Rapid, Accurate Identification of Candida auris by Using a Novel Matrix-Assisted Laser Desorption Ionization–Time of Flight Mass Spectrometry (MALDI-TOF MS) Database (Library)

LETTER

The newly emerging multidrug-resistant yeast Candida auris can cause serious infections and may be underrepresented, as it can be misidentified as other species (e.g., Candida haemulonii, Candida duobushaemulonii, or Saccharomyces cerevisiae) by some biochemical-based testing systems (1–4). Candida auris can be identified using research use only (RUO) libraries on matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) platforms, such as the Biotyper platform (Bruker, Billerica, MA), but may need labor-intensive full-tube extraction procedures (3–5). Our laboratory uses the Biotyper equipped with both FDA-approved and RUO libraries. The RUO library contains three C. auris entries, but the identification of C. auris remains a challenge, and some isolates are never identified by the system. To improve the identification, a novel database, “CMdb,” was developed and evaluated on our two Biotyper systems.

The CMdb was created using internationally collected yeasts from the CDC (6) and one in-house clinical C. auris isolate. Bruker's protocol was used for database creation, and the direct on-plate extraction method was used for target preparation (7). The CMdb was evaluated on 23 clinical C. auris isolates, 20 CDC strains, 52 isolates of 10 other yeast species, and 28 isolates of 16 bacterial species.

The new CMdb database contains 22 mean spectrum projections from C. auris and 4 other related species (C. haemulonii, C. duobushaemulonii, Candida krusei, and Kodameae ohmeri). When we used the CMdb, all 23 clinical C. auris isolates plus the 10 CDC strains were correctly identified (100%) (Table 1); of these isolates, 22 had their identification confirmed by sequencing of the internal transcribed spacer region of their ribosomal DNA (rDNA) (one isolate was lost to contamination, and 22 additional isolates were identified during preparation of the manuscript and are not listed here). The identification log score was consistently greater than 2.40, with an average of 2.50. In comparison, 13 (39%) C. auris isolates were identified by the RUO database, with an average log score of 1.76 (P < 0.001 [R-project.org]), and the rest had log scores below the identification level (1.7). The 4 closely related non-C. auris species were correctly identified by using the same CMdb database. No misidentification was observed with the CMdb when tests were run on other yeasts and bacteria.

When the spectrum-producing targets were counted from C. auris and the 4 other species, all spectra were correctly identified with isolates in the CMdb (100%), while 32% of the spectra were identified with isolates in the RUO library. The low identification rate from the RUO library might have been caused by losing some proteins during its full-tube extraction database creation. Candida auris was well identified from 3 different culture media (Sabouraud dextrose agar [SDA], Columbia nalidixic acid agar, and Trypticase soy agar with sheep blood) but had the highest log scores from isolates grown on SDA. Candida auris was reliably identified on SDA at 20 to 30°C for 3 days.

All 23 clinical isolates (100%) were highly resistant to fluconazole (>256 μg/ml), and two isolates expressed multidrug resistance (Table 2).

The identification of C. auris with this novel CMdb was accurate (100%), quick, and easy, with significantly higher log scores, which will benefit patient care and public health interventions.