Apparent Increased Prevalence of High-Level Aminoglycoside-Resistant Enterococcus durans Resulting from False Identification by a Semiautomated Software System

Abstract

Identification of enterococci by using a semiautomated system (PASCO; Difco Laboratories, Detroit, Mich.) in the AHEPA University Hospital, Thessaloniki, Greece, revealed a high proportion of Enterococcus durans, particularly among isolates highly resistant to gentamicin and streptomycin. When 14 isolates were further tested by a conventional biochemical scheme and by PCR, all were reidentified as Enterococcus faecalis. Antibiotic resistance and pulsed-field gel electrophoresis patterns showed that unrelated strains were misidentified.

Enterococci are significant nosocomial pathogens and have the capacity to develop and transfer antimicrobial resistance. High-level aminoglycoside-resistant enterococci (HLARE) have been increasingly reported from hospitalized patients worldwide (10, 17). The situation is more complicated when high-level resistance to both gentamicin and streptomycin is identified. In such cases, the synergic combination of a cell wall-active agent plus an aminoglycoside cannot be used for the treatment of severe enterococcal infections (7).

High-level aminoglycoside resistance was first described among Enterococcus faecalis strains, but recently Enterococcus faecium strains resistant to high levels of streptomycin as well as to gentamicin have been increasingly reported (12, 19). These two species account for most infections due to HLARE, while other species generally account for less than 5% in most reports (5, 11, 15). However, this percentage may be higher since most conventional methods do not identify uncommon enterococcal species. Previous data from our region have shown that, among clinical isolates of enterococci, 21% exhibited high-level resistance to gentamicin and 26.3% exhibited high-level resistance to streptomycin (1). During 1995 and 1996, we have noted a significant increase in these proportions, the respective figures being 40.2% and 63%, and in many cases resistance to both aminoglycosides was observed. Identification of these isolates by using the semiautomated PASCO MIC/ID gram-positive panel (Difco Laboratories) revealed that high-level aminoglycoside resistance was detected more frequently among isolates of Enterococcus durans; 47% of isolates obtained during 1995 and 1996 were resistant to both aminoglycosides tested. After this observation we evaluated the accuracy of E. durans identification reported by the PASCO system.

Between November 1996 and April 1997, 14 isolates of enterococci identified as E. durans by the PASCO system were further evaluated for the accuracy of the identification. These isolates were nonrepetitive and were from clinical material (four from urine, three from blood, and seven from wounds) in the Clinical Microbiology Laboratory at the AHEPA University Hospital.

Enterococci were isolated from blood agar and identified by the catalase test, the bile esculin test (prepared in-house), growth in 6.5% NaCl broth, and the PASCO identification system. Identification by the automated system was based on the following test reactions: arabinose, cellobiose, lactose, mannitol, ribose, and sucrose fermentation; urea hydrolysis; arginine; esculin; phosphatase; α-glucopyranoside; β-glucuronide; Voges-Proskauer; pyroglutamate; optochin; 6.5% NaCl; bacitracin; novobiocin; catalase; and hemolysis. Antimicrobial susceptibility tests were also performed using the same system with a broth microdilution method by using Mueller-Hinton broth and a final inoculum of approximately 10⁵ CFU/ml. Interpretive criteria for susceptibility status were those of the National Committee for Clinical Laboratory Standards (9). Strains identified by the PASCO system as E. durans were further tested with the conventional test scheme proposed by Facklam and Collins (3), by using key tests prepared in-house for the discrimination of the three major groups of enterococci and the identification of the species within the groups. Additional biochemical tests (ribose fermentation and tellurite reduction) were performed on lactose-negative isolates as previously mentioned (11), in order to distinguish E. faecalis from Enterococcus solitarius. All isolates were further tested with a PCR system that can identify four enterococcal species (E. faecalis, E. faecium, Enterococcus gallinarum, and Enterococcus casseliflavus), while the primer pairs do not amplify DNA from the other species (2).

For 18 isolates (9 identified as E. durans described above and 9 identified as E. faecalis by the PASCO system), all randomly selected, macrorestriction of bacterial genomic DNA was performed by using SmaI. Pulsed-field gel electrophoresis (PFGE) separation was performed with a contour-clamped homogeneous electric field Mapper II system (Bio-Rad, Hemel Hempstead, Hertfordshire, United Kingdom) as previously described (8). Banding patterns were analyzed with the aid of GelCompar software (Applied Maths, Kortrijk, Belgium). The Dice correlation coefficient was applied, and a dendrogram of percent similarity was produced by using the unweighted-pair-group-matching-by-arithmetic-averages algorithm.

During the 5-month study period, 101 enterococci were isolated, 59 (58.4%) of which were resistant to high levels of at least one of the two aminoglycosides gentamicin and streptomycin and 38 (37.6%) of which were resistant to both of them. Fourteen (13.9%) isolates were identified as E. durans by the PASCO system, while 69 (68.3%) isolates were identified as E. faecalis. Among the isolates identified as E. durans, 10 (71.4%) were resistant to at least one aminoglycoside and 7 (50%) were resistant to high levels of both aminoglycosides. The corresponding figures for the isolates identified as E. faecalis were 38 (55.1%) and 24 (34.8%), respectively.

Tests for susceptibility to 15 other antimicrobials showed 11 different antibiotic resistance profiles among the 14 isolates identified as E. durans. When the latter isolates were tested with the conventional scheme, all were identified as E. faecalis. Five of these isolates were lactose negative and according to the scheme proposed by Facklam and Collins (3) would be identified as E. solitarius. However, when the additional biochemical tests of tellurite reduction and ribose fermentation were performed, all 14 isolates were actually identified as E. faecalis. The remaining biochemical reactions in the conventional scheme were similar for all isolates tested, although five were nonreactive with the group D antigen. Further analysis by the PCR method supported the results of the conventional identification, as each isolate gave a specific product corresponding to E. faecalis (Fig. 1). Analysis of macrorestriction fragment length polymorphisms by PFGE revealed that the misidentified isolates did not form a genetically homogeneous group; they belonged to three different clusters, and each cluster contained both correctly identified and misidentified isolates (Fig. 2).

FIG. 1.

PCR analysis of DNA from nine representative isolates identified as E. durans by the PASCO system. Lanes: 1, E. faecalis control; 2, E. faecium control; 3, E. gallinarum control; 4, E. casseliflavus control 24788; 5, negative (water) control; 6, 123-bp ladder (Life Technologies, Paisley, United Kingdom); 7 to 15, isolates identified as E. durans by the PASCO system.

FIG. 2.

Dendrogram of percent similarity of banding patterns generated by PFGE separation of SmaI macrorestriction fragments of genomic DNA of nine E. durans (D) and nine E. faecalis (F) isolates, as identified by the PASCO system.

Emerging drug resistance of the enterococci necessitates a precise identification at the species level. This is more relevant when an outbreak of HLARE has been established, such as in our hospital environment. Accurate identification to species level may facilitate infection control attempts in order to identify correctly the source of the outbreak. However, many species of enterococci present similar biochemical reactions, and incidence of uncommon species may not be correctly estimated (16).

E. durans is one of several species of enterococci that are only reported sporadically in clinical infections and according to the recent literature has not been implicated in clusters of infection or hospital outbreaks (4). Thus, an unexpected high frequency of this species needs further evaluation by alternative methods, such as standard biochemical tests and gene analysis, in order to estimate the accuracy of the system. Our hospital uses a version of the PASCO system with recently updated software, but the system can only discriminate four species of enterococci (E. faecalis, E. faecium, E. durans, and Enterococcus avium) by comparison with an inbuilt database. It is of interest that although some of the biochemical reactions contained in the automated panel (such as mannitol and sucrose fermentation) were typical of E. faecalis, the software analysis characterized the isolates as E. durans based on some reactions (e.g., α-glucopyranoside and hemolysis) which are not mentioned in the conventional scheme of Facklam and Collins (3). Furthermore, isolates identified as E. durans by the PASCO system exhibited various antibiotic resistance patterns and belonged to different PFGE clusters, which suggests that genetically unrelated strains had been misidentified.

It has been shown that E. durans is phylogenetically closer to E. faecium and Enterococcus hirae and that there is a low rRNA homology between E. faecalis and other enterococci, setting the latter species apart from the rest of the genus (18). Thus, in a recent report, isolates misidentified as E. durans by a version of the Vitek card system were identified as E. faecium when a conventional identification system was used (13). Also, when the MicroScan identification system was used, two E. durans isolates were reidentified as E. hirae by a conventional scheme (14). In similar cases of misidentification, the PCR method of identification to species level has shown that isolates sent to a reference laboratory as E. durans were actually E. faecium, several of which had atypical biochemical profiles and were arabinose negative (6). The present report describes the first case where isolates misidentified as E. durans by an automated system gave biochemical profiles and specific products by PCR equivalent to E. faecalis. It is recommended that laboratories using the PASCO system inspect by naked eye the typical reactions of the enterococcal species. Also, the database of the system should be based more on these typical reactions.

Errors in automated identification systems are not easily detected, as species identification is based on an inbuilt database. Misidentification of isolates with properties typical of another species suggests the need for modification of the database of the system. Also, additional panel tests would probably enable the system to discriminate more clinically relevant Enterococcus species.