Identification of Candida auris by Use of the Updated Vitek 2 Yeast Identification System, Version 8.01: a Multilaboratory Evaluation Study

INTRODUCTION

Candida auris was first isolated from an external ear specimen in Japan in 2009 (1). This microorganism distinguishes itself from other Candida species by its ability to acquire resistance to multiple antifungal classes, as well as by its propensity for transmission within hospital facilities (2–6). The latter feature has led to outbreaks of hospital-acquired invasive C. auris infection with high morbidity and mortality (7, 8). These outbreaks have occurred worldwide and have been associated with simultaneous emergence of four distinct genetic clades across different countries (9, 10). C. auris is now recognized as a major global health concern (11). Therefore, it is essential that laboratories be able to rapidly and accurately identify C. auris from clinical specimens in order to institute proper antifungal therapy and infection control measures (12–14). In this context, identification systems are brought to the fore.

Unfortunately, identification of C. auris in the microbiology laboratory using standard methods has proven to be difficult (5). Standard methods involve panels of tests that can include a combination of enzyme colorimetric tests, assimilation tests, and growth tests using carbohydrates and nitrogen compounds. Automated commercial identification systems rely on disposable cards that contain several conventional biochemical tests incorporated into microtiter wells used to identify a microorganism. These systems do not include C. auris in their databases and thus may incorrectly identify C. auris as other yeast species (5, 15). Automated identification systems have reported incorrect identification of C. auris isolates as Candida haemulonii, Candida duobushaemulonii, Candida famata, and Rhodotorula glutinis (16, 17). Phylogenetic studies show close relatedness between C. auris, C. duobushaemulonii, and C. haemulonii (1, 18). In 2017, the Vitek 2 identification software was updated to its 8.01 version, which included the addition of the following four taxa: C. auris, C. duobushaemulonii, C. haemulonii var. vulnera, and Cryptococcus gattii. To our knowledge, the performance of this updated version for identification of C. auris has not been independently evaluated.

MATERIALS AND METHODS

The study was conducted in the following three hospital-based microbiology laboratories in Montreal, Canada: Centre Hospitalier de l’Université de Montréal (lab A), Hôpital Maisonneuve-Rosemont (lab B), and Centre Hospitalier Universitaire Sainte-Justine (lab C).

A panel of 44 isolates of C. auris and related species was used to validate the Vitek 2 automated system. The panel included 35 isolates of C. auris encompassing the four major clades, 5 strains of C. haemulonii, and 4 strains of C. duobushaemulonii. Most (97%, 34/35) C. auris isolates were obtained from the Centers for Disease Control and Prevention (CDC). The isolates were centralized at the Provincial Public Health Laboratory (reference laboratory) and subsequently distributed to each participating hospital laboratory as frozen aliquots of a yeast suspension. Aliquots were preserved in 10% glycerol and stored at −70°C. Upon thawing, isolates were subcultured twice before testing with a Vitek 2 YST ID card. Per local practices, culture was done on Sabouraud dextrose agar (Oxoid) at lab A and B, while inhibitory mold agar (Oxoid) was used at lab C. Identification with Vitek 2 was performed according to the manufacturer’s instructions, using the latest software update of the Vitek 2 identification system (version 8.01). Results were compared to those from sequence-based identification. Correct identification was defined as the Vitek 2 yielding a single taxon concordant (with high confidence) with reference identification. Identification was considered incorrect when the result consisted of a single taxon not matching that identified by the reference method or a low-discrimination (LD) result that did not include the true species identification among possible taxa. Results were considered to be LD when the correct species was included among possible taxa.

All 44 isolates were previously identified by ribosomal DNA (rDNA) sequencing of internal transcribed spacer (ITS)/D1-D2 regions and by matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF) (Vitek mass spectrometer [MS], clinical knowledge database v3.2; bioMérieux), which were 100% concordant. C. auris genetic clade determination was performed by whole-genome sequencing at the CDC (Mycotic Diseases Branch).

Orthologs of the genes from the l-rhamnose catabolic pathway (RHA1, LRA1, LRA2, LRA3, and TRC1) of Scheffersomyces stipites and Aspergillus niger l-rhamnose clusters (19, 20) were identified using a BLAST search against the genomes of C. auris strains representing each clade (B8441, B11221, B11220, and B11243) and those of C. haemulonii, C. duobushaemulonii, and Candida pseudohaemulonii (18).

RESULTS

Correct identification of the isolates from our panel with the Vitek 2 system was calculated at an average of 61% of tested samples. The proportion of correct identification was not statistically different across different centers (chi-square test; P = 0.78). Detailed identification results for each center are presented in Table 1.

DISCUSSION

To our knowledge, this is the first multicenter study to evaluate the performance of the Vitek 2 (8.01 software version) to identify C. auris and closely related species. Our results show that Vitek 2 was able to correctly identify 52% of tested samples of C. auris isolates with a high level of confidence. There were no false-positive results of C. auris that were reported by Vitek 2 among the non-auris Candida strains. Thus, C. auris identification as reported by Vitek 2 appeared to be reliable. This finding is consistent with the CDC’s current identification algorithm, which indicates that no further testing is needed if a C. auris identification is obtained by the Vitek 2 8.01 system (21).

However, our results demonstrated that Vitek 2 had limited ability to distinguish between C. auris and C. duobushaemulonii. This finding highlights the problematic distinction between C. auris and closely related species. C. auris is genetically related to C. haemulonii, C. haemulonii var. vulnera, and C. duobushaemulonii (5, 22). Recent taxonomic studies have reclassified C. haemulonii, C. haemulonii var. vulnera, and C. duobushaemulonii under the same C. haemulonii complex. In the past, distinction between these closely related species has been difficult. Ramos et al. reported a high rate of incorrect identification among these closely related species with the Vitek 2 software version prior to 8.01; among 36 isolates identified as C. haemulonii by Vitek 2, 47% were C. duobushaemulonii, 22% were C. haemulonii, and 14% were C. auris when identified by MALDI-TOF and genome sequencing (23). In the prior Vitek 2 software version, the majority of isolates suspected to be C. auris were identified as C. haemulonii (5). It is unclear if strains incorrectly identified as C. haemulonii in the prior Vitek 2 software version were in fact C. duobushaemulonii strains. The updated version of Vitek 2 provides reliable distinction between C. auris and C. haemulonii, as none of the C. auris or C. duobushaemulonii isolates were incorrectly identified as C. haemulonii. While the updated version of Vitek 2 (8.01 software version) demonstrates improved discriminatory ability for C. haemulonii, its performance remains suboptimal for distinction between C. auris and C. duobushaemulonii. Taken together, our results suggest that a C. duobushaemulonii identification result obtained by Vitek 2 warrants further testing to determine whether it is in fact C. duobushaemulonii or C. auris. Epidemiology of C. duobushaemulonii is not well studied, but cases of vulvovaginitis, wound and cutaneous infections, and invasive infections have been reported in the literature (23). A large survey in Latin America reported that C. haemulonii accounted for less than 0,01% of candidemia episodes at a time when C. duobushaemulonii was not yet described (24). That rate probably overestimates the real prevalence of C. duobushaemulonii but gives a general idea that the prevalence might still be low. In a large descriptive Brazilian study, 14,642 clinical samples that tested positive for yeast culture were collected between 2010 and 2015. C. haemulonii complex accounted for 0.3% of all of the yeast-positive samples, with 32% of C. haemulonii complex samples being blood cultures (25). The SENTRY Antifungal Surveillance Program reported that among 15,308 isolates of Candida species collected worldwide between 2006 and 2016, only 6 isolates were C. auris (26). Similarly to C. auris, C. duobushaemulonii can present high level resistance to azole and amphotericin B (18). Whether C. duobushaemulonii can be transmitted between patients and cause health care outbreaks yet remains to be seen. Hence, the distinction between C. auris and C. duobushaemulonii may have an impact on infection prevention and control measures (23). While the majority of LD results observed in our study were due to difficulty differentiating between C. auris and C. duobushaemulonii, two LD results for C. auris/C. famata and two LD results for C. duobushaemulonii/C. lusitaniae were observed. Previous studies have reported incorrect identification of C. auris as C. famata as well as C. lusitaniae (5, 15). Interestingly, phylogenetic relatedness between C. auris and C. lusitaniae was demonstrated by whole-genome sequencing (18, 27). Therefore, our results suggest that whenever Vitek 2 provides a LD result with C. auris or C. duobushaemulonii and either C. famata or C. lusitaniae, additional testing should be required to rule out a true C. auris isolate.

Lockhart et al. reported that isolates can be grouped into unique clades according to geographic regions. Clades groups were divided as follows: Venezuela (South America), South Africa (Africa), Japan (East Asia), and India and Pakistan (South Asia) (11). An epidemiological survey of 73 clinical isolates of C. auris collected from ten U.S. states reported that 90% of strains belonged to the South Asian clade. Isolates from the three other clades were also reported in U.S. territory but at a lower frequency, which ranged from 1% to 7% (28). Stratification of C. auris by phylogenetic clades demonstrated differences in accurate identification. All strains of the South American clades were correctly identified; strains of the South Asian clade gave mixed results, with correct identification ranging between 54 and 92%, and strains from the African and East Asian clades yielded high rates of LD results or incorrect identifications. Furthermore, all isolates from the East Asian clade were repeatedly identified incorrectly as C. duobushaemulonii. Biochemical profile analysis demonstrated that assimilation of l-rhamnose was specific to the African clade, while the majority of the other clades did not show assimilation. An l-rhamnose gene cluster appears to correlate with these data, as this cluster is absent in the genomes of isolates from the East Asian, South American, and South Asian clades. Similarly, strains of the East Asian clade did not show assimilation of NAG. Despite the fact that no gene cluster has been identified yet, the biochemical difference between the East Asian clade and other clades could explain its high rate of incorrect identification.

Many automated systems that utilize phenotype-based identification methods to identify yeast species failed to correctly identify C. auris. Vitek 2 updated its database in order to improve the correct identification rate of C. auris; however, our study did not show reliability (21). Therefore, alternative methods of identification were evaluated to answer the need to adequately identify C. auris. Gene sequencing and MALDI-TOF technology accurately and reliably identify C. auris (17). However, their widespread use is limited by their high acquisition cost and personnel expertise requirement, and thus their use is limited to that by reference laboratories for confirmation of suspected cases of C. auris (5, 15, 21). Molecular assays have been developed to replace culture-based identification techniques . Leach et al. developed a real-time singleplex PCR assay for detection of C. auris from environmental samples and reported sensitivities of 89% and 100% from swab and sponge samples, respectively (29). In addition, different multiplex quantitative PCRs with the ability to detect C. auris and related species have also been developed (30–32). As such, PCR assays show potential to replace culture-based technique; however, further studies are required to validate the performance of these molecular assays in routine clinical practice.

Our study has several limitations. First, the small number of C. auris isolates limits the power of our analysis. Given the rarity of this new emerging pathogen, the number of available strains was very small. Nevertheless, we obtained strains from the CDC to increase our sample size. Second, the limited number of non-auris Candida (C. haemulonii complex) strains hampered our ability to accurately assess the discriminatory ability of the Vitek 2 version 8.01, potentially resulting in underestimation of the rate of LD or incorrect identification results. Nevertheless, similar results obtained from all three centers demonstrated high concordance, suggesting that our results were reproducible. Each isolate from the panel was tested in triplicate with the Vitek 2 at each center, and no discrepancies were noted among the repeated tests, thus demonstrating the precision and reproducibility of the technology.

In summary, our study shows that the Vitek 2 (software version 8.01) appears to correctly identify about half of C. auris isolates and that such identification with a high level of confidence is accurate and reliable. However, the system’s ability to discriminate between C. auris and C. duobushaemulonii remains low, with a high proportion of LD results or incorrect identification as C. duobushaemulonii. Thus, the identification of C. duobushaemulonii at any level of confidence or an LD result between C. auris and another species should trigger additional testing to rule out C. auris. Although the number of isolates in each clade was limited, Vitek 2 performance appears to differ markedly according to C. auris genetic clades, with acceptable accuracy for isolates from South America clade but poor accuracy for isolates for the East Asian and African clades. Based on our findings, pattern reactions within the YST card could be assessed to find out if erroneous results yielded as C. duobushaemulonii should be instead listed as LD results between C. auris and C. duobushaemulonii. Further studies on the biochemical profiles according to each clade would be desirable.

Supplementary Material