This is the accepted version of the journal article: Sánchez, A.; Benito, Natividad; Rivera, Alba; [et al.]. «Pathogenesis of Staphylo- coccus epidermidis in prosthetic joint infections : can identification of virulence genes differentiate between infecting and commensal strains?». The journal of hospital infection, Vol. 105, Issue 3 (July 2020), p. 561-568. 8 pàg. DOI 10.1016/j.jhin.2020.04.026 This version is available at https://ddd.uab.cat/record/285280 under the terms of the license 1 Pathogenesis of Staphylococcus epidermidis in prosthetic joint infections: Can the 2 identification of virulence genes differentiate between invasive and commensal 3 strains? 4 5 Alba Sánchez a, Natividad de Benito b, Alba Rivera a, Lucas García a, Elisenda Miró a, 6 Isabel Mur b, Yesica Gonzalez a, Cristina Gutierrez a, Pere Colla, Paula Espinal a*†and 7 Ferran Navarro a,c*† 8 a 9 Department of Microbiology, Hospital de la Santa Creu i Sant Pau, Institut 10 d'Investigació Biomèdica Sant Pau (IIB Sant Pau), Barcelona, Spain. b 11 Department of Infectious Diseases, Hospital de la Santa Creu i Sant Pau, Institut 12 d'Investigació Biomèdica Sant Pau (IIB Sant Pau), Barcelona, Spain. c 13 Genetics and Microbiology Department. Universitat Autònoma de Barcelona, 14 Barcelona, Spain. 15 † 16 Contributed equally as senior authors 17 *Corresponding address: Dr. Ferran Navarro and Dr. Paula Espinal, Servei de 18 Microbiologia, Hospital de la Santa Creu i Sant Pau, Institut d'Investigació Biomèdica 19 Sant Pau, Barcelona, Spain., C/Sant Quintí, 89, 08041 Barcelona, Spain; Phone: 20 +34935537297; E-mails: fnavarror@santpau.cat and pespinal@santpau.cat 21 22 Running title: Pathogenesis of Staphylococcus epidermidis in prosthetic joint 23 infections 24 Keywords: Coagulase negative Staphylococci, biofilm, virulence, prosthetic joint 25 infections 26 ABSTRACT 27 Staphylococcus epidermidis is a commensal of cutaneous flora that has emerged as one 28 of the most frequent causative microorganism of prosthetic joint infections (PJIs). The 29 aim of this study was to elucidate possible antimicrobial resistance and virulence 30 markers involved in the pathogenicity of S. epidermidis analyzing the possible 31 differences between infecting and commensal strains. Fifty S. epidermidis strains from 32 PJI patients, 50 from skin of healthy-individuals (HI) and 17 from the surgery-field 33 during a primary arthroplasty (SF) were analyzed in regard to antimicrobial resistance, 34 detection of resistance genes, PFGE, MLST, biofilm formation and detection of 35 virulence genes The PJI strains were significantly more often resistant to antibiotics 36 than HI and SF strains. PFGE defined 56 PFGE types and none was shared between PJI 37 and SF groups. The ST2 was the predominant sequence type and was only present in 38 PJI strains (44%), while HI and SF showed a large variety of STs. All strains were 39 biofilm producers and significant differences were not detected. The genes sdrF, bhp, 40 icaA, icaB, icaD, the complete ica operon and the IS256 were significantly predominant 41 in PJI strains, whereas, embp, hld, ACME prevailed in HI strains. The SF strains 42 behaved similarly to HI strains. In conclusion, there was no single marker to 43 differentiate invasive from commensal S. epidermidis strains, but a combination of them 44 were more common in strains from the PJI group than from the other groups including 45 ica operon, IS256, sdrF, bhp and mecA, higher antimicrobial resistance and a 46 predominance of the ST2. 47 48 49 50 51 INTRODUCTION 52 Prosthetic joint infections (PJIs) are considered serious complications since they are 53 long-lasting and difficult to treat infections with high economic costs that can generate 54 irreversible physical sequelae. (Becker, Heilmann, and Peters 2014b; Parvizi, Fassihi, 55 and Enayatollahi 2016; Runner et al. 2019; Tsai et al. 2015). 56 The majority of PJI cases were caused by coagulase negative staphylococci (39.6%), 57 with S. epidermidis as the more common etiological agent (23.3%) (Becker, Heilmann, 58 and Peters 2014a; Benito et al. 2016) . 59 Staphylococcus epidermidis, is an important commensal specie in the nasal and 60 cutaneous human microbiota, and it has a great clinical impact as an opportunistic 61 pathogen implicated in device related infections. During the last years, S. epidermidis 62 has been considered one of the principal causes of PJIs (Le, Park, and Otto 2018) after 63 S. aureus (Post V, Harris LG, Morgenstern M, Mageiros L, Hitchings MD, Méric G, 64 Pascoe B, Sheppard SK, Richards RG 2017). However, it is not always easy to establish 65 if S. epidermidis isolated in periprosthetic clinical specimens represent true infection or 66 only colonization/contamination. (Karsten Becker, Christine Heilmann, Georg Peters. 67 Coagulase-Negative Staphylococci. Clin Microbiol Rev 2014; 27: 870 –926) 68 The production of a biofilm on surfaces of implanted materials is recognized as the 69 main virulence factor in S. epidermidis (Bengt Hellmark, Soderquist, et al. 2013). 70 During the biofilm formation a set of virulence genes (VG) are involved from the 71 attachment to the surface that they colonize until the aggregation and the subsequent 72 cellular detachment (Büttner et al. 2015). Some genes such as, ica operon, IS256, bhp, 73 aap and embp among others, have been described as possible molecular markers that 74 might differentiate invasive from commensal S. epidermidis strains (Gu et al. 2005; 75 Heilmann, Ziebuhr, and Becker 2018; B. Hellmark et al. 2013; Ortega-Peña et al. 2019). 76 However, to date, no single marker has permitted distinguish both populations. The aim 77 of this study was to elucidate possible genetic markers involved in the pathogenicity of 78 S. epidermidis through molecular characterization, including epidemiological, resistance 79 and virulence features of three populations of S. epidermidis: strains of PJIs, strains of 80 skin or mucous membranes of healthy individuals and strains isolated from the surgery 81 field during a primary arthroplasty (that did not be eventually involved in PJIs). 82 MATHERIALS AND METHODS 83 Bacterial strains 84 A total of 117 S. epidermidis strains from three populations were studied: 50 strains 85 recovered from the collection of all microorganisms causing PJIs during 2013-2019 at 86 the Microbiology Department, Hospital de la Santa Creu i Sant Pau, 50 strains isolated 87 from nasal and skin swabs of healthy individuals (HI) not related with a clinical 88 environment during 2017-2018, and 17 strains isolated from the surgery field (SF), 89 during a primary arthroplasty performed during 2016-2019 in the Hospital de la Santa 90 Creu i Sant Pau and the Hospital del Mar without causing infection (after at least one 91 year follow-up) (Table 1). 92 The participants of the HI group did not consume antibiotics and were not hospitalized 93 in the three months prior to sample collection; whereas patients from the SF group had 94 received as surgical antimicrobial prophylaxis one dose of cefazolin orcefazolin plus 95 gentamycin, (in penicillin-allergic patients, vancomycin or vancomycin plus 96 gentamicin) before obtaining the samples Most patients with IPA did not received 97 antibiotics prior to collect the diagnostic periprosthetic samples (by synovial fluid 98 aspiration or during the surgical treatment). 99 One isolate per patient was included in the study except when the strains differed in the 100 antibiogram. 101 Strains were isolated and identified according to the routine microbiological techniques 102 and confirmed by MALDI TOF MS (Bruker). 103 The study was approved by the Hospital de la Santa Creu i Sant Pau ethics committee 104 (IIBSP-STA-2016-81). 105 Antimicrobial susceptibility testing and resistance genes 106 All strains underwent antimicrobial susceptibility testing using the Sensititre ESTEN2F 107 (ThermoFisher Scientific, Inc.). Results were interpreted according with the Clinical 108 and Laboratory Standards Institute (CLSI 2017)(Clinical and Laboratory Standards 109 Institute (CLSI) n.d.), except for fosfomycin and tigecycline that were interpreted 110 according with the European Committee on Antimicrobial Susceptibility Testing 111 (EUCAST 2017) 112 Total DNA of all strains was extracted using the GenElute-Bacterial Genomic DNA Kit 113 (Sigma-Aldrich). Resistant strains were PCR-screened for a set of antimicrobial 114 resistance genes (AGRs): to oxacillin (mecA), erythromycin (ermA, ermB, ermC, msrA, 115 mphC), clindamycin (lnuA), cotrimoxazole (dfrS), linezolid (cfr, rRNA 23S, L3, L4, 116 L22), aminoglycosides (aacA-aphD and aadD) (Table S1). PCR products were purified 117 using EXOSAP (Isogen Life Science) and sequenced following Macrogen Inc 118 guidelines. Sequences were analysed by BioNumerics v.7.6 (Applied Maths NV) and 119 Basic Local Alignment Search Tool (BLAST). Staphylococcal cassette chromosome 120 mec (SCCmec) typing was performed in all methicillin-resistant S. epidermidis (MRSE) 121 strains by multiplex PCRs (M-PCR1 and M-PCR2) (Kondo et al. 2007) and 122 http://www.sccmec.org (Table S1). 123 Molecular typing of the strains 124 Pulsed-Field Gel Electrophoresis (PFGE) methodology was applied to the analysis of 125 strains from the PJI and SF groups. Strains from the HI group were excluded because 126 the divergence of them (not clinically-epidemiologically related). SmaI enzyme (Sigma- 127 Aldrich) was used for the genomic restriction of the strains following the procedure 128 described (Goering and Fey 2014). Further analysis of the PFGE procedure was 129 obtained with the Bionumerics software v.7.6 using the Dice similarity coefficient and 130 the construction of a dendrogram with the UPGMA algorithm (Unweighted Pair Group 131 Method using Arithmetic averages). The interpretation criteria was based on the 132 previously established (Tenover et al. 1995) (M. Miragaia et al. 2008). 133 Multilocus sequence typing (MLST) was performed in all strains according to the 134 published scheme of Thomas et al. (Thomas et al. 2007). Sequence types (STs) were 135 assigned according to the S. epidermidis MLST data base (http://sepidermidis.mlst.net). 136 Clonal complex was assigned by eBURST (http://eburst.mlst.net) and published reports 137 (Feil EJ, Li BC, Aanensen DM, Hanage WP 2004; M. Miragaia et al. 2007, 2008). 138 The minimum spanning tree (MST) was constructed with the STs of all strains using 139 Bionumerics software v.7.6. 140 Biofilm formation assay 141 S. epidermidis strains were growth in Tryptic Soy Broth supplemented with 0.5% 142 glucose (TSBG, Oxoid). The bacterial suspensions adjusted to OD600 of 1.5 were diluted 143 1:100 with TSBG and inoculated (200 μL) in 96-well microtiter plates for 24 h at 37 ºC. 144 Biofilm quantification was performed by 1% Crystal Violet (CV) stain following 145 (Seyedi-Marghaki et al. 2019). The biofilm formation was calculated by the median 146 absorbance of three replicates of the experiment. Strains were classified as strong (>2), 147 moderate (1-2), weak (0.1-1) and non-biofilm producer (<0.1). The S. epidermidis 148 RP62A (ATCC 35984) and S. epidermidis ATCC 12228 strains were used as biofilm- 149 forming and non-biofilm-forming controls, respectively. 150 Detection of virulence genes 151 Presence of virulence genes (VG) involved in the biofilm formation was determined by 152 PCR of the following genes: atlE, sdrG, sdrF, sesI, embp, ica operon (icaA, icaB, icaC, 153 icaD, icaR), bhp, aap and hld. In addition, the two quorum sensing systems, agr and 154 luxS were analysed (Table S1). 155 The agr type was determined by the analysis of the deduced amino acid sequence 156 (Bengt Hellmark, Söderquist, et al. 2013). 157 Additionally, the mobile elements IS256 and ACME were studied by PCR (Table S1). 158 The ACME type was classified according with the arc, opp3 and kdp genes presence: 159 ACME-I (arc and opp3), ACME-II (arc), ACME-III (opp3), ACME-IV (arc and kdp), 160 ACME-V (arc, opp3 and kdp) and ACME-VI (kdp) (Maria Miragaia et al. 2009) 161 (O’Connor et al., 2018). 162 Data analysis 163 The categorical variables were compared using Chi-square test (χ2 test) or Fisher´s 164 exact test and the ordinal variables by Mann-Whitney or Kruskal-Wallis. All 165 comparisons were performed with IBM SPSS Statistics v25.. A p <0.05 was considered 166 statistically significant. 167 Results 168 Antimicrobial susceptibility testing and antimicrobial resistance genes 169 A total of 117 S. epidermidis strains were analysed (Table 1). The resistance rates to 170 most of the antimicrobials tested were found significantly higher in PJI strains than in 171 HI and SF strains (Table 2). All strains remained susceptible to teicoplanin, vancomycin, 172 daptomycin and tigecycline. Fourty seven percent (55/117) of the strains were MRSE 173 and were predominant in PJI strains (37/50; 74%) versus SF (9/17; 53%) and HI (9/50; 174 18%) (p<0.05) (Table 2A). The mecA gene was detected in 53/55 (96.4%) of MRSE 175 strains. Among erythromicin resistant strains (n=74), mphC and msrA genes and its 176 combination were the predominant ones followed by the ermC and the combination of 177 ermC+mphC+msrA genes (Table 2B). Gentamicin resistance was only present in PJI 178 strains except for two SF strains. This resistance was mainly mediated by aadD (Table 179 2B). The same mutations were observed for both linezolid resistant strains: G2603T 180 mutation in the 23S rRNA gene and Val154Leu amino acid alteration in L3. 181 Clonality of the strains 182 The PFGE analysis of 67 strains from PJI (n=50) and SF (n=17) showed a great 183 diversity and were distributed into 56 PFGE types (Figure 1): 49 single-patterns (32 184 from PJI and 17 from SF), six PFGE types including two strains (4, 7, 26, 27, 31 and 185 42) and the PFGE type 24 that contained six strains recovered from patients P5, P6 and 186 P7. PFGE types 31 and 42 harboured a pair of strains from patients P10 and P13, 187 respectively (each pair of strains differed in the isolation date and/or antibiogram). A 188 clinical epidemiological relationship suggesting cross-transmission between patients of 189 PFGE types 4, 7, 24, 26 and 42 was not found. PFGE types were not shared between PJI 190 and SF strains. 191 The MLST analysis of all strains determined a total of 66 different STs. The sequence 192 type ST2 was the predominant in PJI strains (22/50; 44%) followed by the new ST640 193 and ST5 (6/50; 12% both). In HI and SF strains, a great variability of STs was observed, 194 without a predominant sequence type. Twenty four new STs were described in this 195 study (Figure 2 and S1). 196 The minimum spanning tree (MST) showed that ST35 and ST657 were shared between 197 the three populations; ST5 and ST57 were shared between PJI and HI strains, and, 198 ST17, ST32, ST88, ST89 and ST795 between HI and SF strains (Figure 2). Most of the 199 strains (67; 57.3%) belonged to the clonal complex 2 (CC2). The subgroup CC2-I was 200 predominant in PJI strains (p<0.05) (Table S2). 201 A SCCmec type was not possible to assign for any MRSE strain, showing a great 202 variability of combinations between ccr and mec complexes. The most frequent 203 combinations were: ccr1-ccr2-mec class B (19/55; 34.5%), ccr1-ccr2-ccr5-mec class B 204 and ccr3-ccr4-ccr5-mec class A and B (both 2; 3.6%), all only present in PJI strains. 205 Biofilm formation and virulence genes 206 All strains were biofilm producers and significant differences were not detected 207 between the three populations (Table 3). The majority of PJI and SF strains were 208 classified as strong or moderate biofilm producers (66% and 64.7%, respectively); 209 whereas, contrary to the expected, this percentage was even higher (76%) in HI strains. 210 Different virulence genes content was observed between all strains: sdrF, bhp, the 211 complete ica operon, icaA, icaB, icaD and the IS256 were significantly predominant in 212 PJI strains, whereas, embp, hld and ACME were more prevalent in HI strains. The 213 quorum sensing luxS and agr genes were present in all strains; however, agr-IV and 214 ACME-VI were statistically higher in PJI; agr-III, ACME-IV and ACME-V in HI; and 215 ACME-II in SF strains (Table 3). 216 Comparison of biofilm formation and the presence of VGs determined that icaA 217 positive strains were not associated with strong biofilm formation; only, sdrF, embp and 218 hld genes showed relation to strong biofilm production in HI strains (data not shown). 219 From gentamicin resistant strains, 82.1% (32/39) were positive for IS256 (p<0.001) and 220 all of them belonged to the PJI group. Resistance to gentamicin were not detected in HI 221 and only in two SF strains. 222 Finally, repetitive virulence gene content was observed in strains with specific STs, 223 among them: all ST2 strains belong to PJI and was positive for sesI, sdrF, ica operon, 224 aap, hld, IS256 and agr-I; ST35, present in the three populations, contained embp, sesI, 225 ica operon, aap, hld, ACME and agr-I in all strains. In addition, some STs were 226 associated with agr types, such as, ST89, ST35 and ST153 to agr-I; ST32 and ST59 to 227 agr-II; and ST640 to agr-IV (Figure S2). 228 DISCUSSION 229 Because S. epidermidis is a predominant microorganism in the human skin microbiome, 230 several studies have focused on the search for genetic markers that can differentiate 231 commensal from invasive S. epidermidis strains in PJIs (B. Hellmark et al. 2013; Bengt 232 Hellmark, Soderquist, et al. 2013; Méric et al. 2018; Ortega-Peña et al. 2019). Herein, 233 we present a comparative study of S. epidermidis strains of patients with PJIs, healthy 234 individuals and, as a novelty, a group of strains recovered in the surgery field during a 235 primary arthroplasty that were not eventually involved in PJIs. We consider that the 236 latter is a true control because, in addition to being commensal and not causing 237 infection, these strains are present in a place that could be a route of entry for invasive 238 strains. 239 Strains from PJI were significantly resistant to more antibiotics than HI and SF (Table 240 2), as has been already described (Soraya Cherifi et al. 2014; Ortega-Peña et al. 2019; 241 Salih et al. 2018). (S. Cherifi et al. 2013; B. Hellmark et al. 2013; Rolo, de Lencastre, 242 and Miragaia 2012). Despite SF strains have undergone antibiotic pressure due to 243 prophylaxis before the arthroplasty, they did not show elevate percentage of resistance 244 than PJI strains. 245 Results determined the highly significance of MRSE (73.9%) in PJI strains than in 246 commensals. In addition, MRSE rates remained similar to reported in other studies 247 (Ortega-Peña et al. 2019; Salih et al. 2018)(B. Hellmark et al. 2013; Ortega-Peña et al. 248 2019; Rolo, de Lencastre, and Miragaia 2012). It highlights MRSE is an important 249 feature to distinguish invasive from commensal strains (B. Hellmark et al. 2013). 250 The PFGE analysis showed heterogeneity of S. epidermidis strains (PJI and SF) that 251 could be explained by the increment of horizontal gene transfer and mobile genetic 252 elements during the adaptation to both, community and hospital environments (S. 253 Cherifi et al. 2013; M. Miragaia et al. 2007). Although six PFGE types were found 254 containing closely related strains, it was not possible to define cross transmission among 255 them. Five of them belonged to ST2 and shared some VGs: ica operon, IS256, aap, 256 sdrF, sesI and hld. PFGE type 27 belonged to ST640 and, in contrast, it was ica operon 257 negative, therefore, biofilm formation seems to be mediated by other VGs such as, bhp, 258 sdrF and embp. 259 Moreover, the great diversity of STs found, showed that commensal strains were more 260 divergent than the invasive strains (Figure S1), similar to that found by Du et al. (Du et 261 al. 2013). ST2 was the most frequent and only occurred in PJI strains; as is well 262 known, it has been reported that ST2 in the most widespread hospital-associated ST 263 worldwide (Soraya Cherifi et al. 2014; Du et al. 2013; M. Miragaia et al. 2007; Ortega- 264 Peña et al. 2019; Salgueiro et al. 2017). 265 ST2 strains besides to being present in PJIs were ica operon, IS256, mecA positive and 266 multirresistant. The combination of these pathogenicity determinants contributes to the 267 adaptation and survival of pathogenic clones, such as ST2, to specific physiological 268 conditions during infection, the antibiotic barrier and the host defenses(Christensen and 269 Brüggemann 2014; Du et al. 2013; Post et al. 2017, Meric 2018). 270 Instead, other STs such as ST5, ST35, ST57 and ST657 were shared between strains of 271 different groups (Figure 2), which can be considered true opportunistic pathogens, 272 where strains have the same ability to cause infection (meric 2018). 273 In this study, significant differences were not detected in biofilm production and the 274 presence of icaA, among invasive and commensal strains; similar to other studies 275 (Bengt Hellmark, Soderquist, et al. 2013; Méric et al. 2018; Salgueiro et al. n.d.). 276 Although icaA positive strains have been related to strong biofilm producers (Soraya 277 Cherifi et al. 2014), it was not observed in our study. Thus, from our point of view, 278 biofilm formation is not a feature to discriminate between commensal and pathogenic 279 strains. 280 Finally, embp, hld and ACME were statistically higher in HI. It could be explain due to 281 the importance of these markers for bacterial survival and increase in the ability of S. 282 epidermidis to colonize and spread in different environments (B. Hellmark et al. 2013; 283 Bengt Hellmark, Soderquist, et al. 2013). Although agr-I has been previously associated 284 with invasive strains (B. Hellmark et al. 2013) we did not observe this predominance. 285 The limitations of our study are those implicit to the sample size, mainly in SF and the 286 absence of commensal strains from the skin of PJI patients, but the main strength is that 287 includes a new group of strains to be compared. Until this study, clinical strains were 288 always compared with those isolated from healthy skin, but never compared to strains 289 that could reach the surgical field, but had not caused an infection. Development of the 290 infection depends on the strain, the inoculum of microorganisms that have reached the 291 surgical field, the immune system of the patient and the prophylactic treatment. 292 Despite the low number of strains, our findings suggest that SF behaved similarly to HI 293 strains. This new group of strains allowed it to be defined as a true control to 294 discriminate between the commensal and PJI strains. 295 In conclusion, there was no single marker to differentiate invasive from commensal 296 strains, but the combination of some markers allow to establish differences between 297 populations: PJIs were caused by S. epidermidis strains that belonged more often to the 298 ST2, were positive for ica operon, IS256, sdrF, bhp and mecA, whereas the presence of 299 embp, hld and ACME was found significant in commensal strains. The development of 300 rapid tests in the routine laboratory, such as a multiple PCR, which includes these 301 pathogenicity determinants, could allow rapid differentiation of invasive and 302 commensal strains into PJIs.