Evaluation of Surrogate Tests for the Presence of mecA-Mediated Methicillin Resistance in Staphylococcus capitis, Staphylococcus haemolyticus, Staphylococcus hominis, and Staphylococcus warneri

Testing of staphylococci other than Staphylococcus aureus (SOSA) for mecA-mediated resistance is challenging. Isolates of Staphylococcus capitis, Staphylococcus haemolyticus, Staphylococcus hominis, and Staphylococcus warneri were evaluated by cefoxitin and oxacillin broth microdilution (BMD), disk diffusion (DD), and PBP2a immunoassay, and the results were compared to mecA PCR results. ABSTRACT Testing of staphylococci other than Staphylococcus aureus (SOSA) for mecA-mediated resistance is challenging. Isolates of Staphylococcus capitis, Staphylococcus haemolyticus, Staphylococcus hominis, and Staphylococcus warneri were evaluated by cefoxitin and oxacillin broth microdilution (BMD), disk diffusion (DD), and PBP2a immunoassay, and the results were compared to mecA PCR results. No phenotypic susceptibility test correlated well with PCR results across all species, although the PBP2a immunoassay yielded 100% correlation. Oxacillin BMD testing by current Clinical and Laboratory Standards Institute (CLSI) SOSA breakpoints led to 2.1% very major errors (VMEs) and 7.1% major errors (ME). Adjusting this breakpoint up by a dilution (susceptible, ≤0.5 μg/ml; resistant, ≥1.0 μg/ml) led to 2.8% VMEs and 0.3% MEs. Among species evaluated, S. haemolyticus had unacceptable VMEs with this new breakpoint (6.4%), as did S. hominis (4.0%). MEs were acceptable by this new breakpoint, ranging from 0 to 1.2%. Oxacillin DD yielded high ME rates (20.7 to 21.7%) using CLSI or European Committee on Antimicrobial Susceptibility Testing breakpoints. VMEs ranged from 0 to 5.3%. Cefoxitin BMD led to 4.9% VMEs and 1.6% MEs. Cefoxitin DD performed best when interpreted with the CLSI SOSA breakpoint, with 1.0% VMEs and 2.9% MEs. This study led CLSI to adjust the oxacillin MIC breakpoints for SOSA. Laboratories should be aware that no individual phenotypic test correlates well across all species of SOSA with mecA PCR results. Molecular testing for mecA or evaluation for PBP2a is the preferred approach.

schleiferi subsp. coagulans, respectively (2). Historically, identification of SOSA to the species level by morphological and biochemical testing was difficult, lacked resolution, and was often deemed unnecessary as these were commonly viewed as nonpathogenic contaminants (1). In recent years, identification of the members of the SOSA has become more commonplace, through both a better understanding of their role as opportunistic pathogens (1) and the availability of matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) in clinical laboratories (7). The ability to accurately identify SOSA to the species level has improved understanding of the clinical significance of specific species and is accompanied by a need to understand how to best evaluate these isolates for the presence of mecA and ultimately susceptibility to ␤-lactams.
Accurate detection of mecA (or mecC)-mediated resistance to the penicillinase-stable penicillins (oxacillin, nafcillin, cloxacillin, dicloxacillin, and flucloxacillin) is challenging for the staphylococci due to the heterogenous nature of mecA expression across this diverse group of bacteria. In 2000, the Clinical and Laboratory Standards Institute (CLSI [then the National Committee for Clinical Laboratory Standards]) first adopted SOSA oxacillin breakpoints that differed from those used for Staphylococcus aureus. This study was conducted using 50 isolates, weighted heavily for isolates of Staphylococcus epidermidis (n ϭ 27). Eight isolates of Staphylococcus hominis, 6 of Staphylococcus warneri, 3 of Staphylococcus haemolyticus, 2 each of Staphylococcus lugdunensis and Staphylococcus saprophyticus, and a single isolate each of Staphylococcus capitis and Staphylococcus simulans were also evaluated (8). In this study, good correlation (84 to 94%) between oxacillin MIC values of Յ0.25 g/ml and the absence of mecA was found for all isolates. Similarly, excellent correlation (99 to 100%) between oxacillin MICs of Ն0.5 g/ml and the presence of mecA was found for S. epidermidis. However, poor correlation between oxacillin MICs of Ն0.5 g/ml and the presence of mecA was shown for species other than S. epidermidis, ranging from 53.0 to 66.4% correlation (i.e., 33.6 to 47.0% major errors [MEs]). In 2004, CLSI recommended cefoxitin, a cephamycin, be used as a surrogate test for oxacillin as it was found that cefoxitin disk results correlated better with the presence of mecA for SOSA, particularly isolates with MIC values in the 0.5-to 2.0-g/ml range (9).
In 2016, CLSI formed a coagulase-negative Staphylococcus ad hoc working group to evaluate existing oxacillin and cefoxitin testing recommendations to determine their performance as surrogate tests for mecA/C in staphylococci other than S. aureus and S. lugdunensis (referred to herein as SOSA for simplicity). Prioritization of species evaluated include those with known or suspected testing challenges (e.g., Staphylococcus pseudintermedius) and those with a higher prevalence in human infections (e.g., S. epidermidis). The combined work has led to several changes to recommendations in the M100 standard for these isolates (Table 1), including addition of oxacillin disk breakpoints for Staphylococcus pseudintermedius (10), S. schleiferi (11), and S. epidermidis (12)  and recommendation against use of cefoxitin to predict oxacillin susceptibility for S. pseudintermedius and S. schleiferi (10,11). The present study is a continuation of the work performed by CLSI to reevaluate SOSA breakpoints. Isolates of S. capitis, S. haemolyticus, S. hominis, and S. warneri were evaluated by reference cefoxitin and oxacillin broth microdilution (BMD) and disk diffusion to determine the best phenotypic tests to use as surrogates for the presence of mecA. These isolates were chosen due to their high frequency in human bloodstream infections, ranging from 3.5 to 12% for S. capitis, 2.8 to 36.7% for S. hemolyticus, 3.0 to 18% for S. hominis, and 2.0 to 7.8% for S. warneri (1).

MATERIALS AND METHODS
Bacterial isolates. A total of 198 isolates were evaluated in this study ( Table 2). Isolates included S. capitis (n ϭ 50), S. haemolyticus (n ϭ 50), S. hominis (n ϭ 50), and S. warneri (n ϭ 48). Isolates were collected from multiple geographic locations between 2014 and 2018 ( Table 2) and were chosen to represent a roughly even number of mecA-positive and -negative isolates for each species. The isolates were recovered from a variety of clinical cultures, including 126 from blood cultures (64%), 25 from sterile fluids (12.6%), 20 from skin and soft tissue (10.1%), 7 from genitourinary sources (3.5%), 6 from bone (3.0%), 3 respiratory (1.5%), 2 ocular (1.0%), and 3 from other sources (1.5%). Isolates were identified to the species level at each institution according to their standard operating procedures and confirmed at Accelerate Diagnostics, Inc. (Tucson, AZ), by MALDI-TOF MS (Bruker Daltonics, Inc., Billerica, MA) and at Washington University with Vitek MS (bioMérieux, Durham, NC). mecA/C PCR. All isolates were evaluated for mecA and mecC by colony PCR performed from overnight growth on tryptic soy agar with 5% sheep's blood agar plates (BAPs; Hardy Diagnostics, Santa Maria, CA) as previously described (12).
Antimicrobial susceptibility testing. Isolates were stored as described previously (10) and subcultured twice from frozen stocks on BAPs before testing. Disk diffusion (DD) and BMD were performed as described by CLSI (13,14). DD was evaluated on Mueller-Hinton agar (MHA) plates obtained from 3 vendors: Remel (Lenexa, KS), Hardy Diagnostics, and BD (Sparks, MD). Disks containing 1 g oxacillin and 30 g cefoxitin (BBL, BD) were used. BMD was performed by CLSI reference, frozen-form panels containing cation-adjusted Mueller-Hinton broth (CA-MHB) with cefoxitin or oxacillin. Oxacillin BMD tests were supplemented with 2% NaCl. BMD panels were made by Accelerate Diagnostics, Inc. CA-MHB samples from 3 different manufacturers (Difco [BD], BD, and Oxoid [ThermoFisher Scientific, Lenexa KS]) were evaluated on a single panel. Oxacillin and cefoxitin were tested in 2-fold dilutions at concentrations ranging from 0.016 to 32 g/ml.
For testing, 3 to 5 isolated colonies grown overnight on a BAP at 35 to 37°C in ambient air were used to prepare a suspension equivalent to a 0.5 McFarland standard, using a nephelometer, in 0.85% saline. This suspension was used to inoculate all DD and BMD tests, in parallel (13,14). DD tests were incubated at 35 to 37°C in ambient air, and zones of inhibition were measured at 16 to 18 h for oxacillin and 24 h for cefoxitin. BMD tests were incubated at 35 to 37°C in ambient air and read at 16 to 20 h for cefoxitin and 24 h for oxacillin. Colony counts were performed on each inoculum by subculturing the growth control well for each BMD panel. S. aureus ATCC 25923 and S. aureus ATCC 29213 were used as quality controls (QCs) for DD and BMD, respectively, and quality control testing was performed each day of testing.
PBP2a testing. PBP2a testing was performed using the Abbott PBP2a SA Culture Colony test (Abbott Diagnostics, Inc., Scarborough, ME), according to the package insert instructions for testing S. aureus and from the same plate used to make inocula for BMD and DD. In advance, it was agreed that isolates that displayed a different PBP2a result versus mecA PCR result would be repeated in duplicate and that using growth from around a cefoxitin disk and growth from a BAP would be performed, although no isolates required induction. S. aureus ATCC 43300 and S. aureus ATCC 25923 were used as positive and negative controls, respectively. Data analysis. Zone diameters and MIC values were interpreted using the following breakpoints in the CLSI M100, 29th edition, document (Table 3) (3), Belgium (1), Canada (7), Colombia (1) (16). Results were compared to mecA (and mecC, although all isolates were negative) PCR as the "gold standard" for oxacillin resistance. Categorical agreement (CA), very major errors (VMEs), and major errors (MEs) were calculated as previously described (17). VMEs were for isolates that were mecA positive but were oxacillin or cefoxitin susceptible. MEs were defined as those isolates that were mecA negative but were oxacillin or cefoxitin resistant. Disk-to-MIC correlates were evaluated using dBETs software (https://dbets.shinyapps.io/ dBETS/). As each isolate was evaluated on three brands of MHA, data were pooled such that each isolate yielded n ϭ 3 results.

RESULTS
Ninety-six isolates were positive for mecA, and 102 were negative by PCR (Table 2) S. capitis. Clear separation of the mecA-positive and -negative isolates was observed by both oxacillin and cefoxitin BMDs (see Fig. S1A and S2A in the supplemental material). Most oxacillin MIC values for isolates with mecA were Ͼ32 g/ml (89.3%). Only Sigma CA-MHB yielded oxacillin MICs of Ͻ32 g/ml, with 2 isolates with a MIC of 4 g/ml, 1 at 8 g/ml and 1 at 16 g/ml (Fig. S1A). Both the OX MIC SOSA and OX MIC SAU breakpoints yielded 100% CA (Table 4). Most cefoxitin MIC values for isolates with mecA were Ͼ32 g/ml (85.3%; Fig. S2A), whereas mecA-negative isolates in general had a cefoxitin MIC of 2 or 4 g/ml. Overall, no statistical difference was noted between brands of media with regard to VME or ME rates.
DD performed with a 1-g oxacillin disk similarly split the mecA-positive and -negative isolates (see Fig. S3A in the supplemental material). Using the OX DD SOSA   (0)  breakpoint, 100% CA was seen (Table 4). Using the OX DD EUCAST breakpoint, 3/75 results represented MEs (4%; Table 4). These errors were for a single isolate with a zone of inhibition of 18 mm on all three brands of MHA (Fig. S3A). DD performed with the 30-g cefoxitin disk yielded 100% CA using both the FOX DD SOSA and the FOX DD SAU breakpoints (Table 4).

S. haemolyticus.
No test provided good differentiation of mecA-positive from -negative isolates of S. haemolyticus (see Fig. S1B to S4B in the supplemental material). While over half of the oxacillin MIC values for isolates with mecA were Ͼ32 g/ml (60.3%), 19.2% of isolates with mecA yielded oxacillin MICs of 0.25 to 1 g/ml. Isolate WU-09, which harbored mecA, had an oxacillin MIC value of 0.25 g/ml by all three manufacturers of media (Table 4), yielding 3.8% VMEs by the OX MIC SOSA breakpoint. Both mecA PCR and oxacillin tests were repeated for this isolate and yielded the same results. Additionally, a 1.4% ME rate was observed by this breakpoint ( Table 4). The OX MIC SAU breakpoint yielded 19.2% VMEs and 0% MEs (Table 4). Cefoxitin MIC values ranged from Յ0.016 to 4 g/ml for mecA-negative isolates and 0.25 to Ͼ32 g/ml for mecA-positive isolates, with substantial overlap in MICs between mecA-positive and -negative isolates (Fig. S2B). The VME rate was 7.7% by the FOX MIC SAU breakpoint, with no MEs (Table 4).
DD performed with 1-g oxacillin disks resulted in zones of growth inhibition ranging from 20 to 26 mm for mecA-negative isolates and 6 to 20 mm for mecA-positive isolates (Fig. S3). Using the OX DD SOSA breakpoint, 6.4% VMEs and 0% MEs were observed (Table 4). Using the OX EUCAST breakpoint, 2.6% VMEs were seen and there were no MEs. DD performed with the 30-g cefoxitin disk yielded zones of 26 to Ն30 mm and 6 to 28 mm for mecA-negative and -positive isolates, respectively (Fig.  S4B). Again, only VMEs were observed, at 5.1% by the FOX DD SAU and 3.8% by the FOX DD SOSA breakpoints (Table 4).
S. hominis. The mecA-negative isolates yielded oxacillin MICs of 0.03 to 0.5 g/ml, and mecA-positive isolates had MICs of 0.125 to Ͼ32g/ml (Fig. S1C). Three VMEs were observed by both the OX MIC SOSA and OX MIC SAU breakpoints (Table 4), all with a MIC of 0.125 g/ml for the same isolate (CHLA Shom 14). The only MEs were for oxacillin BMD results interpreted by the OX MIC SOSA breakpoint (Table 4). Cefoxitin MICs ranged from 0.25 to 8 g/ml for mecA-negative isolates and 4 to Ͼ32 g/ml for mecA-positive isolates. The VME rate was 8.0% by the FOX MIC SAU breakpoint, with 6.7% MEs (Table 4).
S. warneri. Isolates without mecA yielded oxacillin MICs of 0.125 to 1 g/ml, and isolates with mecA had MICs of 4 to Ͼ32g/ml (Fig. S1D). Only MEs were observed, at 19% by the OX MIC SOSA breakpoint but 0% by the OX MIC SAU breakpoint (Table 4). Cefoxitin MICs ranged from 1 to 4 g/ml for mecA-negative isolates and 4 to Ͼ32 g/ml for mecA-positive isolates (Fig. S2D). The VME rate was 3.4% by the FOX MIC SAU breakpoint, with no MEs ( Table 4). The two VMEs were for two different isolates, WCM1249671 and SW-02, on two different manufacturers' media, suggesting random errors. However, 27% of results for isolates with mecA were at a MIC value of 8 g/ml at the FOX MIC SAU breakpoint.
DD performed with a 1-g oxacillin disk resulted in zones of growth inhibition ranging from 14 to 22 mm for mecA-negative isolates and 6 to 15 mm for mecA-positive isolates. Most results (94.9%) from mecA-positive isolates generated a 6-mm zone of inhibition (i.e., no zone). In contrast, 29.8% of mecA-negative isolates yielded a zone of inhibition of 19 mm, with a normal distribution of zones for all isolates ranging from 14 to 22 mm. Cefoxitin DD results for mecA-positive isolates ranged broadly from 6 to 22 mm, and the majority of mecA-negative isolates had zones of inhibition of 29 mm (38%) or Ն30 mm (47.6%) (Fig. S4D). Using the OX DD SOSA breakpoint, 0% VMEs and 79.8% MEs were seen (Table 4). Using the OX EUCAST breakpoint, 0% VMEs and 70.2% MEs were seen. CA improved with the cefoxitin disk, with 100% CA for the FOX DD SOSA breakpoint and only 1 VME (1.7%) by the FX DD SAU breakpoint.
Evaluation of results for all species. No single MIC breakpoint performed well across the species of SOSA evaluated (Table 4). When evaluated against CLSI criteria of Ͻ3% MEs and Ͻ1.5% VMEs (18), only the FOX DD SOSA breakpoint had an acceptable VME rate, whereas the OX MIC SAU, FOX MIC SAU, FOX DD SOSA, and FOX DD SAU breakpoints had acceptable ME rates. While some error rates were near acceptance limits for an individual species, each method had at least one species for which performance was particularly poor: OX MIC SOSA with 19% MEs with S. warneri, OX MIC SAU with 19% VMEs for S. haemolyticus, OX DD SOSA with 80% MEs for S. warneri, FOX MIC SAU with 8% VMEs for S. hominis, FOX DD SOSA with 12% MEs for S. hominis, and FOX DD SAU with 5.1% VMEs for S. haemolyticus (Table 4). For each species, however, one or more methods could be considered acceptable (Table 4).
Currently, the CLSI recommends confirmation of oxacillin MICs in the 0.5-to 2.0-g/ml range (resistant by the OX MIC SOSA breakpoints) by PBP2a or mecA detection methods. Historically, use of cefoxitin DD was also considered acceptable. We evaluated isolates with oxacillin VMEs or MEs for which the MIC values that generated the errors were in this range to determine if cefoxitin DD could resolve the error. For S. hominis, all 5 MEs were noted at a MIC of 0.5 g/ml, and all cefoxitin disk zones for these isolates yielded zones of 22 to 23 mm (resistant by FOX DD SOSA but susceptible by FOX DD SAU). For S. warneri isolates with MEs, oxacillin MIC values were 0.5 g/ml (n ϭ 15) and 1 g/ml (n ϭ 1). For the 7 isolates that encompassed these 15 MEs, all disk results obtained were susceptible by the FOX SOSA DD breakpoint. Finally, 3 VMEs and 1 ME were due to S. haemolyticus isolates with oxacillin MIC values in the 0.5-to 2.0-g/ml range. None of the 3 VMEs were resolved by cefoxitin DD, but the 1 ME was. As such, the strategy historically proposed by CLSI performed well for resolving errors for S. warneri, but not the other species evaluated herein.
New proposed breakpoint for oxacillin MIC and evaluation of disk correlates. The majority (21/22 [95.5%]) of OX SOSA MIC MEs in this study occurred due to an oxacillin MIC of 0.5 g/ml (Fig. 1). VMEs for the OX SOSA MIC breakpoint were due to oxacillin MICs of 0.125 g/ml (n ϭ 3) and 0.25 g/ml (n ϭ 3) (Fig. 1). Adjusting the OX SOSA susceptible MIC breakpoint from Յ0.25 g/ml to Յ0.5 g/ml resulted in an overall reduction in MEs from 7.2% to 0.3% and an increase in VMEs from 2.1% to 2.8%, due to two additional VMEs for the same S. haemolyticus isolate ( Fig. 1 and Table 4). The two species at this new breakpoint with unacceptable error rates were S. hominis, with 4.0% VMEs, and S. haemolyticus, with 6.4% VMEs. All three VMEs for S. hominis were for an isolate with an oxacillin MIC of 0.125 g/ml by all three CA-MHB brands evaluated. This isolate also yielded 1 VME by FOX MIC SAU (not shown). For S. haemolyticus, two isolates yielded VMEs-one of these isolates yielded a VME by all breakpoints evaluated in this study but gave a positive PBP2a result (data not shown).
In order to determine if changing the oxacillin breakpoint would impact other species, data from prior studies were reevaluated, including evaluation of S. epidermidis (n ϭ 100 isolates tested across three brands of media [note some isolates did not grow on some media, yielding 291 results]), S. pseudintermedius (n ϭ 111 isolates), and S. schleiferi (n ϭ 90 isolates). This resulted in 0% VMEs and only 3% MEs ( Table 5). The ME was for an isolate of S. epidermidis with an oxacillin MIC value of 4 g/ml but which was mecA negative. The overall MIC distribution is shown in Fig. 1. Combined, this new breakpoint resulted in 1.6% VMEs (8/490 mecA-positive results) and 0.9% MEs (5/562 mecA-negative results). In contrast, the 2019 CLSI OX MIC SOSA breakpoint yielded 1.2% VMEs (6/490) and 4.5% MEs (25/562). The biggest change was for S. warneri, where the ME rate went from 21.3% to 1.3% with this change to the breakpoint.
Disk diffusion breakpoints are traditionally set by correlating zones of growth inhibition to MICs, and this was attempted to see if an improved disk breakpoint could be established with the new oxacillin MIC breakpoint. Disk correlates were evaluated using the dBETS program, which indicated for cefoxitin a susceptible disk breakpoint of Ն25 mm was preferred, with a rate of 0.2% VMEs and 2.4% MEs. This is the same disk diffusion breakpoint currently endorsed by CLSI. Similarly, a susceptible breakpoint of Ն18 mm was calculated for the oxacillin disk, which was associated with a 0.5% VME rate and a 3.0% ME rate compared to the oxacillin MIC. This too is the same disk breakpoint currently endorsed by CLSI. Manual evaluation of the disk results compared to mecA PCR (as opposed to the oxacillin MIC, which is used by dBETs) demonstrated overlap between cefoxitin zones for mecA-positive and mecAnegative isolates of S. haemolyticus (at 27 and 28 mm; Fig. 2) and S. hominis (at 22 and 23 mm; Fig. 2). Overlap of oxacillin zones for mecA-positive and mecA-negative  (Table 5). Oxacillin MICs were determined in CA-MHB plus 2% NaCl. mecA-negative isolates are presented in panel A, and mecA-positive isolates are presented in panel B. The lower end of the MIC distribution for S. pseudintermedius was 0.25 g/ml, and the upper end was 32 g/ml. All other species were tested across the entire oxacillin concentration range shown. The vertical blue line represents the 2020 CLSI oxacillin breakpoint. isolates occurred for S. haemolyticus (at 20 mm), S. warneri (at 15 mm), and S. hominis (at 20 and 22 mm) ( Fig. 3; Fig. S4). Quality control results. All results were within QC ranges, as published in the CLSI M100 standard, 29th edition (15). S. aureus ATCC 25923 cefoxitin DD results ranged from 23 to 28 mm (mode, 25 mm). Oxacillin DDs ranged from 18 to 21 mm (mode,  mecC, and so performance of phenotypic tests for isolates with that resistance factor was not evaluated. It should be noted that the incidence of mecC among SOSA is exceedingly low, with occasional reports of individual isolates of Staphylococcus caprae (22), Staphylococcus xylosus (6,23), Staphylococcus saprophyticus (24,25), Staphylococcus stepanovicii, Staphylococcus edaphicus (3), and S. warneri (22) harboring the gene. Several isolates of Staphylococcus sciuri have been reported to harbor mecC (22). Most laboratories perform MIC susceptibility testing using automated systems, and the breakpoints proposed herein may not correlate with mecA results for these systems. However, strengths of the present study include evaluation of a diverse collection of contemporary isolates from multiple geographic regions.
In conclusion, laboratories should consider performing mecA PCR or a PBP2a test if a penicillinase-stable penicillin, such as oxacillin, is being considered for therapy for serious infections due to SOSA.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.1 MB.