INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

In recent years, enterococci have emerged as important pathogens (10). These organisms are now the third most commonly encountered nosocomial bloodstream infection pathogens (11, 33). Furthermore, the intrinsic antimicrobial resistances possessed by enterococci (3, 16, 35) have limited the number of therapeutic agents. At this time, cell wall-active antimicrobials (penicillins or glycopeptides) are usually administered in combination with an aminoglycoside (15). This synergistic combination therapy is clearly optimal for infections, such as endocarditis and meningitis, which require a bactericidal therapeutic effect (16). The widespread emergence of high-level resistance to aminoglycosides and glycopeptides among enterococci in Europe, North America, and other locations (20, 25, 31) has brought the medical community one step closer to the postantimicrobial era.

Accurate detection of resistant strains is important for microbiologists and clinicians in guiding optimal therapy for enterococcal infections and in implementing interventions to control and prevent nosocomial outbreaks (5). However, several studies have reported that some susceptibility testing methods, especially commercial automated systems and the commonly used disk diffusion test, are unable to detect resistance to key aminoglycosides and glycopeptides (2, 7, 9, 23, 30, 34, 36).

In this study, the ability of MicroScan dried overnight incubation gram-positive MIC antimicrobial combination panel no. 8 (PM-8) to detect vancomycin-resistant (VR) and high-level aminoglycoside-resistant (HLAR) enterococci was compared to those of frozen reference broth microdilution (RBM) trays and Synergy Quad Agar (QA) plates. Only the panel components involved in the susceptibility testing of glycopeptides and aminoglycosides were evaluated. Furthermore, other MicroScan panels may be used for these tests, and the presented results should apply to those panels, as well. The MicroScan panel was evaluated in three phases at each of three medical centers in the United States: Washington University School of Medicine (WUSM)/Barnes Jewish Hospital, St. Louis, Mo.; University of Iowa College of Medicine (UICM), Iowa City, Iowa; and University of Wisconsin Hospital and Clinic (UWHC), Madison, Wis. Molecular gene detection by PCR was used as the reference test. Challenge strains from stock collections and recent clinical isolates of Enterococcus spp. were tested.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Organisms. Fifty recent clinical enterococcal isolates encountered at each participant site were tested at that site in the efficacy phase (EP). All isolates were from systemic infections, and only one isolate per patient was processed. After the testing, these clinical isolates were subcultured and shipped to the UICM for further molecular characterization. In the challenge phase (CP), testing was conducted with two previously characterized sets of strains obtained from the Centers for Disease Control and Prevention (CDC). One set contained 48 strains characterized for phenotypic and genotypic parameters of HLAR, and the second set contained 50 strains with characterized phenotypes and genotypes for VR. Both groups of organisms (masked) were sent by MicroScan to each study site for evaluation. An additional 50 geographically diverse enterococcal isolates collected from previously reported national surveillance studies were tested at UICM only (11). In the reproducibility phase (RP), the abilities of all three methods to provide consistent and reproducible susceptibility test results (categorical and ± 1 log₂ unit dilution) were examined. Reproducibility testing included 10 strains, five selected from each set of CDC challenge strains. This experiment was conducted in triplicate for three consecutive days by each study site (90 total tests per laboratory per test method).

Susceptibility testing methods. All strains were twice subcultured from frozen stocks or from chocolate transport agar slants to sheep blood agar plates prior to being tested. Several morphologically similar colonies were then used to prepare an inoculum suspension equal to a 0.5 McFarland turbidity standard by using a spectrophotometer (MicroScan turbidity meter; Baxter Diagnostics Inc., Deerfield, Ill.). Susceptibility testing was performed on all isolates in accordance with the manufacturer's instruction. PM-8 panels contained 500 µg of gentamicin, 1,000 µg of streptomycin, and 2 to 16 µg of vancomycin per ml. Following inoculation, the panels were incubated in the Walk/Away system and read automatically at 18 and 24 h. The results reported here were those recorded after 24 h of incubation. The panels were subsequently removed from the Walk/Away system, and those indicating susceptibility to streptomycin at 24 h were placed in an ambient air incubator (35°C) and read manually at 48 h. RBM trays (frozen form; MicroScan), which were tested according to National Committee for Clinical Laboratory Standards (17, 18, 28) procedures, contained gentamicin (500 µg/ml), streptomycin (1,000 µg/ml), and vancomycin (2 to 16 µg/ml). QA plates contained gentamicin (500 µg/ml), streptomycin (2,000 µg/ml), and vancomycin (6 µg/ml). For isolates undergoing vancomycin evaluations, a MicroScan European dried overnight gram-positive Combo type 3I panel containing teicoplanin (4 to 16 µg/ml) was tested to characterize the glycopeptide resistance phenotype, and this tray was read manually. All results were validated by PCR analysis (amplification) of the representative resistant gene. Discordance between testing methods was resolved by repeat testing using all systems and by the E-test (AB BIODISK, Solna, Sweden) method (7, 10).

Molecular characterization. The VR genotypes (vanA, vanB, vanC₁, and vanC_2-3) were confirmed by amplifying the respective genes by PCR as described previously (4, 11). Briefly, PCR amplification of the genes of HLAR was performed as follows. (i) Robotically prepared stock quantities of master mixtures included 50 pmol of each oligonucleotide primer for high-level gentamicin resistance (HLGR) testing [AAC(6') plus APH(2")] and for high-level streptomycin resistance (HLSR) testing [ANT(6)-I] per liter. These primers amplified 985- and 597-bp fragments, respectively. Primer sequences were those described by Huycke et al. (8) or were obtained from D. Sahm (26a). (ii) The mixtures contained 200 mmol of dATP, dTTP, dGTP, and dCTP, 50 mmol of NH₄Cl, 1.5 mmol of MgCl₂, and 10 mmol of Tris-HCl buffer (pH 9.0) per liter at room temperature (1). (iii) Aliquots were then robotically loaded into a 96-well polycarbonate plate (Corning), which was kept at 4°C in a minirefrigerator. (iv) The distribution of the master mixture (XP robotic system) included the adding of 10 µl of target template DNA and the pipetting of 35 µl of light mineral oil as an overlay to avoid evaporation and contamination. (v) Two and one-half units of Taq polymerase (Promega) per liter was added to the master mixture immediately prior to the assembling procedure. (vi) Thermal cycling conditions for HLGR testing were as follows: 10 min at 95°C initially, followed by 35 cycles of 1 min at 94°C, 1 min at 55°C, and 3 min at 72°C; those for HLSR testing were as follows: 10 min at 95°C initially, followed by 35 cycles of 1 min at 94°C, 1 min at 58°C, and 1 min at 72°C (extension). PCR products were analyzed by electrophoresis through an agarose gel. Detection was accomplished by staining the products with ethidium bromide.

Data analysis. Results for vancomycin susceptibility testing of each organism-antimicrobial pair by PM-8 and RBM methods were interpreted as susceptible, intermediate, or resistant according to current National Committee for Clinical Laboratory Standards guidelines (18). For the QA plates and for the synergy wells of high-level aminoglycosides in PM-8 panels and RBM trays, only susceptible and nonsusceptible categories were applicable (one drug concentration tested). Very major errors (false susceptible) were defined as susceptible result from any phenotypic method when the organism was characterized by PCR (genotype) as positive for HLAR (HLGR or HLSR) or VR genes (i.e., vanA, vanB, vanC₁, or vanC_2-3). The opposite pattern (false resistant) was considered a major error, and any other discrepancy (a susceptible or resistant result versus an intermediate result) was defined as a minor error. Minor errors are not possible for synergy tests which use a single drug concentration.

Quality control. Enterococcus faecalis ATCC 29212 (susceptible to vancomycin and aminoglycosides) and ATCC 51299 (resistant to high levels of aminoglycosides) and Staphylococcus aureus AmMS 261 (teicoplanin MIC range, 4 to 16 µg/ml; MicroScan internal control) were included as quality control strains on each day of testing (28, 29).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Characterization of strains. The resistance genotypes for all strains tested in this evaluation are presented in Table 1. One of the 150 clinical isolates tested in the EP (from WUSM) and three strains from the UICM-originated challenge set (tested only at UICM in the CP) were excluded from evaluation because mixed cultures produced data for the phenotypic and genotypic tests that were inconsistent despite multiple replicate tests. The distribution of the EP strains was representative of the actual distribution of drug-resistant enterococci encountered at each institution. Overall, 73.2% of strains tested for the VR genes were negative (range, 61.2 to 94.0%), while 23.5% produced a vanA or vanB PCR product, including four strains from WUSM harboring both genes. The remaining 3.4% of strains produced a band consistent with one of the vanC genotypes, usually associated with Enterococcus gallinarum, Enterococcus casseliflavus, and Enterococcus flavescens (12, 18, 25, 31). The percentages of strains lacking the HLSR or HLGR gene were 54.4% (range, 36.0 to 74.0%) and 77.2% (range, 70.0 to 81.6%), respectively.

Detection of HLAR. The results of in vitro susceptibility testing by the three methods in the EP and CP (three participating medical centers) are depicted in Table 2. In the EP, when the results of the three testing methods for the HLAR genotypes were compared, agreement between all methods occurred for 96.0% of the isolates tested for HLSR and for 97.3% of those tested for HLGR. The incidence of very major (false-susceptible or possible synergy) errors was very low (<1%) for each testing method for detecting HLSR and HLGR. The testing accuracy in detecting HLSR was 98.7% for the RBM and QA methods and 98.0% for the PM-8 method. For HLGR, 99.3% accuracy was observed for all methods tested.

Detection of VR. Table 3 summarizes the error rates for the three phenotypic tests for detecting VR in enterococci having vanA or vanB genotypes. These results were compared to those of the reference molecular test for each gene (PCR). In the EP, a very low error rate (1.4%) was observed for all three tests, with a rate of only 0.7% for very major false-susceptible errors. Of note was the observation that the strain producing an error of this type contained the vanA gene but was found to be phenotypically susceptible to both vancomycin and teicoplanin by all test methods, a feature confirmed by the CDC. Only one other interpretive error was observed with each test.

Reproducibility of categorical interpretations. Ten organisms were selected for reproducibility experiments that included nine replicate results produced per strain in each laboratory over a 3-day period (Table 5). The strains included vanA, vanB, vanC₁, vanC_2-3, and HLAR genotypes. Generally all methods performed well (error rates of 0.7, 1.9, and 10.0% for very major, major, and minor errors, respectively), except for the QA result for HLSR (17.2% very major error rate). This high false-synergy rate was explained by the replicate testing of two strains that contributed five of six (83.3%) very major errors (9.4%; Table 2) in the CP experiments for HLSR. Reproducible error by the QA test (2,000 µg of streptomycin per ml) was observed in these strains for HLSR. The HLSR testing in broth (PM-8 panels or RBM trays with 1,000 µg of streptomycin per ml appeared to be very accurate for detecting HLSR, with 1.5, 2.0, and 5.2% total errors in the EP, CP, and RP, respectively (28, 29). When the reproducibility was assessed in terms of the variations from the results for the all-participant mode for each method, the percentages of agreement (± 1 log₂ unit dilution) were as follows: for VR 100.0 (PM-8), 98.9 (QA), and 90.0% (RBM); for HLSR 99.6 (RBM), 98.5 (PM-8), and 82.2% (QA); and for HLGR 99.6 (RBM), 99.3 (PM-8), and 98.1% (QA). All results appear highly reproducible.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Infections caused by enterococci with antimicrobial resistance are being reported with increasing frequency throughout the United States (10, 11, 22), and evidence has been presented documenting intrahospital (11, 21) or interhospital dissemination (11, 24). The latest (1997) surveillance of bloodstream infections in the United States (22) revealed the following resistance patterns among enterococci: HLGR, 33%; HLSR, 43%; and vancomycin nonsusceptible, 18%. The rates of VR were 20.0 and 7.1% for hospital- and community-acquired infections, respectively (22). Serious infections with these antimicrobial agent-resistant organisms may not be treatable with conventional therapy such as synergistic combinations of cell wall-active drugs (penicillins or vancomycin) and aminoglycosides. Therefore, it is critical for clinical microbiology laboratories to provide accurate antimicrobial susceptibility testing results for enterococci, so that effective therapy and infection control measures can be initiated.

Recent reports have demonstrated the inability of many clinical laboratories (2, 23, 34) and commercial susceptibility testing systems (12, 23, 34) to detect glycopeptide resistance, especially patterns of low- or moderate-level resistance to vancomycin in enterococci (strains with a vanC genotype and certain strains of the vanB type). Enterococci with the vanC genotype, i.e., E. gallinarum and E. casseliflavus, which possess an intrinsic property of constitutive low-level VR (13, 19), are infrequently recovered from clinical specimens (11, 32). The clinical significance of resistance expressed by these organisms, as opposed to the significance of that expressed by vanA or vanB, or HLAR enterococci, remains unclear (32), and the implications of susceptible or intermediate vancomycin results for isolates of these species are also uncertain. A laboratory report of a VR isolate will initiate a cascade of infection control events that are both time-consuming and costly (5) and that should be focused on organisms of the vanA or vanB pattern. Thus, an important adjunct to susceptibility testing is the need for accurate identification of enterococcal species to differentiate intrinsic vanC VR from low-level vanB VR in E. faecalis or Enterococcus faecium (4, 26).

The interpretation of susceptibility testing results for enterococci with the vanC and low-level vanB genotypes was difficult in the present and other reported investigations. While the initial broth microdilution results (2 to 4 µg/ml by CDC) for these vanC enterococci (five E. gallinarum isolates and five E. casseliflavus isolates) indicated that the isolates were susceptible, they might be considered as indicating resistant isolates because of the altered ligases mediating decreased vancomycin susceptibility (19). However, both susceptible and intermediate interpretive categories should be considered correct results for these species, and 96.7 and 100% of testing results by the PM-8 and RBM methods, respectively, fell into these interpretive categories. Based on the same reasoning, both susceptible and nonsusceptible interpretative categories were considered correct for the QA method. In fact, the categorical distribution of results for vanC enterococci in the CP was diverse since 63.3 and 70.0% of the PM-8 and RBM results, respectively, were in the intermediate category. Variations beyond essential accord (greater than ± 1 log₂ unit dilution) were rare, i.e., one or two occurrences. However, the 6-µg/ml QA screen produced 86.7% nonsusceptible results, conforming closely to the modal MIC of 8 µg/ml for these species. Testing results for the five strains of vanC enterococci in the EP were distributed between 4 and 8 µg/ml for both the PM-8 and RBM methods, but all strains were nonsusceptible by QA. In the CDC VR enterococci challenge testing, 88.9% of testing results from six strains with low-level resistance (vanB; MIC, 16 to 32 µg/ml by PM-8) fell into the resistant category and the remaining MICs were in the intermediate category. None of the results were outside the limits of essential accord. All QA results were correctly categorized as nonsusceptible (>6 µg/ml), but 2 of 18 (11.1%) results by the RBM test were classified as susceptible, and a trend toward lower MICs was observed. Enterococci with vanA were easily categorized as resistant by all methods.

In this study, testing results for the automated MicroScan PM-8 panels demonstrated their ability to detect HLAR at an accuracy comparable to that of RBM trays in all three testing phases. Testing accuracy percentages for HLSR by the two broth-based methods were 97.2 and 99.0%; for HLGR the accuracy was 98.5%. Very major errors rarely occurred (1.0% for HLSR and 0.7% for HLGR). The most notable observation was the improved sensitivity of the PM-8 MicroScan panels for the detection of HLSR; earlier panels (25, 30, 36) detected isolates with HLSR at rates of only 41 to 90.2%. Reformulation of the panels with a modified basal broth appears to have enhanced the growth of enterococci resulting in improved test accuracy. In spite of the fact that highly accurate (98.6%) results were obtained with the QA method in the EP (6, 25, 27), problems with unacceptably high error rates (12.0 to 17.8%) were encountered in the CP and RP when using QA to detect HLSR. The factor causing different results for the detection of HLSR by QA between the EP and the CP and RP was considered to be the very major errors contributed exclusively by two unusual isolates, not a reagent problem (6).

Studies have shown that prolonged incubation does not have a beneficial effect (30), but there have been reports demonstrating significant improvement of test sensitivity for the detection of HLAR (25, 36). In our current study, prolonging the incubation to 48 h increased the detection rate by 6.2% for occurrences of resistance (CP phase). The inability to read trays on the Walk/Away instrument following extended incubation is the major drawback of the PM8 panel. Some studies showed that visual or manual inspection of MicroScan panels can greatly increase the test sensitivity, especially for detection of HLAR (14, 34). Observations at the UICM showed that the Walk/Away instrument failed to read 22 (7.7%) of 285 tests performed during all three phases. The results were obtained only by visual inspection of the trays at 24 and 48 h. Most of the reading failures were due to skipped wells (17 occurrences) and insufficient growth in the positive control well (3 occurrences), despite the normal appearance of the wells during a visual inspection for aminoglycoside synergy and vancomycin susceptibility. The reason for the frequent skipping of wells and for insufficient growth during the performance of this trial was not ascertained, but the Walk/Away software and/or the reader (34) and inoculum concentrations (30) may be responsible.

The genotypic characterization of the tested stains provides an alternative for the detection and characterization of antimicrobial resistances in enterococci. This is a recent concept made possible by the application of DNA hybridization techniques and PCR (4, 13). A PCR assay to detect genotypes of VR and to identify organisms to the species level offers a specific and moderately rapid method for susceptibility testing, in particular for detection of low-level glycopeptide resistance (4, 26). Molecular methods have also been utilized in clinical and surveillance studies (11, 21), and PCR produced excellent results in the EP of this investigation. There have been few reports of using PCR to detect the HLAR genotype in the clinical evaluation of susceptibility testing methods (25). In this study, we used PCR to identify the genotypes for HLAR and VR and to validate all susceptibility testing results from the MicroScan System, as compared to those from the RBM method and synergy screen panels. The use of molecular methods as the "definitive standard" for resistance testing has great appeal; however, there may be (rare) instances where the resistance genotype is not expressed phenotypically. Discovered strains with a resistance-positive genotype but with a susceptible phenotype (e.g., one strain carrying vanA at the UWHC) or with a vanB genotype and a VanA phenotype should be noted, and investigations of the mechanism should be pursued.

In conclusion, this study evaluated the ability of MicroScan PM-8 panels to detect VR and HLAR in enterococci, compared to those of RBM trays and QA plates. Results were further validated by molecular characterization of genotype. The MicroScan PM-8 panels showed a reliability for detecting HLAR that was comparable to that of the RBM method and superior to that of QA in all three study phases. PM-8 panels also demonstrated excellent ability to detect VR of vanA enterococci and high- and moderate-level VR of vanB enterococci. PM-8 was also slightly superior to RBM in the detection of low-level vanB resistance; however, QA had the best ability to detect VR at a low level. The vanC resistances need further clarification as to clinical significance (13, 19, 32), since there remain testing problems because of their usual intermediate levels of vancomycin susceptibility (4 to 16 µg/ml).