Known mechanisms cannot account for at least one third of reduced susceptibility in a diverse collection of non-aureus staphylococci

Introduction Non-aureus staphylococci (NAS) are implicated in many healthcare-acquired infections and an understanding of the genetics of antimicrobial resistance in NAS is important in relation to both clinical intervention and the role of NAS as a reservoir of resistance genes. Gap statement The burden of antimicrobial resistance in NAS, particularly to clinically relevant antimicrobials, is under recognised. Methodology We sourced 394 NAS isolates from clinical samples, healthy human volunteers, animals and type cultures and subjected them to agar dilution susceptibility testing against eight antimicrobials. Cefoxitin was used to screen for methicillin resistance in S. aureus, as it stimulates expression of mecA. We performed whole genome sequencing on 366 isolates and analysed these genotypically for the presence of genetic mechanisms responsible for the phenotypic levels of reduced antimicrobial susceptibility. Results We observed 175 sequenced isolates with a minimum inhibitory concentration (MIC) of at least 4 μg/ml to cefoxitin, of which 50% (87/175) did not harbour a known mec homologue. Eight clinical NAS isolates displayed high daptomycin MICs (>4 μg/ml), with no known mechanism identified. Differences in MICs against erythromycin were attributable to the presence of different resistance genes (msrA and ermC). In total, 49% (193 /394) of isolates displayed reduced susceptibility to three or more of the antimicrobials tested. Conclusions The widespread presence of reduced antimicrobial susceptibility in NAS is a concern, with an increased likelihood of (1) harder to treat infections caused directly by NAS, and (2) resistance genes being passed on to other bacteria via horizontal gene transfer, both of which have clinical implications for treatment and management of patients.

Gap statement: The burden of antimicrobial resistance in NAS, particularly to clinically 23 relevant antimicrobials, is under recognised. 24 Methodology: We sourced 394 NAS isolates from clinical samples, healthy human volunteers, 25 animals and type cultures and subjected them to agar dilution susceptibility testing against 26 eight antimicrobials. Cefoxitin was used to screen for methicillin resistance in S. aureus, as it 27 stimulates expression of mecA. We performed whole genome sequencing on 366 isolates and 28 analysed these genotypically for the presence of genetic mechanisms responsible for the 29 phenotypic levels of reduced antimicrobial susceptibility. 30 Results: We observed 175 sequenced isolates with a minimum inhibitory concentration (MIC) 31 of at least 4 µg/ml to cefoxitin, of which 50% (87/175) did not harbour a known mec 32 homologue. Eight clinical NAS isolates displayed high daptomycin MICs (>4 µg/ml), with no 33 known mechanism identified. Differences in MICs against erythromycin were attributable to 34

Introduction 42
The non-aureus staphylococci (NAS) represent an important source of nosocomial disease, 43 including prosthetic joint infection (PJI), infective endocarditis and infection in pre-term 44 babies (1). In the UK, over 215,000 joint replacements (hip, knee and shoulder) took place in 45 2016, with a year-on-year increase of 4% (2). Of these replacements, 1.5% require surgical 46 revision due to infection (2). These infections are most commonly caused by Staphylococcus 47 with the S. aureus control. After inoculation, plates were sealed and incubated at 37 o C at 90 180 rpm for a minimum of 10 hours. The experimental design enabled 13.5% of the collection 91 to be tested in duplicate; MIC data was compared and then tabulated (Table S1). 92 Iso-sensitest agar (Oxoid) was prepared in 250 ml aliquots and autoclaved. Antimicrobial 93 stocks were added to obtain the desired final concentrations once the media had cooled to < 94 50 o C. For daptomycin, Ca 2+ was also added at 50µg/ml. The agar antimicrobial mixture was 95 then poured into sterile rectangular plates (Fisher Scientific) and dried. 96 Per strain, a 1:10 dilution of overnight culture was transferred to a 96 well plate and the OD600 97 was measured. An average OD600 was calculated for each column, which was then diluted to 98 approximately OD600 0.6 to generate an inoculum plate for susceptibility testing. 99 Using a 96-pin multi-point inoculator (Denley), ~1 µl of inoculum per isolate was stamped 100 onto the agar containing antimicrobials, from the lowest concentration to the highest. 101 Between inoculum plates, the pins were washed in 70 % ethanol for 30 s and allowed to dry 102 before stamping on an antimicrobial-free plate to confirm sterility. Washes were also carried 103 out between antimicrobials using sterile water. All stamped plates were incubated at 37 o C. 104 Isolates found to have reduced susceptibility to daptomycin had their MICs determined for a 105 second time by spotting 10 μl of culture onto TSA plates containing various daptomycin 106 concentrations (supplemented with Ca 2+ at 50 µg/ml). To increase mecA expression, 14 107 isolates which contained mecA but on initial testing showed susceptibility to cefoxitin (MIC < 108 4 µg/mL) were re-tested on Muller Hinton Agar with 3 % NaCl added alongside a further 6  To determine which antimicrobial resistance genes and associated individual mutations were 149 present in each of our 364 NAS genomes, reference gene sequences were downloaded from 150 CARD v2.0.0 (11) and used as input to ARIBA v2.13.2 (12) which generates local assemblies 151 from sequence reads and reports back which reference genes (and individual mutations) are 152 identified, with a minimum percent identity cut off at 90% (Table S5) The range of antimicrobials tested were selected based upon clinical relevance and availability 171 (Table S1) Cefoxitin is used to screen for methicillin resistance in S. aureus as it induces mecA expression. 175 However, while methicillin resistant S. aureus (MRSA) have a high public profile, much less is 176 known about methicillin resistant NAS (MRNAS). EUCAST guidelines state that for MRSA 177 "cefoxitin is a very sensitive and specific marker of mecA/mecC-mediated methicillin 178 resistance including in heterogeneous expressing strains and is the agent of choice" (14). In 179 this collection, we found 194/394 (49%) displayed reduced susceptibility to cefoxitin with 180 MICs > 4 µg/ml (Table S1). The vast majority of these isolates were from clinical samples (FOX 181 Figure 1, Table S1) but analysis at the nucleotide level (Table S2 and S3) indicated that only 182 88 out of 175 (50 %) sequenced isolates with an MIC > 4µg/ml harboured a known mecA. 183 (Table S4). Other mec elements were also identified (e.g. mecC, mecI and mecR1) but only 185 ever in addition to mecA. Breaking this down by species, 20/21 S. saprophyticus isolates with 186 cefoxitin MIC > 4µg/ml harboured no mecA (Figure 2). No mecA was detected in eleven 187 species with high cefoxitin MICs and for S. hominis, S. warneri and S. haemolyticus, the 188 percentage of the population that exhibited MIC > 4µg/ml with no mecA was between 28 and 189 67% ( Figure 2). Our results support cefoxitin as a good indicator of mecA presence in S. 190 epidermidis (7), but suggest that it performs poorly in the less common, but still clinically 191 relevant NAS. In addition, we observed 14 cases where the presence of mecA did not result 192 in an MIC > 4µg/ml. These were re-tested (alongside 6 others) under conditions designed to 193 encourage mecA expression and resulted in increased MICs in all isolates, however 7/20 194 remained <4 ug/mL (Table S1). 195 196 Reduced susceptibility only partly explained by known mechanisms 197 When exposed to gentamicin and tetracycline, isolates could broadly be divided into two 198 populations, displaying susceptible or reduced susceptibility phenotypes (Figure 1 GEN and  199 TET). In isolates displaying an MIC ≥1 µg/ml for gentamicin, 49/130 isolates harboured aac(6')-200 le-aph(2")-la (Table 2) which is associated with gentamicin resistance in Enterococcus (15, 16) 201 but has also been observed in Staphylococcus (17, 18). A total of 12 isolates had a match for 202 aph(3')IIIa, but only five of them were associated with reduced susceptibility. 203 Six isolates that contained aac(6')-le-aph(2")-la displayed susceptible MICs, making them the 204 equivalent of major errors (MEs) in public health terms, as the isolates were genotypically 205 resistant but phenotypically susceptible (19). Accordingly, the 81/130 isolates with reduced 206 susceptibility (≥1 µg/ml) that harboured no aac(6')-le-aph(2")-la represented the equivalent 207 of very major errors (VMEs) as they were genotypically susceptible but phenotypically 208 resistant (19). This is highly suggestive of novel mechanisms of resistance and was a feature 209 of other antimicrobials tested (Figure 3). 210 To identify whether efflux pumps might play a role these phenotypes, we assessed the ARIBA 211 output for the staphylococcal-specific norABC, mgrA, mepR and qac genes (20) The distribution of erythromycin phenotypes was more complex. With this antimicrobial, we 225 observed both susceptible isolates and those with reduced susceptibility, but the latter 226 appeared to consist of two populations, one with MICs between ≥2-256 µg/ml and one with 227 MICs ≥512 µg/ml (Figure 1 ERY). We had sequence data available from 135 of the ≥2-256 228 µg/ml population and 66 of the ≥512 µg/ml population, and identified the presence of a 229 resistance gene (ermA, ermC, msrA) in 65.9 % (89/135) of the ≥2-256 µg/ml population and 230 55/66 (83.3 %) of the ≥512 µg/ml population (Table 2). Our results indicated that the presence 231 of ermC rather than msrA was the major cause of MICs exceeding 256 µg/ml. Although rare (n=4), harbouring both genes resulted in an MIC ≥ 512 µg/ml in three cases and 128 µg/ml in 233 the other. In isolates with an MIC of ≥ 512 µg/ml, qac was identified 36 times. In 23 of these 234 cases ermC was also present; qac was only found twice with no other known erythromycin 235 resistance mechanisms present. A total of 101 isolates with an MIC ≥2 µg/ did not contain 236  Testing (EUCAST) and The British Society for Antimicrobial Chemotherapy (BSAC) surveillance 243 data. These MICs were repeated a second time and confirmed. This is concerning given that 244 daptomycin is a current therapeutic choice for treating soft tissue infections caused by NAS 245 (21). Seven of these isolates were sequenced and our ARIBA analysis (Table 3)  liaFRS with an MIC of 1 µg/ml, the remaining 169 isolates with an ≥1 µg/ml MIC did not 248 harbour any of these genes. Several mutations or genes are associated with daptomycin 249 resistance in S. aureus (including mprF, and SNPs in rpoC) but none of these were identified 250 in the NAS collection (6, 22). More recently, SNPs in walK have also been associated with 251 daptomycin resistance in S. aureus and S. epidermidis (23). The three S. aureus SNPs are 252 present in CARD and were not identified in our collection. The V500F mutation from S. 253 epidermidis (23) was also not identified in our S. epidermidis with DAP MICs  4 µg/ml. Whilst 254 walK was identified by protein BLAST as present across the NAS collection (as expected for an 255 essential gene (24)), sequence variation was observed at the protein level which prevents SNPs observed in S. aureus or S. epidermidis being extrapolated to NAS. We therefore 257 conclude that there are potentially novel daptomycin resistance mechanisms present in these 258 strains. 259 260

Resistance to vancomycin found in clinical samples 261
Vancomycin is a treatment option in prosthetic joint infection, and 94 % of isolates had an 262 MIC below 4 µg/ml (Figure 1 VAN). However, of the 24 isolates with reduced susceptibility, 263 22 (92 %) came from clinical samples and only 2/24 were found in healthy volunteers. This is 264 indicative of a wider trend, where isolates associated with clinical samples had significantly 265 higher MICs (p < 0.005) than non-clinical isolates for cefoxitin, erythromycin, gentamicin, 266 tetracycline, daptomycin and vancomycin ( Figure S1). Given the importance of NAS in 267 nosocomial infections, this is a worrying prospect both in terms of what is presenting in the 268 clinic and also the possibility of AMR gene transfer into organisms more capable of causing 269 infection, including S. aureus. In addition, no known mechanisms of resistance were identified 270 for vancomycin, rifampicin or teicoplanin (Figure 3 and Table S3). 271

Over half of the NAS collection displayed susceptibility to multiple antimicrobials 273
Out of the all the isolates tested, 48 % of (192/394) had reduced susceptibility to three or 274 more antimicrobials. Twenty-five isolates had reduced susceptibility to six antimicrobials, and 275 three isolates had reduced susceptibility to seven antimicrobials; of these 24/25 and 3/3 were 276 isolated from clinical samples (Table S1). The implications of these are difficult to treat 277 infections and potentially a large reservoir of staphylococcal resistance genes within the 278 patient under antimicrobial treatment. 279

Animal isolates have similar MIC distributions to human isolates 281
It is generally acknowledged that the presence of reduced susceptibility in microorganisms 282 isolated from animals can impact upon public health if those organisms also cause infection 283 in humans (25, 26). In our collection there were 40 NAS isolated from animals (7 NCTC strains), 284 of which we obtained genome sequences from 23. Although in much fewer numbers than the 285 human isolates in the collection, the animal isolates displayed very similar MIC distributions 286 and harboured corresponding genetic mechanisms. This does not rule out the possibility that 287 animals could be a reservoir of AMR for staphylococci. 288