide structure to the MDD focus groups (lasting approximately 1.5 h each), the schizophrenia individual interviews (lasting 45 min each), and the psychiatrist focus group (lasting approximately 1.5 h). All focus groups and interviews were conducted by two PhD-level psychologists. General discussion was followed by targeted questions along with handouts for each participant to provide individualized feedback. The following information was obtained at the group or individual patient level: 1) exhaustive lists of TEAEs experienced, 2) frequency of each TEAE, and 3) bother ranking for the most bothersome TEAEs ("1" for most bothersome TEAE, "2" for the next most bothersome, and so on; up to a number that seemed meaningful to the patient). At the physician level, collected information included: 1) an exhaustive list of TEAEs observed or reported by their patients; 2) the most and least frequently occurring TEAEs; 3) clinically important TEAEs (ranked as 1 for most clinically important); and 4) level of patient-perceived bother for each clinically important TEAE (0 = no bother to 10 = extremely bothered). Physicians were not asked to distinguish between adverse event profiles for patients with MDD and schizophrenia. Analysis {#Sec6} -------- To ensure consistency in organizing and coding TEAEs across patients and physicians, a codebook was developed and applied, providing consensus in TEAE coding decisions by each of the two focus group and interview moderators. In order to organize the rankings and rating scores in a more meaningful manner, a "top 3 box" approach was taken (i.e., collapsing the proportion of participants reporting in the top 3 responses). For example, bother and clinical importance rankings of 1, 2, and 3 were collapsed into the "most" bothersome or "most" clinically-important TEAEs, and ratings of 8, 9, and 10, among physicians in rating patient burden were collapsed into the "most" bothersome to patients. Given the qualitative nature of the study, no formal statistical analyses or comparisons were conducted. Results {#Sec7} ======= Patients and physicians {#Sec8} ----------------------- A total of 42 patients participated in the study---25 patients with MDD and 17 with schizophrenia (Table [1](#Tab1){ref-type="table"}). More than half of the patients with MDD were female (64%) and more than half of the patients with schizophrenia were male (65%). The most common currently prescribed AAPs in patients with MDD were quetiapine and aripiprazole (both 24%) and for patients with schizophrenia, risperidone and olanzapine (both 24%). These AAPs were also the most commonly prescribed to each of the two patient groups within the last year. Of patients with MDD, 44% were diagnosed within the last 10 years while 35% of patients with schizophrenia were diagnosed within that time frame. Of patients with MDD, 48% reported living with a spouse or partner while only 4% reported living with roommates who were not family members. Patients with schizophrenia more often reported living with non-family roommates (29%) than living with a spouse or partner (24%).Table 1Patient characteristics at screeningCharacteristicMDD\ (*n* = 25)Schizophrenia\ (*n* = 17)Total Patients\ (*N* = 42)Sex, *n* (%) Male9 (36)11 (65)20 (48) Female16 (64)6 (35)22 (52)Age, years Mean (range)46.4 (23--70)45.5 (25--59)46.0 (23--70)Race/ethnicity,^a^ *n* (%) White20 (80)9 (53)29 (69) African American3 (12)8 (47)11 (26) Asian1 (4)0 (0)1 (2) Hispanic/Latino1 (4)1 (6)2 (5) Mixed race0 (0)1 (6)1 (2)Current atypical medications, *n* (%)^b^ Quetiapine6 (24)2 (12)8 (19) Aripiprazole6 (24)2 (12)8 (19) Risperidone3 (12)4 (24)7 (17) Olanzapine0 (0)4 (24)4 (10) Lurasidone3 (12)2 (12)5 (12) Clozapine1 (4)2 (12)3 (7) Ziprasidone2 (8)1 (6)3 (7) Paliperidone0 (0)1 (6)1 (2) Fluphenazine0 (0)1 (6)1 (2) Asenapine0 (0)1 (6)1 (2)^a^ Participants could report more than 1 race^b^ Total may not equal 100%. Four patients with schizophrenia and one patient with MDD reported using more than one AAP, and four patients with MDD and 2 patients with schizophrenia had a recent history of AAP use but were not currently using an AAP at screening. Each medication listed may include reference to 1 or more brand names, the chemical or generic name, and/or different formulations Four psychiatrists participated in the study; all were male and they had an average of 21.5 years of experience in practice. In total during the last year, the psychiatrists treated approximately 600 patients with MDD (35% of whom used AAPs) and approximately 300 patients with schizophrenia (57% of whom used AAPs). All of the psychiatrists regularly prescribed quetiapine, aripiprazole, risperidone, olanzapine, lurasidone, and ziprasidone; 75% of them regularly prescribed clozapine and paliperidone; and 50.0% regularly prescribed asenapine. Patient results {#Sec9} --------------- During the MDD group discussions and the schizophrenia interviews, exhaustive lists of TEAEs were developed. The first step was to gather a list of AEs from the patients via spontaneous elicitation, followed by queries to the patients from a target listing of TEAEs. Specific TEAEs reported by more than half of patients overall (across both patient types) included: weight gain (76%), cognitive issues including decreased ability to attend, concentrate, remember, or recall (79%), need to sleep/excessive sleep/excessive sleepiness (71%), low energy (67%), EPS (62%), and anxiety (55%) (Tables [2](#Tab2){ref-type="table"} and [3](#Tab3){ref-type="table"}). The reporting rates of the TEAEs differed between the two groups of patients. For example, patients with MDD were most likely to report cognitive issues as a TEAE (92%) while patients with schizophrenia were most likely to report weight gain as a TEAE (94%). Patients with MDD also commonly reported somnolence (76%), weight gain (64%), low energy (56%), and EPS (52%). Along with weight gain, patients with schizophrenia commonly reported low energy (82%), EPS (77%), somnolence/sedation (71%), and anxiety (65%).Table 2Frequency, bother, and most bothersome atypical antipsychotic AEs reported by patients with MDDMDD, *n* (%)\ (*n* = 25)AE CategoriesFrequencyBotherMost Bothersome^a^**Cognitive issues23 (92)18 (72)13 (52)**Weight changes20 (80)----  **Weight gain and/or increased appetite16 (64)11 (44)8 (32)** Weight loss and/or decreased appetite5 (20)00Somnolence/sedation19 (76)----  **Need to sleep/excessive sleep/excessive sleepiness19 (76)9 (36)7 (28)** Zombie-like/out of it5 (20)2 (8)1 (4)**Low energy14 (56)8 (32)7 (28)**EPS13 (52)5 (20)3 (12)Anxiety^b^12 (48)6 (24)4 (16) Mental anxiety6 (24)2 (8)2 (8) Physical anxiety8 (32)1 (4)0 Social anxiety000Sexual function12 (48)---- Increased sexual desire/activities1 (4)1 (4)0 Reduced sexual desire or performance11 (44)5 (20)2 (8)Anticholinergic-related dryness12 (48)---- Dry eyes3 (12)00 Dry mouth9 (36)3 (12)1 (4) Dry skin2 (8)2 (8)0Disequilibrium11 (44)3 (12)0Insomnia9 (36)3 (12)2 (8)Restlessness/akathisia9 (36)5 (20)2 (8)**Flat/restricted affect8 (32)7 (28)6 (24)**Anger/aggression8 (32)5 (20)3 (12)Irritability7 (28)6 (24)4 (16)Depressive symptoms6 (24)6 (24)4 (16)Hypomania5 (20)4 (16)2 (8)Social withdrawal5 (20)2 (8)2 (8)Bowel/digestive system changes5 (20)---- Constipation1 (4)1 (4)0 Diarrhea3 (12)1 (4)0 Nausea/vomiting2 (8)00Cardiovascular4 (16)---- Hypertension2 (8)1 (4)0 Hypotension1 (4)00 QT prolongation/skipped heartbeat1 (4)1 (4)Visual problems3 (12)2 (8)0Abnormal blood/laboratory test levels3 (12)---- Anemia000 Hyperglycemia1 (4)1 (4)1 (4)Hyperlipidemia2 (8)2 (8)0Hormonal2 (8)2 (8)0Pain2 (8)1 (4)1 (4)Other ^c^2 (8)1 (4)0Major medical (hypothyroidism)2 (8)00Major medical (diabetes)000Major medical (seizures)000Increased schizophrenia positive symptoms000*Abbreviations*: *EPS* extrapyramidal symptoms, *MDD* major depressive disorder, *QT* time between the start of the Q wave and the end of the T wave in the heart's electrical cycle^a^ The top 5 AEs most frequently reported as most bothersome are highlighted in bold^b^ Anxiety is reported as the sum of the participants who endorsed any of the anxiety subcodes and also accounts for participants who reported "anxiety" without noting a subcode^c^ The category of "other" was created to represent select symptoms reported by 2 or fewer participants that were also not reported as most bothersome, most clinically important, or an AE of specific interest Table 3Frequency, bother, and most bothersome atypical antipsychotic AEs reported by patients with schizophreniaSchizophrenia, *n* (%)\ (*n* = 17)AE CategoriesFrequencyBotherMost Bothersome^a^Weight changes16 (94)----  **Weight gain and/or increased appetite16 (94)12 (71)7 (41)** Weight loss and/or decreased appetite4 (24)00**Low energy14 (82)12 (71)6 (35)**EPS13 (77)10 (59)2 (12)Somnolence/sedation12 (71)---- Need to sleep/excessive sleep/excessive sleepiness11 (65)6 (35)1 (6) "Zombie-like"/"out of it"4 (24)2 (12)1 (6)**Anxiety** ^**b**^**11 (65)8 (47)6 (35)Mental anxiety7 (41)6 (35)6 (35)** Physical anxiety4 (24)4 (24)2 (12) Social anxiety1 (6)01 (6)Cognitive issues10 (59)6 (35)3 (18)Sexual function10 (59)---- Increased sexual desire/activities2 (12)1 (6)1 (6) Reduced sexual desire or performance8 (47)4 (24)2 (12)**Increased schizophrenia positive symptoms8 (47)7 (41)4 (24)**Anticholinergic-related dryness7 (41)---- Dry eyes1 (6)00 Dry mouth6 (35)5 (29)3 (18) Dry skin000Visual problems7 (41)4 (24)2 (12)Insomnia6 (35)5 (29)2 (12)Restlessness/akathisia6 (35)3 (18)3 (18)Disequilibrium5 (29)3 (18)1 (6)Cardiovascular5 (29)---- Hypertension4 (24)1 (6)1 (6) Hypotension1 (6)1 (6)1 (6)QT prolongation/skipped heartbeat000Depressive symptoms4 (24)2 (12)2 (12)Pain4 (24)2 (12)1 (6)Bowel/digestive system changes4 (24)---- Constipation2 (12)00 Diarrhea2 (12)1 (6)1 (6) Nausea/vomiting1 (6)1 (6)0Abnormal blood/laboratory test levels4 (24)---- Anemia1 (6)1 (6)1 (6) Hyperglycemia1 (6)1 (6)0 Hyperlipidemia2 (12)1 (6)1 (6)Irritability3 (18)2 (12)1 (6)Anger/aggression2 (12)2 (12)1 (6)Hypomania2 (12)2 (12)0Hormonal2 (12)1 (6)1 (6)Other^c^2 (12)1 (6)0Major medical (diabetes)2 (12)1 (6)1 (6)Major medical (seizures)1 (6)1 (6)0Flat/restricted affect1 (6)00Social withdrawal1 (6)00Major medical (hypothyroidism)000*Abbreviations*: *EPS* extrapyramidal symptoms, *QT* the time between the start of the Q wave and the end of the T wave in the heart's electrical cycle^a^ The top 5 AEs most frequently reported as most bothersome are highlighted in bold^b^ Anxiety is reported as the sum of the participants who endorsed any of the anxiety subcodes and also accounts for participants who reported "anxiety" without noting a subcode^c^ The category of "other" was created to represent select symptoms reported by 2 or fewer participants that were also not reported as most bothersome, most clinically important, or an AE of specific interest Using the exhaustive list of TEAEs generated in each group or interview as a reference guide, patients reported those that they perceived as bothersome and, of those, further delineated the TEAEs they found to be "most bothersome" through a ranking process. Table [2](#Tab2){ref-type="table"} lists the frequencies of TEAEs reported by patients with MDD as well as those identified as bothersome and the bothersome TEAEs ranked as the top 3 "most bothersome". Table [3](#Tab3){ref-type="table"} lists the frequencies of TEAEs reported by patients with schizophrenia. The table includes the TEAEs identified as bothersome, as well as those ranked in the top 3 "most bothersome." Specific TEAEs reported as bothersome across the overall patient sample (across both patient groups) included cognitive issues (57%), weight gain and/or increased appetite (55%), low energy (48%), EPS (36%), and need to sleep/excessive sleep/excessive sleepiness (36%). Again, the pattern of results differed between the two patient groups. Patients with MDD were most likely to include cognitive issues, weight gain, and excessive sleepiness as bothersome issues (72, 44, and 36%, respectively). These same three TEAEs were also most likely to be selected as most bothersome by patients with MDD (52, 32, and 28% of patients, respectively) although the TEAE of low energy was also rated as most bothersome 28% of the time. In contrast, patients with schizophrenia were most likely to include weight gain (71%), low energy (71%), and EPS (59%) on the list of bothersome TEAEs, and then to select weight gain, low energy, and anxiety as the most bothersome symptoms (41, 35, and 35%, respectively). There were other findings that were common across both patient groups. For instance, reduced sexual desire was mentioned as a TEAE by 44% of patients with MDD and by 47.1% of patients with schizophrenia; however, neither group was likely to endorse this TEAE as the most bothersome (8.0% for MDD and 11.8% for schizophrenia). The burden or impact of TEAEs varied in the patient groups, although some generalities could be made. For example, cognitive issues were reported by both groups of patients although patients with MDD reported a more significant impact, including trouble holding conversations, trouble managing work or school, concern about "losing it" or having permanent memory problems, and having poor self-esteem resulting from "feeling stupid". Participants in both groups noted concerns about gaining weight: small amounts gained quickly, and large amounts gained over a longer period. Most patients attributed weight gain to AAP use and patients in both groups noted that the weight gain resulted in physical problems, poor body image, and poor self-esteem. Additionally, whereas patients with schizophrenia verbalized that they would likely discontinue their medications because of significant weight gain and the related concern about the impact on existing and future cardiometabolic TEAEs, patients with MDD noted that they would rather live with the TEAEs associated with the medications than with extreme depressive symptoms. Both groups also noted increased fatigue, low energy, and sleepiness, usually immediately following AAP initiation and, especially in the case of patients with schizophrenia, these TEAEs were noted with each administration of medication. The impact of the somnolence/sedation was significant and similar in the two groups and included missing time with family and friends, missing social activities, lack of energy leading to not eating properly, poor self-esteem, and feelings of sedation that interfered with proper functioning. Patients in both groups reported EPS symptoms, including tremors and irregular jerky movements, but only patients in the schizophrenia group reported them as burdensome and impactful. Impact in these patients included the fear of others noticing tremors in public and interference with job duties. Patients in both groups noted change in sexual desire and functioning, and while most patients who reported this experienced decreased sexual desire effect, a few patients experienced increased desire. However, in both cases, most patients did not report an impact. Physician results {#Sec10} ----------------- As a group, the physicians also participated in the creation of an exhaustive list of TEAEs. Using the exhaustive list as an index, the physicians reported and then rated or ranked both clinically important and bothersome TEAEs (Table [4](#Tab4){ref-type="table"}). Those considered most clinically important by at least 2 of the 4 psychiatrists were metabolic syndrome (100%), weight gain (50%), neutropenia (75%), hyperglycemia (50%), and QT prolongation (50%). TEAEs considered most bothersome to patients by at least 2 of the 4 psychiatrists were weight gain (100%), reduced sexual desire or performance (50%), EPS (50%), akathisia (50%), and hormonal issues (50%).Table 4Atypical antipsychotic AEs reported by physicians as clinically important and/or bothersomeAEs, *n* (%)Clinically important\ (*n* = 4)Most clinically important\ (*n* = 4)Most bothersome to patients\ (*n* = 4)Metabolic syndrome ^a^4 (100)4 (100)n/aWeight gain4 (100)2 (50)4 (100)Reduced sexual desire or performance4 (100)1 (25)2 (50)Neutropenia ^b^3 (75)3 (75)1 (25)Hyperglycemia ^b^3 (75)2 (50)0EPS ^b^3 (75)1 (25)2 (50)Hyperlipidemia3 (75)1 (25)0Akathisia3 (75)02 (50)QT prolongation ^b^2 (50)2 (50)0Major medical (seizures)2 (50)1 (25)1 (25)Hormonal2 (50)02 (50)Hypotension2 (50)01 (25)Cognitive issues2 (50)01 (25)Major medical (diabetes) ^b^1 (25)1 (25)1 (25)Hypertension1 (25)1 (25)0Low energy1 (25)01 (25)Depressive symptoms1 (25)00Flat/restricted affect1 (25)00Somnolence/sedation1 (25)00*Abbreviations*: *AE* adverse event, *EPS* extrapyramidal symptoms, *QT* time between the start of the Q wave and the end of the T wave in the heart's electrical cycle^a^ Reported as "most clinically important" in the group setting but not the individual task^b^ Reported as "most clinically important" in the group setting and in the individual task The two TEAEs discussed most by physicians were weight gain and reduced sexual desire or performance. According to the physicians, weight gain was common and almost immediate, but in contrast to patient opinion, it was not always attributed to medication; rather, it was attributed to poor eating habits that were exacerbated during hospitalization or major depressive episode. The physicians believed that the weight gain created a negative impact that was generally more significant for women, but sometimes created a positive impact for individuals with poor appetite associated with MDD, because it helped achieve healthy body weight. Reduced sexual desire or performance was reportedly mentioned within the first visit following medication change or initiation. The physicians commented that it was not always due to AAP, and it was rare to see an improvement of this TEAE while on medication. Patient and physician summary {#Sec11} ----------------------------- A summary of frequently reported, bothersome, or clinically important AAP TEAEs conveyed by patients and psychiatrists is illustrated in Fig. [1](#Fig1){ref-type="fig"}. This figure shows that patients with MDD and schizophrenia used similar terms to describe their TEAEs and the most frequent and bothersome TEAEs reported by patients were mentioned by the psychiatrists as well. One exception was that the terminology and descriptions of some TEAEs differed between patients and clinicians, such as that of akathisia. Instead of using the clinical term, patients described experiences consistent with akathisia, including "very uncomfortable and shaking inside," "can't get comfortable," "needing to move," "can't sit still," "had to fidget," and "jumping out of my skin." Many TEAEs that were highly bothersome were also noted as frequently occurring, including weight gain, low energy, somnolence, cognitive issues, and EPS. TEAEs considered highly bothersome by patients and clinically significant by psychiatrists included weight gain and cognitive issues (MDD and schizophrenia) and EPS (schizophrenia), although the results indicate that cognitive issues are more bothersome than physicians think, especially for MDD patients. In contrast, many other TEAEs were considered clinically important by psychiatrists but not by patients, including metabolic syndrome, reduced sexual function, QT prolongation, hormonal changes, akathisia, neutropenia, seizures, and hypotension. It should be noted that some of these TEAEs (e.g., QT prolongation, neutropenia) are not amenable to self-reporting by the patient and therefore, should not necessarily be considered unimportant by the patient. However, although the patients cannot self-report these specific TEAEs, they were discussed as part of the exhaustive list.Fig. 1Summary of Frequently Reported, Bothersome, or Clinically Important Adverse Events Reported by Patients and Physicians.^a.^ EPS, extrapyramidal symptoms; MDD, major depressive disorder; TEAE, treatment-emergent adverse event.^a^ "*Frequent*" refers to the overall frequency of report for either patient type or for psychiatrists and refers to TEAEs mentioned by ≥50% of patients or those described as "*frequent*" by psychiatrists; "*Bothersome*" refers to the report of a TEAE as bothersome by ≥30% of either patient type or as "*most*" bothersome by ≥20% of either patient type and for psychiatrists as "*most bothersome*" by 2 of the 4 psychiatrists; "*Important*" refers to TEAEs identified as "*clinically important*" by 2 of the 4 psychiatrists. ^b^ In addition, the following AEs were listed as clinically important by ≥2 physicians: neutropenia, hyperglycemia, hyperlipidemia, QT prolongation, seizures (*major medical*), hypotension, and metabolic syndrome Discussion {#Sec12} ========== Newer, second generation (atypical) antipsychotics demonstrate improved efficacy and reduced side effect burden compared with older, first generation antipsychotic medications \[[@CR11], [@CR12]\]. However, a risk of TEAEs, including weight gain, EPS, sexual dysfunction, and somnolence remains \[[@CR9], [@CR16], [@CR17]\]. This study supports these previous findings and sheds light on TEAE perspectives and preferences that may impact treatment decisions for patients with schizophrenia or MDD being treated with these newer agents. These differences should be taken into account to improve the therapeutic alliance between patients and physicians. This study used focus groups for MDD patients, interviews for schizophrenia patients, and focus groups for physicians to form and rank an exhaustive list of TEAEs to show that patients with MDD and schizophrenia generally used similar words to describe the TEAEs they were experiencing. TEAEs reported as bothersome across both patient groups included cognitive issues, weight gain and/or increased appetite, low energy, EPS, and the need to sleep/excessive sleep/excessive sleepiness. TEAEs considered more bothersome by patients with schizophrenia were weight gain, low energy, EPS, mental anxiety and increased positive symptoms, whereas TEAEs considered more bothersome by patients with MDD were cognitive issues, somnolence/sedation, and flat/restricted affect. Patients' and physicians' perspectives did not align as well as those across the two patient groups. TEAEs considered most clinically important by psychiatrists included metabolic syndrome, weight gain, neutropenia, hyperglycemia, and QT prolongation; those considered most bothersome to patients from physicians' perspectives included weight gain, reduced sexual desire or performance, EPS, akathisia, and hormonal issues. Some of the differences between the perspectives of patients and psychiatrists were related to clinical term use. For example, differences in understanding of akathisia may have been due to differences in terminology as well as differences in description and/or experience, which may explain why clinicians saw it as clinically important and patients did not list it as most bothersome. In addition, due to patients' tendency to combine types of EPS, the TEAEs of tardive dyskinesia, dystonia, and Parkinson symptoms were combined for the purposes of this study. Interestingly, three of the most frequent and bothersome TEAEs reported by patients (low energy, somnolence/sedation, and cognitive issues) were not described by physicians as being clinically important or most bothersome. This may reflect the challenges that physicians have in measuring and recording these events; but it may also reflect an assumption by the physicians that these symptoms are due to the disease while patients ascribe the adverse events to the treatment. This study did not attempt to tease this apart but this question would be useful for further research in support of improving the therapeutic alliance. The results described in this study provide better understanding of the occurrence and burden of TEAEs associated with AAP medications as reported by patients and physicians. It is also interesting to compare the frequency and burden ratings from this study to the adverse event profiles of the drugs as observed in the clinical trials and described in the approved product information (PI) sheets. A survey of the top five product PIs as listed by the patients in this study (risperidone, olanzapine, lurasidone, aripiprazole and quetiapine) reveals both similarities and differences. No adverse event is mentioned in all five PIs (using the section on Adverse Reactions with MDD or schizophrenia. Akathisia is listed in four of the five PIs, which is reflected in the common mention by psychiatrists. Other frequent and burdensome AEs, such as weight gain or increased appetite, somnolence, and lethargy/fatigue are each mentioned in three of the five PIs as a common adverse event. However, dizziness and constipation are also listed in three of five PIs but are not mentioned as often by the patients or physicians in this study. Conversely, the very common and burdensome AE of cognitive issues is not mentioned in any PI, nor are sexual function changes. The mismatch between labelling and patient and physician experience further emphasizes the need to better incorporate patient perspective into adverse event reporting and evaluation. Another way to look at efficacy and tolerability of AAPs is head-to-head clinical trial analysis, as has been shown in a recent mixed-treatment comparison analysis that was used to directly compare the efficacy and tolerability of AAPs \[[@CR27]\]. While useful, these types of trials do not reveal how the TEAEs affect patients. To fully understand AE burden in utilization of antipsychotics, the patient perspective needs to be taken into account. Patient attitudes toward antipsychotics are important for adherence to and eventual outcome of the antipsychotic treatment. Surveys of patient attitude toward antipsychotics, including tolerability of TEAEs, have shown a better attitude toward second-generation antipsychotics compared with first-generation antipsychotics, although the results are still debatable \[[@CR28]\]. A tolerability index would facilitate the evaluation of TEAE burden among patients and would contribute to the process of prescribing appropriate antipsychotics to patients with MDD and schizophrenia. The information gained in the current study on patient perspectives of TEAE burden will assist in the creation of a tolerability index, which may facilitate appropriate prescribing of AAP medications. An important consideration based on this study is that if the most burdensome side effects for patients can be determined, physicians can better select the right treatment from the choices available. A multiple treatment meta-analysis of randomized, controlled schizophrenia trials compared 15 antipsychotic drugs and placebo in the acute treatment of schizophrenia and found that whereas antipsychotics had small but robust differences in efficacy, they differed substantially in side effects \[[@CR12]\]. The findings challenged the routine classification of antipsychotics by first- and second-generation types (ie, typical and atypical types) and suggested that domain hierarchies should be used to help clinicians adapt the choice of an antipsychotic drug to the needs of individual patients \[[@CR13]\]. The results of this study demonstrate the variety of the individual needs of patients, and the potential mismatch between patients' and physicians' perspectives. In future research, it will be interesting to further examine sources of variability in patient and physician perspective. For instance, there may be international differences in the perception of burden of these common adverse events that should be carefully described and taken into account in selection of medication in other settings. This was a patient perspective study, and in addition to the general limitations of a study based on patient reports, it is important to recognize other study limitations. For example, the sample size was small and limited to two geographic locations in the US. A US sample was selected in part because of the wider range of AAPs currently approved for the treatment of MDD in the US, and also because of the recent campaign by the FDA to include patients' perspectives in the overall benefit-risk profile for treatment \[[@CR18]\]. Therefore, the results may not be able to be generalized to a larger or transnational patient or physician population. Further, patient reports proved insufficient for concluding valid results on certain queries. In particular, TEAE onset was investigated because patients are more likely to experience TEAEs early in treatment that may or may not resolve as the patient continues with the treatment. Unfortunately, onset was difficult to determine for most patients and varied greatly in the cases where it was able to be determined. Finally, the data quantitation used a "top 3 box approach." While this approach enabled the intuitive comparison of perspectives across small and varied sample sizes, this method can also cause a loss of precision. Finally, although unlikely to have affected the study outcomes, the study was funded by a pharmaceutical company and the participating patients and physicians received honoraria for their time. Conclusion {#Sec13} ========== The wide range of TEAEs that are both frequent and bothersome and the variation in perceived burden according to the diagnosis highlight the need for a tailored TEAE-awareness approach when choosing an AAP. Following feedback from both patients and psychiatrists, the following TEAEs would probably carry the most weight for a tolerability index measure: weight gain and/or increased appetite, low energy, somnolence/sedation, cognitive issues, EPS, and reduced sexual desire or performance. Other TEAEs that were considered clinically important by psychiatrists included abnormal test results or blood values, cardiovascular issues, major medical issues, hormonal changes, and akathisia. Information gained in this study on TEAE burden associated with AAP use will be helpful in the future development of tools for assessing the overall tolerability of these agents. AAPs : Atypical antipsychotics AE : Adverse event EPS : Extrapyramidal symptoms MDD : Major depressive disorder TEAEs : Treatment-emergent adverse events Editorial support for the preparation of this manuscript was provided by Melissa Kirk, Ph. D. at Scientific Connexions, Lyndhurst, NJ, USA, an Ashfield Company, part of UDG Healthcare plc, and funded by Otsuka America Pharmaceutical, Inc and Lundbeck LLC, Copenhagen, Denmark. Funding {#FPar1} ======= This study was sponsored by Otsuka America Pharmaceutical, Inc., Princeton, NJ, USA, Otsuka Pharmaceutical Development and Commercialization, Inc., Princeton, NJ, USA, and Lundbeck LLC, Copenhagen, Denmark. Availability of data and materials {#FPar2} ================================== Due to the qualitative nature of the study, additional data can be requested from the authors and will be made readily available. Authors' contributions {#FPar3} ====================== CL, AH, DBD, SAK, and CF contributed to study concept and design and P-ML, AH, DBD, SAK, and CF to the development of study methods. TMB and DBD contributed to acquisition of data and P-ML, AH, TMB, DBD, SAK, and CF contributed to analysis and interpretation of data. AH, TMB, DBD, and CF made contributions of administration, technical or material support and AH and CF contributed to study supervision. All authors participated in the writing, review, and revision of the manuscript. All authors read and approved the final manuscript. Competing interests {#FPar4} =================== Pierre-Michele Llorca discloses personal fees for travel reimbursement from Lundbeck LLC related to this study and personal fees from Janssen, Otsuka, and Takeda, not related to this study. Christophe Lancon has nothing to disclose. Ann Hartry and Clement Francois are employees of Lundbeck LLC, and Siddhesh Kamat is an employee of Otsuka America Pharmaceutical, Inc. T. Michelle Brown and Dana B. DiBenedetti are employees of RTI Health Solutions, which has received funds from Lundbeck LLC in connection with the conduction of this study. Consent for publication {#FPar5} ======================= Not applicable. Ethics approval and consent to participate {#FPar6} ========================================== Following a study review and approval by the RTI International institutional review board (approval \#13733), medical recruiters at each facility screened all participants. Informed, written consent was obtained prior to initiation of the study. The study adhered to COREQ guidelines and methodology.

INTRODUCTION {#sec1-1} ============ According to the global burden of disease study, COPD will be the fifth leading cause of disability and the third leading cause of death in the world in the first half of the twenty-first century. For developing countries, COPD is expected to be the fourth leading cause of disability for males and the third for females in 2020.\[[@ref1]\] Chronic obstructive pulmonary disease (COPD) has been defined by the global initiative for obstructive lung disease (GOLD) as a disease state characterized by airflow limitation that is not fully reversible.\[[@ref2]\] Dyspnoea, the hallmark symptom of COPD, is the reason for which most patients seek medical attention and is a major cause of disability and anxiety associated with the disease. Chronic cough, often the first symptom of COPD, to develop.\[[@ref3]\] The diffusing capacity of lung for carbon monoxide (DLCO) is a measure of the ability of gas to transfer from the alveoli across the alveolar epithelium and the capillary endothelium to the RBCs. Gas exchange is impaired by parenchyma destruction, which disrupts the local matching of ventilation and perfusion.\[[@ref4]\] The imbalance of ventilation perfusion may lead to alteration of transfer factor. It is usually due to a change in either or both the volume of blood in the alveolar capillaries and the diffusion capacity of alveolar capillary membrane. In chronic lung disease, diffusion capacity is impaired when there is a reduction in the effective surface area for gas exchange in lung, in disease of lung parenchyma in which there is loss of lung tissue or part of the lung, not ventilated.\[[@ref5]\] A comprehensive yoga program can have a salutary effect on general health and respiratory health and thereby help increase a person\'s ability to perform activities of daily living. COPD is known to increase the level of stress, emotional vulnerability, physical inactivity and muscle wasting. This yogic regimen may change the milieu at the bronchioles and the alveoli particularly at the alveolo-capillary membrane to facilitate diffusion and transport. Hence transfer factor of the lung for carbon monoxide (TLCO) has been included in this proposed study. We have not come across any study regarding the effect of yogic exercises on transfer factor of lung for carbon monoxide; it was therefore decided to scientifically study the effects of yogic exercises in a group of COPD patients by measuring pulmonary function tests along with diffusion capacity. MATERIALS AND METHODS {#sec1-2} ===================== The study was conducted in University College of Medical Sciences (UCMS), Delhi on 60 diagnosed patients of mild (*n*=30) and moderate (*n*=30) COPD patients in the age group of 30-60 years, of either sex, having disease duration of more than one year. The patients were recruited from the medicine outpatient department (OPD) of Guru Teg Bahadur (GTB) hospital, Delhi. The diagnosis was based on three most common symptoms, namely, cough, sputum production and exertional dyspnoea. Inclusion criteria for COPD patients of mild and moderate severity, according to GOLD guidelines: Mild ratio of forced expiratory volume in first second to forced vital capacity (FEV~1~/FVC)\<70% and FEV~1~\>80% predicted, moderate- FEV~1~/FVC\<70% and FEV~1~=50-80% predicted. All patients remained on their prescribed medical treatment during the study. They were on regular conventional treatment with daily inhaled bronchodilator β~2~ agonist; salbutamol 100-200 μg at the interval of 6 h and inhaled anticholinergic; ipratropium bromide 40-80 μg at the interval of 6-8 h. Subjects with a history of an exacerbation or respiratory tract infections, tuberculosis, current smokers, pregnant or lactating women, diabetes or any other disorder were excluded. The medication for COPD was kept the same throughout the study period for both control and the yoga group. The study was explained to the patients and their signed informed consent was taken. Ethical clearance was also obtained from UCMS ethical committee. Selected patients were randomized into two groups:  **Group 1 control group** (*n*=30): This group was further subdivided according to severity: Group 1a Mild COPD: This group was taking conventional treatment with inhaled bronchodilator β~2~ agonist, salbutamol 100-200 μg at an interval of 6 h. Group 1b Moderate COPD: This group was taking conventional treatment with inhaled bronchodilator β~2~ agonist, salbutamol 100-200 μg at an interval of 6 h and inhaled anticholinergic, ipratropium bromide 40-80 μg at an interval of 6 h. **Group 2 Yoga group** (*n*=30): These patients were taught pranayama and asanas and were asked to continue the same medication as group 3. This group was further subdivided according to severity. Group 2a Mild COPD Group 2b Moderate COPD The training in yoga: The yoga practice given to COPD patients (for 2 months) included pranayama and asanas. Before putting the patients on yoga regimen, they were clinically examined to rule out any physical ailments. Patients were asked to perform yoga exercises for 40-50 min everyday for 2 months under the supervision and guidance of a yoga instructor. Yoga includes pranayama (30-35 min), asanas (10 min), meditation (10 min) and life style changes. Pranayama (breathing exercises):\[[@ref6]\] Bhastrika: 5 minAnulom vilom: 15 minKapalbhati: 10 minBhramari: 5 times Asanas (postures): Surya NamaskarTadasanaSukhasanaPaschimotanasanaShavasana: 10-20 min Both the control group and yoga group were matched for age, sex and duration of asthma and COPD. All subjects underwent complete physical examination and clinical assessment. Routine laboratory tests (complete hemogram) were done at the time of commencement of the study. Patients were tested, instructed and followed up in cardiopulmonary laboratory of the Physiology department. Yoga group were explained about yoga and their lifestyle modifications. They were advised about the diet, in which more fruits and vegetables were included. They were instructed to avoid alcohol and smoking and to keep regular working and sleeping hours. Proper yoga training was given by yoga experts. Subsequently they performed yoga in the yoga clinic for 21 days for an average of 45 min daily. Thereafter, they were asked to practice yoga for 45 min daily at home. The subjects were followed in the cardiopulmonary laboratory, after each week, for an evaluation and compliance to see whether they were doing the yogic exercises properly. Subjects maintained daily records of their breathing exercises, asthma symptom severity during the day and night, plus activity limitations due to asthma. They were asked to note down if there was any change in the dose of their medications. The control groups were also asked to maintain records of all events related to disease and medication use. Both the groups were regularly attended by their treating physicians during study evaluation visits. Both the groups filled the daily diary and brought it at each visit. During the follow up period, telephonic support was provided for motivating participants to improve their compliance. All the subjects were evaluated three times, first at the time of recruitment, then after one month and two months. Simultaneously they underwent either conventional treatment or conventional treatment with yogic intervention. The recorded parameters were compared, statistically analyzed and then concluded. All of the patients received the same yoga training. Parameters {#sec2-1} ---------- Standing height and weight of the patients were measured and body mass index (BMI) was calculated. Diffusion capacity was assessed prior to yoga training, at the end of 1 month and after 2 months of yoga therapy. It was carried out on each stable subject using computerized medisoft instrument (HYP'AIR compact-manufacturer- PK MORGON). The patients were acclimatized to the laboratory for 10 min. The level of the mouth piece was adjusted so that the patient was comfortable. Adequate demonstration was given till the subject had comprehended the instructions. Diffusion capacity was carried out in the morning between 9:30 am and 11 am. To make the measurement of transfer factor, the subject was seated upright in front of the apparatus; this is set so that the subject breathes air through the mouth piece. A nose clip was worn. After a few normal breaths, the subject breathed out to residual volume. Then immediately, the subjectrapidly inhaled the test gas to total lung capacity, held the breath for approximately 9-11 s and finally breathed out at a moderately fast rate. After exhalation, a sample of alveolar gas was collected for analysis. By the method of Jones and Meade, the effective duration of breath holding is taken to include two- thirds of time of inspiration and the time of expiration up to half way through the period of sample collection.\[[@ref7][@ref8]\] A total of 3 tests were performed and the best of the three fulfilling the criteria of reproducibility and vitality were considered for analysis. The control group was told to continue the medicines, with no dietary restrictions or daily activities. They were also assessed for diffusion capacity at the time of recruitment, at the end of 1 month and after 2 months. Data were collected, tabulated and analyzed using repeated measures of analysis of variance (ANOVA), followed by Tukey test with *P*\<0.05 as statistically significant. Results are expressed as mean±standard deviation (SD). RESULTS {#sec1-3} ======= The mean±SD for age in group 1a was 39.33±8.415 years and for group 2a was 46±9.449 years. The mean±SD for age in group 1b was 50.87±8.634 years and for group 2b was 52.47±8.911 years. The baseline parameters were comparable between the yoga and the control group of mild COPD patients and moderate patients (*P*\>0.01 for all). Anthropometric variables (body surface area (BSA), body mass index (BMI) were found to be significantly different between group 2a and group 2b \[Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}\] (*P*\<0.05). ###### Anthropometric variables in group 2a  ###### Anthropometric variables in group 2b  Transfer factor of lung for carbon monoxide (TLCO) in group 2a and group 2b showed a statistically significant increasing trend over time from baseline to 1 month and baseline to 2 months (*P*\<0.001) \[Tables [3](#T3){ref-type="table"} and [4](#T4){ref-type="table"}\]. TLCO was higher in group 2a than in group 1a, but the differences were statistically not significant (*P*=0.443 \[[Table 3](#T3){ref-type="table"}\]. Also, TLCO was higher in group 2b than in group 1b, but the differences were statistically not significant (*P*=0.409) \[[Table 4](#T4){ref-type="table"}\]. ###### Transfer factor of lung for carbon monoxide in group 1a and 2a  ###### Transfer factor of lung for carbon monoxide in group 1b and 2b  DISCUSSION {#sec1-4} ========== In the present study, there was significant decrease in weight and body mass index after yoga. The values of TLCO showed a statistically significant improvement after two months of yoga training in mild as well as moderate COPD. Behera D studied the effect of yoga on COPD patients and showed that lung function parameters (forced vital capacity (FVC), forced expiratory volume in first second (FEV~1~)and peak expiratory flow rate(PEFR) improved after the practice of yoga.\[[@ref9]\] COPD patients undergoing conventional drug treatment had no change in TLCO in COPD patients In the earlier stage of COPD a person\'s lung function may be impaired 20% to 30%. The progress of disease causes further impairment, which in turn increases inactivity. Inactivity itself may contribute to the deterioration of the lungs and the total musculature.\[[@ref10]\] Yoga when practiced by patients with COPD results in improvement in their quality of life and lung function on a short-term basis.\[[@ref11]\] Through proper breathing exercises more oxygen is available for the exchange at tissue level, in turn muscles throughout the body to do their job efficiently. Breathing exercises and stretching postures are used to increase respiratory stamina, relaxation of the chest muscles, expansion oflungs, raising energy levels and calming the body.\[[@ref12]\] Yoga improves the blood circulation; there is better perfusion of tissues also, which increases the strength of respiratory muscles. More oxygen binds with hemoglobin. So oxygen delivery increases. The further advantage of yogic breathing lies in the fact that it is more of a vertical breathing. By this vertical breathing, all the alveoli of both the lungs open up evenly. Due to the even expansion of all the alveoli, a vast expanse of alveolar membrane is available for exchange of gases. This surface is about 50 m^2^ in extent, which is 20 times the entire body surface. The larger the surface available for the process of diffusion, the better is the process. The purpose of yoga breathing exercises is to supply the body with oxygen and cleanse it of carbon dioxide and other toxins. Generally, a small portion of lung capacity is been utilized. This inadequate supply of oxygen results in improper waste disposal from the body. The body functions are slowed down and the cells/tissues fail to regenerate themselves due to lack of sufficient energy.\[[@ref6]\] With pranayama practice, there can be an increased intake of oxygen as much as five times. This means five times of carbon dioxide is gottenrid from the body. There can be great improvement in the health by doing pranayama. The controlled breathing in yoga can ease anxiety, achieve relaxation, and provide more oxygen to the blood stream. The exercises help open blocked airways caused by bronchitis or emphysema, which are linked to COPD, and improve the function of air circulation. Simple yoga moves can even aid those with advanced COPD. Vedanathan has observed that one of the most important aspects of yoga for asthma and COPD patients is that "they develop an increased capacity to relax and control their breathing. They learn that they don't have to let their breathing control and that they can take charge of their breathing." As with any technique, they emphasize that yoga requires regular practice in order to be effective.\[[@ref13]\] Pranayama, asana and meditation, through external signals (five sense organs) and internal signals (proprioreceptors, visceroceptors and chemoceptors) modulate the brain\'s cortico-limbic-hypothalamic systems and provide beneficial effects due to better functional coupling of 'autonomic, endocrine and somatic' responses. This could be correlated with homeostatic responses set up to negate the undesirable effects of stress. CONCLUSION {#sec1-5} ========== We conclude that yoga, especially the pranayamic breathing exercises when used adjunctively with standard pharmacological treatment, can significantly improve TLCO in yoga with mild-to-moderate grades of COPD. **Source of Support:** Nil **Conflict of Interest:** None declared

Introduction {#Sec1} ============ Although the US has achieved substantial gains in improving the health of women and children, and infant mortality has reached record low levels, the US presently ranks 27th among established market economies in infant mortality \[[@CR1]\]. Four recurring causes account for more than half of all infant deaths: birth defects, disorders related to short gestation and preterm birth, maternal complications of pregnancy (including complications of the placenta, cord and membranes), and sudden infant death syndrome \[[@CR2]\]. In recent years some causes of infant mortality have increased, particularly in the percentage of births that were preterm and of low birth weight \[[@CR3]\]. In 2002, congenital anomalies, low birth weight, preterm delivery, and maternal complications of pregnancy accounted for 14,263 (50.9%) of the 28,034 infant deaths \[[@CR2]\]. Relatively little is known about the risk factors underlying the continued increase in these adverse outcomes. Adequate prenatal care has long been considered as an opportunity to reduce such risks. However, despite increases in access to and utilization of early prenatal care, interventions and efforts directed at addressing such risk factors fall short of their goal. Indeed, the effects of such efforts may have reached their peak, and new approaches may be necessary. Reviews on selected risk factors indicate that a large proportion of women enter pregnancy with pre-existing risks for adverse pregnancy outcomes. Although some women tend to take action to reduce their risk as soon as they learn that they are pregnant, the extent of pregnancy related change in risk factors varies considerably and often does not occur early in pregnancy when teratogenic effects are more pronounced. Moreover, post-pregnancy relapse is high \[[@CR4]\]. For example in the period 1996--1998 the reported reduction in the use of alcohol, tobacco and illicit drugs in the first trimester was 46, 28, and 28% respectively \[[@CR4]\]. Such information is important to the emerging emphasis on preconception care as a complementary approach to reduce risks to pregnancy. This paper provides nationally-representative estimates on risks during the preconception period and describes the apparent reductions in risk achieved during pregnancy for all known risk factors for which data are available. Data {#Sec2} ==== The estimates presented herein are based on data from the 2004 Behavioral Risk Factor Surveillance System (BRFSS) \[[@CR5]\]. However, because not all risk indicators that were considered were included in the 2004 BRFSS, 2002 data were used for a small number of risk indicators for which the 2002 BRFSS provided the most recently available data. The BRFSS is an ongoing annual telephone survey of the non-institutionalized adult civilian population aged 18 years and older conducted in each state. The survey obtains information on a wide range of modifiable risk behaviors. In 2004 the median response rate for the BRFSS state surveys was 52.7 percent. This rate represents an estimate of the percentage of eligible respondents that completed telephone interviews, and is computed based on procedures recommended by the Council of American Survey Research Organizations (http://www.casro.org/resprates.cfm.) In comparison with other national surveys, BRFSS data appear to be of good quality \[[@CR5]\]. Additional technical information is available online at http://www.cdc.gov/brfss/. The survey includes core modules asked in all states, rotating modules asked every second year in all states, and optional modules that are included only in some states. Being a general survey, the BRFSS lacks information on some factors of interest to the present topic, for example month of pregnancy for pregnant women, or information on prenatal care. Further, the BRFSS is not designed specifically to study all known risk behaviors at preconception or pregnancy. In addition, as an interview survey, the survey relies on self-reported data which contain an unknown level of reporting error. Identification of women in the preconception group in the BRFSS is made possible by questions introduced in the "family planning section" of the questionnaire. The availability of these questions makes it possible to compare risk behaviors among women about to become pregnant with those who are already pregnant on a large and nationally-representative sample of women. Women in the preconception period were identified as those who reported that they wanted a baby in the next 12 months, were not using contraception, were not sterile and were not already pregnant. Women who were pregnant were identified based on the response to the question "To your knowledge are you now pregnant?" Age of gestation was not assessed in this survey. Data were aggregated across states to make national estimates. For 2004 this resulted in a total national U.S. sample of 70,917 women aged 18 to 44 years, of whom 2308 (3.4%) were classified into the preconception period, and 2998 (4.7%) reported that they were pregnant at the time of their interview. The corresponding number of women in the 2002 are 61,284 women 18--44, with 2204 (3.7%) in the preconception period, and 2556 (4.6%) pregnant women. The analysis on folic acid and vitamins are based on surveys in 12 states in 2004. The results for these items are generalizable to the populations of these 12 states (shown in Table [2](#Tab2){ref-type="table"}) but not to the entire United States population. Table 1Comparison of age and race/ethnicity, 2004 Women in preconception and pregnant (BRFSS) and Births (Vital Statistics data)BRFSSPreconception womenPregnant womenVital statisticsPercentCIPercentCIbirths PercentAge group 18--192.00.97.52.27.1 20--2416.33.326.53.026.1 25--2923.43.126.52.727.9 30--3428.43.126.22.824.4 35--3918.52.510.21.512.0 40--4411.42.23.10.92.6100.0100.0Race/ethnicity NH White64.93.960.13.456.8 NH Black11.92.511.51.913.6 Hispanic16.03.523.03.422.5 Others7.22.15.51.67.2100.0100.0100.0*Note.* For BRFSS estimates, 95% confidence interval = Percent +/− CI. Vital statistics data: Hamilton BE, Martin JA, Ventura SJ, Sutton PD, Menacker F. Births: Preliminary data for 2004. National vital statistics reports; vol 54 no 8. Hyattsville, Maryland: National Center for Health Statistics. 2005.Table 2Health risk indicators by pregnancy status Women 18--44, 2004 and 2002 BRFSSPreconception womenPregnant womenPercentCIPercentCI2004 data--nationwide^*a*^ General Health  Poor/fair general health status8.32.16.41.7  14+ days in past month mental health not good^\*^12.82.49.61.9  No health plan^\*^18.83.511.92.3  No dental visit past year28.33.230.23.1  Told had diabetes^\*^2.00.90.70.4 HIV  Don't know about prevention of MTC HIV transmission38.23.534.13.0  Never tested for HIV^\*^34.93.424.43.0  1 or more HIV risk category^\*^4.31.46.91.8 Alcohol/smoking  Any alcohol in past month^\*^53.93.710.71.8  Average 1 or more drink per day, past month^\*^6.02.00.80.5  Binging: Any occasions of 5+ drinks in past month^\*^10.72.11.90.9  Frequent drinking: binging or 1 or more drinks per day^\*^12.92.52.20.9  Current smoker^\*^19.42.78.41.5 Obesity  Overweight, body mass index (BMI) \>2546.03.6NA  Obese, BMI \>3022.43.1NA2004 data, 12 states only^*b,c*^ Folic acid  Don't know about folic acid for birth defects prevention46.16.938.45.6  Don't take vitamins of any kind^\*^36.97.110.23.6  Don't take folic acid or multivitamin^\*^38.47.017.54.5  Don't take folic acid or multivitamin *daily*^\*^44.86.919.94.7Nutrition (2002 BRFSS)^d^ Fewer than 5 servings/fruit and vegetables^\*^74.93.164.73.5 Fewer than 1 servings/fruit and vegetables3.31.12.91.3*Note.* 95% confidence interval = Percent +/− CI.^*a*^Unweighted number of observations: Preconception women (2308), Pregnant Women (2998).^*b*^Unweighted number of observations: Preconception women (607), Pregnant Women (756).^*c*^States: AZ, CO, FL, KY, MN, MT, NC, ND, NE, TX, VA, WI.^*d*^Unweighted number of observations: Preconception women (2204), Pregnant Women (2556).^\*^*p* \< .05, preconception vs. pregnant women.^\*\*^HIV risk: Any statement is true: in past year illegal drug injection, treated for STD, exchange for money/drugs, uprotected anal intercourse. Statistical methods {#Sec3} =================== The percentage of women exposed to 21 risk indicators was identified for women in 2 groups: women in the preconception period, and pregnant women (Table [1](#Tab1){ref-type="table"}). The categories of risk indicators for which data are available include perceived general and mental health, lack of medical insurance, knowledge of HIV prevention and practices, alcohol and tobacco use, obesity, nutrition including folic acid uptake for the prevention of birth defects, and presence of chronic medical conditions. Because these are not longitudinal data it was not possible to observe behavior change in individual women. Rather the prevalence of risk factors was compared between preconception and pregnant women as a proxy measure of pregnancy-related risk reduction \[[@CR6]\], although is is possible that other factors could influence the differences between preconception and pregnant women. The statistical significance between preconception and pregnant women of the percentage reporting each risk factor, was assessed using a *t*-test. All statistical analysis was based on weighting factors designed to produce unbiased estimates, and statistical tests were adjusted for complex sample design using the software for survey data analysis (SUDAAN, Research Triangle Institute, Research Triangle Park, NC). Results {#Sec4} ======= Comparisons with vital statistics data on births in 2004 indicate that the 2004 BRFSS respondents in the preconception stage and pregnant when interviewed were distributed very similarly by age and race-ethnicity to women giving birth in 2004 (Table [1](#Tab1){ref-type="table"}). Overall, for many risk factors, pregnant women reported lower prevalence of risk than preconception women (Table [2](#Tab2){ref-type="table"}). Percentages reporting alcohol consumption and smoking were much lower for pregnant women than preconception women. For example, 53.9% of preconception women reported any use of alcohol in the past month, compared with 10.7% of pregnant women, 19.4% of preconception women reported current smoking in contrast to 8.4% of pregnant women, and, in the 12 states that included the question, 44.8% of preconception women reported not taking vitamins with folic acid versus 19.9% of pregnant women. For some risk indicators, no statistically significant difference was observed between women in the preconception period and pregnant women. The lack of awareness about methods to prevent mother-to-child HIV transmission (PMTCT) was not significantly different between preconception women and pregnant women (38.2% vs. 34.1%). However, a small but significantly *higher* percentage of pregnant women reported HIV risk (6.9%) compared to preconception women (4.3%), which is perhaps related to sexual risk behaviors being causally linked to both HIV and to unplanned pregnancies. Of concern, is that although a lower percentage of pregnant women compared with preconception women had never been tested for HIV, about one-fourth of pregnant women remained untested, despite national recommendations that call for all pregnant women to be tested for HIV \[[@CR7]\]. In the case of chronic diseases such as diabetes, the lower level of disease among pregnant women may indicate self-selection of diabetic women to avoid pregnancy. Of note, nearly one fifth of the women in preconception period were obese (defined as having a body mass index or BMI of greater than 30) and twice that number were overweight (BMI of greater than 25) (Table [2](#Tab2){ref-type="table"}). Most preconception and pregnant women did not report consuming the recommended 5 servings of fruits and vegetables per day \[[@CR8]\], although the percentage was lower for pregnant women (64.7%) compared with women in the preconception period (74.9%). Indicators of general well being and recent dental care were not different between women in preconception period and pregnancy. A composite measure of 3 basic pregnancy risks was formed, frequent alcohol consumption, current smoking and lack of an HIV test, in order to use the BRFSS data to illustrate the size of the target populations for preconception care. These are risk factors for which we have data that are generalizable to the US population, and for which interventions or information are widely available or promoted. The results indicate that 54.5% of preconception women reported one or more of these 3 risk factors, which represents approximately 1 million preconception women (Table [3](#Tab3){ref-type="table"}); 32.0 percent of pregnant women had one or more of these risks, which is equivalent to 835,000 pregnant women. Table [3](#Tab3){ref-type="table"} also shows this composite risk factor by 2 measures of access to health care, whether the woman had a health plan or insurance and whether she had a personal physician. Pregnant women had a lower percent at risk in all categories except for those without health insurance. Among women without insurance, the percentage reporting risk was not significantly lower for pregnant women, compared with preconception women. This suggests that women with poorer access to health care, as indicated by no health insurance, are not as likely as others to adopt healthier behaviors when they become pregnant. Table 3Percentage and estimated number reporting any of 3 risks for pregnancy outcome^*a*^ for preconception and pregnant women aged 18--44 Women 18--44, 2004 BRFSSPreconception women (Estimated number)Currently pregnant women (Estimated number)PercentCI1000sCINPercentCI1000sCINTotal54.5^\*^3.71,02494228632.03.18351012949No health plan/insurance63.411.72225232852.710.315650303Others52.5^\*^3.780279195729.43.2679882646No personal doctor64.1^\*^9.02785944846.39.223473467Others51.6^\*^3.974573183728.62.9601712482*Note.* 95% confidence interval = Percent +/− CI and Estimated number +/− CI. Estimated number is the estimated number of women in the category with one of the risk factors.^*a*^Any of 3 risk factors: frequent drinking, current smoker, no HIV test.^\*^difference between preconception and pregnant women significant, *p* \< .05. Discussion {#Sec5} ========== The 2004 BRFSS provides an opportunity to measure the extent of risk behaviors on a relatively large sample of women who are planning to become pregnant. These nationally-representative survey data indicate that major health risks were reported by substantial proportions of US women in the preconception period. The data suggest that among those at risk during preconception, the majority continued the risk into pregnancy. Women appear to respond to messages regarding behaviors that can improve pregnancy outcomes such as consuming of folic acid, and the reduction in smoking, alcohol consumption, but many remain unaware of the benefits of available interventions to prevent HIV transmission and birth defects. The fact that more than one-third of preconception and pregnant women were not aware of methods to prevent mother-to-child HIV transmission and did not know about the benefits of taking folic acid during pregnancy suggests that education and public information programs could have some effect in reducing risk behaviors. Risk reduction appears to be lower among women who have no health insurance. Other data are not available for comparison of the overall burden of risk for adverse pregnancy outcomes among women in the preconception period or pregnancy. However, data on selected indicators presented here for pregnant women are comparable to the reported range from other studies \[[@CR4], [@CR6], [@CR8]--[@CR11]\]. The finding that the percentage who report some risk behaviors is significantly lower among pregnant women compared with women in the preconception period may indicate the desire of many pregnant women to adopt healthier behaviors to achieve the best possible outcome for their pregnancy. Examination of three well known risk factors for which interventions are available (frequent alcohol consumption, current smoking, or never having been tested for HIV) illustrates that risks for adverse pregnancy outcomes exist among more than half of women in the preconception period and among one-third of pregnant women. Nationwide, this represents approximately one million women in the preconception period and 835,000 pregnant women. The data from the BRFSS are subject to potential errors associated with survey research such as exclusion from the sampling frame, non-response and reporting errors. BRFSS is a telephone-based survey and might not be representative of the small percentage of the population without telephones. The BRFSS methodology contains steps to ensure accurate estimation, including the use of weighting factors to compensate for non-response rates. In addition, estimates from BRFSS data have been found to be consistent with data from other surveys \[[@CR5]\]. The preconception and pregnant women identified were distributed very similarly to women giving birth in 2004 by age group and race/ethnicity, providing support for the representativeness of the data. In this report, the reported risk behavior of women in the preconception stage and pregnant women have been compared as a proxy for the type of behavior change occurring between preconception and pregnancy, but other factors could affect the actual change in behavior among individual women. For known risk factors, progress in efforts to reduce such risks among pregnant women has been slow. For example, as indicated in data collected on earlier rounds of the BRFSS, alcohol consumption among pregnant women has remained relatively steady over the past decade \[[@CR10]\]. Even among those who reduce consumption of alcohol, tobacco, or drugs, post pregnancy relapse of such behaviors have been found to be very high \[[@CR4]\]. Of the 17 maternal and infant health objectives included in the Healthy people 2010 objectives, progress has been made toward the target in 8 objectives \[[@CR12]\]. Little positive progress has occurred in the areas of maternal death, fetal alcohol syndrome, and low birth weight. To close the gap in reduction of poor maternal and child health outcomes, current maternal and child health initiatives should be complemented with alternate approaches, including more emphasis on preconception health promotion strategies among women of childbearing age. Recent experience in the prevention of alcohol-exposed pregnancy and prevention of neural tube defects through folic acid supplementation has shown that such efforts are useful. In 2003, the Project CHOICES Research Group reported the findings of a feasibility study designed to provide prevention counseling to women at high risk for an alcohol--exposed pregnancy \[[@CR13]\]. Brief interventions were found to be effective in reducing hazardous alcohol use in adults in a number of well-controlled studies. Interventions generally consist of advice, feedback, goal setting, and follow-up for further assistance and support \[[@CR14]\]. The use of a counseling style referred to as "motivational interviewing" has also proven to be effective in reducing problem drinking \[[@CR15]\]. Results of the initial follow-up assessment, suggest that this approach can be effective among women of childbearing age \[[@CR13]\], and a larger efficacy trial is now underway. To summarize, our results were consistent with a high prevalence of risk factors in the preconception period, and a high degree continuation of risk into pregnancy. This underscores the need for new approaches to complement current prevention efforts. Given that women of childbearing age in the United States have on an average 6.4 visits to physicians each year \[[@CR16]\], it may be feasible to inform them and engage them for interventions. Such contacts may provide an opportunity to assess and advise on many elements of health mentioned here. Our finding of high risk levels among women who had no insurance or no personal doctor strongly suggests that ensuring access to such services will remain the key determinant of maximum uptake of known intervention services. Medicaid and other funding programs can play an important role to address the access gap---whether during, before, or after pregnancy. Attempts to improve the health of would-be mothers not only improve pregnancy and infant health, but also help reduce the long-term impact of many risk factors on the women themselves.

Introduction {#s1} ============ Over the past few decades, studies have found that regulated intramembrane proteolysis (RIP) plays an important role in various kinds of cellular processes, including cell signaling, gene transcription and apoptosis [@pone.0037452-Brown1], [@pone.0037452-Wolfe1]. Regulated intramembrane proteolysis is a process through which proteases cleave substrates within their transmembrane regions. Three major intramembrane protease families have been identified, including the metalloprotease-type S2P family, the γ-secretase and signal peptide peptidase family and the rhomboid proteases. Rhomboid proteins are a family of widely conserved intramembrane serine proteases found from bacteria to humans [@pone.0037452-Urban1]. In Drosophila, they cleave the epidermal growth factor-like ligands to regulate the EGFR signaling pathway, playing important roles in signal transduction pathways and organism development [@pone.0037452-Lee1], [@pone.0037452-Urban2]. The mitochondrial rhomboid protease PARL has been found to regulate mitochondrial membrane remodeling and apoptosis [@pone.0037452-McQuibban1], [@pone.0037452-Cipolat1]. Rhomboid proteases are also involved in para-invasion and replication [@pone.0037452-Brossier1]. Although significant progress has been made since their discovery approximately ten years ago, the biological functions of most rhomboid proteases remain unclear. Few substrates have been identified for the rhomboid proteases, and the identification of these substrates will play a key role in understanding the function of this intriguing family of intramembrane proteases. Thus, finding novel substrates is an urgent need for the investigation of rhomboid proteases. Furthermore, only single transmembrane proteins have been identified as substrates of rhomboid proteases [@pone.0037452-Strisovsky1], [@pone.0037452-Jin1], [@pone.0037452-Adrain1]. However, because so many transmembrane proteins are multi-pass membrane proteins, there is a question regarding whether any multi-pass transmembrane proteins are substrates for rhomboid proteases. TSAP6/STEAP3 was initially identified as a p53-inducible gene implicated in apoptosis and cell cycle regulation [@pone.0037452-Passer1], [@pone.0037452-Amson1]. TSAP6 is a member of the six-transmembrane epithelial antigen of the prostate (STEAP) family and facilitates ferrireductase iron uptake in erythroid cells [@pone.0037452-Ohgami1], [@pone.0037452-SanchezPulido1]. However, the most intriguing finding is that TSAP6 plays a role in the regulation of nonclassical exosomal secretion, which may have a broad biological significance [@pone.0037452-Amzallag1], [@pone.0037452-Lespagnol1]. Exosomes, which are 30--100 nm in diameter, are derived from the internal vesicles of multivesicular bodies. They are secreted by various types of cells through a nonclassical secretion pathway [@pone.0037452-Nickel1], [@pone.0037452-Keller1]. Recent studies have highlighted the role of exosomes in various kinds of cellular processes, including cell maturation and differentiation, cancer progression and immune regulation [@pone.0037452-Li1], [@pone.0037452-Camussi1]. The function of the exosome is diverse and complicated and is related to its origin, its numerous components, and its targeted cells [@pone.0037452-Hwang1], [@pone.0037452-Zitvogel1], [@pone.0037452-Huber1], [@pone.0037452-Szajnik1]. However, the mechanism by which exosome secretion is regulated is still unclear. Exosomes originate from intracellular membrane vesicles, such as endosomes, lysosomes, and the plasma membrane, and they are secreted through fusion with the plasma membrane [@pone.0037452-Heijnen1], [@pone.0037452-Cocucci1]. Studies have found that various kinds of factors and proteins can influence exosome secretion, including calcium, DGKα, Rab family proteins and p53 [@pone.0037452-Savina1], [@pone.0037452-Alonso1], [@pone.0037452-Yu1], [@pone.0037452-Ostrowski1]. Interestingly, p53 has been found to promote exosome secretion through up-regulating TSAP6 transcription [@pone.0037452-Amzallag1]. We investigated the function of the mammalian rhomboid protein RHBDD1 and found that it could promote the proteolytic cleavage and subsequent proteasomal degradation of the pro-apoptotic Bcl-2 family member Bik [@pone.0037452-Wang1]. We also found RHBDD1 to be implicated in the regulation of spermatogenesis [@pone.0037452-Wang2]. Bik was found to interact with RHBDD1 using a yeast two-hybrid screen with RHBDD1 as bait. In addition to Bik, the multi-transmembrane protein TSAP6 was also identified in the screen. In this study, we investigated whether TSAP6 was a novel substrate of RHBDD1. We found that RHBDD1 induced the cleavage of multi-transmembrane protein TSAP6. The major cleavage site was mapped through mass spectrometry and mutagenesis experiments. We also show that the level of TSAP6-mediated exosomal secretion was elevated when RHBDD1 was mutated. Our findings may help clarify the mechanism by which rhomboid proteases are involved in the cellular secretion pathway in mammals. Results {#s2} ======= RHBDD1 overexpression induces the cleavage of TSAP6 {#s2a} --------------------------------------------------- Because RHBDD1 is a serine-type intramembrane protease, we hypothesize that transmembranal TSAP6 may be a substrate of RHBDD1. We co-transfected a Flag-tagged TSAP6 construct with RHBDD1 or with its catalytically inactive form, RHBDD1^S144A^ [@pone.0037452-Wang1]. When transfected alone, Flag-tagged TSAP6 appeared to be an approximately 55--70 kDa doublet on the immunoblot ([Fig. 1A](#pone-0037452-g001){ref-type="fig"}), probably as a result of posttranscriptional glycosylation, as previously reported [@pone.0037452-Lespagnol1]. However, the pattern of the bands of TSAP6 was different when it was cotransfected with RHBDD1. The intensity of the 55--70 kDa TSAP6 doublet decreased dramatically, and a lower band (approximately 38 kDa) appeared ([Fig. 1A](#pone-0037452-g001){ref-type="fig"}). However, RHBDD1^S144A^ did not show this effect, indicating that the catalytic activity of RHBDD1 is essential for the cleavage of TSAP6. In addition to the 38-kDa cleavage band, two weaker cleavage bands were detected when the blot was exposed for a longer period of time ([Fig. 1A](#pone-0037452-g001){ref-type="fig"}). The two weaker cleavage bands were visible in most cases during our investigation of the cleavage when exposed for longer periods. To determine whether RHBDD1 overexpression was responsible for the cleavage of TSAP6, the effect of the level of RHBDD1 on cleavage was investigated. The proteolytic cleavage of TSAP6 increased gradually when the expression of RHBDD1 was increased ([Fig. 1B](#pone-0037452-g001){ref-type="fig"}). Different parts of RHBDD1 were tested for their ability to induce the cleavage of TSAP6. RHBDD1 is mainly composed of an N-terminal rhomboid domain and a C-terminal domain ([Fig. 1C](#pone-0037452-g001){ref-type="fig"}). The N-terminal region (1--214 aa, mainly composed of the 59--214 aa rhomboid domain) of RHBDD1 was sufficient for the proteolysis of TSAP6, but its activity was much weaker than full-length RHBDD1 ([Fig. 1D](#pone-0037452-g001){ref-type="fig"}). {#pone-0037452-g001} The specificity of RHBDD1 for TSAP6 processing was also evaluated using another mammalian rhomboid protease, RHBDL2, for which several substrates have been identified [@pone.0037452-Adrain1], [@pone.0037452-Lohi1], [@pone.0037452-Pascall1]. TSAP6 was cotransfected with Myc-tagged RHBDD1 or RHBDL2 into HEK-293T cells to determine whether RHBDL2 overexpression might also result in TSAP6 cleavage ([Fig. 1E](#pone-0037452-g001){ref-type="fig"}). No cleaved fragments were detected when TSAP6 was co-expressed with RHBDL2, indicating that not all rhomboid proteases have the ability to induce the cleavage of TSAP6. Because TSAP6/STEAP3 is a member of STEAP family, we determined whether other members of the family (STEAP1 and STEAP4) could be cleaved when RHBDD1 was co-expressed. Because some of the proteins contained N-terminal signal peptides, rendering N-terminal tags undetectable, all of the STEAP members were subcloned into a p3xFlag-CMV-14 construct containing a C-terminal 3xFlag tag. As shown in [Fig. 1F](#pone-0037452-g001){ref-type="fig"}, no cleavage bands were detected when STEAP1 and STEAP4 were co-expressed with RHBDD1. Taken together, these results suggest that RHBDD1 could specifically induce the cleavage of TSAP6, indicating that TSAP6 is a potential target of RHBDD1-mediated proteolysis. RHBDD1-induced cleavage of TSAP6 occurs at multiple sites {#s2b} --------------------------------------------------------- Immunoblotting showed TSAP6 to be a 55--70 kDa doublet. This might be a result of posttranscriptional glycosylation. The posttranscriptional modification of TSAP6 hinders the mapping of TSAP6 cleavage sites, and is probably involved in the regulation of TSAP6 cleavage. In order to map the cleavage sites of TSAP6 and determine whether the posttranscriptional modification is involved in the regulation of TSAP6 cleavage, the type of posttranscriptional modification of TSAP6 first had to be examined. The 70 kDa band seemed to result from N-glycosylation because it faded off when treated with a nonspecific N-glycosidase, glycopeptidase F ([Fig. 2A](#pone-0037452-g002){ref-type="fig"}). Two potential glycosylation sites, Asn256 and Asn344, were predicted according to the NetNGlyc 1.0 server ([www.cbs.dtu.dk/services/NetNGlyc/](http://www.cbs.dtu.dk/services/NetNGlyc/)). To determine whether these two sites were the glycosylation sites of TSAP6, both of the Asn residues were serially mutated to Ile, and the ability of RHBDD1 to mediate cleavage of these mutant proteins was determined ([Fig. 2B](#pone-0037452-g002){ref-type="fig"}). The TSAP6^N256I^ mutation significantly reduced the glycosylated form of TSAP6, and the TSAP6^N256I+N344I^ double mutation completely abolished the glycosylated band. This confirmed that the 70 kDa band resulted from glycosylation. However, when RHBDD1 was co-transfected, the TSAP6 mutant proteins were cleaved normally, indicating that the cleavage did not rely on the N-glycosylation of TSAP6. {#pone-0037452-g002} TSAP6 is a member of the oxydoreductase family, and it contains six transmembrane domains [@pone.0037452-SanchezPulido1]. The total transmembrane domains of TSAP6 were predicted with SMART software (smart.embl-heidelberg.de/) ( [Table 1](#pone-0037452-t001){ref-type="table"}). 10.1371/journal.pone.0037452.t001 ###### Predicted transmembrane domains of TSAP6 using SMART software. {#pone-0037452-t001-1} Predicted transmembrane domains --------------------------------- ----- ----- TM1 209 231 TM2 258 280 TM3 305 327 TM4 359 381 TM5 394 416 TM6 431 453 To locate the cleavage regions of TSAP6, the samples in [Fig. 2B](#pone-0037452-g002){ref-type="fig"} were immunoblotted using an anti-V5 antibody to determine the C-terminal cleavage fragments of TSAP6. When RHBDD1 was co-transfected, three major C-terminal fragments were detected. They were approximately 38 kDa, 28 kDa, and 18 kDa in mass ([Fig. 2C](#pone-0037452-g002){ref-type="fig"}). We have shown that there were also three N-terminal cleavage fragments of TSAP6 detected. They were of approximately 38 kDa, 30 kDa and 22 kDa ([Figs. 1A](#pone-0037452-g001){ref-type="fig"}, [2B](#pone-0037452-g002){ref-type="fig"}). These cleaved fragments were found to be relevant and added up to approximately the molecular weight of the full-length TSAP6 when the N-terminal and C-terminal fragments were added in pairs. Thus, we proposed that these fragments are cleaved fragments within three different regions. The major cleavage site is near the C-terminal one, generating an N-terminal 38-kDa cleaved fragment. This site was chosen for investigation of the mechanism of cleavage. TSAP6 is mainly cleaved around the C-terminal of its TM3 domain {#s2c} --------------------------------------------------------------- To determine the cleavage sites of TSAP6, mass spectrometric analysis was used. Flag-tagged N-terminal and C-terminal cleavage fragments were generated by co-expressing the pcDNA6-Flag-TSAP6 or p3xFlag-TSAP6 constructs with RHBDD1-HA in 293T cells. The fragments were then immunoprecipitated with anti-Flag agarose, separated with SDS-PAGE, and stained with Coomassie blue R-250 (N-terminal fragment, [Fig. 3A](#pone-0037452-g003){ref-type="fig"}; C-terminal fragment, not shown). The purified fragments were digested and subjected to mass spectrometric analysis. As shown in [Fig. 3B](#pone-0037452-g003){ref-type="fig"}, a series of N-terminal peptides (yellow) and C-terminal peptides (green) were identified, with a N-terminal peptide (YSFCLPLR) located in the C-terminal of the third transmembrane domain (TM3). One peptide (YDLVNLAVK) that was adjacent to the TM3 region was identified in C-terminal fragment. The resulting peptides indicated that a cleavage site might lie in the short sequence (RAHR) between the two peptides. {#pone-0037452-g003} In order to confirm that the region contains a RHBDD1-induced cleavage site, we mutated a series of residues of TSAP6 into phenylalanine and leucine ([Fig. 4A](#pone-0037452-g004){ref-type="fig"}). We found that the cleavage was dramatically reduced when residues around the C-terminal region of TM3 were mutated (L325F and 326-3L mutations, [Fig. 4B](#pone-0037452-g004){ref-type="fig"}). However, mutations in other regions did not exert such a significant influence. L325F mutation significantly reduced the cleavage, while cleavage with 326-3L mutant protein was barely detectable, suggesting that the region around the C-terminal of the TSAP6 TM3 domain possesses a RHBDD1-induced cleavage site. {#pone-0037452-g004} The inactivation of endogenous RHBDD1 increases TSAP6-mediated exosome secretion {#s2d} -------------------------------------------------------------------------------- TSAP6 has been reported to regulate the nonclassical exosomal secretion pathway [@pone.0037452-Lespagnol1]. Because RHBDD1 induces cleavage in TSAP6 in cells, we evaluated the involvement of RHBDD1 in the TSAP6-mediated nonclassical secretion pathway. Recently, a recombinant-adeno-associated virus-based homologous recombination-mediated somatic cell knock-in approach was developed to introduce various kinds of alterations to endogenous loci [@pone.0037452-Hirata1]. RHBDD1 possesses a GFSGV motif in its rhomboid domain, and this motif is critical to its enzymatic activity [@pone.0037452-Wang1]. For this reason, we selected the knock-in approach to introduce two point mutations, G142A and S144A, into the endogenous loci of RHBDD1 in HCT116 colon cancer cells ([Fig. 5A](#pone-0037452-g005){ref-type="fig"}). After two rounds of targeting, HCT116 cells with mutant version of both alleles (RHBDD1-mt) were generated. They were verified using PCR analysis and DNA sequencing (data not shown). The total mRNA in HCT116-mt cells was extracted and reverse-transcribed into cDNA. Then, the cDNA encoding RHBDD1 was amplified and sequenced. This confirmed the mutations ([Figure. S1](#pone.0037452.s001){ref-type="supplementary-material"}). When immunoblotted with an anti-RHBDD1 antibody, the RHBDD1 protein was found to be barely detectable in RHBDD1-mt cells ([Fig. 5C](#pone-0037452-g005){ref-type="fig"}). This low level of expression was probably the result of rapid proteasomal degradation, as indicated by the fact that the proteasome inhibitor MG132 partially restored RHBDD1 in the mutated cells (data not shown). To investigate endogenous TSAP6, we generated a monoclonal anti-TSAP6 antibody. However, the antibody was shown to be rather poor in quality and there were non-specific bands in the blot, despite that it did recognize bands corresponding to TSAP6 ([Figure S2](#pone.0037452.s002){ref-type="supplementary-material"}). {#pone-0037452-g005} As shown in [Fig. 5C](#pone-0037452-g005){ref-type="fig"}, the level of TSAP6 was found to be significantly increased in RHBDD1-mt cells, which was probably the result of reduced RHBDD1-mediated proteolysis of endogenous TSAP6. Because the change in the level of TSAP6 could also have resulted from increased gene transcription, the relative mRNA level of TSAP6 was quantified using RT-PCR. No significant differences were found ([Fig. 5D](#pone-0037452-g005){ref-type="fig"}). Similar results were also observed with Bik, and the mRNA of RHBDD1 was found to be slightly elevated in RHBDD1-mt cells. In this way, abrogation of RHBDD1-mediated proteolysis seemed to be responsible for the increase in TSAP6 protein level in RHBDD1-mt HCT116 cells. Next, we analyzed whether RHBDD1 inactivation had an influence on TSAP6-mediated exosome secretion. According to a published protocol for exosome isolation and characterization, we isolated the secreted exosomes in wild-type and RHBDD1-mt HCT116 cells through differential ultracentrifugation approach [@pone.0037452-Thery1]. The secreted exosome proteins were quantified and found to be significantly elevated when RHBDD1 was mutated ([Fig. 5B](#pone-0037452-g005){ref-type="fig"}). The experiment was repeated at least four times, and similar results were obtained. Because some non-specific contaminations might be associated with exosomes purified by differential ultracentrifugation, we further detected two exosomal components, tumor susceptibility gene 101 (Tsg101) and transferrin receptor (Tf-R) to determine whether there was an increase of exosome components in the samples. Gp96 (glycoprotein 96) was used as a control to determine whether there was any contamination with cellular debris in exosome samples ([Fig. 5C](#pone-0037452-g005){ref-type="fig"}). The exosomal components Tsg101 and Tf-R were found to be significantly elevated in exosome samples from RHBDD1-mt cells, indicating that RHBDD1 inactivation could significantly increase the quantity of total secreted exosomes. Besides, the isolated exosome samples were evaluated through sucrose gradient ultracentrifugation, which is an accurate method to define the exosomal fraction [@pone.0037452-Thery1]. Both of the exosomal components were found to accumulate mostly in the 1.08--1.23 g/ml fractions, indicating that the components are located in exosome fractions ([Fig. 5E](#pone-0037452-g005){ref-type="fig"}). In order to determine whether elevated exosome secretion caused by RHBDD1 inactivation was reproducible in other types of cells, we introduced the same mutations into endogenous RHBDD1 in RKO colon cancer cells. Exosome secretion was assayed in wild-type (RKO wt) and RHBDD1-mt (RKO mt) cells ([Fig. 6A](#pone-0037452-g006){ref-type="fig"}). The exosome components, Tsg101 and Tf-R, were also detected in equal amount of exosome samples ([Fig. 6B](#pone-0037452-g006){ref-type="fig"}). Both of total secreted exosomes and the exosomal components were elevated significantly when RHBDD1 was mutated. These findings indicated that RHBDD1 inactivation in HCT116 and RKO cells had very similar influence on exosome secretion. In this way, RHBDD1 mutation was shown to be associated with elevated exosome secretion, indicating that RHBDD1 might have a role in the regulation of exosome secretion. {#pone-0037452-g006} It has been documented that TSAP6 played a role in regulating exosome secretion. Thus, we speculated that RHBDD1 might facilitate the regulation of exosome secretion through modulating the level of TSAP6 in cells. In order to determine whether the elevated exosome secretion in RHBDD1 mutant cells was TSAP6-dependent, we knocked down the expression of TSAP6 in wild-type and RHBDD1-mt HCT116 cells, and assayed exosome secretion in these cells. Exosome secretion was found to be significantly decreased in both wild type and RHBDD1-mt HCT116 cells when TSAP6 was knocked down ([Fig. 6C](#pone-0037452-g006){ref-type="fig"}). More importantly, the ratio of secreted exosomes between RHBDD1-mt and wild-type cells was reduced dramatically when TSAP6 was knocked down, from 2.05 (21.6 µg/10.5 µg) to 1.25 (5.5 µg/4.4 µg), which is consistent with the reduced difference in TSAP6 expression. The differential levels of exosomal components in wild type and RHBDD1-mt cells were also reduced significantly when TSAP6 was knocked down ([Fig. 6D](#pone-0037452-g006){ref-type="fig"}). These findings indicated that the elevation of exosome secretion by RHBDD1 inactivation was very likely to be TSAP6-dependent. RHBDD1 inactivation promotes exosome-induced apoptosis in Jurkat cell {#s2e} --------------------------------------------------------------------- Recent studies have highlighted exosomes as a multifunctional organelle with biological significance. Exosomes have been found to participate in cell-cell communication, tumor progression and immune regulation through specialized exosomal components. Recently, FasL and Trail have been identified as exosomal components secreted from cancer cells. They can cause apoptosis in T lymphocytes through the activation of their corresponding receptors [@pone.0037452-Huber1]. For this reason, we determined whether exosomes secreted from HCT116 cells also contained these ligands and whether exosomes secreted from RHBDD1-mt cells had enhanced biological activity relative to wt HCT116 cells. Jurkat cells were used as the model for T lymphocytes in this assay [@pone.0037452-Huber1]. As shown in [Fig. 7A](#pone-0037452-g007){ref-type="fig"}, FasL and Trail were both detected in the exosome samples. In addition, levels of both ligands were increased in exosomes secreted from RHBDD1-mt cells. Furthermore, the exosomes from RHBDD1-mt cells had significantly enhanced ability to induce apoptosis in Jurkat cells. This ability appeared to be primarily dependent on FasL and Trail because FasL and Trail antibodies dramatically attenuated the apoptotic effect of exosomes ([Fig. 7B](#pone-0037452-g007){ref-type="fig"}). Taken together, these results show that RHBDD1 has influences on the secretion of the exosomal components FasL and Trail, which partly contribute to the role of exosomes in inducing the apoptosis in Jurkat cells. {#pone-0037452-g007} Discussion {#s3} ========== It has been a decade since rhomboid proteases were identified, and great progress has been made since the discovery of this intriguing intramembrane protease family [@pone.0037452-Urban3]. In the present study, we show that TSAP6 was cleaved specifically in response to RHBDD1 overexpression. Further, we identified a cleavage site in the C-terminal region of TSAP6 TM3 domain, which resembles an intramembrane cleavage site. Finally, endogenous TSAP6 became enriched when RHBDD1 was inactivated, indicating that RHBDD1 could restrict the level of endogenous TSAP6 protein in cells. Taken together, these results indicate that TSAP6 is a target of RHBDD1-induced proteolysis. Besides, TSAP6 was identified through yeast two-hybrid assay with RHBDD1 as bait, implicating that TSAP6 might have direct binding with RHBDD1 and was very probably catalyzed directly by the protease ([Fig. 8](#pone-0037452-g008){ref-type="fig"}). However, the substrate-recognition-mechanisms of rhomboid proteases are diverse and complicated and the model for how RHBDD1 recognizes and cleaves its substrates has not yet been proposed. Thus, whether TSAP6 is directly cleaved by RHBDD1 remains to be a question of doubt and needs to be verified when there is better understanding for the protease. {#pone-0037452-g008} To data, all well-confirmed rhomboid substrates are type I membrane proteins. However, evidences indicated that other types of transmembrane proteins might also be the substrates of rhomboid proteases. Type II membrane protein Star is cleaved in transmembrane region in response to Rhomboid-1 overexpression [@pone.0037452-Tsruya1]. GlpG was shown to cleave synthetic multi-pass membrane proteins [@pone.0037452-Erez1]. In this study, intact multi-pass membrane protein was found to might also be cleaved by rhomboids. Intriguingly, both of the proposed multi-pass membrane proteins seemed to be cleaved in type II orientation. These findings indicated that rhomboid proteases might have a broader substrate profile than previously supposed. We found that RHBDD1 inactivation enriched endogenous TSAP6 and increased exosome secretion in cells. Because it has been established that TSAP6 promotes exosome secretion, it is highly possible that the role for RHBDD1 in regulating exosome secretion mainly involves restricting TSAP6 [@pone.0037452-Amzallag1], [@pone.0037452-Lespagnol1]. By knocking down the endogenous TSAP6, the increase in exosome secretion induced by RHBDD1 inactivation was reduced. This was consistent with the idea that TSAP6 is a possible intermediate for RHBDD1\'s regulatory role in exosome secretion. A large percentage of rhomboid proteases are involved in the cellular secretion pathway. Our findings indicate that RHBDD1 is involved in the regulation of exosomal secretion probably through modulating the amount of TSAP6. This might play a role in inducing apoptosis in T lymphocyte. In this respect, the exosomes secreted from cancer cells might contribute to immune suppression and cancer progression, as previously reported [@pone.0037452-Valenti1], [@pone.0037452-Liu1]. However, an increasing number of studies have indicated that exosomes are multi-functional organelles with biological significance. The biological functions of exosomes are complicated and depend on various factors, including the numerous components of the exosomes, the cells from which they originate, and the cells with which they interact. For example, exosomes originating from immune cells, such as dendritic cells and B cells, have been found to play a role in regulating antigen presentation, which might promote immune systems to target cancer cells [@pone.0037452-Denzer1], [@pone.0037452-Kovar1]. In this way, the biological consequence of RHBDD1 in affecting exosome secretion remains unclear. More work must be performed to determine with certainty the biological significance of exosome secretion and how RHBDD1 plays a role in this regard. Materials and Methods {#s4} ===================== Ethics statement {#s4a} ---------------- The Animal Ethics Committee of National Research Institute for Family Planning Beijing approved the animal experimentation protocols. All animal experiments were performed in accordance with the Guidelines for the Care and Use of Laboratory Animals as established by the Chinese Council on Animal Care. All the protocol numbers we received from the Animal Ethicics Committee of National Research Institute for Family Planning are as follows: 20090316, 20091016 and 20100505. Cell Culture, cDNAs, and Transfection {#s4b} ------------------------------------- HEK-293T (Cell Resource Center, PUMC) cells were cultured in Dulbecco\'s modified Eagle\'s medium (Invitrogen) supplemented with 10% fetal calf serum. HCT116 (Cell Resource Center, PUMC) and RKO (Cell Resource Center, PUMC) cells were cultured in Iscove\'s modified Dulbecco\'s medium (Invitrogen) supplemented with 10% fetal calf serum (Invitrogen). The 293T cells were transfected using Entranster D (Engreen Biosystem Co, Ltd.). Jurkat cells (Clone E6-1, Cell Resource Center, PUMC) were cultured in RPMI-1640 medium supplemented with 10% fetal calf serum (Hyclone). The cDNA encoding TSAP6 was amplified from a 293T cDNA library using PCR using the following primers: sense, 5′-ATCAAGCTTACCATGCCAGAAGAGATGGAC-3′, and anti-sense, 5′-ATCCTCGAGGTGACCGTACGTGGCTCGTCTTCTCGGCC-3′. A Flag tag (DYKDDDDK) was added in the NheI and HindIII sites of the pcDNA6-V5/His-B plasmid (Invitrogen). The TSAP6 PCR product was subcloned into the HindIII and XhoI sites of the pcDNA6-Flag/V5 construct. The TSAP6 mutations were constructed using overlapping extension PCR. RHBDD1 and its S144A mutation constructs were described previously [@pone.0037452-Wang1]. Myc-tagged RHBDD1- and RHBDL2-expressing vectors were constructed by inserting the coding sequences into the HindIII and XhoI sites of pcDNA6-Myc/His-B (Invitrogen). The C-terminal 3xFlag-tagged STEAP1, STEAP4 and TSAP6 constructs were generated by inserting the coding sequences into the HindIII and XbaI sites of p3xFlag-CMV-14 (Sigma). Antibodies, Co-immunoprecipitation, and Immunoblotting {#s4c} ------------------------------------------------------ The following antibodies were purchased: rabbit polyclonal anti-RHBDD1 (HPA013972, Sigma), Flag (F3165, Sigma), Myc (sc-40, Santa Cruz), HA (H3663, Sigma), GAPDH (TA-08, ZSbio), Gp96 (bs-0194R, Bios Biotech), Tsg101(bs-1365R, Bios Biotech), Tf-R (sc-32272, Santa Cruz), FasL (\#4273, Cell Signaling, for western blot analysis), FasL (556371, BD Biosciences, for antibody-blocking assays), Trail (550515, BD Biosciences), and HRP-conjugated goat anti-mouse and goat anti-rabbit antibodies (ZSbio). The monoclonal anti-TSAP6 antibody was generated through the immunization of a BALB/c mouse (Laboratory Animal Centre of PUMC, Beijing) with a 6×His-tagged purified recombinant protein corresponding to the N-terminal 1-126-amino acid fragment of human TSAP6. Hybridomas were screened by an indirect enzyme-linked immunosorbent assay (ELISA) using the 6×His-tagged recombinant protein. Immunoblotting assays were performed as described previously [@pone.0037452-Wang1]. Immunoprecipitation and mass spectrometric analysis {#s4d} --------------------------------------------------- 5×10^6^ 293T cells transfected with 8 µg RHBDD1-HA and 8 µg pcDNA6-Flag-TSAP6 or p3xFlag-TSAP6 plasmids were lysed in lysis buffer (20 mM Tris, pH 7.5, 200 mM NaCl, 1% NP-40, 10% glycerol, 1 mM DTT and completed protease inhibitor cocktail (Roche)). They were then immunoprecipitated with 40 µl anti-Flag slurry ( A2220, Sigma). The immunoprecipitates were separated with 10% SDS-PAGE and stained with Coomassie blue R-250. The protein bands were cut and used for in-gel digestion with trypsin. The resulting peptides were analyzed with LTQ Orbitrap Velos mass spectrometer (Thermo Scientific). RNA extraction and relative quantitative real-time PCR {#s4e} ------------------------------------------------------ RNA was extracted from HCT116 cells using TRIzol reagent (Invitrogen) according to the manufacturer\'s instructions. The total RNA was reverse-transcribed into cDNA using ReverTra Ace-α (Toyobo). Relative quantitative RT-PCR was performed using the TaqMan Universal PCR Master Mix (Applied Biosystems) on an iCycler IQ5 PCR machine (Bio-Rad). The primers for RT-PCR were as follow: RHBDD1: sense, 5′-CTCTGGGACCGAGGAAATACC-3′, and anti-sense, 5′-ACCTCACTGGCTATCGAATCTGT-3′.;TSAP6: sense, 5′-TGCAAACTCGCTCAACTGGAG-3′, and anti-sense, 5′-GAAGGTGGGAGGCAGGTAGAA-3′; Bik: sense, 5′-ACCATGGAGGTTCTTGGCA-3′, and anti-sense, 5′-AGGCTCACGTCCATCTCGTC-3′; Actin: sense, 5′-AGGCCAACCGCGAGAAGAT-3′, and anti-sense, 5′-TCACCGGAGTCCATCACGAT-3′. Exosome secretion and isolation {#s4f} ------------------------------- Exosome isolation and characterization assays were carried out in accordance with a published protocol [@pone.0037452-Thery1]. For the exosome secretion assay, 2×10^7^ HCT116 or RKO cells (1×10^7^ cells were used in TSAP6 knockdown assay) were plated and cultured for 24 h. The cells were incubated with exosome-free IMDM medium, which was prepared using ultracentrifugation at 100,000 g for 16 h at 4°C. After 16 h of incubation, the medium was collected and centrifuged twice at 300 g for 5 min to remove cell debris. The medium was further centrifuged at 12,000 g for 30 min and filtered through a 0.22 µm filter. Exosome pellets were isolated using centrifugation at 100,000 g for 2 h at 4°C in a Beckman L-100 XP Ultracentrifuge. The isolated exosomes were washed with a large volume of PBS and resuspended in 100 µl PBS. The amount of exosomes secreted was quantified using a BCA protein assay kit (Thermo Scientific). The sucrose gradient used to purify exosomes was similar to a method described previously [@pone.0037452-Lespagnol1]. Briefly, 40 µg of total exosomes was centrifuged at 100,000 g for 16 h in 2.2 ml of a 0.25 M--2.5 M sucrose gradient at 4°C in a Beckman LT-100 Ultracentrifuge. Each fraction (300 µl) was added with 75 µl of 5× SDS sample buffer, boiled for 5 min and analyzed using western blotting. Deglycosylation assay {#s4g} --------------------- The treatment of TSAP6 with glycopeptidase F (Takara Biotech) was performed under denaturing conditions according to the manufacturer\'s protocol. Exosome-induced apoptosis assay {#s4h} ------------------------------- 1×105 Jurkat cells were incubated with 5 µg of purified exosome samples for 60 h, stained with annexin-V-FITC/PI (Neobioscience) and analyzed using an Accuri C6 flow cytometer (BD Biosciences). For antibody-blocking assays, the exosome samples were pre-incubated with 1 µg/ml anti-FasL and anti-Trail antibodies for 30 min. TSAP6 RNAi assay {#s4i} ---------------- Lentivirus human TSAP6-targetting oligonucleotides were synthesized and packaged by GENECHEM Inc. The TSAP6 targeting sequences was: 5′- GAGGGAGTTCAGCTTCGTTCA -3′. HCT116 colon cancer cells were infected with TSAP6-shRNA-lentivirus or control shRNA virus. At 72 h after infection, the cells were subjected to exosome secretion assay. Supporting Information {#s5} ====================== ###### **Alignment of knock-in cDNA and the coding sequence of RHBDD1.** Note that two mutations were detected in the knock-in cDNA of RHBDD1, which codes an inactivated form of RHBDD1. (DOC) ###### Click here for additional data file. ###### **Verification of TSAP6 antibody with cell lysates.** Protein samples were seperated with SDS-PAGE and blotted with monoclonal anti-TSAP6 antibody. Line 1, HCT116 cell lysate. Line 2, 293T cell lysate transfected with TSAP6-Flag. (TIF) ###### Click here for additional data file. We acknowledge the Core Instrument Facility of Institute of Basic Medical Sciences for instrumental assistance in our research. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by grants from the National Important Research Plan of China (2011CB944302), grant for the youth (81101855) from the National Natural Science Foundation of China, the State Key Laboratory Special fund (2060204) from the Ministry of Science, and the Research Fund for the Doctoral Program of High Education (20111106110023). The experimental study was performed in compliance with the laws of the People\'s Republic of China. 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: LW SM WS. Performed the experiments: CW WS JF. Analyzed the data: CW WS JF. Contributed reagents/materials/analysis tools: YW WS. Wrote the paper: CW WS.

Introduction {#s1} ============ In recent decades, breast cancer cases have increased worldwide [@pone.0053902-Pedraza1] and today are the main cause of cancer mortality and morbidity in women [@pone.0053902-Hortobagyi1]. Breast cancer is caused by environmental and genetic factors [@pone.0053902-Song1]. Two main genes, *BRCA1* and *BRCA2*, are associated with hereditary breast cancer. Sporadic cases of breast cancer, however, may be related to variants in low-penetrance genes such as polymorphisms [@pone.0053902-Martin1], [@pone.0053902-Rodrigues1]. Lifetime exposure to estrogen is another factor that plays an important role in breast cancer. It is known that this hormone is involved both in the development of the mammary gland, as well as in the pathogenesis and progression of breast cancer [@pone.0053902-Germain1]. Based on this, the study of genes related to biosynthesis and the metabolism of estrogen is one way to identify possible candidate genes for breast cancer risk [@pone.0053902-Dumas1]. One of them is the *CYP19* (P450arom) gene. The human *CYP19* gene is located in the chromosome 15q21.2 region and is comprised of a 30 kb coding region [@pone.0053902-Bulun1]. This gene encodes for aromatase, an enzyme whose function is to catalyze the conversion of androgens into estrogens, a reaction known as aromatization. In premenopausal women, the main source of estrogens is the ovaries. Meanwhile, in postmenopausal women, aromatization takes place elsewhere such as in adipose tissue, skin, muscle, and liver cells. There have been several epidemiological studies of polymorphisms on the *CYP19* gene with the aim of finding associations between genetic variations and breast cancer risk. Some have found an association with an increased risk of breast cancer, such as that of a tetra-nucleotide repeat polymorphism in the intron 4 (TTTA)n [@pone.0053902-Fasching1]. Still, other polymorphisms studied have not shown a clear association with breast cancer risk, thus generating a situation of inconsistent results. This is the case of a C/T single nucleotide polymorphism (SNP) located in the 3′ untranslated region (3′-UTR) of the *CYP19* gene (rs10046). Some studies have linked this polymorphism with breast cancer risk [@pone.0053902-Kristensen1], however, others show different results [@pone.0053902-Haiman1], [@pone.0053902-Ralph1], [@pone.0053902-Zhang1]. This discrepancy in results led us to conduct a case-control study of this SNP in a population in Valencia (Spain). Additionally, we performed a meta-analysis of this polymorphism for the first time. It is a powerful tool for overcoming the problems of the small sample size and inadequate statistical power of genetic studies. This approach gives more reliable results than a single case--control study can. The aim was to collect all results published to date about this polymorphism and to obtain conclusive results about their relevance in susceptibility to breast cancer. Materials and Methods {#s2} ===================== Case-control Study {#s2a} ------------------ ### Study population {#s2a1} Association analysis between the rs10046 polymorphism (*CYP19*) and breast cancer disease was performed in a case-control study. The study was performed in a Caucasian Spanish population composed of 522 breast cancer patients and 1221 controls recruited at the Clinic Hospital of Valencia (Spain) with a mean age at diagnosis of 51 years (range 21--89) and 51 years (range 18--86), respectively. The controls were women without malignant pathology recruited at the blood donor bank and women with non-malignant pathology from the menopausal unit of the same hospital. The recruitment of the cases and controls was performed in the same interval of time ±0.5 years. The research protocols were approved by the ethics committee of the Institute of Health Research INCLIVA before the study began. All the participants in the study gave their written informed consent to participate in the study. ### Genotyping {#s2a2} Genomic DNA was extracted from blood samples using the DNeasy tissue kit from Qiagen (Izasa, Madrid, Spain) or DNA Isolation Kit by MOBIO (Carlsbad, CA, USA) using minor modifications to the manufacturer protocol. A final elution volume of 100 µl was established. DNA quantity was measured by absorbance at 260 nm using a NanoDrop spectrophotometer, and DNA purity was evaluated by measurement of the 260/280 absorbance ratio. DNA samples were stored at −20°C. Genotyping analysis was performed by real-time PCR (5 ng/ul DNA), using the TaqMan SNP Genotyping Assays C\_\_\_8234731_30 (Applied Biosystems) according to the manufacturer instructions. Thermal cycling and detection was performed on the ABI Prism 7900 using the Sequence Detection Software (Applied Biosystems). The results were analyzed using the allelic discrimination assay program of Sequence Detection Software version 2.4 (Applied Biosystems). ### Statistical analysis {#s2a3} The SNPs genotype analysis of our study population was done with the SNPstats software [@pone.0053902-Sole1] (allele and genotype distributions, association test, Hardy-Weinberg Equilibrium (HWE)). SNPstats association was based on binary logistic regression according to the response variable providing odds ratios (ORs), the confidence interval (CI), and the p-values for multiple inheritance models (dominant, recessive, over-dominant co-dominant, and log-additive). The lowest Akaike's Information Criterion and Bayesian Information Criterion value indicate the best inheritance genetic model for each specific polymorphism. Meta-analysis {#s2b} ------------- ### Literature search strategy for identification of the studies {#s2b1} We did a literature search in PubMed, Scopus and EBSCO data bases using the terms "breast cancer and rs10046", along with additional terms such as "polymorphisms, SNPs and CYP19", and all possible combinations. The studies for the meta-analysis were selected when they satisfied the following criteria: studies published by March 2012, case-control studies in humans, studies with genotype frequencies or OR data, information about HWE and information about procedure (adjusted or not, subgroups, etc.). In order to search more deeply, we reviewed the references of the selected articles to retrieve data that we could have ignored in the initial search. ### Data extraction {#s2b2} Information was extracted from the articles by two of the authors following the criteria listed above (B.P. and P.E.). Disagreement was discussed and resolved between the two authors. In the event that a study presented subpopulations, these were taken to be different studies. The same was done for studies composed of a first set and a second independent validation set. ### Statistical analysis {#s2b3} Raw data from comparable studies were analyzed jointly using likelihood methods. The estimate of association with breast cancer risk was evaluated using the fixed effect method [@pone.0053902-Mantel1] which calculates the ORs and the corresponding 95% CIs for individual studies and the global association. When the test was heterogeneous, the random-effects method [@pone.0053902-DerSimonian1] was applied. If heterogeneity was not corrected, we performed an influence analysis to determine the study responsible for that variability. Recessive, dominant, co-dominant, additive and over-dominant models were computed. The analysis of heterogeneity between studies was performed by the Q statistic, with p-values \<0.1 indicating significant heterogeneity [@pone.0053902-Lau1]. We also used the I^2^ statistic to quantify heterogeneity; values of 25% correspond to low heterogeneity, 50% to moderate heterogeneity and 75% to high heterogeneity [@pone.0053902-Higgins1]. The meta-regression was performed with the SPSS package (Statistical Package for the Social Sciences, version 19.0), and a variable was considered a source of heterogeneity when the p-value was significant in the ANOVA analysis. Publication bias was assessed by funnel plots of effect sizes versus standard errors to identify significant asymmetry. Additionally, Egger's linear regression test [@pone.0053902-Egger1] and Begg's test [@pone.0053902-Begg1] were performed to evaluate the potential bias. The Gleser-Olkin method [@pone.0053902-Gleser1] was used to estimate the number of unpublished studies. Accumulative meta-analysis [@pone.0053902-Jeng1] was assessed by year of publication to evaluate the possible publication bias by time. Statistical analysis of association was done with SPSS, version 19.0. Results {#s3} ======= rs10046 Genotype in a Spanish Population {#s3a} ---------------------------------------- The distribution of the genotype frequencies in this polymorphism within the control group is in agreement with that expected under HWE with a p-value of \<0.05. We also observed that the frequencies in this study were similar to those previously reported in the European population described by HapMap (<http://hapmap.ncbi.nlm.nih.gov/>). Our results show an association between the rs10046 polymorphism on the *CYP19* gene and breast cancer risk. The carriers of at least one C allele (dominant model) have 1.29 times increased risk of developing breast cancer (95% CI 1.01--1.66, p-value = 0.038) *vs*. non-carriers. The distribution of the rs10046 genotype among cases and controls and risk of breast cancer is listed in [Table 1](#pone-0053902-t001){ref-type="table"} (codominant and dominant models). 10.1371/journal.pone.0053902.t001 ###### Genotypic and allelic frequencies of rs10046 and breast cancer risk. {#pone-0053902-t001-1} Polymorphism Genotype Cases(n = 522) Controls(n = 1221) OR (95%CI)^a^ p-value -------------- ---------- ---------------- -------------------- ------------------- --------- rs10046 TT 109 (20.9%) 311 (25.5%) Reference *CYP19* TC 278 (53.3%) 629 (51.5%) 1.26 (0.97--1.64) 0.094 CC 135 (25.9%) 281 (23.0%) 1.37 (1.02--1.85) 0.038 TC+CC 413 (79.1%) 910 (74.5%) 1.29 (1.01--1.66) Meta-analysis Study Characteristics {#s3b} ----------------------------------- A total of 14 studies were selected from the bibliography search regarding the association between the rs10046 polymorphism on the aromatase *CYP19* gene and breast cancer. Two of these were excluded for not having been case-control studies [@pone.0053902-Colomer1], [@pone.0053902-GarciaCasado1]. Additionally, of the remaining 12 studies, 3 were excluded for not stating the genotype distribution [@pone.0053902-Fasching1], [@pone.0053902-Olson1], [@pone.0053902-Raskin1]. Afterwards, the studies presenting different subpopulations according to ethnicity [@pone.0053902-Iwasaki1] or having a validation set [@pone.0053902-Ralph1] were divided into independent studies. Finally, 12 studies were accepted for the first association analysis using the genotype distribution in cases and controls ([Figure 1](#pone-0053902-g001){ref-type="fig"}). [Table 2](#pone-0053902-t002){ref-type="table"} lists the main characteristics of these studies: first author, publication year, original country, ethnicity, source of controls, genotype distribution of controls and cases, HWE p-value, mean age and menopausal status. {#pone-0053902-g001} 10.1371/journal.pone.0053902.t002 ###### Main characteristics of the meta-analysis papers and genotype distribution of rs10046. {#pone-0053902-t002-2} FIRST AUTHOR YEAR COUNTRY ETHNICITY SOURCE GENOTYPE DISTRIBUTION (CONTROLS) GENOTYPE DISTRIBUTION (CASES) HWE MEAN AGE MENOPAUSAL STATUS -------------------- ------ --------------------------- ----------------- -------- ---------------------------------- -------------- ------------- ------------- -------------- ------------------------------- -------- ---------------------------------------- ------------------------------------------------------------------------------------------- Olson 2006 US (Minnesota) Caucasian (94%) H NA -- Haiman 2002 US (Nurses' Health Study) Caucasian H 167 (27%) 310 (51%) 134 (22%) 118 (26%) 240 (52%) 103 (22%) -- Raskin 2009 Israel (Jews Ashkenazi) Jews H NA \>0.05 Ralph 2007 USA\* Caucasian H 883 (26.8%) 1650 (50.1%) 758 (23.1%) 461 (28.1%) 830 (50.6%) 349 (21%) 0.6 cases 9.8(±9.3) controls 49.8(±9.3) Ralph (validation) 2007 USA\* Caucasian H 274 (27.3%) 503 (50.1%) 222 (22.1%) 142 (28.3%) 231 (46%) 129 (25.7%) \>0.05 Zhang 2008 China Asian H 120 (30.8%) 176 (45.1%) 94 (24.1%) 94 (31.3%) 151 (50.3%) 55 (18.3%) 0.43 cases 52.04±11.81 controls 50.41±9.15 cases controls TT CT CC TT CT CC pre 29% 49% 22% 30% 42% 28% post 34% 52% 14% 31% 50% 18% Kristensen 2000 Norway Caucasian HP 53 (23%) 114 (48%) 69 (29%) 146 (31%) 240 (50%) 95 (19%) \>0.05 Colomer 2008 Spain Caucasian H 17 (26%) 29 (45%) 19 (29%) -- Fasching 2008 Germany Caucasian H 362 (29%) 606 (48.6%) 279 (22.4%) \>0.05 Dunning 2004 Inglaterra Caucasian H 1049 (28.9%) 1773 (48.9%) 808 (22.2%) 739 (28%) 1286 (48.9%) 610 (23.1%) 0.3 cases 22--65 controls 45--74 Yoshimoto 2011 Japan Asian H 60 (21.7%) 120 (43.3%) 97 (35%) 160 (19.3%) 427 (51.7%) 239 (29%) \>0.05 Garcia-Casado 2010 Spain (Valencia) Caucasian H 23 (24.2%) 57 (60%) 15 (15.8%) \>0.05 Chen 2008 China Asian H 277 (31.6%) 436 (49.8%) 163 (18.6%) 178 (29.1%) 308 (50.4%) 125 (20.5%) \>0.05 range 30--64 Iwasaki^1^ 2009 Japan Asian H 69 (17.8%) 194 (50%) 125 (32.2%) 82 (21.1%) 188 (48.5%) 118 (30.4%) \>0.05 cases 53.8 controls 54.0 cases controls pre 12% 9% post 88% 91% Iwasaki^2^ 2009 Brasil (japanese) Asian H 13 (16.5%) 44 (55.7%) 22 (27.8%) 14 (17.7%) 41 (51.9%) 24 (30.4%) \>0.05 cases 56.6 controls 56.5 cases controls pre 8% 8% post 92% 92% Iwasaki^3^ 2009 Brasil Caucasian H 58 (15.3%) 200 (52.8%) 121 (31.9%) 67 (17.7%) 179 (47.2%) 133 (35.1%) \>0.05 cases 52.4 controls 52.5 cases controls pre 11% 10% post 89% 90% Pineda 2012 Spain Caucasian H 311 (23.1%) 629 (51.5%) 281 (25.4%) 109 (20.9%) 278 (53.3%) 135 (25.9%) 0.094 cases 51 (21--89) controls 51 (18--86) cases controls TT CT CC TT CT CC pre 22% 59% 19% 25% 53% 22% post 21% 50% 29% 26% 51% 23% Meta-analysis Results {#s3c} --------------------- When the eligible studies were pooled in the meta-analysis, no significant association with breast cancer risk was found in any of the genetic models. The dominant model did not present homogeneity when either the fixed effect model or the random effect model were applied. We assessed an analysis of influence to discriminate if one of the studies was causing the heterogeneity. The Yoshimoto et al. [@pone.0053902-Yoshimoto1] study shows the highest p-value for the heterogeneity test (p = 0.16), revealing it as the more discordant of all in the comparison. Consequently, we eliminated this study from further analysis. Additionally, the fixed effect model did not show homogeneity for two of the assayed models. To explore whether the sources of heterogeneity were mean age and menopausal status, we performed a meta-regression analysis. Neither of these variables explained the heterogeneity (F = 3.042; p = 0.247). The random effect model was chosen for all the analyses. For the dominant model, the odds ratio obtained was 0.99 (95% CI; 0.91--1.08) ([Figure 2](#pone-0053902-g002){ref-type="fig"}). This result shows no association with the presence of the C allele and a predisposition to breast cancer. {#pone-0053902-g002} Similarly, we proceeded to evaluate the possible association using the co-dominant ([Figures 3](#pone-0053902-g003){ref-type="fig"} and [4](#pone-0053902-g004){ref-type="fig"}), recessive (supplementary [Figure 1](#pone-0053902-g001){ref-type="fig"}), additive (supplementary [Figure 2](#pone-0053902-g002){ref-type="fig"}) and over-dominant models (supplementary [Figure 3](#pone-0053902-g003){ref-type="fig"}). In all cases, the analysis presents an OR, again, nearly 1, and the 95% CI crossed this limit. We can conclude that neither of the models presents an association with the risk to develop breast cancer for the rs10046. {#pone-0053902-g003} {#pone-0053902-g004} Subgroup Analysis {#s3d} ----------------- Considering the possible impact of ethnic variations, we then performed two subgroup analyses in European, Asian and American populations, and in Caucasian, Asian and others ethnicities. The analyses of these groups failed to suggest an association between rs10046 and breast cancer risk in any of the models ([Table 3](#pone-0053902-t003){ref-type="table"}). 10.1371/journal.pone.0053902.t003 ###### Association between rs10046 and breast cancer risk stratified by subgroups. {#pone-0053902-t003-3} GENETIC MODEL ------------------- ------- --------------- -------------- -------------- --------------- --------------- ------------------ -------- ------------ -------- ------------ -------- ------------ 0 4 1.03 0.89--1.18 0.95 0.86--1.06 0.95 0.86--1.05 0.95 0.84--1.08 0.95 0.85--1.06 0.95 0.84--1.07 1 3 0.94 0.70--1.26 1.00 0.76--1.32 0.98 0.72--1.34 0.94 0.61--1.44 1.03 0.84--1.27 1.02 0.93--1.11 2 4 1.08 0.79--1.48 1.01 0.86--1.18 0.99 0.84--1.17 0.94 0.72--1.21 1.02 0.86--1.21 1.15 0.91--1.44 **GENETIC MODEL** **Subgroup B** **n** **RECESSIVE** **DOMINANT** **ADDITIVE** **CC vs. TT** **CT vs. TT** **OVERDOMINANT** **OR** **95% CI** **OR** **95% CI** **OR** **95% CI** **OR** **95% CI** **OR** **95% CI** **OR** **95% CI** 0 6 0.99 0.86--1.14 1.00 0.89--1.13 1.00 0.87--1.14 0.99 0.82--1.20 1.01 0.91--1.11 1.00 0.94--1.07 1 4 1.08 0.79--1.48 1.01 0.86--1.18 0.99 0.84--1.17 0.94 0.72--1.21 1.02 0.86--1.21 1.15 0.91--1.44 2 1 1.15 0.85--1.56 0.84 0.57--1.24 0.87 0.60--1.27 0.95 0.62--1.46 0.77 0.52--1.16 0.80 0.60--1.07 Subgroup A corresponds to American (0: Haiman, Ralph, Ralph validation, Iwasaki3), European (1: Kristensen, Dunning, Pineda) and Asian (2: Chen, Iwasaki1, Iwasaki2, Zhang) populations. Subgroup B corresponds to Caucasian (0: Haiman, Ralph, Ralph validation, Kristensen, Dunning, Pineda), Asian (1: Chen, Iwasaki1, Iwasaki2, Zhang) and other ethnicities (2: Iwasaki3). OR: odds ratio, CI: confidence interval. The Random effect model was used to perform the analysis. Evaluation of Reliability {#s3e} ------------------------- Heterogeneity is a potential problem when performing a meta-analysis. In the present study, we systematically performed an evaluation of the heterogeneity in each model using the Q test and I^2^ statistic. The breach of the homogeneity condition led to the use of a less restrictive model (random effect model instead of fixed effect model). When the analysis was not homogeneous even using the random effect model we performed an influence analysis to discriminate study outliers and a meta-regression to explore whether the source of heterogeneity was the mean age and menopausal status variables. An influent study was detected as the source of heterogeneity and was eliminated from further analysis. Another limitation of meta-analysis is publication bias. To check for potential problems, we performed two statistical analyses: Begg's and Eggeŕs. The Begǵs method checks the correlation between an effect and its variability. This analysis calculates the Kendall's tau (τ) coefficient. The absence of statistical significance suggests no publication bias. Of the three models assayed, only the recessive model showed significance (p-values = 0.14, 0.02 and 0.89 for dominant, recessive and over-dominant models, respectively). To confirm these results, we analyzed the data with the more sensitive Egger's testwhich uses a linear regression between the reduced measure of the effect and the precision. The existence of bias was evaluated by the significance of the ordinate value in the origin for a value p\<0.1. This analysis confirmed there was no publication bias (p-values = 0.40, 0.52, 0.94 for dominant, recessive and over-dominant models, respectively). Using the Gleser-Olkin method, we estimated the number of unpublished studies based on the number of known publications and their p-values. The low limit of the 95% CI was a negative value that was compatible with the hypothesis of absence of publication bias. Furthermore, we performed a cumulative meta-analysis for year of publication to verify the influence of time on the results observed by different groups. There were no tendencies related to time that might affect the data published (data not shown). Discussion {#s4} ========== The importance of breast cancer worldwide has led to a substantial increase in research in this field. The present efforts to fight the illness are focused in its better classification [@pone.0053902-Eroles1] and treatment, but without forgetting the relevance of prevention and early diagnosis. Among the factors to take into consideration in the early detection of breast cancer is the exposure to estrogens and to other hormones. Prospective studies have shown a direct association between circulating sex hormones with the risk of developing breast cancer in postmenopausal women [@pone.0053902-Dunning1]. Furthermore, the circulating levels of estradiol precursors and metabolites have also been related to increased risk of breast cancer. These circulating levels are largely under genetic control and, consequently, can be modified by polymorphisms on genes associated with estradiol. Therefore, it is logical to think that changes in genes that control the levels of estrogen, as in the case of rs10046 (*CYP19*), are potential candidates to predispose for this illness (30). To date, several studies have been performed to evaluate this hypothesis in different populations, with contradictory results. Consequently, we have done a case-control study in a Spanish population with samples paired by age and menopausal status to assure that these variables have not affected the results. We obtained a significant association between carrying at least one C allele (dominant model) and the risk of breast cancer. This result was in agreement with previous works published, where the frequency of the C allele is higher in cases vs. controls [@pone.0053902-Haiman1], [@pone.0053902-Dunning1], [@pone.0053902-Chen1]. Especially relevant is the publication of Dunning et al. [@pone.0053902-Dunning1], where the study was carried out in more than 2000 cases and 3000 controls. Other authors however, have not detected a significant association or even found opposite results. Whereas Kristensen et al. found a predisposition role for the T allele [@pone.0053902-Kristensen1], other authors have not found a clear association with breast cancer risk [@pone.0053902-Ralph1], [@pone.0053902-Zhang1], [@pone.0053902-Iwasaki1]. The reason for these conflicting results could be that the studies were developed in different populations, geographical areas and, additionally, with a variable number of samples. Though some of these cases-control studies did not reach statistical significance, it is possible that the polymorphism contributed to levels of circulating sex hormones. At least some publications claim that this polymorphism is related to the levels of estradiol and the estradiol:testosterone ratio in normal postmenopausal women [@pone.0053902-Dunning1], a factor relevant in the development of breast cancer [@pone.0053902-Germain1]. Unfortunately, in our case-control study, serum samples were not available for the determination of estradiol, which would have added important information. Published expression data show higher estrogen levels with higher repeats in \[TTTA\]n polymorphism on *CYP19* found in linkage disequilibrium with rs10046 [@pone.0053902-Gennari1], [@pone.0053902-Haiman2], [@pone.0053902-Tworoger1]. Additionally, a highly significant relationship between aromatase SNPs and circulating estrogen levels among postmenopausal women has been found by Haiman et al. [@pone.0053902-Haiman3]. Between the highly correlated tagging SNPs, there were polymorphisms in different haplotype blocks, including rs10046. How this SNP can affect estrogen levels is not obvious. Different studies have shown SNPs that can affect phenotypic outcome by altering DNA binding sites [@pone.0053902-Bond1], [@pone.0053902-Lian1], [@pone.0053902-Skorupski1], mRNA stabilization, folding, splicing [@pone.0053902-Chen2], [@pone.0053902-Duan1], [@pone.0053902-Jin1] and modification of mechanisms involving the enhancement of transcription and the posttranslational regulation. All the above highlights the importance of studying in depth the possible association of C\>T rs10046 SNP with breast cancer risk. With that in mind, we have performed a meta-analysis, a reliable analytical tool for comparing the different data related to this polymorphism. The results of our meta-analysis showed no significant association with breast cancer risk in any of the genetic models tested. In all cases, the overall OR calculated is near the value of 1, indicating no existence of a trend or predilection for the rs10046 genetic variants between cases and controls. They were no significant differences in age and menopausal status across genotypes, as the meta-regression analysis revealed. The comparisons of subgroups based on population stratifications showed only slight discrepancies, data that in no case reached significance. The results of our work show that the case-control studies restricted to a limited population provide different results than do those with a wide representation of the population. To extrapolate results and to come to relevant conclusions of the possible influence of a factor in a disease seems essential to realize global studies. The meta-analysis approach allows us to obtain relevant conclusions and simultaneously to summarize and to unify the studies in the field. On the basis of our results, polymorphism rs10046 is not capable of modifying the risk to develop breast cancer. Nonetheless, several studies associated this polymorphism to circulating hormonal levels [@pone.0053902-Dunning1]. Additionally, rs10046 has been related to altered disease free survival in the subgroup of pre-menopausal breast cancer patients [@pone.0053902-Fasching1]. Polymorphisms in linkage disequilibrium with rs10046, as in the case of rs4664, rs700518 and rs700519, has been associated with variable efficacy of treatments [@pone.0053902-Colomer1], [@pone.0053902-GarciaCasado1], [@pone.0053902-Park1] and breast cancer survival [@pone.0053902-Darabi1], [@pone.0053902-Long1]. The data available is not sufficient to affirm that it is an activating polymorphism, but published data suggest that it could be related to an advantage in the protein structure that makes it more active [@pone.0053902-Colomer1]. In the absence of a mechanistic explanation, however, strong linkage disequilibrium with other polymorphisms remains possible. There are some studies about polymorphisms in linkage disequilibrium with rs10046 which describe a significant association with efficacy of the aromatase inhibitor letrozole in patients with breast cancer, as is the case for rs700158 [@pone.0053902-Park1] and rs4646 [@pone.0053902-Colomer1], [@pone.0053902-GarciaCasado1]. The latter SNP has also been reported to be associated with HER2 status of tumors [@pone.0053902-Fasching1], circulating steroid hormones [@pone.0053902-Haiman3] and histological grade and tumor size in postmenopausal women [@pone.0053902-Haiman3], [@pone.0053902-Darabi1]. Haiman et al. reported a significant relationship between this SNP and circulating estrogen levels among postmenopausal women. A haplotype analysis has also been performed on the *CYP19* gene in most of these studies. Some specific haplotypes, including rs10046, were associated with an increased risk of breast cancer with concurrent proliferative fibrocystic conditions [@pone.0053902-Chen1] and with clinical efficacy of letrozole [@pone.0053902-Park1]. Moreover, the haplotype studies conducted by Raskin et al. showed a trend to association with breast cancer risk in BRCA1 carriers aged \<50 years [@pone.0053902-Raskin1]. In our study, we have not done any analysis of polymorphism in linkage disequilibrium with rs10046. This could be a limitation, as it could have provided more information on the role of *CYP19* in breast cancer. Nevertheless, it should be noted that these studies were taken into account when we performed the meta-analysis even though some of them were excluded from the final statistical analysis due to lack of necessary data. In conclusion, despite the limitations, the results of the present meta-analysis suggest that rs10046, by itself, does not directly affect the risk to suffer breast cancer. Further extensive studies to clarify the influence of *CYP19* polymorphisms on estradiol circulation levels are necessary. Supporting Information {#s5} ====================== ###### **Meta-analysis of OR for rs10046 polymorphism associated with breast cancer (recessive model).** (TIF) ###### Click here for additional data file. ###### **Meta-analysis of OR for rs10046 polymorphism associated with breast cancer (additive model).** (TIF) ###### Click here for additional data file. ###### **Meta-analysis of OR for rs10046 polymorphism associated with breast cancer (over-dominant model).** (TIF) ###### Click here for additional data file. The authors thank sample donors and all the field workers who supported the data collection. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Reviewed the manuscript: MAG AC AL. Conceived and designed the experiments: MAG PE. Performed the experiments: BP PE. Analyzed the data: PE AL. Contributed reagents/materials/analysis tools: MAG PE AL AC. Wrote the paper: BP PE.

Introduction ============ The three-year mandate of the European Union Committee of Experts on Rare Diseases \[[@B1]\] drew to a close in July 2013 with an impressive record. These results were achieved thanks to the unique nature of the forum it provided for the discussion of key topics for the rare disease community, bringing together representatives from all 28 European Member States, experts, patient representatives and members of the Industry, as well as the European Commission \[Figure [1](#F1){ref-type="fig"}\]. The enthusiastic participation of the 51 members of the EUCERD allowed the EUCERD to foster exchanges of relevant experience, policies and practices in the field of rare diseases, and help the European Commission and the Member States prepare and implement activities in the in the field of rare diseases. {#F1} Established via the European Commission Decision of 30 November 2009 (2009/872/EC) \[[@B2]\], the EUCERD responded quickly to the evolving needs of the rare disease community \[Figure [2](#F2){ref-type="fig"}\] following the *Commission Communication Rare Diseases: Europe's Challenge* (2008) \[[@B3]\] and the *Council Recommendation on an Action in the Field of Rare Diseases* (2009) \[[@B4]\]. These texts defined a European policy in the field and notably encouraged Member States to elaborate national plans/strategies for rare diseases by the end of 2013 to *"\[guide\] and \[structure\] relevant actions in the field of rare diseases within the framework of their health and social systems"*. Although other non-European countries and world regions have policies in place relative to orphan medicinal products, Europe is the only region with a public health policy in the field of rare diseases. {#F2} The EUCERD selected from the key areas cited by the Commission Communication and Council Recommendation a number of topics which had to be tackled immediately in order to clarify concepts and guide the reflection underway at national level. Most of these selected topics had already been tackled by the European Commission's Rare Diseases Task Force (RDTF), the EUCERD's predecessor, and there was sufficient consensus to elaborate sets of recommendations for the European Commission and Member States. Supported via two consecutive Joint Actions between Member States (N°2008 22 91 and N°2011 22 01), the EUCERD thus organised a number of expert workshops over the years to elaborate the recommendations in areas concerning the organisation of expert services (centres of expertise and European Reference Networks), data collection, and access to medicines. For each topic the process started with the identification of the scope of the issue and a fact-finding mission concerning the state of the art of the topic. This information was used to produce a preliminary report, or a draft recommendation, which provided material in turn for an expert workshop to discuss and elaborate a recommendation which was then submitted for discussion and adoption by the EUCERD \[Figure [3](#F3){ref-type="fig"}\]. The Recommendations have been disseminated by the members of the Committee and through OrphaNews Europe \[[@B5]\], the newsletter of the committee with 14 000 registered readers. {#F3} Centres of expertise : the key stone of national plans and strategies for rare diseases and a step towards European networks ---------------------------------------------------------------------------------------------------------------------------- The first area of the *Council Recommendation* to be dealt with by the Committee concerned the centres of expertise for rare diseases and European Reference Networks for rare diseases. The identification, and creation, of centres of expertise for rare disease is a key element of the *Council Recommendation* and central to national rare disease plans/strategies. There are around 6 000 rare diseases and most are unknown to healthcare professionals so rare diseases patients suffer from not knowing where to consult. To overcome this, some Member States have established centres specialised in some rare diseases/groups of rare diseases which have proven to be very efficient in providing quality of care for patients. The networking of these centres could lead to the gathering of the scarce expertise concerning these diseases at European level, in order to ensure equal access to accurate information, appropriate and timely diagnosis and high quality care for rare disease patients. This area had been previously identified by the former RDTF as a priority topic with a dedicated working group, as networks of centres of expertise had been defined by the High Level Working Group on European Centres of Expertise as an area of European added-value with consensus on the benefits of a cross-border approach \[[@B6]\]. Indeed, the RDTF had previously established a list of criteria for designation of centres of expertise which served as the starting point for the EUCERD's reflection on this subject. Currently, the organisation of centres of expertise varies greatly from country to country: few countries currently have a designation process in place, and the designation criteria vary across these countries, and sometimes even from region to region. By revisiting the work of the RDTF, the EUCERD was able to draft and adopt recommendations relevant to the current situation concerning the definition of the missions, scope and criteria for designation and evaluation of centres of expertise. The EUCERD's *Quality Criteria for Centres of Expertise for Rare Diseases in Member States* (2011) \[[@B7]\] have been very well received in the EU Member States and are being actively used as a starting point in the elaboration of the actions concerning the organisation of care in their national plans/strategies. Member States have appreciated the clarification of concepts provided by the Recommendations which can be tailored to the countries' specificities. It is hoped that the use of these Recommendations will help to harmonise the criteria for designation of centres from country to country, in preparation for the establishment of the future European Reference Networks for rare diseases, for which centres of expertise are the building blocks \[Table [1](#T1){ref-type="table"}\]. ###### Key facts and state of play in European countries in December 2013: Centres of expertise for rare diseases (CE RD)^1^   --- ------------------------------------------------------------------------------------------------------------------------------------------------------------------ • 1 European country with designated CE RD in the scope of a national plan for rare diseases • 5 European countries with officially designated CE RD • 15 European countries with non-designated CE RD acknowledged by health authorities to varying degrees • 9 European countries with CE RD recognised by reputation only • 16 European countries with plans to designate CE RD in their national plans/strategies for RD • *EUCERD Recommendations on quality criteria for centres of expertise for rare diseases in Member States* adopted on 24 October 2011 • Consensus on 45 recommendations covering the mission and scope, criteria for designation, process of designation and evaluation, and European dimension of CE RD ^1^Data as of December 2013, reproduced from S. Aymé, C. Rodwell (eds.), **2014 Report on the State of the Art of Rare Disease Activities in Europe,** to be published in July 2014. European Reference Networks for rare diseases: the European dimension of health care pathways --------------------------------------------------------------------------------------------- Collaboration of centres of expertise through European Reference Networks (ERNs) is highlighted in the *Council Recommendation*, whilst the creation of such European structures is a key element of the *Directive on the application of patients' rights in cross-border healthcare (2011/24/EU)* (9 March 2011) \[[@B8]\], which specifically cites rare diseases as an area of interest for ERNs. The concept of ERNs for rare diseases had been developed over a number of years before the creation of the EUCERD by a dedicated RDTF working group mandated by the HLG. This working group ascertained that the treatment of rare diseases demands multidisciplinary care which is sometimes not available at local or national level and worked on establishing a set of recommendations concerning ERNs \[[@B9]\]. The EUCERD decided, once the concept of centres of expertise was well established, to take stock of this work in the context of the Directive in order to establish a recommendation for both the European Commission and the Member States. The *Recommendations on European Reference Networks for Rare Diseases* (2013) \[[@B10]\] highlights the specificities of rare diseases to be taken into account when considering the scope, mission, governance, designation and evaluation of such ERNs in the future activities planned around the Cross-Border Healthcare Directive. The reflections during the elaboration of the document have helped develop the concept of ERNs in the field of rare diseases. The adopted EUCERD Recommendations have provided guidance to Member States in process of elaborating the healthcare pathways at both the national and European levels in the scope of their national plans/strategies for rare diseases to ensure that this future possibility is envisaged from their inception. The Recommendations have also now fed into the process underway at the European Commission to prepare the legal framework for the implementation of the future ERNs, of which a number, it is hoped, will be dedicated to rare diseases. Stakeholders are now waiting for confirmation of how the sustainability and designation of these ERNs will be organised at European level. This question has been partially answered by the mention of support for projects concerning the modelling and validation of system methodologies for European Reference Networks in the Horizon 2020 call \[[@B11]\] \[Table [2](#T2){ref-type="table"}\]. ###### Key facts : European Reference Networks (ERNs) for rare diseases   --- ---------------------------------------------------------------------------------------------------------------------------------------------------------------- • Rare diseases cited in Directive on the application of patients' rights in cross border healthcare (2011/24/EU) (9 March 2011) as priority area for ERNs • *EUCERD Recommendations on European Reference Networks for Rare Diseases* adopted on 31 January 2013 • Consensus on 21 recommendations covering mission, vision and scope, governance, composition, funding and evaluation, and designation of ERNs for Rare Diseases Patient registration and data collection: Gathering information at European level on rare diseases -------------------------------------------------------------------------------------------------- Patient registries are a key aspect of national plans/strategies for rare diseases and are cited as a crucial source of information on rare diseases, in terms of basic and clinical research as well for epidemiological and public health purposes, to be supported at national and European level in the *Council Recommendation*. Patient registries are a key tool for gathering the scarce knowledge relevant to rare diseases so as to improve the understanding of these conditions and the treatment available to patients, as well as the planning of healthcare services for these diseases. Over 640 rare disease registries exist in Europe according to data extracted from Orphanet \[[@B12]\] in December 2013, with the majority concerning diseases of groups of diseases for which there is an innovative treatment either in development or already on the market. At the national level many countries are considering in the scope of their national plans/strategies the best way to collect data relative to rare disease patients, and at the Community level the European Commission is in the process of establishing a European Platform for Rare Disease Registration. The EUCERD thus decided to build on the previous work of the RDTF on patient registries to elaborate a set of *Recommendations on Rare Disease Patient Registration and Data Collection*\[[@B13]\] with the aim of setting down the consensus reached to date and to guide all stakeholders at this crucial moment in the collective reflection on the topic. The EUCERD's recommendations provide the consensus to date in the field and call for the involvement of all stakeholders in the designing, maintenance and governance of registries in the future, as well as public-private partnerships for long-term sustainability. These recommendations will serve all stakeholders in the field, including Member States establishing their national plans/strategies for rare diseases, the European Commission services in their reflection on the sustainability of registries for rare diseases at European level, and members of the Industry looking to fulfil their post-market authorisation obligations. These recommendations are also feeding into discussions on data collection and registration underway at international level in the working groups of the International Rare Disease Research Consortium (IRDiRC) \[[@B14]\] notably concerning interoperability and pooling of data for research purposes. Stakeholders at European and national level are now awaiting more information from the Commission concerning the mission of the European platform \[Table [3](#T3){ref-type="table"}\]. ###### Key facts and state of play in December 2013: Rare disease patient registration and data collection   --- --------------------------------------------------------------------------------------------------------------------------------------------------------------------- • Around 640 rare disease registries in Europe^1^ • Majority of registries are academic • Some RD have more than one registry, many RD have no registry • *EUCERD Core Recommendations on Rare Disease Patient Registration and Data Collection* adopted on 5 June 2013 • Consensus on 6 main areas : international operability, sources of data, collection of data, good practices, use of data for regulatory purposes, and sustainability ^1^ Data as of December 2013 from **Orphanet** \[<http://www.orpha.net>\] concerning European Member States and surrounding European countries in the Orphanet Consortium. Steps towards better and timely access to orphan medicinal products for rare diseases : Improving the Clinical Added Value of Orphan Medicinal Products information flow ------------------------------------------------------------------------------------------------------------------------------------------------------------------------ The EUCERD also worked, at the request of the European Commission following a Tender report on the subject, on the issue of equitable and timely access to approved orphan medicinal products for rare diseases. This matter is highlighted by the *Communication* and *Council Recommendation* where sharing of relevant knowledge is promoted in order to *"minimise delays in access to orphan drugs for rare disease patients*". Although the European Regulation on Orphan Medicinal Products (141/2000) \[[@B15]\] establishes a centralised procedure for the designation of orphan medicinal products and puts in place incentives for the research, marketing and development of orphan medicinal products, access to these medicines is often difficult and unequal from country to country. In order to advance the reflection on this matter, the EUCERD brought together key stakeholders to discuss these issues and adopted in 2012 a *Recommendation on Improving Informed Decisions Based on the Clinical Added Value of Orphan Medicinal Products Information Flow*\[[@B16]\]*.* The EUCERD's recommendation highlights potential ways to facilitate scientific information exchange in order to support Member States in their processes of making informed decisions on the scientific assessment of the clinical effectiveness of an orphan medicinal product. The EUCERD recommends creating an information flow between individual Member States and between Member States and EU bodies to bridge existing knowledge gaps, especially at the time of market authorisation. Ultimately this will accelerate access to approved orphan medicinal products by providing the most robust set of information possible, while encouraging pricing and reimbursement decisions based on the value of these products and promoting good medical practices throughout the EU. This recommendation is the fruit of dedicated participation from all stakeholders, and has been warmly welcomed by across the board. The EUCERD has submitted this proposal to the European Commission in order to inform the next steps in the process. Stakeholders now await further details concerning pilot initiatives concerning suggested actions at European level \[Table [4](#T4){ref-type="table"}\]. ###### Key facts and state of play in December 2013: The clinical added value of Orphan Medicinal Products^1^   --- -------------------------------------------------------------------------------------------------------------------------------------------------------------------- • 1234 positive opinions for orphan product designation from 1798 applications submitted since 2000 at EU level and a total of 1219 European Commission designations • 85 orphan designated products have received marketing authorisation by end of 2013 at EU level • *EUCERD Recommendation on Improving Informed Decisions Based on the Clinical Added Value of Orphan Medicinal Products Information Flow* adopted in September 2012 • Recommendation proposes four key time points for information sharing to improve the pricing and reimbursement decision process ^1^Data from the report of the European Medicines Agency\'s Committee on Orphan Medicinal Products December 2013 \[<http://www.ema.europa.eu/docs/en_GB/document_library/Committee_meeting_report/2014/01/WC500159429.pdf>\]. Indicators for rare disease national plans/strategies : Monitoring the progress made in national policy ------------------------------------------------------------------------------------------------------- To provide Member States elaborating their national plans/strategies for rare diseases with guidance concerning indicators to help monitor the elaboration and implementation of these plans at European level, the EUCERD has also adopted *Recommendations on core indicators for rare disease national plans/strategies* (2013) \[[@B17]\]. Indicators are vital tools for assessing the outcomes and success of these measures. These indicators were also intended to provide data for the European Commission's report on the implementation of the *Commission Communication* and *Council Recommendation* and to serve as a basis for indicators at national level. Although each country will have to tailor indicators to suit the measures foreseen in their national plan/strategy, the Recommendations may serve as a starting point for some countries. They will be reviewed in time to reflect the experience of Member States in the monitoring of their plans/strategies. Newborn screening: A definition of possible areas of European collaboration --------------------------------------------------------------------------- The EUCERD also decided to investigate areas of potential collaboration at European level in the field of newborn screening, at the request of the European Commission following a Tender report on the subject \[[@B18]\]. Currently, a great heterogeneity of practices can be observed from country to country in the EU and there is no systematic approach to this topic at EU level : this considered, the EUCERD took its cue from the *Council Recommendation* encourages that Member States "*develop European guidelines \[...\] population screening, while respecting national decisions and competences"*. After examining the results of the Tender study, the Committee was able to identify a number of possible topics for collaboration which respected the principle of subsidarity and after discussion on the subject at Committee level an Opinion \[[@B19]\] was adopted in July 2013. It was decided to not prioritise these elements, which were submitted as an Opinion to the European Commission, the Member States and third parties for further consideration. This document is an important first step in this complicated area. The EUCERD State of the Art Report: Monitoring progress and reporting on initiatives ------------------------------------------------------------------------------------ The EUCERD has played an important role in the monitoring and dissemination of the results of measures taken at Community and national level in the field of rare diseases, both through the annual report on the *State of the Art of Rare Disease Activities in Europe*\[[@B20]\] and the bi-monthly newsletter of the EUCERD *OrphaNews Europe*\[[@B5]\]*.* The *State of the Art* report is a comprehensive document updated each year which presents the current state of activities at both European and Member State level in a range of different areas of the rare disease field, such as rare disease national plans/strategies, expert services, patient organisations, research and orphan medicinal product policy. The first volume of the report presents an overview of the field which presents to the general public in an easily accessible manner the achievements to date and challenges for the future, whereas the following volumes present the environment in greater detail. The report is also available in country-specific volumes for reuse at national level to inform stakeholders of the state of play. Both the report and the newsletter are supported by the EUCERD Joint Action and entrusted to the Scientific Secretariat of the EUCERD, with valuable contributions and input from members of the Committee. Notably, the information collected for the *OrphaNews Europe* newsletter and *State of Art* report have contributed to the preparation of the Commission reports on the implementation of the *Commission Communication* and *Council Recommendation*, currently underway. This report from the Commission will help determine what has been achieved to date and the measures to be taken in the future in the field of rare diseases at European level. Promoting cooperation across Europe and beyond ---------------------------------------------- The EUCERD was also charged with assisting the Commission in international cooperation on matters relating to rare diseases. To this end, the EUCERD has welcomed representatives of third countries, such as Japan, to its meetings to partake in discussions and share experiences. The EUCERD has also promoted liaison with other groups implicated in the field of rare diseases (e.g. IRDiRC \[[@B14]\], EUnetHTA \[[@B21]\], European Partnership Action Against Cancer \[[@B22]\], PARENT Joint Action \[[@B23]\] by inviting representatives to their meetings and including them in expert workshops, as well as participating in their meetings and workshops, to ensure that relevant experience is shared across different groups and fields. The EUCERD Joint Action is also working hard to liaise with other EC-funded projects implicated in the field of rare diseases to ensure that results are disseminated and duplication of efforts are avoided. Conclusion ========== The unique forum of stakeholders provided by the EUCERD has been a key factor in the success of the Committee in advancing discussions on important topics for the rare disease community. The Members have shown dedication and perseverance over the past three years, helping the Committee to achieve much in a relatively short space of time. Great advances have been made in developing concepts, reaching consensus and establishing recommendations for a number of key topics which will be at the centre of national and European rare disease policy in the coming years. The successful collaboration and outputs of the EUCERD is a source of inspiration for countries and regions throughout the world. The EUCERD has laid down the foundations, but much more work is needed in the future as we move from the stage of elaboration to the stage of implementation of these initiatives and policies. It is hoped that the European Commission Expert Group on Rare Diseases, which replaces the EUCERD as of January 2014, will be as efficient and productive as the former Committee in their new mandate so as to not lose the momentum gathered under the EUCERD concerning the main questions and problems faced by the rare disease community. Competing interests =================== The authors declare that they have no competing interests. Authors' contributions ====================== SA and CR carried out the analysis of the actions of the EUCERD described in this manuscript, and CR drafted and finalised the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ The authors would like to thank all the Members of the EUCERD for their implication in the work described in this article. The work of the Scientific Secretariat of the European Union Committee of Experts on Rare Diseases was financed by two consecutive Joint Actions between Member States from the European Commission's Executive Agency for Health and Consumers: the Scientific Secretariat of the European Union Committee of Rare Diseases Joint Action N° 2008 22 91 and European Union Committee of Rare Diseases Joint Action: Working for Rare Diseases N° 2011 22 01.

Specifications tableSubject area*Ecology, Conservation Biology*More specific subject area*Urban ecology*Type of data*Tables, graphs*How data was acquired*Floristic inventories and vegetation relevés*Data format*Raw and aggregated*Experimental factors*A stratified sampling design with two types of gardens, and two crossed (independent) gradients: an urban intensity and a garden management intensity/habitat spatial heterogeneity gradient.*Experimental features*Gardens were chosen following a stratified sampling design, based on the urban habitat mapping key of the city of Zurich. The (independent) strata included i) garden type (domestic vs. allotment); ii) a garden spatial heterogeneity/management intensity gradient, ranging from extensively managed gardens with a high vertical vegetation structure and a high proportion of native plant species, to intensively managed gardens with a low vertical vegetation structure and a high proportion of alien plant species; and iii) an urban intensity gradient, which ranged from densely to less densely built-up areas of the city.*Data source location*City of Zurich, Switzerland; 47°22′N, 8°33′E*Data accessibility<https://doi.org/10.17632/452pj39jm2.2>Related research article*Young, C., Frey, D., Moretti, M., & Bauer, N. (2019). Research Note: Garden-owner reported habitat heterogeneity predicts plant species richness in urban gardens. Landscape and Urban Planning, 185, 222--227.*<https://doi.org/10.1016/j.landurbplan.2019.01.013>*.*[@bib1]**Value of the Data**•The data is comprehensive as it describes all vascular plants growing on entirely sampled garden lots with a high taxonomic resolution, and plants growing on standardized sampling plots.•The data can contribute to comparative studies of community assembly rules of spontaneously versus human assembled urban plant communities.•The data can contribute to comparative studies of garden floras to understand mechanisms of plant introductions and invasions.•The data can be used to investigate the effects of garden plants on diversity patterns of species of other trophic levels (e.g. herbivors, pollinators), for which data exists from the same gardens•The data can be used to investigate above-below ground interactions, as biotic and abiotic soil data exists in the same gardens. 1. Data {#sec1} ======= This article presents data of a survey of vascular plants in 85 urban gardens in the city of Zurich, Switzerland. Two garden types were investigated: allotment (*N* = 42) and domestic gardens (*N* = 43). In each garden, we applied two sampling methods: a floristic inventory of entire garden lots (mean area ±SD: 312 ± 155 m^2^) and sampling on plots of a standardized size (2 × 10 m^2^). The two plots were centred within the two main land-use types found in each garden: lawn, meadow, vegetable bed, flower bed or berry patch. We give the origin status for each of the 1081 taxa found, i.e., whether a taxon is considered native or alien to Switzerland ([Appendix](#appsec1){ref-type="sec"}). Similarly, for each taxon and garden, we estimate the origin of each plant or garden population by classifying each taxon as either cultivated or spontaneously growing ([Appendix](#appsec1){ref-type="sec"}). Species richness (i.e. number of taxa) of all (*S*~total~), native (*S*~native~), cultivated (*S*~cultivated~) and spontaneous (*S*~*s*pontaneous~) plants was computed for each garden and overall ([Table 1](#tbl1){ref-type="table"}). In addition, species richness levels of gardens and land-use types were visualized, and results of the two sampling methods were compared ([Fig. 1](#fig1){ref-type="fig"}, [Fig. 2](#fig2){ref-type="fig"}). For each garden, additional environmental data is given, such as garden type, area, and urbanization intensity, which was calculated as the landscape proportion of impervious (i.e. built and paved) surface within a 500-m radius ([Table 1](#tbl1){ref-type="table"}). The data is part of an inter- and transdisciplinary investigation of biodiversity, soil quality, ecosystem services and social value of urban gardens in Switzerland ([www.bettergardens.ch](http://www.bettergardens.ch){#intref0020}). The data can be linked to biotic and abiotic soil data [@bib2], [@bib3], [@bib4], to data of soil surface dwelling and flying arthropods, which were sampled in the same gardens and during the same period, and to arthropod and bird predation data [@bib5], [@bib6]. The raw data are available from Mendeley Data <https://doi.org/10.17632/452pj39jm2.2> [@bib7].Table 1Species richness (i.e. number of taxa) of all (*S*~total~), native (*S*~native~), cultivated (*S*~cultivated~) and spontaneous (*S*~*s*pontaneous~) vascular plants in 85 urban gardens in the City of Zurich, Switzerland. The data are based on a complete floristic inventory of garden lots. For each garden, garden type, area and landscape proportion of sealed surface within a 500-m radius is given. Sealed surface was defined as built and paved land-cover. Note that plants can belong to more than one category (i.e., native, cultivated, spontaneous), since a plant can occur spontaneously in one garden while it is cultivated in another, and both cultivated and spontaneous plants can be native.Table 1Garden IdGarden typeGarden area (m2)Landscape proportion of sealed surface (500-m scale)*S*~total~*S*~native~*S*~cultivated~*S*~spontaneous~1allotment210.30.190171103114572home285.10.39314878109393allotment183.90.6301155278374home399.50.2331458189565home666.80.27216587108576allotment169.50.3901266887397home238.10.671695322478home473.50.428794051289allotment198.80.2009761435410allotment246.10.536179991136611home659.20.4348950543512home476.20.42913065884213allotment197.00.4329156533814allotment172.60.21811361694415allotment256.10.2857843473116allotment307.50.24312059863417home346.90.2107137422918allotment179.60.21010960674219home510.40.81211469753920allotment496.00.20813991786121allotment223.90.273155841282722home170.20.7266452174723allotment212.50.20711355724124allotment205.30.29014883876125home695.70.29513275854726home400.10.55912071675327home135.90.81111770754228home407.50.61311060753529allotment734.50.36211477585630home273.60.7161851241454031allotment167.90.21110451693532home229.70.2849543732233allotment198.70.24811558694634allotment173.10.5396338273635home249.90.7249247563636home255.90.3739652643237home291.00.58213282953738allotment501.20.1599957534639home791.50.1919173236840home294.20.53410558713441home444.20.43013168894242home486.10.2482051691535243home107.40.6165339223144home497.20.51711862774145home355.50.574178951195946allotment212.50.19511346823147home286.80.472186971384848home422.20.72110254673549allotment169.10.32012553883750home150.90.6139237741851allotment184.50.360122491002252home366.70.4971741071383653allotment410.30.81812580715454allotment179.60.1877552314455allotment240.40.11112470715356home438.90.2491711231224957home121.70.8149850722658home551.40.383143102895459home455.30.4417545433260allotment220.20.1299960574261allotment150.40.1098848513762home291.70.66310355604363allotment370.90.83413067864464allotment201.60.548155751124365home479.60.21512168932866allotment232.40.42312350903367allotment507.40.279176871067068allotment494.20.4619461494569allotment200.70.287132501062670allotment242.40.1831921161395371home287.20.215177841314672allotment486.10.35310462584673home277.10.6977646463074allotment150.90.3088342503375allotment177.80.399154701094576allotment180.50.491131471052677allotment231.50.4058632711578home244.80.651173911314279home207.90.6219457524280allotment186.40.25214171875481allotment110.20.4396838422682allotment177.80.42113669973983home288.00.3699070553584home281.60.2568546562985home438.70.21682685329Mean ± SD311.6\ ±155.10.411\ ±0.195118.8\ ±34.766.5\ ±23.277.9\ ±30.840.9\ ±11.4Range107.4--791.50.109--0.83453-20532-16917-15315-70Fig. 1Violin plots illustrating the distribution of plant species richness of all (A), native (B), cultivated (C), and spontaneous (D) taxa growing on 10 m^2^ plots for each of the five land-use types based on all 85 gardens. Note the unequal sampling size among the land-use types: meadows (*N* = 11), lawns (*N* = 56), flower beds (*N* = 35), berry patches (*N* = 15) and vegetable beds (*N* = 47). Only the herbaceous vegetation layer was considered for meadows, lawn, and vegetables, and the tree layer was excluded altogether. Six plots were excluded due to pseudoreplication (e.g. two plots in lawn of the same garden).Fig. 1Fig. 2Comparison between the two sampling methods: Species richness (S) of plants growing in the two 10 m^2^ sampling plots versus plant species richness of the entire garden lot, based on data of all 85 urban gardens. Species richness of all (A), native (B), cultivated (C) and spontaneous (D) taxa are plotted. Note the different scales of the axes.Fig. 2 2. Experimental design, materials and methods {#sec2} ============================================= 2.1. Data source {#sec2.1} ---------------- The data was acquired in the city of Zurich, Switzerland (47°22′N, 8°33′E). Zurich is located in the temperate climate zone of Europe, with a mean annual temperature of 9.3 °C (1981--2010) and mean annual precipitation of 1134 mm [@bib8]. It harbours a population of 0.4 million in an area of approximately 92 km^2^, placing it in the globally most common city class [@bib9]. 2.2. Garden selection {#sec2.2} --------------------- We collected floristic data in 85 urban gardens (43 domestic and 42 allotment gardens). We defined a domestic garden to be a garden directly adjacent to a single-occupancy or terraced house. Our definition of allotment gardens encompasses spatially clustered garden lots on public land, managed by associations and leased to leisure gardeners as lots of 100m^2^--200m^2^ [@bib10]. Gardens were chosen following a stratified sampling design, based on visual criteria defined by the urban habitat mapping key to the city of Zurich [@bib11]. Potentially suitable gardens were identified based on the habitat map of the city of Zurich, aerial images and during field visits. We approached the garden owners initially by letter and thereafter by phone to arrange a visit. If no phone number was available, owners were approached personally. The (independent) strata included i) garden type (domestic vs. allotment), ii) a garden spatial heterogeneity/management intensity gradient, ranging from extensively managed gardens with a high vertical vegetation structure and a high proportion of native plant species, to intensively managed gardens with a low vertical vegetation structure and a high proportion of alien plant species [@bib11], and iii) an urban intensity gradient, which ranged from densely to less densely built-up areas of the city. The urban intensity gradient was quantified as the proportional area of impervious (i.e. built and paved) surface within a 500-m radius around each garden lot ([Table 1](#tbl1){ref-type="table"}). Variance in garden area was kept small and no novel gardens were included. To assure statistical independence among observations, no adjacent garden lots were sampled, and gardens were distributed across the city to include all urban districts. Additionally, with two exceptions, only one garden lot was sampled per allotment garden area. The average pairwise distance between gardens was 4.5 km (SD ± 2.2 km, min.-max. 0.1--11 km). 2.3. Floristic data of garden lots {#sec2.3} ---------------------------------- A complete floristic inventory of each garden lot was made during repeated garden visits in 2015, based on the standard determination literature of the Swiss, resp. European (garden) flora [@bib12], [@bib13], [@bib14], [@bib15], [@bib16], [@bib17]. Potted plants were included in the inventory. Abundance of each taxon was estimated semi-quantitatively on a six-point scale: 1 (1--10 individuals), 2 (11--25 ind.), 3 (26--50 ind.), 4 (51--100 ind.), 5 (101--250 ind.) and 6 (\>250 ind.). To account for the different vegetation periods, gardens were visited three times in March/April, May, and July/August. To standardize the sampling effort among gardens, the duration of each visit was restricted to about 1.5 h. Note that the early spring flowering genus *Crocus* L. was missed. The species richness of each garden is given in [Table 1](#tbl1){ref-type="table"}. The list of all taxa and the number of observations per taxon is given in the Appendix. The raw data is available from Mendeley Data <https://doi.org/10.17632/452pj39jm2.2> [@bib7]. 2.4. Sampling on standardized plots (vegetation relevés) {#sec2.4} -------------------------------------------------------- In each garden, in addition to the floristic inventory, plant species were sampled on two circular plots of 10 m^2^ each during the summer. The methodology was based on the survey of angiosperm diversity of the Swiss Biodiversity Monitoring program [@bib18]. Potted plants were not included. On each plot, three vegetation layers (ground vegetation, shrub and tree layer) were roughly distinguished and within each layer, the percentage cover of each taxon was scored on an ordinal scale: 1 (\<1% cover), 2 (1--5% cover), 3 (\>5--25% cover), 4 (\>25--50% cover), 5 (\>50--75% cover), 6 (\>75--100% cover). Note that due to the heterogenous vegetation structure of gardens, cover sums over 100% within a vegetation layer were allowed. The two plots were centred within the two main garden land-use types found in each garden: lawn, meadow, vegetable bed, flower bed or berry patch. In addition, the levels of soil disturbance were contrasted between the two selected land-use types [@bib2]: in each garden, one of the plots had to be in a low soil disturbance land-use type with mostly perennial vegetation (e.g. lawn), while the other had to be in a high soil disturbance land-use type with mostly annual vegetation (e.g. a vegetable bed). The distribution of species richness in each land use type is given in [Fig. 1](#fig1){ref-type="fig"}. A comparison between the two sampling methods is given in [Fig. 2](#fig2){ref-type="fig"}. The plot-based data can be linked to urban soil data of land-use types [@bib3]. The raw data is available from Mendeley Data <https://doi.org/10.17632/452pj39jm2.2> [@bib7]. 2.5. Taxonomic treatment, origin status and origin of garden populations {#sec2.5} ------------------------------------------------------------------------ Taxonomy largely followed the Checklist of the National Data and Information Centre of the Swiss Flora [@bib19]. In addition, for cultivated ornamental plants, Huxley & Royal Horticultural Society [@bib13], Jäger [@bib14], and Cullen [@bib15] were consulted. Taxa below the species level, taxa within species complexes, and cultivars were not consequently determined at the lowest possible taxonomic rank. They were mostly grouped into aggregates (e.g. *Taraxacum officinale* aggr.), Cultivar Groups (e.g. *Begonia* Semperflorens Cultorum Group), or labeleld CV as cultivars without further distinction (e.g. *Rosa* CV). Juillerat et al. [@bib19] were followed to assess the origin status of species, i.e. whether a taxon is considered native or alien to Switzerland. Our definition of native plants encompasses archeophytes, which are taxa introduced to Switzerland before 1500, and neophytes of European origin that have colonized Switzerland spontaneously. A more detailed description of the origin status is given in the raw data [@bib7]. Cultivar groups not derived from native plants were considered to be alien. The origin of a plant individual or "population" in a garden, i.e. whether a plant or a group of plants was cultivated, or whether it occurred spontaneously, was determined by consulting the Flora of the City of Zurich [@bib20] and/or by asking the garden owners. Intentionally introduced plants, which subsequently formed self-sustaining local populations, were considered to be cultivated. Plants that spontaneously colonized a garden and were subsequently tolerated or even locally favored by the garden owner/tenant were considered to be spontaneous plants. The origin of rare native plants was always verified by asking the garden owners. Meadow and lawn plants originating from seeding were considered to be cultivated plants. In lawns and meadows not deliberately enriched with herbs, only grasses employed in landscaping were considered to be cultivated [@bib13]. It is important to note that the origin of a garden population can not always be unambiguously retraced and should therefore be interpreted with caution: especially in the case of lawn and grassland plants. Note that a taxon can belong to more than one category (native, cultivated, spontaneous), since a species may occur spontaneously in one garden, while it is cultivated in another, and both cultivated and spontaneous plants can be native. {#appsec3} List and number of observations of vascular plants taxa in 85 urban gardens in the City of Zurich, Switzerland, based on the floristic inventory. OS, origin status (A, alien/ N, native); *N*~cult~, number of observations of garden populations of anthropic origin; *N*~spont~, number of observations of spontaneously occurring garden populations; *N*~total~, total number of observations of a taxon. Note that the early spring flowering genus *Crocus* L. was missed.Taxon nameAuthorityFamilyOS*N*~cult.~*N*~spont.~*N*~total~Abelmoschus esculentus(L.) MoenchMalvaceaeA101Abies albaMill.PinaceaeN202Abutilon × hybridumVossMalvaceaeA202Acanthus mollisL.AcanthaceaeA707Acer × conspicuumvan Gelderen & OterdoomSapindaceaeA101Acer campestreL.SapindaceaeN31417Acer palmatumThunb.SapindaceaeA19019Acer platanoidesL.SapindaceaeN167Acer pseudoplatanusL.SapindaceaeN31316Achillea filipendulinaLam.AsteraceaeA505Achillea millefolium aggr.AsteraceaeN14216Achillea ptarmicaL.AsteraceaeN303Acinos arvensis(Lam.) DandyLamiaceaeN101Aconitum napellusL.RanunculaceaeN808Actinidia arguta(Siebold & Zucc.) Planch. ex Miq.ActinidiaceaeA505Actinidia chinensisPlanch.ActinidiaceaeA808Actinidia kolomikta(Rupr. & Maxim.) Maxim.ActinidiaceaeA101Adenophora liliifolia(L.) A. DC.CampanulaceaeN101Adiantum capillus-venerisL.PteridaceaeN101Aegopodium podagrariaL.ApiaceaeN02626Aeonium haworthiiWebb & Berthel.CrassulaceaeA101Aesculus hippocastanumL.SapindaceaeA303Aethusa cynapiumL.ApiaceaeN022Agapanthus praecoxWilld.AmaryllidaceaeA303Agastache mexicana(Kunth) Lint & EplingLamiaceaeA303Agave americanaL.AsparagaceaeA101Ageratum houstonianumMill.AsteraceaeA101Agrimonia eupatoriaL.RosaceaeN044Agrostis capillarisL.PoaceaeN303Agrostis giganteaRothPoaceaeN022Agrostis stoloniferaL.PoaceaeN38038Ailanthus altissima(Mill.) SwingleSimaroubaceaeA022Ajuga reptansL.LamiaceaeN83442Akebia quinata(Houtt.) Decne.LardizabalaceaeA101Alcea roseaL.MalvaceaeA35035Alchemilla mollis(Buser) Rothm.RosaceaeA31031Alchemilla vulgaris aggr.RosaceaeN213Alisma plantago-aquaticaL.AlismataceaeN303Alliaria petiolata(M. Bieb.) Cavara & GrandeBrassicaceaeN02121Allium carinatumL.AmaryllidaceaeN101Allium cepaL.AmaryllidaceaeA20020Allium cristophiiTrautv.AmaryllidaceaeA18018Allium fistulosumL.AmaryllidaceaeA202Allium neapolitanumCirilloAmaryllidaceaeA101Allium nigrumL.AmaryllidaceaeA202Allium porrumL.AmaryllidaceaeA17017Allium schoenoprasumL.AmaryllidaceaeN41041Allium siculumUcriaAmaryllidaceaeA101Allium tuberosumRottler ex Spreng.AmaryllidaceaeA101Allium ursinumL.AmaryllidaceaeN03131Allium vinealeL.AmaryllidaceaeN101Aloe vera(L.) Burm. f.AsphodelaceaeA101Alopecurus pratensisL.PoaceaeN202Aloysia citriodoraPalauVerbenaceaeA707Althaea officinalisL.MalvaceaeA505Amaranthus blitum aggr.AmaranthaceaeN02020Amaranthus caudatusL.AmaranthaceaeA808Amaranthus hybridus aggr.AmaranthaceaeA033Amaranthus retroflexusL.AmaranthaceaeA022Amelanchier alnifolia(Nutt.) Nutt. ex M. Roem.RosaceaeA202Amelanchier lamarckiiF. G. Schroed.RosaceaeA11011Amelanchier ovalisMedik.RosaceaeN202Anagallis arvensisL.PrimulaceaeN02323Anchusa officinalisL.BoraginaceaeN101Androsace septentrionalisL.PrimulaceaeN101Androsace studiosorum(Duby) GovaertsPrimulaceaeA101Anemone blandaSchott & KotschyRanunculaceaeA707Anemone coronariaL.RanunculaceaeA101Anemone hupehensis(Lemoine) LemoineRanunculaceaeA29029Anemone nemorosaL.RanunculaceaeN03232Anemone sylvestrisL.RanunculaceaeN101Anethum graveolensL.ApiaceaeA808Angelica archangelicaL.ApiaceaeA101Angelica sylvestrisL.ApiaceaeN202Anthemis tinctoria(L.) J. GayAsteraceaeN505Anthericum ramosumL.AsparagaceaeN202Anthoxanthum odoratumL.PoaceaeN606Anthyllis vulnerariaL.FabaceaeN101Antirrhinum majusL.PlantaginaceaeA27027Apium graveolensL.ApiaceaeA22022Aquilegia skinneriHook.RanunculaceaeA101Aquilegia vulgarisL.RanunculaceaeN63063Arabidopsis thaliana(L.) Heynh.BrassicaceaeN044Arabis alpina subsp. caucasica(Willd.) Briq.BrassicaceaeA202Araucaria araucana(Molina) K. KochAraucariaceaeA101Arctium lappaL.AsteraceaeN134Argyranthemum frutescens(L.) Sch. Bip.AsteraceaeA303Aristolochia clematitisL.AristolochiaceaeN101Aristolochia macrophyllaLam.AristolochiaceaeA101Aristolochia rotundaL.AristolochiaceaeN101Armeria maritima(Mill.) Willd.PlumbaginaceaeA404Armoracia rusticanaG. Gaertn. & al.BrassicaceaeN16016Aronia melanocarpa(Michx.) ElliottRosaceaeA404Arrhenatherum elatius(L.) J. Presl & C. PreslPoaceaeN101323Artemisia abrotanumL.AsteraceaeA606Artemisia absinthiumL.AsteraceaeN606Artemisia campestrisL.AsteraceaeN101Artemisia dracunculusL.AsteraceaeA101Artemisia tridentataNutt.AsteraceaeA101Artemisia vulgarisL.AsteraceaeN606Arum italicumMill.AraceaeN606Arum maculatumL.AraceaeN088Aruncus dioicus(Walter) FernaldRosaceaeN347Asarum europaeumL.AristolochiaceaeN101Asclepias syriacaL.ApocynaceaeA303Asimina triloba(L.) DunalAnnonaceaeA101Asparagus densiflorus(Kunth) JessopAsparagaceaeA202Asparagus officinalisL.AsparagaceaeN808Asphodeline lutea(L.) Rchb.AsphodelaceaeA202Asplenium ruta-murariaL.AspleniaceaeN022Asplenium trichomanesL.AspleniaceaeN224Aster amellusL.AsteraceaeN101Aster dumosusHoffm.AsteraceaeA202Aster novae-angliaeL.AsteraceaeA15015Aster novi-belgii aggr.AsteraceaeA25025Astilbe Arendsii GroupSaxifragaceaeA17017Astilboides tabularis(Hemsl.) Engl.SaxifragaceaeA101Astrantia majorL.ApiaceaeN202Athyrium filix-femina(L.) RothAthyriaceaeN505Atriplex hortensisL.AmaranthaceaeA404Atropa belladonnaL.SolanaceaeN101Aubrieta deltoidea(L.) DC.BrassicaceaeA14014Aucuba japonicaThunb.GarryaceaeA404Aurinia saxatilis(L.) Desv.BrassicaceaeA11011Avena fatuaL.PoaceaeN011Avena sativaL.PoaceaeA044Baptisia australis(L.) R. Br.FabaceaeA101Barbarea vulgarisR. Br.BrassicaceaeN303Begonia Semperflorens Cultorum GroupBegoniaceaeA10010Bellis perennisL.AsteraceaeN85159Berberis julianaeC. K. Schneid.BerberidaceaeA202Berberis thunbergiiDC.BerberidaceaeA505Berberis verruculosaHemsl. & E. H. WilsonBerberidaceaeA202Berberis vulgarisL.BerberidaceaeN202Bergenia crassifolia(L.) FritschSaxifragaceaeA11011Beta vulgarisL.AmaranthaceaeA34034Betula pendulaRothBetulaceaeN358Bidens ferulifolia(Jacq.) DC.AsteraceaeA10010Bidens frondosaL.AsteraceaeA011Bletilla striata(Thunb.) Rchb. f.OrchidaceaeA303Borago officinalisL.BoraginaceaeA25025Bougainvillea spectabilisWilld.NyctaginaceaeA101Brachypodium pinnatum(L.) P. Beauv.PoaceaeN101Brachypodium sylvaticum(Huds.) P. Beauv.PoaceaeN02525Brachyscome iberidifoliaBenth.AsteraceaeA101Brassica napusL.BrassicaceaeA011Brassica oleraceaL.BrassicaceaeA43043Brassica rapaL.BrassicaceaeA505Briza mediaL.PoaceaeN101Bromus erectusHuds.PoaceaeN202Bromus hordeaceusL.PoaceaeN134Bromus inermisLeyss.PoaceaeA011Bromus ramosusHuds.PoaceaeN011Bromus sterilisL.PoaceaeN01111Bromus tectorumL.PoaceaeN011Brugmansia suaveolens(Humb. & Bonpl. ex Willd.) Bercht. & J. PreslSolanaceaeA303Brunnera macrophylla(Adams) I. M. Johnst.BoraginaceaeA10010Bryonia dioicaJacq.CucurbitaceaeN202Buddleja alternifoliaMaxim.ScrophulariaceaeA101Buddleja davidiiFranch.ScrophulariaceaeA10414Buglossoides purpurocaerulea(L.) I. M. Johnst.BoraginaceaeN202Buphthalmum salicifoliumL.AsteraceaeN303Buxus sempervirensL.BuxaceaeN13013Calamintha nepeta aggr.LamiaceaeN303Calendula arvensisL.AsteraceaeN202Calendula officinalisL.AsteraceaeA32032Callistephus chinensis(L.) NeesAsteraceaeA707Caltha palustrisL.RanunculaceaeN303Calystegia sepium(L.) R. Br.ConvolvulaceaeN05959Camassia cusickiiS. WatsonAsparagaceaeA101Campanula carpaticaJacq.CampanulaceaeA101Campanula mediumL.CampanulaceaeA101Campanula persicifoliaL.CampanulaceaeN12012Campanula portenschlagianaSchult.CampanulaceaeA404Campanula poscharskyanaDegenCampanulaceaeA16016Campanula rapunculoidesL.CampanulaceaeN202Campanula rotundifoliaL.CampanulaceaeN202Campanula tracheliumL.CampanulaceaeN819Campsis radicans(L.) Seem.BignoniaceaeA606Cannabis sativaL.CannabaceaeA404Capsella bursa-pastoris(L.) Medik.BrassicaceaeN011Capsella rubellaReut.BrassicaceaeN022Capsicum annuumL.SolanaceaeA29029Cardamine bulbifera(L.) CrantzBrassicaceaeN101Cardamine flexuosa aggr.BrassicaceaeN01111Cardamine hirsutaL.BrassicaceaeN07373Cardamine impatiensL.BrassicaceaeN022Cardamine pentaphyllos(L.) CrantzBrassicaceaeN011Cardamine pratensisL.BrassicaceaeN52833Cardaria draba(L.) Desv.BrassicaceaeA011Carex digitataL.CyperaceaeN011Carex elataAll.CyperaceaeN202Carex flaccaSchreb.CyperaceaeN055Carex grayiCareyCyperaceaeA202Carex hirtaL.CyperaceaeN077Carex morrowiiBoottCyperaceaeA202Carex muricata aggr.CyperaceaeN02020Carex pendulaHuds.CyperaceaeN224Carex pilosaScop.CyperaceaeN011Carex remotaL.CyperaceaeN022Carex sylvaticaHuds.CyperaceaeN03131Carlina vulgarisL.AsteraceaeN101Carpinus betulusL.BetulaceaeN71118Caryopteris × clandonensisSimmondsLamiaceaeA707Castanea sativaMill.FagaceaeN101Catalpa bignonioidesWalterBignoniaceaeA101Cedrus libaniA. Rich.PinaceaeA101Celtis australisL.CannabaceaeN101Centaurea cyanusL.AsteraceaeN707Centaurea jaceaL.AsteraceaeN202Centaurea montanaL.AsteraceaeN33033Centaurea pseudophrygiaC. A. Mey.AsteraceaeN101Centaurea scabiosaL.AsteraceaeN202Centaurium erythraeaRafnGentianaceaeN101Centranthus ruber(L.) DC.CaprifoliaceaeA12012Cephalaria alpina(L.) Roem. & Schult.CaprifoliaceaeN101Cerastium brachypetalumPers.CaryophyllaceaeN022Cerastium fontanum subsp. vulgare(Hartm.) Greuter & BurdetCaryophyllaceaeN51823Cerastium glomeratumThuill.CaryophyllaceaeN022Cerastium tomentosumL.CaryophyllaceaeA909Ceratophyllum demersumL.CeratophyllaceaeN101Ceratostigma plumbaginoidesBungePlumbaginaceaeA707Cercis siliquastrumL.FabaceaeA303Cerinthe majorL.BoraginaceaeA101Chaenomeles japonica(Thunb.) SpachRosaceaeA707Chaenorrhinum minus(L.) LangePlantaginaceaeN01818Chaenostoma cordatum(Thunb.) Benth.ScrophulariaceaeA101Chamaecyparis lawsoniana(A. Murray bis) Parl.CupressaceaeA505Chamaecyparis pisifera(Siebold & Zucc.) Endl.CupressaceaeA101Chelidonium majusL.PapaveraceaeN02727Chenopodium album aggr.AmaranthaceaeN055Chenopodium bonus-henricusL.AmaranthaceaeN101Chenopodium giganteumD. DonAmaranthaceaeA909Chenopodium polyspermumL.AmaranthaceaeN02828Chlorophytum comosum(Thunb.) JacquesAsparagaceaeA101Chrysanthemum × grandiflorumRamat.AsteraceaeA606Cichorium endiviaL.AsteraceaeA13013Cichorium intybusL.AsteraceaeN808Circaea lutetianaL.OnagraceaeN06464Cirsium arvense(L.) Scop.AsteraceaeN055Cirsium vulgare(Savi) Ten.AsteraceaeN033Cistus CVCistaceaeA202Citrus japonicaThunb.RutaceaeA404Clematis alpina(L.) Mill.RanunculaceaeN303Clematis CVRanunculaceaeA16016Clematis montanaBuch.-Ham. ex DC.RanunculaceaeA606Clematis rectaL.RanunculaceaeN101Clematis vitalbaL.RanunculaceaeN246Cleome hasslerianaSchltdl.CleomaceaeA404Clinopodium vulgareL.LamiaceaeN101Coffea arabicaL.RubiaceaeA101Consolida ajacis(L.) SchurRanunculaceaeA101Convallaria majalisL.AsparagaceaeN39039Convolvulus arvensisL.ConvolvulaceaeN077Convolvulus tricolorL.ConvolvulaceaeA202Conyza canadensis(L.) CronquistAsteraceaeA01414Coreopsis verticillataL.AsteraceaeA011Coriandrum sativumL.ApiaceaeA202Cornus albaL.CornaceaeA202Cornus kousaF. Buerger ex HanceCornaceaeA202Cornus masL.CornaceaeN909Cornus sanguineaL.CornaceaeN15722Cornus sericeaL.CornaceaeA101Cortaderia selloana(Schult. & Schult. f.) Asch. & Graebn.PoaceaeA404Corydalis cava(L.) Schweigg. & KörtePapaveraceaeN099Corydalis lutea(L.) DC.PapaveraceaeN202Corylopsis paucifloraSiebold & Zucc.HamamelidaceaeA505Corylus avellanaL.BetulaceaeN14923Corylus maximaMill.BetulaceaeA505Cosmos bipinnatusCav.AsteraceaeA20020Cosmos sulphureusCav.AsteraceaeA303Cotinus coggygriaScop.AnacardiaceaeN404Cotoneaster bullatusBoisRosaceaeA101Cotoneaster dammeriC. K. Schneid.RosaceaeA808Cotoneaster dielsianusE. Pritz.RosaceaeA101Cotoneaster divaricatusRehder & E. H. WilsonRosaceaeA707Cotoneaster horizontalisDecne.RosaceaeA404Cotoneaster salicifoliusFranch.RosaceaeA202Crataegus monogyna aggr.RosaceaeN8513Crepis biennisL.AsteraceaeN6410Crepis capillarisWallr.AsteraceaeN02020Crocosmia × crocosmiiflora(Lemoine) N. E. Br.IridaceaeA27027Cucumis meloL.CucurbitaceaeA606Cucumis sativusL.CucurbitaceaeA33033Cucurbita maximaDuchesneCucurbitaceaeA14014Cucurbita pepoL.CucurbitaceaeA44044Cuphea igneaA. DC.LythraceaeA101Cyclamen coumMill.PrimulaceaeA101Cydonia oblongaMill.RosaceaeA15015Cymbalaria muralisG. Gaertn. & al.PlantaginaceaeN617Cynara cardunculus subsp. scolymus(L.) BegerAsteraceaeA12012Cynodon dactylon(L.) Pers.PoaceaeN011Cynosurus cristatusL.PoaceaeN202Cyperus involucratusRottb.CyperaceaeA101Cytisus × praecoxWheeler ex. BeanFabaceaeA101Dactylis glomerataL.PoaceaeN111829Dahlia CVAsteraceaeA14014Dahlia merckiiLehm.AsteraceaeA101Daphne mezereumL.ThymelaeaceaeN202Datura innoxiaMill.SolanaceaeA101Datura stramoniumL.SolanaceaeA101Daucus carota subsp. carotaL.ApiaceaeN12416Daucus carota subsp. sativus(Hoffm.) Arcang.ApiaceaeN17017Delphinium × cultorumVoss.RanunculaceaeA606Deschampsia cespitosa(L.) P. Beauv.PoaceaeN044Deutzia gracilisSiebold & Zucc.HydrangeaceaeA505Deutzia scabraThunb.HydrangeaceaeA202Dianthus armeriaL.CaryophyllaceaeN066Dianthus barbatusL.CaryophyllaceaeA13013Dianthus carthusianorumL.CaryophyllaceaeN404Dianthus chinensisL.CaryophyllaceaeA808Dianthus deltoidesL.CaryophyllaceaeN303Dianthus gratianopolitanusVill.CaryophyllaceaeN10010Dicentra spectabilis(L.) FukuharaPapaveraceaeA15015Dictamnus albusL.RutaceaeN101Digitalis grandifloraMill.PlantaginaceaeN202Digitalis luteaL.PlantaginaceaeN404Digitalis purpureaL.PlantaginaceaeA20020Digitaria sanguinalis(L.) Scop.PoaceaeN05353Dionaea muscipulaJ.EllisDroseraceaeA101Diospyros lotusL.EbenaceaeA101Diplotaxis tenuifolia(L.) DC.BrassicaceaeN16016Dipsacus fullonumL.CaprifoliaceaeN11011Dipsacus laciniatusL.CaprifoliaceaeN101Dipsacus pilosusL.CaprifoliaceaeN011Doronicum orientaleHoffm.AsteraceaeA505Dracunculus vulgarisSchottAraceaeA404Dryopteris cycadina(Franch. & Sav.) C. Chr.DryopteridaceaeA101Dryopteris filix-mas(L.) SchottDryopteridaceaeN48048Duchesnea indica(Andrews) FockeRosaceaeA02424Echeveria elegansRoseCrassulaceaeA101Echinacea purpurea(L.) MoenchAsteraceaeA21021Echinochloa crus-galli(L.) P. Beauv.PoaceaeN044Echinops sphaerocephalusL.AsteraceaeN505Echium vulgareL.BoraginaceaeN505Eichhornia crassipes(Mart.) SolmsPontederiaceaeA101Elaeagnus × submacrophyllaServett.ElaeagnaceaeA101Eleocharis palustris aggr.CyperaceaeN101Elodea canadensisMichx.HydrocharitaceaeA101Elymus caninus(L.) L.PoaceaeN022Elymus repens(L.) GouldPoaceaeN02222Epilobium angustifoliumL.OnagraceaeN505Epilobium dodonaeiVill.OnagraceaeN202Epilobium hirsutumL.OnagraceaeN01010Epilobium montanumL.OnagraceaeN03232Epilobium parviflorumSchreb.OnagraceaeN03939Epilobium roseumSchreb.OnagraceaeN03737Epilobium tetragonumL.OnagraceaeN055Epimedium pinnatumFisch.BerberidaceaeA505Equisetum arvenseL.EquisetaceaeN01515Equisetum telmateiaEhrh.EquisetaceaeN011Eragrostis minorHostPoaceaeN066Eranthis hyemalis(L.) Salisb.RanunculaceaeN19019Erica carneaL.EricaceaeN303Erigeron annuus(L.) Desf.AsteraceaeA01717Erigeron karvinskianusDC.AsteraceaeA101Erinus alpinusL.PlantaginaceaeN202Erodium × hybridumSünd.GeraniaceaeA101Erodium × variabileA. C. LeslieGeraniaceaeA101Erophila verna aggr.BrassicaceaeN011Eruca sativa(L.) Cav.BrassicaceaeA909Erysimum cheiri(L.) CrantzBrassicaceaeA808Eschscholzia californicaCham.PapaveraceaeA404Euonymus europaeusL.CelastraceaeN808Euonymus fortunei(Turcz.) Hand.-Mazz.CelastraceaeA707Euonymus japonicusThunb.CelastraceaeA606Eupatorium cannabinumL.AsteraceaeN213Euphorbia amygdaloidesL.EuphorbiaceaeN505Euphorbia characiasL.EuphorbiaceaeA101Euphorbia cyparissiasL.EuphorbiaceaeN404Euphorbia dulcisL.EuphorbiaceaeN022Euphorbia epithymoidesL.EuphorbiaceaeA101Euphorbia helioscopiaL.EuphorbiaceaeN022Euphorbia lathyrisL.EuphorbiaceaeA18018Euphorbia maculata aggr.EuphorbiaceaeA077Euphorbia myrsinitesL.EuphorbiaceaeA101Euphorbia palustrisL.EuphorbiaceaeN101Euphorbia peplusL.EuphorbiaceaeN06262Euphorbia verrucosaL.EuphorbiaceaeN101Eutrema japonicum(Miq.) Koidz.BrassicaceaeA101Fagopyrum esculentumMoenchPolygonaceaeA505Fagus sylvaticaL.FagaceaeN516Fallopia aubertii(L.Henry) HolubPolygonaceaeA202Festuca arundinaceaSchreb.PoaceaeN24024Festuca cinereaVill.PoaceaeA101Festuca gigantea(L.) Vill.PoaceaeN011Festuca ovina aggr.PoaceaeN044Festuca rubra aggr.PoaceaeN67067Ficus caricaL.MoraceaeN17017Filipendula ulmaria(L.) Maxim.RosaceaeN022Filipendula vulgarisMoenchRosaceaeN202Foeniculum vulgareMill.ApiaceaeA26026Forsythia × intermediaZabelOleaceaeA27027Fragaria × ananassa(Weston) RozierRosaceaeA31031Fragaria moschataDuchesneRosaceaeN101Fragaria vescaL.RosaceaeN06666Frangula alnusMill.RhamnaceaeN202Fraxinus excelsiorL.OleaceaeN02424Fritillaria imperialisL.LiliaceaeA404Fritillaria meleagrisL.LiliaceaeN202Fuchsia magellanicaLam.OnagraceaeA21021Fumaria officinalisL.PapaveraceaeN022Gaillardia × grandifloraHort. ex Van HoutteAsteraceaeA101Galanthus elwesiiHook. f.AmaryllidaceaeA404Galanthus nivalisL.AmaryllidaceaeN43043Galeopsis tetrahitL.LamiaceaeN01616Galinsoga quadriradiataRuiz & Pav.AsteraceaeA088Galium aparineL.RubiaceaeN066Galium mollugo aggr.RubiaceaeN51318Galium odoratum(L.) Scop.RubiaceaeN11213Galium verumL.RubiaceaeN404Gaura lindheimeriEngelm. & A.GrayOnagraceaeA606Gazania CVAsteraceaeA202Gentiana acaulisL.GentianaceaeN707Geranium × cantabrigienseP. F. YeoGeraniaceaeA303Geranium × magnificumHyl.GeraniaceaeA16016Geranium × oxonianumP. F. YeoGeraniaceaeA202Geranium dissectumL.GeraniaceaeN011Geranium himalayenseKlotzschGeraniaceaeA101Geranium macrorrhizumL.GeraniaceaeA11011Geranium nodosumL.GeraniaceaeN101Geranium palustreL.GeraniaceaeN101Geranium phaeumL.GeraniaceaeN303Geranium pratenseL.GeraniaceaeN909Geranium pyrenaicumBurm. f.GeraniaceaeN01313Geranium renardiiTrautv.GeraniaceaeA404Geranium robertianum subsp. purpureumVill.GeraniaceaeN011Geranium robertianum subsp. robertianumL.GeraniaceaeN05050Geranium sanguineumL.GeraniaceaeN10010Geranium sylvaticumL.GeraniaceaeN202Geum coccineumSibth. & Sm.RosaceaeA303Geum rivaleL.RosaceaeN022Geum urbanumL.RosaceaeN07272Ginkgo bilobaL.GinkgoaceaeA404Gladiolus × hortulanusL.H.BaileyIridaceaeA21021Gladiolus callianthusKelwayIridaceaeA101Gladiolus communisL.IridaceaeA101Glebionis segetum(L.) Fourr.AsteraceaeA202Glechoma hederaceaL.LamiaceaeN05050Gleditsia triacanthosL.FabaceaeA101Glycine max(L.) Merr.FabaceaeA303Gynostemma pentaphyllum(Thunb.) MakinoCucurbitaceaeA202Gypsophila paniculataL.CaryophyllaceaeA202Gypsophila repensL.CaryophyllaceaeN202Hamamelis mollisOliv. ex F. B. Forbes & Hemsl.HamamelidaceaeA202Hebe odoraCockayneScrophulariaceaeA404Hedera helixL.AraliaceaeN43548Helenium autumnaleL.AsteraceaeA404Helianthemum nummulariumDun.CistaceaeN101Helianthus annuusL.AsteraceaeA31031Helianthus pauciflorusNutt.AsteraceaeA505Helianthus tuberosusL.AsteraceaeA13013Helichrysum bracteatum(Venten.) Willd.AsteraceaeA101Helichrysum italicum(Roth) G. DonAsteraceaeA606Helictotrichon pubescens(Huds.) Pilg.PoaceaeN202Heliopsis helianthoides(L.) SweetAsteraceaeA202Helleborus foetidusL.RanunculaceaeN404Helleborus nigerL.RanunculaceaeN11011Helleborus orientalisLam.RanunculaceaeA32032Hemerocallis fulva(L.) L.AsphodelaceaeA15015Hemerocallis lilioasphodelusL.AsphodelaceaeA101Hepatica nobilisSchreb.RanunculaceaeN202Heracleum sphondyliumL.ApiaceaeN224Herniaria hirsutaL.CaryophyllaceaeN011Hesperis matronalisL.BrassicaceaeA404Heuchera americanaL.SaxifragaceaeA14014Hibiscus syriacusL.MalvaceaeA34034Hibiscus trionumL.MalvaceaeA101Hieracium aurantiacumL.AsteraceaeN17017Hieracium lachenaliiC. C. Gmel.AsteraceaeN011Hieracium murorum aggr.AsteraceaeN022Hieracium pilosellaL.AsteraceaeN347Hieracium piloselloidesVill.AsteraceaeN101Hippocrepis emerus(L.) LassenFabaceaeN202Hippophaë rhamnoidesL.ElaeagnaceaeN505Hippuris vulgarisL.PlantaginaceaeN202Holcus lanatusL.PoaceaeN51217Hordeum murinumL.PoaceaeN011Hordeum vulgare aggr.PoaceaeA011Hosta CVAsparagaceaeA16016Humulus lupulusL.CannabaceaeN448Hyacinthoides non-scripta(L.) Rothm.AsparagaceaeA42042Hyacinthus orientalisL.AsparagaceaeA41041Hydrangea asperaD. DonHydrangeaceaeA101Hydrangea macrophylla(Thunb.) Ser.HydrangeaceaeA32032Hydrangea paniculataSieboldHydrangeaceaeA202Hydrangea petiolarisSiebold & Zucc.HydrangeaceaeA10010Hydrangea quercifoliaW. BartramHydrangeaceaeA101Hypericum androsaemumL.HypericaceaeN202Hypericum calycinumL.HypericaceaeA303Hypericum perforatumL.HypericaceaeN8715Hypericum tetrapterumFr.HypericaceaeN022Hypochaeris radicataL.AsteraceaeN41822Hyssopus officinalisL.LamiaceaeN606Iberis sempervirensL.BrassicaceaeA23023Iberis umbellataL.BrassicaceaeA303Ilex aquifoliumL.AquifoliaceaeN18119Impatiens balfouriiHook. f.BalsaminaceaeA011Impatiens balsaminaL.BalsaminaceaeA101Impatiens parvifloraDC.BalsaminaceaeA011Impatiens wallerianaHook. f.BalsaminaceaeA303Inula conyzae(Griess.) MeikleAsteraceaeN011Inula ensifoliaL.AsteraceaeA101Inula heleniumL.AsteraceaeA404Ipomoea × multifida(Raf.) ShinnersConvolvulaceaeA101Ipomoea batatas(L.) Lam.ConvolvulaceaeA101Ipomoea purpurea(L.) RothConvolvulaceaeA11011Iris × germanicaL.IridaceaeN32032Iris × hollandicahort.IridaceaeA303Iris foetidissimaL.IridaceaeA202Iris pseudacorusL.IridaceaeN415Iris sibiricaL.IridaceaeN606Isatis tinctoriaL.BrassicaceaeN101Ismelia carinata(Schousb.) Sch.Bip.AsteraceaeA303Jasminum nudiflorumLindl.OleaceaeA808Juglans regiaL.JuglandaceaeN6612Juncus effususL.JuncaceaeN011Juncus inflexusL.JuncaceaeN022Juncus subnodulosusSchrankJuncaceaeN011Juncus tenuisWilld.JuncaceaeA011Juniperus chinensisL.CupressaceaeA202Juniperus communisL.CupressaceaeN101Kerria japonica(L.) DC.RosaceaeA12012Kickxia spuria(L.) Dumort.PlantaginaceaeN022Knautia arvensis(L.) Coult.CaprifoliaceaeN404Knautia dipsacifoliaKreutzerCaprifoliaceaeN202Kniphofia CVLiliaceaeA101Kolkwitzia amabilisGraebn.CaprifoliaceaeA808Laburnum anagyroidesMedik.FabaceaeN101Lactuca sativaL.AsteraceaeA29029Lactuca serriolaL.AsteraceaeN022Lamium albumL.LamiaceaeN101Lamium galeobdolon subsp. argentatum(Smejkal) J. Duvign.LamiaceaeA19019Lamium maculatum(L.) L.LamiaceaeN134Lamium purpureumL.LamiaceaeN066Lampranthus spectabilis(Haw.) N. E. Br.AizoaceaeA101Lantana camaraL.VerbenaceaeA202Lapsana communisL.AsteraceaeN03535Larix deciduaMill.PinaceaeN101Lathyrus latifoliusL.FabaceaeN909Lathyrus pratensisL.FabaceaeN224Lathyrus vernus(L.) Bernh.FabaceaeN303Laurus nobilisL.LauraceaeN909Lavandula × intermediaLoisel.LamiaceaeA53053Lavandula stoechasL.LamiaceaeA909Lavatera × clementiiCheekMalvaceaeA101Lavatera trimestrisL.MalvaceaeA707Lemna minorL.AraceaeN404Lemna trisulcaL.AraceaeN101Leontodon hispidusL.AsteraceaeN426Leontopodium alpinum(Cass.) GreuterAsteraceaeN303Leonurus cardiacaL.LamiaceaeN202Lepidium latifoliumL.BrassicaceaeA101Lepidium sativumL.BrassicaceaeA101Lepidium virginicumL.BrassicaceaeA011Leucanthemum vulgare aggr.AsteraceaeN15722Leucojum aestivumL.AmaryllidaceaeN404Leucojum vernumL.AmaryllidaceaeN505Levisticum officinaleW. D. J. KochApiaceaeA22022Liatris spicata(L.) Willd.AsteraceaeA101Ligularia przewalskii(Maxim.) DielsAsteraceaeA606Ligustrum japonicumThunb.OleaceaeA202Ligustrum ovalifoliumHausskn.OleaceaeA14014Ligustrum vulgareL.OleaceaeN16117Lilium CVLiliaceaeA11011Lilium martagonL.LiliaceaeN202Limonium platyphyllumLincz.PlumbaginaceaeA202Limonium sinuatum(L.) Mill.PlumbaginaceaeA101Linaria maroccanaHook. f.PlantaginaceaeA404Linaria purpurea(L.) Mill.PlantaginaceaeA19019Linaria vulgarisMill.PlantaginaceaeN066Linum grandiflorumDesf.LinaceaeA202Linum usitatissimumL.LinaceaeA606Liriodendron tulipiferaL.MagnoliaceaeA303Lobelia erinusL.CampanulaceaeA202Lobelia siphiliticaL.CampanulaceaeA202Lobularia maritima(L.) Desv.BrassicaceaeA808Lolium perenneL.PoaceaeN70070Lonicera × purpusiiRehderCaprifoliaceaeA303Lonicera × tellmannianaMagyar ex H. L. SpäthCaprifoliaceaeA505Lonicera caprifoliumL.CaprifoliaceaeN707Lonicera henryiHemsl.CaprifoliaceaeA303Lonicera japonicaThunb.CaprifoliaceaeA505Lonicera kamtschaticaPojark.CaprifoliaceaeA303Lonicera periclymenumL.CaprifoliaceaeN606Lonicera pileataOliv.CaprifoliaceaeA21021Lonicera xylosteumL.CaprifoliaceaeN505Lotus corniculatusL.FabaceaeN314Lotus tetragonolobusL.FabaceaeA101Lunaria annuaL.BrassicaceaeA303Lunaria redivivaL.BrassicaceaeN404Lupinus polyphyllusLindl.FabaceaeA909Luzula campestris(L.) DC.JuncaceaeN011Luzula nivea(L.) DC.JuncaceaeN303Luzula sylvatica(Huds.) GaudinJuncaceaeN202Lycianthes rantonnetii(Carrière ex Lesc.) BitterSolanaceaeA202Lycium barbarumL.SolanaceaeA202Lycopersicon esculentumMill.SolanaceaeA49049Lysimachia clethroidesDubyPrimulaceaeA303Lysimachia nummulariaL.PrimulaceaeN01111Lysimachia punctataL.PrimulaceaeA17017Lysimachia vulgarisL.PrimulaceaeN505Lythrum salicariaL.LythraceaeN15318Magnolia × soulangeanaSoul.-Bod.MagnoliaceaeA404Mahonia aquifolium(Pursh) Nutt.BerberidaceaeA909Malus pumilaMill.RosaceaeA43043Malva alceaL.MalvaceaeN606Malva moschataL.MalvaceaeN10111Malva neglectaWallr.MalvaceaeN044Malva sylvestrisL.MalvaceaeN18018Malva verticillataL.MalvaceaeA101Mandevilla splendens(Hook. f.) WoodsonApocynaceaeA404Matricaria chamomillaL.AsteraceaeN167Matricaria discoideaDC.AsteraceaeA011Matteuccia struthiopteris(L.) Tod.OnocleaceaeN909Matthiola incana(L.) R. Br.BrassicaceaeA303Meconopsis cambrica(L.) Vig.PapaveraceaeA505Medicago lupulinaL.FabaceaeN01919Medicago sativaL.FabaceaeN224Melampyrum arvenseL.OrobanchaceaeN101Melica nutansL.PoaceaeN224Melilotus albusMedik.FabaceaeN011Melilotus officinalis(L.) Pall.FabaceaeN202Melissa officinalisL.LamiaceaeN42042Melittis melissophyllumL.LamiaceaeN202Melothria scabraNaudinCucurbitaceaeA303Mentha × piperitaL.LamiaceaeA50050Mentha aquaticaL.LamiaceaeN202Mentha longifolia(L.) Huds.LamiaceaeN134Mentha spicataL.LamiaceaeA18018Mentha suaveolensEhrh.LamiaceaeN13013Menyanthes trifoliataL.MenyanthaceaeN404Mercurialis perennisL.EuphorbiaceaeN022Mespilus germanicaL.RosaceaeN101Mirabilis jalapaL.NyctaginaceaeA101Miscanthus sinensisAnderssonPoaceaeA505Moehringia trinervia(L.) Clairv.CaryophyllaceaeN011Molinia caerulea(L.) MoenchPoaceaeN101Monarda didymaL.LamiaceaeA11011Morus nigraL.MoraceaeA303Muehlenbeckia complexaMeisn.PolygonaceaeA101Musa basjooSiebold & Zucc. ex IinumaMusaceaeA202Muscari armeniacumBakerAsparagaceaeA33033Mycelis muralis(L.) Dumort.AsteraceaeN066Myosotis arvensisHillBoraginaceaeN01010Myosotis sylvaticaHoffm.BoraginaceaeN31031Myriophyllum spicatumL.HaloragaceaeN101Myrrhis odorata(L.) Scop.ApiaceaeN101Narcissus cyclamineusDC.AmaryllidaceaeA39039Narcissus jonquillaL.AmaryllidaceaeA101Narcissus poëticusL.AmaryllidaceaeN101Narcissus pseudonarcissusL.AmaryllidaceaeN57057Narcissus tazettaL.AmaryllidaceaeA404Narthecium ossifragum(L.) Huds.NartheciaceaeA101Nepeta × faaseniiStearnLamiaceaeA909Nerium oleanderL.ApocynaceaeA606Nicotiana × sanderaeW. WatsonSolanaceaeA101Nicotiana tabacumL.SolanaceaeA101Nigella damascenaL.RanunculaceaeA20020Nymphaea albaL.NymphaeaceaeN10010Nymphoides peltata(S. G. Gmel.) KuntzeGentianaceaeN101Ocimum basilicumL.LamiaceaeA29029Ocimum tenuiflorumL.LamiaceaeA202Oenothera biennis aggr.OnagraceaeA13233Oenothera fruticosaL.OnagraceaeA101Oenothera parviflora aggr.OnagraceaeA011Olea europaeaL.OleaceaeA404Onobrychis viciifoliaScop.FabaceaeA202Onoclea sensibilisL.OnocleaceaeA101Opuntia humifusa(Raf.) Raf.CactaceaeA101Opuntia phaeacanthaEngelm.CactaceaeA303Origanum majoranaL.LamiaceaeA202Origanum vulgareL.LamiaceaeN41142Orlaya grandiflora(L.) Hoffm.ApiaceaeN101Ornithogalum umbellatumL.AsparagaceaeN202Orobanche hederaeDubyOrobanchaceaeN011Osmanthus heterophyllus(G.Don) P. S. GreenOleaceaeA101Osmunda regalisL.OsmundaceaeN101Oxalis corniculataL.OxalidaceaeN05757Oxalis debilisKunthOxalidaceaeA505Oxalis strictaL.OxalidaceaeA04848Oxalis tetraphyllaCav.OxalidaceaeA505Oxalis triangularisA. St.-Hil.OxalidaceaeA303Pachysandra terminalisSiebold & Zucc.BuxaceaeA101Paeonia × suffruticosaAndrewsPaeoniaceaeA707Paeonia lactifloraPall.PaeoniaceaeA41041Paeonia luteaFranch.PaeoniaceaeA101Paeonia officinalisL.PaeoniaceaeN909Panicum capillare aggr.PoaceaeA01010Panicum miliaceumL.PoaceaeA011Panicum virgatumL.PoaceaeA101Papaver croceumL.PapaveraceaeA101Papaver dubiumL.PapaveraceaeN011Papaver orientaleL.PapaveraceaeA13013Papaver rhoeasL.PapaveraceaeN606Papaver somniferumL.PapaveraceaeA909Parrotia persicaC.A. Mey.HamamelidaceaeA202Parthenocissus quinquefolia aggr.VitaceaeA16016Parthenocissus tricuspidata(Siebold & Zucc.) Planch.VitaceaeA707Passiflora caeruleaL.PassifloraceaeA404Pastinaca sativaL.ApiaceaeN404Pelargonium CVGeraniaceaeA21021Pennisetum alopecuroides(L.) Spreng.PoaceaeA707Perilla frutescens(L.) BrittonLamiaceaeA101Perovskia atriplicifoliaBenth.LamiaceaeA404Persea americanaMill.LauraceaeA101Petasites albus(L.) Gaertn.AsteraceaeN101Petrorhagia saxifraga(L.) LinkCaryophyllaceaeN202Petroselinum crispum(Mill.) FussApiaceaeA25025Petunia × atkinsianaD. Don ex W. H. BaxterSolanaceaeA12012Phacelia campanulariaA. GrayBoraginaceaeA101Phacelia tanacetifoliaBenth.BoraginaceaeA11011Phalaris arundinaceaL.PoaceaeN202Phalaris canariensisL.PoaceaeA011Phaseolus vulgarisL.FabaceaeA36036Philadelphus coronariusL.HydrangeaceaeA18018Phleum pratenseL.PoaceaeN123Phlomis fruticosaL.LamiaceaeA101Phlomis russelianaBenth.LamiaceaeA202Phlox paniculataL.PolemoniaceaeA32032Phlox subulataL.PolemoniaceaeA10010Phoenix roebeleniiO\'BrienArecaceaeA202Phyllitis scolopendrium(L.) NewmanAspleniaceaeN808Phyllostachys aureaRivière & C. RivièrePoaceaeA101Physalis alkekengiL.SolanaceaeA505Physalis peruvianaL.SolanaceaeA13013Physalis philadelphicaLam.SolanaceaeA202Physostegia virginiana(L.) Benth.LamiaceaeA505Picea abies(L.) H. Karst.PinaceaeN10010Picea glauca var. albertianaS. Br.PinaceaeA707Picea omorika(Pancic) Purk.PinaceaeA404Picea pungensEngelm.PinaceaeA202Picris hieracioidesL.AsteraceaeN011Pieris japonica(Thunb.) D. Don ex G. DonEricaceaeA101Pimpinella major(L.) Huds.ApiaceaeN101Pimpinella saxifragaL.ApiaceaeN202Pinus cembraL.PinaceaeN101Pinus mugoTurraPinaceaeN10010Pinus strobusL.PinaceaeA101Pinus sylvestrisL.PinaceaeN303Pisum sativumL.FabaceaeN606Plantago coronopusL.PlantaginaceaeA101Plantago lanceolataL.PlantaginaceaeN102838Plantago major subsp. intermedia(Gilib.) LangePlantaginaceaeN011Plantago major subsp. majorL.PlantaginaceaeN04343Plantago mediaL.PlantaginaceaeN314Platycodon grandiflorus(Jacq.) A. DC.CampanulaceaeA303Plumbago auriculataLam.PlumbaginaceaeA101Poa annuaL.PoaceaeN06666Poa compressaL.PoaceaeN022Poa nemoralisL.PoaceaeN044Poa pratensisL.PoaceaeN31031Poa trivialisL.PoaceaeN104050Polemonium caeruleumL.PolemoniaceaeN303Polygonatum multiflorum(L.) All.AsparagaceaeN31417Polygonum aviculare aggr.PolygonaceaeN077Polygonum bistortaL.PolygonaceaeN101Polygonum persicariaL.PolygonaceaeN01313Polystichum setiferum(Forssk.) Woyn.DryopteridaceaeN10010Pontederia cordataL.PontederiaceaeA202Populus nigra aggr.SalicaceaeN202Portulaca oleraceaL.PortulacaceaeN03232Potamogeton natansL.PotamogetonaceaeN101Potentilla anserinaL.RosaceaeN033Potentilla argenteaL.RosaceaeN202Potentilla fruticosa(L.) Rydb.RosaceaeA505Potentilla rectaL.RosaceaeN101Potentilla reptansL.RosaceaeN03434Potentilla sterilis(L.) GarckeRosaceaeN01818Potentilla vernaL.RosaceaeN202Primula acaulisHuds.PrimulaceaeN85085Primula elatior(L.) L.PrimulaceaeN10111Primula verisL.PrimulaceaeN303Prunella grandiflora(L.) SchollerLamiaceaeN202Prunella vulgarisL.LamiaceaeN83644Prunus armeniacaL.RosaceaeA16016Prunus aviumL.RosaceaeN16117Prunus cerasiferaEhrh.RosaceaeA404Prunus cerasusL.RosaceaeA101Prunus domesticaL.RosaceaeA19019Prunus dulcis(Mill.) D. A. WebbRosaceaeA101Prunus kurilensis(Miyabe) MiyabeRosaceaeA101Prunus laurocerasusL.RosaceaeA17017Prunus lusitanicaL.RosaceaeA202Prunus padusL.RosaceaeN303Prunus persica(L.) BatschRosaceaeA606Prunus serrulataLindl.RosaceaeA303Prunus spinosaL.RosaceaeN505Prunus trilobaLindl.RosaceaeA101Pseudolysimachion spicatum(L.) OpizPlantaginaceaeN303Pseudosasa japonica(Steud.) NakaiPoaceaeA808Pulmonaria obscuraDumort.BoraginaceaeN202Pulmonaria officinalisL.BoraginaceaeN202Pulsatilla vulgarisL.RanunculaceaeN606Punica granatumL.LythraceaeA101Pyracantha coccineaM. Roem.RosaceaeA606Pyrus communisL.RosaceaeA15015Quercus roburL.FagaceaeN257Ranunculus acris subsp. friesianus(Jord.) SymeRanunculaceaeN51318Ranunculus auricomus aggr.RanunculaceaeN011Ranunculus bulbosusL.RanunculaceaeN101Ranunculus ficariaL.RanunculaceaeN06969Ranunculus gramineusL.RanunculaceaeN101Ranunculus linguaL.RanunculaceaeN202Ranunculus repensL.RanunculaceaeN05353Raphanus sativusL.BrassicaceaeA27027Reseda luteaL.ResedaceaeN303Rheum rhabarbarumL.PolygonaceaeA29029Rhodiola roseaL.CrassulaceaeN101Rhododendron Catawbiense GroupEricaceaeA19019Rhododendron luteumSweetEricaceaeA303Rhus typhinaL.AnacardiaceaeA101Ribes × nidigrolariaRud. Bauer & A. BauerGrossulariaceaeA808Ribes alpinumL.GrossulariaceaeN202Ribes aureumPurshGrossulariaceaeA202Ribes nigrumL.GrossulariaceaeA15015Ribes rubrumL.GrossulariaceaeA56056Ribes sanguineumPurshGrossulariaceaeA101Ribes uva-crispaL.GrossulariaceaeN24024Ricinus communisL.EuphorbiaceaeA303Robinia pseudoacaciaL.FabaceaeA235Rodgersia aesculifoliaBatalinSaxifragaceaeA101Rodgersia podophyllaA. GraySaxifragaceaeA101Rorippa sylvestris(L.) BesserBrassicaceaeN022Rosa arvensisHuds.RosaceaeN011Rosa canina aggr.RosaceaeN111021Rosa CVRosaceaeA68068Rosa glaucaPourr.RosaceaeN505Rosa multifloraThunb.RosaceaeA202Rosa pendulinaL.RosaceaeN101Rosa rubiginosa aggr.RosaceaeN303Rosa rugosaThunb.RosaceaeA101Rosa spinosissimaL.RosaceaeN404Rosmarinus officinalisL.LamiaceaeA43043Rubus armeniacusFockeRosaceaeA47148Rubus caesiusL.RosaceaeN01111Rubus idaeusL.RosaceaeN47047Rubus phoenicolasiusMaxim.RosaceaeA101Rudbeckia hirtaL.AsteraceaeA20020Rudbeckia laciniataL.AsteraceaeA303Rumex acetosaL.PolygonaceaeN044Rumex obtusifoliusL.PolygonaceaeN01818Rumex sanguineusL.PolygonaceaeN10010Ruscus aculeatusL.AsparagaceaeN101Ruta graveolensL.RutaceaeN909Sagina procumbensL.CaryophyllaceaeN02020Sagittaria latifoliaWilld.AlismataceaeA101Salix × fragilisL.SalicaceaeA101Salix capreaL.SalicaceaeN41317Salix gracilistylaMiq.SalicaceaeA101Salix integraThunb.SalicaceaeA101Salix matsudanaL.SalicaceaeA101Salix purpureaL.SalicaceaeN101Salvia × jamensisJ. ComptonLamiaceaeA101Salvia × sylvestrisL.LamiaceaeA303Salvia africanaL.LamiaceaeA101Salvia apianaJeps.LamiaceaeA101Salvia elegansVahlLamiaceaeA909Salvia farinaceaBenth.LamiaceaeA101Salvia glutinosaL.LamiaceaeN101Salvia officinalisL.LamiaceaeA33033Salvia patensCav.LamiaceaeA505Salvia pratensisL.LamiaceaeN202Salvia sclareaL.LamiaceaeA404Salvia verticillataL.LamiaceaeA101Sambucus nigraL.AdoxaceaeN271138Sanguisorba minorScop.RosaceaeN808Sanguisorba officinalisL.RosaceaeN303Santolina chamaecyparissusL.AsteraceaeA303Santolina rosmarinifoliaL.AsteraceaeA202Sanvitalia procumbensLam.AsteraceaeA404Saponaria ocymoidesL.CaryophyllaceaeN404Saponaria officinalisL.CaryophyllaceaeN51015Sarracenia purpureaL.SarraceniaceaeA101Satureja hortensisL.LamiaceaeA707Satureja montanaL.LamiaceaeA909Saxifraga × arendsiiEngl.SaxifragaceaeA10010Saxifraga paniculataMill.SaxifragaceaeN505Saxifraga stoloniferaMeerb.SaxifragaceaeA101Saxifraga tridactylitesL.SaxifragaceaeN044Saxifraga umbrosaL.SaxifragaceaeA404Scabiosa caucasica(M. Bieb.) Greuter & BurdetCaprifoliaceaeA404Scabiosa columbariaL.CaprifoliaceaeN404Scabiosa ochroleucaL.CaprifoliaceaeA101Schoenoplectus lacustris(L.) PallaCyperaceaeN101Schoenoplectus mucronatus(L.) PallaCyperaceaeN011Scilla forbesii(Baker) SpetaAsparagaceaeA29029Scilla sibericaHaw.AsparagaceaeA21021Scorzonera hispanicaL.AsteraceaeA404Scrophularia nodosaL.ScrophulariaceaeN044Scutellaria altissimaL.LamiaceaeA101Sechium edule(Jacq.) Sw.CucurbitaceaeA303Securigera varia(L.) LassenFabaceaeN011Sedum acreL.CrassulaceaeN011Sedum albumL.CrassulaceaeN9110Sedum floriferumFisch.CrassulaceaeA14014Sedum hispanicumL.CrassulaceaeN022Sedum rupestre aggr.CrassulaceaeN14014Sedum sexangulareL.CrassulaceaeN11819Sedum sieboldiiRegelCrassulaceaeA202Sedum spuriumM. Bieb.CrassulaceaeA14014Sedum telephiumL.CrassulaceaeN36036Sempervivum arachnoideumL.CrassulaceaeN101Sempervivum montanumL.CrassulaceaeN101Sempervivum tectorumL.CrassulaceaeN18018Senecio cinerariaDC.AsteraceaeA303Senecio jacobaeaL.AsteraceaeN134Senecio vulgarisL.AsteraceaeN03232Setaria italica(L.) P. Beauv.PoaceaeA011Setaria pumila(Poir.) Roem. & Schult.PoaceaeN022Setaria verticillata(L.) P. Beauv.PoaceaeN022Setaria viridis(L.) P. Beauv.PoaceaeN077Sherardia arvensisL.RubiaceaeN022Silene armeriaL.CaryophyllaceaeN101Silene chalcedonica(L.) E. H. L. KrauseCaryophyllaceaeA303Silene coronaria(L.) Clairv.CaryophyllaceaeN11011Silene dioica(L.) Clairv.CaryophyllaceaeN347Silene flos-cuculi(L.) Clairv.CaryophyllaceaeN303Silene nutansL.CaryophyllaceaeN011Silene pratensis(Rafn) Godr.CaryophyllaceaeN011Silene vulgaris(Moench) GarckeCaryophyllaceaeN303Silybum marianum(L.) Gaertn.AsteraceaeA101Sinapis albaL.BrassicaceaeN819Sium sisarumL.ApiaceaeA101Skimmia japonicaThunb.RutaceaeA303Solanum dulcamaraL.SolanaceaeN055Solanum jasminoidesJ. PaxtonSolanaceaeA303Solanum melongenaL.SolanaceaeA11011Solanum muricatumAitonSolanaceaeA101Solanum nigrumL.SolanaceaeN088Solanum sisymbriifoliumLam.SolanaceaeA101Solanum tuberosumL.SolanaceaeA34034Solenostemon scutellarioides(L.) R. Br.LamiaceaeA101Solidago canadensisL.AsteraceaeA15015Solidago giganteaAitonAsteraceaeA707Solidago virgaureaL.AsteraceaeN202Sonchus arvensisL.AsteraceaeN01414Sonchus asperHillAsteraceaeN05656Sonchus oleraceusL.AsteraceaeN02828Sorbus aria(L.) CrantzRosaceaeN101Sorbus aucupariaL.RosaceaeN505Sorbus intermedia(Ehrh.) Pers.RosaceaeA202Sorghum bicolor(L.) MoenchPoaceaeA101Sorghum halepense(L.) Pers.PoaceaeA011Spinacia oleraceaL.AmaranthaceaeA101Spiraea × argutaZabelRosaceaeA101Spiraea × cinereaZabelRosaceaeA303Spiraea × vanhouttei(Briot) ZabelRosaceaeA10010Spiraea betulifoliaPall.RosaceaeA404Spiraea japonicaL. f.RosaceaeA505Spirodela polyrhiza(L.) Schleid.AraceaeN202Stachys affinisBungeLamiaceaeA202Stachys byzantinaK. KochLamiaceaeA404Stachys palustrisL.LamiaceaeN011Stachys rectaL.LamiaceaeN202Stachys sylvaticaL.LamiaceaeN066Staphylea pinnataL.StaphyleaceaeN101Stellaria media(L.) Vill.CaryophyllaceaeN02525Stevia rebaudiana(Bertoni) BertoniAsteraceaeA101Symphoricarpos × chenaultiiRehderCaprifoliaceaeA404Symphoricarpos albus(L.) S. F. BlakeCaprifoliaceaeA909Symphytum asperumLepech.BoraginaceaeA202Symphytum grandiflorumDC.BoraginaceaeA505Symphytum officinaleL.BoraginaceaeN16016Syringa meyeriTurcz.OleaceaeA404Syringa vulgarisL.OleaceaeA30030Tagetes erectaL.AsteraceaeA303Tagetes patulaL.AsteraceaeA21021Tagetes tenuifoliaCav.AsteraceaeA202Tamarix parvifloraDC.TamaricaceaeA303Tanacetum corymbosum(L.) Sch. Bip.AsteraceaeN101Tanacetum parthenium(L.) Sch. Bip.AsteraceaeA19019Tanacetum vulgareL.AsteraceaeN18018Taraxacum officinale aggr.AsteraceaeN08585Taxus baccataL.TaxaceaeN16016Telekia speciosa(Schreb.) Baumg.AsteraceaeA202Tellima grandiflora(Pursh) Lindl.SaxifragaceaeA101Tetragonia tetragonoides(Pall.) KuntzeAizoaceaeA101Teucrium chamaedrysL.LamiaceaeN404Thalictrum aquilegiifoliumL.RanunculaceaeN505Thalictrum minusL.RanunculaceaeN101Thelypteris palustrisSchottThelypteridaceaeN101Thuja occidentalisL.CupressaceaeA20020Thunbergia alataBojer ex SimsAcanthaceaeA303Thymus × citriodorus(Pers.) Schreb.LamiaceaeA303Thymus longicaulisC. PreslLamiaceaeN909Thymus pulegioidesL.LamiaceaeN13013Thymus vulgarisL.LamiaceaeA22022Tiarella cordifoliaL.SaxifragaceaeA101Tilia cordataMill.MalvaceaeN101Tilia platyphyllosScop.MalvaceaeN224Tithonia rotundifolia(Mill.) S. F. BlakeAsteraceaeA101Tofieldia calyculata(L.) Wahlenb.TofieldiaceaeN101Torenia fournieriLinden ex E. Fourn.LinderniaceaeA101Torilis japonica(Houtt.) DC.ApiaceaeN011Trachycarpus fortunei(Hook.) H. Wendl.ArecaceaeA224Trachymene coeruleaGrahamApiaceaeA101Tradescantia pallida(Rose) D. R. HuntCommelinaceaeA101Tradescantia virginianaL.CommelinaceaeA12012Tragopogon porrifoliusL.AsteraceaeA303Tragopogon pratensisL.AsteraceaeN437Tricyrtis hirta(Thunb.) Hook.LiliaceaeA202Trifolium campestreSchreb.FabaceaeA011Trifolium dubiumSibth.FabaceaeN088Trifolium incarnatumL.FabaceaeA101Trifolium pratenseL.FabaceaeN12618Trifolium repensL.FabaceaeN06666Tripleurospermum inodorum(L.) Sch. Bip.AsteraceaeN011Trisetum flavescens(L.) P. Beauv.PoaceaeN325Triticum aestivumL.PoaceaeA033Trollius europaeusL.RanunculaceaeN202Tropaeolum majusL.TropaeolaceaeA31031Tropaeolum tuberosumRuiz & Pav.TropaeolaceaeA101Tulipa CVLiliaceaeA48048Tulipa turkestanica(Regel) RegelLiliaceaeA303Tussilago farfaraL.AsteraceaeN055Typha latifoliaL.TyphaceaeN202Typha minimaHoppeTyphaceaeN101Ulmus glabraHuds.UlmaceaeN022Ulmus minorMill.UlmaceaeN011Urtica dioicaL.UrticaceaeN02929Utricularia vulgaris aggr.LentibulariaceaeN101Vaccaria hispanica(Mill.) RauschertCaryophyllaceaeN303Vaccinium corymbosumL.EricaceaeA23023Vaccinium vitis-idaeaL.EricaceaeN404Valeriana officinalisL.CaprifoliaceaeN8412Valerianella locusta(L.) Laterr.CaprifoliaceaeN32032Verbascum blattariaL.ScrophulariaceaeN101Verbascum lychnitisL.ScrophulariaceaeN101Verbascum nigrumL.ScrophulariaceaeN617Verbascum phoeniceumL.ScrophulariaceaeA101Verbascum thapsusL.ScrophulariaceaeN02424Verbena × hybrida(Groenl. & Rümpler) G. L. Nesom & PruskiVerbenaceaeA303Verbena bonariensisL.VerbenaceaeA505Verbena officinalisL.VerbenaceaeN01010Verbena rigidaSpreng.VerbenaceaeA202Veronica agrestisL.PlantaginaceaeN099Veronica arvensisL.PlantaginaceaeN02626Veronica beccabungaL.PlantaginaceaeN303Veronica chamaedrysL.PlantaginaceaeN93039Veronica filiformisSm.PlantaginaceaeA03131Veronica hederifoliaL.PlantaginaceaeN05959Veronica montanaL.PlantaginaceaeN022Veronica peregrinaL.PlantaginaceaeA02525Veronica persicaPoir.PlantaginaceaeA05252Veronica politaFr.PlantaginaceaeN022Veronica serpyllifoliaL.PlantaginaceaeN44347Viburnum × bodnantenseStearnAdoxaceaeA505Viburnum carlesiiHemsl.AdoxaceaeA202Viburnum lantanaL.AdoxaceaeN606Viburnum opulusL.AdoxaceaeN707Viburnum plicatumThunb.AdoxaceaeA101Viburnum rhytidophyllumHemsl.AdoxaceaeA101Viburnum tinusL.AdoxaceaeA101Vicia craccaL.FabaceaeN022Vicia fabaL.FabaceaeA202Vicia sativaL.FabaceaeN011Vicia sepiumL.FabaceaeN05050Vinca majorL.ApocynaceaeA505Vinca minorL.ApocynaceaeN16016Viola × wittrockianaGamsViolaceaeA808Viola albaBesserViolaceaeN134Viola cornutaL.ViolaceaeA18018Viola elatiorFr.ViolaceaeN101Viola hirtaL.ViolaceaeN044Viola odorataL.ViolaceaeN03333Viola reichenbachianaBoreauViolaceaeN04141Viola rivinianaRchb.ViolaceaeN033Viola tricolor aggr.ViolaceaeN18018Vitex agnus-castusL.LamiaceaeA202Vitis viniferaL.VitaceaeA47047Waldsteinia ternata(Stephan) FritschRosaceaeA202Weigela florida(Bunge) A. DC.CaprifoliaceaeA12012Wisteria sinensis(Sims) SweetFabaceaeA16016Yucca filamentosaL.AsparagaceaeA10010Yucca gloriosaL.AsparagaceaeA202Zantedeschia albomaculata(Hook.) Baill.AraceaeA303Zanthoxylum piperitumMaxim.RutaceaeA202Zea maysL.PoaceaeA17017Zelkova serrata(Thunb.) MakinoUlmaceaeA101Zinnia elegansL.AsteraceaeA19019Total6620348110101 Transparency document {#appsec1} ===================== The following is the Supplementary data to this article:Multimedia component 1 We thank Simon Tresch and Andrea Zanetta for their help with data sampling and field work; the garden owners, the Schweizer Familiengärtner-Verband and their local sections for providing access to the allotment garden areas, Grün Stadt Zürich for their support, Rainer Nowack for kindly sharing literature and Chris Young, Simon Tresch and Robert Home for the helpful comments. The BetterGardens-project ([www.bettergardens.ch](http://www.bettergardens.ch){#intref0040}) is financed by the Swiss National Science Foundation (Sinergia; grant no. 154416). Transparency document associated with this article can be found in the online version at [https://doi.org/10.1016/j.dib.2019.103982](10.1016/j.dib.2019.103982){#intref0045}.

Introduction {#sec1-1} ============ The dexamethasone implant (OZURDEX; Allergan Inc., Irvine, CA, USA) is a biodegradable, sustained-release device injected transconjunctivally into the vitreous cavity, containing 0.35 or 0.7 mg of dexamethasone. The implant has been approved by the Food and Drug Administration for use in noninfectious intermediate and posterior uveitis. \[[@ref1]\] The occurrence of raised intraocular pressure (IOP) as a side effect of the drug is around 27% and is usually transient and can most often be controlled medically. \[[@ref2]\] We report a case of uncontrolled ocular hypertension after OZURDEX implant insertion for posterior uveitis in a young patient which necessitated implant removal along with trabeculectomy. Case Report {#sec1-2} =========== A 32-year-old gentleman, diagnosed with chorioretinitis in the left eye probably due to tuberculosis, was treated elsewhere with intravitreal OZURDEX for control of inflammation 72 days prior. Thereafter, he presented to our hospital with secondary rise in IOP. On examination, his unaided visual acuity for distance was 6/6, N6 for near vision in both eyes. His anterior chamber in both eyes was quiet with a clear lens. His IOP was 14 and 42 mmHg in the right and left eye, respectively. Gonioscopy revealed open angles up to scleral spur in all quadrants in both eyes. Fundus examination of his right eye was within normal limits; in his left eye, anterior vitreous face was quiet, optic disc (vertical cup: disc ratio 0.3) and macula were normal, an isolated chorioretinal patch was seen in the inferotemporal quadrant, and the OZURDEX implant cylinder was seen floating freely in the vitreous cavity \[[Figure 1](#F1){ref-type="fig"}\]. On Optical coherence tomogramphy (OCT) macula, his foveal contour in both eyes was normal. The patient was on maximal medical therapy including oral carbonic anhydrase inhibitors, despite which his IOP was uncontrolled. Therefore, we decided to perform a 25-gauge pars plana removal of the implant along with trabeculectomy with mitomycin C \[[Figure 2](#F2){ref-type="fig"}\]. The postoperative period was uneventful with a minimal manipulation of the bleb by the removal of releasable suture on day 5 when the IOP was 18 mmHg, and the IOP then stabilized at 14 mmHg. When the patient came back for a follow-up after 6 weeks, he had a quiet anterior chamber and vitreous cavity and his IOP of 6 mmHg \[[Figure 3](#F3){ref-type="fig"}\]. {#F1} {#F2} {#F3} Discussion {#sec1-3} ========== Ocular hypertension following the use of dexamethasone implant (OZURDEX) is reported to be relatively less frequent when compared to other forms of intraocular steroids.\[[@ref3]\] The IOP usually reaches its peak at around 60 days after the implant insertion.\[[@ref4]\] Uveitic eyes are more prone to have a steroid response due to the already compromised trabecular meshwork.\[[@ref5]\] Although OZURDEX is considered to have a better safety profile, the SAFEODEX study identified six independent risk factors for secondary ocular hypertension after OZURDEX injection, including younger age \<60 years (odds ratio \[OR\] = 2.94, *P* \< 0.001), eyes with uveitis (OR = 3.26, *P* = 0.017), and preexisting glaucoma (OR = 3.7, *P* = 0.017).\[[@ref6]\] Lowder *et al.*\[[@ref1]\] reported that using 0.35 mg DEX implant, 0.7 mg DEX implant, and sham in patients with noninfectious intermediate or posterior uveitis only \<5% had IOP \>35 mmHg and \<10% had IOP \>25 mmHg in all treatment arms. The majority of the study patients needed only a single IOP-lowering medication for IOP control and none required surgical intervention throughout the follow-up period of 26 weeks. However, our patient required a trabeculectomy and a pars plana vitrectomy was also done simultaneously to remove the residual OZURDEX, for a more effective control of IOP. OZURDEX implant removal for anterior migration in aphakic eyes has been reported in the past, especially in eyes with absent capsular support and prior vitrectomy, but there are not many instances of performing pars plana OZURDEX implant removal for uncontrolled glaucoma.\[[@ref7]\] One similar case report from India describes the removal of OZURDEX implant in a child with uveitis after developing intractable glaucoma. Although later, the child still needed trabeculectomy for control of glaucoma despite the removal of the implant.\[[@ref8]\] This can probably be explained by the resultant damage to the trabecular meshwork by the steroid. The effect of the OZURDEX implant usually lasts for as long as 6 months, and a sizable portion of the implant was still present in the vitreous cavity 2½ months after the insertion; hence, we decided to go ahead and remove the implant.\[[@ref9]\] Trabeculectomy was combined owing to the persistently high IOP for a duration \>2 months, the possible permanent damage to the trabecular meshwork, the need for maximum IOP-lowering medications, and the chances of recurrence of inflammation. There are few studies done on the effects of OZURDEX in the younger age group in which population uveitis often occurs. Young people, especially in the age group of \<6 years, respond more dramatically to steroids.\[[@ref10]\] Further studies are required to evaluate the safety of OZURDEX on young eyes with uveitis and to set guidelines for follow-up. This case report throws light on the possible vision-threatening complication of OZURDEX in a young patient that was managed efficiently by removal of implant along with trabeculectomy. Conclusions {#sec1-4} =========== Steroids must be used with caution especially in young patients, patients with uveitic eyes, steroid responders, and patients with preexisting glaucoma. Young patients with uveitis receiving implant must be evaluated periodically, and prompt treatment must be initiated for raised IOP to prevent permanent optic nerve damage. Declaration of patient consent {#sec2-1} ------------------------------ The authors certify that they have obtained all appropriate patient consent forms. In the form the patient(s) has/ have given his/her/their consent for his/her/their images and other clinical information to be reported in the journal. The patients understand that their names and initials will not be published and due efforts will be made to conceal their identity, but anonymity cannot be guaranteed. Financial support and sponsorship {#sec2-2} --------------------------------- Nil. Conflicts of interest {#sec2-3} --------------------- There are no conflicts of interest.

Human movement, cognition, and learning are inextricably bound. Starting in early life, children act upon and understand the environment using mainly sensorimotor actions ([@B160]). Broad changes in perception, cognition, and behaviour appear with the development of a child's sensorimotor repertoire ([@B131]). Different types of motor experiences are prevalent throughout the development from childhood into adulthood, such as reaching and grasping movements ([@B38]), gross motor patterns of varying complexity mastered in the context of physical activity and sport ([@B36]), and gesturing ([@B70]). All these forms of human movements have been shown to affect cognition and learning. In searching for underlying mechanisms explaining the connexion between physical activity, cognition, and learning, two main explanations deriving from two completely different lines of research can be found in the literature: physiological and cognitive. On the one hand, exercise and cognition research has predominantly referred to the physiological changes induced by single bouts (e.g., enhanced catecholamine levels) or multiple bouts of physical activity (e.g., changed brain structures) as an explanation for the inter-relatedness of physical activity and cognitive functioning (e.g., [@B54]; [@B46]). On the other hand, embodied cognition research has mainly focused on cognitive explanations, discussing them in the context of more subtle movements, such as gestures, and more recently whole-body movements (e.g., [@B101]), influencing cognitive processes and learning (e.g., [@B70]). The main goal of this narrative review is to describe both research traditions (i.e., exercise and cognition research and embodied cognition research), and to review the literature to determine whether combining these two approaches may provide a synergistic benefit for both research and educational practice. The included literature will be described in terms of whether the movements are relevant and/or integrated with the cognitive/learning task. In doing so, one can identify that the first research tradition focuses on physiological explanations for the general health and cognitive benefits of gross motor movements in the form of physical activity or exercise, but without considering the relevance of the movements for the learning task. Typically, gross movements are not integrated (i.e., no temporal connexion of the movements with the learning task) with academic lessons, and are not meaningful or congruent with the task. The second research tradition focuses on cognitive explanations of subtle/fine or whole-body movements that are relevant to the learning task. Researchers in this tradition, however, do not consider the physiological benefits of physical activity or exercise. In describing the two lines of research, the concepts of physical activity, exercise, and physical fitness are based on the definitions provided by the [@B23]: 1. Physical activity: "any bodily movement produced by the contraction of skeletal muscle that increases energy expenditure above a basal level." 2. Exercise: "a subcategory of physical activity that is planned, structured, repetitive, and purposive in the sense that the improvement or maintenance of one or more components of physical fitness is the objective. Exercise and exercise training are frequently used interchangeably and generally refer to physical activity performed during leisure time with the primary purpose of improving or maintaining physical fitness, physical performance, or health." 3. Physical fitness: "the ability to carry out daily tasks with vigour and alertness, without undue fatigue, and with ample energy to enjoy leisure-time pursuits and respond to emergencies. Physical fitness includes a number of components consisting of cardiorespiratory endurance (aerobic power), skeletal muscle endurance, skeletal muscle strength, skeletal muscle power, flexibility, balance, speed of movement, reaction time, and body composition" ([@B23]). Acknowledging the aforementioned definitions, the terms "physical activity" and "movement" can be used interchangeably. However, we will use the term "physical activity" to label those activities which have been investigated in the exercise and cognition research tradition: mainly gross motor movements, which largely increase energy expenditure above a basal level. The term "movement," on the other hand, will be used to describe activities which have been investigated in the embodied cognition research tradition: mainly fine motor movements, such as gestures, which only marginally increase energy expenditure. Finally, we propose the third approach, which combines both traditional approaches by looking at integrated, task-relevant physical activities. Blending the physiological and cognitive research traditions to improve both cognition and learning can potentially provide the field of education with valuable insights that can be used to formulate more concrete guidelines for the effective integration of movements in learning environments. Exercise and Cognition Research =============================== The exercise and cognition research tradition has its roots in ageing research ([@B31]; [@B59]). Interested in understanding how physical exercise could reduce the age-related decline in cognitive functioning, exercise scientists adopting a more medical perspective have tried to find the right dose of exercise to reach the largest cognitive benefits (e.g., [@B26]). Only in the last decades, there has been a shift toward the younger population, and concomitantly, interest in the effects of physical activity on children's and adolescents' cognition and brain development ([@B90]; [@B128]). The historical foundation in ageing research coupled with the medical perspective may explain why this line of research is focusing on cognitive functioning as the main outcome variable, rather than on learning, and why the given explanations are so physiological in nature. Physiological Explanations for the Effects of Physical Activity on Cognition ---------------------------------------------------------------------------- The positive effects of physical activity are widespread across various domains of human life. The benefits of physical activity have been expanded beyond cardiovascular health and obesity prevention, and include improved cognitive functioning, as well as brain structure and activity ([@B90]). Several findings indicated that improving cardiovascular fitness through regular exercise induced morphological changes to the brain and enhanced cognitive functioning in ageing humans ([@B92]; [@B32]). Especially, executive functions (EF) appeared to be more susceptible than other cognitive processes to aerobic exercise ([@B31]). EF is the part of cognition that encompasses effortful and goal-oriented functions, including inhibition, working memory, and cognitive flexibility ([@B8]; [@B118]). These foundational components, form the basis for higher order EFs, such as reasoning, problem solving, and planning ([@B33]; [@B105]). The working memory system is especially influential for memory, perception, and attention ([@B7]). It is comprised of the central executive and two peripheral systems, the visuospatial sketch pad and phonological loop. The central executive is responsible for the attentional control of behaviour, such as processing, storage, and coordination of information coming from the peripheral systems. The phonological loop holds acoustic or speech-based information, and is important for speech perception and production, whereas the visuospatial sketch pad is linked with visual perception and action. Findings in older adults resulted in the EF hypothesis, proposing that exercise affect EF's by inducing vascularisation, neurogenesis, and by altering synaptic processes in neural networks supporting EF, therefore influencing higher order thinking processes ([@B137]; [@B90]; [@B173]). Children's cognitive and neural development, and in particular EF and the supporting brain structures, may also be ameliorated by physical exercise ([@B41]; [@B78]; [@B44]). Higher fitness levels correlate to greater school performance due to physiological alterations in the brain structure, e.g., larger hippocampal volume, neurogenesis, synaptic plasticity, oxygenation, and brain circuit of hormones and neurotransmitters among higher fit preadolescent children of 7--10 years old ([@B24]; [@B90]; [@B25]). Although 95% of the brain size is reached by age 6, grey matter volume in the frontal, parietal, and temporal lobes peak during 10--12 and 16--17 years, respectively ([@B64]; [@B97]; [@B90]). Adolescents experience functional changes in EF, where increased activity has been observed in the prefrontal regions during the performance of social cognitive tasks ([@B15]). Developmental neuroimaging studies have shown a gradually maturing sensorimotor system before the emergence of higher order EF, while neuroscientists have linked developmental changes in the brain with behavioural performance measures (e.g., memory function, task performance; [@B20]). Specifically, regions responsible for primary functions such as motor and sensory systems mature earlier than regions related to higher order association. Because the latter regions integrate these primary functions, the consensus is that EF is crucial for mental and physical health, academic success, cognitive, social, and psychological development ([@B42]). It is considered even "more important for school readiness than intelligence quotient" ([@B18]; [@B44], p. 959), and positively affects math and reading performance throughout education ([@B14]). EF is subsequently correlated with on-task behaviour, aiding self-regulation, behavioural inhibition, and the ability to focus on classroom material despite internal or external distractions, which is essential for successful learning. Children show increased on-task behaviour after physical activity programmes at school ([@B140]), confirming that physical activity can positively affect classroom behaviour. Until now, several meta-reviews in children report a favourable relationship between physical activity and aerobic fitness on the one hand, and cognition and brain function on the other hand. In 1997, Etnier and colleagues examined 200 studies, in which 134 were included. They indicated that acute exercise has a significant small positive effect (Hedge's *g* = 0.36) on cognitive performance with children (6--13 years). Also, later in their meta-analysis with the same age group, [@B155], after having examined 118 studies and analyzed 44 studies, reported a similar overall effect size (Hedge's *g* = 0.32). More recently, [@B46] systematically reviewed the relationship between physical activity, fitness, cognition, and academic achievement and concluded that most research findings support the view that physical fitness, single bouts of exercise, and participation in physical activity programmes are beneficial for children's cognitive functioning. Likewise, overall results of a recent narrative review and meta-analysis of [@B171] showed how chronic physical activity interventions positively impacted children's cognitive functioning (Hedge's *g* = 0.46). Finally, the meta-analysis of [@B173] looked specifically at the relationship between physical exercise and EF in preadolescent children, adolescents and young adults. Overall, positive effects of acute physical exercise were found for both general (Cohen's *d* = 0.52) and domain-related EF (i.e., inhibition and interference control, Cohen's *d* = 0.46; working memory, Cohen's *d* = 0.05). Although no effects of chronic exercise were found on general EF, positive effects were found for planning (Cohen's *d* = 0.16). Type and Duration ----------------- Broadly speaking, there are two different methodological approaches to investigate the relationship between physical activity and cognitive functioning, which has consequently resulted in two distinct lines of research. The first type investigates the effects of chronic exercise (i.e., repeated bouts of exercise such as aerobic training; [@B164]). The main goal of the habitual aerobic exercise programme is to enhance children's cardio-respiratory fitness ([@B164]). This improvement, in turn, may induce indirect yet more enduring effects to cognitive functioning, which are assessed as after the chronic aerobic exercise programme has finished ([@B163]). The second line of research investigates the instant changes in cognitive functioning directly after acute bouts of aerobic activity ([@B162]). Participants' accuracy, response time, and speed on cognitive tests are assessed immediately after intense aerobic training. For instance, single bouts of physical activity may provoke physiological arousal, facilitating the available attentional resources and engagement of cognitive functioning ([@B13]). Chronic aerobic exercise generates functional morphological changes in the brain structures (i.e., larger amplitude and shorter latency of the P3 component in the frontoparietal network) critical to learning and memory, whereas acute exercise improves cognitive performance by activating neurochemical responses ([@B12]; [@B16]; [@B26]; [@B81]; [@B162]). Physical activity can enhance cognitive processes related to faster cognitive processing, and increase the allocation of attentional resources during encoding ([@B81]). [@B52] review supports this view, concluding that greater physical activity and higher fitness levels in children is associated with larger grey matter volume in the hippocampus and basal ganglia, greater white matter integrity, and increased neural efficiency (e.g., improved functional connectivity) in adolescents and older adults. Nevertheless, grey matter reductions may also occur after late adolescence due to maturation, continuing to enhance EF as neural patterns respond to an input-dependent manner based on the environmental influences. Here, undesirable behaviors are pruned away in favor of desirable behaviors ([@B15]). For example, instead of just thoughtlessly initiating learning, students plan what and how they will learn beforehand, continuously monitoring their learning processes (e.g., metacognition; [@B58]; [@B184]). Despite the positive associations among aerobic exercise, fitness, cognition, and academic achievement, there are many quantitative (i.e., type, amount, frequency, and timing) and qualitative aspects (i.e., task complexity, novelty, and diversity/variety, emotional activation, and selection of strategies) of physical activity in relation to cognition that remain to be explored. Here, exercise and cognition research has only recently began to shift the focus from the quantitative to the qualitative exercise characteristics ([@B127]). Based on the simple idea that various physical activities may not only differ in their intensity, duration, and frequency, but also, for example, in their coordinative and cognitive complexity, this new perspective is almost exclusively restricted to children and adolescent samples in previous research ([@B171]). Exercise psychologists working in this area of research are stressing the importance of the cognitive demands inherent to many physical activities ([@B12]). The basic assumption of their cognitive stimulation hypothesis, is that coordinatively demanding and non-automated physical activities activate the same brain regions that are used to control higher order cognitive processes ([@B12]; [@B127]; [@B164]). For the relation between *acute* physical activity and cognition, it is assumed that these cognitive demands are leading to better cognitive performance by pre-activating the same cognitive processes during physical activity as the ones used in a subsequent cognitive task ([@B17]). For example, when playing a physical version of the number connexion test ([@B123]), requiring the subject to discriminate visual stimuli, perform fast mental operations, and react with an appropriate motor response, performance, as a result, is facilitated in a consecutive cognitive test involving the same cognitive processes ([@B148]). However, findings of acute studies explicitly testing this hypothesis remain equivocal, with some revealing positive effects on cognitive performance in favour of the cognitively challenging condition ([@B17]; [@B129]; [@B84]; [@B10]; [@B148]), some finding no difference ([@B13]; [@B85]) and others even reporting detrimental effects compared to physical activity without cognitive challenges ([@B61], [@B60]; [@B51]). When it comes to *chronic* physical activity interventions, there is tentative evidence that cognitively engaging aerobic exercise, in which strategic behaviours, complex motor coordination, and adaptation to changing task conditions is required, benefits children's EF (i.e., working memory, inhibition, and shifting) more than non-engaging simple and repetitive actions ([@B27]; [@B149]; [@B91]; [@B130]; [@B168]). Interestingly, the shift from the quantitative to the qualitative characteristics of physical activity is accompanied by a preference for psychological mechanisms explaining the relationship between physical activity and cognition. Thus, at least in terms of their referred theoretical assumptions, this lastly evolved line of exercise and cognition research is approaching the embodied cognition research, which will be described below. Embodied Cognition Research =========================== The roots of embodied cognition research can be placed in literature related to memory of action events ([@B49]; [@B183]), suggesting that performing actions leads to the construction of rich and elaborative representations that enhance memory recall. The "enactment effect" engages the motor system, in which encoding is facilitated compared to listening or just observing the same events. The enactment effect sets the foundations for the embodied cognition perspective. This notion was further developed in psychological research in different domains, such as neuroscience ([@B112]), memory (e.g., [@B65]), and language research (e.g., [@B67]; [@B99]; [@B185]). It was concluded from this psychological research that the emphasis should be placed on the role of bodily form, real-world action, and environmental influences. In that sense, embodiment includes the actual bodily states and the simulations of the experience linked to the cognitive processes (i.e., action, perception, and introspection; [@B121]). Based on this tenet, a substantial body of research emerged using gestures during learning (e.g., [@B2]). Expressing information in gestures is one specific type of sensorimotor experience that has been shown to be effective for children's learning in different domains, such as mathematics (e.g., [@B34]; [@B122]), science (e.g., [@B133]; [@B102]), and language (e.g., [@B159]; [@B108]). Research into the effects of movements in the form of gestures has mainly used learning as outcome variable, and used cognitive explanations to explain these effects. Cognitive Explanations for the Effects of Movements on Learning --------------------------------------------------------------- Building up a conceptual framework, there are several explanations for the cognitive benefits of the human movement effect on learning. In the educational research field, there is a growing body of research that supports the positive effects of engagement in sensorimotor experiences (i.e., the body) during learning activities on learning (e.g., [@B135]). Central to this respect, is the theoretical framework of embodied cognition, which holds that cognitive processes are profoundly dependent upon body's interactions with the world ([@B181]; [@B9]; [@B69]). Research shows that visual and motor processes in the brain are involved during cognitive tasks, such as text comprehension, mental arithmetic, reasoning, and problem solving, while semantic codes are activated during specific motor actions, illustrating the inter-relatedness of cognitive and sensorimotor processes. Complementary to the embodied cognition theoretical framework is the evolutionary explanation of cognitive load theory, which categorises information into biologically primary and secondary ([@B62], [@B63]; [@B125]): biologically primary knowledge evolves naturally without explicit instruction. For example, the development of native language or the use of movements have been acquired effortlessly and sometimes even unconsciously. Biologically secondary knowledge usually occurs after explicit instruction during formal schooling (e.g., mathematics, science). This type of information can be conveyed through conscious processing, meticulous attention, and effortful deliberate practise. It can be argued that the human motor system, as a form of biologically primary knowledge can be employed to support learning of complex tasks (i.e., biological secondary knowledge). In addition to the embodied and evolutionary approach, the cognitive explanation suggests that including relevant movements during learning rather than only observing or listening to a task creates a richer memory trace in long-term memory, rendering it more accessible during recall ([@B56]; [@B111]). Mental representations consist of an elaborated network of information regarding concepts. New information is processed in the working memory. When this information is embellished with sensorimotor experiences stemming from multimodal resources (i.e., visual, auditory, and kinaesthetic), it leads to the construction of higher quality mental representations or cognitive schemas, and in turn faster and better memory performance ([@B72]; [@B133]; [@B35]). Schema enrichment for students' engagement in action along with perceptual tasks can also be linked to the dual coding theory ([@B29]). This theory emphasizes that students who are involved with motion and perception are able to connect the verbal and visual cues with their kinaesthetic "imagery" ([@B29]). For instance, a mental image for the word "Bunsen burner" comprises a visual image of the object, auditory, and olfactory images for the sound and smell of gas, and motor images for adjusting the flow of gas ([@B29]). Overall, using tactile and kinaesthetic cues can offload some of the working-memory resources, as information is processed simultaneously but not through the same system (for a review see, [@B141]). Studies in the domains of mathematics ([@B72]) and Piagetian conservation tasks ([@B133]) have shown that hand gesturing can reduce a speaker's cognitive load during instruction and problem solving. Finally, in accordance with the mirror neuron system, sensorimotor actions, thoughts, or words activate mirror neurons, and in turn mental representations responsible for movements, such as action and language for example that are both located next to the Broca's area in the brain ([@B6]; [@B5]). The mental imagery of the movement, supplementary to the physical execution of the movement, is an alternative way which can possibly achieve the same gains in cognitive learning. The mirror neuron system has been linked with action understanding as well as the ability to observe and imitate others' actions ([@B142]). However, this notion has been under great dispute ([@B19]; [@B76]). [@B169]; see also [@B125]) have argued that the mirroring capacity of the human brain can be used in learning when a motor component is encompassed in the cognitive tasks. Motor imagery could be combined with a concurrent demonstration of the same action, enforcing the involvement of the mirror neuron system, to produce more effective motor learning outcomes ([@B50]). However, more research is needed to shed light on this notion. Nevertheless, solid empirical evidence attests various concrete examples of the aforementioned explanations for the benefits of embodiment on learning. Most of the existing research has focused on subtle movements, such as gestures, with fewer examples existing of research looking at whole-body movements. These movements do not provoke any physical exhaustion, but they are a significant adjunct of the learning process. Gestures can be distinguished into beats (non-representational gestures), iconic (implying a perceptual relationship between concrete concepts), metaphorical (having a narrative character for abstract concepts), and deictic gestures (when the speaker points to actual objects; [@B144]; [@B82]). An illustrative example of a gesture study using iconic gestures, is a study by [@B34]. They reported that children who were instructed to gesture meaningfully (i.e., children first place their hand on the left side of the equation, then they pose, and finally put their hands under the right side), while practising solving mathematics problems, increased their learning relative to children who were told to only speak during the practise phase. These results provide support for a causal role for gesturing in learning. Moreover, [@B93] showed that mimicking gestures were beneficial for learning to write foreign language characters, both in instructional designs, including animations and static graphics. Although the exact mechanism is not completely understood, these findings are in line with the embodied cognition hypothesis, that expressing information in multiple modalities can promote the construction of higher quality cognitive schemas more than conveying the same information in speech alone. Another demonstrations of the benefits of embodiment on learning comes from a study by [@B72] who explored how gesturing during one task (i.e., explaining solution to math problems) impacted the performance on another task (i.e., remembering letters/words). Both children and adults who were given the dual task of remembering letters (for adults) or words (for children) while explaining a difficult math problem remembered more items when gesturing than when not gesturing ([@B72]). The higher score on the secondary task after explaining the math problem solution with gestures indicates that gesturing reduced the cognitive load imposed by math explanation freeing up resources that can be used on a secondary memory task. It is important to note that only task-relevant gestures, defined as movements that are in coordination with the content of the speech, reduce the cognitive load of explanation ([@B35]). Further evidence for the relation between gesturing and cognitive load is provided by studies that found a positive relation between cognitive task complexity and gesture frequency, and studies that found gesturing to be especially beneficial for individuals with low working memory capacity ([@B28]; [@B134]). Finally, some technology-based educational studies have not only used gestures, but also whole-body movements to enhance learning. These studies are based on the theory that knowledge is grounded in body-based actions, where the body works as a sensorimotor metaphor, converting the abstract concepts into tangible examples (e.g., [@B87]). For example, in a recent study, a group of middle school students were put in the role of an asteroid. By using their bodies in an immersive, whole-body, interactive environment they learned about planetary motion and gravitational forces ([@B100]). Another group of students learned about planetary motion using a desktop version of the same simulation. Results of the study showed that the group of students in the interactive digital environment having an embodied learning experience had significantly higher learning performance, higher levels of engagement, and more positive attitudes toward science than the other group using a desktop computer. A Matrix From the Prism of Task Relevance and Integration Levels ================================================================ Although this is a narrative review, given the very broad inclusion of (1) physical activity tasks, (2) cognitive/learning outcomes, (3) length of interventions and of single activity bouts, and (4) intervention settings, arising from the two research traditions (i.e., exercise and cognition research and embodied cognition research), a minimum inclusion criteria was applied: acute and chronic physical activity interventions, gross and fine-motor training tasks, and age (children, adolescents, and adults). The second part of this review will categorise the selected studies on a continuum based on the relevance of the movements for the cognitive/learning task and the integration of the movements into the cognitive/learning task (see Figure [1](#F1){ref-type="fig"}): {#F1} 1. Relevance of the movement for the cognitive/learning task: this categorisation refers to the level of embodiment or relatedness of the physical with the cognitive task. Of a dominant role to this continuum is the factor of embodiment, referring to the enactment of concepts using the body ([@B100]). These range from no embodiment, where the movements are not related with the cognitive tasks, to high embodiment, in which whole-body movements are engaged, meaningfully related with the learning tasks. For example, performing locomotor skills such as skipping, sliding, walking, and jumping, while avoiding obstacles when working on math problems can be considered as non-task-relevant movements ([@B172]), whereas dance movements that children perform when learning the foreign language word for dance ([@B113]) can be considered as task-relevant movements. Both activities, however, are integrated into the corresponding cognitive/learning task. 2. Integration of the movement into the cognitive/learning task: this categorisation refers to the temporal connexion of the movements with the learning task. If the movements are performed before or after the learning task with an interval in between, the integration is low. If movements are performed during the learning task, the integration is high. For example, performing fluid arm movements immediately before being tested in creativity, can be considered as being task relevant, but it is not integrated into the cognitive/learning task ([@B156]). At the other end, an additional after-school physical education programme has low integration level, as the physical activities occurred beyond the academic instruction time and are considered as having low task relevance ([@B30]). A continuum exists ranging from non-integrated movements, where there is no temporal overlap between movements and the learning task, to integrated movements, where the movements are connected and included during learning. Figure [1](#F1){ref-type="fig"} plots four examples chosen for each quadrant of the 2 × 2 matrix based on their relevance and integration with the cognitive/learning task. Research on exercise and cognition (i.e., physiological mechanisms) falls in the low-relevant quadrants, whereas research on integrated movement studies (i.e., embodied learning) falls into the high-relevant quadrants. A discussion of the studies belonging to each of the quadrant is presented below. Bottom Left Quadrant: Body Movements Not Relevant for, and Not Integrated Into the Cognitive/Learning Task ---------------------------------------------------------------------------------------------------------- Research of this area involves studies examining the acute effects of exercise on cognition ([@B80]), activity break studies, and afterschool programmes focusing on physical fitness and its indirect relationship with cognitive and academic performance ([@B21]; [@B39]), Starting with the acute exercise studies, research has suggested that a single bout of exercise improves cognitive performance on attention demanding tasks in preadolescent children. For instance, [@B80] found that single, acute bouts of moderate treadmill walking improved the cognitive control of attention in children as measured with a modified flanker task. Applied aspects of cognition involved in school-based academic performance improved from this single bout of exercise also. In another similar study, preadolescent children performed better on cognitive tasks measuring attention and inhibitory control after short bouts of 20-min walking compared with children who remained seated (either walking or seating breaks were incorporated in between the cognitive tasks for three sessions; [@B48]). Also in the classroom itself, non-integrated and non-relevant activity can take place. Activity breaks, for example, are interspersed between phases of learning, but they do not overlap in time with the learning tasks in the classroom. Several studies investigated forms of physical activity that occurred in the classroom aside from gym classes, recess, or breaks ([@B88]; [@B158]; [@B83]). Generally, these studies investigated the effects of short physical activity breaks (most commonly aerobic routines for 5--20 min), or ways to incorporating physical activity into the learning activity that were either intended to increase learning through motor actions, or to provide purely an exercise moment for children to increase energy expenditure. The studies examined whether and how introducing activity breaks in a classroom environment, impacted health, cognitive skills ([@B77]), attitudes (mood, motivation; [@B170]), academic behaviours (i.e., on-task behaviour, concentration, motivation; [@B73]; [@B176]), and academic achievement (i.e., reading literacy scores or math fluency scores; [@B88]) of children. In general, reviews on the relationship between activity breaks with aspects of academic performance show that activity breaks either have positive effects or do not adversely impact cognitive function and academic performance ([@B22]). For example, research revealed that 4 min of high-intensity interval activity in class enhances selective attention in children aged 9--11 years ([@B106]). Likewise, running games or performing fundamental movement skills such as hopping, skipping, and jumping in the classroom could enhance fluid intelligence and academic achievement in 9--11 years old children ([@B138]). Finally, afterschool physical activity programmes fall under the category non-integrated, non-relevant physical activity. For example, the 9-month afterschool physical activity programme "FITkids" for children 7--9 years old aiming to ameliorate brain health and cognitive performance found improvements in heart rate as a measurement of physical activity, physical fitness, inhibition, and cognitive flexibility in children who participated in the afterschool intervention group, as compared to the control group ([@B21]; [@B79]). The physical activity intervention offered two hours of physical activity each day after school, in which children were requested to participate for up to 40 min. In the study of [@B30], students were enrolled into a physical education programme (including activities such as basketball, soccer, softball, and baseball), or an extra instruction class either in the first or second semester for 55 min every day. Their academic achievement was measured by students' grades on four core courses (mathematics, science, English, and world studies), and a standardised test. Academic achievement was not altered by the moderate intensity in the physical education class; however, children who partly or fully met the guidelines of 30 min moderate activity per day for at least 5 days per week, or 20 min of vigorous activity per day for at least 3 days per week, had higher grades than students who were not engaged in vigorous physical activity in both semesters. [@B39] assessed 7- to 11-year old children's EF after randomly assigning them to a 40 or 20 min/day exercise programme, or a no exercise control condition for 3 months. Children in the high dose exercise group completed two 20-min bouts per day, whereas children in the low dose group completed a 20-min bout of exercise and 20-min sedentary activities (e.g., drawing, board and card games). Standardised cognitive assessments, achievement measures, and fMRI were used in the study. Results demonstrated that regular aerobic exercise advanced cognitive functioning in both exercise groups compared to the control group, but with more pervasive effects observed in the high dose exercise group. The high dose exercise group also outperformed the other groups in math scores. Top Left Quadrant: Body Movements Not Relevant for, but Integrated Into the Cognitive/Learning Task --------------------------------------------------------------------------------------------------- These studies used active workstations (e.g., standing and treadmill desks, cycling workstations) while completing cognitive tasks, physical activity learning games, and exergaming in the classroom. The first type of integrated, but non-relevant, physical activity can be identified in a new area of research focusing on the effects of active workstations on cognition. Several meta-reviews report a positive relationship between active workstations, mostly examined in office environments, thereby resulting in reductions in sitting time for adults, increments in energy expenditure, and improved health ([@B165]; [@B110]). Notwithstanding the health benefits, research into the cognitive effects and computer task performance when using these active workstations remains limited. Further investigation is needed before conclusions can be drawn. Research to-date seems to indicate that low-intensity exercise does not compromise cognitive functions (e.g., [@B146]). For example, [@B132] assessed undergraduate students' performance, motivation (self-reported answers regarding enthusiasm, energy, drive, eagerness, and morale), and engagement (self-reported responses regarding subjective performance, attention, and absorption), when they used a stationary desk bike with a desk top and a traditional desk with chair during the completion of cognitive tasks. Although cognitive performance did not differ, cycling desks improved positive affect, motivation, and morale. By contrast, studies of Schmidt-Kassow and colleagues showed that treadmill walking ([@B151]) and cycling ([@B150]) during vocabulary encoding improved subsequent recall. The authors attributed this improved performance to the temporal alignment of stimulus presentation and motor activity. Since active workstations are a novelty, particularly within a school setting, there is currently only one systematic review that has investigated the physiological effects of standing desks interventions within the classroom setting ([@B152]), suggesting that energy expenditure is increased when using standing desks within the classroom. Overall, implementing active workstations in classrooms could be used to decrease sedentary behaviour with no detriment of cognitive performance; however, more research is needed to glean insight in active workstations' potential as cognitive enhancers. Other types of classroom-based physical activity can also be gathered under the category of non-relevant but integrated movements. For example, [@B172] integrated a 10-min bout of acute aerobic exercise into math lessons. In the integrated condition, preadolescent children performed locomotor skills such as skipping, sliding, walking, hopping, leaping, bear and crab walking, and jumping to avoid obstacles when working on math problems. For example, when the answer was an odd number, students were crab walking, otherwise in case of an even number, they were bear walking. In the seated math practise condition, no bodily movements were involved. They found that elementary children in the integrated physical activity group showed significant improvements on accuracy while performing a cognitive task for inhibitory control, and higher scores on enjoyment compared with the seated math practise group. The authors concluded that classroom-based physical activity are an enjoyable and realistic strategy to increase physical activity and facilitate EF in children. Finally, an exercise and game-based learning approach linking digital technology with exercise and learning, also known as exergaming, can be placed in this quadrant. Exergaming is a developing area of research with yet to be established results. It seems a promising way to increase physical activity, improve general coordination skills, motivation, and cognitive outcomes (e.g., improved attention and visual--spatial skills) through linking exercise with digital technology and learning ([@B157]). Exergaming has also been implemented in elementary school for different learning contents such as math, history, and languages, and with varying difficulty levels ([@B104]). Higher scores in cued recall were found in the experimental condition, in which children had to jump and move as quickly as possible on a sensor pad when playing a vocabulary game compared to the traditional sedentary condition. The integration of physical activity into learning games seems to be enjoyable and engaging, holding promise for learning in children. It should also be noted that these studies had high ecological validity as they took place in real-life situations and existing school lessons. To conclude, it is premature to infer about the effectiveness of active video games on augmenting energy expenditure and learning; however, active video games may have an additional benefit on engaging players with light-to-moderate physical activity compared to other sedentary behaviours such as passive video games and rest ([@B95]). Given the length of time and relatively high-frequency that children and adolescents spend on video gaming (at least 1 h per day on the weekdays and 1.5 h on the weekend for children 10--19 years; [@B37]), successfully designed exergaming is a promising method to positively impact physical activity levels and cognition in children. Bottom Right Quadrant: Body Movements Relevant for, but Not Integrated Into the Cognitive/Learning Task ------------------------------------------------------------------------------------------------------- This kind of research can both be found in the embodied cognition and in the exercise and cognition research tradition. Whereas the studies from the embodied cognition literature involve effects of bodily movements on several measures of cognition such as insight problems or creativity (mostly arm movements at a very basic physical activity level), studies out of the exercise and cognition literature predominantly investigate the effects of whole-body movements on cognitive measures, such as EF. The movements occurred prior or after the cognitive task, with the key concept being the accordance between the movement and the cognitive task. Thus, these movements function as simulated actions, indirectly fostering information processing ([@B117]; [@B1]). It is argued that the body can work as a scaffold, or conceptual metaphor, to abstract cognitive contents ([@B180]; [@B94]), guiding higher order cognitive processing ([@B161]). [@B161] allocated university students to exercise breaks either consistent with (swing group) or inconsistent with (stretch group) an insight problem that they had to solve immediately afterward. They had to solve Maier's two-string problem, which they could only solve by swinging one of the ties with an object attached. The experiment consisted of eight attempt intervals with 20-s exercise and 100-s problem-solving periods. It was found that the swing group was better in solving the two-string problem than the stretch group, without being aware of the swinging arm movements as overt hints to problem-solving. Consistent with these results, [@B178] found a movement-specific influence on participants' solution of two-string and water-jam problems. They assigned adult participants to two movement groups that were congruent with the problem solutions or a control group. To this end, [@B136] suggested that movement goals rather than the arm position cause activation and internal evaluation process, affecting performance in a word association task. Another study compared effects of fluid with non-fluid movements on creative thinking ([@B156]). A set of drawings was designed in which undergraduate students had to trace either arm movements with or without line curvature (fluid and non-fluid movements, respectively). Subsequently, participants were assessed in creative generation, cognitive flexibility, and remote associations. Fluid movement enhanced creativity in all three domains, even though participants did not consciously perceive this positive affect. [@B68] provided evidence on the causal link between language comprehension and the motor system. Participants were required to perform a 20-min repetitive transfer motor task in which they had to move 600 beans from a wide-mouthed container to a narrow-mouthed container. After moving the beans, they had to read sensible and non-sensible sentences describing transfer events with abstract (e.g., "responsibilities") and concrete (e.g., "cards") objects toward or away from themselves. It was found that the participants' concrete and abstract language comprehension differed based on the motor task (i.e., arm movements toward vs. away from the body, and right-hand vs. left-hand index finger). A last series of studies in which the adopted movements can be considered as being relevant for, but not integrated into the cognitive task, are the ones searching for identical elements ([@B182]) between the qualitative characteristics of specific physical activities and the cognitive task used to measure the dependent variable, such as attention or EF ([@B84], [@B85]; [@B10]; [@B149], [@B148]; [@B51]). For example, in a group randomised controlled trial, [@B149] designed a 6-week intervention in physical education enriching traditional team games, e.g. basketball, with additional instructions in which the children had to keep in mind different rules (updating), react appropriately to acoustic cues by inhibit pre-potent movements (inhibition), as well as switch between different situations and rules (shifting). At the end, they compared this cognitively challenging intervention to an aerobic as well as an active control condition, measuring their core EF (updating, inhibition, and shifting). Despite similar benefits in aerobic fitness of the children of both the aerobic and the enriched condition, only the children of the enriched condition displayed higher gains in shifting performance. Top Right Quadrant: Body Movements Relevant for, and Integrated Into the Cognitive/Learning Task ------------------------------------------------------------------------------------------------ The studies included almost exclusively address years of children's development in the school context. First, subtle movements, mainly gestures, are described to continue with studies involving whole-body movements, and finally gross motor movements in the form of physical activity. ### Subtle Movements -- Gestures Well known in this field are the mathematical studies in which 9- to 10-year-old children were taught to produce an abstract gesture as a tool for solving new mathematical-equivalence problems in the form of 5 + 2 + 4 = \_\_ + 4 ([@B72], [@B71]). In solving these problems, children first had to produce a V-point gesture to the first two numbers to represent addends of an equation, and then point at the blank were the sum of the equation should be put. Children who were instructed to make these gestures during a math lesson benefited more from math instruction than children who were not prompted to make these hand movements. They retained more of the knowledge and were able to extract the underlying grouping strategy, although they were never explicitly told what the gestures stood for. Gestures as a form of indexing and enactment have also been shown to facilitate learning in domains of language comprehension, and primary and secondary language acquisition ([@B89]; [@B109]). For example, in case of vocabulary learning, [@B107] taught undergraduate students novel words from an artificial language. Participants enrolled in the gestural condition in which they had to read, repeat the words they heard, and imitate the researcher who was enacting the words with symbolic gestures, were able to recall more words and for a longer period of time than those who had just read, heard, and repeated the same words. Similarly, gestures can be used to enact sentences or storeys to improve reading comprehension: the process of reading a text and understanding its meaning ([@B57]; [@B40]). For example, [@B66] found that children who were manipulating (actual and imaginative) toys referred to in the text they were reading (e.g., farm animals) remembered more action sentences from the storey and showed a better comprehension of the text than children who only read the storey. In another recent study on reading comprehension, [@B11] compared the effect of indexing (i.e., mapping words to objects) and enactment on memory of a narrative. Children in the indexing condition used play mobile figures to perform the storey while children in the enactment condition took on the role of the characters and acted out the storey with gestures and emotional expressions. The results revealed that children in the enactment condition remembered more descriptive parts of the storey than children in the read only condition. Moreover, both enactment and indexing benefited children with poor reading ability. Finally, iconic gesturing in the form of actions or object-manipulations has been proven useful for learning, facilitating problem-solving and vocabulary retrieval by activating relevant perceptual-motor information (e.g., [@B109]). ### Whole-Body Movements In the area of math, it has been shown that whole-body movements can help children adequately develop a spatial representation of number magnitude. For example, [@B145] examined the process of number building (two-digit numbers) in first-grade children by making steps on a ruler across the floor. In the two-movement condition, children made small, medium, and large steps representing different number units of 1, 5, and 10, respectively, whereas in the two control conditions, children had to verbally construct the two-digits numbers. Results showed higher test performance when children were engaged in full-body movements. [@B154] examined the effects of cooperative active learning when learning angles in geometry class. Second- and third-grade students were enrolled either in an experimental or a control condition. In the experimental condition, children collectively formed a circle with their bodies to learn about the circle, whereas children in the control condition learned about the subject geometry through the sedentary conventional method. It was found that the experimental group using movements in a cooperative learning setting reached better results than the conventionally taught group without movements. Finally, a recent study looked at the effects of whole-body movements and gestures on learning foreign language vocabulary in 4-year-old children ([@B166]). First, children were shown flashcards with animal names. Children were assigned to one of three groups in which they had to recall animal words (a) through performing physical activities and gestures relevant to the animal words to be learned, or (b) gesturing related to the animal words, or (c) through the conventional way. Results showed that learning a foreign language vocabulary through physical activities and gestures was considered the most enjoyable way of learning, and resulted in the highest learning outcomes. In this study, no physiological outcomes were measured. ### Blended Approach: Combining Gross Motor Movements With Learning Studies consisting of whole-body movements in the form of classroom-based physical activity are presented, measuring both cognitive and physical activity outcomes. For instance, during the daily 15-min of classroom-based physical activity, elementary school children could learn geometry, by forming different shapes with their bodies (e.g., squares or triangles) while walking or hopping on an outdoor playfield, geography, by running to the appropriate area allocated for one of the cardinal directions, or spelling, and by hopping onto a floor mat with alphabet letters onto it ([@B47]). The results revealed significant improvements in academic achievement as well as a significantly lower increase in body mass index among children in the experimental compared to the control classrooms. A series of intervention studies integrating physical activity into learning tasks lasting 10--15 min per day also demonstrated prominent effects in early childhood ([@B113], [@B115], [@B116], [@B114]). Children were enrolled in the integrated condition in which children were engaged in meaningful, task-relevant physical activities (e.g., dancing while learning the word dance; imitating animal movements relevant to animals living in each continent while learning about the continents and animals living there; moving from the Sun to Mercury and repeat the same process for all planets while learning about the planets' names and their distance from the Sun, counting numbers while walking on foam blocks of numbers), a condition involving task non-relevant physical activities (e.g., running around the room before the learning task), or control condition (sedentary way of learning). Overall, the integrated condition had higher learning scores, was more physically active and enjoyed learning the most, compared to the control condition. Discussion ========== Converging Different Lines of Research Toward a Blended Approach ---------------------------------------------------------------- This review summarised theoretical and empirical evidence, connecting action with perception, cognition, and learning. Previous research has revealed the positive association of physical activity and mental health ([@B179]), as well as the synergistic effects of physical activity, and fitness on cognition and academic performance ([@B96]). Concomitantly, research about embodiment evaluates the role of the body (e.g., gestures, object manipulation) during the learning process, arguing that embodying knowledge through task-relevant movements can positively impact learning ([@B9]). Overall, a new view of embodied cognition can be supported in this review, in which motor and cognitive control, are related to the same cognitive processes ([@B128]). Summarizing the cognitive explanations for including movements during learning suggested in this work, the benefits lie in their associative activation of imagery. Imagery is one way to strengthen the motor representation of the task, which can improve motor performance, and may presumably have a potential benefit on learning. In fact, mental imagery (consisting of visual, motor, or kinaesthetic imagery, as related to our senses) has been linked to better motor performance, general exercise experience, and a variety of mental skills (such as anxiety management, confidence, and concentration; [@B153]; [@B177]). Motor imagery ability may be a key function in determining the extent to which a learner may prepare, rehearse, and subsequently "embody" the activity, long after the physical execution of the task itself. [@B86] suggests that the motor representation produced can be achieved via action observation, and can be functionally equivalent to motor imagery. To this end, a study in adult participants using fMRI explored the cerebral structures engaged in visual and kinaesthetic imagery, concluding that overlapping similar activations (i.e., in motor-related regions and superior and inferior parietal lobules) occurred during physical executing the task, and both visual and kinaesthetic imagery, in comparison with the perceptual condition ([@B75]). However, different brain patterns were activated during visual and kinaesthetic imagery, with kinaesthetic imagery engaging more motor-associated structures. A study in 120 school children (9--10 years) examined whether participation in physical activity and movement imagery ability can predict active play imagery ([@B74]). It was found that active play imagery can be determined by age, participation in physical activity, and ability to use external visual imagery. On the other hand, examining the physiological mechanisms of movements on cognition and learning performance, research in physical activity, exercise, and fitness, vary vastly in intensity, duration of bouts or interventions, cognitive challenges, and time relation between the physical and cognitive engagement. Several types of physical activity interventions such as enhanced or enriched physical education, classroom-based physical activity, activity breaks or active play during recess, extracurricular physical activity interventions, or after-school programmes were included in this review. The contextual factors of physical activity inclusion (e.g., place, type, and duration) seem to be determinant when inferring about the association among physical activity and cognition (i.e., EFs such as working memory, inhibition, and cognitive flexibility, metacognitive functions such as abstract reasoning, problem-solving, and cognitive life skills such as self-regulation, goal setting), academic achievement (i.e., mathematics, language scores), and academic behaviours (i.e., on-task behaviour; [@B3]). Physical activity so far revealed a potential to elicit improvements in cognitive performance. However, there are substantial differences concerning the effects of effortful and prolonged bouts of physical activity on cognition, and those of cognitively engaging movements of low intensity and duration. For example, negligible effects of low-intensity and short physical activity tasks are reported in reviews on acute exercise and cognition ([@B162]; [@B26]; [@B171]). By contrast, more pronounced effects can be observed during cognitively engaging physical tasks (meta-analysis of [@B171]). The mentally enriched and cognitively engaging physical activities offer a range of different inherent qualitative characteristics (i.e., task complexity, novelty, and diversity/variety, emotional activation, selection of suitable mental strategies). These characteristics act as brain stimulators contributing to enhancements in children's executive functioning ([@B45]; [@B171]). Thus, it can be speculated that if these qualitative characteristics are connected with the learning material, the highest learning effects can be expected. Converging the cognitive and physiological mechanisms of movements, we propose an innovative instructional method that combines task-relevant physical activities integrated with learning tasks (Figure [2](#F2){ref-type="fig"}). We believe that by adopting this approach, children could benefit from the combined physiological (e.g., increased arousal, neurological alterations in the brain; [@B12], [@B13]) and cognitive effects (e.g., embodied learning; [@B100]) on cognition and learning. Gathering support toward the integrated task-relevant physical activity programmes, they may offer paramount health and cognitive outcomes (meta-analysis of [@B55]; systematic review of [@B103]; systematic review and meta-analysis of [@B120]; systematic review and meta-analysis of [@B124]). Existing programmes have already received positive social support and feedback within the school environment, because their versatility takes into account time and budget constraints, and limitations in teachers' experience ([@B175]). Integrated physical activities offer different possibilities and variations based on the level of task complexity, children's age group, and syllabus' restrictions. This flexibility promotes high ecological validity including a universal applicability to all classrooms settings in which students are engaged in high quality learning activities in an engaging, motivating, and amusing way. Hence, it is suggested that the classroom instruction integrated with physical activities does not have a detrimental effect on academic time ([@B147]; [@B174]), but rather positions academic content centrally, empowering it. {#F2} Notwithstanding, valuable efforts have been taken to increase physical activity through stealth interventions ([@B143]), mostly in school years. Nevertheless, effective programmes in the early years can improve children's EF ([@B43]). Coupling physical with cognitive tasks fosters cognitive development, motivation, and engagement ([@B98]; [@B119]). Overall, the integration of movement experiences to learning areas results in children being more physically active, enthusiastic, and attentive to learning tasks, such as math or language ([@B167]; [@B115], [@B116], [@B114]; [@B139]). We argue that similar efforts should target all age groups in a way that time spent in schools contributes to children's physical and mental wellbeing, cognitive and socio-emotional functioning, and the development of healthy lifestyles, with ample directions toward prevention of later cognitive delays and building of school readiness ([@B4]; [@B126]; [@B53]). Conclusion ========== Blending Physical Activities With Classroom Instruction ------------------------------------------------------- In this review, we have provided an overview of studies dealing with the effects of physical activity on cognition and learning. We identified two different research traditions, which have evolved unconnected and side by side, based on completely different theoretical backgrounds: Exercise and cognition research and embodied cognition research. In categorising the studies with respect to the *relevance for* and the *integration into* the cognitive/learning task, it became obvious that in these two lines of research various methodological approaches are used. Whereas acute exercise and cognition research focuses on the immediate impact of gross motor physical activities on basic cognitive processes, such as attention or EFs, for example, embodied cognition research is more interested in affecting the learning process itself by fine motor movements, such as gestures. Although considering research into the relationship between physical activity, cognition, and learning through the lens of relevance and integration provides a rough, yet further to be developed categorisation, it offers some insight in how both lines of research could profit from each other. For example, embodied cognition research conducted with children in the educational setting revealed that task-relevant whole-body movements can facilitate foreign vocabulary learning ([@B113]). However, in these studies, attentional performance immediately after single learning sessions has not been considered. For the educational setting, e.g., lesson scheduling, it could be interesting to know if children's attention is enhanced or deteriorated after such interventions. On the other hand, exercise and cognition research is so focused on constructs such as attention and EF that, to the best of our knowledge, there are few studies testing whether the effects on these variables also impact learning as a consequence. Therefore, for both lines of research, there might be an added value in considering each other's theoretical and methodological approaches. In education, physical activity and cognitive activity are typically treated as unrelated processes. By contrast, a more integrated approach is recommended for most effective health and learning outcomes. The current review in no way exhausts the existing literature but rather uses selective examples to draw conclusions and suggest a new instructional method in which physical activities are intermingled with classroom instruction. Results of recent studies confirm that integrated, task-relevant physical activities have paramount effects on learning. Future research is needed to shed light on the required frequency and duration of the classroom-based physical activity programmes, taking into account different age and target groups (including minority or low socioeconomic status children, children of typical development, and diagnosed with developmental disorders), and the feasibility of their implementation in "real-world" settings (including teachers' preferences, sustainable resources, construction of "user-friendly" manuals, and guidelines for teachers). Author Contributions ==================== MM and MR equally contributed to the completion of the manuscript and the development of the original concept (shared first authorship). MS, SL, AO, PC, and FP contributed to the development of the original concept, supervised drafting of the manuscript, and reviewed it for important intellectual content. 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 research was partly funded by the Research Excellence Initiative grant from the Erasmus University Rotterdam awarded to the Educational Psychology section. We would like to thank Assoc. Prof. Caterina Pesce and Amie Wallman-Jone for the fruitful feedback and proofreading the manuscript respectively. [^1]: Edited by: Meryem Yilmaz Soylu, University of Nebraska-Lincoln, United States [^2]: Reviewed by: Daniel Lloyd Eaves, Teesside University, United Kingdom; Gökhan Dağhan, Hacettepe University, Turkey; Paul Ayres, University of New South Wales, Australia [^3]: ^†^Shared first authorship [^4]: This article was submitted to Educational Psychology, a section of the journal Frontiers in Psychology

Background ========== Physical exercises are prescribed in some institutions for patients that wear braces for adolescent idiopathic scoliosis \[[@B1]-[@B3]\], but there is no scientific evidence for this practice \[[@B4],[@B5]\]. The possible useful effects of an exercise regimen for a patient with adolescent idiopathic scoliosis wearing a corrective brace have been divided theoretically into the two categories of general and specific effects \[[@B6],[@B3],[@B11]\]. The former includes all beneficial modifications (obtained through the activation of muscles, the stimulation of ventilatory exchanges and psychological help) that physical activity provides the patient, while reducing the impairments and disabilities induced by wearing the orthosis. Let us look at these singly: 1\. Activation of muscles In braced patients, it is normally thought that the supporting action of trunk muscles is reduced \[[@B9],[@B10],[@B12]-[@B15]\]. Exercises are proposed to: a\. avoid this effect, that could be more pronounced in adolescent patients using braces all day long; b\. have the effect of stabilizing the spine when the brace is removed. 2\. Stimulation of ventilatory exchanges Vital capacity and VO2 max are often reduced in patients whose scoliotic curvatures exceed 30 degrees Cobb \[[@B16],[@B17]\]. The VO2 max is usually reduced beyond a level that might be explained by a decrease in vital capacity alone. The reduction has been attributed to lack of physical exercise \[[@B18],[@B19]\]. Exercises are proposed to: a\. increase vital capacity; b\. train the patient so that both the cardiovascular and the musculoskeletal systems have an increased capacity to use oxygen; c\. improve respiratory ability from a neuromuscular standpoint. 3\. Psychological help Braces may induce a \"negative body image\" \[[@B20]\] in a growing child, that could, in turn, lead to an immature personality in adulthood. Exercises are proposed to reduce the psychological disability induced by the brace (which is not as great as that induced by the impairment itself), which may include a feeling of inferiority of the patient with respect to his/her friends. Conversely, the category of specific effects relates to the pressure that braces exert on the spine through the soft tissues. Specific exercises have been proposed by Stagnara \[[@B3],[@B2]\] and many others \[[@B6],[@B7],[@B13],[@B10],[@B12],[@B13],[@B15],[@B21],[@B22]\] with the rationale of increasing the corrective forces applied by the brace -- somehow using movements as \"dynamic tools\" to amplify the corrective \"static\" forces applied by the orthosis. Obviously such movements are instantaneous, but the immobilization of the ribs and spine that they induce (the former having modelling, and the latter, derotatory and deflective effects), could, in time and with repetition, play a major role in bringing about a positive effect of the brace \[[@B3],[@B2]\]. Moreover, it is necessary to consider the following assumptions: 1\. According to many etiological theories, the central nervous system could play an important role in the origin of the deformity \[[@B23],[@B24]\]; 2\. It has been supposed that the soft tissues are not able to withstand passively the forces that should be applied by a brace in order to correct a scoliosis \[[@B25]\]; 3\. The brace corrective effect with respect to Cobb angle reduction is strongly correlated with the pressure exerted by the pads \[[@B26]\]; 4\. The strap tension should be set as high as possible for right thoracic curvatures \[[@B27]\]; 5\. Muscular contraction has been supposed to play a major role in the effect of braces \[[@B25],[@B28]\]. Esthesio- and proprioceptive stimuli have also been considered important in bringing about a rearrangement of the postural system \[[@B1],[@B23]\]; 6\. The electronic pressure sensor may be the best way to quantify the effectiveness of brace pad placement \[[@B29]\]. On the basis of these hypotheses, the forces applied during specific exercises and physical activities are important not only from a biomechanical standpoint, but also from a neurological perspective that aims to help the patient to develop a new spinal behavior. Thus, exercises that dually act on the spine -- both to increase the forces of the brace and to drive vertebrae in the direction of the correction by means of the \"escape from the pad\" movement \[[@B3],[@B30]\] -- could be extremely useful. Focusing on the specific effect of physical activity, the aims of our study were: 1\. to ascertain the presence of muscle action and, if present, determine its importance; 2\. to evaluate the theoretical possibility of developing a best \"corrective exercise\" in a fiberglass brace (in terms of forces exerted during the movement); and 3\. to compare the exercises most commonly prescribed to achieve specific effects, such as kyphotization, rotation \[[@B9],[@B10],[@B3]\] and \"escape from the pad\" \[[@B3],[@B30]\]. Materials and methods ===================== Subjects -------- We examined 17 patients consecutively admitted to Don Gnocchi Foundation Orthopaedic Department to undergo treatment for idiopathic scoliosis. This treatment was based on two or three (depending on the degrees Cobb of the curvature) fiberglass non-removable braces (modified Risser type) \[[@B1]\]. The inclusion criteria were: 1\. a diagnosis of adolescent idiopathic scoliosis; 2\. the presence of a right thoracic or thoracolumbar curvature; 3\. no significant pain during movements in brace; 4\. second or third Risser fiberglass brace (so that, before the start of our evaluation, the patients involved in the study were accustomed to movements inside the brace); and 5\. good execution of the proposed movements, as judged by a single, well-qualified physical therapist (MR). Fourteen female and three male patients were included in the study. Age ranged from 12 to 17 years; curvatures from 36 to 52 degrees Cobb; 13 of 17 patients had double major curvatures and all patients had classic idiopathic right thoracic, left lumbar curvature patterns. Instrument ---------- A pressure distribution measurement system (F-Scan System™- Tekscan, Inc. Boston, MA 02210 USA) was utilized which detects pressure by means of a sensor and handles data through a dedicated personal computer and software. The sensor is a rectangular plastic sheet 7.62 cm wide, 20.32 cm long and 0.1016 mm thick. It is a resistive-based device made up of a conductive paste interposed between two insulating sheets. The paste is applied so as to form a 6 × 16 cell grid on one side and a 6 × 4 cell grid on the other. The cells are connected to one another by conductive wires which guide the variation of resistance from the pressure applied to the device for measurement. The sensor is connected to a personal computer by an A/D unit able to control the scanning process. Important characteristics of the software include its ease of use and the capability of viewing the acquisition process in real time. The repeatability of the F-Scan system has been verified in other applications \[[@B31],[@B32]\]. We have already studied the capacity of the system to evaluate the forces applied by Fiberglass orthoses, and revealed how certain parameters influence data acquisition \[[@B32]\]. The most important of these are: 1\. the size of the forces to be measured; 2\. the temperature during measurements; 3\. the time taken to perform measurements; 4\. the interval between two consecutive measurements; and 5\. sensor wear and tear. The foregoing perturbations can be limited by calibrating the sensor carefully before the start of data acquisition. For comparison with the data obtained *in vivo*, a measurement is required that must be calculated when all the parameters are known and controlled \[[@B32]\]. Brace ----- The Risser cast has a long tradition in scoliosis treatment \[[@B33],[@B34]\] and is still used in some scoliosis centers around the world\[[@B1]\]. The brace utilized in this study was modified by using a different material (fiberglass) to avoid the last (third) phase of manufacturing \[[@B1],[@B35]\]. The patient lays supine on a modified Risser bed, in traction. The following corrective actions are performed on the patient\'s trunk: • passive deflection; • derotation with a tissue band; and • direct push on the rib hump by two pads that exert a strong pressure through a screw mechanism. The brace is modelled directly on the patient\'s trunk with fiberglass bands. Two windows are cut to allow for expansion of the concave aspect of the scoliotic deformity once the material hardens. Movements evaluated ------------------- We evaluated those movements which can be performed in a seated position and which were thought to be capable of varying the corrective forces exerted on the scoliotic curvatures and rib humps by the brace \[[@B9],[@B10],[@B3]\]. The movements could either enhance or reduce the pressure of the pad. The following movements were tested: 1\. kyphotization: a generalized flexion of the whole dorsal spine; 2\. rotation against the pad of the brace: a movement of the apical aspect of the curvature against the pad of the brace; 3\. \"escape from the pad\" movement \[[@B3],[@B30]\]: a movement of the apical aspect of the curvature in the same direction as that of the passive corrective forces (*i.e*., away from the pad of the brace). The movement is proposed as a means of obtaining an active alignment of the spine. The way the movements were performed are detailed below. ### Kyphotization in sitting position (Figure [1](#F1){ref-type="fig"}) {#F1} a\. the patient is seated; b\. the arms and elbows are flexed at 90°; c\. the elbows are pushed forward. ### On all four kyphotization (Figure [2](#F2){ref-type="fig"}) {#F2} a\. the patient is positioned on all fours; b\. trunk is flexed. ### Rotation (simple) (Figure [3](#F3){ref-type="fig"}) {#F3} a\. the patient is seated; b\. the trunk is rotated against the pad. ### Rotation against a wall (Figure [4](#F4){ref-type="fig"}) {#F4} a\. the patient is seated near a wall; b\. the trunk is rotated against the pad. ### \"Escape from the pad\" (Figure [5](#F5){ref-type="fig"}) {#F5} a\. the patient is seated; b\. the rib hump is detached as far as possible from the pad. ### Reference positions For reference, we also evaluated the pressure values obtained at rest in the seated, supine and \"on all fours\" positions. Methodology ----------- During the making of the fiberglass brace, the sensor is planted between the right (thoracic or thoracolumbar) rib hump and the pad, almost parallel to the spine. Twenty-four hours after the brace is made, the physical therapist reminds the patient of the movements required and watches to see that they are executed correctly. At this point, the pressure data is collected. At the end of the data acquisition process the sensor is removed. Each recording lasts ten seconds and has a frequency of five frames per second. Data was collected following the sequence outlined below: 1\. neutral position; 2\. kyphotization; 3\. simple rotation; 4\. rotation against the wall; 5\. \"escape from the pad\". Analysis of data ---------------- The results (pressure values and area stimulated) refer to the average of the 10 frames in which the maximum pressure on the sensor was obtained. The stimulated area was calculated by multiplying N, *i.e*., the number of activated cells (those whose value is over zero), by β = 1.6129 (the area in cm2 of a single cell). The pressure value corresponds to the average of the measurements obtained over the whole stimulated area of the sensor. The results refer to the full sensor area and data were analyzed using Excel 5.0, Statgraphics Plus 3.1 and Matlab 5.0 for Windows. Statistical analysis was nonparametric, due to the non-Gaussian distribution of the data (Shapiro-Wilks W test). Significance was reached at p \< 0.05. The tests used for statistical analysis were the Wilcoxon \[[@B36]\] test, the Friedman \[[@B36]\] test for repeated measures and a multiple comparison procedure \[[@B37]\]. Results ======= Corrective pressure: absolute values ------------------------------------ In the reference positions, the average pressure values applied by the fiberglass brace itself were as follows: 1\. seated: 264.3 g/cm2 (range 191--436); 2\. supine: 262.5 g/cm2 (range 142--367); 3\. on all fours: 267.3 g/cm2 (range 186--381). Given that the total surface area of the stimulated portion of the sensor averaged 105 cm2, it can be postulated that, without exerting any additional force through movements, there was an average overall force of 28 kg on the rib hump. Through exercises, the force on the rib hump in the sitting position can be intensified as follows: 1\. kyphotization: 47 kg; 2\. simple rotation: 39 kg; 3\. rotation against the wall: 40 kg; 4\. \"escape from the pad\" movement: 32 kg. The average pressure values during kyphotization in different positions are detailed below: 1\. seated: 420 g/cm2 (range 303--958), increase in pressure of 58.9%; 2\. supine: 423 g/cm2 (range 282--899), increase in pressure of 61.1%; 3\. on all fours: 431 g/cm2 (range 283--882), increase in pressure of 61.2%. With reference to the other exercises, the pressure values were lower: 1\. simple rotation: 343 g/cm2 (range 238--809), increase in pressure of 29.8%; 2\. rotation against the wall: 332 g/cm2 (range 222--797), increase in pressure of 25.6%; 3\. \"escape from the pad\" movement: 309 g/cm2 (range 205--553), increase in pressure of 16.9%. Pressure -------- In the sitting position significant overall differences were found between exercises. When the sensor was divided in the latero-lateral direction, no differences emerged between exercises with the exception of kyphotization. When the sensor was divided in the cranio-caudal direction, differences emerged mainly in the proximal part of the pad. The central part registered increased pressure only during kyphotization, while no statistical differences emerged in relation to the distal part. In the sitting position, only the \"escape from the pad\" movement exercise failed to produce differences in relation to the reference position, while the biggest variations were found during kyphotization. In the \"on all fours\" position, kyphotization nearly always increased pressure values on the pad; only in the proximal area did we fail to observe this difference In the supine position there was an overall increase in pressure. When the sensor was divided in the cranio-caudal direction, an increase in pressure during the exercise was found in the central area, but no differences emerged in relation to the latero-lateral direction. Area ---- No statistically significant differences emerged between the area values during different exercises, with the exception of the paracostal and central values in the sitting position. (Submitted to the Friedmann test, it was not possible to ascertain which exercise might be able to determine such differences.) Reference positions ------------------- We found no statistically significant difference, in relation either to pressure values or area, when comparing the reference positions. This was also the case when the sensor was divided into different regions. Discussion ========== Reference positions ------------------- The absence of differences in pressure values applied by the fiberglass brace on the rib hump in different positions is interesting. In fact, we could interpret the three reference positions in relation to the forces which act passively on the brace-body-spine system. In the sitting position, the action of the force of gravity compresses the spine from above, increasing the kyphosis and pushing the posterior part of the trunk toward the pad. In the supine position, the brace is compressed between the body and the ground surface, while in the\"on all fours\" position there is no such compression. The differences among these forces did not influence the force administered by the brace itself. Different exercises ------------------- All exercises done to increase the forces between the rib hump and the brace increased the pressure significantly as expected. The increase in pressure reached a doubling in many cases. There was no variation in the area stimulated, which means that the area depends on something different from exercise that is assumed to be determined during the making of the brace (presumably the shape of the body and the pad, and/or the interaction between them). In terms of pressure exerted, the type of exercise done was more important than the position or way in which the exercise was performed. All kyphotization movements were more effective in terms of the amount of pressure exerted than the rotation movements. The more pronounced effect of the kyphotization as compared to the rotation exercises was surprising, as rotation was thought to be a more specific movement against the pad. Our results can presumably be explained by the strength of the muscles involved (flexors versus rotators of the spine) and by the movement itself (easier in sagittal than in the horizontal plane). No statistically significant difference was found between the neutral reference position and the \"escape from the pad\" exercise -- a finding which was most peculiar and completely discordant with the supposed aim of the exercise \[[@B3]\]. Moreover, the absolute values of pressure were on average found to be higher in the \"escape\" position than in reference position. A partial explanation for this unexpected phenomenon is that the exercises were performed the day after the making of the second/third fiberglass brace of the sequence -- a stage at which it is not easy to detach the skin from the pad. Presumably, the same study performed at a different moment in time would produce different results. Even though it was not found to be useful in term of forces, the \"escape from the pad\" movement may nevertheless be a helpful way of indicating to the central nervous system the direction in which the spine should go in order to correct the deformity. The described exercises must be considered only with respect to wearing braces like the one proposed in this paper (with a direct push on the rib hump). While the three-dimensional combination of planar movements inside the brace may produce positive results, kyphotization without the constraint of the brace in the frontal and horizontal planes could result in a rib hump increase. Different positions during kyphotization ---------------------------------------- Kyphotization was the only exercise analyzed in all positions. It always revealed an increase in pressure -- a finding which also emerged when considering the latero-lateral division of the sensor. The analysis of the cranio-caudal division was interesting, as differences were found in the behavior of the spine during the same exercise. There appears to be, in the sitting position, an extension of the distal part of the spine with a correspondent increase of pressure on only the proximal part of the sensor, while in the \"on all fours\" position the phenomenon is reversed, with an increase in pressure evidenced in only the distal part of the sensor. In both of these situations, the brace lacked the external support provided by the ground in the supine position; indeed in supine, kyphotization increased pressures only in the central part of the pad. Rotation and \"escape from the pad\" ------------------------------------ There were no differences between the two types of rotation exercises; the presence or absence of the support of the wall was not significant, nor was the behavior of the spine with regard to the division of the sensor into different areas as previously described. Even though no significant differences emerged, an analysis of the behavior of the forces on different parts of the pad during the \"escape from the pad\" movement suggested that the action produced here was simply an extension of the spine, with a decrease of pressure in the distal region of the sensor. It is possible that the patient was unable to follow and extend the action of the brace actively while constrained inside the fiberglass brace utilized; a different result might have emerged from a study utilizing a different type of brace, such as the Chêneau 2000 brace \[[@B38]-[@B40]\] that incorporates empty chambers to allow for greater spinal mobility into targeted regions. Although the brace used in this study is thought to be comparable in its shaping action to braces built in the Chêneau 2000 mode, the relative ability of the patient to actively adopt the corrective posture within the brace can perhaps best be assessed by the utilization of the electronic pressor sensor method. This study was performed in 1996 and has not been published previously. Although the Risser braces are no longer in use by the authors, new concepts of treatment have since been developed \[[@B41]\] that continue to apply forces on the rib humps in different ways. Through the indirect measurement of forces applied to the spine, the utilization of the electronic pressure sensor method presents an opportunity to assess the effectiveness of new treatments. Clinical and future research implications ----------------------------------------- We recognize that today there is no standard of bracing for scoliosis and that different schools are competing and comparing notes. The Chêneau and Rigo \[[@B38]-[@B40],[@B42]\], Risser or Lyon \[[@B1],[@B2]\] and SPoRT \[[@B41]\] schools propose different concepts of correction. The exercises investigated herein are rooted in all three groups and follow their theoretical approaches. The results of this study indicate that therapeutic exercises can increase the forces that are produced by the brace and may thereby help to increase the corrective action of the orthosis. This study was performed in the first 24 hours of bracing, and the results could change in the long term (such being one of the reasons why braces are renewed). Future research could address the variations of these effects over time. Exercises and exercise positions adopted can be selected depending on the desired action on the spine. On the basis of our experience and on the minor pressure differences found, we suggest varying the stimuli on the trunk of the patient. Varying the truncal stimuli would be expected to produce not only a better biomechanical action on the spine, but also a better neurological stimulation of the brain to elicit motor learning. The latter consideration may not apply in cases of severe rib hump or when a greater aesthetic improvement is required, as the reduction of the rib hump is primarily a mechanical matter. To our knowledge, this is the first study showing the effects of physical exercises in orthosis on idiopathic scoliosis. Braces are effective in idiopathic scoliosis treatment \[[@B4],[@B43]-[@B45]\], and thus the search for the most useful way of helping patients when they wear braces must be intensified. Physical exercises could serve not only to reduce the secondary effects of braces (muscular, respiratory and psychological impairment) but also in an attempt to improve the action of the orthosis itself \[[@B3],[@B46],[@B47]\]. Future research could address the same topic in classic plastic braces or in traditional casts, and in cases of less severe curvatures in which applied forces may be less significant. Our results suggest that exercises may prove helpful in these cases as well. Conclusion ========== We are able to conclude that: 1\. in static conditions, the position of the patient does not interfere with the forces exerted by the fiberglass brace; 2\. all exercises produce a significant increase in the mechanical forces exerted at rest by the fiberglass brace, with the exception of the\"escape from the pad\" movement, whose action should be exactly the opposite; 3\. the \"escape from the pad\" movement inside the fiberglass brace used is not so; it was not possible to demonstrate any decrease of pressure on the pad; 4\. although the \"escape from the pad\" exercise does not represent an actual mechanical \"escape from the pad\" movement inside of the fiberglass brace, this is apparently how it is perceived mentally, as evidenced by the extension movement observed; 5\. the strongest mechanical forces are produced by kyphotization exercises; 6\. rotation exercises are mechanically less efficient than kyphotization; and 7\. through kyphotization exercises it is possible to select different localized actions on the spine, depending on the patient\'s requirements. The method we used was found to be a useful instrument for evaluating forces exerted during exercises performed in the fiberglass brace. Although a common aim during exercises of this type is to obtain maximum corrective forces, it is also fundamentally important to remember that physical activity plays an important role in the prevention and reduction of impairments and disabilities -- a role that extends beyond the mere exertion of mechanical forces. For this reason, it is important for patients to exercise in their brace, to continue practicing sports and to be encouraged to continue exercising, in and outside of school. Since this study, kyphotization exercises have become a welcome addition to the in-brace SEAS (Scientific Exercises Approach to Scoliosis) protocol, while the already established recommendation of sports activities in and out of the brace has been maintained. Acknowledgements ================ We express our gratitude to the prescribing physician and to those who assisted in the Familaid (counselling for patient and family) phase.

Background ========== Human parvovirus B19 is considered an important trigger of aplastic crisis in patients with chronic congenital hemolytic disorders. Case report =========== We describe the case of a young adult known with hereditary microspherocytosis who presented with fever, marked agitation and confusion and a slight left hemiparesis. The cerebral imaging revealed multiple giant cystic Virchow-Robin spaces, but no other abnormalities. The complete blood counts showed pancytopenia and the peripheral blood smear revealed reticulocytopenia, confirming the diagnosis of aplastic crisis. Human parvovirus B19 infection was proven by the detection of serum DNA using PCR technique. After the initiation of packed red cells transfusion a favorable outcome was seen and his neurologic symptoms fully remitted. Conclusion ========== Direct invasion of human parvovirus B19 may induce both aplastic crisis and acute encephalopathy in patients with hereditary spherocytosis. Human parvovirus B19 may induce aplastic crisis but also acute encephalopathy through direct invasion, although in our case the transient neurological symptoms were rather an effect of hypoxia.

Abbreviations used in this paper: α1-AT, α1-antitrypsin; COPII, coat protein II; ERGIC, ER Golgi intermediate compartment; MCFD2, multiple coagulation factor deficiency protein 2; MEF, mouse embryonic fibroblast; PCA, protein fragment complementation assay. Introduction ============ The lumen of the ER provides a unique oxidizing environment for protein folding and modification. One third of all newly synthesized proteins are translocated into the ER and processed by multiple ER resident enzymes. An elaborate quality-control system monitors the conformational state of the nascent proteins and retains them in the ER during the folding process ([@bib10]). Upon correct folding, proteins are exported from the ER in coat protein II (COPII) vesicles ([@bib13]). Although transmembrane proteins can directly interact with the cytosolic COPII coat ([@bib6]), some soluble luminal proteins require cargo receptors for their selective recruitment into COPII vesicles ([@bib7]; [@bib5]). The best characterized mammalian cargo receptor is the oligomeric type I membrane lectin ERGIC-53 ([@bib27]). ERGIC-53 cycles between the ER and ER Golgi intermediate compartment (ERGIC), captures soluble glycoproteins in the lumen of the ER, and binds COPII by means of a dihydrophobic ER export motif in its cytosolic tail ([@bib12]; [@bib3]). As a cargo receptor, ERGIC-53 mediates ER export of coagulation factors V and VIII and cathepsin C and Z ([@bib17]; [@bib30]; [@bib2]). Furthermore, ERGIC-53 was recently shown to assist the assembly of IgM polymers in the ER ([@bib1]). With only few ERGIC-53 cargo proteins identified, however, it is not known whether receptor-mediated ER export is the predominant mechanism for the transport of soluble proteins. A major problem in identifying cargo--receptor complexes is their transient nature in a unique ionic and oxidizing environment. Further, existing proteomic techniques for identifying protein--protein interactions (e.g., affinity purification/mass spectrometry or the yeast two-hybrid system) are not easily applicable or are not appropriate to the study of membrane proteins. However, protein fragment complementation assays (PCAs) do allow for detection of both transient and dynamic protein--protein interactions in intact living cells, including those for membrane proteins ([@bib24]; [@bib15]). A PCA based on a YFP variant (citrine) has been demonstrated to be applicable to protein--protein interactions inside the lumen of the secretory pathway ([@bib19]). In the YFP PCA, nonfluorescent N- and C-terminal fragments of YFP (YFP1 and YFP2) are individually fused to the coding sequences of two separate proteins and expressed in mammalian cells. If the two fusion proteins interact, the fragments of YFP are brought into proximity, permitting folding and reconstitution of fluorescent YFP in vivo. Using this specific and sensitive assay, we have demonstrated that the oligomerization of ERGIC-53 and its interactions with multiple coagulation factor deficiency protein 2 (MCFD2) and cathepsin C and Z were readily visible in living cells. Here, we describe a YFP PCA--based cDNA library screening strategy for the identification of novel ERGIC-53 cargo proteins. The screening of a human adult liver cDNA library identified α1-antitrypsin (α1-AT) as novel interaction partner of ERGIC-53 and our validation studies establish ERGIC-53 as a transport receptor of α1-AT. Results and discussion ====================== Based on a general PCA cDNA library screening strategy ([@bib22],[@bib23]), we developed a screening approach designed to identify proteins that bind to cargo receptors in the ER lumen. ERGIC-53 was used as bait and was N-terminally tagged with YFP2, which localizes YFP2 to the luminal side of the membrane and allows screening for protein--protein interactions inside the lumen of the ER. Prey proteins were C-terminally tagged with YFP1 and expressed from a cDNA-YFP1 fusion library ([Fig. 1](#fig1){ref-type="fig"}).Figure 1.**Strategy of the ERGIC-53 cargo hunt based on YFP PCA.** (A) Schematic representation of the YFP PCA approach to screen for ERGIC-53 cargo proteins in the lumen of the secretory pathway. YFP PCA is based on the reconstitution of functional YFP from two nonfluorescent fragments that are brought into close proximity by two interacting proteins. YFP2 was fused to the ERGIC-53 bait protein, whereas YFP1 was fused to a human liver cDNA library to express YFP1-tagged prey proteins. (B) The YFP2--ERGIC-53 bait protein was expressed from the kanamycin-resistant pCMV vector, whereas the cDNA-YFP1 library was constructed in the ampicillin-resistant pcDNA3 vector. (C) Flow chart of the screening strategy. This fusion orientation assures that library-encoded secretory and membrane proteins reach the lumen of the ER because N-terminal signal sequences are not perturbed. However, we do not expect to identify known ERGIC-53 interaction partners with this cDNA-YFP1 library because of specific characteristics of these proteins or because of the specific orientation of the prey YFP1 toward cytosol. For example, C-terminal tagging cathepsin Z and C impedes the interaction with ERGIC-53 ([@bib19]), whereas MCFD2-YFP1 is unlikely to be identified because its interaction with ERGIC-53 strictly depends on the full-length MCFD2 protein ([@bib21]). Further, the interaction of ERGIC-53 with itself cannot be detected because the oligomerization of ERGIC-53 requires its transmembrane domain ([@bib18]), which localizes the YFP1 of all oligomerization-competent YFP1--ERGIC-53 prey proteins to the cytosol. This compartment-specific topological requirement is a unique feature of the PCA approach, providing information about the orientation of membrane-associated prey proteins. Our C-terminal tagging strategy requires that library inserts lack their stop codon to avoid termination of translation in front of the YFP1 sequence. To this end, we used cDNAs generated by randomly primed reverse transcription of mRNA, which enriches the 5′ ends. cDNA inserts were subcloned from a human adult liver cDNA library into mammalian pcDNA3 expression vectors containing YFP1 in all three reading frames. The resulting cDNA-YFP1 library contained ∼10^6^ clones and inserts ranged from ∼1 to ∼2.5 kb in size. As expected for a library generated from a secretory tissue like liver, cDNAs encoding secretory proteins were well represented (Fig. S1, available at <http://www.jcb.org/cgi/content/full/jcb.200709100/DC1>). The YFP PCA--based cDNA library screen was performed in COS-1 cells, which express the large T antigen and replicate plasmids containing the SV40 eukaryotic origin of replication. Because of this feature, cDNA-YFP1 library plasmids will be replicated upon transfection, which ensures sufficient expression of individual prey proteins and simplifies recovery and analysis of positive clones. Coexpression of YFP2--ERGIC-53 and the cDNA-YFP1 library resulted in the detection of 1.63% yellow fluorescent COS-1 cells, and YFP PCA can fully account for the detected signal because few positive cells were obtained upon expression of YFP2--ERGIC-53 or the cDNA-YFP1 library alone ([Fig. 2 A](#fig2){ref-type="fig"}).Figure 2.**Screening of the cDNA-YFP1 library for ERGIC-53 interaction partners.** (A) COS-1 cells were transfected with the indicated bait and prey constructs and analyzed by FACS as described in Materials and methods. Coexpression of YFP2--ERGIC-53 and the cDNA-YFP1 library resulted in the specific detection of 1.63% YFP positive (YFP^+^) cells. In a nonsaturating screen, ∼500 fluorescent cells were collected by FACS and total DNA was extracted and transformed into bacteria, which resulted in the recovery of several hundred prey clones. (B) 48 prey clones were randomly selected and plasmids were isolated and individually assayed by YFP PCA with YFP2--ERGIC-53 in COS-1 cells. Fluorometric analysis revealed that prey plasmids 17, 32, 33, and 44 (indicted by red boxes) were positive and reconstitute fluorescent YFP when expressed with YFP2--ERGIC-53. Bars represent fluorometric values of a single screening experiment. The threshold for a positive hit was set to 250 arbitrary fluorescence units, which corresponds to a ∼1.5-fold induction in YFP fluorescence in comparison to untransfected cells. (C) Sequence analyses identified α1-AT as a cDNA insert in all four positive prey plasmids. In the screen, YFP2--ERGIC-53 and the cDNA-YFP1 library were transfected at a ratio of 10:1 to reduce the number of library plasmids transfected per cell. This lowered the percentage of positive cells to 0.12% (unpublished data). Several hundred YFP-positive cells were collected by FACS and prey clones were recovered, 48 of which were individually reanalyzed by YFP PCA in COS-1 cells. Coexpression with YFP2--ERGIC-53 resulted in positive YFP signals for prey plasmids numbered 17, 32, 33, and 44 ([Fig. 2 B](#fig2){ref-type="fig"}). The recovery of only 4 positives out of 48 reanalyzed prey plasmids is not because of the unspecificity of our assay but rather the uptake of several prey plasmids per transfected cell. A YFP-positive cell contains the positive prey plasmid as well as many cotransfected negative plasmids, which will be subsequently also recovered (Fig. S2, available at <http://www.jcb.org/cgi/content/full/jcb.200709100/DC1>). DNA sequence analysis identified α1-AT as the corresponding cDNA insert in prey plasmids 17, 32, 33, and 44. Two types of α1-AT inserts were found, a 1294-bp and a 1297-bp variant. Both inserts were in frame and covered the complete coding sequence of α1-AT with the exception of the last two C-terminal amino acids ([Fig. 2 C](#fig2){ref-type="fig"}). The ERGIC-53--α1-AT interaction was validated by YFP PCA in HeLa cells using full-length α1-AT N-terminally tagged with YFP2 ([Fig. 3 A](#fig3){ref-type="fig"}).Figure 3.**ERGIC-53 captures α1-AT in a carbohydrate- and conformation-dependent manner.** The indicated YFP PCA constructs were expressed in HeLa cells for 24 (A) and 48 h (B). 20 μM Lactacystin (lact) and 200 μM kifunensine (kif) were applied where indicated for 24 h. YFP complementation was analyzed by fluorometric analysis of cell suspensions in microtiter plates and expression levels of the different constructs were probed by Western blotting (WB) using polyclonal antibodies against ERGIC-53 and α1-AT. Endogenous ERGIC-53 functions as an input control. Bars represent mean ± SD (*n* = 3). α1-AT is a liver glycoprotein of 52 kD carrying three N-linked glycans. After its synthesis in hepatocytes, α1-AT is secreted into the blood, where it acts as a serine protease inhibitor, mainly against neutrophil elastase. As one of the major glycoproteins synthesized in the liver, α1-AT is likely to be highly abundant in our nonnormalized library. When YFP2--ERGIC-53 and the cDNA-YFP1 library were expressed in COS-1 cells after siRNA-mediated silencing of α1-AT, a 30--50% reduction in YFP-positive cells was observed (unpublished data). Hence, α1-AT seems responsible for up to half of all observed YFP-positive cells in our screen. N-linked glycosylation of α1-AT is strictly required for its secretion ([@bib11]) and it was suggested that the N-glycans may serve as recognition sites for an ER-to-Golgi transport receptor ([@bib14]). Is ERGIC-53 an intracellular transport receptor of α1-AT? We first analyzed the carbohydrate specificity of the ERGIC-53--α1-AT interaction by performing a YFP PCA with the N156A mutant of ERGIC-53 and a triple N70,107,271Q mutant of α1-AT. ERGIC-53~N156A~ is unable to bind to N-linked oligosaccharides on glycoproteins ([@bib2]) and α1-AT~N70,107,271Q~ remains unglycosylated because of site-directed mutagenesis of all three N-glycosylation consensus sites. In comparison to their wild-type counterparts, ERGIC-53~N156A~ and α1-AT~N70,107,271Q~ showed a markedly reduced YFP PCA signal ([Fig. 3 A](#fig3){ref-type="fig"}). These findings suggest that ERGIC-53 binds α1-AT in a carbohydrate-dependent manner. We also analyzed N70Q, N107Q, and N271Q single mutants of α1-AT and found that mutagenesis of the second N-glycosylation site (N107Q) has the strongest effect on the ERGIC-53--α1-AT interaction ([Fig. 3 A](#fig3){ref-type="fig"}). This result is in full agreement with a previous study showing that the N107Q mutant of α1-AT has a significantly longer intracellular retention time than the other mutations ([@bib25]). Interestingly, all three N-glycosylation single site mutants have similar degradation rates and solubility ([@bib25]). Hence, the carbohydrate attached at N107 is particularly important for ER export. Next, we addressed the conformation specificity of the ERGIC-53--α1-AT interaction by analyzing two misfolded versions of α1-AT known as Z (α1-AT~Z~) and Hong Kong (α1-AT~HK~) mutants ([@bib28]). Although α1-AT~HK~ is soluble and degraded by ER-associated degradation involving the proteasome, α1-AT~Z~ aggregates in the ER and is subjected to proteasomal as well as lysosomal degradation ([@bib9]; [@bib29]). Strikingly, in our YFP PCA, neither α1-AT~HK~ nor α1-AT~Z~ bound to ERGIC-53 ([Fig. 3 B](#fig3){ref-type="fig"}). YFP2--α1-AT~Z~ was expressed at a similar intracellular level as YFP2--α1-AT~WT~ and was not affected by the two inhibitors of proteasomal degradation lactacystin and kifunensine. We cannot rule out the possibility that aggregation decreases the pool of soluble, ERGIC-53 binding-competent α1-AT~Z~, thereby lowering the YFP PCA signal. In the case of α1-AT~HK~, inhibition of proteasomal degradation by lactacystin and kifunensine clearly increased intracellular levels of YFP2--α1-AT~HK.~ Although the amount of YFP2--α1-AT~HK~ was thereby restored to that of the wild-type construct, the YFP PCA signal was not enhanced ([Fig. 3 B](#fig3){ref-type="fig"}). These data suggest that the ERGIC-53--α1-AT interaction is conformation-dependent and supports a function of ERGIC-53 in secondary quality control by capturing only native cargo proteins for ER export ([@bib10]; [@bib4]). To confirm that ERGIC-53 captures α1-AT for ER export, we analyzed the transport of endogenous α1-AT in HepG2 cells in which ERGIC-53 was knocked down by siRNA. Silencing of ERGIC-53 for 96 h reduced total ERGIC-53 protein levels to below 20% ([Fig. 4 A](#fig4){ref-type="fig"}).Figure 4.**α1-AT secretion is impaired in ERGIC-53 knockdown cells.** HepG2 cells were transiently transfected for 96 h with control and ERGIC-53--specific siRNA duplexes. (A) Western blotting using monoclonal anti--ERGIC-53 and anti--CLIMP-63 and polyclonal anti--α1-AT antibodies. ERGIC-53 is efficiently silenced, which results in intracellular accumulation of α1-AT. CLIMP-63, an ER-resident membrane protein, functions as an input control. (B) \[^35^S\]methionine pulse chase of endogenous α1-AT and albumin. HepG2 cells were labeled for 15 min with \[^35^S\]methionine and chased for the indicated times, and α1-AT and albumin were recovered by immunoprecipitation from cell lysates (intracellular) and conditioned medium (extracellular) using anti--α1-AT and anti-albumin antibodies, respectively. (C) The amount of intracellular and extracellular α1-AT and albumin was quantified by densitometric scanning of the band intensities and the secreted fraction of total protein \[extracellular/(intracellular + extracellular)\] was determined for each time point. Bars represent mean ± SD (*n* = 3). Results analyzed by paired *t* test: NS, P \> 0.05; \*\*, P \< 0.01. Intriguingly, the knock down of ERGIC-53 led to substantial steady-state accumulation of α1-AT inside the cell ([Fig. 4 A](#fig4){ref-type="fig"}). This effect was further analyzed by studying transport and secretion of α1-AT in pulse chase experiments using \[^35^S\]methionine. α1-AT is synthesized as high-mannose glycoprotein, undergoes complex glycosylation in the Golgi, and is subsequently secreted into the culture medium. In HepG2 cells transfected with control siRNA, high-mannose α1-AT was rapidly converted to its complex glycosylated form and about half of the protein was already secreted after a 30-min chase. In contrast, in ERGIC-53--silenced cells, α1-AT remained considerably longer in its high-mannose form and only ∼15% of the protein was secreted after 30 min ([Fig. 4, B and C](#fig4){ref-type="fig"}). The inefficient secretion of α1-AT is not caused by a general secretion defect because the secretion of endogenous albumin is unchanged ([Fig. 4, B and C](#fig4){ref-type="fig"}). Furthermore, previous studies in HeLa cells have already shown that neither depletion nor mislocalization of ERGIC-53 change the morphology of the secretory pathway or affect overall protein secretion ([@bib30]; [@bib20]). Hence, reduced levels of ERGIC-53 significantly delay the secretion of α1-AT in a specific manner. To determine if the remaining secretion of α1-AT is caused by residual ERGIC-53, we studied the transport of α1-AT transfected into mouse embryonic fibroblasts (MEFs) derived from ERGIC-53 knockout (−/−) and wild-type (+/+) mice. Pulse chase experiments revealed a half-time of secretion of ∼60 min in +/+ MEFs. ERGIC-53 knockout MEFs secreted in the same time only ∼25% of the newly synthesized α1-AT and showed significantly slower conversion of high-mannose to complex glycosylated α1-AT ([Fig. 5, A and B](#fig5){ref-type="fig"}).Figure 5.**ERGIC-53 is an intracellular transport receptor of α1-AT.** α1-AT was transfected into wild-type (+/+) and ERGIC-53 knockout (−/−) MEFs and expressed for 24 h. (A) \[^35^S\]methionine pulse chase of α1-AT and endogenous fibronectin. MEFs were labeled with \[^35^S\]methionine for 15 min and chased for the indicated times, and α1-AT and fibronectin were recovered by immunoprecipitation from cell lysates (intracellular) and conditioned medium (extracellular) using anti--α1-AT and anti-fibronectin antibodies, respectively. (B) The secreted fraction of α1-AT and fibronectin was quantified as described in [Fig. 4 C](#fig4){ref-type="fig"}. (C) α1-AT was coexpressed with human ERGIC-53 in ERGIC-53 knockout MEFs (MEF −/− plus ERGIC-53). Western blotting was performed using polyclonal antibodies against ERGIC-53, α1-AT, and CLIMP-63. (D) \[^35^S\]methionine pulse chase of α1-AT. (E) The secreted fraction of α1-AT after a 60-min chase was quantified by densitometric scanning and calculated as in [Fig. 4](#fig4){ref-type="fig"} \[extracellular/(intracellular + extracellular)\]. Bars represent mean ± SD (*n* = 3). Results analyzed by paired *t* test: NS, P \> 0.05; \*, P \< 0.05. A general secretion defect can again be excluded because ERGIC-53 knockout MEFs secreted endogenous fibronectin as efficiently as wild-type MEFs ([Fig. 5, A and B](#fig5){ref-type="fig"}). These results demonstrate that α1-AT secretion is significantly delayed without ERGIC-53 but that alternative, less efficient ER export pathways exist. If ERGIC-53 is a transport receptor of α1-AT, reintroduction of ERGIC-53 should correct the secretion defect of α1-AT in ERGIC-53 knockout cells. Indeed, when human ERGIC-53 was coexpressed with α1-AT, intracellular α1-AT was considerably reduced in −/− MEFs at a steady state ([Fig. 5 C](#fig5){ref-type="fig"}). Moreover, ERGIC-53 increased the secretion of α1-AT in −/− MEFs to the level of +/+ MEFs after a 1-h chase ([Fig. 5, D and E](#fig5){ref-type="fig"}). Hence, overexpression of ERGIC-53 in ERGIC-53 knockout MEFs can completely restore α1-AT secretion. These rescue experiments clearly establish ERGIC-53 as an intracellular transport receptor of α1-AT. Previous studies on ERGIC-53 function were mainly based on genetic analysis ([@bib17]), transport studies with a dominant-negative ER-retained mutant of ERGIC-53 ([@bib30]; [@bib2]), or the characterization of ERGIC-53--cargo interactions ([@bib2]; [@bib4]; [@bib19]; [@bib31]). The current study provides now direct evidence for active receptor-mediated ER export of a soluble secretory protein in mammalian cells. Furthermore, our data indicate that transport receptors render cargo transport faster and more efficient, but cargo transport is not entirely blocked in their absence. α1-AT may possess a second transport receptor or may exit the ER by bulk flow to a certain extent. The screening approach we described here could be equally applied using α1-AT as bait to identify novel cargo receptors. With α1-AT, we now identified an attractive model secretory protein to study the mechanism of receptor-mediated protein export from the ER in detail. In conclusion, this study not only identifies an intracellular transport receptor of α1-AT but also opens an unprecedented avenue to genome-wide screening for protein--protein interactions in the secretory pathway. The successful identification of a novel ERGIC-53 cargo protein by YFP PCA--based screening of a complex cDNA-YFP1 fusion library provides a general strategy to identify novel luminal protein complexes. With normalized libraries and optimized transfection conditions, saturating and genome-wide screens will be feasible in the near future. Further, YFP PCA has a promising potential for high-throughput screening of chemical and molecular chaperones ([@bib8]) that can rescue conformational defects of α1-AT mutants and render them secretion-competent by promoting their interaction with ERGIC-53. Materials and methods ===================== Antibodies and inhibitors ------------------------- The following antibodies were used: mouse mAbs against human ERGIC-53 (G1/93; Qbiogene), human CLIMP-63 (G1/296; Qbiogene) and GFP (Roche), rabbit polyclonal antibodies (pAbs) against ERGIC-53 ([@bib27]), human CLIMP-63, and horse fibronectin (provided by M. Chiquet, Friedrich Miescher Institute, Basel, Switzerland), goat pAb against human α1-AT (MP Biomedicals) and sheep pAb against human albumin (The Binding Site). Lactacystin and kifunensine were obtained from EMD. cDNA-YFP1 library construction ------------------------------ The cDNA-YFP1 library was constructed in the pcDNA3 vector (Invitrogen) into which the coding sequences for the (GGGGS)~2~ linker followed by YFP1 (amino acids 1--158 of YFP) was inserted. Note that all three reading frames of YFP1 were covered and that YFP1 contains the citrine mutation (Q69M; [@bib19]). Two SfiI restriction sites were introduced in front of the linker-YFP1 sequences by inserting a pair of annealed oligonucleotides. The pcDNA3(SfiI-linker-YFP1) vectors used for library construction are illustrated in Fig. S1. cDNA inserts were excised from a human adult liver cDNA-NubG library (Dualsystems Biotech) by SfiI restriction digestion and separated by agarose gel electrophoresis. The major cDNA insert fraction, ranging from ∼1 to ∼2.5 kb in size, was purified from the agarose gel and ligated via the SfiI restriction sites into the pcDNA3(SfiI-linker-YFP1) vectors. Ligation products were ethanol precipitated and electroporated into competent MC1061 bacteria. Transformed bacteria were grown overnight at 30°C on ampicillin-containing plates and harvested by scraping into liquid broth medium, and plasmids were isolated using the Plasmid Maxi kit (QIAGEN). YFP PCA plasmids ---------------- The YFP2--ERGIC-53 bait was generated by subcloning SS--YFP2--ERGIC-53 ([@bib19]) into the pCMV-Script vector (Stratagene). YFP fragment 2 contains amino acids 159--239 of YFP. YFP1-MCFD2 was constructed by inserting MCFD2 without its signal sequence into pcDNA3(SS-YFP1) ([@bib19]). YFP2--α1-AT~WT~ and YFP2--α1-AT~Z~ were generated by inserting the corresponding cDNAs, PCR-amplified without the endogenous signal sequences from pECEM(A1Pi) and pECEM(A1PiZ) (provided by M. Spiess, Biozentrum, University of Basel, Basel, Switzerland) into pcDNA3(SS-YFP2) ([@bib19]). YFP2--α1-AT~HK~, YFP2--α1-AT~N70Q~, YFP2--α1-AT~N107Q~, YFP2--α1-AT~N271Q~, and YFP2--α1-AT~N70,107,271Q~ were generated by QuickChange site-directed mutagenesis (Stratagene). Cloning of HA-MCFD2, YFP1-ERGIC-53~WT~, and YFP1-ERGIC-53~N156A~ has been described previously ([@bib19], [@bib20]). cDNA library screening ---------------------- COS-1 cells were grown in 100-mm dishes and transfected with 2.5 μg DNA and 7.5 μl FuGENE6 (Roche). pCMV(SS--YFP2--ERGIC-53) and the cDNA-YFP1 library were cotransfected at a ratio of 10:1. 48 h after transfection, cells were harvested in PBS containing 0.1% bovine serum albumin and 5 mM EDTA, and YFP-positive cells were sorted using a FACS Vantage SE (Becton Dickinson). Fluorescence was excited at a wavelength of 488 nm and recorded at 530/30 nm (YFP signal) and 575/26 nm. In a forward versus side scatter linear dot plot, the cell population of interest was defined as region R1 and gated into a 530/30 nm versus 575/26 nm logarithmic dot plot. Plotting of 530/30 nm versus 575/26 nm fluorescence resulted in autofluorescent cells lying in the diagonal of the plot, whereas cells expressing YFP were exclusively found in region R2 ([Fig. 2 A](#fig2){ref-type="fig"}). Cells in R2 were sorted into PBS and total DNA was isolated using the DNeasy Tissue kit (QIAGEN), ethanol precipitated, and transformed into XL-10 gold ultracompetent bacteria (Stratagene), which were then selected on ampicillin-containing plates. Note that transformants that take up genomic DNA or pCMV(SS--YFP2--ERGIC-53) cannot grow on ampicillin. Plasmids were isolated from liquid overnight cultures of single bacterial transformants using the Plasmid Mini kit (Sigma-Aldrich) and were individually cotransfected with pCMV(SS--YFP2--ERGIC-53) into COS-1 in 6-well plates. Fluorometric analysis of COS-1 cells was performed 48 h after transfection as described previously ([@bib19]) and plasmids resulting in a positive YFP signal were subjected to nucleotide sequence analyses. Cell culture and transfection ----------------------------- HeLa cells (CCL-2; American Type Culture Collection) and MEFs were grown in DME supplemented with 10% fetal bovine serum, 1× nonessential amino acids, and antibiotics. COS-1 cells (CRL-1650; American Type Culture Collection) were grown in DME supplemented with 10% fetal bovine serum and antibiotics. HepG2 cells (HB-8065; American Type Culture Collection) were grown in MEM supplemented with 10% fetal bovine serum and antibiotics. HeLa and COS-1 cells were transfected using FuGENE6 (Roche). HepG2 cells were transfected with control and ERGIC-53 siRNA using HiPerfect (QIAGEN) as described previously ([@bib20]). MEFs were transfected by electroporation using the Amaxa Nucleofector, program T-20, and MEF solution 2 (Amaxa Biosystems). α1-AT and ERGIC-53 were expressed from pECEM(A1Pi) and pECE(ERGIC-53) ([@bib26]), respectively. Immunoblotting -------------- Protein samples were prepared by boiling cell suspensions in protein sample buffer and were separated by SDS-PAGE, transferred to nitrocellulose membranes, immunoblotted with the indicated antibodies, and visualized by enhanced chemiluminescence (GE Healthcare). PCR screening ------------- DNA from heat-killed bacterial transformants was amplified with two primers that anneal onto the pcDNA3 vector backbone and amplify ∼1,150 and ∼650 bp fragments for pcDNA3(SS-YFP1-MCFD2) and pcDNA3(SS-HA-MCFD2), respectively. Preparation of wild-type and ERGIC-53 knockout MEFs --------------------------------------------------- MEFs were generated from embryos at 13.5 d postcoitus as described previously (A. [@bib16]) by mating heterozygous mice with an ERGIC-53 knockout allele and transformed with the SV40 large T antigen, and immortalized cell lines were established. \[^35^S\]methionine metabolic labeling -------------------------------------- Cells were deprived of [l]{.smallcaps}-methionine for 30 min, pulsed for 15 min with 100 μCi \[^35^S\]methionine (Perkin Elmer) and chased for the indicated times in culture medium containing 10 mM [l]{.smallcaps}-methionine. Cells were lysed in 1% Triton X-100, 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM CaCl~2~, and PMSF, and lysates were cleared by centrifugation at 100,000 *g* for 1 h. The chase medium was cleared from cell debris by centrifugation at 20,000 *g* for 5 min. Cleared samples were immunoprecipitated with the indicated antibodies, immunoprecipitates were separated by SDS-PAGE, and radiolabeled bands were imaged and quantified using a phosphorimager (Molecular Dynamics). Statistical analysis -------------------- Mean, standard deviation, and *t* test were calculated by using Excel 2002 (Microsoft). A two-tailed, paired *t* test was used to calculate the statistical significance between two indicated samples. P-values \>0.05 were considered not significant and p-values \<0.05 and \<0.01 were considered significant. Online supplemental material ---------------------------- Fig. S1 shows the assessment of the human adult liver cDNA-YFP1 library. Fig. S2 shows the assessment of plasmid transfection and recovery. Online supplemental material is available at <http://www.jcb.org/cgi/content/full/jcb.200709100/DC1>. Supplementary Material ====================== ###### Supplemental Material Index We thank Käthy Bucher for expert technical assistance and Verena Jäggin for professional assistance in FACS analysis and sorting. This work was supported by the University of Basel, the Swiss National Science Foundation, and the Roche Research Foundation.

1. Introduction {#sec1} =============== *Orostachys japonicus* (*O. japonicus*) is traditionally used as an inflammatory agent, antifebrile, homeostatic agent, and antidote and anticancer agent \[[@B1]\]. The methanol extract of*Orostachys japonicus*is thought to contain several different classes of phytochemicals, including triterpenes, sterols, and flavonoids \[[@B2]\]. Inflammation is caused by a variety of factors, including physical and chemical agents, the immune response, and tissue necrosis \[[@B3]\]. However, further studies on*O. japonicus* are required due to the lack of information on signaling pathways and physiological activity. Immune cells can recognize pathogen-associated molecules, such as the lipopolysaccharide (LPS) of Gram-negative bacteria, the peptidoglycan (PGN) of Gram-positive bacteria, and mannans of yeast cells through toll-like receptors (TLRs) expressed on the cell surface \[[@B4]\]. Humans have various pathogen-recognition receptors including TLRs, nucleotide-binding oligomerization domain- (NOD-) like receptors (NLRs), and retinoic acid-inducible gene-1- (RIG-1-) like receptors \[[@B5]--[@B7]\]. These receptors transduce signals to activate nuclear factor-*κ*B (NF-*κ*B), which subsequently drives the induction of several proinflammatory cytokines and chemokines \[[@B8]--[@B10]\]. TLRs are an integral component of the inflammation process. TLR2 and TLR4, along with their ligands, are best characterized in terms of innate responses to bacteria, including*Chlamydia*. TLR2 is involved in the recognition of a broad range of microbial products, and TLR4 is the signal-transducing receptor for LPS \[[@B11]\]. NOD1 and NOD2 are involved primarily in mediating antibacterial defenses \[[@B12]\]. NOD1 recognizes mainly Gram-negative bacteria, whereas NOD2 recognizes most Gram-positive and Gram-negative bacteria \[[@B13]\]. Inducible nitric oxide synthase (iNOS) is expressed widely in various cell types and is highly expressed in LPS-activated macrophages \[[@B14]\]. Expression of the iNOS gene in macrophages is regulated mainly at the transcriptional level. NF-*κ*B is a pivotal regulator of important immunoregulatory genes involved in immune and inflammatory responses, including iNOS \[[@B15]\]. Cyclooxygenase-2 (COX-2) is expressed in the presence of many proinflammatory mediators, including LPS, interlukin-1*β* (IL-1*β*), and tumor necrosis factor-*α* (TNF-*α*), through which high concentrations of prostaglandin E~2~ (PGE~2~) are produced \[[@B16]\]. iNOS and COX-2 expression are regulated by NF-*κ*B. NF-*κ*B is a transcription factor that regulates several genes, including iNOS, COX-2, IL-1*β*, IL-6, and TNF-*α*, which are important for immunity and LPS-induced inflammation \[[@B17]\]. NF-*κ*B is activated by phosphorylation of inhibitory *κ*B*α* (I*κ*B*α*) through activation of mitogen-activated protein kinases (MAPKs), such as c-Jun N-terminal kinase (JNK), p38, and extracellular signal-regulated kinase (ERK)1/2 \[[@B18], [@B19]\]. In the present study, we investigated the effects of*O. japonicus* on the expression of genes encoding pathogen-recognition receptors (TLR2, TLR4, NOD1, and NOD2) and proinflammatory factors (iNOS, COX-2, and cytokines) in LPS-stimulated PMA-differentiated THP-1 cells, as well as the NF-*κ*B and MAPK pathways. 2. Materials and Methods {#sec2} ======================== 2.1. Extraction of*O. japonicus* {#sec2.1} -------------------------------- *O. japonicus* (20 g) was extracted by overnight incubation at 60°C in 500 mL of 80% methanol. The solution was filtered through Whatman No. 1 filter paper and concentrated using a rotary evaporator (Buchi, Flawil, Switzerland). The concentrated extract was freeze-dried (EYELA, Tokyo, Japan) and stored at 4°C in a vacuum container until use. 2.2. Cell Culture {#sec2.2} ----------------- Human monocytic leukemia THP-1 cells were supplied by the Korean Cell Line Bank. Cells were cultured in RPMI 1640 medium (GIBCO, Grand Island, NY, USA) containing 10% fetal bovine serum and antibiotics. Cells were incubated at 37°C in a humidified atmosphere of 5% CO~2~ in 95% air. THP-1 cells were treated with 100 nM of phorbol myristate acetate (PMA, Sigma-Aldrich Co., St. Louis, MO, USA) for 72 h to induce differentiation into macrophages. After differentiation, nonattached cells were removed by aspiration and adherent macrophages were washed with RPMI 1640 medium three times and then incubated in cell culture medium at 37°C. 2.3. Cell Viability {#sec2.3} ------------------- Cell proliferation was measured with CellTiter 96 Aqueous One Solution (Promega, Madison, WI, USA). Cells were seeded at 1 × 10^4^ per well in 96-well plates and incubated with different concentrations of*O. japonicus* at 37°C for 24 h, respectively. Cell viability was determined using a colorimetric assay with PMS/MTS solution. The absorbance was determined at 490 nm with background subtraction at 650 nm. 2.4. Treatment with*O. japonicus* {#sec2.4} --------------------------------- THP-1 cells were pretreated for 2 h in serum-free medium with*O. japonicus* (0--25 *μ*g/mL) and then incubated with LPS (1 *μ*g/mL) for 4 h (for mRNA expression) and 20 h (for protein expression). At each time point, total RNA and protein were isolated from the cultured THP-1 cells. 2.5. RNA Extraction and Real-Time PCR {#sec2.5} ------------------------------------- Total RNA was purified from cultured cells using the TRIzol reagent following the manufacturer\'s protocol (Invitrogen, Carlsbad, CA, USA). First-strand cDNA synthesis was performed with 1 *μ*g of total RNA and it was transcribed to cDNA using a reverse transcription system with random hexamers (Promega) according to the manufacturer\'s protocol. The sequences for gene-specific primers were as follows: TLR2, 5′-TCTCCCATTTCCGTCTTTTT-3′ and 5′-GGTCTTGGTGTTCATTATCTTC-3′ (125 bp); TLR4, 5′-GAAGCTGGTGGCTGTGGA-3′ and 5′-TGATGTAGAACCCGCAAG-3′ (213 bp); NOD1, 5′-GTCACTGAGGTCCATCTGAAC-3′ and 5′-CATCCACTCCTGGAAGAACCT-3′ (363 bp); NOD2, 5′-CATGTGCTGCTACGTGTTCTC-3′ and 5′-CCTGCCACAATTGAAGAGGTG-3′ (226 bp); iNOS, 5′-TGGATGCAACCCCATTGTC-3′ and 5′-CCCGCTGCCCCAGTTT-3′ (59 bp); COX-2, 5′-CAAATCCTTGCTGTTCCCACCCAT-3′ and 5′-GTGCACTGTGTTTGGAGTGGGTTT-3′ (173 bp); IL-1*β*, 5′-TGATGGCTTATTACAGTGGCAATG-3′ and 5′-GTAGTGGTGGTCGGAGATTCG-3′ (140 bp); IL-6, 5′-GTGTTGCCTGCTGCCTTC-3′ and 5′-AGTGCCTCTTTGCTGCTTTC-3′ (194 bp); IL-8, 5′-GACATACTCCAAACCTTTCCAC-3′ and 5′-CTTCTCCACAACCCTCTGC-3′ (160 bp); TNF-*α*, 5′-ATCTTCTCGAACCCCGAGTG-3′ and 5′-GGGTTTGCTACAACATGGGC-3′ (51 bp); *β*-actin, 5′-GCGAGAAGATGACCCAGATC-3′ and 5′-GGATAGCACAGCCTGGATAG-3′ (77 bp). Real-time PCR was performed on the StepOneplus real-time PCR system with Power SYBR Green PCR Master Mix (Applied Biosystems, Foster, CA, USA). PCR was performed with 1 *μ*L of cDNA in 20 *μ*L reaction mixtures that consisted of 10 *μ*L Power SYBR Green PCR Master Mix, 2 *μ*L primers, and 7 *μ*L PCR-grade water. The reactions were performed with a denaturation step at 95°C for 10 min, followed by 40 cycles of 95°C for 15 sec and 60°C for 1 min. The crossing point of target genes with *β*-actin was calculated using the formula 2^−(target  gene−*β*-actin)^, and the relative amounts were quantified. 2.6. Western Blot Analysis {#sec2.6} -------------------------- Cells were collected and washed with cold PBS and then lysed using lysis buffer (20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM Na~2~EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM *β*-glycerophosphate, 1 mM Na~3~VO~4~, and 1 *μ*g/mL leupeptin) containing 1 mM PMSF (Cell Signaling Technology, Inc., Boston, MA, USA). The protein concentration was determined using a BCA protein assay according to the manufacturer\'s protocol. Protein (30 *μ*g) was fractionated by 12% SDS-PAGE and transferred by electrophoresis to nitrocellulose membranes. The membranes were blocked with 5% nonfat dry milk for 1 h at room temperature and then incubated overnight with antibodies against COX-2 (Santa Cruz Biotechnology, Santa Cruz, CA, USA), iNOS, NF-*κ*B p65, phospho-NF-*κ*B p65, I*κ*B*α*, phospho-I*κ*B*α*, JNK, phospho-JNK, p38 MAPK, phospho-p38 MAPK, ERK, phospho-ERK1/2 (Cell Signaling Technology), and *β*-actin (Sigma-Aldrich Co.), which were diluted to 1 : 1,000 with Tris-buffered saline containing 0.05% Tween 20 (TBS-T). After washing with TBS-T for 1 h, the membranes were incubated for 1 h at room temperature with horseradish peroxidase-conjugated secondary antibodies diluted to 1 : 2,500 in TBS-T. The membranes were subsequently washed with TBS-T for 1 h and proteins were detected using an Enhanced Chemiluminescence Kit (Thermo Scientific, Protein biology, IL, USA). Protein expression was analyzed using a Davinch-Chemi Chemiluminescence Imaging System (Davinch-K Co., Ltd., Seoul, Korea). 2.7. Statistical Analysis {#sec2.7} ------------------------- Values were expressed as means ± SD. Student\'s *t*-test was used to evaluate differences between the LPS-only treated samples and LPS plus*O. japonicus* treated samples. ^\*^ *P* \< 0.05 and ^\*\*^ *P* \< 0.01 were considered to indicate statistical significance. 3. Results {#sec3} ========== 3.1. *O. japonicus* Suppressed Cell Viability {#sec3.1} --------------------------------------------- The cytotoxicity of*O. japonicus* to THP-1 cells was examined by exposing cells to various concentrations of*O. japonicus* for 24 h. Cell viability was measured using the PMS/MTS assay.*O. japonicus* showed no toxicity at low concentrations ([Figure 1](#fig1){ref-type="fig"}). 3.2. *O. japonicus* Inhibited the LPS-Induced Expression of the TLR2, TLR4, NOD1, and NOD2 Genes {#sec3.2} ------------------------------------------------------------------------------------------------ We investigated the effect of*O. japonicus* on the expression of TLR2, TLR4, NOD1, and NOD2 in THP-1 cells. mRNA levels were analyzed using real-time PCR. LPS induced TLR4 and NOD2 expression compared to the control.*O. japonicus* suppressed LPS-induced TLR4 and NOD2 transcription (Figures [2(b)](#fig2){ref-type="fig"} and [2(d)](#fig2){ref-type="fig"}). In contrast, TLR2 and NOD1 mRNA levels were not significantly affected after LPS stimulation (Figures [2(a)](#fig2){ref-type="fig"} and [2(c)](#fig2){ref-type="fig"}). 3.3. *O. japonicus* Inhibited LPS-Induced iNOS and COX-2 Gene Expression {#sec3.3} ------------------------------------------------------------------------ We explored the effect of*O. japonicus* on inflammatory mediator expression. As shown in [Figure 3](#fig3){ref-type="fig"}, the response was dose-dependent, and the effect was significant at \>10 *μ*g/mL*O. japonicus*. We evaluated whether*O. japonicus* affected iNOS and COX-2 mRNA transcription and protein levels. mRNA and protein levels were measured using real-time PCR and western blot analysis, respectively. LPS induced iNOS and COX-2 gene expression compared to the control.*O. japonicus* suppressed LPS-induced iNOS and COX-2 transcription (Figures [3(a)](#fig3){ref-type="fig"} and [3(b)](#fig3){ref-type="fig"}) and translation (Figures [3(c)](#fig3){ref-type="fig"} and [3(d)](#fig3){ref-type="fig"}). 3.4. *O. japonicus* Inhibited LPS-Induced IL-1*β*, IL-6, IL-8, and TNF-*α* Expression {#sec3.4} ------------------------------------------------------------------------------------- We evaluated proinflammatory cytokine transcription in LPS-stimulated THP-1 cells. IL-1*β*, IL-6, IL-8, and TNF-*α* mRNA levels were evaluated using real-time PCR. LPS induced IL-1*β*, IL-6, IL-8, and TNF-*α* mRNA levels compared to the control.*O. japonicus* suppressed the LPS-induced IL-1*β* ([Figure 4(a)](#fig4){ref-type="fig"}), IL-6 ([Figure 4(b)](#fig4){ref-type="fig"}), IL-8 ([Figure 4(c)](#fig4){ref-type="fig"}), and TNF-*α* ([Figure 4(d)](#fig4){ref-type="fig"}) transcription. 3.5. *O. japonicus* Inhibited LPS-Induced NF-*κ*B Signaling and I*κ*B*α* Degradation {#sec3.5} ------------------------------------------------------------------------------------ We next investigated the inhibition of NF-*κ*B p65 activity. The effects of*O. japonicus* on I*κ*B*α* degradation were examined. NF-*κ*B p65 and I*κ*B*α* protein levels were determined by western blot analysis. Treatment with LPS alone increased the phosphorylation of NF-*κ*B p65 and I*κ*B*α*, while*O. japonicus* inhibited LPS-induced phosphorylation of NF-*κ*B p65 (Figures [5(a)](#fig5){ref-type="fig"} and [5(c)](#fig5){ref-type="fig"}) and I*κ*B*α* (Figures [5(b)](#fig5){ref-type="fig"} and [5(d)](#fig5){ref-type="fig"}) in a dose-dependent manner. 3.6. *O. japonicus* Inhibited LPS-Induced Phosphorylation of JNK, p38 MAPK and ERK1/2 {#sec3.6} ------------------------------------------------------------------------------------- To investigate the molecular mechanism underlying NF-*κ*B inhibition by*O. japonicus* in LPS-stimulated cells, we examined the effects of*O. japonicus*on the activation of JNK, p38 MAPK, and ERK1/2. MAPKs were assessed by western blot analysis. The phosphorylation of JNK, p38 MAPK, and ERK1/2 increased in cells treated with LPS alone.*O. japonicus* inhibited the phosphorylation of JNK (Figures [6(a)](#fig6){ref-type="fig"} and [6(d)](#fig6){ref-type="fig"}), p38 MAPK (Figures [6(b)](#fig6){ref-type="fig"} and [6(e)](#fig6){ref-type="fig"}), and ERK1/2 (Figures [6(c)](#fig6){ref-type="fig"} and [6(f)](#fig6){ref-type="fig"}) in LPS-stimulated cells in a dose-dependent manner. 4. Discussion {#sec4} ============= In the present study, we demonstrated that*O. japonicus*inhibits LPS-induced inflammatory signals in the PMA-differentiated human THP-1 cells. To evaluate the anti-inflammatory effects of*O. japonicus*, cells were pretreated with*O. japonicus* and then stimulated with LPS.*O. japonicus*exhibited dose-dependent cytotoxicity. Cell viability was not affected by 24 h treatment with less than 50 *μ*g/mL*O. japonicus*. Therefore, in the subsequent experiments 10--25 *μ*g/mL*O. japonicus*was used. TLRs are expressed predominantly in monocytes/macrophages and neutrophils \[[@B20]\]. TLR2 and TLR4 are transmembrane receptors that transmit LPS signals to intracellular components in signal transduction pathways and play important roles in the immune system. TKR4 is associated with the recognition of Gram-negative bacterial LPS, and TLR2 is considered the receptor for Gram-positive bacteria \[[@B21]\]. While*O. japonicus*was shown to express both TLR2 and TLR4 in the present study, in the presence of LPS, TLR2 expression was higher than that of TLR4. Our findings suggest that*O. japonicus* could elicit inflammatory reactions in PMA-activated THP-1 cells and contribute to inflammatory processes, a process mediated by TLR4 and, to a lesser extent, TLR2. The peptidoglycan subunits are recognized by the NOD family proteins, in particular by NOD2. NOD2 is an intracellular protein involved in innate immunity and is associated with chronic inflammatory diseases in humans \[[@B22]\]. We found that*O. japonicus*downregulated LPS-induced NOD2 expression, with no effect on NOD1. Taken together, these results suggest that*O. japonicus*mediated inflammatory reactions through TLR4 and NOD2. iNOS and COX-2 are important enzymes that mediate inflammatory processes and have been associated with the pathogenesis of certain types of human cancers, as well as inflammatory disorders \[[@B23]\]. Since proper regulation of iNOS and COX-2 expression could provide an effective and promising approach to treat inflammation related to diseases, much effort has been made to identify iNOS and COX-2 modulators, especially from plant sources \[[@B24], [@B25]\]. The present study demonstrates that*O. japonicus* extract can effectively suppress transcription and translational levels of iNOS and COX-2 expression. In addition,*O. japonicus* inhibited cytokine (IL-1, IL-6, IL-8, and TNF-*α*) expression in THP-1 cells stimulated by LPS. We evaluated transcription of proinflammatory cytokines, including IL-1*β*, IL-6, IL-8, and TNF-*α*, which play pivotal roles in the development and progression of inflammation, in LPS-stimulated cells. Our findings further suggest that*O. japonicus*possesses potent anti-inflammatory activity.*Harpagophytum procumbens*suppresses LPS-stimulated expression of iNOS and COX-2 in the fibroblast cell line L929 \[[@B26]\] and inhibits LPS-induced release of cytokines (IL-1*β*, IL-6, and TNF-*α*) and PGE~2~ from human monocytes \[[@B27]\].*Arisaema*cum bile extract inhibits the production of proinflammatory cytokines, including IL-1, IL-6, and TNF-*α*, and also inhibits iNOS and COX-2 expression, which are responsible for the production of NO and PGE~2~ in THP-1 cells \[[@B28]\]. The dichloromethane fraction from*O. japonicus* (OJD) inhibited NO production and TLR4, IL-1*β*, iNOS, and COX-2 expression in LPS-stimulated murine RAW 264.7 macrophage cells and inhibited LPS-induced NF-*κ*B p65 activation by suppressing I*κ*B*α* phosphorylation. However, phosphorylation of JNK and p38 MAPK was suppressed by OJD in a dose-dependent manner in LPS-stimulated cells \[[@B29]\]. Harpagoside inhibited LPS-stimulated NF-*κ*B promoter activity based on a gene reporter in RAW 264.7 cells, indicating that harpagoside interfered with the activation of gene transcription. These results suggest that the inhibition of iNOS and COX-2 expression by harpagoside suppresses NF-*κ*B activation \[[@B30]\]. Inhibition of the NF-*κ*B and MAPK pathways has been proposed to be a major mechanism underlying the attenuation of LPS-induced inflammatory cytokine production. NF-*κ*B plays a crucial role as the transcription factor in regulating many of the proinflammatory cytokine genes. LPS stimulation elicits a cascade leading to the activation of NF-*κ*B \[[@B31]\]. Cryptotanshinone suppressed LPS-induced production of IL-6 and TNF-*α* by inhibiting the activation of NF-*κ*B and MAPKs \[[@B32]\]. MAPKs, such as JNK, p38 MAPK, and ERK, mediate the signal transduction involved in cell proliferation, differentiation, transformation, survival, and death \[[@B33]\]. Expression of the NO, TNF-*α*, and IL-6 genes is dependent on activation of the transcription factor NF-*κ*B, which plays a crucial role in immune and inflammatory responses \[[@B34]\]. Activation of NF-*κ*B requires phosphorylation and proteolytic degradation of the inhibitory protein I*κ*B*α*, an endogenous inhibitor that binds to NF-*κ*B in the cytoplasm \[[@B35]\]. Upon stimulation with LPS, NF-*κ*B is activated and translocated into the nucleus as a result of phosphorylation-mediated degradation of I*κ*B*α* proteins in the lung of AL1 mice. However, pretreatment with a suitable drug could decrease the degradation of I*κ*B*α* and nuclear translocation of NF-*κ*B p65 and, therefore, downstream TNF-*α* and IL-6 production. However, MAPKs including JNK, p38 MAPK, and ERK play an important role in signal transduction pathways and regulate cytokine release \[[@B36]\]. In this study, MAPK was activated in LPS-induced THP-1 cells. However, drug treatment markedly suppressed LPS-induced phosphorylation of JNK, p38 MAPK, and ERK. The inhibition of IL-1*β*, IL-6, IL-8, and TNF-*α* production by*O. japonicus*occurs through pathways that converge on p38 MAPK and I*κ*B*α* activation since these kinases are known to regulate cytokine production in LPS-induced THP-1 cells.*O. japonicus*inhibits anti-inflammatory responses by inhibiting the degradation of I*κ*B*α* and nuclear translocation of NF-*κ*B and downstream cytokine expression. These results suggest that drug activity was dependent in part on the inhibition of MAPK and NF-*κ*B signaling pathways. 5. Conclusion {#sec5} ============= In this study, we found that treatment with*O. japonicus* blocked the activation of JNK, p38 MAPK, and ERK1/2, suggesting that*O. japonicus* suppresses LPS-induced NF-*κ*B translocation by inhibiting the activation of these intracellular signaling cascades and reducing iNOS and COX-2 expression. Conflict of Interests ===================== The authors declare that there is no conflict of interests regarding the publication of this paper. {#fig1} {#fig2} {#fig3} {#fig4} {#fig5} {#fig6} [^1]: Academic Editor: Alfredo Vannacci

Introduction {#sec1-2331216515569792} ============ Cochlear implants (CIs) are neural prostheses for individuals who have severe or profound hearing loss. These devices consist of an array of electrodes that are inserted into the inner ear to stimulate the auditory nerve. The auditory nerve has a tonotopic organization, such that the frequency information carried by the nerve fibers changes systematically and in a topographic manner, with high frequencies represented by neurons that synapse near the basal end of the spiral cochlea and low frequencies represented by neurons that synapse closer to the apex. As such, different electrodes along the array are intended to activate specific populations of auditory nerve cells. By presenting each electrode with information from a limited frequency range, the prosthesis ideally recreates a tonotopic representation of sound in the implanted ear. To adjust the dynamic range of the electrical current delivered to each electrode to the dynamic range of the acoustic input, audiologists have typically used psychophysical measures of the lowest detectable current level, called threshold, and the current level that is most comfortable for listening, called most comfortable level (MCL). Recently, however, some manufacturers have moved away from recommending threshold measurements, in part because the measurement procedure is considered too time consuming, and because thresholds are often quite uniform across the electrode array. One reason for the relatively uniform absolute thresholds across the electrode array is the broad electrical field produced by the monopolar (MP) stimulation used in most commercially available implant systems. The broad stimulation fields produce uniform thresholds but also lead to poor tonotopic representation (e.g., [@bibr5-2331216515569792]). The poor tonotopic representation is thought to underlie at least some of the difficulties faced by CI users in understanding speech, especially in noisy environments (e.g., [@bibr7-2331216515569792]; [@bibr12-2331216515569792]; [@bibr15-2331216515569792]). In contrast to the single active electrode utilized for MP stimulation, focused stimulation can be achieved with CIs by manipulating the spatial arrangement of current delivery across the electrode array. Several such electrode configurations have been proposed over the past 20 years, including bipolar (BP), tripolar (TP), and partial tripolar (pTP), as well as the quadrupolar (QP) configuration employed in this study ([@bibr8-2331216515569792]; [@bibr18-2331216515569792]). These configurations have been shown to produce narrower excitation patterns than MP stimulation, both physiologically ([@bibr5-2331216515569792]; [@bibr9-2331216515569792]; [@bibr21-2331216515569792]) and psychophysically ([@bibr4-2331216515569792]; [@bibr10-2331216515569792]; [@bibr13-2331216515569792]; [@bibr14-2331216515569792]). Although early studies failed to find clear perceptual benefits of focused stimulation ([@bibr16-2331216515569792]; [@bibr24-2331216515569792]; [@bibr26-2331216515569792]), at least one recent study has shown improved understanding of speech in noise using pTP stimulation when compared with MP stimulation ([@bibr22-2331216515569792]). Because focused stimulation produces narrower excitation patterns, it also commonly leads to more variable thresholds between electrodes. This is presumably because the density of surviving neurons and the distances between electrodes and those neurons are typically not uniform along the cochlear array ([@bibr3-2331216515569792]; [@bibr13-2331216515569792]; Teymouri et al., 2011). Therefore, if the use of focused stimulation techniques becomes more commonplace, threshold measurements may be reinstituted as a necessary step in fitting CIs in the clinic. In acoustic hearing, fast methods for measuring thresholds and psychophysical tuning curves have been developed and validated for similar purposes. These methods are based on a variant of the Bekesy tracking technique (e.g., [@bibr25-2331216515569792]), in which a listener adaptively adjusts the level of a tone, which glides slowly upward or downward in frequency, in order to maintain a threshold level of perception ([@bibr19-2331216515569792]; [@bibr20-2331216515569792]). With these variants, the listener presses a button to indicate a tone is audible while the frequency of the tone is gliding. The level of the sound increases when the button is released and decreases when it is pressed. At face value, these methods appear difficult to implement in CIs because of the discrete nature of electrodes that represent sound frequency. However, technology in a subset of currently available CIs allows for quasi-continuous sweeps by using current steering---the stimulation of two neighboring electrodes simultaneously with a variable ratio of current between them, allowing for pitch percepts that are intermediate to those produced by the stimulation of the electrodes individually. This study used virtual current sweeps to implement a procedure based on the Bekesy tracking method, thereby providing a rapid and clinically feasible measure of threshold across the electrode array. Thresholds obtained with this new method were compared with thresholds obtained using a traditional adaptive forced-choice procedure. The test--retest reliability of the two procedures was assessed, along with the respective test times. All procedures were performed with both a broad MP and focused QP electrode configuration. Methods {#sec2-2331216515569792} ======= Subjects {#sec3-2331216515569792} -------- Fifteen postlingually deafened adults participated; all were implanted with the Advanced Bionics (Valencia, CA) HiRes90K CI. One subject was bilaterally implanted (S23, left ear and S36, right ear) to give a total of 16 ears tested. Details are shown in [Table 1](#table1-2331216515569792){ref-type="table"}. Ten of the subjects were tested at the University of Washington in Seattle (subject identifiers with an "S") while the remaining five subjects were tested at the University of Minnesota in Minneapolis (subject identifiers with a "D"). The respective Human Subjects Review Boards approved all procedures, and all subjects provided written informed consent. Table 1.Subject Demographics.Subject codeGenderAge at start of experiment (Years)CI use prior to experiment (Years)EtiologyDuration bilateral severe-profound hearing loss prior to implant (Years)D24M64.36.9Unknown progressive27D26F55.05.5Unknown11D28F65.611.6Familial progressive SNHL7D33M74.71.3Noise exposure; Trauma\<1D38F32.61.3Sudden SNHL\<1S22F73.66.9Unknown progressive11.7S23M69.07.0Unknown progressive, sudden loss 10 years ago7.9S28F74.75.0Autoimmune disease18.7S29M82.05.0Unknown progressive34.0S30F49.010.0Unknown progressive16.0S36^[a](#table-fn1-2331216515569792){ref-type="table-fn"}^M69.14.5Unknown progressive, sudden loss 8 years ago3.5S38M48.83.6Otosclerosis18.3S40M51.71.3Enlarged vestibular aqueduct46.4S41M48.65.6Maternal rubella1.2S42M63.812.6Sudden loss33.5S43M68.40.6Unknown progressive17.9[^1] Stimuli {#sec4-2331216515569792} ------- Biphasic, charge-balanced, cathodic-phase-first pulse trains were used. Phase durations were 97 µs and the pulse rate was 997.9 pulses per second. Each pulse train was 200.4 ms in duration and was presented either in MP or QP configuration (see [Figure 1](#fig1-2331216515569792){ref-type="fig"}). All stimuli were presented and controlled using research hardware and software ("BEDCS") provided by the Advanced Bionics Corporation (version 1.18). Programs were written using the MATLAB programming environment, which controlled low-level BEDCS routines. Identical software and hardware were used at both testing sites (Minneapolis and Seattle). Figure 1.Schematic of steered QP (left) and MP (right) electrode configurations. QP consists of four adjacent electrodes; two active electrodes and two flanking return electrodes that share a fraction of the return current. The remaining current is delivered to an extracochlear ground electrode. Alpha represents the ratio of current for the two active electrodes. An alpha of 0 (a) indicates all stimulating current is delivered to the more apical electrode (panel a), while an alpha of 1 (c) indicates all current is directed to the basal electrode (panel c). Alpha equal to .5 (b) indicates half of the stimulating current is applied to each of the two active electrodes (panel b). In the MP configurations (right), all of the return current is delivered to the extracochlear ground electrode. Alpha nomenclature is the same for both configurations, and electrodes more apical are to the right. The stimulation procedures for current focusing and steering with the QP configuration are illustrated in [Figure 1](#fig1-2331216515569792){ref-type="fig"}. The QP configuration consists of two active electrodes and two adjacent flanking electrodes. With full QP, two flanking electrodes carry all of the return current evenly, which theoretically produces the most restricted current spread. With partial QP (left column of [Figure 1](#fig1-2331216515569792){ref-type="fig"}), a fraction of return current, sigma (σ), is delivered to the flanking electrodes, and the remainder is delivered to an extracochlear ground. The broadest stimulation, corresponding to the MP configuration (right column), is achieved when all of the return current flows to the distant ground (σ = 0). In this experiment, the value of σ was fixed at .8 for one subject (D24) and at .9 for all other subjects. A σ value of .9 was selected because it allowed the measurement of suprathreshold stimulation while maintaining current levels lower than the voltage compliance limits of the device. For subject D24 a smaller σ value of .8 was necessary to remain below compliance limits. The illustrations from top to bottom show the effects of changing the steering coefficient, alpha (α). At the top (panel a), all current is delivered through the more apical of the two active electrodes (α = 0); in the middle (panel b), current is distributed equally (α = .5); and at the bottom (panel c), all current is steered through the most basal active electrode (α = 1). A particular set of electrodes and steering parameter defines a *channel*. Whether stimulated with a broad or focused electrode configuration, it is presumed that every channel produces a different spatial pattern of current in the cochlea. Procedures {#sec5-2331216515569792} ---------- ### Sweep thresholds {#sec6-2331216515569792} Analogous to an upward acoustic frequency sweep, pulse trains were presented at regular time intervals while the alpha value was increased from 0 to 1 in .1 steps beginning with the most apical set of QP electrodes. Each 200.4 ms pulse train was followed by a 300 ms silent interval, for a repetition period of about 500 ms. The value of α was incremented every 1,000 ms, such that the same electrode and α combination occurred twice in succession during the sweep. This process was repeated without interruption for the next, more basal set of electrodes until all available sets were tested (active electrodes 2--15), resulting in a single *forward sweep*. At the beginning of the sweep, the signal was presented at a level of at least 3 dB below MCL. During the sweep, the listener was directed to hold the spacebar of a computer keyboard down when the signal was audible, causing the current level to decrease, and to release the spacebar when it became inaudible, causing the current level to increase. The current step size was 1 dB. The combination of an alpha value of .1 and a tracking step size of 1 dB was considered a good compromise between accuracy and duration, based on pilot data. A similar *reverse sweep* was obtained in the same manner but starting with the most basal set of QP electrodes and ending with the most apical. Current levels from one forward and one reverse sweep constituted a complete *run*, and two such runs were obtained for each subject. Subjects were often provided with a brief practice run that included only two or three electrodes. After this brief exposure, subjects were generally able to complete the task, although some noted that the task seemed difficult. An example of a forward sweep for one subject is shown in [Figure 2A](#fig2-2331216515569792){ref-type="fig"}. The graph depicts a portion of the data sequence beginning 40 s from the start of the sweep and encompassing active electrode pair 5/6, α = .5, to electrode pair 6/7, α = .5. Note that each electrode and α combination, or channel, is represented twice, because during a sweep each stimulus is repeated before proceeding to the next set of parameters. Figure 2.A sample of sweep data over time (A) and across channel numbers (B) for S29. (A) Stimulus level and active electrode number during a portion of a forward sweep. The *x*-axis is time in seconds and the *y*-axis is stimulus level in dB re 1 µA. The active electrodes and alpha values are indicated by each pair of symbols. (B) Stimulus level as a function of active channel number for a forward (blue) and reverse (red) sweep. The *x*-axis is channel number and the *y*-axis is stimulus level in dB re 1 µA. The gray bar centered at Channel 6 indicates the range of data used for the average threshold estimate (black triangle). The corresponding numbers indicate the weights applied to the data for each channel. Final threshold estimates for a run were obtained by a weighted averaging of consecutive current levels along the forward and reverse sweeps. This procedure is illustrated in [Figure 2B](#fig2-2331216515569792){ref-type="fig"}, based on the same forward sweep as [Figure 2A](#fig2-2331216515569792){ref-type="fig"} in addition to the complementary reverse sweep. First, a channel number was assigned to each stimulus by adding the alpha fraction to the number of the apical active electrode. In [Figure 2](#fig2-2331216515569792){ref-type="fig"}, for example, a stimulus with active electrodes 5/6 and α parameter .8 is assigned a channel number of 5.8. The forward and reverse data points were then assigned to the appropriate location along the ordinate channel axis, as depicted by the blue and red points in [Figure 2B](#fig2-2331216515569792){ref-type="fig"}. Note that "nonsteered" channels occurring at integer channel numbers (e.g., 6.0 in [Figure 2B](#fig2-2331216515569792){ref-type="fig"}) have twice the number of data points as other channels; this is because these channels are created from two different sets of electrodes, one set with α = 1 and a more basal set with α = 0, such that the same electrode carries all of the active current for both channels (Electrode 6 in this example). Next, the data points corresponding to the individual channels were averaged together to produce a sequence of intermediate threshold estimates (not shown). Finally, a weighted average was calculated using a 5-point hamming window of length ±.2 centered on the current value of α; the five channel weights, from apex to base, were .083, .25, .33, .25, and .083. The window length of ±.2 was selected because it better matched the thresholds estimated with the standard procedure than a window length of ±.1. In [Figure 2B](#fig2-2331216515569792){ref-type="fig"}, the extent of the averaging window is indicated by the horizontal and dotted lines, and the resulting threshold estimate for channel 6.0 is shown by a black triangle. Averaging in this manner reduces the effect of hysteresis observed in unidirectional sweeps ([@bibr19-2331216515569792]). Thresholds were calculated in this manner for every nonsteered integer-numbered channel, for direct comparison to thresholds estimated using the adaptive forced-choice method, described below. Adaptive forced-choice thresholds {#sec7-2331216515569792} ================================= Clinically, when audiologists measure threshold behaviorally with CI listeners, a procedure is used that is similar to how pure-tone thresholds are measured for an audiogram, except that the listener is sometimes asked to count the beeps. Because the clinical procedures are not standardized across devices, a more scientific and precise procedure was selected for comparison to the sweep method described here. For each subject, detection thresholds were measured using a two-down, one-up, two-interval two-alternative forced-choice (2AFC) procedure converging on the 70.7% correct point of the psychometric function ([@bibr13a-2331216515569792]). Each run began with the stimulus level at least 3 dB below MCL, and the level was initially increased or decreased by 2 dB depending on the listener's responses. After the first two reversals, the step size was decreased to .5 dB, where it remained for the final four reversals. The average stimulus level at the last four reversals was defined as threshold for that run. A final threshold was obtained as the average over four runs, with a fifth added if the standard deviation for the last four reversals was greater than 1.5 dB. Thresholds were measured for all available electrodes from 3 to 15 using the non-steered α = 1.0, and for Electrode 2 using an alpha of 0. The same MP and QP configurations were used as for the sweep thresholds. The measurement order for electrode and stimulation method (2AFC and channel sweep) was randomized across repetitions and subjects. Results {#sec8-2331216515569792} ======= [Figure 3](#fig3-2331216515569792){ref-type="fig"} shows an example of one forward (blue) and one reverse (red) sweep, with current level plotted as a function of channel number from apex to base for subject S29. Black triangles depict the final threshold estimates of the run for each electrode-centered channel, following the weighted averaging procedure. Figure 3.A full example of one forward (blue) and one reverse (red) sweep for S29. The *x*-axis is channel number and the *y*-axis is stimulus level in dB re 1 µA. The black triangles represent the average threshold estimate for each of the cardinal channel numbers. Comparison of Threshold Measures {#sec9-2331216515569792} -------------------------------- Thresholds obtained with the traditional two-interval forced-choice (open symbols) and the channel sweep procedures (filled symbols) are shown for all individual subjects in [Figure 4](#fig4-2331216515569792){ref-type="fig"}. The average thresholds for the QP (triangles) and MP configurations (circles) are plotted, along with the standard deviations estimated from the repeated measures (with *N* = 2 in the case of the sweep, and *N* = 4 or 5 in the case of the forced-choice procedure). Consistent with previous findings, thresholds in the more focused QP stimulation mode are higher and more variable than thresholds in the MP mode ([@bibr2-2331216515569792]; [@bibr13-2331216515569792]; [@bibr17-2331216515569792]). Overall, the channel-to-channel threshold patterns of the two procedures match quite closely. The similarity of threshold estimates between the two procedures can also be observed in the mean data, shown in [Figure 5](#fig5-2331216515569792){ref-type="fig"}. The scatter plot in [Figure 6](#fig6-2331216515569792){ref-type="fig"} compares thresholds from the individual subjects obtained using the two different procedures. Any systematic bias of the sweep procedure, relative to the forced-choice procedure, would be reflected in a tendency for the data points to fall predominantly above or below the major diagonal. No such bias is discernable from the plot. To investigate the possibility of systematic biases more rigorously, a within-subjects repeated-measures analyses of variance (ANOVAs) was carried out with threshold as the dependent variable and electrode number, stimulation mode (MP or QP), and procedure (sweep or forced-choice) as within-subjects factors. (All reported ANOVA results include a Huynh--Feldt correction for lack of sphericity where applicable.) The ANOVA revealed a significant main effect of stimulation mode (*F*(1, 14) = 210.75, *p* \< .001), but no main effect of either electrode number or procedure. However, there were significant interactions between procedure and stimulation mode (*F*(1, 14) = 11.07, *p* = .005), as well as procedure and electrode number (*F*(8.4, 118) = 5.001, *p* \< .001). The interaction between procedure and stimulation mode may reflect the trend apparent in [Figure 5](#fig5-2331216515569792){ref-type="fig"} for thresholds to be lower in the sweep than in the forced-choice procedure in the MP but not the QP mode. The interaction between procedure and electrode number is more difficult to interpret and may reflect the apparently different patterns observed at different electrode sites (e.g., sweep thresholds tend to be higher than forced-choice thresholds on Electrode 2, whereas the opposite is true for Electrode 9). Nevertheless, the relatively small (∼ 1 dB) difference, along with a lack of a main effect of procedure, is consistent with the hypothesis that the sweep procedure does not introduce any systematic biases, relative to the thresholds obtained in the more standard adaptive forced-choice procedure. Figure 4.Detection thresholds measured with 2IFC methods (open) and sweep methods (filled) for all subjects individually. Thresholds for QP are indicated by triangles and MP by circles. Error bars represent ±1 standard deviation and are shown when they exceed the symbol size. The *y*-axis is stimulus current (dB re: 1 µA) and the *x*-axis is CI channel from apical to basal. Figure 5.Detection thresholds averaged across all subjects as a function of electrode number for each stimulation mode and procedure. Error bars represent ±1 standard error of the mean. The *y*-axis is stimulus current (dB re: 1 µA) and the *x*-axis is CI channel from apical to basal. Figure 6.Comparison of thresholds estimated with 2IFC and sweep methods across all subjects. Each panel plots thresholds estimated by sweep method (*y*-axis) as a function of 2IFC method (*x*-axis) for the MP configuration data (left) and QP data (right). *R* values are based on Pearson's correlation analysis and were significant with a *p* \< .001. Reliability of Threshold Measures {#sec10-2331216515569792} --------------------------------- At least four estimates of threshold in the adaptive forced-choice procedure were made in order to provide a "gold-standard," with which to compare the new channel sweep threshold estimates, as described earlier. However, to assess the test--retest reliability of the two procedures, an equal number of runs were compared. As only two runs were collected in each subject for the channel sweep procedure (each consisting of one forward and one reverse sweep), only the first two of the adaptive forced-choice runs were included in this assessment. The comparisons between the first and second threshold estimates are shown as scatter plots in [Figure 7](#fig7-2331216515569792){ref-type="fig"}, with the adaptive forced-choice and channel sweep procedures displayed in the top and bottom rows, respectively, and the MP and QP data in the left and right panels, respectively. In all panels, the first and second run threshold estimates qualitatively appear similar, with no apparent upward or downward biases. Test--retest reliability was evaluated quantitatively by subjecting the squared difference between the two estimates to a repeated-measures ANOVA, with electrode number, stimulation mode (MP or QP), and procedure as factors. The analysis revealed a significant main effect of procedure (*F*(1, 12) = 11.36, *p* = .006, partial η^2 ^= .486), as well as a marginally significant three-way interaction between stimulation mode, procedure and electrode number (*F*(4.8, 57.3) = 2.63, *p* = .035, partial η^2 ^= .179). The main effect of procedure is reflected in the square root of the mean squared (rms) difference between the two runs (averaged across electrode number, stimulation mode, and subject), which was 1.68 dB for the adaptive forced-choice procedure and 2.64 dB for the sweep procedure. Thus the test--retest reliability of the traditional method is better than that of the new sweep method, although the difference in rms error is less than 1 dB. There is no clear interpretation of the three-way interaction term and, given its small effect size and marginal significance, it is unlikely to be of importance. Figure 7.Test--retest reliability for 2IFC (top panels) and sweep (bottom panels) threshold estimates. Threshold estimates from the last two or three runs of the 2IFC procedure (*y*-axis) are plotted as a function of the first two 2IFC estimates (*x*-axis) for MP (top left) and QP (top right). Threshold estimates from the second forward and reverse sweep (*y*-axis) are plotted as a function of the first sweeps (*x*-axis) for MP (bottom left) and QP (bottom right) configurations. *R* values are based on Pearson's correlation analysis and were significant with a *p* \< .001. Testing Time {#sec11-2331216515569792} ------------ A primary goal of this research was to test whether the threshold procedure can be performed within a timeframe that makes it clinically feasible. The duration of testing time was substantially shorter for the channel sweep procedure than for the adaptive forced-choice procedure. The total testing time, in minutes, to complete the first two runs for the traditional (open) and sweep (filled) procedures is shown in [Figure 8](#fig8-2331216515569792){ref-type="fig"}. The average time to complete two runs per electrode for the adaptive procedure was 40.8 min for the MP mode and 49.5 min for the QP mode, while the time to complete two runs (two upward and two downward sweeps) for the channel sweep procedure was 11.4 and 12.1 minutes for the MP and QP stimulation modes, respectively. This is an improvement in testing speed by approximately a factor of 4. This factor would be expected to be similar if only one run of each were performed; in that case, the time for the sweep method could be reduced to around 6 min. A repeated-measures ANOVA was performed with the log-transformed completion time as the dependent variable and procedure and stimulation mode as factors. As expected, there was a significant main effect of procedure (*F*(1, 15) = 779, *p* \< .001, partial η^2 ^= .981). The main effect of stimulation mode was also significant (*F*(1, 15) = 15.0, *p* = .001, partial η^2 ^= .501), as was the interaction between factors (*F*(1, 15) = 5.73, *p* = .03, partial η^2 ^= .276). Figure 8.Duration of testing for sweep (open) and 2IFC (filled) methods. Each pair of bars represent data for one subject and data averaged across subjects is the rightmost pair of each panel. Error bars represent the standard deviation. The height of the bar indicates the time it took (in minutes) to complete threshold estimates for each subject using the sweep and 2IFC procedures. Data are shown for MP (top) and QP (bottom). Comparison of Different Focused Stimulation Methods {#sec12-2331216515569792} --------------------------------------------------- Previous research on focused stimulation conducted at the University of Washington laboratory was based on the more common pTP electrode configuration. We compared thresholds obtained with the QP configuration with those using the pTP configuration for the nine subjects tested at the University of Washington. The pTP mode consists of a single active electrode and two flanking electrodes that share the return current. A fraction of 10% of the return current was delivered to the extracochlear ground electrode, corresponding to pTP stimulation with a sigma of .9. Traditional thresholds were compared between pTP and the two nonsteered QP configurations corresponding to the same active electrode with alpha at 0 and 1. For example, if Electrode site 4 serves as the active electrode for the pTP configuration, that same site would also be the active electrode for both the QP 3/4 electrode pair (with α = 1) and the 4/5 electrode pair (with α = 0). For this analysis, only three active electrodes were tested in each subject (4, 9, and 13) and the data are shown in [Figure 9](#fig9-2331216515569792){ref-type="fig"}. The thresholds were highly significantly correlated among the pTP and two QP electrode configurations. This suggests that the QP sweep thresholds, which were measured using a constantly changing steered configuration, are effectively equivalent to those obtained with a static nonsteered configuration. Moreover, the absolute similarity of the thresholds suggests that the focusing capabilities of the two configurations are comparable. Figure 9.Comparison of threshold estimates using the 2IFC procedure with steered QP and pTP configurations. Left panels represent pTP thresholds (*x*-axis) versus QP thresholds with an alpha of one (top) and zero (bottom). The lower right panel shows threshold estimates for steered QP with alpha of 1 (*x*-axis) and alpha of zero (*y*-axis). *R* values are based on Pearson's correlation analysis and were significant (*p* \< .001). Discussion {#sec13-2331216515569792} ========== Consistent with previous studies, psychophysical thresholds are variable across CI subjects and within subjects from electrode to electrode. The variability is particularly large when focused electrical fields are used, such as those obtained with QP stimulation (e.g., [@bibr2-2331216515569792]; [@bibr17-2331216515569792]). As a check, thresholds with QP were found to be nearly identical to those obtained with the more widely tested pTP configuration (e.g., [@bibr4-2331216515569792]; [@bibr6-2331216515569792]). Inspired by procedures devised for the acoustic domain (e.g., [@bibr20-2331216515569792]), we developed and tested a method to obtain thresholds in CI listeners by sweeping the centroid of electrical stimulation across the cochlea while varying current level adaptively. The average thresholds and trial-to-trial variability using the novel sweep method were similar to those obtained with the more conventional adaptive forced-choice procedure, for both focused and broad electrode configurations. Although the electrodes at the end of the array were not significantly different with the two procedures, future versions of the software will require one reversal for the end electrodes before beginning the sweep. The results also showed that the new sweep method is approximately a factor of 4 times faster for the same number of threshold estimates, bringing it within the realm of possibility for clinical use. Thus, this method could be applied clinically in devices that support current steering (i.e., those with multiple independent current sources that can be controlled in software). [@bibr22-2331216515569792] have recently demonstrated improved speech perception in noise when employing focused stimulation. As investigators in this field further explore how best to program implants with focused electrical fields ([@bibr1-2331216515569792]; [@bibr22-2331216515569792]), the sweep method could be implemented clinically to assist in the fitting procedure. When programming focused strategies, because the thresholds can be highly variable from channel-to-channel, it may be critical to measure individual channel thresholds. If the channel-to-channel variability is not addressed with the measurement of thresholds, stimulation produced by some sounds will be inaudible while others could be mapped too high in the electrical dynamic range producing a distorted signal. In addition to the potential benefits of focused stimulation discussed thus far, thresholds might be used for diagnostic purposes, specifically for evaluating the electrode--neuron interface ([@bibr3-2331216515569792]; [@bibr13-2331216515569792]). For instance, it has been reported that channels with high thresholds have broader psychophysical tuning curves ([@bibr4-2331216515569792]) and smaller dynamic ranges ([@bibr6-2331216515569792]) than channels with low thresholds. The methodology described here provides an efficient way to obtain these important measurements, which could ultimately lead to improvements in fitting strategies. The authors thank the CI subjects who ever so patiently participated in this study. The authors also thank Emily Ellis, AuD, for her assistance with subject recruitment, data collection, and figure preparation, and Leonid Litvak and Kanthiah Koka from Advanced Bionics for technical support. For those interested in obtaining the software used in this study, please go to <https://sites.google.com/site/biererlab/bierer-lab-resources>. Declaration of Conflicting Interests {#sec14-2331216515569792} ==================================== The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding {#sec15-2331216515569792} ======= The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by NIDCD grants R01 DC012142 (J. A. B.) and R01 DC012262 (A. J. O.). [^1]: ^a^Indicates the second ear of S23.

Sevoflurane is a volatile anesthetic with a low blood--gas partition coefficient and produces a rapid induction and recovery of anesthesia \[[@r16]\]. During last decade, clinical use of sevoflurane has been spreading in veterinary practice. Sevoflurane is minimally metabolized and easily cleared in animals, however, it should be remembered that sevoflurane has a dose-dependent depressant effect on cardiorespiratory function in dogs \[[@r22]\]. Because of these side effects, sevoflurane must be carefully titrated, and vigilant monitoring should be employed to avoid excessive anesthetic depth. The term of balanced anesthesia usually refers to the use of different drugs in combination to provide hypnosis, analgesia and muscle relaxation \[[@r25]\]. Administration of analgesic drugs as a part of the balanced anesthesia is sparing with requirements of anesthetic drugs, including sevoflurane, thereby reducing their depressant effects on the cardiovascular system and improving the quality of anesthesia \[[@r24]\]. Minimum alveolar concentration (MAC) is defined as the partial pressure of a gas that produces immobility in 50% of individuals exposed to a supramaximal noxious stimulation \[[@r12]\]. The MAC is the standard measure to evaluate inhaled anesthetic potency \[[@r27]\]. Lidocaine is a sodium channel blocker, and its intravenous infusion can be spared with anesthetic requirements in dogs undergoing surgery \[[@r9], [@r19], [@r31]\]. It is also reported that the intravenous infusion of lidocaine decreases the MAC of isoflurane \[[@r21], [@r36]\] and sevoflurane \[[@r7], [@r11], [@r20]\] in a dose-dependent manner in dogs. Studies in humans and in dogs suggest that low doses of alpha~2~-adrenergic agonists may produce the sparing effects of anesthetic requirements and analgesia with minimal impact on cardiovascular function \[[@r1],[@r2],[@r3], [@r24]\]. In dogs, dexmedetomidine has been shown to reduce the anesthetic requirement for induction and maintenance of general anesthesia \[[@r2], [@r8], [@r24], [@r34]\]. Therefore, it is expected that the balanced anesthesia using a combination of lidocaine and dexmedetomidine infusion may decrease sevoflurane requirement and therefore may reduce the incidence of its side effects. The aim of the present study was to evaluate the effects of constant rate infusion (CRI) of a combination of lidocaine and dexmedetomidine on the MAC of sevoflurane in dogs. The authors suggest the hypothesis that the combination of lidocaine with dexmedetomidine significantly reduces the MAC of sevoflurane in dogs. MATERIALS AND METHODS {#s1} ===================== *Animals and experimental protocol*: Seven adult mixed-breed neutered dogs, age 1--2 years, three males and four females and body weight (mean ± SD) 18.1 ± 9 kg, were included in a prospective randomized cross-over experiment with a 2-week washout period between treatments. Dogs were considered to be healthy on the basis of medical history, physical examination, complete blood count (CBC) and serum biochemical analysis. Food but not water was withheld 8 hr prior to each anesthetic procedure. This study was planned as a randomized crossover trial. Each dog was anesthetized three times and received one of the following three treatments: 1) an intravenous (IV) loading dose (LD) of 2 mg/kg lidocaine (Lidocaína 2% Inyectable: Pisa, México) followed by lidocaine 6 m*g*/kg/hr CRI (LIDO), 2) LD of 2 *µ*g/kg IV dexmedetomidine (Dexdomitor, Orion Corporation, Espoo, Finland, 0.5 mg/m*l*) followed by dexmedetomidine 2 *µ*g/kg/hr CRI (DEX) and 3) LDs of lidocaine 2 mg/kg IV and dexmedetomidine 2 *µ*g/kg IV followed by lidocaine 6 mg/kg/hr and dexmedetomidine 2 *µ*g/kg/hr CRI (LIDO--DEX). Loading doses were diluted up to a final volume of 3 m*l* with sterile water and administered IV over 1 min. Treatments were diluted up to 60 m*l* with saline 0.9% and delivered as a CRI accordingly. All CRIs were started immediately after bolus administration using a syringe infusion device (Colleague, Baxter Healthcare Corporation Medication Delivery, Deerfield, IL, U.S.A.). The sevoflurane MAC was determined before (SEV-MAC~BASAL~) and during one of the three CRI treatments (SEV-MAC~LID~, SEV-MAC~DEX~ and SEV-MAC~LID-DEX~) in each dog. This study was approved by the animal research ethics committee of the Universidad Autonoma de Mexico with protocol number 2267/2010. *Anesthetic procedure and instrumentation*: A 20-gauge catheter was aseptically placed into the cephalic vein. Anesthesia was induced with an intravenous administration of propofol (Fresofol 1%, Fresenius Kabi, Pimble, Australia) at a dose of 6 mg/kg. Orotracheal intubation was performed in all dogs with an appropriately sized, cuffed endotracheal tube that was attached to a circle anesthetic rebreathing system (Fabius Dragër Medical GmbH 23542, Lübeck, Germany). Anesthesia was maintained with sevoflurane (Sevorane Abbott Laboratories, Bogota, Colombia) vaporized in 100% oxygen with a flow rate of 2 l/min (Dragër medical, AG&CO, KGaA, Lubeck, Germany, Dragër Vapor 2000). All dogs were administered lactate Ringer's solution at a flow rate of 3 m*l*/kg/hr through the catheter by the use of an infusion pump (Colleague, Baxter Healthcare Corporation Medication Delivery) and mechanically ventilated with intermittent positive-pressure ventilation (IPPV) (Fabius Dragër Medical GmbH Lübeck) to maintain eucapnia (35--40 mmHg of end-tidal CO~2~) during the anesthesia. End-tidal concentration of sevoflurane SEVO (ET~SEV~) and ETCO~2~ was continuously monitored by a side-stream infrared gas analyzer (Dräger Vamos, Dräger Medical GmbH). Dogs were placed in lateral recumbency, and a 22-gauge catheter was aseptically placed in the dorsal metatarsal artery and attached to an electrical transducer (DTX Plus DT-4812, Becton Dickinson Critical Care Systems Pte Ltd., Singapore) connected to a multi-parameter monitor (WL Surgivet V9212SR 2009-01, Smith Medical PM Inc., Waukesha, WI, U.S.A.). Systolic, diastolic and mean arterial blood pressures (SAP, DAP and MAP, respectively) were continuously monitored via a blood-pressure transducer system connected to the dorsal pedal artery (DTX plus^®^ DT 4812, Becton Dickinson Critical Care Systems Pte Singapore Ltd.). The zero reference point of the pressure transducer was set at the level of the heart. Heart rate and rhythm (EGG lead II) and pulse oximetry were also continuously monitored by placing the electrodes at the level of the elbows and left patella and an infrared sensor attached to the dog's tongue, respectively (WL Surgivet V9212SR 2009-01, Smith Medical PM Inc.). A circulating warm-water blanket was used to maintain the esophageal temperature between 37.5 and 38.5°C. *MAC determination*: Following the propofol induction, the dogs had been anesthetized for at least 90 min as an initial equilibration period at an ET~SEV~ of 2.7% to minimize the effects of propofol. The determination of SEV-MAC~BASAL~ for each dog was started after the initial equilibration period. Once the SEV-MAC~BASAL~ was determined, dogs were received the CRI treatment of lidocaine, dexmedetomidine or combination. The SEV-MAC~LID~, SEV-MAC~DEX~ and SEV-MAC~LID-DEX~ were determined after 45 min equilibration period of the CRI treatments \[[@r15], [@r36]\]. Cardiovascular parameters and other variables were recorded immediately before the determination of minimum alveolar concentration (MAC) of sevoflurane. MAC was determined by use of a previously described technique. Noxious stimulation was applied by clamping a paw of the third or fourth digits. The clamping technique was performed with 24-cm sponge forceps (with protective plastic tubing on each jaw) clamped to the first notch until gross purposeful movement was detected or a period of 60 sec elapsed \[[@r35]\]. A negative response included the lack of movement of head and limbs, muscle rigidity, shivering, tail movement, couching, swallowing or an increase in spontaneous respiratory efforts during controlled ventilation. When a positive response was elicited, the ET~SEV~ was increased by 0.1% and maintained at this concentration for at least 20 min, and the noxious stimulus procedure was repeated. When a negative response was detected, the ET~SEV~ was decreased by 0.1% and maintained at this concentration for at least 20 min, and the noxious stimulus procedure was repeated. The procedure was continued until purposeful movement ceased (increase in anesthetic concentration) or returned (decrease in anesthetic concentration). The sevoflurane MAC was calculated as a mean value between the highest ET~SEV~ at which the purposeful movement was detected and the lowest ET~SEV~ at which the purposeful movement was not detected. In each dog, the sevoflurane MAC was evaluated in duplicate. The sevoflurane MAC values were corrected to sea level by use of the formula (barometric pressure of location/760 mmHg) × obtained MAC value. The mean barometric pressure was obtained from the official city meteorological station for the altitude at which the experiment was performed (2,680 m above sea level) and was 556 mmHg. Once the experiment had been finalized, the dogs were recovered from anesthesia and administered 4 mg/kg carprofen (Rimadyl, Pfizer Animal Health BV, Capelle a/d I Jssel, The Netherlands) subcutaneously every 24 hr for 2 days. *Statistical analysis*: Statistical analysis was performed using computer software (SigmaStat 3.5 program, Systat Software Inc., Point Richmond, CA, U.S.A.). The Shapiro-Wilk test was used for assessment of data normality. Data are reported as means ± standard deviations (SD). A repeated-measures ANOVA was used to evaluate percentage change in sevoflurane MAC before and after the CRI treatments, time to MAC determination and extubation time. A post-hoc Tukey test was used where appropriate. Values were considered significantly different when *P*\<0.05. The interaction of lidocaine and dexmedetomidine was used to evaluate whether change in cardiorespiratory data and MAC values departed from an additive model. The changes observed in dogs treated with lidocaine and dexmedetomidine (SEV-MAC ~LID-DEX~) were compared with those observed in groups (SEV-MAC~LID~ and SEV-MAC~DEX~) by use of two-way repeated-measures ANOVA, if a significant difference was obtained with a significant interaction, the drug interaction between SEV-MAC ~LID-DEX~ and SEV-MAC~LID~, SEV-MAC~DEX~ was judged to be synergistic. If the interaction term was not significant, the main effects of lidocaine and dexmedetomidine was judged to be additive. For all analyses, values of *P*\<0.05 were considered significant \[[@r18], [@r39]\]. Values are expressed as mean ± SD. RESULTS {#s2} ======= Times to SEV-MAC~BASAL~ determination were 174 ± 24 min, 196 ± 18 min and 181 ± 19 min for dogs receiving the LIDO, DEX and LIDO--DEX CRI treatments, respectively. Times to SEV-MAC~LID~, SEV-MAC~DEX~ and SEV-MAC~LID-DEX~ was 172 ± 24 min, 194 ± 17 min and 181 ± 18 min, respectively. These times were not significantly different when groups were compared. The LIDO, DEX and LIDO--DEX CRI treatments significantly decreased the sevoflurane MAC ([Table 1](#tbl_001){ref-type="table"}Table 1.Mean ± standard deviations of the minimum alveolar concentration (MAC) of sevoflurane and percentage of MAC reduction recorded in dogs after a constant-rate infusion (CRI) of lidocaine or dexmedetomidine or their combinationThe CRI treatmentSEV-MAC~BASAL~ (%)SEV-MAC during the\ CRI treatment (%)MAC reduction (%)LIDO1.90 ± 0.2%1.38 ± 0.08\*27.3 ± 8.0\*DEX1.82 ± 0.17%1.10 ± 0.23\*40.5 ± 12.0\*LIDO-DEX1.82 ± 0.17%0.78 ± 0.14**^\*^**^¥^54.1 ± 8.0**^\*¥^**MAC of sevoflurane was determined after 90 min equilibration period in the dogs (SEV-MAC~BASAL~). Then, sevoflurane MAC was determined again in the dogs after 45 min equilibration period of one of the following treatments: an intravenous loading dose of lidocaine 2 mg/kg followed by 6 mg/kg/hr CRI (LID); an intravenous loading dose of dexmedetomidine 2 *µ*g/kg followed by 2 *µ*g/kg/hr CRI (DEX); or their combination (LID-DEX). The % reduction in the sevoflurane MAC after treatments was calculated from (SEV-MAC during the CRI treatment -- MAC~BASAL~ × 100). \*Significantly different from MAC~BASAL~ (*P*\<0.05). ¥ Significantly different from LIDO and DEX treatments (*P*\<0.05). Sparing effect of these combination (LIDO-DEX) on sevoflurane MAC was additive.). The SEV-MAC~BASAL~ of all treatments was 1.82 ± 0.06%. The SEV-MAC~LID~ was 1.38 ± 0.08%, SEV-MAC~DEX~ was 1.22 ± 0.10%, and SEV-MAC~LID-DEX~ was 0.78 ± 0.06%. The SEV-MAC~LID-DEX~ was significantly lower compared with the SEV-MAC~LID~ or SEV-MAC~DEX~ (*P*\<0.05). The LIDO, DEX and LIDO--DEX CRI treatments significantly decreased the sevoflurane MAC by 27.3 ± 8.0%, 41.0 ± 12.0% and 54.14 ± 8.0% for all treatments, when compared with SEV-MAC~BASAL~ (*P*\<0.05). Therefore, it indicates that sparing effects of the combination lidocaine with dexmedetomidine were additive. All dogs recovered smoothly from anesthesia and were extubated within 10 min after the discontinuation of sevoflurane anesthesia. Extubation time was 361 ± 17 min, 384 ± 20 min and 372 ± 19 min for LIDO, DEX and LIDO-DEX, respectively. These values were not significantly different when groups were compared. The statistical interaction of change in heart rate between LIDO+DEX and LIDO, DEX groups was significant (*P*\<0.0001) ([Table 2](#tbl_002){ref-type="table"}Table 2.Cardiorespiratory parameters observed during the determination of minimum alveolar concentration (MAC) of sevoflurane in dogsVariablesMAC~B~MAC~T1~MAC~T2~MAC~T3~Heart rate (beats/min)118 ± 5110 ± 5\*83 ± 9\***^+^**76 ± 22\***^+^**Systolic blood pressure (mmHg)99 ± 8106 ± 399 ± 8104 ± 7Diastolic blood pressure (mmHg)68 ± 571 ± 473 ± 379 ± 3\***^+^**Mean arterial pressure (mmHg)78 ± 483 ± 282 ± 387 ± 4\***^++^**Oxygen saturation (SpO~2~, %)97.4 ± 197 ± 297.5 ± 195.5 ± 0.2Esophageal temperature (°C)38.1 ± 0.537.8 ± 0.538.4 ± 0.538.2 ± 0.2End-tidal CO~2~ (mmHg)36.2 ± 235.8 ± 1.036.5 ± 136.0 ± 1Data are expressed as mean ± standard deviation for n=7 dogs. Cardiovascular parameters and other variables were recorded immediately before the determination of minimum alveolar concentration (MAC) of sevoflurane. The time determination for the MAC basal (MAC~B~) was 174 ± 24 min, 196 ± 18 min and 181 ± 19 min, respectively, for lidocaine, dexmedetomidine or the combination. The time determination for the MAC treatments groups (MAC~T~) was 172 ± 24 min, 194 ± 17 min and 181 ± 18 min, respectively. Treatment 1 (T~1~) lidocaine, treatment 2 (T~2~) dexmedetomidine and treatment 3 (T~3~) lidocaine and dexmedetomidine combination. The MAC~T1~ was 1.38 ± 0.08%, the MAC~T2~ 1.10 ± 0.23% and the MAC~T3~ 0.78 ± 0.14%. \*Significant differences compared to baseline. **+**Significantly differences compared to LIDO group. **++**Significantly differences compared to DEX group. Statistical significance (*P*\<0.05).). Therefore, it indicates that effects of the combination of lidocaine with dexmedetomidine were additive. The statistical interaction of the% change in MAC was significant between LIDO-DEX and LIDO, DEX groups. Therefore, it indicates that sparing effects of the combination with lidocaine and dexmedetomidine on MAC were additive. DISCUSSION {#s3} ========== In this study, the CRI treatments using lidocaine (2 mg/kg IV followed by 6 mg/kg/hr CRI) or dexmedetomidine (2 *µ*g/kg IV followed by 2 *µ*g/kg/hr) produced significant and clinically meaningful reductions in the sevoflurane MAC~.~ In particular, the CRI combination of lidocaine and dexmedetomidine was synergic and reduced the sevoflurane MAC by half in dogs. In addition, the recovery from anesthesia was smooth and uneventful in all dogs. Therefore, the combination of lidocaine and dexmedetomidine infusions is expected to provide a clinically useful balanced anesthesia in dogs anesthetized with sevoflurane. However, the combination of lidocaine and dexmedetomidine infusions coincidently produced significant cardiovascular changes, such as a decrease in heart rate and an increase in blood pressure. A further study will be necessary to clarify the cardiovascular effects in dogs receiving the CRI combination of lidocaine and dexmedetomidine. The sevoflurane MAC in dogs is reported at a range from 2.1% to 2.4% in most previous studies \[[@r13],[@r14],[@r15], [@r17], [@r20], [@r38]\]. The baseline MAC of sevoflurane in our study was 1.82% (i.e. SEV-MAC~BASAL~), which is approximately 14--20% less than that reported in these previous studies, but very similar to that reported by Seddighi *et al*. \[[@r30]\] (1.78% of the sevoflurane MAC) and Wilson *et al.* \[[@r38]\] (1.9% of the sevoflurane MAC). However, the MAC of an inhalational anesthetic can differ substantially among animals of the same species \[[@r34]\]. Factors affecting variability in MAC include the type of noxious stimulus, subjectivity in interpretation of purposeful movement, differences in the anatomical site of stimulation and differences in physiological variables, such as PaCO~2~, body temperature, arterial blood pressure and age of the test subjects \[[@r26], [@r39], [@r40]\]. Variation within this study was minimized by using a one observer (MRS) and maintaining temperature, ETCO~2~ and arterial blood pressure within the physiological range. In dogs, continuous infusion of lidocaine has been shown to reduce the MAC of inhalational anesthetics, such as isoflurane and sevoflurane, in a dose related fashion \[[@r20], [@r36]\]. Valverde *et al.* \[[@r36]\] reported that lidocaine infusion reduced the isoflurane MAC by 18.7% at an infusion rate of 3 mg/kg/hr CRI and 43.3% at an infusion rate of 12 mg/kg/hr CRI. Matsubara *et al.* \[[@r20]\] reported that lidocaine infusion reduced the sevoflurane MAC by 15% at an infusion rate of 3 mg/kg/hr CRI and 37% at an infusion rate of 12 mg/kg/hr CRI. In the present study, the lidocaine infusion at 6 mg/kg/hr CRI reduced the sevoflurane MAC by 26.1%. This is similar to a previous study reported by Wilson *et al.* \[[@r38]\] where lidocaine infusion at 6 mg/kg/hr CRI reduced the sevoflurane MAC by 29%. While the mechanisms of MAC reduction by lidocaine infusion are still not well known, there are some possibilities. Analgesia produced by lidocaine infusion may be due to a mechanism at the level of the supraspinal or spinal cord \[[@r4]\], which may be expected to cause a decline in the MAC with inhalant anesthetics. Another serious theory is that the mechanism acts at the level of the voltage-dependent sodium channels in the central nervous system \[[@r26], [@r27]\]. In addition, there are some reports of inhibition of potential action on the excitability of the cells in the nervous system \[[@r6]\], which may explain both the analgesic properties and the ability of lidocaine to reduce the MAC \[[@r6]\]. The mechanism for MAC reduction with lidocaine is unclear. Whether it is associated with the analgesic or sedative effects of lidocaine is unknown. While lidocaine's analgesic effects may be responsible for the sparing effect on volatile anesthetic MAC, it is also possible that the MAC reduction resulted from the sedative effects of lidocaine as drugs with sedating actions, such as acepromazine, also reduce MAC \[[@r10], [@r27]\]. Dexmedetomidine infusion reduces the MAC of isoflurane in dogs \[[@r11], [@r24]\]. Pascoe *et al.* \[[@r24]\] reported that dexmedetomidine infusion reduced the isoflurane MAC by 18% at an infusion rate of 0.5 *µ*g/kg/hr CRI following a loading dose of 0.5 *µ*g/kg IV and 59% at an infusion rate of 3 *µ*g/kg/hr CRI following a loading dose of 3*µ*g/kg IV. Ebner *et al.* \[[@r11]\] reported that dexmedetomidine infusion reduced the isoflurane MAC by 30% at an infusion rate of 0.5 *µ*g/kg/hr CRI without loading dose. As so far the authors know, there is no study investigating the effects of dexmedetomidine infusion on the sevoflurane MAC in dogs. In our study, dexmedetomidine infusion at 2 *µ*g/kg/hr CRI following a loading dose of 2 *µ*g/kg IV produced 43.6% of reduction in the MAC of sevoflurane in dogs. This is similar to the results of a previous study reported by Pascoe *et al.* \[[@r24]\] that investigated the effects of dexmedetomidine infusion on isoflurane MAC in dogs. The possible central mechanism that explains the reduction of MAC inhaled anesthetic induced by α2 agonists is the reduction in the releasing of nor- epinephrine in the CNS caused by presinapsis stimulated by α2-adrenergic receptors and neural hyperpolarization induced by post-synaptic activation of α2-adrenergic receptors \[[@r28], [@r29], [@r32], [@r33]\]. Lidocaine is a sodium channel blocker that produces analgesia by inhibition of potential action on the excitability of the cells in the nervous system \[[@r6]\]. Dexmedetomidine is an alpha~2~-adrenergic agonist that produces analgesia by activation of dorsal horn alpha~2~-receptors \[[@r1], [@r2]\]. Because of the different mechanisms of analgesic properties, it is expected that the combination of lidocaine and dexmedetomidine may produce an additive analgesic effect. In the present study, the combination of lidocaine (2 mg/kg IV followed by 6 mg/kg/hr CRI) and dexmedetomidine (0.5 *µ*g/kg IV followed by 2 *µ*g/kg/hr CRI) infusions provided a significant reduction in the sevoflurane MAC by 54.4% in the dogs. As mentioned above, we also observed that the lidocaine infusion alone (2 mg/kg IV followed by 6 mg/kg/hr CRI) reduced the MAC by 26.1% and the dexmedetomidine infusion alone (0.5 *µ*g/kg IV followed by 2 *µ*g/kg/hr CRI) reduced the MAC by 43.6% in the same dogs \[[@r18]\]. The interaction between lidocaine and dexmedetomidine infusions on the sevoflurane MAC reduction was judged to be additive in dogs. It is considered that the combination of lidocaine and dexmedetomedine infusions produces an additive interaction on the anesthetic requirements in dogs. The cardiovascular effects observed in our dogs were similar to those in previous reports in dogs \[[@r5], [@r26], [@r37]\]. Valverde *et al.* \[[@r36]\] reported that lidocaine infusions (3 and 12 mg/kg/hr CRI following 2 mg/kg IV) did not induce clinically significant changes in heart rate and arterial blood pressure in dogs anesthetized with isoflurane. Nunes de Moraes *et al*. \[[@r23]\] reported there were not any detrimental cardiovascular effects related to an infusion of lidocaine at 7.2 mg/kg/hr during isoflurane anesthesia in healthy dogs or dogs with aortic stenosis. Matsubara *et al.* \[[@r20]\] reported that lidocaine infusions (3 and 12 mg/kg/hr CRI following 2 mg/kg IV) did not induce clinically significant changes in heart rate and arterial blood pressure in dogs anesthetized with sevoflurane. In the present study, minimal effects were observed on heart rate and blood pressure in dogs receiving the lidocaine infusion alone. Therefore, it is considered that the lidocaine infusion has minimal side-effects on cardiovascular function of dogs anesthetized with isoflurane or sevoflurane. On the other hand, an administration of dexmedetomidine resulted in a decrease in heart rate and cardiac output an increase in systemic vascular resistance in dogs \[[@r37]\]. Pascoe *et al.* \[[@r24]\] reported that heart rate decreased with increasing doses of dexmedetomidine infusion, while blood pressure increased in dogs anesthetized with isoflurane. Ebner *et al.* \[[@r11]\] also reported that dexmedetomidine infusion (0.5 *µ*g/kg/hr CRI) induced decreases in heart rate and cardiac output and increases in arterial blood pressure and systemic vascular resistance in dogs anesthetized with isoflurane. In the present study, it was also observed that the dexmedetomidine infusion induced a decrease in heart rate; however, the increase in arterial blood pressure was only observed in the LIDO-DEX group. A decrease of heart rate is commonly observed after the administration of dexmedetomidine due to an increased systemic vascular resistance induced by alpha~2~-adrenergic receptor. Pypendop & Verstegen (1998) \[[@r25]\], investigating the dose dependency of these effects in dogs, found that medetomidine caused qualitatively similar hemodynamic changes, irrespective of dose between 1 and 20 *µ*g kg^--1^ IV, although these changes were less at doses of 1 and 2 *µ*g/kg. This could explain, because, in our study not observed a significant increase in blood pressure. As so far the authors know, the interaction between systemic lidocaine and dexmedetomidine on cardiovascular function has not been clarified in dogs. In the present study, the combination of lidocaine and dexmedetomidine infusions produced cardiovascular changes, such as a decrease in heart rate and an increase in arterial blood pressure, compared to the CRI administration of lidocaine alone. Sevoflurane has a dose-dependent depressant effect on cardiovascular function, such as a decrease in cardiac output in dogs \[[@r20]\]. We speculate that these cardiovascular changes induced by a preservation of cardiac output as a benefit from the additive sparing effect on the sevoflurane MAC were produced by the combination of lidocaine and dexmedetomidine infusions. The preserved cardiac output might cause a significant increase in arterial blood pressure and a baroreflex followed by a significant decrease in heart rate. However, we did not measure cardiac output in the present study. A further study will be necessary to clarify the cadiovascular effects in dogs receiving the CRI combination of lidocaine and dexmedetomidine. In conclusion, the combination of lidocaine and dexmedetomidine infusions is expected to provide a clinically useful balanced anesthesia in dogs anesthetized with sevoflurane. The sparing effects of lidocaine with dexmedetomidine on sevoflurane MAC reduction in dogs were additive. However, significant cardiovascular changes were coincidently observed in our dogs receiving the CRI combination. A further study will be necessary to clarify the cardiovascular effects in dogs receiving the CRI combination of lidocaine and dexmedetomidine. To the Mexican National Center for Science and Technology (CONACYT) for funding this research for MSc studies.

Background {#Sec1} ========== Clear cell sarcoma of soft tissue (CCSST) is a malignant mesenchymal tumor that mostly affects young adults and tends to affect the lower extremities, close to the tendon and aponeuroses \[[@CR1]\]. Histologically, CCSSTs have epithelioid tumor nests accompanied by some spindling areas, and wreath-like multinucleated giant cells. CCSSTs present with a melanocytic differentiation and often express melanocytic markers including S-100 protein, melanoma antigen (Melan-A), human melanoma black 45 (HMB45), microphthalmia-associated transcription factor (MITF), and SRY-Box 10 (SOX-10) on immunohistochemistry (IHC). Ultrastructurally, CCSST has premelanosomes in the cytoplasm of tumor cells and shares some characteristic features with malignant melanomas (MMs). MMs genetically have BRAF mutations, although CCSST lacks this mutation. Clear cell sarcoma-like gastrointestinal tumor (CCSLGT) is also a malignant mesenchymal tumor that shares some pathological features with CCSST and arises from the gastrointestinal tract, such as the small and large intestine, and stomach. CCSLGT was originally reported to be an "osteoclast-rich tumor of the gastrointestinal tract with features resembling clear cell sarcoma of the soft parts" \[[@CR2]\] and the first case of CCSLGT was reported by Alpers et al. \[[@CR3]\] as a "malignant neuroendocrine tumor of the jejunum with osteoclast-like giant cells" in 1985. Subsequently, the term CCSLGT was first used by Kosemehmetoglu et al. \[[@CR4]\] in their review, which included 13 CCSLGT cases. However, some authors have proposed using the term "malignant gastrointestinal neuroectodermal tumor," because CCSLGTs lack melanocytic differentiation on IHC and ultrastructural examination and appear to have poorer prognosis \[[@CR5]\]. Although CCSLGT has a similar histology to CCSST in some respects, such as a clear cytoplasm and epithelioid cells, there are some differing characteristics. CCSLGT has a pseudo-papillary growth pattern and many osteoclast-type giant cells, and the tumor cells tend to be positive for S-100 protein but show less expression of melanocytic markers on IHC \[[@CR6]\]. Genetically, CCSST and CCSLGT usually have characteristic chimeric fusions of Ewing sarcoma breakpoint region 1 (*EWSR1*) with cAMP response element-binding protein (*CREB*) gene family members, *EWSR1-*activating transcription factor 1 (*ATF1*) and *EWSR1-CREB1*, which were derived from each translocation of t(12;22)(q13;q12) and t(2;22)(q34;q12), respectively \[[@CR7]--[@CR10]\]. *EWSR1-ATF1* fusion is much more frequent than *EWSR1-CREB1* fusion, but *EWSR1-CREB1* fusion of CCSLGT is comparatively often observed. In this study, we used fluorescence *in situ* hybridization (FISH) and reverse transcription polymerase chain reaction (RT-PCR) to perform genetic analyses of 22 cases of CCSSTs and CCSLGTs, and compared their different chimeric fusion types. Methods {#Sec2} ======= Case selection {#Sec3} -------------- The study protocol for the collection of tumor samples and clinical information were approved by the Institutional Review Board of Sapporo Medical University Hospital (Sapporo, Japan; No. 292--3012). We selected 22 cases of clear cell sarcoma (CCS) including 16 CCSST and 6 CCSLGT cases from the clinicopathological archive at the Department of Surgical Pathology, Sapporo Medical University Hospital. We reviewed all hematoxylin and eosin-stained sections and confirmed that each case fulfilled the histologic criteria of CCSST and CCSLGT. Immunohistochemistry {#Sec4} -------------------- We evaluated previously reported IHC findings of melanocytic markers, including S-100 protein, Melan-A, HMB45, and SOX-10, and assessed their positivity. We also performed additional IHC using representative sections from formalin-fixed and paraffin-embedded tissues in some cases. These tissues were sliced into 3-mm-thick sections and examined with an automated IHC system at Sapporo Medical University Hospital. All slides were loaded into a PT Link module (Agilent Technologies, Santa Clara, CA) and subjected to a heat-induced antigen-retrieval protocol with EnVision FLEX Target Retrieval Solution (Agilent Technologies) before being transferred to the Autostainer Link 48 instrument (Agilent Technologies) and Dako Omnis (Agilent Technologies). We used antibodies against the following antigens: S-100 protein (polyclonal; Agilent Technologies), Melan-A (A103; Agilent Technologies), HMB45 (HMB45; Agilent Technologies), and SOX-10 (N-20; Santa Cruz Biotechnology, Santa Cruz, CA). Fluorescence *in situ* hybridization {#Sec5} ------------------------------------ We performed FISH using the specimens obtained from tumor materials and 4 μm slices on glass slides. First, we selected an area showing representative histology and marked a 5-mm-diameter circle with a marker on the glass slides for FISH analyses. We performed FISH using dual color break apart probe for EWSR1 (Abbott, Abbott Park, IL), ATF1 (Empire Genomics, Buffalo, NY), CREB1 (Empire Genomics), and CREM (Empire Genomics). FISH was conducted as previously described \[[@CR11]\], with the following modifications: baking (1 h at 60 °C), deparaffinization, target gene activation (20 min with 0.2 M HCl followed by 30 min with pretreatment solution at 80 °C), enzyme treatment (60 min with protease solution at 37 °C), re-fixation (10 min in 10% formalin neutral buffer solution), denaturation (5 min with denaturation solution at 72 °C), washing and dehydration (1 min each in 70%, 85%, and 100% ethanol), hybridization with 10 mL DNA probe solution (5 min at 90 °C followed by 48 h at 37 °C), and washing with post-hybridization wash buffer (2 min at 72 °C). As a counterstain, 10 μL 4,6-diamidino-2-phenylindole was added. Slides were coverslipped for viewing under a fluorescence microscope. To detect the presence of *EWSR1*, *ATF1, CREB1*, and *CREM* rearrangements, we counted 50 nuclei in tumor cells that showed a pair of fused and split signals, and calculated the percentage of split signals. The signals were considered split when the distance between the red and green signals was at least twice the estimated signal diameter. We did not evaluate any truncated and overlapping cells in FISH analysis. We considered the specimen to be "split positive" if split signals were observed in more than 10% of tumor cells \[[@CR12]\]. Reverse transcription-polymerase chain reaction {#Sec6} ----------------------------------------------- We detected chimeric fusions by RT-PCR using fresh tumor samples in several available cases. RT-PCR analysis was performed for *EWSR1-ATF1* and *EWSR1-CREB1* fusions. For RT-PCR detection of the *EWSR1-ATF1* and *EWSR1-CREB1* fusions, we used the forward primer EWSex7-F1 with either the CREB1ex7-RevA primer (binds both *CREB1* and *ATF1*; sequence: TCCATCAGTGGTCTGTGCATACTG) or the CREB1ex7-RevC primer (specific for *CREB1*; sequence: GTACCCCATCGGTACCATTGT) \[[@CR1], [@CR7], [@CR13]\]. Results {#Sec7} ======= Clinical findings {#Sec8} ----------------- This study involved 8 male and 14 female patients with a mean age of 40 years (range, 8--78 years). Mean tumor size was 4.6 cm (range, 2--10). The anatomical locations were deep soft tissue of the upper (*n* = 6) and lower (*n* = 8) extremities, esophagus (*n* = 1), small intestine (*n* = 5), abdominal wall (n = 1), and skin (n = 1). The primary site in Case 13 was the upper extremity, but a specimen was not available, so we used lymph node specimens of metastatic lesions. Mean follow-up duration was 38 months (range, 3--249 months; Table [1](#Tab1){ref-type="table"}).Table 1Summary of clinical, immunohistochemical, and genetic findings of CCSST and CCSLGT casesNo.Age (years) /SexLocationTumor size (cm)Outcome (months)ImmunohistochemistryFluorescence in situ hybridization (%)Expected fusion genes by FISHRT-PCR findingsS-100 proteinMelan-A and/or HMB45SOX-10EWSR1ATF1CREB1CREM129/FLeg5DOD (73)++NP72482NP*EWSR1-ATF1*NP241/MLeg4DOD (62)++NP68NDNDNP*EWSR1*- (unknown partner)NP325/MLeg8DOD (29)++NP50324NP*EWSR1-ATF1EWSR1* exon 8-*ATF1* exon 4462/FLeg2NED (6)++NP860NPUnknown*EWSR1* exon 8-*ATF1* exon 4533/MLeg2.5DOD (8)++NP1600NP*EWSR1*- (unknown partner)*EWSR1* exon 10-*ATF1* exon 5634/FLegNAAWD (16)++NP62320NP*EWSR1-ATF1*NP734/MLeg6AWD (9)++NP76666NP*EWSR1-ATF1*NP812/FLeg2NED (96)+----36022*EWSR1*-(unknown partner)NP939/FArm3DOD (10)++NP7454NDNP*EWSR1-ATF1*NP1069/FArm2DOD (26)++NP50580NP*EWSR1-ATF1EWSR1* exon 8-*ATF1* exon 41141/MArm2.5AWD (11)++NP58582NP*EWSR1-ATF1EWSR1* exon 10-*ATF1* exon 51240/MArm5.5AWD (23)++NP70368NP*EWSR1-ATF1EWSR1* exon 8-*ATF1* exon 51349/FArm4AWD (24)++NP708436*EWSR1-CREM*\*1456/FArm3.9AWD (3)++NP808620*EWSR1-ATF1*NP158/FAbdominal wall9.5DOD (249)++NP50360NP*EWSR1-ATF1*NP1643/MSkin3NED (29)+--NP84760NP*EWSR1-ATF1*NP1741/FIleum4NED (28)+--NP72582NP*EWSR1-ATF1*NP1878/FIleum9NED (48)+--+22012(unknown partner)-*CREM*NP1938/FSmall intestine10DOD (17)+--NP56642NP*EWSR1-CREB1*NP2020/FIleum4NED (7)+--+80944NP*EWSR1-ATF1EWSR1* exon 8-*ATF1* exon 42147/FIleum4.5AWD (34)+--+74062NP*EWSR1-CREB1EWSR1* exon 7-*CREB1* exon 72257/MEsophagus3NED (18)+--+2208UnknownNP*+* Positive, − Negative, *AWD* Alive with disease, *NED* No evidence of disease, *DOD* Died of disease, *NA* Not available, *ND* Not detected, *NP* Not performed, *FISH* Fluorescence in situ hybridization, *RT-PCR* Reverse transcription PCR; \*, fusion gene was not detected by RT-PCR in formalin-fixed, paraffin-embedded materials Histological and IHC findings {#Sec9} ----------------------------- Histologically, the majority of CCSSTs showed sheet-like and nested growth patterns of epithelioid and/or spindle tumor cells that had round to mildly irregularly shaped nuclei with conspicuous nucleoli and clear to pale eosinophilic or amphophilic cytoplasm (Fig. [1a,](#Fig1){ref-type="fig"}and [b](#Fig1){ref-type="fig"}). Among the 16 CCSSTs, 13 cases showed predominant epithelioid cytology and 3 cases exhibited predominant spindle cytology. Multinucleated giant cells were sparsely found in 14 cases. Visible melanocytic differentiation (melanin pigmentation) was observed in 4 cases. The predominant architecture was sheet-like and nested in 12 cases, fascicular in 3 cases, and pseudo-papillary in 1 case. Nucleolar prominence was found in 14 cases. Case 13, which had both *EWSR1* and *CREM* rearrangements (metastatic lesions in lymph nodes), showed sheet-like and fascicular proliferation of epithelioid and spindle tumor cells that had round nuclei with conspicuous nucleoli and abundant pale eosinophilic cytoplasm (Fig. [2a,](#Fig2){ref-type="fig"}and [b](#Fig2){ref-type="fig"}); no myxoid change was found. On the other hand, CCSLGTs often exhibited pseudo-papillary patterns of epithelioid tumor cells, with round to irregular-shaped nuclei showing a coarse chromatin pattern and having a slightly eosinophilic to less clear cytoplasm (Fig. [3a,](#Fig3){ref-type="fig"}and [b](#Fig3){ref-type="fig"}). All 6 CCSLGT cases showed predominant epithelioid cytology and did not exhibit visible melanocytic differentiation. The predominant architecture was pseudo-papillary in 3 cases and sheet-like in 3 cases. Nucleoli were less conspicuous than those from CCSSTs (Fig. [3c](#Fig3){ref-type="fig"}), and nucleolar prominence was observed in 2 cases. Osteoclast-type giant cells were scattered in 5 cases (Fig. [3d](#Fig3){ref-type="fig"}). Case 18, which had *CREM* rearrangement, showed pseudo-papillary proliferation of epithelioid cells that had round nuclei without conspicuous nucleoli and pale eosinophilic cytoplasm. There were no histologic patterns indicating any correlation between *ATF1* and *CREB1* or indicating cases with *CREM* rearrangement.Fig. 1Pathological findings of CCSST with *EWSR1* and *ATF1* rearrangements. **a** CCSSTs showed sheet-like and nested growth patterns of polyhedral tumor cells (200×). **b** Tumor cells had round nuclei with conspicuous nucleoli and clear to pale eosinophilic cytoplasm (400×). **c** FISH of *EWSR1* split signals. Tumor cells showed *EWSR1* split signal with a pair of fused (arrow) and split (arrow head) patterns (1000×). **d** FISH of *ATF1* split signals. Tumor cells showed *EWSR1* split signal with a pair of fused (arrow) and split (arrow head) patterns (1000×)Fig. 2Pathological findings of CCSST with *EWSR1* and *CREM* rearrangements. **a** Metastatic CCSST of the lymph nodes (100×). **b** Tumor cells were polyhedral to spindle-shaped and had oval to round nuclei with pale to clear and eosinophilic cytoplasm (400×). **c** FISH of *EWSR1* split signals. Tumor cells showed *EWSR1* split signal with a pair of fused (arrow) and split (arrow head) patterns (1000×). **d** FISH of *CREM* split signals. Tumor cells showed *CREM* split signal with a pair of fused (arrow) and split (arrow head) patterns (1000×)Fig. 3Pathological findings of CCSLGT with *EWSR1* and *CREB1* rearrangements. **a** CCSLGTs often exhibited sheet-like proliferation of polyhedral and epithelioid tumor cells (200×). **b** Areas of pseudo-papillary growth pattern around the vasculatures were also observed (200×). **c** Tumor cells had mildly irregular-shaped round nuclei showing a coarse chromatin pattern and lightly eosinophilic to less frequently clear cytoplasm. Nucleoli were inconspicuous (400×). **d** Osteoclast-type giant cells were scattered in the tumor (400×). **e**. FISH of *EWSR1* split signals. Tumor cells showed *EWSR1* split signal with a pair of fused (arrow) and split (arrow head) patterns (1000×). **f**. FISH of *CREB1* split signals. Tumor cells showed *CREB1* split signal with a pair of fused (arrow) and split (arrow head) patterns (1000×) We performed IHC of melanocytic markers, including S-100 protein, Melan-A, HMB45, and SOX-10, and assessed their positivity (Table [1](#Tab1){ref-type="table"}). In Cases 8 and 16, CCS of the deep soft tissue and skin showed no reactivity with Melan-A and HMB45 and Case 13 with both *EWSR1* and *CREM* rearrangements was negative for Melan-A, but almost all of the CCSST cases were positive for all melanocytic markers. In contrast, CCSLGTs showed no reactivity with any melanocytic markers and were positive for only S-100 and SOX-10. All of the IHC results were compatible with the pathological diagnosis of CCSST and CCSLGT. Fluorescence *in situ* hybridization {#Sec10} ------------------------------------ As shown in Table [1](#Tab1){ref-type="table"}, 15 of the 16 CCSST cases (93.8%) had *EWSR1* rearrangement, of which 11 (68.8%) also showed *ATF1* rearrangement (Fig. [1c,](#Fig1){ref-type="fig"}and [d](#Fig1){ref-type="fig"}), suggestive of the presence of *EWSR1-ATF1* fusion. One CCSST (Case 13) (6.3%) exhibited *EWSR1* and *CREM* rearrangements (Fig. [2c,](#Fig2){ref-type="fig"}and [d](#Fig2){ref-type="fig"}), indicating *EWSR1-CREM* fusion although no fusion gene was proven by RT-PCR using formalin-fixed, paraffin-embedded materials (data not shown). Two CCSSTs with *EWSR1* rearrangement had no partner genes: one (Case 8) showed no rearrangement of *ATF1*, *CREB1*, or *CRE*M, and one (Case 2) did not exhibit any rearrangement. Four of 6 CCSLGT cases (66.7%) exhibited *EWSR1* rearrangement: 2 (33.3%) showed *ATF1* rearrangement and the other 2 (33.3%) showed *CREB1* rearrangement (Fig. [3e,](#Fig3){ref-type="fig"}and [f](#Fig3){ref-type="fig"}), suggestive of the presence of *EWSR1-ATF1* and *EWSR1-CREB1* fusions, respectively. One case (Case 18) was positive for split *CREM* signals, although no partner genes were detected. One case (Case 22) showed no rearrangement of *EWSR1*, *ATF1*, *CREB1*, or *CREM*. Reverse transcription-polymerase chain reaction {#Sec11} ----------------------------------------------- RT-PCR was performed in 8 available cases including 6 CCSSTs and 2 CCSLGTs (Table [1](#Tab1){ref-type="table"}). All 6 CCSST cases showed *EWSR1-ATF1* fusion between *EWSR1* exon 8 and *ATF1* exon 4, *EWSR1* exon 10 and *ATF1* exon 5, or *EWSR1* exon 8 and *ATF1* exon 5. Although *EWSR1-ATF1* fusion was confirmed by RT-PCR in 2 cases (Cases 4 and 5), adequate *EWSR1* or *ATF1* split signals were not detected by FISH. Among the 2 CCSLGT cases, one had *EWSR1-ATF1* fusion between *EWSR1* exon 8 and *ATF1* exon 4, and the other had *EWSR1-CREB1* fusion between *EWSR1* exon 7 and *CREB1* exon 7. Discussion {#Sec12} ========== Although CCSST and CCSLGT share similar pathological findings, there are apparent morphological and immunohistochemical differences between the two tumor types. We confirmed the differences in histology and IHC results in our cohort cases. Histologically, the cytological findings of tumor cells and architectural proliferation pattern differed. The tumor cells of CCSST were polyhedral to epithelioid, and spindle-shaped with round to mildly irregular-shaped nuclei and conspicuous nucleoli. In contrast, the tumor cells of CCSLGT had epithelioid tumor cells with irregular-shaped nuclei showing a coarse chromatin pattern and more eosinophilic cytoplasm. Nucleoli were not remarkable in CCSLGT compared to CCSST. CCSST exhibited sheet-like, solid, and nested tumor cell proliferation. In contrast, CCSLGT additionally showed a pseudo-papillary growth pattern. The existence of scattered osteoclast-type giant cells was also characteristic of CCSLGT. CCSSTs were positive for Melan-A and/or HMB45 melanocytic markers in addition to S-100 protein, as determined by IHC. On the other hand, CCSLGTs were not reactive for any melanocytic markers, with the exception of S-100 protein and SOX-10. As in previously reported studies, *EWSR1-CREB1* fusion tended to be detected in CCSLGTs. This genetic tendency might reflect the morphological and immunohistochemical differences between the two tumor types. The novel finding of the study was that we discovered *CREM* rearrangement in a few CCSs. Kao et al. \[[@CR14]--[@CR16]\] stated that an *EWSR1-CREM* fusion was previously detected by RNA sequencing in 2 melanoma cell lines (CHL-1 and COLO 699) and proposed that these cell lines may have originated from CCS because of the histological and immunohistochemical overlap between malignant melanoma and CCS. On the other hand, *EWSR1-CREM* fusion was found in a unique myxoid mesenchymal tumor that was recently described as a new entity \[[@CR14], [@CR17]\]. This myxoid tumor is thought to have an intracranial location, and 8 cases have previously been reported, of which 7 occurred in intracranial lesions like meninges, brain tumors, and ventricles, and one case arose in the pelvic/perirectal region. A genetic study revealed *EWSR1* fusions with *CREB* family genes in all of these tumors. Among the 8 tumors, 3 had *EWSR1-CREM* fusion, 4 had *EWSR1-CREB1* fusion, and one tumor showed *EWSR1-ATF1* fusion. However, histological and immunohistochemical findings completely differed between this particular myxoid mesenchymal tumor and CCSST/CCSLGT, and interestingly, these genetic results corresponded to those of CCSST and CCSLGT. CCSLGT was originally reported as an "osteoclast-rich tumor of the gastrointestinal tract with features resembling clear cell sarcoma of the soft parts" \[[@CR2]\]. However, some authors prefer to refer to CCSLGT as a "malignant gastrointestinal neuroectodermal tumor" (GNET), because these tumors lack evidence of melanocytic differentiation \[[@CR5]\]. A recent review discussed the relationship between clear cell sarcoma of the gastrointestinal tract (CCS-GIT) with GNET. There were differences in morphology and IHC findings between CCS-GIT and GNET. GNET tended to show a wider spectrum of growth patterns, including a pseudo-papillary growth pattern, and exhibited no evidence of melanocytic differentiation. Clinically, CCS-GIT affected males more often than females, and GNET occurred in younger patients although no significant differences existed in their biological behaviors \[[@CR18]\]. It has been reported that GNET has poorer prognosis than CCS-GIT \[[@CR5]\], but additional studies are needed to clarify the relationship between the two entities. While a previous study revealed that *EWSR1-ATF1* fusion was identified by RT-PCR in 91% of CCSST cases \[[@CR1]\], we detected *EWSR1-ATF1* fusion in 13 of 16 CCSST (81.3%) both by FISH and RT-PCR. Moreover, after excluding 1 case of *EWSR1-CREM* fusion, 13 of 15 CCSST cases (86.7%) had *EWSR1-ATF1* fusion. The percentage of positive cases in the present study nearly reached that of the previous study. CCSLGT has been identified as having *EWSR1-ATF1* or *EWSR1-CREB1* fusion, with CCSLGT more frequently having *EWSR1-CREB1* fusion. These fusions have also been detected in angiomatoid fibrous histiocytoma and primary pulmonary myxoid sarcoma despite the morphological and immunohistochemical differences between CCSST/CCSLGT and these tumors. To detect specific fusions, FISH is an effective and useful tool using formalin-fixed, paraffin-embedded sections in routine pathological work. In this study, we successfully detected *EWSR1*, *ATF1*, *CREB1*, and *CREM* rearrangements by FISH in the majority of cases. Some cases showed only one specific rearrangement and did not reveal any rearrangement of partner genes. In such cases, it is expected that certain unknown partner genes can form some novel chimeric genes, and that powerful analytic tools such as next-generation sequencing will be useful for detecting these novel fusions. Conclusions {#Sec13} =========== Rearrangements of *EWSR1* and *ATF1* or *EWSR1-ATF1* fusion were predominantly found in CCSST, whereas those of *EWSR1* and *CREB1* or *EWSR1-CREB1* tended to be detected in CCSLGT. We detected a novel *CREM* rearrangement in surgical CCSST specimens by FISH. Although this novel rearrangement occurred in a minority of CCSST cases, further studies are needed to elucidate its pathological significance. CCS : Clear cell sarcoma CCSLGT : Clear cell sarcoma-like gastrointestinal tumor CCSST : Clear cell sarcoma of soft tissue FISH : Fluorescence *in situ* hybridization IHC : Immunohistochemistry The authors thank the following pathologists for kindly contributing case materials and clinical follow-up information: Akiko Tonooka, Department of Pathology, Tokyo Metropolitan Cancer and Infectious Diseases Center Komagome Hospital, Tokyo, Japan; Misa Ishihara and Kimio Hashimoto, Department of Pathology, Nishi-Kobe Medical Center, Hyogo, Japan; Shigeo Hara, Department of Diagnostic Pathology, Kobe City Medical Center General Hospital, Hyogo, Japan; Koki Moriyoshi, Division of Clinical Pathology, National Hospital Organization Kyoto Medical Center, Kyoto, Japan; Shin Ichihara, Department of Surgical Pathology, Sapporo Kosei General Hospital, Hokkaido, Japan; Yukio Morishita, Department of Pathology, Tokyo Medical University Ibaraki Medical Center, Ibaraki, Japan; and Atsushi Uchida, Department of Pathology, Tsukuba Medical Center Hospital, Ibaraki, Japan. Availability of data and material {#FPar1} ================================= All data were presented in this paper, and there are no additional supporting files. KS participated in the design of the study, performed the pathological analysis, and drafted the manuscript. SS, TH, TS, YI, MT, and HF helped with the pathological analysis. TA and HA conducted the fluorescence *in situ* hybridization. TK performed genetic analysis using available surgical materials. ME examined the clinical data of cases. SS conceived the study, participated in its design and coordination, and helped draft the manuscript. All authors read and approved the final manuscript. Ethics approval and consent to participate {#FPar2} ========================================== All patients, including legal parent or guardian of minor patients, gave general consent for the use of their tissue/data in research as authorized by the Institutional Review Board of Sapporo Medical University Hospital (No. 292-3012). Consent for publication {#FPar3} ======================= Not applicable. Competing interests {#FPar4} =================== The authors declare that they have no competing interests. Publisher's Note {#FPar5} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

T. W. Ford, P. A. Kirkwood. Sympathetic discharges in intercostal and abdominal nerves. Physiol Rep, 6 (11), 2018, e13740, <https://doi.org/10.14814/phy2.13740> **Funding Information** This experimental work was supported by the Jeanne Anderson Fund (Institute of Neurology), the Wellcome Trust (038027/Z/93/Z/1.5), the Canadian MRC and the International Spinal Research Trust (STR 032). Introduction {#phy213740-sec-0001} ============ It is well recognized that the naturally occurring efferent discharges of intercostal nerves in an anesthetized or decerebrate animal may include discharges of both alpha and gamma motoneurones (Eklund et al. [1964](#phy213740-bib-0008){ref-type="ref"}; Sears [1964b](#phy213740-bib-0027){ref-type="ref"}; De Troyer et al. [2005](#phy213740-bib-0006){ref-type="ref"}). However, there is a third category of efferent discharges usually present in muscle nerves, those arising from the sympathetic innervation. To our knowledge, the discharges of these efferents have not been described for the intercostal nerves. Furthermore, there are few descriptions of muscle nerve sympathetic discharges recorded during ongoing somatic motoneurone activity (for refs see Boulton et al. ([2016](#phy213740-bib-0004){ref-type="ref"})). The intercostal nerves provide a unique set of nerve branches for investigation of sympathetic activity under such conditions. These branches, in a range of sizes, innervate various intercostal and abdominal muscles with different patterns of respiratory drive. The nature of the intercostal nerve sympathetic discharges became a matter of interest for us during current work, in which we are reanalyzing cross‐correlation measurements of connections from expiratory bulbospinal neurones to thoracic motoneurones following spinal cord lesions (Ford and Kirkwood [1995](#phy213740-bib-0010){ref-type="ref"}), with a view to separately identifying connections to gamma and alpha motoneurones on the basis of efferent spike size. Separation of gamma from alpha discharges (i.e., the choice of the upper border of spike sizes for the gamma population) may be made by reference to the criteria used by Sears ([1964b](#phy213740-bib-0027){ref-type="ref"}), but the lower size limit for the gamma population seems to be more of a problem. Inspection of recordings of the discharges at high gain revealed a dense population of very small spikes, down to the limit of discrimination from noise. It seemed likely that these discharges were those of postganglionic sympathetic efferents, but since our investigations were being made post hoc, from existing recordings, there was no possibility of identifying them by any experimental procedure. We have therefore looked for their cardiac modulation. This is a well‐known property, indeed often an identifying criterion, in other studies of muscle nerve sympathetic outflow (e.g., Boulton et al. [2014](#phy213740-bib-0003){ref-type="ref"}). We find that such a modulation was indeed present for these spikes for all the categories of nerves, though of a relatively modest strength. A particularly interesting feature of the presumed sympathetic discharges thus identified was that their amplitudes for the largest nerves were found to approach those of the gamma spikes. Material and Methods {#phy213740-sec-0002} ==================== The recordings examined came from experiments for which the main results have already been reported and were conducted according to UK legislation \[Animals (Scientific Procedures) Act 1986\] under Project and Personal Licences issued by the UK Home Office. The data came from 17 cats of either sex, weighing 2.5--3.7 kg. Twelve of these (vagotomized) came from the experiments reported by Saywell et al. ([2007](#phy213740-bib-0024){ref-type="ref"}, [2011](#phy213740-bib-0025){ref-type="ref"}) and the other 5 (vagus nerves intact) from those reported by Road et al. ([2013](#phy213740-bib-0022){ref-type="ref"}). The animals were anesthetized with sodium pentobarbitone (initial dose 37.5 mg kg^--1^ I.P., then I.V. as required). Neuromuscular blockade was achieved by using gallamine triethiodide (subsequent to surgery, I.V., repeated doses 24 mg as required) and the animals were artificially ventilated via a tracheal cannula with oxygen‐enriched air, to bring the end‐tidal CO~2~ fraction initially to about 4%. CO~2~ was then added to the gas mixture to raise the end‐tidal level sufficient to give a brisk respiratory discharge in the mid‐thoracic intercostal nerves (typically 6--7%). During neuromuscular blockade, anesthesia was assessed by continuous observations of the patterns of the respiratory discharges and blood pressure together with responses, if any, of both of these to a noxious pinch of the forepaw. Only minimal, transient responses were allowed before supplements (5 mg kg^--1^) of pentobarbitone were administered. The animal was supported by vertebral clamps, a clamp on the iliac crest and a plate screwed to the skull. Rectal temperature was maintained between 37°C and 38°C by a thermostatically controlled heating blanket. Mean blood pressures, measured via a femoral arterial catheter, were above 80 mmHg throughout. Nerve recordings {#phy213740-sec-0003} ---------------- These were originally made either during spike‐triggered averaging measurements or for cross‐correlation measurements to investigate the connections from expiratory bulbospinal neurones to motoneurones or to spinal interneurones. The recordings were made from the cut central ends of selected nerves via pairs of platinum wire electrodes and with conventional amplification (filter settings 300 Hz to 3 kHz). Thoracic nerves were maintained in a single paraffin oil pool constructed from skin flaps and the lumbar nerves were recorded under petroleum jelly, most often with a piece of thin plastic film separating the electrodes from underlying muscle. These recordings were stored on magnetic tape and subsequently acquired for computer analysis via a 1401 interface and Spike2 software (Cambridge Electronic Design, Cambridge, UK). Sampling rates were around 8--10 kHz, varying according to the number of channels sampled. In the first group of 12 cats, the recordings examined consisted of one external intercostal nerve in either T5 or T6, from one experimental run in each cat, each run being a period (no stimuli being delivered) used for intracellular recording from a motoneurone or an interneurone (1--23 min). To investigate cardiac modulation, an ECG signal was obtained from a cord dorsum recording, originally used for monitoring afferent nerve volleys during the initial testing of the motoneurones or interneurones (Saywell et al. [2011](#phy213740-bib-0025){ref-type="ref"}; Ford et al. [2014](#phy213740-bib-0012){ref-type="ref"}). In the second group of five cats, longer runs of data (52--100 min) originally obtained for cross‐correlation measurements, were available (one run from each cat). Recordings from one external intercostal nerve (T5 or T6) and 3--6 internal intercostal nerves or nerve branches (T8--T11) from the left side of each cat were included. These branches included the following nerves on the left side of T8 and/or T9 (four cats): (1) one of the filaments of the internal intercostal nerve, which are the naturally occurring branches that leave the nerve at intervals to innervate the internal intercostal muscle layer (Sears [1964a](#phy213740-bib-0026){ref-type="ref"}); (2) the lateral branch of the internal intercostal nerve, which innervates external abdominal oblique; (3) the distal remainder of the internal intercostal nerve, which innervates the more distal part of the internal and parasternal intercostal muscles, transversus abdominis, and rectus abdominis (see Meehan et al. [2004](#phy213740-bib-0018){ref-type="ref"} for refs). In the fifth cat, the whole internal intercostal nerves of T9 and of T11 were included, together with those of T9, T10 and T11 on the right side. In addition, in two of the animals, recordings from branches of the L1 ventral ramus were also available, including (in both animals) a branch innervating internal abdominal oblique and a distal remainder (for more details see Road et al. [2013](#phy213740-bib-0022){ref-type="ref"}). In only one of this group of five animals was there a cord dorsum recording that could be used to provide an ECG signal, but in the other four an ECG signal could be derived from one of the nerve recordings, in each case a nerve with a relatively modest efferent discharge. Analysis {#phy213740-sec-0004} -------- This consisted firstly of visual inspection of the efferent discharge signals, concentrating on the smallest spikes viewed at high gain, secondly of the construction of cross‐correlation histograms between efferent spikes in different spike‐amplitude ranges, with *R*‐wave events as reference, over a lag range of ±0.6 sec. *R*‐wave events were obtained by triggering from the ECG signal. For more details, see Ford and Kirkwood ([2018](#phy213740-bib-0011){ref-type="ref"}), which reports similar analyses for the alpha motoneurone spikes in the same recordings. Bin‐widths were usually 4 msec, but in a few instances, where the run length was short or the number of efferent spikes was low, a bin‐width of 10 msec was used instead, and for subsequent quantification, one of 20 msec. Results {#phy213740-sec-0005} ======= External intercostal nerves {#phy213740-sec-0006} --------------------------- An example of an external intercostal nerve recording and its analysis is illustrated in Figure [1](#phy213740-fig-0001){ref-type="fig"}. The main bursts of spikes define inspiration, the largest spikes being derived from alpha motoneurones. Following Sears ([1964b](#phy213740-bib-0027){ref-type="ref"}) gamma motoneurone spikes are also present and need to be differentiated from the alphas, to enable one of the comparisons we wish to make, that between the spikes of sympathetic and of gamma discharges. The criterion we used was a spike amplitude level between alpha and gamma spikes as indicated in Figure [1](#phy213740-fig-0001){ref-type="fig"}, a little below the amplitude of early inspiratory spikes \[Fig. [1](#phy213740-fig-0001){ref-type="fig"}A and B, level 1\]. This criterion is justified in Ford and Kirkwood ([2018](#phy213740-bib-0011){ref-type="ref"}), where exactly the same levels were used in the analyses of the alpha discharges. Seventeen recordings of the discharges from these nerves, from either T5 or T6, were analyzed (each from one cat). {ref-type="ref"}) for more detail). (F), cross‐correlation analyses from the data in (A--E) a, e, cross‐correlation histograms (*R*‐wave reference) for the sympathetic and gamma ranges respectively; b, auto‐correlation histogram for the *R*‐wave events (*R*‐wave period 250 msec); c, central part of the histogram shown in a, for comparison with the *R*‐wave‐triggered average of the nerve signal (d). Number of *R*‐wave events, 3309. Ordinate ranges for a, e, 0.85--1.15 m (m is the mean count calculated over 0.6 sec). Bin widths, 4 msec. The calibrations for (A--E) and for Fd are in the same arbitrary units. The same recording was illustrated in Ford and Kirkwood ([2018](#phy213740-bib-0011){ref-type="ref"}), their Figure [1](#phy213740-fig-0001){ref-type="fig"}, from which Fb and Fd here are reproduced.](PHY2-6-e13740-g001){#phy213740-fig-0001} A second spike amplitude level was also required, to separate gamma spikes from sympathetic discharges, and was also indirect. It was based on the assumption, derived from the experience in this laboratory, that with elevated levels of CO~2~, the gamma discharges in external intercostal nerves cease soon after the start of expiration, that is, very soon after any postinspiratory discharge, which itself sometimes includes a few alpha spikes. We therefore set the level for separating gamma from sympathetic discharges as illustrated in the high gain recording in Figure [1](#phy213740-fig-0001){ref-type="fig"} (B--E, level 2). Note that there was a high density of very small spikes (or spike‐like deflections) with amplitudes below this level, the amplitudes and frequency of these gradually decreasing as expiration progressed. There was also a subjective element in choosing this level. Although, in general, the spike trains of individual units were not separable within the gamma range, from time to time such trains were apparent, especially as the discharges declined in postinspiration. This was never the case within the lower range, which we designated as sympathetic. The spike‐like deflections in this range were not noise transients (if so, they would not have shown the modulation through expiration) and had a relatively consistent time‐course, being diphasic, starting positive going, as for the alpha and gamma deflections. Nevertheless, as one considers progressively smaller amplitudes for these spikes, one gets to the point where they merge into the baseline noise. Another arbitrary choice was therefore made in order to select spikes in this smallest range, as indicated, selecting a level by eye where most level crossings consisted of spike‐like waveforms (Fig. [1](#phy213740-fig-0001){ref-type="fig"}D and E, level 3). For some of the recordings, a more formal justification for the choice of levels was possible and will be described in the final section of Results. The external intercostal nerve is a pure muscle nerve, for which a defining characteristic of its sympathetic discharges is said to be strong cardiac modulation (Jänig et al. [1983](#phy213740-bib-0014){ref-type="ref"}). However, such modulation was not obvious by eye in 16/17 of the recordings, including Figure [1](#phy213740-fig-0001){ref-type="fig"}. We therefore constructed cross‐correlation histograms between the *R*‐wave events and the events defined by the spike‐like deflections with peaks within the sympathetic range (between levels 2 and 3 in Fig. [1](#phy213740-fig-0001){ref-type="fig"}D and E). As illustrated in Figure [1](#phy213740-fig-0001){ref-type="fig"}Fa, a clear modulation with the cardiac cycle, approximately sinusoidal at the cardiac frequency (see the auto‐correlation of the *R*‐wave events in Fig. [1](#phy213740-fig-0001){ref-type="fig"}Fb), was almost always present. In analyses such as these, when events were detected by level crossings close to the baseline, an obvious concern is that an ECG signal could be present in the nerve recording and might contribute to level crossings. In fact, for the 12 experiments of the first group, this was never a problem. An ECG effect was only detected in 3 of these 12 recordings, one of which (typical of these 3) was that illustrated in Figure [1](#phy213740-fig-0001){ref-type="fig"}. The ECG was not visible in the raw data, though it could be detected by averaging the nerve signal, triggered by the *R*‐wave events (Fig. [1](#phy213740-fig-0001){ref-type="fig"}Fd). The *QRS* signal detected was very small, entirely within the baseline noise in Figure [1](#phy213740-fig-0001){ref-type="fig"} (note the voltage scale, the arbitrary units in Fd, being the same as shown for the raw data in A--E). Nevertheless, an effect of this signal in the histogram for the sympathetic discharges appeared as a triphasic feature, corresponding closely to the averaged ECG waveform (see expanded trace in Fc). Although clear, this was a negligible effect, and was one that was easily separable from the rest of the histogram time‐course, as was the case in the other two recordings where a similar artifact appeared. In the second group of five experiments (multiple nerve recordings, no intracellular recordings), ECG waveforms were more common and were larger. It is not known why the two series of experiments differed in this regard. Perhaps it was a result of different earthing (grounding) arrangements in the two series. Despite the larger amplitudes of the ECG waveforms and their much larger effects in the histograms, it remained possible to separate the ECG effects from the near sinusoidal component in the histograms, as is described in a later section (*Internal intercostal nerve filaments*). When this was done, all five of the histograms from the external intercostal nerves in this second group showed a near sinusoidal component, as did 11/12 from the first series. In addition, we routinely constructed similar histograms from spikes with amplitudes in the gamma range. These also frequently demonstrated cardiac modulation, but this always, like the example in Figure [1](#phy213740-fig-0001){ref-type="fig"}Fe, had a lower amplitude than the histogram for the sympathetic range. In Figure [1](#phy213740-fig-0001){ref-type="fig"}Fe, the near sinusoidal waveform also appears noisier. This occurred partly as a result of the lower bin counts, but also because there was a minor contribution of relatively narrow peaks, representing some overlap from the alpha range (Ford and Kirkwood [2018](#phy213740-bib-0011){ref-type="ref"}). Internal intercostal nerves {#phy213740-sec-0007} --------------------------- All of the five recordings from whole internal intercostal nerves that were analyzed came from a single experiment. Figure [2](#phy213740-fig-0002){ref-type="fig"}A and B shows two of these recordings, along with a simultaneous recording from an external intercostal nerve, whose main burst of discharge defined inspiration. As in Kirkwood ([1995](#phy213740-bib-0015){ref-type="ref"}), four out of five of these recordings showed two bursts of discharges, one in expiration, another (rather weaker) in inspiration, together with an apparent tonic background, assumed to include tonic gamma discharges. The spike amplitude levels for the alpha/gamma border were assigned as described by Ford and Kirkwood ([2018](#phy213740-bib-0011){ref-type="ref"}). Setting the gamma/sympathetic border in this animal was then helped by the fact that three of the internal intercostal nerves (including the two illustrated) and the external intercostal nerve all showed cardiac modulation of the sympathetic discharges, clearly visible in the raw data (Fig. [2](#phy213740-fig-0002){ref-type="fig"}B). {ref-type="fig"}. (C) cross‐correlation histograms for the sympathetic range (levels 3--2), as in Fig. [1](#phy213740-fig-0001){ref-type="fig"}Fa. Top two histograms apply to the nerve discharges illustrated in A, B. Bottom two histograms apply to simultaneous recordings from two other internal intercostal nerves on the right side of the animal. 8698 *R*‐wave events. *R*‐wave period, 271 msec. Ordinate ranges 0.8--1.2 m for all histograms. Bin widths 4 msec. The same recordings were illustrated in Ford and Kirkwood ([2018](#phy213740-bib-0011){ref-type="ref"}), their Fig. [2](#phy213740-fig-0002){ref-type="fig"}, where A and B are the same as A and B here, except that different spike amplitude levels were indicated.](PHY2-6-e13740-g002){#phy213740-fig-0002} Cross‐correlation histograms were constructed for each nerve between the *R*‐wave marker and the two spike amplitude ranges, as previously. Each of the four nerves showed cardiac modulation for the sympathetic range, though for one of them the amplitude of this was very small (Fig. [2](#phy213740-fig-0002){ref-type="fig"}C). As was the case for the external intercostal nerves, the histograms for the gamma ranges (not illustrated) showed lower amplitudes of modulation than those for the sympathetic ranges. The fifth internal intercostal nerve recording in this animal (T9 right, not illustrated) showed a tonic discharge with only a very modest respiratory discharge and probably no alpha spikes. It was not considered further. Lateral branch {#phy213740-sec-0008} -------------- Recordings from six of these branches of the internal intercostal nerve (from four cats) were analyzed. These branches innervate external abdominal oblique muscle and have a cutaneous component. Note that in the example shown in Figure [3](#phy213740-fig-0003){ref-type="fig"}A (third trace), as in another 3 of these recordings, the alpha and gamma spikes were readily distinguishable, as in the original illustrations for the filament discharges in Sears ([1964b](#phy213740-bib-0027){ref-type="ref"}). Also, in this particular recording, the gamma discharges were readily distinguished from a different population of smaller spike‐like deflections (not illustrated), which were taken as sympathetic. In this example, cardiac modulation of these events was not obvious by eye, and the histogram (Fig. [3](#phy213740-fig-0003){ref-type="fig"}Ca) shows a relatively low amplitude of cardiac modulation. A similar result was seen for two more of the recordings. However, in two of the other examples (in one animal) cardiac‐related bursts were clear (like those in the distal branch described below) and the relevant histograms showed strong modulation. Sympathetic discharge was not obvious in the remaining recording, but the smallest range of spikes was designated as sympathetic and cardiac modulation observed for this range. In four instances, the gamma ranges gave cardiac modulation (of smaller amplitudes than for their corresponding sympathetic range) and the other two gave noisy histograms (relatively low counts), with weak peaks probably corresponding only to those in their alpha ranges. {ref-type="fig"}. Note cardiac‐related bursts of small spikes in the 4th trace, particularly in late expiration and early inspiration, and in B (\*). (C) cross‐correlation histograms for the sympathetic ranges for each of the three internal intercostal nerve branches illustrated above: a, T8 Lat; b, T8 Dist; c, T8 Fil. 14647 *R*‐wave events. *R*‐wave period, 275 msec. Ordinate ranges for histograms: a, 0.8--1.2 m; b, 0.6--1.6 m; c, 0.3--2.2 m;. Bin widths 4 msec. Some of the same recordings here were illustrated in Ford and Kirkwood ([2018](#phy213740-bib-0011){ref-type="ref"}), their Figure [3](#phy213740-fig-0003){ref-type="fig"}, where the first 4 traces in A here also appeared.](PHY2-6-e13740-g003){#phy213740-fig-0003} Distal branch {#phy213740-sec-0009} ------------- Six recordings from the distal remainder of the internal intercostal nerve (Saywell et al. [2007](#phy213740-bib-0024){ref-type="ref"}) were examined, in four cats. Several intercostal and abdominal muscles are innervated by this branch (see Methods), which also has a cutaneous component. Both an inspiratory burst of spikes, assumed to be destined for the parasternal muscle (Taylor [1960](#phy213740-bib-0029){ref-type="ref"}; De Troyer et al. [2005](#phy213740-bib-0006){ref-type="ref"}), and an expiratory burst were usually present, but the distinction between alpha and gamma recordings was generally less clear than for the lateral branch. In four of these recordings (two in each of two cats), very clear cardiac‐related bursts were visible in the smallest range of spikes. As was typical, in the example shown in Figure [3](#phy213740-fig-0003){ref-type="fig"}, the largest of the spike‐like waveforms in the cardiac bursts had amplitudes not much below the level assumed for the alpha/gamma border. The lower level for the gamma range was set a little below the largest events in the cardiac bursts, some overlap between the sympathetic spikes and the gamma spikes being accepted. The histogram in Figure [3](#phy213740-fig-0003){ref-type="fig"}Cb shows, not surprisingly, strong cardiac modulation for the sympathetic range. The four recordings with visible cardiac bursts all gave similar strong modulation for the sympathetic range and the other two rather weaker effects. The gamma ranges again showed weaker effects than for the sympathetic range. Lumbar (L1) nerves {#phy213740-sec-0010} ------------------ An ECG signal was obtainable in two cats investigated for L1 discharges, allowing analyses of recordings from 2 nerve branches in each, one innervating internal abdominal oblique muscle and one a more distal remainder (Road et al. [2013](#phy213740-bib-0022){ref-type="ref"}). Phasic expiratory alpha and gamma spikes were distinguished for each of these, the distinction being particularly clear for the internal abdominal oblique branch, like the T8 lateral branch in Figure [3](#phy213740-fig-0003){ref-type="fig"}. In all four recordings, smaller spikes (presumed sympathetic) were also visible, so three ranges similar to those used for the intercostal nerves were defined for all nerves. For two of the nerves (both in one cat) no modulation was detectable in the histograms for the possible sympathetic level, but the other two both gave clear effects. For the gamma ranges, the two nerves that showed cardiac modulation showed weaker versions, the other two showed none (except for one, an effect similar to its alpha range, but weaker). Internal intercostal nerve filaments {#phy213740-sec-0011} ------------------------------------ These are of particular interest because of the original observations of Sears ([1964b](#phy213740-bib-0027){ref-type="ref"}), where the efferent spikes were noted as having "two distinct sizes", neither comprising sympathetic efferents, because they survived section of the gray ramus communicans. The records analyzed here were also composed of spikes of two distinct size ranges (Figs. [3](#phy213740-fig-0003){ref-type="fig"}, [4](#phy213740-fig-0004){ref-type="fig"}). Further, at the normal recording gain, in contrast to the other recordings described above, it was hard to detect any smaller spikes above the baseline noise. However, when the recordings were examined at higher gain, very small spikes became apparent, including cardiac‐related bursts (Fig. [3](#phy213740-fig-0003){ref-type="fig"}B), and were designated as sympathetic (see indicated amplitude ranges). In the histogram for this example, this range gave a clear cardiac modulation, together with an artifactual peak at zero lag, related to the *R*‐wave of the ECG (Fig. [3](#phy213740-fig-0003){ref-type="fig"}Cc). The ECG itself was not visible in the raw data from this filament, but was apparent in the *R*‐wave related average (not illustrated). The rather sharp top of the waveform probably also represents an artifact related to the *P*‐wave (see below, this section), but this effect is minor. {ref-type="fig"}. (C) a, c--e, cross‐correlation histograms from the data of A,B: a, sympathetic range (levels 3--2); c, gamma range (levels 2--1); d, histogram derived from negative‐going spikes (levels 4--5): note that the ECG artifacts in this histogram are sign‐reversed compared to those in a, including components related to both the *P*‐ and *T*‐waves (arrowed). e, average of the histograms in a and d: note that the components for the *P*‐ and *T*‐ waves cancel out almost entirely, revealing the underlying sinusoid‐like modulation. b, *R*‐wave‐triggered average of the filament recording: the *P*‐wave is clear, the *T*‐wave just detectable. Calibration in the same arbitrary units as in (A) and (B). 6696 *R*‐wave events. *R*‐wave period, 254 msec. f, cross‐correlation histogram from the gamma range of the filament recording in Figure [3](#phy213740-fig-0003){ref-type="fig"} (typical for the gamma range of filament recordings). Histogram ordinate ranges, 0.6--1.5 m. Bin widths 4 msec.](PHY2-6-e13740-g004){#phy213740-fig-0004} Six recordings of filament discharges (including this one) were analyzed. Four (in two cats) were similar to this one, with very small spikes showing cardiac bursts and strong cardiac modulation in the histograms for the sympathetic range. The other two recordings showed very small spikes without visible cardiac bursts (one described below) both giving relatively weak cardiac modulation. For the gamma ranges, five recordings gave flat or almost flat histograms and the other is described below. Figure [4](#phy213740-fig-0004){ref-type="fig"} shows the analysis of this other filament recording, which is of interest firstly because it is one of the recordings where an ECG waveform was clear in the original recording. At high gain, this may be seen to have an amplitude comparable to that of the very small spikes, which this time did not show obvious cardiac bursts. Nevertheless, cardiac modulation of these very small spikes (see illustrated levels) could be demonstrated in their histogram, though to do this, care was needed to separate rather strong ECG‐related artifacts. The *QRS* artifact is obvious (−4 to +12 msec, truncated in Fig. [4](#phy213740-fig-0004){ref-type="fig"}Ca), but there were also artifactual components related to the *P* and the *T* waves (compare latencies of the marked components in the histogram with the *R*‐wave triggered average waveform in Fig. [4](#phy213740-fig-0004){ref-type="fig"}Cb). To differentiate the cardiac modulation of the nerve discharges from these two, a second histogram was calculated (Fig. [4](#phy213740-fig-0004){ref-type="fig"}Cd), selecting spikes by triggering with the opposite polarity and choosing an amplitude range as near as possibly equivalent to the previous triggers (see levels in Fig. [4](#phy213740-fig-0004){ref-type="fig"}B), so as to give approximately the same value of mean counts in the histogram. The biphasic artifact corresponding to the *P*‐wave and the monophasic one relating to the *T*‐wave were reversed in this new histogram, whereas the underlying wide peak was more‐or‐less constant. The process generating these artifacts was reasonably linear, so that when an average of the two histograms (a and d) was constructed (e), the artifacts almost disappeared, leaving a clear representation of the wide peak (plus the *QRS* artifact). Similar discrimination against ECG artifacts was required for the sympathetic range for five of the filament recordings, for 3 of the major internal intercostal nerve branches and for two of the external intercostal nerves. This recording is also of interest because the histogram for the gamma range here (Fig. [4](#phy213740-fig-0004){ref-type="fig"}Cc) showed clear cardiac modulation. In many of the recordings described above, the gamma range may have shown similar modulation, but because it was weaker than for the accompanying modulation for the sympathetic range, it would be easy to hypothesize that the occurrence of gamma spikes was independent of the cardiac cycle and the modulation was present only because of overlapping amplitudes for the spikes of these two ranges. This would be consistent with the more‐or‐less flat histograms for five‐sixth of the filament recordings (e.g. Fig. [4](#phy213740-fig-0004){ref-type="fig"}Cf). However, for the filament recording represented in the rest of Figure [4](#phy213740-fig-0004){ref-type="fig"}, the gamma spikes were very distinctly discriminated from both the sympathetic and alpha ranges (Fig. [4](#phy213740-fig-0004){ref-type="fig"}A and B). Nevertheless, the cardiac modulation for the gamma range (Fig. [4](#phy213740-fig-0004){ref-type="fig"}Cc) was of similar amplitude to that for the sympathetic range. It is hard to imagine that this effect for the gamma range could have arisen via what could only have been a low level of contamination from the sympathetic range. Quantitative summaries {#phy213740-sec-0012} ---------------------- Lags and amplitudes of the histogram waveforms for the sympathetic ranges were measured using a bin‐width of 20 ms, to give some smoothing of an otherwise noisy signal. An exception to this was made for those like Figure [4](#phy213740-fig-0004){ref-type="fig"}, where a by‐eye interpolation on the histograms in 4 msec bins was used instead, so as to maintain discrimination against residual ECG artifacts. Amplitudes were measured trough‐to‐peak during the first cardiac cycle with a positive lag and expressed as a multiple of the mean count (*m*), measured over ±0.6 sec. Lags were measured to the maximum value for that cycle. The procedure is illustrated in Figure [5](#phy213740-fig-0005){ref-type="fig"}A and the measurements summarized in Figure [5](#phy213740-fig-0005){ref-type="fig"}B. The cardiac cycle times, estimated from the *R*‐wave autocorrelation histograms, ranged from 232 to 330 msec. Lags were relatively consistent, with two exceptions. First, when the two groups of experiments for the external intercostal nerve were considered separately (first two rows in Fig. [5](#phy213740-fig-0005){ref-type="fig"}B), the lags for the first group (vagotomised, first row, median 230 msec) tended to be longer than those for the second group (vagi intact, second row, median 190 msec). Second, the lags for the two examples from the lumbar nerves (330 and 360 msec) were the greatest observed. These compare with the mean value (±SD) for the rest of thoracic population of 200 ± 22 msec (*n* = 27, vagi intact preparations only) or with the mean for the thoracic nerves in the same animal (215 msec). {ref-type="fig"} used for the example in A.](PHY2-6-e13740-g005){#phy213740-fig-0005} The amplitudes showed considerable variation, 0.02*--*1.45 m (note the log scale for amplitude in Fig. [5](#phy213740-fig-0005){ref-type="fig"}B). The recordings where cardiac modulation was visible by eye showed the largest amplitude (filled areas). There were no obvious differences between the values for the different nerves, with the one exception that a difference was detected for the external intercostal nerve for the two different preparations investigated. The amplitudes for the first group (vagotomised, top row in Fig. [5](#phy213740-fig-0005){ref-type="fig"}B, median 0.122 m), were smaller than those for the second group (second row, median 0.179 m). The difference was significant for *P* \< 0.05, (Mann--Whitney, single‐sided). During the course of the above analyses, it became obvious that the spike amplitude levels chosen to separate the different efferent groups varied systematically with the nerve categories. These levels are summarized in Figure [6](#phy213740-fig-0006){ref-type="fig"}A, measured from the mean of the baseline recording, then expressed as a percentage of the amplitude of the largest alpha spikes. It should be emphasized that, at this stage, this is not strictly an experimental result, but a summary of the values chosen on the basis of subjective judgements, as described above. Also, the values are approximate, since the amplitude of the largest alpha spikes varied considerably from breath to breath and sometimes also more slowly with time. The few individual outlying spikes with especially large amplitudes (most likely arising by superposition) were excluded and we took an approximate by‐eye average to allow for the rest of the variation. For any one nerve category, the three boundary amplitude levels appeared to vary to a great extent in parallel between the different recordings, suggesting an underlying strong effect from the largest alpha spike amplitude. This probably represents variation in recruitment of alpha motoneurones according to anesthetic or CO~2~ levels. Note that the alpha motoneurone discharges in these preparations would generally be considered as strong, on account of the moderate hypercapnia. {ref-type="fig"}, but the plots are arranged in descending order of nerve diameter. (B) distributions of the ratios of the upper limit of sympathetic range to the upper limit of the gamma range. Filled areas, recordings with visible cardiac bursts. (C) values of the ratios shown in B (means ± SD) plotted against nerve diameters measured from Sears ([1964a](#phy213740-bib-0026){ref-type="ref"}) and/or Duron et al. ([1978](#phy213740-bib-0007){ref-type="ref"}) (see text). Although it was possible to put the L1 nerves into the order of diameters (from visual assessment of their diameters during the dissections), no histology was available for these branches, so they do not contribute to the plot in C.](PHY2-6-e13740-g006){#phy213740-fig-0006} It is also obvious that there are differences between the different nerve categories, which have now been arranged in descending order of nerve diameter. One of these has a trivial cause. The thicker nerves must give a smaller spike size, via Ohm\'s law, so the low level for the sympathetic range (the bottom of each bar in Fig. [6](#phy213740-fig-0006){ref-type="fig"}A), which mostly represents discrimination against amplifier noise, appears higher in this plot for the thicker nerves. The more interesting result is that the proportion of total range occupied by the sympathetic range also varies with the nerve category. To escape (at least in part) the influence of the variability in the populations of alpha spikes, we have compared the ratios of the maximum sizes of the spikes in the sympathetic and the gamma ranges across the different nerve categories (Fig. [6](#phy213740-fig-0006){ref-type="fig"}B). A clear sequence is present, where the mean ratio for the internal intercostal nerve group is more than 3 times larger than for the internal intercostal nerve filaments. To help reduce the subjectivity involved in this plot, we have indicated the recordings where cardiac bursts were visible in the raw data (filled bars in B), which are the examples which involved the lowest degree of subjectivity in choosing the amplitude level for the gamma/sympathetic border. These examples are nevertheless typical of the rest in this plot. These results were then further quantified by relating the same measurements to measurements of nerve diameter (Fig. [6](#phy213740-fig-0006){ref-type="fig"}C). Diameters were measured from the illustrations in Sears ([1964a](#phy213740-bib-0026){ref-type="ref"}) or, for two instances (whole internal intercostal and external intercostal nerves), using an average value from Sears ([1964a](#phy213740-bib-0026){ref-type="ref"}) and from Duron et al. ([1978](#phy213740-bib-0007){ref-type="ref"}). Further checks on the appropriateness of the sympathetic amplitude ranges {#phy213740-sec-0013} ------------------------------------------------------------------------- The relationship in Figure [6](#phy213740-fig-0006){ref-type="fig"}C is very clear. Nevertheless, to reduce the subjectivity still further, the cross‐correlation measurements were repeated in selected data, to check how closely the near‐sinusoidal cardiac modulation of the sympathetic range was limited to the particular amplitude ranges chosen. This was done using narrower ranges of spike amplitude, which together covered the previous sympathetic range. Filament recordings were excluded, because the sympathetic spikes were too close to the amplifier noise for such an analysis, but such analysis was in any case unnecessary for these recordings, because a clear separation was obvious between the spikes assigned to the sympathetic or to the gamma ranges. We further excluded: (1) the other recordings like that in Figure [4](#phy213740-fig-0004){ref-type="fig"}, where large ECG artifacts were present; (2) a number of other runs where either the activity levels or the shortness of the runs meant that the histograms derived from narrow spike amplitude ranges would have had too few counts and/or where the cardiac modulation was weak, any of which would have made the histograms too noisy. There then remained 17 recordings to be analyzed, 5/17 of the external intercostal nerve recordings, together with 12/20 from the other intercostal or abdominal nerves. A bin‐width of 20 msec was used, as in Figure [5](#phy213740-fig-0005){ref-type="fig"}, but to reduce noise further, the individual histograms were smoothed by 3‐point adjacent averaging before measuring their modulation amplitudes. Examples of the histograms obtained in this way from one recording (that from the T8 distal branch in Fig. [3](#phy213740-fig-0003){ref-type="fig"}) are included in Figure [7](#phy213740-fig-0007){ref-type="fig"}A and a plot of the modulation amplitudes, related to the spike amplitude levels used, is shown in Figure [7](#phy213740-fig-0007){ref-type="fig"}Ba. The spike amplitudes are normalized by the maximum gamma spike amplitude, that is, by the upper limit of the gamma range. It may be seen that the spike amplitude range where the modulation amplitude is substantial is quite similar to the range originally assigned to be sympathetic (joined thin arrows), though extending to slightly larger spike amplitudes. The same is true for a second example (Fig. [7](#phy213740-fig-0007){ref-type="fig"} Bb, from the T9L recording in Fig. [2](#phy213740-fig-0002){ref-type="fig"}), whereas in a third example (Fig. [7](#phy213740-fig-0007){ref-type="fig"}Bc, from the external intercostal nerve recording in Fig. [1](#phy213740-fig-0001){ref-type="fig"}), the bias is in the opposite direction. Here, the higher modulation amplitudes are also concentrated within the original sympathetic spike amplitude range, but toward the lower end. Not surprisingly, given the previous observations of some cardiac modulation for the gamma range, the modulation was never seen to be zero in these examples. We could not test at higher spike amplitude levels without interference via effects from the (rather different) modulation seen for alpha discharges (Ford and Kirkwood [2018](#phy213740-bib-0011){ref-type="ref"}). {ref-type="fig"} (T8 Dist). Histograms for nonoverlapping ranges are shown, with increasing spike amplitudes going from a to f. Bin‐widths, 20 msec, histograms smoothed by 3‐point adjacent averaging. Histogram ordinate ranges, 0.5--2.0 m. (B): a, plot of modulation amplitude (*q/m*, as in Fig. [5](#phy213740-fig-0005){ref-type="fig"}) measured from the series in A (the six bins in this histogram indicate the ranges of amplitudes used for the six cross‐correlation histograms in (A); b, similar plot from the data of Figure [2](#phy213740-fig-0002){ref-type="fig"} (T9L Int); c, similar plot from the data of Figure [1](#phy213740-fig-0001){ref-type="fig"}. The abscissae show spike amplitudes normalized to the amplitudes at the alpha/gamma border. The ranges originally assigned as sympathetic are indicated by narrow arrows joined by a line. Revised values for the sympathetic/gamma ratio (see text) were estimated as the spike amplitudes in these plots where the cardiac modulation was reduced to 50% of its maximum value (thick arrows). The spike amplitude ranges needed to be wider for the larger amplitudes in order to obtain enough counts in the histograms to maintain a reasonable signal‐to‐noise ratio, the spike rates being lower for the larger amplitude spikes. (C) revised versus original values for the sympathetic/gamma ratio. Nerve designations (as in Fig. [5](#phy213740-fig-0005){ref-type="fig"}) indicated by different symbols. Filled symbols, recordings with visible cardiac bursts. (D) same plot as in Figure [6](#phy213740-fig-0006){ref-type="fig"}C, but with revised values for the sympathetic/gamma ratio substituted for the original ones for the 17 recordings represented in B.](PHY2-6-e13740-g007){#phy213740-fig-0007} The spike amplitude level where the modulation amplitude had fallen to 50% of its maximum (thick arrows in Fig. [7](#phy213740-fig-0007){ref-type="fig"}B) was then chosen as an index, so as to make a general comparison of these analyses with the original sympathetic/gamma ratios used in Figure [6](#phy213740-fig-0006){ref-type="fig"}. This index, "Symp/gamma ratio (revised)" is plotted in Figure [7](#phy213740-fig-0007){ref-type="fig"}C against the original ratios for all 17 of the analyses. For all but 4 of these, the two measures are very close to each other (the line is the line of equality), including examples from all of the nerve categories and including examples both with (filled symbols) and without (open symbols) visible cardiac modulation in the raw data. The other four analyses, all involving high values of the ratio and three of them with visible cardiac modulation in the raw data, showed even higher values for the revised values than for the original assumptions. A resulting re‐plot of Figure [6](#phy213740-fig-0006){ref-type="fig"}C, now using the revised values for the 17 reanalyzed examples, but otherwise the original values, is shown in Figure [7](#phy213740-fig-0007){ref-type="fig"}D. The standard deviations are higher for the internal intercostal nerves and their major branches than previously, but the overall relationship between the nerve diameter and the sympathetic/gamma ratio is just as clear. The slope is rather steeper. Discussion {#phy213740-sec-0014} ========== As far as we know, the direct comparisons here between the discharges of sympathetic and somatic efferent discharges in the same nerves comprise a new result. We need to be certain therefore that the sympathetic discharges are what we claim them to be. First, the general appearance of obvious cardiac bursts in all categories of nerves, are very similar to those reported extensively in the literature. Then, the histograms from those recordings without obvious cardiac bursts were very similar to those with the bursts (for instance the peaks in these two groups had very similar lags and, for each category of nerve, the spikes had similar amplitudes). The analyses in Figure [7](#phy213740-fig-0007){ref-type="fig"} then show that the spike amplitude level used to define the sympathetic discharges were reasonably appropriate (assuming that cardiac modulation is a sufficient identifier), both for the nerves with and the nerves without obvious bursts. Further evidence consistent with our identification of sympathetic discharges is the reduced strength of the modulation for the external intercostal nerves in the first (vagotomised) group of experiments as compared to the second, likely resulting from lower baroreceptor inhibition in the first group, as a consequence of the loss of aortic baroreceptor input. Additional support comes from the measured lags for the histograms from the two abdominal nerves, which were longer than those for the thoracic nerves. About half of the difference, 71 msec, can be ascribed to conduction delays. These were calculated as follows: 50 msec for an extra 50 mm peripheral conduction at 1 msec^‐1^, plus 21 msec for extra central conduction (T8 to L1 being 83 mm in this cat) at 4 msec^‐1^ (McAllen [1986](#phy213740-bib-0016){ref-type="ref"}). The remaining difference should then be ascribed to a longer central delay and/or to even slower peripheral conduction if, as was likely, the temperature of the peripheral nerves in question was appreciably lower than the core temperature (Road et al. [2013](#phy213740-bib-0022){ref-type="ref"}). A somewhat (at first sight) surprising result is the range of spike sizes observed. The very small size of spikes seen in the filaments is close to what one might expect for single C‐fibers, as supported by the following calculation. Stein and Pearson ([1971](#phy213740-bib-0028){ref-type="ref"}), gave a theoretical formula for the amplitude of a spike in an umyelinated fiber as *V(a/b)* ^*2*^, where *V* is the amplitude of the intracellular action potential, *a* is the diameter of the axon and *b* is the diameter of the nerve, this formula being empirically supported by direct measurements (Pearson et al. [1970](#phy213740-bib-0020){ref-type="ref"}). For an action potential of 100 mV in a 1 *μ*m fiber in a filament of 0.1 mm, this predicts a spike of 100 × 10^−4^ mV = 10 *μ*V. However, this would be for a nerve consisting of only unmyelinated fibers. Here we may modify this by assuming that the extracellular current will not flow through the myelinated fibers. We measured the proportion of the nerve cross‐sectional area that these occupy from the fiber diameter spectra for the equivalent nerves illustrated in Sears ([1964a](#phy213740-bib-0026){ref-type="ref"}). For the 4 nerve categories included in Figure [6](#phy213740-fig-0006){ref-type="fig"}C (excluding the filament), the average proportional area of these was 66%. The predicted spike amplitude should therefore be increased by a factor of 3, that is to 30 *μ*V. The amplitudes of the gamma spikes in the recordings of Sears ([1964b](#phy213740-bib-0027){ref-type="ref"}), made under identical conditions to ours, were around 200 *μ*V, meaning that the value for the maximum sympathetic amplitude for filaments here, 0.18 times the gamma amplitude (indicated by the lowest point in Fig. [6](#phy213740-fig-0006){ref-type="fig"}C or [7](#phy213740-fig-0007){ref-type="fig"}D), corresponds to around 36 *μ*V, close to the prediction. The apparently much larger spike‐like events in the other nerve recordings, which overlapped in amplitude with the gamma spikes are then most likely to have arisen by superposition of unit spikes. This is well‐recognized to occur for sympathetic discharges. In the very first recordings, Adrian et al. ([1932](#phy213740-bib-0001){ref-type="ref"}) referred to these discharges as "waves", assuming a multi‐unit origin. McAllen and Malpas ([1997](#phy213740-bib-0017){ref-type="ref"}) also pointed this out, writing, "Fortunately, however, their ongoing activity can be measured from whole‐nerve recordings because large numbers of nerve fibers fire action potentials at the same time to give "bursts" or waves of summed spikes". Other early authors illustrated the phenomenon, a particularly clear example being Taylor and Gebber ([1975](#phy213740-bib-0030){ref-type="ref"}), where their Figure [1](#phy213740-fig-0001){ref-type="fig"} shows how, for one such single burst, the commonly recorded high‐pass filtered signal, which has the appearance of a burst of spikes, actually represents the filtered noise on the summit of a relatively smooth wave formed from a very much greater number of summed spikes. This must have been the same in our recordings, although we are now making this assertion with respect to a tonic discharge, rather than a burst (the tonic rate nevertheless modulated by the cardiac and respiratory rhythms). Thus, we are saying that the spike‐like events in Figure [1](#phy213740-fig-0001){ref-type="fig"}, whose amplitude and rate decrease during the course of expiration, are not unit spikes but (neural) noise transients made to look like spikes by the filters in our amplifiers. In these terms, the relationship between the relative amplitudes of these events and the nerve diameters can be predicted, as follows. We assume each sympathetic axon action potential gives a unit‐sized current flowing in the extracellular tissue between the electrodes, and that large numbers of these units summate with random occurrence. This summated signal will have Poisson‐like statistics, so the standard deviation of the amplitudes of the resulting transients in the signal may approximate to being proportional to the square root of the mean number of events in any time interval. The amplitudes we chose as representing the maximum for the sympathetic range are most likely to represent a multiple of the standard deviation. Assuming similar firing rates in the different fibers, the "maximum" amplitudes of these current transients will therefore be proportional to the square root of the number of sympathetic fibers. However, the voltage signal created by these current transients will be reduced according to the thickness of the nerve. Assuming now that the proportion of sympathetic fibers is constant in the different nerves means that the extracellular resistance of the nerve between the electrodes will be inversely proportional to the number of these fibers. Thus, the "maximum" voltage amplitude of the sympathetic transients should be inversely proportional to the square root of number of fibers. In contrast, for individual somatic spikes, such as the gamma spikes, their amplitudes will be proportional to the resistance and thus inversely proportional to the number of fibers. Therefore, the ratio of maximum amplitudes of the sympathetic transients and the gamma spikes should be proportional to the square root of the number of fibers. Since the number of fibers should be proportional to the cross‐sectional area of the nerve, then this ratio should be proportional to the square root of the area, i.e. should be directly proportional to the diameter of the nerve. There are several approximations and assumptions in this calculation, but the end result is remarkably similar to what was observed (Figs. [6](#phy213740-fig-0006){ref-type="fig"}C and [7](#phy213740-fig-0007){ref-type="fig"}D). The calculations above thus can explain the observations that the sympathetic spike‐like events overlap in size with the gamma spikes. They also add a little more confidence that the activity we have assumed as sympathetic really does have that identity, and they emphasize what an enormous barrage of sympathetic efferent activity must be present in the thoracic nerves. Individual sympathetic efferents fire at rates of no more than 2 sec^‐1^, suggesting that there must be several hundred, if not more, tonically active sympathetic fibers in each nerve, in order to give the summated activity we have observed. This view is supported by the large number of unmyelinated fibers present in the gray ramus communicans for these segments in the cat. Coggeshall et al. ([1976](#phy213740-bib-0005){ref-type="ref"}) counted about 5000 in each segment. Interestingly, these authors also described the presence of about 20 myelinated fibers, up to 12 *μ*m in diameter, with evidence that these, too, were postsynaptic sympathetic efferents. If these project into the intercostal nerve and are active, they might provide an alternative explanation for the relatively large sympathetic spikes that we observed in the whole internal intercostal nerves. The absence of these spikes in the filaments might then occur simply on the basis of there being a low probability of one of the few large fibers projecting into any one filament. However, it is hard to see how this explanation would predict the graded effect in Figure [6](#phy213740-fig-0006){ref-type="fig"}C, so we think that the spike summation explanation remains, at present, the best explanation. It should also be emphasized, following Sears ([1964b](#phy213740-bib-0027){ref-type="ref"}), that all the above discussion on spike sizes is contingent on the fact that all our recordings were made from intact, undivided nerve branches that contained, we presume, nerve fibers that were (apart from their cut ends) undamaged. This presumption would not apply to teased preparations. Although supported by our calculations above, our interpretation of the overlap in spike amplitudes between the sympathetic and gamma populations still depends critically on cardiac modulation identifying discharges of a given amplitude as sympathetic. The alternative would be that the firing of the gamma motoneurones themselves shows cardiac modulation. The best evidence against this is the lack of modulation for the gamma discharges in 5/6 of the filaments, the filaments being where the gamma spikes are most clearly separated by spike amplitude. The cardiac modulation seen in the gamma ranges for the other nerves should then be ascribed entirely to effects spreading from the sympathetic range. Actual overlap in spike amplitudes is not actually necessary for this. Assuming that extensive superposition is required to give the larger amplitude sympathetic discharges, then further superposition effects would also inevitably take place between the sympathetic and gamma discharges and would provide the explanation for why the plots in Figure [7](#phy213740-fig-0007){ref-type="fig"}B never approached zero. The cardiac modulation observed in one filament recording that was as strong as in the sympathetic range (Fig. [4](#phy213740-fig-0004){ref-type="fig"}) would then have to remain unexplained -- an outlier. Methodological implications {#phy213740-sec-0015} --------------------------- The considerable overlap in amplitudes between the sympathetic and gamma discharges is an important result for any experimenters measuring activity in efferent neurograms, at least for certain sizes of nerves. Efferent neurograms are frequently used in animal experiments as measures of motor outputs, especially where EMG might be impractical or where fictive behavior is to be studied. In such studies the possibility of gamma, as much as alpha motoneurone activity being involved should always be considered, although this is frequently ignored. The present result, with the amplitude of the sympathetic discharges often being larger than 50% of that of the gamma spikes, means that this third category of efferent activity should now additionally be considered. The recordings where this would be of most concern are those where normal activity has been eliminated or reduced by procedures such as experimental injury, and then restoration of function (often modest) is being sought. For example, one result that might need some re‐interpretation featured in a recent publication from our own laboratory (de Almeida and Kirkwood [2013](#phy213740-bib-0002){ref-type="ref"}), where the strength of cross‐correlations between the spike trains of bulbospinal neurones and motoneurone efferent discharges in the rat was found to be strongly dependent on the efferent spike size. It was suggested in that paper that part of the explanation could be an overlap in spike sizes, consequent on the known overlap in alpha and gamma conduction velocities in this species. One would now need to suggest that, for the lowest part of the range, the effect could also involve an overlap with sympathetic discharges. The assumption in either case is that the connections represented in the cross‐correlations were made mainly to alpha motoneurones, and that a proportion of discharges from the other categories of efferents within the supposed alpha ranges had a dilution effect in the cross‐correlations. If, on the other hand, one chose to interpret our observations as resulting from a lower degree of overlap between the discharges, but from the occurrence of cardiac modulation in gamma motoneurones, then yet more ambiguities could arise. Care would then be needed to exclude gamma motoneurone spikes in measurements of cardiac modulation of sympathetic discharges, including in human microneurography experiments. Functional issues {#phy213740-sec-0016} ----------------- Generally, we have used the similarity between the sympathetic discharges described here and others reported in the literature to help in the identification of the discharges here as sympathetic. Nevertheless, one difference is apparent: our histograms suggest weaker cardiac modulation than is commonly described for muscle nerves. To put this into context, when our results are compared to other studies using a similar method (cross‐correlation of *R*‐wave with sympathetic spikes or spike‐like events), our strongest examples of cardiac modulation were similar in strength to those illustrated as typical results in previous studies (Jänig et al. [1983](#phy213740-bib-0014){ref-type="ref"}; Sato and Schaible [1987](#phy213740-bib-0023){ref-type="ref"}; Fatouleh and Macefield [2013](#phy213740-bib-0009){ref-type="ref"}). As pointed out by McAllen and Malpas ([1997](#phy213740-bib-0017){ref-type="ref"}), there are different methods available to quantify sympathetic discharges each with its own advantage. Here, a particular advantage of using whole nerve recordings is that we have avoided the possible selectivity involved either in preparing teased peripheral nerves or in selecting fascicles for recording by microneurography. Such selectivity could create a bias toward the most phasically active fascicles. Indeed, such bias is often explicit (e.g. Boulton et al. [2016](#phy213740-bib-0004){ref-type="ref"}). A disadvantage of our method is that we cannot necessarily exclude the presence of gammas as contributing to the apparent tonic background. This is quite likely for records such as those in Figure [2](#phy213740-fig-0002){ref-type="fig"}, where for the whole internal intercostal nerves, the overlap in amplitudes between the two populations of spikes was deemed to be high. However, this is likely to be much less so for the large majority of the external intercostal nerves. In the example of Figure [1](#phy213740-fig-0001){ref-type="fig"}, during most of expiration (a high proportion of each respiratory cycle), we think the gamma spikes are largely absent, yet, as was generally the case for the external intercostal nerves, the cardiac modulation in sympathetic discharges was relatively modest and could not be seen in the raw recordings. Assuming this is a real result, it is worth asking whether it is a consequence of something particular to the conditions of our experiments or to the intercostal (and abdominal) nerves. One possibility is our use of an elevated level of CO~2~. This is most often regarded as promoting stronger sympathetic output (Millar and Biscoe [1966](#phy213740-bib-0019){ref-type="ref"}; Gregor and Jänig [1977](#phy213740-bib-0013){ref-type="ref"}; Priess and Polosa [1977](#phy213740-bib-0021){ref-type="ref"}), but this doesn\'t help the particular question. Another possibility is that cutaneous sympathetic fibers show much less cardiac modulation than do those in muscle nerves (Jänig et al. [1983](#phy213740-bib-0014){ref-type="ref"}; Fatouleh and Macefield [2013](#phy213740-bib-0009){ref-type="ref"}), so one might suggest that the modest level of modulation in our recordings was the result of recording from mixed nerves. However, this is unlikely to be the reason, because the recordings from the external intercostal nerve, a pure muscle nerve, comprised the group with, if anything, the lowest level of cardiac modulation, even discounting the effect of vagotomy. One might then also ask why the mixed nerves (internal intercostal and its branches) did not show even weaker cardiac modulation. This could be the result of the inhibitory influence of hypercapnia on cutaneous vasoconstrictor fibers (Gregor and Jänig [1977](#phy213740-bib-0013){ref-type="ref"}). A third overall possibility might be that the muscles innervated by the recorded nerves were more‐or‐less continuously, even if fictively, active. Muscle nerve sympathetic discharges have been reported to either decrease (Wallin et al. [1992](#phy213740-bib-0031){ref-type="ref"}) or increase (Boulton et al. [2014](#phy213740-bib-0003){ref-type="ref"}, [2016](#phy213740-bib-0004){ref-type="ref"}) with centrally commanded muscle activation, but in all the studies so far, tonic sympathetic activity has been excluded because of the difficulty in microneurographic recordings to distinguish changes in tonic activity of sympathetic efferents, if present, from the undoubted changes in tonic afferent or somatic efferent discharges or from artifactual changes resulting from electrode movement. The tonic background that we suggest was present in our sympathetic discharges may therefore have occurred specifically as an accompaniment to somatic efferent activity. A number of authors have suggested possible functional advantages in synaptic transmission for sympathetic efferents to fire in bursts (e.g. McAllen & Malpas, [1997](#phy213740-bib-0017){ref-type="ref"}). We do not need to reiterate these arguments here, but merely emphasize that, for some situations, such as the conditions of our experiments, perhaps considerations of firing in bursts should not be the whole story, and that the functional effects of a considerable tonic background should not be ignored. Conflict of Interest {#phy213740-sec-0018} ==================== No conflicts of interest exist with respect to this report. The authors thank J.D. Road, the late S.A. Saywell, N.P. Anissimova, and C.F. Meehan for their participation in the original experiments.

A hallmark of tumor cells is their ability to grow anchorage independent. Proliferation and survival of tumor cells, determining progression of solid tumors, are independent of signals elicited by interactions with the surrounding extracellular matrix (ECM^1^; [@B17]). In contrast, normal diploid cells require anchorage to the ECM for proliferation as well as survival ([@B9]). Several lines of direct evidence show that integrins transduce these signals ([@B42]). Integrins are the most important family of cell surface receptors that mediate cell--matrix interactions ([@B26]). They are heterodimers of noncovalently linked α and β subunits. So far 15 different α subunits and 8 different β subunits are known. The β1 subunit can associate with at least 10 different α subunits forming the largest subfamily of integrins. Members of the β1 integrin subfamily primarily bind to components of the ECM such as fibronectin, collagens, and laminins, but some of them also participate in direct cell--cell adhesion ([@B26]; [@B22]). The cytoplasmic domain of β1 integrin can directly interact with cytoskeletal proteins such as talin and α-actinin and with signal transducing proteins such as focal adhesion kinase (FAK; [@B37]) and integrin-linked kinase ([@B23]). Integrin engagement and clustering regulate shape, motility, survival, and proliferation of cells. These events are executed by integrin-mediated cascades of intracellular signals that include tyrosine phosphorylation of FAK ([@B21]), increases in intracellular Ca^2+^ levels ([@B38]), intracellular pH ([@B39], [@B40]), inositol lipid synthesis ([@B30]), and expression of cyclins ([@B20]). Furthermore, it has been demonstrated that integrins can also mediate the activation of protein kinase C ([@B45]), mitogen-activated protein kinase ([@B32]) and NF-κB ([@B46]). In addition to these adhesion-mediated signaling pathways, many cells depend on growth factor--mediated signals for appropriate cell cycle progression and proliferation. In the present study we have used β1 integrin--deficient embryonic stem (ES) cells (Fässler al., 1995) to induce teratomas in syngeneic mice. ES cells as well as pre- or early postimplantation embryos of most mouse strains develop into tumors when transplanted into an ectopic location of syngeneic animals ([@B8]; [@B7]). These tumors are composed of various differentiated somatic tissues and are called teratomas. We show that β1-null ES cells give rise to either very small or no teratomas. The most prominent changes that are associated with the impaired growth in β1-null teratomas are abnormal depositon of ECM proteins and various defects in basement membranes. Furthermore, β1-null teratomas showed an inefficient angiogenesis. A number of studies have demonstrated convincingly that tumor growth is dependent on angiogenesis ([@B16]). Tumor angiogenesis is regulated by factors produced by tumor cells as well as by cell adhesion molecules expressed on endothelial cells. Systemic or local administration of antibodies or cyclic RGD peptides blocking αvβ3 integrin function inhibits tumor angiogenesis and as a consequence promotes tumor regression ([@B4]). This anti-angiogenic effect of the αvβ3 antagonists results from the activation of apoptosis of newly sprouting blood vessels ([@B5]). We now show that β1 integrin plays an essential role during angiogenesis in teratomas. In normal teratomas both host- and ES cell--derived endothelial cells contribute to angiogenesis. In contrast, in β1-null teratomas, all vascular cells are exclusively derived from the host. Furthermore, we report that vascular endothelial growth factor (VEGF) treatment--induced proliferation of endothelial cells and extensive branching of blood vessels in normal but not in β1-null embryoid bodies. MATERIALS AND METHODS ===================== Cells and Cell Culture ---------------------- The following ES cells were used to induce teratomas: wild-type ES cell line D3 (+/+; [@B11]); G119 (+/−), which is heterozygous for the β1 integrin gene mutation; G101 (+/+), which is wild-type for β1 integrin gene but mock transfected; and G201 (−/−) which is β1 integrin deficient ([@B14]). In the cell clones G119 and G201, a fusion DNA of β-galactosidase and neomycin is inserted in frame with the ATG of the β1 integrin gene ([@B14]). The cell clone G101 contains a randomly integrated β-galactosidase gene that is ubiquitously expressed. ES cells were cultured in the absence of a fibroblast feeder layer in DME, supplemented with 20% heat-inactivated FCS (GIBCO BRL, Gaithersburg, MD), 0.1 mM β-mercaptoethanol (Sigma Chemical Co., St. Louis, MO), 1× non-essential amino acids (GIBCO BRL), and 1,000 U/ml recombinant leukemia inhibiting factor (GIBCO BRL). For differentiation, ES cells were cultured in hanging drops as described previously ([@B15]). Briefly, 600 cells were cultured in 20 μl of DME, supplemented with 20% FCS hanging from the lid of the culture dish for 5 d, which allows the formation of cell aggregates (embryoid bodies). Subsequently, the aggregates were plated on Tissue Tek chambers (Nunc, Wiesbaden, Germany) and incubated for 7, 15, or 20 d, respectively, fixed in 4% paraformaldehyde and immunostained for von Willebrand Factor (Behringwerke Ag, Marburg Lahn, Germany) and platelet endothelial cell adhesion molecule (PECAM) ([@B43]). Alternatively, after outgrowth of cell aggregates for 20 d, cells were trypsinized and cultured on gelatin-coated glass cover slips for 24 h, fixed, and stained for von Willebrand Factor (vWF) or PECAM. Positive cells were counted using an Axiophot fluorescence microscope (Carl Zeiss, Oberkochen, Germany). For VEGF treatment, ES cells were cultured in hanging drops for 2 d, in bacteriological dishes for 3 d, and then plated on gelatinized glass cover slips for another 12 d. The culture medium was DME, containing 20% FCS with either 10 or 20 ng/ml VEGF (R&D Sytems, Wiesbaden, Germany). Antibodies ---------- The following primary antibodies were used: rabbit anti--rat β1 integrin; hamster anti--mouse β3 integrin; hamster anti--mouse αv integrin (both from PharMingen, San Diego, CA); hen anti--human FN ([@B28]); rabbit anti--mouse FN (GIBCO BRL); goat anti-collagen type I; goat anti-collagen type III (both from Southern Biotechnology Associates Inc., Birmingham, AL); rat anti--mouse nidogen ([@B12]); rat anti--mouse PECAM-1 ([@B43]); and rabbit anti-- mouse vWF (Behringwerke). The following secondary antibodies were used: goat anti--rabbit FITC; goat anti--rabbit Cy3; rabbit anti--hamster FITC; goat anti--chicken FITC; donkey anti--goat Cy3; goat anti--rat Cy3 (Jackson Immunoresearch Lab. Inc., West Grove, PA), rabbit anti-digoxygenin FITC (Boehringer Mannheim, Mannheim, Germany), biotinylated goat anti--rabbit Ig (Vector Laboratories, Inc., Burlingame, CA), biotinylated goat anti--rat Ig (Amersham Intl., Little Chalfont, UK), and streptavidin-horseradish peroxidase conjugate (Amersham Intl.). Teratoma Induction ------------------ 10^7^ ES cells were trypsinized, washed twice, suspended in 100 μl PBS, and then injected subcutaneously on the back of syngeneic 129/SV male mice. After 21 or 28 d, tumors were surgically removed and frozen in ice-cold isopentan. To analyze cell proliferation, 25 mg per 100 g body weight of the thymidine analogue bromodeoxyuridine (BrdU) was injected intraperitoneally 2.5 h before the excision of the tumors. Microscopical Analysis of Embryoid Bodies and Tumor Tissue ---------------------------------------------------------- ### Light microscopy. For light microscopical examination, small pieces of teratomas and 20-μm-thick immunostained cryosections were dehydrated conventionally in a graded ethanol series, and finally infiltrated with and embedded in araldite (Serva, Heidelberg, Germany). Semithin sections of 1--2 μm were analyzed using a Zeiss Axiophot microscope (Carl Zeiss), with or without methylene blue staining. ### Histochemistry. Teratomas were surgically removed, frozen in ice-cold isopentan, and stored at −80°C. Tissue specimens were cut into 10-μm-thick sections and collected on glass slides (Shandon, Frankfurt, Germany). Hematoxylin/Eosin and lacZ staining followed published protocols ([@B14]). Stained sections were analyzed using a Zeiss Axiophot microscope. ### Cell proliferation. Incorporation of intraperitoneally injected BrdU into the DNA of replicating teratoma cells was analyzed using anti-BrdU monoclonal antibodies, following the protocol supplied by the manufacturer (Boehringer Mannheim). Similarly, proliferation of cells in embryoid bodies was assayed after addition of BrdU to the culture medium and incubation for another 2 h. Embryoid bodies were fixed as described above, and then processed following the manufacturer\'s protocol. For development of the immunostaining 3,3-diaminobenzidine-tetrahydrochloride (DAB) with nickel enhancer (12 μl 0.1 PBS, 300 μl DAB, 300 μm NH~4~Cl, 300 μl NiSO~4~, 10% glucose, and 50 μl glucose oxidase) was used. This gives rise to a black reaction product that is localized in the nuclei of proliferating cells. Subsequently, sections were incubated with vWF antibody or PECAM-1 antibody (1:800) and processed as described below. Counterstaining was done with methyl green for 35 s. Afterwards, embryoid bodies were dehydrated in a graded ethanol series, coverslipped, and analyzed using a Zeiss Axiophot microscope. ### Apoptosis. Cell death was analyzed following the protocol supplied by the manufacturer (Oncor, Heidelberg, Germany). Briefly, tissue sections were fixed in 4% paraformaldehyde in PBS, postfixed in ethanol/acetic acid (2:1), and incubated with terminal deoxynucleotidyl transferase, which catalyzes the addition of digoxygenin-conjugated nucleotides to free 3′-hydroxyl groups in apoptotic cells. Incorporation of labeled nucleotides was detected by FITC-conjugated, anti-digoxygenin antibodies. The number of apoptotic cells was determined from random sections of three normal and four β1-null teratomas. ### Immunofluorescence. Tumor sections, which were 6-μm thick, were incubated with primary antibody in PBS, supplemented with 1% BSA at dilutions recommended by the suppliers for 1 h, at room temperature. Subsequently, the sections were washed three times in PBS and incubated with fluorescence-labeled secondary antibodies at dilutions recommended by the manufacturer for 1 h, at room temperature. After washing three times in PBS, the fluorescent specimens were mounted using elvanol ([@B14]). For analyzing the expression of αv and β3 integrin on endothelial cells, embryoid bodies were fixed in 4% paraformaldehyde and indirectly double labeled with hamster antibodies against αv or β3 integrin (PharMingen), respectively, and rat antibodies against PECAM-1 ([@B43]). Subsequently, embryoid bodies were washed three times in PBS and incubated with fluorescence-labeled secondary antibodies at dilutions recommended by the manufacturer for 1 h at room temperature. After washing three times in PBS, the fluorescent specimens were mounted using elvanol ([@B14]). Analysis was performed using a Zeiss Axiophot fluorescence microscope. ### Immunohistochemistry. For immunohistochemistry, 20-μm-thick cryosections of sucrose-embedded teratomas prefixed with 4% paraformaldehyde in 0.1× PBS, and paraformaldehyde-fixed embryoid bodies were stained. Endogenous peroxidase was blocked by 3% H~2~O~2~ in 60% methanol. Subsequently, cells were permeabilized with 0.2% Triton X-100 (Sigma Chemical Co.) in 0.1× PBS. The sections were incubated with von Willebrand antibody (1:1,000) or PECAM antibody (1:800) in PBS. Embryoid bodies were incubated in PBS, supplemented with 0.8% BSA overnight at 4°C. Subsequent sections and embryoid bodies were washed four times in PBS and incubated with biotinylated secondary antibodies for 1 h at room temperature. A streptavidin--horseradish peroxidase complex was then used as a detection system (1:200; incubation for 1 h at room temperature). The staining was developed for 15 min with DAB in 5 mM Tris-HCl buffer, pH 7.4, supplemented with 0.1% H~2~O~2~. Analysis was performed using a Zeiss Axiophot microscope. Morphometrical analyses of vessels identified by PECAM antibody were performed with Optimas 5.2 (Optimas Corporation, Bothwell, WA). Electron Microscopy ------------------- Tumor tissue was obtained from 6-wk-old chimeric animals, which were killed by cervical dislocation and subsequently transcardially perfused with a 0.1 mol/liter cacodylate buffer containing 2% paraformaldehyde and 2% glutaraldehyde at a perfusion pressure of 60 cm H~2~O. Tumors were removed, fixed for an additional 4 h in the same fixative, and then cut into small pieces and fixed in 0.1 mol/liter cacodylate buffer, pH 7.3, containing 2% osmium tetroxide for 2 h at 4°C. Tissue pieces were rinsed three times in cacodylate buffer, block stained for 8 h in 70% ethanol containing 1% uranyl acetate, dehydrated in a series of graded ethanol, and embedded in araldite. Ultrathin sections (30--60 nm) obtained with a diamond knife on a Reichert ultramicrotome (Reichert, Bensheim, Germany) were placed on copper grids and examined with a Zeiss EM 902A electron microscope. Embryoid bodies were fixed in a 100 mM Hepes/Pipes buffer, pH 7.35, containing 1.75% paraformaldehyde, 2% glutaraldehyde, and 15% picric acid for 1 h at room temperature. Afterwards embryoid bodies were treated with 100 mM Hepes/Pipes buffer containing 1% tannic acid for 30 min at room temperature and finally osmificated with 0.5% OsO~4~. Before embedding in Epon resin (Agar Scientific, Stansted, UK) embryoid bodies were dehydrated in a graded series of ethanol. For light microscopy, semithin sections were stained with methylene blue. For electron microscopy, ultrathin sections (30--60 nm) were mounted on formvar-coated copper grids, stained with 0.2% uranyl acetate and lead citrate, and then examined with a Zeiss EM 902A electron microscope. RESULTS ======= ES Cells Lacking β1 Integrin Form No or Only Small Teratomas ------------------------------------------------------------ To test whether the lack of β1 integrin has an influence on the development of teratomas normal and β1-null ES cells were injected subcutaneously into syngeneic male mice. After 21 d of incubation, wild-type tumors were clearly visible on the back of the mice. In contrast, β1-null tumors either did not form or were significantly reduced in size (Fig. [1](#F1){ref-type="fig"}). To investigate the growth of β1-null tumors in more detail, mice were killed by cervical dislocation and examined under the stereomicroscope. In 9 out of 18 animals injected with β1-null ES cells, no tumor could be detected either at the site of injection, or in distant organs. In the remaining nine animals, a small tumor had formed, which was significantly reduced in weight (mean value 47 mg) when compared with tumors derived from normal (0.93 g) or β1 integrin--heterozygous (1.18 g) ES cells (Fig. [2](#F2){ref-type="fig"}). No difference in tumor growth was observed in mice carrying either a β1-null tumor alone or together with a normal or heterozygous tumor. Cell Proliferation and Apoptosis Is Not Significantly Altered in β1-null Teratomas ---------------------------------------------------------------------------------- To test whether cell proliferation is reduced in β1-null tumors BrdU was injected into tumor-bearing mice. Detection of BrdU in tissue sections with specific antibodies showed that the distribution of replicating cells is similar in normal and β1-null teratomas. Some areas in both tumor types contained nests of many proliferating cells, whereas other areas show a homogenous distribution of a few proliferating cells (Fig. [3](#F3){ref-type="fig"}). This heterogeneity of the tumor architecture results from the wide range of different tissues that form in teratomas. As a consequence, the number of proliferating as well as apoptotic cells varied tremendously from one area to another; normal tumors grown for 10 d had between 84 and 176 BrdU-postive cells per mm^2^ tissue section and tumors grown for 21 d had between 220 and 279 BrdU-positive cells per mm^2^ tissue section. In corresponding sections of β1-null teratomas, the number of proliferating cells ranged between 42 and 116 cells in 10-d-old teratomas, and between 156 and 265 cells in 21-d-old teratomas. This large range in proliferating cells in both tumor types did not provide a statistically significant difference but a tendency towards a reduced proliferation rate in β1-null teratomas. Similarly, the number of apoptotic cells was very inhomogenously distributed in the tissue of normal and β1-null teratomas: β1+/+, β1+/−, and β1−/− teratomas showed higher apoptosis rates in areas that were lacZ-positive, and thus ES cell--derived and lower rates in areas that were lacZ-negative, and thus host cell--derived (data not shown). Both tumor types showed areas with small cell clusters with a high rate of apoptosis, whereas in other areas the apoptotic cells were evenly distributed (Fig. [3](#F3){ref-type="fig"}). The number of apoptotic cells ranged between 250 and 712 cells per mm^2^ tissue in normal teratomas, and between 249 and 805 cells per mm^2^ tissue section in β1-null teratomas. Also here, the large variation did not provide a statistically significant difference but again a tendency towards a higher apoptosis rate in β1-null teratomas was apparent. β1-Null Teratomas Contain Differentiated Cells but Have an Altered Tissue Architecture with Less Host-derived Cells ------------------------------------------------------------------------------------------------------------------- Hematoxylin/Eosin staining revealed that β1 +/+ teratomas (G101) (Fig. [4](#F4){ref-type="fig"} *A*) and β1 +/− teratomas (not shown) are composed of a variety of differentiated cells and tissues, e.g., glandular epithelium, cartilage, connective tissue, and neuronal cells. β1-Null tumors appeared cell dense and contained fewer cords of fibrotic material (Fig. [4](#F4){ref-type="fig"} *B*). Histochemical analysis of several β1-null tumors revealed that many differentiated cells are present in β1-null teratomas. ES cell--derived cells could be differentiated by lacZ activity, since the lacZ gene was introduced in the β1 integrin gene and in the genome of the mock-transfected ES cells ([@B14]; and data not shown). Mock-transfected, as well as β1 integrin--heterozygous tumors contained large areas devoid of lacZ activity proving their host-derived origin (Fig. [4](#F4){ref-type="fig"} *C*). In contrast, the number of lacZ-negative cells was low in all β1-null tumors analyzed (Fig. [4](#F4){ref-type="fig"} *D*) indicating that the migration of normal host cells into mutant tumor tissue is reduced. Deposition of Extracellular Proteins Is Altered in β1-null Teratomas -------------------------------------------------------------------- To test whether the deposition of matrix proteins is affected by the lack of β1 integrin, the distribution of fibronectin, collagen I, and nidogen were compared between normal and β1-null tumors. In normal tumors most of the fibronectin (Fig. [5](#F5){ref-type="fig"} *C*) was associated with cells expressing large amounts of β1 integrin (Fig. [5](#F5){ref-type="fig"} *A*). Only in a few areas fibronectin staining did not colocalize with β1 integrin staining (data not shown). In β1-null tumors, on the other hand, fibronectin was deposited throughout the tumor tissue (Fig. [5](#F5){ref-type="fig"} *D*) and did not correlate with host cell--derived β1 integrin expression (Fig. [5](#F5){ref-type="fig"} *B*). Similarly, in sections derived from β1-null tumors, collagen I (Fig. [5](#F5){ref-type="fig"} *F*) and nidogen (not shown) were diffusely distributed. This was not the case in normal tumors in which collagen I (Fig. [5](#F5){ref-type="fig"} *E*) and nidogen (not shown) were mainly restricted to areas with strong β1 integrin staining. To evaluate whether the ultrastructural morphology of basement membranes is altered by the lack of β1 integrin expression, electronmicroscopy was performed with tumor tissues and embryoid bodies derived from normal and β1-null ES cells. In normal tumors and normal embryoid bodies the basement membrane was continuously present along the basal surface of epithelial cells (Fig. [6](#F6){ref-type="fig"}, *A* and *D*) or around muscle cells (not shown) with a constant width that never exceeded 30 nm. In contrast, most cells in β1-null teratomas and embryoid bodies were covered by partially detached basement membranes, which resulted in the formation of irregular clefts between plasma membrane and lamina densa (Fig. [6](#F6){ref-type="fig"}, *B* and *F*). In other areas, cells were covered by several layers of basement membranes (Fig. [6](#F6){ref-type="fig"} *C*) or by only partially developed basement membranes (Fig. [6](#F6){ref-type="fig"}, *E* and *F*). Blood Vessels of β1-null Teratomas Are Small and Irregular in Shape and Lack ES Cell-derived Endothelial Cells -------------------------------------------------------------------------------------------------------------- Hematoxylin/Eosin staining, as well as immunostaining for the endothelial markers vWF and PECAM revealed a large number of circular vessels of various sizes in β1 +/+ teratomas (Fig. [7](#F7){ref-type="fig"} *A*). In contrast, in β1-null tumors, vessels were smaller in size and of irregular shape (Fig. [7](#F7){ref-type="fig"} *B*). To test whether β1-null ES cells can contribute to the formation of blood vessels, endothelial cells were characterized in β1 +/+ and −/− teratomas. In β1 +/+ and +/− teratomas, all cells that expressed vWF or PECAM showed strong staining for β1 integrin (Fig. [7](#F7){ref-type="fig"} *C*). Furthermore, many cells that expressed vWF were also lacZ-positive (Fig. [8](#F8){ref-type="fig"}, *A* and *C*) clearly indicating that ES cells can contribute to angiogenesis in developing teratomas. In β1-null tumors, however, cells that expressed vWF (Fig. [7](#F7){ref-type="fig"} *B*) always expressed high levels of β1 integrin, indicating their host-derived origin (Fig. [7](#F7){ref-type="fig"} *D*). These results were confirmed in serial sections of six β1-null teratomas, which were analyzed by lacZ and vWF stainings. In all β1-null teratomas analyzed so far, cells that were positive for vWF staining (Fig. [8](#F8){ref-type="fig"} *B*) were negative for lacZ expression (Fig. [8](#F8){ref-type="fig"} *D*). During the analysis of tissue sections derived from β1-null teratomas it became evident that also the host cell--derived vasculature has abnormalities. Whereas blood vessels in normal teratomas have a smooth inner surface and are tightly anchored to the surrounding connective tissue (Fig. [9](#F9){ref-type="fig"} *A*), blood vessels in β1-null tumors have an irregular inner surface and display large gaps between endothelial cells and the surrounding tissue (Fig. [9](#F9){ref-type="fig"} *B*). β1-Null ES Cells Can Differentiate into Endothelial Cells In Vitro but Formation of Vessels Is Delayed ------------------------------------------------------------------------------------------------------ ES cells were cultured for 5 d in suspension to form embryoid bodies, and then were plated on tissue culture dishes. Initially, PECAM-positive cells were aggregated in dense clusters (Fig. [10](#F10){ref-type="fig"} *A*). These clusters then started to form thin tubes and branched. After 7 d on the culture dish about half of the PECAM-positive cells were part of blood vessels in embryoid bodies derived from normal ES cells. After 15 d of incubation, PECAM-positive cell clumps were no longer visible, the blood vessels were significantly grown in diameter, extensively branched (Fig. [10](#F10){ref-type="fig"} *B*), and often contained many blood cells (Fig. [10](#F10){ref-type="fig"} *C*). When β1-null embryoid bodies were analyzed 7 or 15 d after plating, only PECAM-positive cell clusters were found without any signs of vessel formation (Fig. [10](#F10){ref-type="fig"} *D*). After 20 d in culture, a fine network of small vessels began to form in some areas of the embryoid bodies (Fig. [10](#F10){ref-type="fig"} *E*). These vessels were small in diameter and never contained blood cells. In other areas, PECAM-positive cells were still organized in small cell patches, often with loose cell-- cell contacts (Fig. [10](#F10){ref-type="fig"} *F*). The mean value of the diameter of normal and β1-null vessels in seven embryoid bodies after a 3-wk culture period was 18.1 ± 3.0 μm and 6.8 ± 2.1 μm, respectively (Fig. [11](#F11){ref-type="fig"}). To determine whether the absence of β1 integrin expression interferes with the differentiation of endothelial cells, we counted the number of PECAM-positive cells in normal as well as β1-null embryoid bodies. After a plating period of 28 d, 100 embryoid bodies (derived from normal and β1-null ES cells, respectively) were trypsinized; single cells were plated and cultured overnight, and then stained for PECAM expression. Out of 9,600 cells counted, 96 PECAM- positive cells were found in normal embryoid bodies and 9 cells in β1-null embryoid bodies. This 10-fold decrease of PECAM-positive cells in the β1-null embryoid bodies indicates a delayed or less efficient differentiation of β1-null endothelial cells. Expression of β3 and αv Integrin Is not Significantly Altered on β1-null Endothelial Cells ------------------------------------------------------------------------------------------ Proliferating endothelial cells express large amounts of αvβ3 and αvβ5 integrins. To determine whether the expression of β3 or αv integrin is altered on β1-null endothelial cells, normal and β1-null embryoid bodies were double-immunostained for either PECAM and β3 integrin (Fig. [12](#F12){ref-type="fig"}, *A--D*), or PECAM and αv integrin (Fig. [12](#F12){ref-type="fig"}, *E--H*), respectively. Vessels in embryoid bodies derived from normal and β1-null ES cells express similar amounts of β3 (Fig. [12](#F12){ref-type="fig"}, *B* and *D*) and αv integrin (Fig. [12](#F12){ref-type="fig"}, *F* and *H*). These data indicate that the absence of β1 integrin is not associated with an altered expression of other endothelial cell integrins. VEGF Does Not Promote Angiogenesis in β1-null Embryoid Bodies ------------------------------------------------------------- VEGF induces proliferation of endothelial cells and sprouting of vessels and therefore is a potent promoter of angiogenesis. To test whether VEGF can overcome abnormal angiogenesis in β1-null embryoid bodies, normal and β1 integrin-deficient ES cells were differentiated in the presence of 10 or 20 ng/ml VEGF, respectively. After 5 d in suspension culture and 12 d on gelatinized cover slips, embryoid bodies were treated with BrdU for 2 h and subsequently stained for PECAM expression and BrdU incorporation. Treatment of normal embryoid bodies with either 10 or 20 ng/ml VEGF significantly increased the number and the extent of blood vessel branching (Fig. [13](#F13){ref-type="fig"}, *A* and *C*; and Table [I](#TI){ref-type="table"}). In contrast, VEGF at both concentrations had no significant effect on blood vessel branching in β1-null embryoid bodies (Fig. [13](#F13){ref-type="fig"}, *B* and *D*; and Table [I](#TI){ref-type="table"}) indicating that VEGF cannot overcome the lack of β1 integrin function. VEGF treatment also increased the number of BrdU-positive cells in normal (Fig. [13](#F13){ref-type="fig"}, *A* and *C*, *arrows*) but not in β1-null embryoid bodies (Fig. [13](#F13){ref-type="fig"}, *B* and *D*, *arrows*). DISCUSSION ========== In the present study we show that the growth of ES cell-- derived teratomas is dependent on the expression of β1 integrin. Whereas subcutaneous injection of normal ES cells under the skin of the back of syngeneic mice gave rise to large teratomas, β1-null ES cells either produced no, or only very small tumors. This reduced growth of β1-null teratomas suggests a growth-supporting role of β1 integrin. These results were unexpected since many tumors lack α5β1 integrin function, a phenomenon that has been linked with anchorage-independent growth and high tumorigenicity ([@B25]; [@B34]; [@B41]). Conversely, overexpression of α5β1 integrin in Chinese hamster ovary cells is accompanied by a loss of tumorigenicity and reduced proliferation in vitro ([@B19]). A similar growth-suppressing role was recently demonstrated for α2β1 integrin, which is frequently absent in breast cancer cells. Reexpression of α2β1 integrin in such cells also leads to a diminution of the malignant phenotype ([@B48]). Several explanations could account for this discrepancy. First, ES cells are normal embryonal cells that differentiate into many somatic tissues and thus give rise to a tumor called teratoma ([@B8]; [@B7]). Since these cells are not transformed, they may still be dependent on anchorage to the extracellular matrix for growth ([@B9]) and survival ([@B31]; [@B3]). Integrins execute cell-- matrix interactions and lack of anchorage in β1-null ES cells could therefore be an explanation for the reduced tumor growth. To test this hypothesis we analyzed cell proliferation by injecting BrdU into teratoma-bearing mice and stained for apoptosis. Extensive analysis of normal and β1-null teratomas revealed that cell proliferation and apoptosis are not significantly affected in β1-null tumors. It has to be noted, however, that the extents of proliferation and apoptosis were very different between different sections of the same tumor, as well as between individual tumors. This heterogeneity is due to the ability of ES cells to differentiate into a wide range of somatic tissues with different apoptosis and proliferation rates. For this reason we obtained, in normal as well as β1-null teratomas, enormous variations in the numbers of proliferating and apoptotic cells. Despite this variation, there were more areas with higher numbers of proliferating and lower numbers of apoptotic cells in normal teratomas when compared with β1-null teratomas. Such alterations would be in agreement with reports that have linked cell survival with integrin function (for review see Ruoslahti and Reed, 1994). The binding of α5β1 integrin to fibronectin is inducing bcl-2 expression, which is one of the best studied cell survival factors ([@B47]). Other studies showed that disrupting the binding to basement membranes can induce apoptosis ([@B3]; [@B6]). Biochemical studies of β1-null teratomas (Sasaki, T., E. Forsberg, W. Bloch, K. Addicks, R. Fässler, and R. Timpl, submitted for publication), together with our electronmicroscopical analysis indicate that the integrity of basement membranes is severely affected by the absence of β1 integrin. Whereas epithelial cells of normal teratomas are tightly attached to a basement membrane, β1-null cells have either partly or completely lost this tight attachment leading to the dislocation of basement membranes into the interstitium. This absence of normal interaction between cells and basement membranes could not only explain the increased cell death in teratomas, but also the altered differentiation of keratinocytes ([@B2]) and cardiac muscle cells ([@B15]). Similar defects in basement membrane morphology have been reported in α3-null mice, which have disorganized and fragmented glomerular basement membranes, and podocytes that completely lack foot processes ([@B29]). In addition, these mice develop skin blisters that are due to an inefficient maintenance of the epidermal basement membrane ([@B10]). A second explanation for the reduced tumor growth could be the altered composition or distribution of ECM proteins in β1-null teratomas. The ECM has an important role as reservoir for growth factors, proteases, and protease inhibitors ([@B1]; [@B33]). Basement membranes, for example, can store and release many growth factors. The various abnormalities of basement membranes as seen in β1-null teratomas and β1-null embryoid bodies may contribute to the diminished growth and to defects in differentiation ([@B15]). In addition, the disturbed matrix deposition can alter mechanochemical properties, which may reduce the growth rate of β1-null teratomas ([@B27]). Indeed, immunostainings for fibronectin and collagens revealed an altered distribution in β1-null teratomas. In normal teratomas both matrix proteins concentrate in streaklike structures and often colocalize with high β1 integrin staining. In contrast, both proteins are diffusely deposited in β1-null teratomas and never concentrate along β1 integrin-positive, host-derived cells. This striking finding was also supported by routine histological analysis. Tissue specimens of β1-null teratomas contained little ECM, with many mutant and only a few normal, host-derived (lacZ negative) cells. A third explanation for the formation of small teratomas could be abnormal angiogenesis in the absence of β1 integrin. In contrast to tumors derived from somatic tissue, the vasculature of teratomas is derived from the host and from the ES cells ([@B35]; [@B44]). Immunostaining demonstrated host-derived endothelial cells expressing β1 integrin in normal as well as β1-null tumors. Whereas lacZ staining indicated ES cell--derived (lacZ positive) endothelial cells in β1 integrin--heterozygous as well as mock-transfected tumors, no ES cell--derived endothelial cells were detected in β1-null teratomas. This complete absence of β1-null endothelial cells is also seen in the highly vascularized fetal and adult liver of β1-null chimeric mice ([@B13]; [@B24]). These data suggest, that β1-deficient ES cells either do not differentiate in vivo into endothelial cells or that they differentiate into endothelial cells but cannot organize into blood vessels. A detailed inspection of the vasculature revealed furthermore that blood vessels were significantly smaller in β1-null teratomas as compared to normal teratomas, which may result from the small tumor size. In addition, however, we found that the shape of smaller vessels is also different in β1-null tumors: the inner surface is irregular and the vessels are loosely embedded into the surrounding tissue. Apparently, even endothelial cells expressing β1 integrin are not able to make normal blood vessels in β1-null teratoma, indicating lack of an essential interaction between endothelial cells and the surrounding cells or the extracellular matrix. To test whether the β1-null ES cells have a cell autonomous impairment of differentiation into endothelial cells, we differentiated ES cells into embryoid bodies in vitro and assessed for the presence of endothelial cells ([@B15]). To our surprise, we found that differentiation of endothelial cells occurs in normal and β1-null embryoid bodies. After 12 d (5 d in suspension culture, and 7 d on plastic surface) normal endothelial cells start to sprout and to form a large network of vessels of various diameters that are also often filled with blood cells. In contrast, β1-null endothelial cells initially form compact cell nests with tight cell--cell contacts. Only after a culture period of 3 wk, β1-null endothelial cells start to migrate and to form small vessels. The ability of β1-null endothelial cells to differentiate may also occur in vivo. This hypothesis is supported by the presence of β1-null hematopoietic stem cells in chimeric mice ([@B24]) that, together with endothelial cells, develop from a common precursor cell called hemangioblast. Currently, we have no experimental explanation for the complete lack of β1-null endothelial cells both in teratomas and in livers of chimeric mice ([@B13]), despite the fact that they are readily formed in vitro. It is possible that the in vitro environment favors the survival of these cells, which is not the case in vivo. An important role for angiogenesis has been reported for αvβ3 and αvβ5 integrins ([@B42]). Immunostaining of β1-null embryoid bodies with antibodies detecting αv or β3 integrin clearly revealed that both integrin subunits are not downregulated. Therefore, we can exclude that the absence of β1 integrin expression influences endothelial cell function by altering the expression of other integrins. Growth factors and angiogenic signals can upregulate αvβ3 expression on endothelial cells and induce blood vessel branching ([@B4]; [@B18]). Moreover, several growth factors were shown to exert their angiogenic signals in concert with endothelial integrins. For example, tumor necrosis factor-α and basic FGF depend on αvβ3 integrin, and VEGF and TGF-α depend on αvβ5 integrin ([@B18]). VEGF is the most potent stimulator of neovascularization by regulating proliferation of endothelial cells and branching of blood vessels. Testing whether VEGF can override the defective branching of β1-null endothelial cells, we found that VEGF had no effect on β1-null endothelial cells: neither the proliferation rate nor the branching could be enhanced in β1-null embryoid bodies. Endothelial cells in normal embryoid bodies, however, responded to VEGF by an increased proliferation rate and by an extensive formation of new vessel branches. These data demonstrate for the first time that β1 integrins play an important role during angiogenesis and moreover, that VEGF signaling depends on the presence of β1 integrin function. This finding is unexpected since earlier reports showed that β1 integrin function is not essential for tumor-induced angiogenesis ([@B5]). [@B5] treated chicken embryos that were exposed to various tumors with anti-β1 integrin antibodies and were unable to influence tumor growth or angiogenesis. One explanation for this discrepancy may be an incomplete inhibition of β1 integrin function with a single injection of antibodies. An alternative explanation could be the ubiquitous expression of β1 integrin. Antibodies against β1 integrin find their antigen on almost all cells and may be eliminated before endothelial cells proliferate and form new blood vessels. We thank M. Grüne and S. Benkert for technical assistance and E. Dejana for antibody gifts. This work was supported by the Deutsche Forschungsgemeinschaft (Fa 296/1-2) and the Hermann and Lilly Schilling Stiftung. W. Bloch and E. Forsberg have contributed equally to this work. Address all correspondence to Reinhard Fässler, Max Planck Institut für Biochemie, am Klopferspitz 18 A, 82 152 Martinsried, Germany. Tel.: (49) 89-8578-2215. FAX: (49) 89-8578-2422. e-mail: <faessler@biochem.mpg.de> BrdU : bromodeoxyuridine ECM : extracellular matrix ES : embryonic stem PECAM : platelet/endothelial cell adhesion molecule VEGF : vascular endothelial growth factor vWF : von Willebrand Factor {#F1} {#F2} {#F3} {#F4} {#F5} {#F6} {#F7} ###### Localization of vWF and lacZ expression in normal (+/+) and β1-null (−/−) teratomas. Teratomas were removed 21 d after inoculation of ES cells and sectioned. Sections were stained overnight for lacZ expression and subsequently immunostained for vWF expression. In normal teratomas, vWF-positive areas (*A*, *arrows*) are found to be positive and negative for lacZ expression (*C*, *corresponding arrows*) indicating that endothelial cells are both host- and ES cell--derived. In contrast, in β1-null teratomas, all vWF-positive areas (*B*, *arrows*) are lacZ-negative (*D*, *corresponding arrows*) indicating that endothelial cells are solely host derived. Many other areas, however, that are vWF-negative (*B*, *arrowhead*) are clearly lacZ-positive (*D*, *corresponding arrowhead*) indicating that many other regions in β1-null teratomas are ES cell--derived. Bar, 100 μm.   {#F9} {#F10} {#F11} {#F12} {#F13} ###### Quantification of Branches of Blood Vessels in Normal (+/+) and β1-null (−/−) Embryoid Bodies Treated Without or With VEGF Without VEGF 10 ng/ml VEGF 20 ng/ml VEGF ------- -- --------------------- -- --------------------- -- -------------------- (1/1) 273 ± 103 (*n* = 9) 363 ± 103 (*n* = 8) 372 ± 86 (*n* = 7) (2/2)  41 ± 30 (*n* = 10)  32 ± 20 (*n* = 8)  76 ± 85 (*n* = 7) Embryoid bodies were cultured 5 d in suspension and 12 d on gelatinized glass cover slips in the absence or presence of 10 or 20 ng/ml VEGF, respectively. After the end of the culture period, embryoid bodies were fixed and stained for PECAM expression. Data show all vessel branches per embryoid body and are expressed in mean + SD. *n* is number of embryoid bodies analyzed.  

Introduction {#section1-1176934319838818} ============ Lily is a perennial bulbous flower in the monocotyledon subclass, *Liliaceae* family and *Lilium* genus, with an extremely high ornamental value. China is a large producer of lily cut flowers. As the planting of lily continues and the planting area of lily expands, lily diseases are becoming more serious. Lily basal rot (also known as root rot or stem rot) has become one of the most important diseases endangering lily. It is a soil-borne disease that mainly harms the underground part of the plant, resulting in wilting, yellowing, and rotting of the bulbs. This disease seriously affecting the yield and quality of lily and resulting in huge economic losses.^[@bibr1-1176934319838818]^ *Fusarium oxysporum* f. sp. *lilii* is the main pathogen that causes lily basal rot.^[@bibr2-1176934319838818]^ *F. oxysporum* can survive for a long time as mycelium in the bulbs or as mycelium, chlamydospore, or sclerotium with diseased residue in the soil. At present, chemical control methods are the main measures to prevent and control lily basal rot. However, due to the soil-borne nature and the development of pathogenic resistance of this disease, chemical control methods have shown dwindling effects.^[@bibr3-1176934319838818]^ At the same time, long-term, large-scale, and repeated use of chemical pesticides causes environmental pollution. Therefore, selective breeding of resistant varieties is the preferred choice to control lily basal rot. The germplasm resources of *Lilium* species highly resistant to *F. oxysporum* have an important role in cultivating lily varieties against this disease. To date, no wild *Lilium* species or cultivars showing complete resistance to lily basal rot have been found, but several wild *Lilium* species and cultivars are highly resistant to it.^[@bibr4-1176934319838818]^ Among lily cultivars, Asian lily hybrid is highly resistant and Oriental lily is the least resistant.^[@bibr4-1176934319838818]^ The lily germplasm resources in China are abundant. Some wild species with strong disease resistance, such as *Lilium henryi, Lilium pumilum*, and *Lilium regale*, have been widely used in disease resistance breeding.^[@bibr4-1176934319838818]^ Although the germplasm resources of *Lilium* species resistant to *F. oxysporum* have been used in disease-resistant breeding, the molecular mechanism of disease resistance is rarely reported. In 2013, a cDNA library of the root of *F. oxysporum*--infested *Lilium regale* was constructed by suppression subtractive hybridization and 180 sequences homologous to known proteins in the model plant by sequencing analysis, most of which belonged to pathogenesis-related (PR) 3, PR10, glutathione S-transferase, cytochrome P450, antioxidant enzymes, and peroxidases, were obtained.^[@bibr5-1176934319838818]^ In 2012, *Lilium leichtlinii* var. *maximowiczii* Baker was used to construct suppression subtractive hybridization library of lily after induced by *F. oxysporum*. They identified 6 types of disease-resistant expressed sequence tags (ESTs), including serine/threonine protein kinase, glutathione S-transferase, peroxidase, and cyclophilin homologs.^[@bibr6-1176934319838818]^ The above 2 studies identified some genes of lily in response to *F. oxysporum* infection, but the mechanism of this resistance has not been investigated in depth. Some studies have used the molecular marker system to construct the genetic map of Asian lily backcross populations. To date, a total of 10 potential quantitative trait loci (QTLs) for *F. oxysporum* resistance have been mapped.^[@bibr7-1176934319838818]^,^[@bibr8-1176934319838818]^ However, none of these studies identified the specific genes involved in *F. oxysporum* resistance. Therefore, the molecular mechanism of resistance to *F. oxysporum* in lily needs to be further studied. In this study, *L. pumilum* that was highly resistant to *F. oxysporum* was used as the experimental material. High-throughput transcriptome sequencing technology combined with bioinformatics method was applied to identify genes related to *F. oxysporum* resistance in *L. pumilum*, understand the function of these genes, and fully integrate the regulatory network, thereby clarifying the molecular mechanism of lily resistance-associated genes in response to pathogen invasion. Our findings will lay a theoretical foundation for the cloning of disease-resistance-associated genes and for disease-resistance breeding. Materials and Methods {#section2-1176934319838818} ===================== The tissue culture seedlings of wild *L. pumilum* with the same genetic background were used in the experiment. The pathogen *F. oxysporum* was isolated from the plants with the symptom of lily basal rot in the Greenhouse of Beijing University of Agriculture. *L. pumilum* and *F. oxysporum* were stored in the Lily Breeding Laboratory, College of Landscape Architecture, Beijing University of Agriculture. *F. oxysporum* was inoculated on *potato dextrose agar* (PDA) medium and cultured at 27°C for 7 days. The spores were washed with sterile water and adjusted to 1 × 10^6^ spores/mL for the subsequent inoculation. The tissue culture seedlings of *L. pumilum* grew at 25°C under a 16/8 h light/dark period. The tissue culture seedlings of *L. pumilum* with a bulb diameter 1 to 1.5 cm and strong root tissue were selected for *F. oxysporum* inoculation. For each treatment, 3 bottles with 3 plant tissue culture seedlings in each bottle were used for inoculation, and each seedling was inoculated with 600 µL of spore suspension. *L. pumilum* tissue culture seedlings were inoculated with sterile water at the same time as the control group. The roots of the tissue culture seedlings in the treatment group and the control group were collected at 6, 12, or 24 h after inoculation and immediately frozen in liquid nitrogen. The samples were stored at −80°C for later use. The 3 time-point samples for *F. oxysporum* inoculation were named F6h, F12h, and F24h, respectively, and the corresponding samples for the control group were named M6h, M12h, and M24h. The total RNA of the samples was extracted using the RNAprep Pure Plant Kit (TIANGEN Ltd; Beijing, China). After the obtained sample was tested, the eukaryotic mRNA was enriched using magnetic beads with Oligo (dT). Subsequently, fragmentation buffer was added to break the mRNA into short fragments. The mRNA was used as the template to synthesize a single strand of cDNA, followed by the addition of buffer, dNTPs, DNA polymerase I, and RNase H to synthesize the double-stranded cDNA, which was purified with AMPure XP beads. The purified double-stranded cDNA was first subjected to end repair by adding poly-A tails and linking to the sequencing adaptor, and then AMPure XP beads were used for the selection based on fragment size. Finally, polymerase chain reaction (PCR) amplification was performed, and the PCR product was purified with AMPure XP beads to obtain the final library. After the library was constructed, Qubit 2.0 was used for the preliminary quantification, and the library was diluted to 1.5 ng/µL. The insert size of the library was detected using an Agilent 2100 Bioanalyzer. When the expected insert size was detected, the effective concentration of the library was accurately quantified (effective concentration of the library \>2 nM) to ensure the quality of the library. The constructed cDNA library was sequenced by Illumina HiSeq 2500 sequencing platform (Illumina, Inc, San Diego, CA, USA) from Novogene. The raw reads from sequencing were filtered. Clean reads were collected after removing reads with the adaptor and those of low quality. The clean reads were spliced by Trinity.^[@bibr9-1176934319838818]^ The resulting transcript sequence was used as a reference for the subsequent analysis. The longest transcript in each gene was used as a Unigene. Functional annotation was carried out for the obtained Unigenes with 7 major databases: National Center for Biotechnology Information (NCBI), non-redundant protein (Nr), NCBI nucleotide sequences (Nt), Protein family (Pfam), euKaryotic Ortholog Groups (KOG), Swiss-Prot, Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Ontology (GO). Using Bowtie2 software, the clean reads were aligned to the assembled Unigene library, and the results were statistically analyzed using RNA-Seq by Expectation Maximization (RSEM) software to further obtain the number of read counts for each gene sample corresponding to each gene.^[@bibr10-1176934319838818]^,^[@bibr11-1176934319838818]^ The expression of the Unigene was calculated using the fragments per kb per million fragments (FPKM) method.^[@bibr12-1176934319838818]^ The read-count data were first standardized using trimmed mean of M values (TMM), followed by the differential analysis using DEGSeq, with the screening threshold *q* \< 0.005 and \|log2FoldChange\| \> 1.^[@bibr13-1176934319838818]^ The obtained differentially expressed genes were subjected to GO functional enrichment analysis and KEGG pathway analysis. Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis was used to analyze the expression of genes responsive to *F. oxysporum* identified by RNA-seq. Total RNA of roots of the tissue culture seedlings were extracted according to the above methods. Total RNA was used for synthesizing reverse transcripts using the PrimeScript RT reagent Kit (TaKaRa, Dalian, China) according to the manufacturer's instructions. cDNA was amplified using the following primer pairs: UTP-glucose-1-phosphate uridylyltransferase, 5P (5′-AAAGGTTTCGGGTGGTTGTG-3′) and 3P (5′-TCCGACGACGGATACAACTG-3′); alpha-1,4 glucan phosphorylase, 5P (5′-CTCCGGCGTAGAAGGTATCG-3′) and 3P (5′-CGAGGATCCAAGGCAAAGAC-3′); phosphoglucomutase, 5P (5′-CGTCAACGATCCCTTCATCA-3′) and 3P (5′-AGGGTTCCTGCATCCAAAGA-3′); NADP-dependent glyceraldehyde-3-phosphate dehydrogenase (GAPDH), 5P (5′-CCCCTCCTCATCGACATAGC-3′) and 3P (5′-GCTCCTTTGTTCCCCAACAC-3′); and 18 rRNA, 5P (5′-GCCTGAGAAACGGCTACCAC-3′) and 3P (5′-ACCAGACTTGCCCTCCAATG-3′). Quantitative reverse transcription polymerase chain reaction was performed using SYBR Premix Ex Taq II (TaKaRa) under the following conditions: 40 cycles at 95°C for 15 s, 60°C for 30 s, and 72°C for 30 s. The qRT-PCR reactions were performed on the BIO-RAD iQ5 (Applied Biosystems, Foster City, CA, USA). Each sample was processed in triplicate, and the relative expression was calculated using 2^−ΔΔCT^.^[@bibr14-1176934319838818]^ The 18S rRNA was used as references to normalize the expression level of genes. Results and Discussion {#section3-1176934319838818} ====================== High-throughput transcriptome sequencing technology is an effective means of studying the molecular mechanisms of the plant without whole-genome information. To date, transcriptome sequencing has been used in studies of genetic divergence,^[@bibr15-1176934319838818]^ flavonoid biosynthesis,^[@bibr16-1176934319838818]^ carbohydrate metabolism,^[@bibr17-1176934319838818]^ cold response,^[@bibr18-1176934319838818]^ and vernalization^[@bibr19-1176934319838818]^ in lily. In this study, transcriptome sequencing was used to identify the genes of *L. pumilum* that were involved in responses to *F. oxysporum* infection. Transcriptomes of the root tissues of the *L. pumilum* at 6, 12, and 24 h after inoculation with *F. oxysporum* or sterile water (control) were sequenced. A total of more than 50 GB of raw reads were obtained in this study. After removing adaptors, low-quality reads, and contaminants, over 7 GB of clean reads were obtained from each sample ([Supplementary Table 1](https://journals.sagepub.com/doi/suppl/10.1177/1176934319838818)). The raw data (Accession Number: SRA633315) used in the study can be obtained from the sequence read archive (SRA) database. A total of 217 098 Unigenes with the average length of 517 bp and the N50 of 664 bp were obtained after the de novo assembly of the obtained clean reads ([Supplementary Table 2](https://journals.sagepub.com/doi/suppl/10.1177/1176934319838818)). The obtained Unigenes were aligned in 7 large databases, and 83 289 of the Unigenes were annotated in more than 1 database, accounting for 38.36% of the total Unigenes ([Supplementary Table 3](https://journals.sagepub.com/doi/suppl/10.1177/1176934319838818)). The number of Unigenes with successful annotation was the highest in Nr, accounting for 23.06% of the total Unigenes, and lowest in KOG in the KOG database, only 7.89% of Unigenes annotated. The analysis of differential gene expression at 6, 12, or 24 h after inoculation with *F. oxysporum* and sterile water was performed. There were 111 differentially expressed genes in F6h vs M6h, of which 61 were upregulated and 50 were downregulated. There were 254 differentially expressed genes in F12h vs M12h, of which 188 were upregulated and 66 were downregulated. There were 2500 differentially expressed genes in F24h vs M24h, of which 2009 were upregulated and 491 were downregulated ([Figure 1](#fig1-1176934319838818){ref-type="fig"}). {#fig1-1176934319838818} The results of GO enrichment analysis of the differentially expressed genes showed that the differentially expressed genes of F6h vs M6h were enriched in the category of molecular function (MF) with only 1 subset annotated as catalytic activity. The differentially expressed genes of F12h vs M12h were enriched in the categories of MF and BP (biological process), including 19 subsets with distribution mainly in catalytic activity, single-organism metabolic process, oxidation-reduction process, oxidoreductase activity, carbohydrate metabolic process, and riboflavin metabolic process. The differentially expressed genes of F24h vs M24h were enriched in the categories of BP, CC (cellular component), and MF, including 196 subsets, which were mostly annotated in the BP categories of metabolic process, single-organism metabolic process, and biosynthetic process; the CC categories of macromolecular complex, cytoplasm, and non-membrane-bound organelle; and the MF categories of catalytic activity, oxidoreductase activity, and structural molecule activity ([Figure 2](#fig2-1176934319838818){ref-type="fig"}). {#fig2-1176934319838818} The results of KEGG enrichment analysis of the differentially expressed genes showed that there were 7 KEGG pathways enriched among the differentially expressed genes of F6h vs M6h. The upregulated genes were significantly enriched in the metabolic pathway of flavonoid biosynthesis ([Supplementary Table 4](https://journals.sagepub.com/doi/suppl/10.1177/1176934319838818)). There were 18 KEGG pathways enriched among the differentially expressed genes of F12h vs M12h. The upregulated genes were significantly enriched in 13 metabolic pathways, mainly including vitamin B6 (VB6) metabolism, flavonoid biosynthesis, phenylalanine metabolism, phenylpropanoid biosynthesis, plant hormone signal transduction, and tyrosine metabolism ([Supplementary Table 4](https://journals.sagepub.com/doi/suppl/10.1177/1176934319838818)). There were 19 KEGG pathways enriched among the differentially expressed genes of F24h vs M24h. The upregulated genes were significantly enriched in 3 metabolic pathways, including biosynthesis of amino acids, proteasome, and ribosome ([Supplementary Table 4](https://journals.sagepub.com/doi/suppl/10.1177/1176934319838818)). Previous studies have shown that starch and sucrose metabolism changes when pathogens infect plants.^[@bibr20-1176934319838818]^,^[@bibr21-1176934319838818]^ Gómez-Ariza reported that defense-related genes were upregulated systemically in rice leaves after adding sucrose to the roots.^[@bibr22-1176934319838818]^ Sucrose can promote the lupine defense responses against *Fusarium*.^[@bibr23-1176934319838818]^ In this study, starch and sucrose metabolism was the pathway with the most upregulated genes at 12 h after inoculation, and a total of 9 genes were upregulated, including *sucrose-phosphate synthase* (*c117443_g1*); *UTP-glucose-1-phosphate uridylyltransferase* (*c95325_g1*); *aglucan phosphorylase* (*c112961_g2*); *alpha-1,4 glucan phosphorylase* (*c121101_g1*); *1,4-alpha-glucan branching enzyme* (c108888_g1); *phosphoglucomutase* (*c116359_g1*); *glucose-1-phosphate adenylyltransferase* (*c108264_g1* and *c117572_g1*); and *fructokinase-2-like* (*c94420_g1*) ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). The upregulation of these genes is conducive to transforming starch and glycogen into sucrose. At 24 h after inoculation, 9 genes in the starch and sucrose metabolism pathway were downregulated, including *sucrose-phosphate synthase* (*c113968_g1*), *pectinesterase* (*c74217_g1*), *glucose-1-phosphate adenylyltransferase* (*c108264_g1*, *c117572_g1*, and *c115394_g1*), *granule-bound starch synthase 2* (*c114915_g1*), *phosphoglucomutase* (*c116359_g1*), *alpha-amylase* (*c117598_g1*), and *beta-fructofuranosidase* (*c119079_g2*) ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). {#fig3-1176934319838818} Glyceraldehyde-3-phosphate dehydrogenase, the key enzyme in the glycolysis/gluconeogenesis pathway, is a multifunctional enzyme that is involved in the regulation of reactive oxygen species (ROS), autophagy, and plant immune responses.^[@bibr24-1176934319838818]^ The knockout of GAPDH in *Arabidopsis* enhances the resistance to the incompatible pathogen *Pseudomonas syringae* pv. *Tomato*.^[@bibr24-1176934319838818]^ In the present study, 5 genes in the glycolysis/gluconeogenesis pathway were upregulated at 12 h after inoculation with *F. oxysporum*, including *GAPDH* (*c103781_g1*), *NADP-dependent GAPDH* (*c105111_g1*), *phosphoglucomutase* (*c116359_g1*), *glucose-6-phosphate 1-epimerase* (*c111964_g1*), and *phosphoglycerate kinase* (*c109005_g2*). The expression levels of these genes were increased by 2.5-fold and 4.9-fold. The expression level of *GAPDH* (*c103781_g1*) was 2.5 times the levels of the control, but there was no significant difference between the expression of these genes and the control at 24 h after inoculation ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). In the present study, 6 genes in the phenylpropanoid biosynthesis pathway were highly expressed at 12 h after inoculation with *F. oxysporum*, including *4-coumarate-CoA ligase 2* (*c115266_g1*), *caffeoyl-CoA O-methyltransferase* (*c105247_g1*), *cinnamyl-alcohol dehydrogenase* (*c119082_g1*), and *peroxidase* (*c118468_g3*, *c115650_g1*, and *c120335_g1*). At 24 h after inoculation, the expression levels of 13 genes in this pathway were downregulated, including *phenylalanine ammonia-lyase* (*c103771_g1*), *cinnamyl-alcohol dehydrogenase* (*c117898_g1*, *c118292_g2*, and *c119633_g1*), *peroxidase* (*c115552_g2*, *c101935_g1*, *c116520_g1*, *c101806_g1*, *c96040_g1*, *c116501_g1*, *c120735_g1*, and *c103286_g1*), and *cationic peroxidase 1* (*c80387_g1*) ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). Flavonoids are involved in the resistance of wheat against *F. graminearum*^[@bibr25-1176934319838818]^ and the resistance of *Brassica napus* against *Xanthomonas campestris* pv. *Campestris*.^[@bibr26-1176934319838818]^ In the present study, the expression levels of 3 *chalcone synthase* genes (*c113690_g1*, *c113916_g1*, and *c113916_g2*) in the flavonoid biosynthesis pathway were upregulated at 6 and 12 h after inoculation with *F. oxysporum. Caffeoyl-CoA O-methyltransferase* (*c105247_g1* and *c105247_g2*), which is involved in the synthesis of feruloyl-CoA, was upregulated at 12 h after inoculation ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). Caffeoyl-CoA O-methyltransferase is also involved in the phenylpropanoid pathway and lignin production, playing an important role in the resistance process of maize against *Cochliobolus heterostrophus, Cercospora zeae-maydis*, and *Setosphaeria turcica*.^[@bibr27-1176934319838818]^ At 12 h after inoculation with *F. oxysporum*, some genes in the jasmonic acid (JA), abscisic acid (ABA), and auxin signaling pathways were induced. Jasmonic acid is an endogenous growth regulator in higher plants and a stress signal molecule that will accumulate rapidly when the plant tissues are attacked by pathogens or insects. The transcription factor *MYC4* (*c114016_g1* and *c113516_g3*) of the JA signaling pathway was upregulated in the present study. Abscisic acid can negatively regulate the disease resistance in plants by inhibiting salicylic acid (SA) signaling.^[@bibr28-1176934319838818][@bibr29-1176934319838818]--[@bibr30-1176934319838818]^ In the present study, the expression of *Protein phosphatase 2C 51* (*c115633_g1*), which negatively regulates ABA signaling, was significantly upregulated.^[@bibr31-1176934319838818]^ *GH3* in the auxin signaling pathway can positively regulate SA-mediated plant disease resistance.^[@bibr32-1176934319838818]^,^[@bibr33-1176934319838818]^ The gene of auxin-responsive *GH3* gene family (*c117107_g1*) was significantly upregulated in this study ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). Vitamin B6 participates in the resistance process of *Arabidopsis thaliana* against *Pseudomonas syringae* pv. *Tomato* DC3000 and *Botrytis cinerea*.^[@bibr34-1176934319838818]^ Pyridoxal phosphate is the enzyme in the last step of the process of VB6 biosynthesis. In the present study, the expression levels of *pyridoxal phosphate* (*c121490_g2*, *c121490_g1*, and *c108372_g2*) at 12 h after inoculation with *F. oxysporum* were 14.9, 13, and 119.4 times those of the control, respectively ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). Inoculation of pathogens can result in the accumulation of different amino acids in the host plant.^[@bibr35-1176934319838818]^ The inoculation of pathogens in *Arabidopsis* can result in changes in the expression of some enzymes in the biosynthesis of amino acids.^[@bibr36-1176934319838818]^,^[@bibr37-1176934319838818]^ Homoserine is a precursor of threonine, isoleucine, and methionine. The mutation of *downy mildew resistant* (*DMR1*)--encoded homoserine kinase leads to the accumulation of homoserine. Mutations in *dihydrodipicolinate synthase 2* (*DHDPS2*) and *aspartate kinase 2* (*AK2*) genes can lead to the accumulation of the amino acids threonine, methionine, and isoleucine. Mutations in the *DMR1, DHDPS2*, and *AK2* genes are reported to enhance resistance to *Hyaloperonospora* in *Arabidopsis*.^[@bibr38-1176934319838818]^,^[@bibr39-1176934319838818]^ In this study, a total of 69 synthase genes were upregulated in acid biosynthesis pathways at 24 h after inoculation with *F. oxysporum*. These enzymes are involved in the synthesis of multiple amino acids, including threonine, isoleucine, tryptophan, tyrosine, lysine, valine, leucine, phenylalanine, arginine, and proline ([Figure 4](#fig4-1176934319838818){ref-type="fig"}). {#fig4-1176934319838818} The ubiquitin-26S proteasome system is an important post-translational protein regulation system in eukaryotes. It plays an important role in pathogen defense,^[@bibr40-1176934319838818]^,^[@bibr41-1176934319838818]^ abiotic stress tolerance,^[@bibr42-1176934319838818]^ hormone signaling,^[@bibr43-1176934319838818]^ morphogenesis,^[@bibr44-1176934319838818]^ and chromatin modification^[@bibr45-1176934319838818]^ in plants. In the present study, the expression levels of 24 subunits of the 26S proteasome in lily root were significantly increased to more than 32 times the levels of the control, and the expression levels of all 14 subunits of the 20S proteasome were significantly increased at 24 h after inoculation with *F. oxysporum* ([Figure 5](#fig5-1176934319838818){ref-type="fig"}). {#fig5-1176934319838818} Ribosomal proteins can maintain the stability of the ribosomal complex and play an important role in the synthesis of proteins. Several ribosomal proteins are involved in the disease resistance of plants.^[@bibr46-1176934319838818]^,^[@bibr47-1176934319838818]^ In this study, 86 ribosome protein-coded genes were significantly upregulated 24 h after inoculation, including 37 genes coding for small-subunit proteins and 39 genes coding for large-subunit proteins. All 86 ribosome protein genes were upregulated by more than 32-fold, and most of the genes were upregulated by approximately 1000-fold ([Figure 6](#fig6-1176934319838818){ref-type="fig"}). {#fig6-1176934319838818} Transcription factors can regulate target genes by binding to specific *cis*-regulatory elements in their promoters.^[@bibr48-1176934319838818]^ The transcription factors of the WRKY, AP2/ERF, MYC, bZIP, and MYB families are involved in the regulation of defense gene expression when attacked by pathogens.^[@bibr49-1176934319838818]^,^[@bibr50-1176934319838818]^ MYC transcription factors are important regulators of JA-responsive genes, negatively regulating the expression of the defense genes in *Arabidopsis*.^[@bibr49-1176934319838818]^,^[@bibr51-1176934319838818]^ ERF is a transcription factor of the AP2/ERF subfamily and is involved in the defense responses of plants to pathogenic bacteria via the SA and Ethylene (ET)/JA-dependent signal transduction pathways.^[@bibr52-1176934319838818]^,^[@bibr53-1176934319838818]^ The transcription factors of the WRKY family are involved in the protection response of plants against various pathogens.^[@bibr54-1176934319838818]^ In this study, 20 differentially expressed transcription factors at 3 time points were identified ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). The number of differentially expressed transcription factors was significantly increased at 24 h after inoculation, and all 12 transcription factors in the WRKY and ERF families had upregulated expression. At 6 h after inoculation, 2 *transcription factors MYC4* (*c113516_g3* and *c114016_g1*) were upregulated. The c113516_g3 was upregulated 97-fold at 6 h after inoculation, but no change in the gene expression was observed at 12 and 24 h after inoculation. At 24 h after inoculation, 7 ERF members were upregulated, including *Ethylene-responsive transcription factor 1A* (*c118561_g1*) by 24.3-fold. *WRKY transcription factor 33* (*c96521_g1*) was upregulated by 2.7-fold and 12.1-fold at 12 h and 24 h after inoculation respectively. *WRKY transcription factor 33* (*c113118_g1*) and *WRKY transcription factors 2, 70*, and *40* were upregulated by 3.7-fold and 8.6-fold at 24 h after inoculation ([Figure 3](#fig3-1176934319838818){ref-type="fig"}). Therefore, we speculate that WRKY and ERF proteins play important roles in the resistance of *L. pumilum* to *F. oxysporum*. The role of MYC in the responses of *L. pumilum* to *F. oxysporum* needs to be further investigated. To validate the results obtained by high-throughput sequencing, 4 genes were selected for further confirmation by qRT-PCR. Our results showed that the 4 genes including *UTP-glucose-1-phosphate uridylyltransferase* (*c95325_g1*), *alpha-1,4 glucan phosphorylase* (*c121101_g1*), *phosphoglucomutase* (*c116359_g1*), and *NADP-dependent GAPDH* (*c105111_g1*) showed significantly higher expression in 12 h after inoculation with *F. oxysporum* ([Figure 7](#fig7-1176934319838818){ref-type="fig"}). Thus, our qRT-PCR results were consistent with those using Illumina sequencing method. {#fig7-1176934319838818} In this study, transcriptome sequencing was used to study resistance responses during early stages of *L. pumilum* infected by lily pathogen *F. oxysporum*. Biological pathways and transcription factors related to the responses of *L. pumilum* to *F. oxysporum* were identified using bioinformatics analysis. Our research greatly improves the current understanding of molecular mechanism in *Lilium* to resist *F. oxysporum*. Supplementary Material ====================== ###### Supplementary material ###### Supplementary material ###### Supplementary material ###### Supplementary material The authors thank Zhengping Liu for her assistance in the isolation of *Lilium* pathogen *F. oxysporum*. **Funding:**The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Outstanding Talent Project of the Organizational Department of the Beijing Municipal Committee (2014000020124G079), the Beijing Municipal Education Commission (CEFF-PXM2017_014207_000043 and SQKM201810020009), and Building Project of Beijing Laboratory of Urban and Rural Ecological Environment (PXM2015-014207-000014). **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. **Author Contributions:** XH, SX, and WW conceived and designed the experiments; WL, WZ, and XJ performed the experiments; XH and WL analyzed the data; and XH, SX, WW, AGS, and TA wrote the paper. **Supplemental Material:** Supplemental material for this article is available online. **ORCID iD:** Tessema Aynalem  <https://orcid.org/0000-0003-1709-3234>

**A Primer of Ecology with R** M. Henry H. Stevens New York:Springer, 2009. 388 pp. ISBN: 978-0-387-89881-0, \$64.95 **Acute Exposure Guideline Levels for Selected Airborne Chemicals: Vol. 8** Committee on Acute Exposure Guideline Levels; Committee on Toxicology; National Research Council Washington, DC:National Academies Press, 2009. 434 pp. ISBN: 978-0-309-14515-2, \$84.50 **Carbon Sinks And Climate Change Forests in the Fight Against Global Warming** Colin A.G. Hunt Northhampon, MA:Edward Elgar Publishing, 2009. 256 pp. ISBN: 978-1-84720-977-1, \$110 **Contemporary Issues In International Environmental Law** Malgosia Fitzmaurice Northhampon, MA:Edward Elgar Publishing, 2009. 256 pp. ISBN: 978-1-84542-283-7, \$99 **Ecological Economics, 2nd ed.: Principles and Applications** Herman Daly, Joshua Farley Washington, DC:Island Press, 2009. 488 pp. ISBN: 978-1-59726-681-9, \$75 **Environmental Law Handbook, 20th ed.** Daniel M. Steinway, et al. Cincinnati, OH:ACGIH, 2009. 975 pp. ISBN: 978-1-60590-278-4, \$99 **Global Sources of Local Pollution: An Assessment of Long-Range Transport of Key Air Pollutants to and from the United States** Committee on the Significance of International Transport of Air Pollutants; National Research Council Washington, DC:National Academies Press, 2009. 250 pp. ISBN: 978-0-309-14398-1, \$42 **Governing The Environment: Salient Institutional Issues** Albert Breton, Giorgio Brosio, Silvana Dalmazzone, Giovanna Garrone, eds. Northhampon, MA:Edward Elgar Publishing, 2009. 978 pp. ISBN: 978-1-84720-397-7, \$140 **Inorganic Chromium (III) Compounds** World Health Organization Geneva:WHO Press, 2009. 102 pp. ISBN: 978-9-2415-3076-7, \$20 **International Handbook On The Economics of Energy** Joanne Evans, Lester C. Hunt, eds. Northhampon, MA:Edward Elgar Publishing, 2009. 848 pp. ISBN: 978-1-84720-352, \$330 **Night Noise Guidelines for Europe** WHO Regional Office for Europe Geneva:WHO Press, 2009. 180 pp. ISBN: 978-9-2890-4173-7, \$40 **Our Choice: A Plan to Solve the Climate Crisis** Al Gore London:Bloomsbury Publishing, 2009. 416 pp. ISBN: 978-0-7475-9098-9, \$20.90 **Technonatures: Environments, Technologies, Spaces, and Places in the 21st Century** Damian F. White, Chris Wilbert, eds. Waterloo, ON:Wilfrid Laurier University Press, 2009. 282 pp. ISBN: 978-1-55458-150-4, \$38.95

All protein array data files are available from the Array Express database (accession number E-MTAB-7622). Introduction {#sec001} ============ The absence of Reelin--a glycoprotein crucial for physiological retinogenesis--has been recently associated with changes in both Nerve Growth Factor (NGF) protein and mRNA in the retina \[[@pone.0212732.ref001]--[@pone.0212732.ref003]\]. NGF and Reelin have been reported to take actively part in neurogenesis and retinogenesis \[[@pone.0212732.ref004]--[@pone.0212732.ref006]\]. NGF has been hypothesized to interact with Reelin by modulating neuronal migration, neurodevelopment and other physiological processes in the central nervous system and retina \[[@pone.0212732.ref001],[@pone.0212732.ref002],[@pone.0212732.ref007]\]. NGF activities encompass cell proliferation, cytoskeletal reorganization, migration, differentiation, survival and/or apoptosis \[[@pone.0212732.ref004],[@pone.0212732.ref008]\]. In the retina, NGF modulates retinal cell development, differentiation and functional activity, and promotes survival/recovery of Retinal Ganglion Cells (RGCs), photoreceptors and optic axons after experimental injuries as well as normal functional and anatomical development of visual acuity and binocularity \[[@pone.0212732.ref004],[@pone.0212732.ref008]--[@pone.0212732.ref010]\]. The neurodegenerative process occurring in *Reeler* retina evolves through a series of changes at different cell types (neural, vascular and glial cells) and comprises several overlapping/interrelated molecular pathways \[[@pone.0212732.ref005]\]. Glial cell activation (astrocytes, Müller cells and resident microglia) represents a crucial step for protecting neurones from degeneration \[[@pone.0212732.ref007]\]. Müller cells work in concert with other glial cells and neurones to guarantee optimal development of retinal structure \[[@pone.0212732.ref009]--[@pone.0212732.ref012]\]. An open question regards the glia cell activation upon Reelin deficiency and the possibility for NGF to maintain retinal homeostasis via glial cell activation, as observed in other neuronal degenerating tissues \[[@pone.0212732.ref013],[@pone.0212732.ref014]\]. Therefore, the aim of this study was to look for some proinflammatory/profibrogenic mediator changes in the vitreous and retina as well as glial cell activation in the retina of *Reeler* mice. Materials and methods {#sec002} ===================== Animals and ethical approval {#sec003} ---------------------------- Eighteen (18) animals were used for the study, including 9 *Reeler* (RELN^-/-^; 9--11 gr body-weight) and 9 WT (RELN^+/+^; B6C3Fe; 12--14 gr body-weight) mice (Charles River, Calco, Como). Experimental procedures were approved by the Ethical Committee of Tor Vergata University (Rome, Italy), according with ethical standards stated in the Declaration of Helsinki and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. All the steps in the procedure were in compliance with the directive 2010/63/EU guidelines, under the authorization provided by the Italian Ministry of Health. All efforts were made to minimize suffering. All analytic and molecular grade reagents were purchased from Carlo Erba (Milan, Italy), Euroclone (Milan, Italy) and Sigma (Milan, Italy), otherwise specified in the text. Daily produced MilliQ RNAse-free water was provided for biochemical and molecular purposes (Direct Q5; Millipore Corp., Billerica, MA). Experimental procedure: Vitreous and retina {#sec004} ------------------------------------------- At postnatal day (p) p14, p21 and p28, mice were anaesthetized by intraperitoneal injection of 2 mg/mL ketamine (0.2 mL/10 gr body-weight; Ketavet, Gellini Pharmaceutics, Italy) and 0.23 mg/mL medetomidine (0.24 mL/10 gr body-weight; Domitor, Orion Corp., Espoo, Finland) mixture. Sampling was carried out under a dissecting microscope (SMZ645; Nikon, Tokyo, Japan) equipped with cold-light optic fibers (PL2000 photonic; Axon, Vienna, Austria), as previously reported with slight modifications \[[@pone.0212732.ref002]\]. A corneal incision was produced and lens, retina and vitreous were collected in a microvial with separating membrane. Centrifugation (13.000rpm/15min) was performed to detach vitreous from retina and lens. Vitreous (left/right eyes) and retina (right eye) were appropriately stored for biochemical and molecular studies. Other retinas (left eye) were used for immunofluorescent analysis. Vitreous and retina (n = 3/time-point; *Reeler* and WT mice) were diluted / extracted in modified RIPA Buffer (50mM Tris-HCl, 150mM NaCl, 1% Triton-X100, 5mM EDTA, 100mM NaF and 1mM PMSF; pH 7.5) and finally sonicated (VibraCell Sonics, Inc., Newtown, USA), according to a standard procedure \[[@pone.0212732.ref002]\]. Total proteins were quantified with Nanodrop Spectrophotometer (A1000, Celbio, Milan, Italy). Protein array {#sec005} ------------- A customised chip-based array was used to quantify inflammatory/profibrogenic factors in vitreous and retinal lysates, between a list of potential candidates (G-series arrays; Ray Biotech, Norcross, CA). Each glass-slide comprised 14 identical sub-arrays containing 50 factors (antibody spots in duplicate) retrospectively selected by literature search \[[@pone.0212732.ref015]--[@pone.0212732.ref018]\]. *Reeler* and WT samples were processed simultaneously. Briefly, normalized protein extracts (50μg total protein; 70μL per well) were diluted in appropriate buffer and hybridized in sub-arrays. Washing, detection and labelling steps were performed according to the manufacturer's recommendation. Spin-dried slides were scanned in a GenePix 4100A Microarray platform (Molecular Devices LLC, Sunnyvale, Silicon-Valley, CA). Capturing conditions and image digital acquisitions were done as previously reported \[[@pone.0212732.ref019]\]. Images were uniformly adjusted for size, brightness, contrast and chip-to-chip comparisons by the software and provided as 8-bit Tiff format (Axon GenePix Pro 6.0.1.25 software; Molecular Devices). Inter-assay normalization was guaranteed by the presence of multiple internal controls for each sub-array. The sensitivity range was 3.8--56 pg/mL, as provided by the manufacturer. Microarray data are available in the ArrayExpress database (<http://www.ebi.ac.uk/arrayexpress>) under accession number E-MTAB-7622. NGF ELISA {#sec006} --------- Samples were further diluted 1:2 in assay diluent (10mM PB, 150mM NaCl, 0.5% BSA, 0.1% Triton X100 and 1x protease inhibitor cocktail (Pierce---Thermo Fisher Scientific Inc.; Waltham, MA USA); Ph 7.5). Briefly, 96-well Maxisorp enzyme-linked immunoassay plates (Nunc, Roskilde, Denmark) were precoated with monoclonal anti-βNGF antibodies (0.4 μg/mL; MAB256; R&D Systems Inc, Minneapolis, Minnesota, USA). Standards (0.15 pg/mL to 1 ng/mL βNGF; Alomone Labs, Jerusalem, Israel) and samples were incubated at 4°C for 18 hours. ELISA was developed by using polyclonal biotinylated anti-NGF antibodies (0.15 μg/mL, BAF256; R&D), horseradish peroxidase streptavidin (1:300; DY998, R&D) and the ready-to-use TMB substrate (eBioscience, San Diego, CA, USA). Under these conditions, no cross reactivity with Brain Derived Neutrophic Factor (BDNF) or Neutrophins 3/4/5 was observed. The colorimetric signals (Optic Density, OD; λ450-570nm) were quantified using the Sunrise plate reader (Tecan Group Ltd., Männedorf, Switzerland) and the related mean values (pg/mL) were produced according to 3^rd^ grade polynomial standard curve and normalized to total protein content (A280; Nanodrop). Total RNA extraction, cDNA synthesis and real-time PCR analysis {#sec007} --------------------------------------------------------------- Total RNA was extracted from retinas according to the TRIfast procedure (EuroClone) and rehydrated in 10μL fresh RNAse free MilliQ water, before treating with RNase-Free DNaseI (2U/μL; AM-1907; Turbo DNA free kit; Ambion Ltd., Huntingdon, Cambridgeshire, UK). Quantity and purity (\>1.8; A280 program, Nanodrop) as well as sign of RNA degradation (1% agarose gel analysis) were checked. cDNAs were generated from normalised templates (1μg RNA) (ImProm-II Reverse Transcription System; Promega Corp., Madison, USA) in a One Cycler programmable thermocycler (PeqLab Biotech, Erlangen, Germany) and amplified using the SYBR Green PCR core reagent kit (Applied Biosystems, Foster City, CA) in Eco Real Time PCR thermocycler equipped for 48-well plate (Illumina, MA, USA). Specific primers and amplification procedure are summarised in **[Table 1](#pone.0212732.t001){ref-type="table"}**. Samples were amplified in duplicate and in parallel with negative controls (either without template or with mRNA as template). Real Cycle numbers (Cn) were recorded and normalized for referring genes run in parallel (nCq = Cq~target~---Cq~referring~). The Cq averages were calculated from these replicates and showed as expression ratios (log2-scale) of a normalized target gene REST analysis \[[@pone.0212732.ref020]\]. 10.1371/journal.pone.0212732.t001 ###### Primers for real time PCR. {#pone.0212732.t001g} -------------------------------------------------------------------------- TARGET Sequence[^\#^](#t001fn002){ref-type="table-fn"} Amplicon\ Length ------------ ------------------------------------------------- ----------- **IL33** F: `5’-TGAGTCTCAACACCCCTCAA-3’` 169 **MIP3α** F: `5'-CTCCTGGCTGCTTTGATGTC-3'` 151 **MIP3β** F: `5’-GTGCCTGCTGTAGTGTTCACC-3’` 133 **GS** F: `5’-GCCCCCTATCAAGGAACTT-3’` 173 **Nestin** F: `5’-GGCCATGACTCTGACCTCTC-3’` 190 **GAPDH** F: `5’-GTGGACCTCATGGCCTACAT-3’` 117 -------------------------------------------------------------------------- Amplification procedure: initial hot start activation (95°C/15min) followed by 39 cycles of Denaturation (94°C/10s) / Annealing (58°C/15s)/Extension (75°C/10s) and melting curve generation (55°C---95°C with one fluorescence reading every 0.5°C). ^\#^Forward sequences are reported. Epifluorescent analysis and integrated optical densitometry {#sec008} ----------------------------------------------------------- Post-fixed and cryoprotected eyes were quickly frozen in dry-ice, embedded in OCT medium (TissueTek; Leica, Heidelberg, Germany) and sectioned (CM3050 cryostat; Leica Microsystems, Rijswijk, Netherland). Serial sections (10μm) were placed on BDH slides (Milan, Italy) and stored at -20°C. Antigen retrieval (0.05% trypsin-EDTA solution, 2min) and blocking/permeabilizing (1% BSA and 0.5% Triton X100 in PBS, 15min) steps were performed. The specific antibodies were anti-GFAP (1:100; G3893; Sigma-Aldrich, St. Louis, MO, USA; 1:50; \#AB5804; Merck-Millipore, Darmstadt, Germany), anti-Nestin (1:500; NB300-266; Novus Biologicals, Littleton, CO, USA) and anti-CD45 (1:100; sc-1178; Santa Cruz Biotech, Santa Cruz, CA, USA). The secondary antibodies were Cy2 (green) and Cy3 (red) conjugated species-specific antibodies (1:1000; donkey; Jackson ImmunoResearch, Europe Ltd, Suffolk, UK). DAPI was used for nuclear counterstaining (D9542; Sigma-Aldrich, St. Louis, MO, USA). Negative controls (isotypes) were carried out in parallel with the omission of primary antibodies and used for appropriate background subtractions. Serial images were acquired by NIS software connected to Epifluorescent direct microscope (Eclipse N*i*; Nikon, Tokyo, Japan). Data management and statistical analysis {#sec009} ---------------------------------------- Graphics were assembled by using the Prism5software (GraphPad software Inc., La Jolla, CA, USA), while statistical analysis was carried out with the StatView II software for PC (Abacus Concepts. Inc., Barkley, CA, USA). A p\<0.05 value was considered statistically significant. For chip-based array, the Fluorescent Intensity (FI) values were obtained by subtracting the background signal (GenePix Pro 6.0.1.25 software--Molecular Devices). Single FI values were entered into a Microsoft Excel database (Microsoft, Redmond, WA, USA) and duplicate spots outside the 10% coefficient of variability were refused from the statistical analysis. FDR value of 0.01 was set. FI averages (means±SD) were calculated from replicates (2 spots) of not-pooled samples. The following cut-offs were used: Fold Changes ≥ or ≤ 2 (FC; herein defining a given candidate factor with respect to control) and pValues ≤ 0.001 for multiple testing (p = 0.05/50 targets; two tails T-test followed by Bonferroni Correction). For Integrated optical Density (IntDen), the 8-bit TIFF saved digital images (512x512 or 1024x1024 dpi; n = 5 sections/slide; x40/dry 0.75 DIC M/N2) were subjected to single analysis with the ImageJ v1.43 (NIH-<http://rsb.info.nih.gov/ij/>). IntDen data (mean±SD per retina field) were calculated, grouped and subjected to statistical analysis. Results {#sec010} ======= A specific protein profile characterizes vitreous at p14, p21 and p28 {#sec011} --------------------------------------------------------------------- To characterize the pro-inflammatory/fibrogenic profile, both vitreous and retinal extracts were analyzed for protein array (see above). The Volcano plots in **[Fig 1A](#pone.0212732.g001){ref-type="fig"}** highlight the changes of interest as detected at all time-points investigated. At p14 (left panel), changes for NGF (-4.45 FC; p\>0.05) and IL33 (-3.57 FC; p\>0.05) expression were observed in *Reeler* vitreous with respect to WT. At p21 (middle panel), MIP3β (-3.15 FC; p\>0.05) and MIP3α (-2.36 FC; p\>0.05) were decreased while MMP7 (3.79 FC; p\>0.05), MMP13 (3.53 FC; p\>0.05), NGF (2.73 FC; p\>0.05) and IL33 (2.60 FC; p\>0.05) were increased in *Reeler* vitreous with respect to WT. At p28 (right panel), NGF (7.23 FC; p\<0.001), IL33 (4.96 FC; p\<0.001) and TIMP1 (4.79; p\<0.001) continued to be increased in *Reeler* vitreous while NT3 (-4.32 FC; p\<0.05) was significantly decreased in *Reeler* vitreous as compared to WT. A trend to an increase was observed for MMP9 (2.76 FC; p\>0.05) and as well as for sTNF-RI/II (2.12 and 2.11; p\>0.05). Significant results are listed in **[Table 2](#pone.0212732.t002){ref-type="table"}**. {#pone.0212732.g001} 10.1371/journal.pone.0212732.t002 ###### Sketch of significant inflammatory/profibrogenic mediators in p28 *Reeler* vitreous. {#pone.0212732.t002g} Mediator Function FC (Sign) Cell origin in the retina \[ref\] ----------- ---------------------------------------------- --------------- --------------------------- -------------------------------------------------------------------------- **βNGF** Growth factor **7.23 (\*)** **all cells** \[[@pone.0212732.ref001],[@pone.0212732.ref002],[@pone.0212732.ref050]\] **IL33** Pro-inflammatory and pro-fibrogenic cytokine **4.96 (\*)** **glial cells** \[[@pone.0212732.ref035],[@pone.0212732.ref040]\] **TIMP1** Metalloproteinase inhibitor **4.79 (\*)** **glial cells** \[[@pone.0212732.ref052]\] Values obtained from protein chip array showing the fold changes (FC) in the expression of some inflammatory/profibrogenic mediators at p28 *Reeler* retina as compared to WT. Significant values (Sign) are labeled by asterisk (\*, p\<0.05) or ns (not significant). As shown in **[Fig 1B](#pone.0212732.g001){ref-type="fig"}**, the comparison between p28 and p14 *Reeler* protein profiles confirmed the significant increase of NGF expression in the vitreous. Interesting, the NGF increase was time-point dependent, as validated by ELISA (**[Fig 1C](#pone.0212732.g001){ref-type="fig"}**; Particularly, NGF levels were significantly increased at p28 (24.663±0.045 *vs*. 15.770±0.050 pg/mL; p\<0.01, *Reeler vs*. WT) with respect to p21 (16.049±0.034 *vs*. 12.456±0.016 pg/mL; p\<0.05, *Reeler vs*. WT), both as compared to related WT counterparts. Only a trend to a decrease was detected for NGF at p14 (9.785±0.063 *vs*. 11.475±0.094 pg/mL; p\>0.05, *Reeler vs*. WT). IL33, MIPs, Glutamine Synthetase (GS) and Nestin mRNAs' expression in *Reeler* retinas {#sec012} -------------------------------------------------------------------------------------- MIPs and IL33 changes were verified p28 *Reeler* and related WT retinal extracts by real time PCR. As shown in **[Fig 2A](#pone.0212732.g002){ref-type="fig"}**, a significant upregulation of IL33 transcript was detected in *Reeler* retinas (3.125±0.227 ~2log-ratio~; p\<0.01, *Reeler vs*. WT). By contrary, MIP3α (-1.512±0.691 ~2log-ratio~) and MIP3β (-2.678±0.335 ~2log-ratio~) transcript downregulations were observed in *Reeler* retinas (p\>0.05; *Reeler vs*. WT). {#pone.0212732.g002} Müller cell activation was investigated by means of GS and Nestin mRNA expression. A significant upregulation of mRNAs' specific for GS (p14: 0.743 ± 0.103 ~2log-ratio~; p21: 3.041 ± 0.075 ~2log-ratio~ and p28: 2.694 ± 0.210 ~2log-ratio~; p\<0.05, *Reeler vs*. WT) and Nestin (p14: 1.774 ± 0.124 ~2log-ratio~; p21: 4.130 ± 0.072 ~2log-ratio~ and p28: 3.423 ± 0.085 ~2log-ratio~; p\<0.01, *Reeler vs*. WT) was observed as displayed in **[Fig 2B](#pone.0212732.g002){ref-type="fig"}**. Reactive müller cells populate *Reeler* retinas at p28 {#sec013} ------------------------------------------------------ Serial sections were immunostained for GFAP, Nestin and CD45 (leukocyte common antigen expressed by T/B lymphocytes, granulocytes, monocytes/macrophages and retinal microglia) specific antibodies \[[@pone.0212732.ref021]\]. Vimentin and GS were not investigated by immunofluorescence. An increased immunoreactivity for GFAP (red) was observed in *Reeler* retinas (**[Fig 3B](#pone.0212732.g003){ref-type="fig"}**) with respect to WT (**[Fig 3A](#pone.0212732.g003){ref-type="fig"}**), as merged on a blue nuclear staining. Increased GFAP immunoreactivity was mainly localized at the Ganglion Cellular Layer (GCL), although a slight immunostaining was also detected at the inner nuclear layer (INL; see arrows). A co-expression of GFAP (red) and Nestin (green) was observed in *Reeler* retinas, as pointed by arrow in the white frame (**[Fig 3B](#pone.0212732.g003){ref-type="fig"}**), with respect to WT retina (**A**). This GFAP-Nestin co-expression in *Reeler vs*. WT retinas is better shown in related magnifications (**[Fig 3D](#pone.0212732.g003){ref-type="fig"}** *vs*. **[3C](#pone.0212732.g003){ref-type="fig"}**). A positive correlation between GFAP and Nestin was provided by Kendall analysis (**[Fig 3E](#pone.0212732.g003){ref-type="fig"}**). The histogram showing the densitometric analysis for GFAP (28.25±3.32 *vs*. 13.29±1.30 IntDen; p\<0.05, *Reeler vs*. WT) and Nestin (43.66±9.24 *vs*. 12.73±1.23 IntDen; p\<0.05, *Reeler vs*. WT) is displayed in **[Fig 3F](#pone.0212732.g003){ref-type="fig"}**. Either alone or in combination with GFAP, not significant changes were observed for CD45 immunoreactivity between *Reeler* and WT retinas (18.19±3.38 vs. 17.30±1.71 IntDen; p\>0.05, Reeler vs. WT). {#pone.0212732.g003} Discussion {#sec014} ========== Herein, we show for the first-time changes in vitreous proteins and the presence of reactive Müller cells in the retinas of *Reeler* mice. The results are below discussed. Our previous studies highlighted the overexpression of NGF and NGF receptors (mRNA/protein) in *Reeler* retina at p14, p21 and particularly p28, postulating a potential compensatory NGF activity in Reelin deficiency \[[@pone.0212732.ref001],[@pone.0212732.ref002]\]. Albeit Reelin deficient retina does not show any crucial inflammatory state, the presence of several cell-to-cell and cell-to-mediator events cannot be excluded during the neurodegenerative process that might include the expression and/or modulation of gene and related products \[[@pone.0212732.ref001],[@pone.0212732.ref003]\]. Therefore, to better characterize retinal microenvironment and provide further information on NGF interplays in Reelin-impaired retinogenesis, possible mediator changes were investigated in vitreous and retina from *Reeler* and WT mice, at the same time-points (p14-p21-p28). First, increased expression for NGF, IL33 and TIMP1 proteins as well as a trend to an increase for MMP9 and TNFR-I/II were quantified in p28 *Reeler* vitreous. Vitreous represents a precious tool to support retinal disorder management (personalized medicine), being a reservoir of active mediators and an indicator of the underneath suffering retina, especially as vitreal reflux \[[@pone.0212732.ref016],[@pone.0212732.ref022],[@pone.0212732.ref023]\]. Vitreous is a normal clear jelly-form ocular fluid composed of collagens, sulphated proteoglycans and hyaluronan \[[@pone.0212732.ref024]\]. Widely, pathological vitreous is filled of inflammatory/angiogenic factors (cytokines, chemokines, growth factors and enzymes of extracellular matrix), depending on the grade of retinal inflammation/degeneration conditions or even microcirculation / systemic influences \[[@pone.0212732.ref025],[@pone.0212732.ref026]\]. The detection of changes in NGF, IL33 and TIMP1 as well as MMP9 and soluble TNFR-I/II proteins in *Reeler* vitreous suggest the presence of a "low" proinflammatory/profibrogenic profile, in line with previous studies \[[@pone.0212732.ref027]\]. This *Reeler*-associated protein profile might be consistent with the degenerative condition due to neuron-to-neuron impairments, morphological changes and reduced retinal cell migration (rod bipolar cells) \[[@pone.0212732.ref005],[@pone.0212732.ref028]\]. The overexpression of NGF in *Reeler* vitreous at p21 and p28 is in line with our previous studies on *Reeler* retina and other systems \[[@pone.0212732.ref001],[@pone.0212732.ref002],[@pone.0212732.ref009]\]. With respect to MMP9 and TIMP1 changes in *Reeler* vitreous, a possible explanation might lay in a steady MMP9/TIMP1 ratio known to be crucial for the extracellular matrix stability and particularly during development and synaptic plasticity \[[@pone.0212732.ref029]\]. Furthermore, the increased TIMP1 levels in *Reeler* vitreous might be an attempt to counteract the lack of Reelin by inhibiting the MMPs that are involved in the cleavage of Reelin glycoprotein, as observed in other experimental models \[[@pone.0212732.ref030]\]. Although needing further investigations, the increased MMP7 and MMP13 levels might participate in tissue remodeling and protective effects \[[@pone.0212732.ref030]\]. Although not significant, the overexpression of TNF-RI/II might be consistent with the recent evidence that a soluble counterpart exists for a number of signaling receptors (for review, see \[[@pone.0212732.ref031]\]). Soluble receptors participate to cell apoptosis, survival, proliferation and differentiation, and particularly the "receptor-shedding" mechanism seems to modulate or inhibit ligand activity/function by preventing the interaction with proper cellular targets \[[@pone.0212732.ref032]\]. The consistent overexpression of vitreal IL33 at p28 appears of great interest. A possible explanation might be found in the pathogenic IL33 route during the impaired retinogenesis occurring in Reelin-deficient cells/retina. Liu and coworkers associated IL33 human expression with retinal lesions occurring in Age related Macular Disease as well as with the loss of photoreceptor and retinal ganglion cells in the experimental model \[[@pone.0212732.ref033]\]. In other systems, the constant "bright light exposure"---a condition that can trigger photoreceptor lost---was associated with IL33 overexpression and explained as an attempt to balance the minor cell damage and glial cell activation \[[@pone.0212732.ref034],[@pone.0212732.ref035]\]. Pathological vitreous retains most of retina-derived products, representing an indirect source of retina microenvironment information in vitreoretinal practice \[[@pone.0212732.ref025]\]. The majority of vitreous-detectable soluble proteins originate from vitreous itself (hyalocytes) or the surrounding tissues (ciliary body and retina), while albumin might be also of plasma source \[[@pone.0212732.ref026]\]. Therefore, a molecular approach was performed on retinal total RNA extracts to validate some protein changes quantified in the vitreous. While the biomolecular expression of NGF in the retina was previously investigated, the herein observed IL33mRNA overexpression in retinal extracts corroborated the biochemical data on vitreous, suggesting a potential IL33 driven activity on *Reeler* retinal cells. The significant downregulation of MIP3α and MIP3β mRNAs in *Reeler* retinas can sustain the absence of inflammation and local macrophage recruitment \[[@pone.0212732.ref001],[@pone.0212732.ref002],[@pone.0212732.ref036]\]. The observation of increased IL33mRNA expression in the retina might provide explanation for increased IL33 in the vitreous and suggest *in situ* glial cell (Müller cell) activation. In rodent and human retina, IL33 (IL1 family) is locally expressed (epithelial, endothelial, glial and Müller cells), modulates immune cells (T-helper, macrophages, eosinophils and mast cells), recruits innate cells (neutrophils, macrophages, dendritic cells and eosinophils), induces hematopoietic stem / progenitor cell mobilization and even triggers the Th2-cytokine pattern \[[@pone.0212732.ref037]--[@pone.0212732.ref040]\]. Of interest, Müller cells (i.) are crucial in maintaining laminar structure, neuronal survival, metabolic homeostasis and retinal regeneration; (ii.) participate in reactive gliosis in response to injury and interesting (iii.) synthesize/release IL33 *in vivo* under stressing conditions \[[@pone.0212732.ref035],[@pone.0212732.ref041]\]. To verify the presence of reactive Müller cells, we checked for GFAP and GS mRNAs in retinal extracts as well as GFAP and Nestin immunoreactivity in retina sections, all known as Müller cell markers in aged, damaged, injured and stressed retinas \[[@pone.0212732.ref042]\]. Merely, GS and Nestin represent well-known marker for Müller cell identification *in situ* while GFAP represents a marker of activated Müller cells \[[@pone.0212732.ref043]\]. A significant increase of GS mRNA expression and immunoreactivity (cell number and protein accumulation in Müller cells) occurred in *Reeler* retinas. Since GFAP is a marker of both astrocytes and activated Müller cell, the observation of increased GFAP-positive and GFAP/Nestin-positive cells would suggest the presence of astrocytes and activated Müller cells \[[@pone.0212732.ref044]\]. The absence of GFAP/CD45-immunoreactivity would exclude the microglia involvement \[[@pone.0212732.ref045]\]. Newly generated GFAP-positive glial cells (gliogenesis), as well as their local differentiation into astrocytes, have been previously reported in *Reeler* mice and this increase might reflect the "misorganization" of retinal layers \[[@pone.0212732.ref005],[@pone.0212732.ref028]\]. According to literature, gliosis and reactive Müller cells occur in response to injury and/or upon cytokine as well as growth factor stimulation (VEGF, NGF, ...) \[[@pone.0212732.ref041]\]. On the other hand, both astrocytes and Müller cells are source of NGF and in turn can utilize NGF by means of trkA^NGF^ expression, as reported in human retinal diseases and experimental models \[[@pone.0212732.ref043],[@pone.0212732.ref046]\]. Therefore, the increased NGF expression described in *Reeler* retina and the observation of increased NGF level in *Reeler* vitreous might be view as a direct product of astrocytes/Müller cell activation, in addition to the neuronal task \[[@pone.0212732.ref001],[@pone.0212732.ref002],[@pone.0212732.ref047]\]. Corroborating data indicate that Müller cells are active players in retinal injury and chronic inflammatory/autoimmune mediated retinal disorders \[[@pone.0212732.ref048]\]. Long-term gliosis releases a plethora of inflammatory cytokines that lead to secondary injury, exacerbating the inflammatory reaction \[[@pone.0212732.ref047]\]. Local reactive gliosis might be view as an attempt to protect neuronal tissue from further damage \[[@pone.0212732.ref049]\]. Conclusions {#sec015} =========== Taken together, the finding herein reported reinforce the observation that NGF might be a compensatory effector of Reelin deficiency during postnatal brain and retinal development. NGF appears to function as an intrinsic determinant of cell migration, acting as a compensative regulator of Reelin expression \[[@pone.0212732.ref050],[@pone.0212732.ref051]\]. Cumulating, the NGF and IL33 overexpression in *Reeler* vitreous and the Müller cell activation in the retina would strength the presence of a protective mechanism to prevent damage and/or promote tissue repair. According to literature, the proposed (neuro)-protective attempt might be not enough due to the activation of other proinflammatory target genes (cytokines, chemokines, growth and neurotoxic factors) \[[@pone.0212732.ref002],[@pone.0212732.ref009]--[@pone.0212732.ref010],[@pone.0212732.ref018],[@pone.0212732.ref025],[@pone.0212732.ref033],[@pone.0212732.ref041],[@pone.0212732.ref051]\]. On the other hand, the overexpression of NGF, IL33, TIMP1, MMP9 and sTNFR-I/II factors to counteract/delay failures in retinogenesis and the potential NGF modulation of Müller cell activation represent interesting points that deserve further investigation for alternative strategies not restricted to reelin deficiency. BOB, GE and AM are grateful to Fondazione Roma for their continuous Intramural/non specific support and to Angelica Napoli for her contribution in drawing the retina for the striking image. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.

Background {#Sec1} ========== The number of total hip arthroplasties (THA) carried out in Germany has increased in recent years up to 160,000 annually \[[@CR1]\]. The main reasons for this are the aging of the population and technical improvements in orthopedic surgery. The current rate of prosthetic joint infection (PJI) in Europe and the United States is around 0.6--1.0% for THA \[[@CR1]--[@CR3]\]. This indicates that not only the number of THA increasing, but also the absolute numbers of infections. Treatment for PJI depends on the duration of symptoms. Early or hematogenous infections, with symptoms for less than 4 weeks, can be treated with local debridement and retention of the implant. In patients with late infections, one-stage or two-stage procedures are necessary \[[@CR4]\]. As success rates are above 90%, two-stage revision is the first choice for patients with late PJI in Germany \[[@CR5], [@CR6]\]. Either cemented or cementless revision systems can be used \[[@CR5]--[@CR9]\]. Cemented fixation was used for reimplantation in the past, because it allows antibiotics to be used in the cement to reduce the risk of recurrent infection \[[@CR7], [@CR8]\]. However, following good results with cementless fixation in aseptic surgery \[[@CR10], [@CR11]\], there have been increasing numbers of publications describing successful cementless fixation in two-stage revision procedures \[[@CR3], [@CR5], [@CR12]--[@CR17]\]. The aim of the present study was to evaluate clinical and functional outcomes with the Modular Universal Tumor And Revision System (MUTARS) RS stem, which has been used in our department for more than 10 years, in patients undergoing two-stage revisions. Methods {#Sec2} ======= Patients {#Sec3} -------- The functional and clinical outcomes for 43 patients who had undergone two-stage revision procedures for PJI were analyzed in a retrospective study. The patients had been treated between 2000 and 2012 in a university orthopedic department. We had an ethics approval of the local ethic committee of the University of Münster (2014-324-f-N). Every patient were informed about the study and agreed to publish their data. A consent statement was signed. They included 21 men and 22 women, with an average age of 66 years (range 40--84 years). The minimum follow-up period was 24 months. Shorter follow-up periods were only observed when there were complications such as loosening or recurrent infection. The mean follow-up period was 3.86 years (range 7 months to 11.6 years). The indication for THA was primary osteoarthritis (OA) in 31 cases, femoral neck fracture in six cases, secondary OA after acetabular fracture in four cases, rheumatoid arthritis in one case, and necrosis of the femoral head in one case. The last operation for the explantation of the prosthesis was in mean 43,5 months (1 months -- 24 years). Clinical and functional follow-up {#Sec4} --------------------------------- All of the patients received radiographic and clinical follow-up examinations. In case of death (n = 2), the last clinical and radiographic examination was evaluated, and inquiries were also made of the patient's relatives. The Della Valle--Paprosky classification was used to classify femoral bone defects \[[@CR18]\]. The functional outcome was evaluated at the outpatient examinations using the Harris Hip Score \[[@CR19]\]. End points and definitions {#Sec5} -------------------------- The inclusion criterion for the patients was a minimum follow-up period of 24 months, or less in case of early loosening of the stem or infection. PJI was diagnosed if at least one diagnostic method was positive in accordance with the Centers for Disease Control criteria \[[@CR20]\]. The primary end points of the study were successful treatment for infection or reinfection with loosening of the prosthesis. Clinical cure, with no clinical signs of inflammation and negative C-reactive protein findings, was assessed by the treating clinician at the date of the last available follow-up. Aseptic loosening of the stem was a secondary end point. Surgical treatment {#Sec6} ------------------ If at least one Centers for Disease Control criterion \[[@CR20]\] was positive, a two-stage revision was performed (Figure [1](#Fig1){ref-type="fig"}). All of the patients were treated with an antibiotic-loaded polymethylmethacrylate (PMMA) spacer. We collected at least three biopsies for microbiological examination. The composition of the antibiotics in the spacer was adapted to the bacterial resistance (Table [1](#Tab1){ref-type="table"}). Between 2000 and 2004, seven patients were treated with a short period of parenteral antibiotic therapy for a mean of 20 days (range 17--26 days) between both operations. Since 2005, all patients have been treated with parenteral antibiotic therapy for at least 2 weeks, followed by oral antibiotic therapy for at least 4 weeks. If the bacteria involved were not identified, calculated antibiosis with a cephalosporin and clindamycin was administered. In other cases, specific antibiotic therapy was used.Figure 1**Two-stage procedure in a 74-year-old-woman. a**: Septic loosening of the hip. **b**: Explantation of the hip prosthesis and spacer implantation. **c**: Reimplantation of a hip prosthesis with the MUTARS RS stem and a cementless cup. ###### **Antibiotic combinations in antibiotic-loaded polymethylmethacrylate (PMMA) spacers** Antibiotic combination Patients (n) --------------------------------------- -------------- Gentamicin 11 Gentamicin/vancomycin 4 Gentamicin/vancomycin/clindamycin 16 Gentamicin/clindamycin 11 Gentamicin/clindamycin/flucloxacillin 2 Before replantation we paused the antibiotic therapy for at least two weeks. If the C-reactive protein was less than 2 mg/dl we do the reimplantation. The MUTARS RS stem (Implantcast Ltd., Buxtehude, Germany) was used for reimplantation. It is a modular revision stem with a hexagonal stem design \[[@CR21]\] and has included a hydroxyapatite layer since 2006. For acetabular reconstruction we used in three cases a cemented dual mobility cup, in eight cases a cemented polyethylene, in 22 cases a cementless cup and in 10 cases a reconstruction with an antiprotrusio cage. In three cases we used the cage in combination with spongiosa chips of an allograft. After reimplantation, 2 weeks of the parenteral antibiotic therapy was administered, followed by 4 weeks of oral antibiotic therapy. Statistical analysis {#Sec7} -------------------- Statistical analysis was performed with IBM SPSS Statistics for Windows, version 21.0 (IBM Corporation, Armonk, New York, USA). Results {#Sec8} ======= Diagnosis of the periprosthetic infection {#Sec9} ----------------------------------------- The leading indication for explantation was bacteria in the aspiration fluid in 18 cases, fistula in 12 cases, purulent synovial fluid in eight cases, elevated leukocyte counts (\>4500) in the aspiration fluid in combination with elevatd C-reactive protein in four cases, and positive leukocyte scintigraphy in combination with elevated C-reactive protein in one case. Microorganisms were identified in 88.3% of cases (Table [2](#Tab2){ref-type="table"}). *Staphylococcus aureus* was present in most cases.Table 2**Microorganisms identified in soft-tissue samples**MicroorganismPatients (n)*Staphylococcus aureus*13\**Staphylococcus epidermidis*12*Enterococcus faecalis*3*Streptococcus gallolyticus*2*Corynebacterium*2*Escherichia coli*2*Staphylococcus haemolyticus*2*Enterobacter cloacae*1*Neisseria* species1*Micrococcus luteus*1*Serratia marcescens*1*Staphylococcus capitis*1*Staphylococcus cohnii* species1*Staphylococcus hominis*1*Staphylococcus warneri*1*Streptococcus agalactiae*1*Streptococcus bovis*1*Streptococcus pluranimalium*1*Peptococcus* species1*Propionibacterium* species1*Pseudomonas aeruginosa*1Sterile7Patients with two species6Patients with eight species1\*Two were methicillin-resistant. Infection therapy {#Sec10} ----------------- A spacer exchange and repeated complete PJI management were necessary in four cases. The cause of persisting infection was a nonsensitive spacer in a patient with methicillin-resistant *S. aureus* (MRSA) in one case, persistent fistula in two cases, and persistently high C-reactive protein levels in one case. The mean C-reactive protein level was 0.8 mg/dL (0.5--2.0 mg/dL) at the time of reimplantation. The success rate with infection control for PJI was 93%. Reinfection occurred in four cases (7%) (Table [3](#Tab3){ref-type="table"}). Another two-stage revision with a cemented prosthesis was carried out in two cases, and resection arthroplasty was performed in two patients. Two patients with reinfection had MRSA infections. The risk of reinfection after MRSA infection was 20.5 times greater (odds ratio; *P* \>0.01) than with sensitive or unknown bacteria.Table 3**Patients with recurrent infection**Patient no.Original indicationRisk factorsRevisionsMicroorganismSpacer period (weeks)ComplicationsTime to reinfection (months)1Secondary coxarthrosis (acetabular fracture)Fistula, malignancy, diabetes, two-stage revisionSpacer exchangeMRSA24--16.82Primary coxarthrosis--Aseptic stem exchange--8--32.03Femoral neck fracture----MRSA77--30.04Primary coxarthrosisFistula--*S. epidermidis*4.7--22.0MRSA, methicillin-resistant *Staphylococcus aureus.* Della Valle--Paprosky classification of the femur and radiographic analysis {#Sec11} --------------------------------------------------------------------------- The femoral defects were classified using the Della Valle--Paprosky classification \[[@CR18]\]. There were nine patients with type II, 30 with type IIIa, and four with type IIIb femoral defects. A MUTARS RS stem was used in all cases. In two cases there was a subsidence of the stem of 5 mm and 3 mm. Both patients had no sign of loosening after a follow-up of three years. Comorbidities and risk factors {#Sec12} ------------------------------ Comorbid conditions that were documented consisted of obesity (body mass index \>30) in 11 patients, diabetes mellitus in eight patients, malignancy in five patients, long-term cortisone therapy in three patients, endocarditis in two patients, chronic obstructive lung disease in two patients, osteomyelitis of the femur in one patient, PJI of the contralateral knee endoprosthesis in one patient, hepatitis B in one patient, hepatitis C in one patient, and long-term methotrexate therapy in one patient. Four patients had more than one comorbidity. The number of surgical procedures that had been carried out before the two-stage procedure was also documented as an additional potential risk factor. A two-stage procedure had already been performed in three patients; the infection occurred after the primary operation in 18 patients; 16 patients had one aseptic revision, five patients had two aseptic revisions, and one patient had one septic revision with retention of the prosthesis. Complications unrelated to infection {#Sec13} ------------------------------------ Two complications associated with the stem occurred. Both involved aseptic loosening after 7 and 20 months, respectively, with a type IIIa Della Valle--Paprosky defect in one case and a type IIIb defect in the other (Figure [2](#Fig2){ref-type="fig"}). The preoperative C-reactive protein findings were negative and no bacteria were found in the intraoperative tissue samples. Revision surgery with a long cemented stem was carried out in both cases.Figure 2**Aseptic loosening of a MUTARS RS stem in case of a Paprosky IIIa defect.** **a**: postoperative x-ray. **b**: 18 months x-ray control with aseptic loosening. **c**: Revision with a cemented stem. Wound healing problems occurred in four cases. Healing followed after local debridement in all of these patients. There was a trend for patients with diabetes mellitus to have a 4.5 times (odds ratio) greater risk of developing wound healing problems. One patient developed peroneal paralysis postoperatively; immediate revision surgery with shortening of the femoral head was necessary. One patient had a periprosthetic fracture after a fall; osteosynthesis was carried out successfully. Another patient developed dislocation of an inlay, and the cup had to be exchanged. One patient developed recurrent dislocation after reimplantation; the inlay was changed to a 10° insert and the antetorsion of the stem was altered. Another patient had aseptic loosening of an antiprotrusio cage, and an exchange was necessary (Table [4](#Tab4){ref-type="table"}).Table 4**Patients with complications unrelated to the infection**ComplicationsPatients (n)*Wound healing problems*4*Aseptic loosening of the stem*2*Recurrent dislocations*1*Aseptic loosening of the cup*1*Periprosthetic fracture after collaps*1*Peroneal palsy*1*Inlaydislocation*1 Functional outcome {#Sec14} ------------------ Evaluation of the postoperative Hip Society Score was possible in 37 cases. Aseptic or septic exchanges of the stem had already been performed in three cases already, two of the patients had died, and it was not possible to contact the patient for the questionnaire in one case. The average Harris Hip Score was 80 (range 62--93). Excellent results were seen in four cases, good results in 19, fair results in 11, and poor results in three cases. There was no difference in the functional outcome depending on the bony defect. Discussion {#Sec15} ========== Two-stage revision arthroplasty is today regarded as the gold standard treatment method for deep prosthetic joint infection \[[@CR22], [@CR23]\]. However, there has been debate on whether cemented or cementless reimplantation is better, in view of the lack of local antibiotic protection with the latter method \[[@CR5]--[@CR7], [@CR9], [@CR12]--[@CR15], [@CR24]--[@CR26]\]. In earlier studies on cementless two-stage revision, high rates of loosening and reinfection were reported \[[@CR25], [@CR26]\]. However, more recent studies have described good eradication rates and good prosthetic survival \[[@CR5], [@CR6], [@CR14], [@CR24]\]. The present study focuses on the reinfection rate and early loosening of the MUTARS RS stem (Implantcast Ltd., Buxtehude, Germany). Limitation of this study are the retrospective nature of the study and the relatively short follow-up period of 2 years in a few cases. However, most cases of reinfection occur in the first few years after reimplantation \[[@CR15], [@CR24]\] and most studies on cementless two-stage revision have reported follow-up periods of 2 years \[[@CR5], [@CR6], [@CR12], [@CR24], [@CR25]\]. Subsidence of the stem and early loosening of the components would also occur within the first few years \[[@CR27]--[@CR30]\]. The eradication rates achieved ranged from 82% to 100% with cementless two-stage revisions \[[@CR5], [@CR6], [@CR12]--[@CR15], [@CR17], [@CR24]--[@CR26]\] and from 84% to 100% with cemented two-stage revisions \[[@CR31]--[@CR35]\]. The eradication rate observed in the present study was 93%, including multiresistant bacteria. The treatments administered in other studies vary widely. Haddad et al. \[[@CR6]\] used a course of 5 days of parenteral antibiotics, followed by oral antibiotics for 3 weeks in the interval up to reimplantation, with an eradication rate of 92%. Other studies have reported administering intravenous antibiotics for up to 6 weeks \[[@CR5], [@CR12]--[@CR14], [@CR24]\]. Another limitation in the present study is that we had two different treatment protocols. Initial we treated seven patients in mean only 20 days parenteral between both operations. We had no reinfection with this kind of treatment. To our knowledge there is no study describing the gold standard for the perioperative treatment \[[@CR23], [@CR36]\]. Since 2004 we treated with a periods of 2 weeks with parenteral antibiotic treatment and 4 weeks of oral antibiotic therapy were used in the interval to reimplantation, followed by a further course of antibiotic therapy with 2 weeks of parenteral treatment followed by 4 weeks of oral treatment after reimplantation. Fink et al. reported a 100% eradication rate with non-multiresistant bacteria using this sequence \[[@CR5]\]. The 2-week period of parenteral antibiotic therapy appears to be short, but it is consistent with the recommendations of Zimmerli et al. \[[@CR4], [@CR37]\]. The 3-month period of antibiotic treatment is also in agreement with the recommendations \[[@CR4], [@CR37]\]. Local methods of treatment for PJI also vary widely. In earlier studies, resection arthroplasty \[[@CR25], [@CR26]\], antibiotic beads \[[@CR6], [@CR12]\], and articulating spacers \[[@CR15]\] were used. Gentamicin, alone or in combination with vancomycin, was often used \[[@CR6], [@CR12]\]. However, it was not possible to successfully treat gram-negative bacteria in particular with this combination. Masri et al. and Fink et al. were the first to describe specific local antibiotic therapy \[[@CR5], [@CR24]\], which makes it possible to treat resistant bacteria successfully, on the one hand, while on the other the development of new resistance can be avoided. Specific antibiotic therapy was therefore always used in the present study when possible. In cases of unknown bacteria, gentamicin alone was used. The results, with an eradication rate of 93%, provide support for this form of local therapy. Multiresistant bacteria such as methicillin-resistant *S. aureus* (MRSA) are an increasing problem \[[@CR38]\] and are associated with higher reinfection rates \[[@CR39]\]. Kilgus et al. reported that the reinfection rate in cases of multiresistant *S. aureus* and *S. epidermidis* has increased from 19% to 52% \[[@CR40]\]. Lim et al. compared 24 patients with resistant microorganisms with 13 patients with nonresistant microorganisms \[[@CR41]\] and noted a recurrence rate of 33% in the resistant group, in comparison with no failures in the nonresistant group. Similar results were observed in the present study. Only two patients had MRSA infections, but reinfection occurred in both cases. One had already undergone a two-stage revision due to MRSA infection. The reinfection rate was significantly higher in patients with MRSA infection. Aseptic loosening occurred in 4.6% of the cases. In a review, Wirtz et al. reported mean loosening rates of 4.4% with cementless stems in cases of aseptic revision and 21.2% with cemented stem revisions \[[@CR42]\]. Dohmae et al. \[[@CR43]\] demonstrated that there is a 70% lower shear force capacity in revision cases in comparison with primary implantations. This may be the reason why the cement does not fuse with spongy bone. It is therefore necessary for the revision implant to bridge the bone affected by the original femur implant. In two-stage revisions, Nestor et al. \[[@CR25]\] reported an 18% rate of loosening using a nonmodular, proximally coated stem, while other studies have reported loosening rates of 0% using nonmodular or modular stems \[[@CR5], [@CR6], [@CR12]--[@CR14], [@CR17], [@CR24]\]. Most of the studies do not report any classification of the defects. The loosening rates observed with two-stage revisions are equal to or better than those with aseptic revisions \[[@CR42]\]. It was therefore concluded that the loosening rate depends on the defects involved, rather than on the two-stage procedure. Other risk factors have also been identified that have an unfavorable influence on the course of salvage procedures following infection. Particularly with primary hip arthroplasty, these factors include postoperative osteoarthritis, multiple surgical revisions, cutaneous and urinary tract infections, chronic liver disease, inadequate antibiotic prophylaxis, and malignancies \[[@CR44]\]. Further risk factors include rheumatoid arthritis, steroid therapy, diabetes mellitus, immunosuppressant therapy, and nosocomial infections \[[@CR45]\]. The only risk factor for reinfection identified in the present data was the presence of multiresistant bacteria. In fact, the same risk factors appear to be present with two-stage revisions as in primary arthroplasty. It is possible that the number of patients included was too small for statistical significance to be reached. Conclusions {#Sec16} =========== This retrospective study showed a 93% rate of eradication using specific antibiotic therapy in accordance with the recommendations made by Zimmerli et al. \[[@CR4]\] and Trampuz et al. \[[@CR46]\]. With the modular MUTARS RS stem, there was a low rate of aseptic loosening of 4.6%. MRSA infection was identified as a risk factor for reinfection. The two-stage procedure used with the modular cementless implants is therefore appropriate for treating periprosthetic infections associated with hip endoprostheses. Authors' original submitted files for images {#Sec17} ============================================ Below are the links to the authors' original submitted files for images.Authors' original file for figure 1Authors' original file for figure 2Authors' original file for figure 3Authors' original file for figure 4Authors' original file for figure 5Authors' original file for figure 6 **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** RD was responsible for the design of the study and for the literature search, data extraction, and methodological appraisal of the study. He performed the statistical analyses, led the interpretation of the results, and wrote and drafted the manuscript. DS, GG, KB, KD, SA, JD were involved in data collection. SH participated in the design of the study and writing of the manuscript. All of the authors have read and approved the final manuscript. The authors are grateful for support to the *Deutsche Forschungsgemeinschaft* and the Open Access Publication Fund of the University of Münster.

Background {#Sec1} ========== Treatment of horses for infection of a limb with *Pythium insidiosum* is challenging, and its effectiveness appears to depend on such factors as size and site of the lesion, duration of infection, immunocompetence, and the type of treatment. Systemically administered antifungal drugs, such as potassium or sodium iodide, ketoconazole, miconazole, fluconazole, itraconazole, and amphotericin B, have been administered with or without the surgical excision of the lesion to improve outcomes \[[@CR1]--[@CR3]\]. However, these drugs are considered hazardous and are expensive when administered systemically to horses, and because *Pythium insidiosum* is not a true fungus, this protistal organism has increased resistance to most available antifungal agents \[[@CR3], [@CR4]\]. Dória et al. \[[@CR5]\] reported that intravenous regional limb perfusion (IRLP) with amphotericin B is effective for treating horses with a cutaneous lesion of pythiosis of a limb, resolving the infection with manageable complications, such as oedema of the limb, signs of pain during palpation of the limb, and inflammation at the site of venipuncture for IRLP. Their study demonstrated that 92 % of the horses (*n =* 11) affected by *Pythium insidiosum* and presenting with exuberant granulation tissue in the distal aspect of a thoracic or pelvic limb that were treated with intravenous regional perfusion with amphotericin B had complete resolution of their lesions 35 days after one treatment or 60 days after two treatments. Dória et al. \[[@CR5]\] considered the intravenous regional perfusion of amphotericin B to be an effective adjunct therapy to surgical excision and thermocauterization in treating horses for pythiosis of the distal portion of the limb. We hypothesized that better results with fewer complications might be achieved if dimethylsulfoxide (DMSO) is added to the perfusate when amphotericin B is administered by IRLP as a treatment for pythiosis of a limb. We reasoned that adding DMSO to the perfusate would achieve a higher concentration of amphotericin B within the infected tissues, and because DMSO has been shown to be an effective anti-inflammatory agent, its addition to the perfusate would reduce inflammation at the site of intravenous administration. It may also possess some inhibitory effects on the growth of a variety of fungi as a result of its effect on the immune response and as a result of reducing endotoxin-induced tissue damage. Studies have shown that intravenous administration of DMSO results in vascular dilation and increased flow of blood through experimentally created cutaneous flaps \[[@CR6]--[@CR9]\]. Such properties associated with its ability to penetrate biological membranes provide a rationale for its use in conjunction with an antifungal drug for regional limb perfusion \[[@CR10]\]. Thus, we evaluated the effects of administering amphotericin B in a 10 % solution of DMSO by IRLP as an adjunct therapy to surgical excision, hoping to diminish the detrimental vascular effects associated with administration of amphotericin B, to treat horses for pythiosis of a limb. Methods {#Sec2} ======= This study was approved by the University of Cuiabá Animal Care and Ethics Committee, under protocol number 2009--228. Fifteen horses (8 males and 7 females, age 4 months to 15 years, weighing 100--420 kg) diagnosed with pythiosis were studied. This research was carried out on horse farms located in Tocantins state (*n* = 1), São Paulo state (*n =* 2) and Mato Grosso state (*n =* 12), Brazil. The presumptive diagnoses were made on the basis of historical data, the gross appearance of the granulomatous lesion, and the histological appearance of the lesion. Lesions were located distal to the cubital joint (elbow) or the femorotibial joint (stifle). In preparation for the treatment, feed was withheld for 12 h before treatment was administered. A catheter was placed into a jugular vein, and the horses were tranquillized with acepromazine (0.1 mg/kg) and then anesthetized with guaiacol glycerol ether (100 mg/kg), ketamine (2 mg/kg), and midazolam (0.1 mg/kg). Horses were positioned in lateral recumbency with the lesion uppermost. The site of the lesion was prepared for surgery by scrubbing the site with soap and applying dilute povidone-iodine solution and 70 % isopropyl alcohol. An Esmarch tourniquet was applied proximal to the lesion to prevent hemorrhaging, making it possible to surgically remove the exuberant granulation tissue and kunkers safely avoiding exposing the bone or entering a synovial structure. Tissues (1 cm^3^) were collected and fixed in 10 % formalin for histological (hematoxylin and eosin; Grocott's methenamine silver) and immunohistochemical (the labeled streptavidin biotin method) analyses, performed as previously reported \[[@CR11]--[@CR13]\]. After surgery, the Esmarch tourniquet was partially released to identify bleeding vessels for cauterization with electric thermocautery, and the site of the lesion was bandaged. The Esmarch tourniquet was repositioned and retightened, and a superficial vein (the cephalic, palmar digital, saphenous, plantar digital, or dorsal digital vein) next to the lesion and distal to the tourniquet was catheterized using a 20-, 22-, or 24-ga catheter for administration of the drugs. The catheterized vein was injected with 60 mL of a solution containing 50 mg (10 mL) of amphotericin B[1](#Fn1){ref-type="fn"}, 6 mL of medical grade DMSO[2](#Fn2){ref-type="fn"} and 44 mL of lactated Ringer's solution. The injection was delivered using finger pressure over 5 min using a 60-mL syringe connected to an extension line. The catheter was removed after the administration, and firm pressure was applied manually to the site of venipuncture. The tourniquet was released 45 min after the administration of the solution of amphotericin B and DMSO. Lesions were evaluated before treatment (Day 0) and then weekly (1 to 9 weeks) after the administration of the solution of amphotericin B and DMSO until the lesions were completely healed. At the same time points, blood samples were collected for haematological and biochemical analyses using standard techniques \[[@CR14]\]. The lesion's size (length and width) was determined at day 0 using a ruler. The lesions were assessed qualitatively, and the response to therapy was determined by clinical observation (i.e., gross appearance of the lesion). The appearance of the lesion was compared to the photographic appearance of the lesion at the times of previous evaluations. A wound was considered completely healed when it was completely covered with epithelium. The affected limb was evaluated for regional swelling, sensitivity to palpation, and signs of inflammation at the site of venipuncture. The data were analyzed with the Statistical Analysis System (SAS.2011). Normality of the data distribution was evaluated by the Shapiro-Wilk test. A one-way analysis of variance with repeated measures, followed by a Tukey test, was used for the comparisons of the weekly data. The differences were considered to be statistically significant when *P* ≤ 0.05. Results {#Sec3} ======= The pythiosis lesions were located in different regions, including the radius (20 %), carpus (6.66 %), metacarpus (6.66 %), tibia (6.66 %), tarsus (6.66 %), metatarsus (6.66 %), fetlock and pastern (46.7 %). At day 0, all of the horses had large lesions ranging from 12 to 50 cm in length (mean length 27 cm) and 15 to 40 cm in width (mean width 25 cm) and surrounded 90° (4 horses), 180° (7 horses) or 360° (4 horses) of the affected areas. The lapse of time between emergence of the lesion and the beginning of treatment with amphotericin B and DMSO ranged from two to five months (average, three months). Before treatment (D0), the affected limb was swollen, and the lesion was filled with exuberant, ulcerated granulation tissue and surrounded by oedematous tissue. The surface of the granulation tissue was nodular and covered with a mucosanguineous viscous exudate. The tissue sections had sinuses containing kunkers, usually surrounded by seropurulent discharge. The horses showed signs of intense pruritus that was characterized by self-mutilation. The histopathological diagnosis was pyogranulomatous dermatitis associated with the pseudo hyphae characteristic of *Pythium insidiosum* (subcutaneous pythiosis), confirmed by the immunohistochemistry analysis. Surgical excision of the exuberant granulation tissue and IRLP administration of amphotericin B and DMSO resulted in complete resolution of the lesions. All 15 lesions regressed after a single administration of a solution containing amphotericin B and DMSO. By one week after a single treatment, exudate, sinuses, and kunkers had disappeared from the lesions, and the granulation tissue had regressed and turned from white to pink-yellow. In subsequent weeks, the granulation tissue became pink and flat, and a margin of epithelialization could be seen moving centripetally. Epithelialization was complete by 6 weeks in 5 horses, 4 of which had lesions surrounding 90° of the affected area and 1 of which surrounded 180° of the affected area; 8 weeks in 5 horses of which had lesions encompassing 180° of the limb; and 9 weeks in 5 horses, 1 of which had lesions encompassing 180° of the limb and 4 of which encompassed 360° of the limb (Fig. [1](#Fig1){ref-type="fig"}). No complications at the site of venipuncture for RLP were observed in any of the horses, and no horse had recurrence of pythiosis at the time of follow-up, one year after resolution of disease.Fig. 1Healing of pythiosis lesions treated with amphotericin B and DMSO by IRLP in horses. Progression of healing of pythiosis lesions in the limb of horses treated with a single administration of amphotericin B in a 10 % DMSO solution by intravenous regional perfusion. Note lesions before (Day 0; left; **a**), 4 (middle; **b**), and 8 (right; **c**)weeks after the treatment Haematology showed that there was an increase in the RBC and PCV after treatment. Anemia resolved by 2 weeks after IRLP. The total and differential leucocyte counts were significantly increased before treatment, with neutrophilia and eosinophilia. The neutrophil count and the eosinophil count decreased and returned to normal values by 2 weeks after treatment. Increased serum concentration of fibrinogen found when the horses were admitted had decreased to normal values by 2 weeks after treatment. No significant abnormalities were noted during examination of the results of other biochemical assays (Table [1](#Tab1){ref-type="table"}).Table 1Haematology and biochemistry in horses with pythiosis treated by IRLP with amphotericin B and DMSO. Mean values of haematology and selected blood biochemistry tests before (week 0) and after (1 to 9 weeks) intravenous regional perfusion with amphotericin B in a 10 % DMSO solution in 15 horses with pythiosisHaematologyBiochemistryWeeksRBC (x10^6^/μL)PCV\* (%)TLC\* (x10^3^/μL)SN\* (x10^3^/μL)E\* (/μL)Fibrinogen (g/L)TP (g/L)AST (UI/L)GGT (UI/L)ALP (UI/L)CR\* (μmol/L)U (mmol/L)06.1726^ab^21^a^15.38^a^949.83^a^5.274^d^192^d^17.30^d^306.53^d^112.27^ad^13.24^d^15.5424^a^14.32^ab^10.15^ab^827.20^ab^4.675^d^163^d^20.78^d^213.33^d^74.26^bd^13.02^d^26.65^d^28^ab^11.54^bcd^7.41^bcd^577.43^abd^4^d^77^d^167^d^18.49^d^254.12^d^79.56^bd^14.40^d^36.74^d^29^ab^10.77^bcd^6.24^bcd^571.83^abd^3.7^d^75^d^181^d^15.38^d^222.53^d^79.56^bd^13.97^d^47.25^d^31^b^10.48^bcd^5.98^bcd^438.33^abd^2^d^75^d^194^d^13.77^d^273.51^d^78.68^bd^16.25^d^57.45^d^31^b^9.28^bcd^5.32^bcd^346.67^abd^2.7^d^72^d^214^d^15.94^d^268.19^d^80.44^bd^17.04^d^67.34^d^32^bd^8.67^bcd^4.55^cd^334.17^abd^3^d^72^d^184^d^17.85^d^210.73^d^80.44^bd^18.58^d^77.38^d^31^b^8.38^cd^5.04^bcd^287.17^bd^2^d^72^d^196^d^14.17^d^211.73^d^94.59^abd^17.83^d^87.80^d^33^bd^9.13^bcd^5.06^bcd^315^abd^2^d^72^d^195^d^14.40^d^200.18^d^84.86^abd^13.54^d^97.66^d^32^abd^9.13^bcd^5.41^bcd^189^abd^2^d^69^d^221^d^13.80^d^277.28^d^83.98^abd^13.05^d^Reference range^22^6.4--10.032--475.2--13.92.26--8.580--6001--457--79152--2949--25143--39588.4--167.969.21--19.96*RBC* red blood cell count, *PCV* packed cell volume, *TLC* total leukocytic count, *SN* segmented neutrophils, *E* eosinophils, *F* fibrinogen, *TP* total protein, *AST* aspartate aminotransferase, *GGT* gamma-glutamyltransferase, *ALP* alkaline phosphatase, *CR* creatinine, *U* urea nitrogen\*Significant at probability ≤ 0.05^a^; ^b^; ^c^ - means with different letters are significantly different (Tukey's test; *P* ≤ 0.05); d Mean value inside reference range Discussion {#Sec4} ========== We report the successful treatment of horses for cutaneous pythiosis of the limb after IRLP using amphotericin B and DMSO in combination with excision of the exuberant granulation tissue. Our results indicate that the treatment we administered is successful, cost-effective, and easily performed. The challenge of treating pythiosis is characterized by the severity of the disease and by the absence of gold standard chemotherapy. Complete surgical excision is the treatment of choice, but the disease is often too extensive at the time of diagnosis to allow complete resection. Immunotherapy and/or the systemic, intralesional, and topical administration of antifungal drugs has been combined with surgical excision to improve therapeutic efficacy \[[@CR3], [@CR15]\]. Amphotericin B, when administered systemically, does not reach a therapeutically effective concentration at the site of infection with *Pythium insidiosum* and can be associated with adverse effects such as nephrotoxicity and anaemia \[[@CR16], [@CR17]\]. However, IRLP has been shown to be an effective technique for obtaining high drug concentrations in tissue \[[@CR18]--[@CR21]\], and the high concentration of amphotericin B in the infected tissue enhances the efficacy of the drug against *Pythium insidiosum*. Increased concentrations of amphotericin B impair or inhibit this oomycete's essential metabolic processes and result in the death of the microorganism \[[@CR5]\], as demonstrated in this study. Systemically administered medical treatment of horses for pythiosis has been reported to be marginally effective when the lesions were located in the limbs \[[@CR22]\]. Local thrombosis, ischemia, tissue necrosis, and abscess formation, the typical characteristics of pythiosis lesions, decrease the local blood supply and thus decrease the effective therapeutic drug concentration that reaches *Pythium insidiosum* \[[@CR23], [@CR24]\]. The inability to completely resect a lesion because of its proximity to joints or bones adds to the challenge of treating horses for cutaneous pythiosis and favours the persistence of the infection. According to some previous reports, only horses with small, superficial lesions of pythiosis can be treated successfully \[[@CR25], [@CR26]\]. Dória et al. \[[@CR5]\], however, were able to resolve the lesion caused by pythiosis in 92 % of affected horses, regardless of the size of the lesion, by using a combination of either complete or partial surgical excision, thermocautery, and administration of one or two IRLP of amphotericin B. Complications associated with IRLP of amphotericin B, such as oedema of the limb, signs of pain during palpation of the limb, and inflammation at the site of venipuncture, were local and resolved. Similarly, the current study demonstrates that the combination of amphotericin B and DMSO provides excellent results. All 15 of the horses (100 %) showed complete resolution of their lesion after a single intravenous regional perfusion, without systemic or local complications, and all lesions were large, some involving the complete circumference of the limb. Dória et al. \[[@CR5]\] reported that excision of the exuberant granulation tissue alone is not sufficient to resolve an infection of a limb caused by *Pythium insidiosum*. In chronic conditions, there is more fibrosis and scarring, which further isolate the protistal organism from the body's defense mechanisms as well as from systemically administered antifungal drugs \[[@CR24], [@CR27]\]. A high concentration of an antifungal drug can be achieved in chronically infected tissue by intravenous regional perfusion \[[@CR24], [@CR28]\], as demonstrated in this study. All of the horses treated with IRLP of amphotericin B in a 10 % DMSO solution had complete resolution of their lesion with no recurrence at the time of follow-up one year later, even though excision of the infected tissue was incomplete. We observed, 1 week after treatment, that the granulation tissue was devoid of exudate and that the sinuses and kunkers were no longer present, indicating that infection by *Pythium insidiosum* had resolved completely. Healing period was related to the size of the lesion. The larger the lesion, the longer the time the lesion took to heal. The horses had anaemia before surgery and this has been observed previously \[[@CR29]--[@CR31]\]. There was also marked leukocytosis with neutrophilia and eosinophilia before treatment \[[@CR2], [@CR16]\]. Administration of a combination of amphotericin B and DMSO by IRLP was effective in treating horses for cutaneous pythiosis of the limb and evidence of resolution of infection included a marked decrease in the total leukocyte count by 2 weeks and return of other haematological values, such as plasma concentrations of acute phase protein and fibrinogen, to normal by 2 weeks after treatment. No changes were observed in the serum enzyme activities after the IRLP of amphotericin B and DMSO suggesting that IRLP with amphotericn B did not cause systemic complications, such as liver and/or kidney dysfunction, which have been reported to occur when amphotericin B is administered systemically \[[@CR32]--[@CR35]\]. Studies have found that by administering a drug by IRLP, the concentration of the drug at the site of the disease can be increased many fold without risk of reaching a toxic concentration of the drug systemically \[[@CR19], [@CR24], [@CR36]--[@CR38]\]. IRLP allows a wider use of drugs that are considered hazardous and expensive when systemically administered to horses \[[@CR39], [@CR40]\], such as amphotericin B \[[@CR4], [@CR17], [@CR34], [@CR35], [@CR41]\]. We administered 50 mg amphotericin B in lactated Ringer's solution by IRLP, as reported by Dória et al. \[[@CR5]\] but added 6 mL DMSO (10 % of the total volume) to compose the final solution. The dose and the total volume of amphotericin B (0.83 mg/mL/60 mL) and DMSO (6 mL/60 mL) were standardized in this study and were administered irrespective of the horse's age and body weight, the size of the lesion, and the site on the limb at which the tourniquet was applied. We believe that the addition of DMSO to the perfusate prevented vascular inflammation typically induced by IRLP of amphotericin B and reported by Dória et al. \[[@CR5]\]. DMSO is a potent anti-inflammatory agent that acts, through the activity of superoxide dismutase, to suppress production of prostaglandins, to limit production of free radicals, and to scavenge free radicals \[[@CR9], [@CR42]\]. DMSO has multiple properties that can be beneficial when used as adjunctive treatment for sepsis. DMSO reduces platelet aggregation and thus decreases the incidence of thrombus formation in small vessels. This activity helps to normalize tissue perfusion in the face of the vascular insults that often accompany septic conditions of the equine extremities. The inhibitory effects of DMSO on the growth of a variety of bacteria, viruses, and fungi have been documented by a number of studies \[[@CR6], [@CR7]\]. The bacteriostatic and bactericidal effects of DMSO may be the result of the drug's ability to penetrate biological membranes \[[@CR8], [@CR9]\]. A drug that possesses these qualities is likely to attenuate some of the deleterious effects of vascular injury and free radical production inherent in septic conditions of the limb by affording a potent anti-inflammatory effect and aiding in tissue perfusion \[[@CR10]\]. It is noteworthy that DMSO is not a licensed drug therefore it is not allowed for treatment in many countries. We assumed that by using an appropriate rate of infusion, the increased intravascular pressure would favour the diffusion of amphotericin B through the tissues and enhance the drug's concentration in the target tissues \[[@CR40], [@CR43], [@CR44]\] and that the combination with DMSO would increase the drug's diffusion through the affected tissues, providing better resolution of infection than when amphotericin B is administered alone. In contrast with the study by Dória et al. \[[@CR5]\], here only one IRLP with amphotericin B and DMSO was necessary to resolve a large wound on a limb caused by chronic pythiosis. Administering DMSO and a low dosage of amphotericin B by IRLP caused no systemic or local complications and is likely to be safer and more economical than administering amphotericin B systemically. Conclusion {#Sec5} ========== This small case series demonstrates that in association with surgical excision of the exuberant granulation tissue caused by *Pythium insidiosum* infection, a single IRLP administration of 50 mg of amphotericin B and DMSO (10 % of the total solution) was an effective technique for treating horses for pythiosis of a limb that produced no noticeable local or systemic complications. IRLP : Intravenous regional limb perfusion DMSO : Dimethylsulfoxide RBC : Red blood cells PCV : Packed cell volume Anfotericin B®, Cristália, São Paulo, Brazil. Dimesol®, Marcolab, São Paulo, Brazil. **Competing interests** The authors have declared no competing interests. None of the authors has any financial or personal relationships with other people or organizations that could inappropriately influence or be perceived to influence the content of the paper. **Authors' contributions** RGSD and SHF participated in the design of the study, supervised the experiments, revised and submitted the final manuscript. MBC, LML, MAGS participated in the study from the early phase up to the end of data collection. FSM and EMC carried out the histopathological and immunohischemistry studies. RG and PFN performed statistical analysis and prepared the final draft of the manuscript. All authors read and approved the final manuscript. This study was performed with financial support provided by Fundação de Amparo à Pesquisa do Estado de Mato Grosso (FAPEMAT). Cristália Chemical Products and Pharmaceutics provided the Anfotericin B.

Introduction {#Sec1} ============ Physical inactivity has been recognized as a major modifiable risk factor for non-communicable diseases (NCDs) since the 1950s \[[@CR1]\]. Recent reports have equated the impact of physical activity (PA) to that of smoking with respect to the worldwide burden of NCDs \[[@CR2]\]. Physical activity is a challenging variable to measure, on account of the inherent complexity and diversity of human behavior. Traditionally, tools for measuring physical activity have been divided into subjective and objective methods. While the use of objective methods generally provides more accurate estimates of physical activity, these methods are cumbersome and impractical for use outside the setting of specialized research units. Subjective (self-reported) methods involving the use of physical activity questionnaires (PAQs) have therefore become the preferred method of assessing physical activity in epidemiological studies. A number of PAQs have been described in the literature, most of which have been designed for use, and validated in, developed countries. Several factors mitigate against the use of these questionnaires in low and middle income countries like India. A major drawback of these PAQs in the Indian context, is the importance given to leisure time physical activity (LTPA). While LTPA contributes significantly to total physical activity in Western populations, studies from India show that less than 10% of the population performs any LTPA at all \[[@CR3]\]. Also, the use of many of these PAQs demands a certain level of literacy in the respondents, which may not be the case in developing countries like India. In recent years, international questionnaires such as the Global Physical Activity Questionnaire (GPAQ) \[[@CR4]\] and International Physical Activity Questionnaire (IPAQ) \[[@CR5]\] have been validated in several populations, including those of developing nations. Many of these questionnaires, though valid and reliable, do not permit collection of information on region-specific and culturally relevant activities across different domains. These questionnaires assess physical activity over the week prior to administration and may not be suited for use in individuals with varied educational levels as seen in India, as they require the respondent to self-rate their own level of activity intensity, which has been shown to be difficult in the Indian setting. The Indian Migration Study (IMS) questionnaire \[[@CR6]\] was developed as an alternative to the international questionnaires for use in India. While the IMS questionnaire is reliable, valid and culturally relevant, it only collects information pertaining to the month immediately preceding its administration. Also, the IMS questionnaire does not address the aspect of seasonality of occupations and variations in physical activity in individuals holding multiple jobs at the same time. Therefore, we attempted to develop a PAQ for use in India, that would measure habitual, culturally relevant activities in various domains (occupational, transport, recreational, activities of daily living and weekend activities) over a year and which would be valid for use in adults of different age groups with varying levels of activity in urban as well as rural settings. The present paper aims to assess the reliability and validity of this new PAQ- termed the Madras Diabetes Research Foundation- Physical Activity Questionnaire (MPAQ). Methodology {#Sec2} =========== The MPAQ was developed (Additional file [1](#MOESM1){ref-type="media"}) after reviewing various published validated physical activity questionnaires both in India and abroad. In addition, 24 hour physical activity recalls encompassing a weekday and weekend were collected from 50 volunteers across all ages and occupations. From these 24 hr recalls, the various activities reported across all domains were listed in the MPAQ and similar activities were grouped together and further truncated based on the average energy cost, as the Physical Activity Ratio (PAR) provided by the WHO/FAO 2001 \[[@CR7]\]. The MPAQ was designed to capture frequency and duration of habitual obligatory and discretional activities by means of a mix of open and closed-ended questions arranged in four domains viz. work-related activity (work domain), activities of daily living \[general activity domain which includes sleep (daytime napping and sleep at night), personal care and domestic chores\], transport-related activities (transport domain) and recreational activities (recreational domain). In all domains, options are provided to capture both seasonal and non-seasonal activities. The questionnaire captures details of up to two jobs and elicits information on time spent sitting, standing, walking and climbing stairs in each of these jobs, providing insight into the nature of the job and intensity of work activity. In the recreational domain, the questionnaire elicits information on sedentary behavior (including TV viewing, chatting with friends, listening to music etc.) as well as light, moderate and vigorous activities on a daily, weekly or monthly basis. In addition, there is provision for recording the extra activities or extra hours of sedentary behavior that happen during the weekend. The questionnaire enables calculation of physical activity for an "average" day by summing up activities in various domains for a 24-hour period. Similarly, weekly and monthly calculations can also be done and information aggregated to compute activity for a year. Total energy expenditure can be estimated through factorial calculations recommended by a joint FAO/WHO/UNU expert consultation \[[@CR7]\]. The factorial calculations are based on the time spent on various activities in the multiple domains and the energy cost of these activities. Energy cost is reported as a multiple of Basal Metabolic Rate (BMR) and called Physical Activity Ratio (PAR). Total time spent on habitual activities is multiplied by PAR to derive the total energy expenditure (TEE) of 24 hours. The physical activity level (PAL) can then be calculated as TEE/BMR for 24 hours. Based on the PAL values \[[@CR7]\], individuals can be divided into three categories: Sedentary (1.40 -- 1.69), moderately active (1.70-1.99) and vigorously active (2.00-2.40) \[[@CR7]\]. Written informed consent was obtained from each participant before start of the reliability and validity studies. Institutional Ethics Committee approval was obtained from the Ethics Committee at MDRF. Reliability study {#Sec3} ----------------- The MPAQ was administered by trained interviewers to individuals of either gender aged 20 years and above from 10 states in India namely Tamilnadu, Gujarat, Maharashtra, Jharkhand, Haryana, Bihar, Chandigarh, Assam, Tripura and Arunachal Pradesh. The states were so chosen as to be representative of the country in terms of geography, socioeconomic status, variability of occupations and climatic conditions. From one district in each state, two census enumeration blocks (CEBs) in urban areas and three villages in rural areas were randomly selected (Figure [1](#Fig1){ref-type="fig"}). In each CEB or village, 10 households were randomly selected and in each household, one individual was selected to participate in the study. Thus, 50 individuals were selected from each state, and 500 for the entire study. In addition, five individuals from each state were recruited to allow for non-response over time. Hence a total of 550 individuals were initially invited for the study, of whom 543 individuals participated and had all required information at baseline. The participants were sampled so as to obtain individuals across all age categories and both genders with varying literacy levels and engaged in a wide range of occupations so as to test the ability of the MPAQ to measure physical activity of individuals from all walks of life.Figure 1Selection of study subjects. Demographic details and information on smoking and alcohol use were obtained from all participants as well as height, weight, waist and blood pressure (BP) measurements, assessed using standardized techniques. Weight (in kilograms-kg) was measured with the subjects wearing light clothing after having removed shoes and heavy jewelry. Height was measured to the nearest centimeter (cm) using a stadiometer with the subjects standing erect without shoes. Body mass index (BMI) was calculated as the weight (kg) divided by the height (in meters squared). Waist circumference was measured using a non-stretchable tape, as the mean of two measurements of the smallest horizontal girth between the costal margins and the iliac crests at minimal respiration. BP was recorded in the sitting position in the right arm using an electronic instrument (Model: HEM- 7101, Omron Corporation, Tokyo, Japan). Two readings were taken 5 minutes apart and the mean of 2 readings was taken as the blood pressure. The baseline administration of MPAQ was performed from May to August 2011 in all the states. This was followed by a repeat administration within a month for assessing reliability. The interval of one month was chosen based on a previously published study from India \[[@CR6]\], and was deemed most appropriate to eliminate recall of previous responses by the participants as well as any possibility of physical activity patterns having significantly altered in the interim. Validity studies {#Sec4} ---------------- For assessing relative validity, the MPAQ and the GPAQ were administered in a randomized order by trained interviewers to all selected participants across the 10 states one day apart so as to avoid questionnaire fatigue. The GPAQ was chosen because it is a widely used global PAQ which has been validated in India \[[@CR8]\]. The test questionnaire took on an average, 10 minutes (±5) to administer. Subject acceptability and co-operation were good with both questionnaires but the subject understanding was better with the test questionnaire. Construct validity indicates the consistency or the relationship between the activity instrument (MPAQ) and the physiological variable such as BMI. This was tested by plotting time spent in sitting and moderate and vigorous physical activity (MVPA) (measured as minutes/day) against BMI and waist circumference measured at baseline. For assessing criterion validity, 107 individuals of either gender aged 20 years and above were recruited from Chennai city in Tamilnadu. The sample was so chosen as to get individuals across a wide age range, both genders and all categories of activity: At the start of the study, information on demographic parameters, height, weight and occupation were obtained as described above. Criterion validity was assessed using the Actigraph (Actilife 5) GT3X+ Triaxial Accelerometer (Actigraph, Pensacola, Florida, USA). Participants were asked to wear the accelerometer for 7 days during waking hours; however, the device was allowed to be removed while bathing or swimming. The device was worn on the hip of the dominant side (right in most cases). The device was worn either above or beneath clothing and not necessarily in contact with skin; however, a snug fit against the body was ensured to avoid erroneous readings. Accelerometers were initialized to monitor and record data in 60- second "epochs" as "activity counts" and sample frequency at 100 Hz. The start date and time and stop date and time were used for the start and stop of data collection. While initializing, each device was given a unique number denoting the individual participant with their age, gender, height, weight, date of birth and race. The GT3X+ device collects data from all three axes of movement regardless of the configuration, with Axis 1 collecting the vertical axis acceleration activity data, Axis 2 the horizontal axis data and Axis 3 the perpendicular axis data. The duration (minutes per day) spent in different intensity activities- light (1.5-3 METS, 100 ≤ 1951 counts), moderate (3--6 METS, 1952--5724 counts) and vigorous (\>6 METS, ≥ 5725 counts) were determined according to published data \[[@CR9],[@CR10]\]. The MPAQ was administered anytime during the period the individual was wearing the accelerometer. Data from the MPAQ was computed for a typical week, and then converted to minutes/week, so as to make comparisons with the accelerometer data more realistic. Inter-rater reliability {#Sec5} ----------------------- MPAQ being an interviewer administered questionnaire, inter rater reliability was measured to assess the agreement between the interviewers. One interviewer administered the questionnaire to the participant while the rest of the interviewers passively observed and rated participant's response independently. This procedure was completed for a total of 20 participants by all 20 interviewers who collected the questionnaires across the 10 states. A kappa value of 0.83 indicated good agreement among the interviewers. Statistical analyses {#Sec6} -------------------- Statistical analyses were performed using a SAS (Statistical Analysis System) statistical package (version 9.0; SAS Institute, Inc., Cary, NC). The results are expressed as mean ± standard deviation or proportions. Reliability of the MPAQ was examined by calculating the intra class correlation coefficient (ICC) of the activities reported and presented by urban/rural status and gender. ICC values of \<0.40 were considered as poor agreement, 0.40-0.59 as fair, 0.60-0.74 as good and 0.75-1.0 as excellent agreement \[[@CR11]\]. Relative validity between the GPAQ and MPAQ was also assessed using ICC. Construct validity was used to assess the degree to which a measure (in this case, the MPAQ) compares with an underlying theoretical construct (a latent variable in this context such as BMI and waist circumference). Linear regression models were used to assess the association of deciles/tertiles of activity with BMI and waist circumference after adjusting for age and gender. For assessing criterion validity, the MPAQ was compared against the triaxial accelerometer as a criterion. Spearman's correlation coefficients and 95% CI were used for comparisons. Total duration (minutes/week) of time spent in sedentary and MVPA as estimated from the MPAQ were compared against those recorded by the accelerometer using established cut-points \[[@CR10]\]. As the accelerometer measured data for a week, the data obtained from the MPAQ was also computed for a week so as to make it comparable. Accelerometer data were initially downloaded and processed using customized software viz. Actilife Data Analysis Software \[Version 5.0\], prepared by Actigraph R&D and Software department (Florida, USA). For the purposes of this study, correlation coefficient values: \< 0.20, 0.21-0.40, 0.41-0.60, 0.61-0.80 and 0.81-1.0 were considered as weak, fair, moderate, strong and very strong correlation respectively \[[@CR6]\]. Bland and Altman plots were used to assess the agreement between data obtained using the MPAQ and accelerometer (within the 95% limits). In addition this plot also indicates the random and systematic errors of the data. The mean difference (bias) between accelerometer and the MPAQ, of sedentary activity/ week were plotted (y- axis) against the mean of estimated sedentary minutes/week obtained from the accelerometer and MPAQ \[[@CR12]\]. A similar plot was constructed for MVPA minutes as well. A p-value \<0.05 was considered significant for all statistical measures. Results {#Sec7} ======= Of the 543 selected participants in the reliability study, 288 were male (53%). Tables [1](#Tab1){ref-type="table"} and [2](#Tab2){ref-type="table"} show the baseline characteristics of the study population state-wise, in urban and rural areas respectively. In urban areas, the overall mean age was 44 ± 14 years, BMI was 23.7 ± 4.2 kg/m^2^, waist 83.5 ± 12.1 cm, SBP 134 ± 20 mm Hg and DBP 81 ± 12 mmHg. Overall, 82% of the population was literate, 19% smoked and 21% consumed alcohol. In rural areas, the mean age was 42 ± 13 years, BMI 21.9 ± 4.2 kg/m^2^, waist 78.3 ± 11 cm, systolic BP 129 ± 19 mm Hg and diastolic BP 79 ± 11 mm Hg. 74% were literate, 16% smoked and 20% consumed alcohol.Table 1**Baseline characteristics of the urban population studied (State Wise)VariablesChandigarhHaryanaBiharArunachal PradeshTripuraAssamJharkhandGujaratMaharashtraTamil NaduOverall**N20202116171921202035209Age (years)42.8 ± 3.944.3 ± 3.840.8 ± 2.945.1 ± 3.245.4 ± 2.844.2 ± 4.046.2 ± 3.937.8 ± 5.646.4 ± 4.848.3 ± 4.744.4 ± 14.2BMI (kg/m^2^)24.5 ± 11.425.9 ± 12.422.4 ± 13.623.1 ± 14.122.8 ± 12.722.1 ± 17.724.3 ± 14.825.5 ± 12.721.7 ± 14.824.2 ± 15.723.7 ± 4.2Waist (cms)87.0 ± 11.086.6 ± 11.281.7 ± 9.980.4 ± 7.984.8 ± 8.980.4 ± 9.989.4 ± 14.382.6 ± 15.678.0 ± 11.783.5 ± 13.783.5 ± 12.1SBP (mmHg)139 ± 11134 ± 9131 ± 13134 ± 13130 ± 12136 ± 7134 ± 11132 ± 11123 ± 13138 ± 16134 ± 20DBP (mmHg)80 ± 2081 ± 1581 ± 2284 ± 2383 ± 1579 ± 2278 ± 2080 ± 1576 ± 1684 ± 2581 ± 12Literate n(%)15 (75.0)18 (90.0)12 (57.1)10 (62.5)14 (82.4)17 (89.5)17 (81.0)17 (85.0)18 (90.0)33 (94.3)171 (81.8)Smoking n(%)3 (15.0)7 (35.0)7 (33.3)3 (18.8)5 (29.4)1 (5.3)4 (19.0)1 (5.0)2 (10.0)6 (17.1)39 (18.7)Alcohol n(%)6 (30.0)3 (15.0)4 (19.0)7 (43.8)1 (5.9)2 (10.5)6 (28.6)0 (0.0)8 (40.0)6 (17.1)43 (20.6)SBP- Systolic blood pressure; DBP- Diastolic blood pressure; BMI-Body Mass Index.Table 2**Baseline characteristics of the rural population studied (State Wise)VariablesChandigarhHaryanaBiharArunachal PradeshTripuraAssamJharkhandGujaratMaharashtraTamil NaduOverall**N30303137262432303064334Age (years)43.7 ± 5.144.0 ± 5.839.9 ± 3.038.5 ± 3.740.7 ± 3.737.9 ± 2.945.8 ± 3.639.5 ± 3.445.1 ± 3.741.7 ± 4.141.7 ± 13.4BMI (kg/m^2^)24.2 ± 11.623.5 ± 13.720.8 ± 15.822.7 ± 11.221.7 ± 14.519.7 ± 11.820.1 ± 11.922.1 ± 10.920.1 ± 16.322.6 ± 14.121.9 ± 4.2Waist (cms)84.0 ± 13.583.9 ± 14.775.6 ± 10.580.6 ± 7.079.4 ± 10.072.6 ± 6.874.8 ± 10.079.4 ± 9.372.1 ± 8.979.1 ± 11.878.3 ± 11.2SBP (mmHg)128 ± 12130 ± 13130 ± 11130 ± 12127 ± 9132 ± 12127 ± 10127 ± 11125 ± 11132 ± 11.0129 ± 18.8DBP (mmHg)79 ± 1580 ± 2377 ± 2384 ± 2176 ± 1580 ± 2377 ± 1581 ± 1379 ± 1779 ± 2079 ± 11.4Literate n (%)25 (83.3)23 (76.7)13 (41.9)26 (70.3)20 (76.9)18 (75.0)19 (59.4)22 (73.3)22 (73.3)60 (93.8)248 (74.3)Smoking n (%)5 (16.7)5 (16.7)1 (3.2)10 (27.0)9 (34.6)2 (8.3)5 (15.6)5 (16.7)6 (20.0)6 (9.4)54 (16.2)Alcohol n (%)10 (33.3)5 (16.7)2 (6.5)16 (43.2)7 (26.9)3 (12.5)11 (34.4)1 (3.3)8 (26.7)2 (3.1)65 (19.5)SBP- Systolic blood pressure; DBP- Diastolic blood pressure; BMI-Body Mass Index. Table [3](#Tab3){ref-type="table"} shows the results of the reliability study. The maximum time spent was in the work domain both in urban and rural areas as well as among males and females. Out of the 543 participants at baseline, 520 were available in the 1st month (95.7%). Overall the ICC for TEE and PAL between the baseline and the 1st month were 0.82 and 0.73 respectively, demonstrating good reliability of the MPAQ. The table also presents the ICC domain wise. The lowest ICC was for recreational activity (0.58), and the highest for sitting (0.81). The reliability was good in both males and females and in both urban and rural areas.Table 3**Reliability study: ICC of MPAQ at baseline and 1 monthDomains/ActivitiesOverallMaleFemaleUrbanRuralTime spent (mins/d)ICCTime spent, (mins/d)ICCTime spent, (mins/d)ICCTime spent, (mins/d)ICCTime spent**\ **(mins/d)ICCBase line, N = 5431 month, N = 520Base line, N = 2881 month, N = 277Base line, N = 2551 month, N = 243Base line, N = 2091 month, N = 199Base line, N = 3341 month, N = 321**Work507.1 ± 212.2490.5 ± 222.70.72540.4 ± 191.8524.2 ± 210.60.74428.4 ± 237408.8 ± 231.10.63546.5 ± 258.1535.2 ± 268.80.79482.2 ± 173.3463 ± 184.40.61Transport135.3 ± 59.2130.6 ± 66.40.67159.5 ± 61.0149.2 ± 71.30.62108.0 ± 43.0109.2 ± 52.70.61123.5 ± 53.4119.9 ± 56.10.63142.7 ± 61.5137.1 ± 71.30.72Sleeping370.2 ± 55.7358.06 ± 56.80.63368.9 ± 55.3362.5 ± 56.10.60371.7 ± 56.3353 ± 57.20.67380.2 ± 47.5371.9 ± 53.40.64364 ± 59.5349.5 ± 57.20.61General activity\*199.9 ± 116.3173.8 ± 97.90.75132.3 ± 66.2118.0 ± 66.40.59276.2 ± 113.4237.5 ± 88.70.61211.1 ± 119.2178.5 ± 106.50.75192.9 ± 114171 ± 92.20.74Recreation\#254.7 ± 111.1241.4 ± 115.80.58238.0 ± 106.7237.3 ± 111.40.61273.6 ± 113.1246.2 ± 120.70.57256.7 ± 115.3240.1 ± 115.60.64253.5 ± 108.6241.7 ± 116.10.54Walking as an exercises48.4 ± 38.452.4 ± 45.40.6853.7 ± 43.158.8 ± 47.90.7242.4 ± 32.745.1 ± 42.60.5430.9 ± 24.837.38 ± 29.20.7159.3 ± 42.161.6 ± 51.60.65Sitting176.9 ± 152.7190.8 ± 163.80.81213.5 ± 160.1229.3 ± 166.40.83134.8 ± 132.0146.7 ± 149.40.74224.0 ± 187.2231 ± 192.60.85148.01 ± 118.3165.9 ± 137.70.73TV viewing127.4 ± 55.6120.7 ± 57.90.67119 ± 53.4118.7 ± 55.70.71136.8 ± 56.7123.1 ± 60.30.63128.4 ± 57.6120.5 ± 57.80.78126.7 ± 54.3120.9 ± 58.00.60Total Energy Expenditure (TEE)36.9 ± 11.435.1 ± 12.10.8239.6 ± 11.738.7 ± 12.60.8633.8 ± 10.431.1 ± 10.00.7436.6 ± 11.734.9 ± 12.70.8237.04 ± 11.335.20 ± 11.70.81Physical Activity Level (PAL)^\$\$^1.56 ± 0.471.47 ± 0.490.731.67 ± 0.481.61 ± 0.530.751.44 ± 0.431.30 ± 0.400.651.56 ± 0.471.47 ± 0.500.781.57 ± 0.471.47 ± 0.510.70\*Include obligatory personal and domestic chores ^\$\$^PAL- Physical Activity Level. ^\#^Recreation includes sitting, TV viewing and walking as an exercises. The relative validity between the already validated GPAQ and MPAQ was evaluated. Of the 543 participants administered the MPAQ, data from GPAQ was available in 440. The ICC between GPAQ and the MPAQ was 0.40 overall, indicating fair correlations between the questionnaires. The ICC was found to be highest for sitting (0.57) followed by MVPA recreational (0.50), work vigorous (0.46) and transport (0.44). (Data not shown). Figure [2](#Fig2){ref-type="fig"} shows the construct validity of the MPAQ. Figure [2](#Fig2){ref-type="fig"}a to d show the association of sitting and MVPA measured by the MPAQ with BMI and waist circumference independent of age and gender. Individuals in the lowest decile of sitting had the lowest waist circumference (mean difference between highest decile and lowest decile - 32.1 cm, 95%CI −35.0 to - 29.1; p value \<0.001) (Figure [2](#Fig2){ref-type="fig"}a) and BMI (mean difference between highest decile and lowest decile −9.3 kg/m^2^, 95%CI −10.42 to −8.15; p value \<0.001) (Figure [2](#Fig2){ref-type="fig"}b) compared to those in the highest decile whereas those in the highest tertile of MVPA had the lowest waist circumference (highest tertile versus lowest tertile mean difference in WC 3.47 cm, 95%CI 0.50-6.44; p value =0.02), a 2.5% decrease in WC (Figure [2](#Fig2){ref-type="fig"}c) and lowest BMI (highest tertile versus lowest tertile mean difference in BMI 2.3 kg/m^2^, 95%CI 1.41-3.20; p value \<0.001), a 9.5% decrease in BMI (Figure [2](#Fig2){ref-type="fig"}d).Figure 2Construct validity of the MPAQ (a-d). **a** Association of waist circumference with deciles of time spent in sitting(means and 95% confidence intervals), \*p value adjusted for age and sex. **b** Association of BMI with deciles of time spent in sitting(means and 95% confidence intervals), \*p value adjusted for age and sex. **c** Association of waist circumference with tertiles of time spent in moderate/vigorous activity (means and 95% confidence intervals), \*p value adjusted for age and sex. **d** Association of BMI with tertiles of time spent in moderate/vigorous activity (means and 95% confidence intervals) \*p value adjusted for age and sex. For criterion validity, after excluding individuals with missing data and incorrect usage of accelerometer, data from 103 individuals was available for analysis. Data was considered acceptable if it covered activity over five full days (10 hours/day) including a weekend day. Of the 102 participants included in the criterion validity study, 54 (52.4%) were male. The mean age of the participants was 32 ± 8.7 years and the mean BMI, 25.2 ± 4.5 kg/m^2^. Table [4](#Tab4){ref-type="table"} shows the correlation between physical activity assessed by the MPAQ and measured by the accelerometer. The Spearman's correlation coefficients for sedentary activity, MVPA and overall physical activity for MPAQ against the accelerometer were 0.48 (95% CI-0.32-0.62), 0.44 (0.27-0.59) and 0.46 (0.29-0.60) respectively, showing modest correlation of the MPAQ with the reference method (accelerometer).Table 4**Spearman's correlation between accelerometer and MPAQ (n = 103)VariablesAccelerometer (mins/week)MPAQ (mins/week)Spearman's correlationMean ± SDMean ± SDr95% CI**Sedentary activity5009 ± 7854965 ± 8690.4840.32-0.62Moderate and Vigorous physical activity252 ± 208350 ± 2060.4430.27-0.59Total physical activity7120 ± 76.87110 ± 1970.4580.29-0.60 Figure [3](#Fig3){ref-type="fig"}a and b present the Bland and Altman plots representing the agreement between sedentary activity and MVPA obtained from the MPAQ (measured as minutes/week) against the accelerometer (also measured as minutes/week). Figure [3](#Fig3){ref-type="fig"}a shows that the agreement between the MPAQ and the accelerometer for sedentary behavior was good \[mean bias = 44.4 minutes/week, ±2SD −1599 to 1688\] min/week. A similar plot was seen for MVPA \[mean bias = 97.8, ±2SD −502.6 to 307.1\] min/week (Figure [3](#Fig3){ref-type="fig"}b).Figure 3Bland-Altman plots showing the agreement between MPAQ and accelerometer. **a:** Agreement for sedentary activity (n=103). **b:** Agreement for moderate-to-vigorous activity (n=103). Discussion {#Sec8} ========== The MPAQ was developed with the objective of producing an instrument which can be used widely in epidemiological studies in India. One of the main advantages of the MPAQ is that it assesses physical activity patterns over the period of a year, enabling it to take into consideration, the variability in activity patterns that occurs over this time period. Other physical activity questionnaires capture activities over a typical week or month, which may be unduly influenced by the time at which the questionnaire is administered and may not be truly representative of the individual's overall levels of physical activity over longer periods of time. This is especially so in rural areas where agricultural activities are inherently seasonal in nature. Even in urban areas, individuals could have more than one job at a time, either full-time or part-time. Others might pursue one form of occupation during the week and another during the weekend. The MPAQ attempts to address the issue of variability in physical activity patterns by asking specifically about seasonality and nature of occupations performed and listing in detail the number of months in a year, weeks in a month, days in a week and hours in a day each specific job is performed (for up to two jobs). It also specifically looks at time spent in walking, sitting, standing etc. for each of the jobs performed. The MPAQ also had good reliability at 1 month from baseline with an ICC of 0.82 which is comparable to other widely accepted PAQs, such as the EPIC- Norfolk questionnaire, GPAQ, IPAQ and IMS questionnaires, all of which have reported reliability values varying from 0.67 to 0.88 \[[@CR6],[@CR8],[@CR13]-[@CR16]\]. Questionnaires assessing past year activity have also shown similar estimates of repeatability \[[@CR17]-[@CR20]\]. The reliability was best for sitting and lowest for the recreational domain, which could perhaps be a reflection of the magnitude of the activities performed. In the relative validity study, the MPAQ showed moderate correlation (ICC 0.40) with the GPAQ in all the domains. The absence of a stronger correlation is not surprising since the two questionnaires are inherently different in many respects. For instance, the GPAQ does not capture details on many activities culturally relevant to India that are included in the MPAQ. Even in domains that are common (i.e., transport domain), walking and cycling are captured together as one question in GPAQ in contrast to the MPAQ, where these are captured separately, thus making comparisons difficult. The construct validity study used categories of physical activity based upon reported time in different activity intensities to predict association with anthropometric indices such as waist circumference and BMI. Individuals who performed the most physical activity had the lowest waist circumference and BMI, thereby providing a more robust assessment of validity of the MPAQ. However, as in any cross-sectional study, these results must considered in light of the probability of reverse causation, that is, individuals becoming less active on account of the physical and psychological constraints imposed by overweight and obesity. For measuring criterion validity, we used the triaxial accelerometer as the reference method. While tools such as indirect calorimetry, doubly labelled water method and heart rate monitoring have been considered as the gold standard for measures of energy expenditure, logistics precluded the use of these techniques in the present study. Accelerometers are a less expensive and more convenient alternative for the objective measurement of physical activity and have been widely used for this purpose \[[@CR8],[@CR13]\]. Our results show that the relationship was good for comparisons of total activity, sedentary behaviour and MVPA measures obtained using the MPAQ and the triaxial accelerometer. This is in line with other questionnaires such as the EPIC-Norfolk questionnaire, GPAQ, IPAQ, IMS questionnaire etc. which have been evaluated for criterion validity against different reference methods and have correlation coefficients ranging from 0.28 to 0.48 \[[@CR6],[@CR9],[@CR14]\]. This newly developed physical activity questionnaire provides a practical method to assess physical activity for risk stratification of chronic diseases in large observational studies and epidemiological surveillance. This questionnaire captures various dimensions of physical activity such as the type, intensity and duration and estimates the time spent in these activities, and hence could rank individuals based on the physical activity level. Thus this questionnaire could be recommended for studies that assess the health consequences and correlates of physical activity or lack thereof. The MPAQ also has some limitations. Firstly, as for any questionnaire, it is subject to recall bias, which could lead to overestimation of activity levels in some domains and underestimation in others. Also, recent events are likely to be recalled more accurately than those in the distant past, introducing another element of bias. The questionnaire only calculates physical activity for an "average" day, week or month; physical activity assessments for a specific day, week or month by month cannot be obtained. In addition, in the Bland Altman- the mean values were a small fraction of the magnitude of the measurement, indicating that the MPAQ is a better tool for large epidemiological studies rather than for individual assessments. However, it was also seen that the MPAQ over-reports MVPA and under-reports sedentary behavior, to a minimal extent. This observation is perhaps due to the small number of subjects engaged in MVPA and the relatively low volume of MVPA performed by these subjects. Moreover, physical activity questionnaires have been shown to be prone to both random and systematic errors, with respondents tending to over-report physical activity and under-report sedentary behavior that are influenced by cultural and social desirability factors \[[@CR15]\]. In additions the questionnaire also requires adequate training and understanding in order to obtain good quality data. To conclude, our results show that the MPAQ is a relevant, acceptable, reproducible and valid instrument, which, in a single administration, captures data from multiple activity domains over the period of a year from adults of both genders and varying ages in various walks of life residing in urban and rural areas of different regions of the country. The MPAQ is also the first to have been validated in the north-eastern region of India, which differs significantly from the rest of the country in terrain, climatic conditions and lifestyles, all of which can alter physical activity patterns. In addition, the MPAQ is also easy to administer and well-understood by the study subjects. The MPAQ can therefore be considered an important addition to the physical activity researcher's armamentarium in that it helps to overcome many of the challenges of measuring physical activity in low and middle income countries like India, making assessment of this important risk factor easier and perhaps, more accurate. Additional file {#Sec9} =============== Additional file 1:**MDRF Physical activity questionnaire (MPAQ).** **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** RMA conceived the study, its design, and was involved in implementation of the study, interpretation of the data and drafting the manuscript. SV, LG, RG and MRB were involved in coordination of the study, helped in the execution of the study. LN, SS and VSB were responsible for data management and statistical analysis. RP and DM were responsible for supervision of the study in their respective states. AVK, RU, and VM were part of the study expert committee and helped with the conception and design of the study and revising the manuscript critically for important intellectual content. All authors read and approved the final manuscript.

Learning pointsAs the number of transcatheter aortic valve implantation (TAVI) has increased in the last decade, a unique set of post-intervention events and complications have been identified.This is the first clinical case report to describe the occurrence of acute necrotizing oesophagitis (ANE) as a possible complication of TAVI.Transcatheter aortic valve implantation patients are prone to ANE due to their advanced age and frailty, comorbidities, ischaemic insult, and anti-platelet therapy.Acute necrotizing oesophagitis is a rare entity with high rate of acute and long-term complications and successful management requires early detection and multidisciplinary care. Introduction ============ Transcatheter aortic valve implantation (TAVI) has revolutionized the management of aortic stenosis in the last decade.[@ytz069-B1] However, as the number of TAVI has increased, a unique set of post-intervention events and complications have been identified. Bradyarrhythmias, vascular access complications, acute cerebral vascular accidents, acute renal failure, cardiac tamponade, aortic root rupture, and aortic regurgitation are some of the most common described complications.[@ytz069-B2] Their recognition is critical to improve patients' outcomes, often involving a multidisciplinary team. In the present article, we report the occurrence of acute necrotizing oesophagitis (ANE), an undescribed complication possibly associated with TAVI. Timeline ======== Time Description of relevant clinical information Intervention required ----------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------ December 2016 Evaluation in outpatient clinic: complains of fatigue on moderate exertion in the last 4 months. Performing of diagnostic tests in ambulatory. January 2017 Severe aortic valve stenosis diagnosis (mean gradient of 52 mmHg and valve area of 0.87 cm^2^). Discussion in Heart Team. March 2017 Accepted to transcatheter aortic valve implantation (TAVI). TAVI scheduling. Day 0 TAVI procedure, transfemoral approach, without acute complications. Admission to intensive care unit. Dual antiplatelet therapy (DAPT) initiation. Day 1 Favourable evolution (without access complications or rhythm disturbances and with haemodynamic stability). Day 2 Acute abundant haematemesis, haemodynamic instability, and haemoglobin drop. Endoscopy showed lesions compatible with acute necrotizing oesophagitis diagnosis. Fluid therapy and replacement of blood products. Proton-pump inhibitor (IBP) therapy, domperidone and sucralfat. DAPT suspended. Day 3 Haematemesis in small quantity. Therapeutic maintenance. Day 4 Favourable evolution, with haemodynamic stabilization and without haematemesis. Therapeutic maintenance. Day 5 Endoscopy showing resolution of the previously described lesions. Oral diet was progressively initiated. Day 6 IBP was switched to oral formulation. Day 7 DAPT was re-instituted. Day 8 Patient was discharged clinically well and without any symptoms. August 2017 Follow-up in outpatient clinic, without symptoms or complications. Case presentation ================= We present the case of an 86-year-old woman with a past medical history of endometrial cancer, treated by hysterectomy, oophorectomy and radiotherapy, and with consequent radiation cystitis; she also had dyslipidaemia and osteoporosis, associated with thoracic kyphosis, chronic back pain and mobility impairment (need of walking stick support, help with daily activities and house-keeping). The patient presented with fatigue on moderate exertion with progressive aggravation in the last 4 months. On cardiac auscultation, a mid-systolic murmur was audible along the upper right sternal border. The electrocardiogram revealed sinus rhythm with left axis deviation and repolarization abnormalities (*Figure*[*1*](#ytz069-F1){ref-type="fig"}). Transthoracic echocardiography (TTE) revealed: left ventricular (LV) concentric hypertrophy (LV mass index of 123 g/m^2^; relative wall thickness of 0.49); non-dilated LV (end-diastolic volume of 45 mL/m^2^), normal ejection fraction (57%), and diastolic dysfunction (*e*/*a* ratio of 0.5; *e*/*e*′ of 13; left atrium volume index of 44 mL/m^2^; and tricuspid regurgitation velocity of 2.9 m/s); ascending aorta dilatation (49 mm); tricuspid aortic valve, severely calcified, with severe stenosis (sAS)---mean gradient of 52 mmHg and valve area of 0.87 cm^2^. Blood tests showed N-terminal prohormone of brain natriuretic peptide of 1302 pg/mL (0--300 pg/mL), without other significant abnormalities (Table [1](#ytz069-T1){ref-type="table"}). Coronary angiography was normal. Due to symptomatic sAS, the case was discussed in Heart Team, in order to decide the best approach to aortic valve replacement. Considering the advanced age, moderate frailty (Clinical Frailty Scale assessment of 6,[@ytz069-B5] normal body mass index of 22 kg/m^2^), restricted mobility, and intermediate surgical risk \[Society of Thoracic Surgeons (STS) score: 4.2% risk of mortality\], TAVI was considered the best option. Pre-TAVI computed tomography angiography was performed to determine the best vascular access and the appropriate dimensions of the aortic valve prosthesis. It confirmed dilatation of the ascending aorta (50 mm) and diffuse atherosclerotic disease, without any obstructive lesions. Elective transfemoral TAVI was scheduled. {#ytz069-F1} ###### Blood tests results Parameters (unit) Pré-TAVI First day Post-TAVI Second day Post-TAVI Fourth day Post-TAVI At discharge Normal range ---------------------------------- ---------- --------------------- ---------------------- ---------------------- -------------- -------------- Haemoglobin (g/dL) 12.7 11.4 9.6 9.9 10.4 12--15.3 Leucocytes (×10^9^/L) 4.0 6.5 3.04 3.0 3.0 4.0--11.0 Platelets (×10^9^/L) 181 115 119 146 170 150--450 Partial thromboplastin time (s) 27.3 31.0 Prothrombin time (s) 11.3 11.6 Urea (mg/dL) 52 48 36 22 21 16--49 Creatinine (mg/dL) 0.99 0.9 0.89 0.75 0.83 0.51--0.95 Sodium (mmol/L) 139 140 143 139 139 135--145 Potassium (mmol/L) 4.8 4.1 3.9 3.8 4.2 3.5--5.1 C-reactive protein (mg/dL) 0.09 7.49 7.55 2.4 1.51 \<0.5 Aspartate aminotransferase (U/L) 18 16 13 15 0--32 Alanine aminotransferase (U/L) 6 6 5 5 0--33 Gamma glutamyl transferase (U/L) 12 9 9 0--40 Alkaline phosphatase (U/L) 99 63 65 35--105 Bilirubin (mg/dL) 0.32 0.32 \<1.2 NTproBNP (pg/mL) 1302 580 \<300 The procedure was performed under general anaesthesia and guided by transoesophageal echocardiogram (TOE), via the right femoral route. Balloon valvuloplasty (20 × 40 mm balloon) was performed prior to aortic valve prosthesis implantation (Edwards Sapiens 3™, 23 mm, through a NovaFlex™ delivery system). Valvuloplasty and prosthesis deployment were accomplished during a burst of rapid ventricular pacing, using a temporary pacemaker, with 30--40 s of intentional low cardiac output state---systolic blood pressure (SBP) of 40--50 mmHg. After valve deployment, the patient presented a short period of hypotension (SBP 60--70 mmHg and diastolic blood pressure ±20 mmHg; heart rate ±100 b.p.m.), with spontaneous recovery in less than 3 min. The final result was satisfactory, with a slight leak in relation to the native left coronary cusp, a maximum transvalvular gradient of 12 mmHg, and a mean gradient of 5.3 mmHg. The patient was extubated at the end of the procedure and was admitted to the intensive care unit, with maintained haemodynamic stability and on spontaneous ventilation. Dual antiplatelet therapy (DAPT)---acetylsalicylic acid 100 mg and clopidogrel 75 mg---was initiated, in addition to the patient's usual medication (otilonium bromide, trospium chloride, nitrofurantoin, simvastatin, risedronate, calcium, and vitamin D). Despite the favourable initial evolution, during the second day after TAVI the patient presented with acute abundant haematemesis (coffee ground colour vomitus) associated with hypotension (decrease of SBP from 130 mmHg to 90 mmHg) and haemoglobin drop from 12.7 g/dL to 9.6 g/dL (12--15.5 g/dL). C-reactive protein was slightly elevated (7.55 mg/dL; reference range 0--0.5 mg/dL), as usual following an invasive procedure (*Table*[*1*](#ytz069-T1){ref-type="table"}). Intravenous pantoprazole infusion was administered, the patient was kept nil by mouth, intensive crystalloid fluid therapy was initiated, and an emergency upper gastrointestinal endoscopy was performed. Endoscopy showed striking diffuse, circumferential, black-appearing oesophageal mucosa with signs of friability, and under-distension of the lumen, involving the distal oesophagus (*Figure*[*2*](#ytz069-F2){ref-type="fig"}). These findings were compatible with ANE, also known as *black oesophagus*. Accordingly, haemodynamic stabilization measures (fluid therapy and replacement of blood products) and proton-pump inhibitor (PPI) therapy were maintained, domperidone and sucralfat were initiated, and DAPT was temporarily suspended. Haematemesis stopped and the patient was stabilized 48 h after the acute event. Computed tomography angiography was performed to evaluate any abnormality in oesophageal vascularization or aortic lesion and to exclude complications: it showed wall thickness of lower oesophagus with mucosal enhancement and densification of perioesophageal fat, without other abnormalities. Upper endoscopy was repeated 72 h later, showing resolution of the previously described lesions (*Figure*[*3*](#ytz069-F3){ref-type="fig"}). Afterwards, oral diet was progressively initiated, PPI was switched to oral formulation, and DAPT was re-instituted under surveillance, with good tolerance. Oral iron supplementation was initiated. The patient was discharged on the eighth day after TAVI, clinically well. She was referred for short-term outpatient clinic evaluation by Cardiology and Gastroenterology. {#ytz069-F2} {#ytz069-F3} At follow-up, the patient remained without symptoms of heart failure or related to ANE complications. Transthoracic echocardiography evaluation 3 months after discharge showed maintenance of a very slight leak in relation to the native right coronary cusp, with no significant changes in all other parameters. Endoscopic reassessment was not performed, since the previous showed good evolution and the patient had no symptoms or signs of stenosis or relapsing necrosis. Discussion and conclusions ========================== To the best of our knowledge, this is the first clinical case report to describe the occurrence of ANE as a possible complication of TAVI. There is a single case of isolated proximal *black oesophagus* reported in a patient who underwent cardiac catheterization, an event that was possibly related to the procedure.[@ytz069-B6] The ANE is an entity first described in 1990 by Goldenberg *et al*.[@ytz069-B7] It is a very rare disease, affecting men four times more commonly than women, with undetermined incidence and is often under-diagnosed; in endoscopic studies the prevalence of ANE ranges from 0.01% to 0.2%.[@ytz069-B8] The most frequent clinical manifestations are the occurrence of upper gastrointestinal bleeding (90% of the cases), haematemesis or melaenas, and chest pain.[@ytz069-B11] The usual triggering factor is an episode of ischaemic insult---haemodynamic instability, shock and/or hypoxia.[@ytz069-B14] Aggression or fragilization of the oesophageal mucosa by internal or external factors, and the existence of associated medical comorbidities are recognized predisposing factors.[@ytz069-B7] It has a typical endoscopic presentation, previously described in this case, primarily affecting the distal region of the oesophagus, which is related to the scarce vascularity of this region.[@ytz069-B14] Biopsy is usually unnecessary, but it can be performed specially when it is important to make a differential diagnosis with melanoma, acanthosis nigricans, and pseudo-melanosis of the oesophagus. Treatment of ANE should focus on supportive care to maximize organ perfusion, treatment of underlying medical conditions, and intravenous PPI. Adequate monitoring to detect and properly treat complications is also essential.[@ytz069-B15] The importance of recognition of this entity is related to the high rate of acute and long-term complications---occurrence of transmural necrosis with perforation, mediastinitis, mediastinal abscess, sepsis, and oesophageal stenosis (10--25% at long-term follow-up). Mortality depends on the underlying condition and it is variably reported in the literature, between 6% to as high as 36%, when associated with other comorbidities.[@ytz069-B10]^,^[@ytz069-B12] There are limited data to guide the management of ANE, and it is largely based upon clinical experience. Repeat endoscopic evaluation should be done to assess mucosal healing. With supportive care, resolution of endoscopic findings occurs in most patients within 72 h to 2 weeks. Symptoms of oesophageal stenosis should be monitored in the long term.[@ytz069-B16] In this clinical case, the transient low output state during TAVI was probably the ischaemic insult that triggered ANE. The period of hypotension was short (\<4 min), without evidence of other end-organ hypotensive damage. However, the insult of TOE during TAVI and DAPT may have led to oesophageal mucosa fragilization, making it especially vulnerable to the ischaemic injury. Patient's age, frailty and diffuse atherosclerotic disease are additional predisposing factors of ANE. The reason why this specific patient had this manifestation, when most TAVI patients share these risk factors, is not known. The use of conscious sedation, with local anaesthesia and TTE guidance, could minimize the risk of this complication, avoiding the additional aggression of the oesophageal mucosa by the TOE probe. In addition to demonstrating, for the first time, the occurrence of a rare complication with a possible association with TAVI, this clinical case is also paradigmatic and exemplary of the importance of the multidisciplinary approach of these complex patients that extends even beyond the concept of the Heart Team. Ethical disclosures =================== The authors declare that they have followed the protocols of their work centre on the publication of patient data. The authors have obtained the written informed consent of the subject mentioned in the article. Lead author biography ===================== {#ytz069-F4} Inês S. Gonçalves completed her Medical Degree in 2012, School of Medicine of the University of Minho, Portugal. She underwent general internship in 2013 and specific residency training in Cardiology between 2014 and 2018 at the Cardiology Department of Santa Maria University Hospital---CHULN E.P.E, Lisbon, where she is currently working. She is a researcher at Cardiovascular Centre of the University of Lisbon (CCUL) in the areas of *Cardiac Rhythm Abnormalities* and *Heart Failure and Cardiomyopathies*, her main areas of interest. She has taken part in several clinical trials as co-investigator and has 150 scientific communications at national and international meetings. **Slide sets:** A fully edited slide set detailing this case and suitable for local presentation is available online as [Supplementary data](#sup1){ref-type="supplementary-material"}. **Consent:** The author/s confirm that written consent for submission and publication of this case report including image(s) and associated text has been obtained from the patient in line with COPE guidance. **Conflict of interest:** none declared. Supplementary Material ====================== ###### Click here for additional data file.

Introduction {#S0001} ============ Infections owing to soil-transmitted helminths and foodborne trematodes are serious public health problems in Laos.[@CIT0001]--[@CIT0003] Previous studies have reported a high prevalence of intestinal parasitic infections among school-aged children and young people in Laos. For example, the reported prevalence of intestinal helminths among children aged 6--11 years is 70.3%, and this is 71.2% among young people aged 12--19 years in northern Laos.[@CIT0004] Among children under 15 years of age in Bolikhamxay Province, this prevalence is 56.7%.[@CIT0005] Recent studies show that the main parasitic infections among the general population are hookworm and *Opisthorchis viverrini* infections.[@CIT0002],[@CIT0006],[@CIT0007] This tendency has also been observed among children and young people, with 80.0% of the children aged 6--15 years infected by hookworm and 64.6% by *O. viverrini* in Champasack Province,[@CIT0008] and 87.0% infected by hookworm and *O. viverrini* in Saravane Province.[@CIT0002] The impact of helminth infections is significant predictors of malnutrition, iron deficiency anemia, and poor academic performance among schoolchildren.[@CIT0009]--[@CIT0012] Additionally, *O. viverrini* infection is associated with cholangiocarcinoma and bile duct cancer.[@CIT0013],[@CIT0014] Therefore, a successful parasite control program for students at school is necessary to improve their health status, increase their chances of optimal growth and learning,[@CIT0015] and avoid future early death. According to the National School Health Policy formulated by the Ministry of Education (MoE) and Ministry of Health (MoH) in Laos, the school health program consists of five components: personal health and life skills, healthy school environment, health and nutrition services, control and prevention of common diseases, and school and community partnership.[@CIT0016] Therefore, helminths control is also included in the program and has been implemented as the school deworming campaign for all primary school-aged children since 2005.[@CIT0017] This campaign has covered 93% of the country.[@CIT0018] The school-based program is expected to provide schoolchildren with proper knowledge, attitudes, and practices for maintaining better health status throughout their entire life. However, the influence of this program during primary school and the factors related to parasitic infection among secondary school students has not been sufficiently investigated. In the present study, we sought to identify factors, including primary school health programs, related to infection with hookworm and *O. viverrini* among secondary school students in a large city of Laos. Material and methods {#S0002} ==================== Study area {#S0002-S2001} ---------- The study was conducted in Vientiane, the largest city in Laos. Vientiane is located on the banks of the Mekong River near the border with Thailand, at elevation 174 m above sea level. Vientiane has a tropical monsoon climate, with the rainy season occurring between May and October.[@CIT0019] Vientiane has nine districts. The study site was a secondary school in Sikottabong District, which is located at about 9 km from the center of the city and urban area. The students of the target school were mainly from nine villages located in the vicinity of the school. The total number of students at the school was around 230. The gross enrollment rate in secondary schools in Laos is 66.5%.[@CIT0020] Study population and sample size {#S0002-S2002} -------------------------------- As of the 2015 census, Vientiane city had a population of 820,000, with 8.9% of the population aged 10--14 years.[@CIT0019] We selected one secondary school located in an area where we have implemented a school health program since 2002. The sample size was determined using the A-priori Sample Size Calculator for Multiple Regression (available from <http://www.danielsoper.com/statcalc>) with a 0.15 anticipated effect size, 80% statistical power, 15 predictors, and 0.05 probability level. The calculated minimum sample size was 139. Study design {#S0002-S2003} ------------ We conducted a cross-sectional survey in September 2013. First, schoolteachers and students were informed about the study objectives and procedures. A total of 197 students agreed to voluntarily participate and signed consent forms. Students in first grade were not included because they were in the process of admission during the study period. Questionnaire survey {#S0002-S2004} -------------------- The questionnaire included sociodemographic characteristics, perception of the primary school health program, and status of water and sanitation in the student's home. This questionnaire was translated into the Lao language, and any discrepancies in terms of wording were settled through discussions with Laotian senior researchers from the Lao Tropical and Public Health Institute. Finally, the translated questionnaires were pre-tested. To collect data, first, a Laotian researcher explained the questionnaire to the students and instructed them on how to complete the questionnaire. The students then completed the self-administered questionnaires in the Lao language. Sample collection and stool examination {#S0002-S2005} --------------------------------------- All students participated were provided with a stool container labeled with an identification number on the day before the survey. Students were requested to provide their own fresh stool sample on the sample collection day. The research team visited classrooms at the target school between 8:00 a.m. and 9:00 a.m. to collect the stool samples together with the completed questionnaire. All collected samples were kept in a cool box and were transferred by car to the laboratory of the Lao Tropical and Public Health Institute within an hour after collection. For each stool sample, two Kato--Katz thick-smear slides were prepared, using standard 41.7 mg templates. After a clearing time of 30 mins, the slides were examined under a light microscope (100× magnification). All samples were examined on the day of collection. Data analysis {#S0002-S2006} ------------- In total, 178 participants submitted a stool sample, and samples were analyzed for the prevalence of helminths among study participants. Among 178 students, 164 completed the questionnaire. The data from these 164 participants were used to analyze associations among sociodemographic characteristics, primary school health program score, status of hygiene and sanitation in the student's home, and parasitic infection. The items addressing the perception of the school health program were from a previous study on school health program evaluation in Laos.[@CIT0021] Response options to the survey questions were "yes" or "no", and "yes" responses were summed to obtain the overall primary school health program score ([Table 1](#T0001){ref-type="table"}). The intensity of helminth infections was expressed in fecal eggs per gram (EPG). According to the previous studies,[@CIT0002],[@CIT0006],[@CIT0022] for hookworm and *O. viverrini* infections, the following light-, moderate-, and high-intensity groups were established based on the EPG counts: hookworm; 1--1,999 EPG, 2000--3,999 EPG, and ≥4,000 EPG and *O. viverrini* and *T. trichiura*: 1--999 EPG, 1,000--9,999 EPG, and ≥10,000 EPG.Table 1Questionnaire items for primary school health program scoreItems%There was water available in primary school70.6There was boiled or bottled water for students to drink87.1There was a latrine in primary school93.3The latrine was always kept tidy78.5The primary school had a good fence to prevent animals from entering the schoolyard91.4Health personnel visited the primary school to check students' health condition87.1Primary school teachers instructed students to wash hands before eating98.2Primary school teachers instructed students to wash hands after using the latrine95.1Primary school teachers instructed students to wash hands after returning from outside69.9Primary school teachers instructed students on how to clip fingernails97.5Primary school teachers instructed students on how to keep eyes and ears clean88.3Mean (SD)Total score (range: 5--11)9.57 (1.33)[^1] Data analysis was performed using IBM SPSS, version 22.0 (IBM Corp., Armonk, NY, USA). The association among parasite infections (hookworm and *O. viverrini*), sociodemographic characteristics, and the school health program available during primary school were assessed in a univariate logistic regression analysis. Predictors with *p*\<0.25 were retained in a multivariate logistic regression model. Odds ratios (ORs) and 95% confidence intervals (CIs) were reported. The significance level was set at *p*\<0.05. Ethics approval and consent to participate {#S0002-S2007} ------------------------------------------ Ethical approval was obtained from the Tenri Health Care University Ethics Committee (Project no. 27) and the National Ethics Committee on Health Research, MoH, Lao PDR (Ethical Clearance No. 045/2013 NECHR). Permission for the survey was obtained from the MoH, as well as the school committee and director. Meetings were held with students' parents or guardians in which they were given a detailed explanation of the study aims, procedures, potential risks, and benefits. Written informed consent was obtained from the parents or guardians of participants. In cases where a parent or guardian was illiterate, students signed the consent form, witnessed by a parent or guardian." Participants were free to withdraw from the study at any time. Results {#S0003} ======= Sociodemographic characteristics of participants {#S0003-S2001} ------------------------------------------------ The data of 164 students were included in the analysis; mean student age was 13.7 years (SD 1.3; range 10--18 years). The mean student age when entering primary school was 6.0 years (SD 1.0; range 3--9 years), and the mean age when entering secondary school was 11.5 years (SD 1.0; range 9--15 years). A total of 49.4% of the students were male, and most students (86.0%) attended primary school in Vientiane. Among the total, 41.4% of the students' fathers and 52.4% of the students' mothers had fewer than 6 years' education. The most common occupations among fathers were farmer (33.5%) and government employee (24.5%); the most common jobs among mothers were farmer (28.1%) and trader (26.8%). A total of 86.0% of the students reported having a latrine at home and 87.2% of the used a water-sealed latrine; only 21.3% of the students had tap water available at home. Among the total, 89.0% of the students belonged to the Lao-Thai ethnic group ([Table 2](#T0002){ref-type="table"}).Table 2Sociodemographic characteristics of participantsVariablesn (Mean)% (SD)(Range)Age, y13.71.310--18Age entering primary school, y6.01.03--9Age entering secondary school, y11.51.09--15Location of primary school Vientiane city14186.0 Outside of Vientiane city2012.2 Missing31.8Student's sex Male8149.4 Female8350.6Father\'s education Primary school or no schooling6841.4 Lower secondary school or higher9256.0 Missing42.4Mother\'s education Primary school or no schooling8652.4 Lower secondary school or higher7545.7 Missing31.8Father\'s occupation Unemployed42.4 Farmer5533.5 Trader1710.4 Government employee4024.5 Driver95.5 Other3823.2 Missing10.6Mother\'s occupation Unemployed3118.9 Farmer4628.1 Trader4426.8 Government employee148.5 Seamstress116.7 Other1710.4 Missing10.6Latrine Latrine present at home14186.0 Share latrine with neighbors2012.2 No latrine31.8Type of latrine Water-sealed latrine14387.2 Pit latrine or other2112.8Source of water for daily use Tap water3521.3 Other (river/well/rainwater)12978.7Ethnicity Lao-Thai14689.0 Hmong138.0 Other31.8 Missing21.2[^2] Parasitic infection {#S0003-S2002} ------------------- The most frequently encountered parasitic infection was *O. viverrini*, with a positive rate of 39.0%. Hookworm, *Trichuris trichiura*, and *Taenia* spp. were found in 36.0%, 4.3%, and 2.4% of the participants, respectively. Most cases of *O. viverrini*, infection were categorized as light intensity, with no cases of heavy intensity infection found. Most cases of hookworm infection were also light intensity; in addition, three cases were classified as moderate intensity and one case as heavy intensity ([Table 3](#T0003){ref-type="table"}).Table 3Prevalence and intensity of parasitic infection by gender and age group (n=164)GenderAge groupsn (%)FemaleMale10--1112--1314--1516--18*Opisthorchis viverrini* Intensity of infection:  Negative100 (61.0)45 (54.2)55 (65.4)3 (50.0)48 (66.7)43 (57.3)6 (54.5)  Light (1--999 EPG)62 (37.8)37 (44.6)25 (30.9)3 (50.0)23 (31.9)31 (41.3)5 (45.5)  Moderate (1,000--9,999 EPG)2 (1.2)1 (1.2)1 (1.2)0 (0.0)1 (1.4)1 (1.3)0  Heavy (≥10,000 EPG)0000000Hookworm Intensity of infection:  Negative105 (64.0)52 (62.7)53 (65.4)6 (100.0)56 (77.8)37 (49.3)6 (54.5)  Light (1--1,999 EPG)55 (33.5)29 (34.9)26 (32.1)016 (22.2)36 (48.0)3 (27.3)  Moderate (2,000--3,999 EPG)3 (1.8)1 (1.8)2 (2.5)001 (1.3)2 (18.2)  Heavy (≥4,000 EPG)1 (0.6)1 (1.8)0001 (1.3)0*Trichuris trichiura* Intensity of infection:  Negative157 (95.7)79 (95.2)78 (96.3)6 (100.0)69 (95.8)72 (96.0)10 (90.9)  Light (1--999 EPG)6 (3.7)3 (3.6)3 (3.7)03 (4.2)2 (2.7)1 (9.1)  Moderate (1,000--9,999 EPG)1 (0.6)1 (1.2)0001 (1.3%)0  Heavy (≥10,000 EPG)0000000*Taenia* spp.4 (2.4)2 (2.4)2 (2.5)02 (2.8)2 (2.7)0[^3] Factors related to hookworm infection {#S0003-S2003} ------------------------------------- Univariate logistic regression analysis resulted in 10 predicted variables: age, age entering primary school, age entering secondary school, primary school health program score, father's education level, father's occupation; mother's occupation, type of latrine, source of daily-use water, and frequency of deworming. The results of multivariate logistic regression analysis showed that two variables were identified as significantly related to hookworm infection; older age (OR=1.55, 95% CI: 1.01--2.37, *p*=0.046) was significantly associated with higher infection risk and the student's father having an occupation with regular income was associated with significantly lower risk (OR=0.47, 95% CI: 0.13--0.93, *p*=0.036) for hookworm infection. The primary school health program score was not significant in the multivariate model (OR=0.76, 95% CI: 0.56--1.02, *p*=0.068) although it was significant (OR=0.69, 95% CI: 0.53--0.88, *p*=0.003) in univariate logistic regression ([Table 4](#T0004){ref-type="table"}).Table 4Factors associated with hookworm infection in logistic regressionUnivariateMultivariateIndependent variablesOR95% CI*p*-valueOR95% CI*p*-valueAge1.581.19--2.090.0021.551.01--2.370.046Age entering primary school1.491.04--2.130.0311.370.82--2.300.234Age entering secondary school1.290.93--1.810.1320.680.40--1.160.154Primary school health program score0.690.53--0.880.0030.760.56--1.020.068Location of primary school attended Vientiane1.00 Outside of Vientiane0.570.20--1.660.303Sex Male1.00 Female1.130.60--2.140.711Father's education 6 years or more1.00 Less than 6 years1.921.00--3.680.0511.030.44--2.420.943Mother's education 6 years or more1.00 Less than 6 years1.390.73--2.660.321Father's occupation Unemployed/farmer1.00 Job with regular income0.470.24--0.910.0250.340.13--0.930.036Mother's occupation Unemployed/farmer1.00 Job with regular income0.680.36--1.290.2391.110.43--2.910.828Ethnicity Lao-Thai1.00 Hmong/other0.770.26--2.350.651Latrine at home Latrine available in the home1.00 Share with neighbors/no latrine0.940.37--2.370.898Type of latrine Water-sealed latrine1.00 Pit latrine or other1.920.72--5.140.1951.730.51--5.870.376Source of daily-use water Tap water1.00 Other (river/well/rainwater)1.830.79--4.220.1571.250.43--3.630.679Frequency of deworming Regular deworming medication1.00 No deworming medication3.561.13--11.200.0301.980.53--7.490.313[^4] Factors related to *O. viverrini* infection {#S0003-S2004} ------------------------------------------- Univariate logistic regression analysis resulted in nine predicted variables; school health score, location of primary school attended, student's sex, father's education, mother's education, father's occupation, mother's occupation, ethnicity, and source of daily-use water. The results of multivariate logistic regression analysis showed that only the mother having an occupation with regular income was significantly associated with lower risk (OR=0.30, *p*=0.005) for *O. viverrini* infection ([Table 5](#T0005){ref-type="table"}).Table 5Factors associated with *Opisthorchis viverrini* infection in logistic regressionUnivariateMultivariateIndependent variablesOR95% CI*p*-valueOR95% CI*p*-valueAge1.080.8--1.390.542Age entering primary school0.920.66--1.280.628Age entering secondary school1.160.84--1.610.358Primary school health program score1.170.92--1.500.2071.200.90--1.610.221Location of primary school attended Vientiane1.00 Outside of Vientiane0.340.11--1.060.0630.420.10--1.760.236Student's sex Male1.00 Female1.790.95--3.370.0741.290.62--2.680.496Father's education 6 years or more1.00 Less than 6 years1.500.79--2.840.2161.120.47--2.650.803Mother's education 6 years or more1.00 Less than 6 years1.490.79--2.820.2191.050.44--2.490.909Father's occupation Unemployed/farmer1.00 Job with regular income0.420.22--0.810.0100.770.33--1.810.550Mother's occupation Unemployed/farmer1.00 Job with regular income0.290.15--0.55\<0.0010.300.13--0.690.005Ethnicity Lao-Thai1.00 Hmong/other0.490.15--1.600.2380.810.19--3.450.780Latrine at home Latrine available in the home1.00 Shares with neighbors/no latrine1.010.41--2.480.991Type of latrine Water-sealed latrine1.00 Pit latrine or other0.560.19--1.670.300Source of daily-use water Tap water1.00 Other (river/well/rainwater)2.591.09--6.130.0312.170.80--5.850.127Frequency of deworming Regular deworming medication1.00 No deworming medication1.180.39--3.570.774[^5] Discussion {#S0004} ========== In this study, we found that infection with *O. viverrini* and hookworm were the most prevalent parasitic infections among secondary students in Vientiane, Laos. Our findings confirmed previous studies.[@CIT0002],[@CIT0007],[@CIT0008],[@CIT0022] Compared with countries surrounding Laos, the prevalence rates of hookworm infection (36.0%) in our study were similar to those in Cambodia, with 34.9% among schoolchildren[@CIT0023] and 34.8% among young people aged 11--16 years.[@CIT0024] The prevalence among schoolchildren was slightly lower in Vietnam, 22.8% in the southern part of the country[@CIT0025] and 5.3% in the northern part.[@CIT0026] In contrast, only 0.5% of the schoolchildren in Bangkok, Thailand have hookworm infection.[@CIT0027] Risk factors of hookworm infection include walking barefoot in contaminated soil. In our study, students usually use beach sandals, which cover the soles of their feet. Although 87.2% of the participants had a water-sealed latrine in their dwelling, every rainy season, these flood and overflow with waste, which contaminates the soil. In addition, even in urban areas of Laos, only the main roads are paved and other areas where students walk and play are unpaved. These factors lead to the high prevalence rate of hookworm infection in Vientiane. The prevalence rates of *O. viverrini* (39.0%) in our study were also at the same levels as those in Cambodia, with 40.0% among young people aged 10--19 years.[@CIT0028] However, in northeastern Thailand, only 0.0--2.1% of those aged less than 20 years are infected.[@CIT0029],[@CIT0030] *O. viverrini* infection is caused by ingesting raw or inadequately cooked fish. Suburban areas of Vientiane city still have fishponds and people catch fish for daily consumption. This may lead to frequent raw fish consumption and the subsequent high prevalence of *O. viverrini* infection. The World Health Organization (WHO) categorizes populations with prevalence rates between 10% and 50% as requiring preventive chemotherapy.[@CIT0016] Most participants in this study were covered by a national deworming program. The intensity of hookworm infection showed that most infected cases were categorized as light intensity. This suggests that deworming programs can maintain lower hookworm infection intensity. However, reinfection after mass deworming programs has been frequently reported.[@CIT0031]--[@CIT0033] The medication that was distributed in primary schools for a long time was single-dose mebendazole 500 mg, which has low efficacy against hookworm.[@CIT0034] The study in Laos by Soukhathammavong et al, showed that the cure rate and egg reduction rate with single-dose mebendazole treatment was 17.6% and 76.3%, respectively; these rates with albendazole treatment were 36.0% and 86.7%, respectively.[@CIT0007] A study in China showed that triple-dose albendazole treatment has high efficacy against hookworm, with a 96.7% cure rate, a 99.1% egg reduction rate, and a low reinfection rate (5.1%) at 6 months after treatment.[@CIT0035] To reduce the infection rate and number of cases with heavy intensity, the medication used in deworming programs should be reconsidered. Additionally, mebendazole is less effective for *O. viverrini* infection.[@CIT0004],[@CIT0007] Praziquantel is the only medicine recommended by the WHO for the treatment of *O. viverrini* infection and it should be administered in certain doses according to students' weight.[@CIT0036] Such administration is not easy for schoolteachers; therefore, measures for controlling *O. viverrini* infection is not included in national school health programs in Laos. For safety, the MoH in Laos recommends that medical personnel should supervise and monitor treatment and follow-up when praziquantel is administered. Therefore, good collaboration between MoH and MoE is needed if praziquantel is to be introduced in the school health program. The results of univariate logistic regression in this study showed that a lower primary school health program score was a risk factor of hookworm infection. The primary school questions in our study mainly queried participants about the contents of health education and the school environment during their primary school years. This score possibly includes recall bias; however, hygiene education such as washing hands, using the toilet, eating clean food, wearing clean clothes, and cleaning the environment, may increase students' awareness and improve behaviors, such as the use of regular medication and hygiene practices to prevent hookworm infection. In the other hand, the multivariate model showed that primary school health program scores were not significant factors but sociodemographic factors had a greater influence on hookworm infection. Our findings showed that older students were at greater risk of hookworm infection. This tendency was also found in the previous studies from southern Laos.[@CIT0007],[@CIT0009],[@CIT0037],[@CIT0038] Older students may engage in more agriculture work to help their parents, resulting in greater hookworm infection. The data in our study that students whose father was unemployed or a farmer had a greater risk of hookworm infection also support this result. With respect to the father's occupation, the percentage of students whose father worked as a farmer was the highest in this study, even in Vientiane city. Our study findings suggested that having a mother who was employed and had regular income was associated with lower risk of *O. viverrini* infection. Several studies in Laos have shown that low economic status is associated with *O. viverrini* infection.[@CIT0007],[@CIT0037] In addition, previous studies on female economic empowerment have demonstrated that women who have financial resources tend to spend their income on their children's diet, education, and health.[@CIT0039],[@CIT0040] The main cause of *O. viverrini* infection is eating raw or partially cooked fish. Fish is a low-cost food, in comparison with meat. Therefore, a mother may spend her money to buy meat rather than fish and her children may thus have a lower risk of infection with *O. viverrini*. Our study clarified the situation of *O. viverrini* and hookworm infections, as well as social factors predicting both infections, among adolescents in Vientiane, the largest city in Laos. However, this study has several limitations. First, the participants in our study were from a purposely selected school in an urban area; therefore, generalization of the study findings is limited. Second, our study was intended to evaluate the effect of a school health program conducted when participants were in primary school; therefore, recall bias was unavoidable. More accurate measures must be developed to evaluate school health programs for preventing hookworm and *O. viverrini* infection. Third, our study could not determine students' habits of raw fish consumption. Because participants' daily food is prepared by parents or guardians, students' habits may be influenced by those of their parents or guardians. Further studies are required to determine whether there is an association between *O. viverrini* infection in children and the consumption frequency of raw fish prepared by parents or guardians. Fourth, our study was conducted in 2013; therefore, some variables might not indicate the present situation, especially parents' occupation and hygienic and sanitary facilities at home, owing to rapid economic development in Laos. Further evaluation is needed in future studies. Conclusions {#S0005} =========== This study highlights that *O. viverrini* and hookworm infection are prevalent among secondary school students in Vientiane city. Improving primary school health programs is effective in reducing the risk of hookworm infection among secondary students, although sociodemographic factors have an influence on both infections. *O. viverrini* infection is associated with mothers having an occupation with regular income, which is a factor associated with lower risk. Hookworm infection is associated with age and father's occupation. Older students have greater infection risk, and students whose father is employed and has regular income have lower risk. Current primary school health programs may be insufficient to reduce *O. viverrini* infections, and different approaches may be needed, such as encouraging mothers to seek jobs that provide regular income for their family. We are grateful to all participants. Our sincere thanks go to the staff at the National Institute of Tropical and Public Health Lao PDR for their support and fieldwork assistance. We also thank Analisa Avila, ELS, of Edanz Group ([www.edanzediting.com/ac](http://www.edanzediting.com/ac)) for editing a draft of this manuscript. This work was supported by a research grant from Tenri Health Care University in 2014. Disclosure {#S0006} ========== The authors report no conflicts of interest in this work. [^1]: **Abbreviation:** SD, standard deviation. [^2]: **Abbreviation:** SD, standard deviation. [^3]: **Abbreviation:** EPG=egg per gram of stool. [^4]: **Abbbreviations:** OR, odds ratio; CI, confidence interval. [^5]: **Abbreviations:** OR, odds ratio; CI, confidence interval.

Background {#Sec1} ========== Adverse food reactions (AFRs) are defined as abnormal responses to an ingested food or food additive \[[@CR1], [@CR2]\]. When an AFR is suspected, its diagnosis should always be based on a complete and accurate dietary history, clinical signs and, in particular, on the results of a dietary elimination test, i.e., an "elimination and re-introduction diet" \[[@CR3]\]. Indeed the gold standard method to diagnose an AFR consists of feeding the animal a limited-antigen diet until the abatement of clinical signs (elimination diet) and then reintroducing the diet previously fed to demonstrate the recurrence of symptoms (provocation diet) \[[@CR3]\]. Other tests, such as skin testing, skin patch testing, serologic tests measuring food allergen-specific serum IgE, and gastric biopsy tests have already been shown to have low sensitivity and specificity \[[@CR3]--[@CR5]\]. The elimination diet should contain protein and carbohydrate sources that the animal has never eaten before the trial and should be administered for a period of at least 8--10 weeks \[[@CR6]\]. Clinical signs of gastrointestinal disease usually improve within 2 weeks, whereas cutaneous clinical signs may take up to 8 to 12 weeks to respond to dietary change. For the elimination diet, veterinarians can use either commercial veterinary prescription diets, such as limited-antigen diets or hydrolyzed diets, or home-cooked diets \[[@CR6]\]. Commercially available limited-antigen diets typically include proteins from venison, quail, rabbit and duck and are generally combined with alternative carbohydrate sources, such as green peas, rice and potatoes. A hydrolyzed diet can provide an alternative to the novel protein diet; they contain small peptides of molecular weights below 3 kDa, which are easily digestible and associated with low antigenic stimulation \[[@CR7]\], but their efficacy has not always been confirmed \[[@CR6], [@CR8]--[@CR12]\]. In the case of no evident improvement in clinical signs during a dietary trial, AFR cannot be excluded, because residual allergenic fragments may remain in the hydrolyzed diet and contamination may have occurred during the production process of limited-antigen diets \[[@CR13]\]. In each of these cases, the correct diagnosis of AFR would be compromised. Many studies have already demonstrated the presence of undeclared ingredients in limited-antigen dry diets, dry and wet physiological foods, vegetarian and vegan diets, supplements and treats for pets, such as microscopic bone fragments, proteins, detected by enzyme-linked immunosorbent assay (ELISA), and deoxyribonucleic acid (DNA), detected by polymerase chain reaction (PCR), reverse transcription PCR, multiplex PCR or PCR-restriction fragment length polymorphism \[[@CR14]--[@CR24]\]. Considering that the production technologies involved in dry and wet pet food manufacturing are distinct, the aim of the present study was to investigate whether the high rate of contamination found in commercial limited-antigen dry diets \[[@CR13]\] is also applicable to the antigen-limited wet diets often used in dietary elimination trials for the diagnosis and therapy of AFR in pets. Methods {#Sec2} ======= Samples {#Sec3} ------- Eleven canine and feline antigen-limited wet diets (10 novel protein diets and 1 hydrolyzed diet) produced by 5 different pet food manufacturers (2 Italian and 3 international) were obtained from veterinary clinics. All of the products used in this study were declared as dietetic complete pet food, specifically intended for particular animal nutritional purpose of reduction of ingredients and nutrient intolerances, according to the European law 2008/38/EC \[[@CR25]\]. For each sample, the product's label was carefully studied to identify all protein sources of animal origin and of vegetal origin in the case that the product was declared to be vegetarian. The product's brand, description, therapeutic indications, animal species, feed material declaration, additives lists, analytical constituents, instruction for proper use, net weight, lot number, expiry date, and the name and address of the business operator responsible for its labelling were recorded. Each pet food product was randomly assigned a unique three-digit sample identification number. Samples were then submitted to the polymerase chain reaction (PCR) to identify DNA belonging to 3 zoological classes (mammalian, avian and fish). Based on these results, PCR analyses for specific animal species (quail, rabbit, duck, horse, deer, turkey, chicken, domestic ruminant, tuna and swine) were performed according to the zoological classes testing positive. The limited-antigen diet composed of vegetal protein only was also tested using PCR to identify vegetal protein origin. DNA extraction {#Sec4} -------------- To extract DNA from pet food samples, the DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) was used following the manufacturer's instructions. Negative controls (water) were included in each extraction. DNA concentration was determined by spectrophotometry (Nanodrop ND1000; NanoDrop Technologies Inc., Wilmington, DE, USA). Species identification protocol {#Sec5} ------------------------------- Previously published PCR primers (Table [1](#Tab1){ref-type="table"}) designed to recognise species-specific regions of mitochondrial DNA were used to test for the presence of quail, rabbit, duck, horse, turkey, chicken, ruminant and swine products. Deer and tuna were identified using minisequencinq protocols (Table [1](#Tab1){ref-type="table"})**.** For horse, a specific real time PCR assay was run in a 7300 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The PCR amplifications were performed in an ABI 2720 thermocycler (Applied Biosystems, Foster City, CA,USA). Minisequencing was performed on an ABI 310 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). All amplification protocols were performed following the published protocols (Table [1](#Tab1){ref-type="table"}).Table 1Primers used for the detection of species specific and zoological class DNAPrimerPCR product (bp)ReferencesDetection of species-specific DNA Chicken95Martín et al., 2007 \[[@CR32]\] Deer232La Neve et al., 2008 \[[@CR33]\] Duck64Martín et al., 2007 \[[@CR32]\] Horse147Kesmen et al., 2009 \[[@CR34]\] Quail129Rojas et al., 2010 \[[@CR35]\] Rabbit160Walker et al., 2004 \[[@CR36]\] Domestic Ruminant104Dalmasso et al., 2004 \[[@CR15]\] Swine108Meyer et al., 1995 \[[@CR37]\] Tuna132Bottero et al., 2007 \[[@CR38]\] Turkey122Martín et al., 2007 \[[@CR32]\]Detection of zoological class DNA MAMMALIAN117Chiappini et al., 2005 \[[@CR39]\] FISH224Dalmasso et al., 2004 \[[@CR15]\] POULTRY183Dalmasso et al., 2004 \[[@CR15]\] VEGETABLE132Little DP, 2014 \[[@CR39]\] Results {#Sec6} ======= Eleven canine and feline antigen-limited wet diets were collected and submitted to PCR to test for the presence of DNA from 3 different zoological classes (mammalian, avian and fish), and subsequently for DNA from specific animal species as well as for DNA of vegetal origin. Table [2](#Tab2){ref-type="table"} shows the animal protein sources as listed on the product labels and the results obtained by species-specific PCR. Discrepancies between the results expected and those obtained from PCR were observed in 6 out of the 11 sampled diets. Of the 6 showing discrepancies, 5 were contaminated with protein from additional animal species and 1 was wholly composed of animal protein sources completely unrelated to those declared on the label. Of these 5 contaminated with additional animal species, 2 contained DNA belonging to 2 different zoological classes, i.e., inter-zoological class contaminations: Sample n.6 declared rabbit as its unique animal protein source on the label, but PCR detected DNA belonging to 2 different zoological classes: 2 avian (turkey and chicken) and 1 mammalian (rabbit). Sample n.11 declared duck as its unique animal protein source on the label, but PCR again detected DNA belonging to 2 different zoological classes: 3 avian (duck, chicken, and turkey) and 1 mammalian (pork).. In the remaining 3 samples contaminated with additional animal species, the contaminating species belonged to the same zoological class. Sample n.1 declared turkey as its unique animal protein source on the label, whereas PCR analysis detected DNA from turkey and chicken. Sample n.3 declared deer as its unique animal protein source on the label, but PCR analysis detected DNA from another domestic ruminant in addition to deer. Sample n.8 declared quail as its unique animal protein source on the label, but PCR analysis confirmed the presence of another avian species (not covered by the species-specific PCR primers listed in Table [1](#Tab1){ref-type="table"}) in addition to quail. Finally, the label for sample n.7 indicated duck as its unique animal protein source, but PCR analysis was only able to detect DNA from turkey, chicken and horse, (and thus a case of inter zoological class contamination); no traces of duck DNA were found. In the remaining 5 sample diets, PCR analysis confirmed the animal protein species sources declared on the label.Table 2List of declared animal protein sources and PCR results expressed as zoological class and species in 11 antigen-limited wet dietsSamplesDeclared animal protein source- Zoological classDeclared animal protein source-speciesDeclared fat sourcePCR analysis results- Zoological classPCR analysis results-Species1ATurkeyVegetalATurkey\ Chicken2MHorseChickenMHorse3MDeerFishMDeer\ Domestic Ruminant4NoneNoneVegetalNoneVegetal^a^5FTunaFishFTuna6MRabbitVegetalM\ ARabbit\ Chicken\ Turkey7ADuckVegetalA\ MTurkey\ Chicken\ Horse8AQuailVegetalAQuail\ Another Avian9MHorseVegetalMHorse10AHydrolyzed ChickenFish and VegetalAHydrolyzed Chicken11ADuckFish and VegetalA\ MDuck\ Turkey\ Chicken\ Swine*A* avian, *F* fish, *M* mammalian^a^ Diet declared as "vegetarian", with no animal species declared, but positive results for a vegetal primer Discussion {#Sec7} ========== The elimination diet is currently the most important and reliable diagnostic test for evaluating and diagnosing AFRs in dogs and cats \[[@CR26]\]. The test comprises two phases: 1) the accurate selection by veterinarians of a specific limited-antigen diet (commercial or homemade, based on a novel or hydrolyzed protein), avoiding ingredients previously fed; 2) the meticulously administration of this selected diet by the pet's owner. Limited-antigen diets should have the following characteristics: eliminate the potential allergens from the diet, be easy to use, readily available and reasonably priced \[[@CR8]\]. Commercial limited-antigen diets are most frequently used due to their convenience and relative low cost, but unfortunately a recent study reports that limited-antigen dry diets should not be considered reliable due to their high degree of contamination with other sources of animal protein not declared on the label \[[@CR14]\]. The authors found more than 80% of such diets to be contaminated \[[@CR14]\]. The present study used PCR analysis to assess for the potential cross-contamination of antigen-limited wet diets with animal proteins not mentioned on the label and confirmed a high rate of contamination (54.5%) despite the different technologies involved in wet animal food production and the different feed materials commonly used. Discrepancies between the labelling declarations and the results obtained by PCR analyses were observed in 6 out of the 11 diets sampled; in particular, of these 6 contaminated diets, 5 were contaminated with additional animal species and 1 was composed of animal protein sources completely unrelated to that declared on the label. These results provide an important input to the debate on the potential causes behind why elimination diets may fail; i.e., if animal protein contaminants are widespread in the commercial dietetic limited-antigen pet foods, inadequate practices in the pet food industry could be implicated as a cause of AFRs. The pet food industry has a legal obligation to produce safe food for its consumers (Article 1, Regulation No. 767/2009) \[[@CR27]\]. Based on the Directive 2008/38/EC, dietetic limited-antigen diets are indicated for the reduction of ingredient and nutrient intolerances in dogs and cats and for that purpose they should be formulated with selected protein sources and/or carbohydrate sources; in the current EU legislation, the tolerance of analytical traces of other animal proteins is not discussed \[[@CR25]\]. Regarding allergens, the pet food industry has not been obliged to embrace allergen risk management, as the human food industry has had to. Food allergen risk should be considered in parallel with physical, chemical and microbiological risks. For example, in the manufacturing of baby foods intended for human consumption, production lines are often dedicated to the manufacture of a single product composed of specific antigenic ingredients. Furthermore, the labelling of human foodstuffs is obliged to inform the consumer of the possible presence of potential allergens, and allergen management has become an integral part of the Hazard Analysis and Critical Control Points (HACCP). With regard to human food, the EU has implemented regulations that require the presence of certain potential allergens, which are present as ingredients, to be declared by manufacturers (EU Directive 2007/68/EC and EU Regulation 1169/2011, which came into force in 2014) \[[@CR28], [@CR29]\]. Thus, the management of animal protein traces in the pet food industry providing dietetic limited-antigen foods should be seen as an integral part of the management of existing food safety procedures and should take into account all the operations, from feed materials to manufacturing and the packaging of the final products. The adoption of good practices in pet food manufacturing will allow veterinarians to make more informed choices in relation to elimination trials, which is essential for the correct diagnosis and treatment of AFRs in dogs and cats. Indeed, the high prevalence of AFRs in dogs and cats is an important reason why allergen management should become an integral part of the food safety management system in the pet food industry. In canine populations, AFRs were diagnosed in 1.69% of dogs presented to a veterinary teaching hospital in a single year \[[@CR30]\]. In cats, the prevalence of AFRs was described to be less than 1%, but higher in cats with skin diseases, for whom it ranged from 3 to 6% \[[@CR31]\]. Thus the high incidence of animal protein cross-contamination in dietetic pet foods, highlighted by our results and by many previous studies \[[@CR14]--[@CR24]\], indicates a highly unsatisfactory situation. Animal protein cross-contamination could occur at two levels in the pet food industry: i) during the production of the feed materials (animal by-products); or ii) during the actual production of pet food (limited-antigen diets). The possibility of intentional cross-contamination of either feed materials or final products is unlikely considering the similar or sometimes higher cost of the undeclared animal species identified in our samples, such as the horse protein found in sample n.7 which declared duck as its unique animal protein source. Intentional cross-contamination might occur in situations of limited availability of some selected protein sources, such as duck, for which availability cannot always be guaranteed for pet food production with high meat inclusion rates. In order to ensure a high level of feed safety and to improve transparency, provisions should be added as an integral aspect of good practice. Although every effort could be taken by the pet food industry to eliminate the risk posed by the unintended presence of food allergens, in many businesses it is virtually impossible to produce a zero risk product. In many manufacturing premises, production lines dedicated to the manufacture of a single product are not always feasible or practical for economic or logistical reasons. Therefore, the cleaning of shared equipment, processing lines and the local environment becomes a key element in allergen control. Manufacturers are presently required to define the cleaning procedures and cleaning schedules appropriate for their facilities. But they are not required to ensure or to demonstrate that the cleaning programme is effective and performing properly through the use of specific and accurate analyses, such as PCR-food analysis. Multiple lines are more frequently used for the production of commercial dry pet food than for the production of wet food. This could justify the lower contamination rate of our wet samples (54.5%) compared with that for dry foods analysed by other authors, described as being over 80% \[[@CR14]\]. Our methodology was also more sensitive, involving the use of species-specific primers rather than zoological class primers only, as used by previous authors \[[@CR14]\]. Had we used zoological class primers only, the apparent rate of contamination would have been greatly underestimated: lower than 20% instead of 54.5% as obtained using species-specific primers. Indeed, had we used zoological class primers, sample 7 would not have resulted as being a totally misleading diet, and samples 1, 3, 8 and 11 would not have resulted as being contaminated at all. One of the most interesting results obtained in our study using the species-specific PCR was that diets with protein sources derived from horse and fish (tuna) were always clean (100% as declared, with no contaminations). These data might imply that the different feed material production techniques affect the potential for cross-contamination by other animal protein sources; moreover, they highlight and shift the focus from the pet food manufacturers to the suppliers of feed materials (animal by-products). It seems that when separate lines are required for the slaughter, transformation and distribution of animal products (as is the case for horse meat and fish), the final by-products are also cleaner. In the past, some authors also focused their attention on lipid sources of animal origin. In the present study, the animal lipid sources declared on the label were never found to be species-specific or zoological class PCR positive \[[@CR14]\]. It would be interesting if processing standards similar to those used for hydrolyzed protein production (Section 5 D Reg. 142/2011) were also applied in manufacturing plants processing animal proteins intended for limited-antigen diets \[[@CR28]\]. Adherence to these specific processing requirements could help avoid cross-contaminations between different animal protein sources. Good practices that would help avoid cross-contaminations could include: the use of dedicated containers and means of transport maintained in a clean state (cleaned, washed and/or disinfected after each use) for the carriage of animal protein sources of a single animal species; the clear and total separation of the manufacturing plant areas involved in the processing of animal proteins intended for limited-antigens diets, from reception until dispatch; the correct setting up and management of the equipment used for product processing, including separate processing lines and cleaning procedures to exclude the risk of cross-contamination. The results of this study are very important for pet food producers of limited-antigen diets. They indicate the need for improvements to be made in the selection of feed material suppliers and the type of checks installed throughout the production chain before final products are introduced onto the market. Importantly, our results indicate that the type of checks used should involve PCR analyses for species-specific classes and not just zoological class. Conclusion {#Sec8} ========== In conclusion, the high incidence of contaminations found in the antigen-limited wet diets analyzed in our study confirm the severity of the problem as previously highlighted for dietetic limited-antigen dry diets \[[@CR14]\]. These results provide a new slant on the debate regarding the causation of elimination diet failure as they suggest that the practices in place in the pet food industry lead to the presence of animal protein contaminants in commercial dietetic limited-antigen diets, which could, in turn, result in AFRs. In order to ensure food safety standards of these commercial diets and guarantee the efficacy of the diagnosis and treatment of AFRs in dogs and cats, general best practices should be adopted by the pet food industry. The management of cross contaminant animal proteins in limited-antigen diets may be addressed at three levels: (1) The feed materials and supply chain: establish an appropriate policy for assessing the allergen status of the feed materials, such as the identification of species-specific allergens by PCR analysis in feed materials received from suppliers before their unloading in the factory; (2) manufacturing equipment and processes: integrate cross-contamination risk assessment management into production processes at different levels; this should involve cleaning programs for staff, equipment, production and packaging lines and a method for validating efficacy that involves PCR analyses and clearly identification of the re-work programs in order to be tracked; (3) labelling: in the interests of transparency, information should be added to the label regarding the possibility of cross contamination antigens. This additional information should be mandatory for dietetic antigen-limited foods and include the analytical limit of detection and testing method. A : Avian AFR : Adverse food reactions DNA : Deoxyribonucleic acid F : Fish FA : Food allergy M : Mammalian PCR : Polymerase chain reaction The authors thank the staff of the pet nutrition consulting service for their participation in the study's design. Availability of data and materials {#FPar1} ================================== The datasets supporting the conclusions of this article are included within the article. EP and LP carried out the studies and drafted the manuscript. CL participated in the design of the study. DM performed the PCR analysis. MT and AD carried out the interpretation of the PCR analysis. AS helped draft the manuscript. All authors have read and approved the final manuscript. Ethics approval and consent to participate {#FPar2} ========================================== Not applicable. Competing interests {#FPar3} =================== The authors declare that they have no competing interests. Publisher's Note {#FPar4} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Introduction {#Sec1} ============ In recent years, inequalities with power-exponential functions have been intensively studied \[[@CR1]--[@CR7]\]. They have many important applications. For example, they can be found in mathematical analysis and in other theories like mathematical physics, mathematical biology, ordinary differential equations, probability theory and statistics, chemistry, economics. For more details, a literature review and the history of inequalities with power-exponential functions, see \[[@CR2]\]. Cîrtoaje, in \[[@CR1]\], has introduced the following interesting conjecture on the inequalities with power-exponential functions. The inequality is similar to the reverse arithmetic-geometric mean inequality where its terms were rearranged. Conjecture 1 {#FPar1} ------------ *If* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a,b \in(0,1]$\end{document}$ *and* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r \in(0,e]$\end{document}$, *then* $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ 2\sqrt{a^{ra}b^{rb}}\geq a^{rb}+b^{ra}. $$\end{document}$$ The conjecture was proved by Matejíčka \[[@CR3]\]. Matejíčka also proved ([1](#Equ1){ref-type=""}) under other conditions in \[[@CR4], [@CR5]\]. For example, he showed that ([1](#Equ1){ref-type=""}) is also valid for $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a,b,r \in (0,e]$\end{document}$. In \[[@CR5]\], one interesting property of the generalized Cîrtoaje's inequality (CI) was found. In \[[@CR6]\], a classification of sets of solutions of (CI) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ n\sqrt[n]{\prod_{i=1}^{n}x_{i}^{rx_{i}}} \geq x_{n}^{rx_{1}}+ \sum_{i=1}^{n-1}x_{i}^{rx_{i+1}} $$\end{document}$$ was made. Methods {#Sec2} ======= In this paper, methods of mathematical and numerical analysis are used. We make a classification of sets of solutions of the other generalization of (CI). Let *φ*, *ψ* be functions from $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\{1,\ldots,n\}$\end{document}$ to $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\{1,\ldots,n\}$\end{document}$, where $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n\in N$\end{document}$. Put $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ F(r)=\ln n+\frac{r}{n} \Biggl(\sum_{i=1}^{n}x_{\varphi (i)} \ln x_{i} \Biggr)-\ln \Biggl(\sum_{i=1}^{n}e^{rx_{\psi(i)}\ln x_{i}} \Biggr). $$\end{document}$$ The function $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r)$\end{document}$ is defined on $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$R^{n}_{+}$\end{document}$ where $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n\in\mathbf{N}$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r\geq0$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$R^{n}_{+}=\{(x_{1},\ldots,x_{n}), x_{i}>0, i=1,\ldots,n\}$\end{document}$. We note that $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r)\geq0$\end{document}$ is equivalent to the following generalization of Cîrtoaje's inequality (I): $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ n\sqrt[n]{\prod_{i=1}^{n}x_{i}^{rx_{\varphi(i)}}} \geq \sum_{i=1}^{n}x_{i}^{rx_{\psi(i)}}. $$\end{document}$$ The reverse inequality to (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ n\sqrt[n]{\prod_{i=1}^{n}x_{i}^{rx_{\varphi(i)}}}< \sum_{i=1}^{n}x_{i}^{rx_{\psi(i)}} $$\end{document}$$ we denote by (RI). Results and discussion {#Sec3} ====================== We remark that in \[[@CR6]\] the special case of our classification for ([4](#Equ4){ref-type=""}) was presented, where $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(i)=i$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(i)=i+1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$i=1,\ldots,n-1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(n)=n$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(n)=1$\end{document}$. The following functions: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$\begin{aligned}& g(x_{1},\ldots,x_{n}) = \frac{1}{n}\sum _{i=1}^{n}x_{\varphi(i)}\log (x_{i})-m_{x}, \end{aligned}$$ \end{document}$$ $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$\begin{aligned}& \text{where}\quad m_{x} = \max_{1\leq m\leq n}\bigl\{ x_{\psi(m)}\log (x_{m})\bigr\} , \\& h(x_{1},\ldots,x_{n}) = \sum _{i=1}^{n} (x_{\varphi(i)}-x_{\psi (i)} ) \log(x_{i}), \end{aligned}$$ \end{document}$$ we will call characteristic functions of (I). Put $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ S^{n} =\bigl\{ (x_{1},\ldots,x_{n})\in R^{n}_{+}; x_{i}=x_{j}, i,j=1,\ldots,n \bigr\} . $$\end{document}$$ We prove the following lemma. Lemma 1 {#FPar2} ------- *Let* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r)$\end{document}$ *be defined by* ([3](#Equ3){ref-type=""}). *Let* *φ*, *ψ* *be arbitrary functions from* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\{1,\ldots,n\}$\end{document}$ *to* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\{ 1,\ldots,n\}$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n\in N$\end{document}$. *Then* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r)$\end{document}$ *is a concave function for each* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A\in R^{n}_{+}-S^{n}$\end{document}$, *and* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(0)=0$\end{document}$. *If there is* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$i\neq j$\end{document}$; $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$i,j\in N$\end{document}$ *such that* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$x_{\psi(i)}\ln x_{i}\neq x_{\psi(j)}\ln x_{j}$\end{document}$, *then* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r)$\end{document}$ *is a strongly concave function in* *A*. Proof {#FPar3} ----- $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(0)=0$\end{document}$ is evident. Easy calculation gives $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ F'(r)=\frac{1}{n} \Biggl(\sum_{i=1}^{n}x_{\varphi(i)} \ln x_{i} \Biggr)-\frac{\sum_{i=1}^{n}e^{rx_{\psi(i)}\ln x_{i}}x_{\psi(i)}\ln x_{i}}{ \sum_{i=1}^{n}e^{rx_{\psi(i)}\ln x_{i}}} $$\end{document}$$ and $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ F''(r)=\frac{-L(r)}{ (\sum_{i=1}^{n}\exp (rx_{\psi(i)}\ln x_{i} ) )^{2}} $$\end{document}$$ where $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document} $$\begin{aligned} L(r)&=\sum_{i=1}^{n}\sum _{j=1}^{n}\exp (rx_{\psi(i)}\ln x_{i}+rx_{\psi(j)}\ln x_{j} )x_{\psi(i)}^{2} \ln^{2} x_{i} \\ &\quad {}-\sum_{i=1}^{n}\exp (rx_{\psi(i)}\ln x_{i} )\sum_{j=1}^{n}\exp (rx_{\psi(j)}\ln x_{i} ) (x_{\psi(i)}\ln x_{i})x_{\psi(j)}\ln x_{j} \\ &=\frac{1}{2}\sum_{i=1}^{n}\sum _{j=1}^{n}\exp (rx_{\psi(i)}\ln x_{i}+rx_{\psi(j)}\ln x_{j} )x_{\psi(i)}^{2} \ln^{2} x_{i} \\ &\quad {}+\frac{1}{2}\sum_{i=1}^{n}\sum _{j=1}^{n}\exp (rx_{\psi(i)}\ln x_{i}+rx_{\psi(j)}\ln x_{j} )x_{\psi(j)}^{2} \ln^{2} x_{j} \\ &\quad {}-\sum_{i=1}^{n}\exp (rx_{\psi(i)}\ln x_{i} )\sum_{j=1}^{n}\exp (rx_{\psi(j)}\ln x_{i} ) (x_{\psi(i)}\ln x_{i})x_{\psi(j)}\ln x_{j} \\ &=\sum_{i=1}^{n}\sum _{j=1}^{n}\exp (rx_{\psi(i)}\ln x_{i}+rx_{\psi (j)}\ln x_{j} ) (x_{\psi(i)}\ln x_{i}-x_{\psi(j)}\ln x_{j} )^{2}\geq0. \end{aligned}$$ \end{document}$$ The proof is completed. □ Now we prove the following lemma. Lemma 2 {#FPar4} ------- *Let* *g*, *h* *be defined by* ([6](#Equ6){ref-type=""}), ([7](#Equ7){ref-type=""}). *Let* *φ*, *ψ* *be arbitrary functions from* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\{1,\ldots,n\}$\end{document}$ *to* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\{1,\ldots,n\}$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n\in N$\end{document}$. *Then there are five cases*. *If* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$h(x_{1},\ldots,x_{n})=\sum_{i=1}^{n} (x_{\varphi (i)}-x_{\psi(i)} )\log(x_{i})<0$\end{document}$ *for* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$ *then* (RI) *is valid for all* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r>0$\end{document}$ *in* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$.*If* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$h(x_{1},\ldots,x_{n})=\sum_{i=1}^{n} (x_{\varphi (i)}-x_{\psi(i)} )\log(x_{i})=0$\end{document}$ *and* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$g(x_{1},\ldots,x_{n})= \frac{1}{n}\sum_{i=1}^{n}x_{\varphi(i)}\log (x_{i})-\max_{1\leq m\leq n}\{x_{\psi(m)}\log(x_{m})\}<0$\end{document}$ *for* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$ *then* (RI) *is valid for all* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r>0$\end{document}$ *in* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$.*If* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$h(x_{1},\ldots,x_{n})=\sum_{i=1}^{n} (x_{\varphi (i)}-x_{\psi(i)} )\log(x_{i})=0$\end{document}$ *and* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$g(x_{1},\ldots,x_{n})= \frac{1}{n}\sum_{i=1}^{n}x_{\varphi(i)}\log (x_{i})-\max_{1\leq m\leq n}\{x_{\psi(m)}\log(x_{m})\}=0$\end{document}$ *for* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$ *then* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r)=0$\end{document}$ *for* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r\geq0$\end{document}$ *in* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$.*If* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$h(x_{1},\ldots,x_{n})=\sum_{i=1}^{n} (x_{\varphi (i)}-x_{\psi(i)} )\log(x_{i})>0$\end{document}$ *and* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$g(x_{1},\ldots,x_{n})= \frac{1}{n}\sum_{i=1}^{n}x_{\varphi(i)}\log (x_{i})-\max_{1\leq m\leq n}\{x_{\psi(m)}\log(x_{m})\}\geq0$\end{document}$ *for* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$ *then* (I) *is valid for all* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r\geq0$\end{document}$ *in* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$.*If* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$h(x_{1},\ldots,x_{n})=\sum_{i=1}^{n} (x_{\varphi (i)}-x_{\psi(i)} )\log(x_{i})>0$\end{document}$ *and* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$g(x_{1},\ldots,x_{n})= \frac{1}{n}\sum_{i=1}^{n}x_{\varphi(i)}\log (x_{i})-\max_{1\leq m\leq n}\{x_{\psi(m)}\log(x_{m})\}<0$\end{document}$ *for* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$ *then there is* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r_{0}>0$\end{document}$ *such that* (I)*is valid for* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r\in(0,r_{0}]$\end{document}$ *and* (RI) *is valid for* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r\in(r_{0},\infty)$\end{document}$ *in* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A=(x_{1},\ldots,x_{n})\in R^{n}_{+}$\end{document}$. Proof {#FPar5} ----- The proof is evident. It follows from Lemma [1](#FPar2){ref-type="sec"}. □ Note 1 {#FPar6} ------ It is easy to see that if $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$g(x_{1},\ldots,x_{n})=0$\end{document}$ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$h(x_{1},\ldots,x_{n})=0$\end{document}$ then $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r)=0$\end{document}$ for all $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r\geq0$\end{document}$. Really, from Lemma [1](#FPar2){ref-type="sec"} we have $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F'(0)=0$\end{document}$ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\lim_{r\rightarrow\infty }F'(r)=0$\end{document}$. If $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r_{1})\neq0$\end{document}$ for some $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r_{1}>0$\end{document}$ then $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r_{1})<0$\end{document}$. Then there exists *z* such that $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F(r_{1})-F(0)=F'(z)r_{1}$\end{document}$ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$0< z< r_{1}$\end{document}$. It implies $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F'(z)<0$\end{document}$. Because of $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F''(r)\leq0$\end{document}$ we get $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F'$\end{document}$ is non-increasing for $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r\geq0$\end{document}$. For $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r>z>0$\end{document}$ we obtain $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$F'(r)\leq F'(z)$\end{document}$ so $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\lim_{r\rightarrow\infty}F'(r)\leq F'(z)<0$\end{document}$. This is a contradiction. Conclusion {#Sec4} ========== In this paper, we showed the following. If (I) is valid in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$(x_{1},\ldots,x_{n})$\end{document}$ for some $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r_{0}>0$\end{document}$ then (I) is valid in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$(x_{1},\ldots,x_{n})$\end{document}$ for all $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$0< r\leq r_{0}$\end{document}$. Similarly, if (RI) is valid in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$(x_{1},\ldots,x_{n})$\end{document}$ for some $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r_{0}>0$\end{document}$ then (RI) is valid in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$(x_{1},\ldots,x_{n})$\end{document}$ for all $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r>r_{0}$\end{document}$. We think that the way how to classify sets of solutions of the power-exponential inequalities could be used for other suitable inequalities. Now we give examples of concrete applications of our results. We make the complete classification of sets of solutions for (I) and (RI) inequalities where $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$. Using Matlab for plotting graphs of the solution curves for the characteristic equations $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$g(X)=0$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$h(X)=0$\end{document}$ we obtain the following figures for (I) and (RI). In the figures we denote by $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$I+\mathit{RI}$\end{document}$ the points where (I) and also (RI) are locally valid. By *I* we denote points where (I) is valid for all $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r>0$\end{document}$ and by *RI* we denote points where (RI) is valid for all $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$r>0$\end{document}$. It is easy to show that for $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$ there are only 12 basic cases of inequalities (I). The other four cases of (I) can be transformed to the previous cases. Example 1 {#FPar7} --------- Let $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=2$\end{document}$. Then we have (I): $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{1}}x_{2}^{rx_{2}}} \geq x_{1}^{rx_{1}}+x_{2}^{rx_{2}}, \\ h(x_{1},x_{2})=(1/2) (x_{1}-x_{1} ) \log(x_{1})+(1/2) (x_{2}-x_{2} ) \log(x_{2})=0, \\ g(x_{1},x_{2})=(1/2) \bigl(x_{1} \log(x_{1})+x_{2}\log(x_{2}) \bigr)-\max \bigl\{ x_{1}\log(x_{1}),x_{2}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [1](#Fig1){ref-type="fig"}. Figure 1**Solution points for inequalities Examples** [**1**](#FPar7){ref-type="sec"} **,** [**7**](#FPar13){ref-type="sec"} **,** [**8**](#FPar14){ref-type="sec"} **,** [**11**](#FPar17){ref-type="sec"} **.** Example 2 {#FPar8} --------- Let us consider that $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi (1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=2$\end{document}$. Then by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{2}}x_{2}^{rx_{1}}} \geq x_{1}^{rx_{1}}+x_{2}^{rx_{2}}, \\ h(x_{1},x_{2})=\frac{1}{2} (x_{2}-x_{1} )\log(x_{1})+\frac {1}{2} (x_{1}-x_{2} ) \log(x_{2})=\frac{1}{2} (x_{2}-x_{1} )\ln \biggl(\frac{x_{1}}{x_{2}} \biggr), \\ g(x_{1},x_{2})=(1/2) \bigl(x_{2} \log(x_{1})+x_{1}\log(x_{2}) \bigr)-\max \bigl\{ x_{1}\log(x_{1}),x_{2}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [2](#Fig2){ref-type="fig"}. Figure 2**Solution points for inequalities Example ** [**2**](#FPar8){ref-type="sec"} **.** Example 3 {#FPar9} --------- Put $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=1$\end{document}$. Then we obtain by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{1}}x_{2}^{rx_{2}}} \geq x_{1}^{rx_{1}}+x_{2}^{rx_{1}}, \\ h(x_{1},x_{2})=\frac{1}{2} (x_{1}-x_{1} )\log(x_{1})+\frac {1}{2} (x_{2}-x_{1} ) \log(x_{2}), \\ g(x_{1},x_{2})=(1/2) \bigl(x_{1} \log(x_{1})+x_{2}\log(x_{2}) \bigr)-\max \bigl\{ x_{1}\log(x_{1}),x_{1}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [3](#Fig3){ref-type="fig"}. Figure 3**Solution points for inequalities Example ** [**3**](#FPar9){ref-type="sec"} **.** Example 4 {#FPar10} --------- Let us consider $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=1$\end{document}$. Then we get by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{2}}x_{2}^{rx_{1}}} \geq x_{1}^{rx_{1}}+x_{2}^{rx_{1}}, \\ h(x_{1},x_{2})=(1/2) (x_{2}-x_{1} ) \log(x_{1})+(1/2) (x_{1}-x_{1} ) \log(x_{2}) \\ \hphantom{h(x_{1},x_{2})}=(1/2) (x_{2}-x_{1} )\log(x_{1}), \\ g(x_{1},x_{2})=(1/2) \bigl(x_{2} \log(x_{1})+x_{1}\log(x_{2}) \bigr)-\max \bigl\{ x_{1}\log(x_{1}),x_{1}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [4](#Fig4){ref-type="fig"}. Figure 4**Solution points for inequalities Example ** [**4**](#FPar10){ref-type="sec"} **.** Example 5 {#FPar11} --------- Let $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=2$\end{document}$. Then we have by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{1}}x_{2}^{rx_{1}}}\geq x_{1}^{rx_{2}}+x_{2}^{rx_{2}}, \\ h(x_{1},x_{2})=(1/2) (x_{1}-x_{2} ) \log(x_{1})+(1/2) (x_{1}-x_{2} ) \log(x_{2}) \\ \hphantom{h(x_{1},x_{2})}=(1/2) (x_{1}-x_{2} )\log(x_{2}x_{1}), \\ g(x_{1},x_{2})=(1/2) \bigl(x_{1} \log(x_{1})+x_{1}\log(x_{2}) \bigr)-\max \bigl\{ x_{2}\log(x_{1}),x_{2}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [5](#Fig5){ref-type="fig"}. Figure 5**Solution points for inequalities Example ** [**5**](#FPar11){ref-type="sec"} **.** Example 6 {#FPar12} --------- Put $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=1$\end{document}$. Then by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{1}}x_{2}^{rx_{2}}} \geq x_{1}^{rx_{2}}+x_{2}^{rx_{1}}, \\ h(x_{1},x_{2})=(1/2) (x_{1}-x_{2} ) \log(x_{1})+(1/2) (x_{2}-x_{1} ) \log(x_{2}) \\ \hphantom{h(x_{1},x_{2})}=(1/2) (x_{2}-x_{1} )\log \biggl(\frac{x_{2}}{x_{1}} \biggr), \\ g(x_{1},x_{2})=(1/2) \bigl(x_{1} \log(x_{1})+x_{2}\log(x_{2}) \bigr)-\max \bigl\{ x_{2}\log(x_{1}),x_{1}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [6](#Fig6){ref-type="fig"}. Figure 6**Solution points for inequalities Example ** [**6**](#FPar12){ref-type="sec"} **.** Example 7 {#FPar13} --------- Let us consider $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=1$\end{document}$. Then we obtain by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{1}}x_{2}^{rx_{1}}}\geq x_{1}^{rx_{1}}+x_{2}^{rx_{1}}, \\ h(x_{1},x_{2})=(1/2) (x_{1}-x_{1} ) \log(x_{1})+(1/2) (x_{2}-x_{2} ) \log(x_{2})=0, \\ g(x_{1},x_{2})=(1/2) \bigl(x_{1} \log(x_{1})+x_{1}\log(x_{2}) \bigr)-\max \bigl\{ x_{1}\log(x_{1}),x_{1}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [1](#Fig1){ref-type="fig"}. Example 8 {#FPar14} --------- Put $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=2$\end{document}$. Then by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{2}}x_{2}^{rx_{2}}} \geq x_{1}^{rx_{2}}+x_{2}^{rx_{2}}, \\ h(x_{1},x_{2})=(1/2) (x_{1}-x_{1} ) \log(x_{1})+(1/2) (x_{2}-x_{2} ) \log(x_{2})=0, \\ g(x_{1},x_{2})=(1/2) \bigl(x_{2} \log(x_{1})+x_{2}\log(x_{2}) \bigr)-\max \bigl\{ x_{2}\log(x_{1}),x_{2}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [1](#Fig1){ref-type="fig"}. Example 9 {#FPar15} --------- Let us consider $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=2$\end{document}$. Then we get by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{1}}x_{2}^{rx_{1}}}\geq x_{1}^{rx_{1}}+x_{2}^{rx_{2}}, \\ h(x_{1},x_{2})=\frac{1}{2} (x_{1}-x_{1} )\log(x_{1})+\frac {1}{2} (x_{1}-x_{2} ) \log(x_{2})=\frac{1}{2} (x_{1}-x_{2} )\ln x_{2}, \\ g(x_{1},x_{2})=(1/2) \bigl(x_{1} \log(x_{1})+x_{1}\log(x_{2}) \bigr)-\max \bigl\{ x_{1}\log(x_{1}),x_{2}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [7](#Fig7){ref-type="fig"}. Figure 7**Solution points for inequalities Example ** [**9**](#FPar15){ref-type="sec"} **.** Example 10 {#FPar16} ---------- Let $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=1$\end{document}$. Then by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{1}}x_{2}^{rx_{1}}}\geq x_{1}^{rx_{2}}+x_{2}^{rx_{1}}, \\ h(x_{1},x_{2})=\frac{1}{2} (x_{1}-x_{2} )\log(x_{1})+\frac {1}{2} (x_{2}-x_{2} ) \log(x_{2})=\frac{1}{2} (x_{1}-x_{2} ) \log(x_{1}), \\ g(x_{1},x_{2})=(1/2) \bigl(x_{1} \log(x_{1})+x_{1}\log(x_{2}) \bigr)-\max \bigl\{ x_{2}\log(x_{1}),x_{1}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [8](#Fig8){ref-type="fig"}. Figure 8**Solution points for inequalities Example ** [**10**](#FPar16){ref-type="sec"} **.** Example 11 {#FPar17} ---------- Put $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=1$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=1$\end{document}$. We obtain by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{2}}x_{2}^{rx_{1}}} \geq x_{1}^{rx_{2}}+x_{2}^{rx_{1}}, \\ h(x_{1},x_{2})=(1/2) (x_{1}-x_{1} ) \log(x_{1})+(1/2) (x_{2}-x_{2} ) \log(x_{2})=0, \\ g(x_{1},x_{2})=(1/2) \bigl(x_{2} \log(x_{1})+x_{1}\log(x_{2}) \bigr)-\max \bigl\{ x_{2}\log(x_{1}),x_{1}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [1](#Fig1){ref-type="fig"}. Example 12 {#FPar18} ---------- Let $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$n=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\varphi(2)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(1)=2$\end{document}$, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\psi(2)=1$\end{document}$. Then we have by (I) $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} 2\sqrt{x_{1}^{rx_{2}}x_{2}^{rx_{2}}} \geq x_{1}^{rx_{2}}+x_{2}^{rx_{1}}, \\ h(x_{1},x_{2})=(1/2) (x_{2}-x_{2} ) \log(x_{1})+(1/2) (x_{2}-x_{1} ) \log(x_{2})=(1/2) (x_{2}-x_{1} ) \log(x_{2}), \\ g(x_{1},x_{2})=(1/2) \bigl(x_{2} \log(x_{1})+x_{2}\log(x_{2}) \bigr)-\max \bigl\{ x_{2}\log(x_{1}),x_{1}\log(x_{2}) \bigr\} . \end{gathered} $$\end{document}$$ See Figure [9](#Fig9){ref-type="fig"}. Figure 9**Solution points for inequalities Example ** [**12**](#FPar18){ref-type="sec"} **.** Dedicated to my wife Emília Matejíčková **Competing interests** The author declares that he has no competing interests. **Author's contributions** All authors read and approved the final manuscript. **Publisher's Note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. The work was supported by VEGA grant no. 1/0649/17 and KEGA grant no. 007TnUAD-4/2017. The author thanks Professor Ondrušová, Dean of the faculty FPT TnUAD in Púchov, Professor Vavro and Ing. Balážová for their kind support and he is deeply grateful to the unknown reviewer for valuable remarks and suggestions.

INTRODUCTION {#s1} ============ Natural Killer (NK) cell neoplasms are a rare and heterogeneous group of disorders characterized by excessive proliferation of cytotoxic CD3^−^ CD16/56^+^ NK cells. This group comprises two subtypes: aggressive NK leukemia (ANKL) and chronic lymphoproliferative disorder (CLPD) \[[@R1], [@R2]\]. ANKL is an Epstein Barr virus (EBV)-associated tumor most prevalent among Asian young adults (median age of 42 years). It has a fulminant clinical course, frequently resulting in death within two months. In contrast, CLPD has no demonstrable association with EBV and tends to occur in older adults (median age of 60 years). The clinical course is typically indolent, similarly to T-cell large granular lymphocytes leukemia (T-LGLL) \[[@R3]--[@R5]\]. Since no standard therapies for aggressive NK cell neoplasms have been established so far and the overall outcomes are dismal, new therapeutic options are needed. Fyn, a tyrosine-specific phospho-transferase, is a member of Src family kinases which includes c-Src, Yes, Lck, Lyn, Hck, Fgr and Blk \[[@R6]--[@R8]\]. It phosphorylates a variety of target proteins involved in different signaling pathways \[[@R6]\]. Moreover, it regulates several biological functions, including growth factor and cytokine receptor signaling, cell-cell adhesion, integrin-mediated signaling, ion channel function, platelet activation, T and B-cell receptor signaling, axon guidance, mitosis, differentiation of NK cells \[[@R9]--[@R11]\]. In the last decade, the implication of Fyn in cancer biology and in hematologic malignancies has become more apparent. In chronic myeloid leukemia (CML) Fyn is up-regulated and its activation seems to be important in imatinib resistance \[[@R12], [@R13]\]. It is notably involved in the pathogenesis of peripheral T cell lymphomas \[[@R14]\] and in acute myeloid leukemia its higher expression, combined to FLT3-Internal Tandem Duplication (ITD), is correlated with poor prognosis \[[@R15]\]. The tissue-specific pattern of Fyn mRNA indicates that it is more expressed in normal NK and T cells respect to other human tissues \[[@R16]\]. Tintori *et al.,* by a structure-based drug design protocol and following hit-to-lead optimization, found 4c pyrazolo\[3,4-*d*\]pyrimidine compound driving inhibition of Fyn phosphorylation with a nanomolar range in an enzymatic cell-free assay. Moreover, the compound showed anti-proliferative activities against different cancer cell lines \[[@R17]\]. Therefore, in the present study we investigated the expression of Fyn in NK leukemic cells and the effect of 4c pyrazolo\[3,4-*d*\]pyrimidine compound in NK cell lines and in primary cells from chronic leukemic neoplasms. RESULTS {#s2} ======= Fyn is highly expressed in NK leukemic cells {#s2_1} -------------------------------------------- In order to quantify the presence of Fyn in NK leukemic cells, we firstly assessed its mRNA expression levels in peripheral blood mononuclear cells (PBMCs) from 10 healthy donors (HDs) and 8 CLPD patients by quantitative Real Time-Polymerase Chain Reaction (qRT-PCR). *Fyn* level was significantly up-regulated in PBMCs from CLPD patients compared to HDs (*p* \< 0.001; Figure [1A](#F1){ref-type="fig"}). We also analyzed Fyn protein level by western blotting (WB) in PBMCs of 3 HDs, 3 CLPD patients and in two NK cell lines, KHYG1 and NK92. We observed a high level of Fyn protein in PBMCs from chronic patients and in NK leukemic cell lines respect to PBMCs from HDs (Figure [1B](#F1){ref-type="fig"}). {#F1} 4c pyrazolo\[3,4-*d*\]pyrimidine compound reduced cell viability, induced apoptosis and cell cycle arrest in NK leukemic cells {#s2_2} ------------------------------------------------------------------------------------------------------------------------------ We treated two cell lines (KHYG1 and NK92), 3 PBMCs from HDs and NK cells isolated from 3 HDs with 4c pyrazolo\[3,4-*d*\]pyrimidine compound or with dimethyl sulfoxide (DMSO) vehicle control at different concentrations (2--10 μM) for 24, 48 and 72 hours, cultured with IL2. After 4c treatment, we performed viability test which showed that 4c compound reduced viability of both cell lines in a dose-dependent manner. The effect was observed at 24 hours and remained constant in the other time points (Figure [2A](#F2){ref-type="fig"}). Importantly, both cell lines needed a significantly lower drug concentration (*p* \< 0.01) to reach 50% reduction of viability (EC~50~) (Table [1](#T1){ref-type="table"}). Interestingly, 4c compound had negligible effect in PBMCs and in purified NK cells from HDs (Figure [2A](#F2){ref-type="fig"}). Same results were obtained in primary NK cells from HDs treated with 4c compound and cultured without IL2 ([Supplementary Figure S1](#SD1){ref-type="supplementary-material"}). {#F2} ###### EC~50~ obtained in two cell lines and PBMCs and in purified NK cells from HD samples after 4c compound treatment (*P* value is \< 0.01 for both cell lines vs HDs) (n.c.: not calculable) EC~50~ (μM) ------------- ------------- KHYG1 5.4 NK92 10.6 HDs 59 HD-NK cells n.c. To further investigate cell death mechanism induced after treatment, we performed apoptosis and cell cycle analysis on KHYG1 cell line by cytometric analysis of Annexin V/propidium iodide (PI) and PI, respectively. After treatment with 4c compound at 4 μM for 24 hours, we observed a significant increase of apoptotic cells (*p* \< 0.001) and cell cycle arrest in G2/M phase in treated KHYG1 respect to their control (Figure [2B--2C](#F2){ref-type="fig"}). Fyn phosphorylation is reduced after 4c compound treatment and it decreased Akt and P70 S6 kinase activation {#s2_3} ------------------------------------------------------------------------------------------------------------ To verify Fyn inhibition we performed its immunoprecipitation in KHYG1 cell line treated with 4c compound or with DMSO vehicle control and we detected its phosphorylation. We observed that Fyn phosphorylation significantly decreased after treatment (*p* \< 0.01; Figure [3A](#F3){ref-type="fig"}). {#F3} We also explored, by WB, the activation of two protein involved in Fyn pathway, Akt and P70 S6 kinase. Our data showed that there was a decrease of phosphorylation of Akt and P70 S6 kinase after treatment with 4c compound (*p* \< 0.05; Figure [3B](#F3){ref-type="fig"}). Gene expression and protein profile of treated NK leukemic cells showed the activation of apoptotic pathways {#s2_4} ------------------------------------------------------------------------------------------------------------ We performed gene expression profile (GEP) analysis of KHYG1 cell line after 4 μM drug incubation at 24 hours compared to control. A total of 697 genes (297 up-regulated vs 400 down-regulated genes) were differentially expressed. Ingenuity Pathway Analysis (IPA) gene ontology analysis demonstrated that most of genes were involved in cellular growth, death, development and cell cycle (Table [2](#T2){ref-type="table"}). Moreover, analysis carried out with DAVID let us to cluster all genes in the same functions indicated by IPA. In fact, the first up-regulated cluster genes \[e.g. *BCL2-like 13, Caspase-Related Inducer of Apoptosis* (*CFLAR*), *DNA Fragmentation Factor Alpha* (*DFFA*) and *Fas Ligand*\] included pro-apoptotic functions (Figure [4A](#F4){ref-type="fig"} and [Supplementary Material, Table S1](#SD1){ref-type="supplementary-material"}); conversely, the first down-regulated cluster genes \[e.g. *Survivin*, *Cell Division Cycle 34*, *Aurora Kinase A*\] included pro-survival functions such as cell cycle progression, cell division, mitosis (Figure [4B](#F4){ref-type="fig"} and Supplementary Material, [Table S2](#SD2){ref-type="supplementary-material"}). ###### Molecular and cellular functions of up- and down-regulated genes resulting from IPA gene ontology analysis in GEP analysis of 4c compound vs DMSO treated KHYG1 cells Molecular and Cellular Functions *p*-value range Molecules ----------------------------------- ------------------- ----------- Cellular Growth and Proliferation 7.29E-04/5.16E-15 259 Cellular Function and Maintenance 6.79E-04/5.82E-14 126 Cell Death and Survival 7.81E-04/6.66E-14 239 Cellular Development 5.67E-04/6.26E-13 227 Cell Cycle 7.64E-04/2.48E-11 126 {#F4} We validated GEP data by qRT-PCR confirming the down-regulation of anti-apoptotic gene, *Survivin* (*p* \< 0.05), and the up-regulation of two pro-apoptotic genes, CFLAR (*p* \< 0.05) and *DFFA* (*p* \< 0.001), in KHYG1 treated vs control (Figure [4C](#F4){ref-type="fig"}). Subsequently, to corroborate apoptosis and GEP data, we also performed an array analysis of multiple apoptotic proteins after treatment. As expected, we observed a low expression of anti-apoptotic proteins as cellular Inhibitors of Apoptosis Protein 1 (cIAP1), Claspin, Clusterin, Heat Shock Protein32 (HSP32), Livin, and high expression of apoptotic ones as cleaved caspase 3, TRAIL receptor 2 (TRAIL-R2) and Fas (Figure [4D](#F4){ref-type="fig"}). In order to validate protein array, we showed a significantly increase of cleaved caspase 3 level (***p*** \< 0.05) in 4c compound treated cell line by WB (Figure [4E](#F4){ref-type="fig"}). 4c pyrazolo\[3,4-*d*\]pyrimidine compound induced cytotoxic effect and cell cycle arrest in primary NK leukemic cells {#s2_5} --------------------------------------------------------------------------------------------------------------------- In order to evaluate Fyn inhibitor effects on primary cells, we exposed PBMCs from CLPD patients to 4 μM of 4c compound for 24 hours and we evaluated cell viability by trypan blue count. We observed a decrease of 30% of viable cells in treated PBMCs respect control (Figure [5A](#F5){ref-type="fig"}). Notably, in treated NK cells from leukemic patients there was an increase of active caspase 3/7 level indicating an apoptosis caspase 3-mediated (*p* \< 0.01, Figure [5B](#F5){ref-type="fig"}). {#F5} To better elucidate 4c effects on primary NK cells, we sorted CD56^+^ NK cells from 3 PBMCs of CLPD patients and we exposed them to 4c compound for 24 hours. After incubation time, we evaluated viability by trypan blue count, proliferation by PKH67 labeling, apoptosis and cell cycle by Annexin V/7-Amino-Actinomycin D (7ADD) and PI respectively. Firstly, we observed a decrease of 37% of viable cells (Figure [6A](#F6){ref-type="fig"}). Furthermore, we analyzed proliferation, apoptosis and caspase 3/7 rate both on CD56^+^/CD16^+^ and CD56^+^/CD16^−^ after 4c treatment. We noted a reduction of proliferation in both CD56^+^/CD16^+^ and CD56^+^/CD16^−^cells (Figure [6B](#F6){ref-type="fig"}). Moreover, in both treated populations, we observed a significantly increased apoptosis compared to their respective control (+ 17% in CD56^+^/CD16^+^ cells and + 5.4% in CD56^+^/CD16^−^ cells; *p* \< 0.01 and *p* \< 0.05 respectively) (Figure [6C](#F6){ref-type="fig"}). The 4c effect on apoptosis was also confirmed by increased expression of caspase3/7 levels in both population (+ 2.5% in CD56^+^/CD16^+^ cells and + 8% in CD56^+^/CD16^−^ cells; *p* \< 0.01 and *p* \< 0.05 respectively) (Figure [6D](#F6){ref-type="fig"}). Cell cycle analysis after 4c treatment showed that there was an increase of number of NK primary cells in G0/G1 phase (Figure [6E](#F6){ref-type="fig"}). {#F6} 4c compound treatment induced phenotype changes in NK primary cells {#s2_6} ------------------------------------------------------------------- To evaluate if 4c compound influence NK cell activation, we sorted CD56^+^ NK cells from 3 PBMCs of CLPD patients and we exposed them to 4c compound for 24 hours to study cell phenotype changes. We analyzed the expression of two markers related to NK activation, CD38 and CD25. We observed that there was a decrease of CD38 expression and similar expression of CD25 after treatment respect to control in NK cells (Figure [7A--7B](#F7){ref-type="fig"}). {#F7} DISCUSSION {#s3} ========== NK cell-type large granular lymphocyte leukemia is a rare neoplasm divided in two subtype which could be considered different and independent neoplasms because they have distinct clinical course and management. Chronic lymphoproliferative subtype is characterized by indolent clinical course lacking effective therapeutic approaches \[[@R18]\]; the number of circulating NK cells remains stable for long periods of time and some cases have been reported to show spontaneous regression. On the other hand, aggressive NK leukemia subtype is a fulminant disorder which is not curable. By tissue-specific pattern of mRNA \[[@R16]\], it was identified Fyn like as a gene more expressed in normal NK cells respect to other tissues. It is required to control cytotoxicity and cytokine production in NK cells \[[@R19], [@R20]\], and it has a critical role in NK and T-cell development and activation \[[@R9], [@R21]\]. Since NK leukemia is characterized by an outgrowth of CD3^−^ CD16/56^+^ cells, Fyn could be a possible target in this hematological disease. In fact, analyzing its expression in NK leukemic cells, we found an increased Fyn transcript level in PBMCs from CLPD patients respect to HDs. Interestingly, Fyn protein levels was higher in CLPD patients and in NK cell lines respect to HDs. In these last samples Fyn protein level was not detectable, perhaps for the lower percent of NK cells in HD-PBMCs (\< 10%) respect to CLPD patients (20--40%), to equal protein loaded lysate in western blotting assay. The over-expression of Fyn is documented in different cancers, such as CML, brain, prostate and breast tumors where it facilitates the growth, the migration of tumor cells and genomic instability \[[@R22]--[@R26]\]. Recently, *FYN* mutations were found in peripheral T-cell lymphomas where they impaired DNA damage response and escape from immune surveillance \[[@R14]\]. Our previous data showed that 4c pyrazolo\[3,4-*d*\]pyrimidine compound is able to inhibit Fyn kinase in a cell-free assay and to induce apoptosis and cell cycle arrest in CML cell line \[[@R17]\]. In this paper, for the first time, we investigated the effect of this compound on NK cell leukemia. First of all, we demonstrated that 4c reduced cell viability in NK cell lines. Interestingly, it had no effect on both PBMCs and NK purified cells from HDs. Since it was discovered previously that exposure to stimulatory factors such as the cytokine IL2 enhanced NK cell potency significantly \[[@R27]\], we cultured primary HD-NK cells with and without IL2 in order to evaluate if 4c compound could have diverse effect on NK cells differently activated. We observed that 4c compound had no effect on HD-NK cell viability in both culture conditions. So the presence of IL2 not influenced response to 4c treatment. These results allowed us to conclude that 4c compound acted only on NK leukemic cells and not on healthy NK cells. Moreover, it not influenced viability of other PBMC population. We obtained different percent of viability reduction in cell lines after 4c treatment. More precisely, the major effect was observed in KHYG1 respect to NK92. The diverse response to 4c compound of the two cell lines could be due to their different origin. In particular, KHYG1 was established from peripheral blood of a woman with aggressive NK leukemia at diagnosis, while NK92 was established from a man with non-Hodgkin\'s lymphoma with large granular lymphocytes, more corresponding to chronic subtype. We decided to study in deep the viability reduction in NK aggressive KHYG1 cells and demonstrated that 4c induced a G2/M cell cycle arrest and a significant apoptosis. Recent studies show the possibility to conduct NK malignant cells to apoptosis thus supporting a possible therapeutic option. Quoc Trung L *et al*. showed that resveratrol induced G0/G1 cell cycle arrest and apoptosis in NK leukemic cell lines by inhibiting STAT3 signaling \[[@R28]\]; similar results were obtained by Sun X *et al*. using a combination of drugs, vorinostat and cladribine \[[@R29]\]. To demonstrate that effects on cell viability were due to Fyn inhibition, we confirmed its reduced phosphorylation. These findings are in agreement with Ninio-Many L *et al*. who revealed that miR-125a-3p causes an arrest of cell cycle at G2/M stage and a decreases cell viability in a Fyn-dependent manner \[[@R30]\]. Other studies, in glioma and prostate cells, revealed similar results after Fyn down modulation \[[@R23], [@R31]\]. All this confirms the fact that Fyn is considered an important regulator of cell cycle, growth and proliferation \[[@R32]--[@R34]\]. To the best our knowledge, this is the first study concerning the Fyn targeting in NK malignancies. Moreover, we investigated Fyn downstream pathway. Since some studies have shown that Src kinases, including Fyn, may be required for Akt activation by a tyrosine phosphorylation-dependent mechanism \[[@R35], [@R36]\], we also investigated the Akt behavior after 4c treatment and we observed that Fyn inhibition reduces Akt phosphorylation. Additionally, our data showed a reduction of P70 S6 kinase phosphorylation, an important downstream target of Akt involved in cell survival and in apoptosis inhibition \[[@R37], [@R38]\]. Furthermore, GEP analysis of 4c treated KHYG1 cells confirmed the down-regulation of pro-survival gene pathways and the up-regulation of apoptotic functions. In particular, we validated by qRT-PCR the reduction of *Survivin*, and the increase of *CFLAR* and *DFFA* mRNA. Survivin, a member of the inhibitor of apoptosis protein (IAP) family, regulates leukemic NK cell survival via ERK/MAPK signaling \[[@R39]\]. Interestingly, Liu *et al*. demonstrated that targeting of survivin may be a possible therapeutic approach for NK leukemia because it is highly expressed in both aggressive and chronic leukemic NK cells respect to normal ones \[[@R40]\]. The first up-regulated gene, *CFLAR*, also known as Casper/cFLIP, is a FADD- and Caspase-Related inducer of apoptosis and its over-expression induces apoptotic cell death \[[@R41]--[@R43]\]. It also play a critical role in autophagy, necroptosis and apoptosis in T-lymphocytes \[[@R44]\] and it has prognostic and therapeutic relevance in acute myeloid leukemia (AML) \[[@R45]\] as well as in development of drug resistance \[[@R46]\]. *DFFA* gene is the substrate for caspase 3 and triggers DNA fragmentation during apoptosis \[[@R47], [@R48]\]. In addition, the protein array confirmed the down modulation of survivin, of other members of IAP family, such as cIAP-1, XIAP, livin \[[@R49], [@R50]\] and of cell cycle regulators such as Cluspin \[[@R51]\], and the high modulation of pro-apoptotic ones, such as cleaved caspase 3. Interestingly, dot blot assay showed a decrease of Cytochrome C and Smac/DIABLO and an increase of Fas levels suggesting the activation of extrinsic apoptotic pathway \[[@R52]\]. We also confirmed, by western blotting (WB), the presence of cleaved caspase 3 after treatment letting us to assert that 4c Fyn inhibitor compound is able to induce apoptotic death caspase 3-mediated in NK malignant cells. Finally, we demonstrated that 4c pyrazolo\[3,4-*d*\]pyrimidine compound reduced viability and induced apoptosis in PBMCs and in NK cells from patients with NK leukemia. In human PBMCs the NK cells can be divided on the basis of the expression of CD56 and CD16 markers. CD56^+^/CD16^+^ and CD56^+^/CD16^−^ populations show different proliferation, redox state and function. Specifically, CD56^+^/CD16^−^ cells are responsible for immunoregulatory cytokines production, CD56^+^/CD16^+^ cells are responsible for the cytotoxic lysis of target cells \[[@R53]--[@R54]\]. On the basis of these considerations, we analyzed the 4c effects on both NK purified population. We demonstrated that there was a reduction of viability after treatment but also a reduction of proliferation and an increase of apoptosis caspase dependent in both population CD56^+^/CD16^+^ and CD56^+^/CD16^−^. From cell cycle analysis, 4c compound resulted to be cytostatic, in fact there was an increase of G0/G1 phase. In order to study the 4c effect on NK cell phenotype, we evaluated two important surface molecule involved in NK cells activation, such as CD38 and CD25 \[[@R55]\]. We demonstrated that 4c treatment reduced the expression of CD38, while CD25 expression remained constant. It means that NK treated cells could be slightly active.In summary, the present study demonstrated that Fyn kinase has a role in the pathogenesis of NK leukemia. Moreover, inhibition of Fyn phosphorylation by 4c pyrazolo\[3,4-*d*\]pyrimidine compound induces apoptosis and cell cycle arrest in NK leukemic cells. Fyn targeting by 4c compound could represent a potential and possible strategy for NK leukemia treatment. MATERIALS AND METHODS {#s4} ===================== Patients, cell lines and chemical {#s4_1} --------------------------------- Peripheral blood samples from HDs and NK-CLPD (patient characteristics are shown in Table [3](#T3){ref-type="table"}) were obtained with informed consent. PBMCs were isolated by Ficoll-hypaque gradient separation and NK cells were isolated by cell sorter MoFlo Atrios (Beckman Coulter, Brea, California, USA) by labeling with anti-CD56 phycoerythrin (PE) (Becton Dickinson, BD, Franklin, NJ, USA). Human NK cell lines, KHYG1 and NK92 \[[@R56], [@R57]\], were purchased from Leibniz-Institut DSMZ -- Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH and American Type Culture Collection, respectively. NK primary cells and cell lines were cultured in RPMI 1640 (Gibco, Life technologies, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (FBS, Gibco), 1% of penicillin-streptomycin (Gibco) and 10 ng/ml interleukin-2 (IL-2, Miltenyi Biotec, Auburn, CA) at 37°C and 5% CO~2~. ###### Clinical characteristics of NK-CLPD patients UPN Gender Age NK cells ----- -------- ----- ---------- 1 F 65 \> 60% 2 M 63 \> 60% 3 M 49 40--60% 4 M 49 40--60% 5 M 45 40--60% 6 F 77 30--40% 7 M 45 30--40% 8 M 51 20--30% Abbreviations: UPN, unique patient number; F, female; M, male. 4c pyrazolo\[3,4-*d*\]pyrimidine compound, given by Lead Discovery Siena s.r.l. (patent: WO2016066755), was dissolved in DMSO (Sigma Aldrich, St Louis, MO, USA) and diluted in FBS for cell treatments. RNA extraction and qRT-PCR {#s4_2} -------------------------- Total RNA was extracted with AllPrep DNA/RNA/miRNA Universal Kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer\'s instructions. Reverse transcription was performed using Transcriptor First Strand cDNA Synthesis kit (Roche, Indianapolis, IN, USA). mRNA expression was evaluated by qRT-PCR, performed on Light Cycler 480 II (Roche) with intercalating dye SYBR Green I Master Mix (Roche) using 100 ng cDNA at the following conditions: 95°C for 10 min, 45 cycles at 95°C for 10 sec, 60°C for 10 sec, 72°C for 15 sec. Each sample was analyzed in triplicate. Relative mRNA expression values were normalized using GAPDH as reference gene and Universal Human Reference RNA (Qiagen) as calibrator and calculated on the basis of the E^-ΔΔCp^ method. The primer sequences of human *Fyn*, *GAPDH*, *Survivin*, *CFLAR* and *DFFA* were as follows: Fyn forward 5′-AGATTGCTGACTTCGGATTG-3′, Fyn reverse 5′-CAGACTTGATTGTGAACCTC-3′, GAPDH forward 5′-AGGCTGAGAACGGGAAGC-3′, GAPDH reverse 5′-CCATGGTGGTGAAGACGC-3′, Survivin forward 5′-AGAACTGGCCCTTCTTGGAGG-3′, Survivin reverse 5′-CTTTTTATGTTCCTCTATGGGGTC-3′, CFLAR forward 5′-TGGTAAGGCATGCTTCCAGA-3′, CFLAR reverse 5′-ACAGTATCAGAAGGTGGGGC-3′, DFFA forward 5′-CACTCCAACAGGTGCTTGAC-3′, DFFA reverse 5′-AGTGCAGTAAGGATGTGGCT-3′. Gene expression profile analysis {#s4_3} -------------------------------- Total RNA was quantified with a NanoDrop 2000c spectrophotometer (Thermo Scientific, Wilmington, DE, USA) and its quality was assessed by capillary electrophoresis on an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc, Santa Clara, CA) using RNA 6000 Nano Assay Kit (Agilent). Only samples with RNA integrity number (RIN) \> 7 were used. Samples preparation, hybridization, staining and scanning of Illumina HumanHT12 v4.0 Expression BeadChip array (Illumina Inc., San Diego, CA, USA) on HiScanSQ system (Illumina Inc.) was performed as described before \[[@R58]\]. The intensity files were loaded into the Illumina Genome Studio software for quality control and gene expression analysis. Quantile normalization algorithm was applied on the data set to correct systematic errors, values below a detection score of 0.05 were filtered out and missing values were imputed. Differently expressed genes (DEGs) were selected with differential score (DiffScore) cutoff set at ± 13 (*p* \< 0.05). Microarray data were submitted to Array Express under accession number E-MTAB-4536. DEGs list was used to evaluate the functional behavior in terms of Biological Processes performing an enrichment analysis with IPA (Ingenuity Systems; Mountain View, CA, USA). DEGs list was uploaded also into DAVID bioinformatic tool \[[@R59]\] and functional annotation clustering was evaluated separately for up- and down-regulated genes. Western blotting analysis {#s4_4} ------------------------- Cells were lysed as previously reported \[[@R60]\]. 80 μg was subjected to sodium dodecyl sulfate polyacylamide gel electrophoresis on a 10% gel under reducing conditions and then electrotransferred onto a polyvinylidene difluoride membranes using Trans Blot Turbo Transfer System (BioRad, Hercules, CA, USA). Membranes were probed with primary antibodies directed against Fyn, phospho-Src Family (Tyr416), phospsho-AKT (Ser473), Akt, phospho-P70 S6 kinase (Thr389), P70 S6 kinase, β actin (Cell Signaling, Beverly, MA, USA), active caspase 3 (Abcam, Cambridge, UK) and then incubated with secondary antibody (horseradish peroxidise-conjugated goat anti-mouse or anti-rabbit; Cell Signaling). Immune complexes were detected by ECL chemiluminescence system (Bio-Rad Laboratories), as recommended by the manufacturer. Densiometric analysis was performed using BioRad Image Lab 4.1 software. The intensity of bands of all proteins was normalized to the β actin signal. Immunoprecipitation {#s4_5} ------------------- Protein immunoprecipitation was carried out starting from 1 mg of total protein extracts. Lysates were pre-cleared by incubating with protein A/G-Agarose (Santa Cruz Biotechnology) for 1 hour, subsequently incubated with anti-Fyn antibody for 18 hours and then with fresh A/G-Agarose for 2 hours. All incubation were conducted at 4°C. The product was collected by centrifugation and washed twice. WB analysis was performed as previously reported. Proteome profiler array {#s4_6} ----------------------- Protein extract of KHYG1 cell line, treated with 4c compound at 4 μM or with DMSO vehicle control for 24 hours, were subjected to the human apoptosis array following the manufacturer\'s instructions (Human Apoptosis Array kit, R&D Systems, Abingdon, UK). Cell viability {#s4_7} -------------- KHYG1 and NK92 cell lines, PBMCs and NK cells from HDs were seeded into 96-well plates (3 × 10^4^ cells/100 μl) and incubated with 4c compound at increasing concentrations (2--10 μM) for 24, 48 and 72 hours. Cells treated with DMSO vehicle were used as control. Cell viability was determined using the CellTiter 96 Aqueous One Solution assay kit (MTS, Promega, Madison, WI, USA). The optical density was measured at 492 nm. Cellular viability was calculated as percentage of viable cells compared with control. All experiments were conducted in triplicate. EC~50~ values were obtain by GraphPad Prism (GraphPad Prism, San Diego, CA). Viability of cell lines and primary NK cells from NK-CLPD was assessed by Trypan blue dye count method. Functional tests {#s4_8} ---------------- KHYG1 and NK cells from NK-CLPD samples were treated with 4 μM of 4c compound or with DMSO vehicle control for 24 hours (cell density 3 × 10^5^ cells/ml) and used in: Apoptosis assay {#s4_9} --------------- Apoptosis of KHYG1 was evaluated by cytometric analysis of Annexin V and Propidium Iodide (PI)-stained cells using fluorescein isothiocyanate (FITC) Annexin V Apoptosis Detection kit I (BD). NK cells isolated from NK-CLPD samples were firstly labeled with anti CD56 PE and CD16 allophycocyanin (APC), than labeled with Annexin V and 7-Amino-Actinomycin D (7ADD, BD) to evaluate apoptosis. Stained cells were acquired using FACSCalibur flow cytometer and analyzed by CellQuest Pro software (BD). Single positive for Annexin V and double positive for Annexin V and PI/7ADD cells were interpreted as signs of early and late phases of apoptosis respectively. Cell cycle analysis {#s4_10} ------------------- After treatment KHYG1 cells were fixed in cold ethanol 70% for 1 hour, then labeled with PI/RNase staining solution for 30 minutes. Instead, after treatment, NK cells from NK-CLPD samples were labeled with anti-CD56 PE and anti-CD16 APC, fixed and permeabilzed by Intracell kit (Immunostep, S.L. Avda), then labeled with 7ADD/RNase staining solution for 30 minutes. Samples were acquired by FACSCalibur (BD). Data were analyzed by ModFit LT Software (Verity Software House). Caspase 3/7 activity assay {#s4_11} -------------------------- NK primary cells, treated with 4c compound or with DMSO vehicle for 24 hours (cell density 1 ^×^ 10^6^/ml), were labeled with anti-CD56 PE and anti-CD16 APC, then incubated with 1 μM of CellEvent caspase3/7 Green Detection Reagent (Life technologies, Carlsbad, CA, USA) in complete medium at 37°C for 30 minutes as manufacturer\'s protocol. Stained cells were analyzed by FACSCalibur cytometer (BD). Proliferation assay {#s4_12} ------------------- NK primary cells were stained with PKH67 (PKH67 Green Fluorescent Cell Linker Kit for General Cell Membrane Labeling, Sigma Aldrich) as manufacturer\'s protocol, treated with 4c compound or with DMSO vehicle control for 24 hours and analyzed by FACSCalibur cytometer (BD). Analysis of NK phenotype {#s4_13} ------------------------ NK primary cells, treated with 4c compound or with DMSO vehicle control for 24 hours, were labeled with anti-CD25 FITC, anti CD38-APC (BD) and analyzed by FACSCalibur cytometer (BD). Statistical analysis {#s4_14} -------------------- Statistical significance was determined using a paired or unpaired t test by GraphPad Prism. All error bars represent the standard deviation (SD) of the mean. SUPPLEMENTARY MATERIALS FIGURE AND TABLES {#s5} ========================================= This work was supported by Italian Ministry of Health, Current Research Founds for IRCCS, CUP E66J12000230001. The authors acknowledge Lead Discovery Siena s.r.l. that holds intellectual property of patent \[WO2016066755\]. **CONFLICTS OF INTEREST** The authors declare no competing financial interest.