Optimized In Vitro Antibiotic Susceptibility Testing Method for Small-Colony Variant Staphylococcus aureus

Abstract

Staphylococcus aureus small-colony variants (SCVs) emerge frequently during chronic infections and are often associated with worse disease outcomes. There are no standardized methods for SCV antibiotic susceptibility testing (AST) due to poor growth and reversion to normal-colony (NC) phenotypes on standard media. We sought to identify reproducible methods for AST of S. aureus SCVs and to determine whether SCV susceptibilities can be predicted on the basis of treatment history, SCV biochemical type (auxotrophy), or the susceptibilities of isogenic NC coisolates. We tested the growth and stability of SCV isolates on 11 agar media, selecting for AST 2 media that yielded optimal SCV growth and the lowest rates of reversion to NC phenotypes. We then performed disk diffusion AST on 86 S. aureus SCVs and 28 isogenic NCs and Etest for a subset of 26 SCVs and 24 isogenic NCs. Growth and reversion were optimal on brain heart infusion agar and Mueller-Hinton agar supplemented with compounds for which most clinical SCVs are auxotrophic: hemin, menadione, and thymidine. SCVs were typically nonsusceptible to either trimethoprim-sulfamethoxazole or aminoglycosides, in accordance with the auxotrophy type. In contrast, SCVs were variably nonsusceptible to fluoroquinolones, macrolides, lincosamides, fusidic acid, and rifampin; mecA-positive SCVs were invariably resistant to cefoxitin. All isolates (both SCVs and NCs) were susceptible to quinupristin-dalfopristin, vancomycin, minocycline, linezolid, chloramphenicol, and tigecycline. Analysis of SCV auxotrophy type, isogenic NC antibiograms, and antibiotic treatment history had limited utility in predicting SCV susceptibilities. With clinical correlation, this AST method and these results may prove useful in directing treatment for SCV infections.

INTRODUCTION

Staphylococcus aureus causes diverse infections ranging in severity from benign to life-threatening. Persistent and relapsing S. aureus infections often occur despite prolonged antimicrobial therapy and have been linked to the emergence of small-colony variants (SCVs). S. aureus SCVs have been associated with a poor response to antibiotic treatment (1, 2) and are often recovered from many chronic infections, including endocarditis (3, 4), osteomyelitis (5), device infections (3), soft tissue infections (6), and airway infections (7, 8). SCV airway infections are particularly prevalent among people with the genetic disease cystic fibrosis (CF), affecting between 8 and 33% of CF patients, often following prolonged antibiotic treatment (1, 7). In children with CF, SCV respiratory infection was found to be independently associated with worse lung function and faster lung function decline (9).

SCVs are typically identified by distinctive phenotypic traits when grown on most agar-based media. SCVs produce colonies approximately 1/10 the size of normal-colony (NC) S. aureus isolates (Fig. 1), and they are characteristically nonhemolytic and nonpigmented and have diminished coagulase production (10). SCVs usually carry mutations in one of a few, specific metabolic pathways; however, even SCVs that share similar colony morphologies on agar surfaces can have different metabolic defects. The pathways most commonly affected are important for electron transport or DNA biosynthesis, resulting in auxotrophy for specific nutrients. Accordingly, supplementation of the growth media with menadione and hemin, cofactors in menaquinone and cytochrome synthesis, respectively, complements electron transport-deficient SCVs unable to synthesize these compounds (11). Thymidine supplementation complements SCVs harboring mutations in thyA, which encodes thymidylate synthase (12). SCVs have been isolated both in vivo and in vitro after antibiotic exposure (7, 13, 14). For example, trimethoprim-sulfamethoxazole (SXT) usage has been associated with the recovery of SXT-resistant, thymidine-dependent SCVs from CF airway secretions (15), while hemin- and menadione-dependent SCVs can emerge following aminoglycoside exposure (14, 16).

FIG 1.

Distinguishing phenotypic features of SCVs. In vitro growth comparison of an NC isolate (left) and a genetically related SCV (right) on blood agar showing characteristics that commonly distinguish SCVs, including reduced hemolysis, a small colony morphology, and decreased pigmentation.

The impaired in vitro growth exhibited by SCVs presents two important challenges to the clinical microbiology laboratory: SCVs are difficult to detect using conventional approaches, and there are no approved methods for SCV antimicrobial susceptibility testing (AST), hindering the selection of appropriate treatments. AST was developed for rapidly growing, aerobic bacteria analyzed within 24 h of collection using a standard medium (Mueller-Hinton medium for S. aureus); however, the growth of SCVs under these conditions is generally insufficient to provide meaningful results. As such, a variety of different methods using diverse media (see Table S1 in the supplemental material) and extended incubation have been utilized for AST of SCVs. These conditions were often chosen empirically for the specific SCVs being characterized. To further complicate testing, SCVs frequently revert to NC phenotypes during in vitro growth, potentially altering susceptibilities and limiting interpretation of the results. The primary goal of this study was to identify in vitro conditions that would adequately support growth, minimize reversion to NCs, and allow measurement of the in vitro susceptibilities of a diverse collection of S. aureus SCV clinical isolates. We also sought to determine whether the in vitro susceptibilities of SCV isolates determined by this method could be reliably predicted from (i) SCV auxotrophy type and methicillin resistance, (ii) the susceptibilities of clonally related NC isolates from the same patients, or (iii) the source patient's antibiotic treatment history either as a complement or as an alternative to SCV AST. Our results indicate that this AST method, if validated, could be useful in directing antibiotic treatment for S. aureus SCV infections.

MATERIALS AND METHODS

S. aureus strains and isolates.

The S. aureus isolates tested in this study comprised 85 SCVs and, for each SCV type defined by pulsed-field gel electrophoresis (PFGE), at least 1 previously isolated or coisolated NC isolate (n = 27) (see Table S2 in the supplemental material). All clinical isolates were obtained from a single-center study of children (n = 23) with CF performed from 2008 to 2010; details regarding the source population were described previously (9). Since a menadione-dependent clinical SCV was not available, a menadione-dependent SCV selected in vitro from S. aureus Newman (16) was used. S. aureus ATCC 25923 and ATCC 29213 were used as controls in AST, and USA300 methicillin-resistant S. aureus (MRSA) strain JE2 (17) served as a positive control for mecA PCR and β-lactam resistance in AST. This study was approved by the Seattle Children's Hospital Human Subjects Institutional Review Board.

Media.

The growth media used included Luria-Bertani agar (LB) (Becton Dickinson [BD], Sparks, MD), Trypticase soy broth and agar (TSB and TSA, respectively; BD), TSA with 5% sheep's blood (blood agar plate [BAP]; Remel, Lenexa, KS), mannitol salt agar (BD), Mueller-Hinton agar (MHA; Oxoid, Basingstoke, Hampshire, United Kingdom), brain heart infusion (BHI; BD [BD-BHI]) agar, and chocolate agar (Remel). LB and BD-BHI agars were prepared in-house using their respective broth powders and 1.5% (wt/vol) Bacto agar (BD). Commercially available BHI agar was also obtained from Hardy Diagnostics (Santa Maria, CA) and Teknova (Hollister, CA). A synthetic CF sputum agar medium (SCFM), formulated to mimic the nutritional composition of CF sputum, was prepared as described previously (18) with the addition of 0.1 mg/ml herring sperm DNA (SCFMD; Sigma).

Some media (TSA, BHI agar, MHA, and SCFMD) were supplemented (indicated by the prefix “sup-”) with the following SCV growth-supportive nutrients: 5 μg/ml of thymidine, 1 μg/ml of hemin, and 1 μg/ml of menadione. Final concentrations were chosen empirically by evaluating the growth of thymidine-, hemin-, and menadione-dependent SCVs on MHA with increasing supplement concentrations (see Fig. S1 in the supplemental material). SCVs from freezer stocks (TSB with 15% glycerol) were inoculated onto agar media. The supplemented media contained each supplement at the lowest concentration that both (i) significantly improved the in vitro growth of SCVs auxotrophic for that supplement and (ii) did not inhibit the growth of NC S. aureus and SCVs of other auxotrophic types (see Fig. S1 in the supplemental material). Previous studies indicated that 5 μg/ml thymidine would approximate the sum of the concentrations of thymidine and its biologically active metabolite, dTMP, in CF sputum (19, 20). sup-MHA was also custom ordered from Teknova and prepared using the same protocol used to prepare our in-house sup-MHA, referred to here as THM-MHA (thymidine-hemin-menadione Mueller-Hinton agar).

Evaluation of SCV growth.

The growth of a subset of SCVs (n = 10, including 1 menadione-dependent SCV, 2 hemin-dependent SCVs, and 7 thymidine-dependent SCVs) and paired, isogenic NCs (n = 10) was assessed. SCVs and paired, isogenic NCs were cultured directly from frozen stocks (TSB with 15% glycerol) onto 11 different agar media: LB, TSA, BAP, BHI (from BD, Hardy, and Teknova), mannitol, MHA, SCFMD, sup-TSA, sup-BD-BHI, sup-MHA, and sup-SCFMD. Plates were incubated for 24 h at 35°C under both aerobic and microaerobic (5% CO₂) conditions. SCV colonies similar in size and appearance to NCs (diameter, ≥1 mm) when grown on the same media were considered to have adequate growth. SCV colony diameters of ≤1 mm indicated poor growth.

Evaluation of SCV phenotypic stability.

SCV phenotypic stability on SCV growth-supportive media was evaluated in two ways: (i) assigning a stability score (specific to each growth medium) determined by visual inspection of colony morphology variability and (ii) measuring the frequency of reversion of SCVs to stable NCs. Specifically, SCVs from frozen stocks were plated onto agar media that adequately supported SCV growth (sup-MHA, BD-BHI agar, sup-BD-BHI agar, TSA, sup-TSA) and incubated at 35°C for 24 h, and the colonies were inspected for consistency. Stability scores reflected the number of colony types observed on each medium: a score of 1 indicated a single colony type, a score of 2 indicated two colony types, etc. Media with one colony type (sup-MHA, BD-BHI agar, sup-BD-BHI agar) for each SCV were chosen for subsequent experiments. The frequency of reversion was determined by subculturing 150 randomly chosen colonies from the test medium onto LB agar, which identifies SCVs (see Fig. S2 in the supplemental material); incubating at 35°C for 24 h; and measuring the colony size. The frequency of reversion was calculated by dividing the number of colonies that grew ≥1 mm in diameter (NC size) on LB agar by 150, the total number of colonies.

Antimicrobial susceptibility testing.

Disk diffusion was conducted on all SCVs and paired, isogenic NCs using 24 antibiotics (see Table 1 for the list of the 22 tested antibiotics for which there are established disk diffusion breakpoints; we also tested fosfomycin and vancomycin, for which there are no established disk diffusion breakpoints) according to Clinical and Laboratory Standards Institute (CLSI) guidelines (21), except that BD-BHI agar and sup-MHA were substituted for MHA. S. aureus ATCC 25923 and NC isolates were also tested on MHA, BD-BHI agar, and sup-MHA to compare the effects of the media and supplementation on AST. For a subset of isolates (26 SCVs, 13 corresponding NCs, and ATCC 25923), disk diffusion was also performed under microaerobic conditions (5% CO₂). AST results were interpreted according to CLSI guidelines (21, 22) for all antibiotics except fusidic acid, for which EUCAST guidelines were used (23). CLSI guidelines formulated in 2009 (22) were used to interpret susceptibilities to β-lactams, since these breakpoints are absent from the 2014 guidelines.

TABLE 1.

Susceptibility profiles of NC and SCV clinical isolates determined by disk diffusion^b

Antibiotic(s)c	% susceptible
	NC MSSAa (n = 19), sup-MHA and BHI	TD-SCV MSSAa (n = 48), sup-MHA and BHI	HD/MD-SCV MSSA (n = 3)		ND-SCV MSSAa (n = 1), sup-MHA and BHI	NC MRSA (n = 9)		TD-SCV MRSAa (n = 31), sup-MHA and BHI	HD-SCV MRSAa (n = 3), sup-MHA and BHI
			sup-MHA	BHI		sup-MHA	BHI
Cefoxitin	100.0	100.0	100.0	100.0	100.0	0.0	0.0	3.2	0.0
Oxacillin	100.0	100.0	100.0	100.0	100.0	22.2	33.3	32.3	33.3
Piperacillin-tazobactam	100.0	100.0	100.0	100.0	100.0	NT	NT	NT	NT
Amoxicillin-clavulanate	100.0	100.0	100.0	100.0	100.0	NT	NT	NT	NT
Cefazolin	100.0	100.0	100.0	100.0	100.0	NT	NT	NT	NT
Meropenem	100.0	100.0	100.0	100.0	100.0	NT	NT	NT	NT
Ticarcillin-clavulanate	100.0	100.0	100.0	100.0	100.0	NT	NT	NT	NT
Levofloxacin	89.5	83.3	100.0	100.0	100.0	22.2	22.2	9.7	66.7
Moxifloxacin	89.5	83.3	100.0	100.0	100.0	22.2	22.2	9.7	66.7
Tobramycin	68.4	70.8	66.7	0.0	100.0	0.0	0.0	0.0	0.0
Amikacin	73.7	68.8	33.3	0.0	100.0	0.0	0.0	3.2	0.0
Trimethoprim-sulfamethoxazole	100.0	0.0	66.7	66.7	0.0	100.0	100.0	0.0	100.0
Minocycline	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Chloramphenicol	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Azithromycin	63.2	37.5	100.0	100.0	100.0	11.1	11.1	0.0	0.0
Clindamycin	94.7	85.4	100.0	100.0	100.0	66.7	66.7	45.2	66.7
Quinupristin-dalfopristin	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Telithromycin	94.7	83.3	100.0	100.0	100.0	66.7	66.7	45.2	66.7
Linezolid	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Rifampin	94.7	87.5	100.0	100.0	100.0	44.4	44.4	12.9	66.7
Tigecycline	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Fusidic acid	100.0	100.0	100.0	100.0	100.0	88.9	88.9	93.5	100.0

^aThese isolates displayed identical percent susceptibilities on both sup-MHA and BD-BHI agar (simplified to BHI in table); therefore, percent susceptibilities represent those for both media and the data are presented in a single column.

^bDiscrepancies between media for the same group of isolates are indicated by bold and underlined type and are discussed in the text. TD-SCV, thymidine-dependent SCV; HD/MD-SCV, hemin- and menadione-dependent SCV; ND-SCV, SCV with a nondefined auxotrophy; HD-SCV, hemin-dependent SCV; NT, the activities of the β-lactam drugs against the corresponding MRSA isolates were not tested.

^cFosfomycin activity and vancomycin activity were also tested against the full set of isolates using disk diffusion; however, because there are no established disk diffusion breakpoints for these antibiotics, they are not listed here. Results are discussed in the text.

MICs were determined by Etest for 13 antibiotics (see Table 2 for the list of antibiotics) using a subset of isolates (26 SCVs and 24 corresponding NCs) and ATCC 29213 on THM-MHA (Teknova). In parallel, ATCC 29213 MICs were determined on standard MHA from Teknova to compare the effects of supplementation on susceptibility.

TABLE 2.

Susceptibility profiles of NC and SCV clinical isolates determined by Etest^a

Antibiotic(s)	% susceptible
	NC MSSA (n = 17)	TD-SCV MSSA (n = 15)	HD-SCV MSSA (n = 2)	MD-SCV MSSA (n = 1)	NC MRSA (n = 7)	TD-SCV MRSA (n = 6)	HD-SCV MRSA (n = 2)
Clindamycin	94.1	80.0	100.0	100.0	57.1	33.3	50.0
Trimethoprim-sulfamethoxazole	100.0	0.0	50.0	100.0	100.0	0.0	100.0
Levofloxacin	88.2	80.0	100.0	100.0	14.3	0.0	50.0
Chloramphenicol	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Linezolid	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Vancomycin	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Minocycline	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Quinupristin-dalfopristin	100.0	100.0	100.0	100.0	100.0	100.0	100.0
Cephalothin	100.0	100.0	100.0	100.0	NT	NT	NT
Amoxicillin-clavulanate	100.0	100.0	100.0	100.0	NT	NT	NT
Imipenem	100.0	100.0	100.0	100.0	NT	NT	NT
Piperacillin-tazobactam	100.0	100.0	100.0	100.0	NT	NT	NT
Tobramycin	64.7	73.3	50.0	100.0	0.0	0.0	0.0

^aAll testing was completed on THM-MHA manufactured by Teknova. TD-SCV, thymidine-dependent SCV; HD-SCV, hemin-dependent SCV; MD-SCV, menadione-dependent SCV; NT, the activities of the β-lactam drugs against the corresponding MRSA isolates were not tested.

SCVs, paired isogenic NC isolates, and ATCC control strains were subcultured onto chocolate agar plates from frozen stocks and grown aerobically at 35°C overnight. Chocolate agar was validated to support both SCV growth and stability in preliminary, control experiments (not shown). Colonies of individual isolates were suspended in 0.85% saline to a 0.5 McFarland standard (∼1 × 10⁸ CFU/ml), and samples of this suspension were swabbed onto the surfaces of either sup-MHA, BD-BHI agar, THM-MHA, or MHA plates (MHA was used only for NC and control strains). Antibiotic disks or Etest strips were added as indicated below. Plates were incubated aerobically or microaerobically (with 5% CO₂) at 35°C for 18 to 24 h. All antibiotic disks except the fusidic acid disk were obtained from BD; the fusidic acid disk was prepared with powder (Sigma, St. Louis, MO) and filter disks (BD) at 10 μg/disk. Etests were obtained from bioMérieux, Durham, NC.

Antimicrobial susceptibility testing with thymidine in AST media.

Because sup-MHA contains antagonists of trimethoprim-sulfamethoxazole (SXT) activity (e.g., thymidine), CLSI recommends that slight growth (≤20% of lawn growth) around an SXT disk or strip, if present, should be ignored and measurements should be made at the margin of the heaviest growth (21). SXT-susceptible, control organisms (ATCC 25923 and ATCC 29213) displayed slight growth around SXT disks/strips on sup-MHA and BD-BHI agar, the latter of which presumably contains thymidine, and so we applied the ≤20% guideline to both media (see Fig. S3 in the supplemental material).

MRSA detection.

MRSA was identified using a PBP 2′ latex agglutination kit (Oxoid), with confirmation being performed by mecA gene detection using PCR as described previously (24, 25).

PFGE.

The genetic relatedness of clinical isolates was determined by PFGE as described previously (26). Briefly, chromosomal DNA was digested with 50 U of SmaI (Roche) at room temperature for 3 h, and DNA restriction fragments were separated in a 1% SeaKem Gold agarose gel using a CHEF DR-III apparatus (Bio-Rad, Hercules, CA). Dendrograms were constructed by the unweighted pair group method with arithmetic mean (UPGMA) clustering method.

Correlation of susceptibilities with antibiotic exposure of source patients.

The relationship between antibiotic exposure during the 90 days prior to isolation of an SCV and susceptibility to that antibiotic was determined using antibiotic treatment data from the source study (9). Antibiotic exposure was identified if the end date for treatment with that antibiotic occurred within 90 days prior to the date that the sample for culture was obtained. Antibiotic end dates were missing for 15 of 407 antibiotic courses identified; in those cases, end dates were assigned a value equal to the start date plus 28 days for inhaled tobramycin, the start date plus 180 days for maintenance oral azithromycin, or the start date plus 14 days for the other antibiotics tested. Logistic regression models accounted for multiple isolates per subject and used robust variance estimates. The odds ratio (OR) with the 95% confidence interval (CI) was calculated as the measure of association between the predictor (antibiotic exposure within 90 days prior to the date that the sample for culture was obtained) and the outcome (the isolate was not susceptible to the antibiotic tested). Models could not be fit when the antibiotic exposure perfectly predicted the isolate's status as susceptible or not susceptible (these observations were dropped from the model, leaving no variability in the predictor variable). This analysis excluded 3 replicate SCVs from the same culture as another SCV and 11 isolates for which antibiotic treatment data were incomplete, leaving 71 isolates from 21 subjects as the analyzed data set.

RESULTS

SCV growth and phenotypic stability on various media.

All SCVs grew poorly (menadione-dependent and hemin-dependent SCVs) or not at all (thymidine-dependent SCVs) on MHA, the CLSI-recommended AST medium (21) (Fig. 2). By comparison, all SCVs grew similarly to isogenic NC isolates on 5 media: sup-TSA, TSA, BD-BHI agar, sup-BD-BHI agar, and sup-MHA. Medium supplementation (denoted “sup-”) indicates the addition of SCV-growth supportive nutrients: thymidine, hemin, and menadione (see Fig. S1 in the supplemental material). Surprisingly, thymidine-dependent SCVs did not grow on multiple lots each of BHI agar from two other companies (Hardy and Teknova), indicating manufacturing differences. The medium formulated to mimic CF sputum (SCFMD) did not support the growth of all SCVs, even when it was supplemented with thymidine, hemin, and menadione (Fig. 2). SCV growth on the tested media was comparable under aerobic and microaerobic atmospheric conditions (not shown).

FIG 2.

Quality of SCV growth determined by visual inspection of colony size on agar-based media. Ten SCVs were characterized by colony size as an indicator of adequate growth on various media. *, SCV and NC colony growth was tested at a later date on THM-MHA (which is chemically identical to sup-MHA but which is manufactured by Teknova) and Teknova-manufactured unsupplemented MHA (Teknova-MHA) to confirm that isolates grew comparably on commercially available and in-house-prepared supplemented and unsupplemented MHA. Abbreviations: LB, Luria-Bertani agar; BAP, Trypticase soy agar plus 5% sheep blood; MHA, Mueller-Hinton agar; TSA, Trypticase soy agar; BHI, brain heart infusion agar; SCFMD, synthetic CF sputum agar medium supplemented with herring sperm DNA; sup-, supplementation of base media (MHA, TSA, BHI agar, SCFMD) with SCV growth-supportive nutrients thymidine, hemin, and menadione.

Improved SCV growth may be due to either the presence of growth-supportive nutrients or genetic reversion. In the first case, the in vitro susceptibilities of SCVs on a medium would be measured under chemically complementary conditions. In the latter case, the results would reflect the susceptibilities of in vitro-selected NC revertants rather than those of SCVs. SCVs grew with substantial morphological variability (stability score, >1) on TSA and sup-TSA, indicative of genetic instability and reversion, while the other media tested (e.g., sup-MHA) supported uniform growth (Fig. 3A). To determine whether uniform colony sizes indicated SCV stability, we compared the reversion frequencies for each SCV on the media that supported uniform growth and on blood agar (BAP), a medium previously used for SCV AST (7, 19). The reversion frequencies were the lowest on sup-MHA under both aerobic and microaerobic conditions (Fig. 3B). BAP yielded the highest reversion rates under both atmospheric conditions, and on this medium, thymidine-dependent SCVs exhibited a distinctive, anucleate cellular morphology upon staining with Gram stain (27), indicating poor growth (not shown). sup-MHA and BD-BHI agar were therefore chosen for AST, as they best supported growth and maintained SCV phenotypic stability. While SCVs grow to sizes approximately the same as NC sizes on these media, these isolates retain SCV phenotypes upon subculture to LB agar, and for simplicity, we refer to these isolates as SCVs even when they are cultivated on growth-complementing media.

FIG 3.

Morphological variability and quantification of reversion frequency as measures of SCV stability. (A) Thymidine-dependent SCVs grown on TSA (left), sup-TSA (middle), and sup-MHA (right) displayed variable colony morphologies on TSA and sup-TSA, indicative of reversion. However, on sup-MHA, a single uniform colony morphology was observed (the morphology of a representative thymidine-dependent SCV is shown). All SCVs tested showed comparable growth on sup-MHA, sup-BHI, and BHI (of these, growth on only sup-MHA is shown in the right panel for simplicity). (B) SCV isolates were grown overnight on the media listed on the x axis, after which colonies selected at random were subcultured onto LB agar (discriminatory medium) to distinguish between reversion to NC growth and complementation by the chemical content of the medium. The percentage of colonies exhibiting reversion is shown on the y axis. *, commercially prepared sup-MHA (THM-MHA) was validated at a later date.

AST. (i) Aerobic disk diffusion.

To determine how closely the disk diffusion zone diameters on our test media would overlap the acceptable range limits defined by CLSI, we tested the susceptibility of control strain ATCC 25923 on MHA (recommended by CLSI), sup-MHA, and BD-BHI agar. The average and standard deviation zone diameters for ATCC 25923 fell within CLSI-accepted range limits for each antibiotic on all media, with a few exceptions on BD-BHI agar (Fig. 4). The susceptibilities of a subset of 14 NC clinical isolates (8 methicillin-susceptible S. aureus [MSSA] isolates and 6 MRSA isolates) were also identical on MHA, sup-MHA, and BD-BHI agar (not shown). Therefore, supplementation of MHA with thymidine, hemin, and menadione did not significantly alter the antibiotic susceptibilities of the control isolates, suggesting that this medium is reliable for S. aureus AST. Although the results for the control strain fell outside acceptable limits for several antibiotics on BD-BHI agar, susceptibility testing was conducted with both BD-BHI agar and sup-MHA for the complete set of isolates for comparison.

FIG 4.

Average zone diameters for strain ATCC 25923 in air. The average zone size for control strain ATCC 25923 fell within CLSI-accepted ranges (horizontal red dashed lines) for MHA (black circles) and sup-MHA (light gray squares) for all antibiotics tested and for the majority of antibiotics tested on BD-BHI agar (dark gray triangles). Error bars depict the standard deviations for 15 replicates. The average zone diameter for tigecycline and the standard deviations for levofloxacin, cefoxitin, and oxacillin fell outside the CLSI-defined acceptable range limits on BD-BHI agar. Fusidic acid does not have CLSI-defined range limits; therefore, no red lines are shown. Although for ATCC 25923 a range limit has not been reported for fusidic acid, average zone diameters and standard deviations for this antibiotic were comparable on all three media and well above the EUCAST breakpoint (≥24 mm).

AST was performed on 112 clinical isolates (including SCVs and isogenic NCs), a menadione-dependent SCV selected in vitro from S. aureus strain Newman (16), and S. aureus Newman (see Table S2 in the supplemental material). The clinical isolates belonged to 15 distinct PFGE groups and included both MSSA (n = 71) and MRSA (n = 43) isolates. The MSSA isolates belonged to 9 different PFGE groups, and the MRSA isolates comprised 6 distinct PFGE groups, with just over half of the MRSA isolates (n = 25) belonging to the USA100 pulsotype (see Fig. S4 in the supplemental material). A single isogenic MRSA NC/SCV pair belonged to the USA700 pulsotype.

All isolates grew on BD-BHI agar and sup-MHA, and zone diameters were measurable after 18 to 24 h. No discrepancies in AST results were observed between BD-BHI agar and sup-MHA for thymidine-dependent SCVs (MSSA and MRSA). All thymidine-dependent SCVs were resistant to trimethoprim-sulfamethoxazole, while NC isolates were susceptible (Table 1). Collectively, thymidine-dependent SCVs tended to be nonsusceptible to azithromycin. Small percentages (∼12 to 30%) of thymidine-dependent MSSA SCVs were nonsusceptible to fluoroquinolones, aminoglycosides, clindamycin, telithromycin, and rifampin. Fewer thymidine-dependent MRSA SCVs than thymidine-dependent MSSA SCVs were susceptible to these antibiotics (Table 1).

In contrast to the thymidine-dependent SCVs, hemin- and menadione-dependent MSSA SCVs exhibited different susceptibilities on BD-BHI agar and sup-MHA. While these SCVs were generally susceptible to most antibiotics, they were all nonsusceptible to the aminoglycosides tobramycin and amikacin on BD-BHI agar (Table 1). However, these SCVs exhibited slightly larger zone diameters (by 2 to 8 mm) on sup-MHA, which, in some cases, rendered them aminoglycoside susceptible. Hemin-dependent MRSA SCVs were uniformly nonsusceptible to both aminoglycosides (Table 1). An SCV of undetermined auxotrophy was resistant only to trimethoprim-sulfamethoxazole (Table 1).

MRSA control strain JE2 and all S. aureus isolates were screened for methicillin resistance using oxacillin and cefoxitin disks. JE2 was resistant to both antibiotics, with the zone diameters on MHA, BD-BHI agar, and sup-MHA being comparable (±1 mm). Cefoxitin resistance was detected for all NC mecA-positive isolates on these media. Oxacillin resistance was detected in 67% and 77% of these isolates on BD-BHI agar and sup-MHA, respectively. Similarly, 97% of mecA-positive thymidine-dependent SCVs were resistant to cefoxitin and 67.7% were resistant to oxacillin on both media. A single mecA-positive, thymidine-dependent SCV was susceptible to both antibiotics on BD-BHI agar, sup-MHA, and BAP. Similar results were observed for hemin-dependent MRSA SCVs (Table 1). The mecA-negative MSSA SCVs were β-lactam susceptible. Therefore, MRSA status correlated well with cefoxitin resistance for NCs and SCVs.

All isolates were susceptible to minocycline, chloramphenicol, quinupristin-dalfopristin, linezolid, and tigecycline (Table 1). All but 3 MRSA isolates (1 NC and 2 SCV isolates) were susceptible to fusidic acid. While methicillin resistance status correlated well with β-lactam resistance and thymidine auxotrophy correlated well with trimethoprim-sulfamethoxazole resistance, susceptibility to other clinically useful classes (including aminoglycosides, fluoroquinolones, macrolides, clindamycin, and rifampin) could not be reliably predicted from those two characteristics. There are no established disk diffusion breakpoints for fosfomycin and vancomyin; therefore (although testing on the full set of isolates was completed), percent susceptibilities are not reported in Table 1.

(ii) Microaerobic disk diffusion.

The CF lung has areas with limited oxygen (28), and SCVs frequently exhibit metabolic characteristics (29) that could impact their growth under different atmospheric conditions. Therefore, we compared AST results performed under microaerobic versus aerobic conditions. ATCC 25923 gave similar results under both conditions on BD-BHI agar. On MHA and sup-MHA, zone sizes for cefoxitin, chloramphenicol, oxacillin, and tigecycline fell outside the CLSI ranges only under microaerobic conditions (Fig. 4; see also Fig. S5 in the supplemental material).

The susceptibilities of the vast majority of clinical isolates tested (26 SCVs and 13 paired isogenic NCs) under microaerobic conditions were identical to those of the clinical isolates tested under aerobic conditions, with one exception. Four clonally related, thymidine-dependent, mecA-positive isolates were nonsusceptible to cefoxitin and oxacillin under aerobic conditions but were susceptible to both under microaerobic conditions (not shown).

Comparison of disk diffusion AST on in-house-prepared and commercially prepared sup-MHA.

To this point, method development was completed with in-house-prepared sup-MHA. To test whether commercially prepared sup-MHA would perform similarly to our own, we asked Teknova to prepare sup-MHA (referred to here as THM-MHA) using our supplementation protocol. As a quality control, disk diffusion antibiograms for a subset of 5 clinical SCVs, the corresponding isogenic NC isolates, and control strain ATCC 25923 were determined on Teknova MHA (NC and the control strain only) and THM-MHA (all 11 isolates). The susceptibilities of strain ATCC 25923 and all 11 isolates tested were identical between sup-MHA and THM-MHA (not shown).

AST by Etest.

Etest was performed using THM-MHA to obtain MICs. As a quality control step, we compared the susceptibilities of S. aureus ATCC 29213 on THM-MHA and Teknova MHA. The MICs fell within the acceptable range limits of CLSI for all drugs tested (Fig. 5) on both media. Thus, supplementation of the CLSI-recommended base medium did not influence the susceptibilities of the S. aureus control strain.

FIG 5.

Median MICs for strain ATCC 29213. The median MICs were determined by the Etest for control strain ATCC 29213 and fell within CLSI-acceptable ranges (horizontal red dashed lines) for MHA (black circles) and sup-MHA (gray squares) for all the antibiotics tested. Errors bars depict the lower and upper range limits detected for 4 replicates.

We performed Etest on a subset of the isolates (24 NC and 26 SCV isolates) previously analyzed by disk diffusion. All isolates grew on THM-MHA, and MICs were measurable after 24 h. As cefazolin and meropenem Etest strips are not available, cephalothin and imipenem Etest strips, respectively, were used as surrogates. Vancomycin susceptibility testing was performed only by Etest because there are no established vancomycin disk diffusion breakpoints; all isolates were vancomycin susceptible (Table 2). No discrepancies between disk diffusion and Etest results were observed.

NC susceptibilities as predictors of SCV susceptibilities.

We evaluated whether the in vitro susceptibilities of SCVs could be predicted from the antibiograms of isogenic NC S. aureus isolates recovered before SCVs from the same patients were recovered or concurrently with SCVs from the same patients. As stated in the introduction, exposure to either trimethoprim-sulfamethoxazole or aminoglycosides (amikacin or tobramycin) has been linked to SCV emergence, and thus, differences in susceptibility between NCs and SCVs were expected for these drugs. However, some SCV/NC pairs also differed in their susceptibilities to other antibiotics (Table 3), and the susceptibilities of some genetically related SCVs from the same patient differed as well (see Table S3 in the supplemental material). Therefore, the susceptibilities of SCVs cannot be consistently predicted by the susceptibilities of isogenic isolates from the same patients, indicating that in vitro susceptibilities for an SCV can be reliably defined only by testing that isolate.

TABLE 3.

Comparison of SCV and NC susceptibility profiles

^a The isolate collection was primarily composed of thymidine-dependent SCVs. Hemin- and menadione-dependent SCVs made up the minority of clinical isolates.

^b Exposure to these drugs has been linked to the emergence of SCVs (tobramycin and amikacin with hemin- and menadione-dependent SCVs and trimethoprim-sulfamethoxazole with thymidine-dependent SCVs), and therefore, discrepancies between NC and SCV susceptibilities are not surprising for these drugs.

^c Shading highlights any value(s) indicating less than 100% similarity between SCV and NC susceptibilities to an antibiotic. TD-SCV, thymidine-dependent SCV; HD/MD-SCV, hemin- and menadione-dependent SCV; HD-SCV, hemin-dependent SCV.

Antibiotic exposure as a predictor of SCV susceptibilities.

To determine whether a history of recent exposure to a specific antibiotic reliably predicts SCV resistance to that antibiotic (or others in its class), we analyzed the relationship between SCV in vitro susceptibilities on sup-MHA and antibiotic treatment data from the parent study (9). As shown in Table 4, SCVs were frequently, but not always, nonsusceptible to antibiotics to which the source patient was recently exposed. Exposures to tobramycin and trimethoprim-sulfamethoxazole were each associated on logistic regression with a significantly higher odds that subsequent SCVs would be nonsusceptible to those antibiotics (OR, 18.1 [95% CI, 3.2, 100.9] for tobramycin; OR, 2.5 [95% CI, 0.3, 20.6] for trimethoprim-sulfamethoxazole). However, the correlation was not 100%. While treatment with other antibiotics prior to SCV isolation was relatively less frequent, SCVs isolated after azithromycin, clindamycin, or rifampin exposure were invariably resistant to those drugs; in contrast, SCVs isolated after fluoroquinolone treatment were uniformly susceptible to levofloxacin. There were no other antibiotic exposures identified in SCV-positive subjects. Therefore, a history of exposure to all antibiotics mentioned above except fluoroquinolones may provide a clinically meaningful rationale for avoiding those drugs when treating SCV infections, assuming that in vitro susceptibilities accurately predict the clinical response. Nevertheless, these results indicate that treatment history, as for other candidate predictors, has limited utility in predicting in vitro susceptibility results.

TABLE 4.

Relationships between antibiotic exposure and in vitro susceptibilities^c

Antibiotic(s) tested	Susceptibility findings for 71 isolates in the analysis data set	Antibiotic exposure(s) (when present) within 90 days prior to culture datea	Logistic model resultsb
Amikacin	29 susceptible, 42 not susceptible	Amikacin (no instances of amikacin exposure)	Not applicable
		Tobramycin (27 isolates with prior tobramycin exposure; 25 isolates were not susceptible)	OR = 19.9 (95% CI, 3.5, 111.9; P = 0.001)
Azithromycin	13 susceptible, 58 not susceptible	Clarithromycin (no instances of clarithromycin exposure)	Not applicable
		Azithromycin (8 isolates from 5 subjects with prior azithromycin exposure; all 8 isolates were not susceptible); azithromycin or clindamycin (11 isolates from 6 subjects with prior azithromycin or clindamycin exposure; all 11 isolates were not susceptible)	Unable to fit
Clindamycin	49 susceptible, 22 not susceptible	Clindamycin (3 isolates from 1 subject with prior clindamycin exposure; all 3 isolates were not susceptible)	Unable to fit
		Azithromycin or clindamycin (11 isolates with prior azithromycin or clindamycin exposure; 7 isolates were not susceptible)	OR = 5.2 (95% CI, 0.8, 34.0; P = 0.08)
Levofloxacin	38 susceptible, 33 not susceptible	Quinolones (ciprofloxacin, levofloxacin, or moxifloxacin; 8 isolates with prior quinolone exposure; 7 isolates were susceptible)	OR = 0.14 (95% CI, 0.02, 1.1; P = 0.06)
Rifampin	41 susceptible, 30 not susceptible	Rifampin (12 isolates from 4 subjects with prior rifampin exposure; all 12 isolates were not susceptible)	Unable to fit
Tobramycin	28 susceptible, 43 not susceptible	Tobramycin (27 isolates with prior tobramycin exposure; 25 isolates were not susceptible)	OR = 18.1 (95% CI, 3.2, 100.9; P = 0.001)
Trimethoprim-sulfamethoxazole (SXT)	4 susceptible, 67 not susceptible	SXT (50 isolates with prior SXT exposure; 48 isolates were not susceptible)	OR = 2.5 (95% CI, 0.3, 20.6; P = 0.39)

^aPrevious antibiotic exposure was determined by evaluating antibiotic start and end dates in relation to the date that the sample for culture was obtained. Antibiotic exposure was defined as being present if the antibiotic was used within the 90 days prior to the date that the sample for culture was obtained. None of the subjects were exposed to amikacin or clarithromycin during the study, as indicated.

^bLogistic regression models accounted for multiple isolates per subject and used robust variance estimates. Logistic models were unable to be fit when the presence of prior antibiotic exposure perfectly predicted the isolate susceptibility category.

^cAll antibiotics to which subjects who were culture positive for SCVs during the study reported that they were exposed within 90 days prior to the collection of samples for culture are included.

DISCUSSION

SCVs have been selected in vitro following antibiotic exposure, and antibiotic treatment has been associated with the subsequent detection of SCVs (10, 30). Therefore, antibiotic selection plays key roles in SCV emergence and subsequent SCV susceptibilities. While AST is the standard method for identifying appropriate therapies, the growth characteristics of SCVs render them incompatible with recommended AST methods. Several studies have reported the susceptibilities of limited numbers of SCVs using various in vitro methods (see Table S1 in the supplemental material), but media vary substantially in their abilities to support SCV growth and minimize reversion (including the same medium from different manufacturers, as shown here). We found that two media, BD-BHI agar and sup-MHA (commercially, THM-MHA), allowed reproducible AST determination within 24 h while minimizing reversion to the NC phenotype. The resulting in vitro susceptibilities could not be reliably predicted from clinical or biochemical characteristics or from the susceptibilities of genetically related NC isolates from the same patients.

As expected, all thymidine-dependent SCVs were resistant to trimethoprim-sulfamethoxazole on both media, and hemin- and menadione-dependent SCVs were nonsusceptible to aminoglycosides (at least on BD-BHI agar). Similarly, MRSA isolates were generally resistant to β-lactams. Therefore, our results indicate that susceptibility to certain antibiotics may be reliably predicted from either SCV auxotrophy or MRSA status. However, discrepancies existed even for these simple comparisons (discussed in detail below). The antibiograms of SCVs and their genetically related NCs were usually similar, but this was not always the case, indicating that NC isolate susceptibilities may be used as a rough guide for the appropriate treatment for SCVs at best. Our results also indicate that SCVs from the same patient can exhibit diverse susceptibilities (see Table S3 in the supplemental material). A similar diversity of Pseudomonas aeruginosa isolates from the same CF patient has been identified (31, 32). While we did not determine whether the isolates exhibited hypermutability, which has been found in S. aureus isolates from CF patients (33), this characteristic could impact the tempo of this diversification and adaptation, including the development of resistance and SCV emergence. Similarly, genotyping our isolates with multilocus sequence typing would identify clonal complexes, an additional, portable candidate predictor of susceptibilities. These analyses are therefore of interest for future studies.

While azithromycin resistance was common among SCVs regardless of auxotrophy, prior work found no association between azithromycin usage and SCV detection (34), indicating that SCVs were not likely selected by azithromycin. Rather, these results probably reflect common macrolide usage in the CF population (33). Similarly, MRSA SCVs were more often multidrug resistant than MSSA SCVs, likely reflecting the higher antibiotic exposure intensities among MRSA-positive CF patients (35, 36).

Identification of MRSA is an important function of the clinical laboratory, as infections caused by these isolates are difficult to treat and may be associated with worse patient outcomes (36). Disk diffusion testing on BD-BHI agar and sup-MHA with cefoxitin disks correctly identified MRSA SCVs. False-negative results were encountered when oxacillin was used. During manuscript preparation, another study reported the detection of thymidine-dependent MRSA SCVs through the use of agar medium supplemented with 10 μg/ml thymidine (37). Our study supports this approach (albeit with 5 μg/ml thymidine) and extends it to other auxotrophic types by supplementation with menadione and hemin. Our results further show that AST with supplemented media is reproducible and useful beyond the detection of methicillin resistance.

While SCVs generally grew well on BD-BHI agar and sup-MHA with minimal reversion, each medium had notable limitations. The control strain, ATCC 29523, did not always exhibit susceptibilities within the CLSI-acceptable ranges for certain antibiotics on BD-BHI agar, including tigecycline, and this strain was often more susceptible to specific drugs on BD-BHI agar than on sup-MHA. Thymidine-dependent SCVs grew on BHI agar from one manufacturer but not BHI agars from two others, suggesting that formula variations impacted growth. The media developed here (sup-MHA and THM-MHA) used the CLSI-recommended base medium, MHA, supplemented with compounds required by SCVs for optimal growth. Concentrations were chosen on the basis of the best available knowledge about their abundance in CF sputum (19, 38) as well as their empirically determined abilities to support SCV growth and stability (see Fig. S1 in the supplemental material). Manufacturer-to-manufacturer differences were eliminated by use of this medium. Most importantly, supplementation of MHA did not impact the susceptibilities of ATCC 29523 (disk diffusion) or ATCC 29213 (Etest); zone diameters and MICs fell within CLSI-acceptable ranges and were comparable to those on unsupplemented MHA (Fig. 4 and 5). AST results were comparable on MHA and sup-MHA for the NC isolates tested. The only susceptibility differences observed between BD-BHI agar and sup-MHA were for aminoglycosides among hemin- and menadione-dependent SCVs, both of which have defects in electron transport and, consequently, in aminoglycoside uptake (29). Therefore, these susceptibility differences likely reflect different hemin and menadione concentrations.

One could argue that stimulating the growth of SCVs through medium supplementation masks true susceptibilities. However, SCVs reach sputum densities in children with CF similar to those of NC S. aureus (9), and thymidine has been detected in the sputum of CF patients at concentrations similar to those found in sup-MHA (19, 38). In contrast to thymidine, the concentrations of hemin and the naturally occurring menadione analogs (menaquinones) have not been defined for the sputum of CF patients, but they may be present in the airways of CF patients in amounts sufficient to support SCV growth. Thus, while the in vivo growth characteristics and antibiotic susceptibilities of SCVs are not known, these behaviors could be approximated by growth on supplemented media.

While the AST method developed here was applied to respiratory SCVs from CF patients, this methodology could be extended to diverse S. aureus SCV infections. This method may or may not reflect the in vivo antibiotic response of SCVs, but we identified 100% susceptibility among a large collection of clinical SCVs to several antibiotics (linezolid, vancomycin, tigecycline, chloramphenicol, minocycline, and quinupristin-dalfopristin) that may prove useful as empirical choices for treating SCV infections in the absence of AST. We suggest that the antibiotic treatment of S. aureus SCV infection should be the focus of prospective clinical trials, preferably measuring both clinical and microbiological outcomes, using the results (for testing empirical choices) and methods (for testing AST-guided choices) we describe here.

ADDENDUM IN PROOF

After acceptance of our manuscript, the manufacturer of Etest strips notified us of a recall of piperacillin-tazobactam Etests due to a manufacturing defect; retesting of all strains and isolates with new strips yielded the same results as the original tests.

Funding Statement

Seattle Children's Research Institute provided funding to Lucas R. Hoffman and Jane L. Burns through Ignition Funds. NIH provided funding to Lucas R. Hoffman and Jane L. Burns under grant number P30DK089507. NIH also provided funding to Lucas R. Hoffman and Mimi R. Precit under grant numbers K02HL105543-01 and ESR01A29422, respectively. The Cystic Fibrosis Foundation provided funding to Lucas R. Hoffman under grant number HOFFMA07P0. The National Science Foundation provided funding to Mimi R. Precit under a Graduate Research Fellowship. The American Thoracic Society (ATS) provided funding to Lucas R. Hoffman under grant number CF-07-003.