Validation of a Gradient Diffusion Method (Etest) for Testing of Antimicrobial Susceptibility of Aerococcus urinae to Fluoroquinolones

ABSTRACT

Aerococcus urinae is a urinary pathogen with well-described resistance to fluoroquinolones. This study aimed to validate the gradient diffusion (GD) method (Etest) on cation-adjusted Mueller-Hinton agar with 5% sheep blood for testing the susceptibilities of Aerococcus urinae to the antimicrobial agents ciprofloxacin and levofloxacin and to compare the Etest to the broth microdilution (BMD) method from CLSI document M45-A3. Agar dilution (AD), as recommended by EUCAST, was used as an alternative reference method to arbitrate discrepancies or address technical issues. Aerococcus urinae isolates from urinary specimens were prospectively collected between June 2016 and December 2017 from six hospitals in Quebec, Canada, and identifications were confirmed using Vitek MS with the IVD 3.0 database. Of the 207 isolates tested using BMD, 37 (17.9%) showed trailing and 19 (9.2%) showed insufficient growth; these were tested using AD. Also, 38 isolates (18.4%) for ciprofloxacin and 13 isolates (6.3%) for levofloxacin showed a lack of essential or categorical agreement between the Etest and BMD and were also tested by AD. By use of a combined reference method (BMD or AD), the susceptibility rates of Aerococcus urinae were 82.6% and 81.6% for ciprofloxacin and levofloxacin, respectively. Categorical agreement between GD and the combined reference methods was 95.2% for ciprofloxacin and 97.1% for levofloxacin, with no very major error identified. Major and minor error rates were 0.6% and 4.3% for ciprofloxacin and 1.2% and 1.9% for levofloxacin. Overall, antimicrobial susceptibility testing (AST) using the Etest on sheep blood agar showed good agreement with the reference methods and can be considered by clinical laboratories wishing to perform AST on Aerococcus urinae isolates.

INTRODUCTION

Aerococcus urinae is an emerging urinary tract pathogen that was first described by Christensen et al. in 1989 (1). It is the Aerococcus species most commonly isolated from human clinical specimens and is found in about 0.2 to 0.8% of urinary tract specimens (2). This pathogen is responsible for a wide range of infections, from uncomplicated urinary tract infections (UTIs) to potentially fatal endocarditis (3, 4), and is usually found in elderly patients or patients with predisposing factors for UTI, such as prostate hypertrophy or a Foley catheter (2). Traditional identification methods can easily misidentify it as alpha-hemolytic Streptococcus or an Enterococcus species, and it has likely been largely underreported in past decades (5). However, matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) methods now allow for quick and accurate identification of this organism to the species level in most clinical laboratories.

Aerococcus urinae isolates are usually susceptible to beta-lactams, vancomycin, and nitrofurantoin (for UTI isolates) and are considered to be intrinsically resistant to trimethoprim-sulfamethoxazole (TMP-SMX) (6). However, susceptibility to fluoroquinolones is less predictable; resistance rates between 4 and 33%, as determined by use of various antimicrobial susceptibility testing (AST) methods, have been reported in the literature (7–13). Since this class of antibiotics is often used for empirical treatment of UTIs, a reliable AST method is needed for these antibiotics.

The Clinical and Laboratory Standards Institute (CLSI) first published clinical breakpoints for Aerococcus species in 2016 (6). The proposed AST method relies on broth microdilution (BMD) with cation-adjusted Mueller-Hinton broth (CAMHB) supplemented with lysed horse blood (LHB). However, the BMD method is complex and time-consuming and hence is generally restricted to reference clinical microbiology laboratories. The CLSI has yet to provide criteria for widely available susceptibility testing with disk diffusion on solid media. Since 2017, EUCAST (the European Committee on Antimicrobial Susceptibility Testing) has provided clinical breakpoints with a similar reference method using BMD with CAMHB, LHB, and β-NAD. EUCAST also provided clinical breakpoints for disk diffusion relying on Mueller-Hinton agar (MHA) supplemented with defibrinated horse blood and β-NAD. For historical reasons, MHA supplemented with horse blood is seldom available in North America (14). It remains unclear whether EUCAST recommendations can be directly applied to agar-based AST relying on a sheep blood-supplemented medium, since it has already been shown that different results could be obtained in horse blood- versus sheep blood-based media (15). Two recent studies evaluated gradient diffusion (GD) using Etest strips (bioMérieux, Marcy-L’Étoile, France) to test for susceptibility to fluoroquinolones but were conducted on small numbers of isolates and did not compare GD with any reference methods (12, 16).

The goal of this study was to validate the use of the GD AST method on MHA supplemented with 5% sheep blood for testing the susceptibility of Aerococcus urinae to fluoroquinolones and to compare this method to the BMD method described in CLSI guideline M45-A3 (6) as the reference method. We also relied on agar dilution (AD), as recommended by EUCAST, as an alternative reference method to account for technical issues or to arbitrate discrepancies (17).

MATERIALS AND METHODS

Study isolates.

The American Society for Microbiology’s Cumitech 31A (18) recommends the inclusion of at least 35 resistant isolates to validate susceptibility testing methods. Therefore, we aimed to collect 200 isolates in order to adequately assess resistance to fluoroquinolones, assuming a resistance rate of approximately 15%, as described in the literature (11). Only isolates from urinary tract specimens were included in this study. Isolates were collected between June 2016 and December 2017 from hospitals in Montreal (Hôpital du Sacré-Coeur de Montréal) and Quebec City (Hôpital de L’Enfant-Jésus, Hôpital du Saint-Sacrement, Centre Hospitalier de l’Université Laval, Hôpital Saint-François d’Assise, and Hôtel-Dieu de Québec) in Canada. Organisms were initially identified by routine methods in each laboratory: traditional phenotypic identification, the API gallery, the Vitek 2 identification system, or MALDI-TOF technology on either the MicroFlex (Bruker, Billerica, MA, USA) or the Vitek mass spectrometry (MS) (bioMérieux, Marcy-L’Étoile, France) system. Isolates were frozen at −80°C for further use. The identification of all isolates to the species level was later confirmed using Vitek MS with the IVD 3.0 database at the Laboratoire de Santé Publique du Québec (LSPQ) reference laboratory, Ste-Anne-de-Bellevue, Canada.

Susceptibility testing.

Levofloxacin and ciprofloxacin were tested by both the GD method under evaluation and the gold standard (see below) for each isolate. Technical work was divided between the Hôpital de l’Enfant-Jésus (GD) and the LSPQ (BMD and agar dilution). Streptococcus pneumoniae ATCC 49619 was used as the quality control (QC) strain for all techniques. Staphylococcus aureus ATCC 29213 was used as an additional QC strain for both AD and BMD.

The GD method was performed using Etest strips on MHA plus 5% sheep blood incubated under 5% CO₂ for 20 to 24 h. Plates were inoculated with a bacterial suspension at a 0.5 McFarland standard using the Etest Inoculator Retro C80 device (bioMérieux, Marcy-L’Étoile, France). MICs were assessed according to the package insert. BMD was performed by using the Sensititre system (Thermo Fisher, Waltham MA, USA) according to the manufacturer’s instructions and following the method described in CLSI guideline M45-A3 (6). Concentrations of antibiotics ranging from 0.016 mg/liter to 16 mg/liter were used. Isolates showing insufficient growth at 24 h by either GD or BMD were reincubated and read at 48 h. To ensure the reliability and reproducibility of results for BMD, high-resolution photographs were taken, and MICs were interpreted according to CLSI M45-A3 breakpoints (6). All isolates that failed to grow or that created an obvious trailing effect on BMD for either levofloxacin or ciprofloxacin were retested using the EUCAST reference method, agar dilution (AD) using MHA supplemented with 5% defibrinated horse blood and 20 mg/liter β-NAD incubated under 5% CO₂ for 48 h, to determine the MIC (0.25 to 8 mg/liter); this was then considered to be the alternative gold standard. Moreover, recognizing that BMD reading and interpretation were challenging for this species (see Results), we decided that all isolates showing a discrepancy (lack of either essential or categorical agreement) between GD and BMD would be further tested by AD to resolve the conflict. In these cases, the AD result was always chosen over the BMD result. Therefore, a combination of BMD and AD was used as the gold standard. Considering this gold standard, agreement with the GD method, including essential (i.e., within 1 dilution) and categorical (susceptible [S], intermediate [I], or resistant [R]) agreement, as well as minor, major, and very major error percentages, were calculated according to Cumitech 31A guidelines (18). MICs were interpreted using CLSI M45-A3 breakpoints (6).

RESULTS

Over the study period, 208 urinary tract isolates were prospectively collected and frozen for further characterization. Of these, 207 could be retrieved for confirmation of identification and AST at the time of this study. All isolates (100%) were confirmed to be Aerococcus urinae by Vitek MS MALDI-TOF technology with the IVD 3.0 database. Detailed MIC data for all 207 isolates are presented in Data Set S1 in the supplemental material. QC strains were within the expected range for all methods based on CLSI guideline M45-A3 (6; also the supplemental material). The distributions of MICs for GD and the reference method are presented in Fig. 1. Overall, 37 (17.9%) isolates showed clear trailing and 19 (9.2%) isolates showed insufficient growth for both levofloxacin and ciprofloxacin. Discrepancies (lack of either categorical agreement or essential agreement) were observed for 40 (19.3%) isolates for ciprofloxacin and 13 (6.3%) isolates for levofloxacin. Therefore, a total of 96 (37 + 19 + 40) (46.4%) isolates were considered for AD for ciprofloxacin, of which 2 could not be retested (1 lost, 1 showing no growth) and therefore retained the initial value provided by the BMD method (with minor errors relative to GD), leaving 94 isolates with an AD result for ciprofloxacin (Table 1). For levofloxacin, a total of 69 (37 + 19 + 13) (33.3%) isolates were successfully retested (Table 1). After the two reference methods were compared, it was observed that for some isolates, BMD was more difficult to interpret because growth was not constant or gave the impression of a trailing effect with very granular growth in the wells (Fig. 2 shows an example). All the isolates showing this trailing effect were found to be susceptible when retested with AD, suggesting that no specific resistance pattern was associated with this phenotype (Table 1). Overall, GD results were easier to interpret, but growth was sometimes slow and required additional incubation, for as long as 48 h.

FIG 1.

Distributions of ciprofloxacin and levofloxacin MICs according to GD and the reference method (either broth microdilution or agar dilution). *, The lowest MIC tested for agar dilution was 0.25 mg/liter; therefore, all isolates with a MIC of ≤0.25 mg/liter by either BMD or AD were included in this category. **, The highest MIC tested for agar dilution was 8 mg/liter; therefore, all isolates with a MIC of ≥16 mg/liter by either BMD or AD were included in this category.

TABLE 1.

Interpretation of susceptibility testing for ciprofloxacin and levofloxacin according to the method used

Method	No. of isolates with the indicated reaction (MIC [mg/liter]) to:
	Ciprofloxacin				Levofloxacin
	S (≤1)	I (2)	R (≥4)	Total	S (≤2)	I (4)	R (≥8)	Total
Etest	168	15	24	207	167	2	38	207
BMD	88	4	21	113	106	2	30	138
Agar dilution (trailing)	37	0	0	37	37	0	0	37
Agar dilution (insufficient growth)	17	1	1	19	17	1	1	19
Agar dilution (disagreement)	29	3	6	38	9	3	1	13
Reference method—total	171	8	28	207	169	6	32	207

FIG 2.

Trailing effect on Sensititre broth microdilution. Shown is an example of the trailing effect observed when BMD was used for testing the susceptibility of Aerococcus urinae to ciprofloxacin. This pattern with granular growth was observed after 48 h of incubation at 35°C under a 5% CO₂ atmosphere. Ciprofloxacin concentrations ranged from 16 mg/liter (well 1) to 0.016 mg/liter (well 11). Well 12 contains the growth control.

MIC interpretations according to AST methods are presented in Table 1. The results obtained by AD were frequently in agreement with the GD results, especially in the lower range of MICs, whereas BMD had a tendency to overestimate the MIC produced by the two other methods by 1 or 2 dilutions. For example, of the 38 discrepancies that could be solved with AD for ciprofloxacin, 26 (68.4%) were either minor errors (S versus I) or a lack of essential agreement within the susceptible range, where isolates had a higher MIC with BMD and were susceptible or had a lower MIC when tested with either GD or AD. This phenomenon was much less common with levofloxacin, where most of the discrepancies (8 out of 13) were within the susceptible range and where GD indicated lower MICs than BMD.

By use of a combined reference method (either BMD or AD), agreement with GD was found to be acceptable, without very major errors, and with a major error rate below 3% for both fluoroquinolones and an overall error rate below 7% (18). The rate of GD agreement with the reference method is presented in Table 2. According to the reference method, cross-resistance between ciprofloxacin and levofloxacin was high (Table 3): among 171 ciprofloxacin-susceptible isolates, 3 (1.8%) were not susceptible to levofloxacin and only 1 (0.6%) strain was fully resistant. Conversely, only 1 (0.6%) of 169 levofloxacin-susceptible isolates was nonsusceptible (intermediate) to ciprofloxacin.

TABLE 2.

Agreement of Etest with a reference method (either AD or BMD)

Agreement or error	Value (%)a
	Ciprofloxacin	Levofloxacin
Categorical agreement	95.2	97.1
Essential agreement	94.7	97.1b
Very major errors	0	0
Major errors	0.6 (1/171)	1.2 (2/169)
Minor errors	4.3 (9/207)	1.9 (4/207)

^aFor major and minor errors, the number of isolates with errors/total number of isolates is given in parentheses.

^bOne isolate had a MIC of 8 mg/liter by GD and ≥32 mg/liter by BMD and could not be retested by AD (did not grow); it was considered to have categorical agreement (within the resistant range) but not essential agreement.

TABLE 3.

Cross-resistance observed between ciprofloxacin and levofloxacin (using the reference method)

Levofloxacin result	No. of isolates with the following ciprofloxacin result:			Total no. of isolates
	S	I	R
S	168	1	0	169
I	2	4	0	6
R	1	3	28	32
Total	171	8	28	207

DISCUSSION

To our knowledge, this study represents the largest collection of Aerococcus urinae isolates tested for fluoroquinolone susceptibility. Overall, we found that the rate of A. urinae susceptibility to fluoroquinolones was comparable to that described in the literature (82.6% for ciprofloxacin and 81.6% for levofloxacin) (13). GD showed good agreement with reference methods, with no very major errors and very good categorical and essential agreement. However, our sample fell short of the 35 resistant isolates recommended by Cumitech 31A, lacking 7 isolates resistant to ciprofloxacin and 3 isolates resistant to levofloxacin (18). Also, for a significant portion of the isolates (n = 56 [27%]), BMD showed either growth problems or trailing. This phenomenon was also observed when we sent isolates to the Clinical and Laboratory Standards Institute (USA) to have the procedure repeated. Aerococcus growth seems easier on agar-based media than in broth-based media. An explanation for growth and reading issues in microdilution broth could be the ability of certain strains to exhibit an aggregative phenotype in vitro in liquid-based media, as described recently by Hilt and al. (19). In general, issues can also be seen with some isolates of other fastidious organisms, suggesting that a subset of isolates may need some unknown nutrient factors that are lacking in broth-based media (20). We therefore strongly agree with EUCAST’s recommendation stating “For fluoroquinolones, AD may produce clearer endpoints” (17) and would suggest using AD methods as a reference in any further studies.

Although our study presented a relatively low number of resistant isolates, with regard to the reference method MIC, there was 96.6% categorical agreement between ciprofloxacin and levofloxacin, and levofloxacin could accurately predict resistance to ciprofloxacin, with no very major error observed. Therefore, it is likely that levofloxacin could be used as a surrogate marker for ciprofloxacin resistance (21).

This study presents several limitations. Our initial gold standard (BMD) presented many technical problems, and we needed to rely on a second reference method (AD) for many isolates. The fact that the GD method was performed in a clinical laboratory while the BMD and AD methods were performed in a reference laboratory could introduce another source of variability in our results. However, we would not expect this to artificially increase the agreement level. Moreover, this multisite approach reflects a “real-life” setting that many laboratories will encounter: initial testing with a GD method in the clinical laboratory and later confirmation with a reference method performed elsewhere. Our isolates came exclusively from urinary tract specimens, making it unclear if these findings would be applicable to a subset of isolates from blood cultures, although there is no evidence to the contrary. Clinical information was not collected to differentiate true urinary pathogens from potential colonizers. Fewer resistant isolates than expected were included, which limited this study’s ability to identify very major errors related to GD.

Conclusions.

In this study, we demonstrated that the susceptibility of Aerococcus urinae to fluoroquinolones was near 80% in our area of Canada. Laboratories wishing to test fluoroquinolones on urinary tract isolates can use a GD method with Etest strips on sheep blood agar as a quick and easy way to perform AST on these isolates. This method showed excellent agreement with reference methods and is expected to generate fewer indeterminate results than BMD.