ABSTRACT
We evaluated the use of matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) for the rapid identification of yeast species. Using Bruker Daltonics MALDI BioTyper software, we created a spectral database library with m/z ratios of 2,000 to 20,000 Da for 109 type and reference strains of yeast (44 species in 8 genera). The database was tested for accuracy by use of 194 clinical isolates (23 species in 6 genera). A total of 192 (99.0%) of the clinical isolates were identified accurately by MALDI-TOF MS. The MALDI-TOF MS-based method was found to be reproducible and accurate, with low consumable costs and minimal preparation time.
Invasive fungal infections due to opportunistic pathogens are a significant cause of morbidity and mortality (2, 5, 8). The current rise in fungal infections correlates with the widespread use of broad-spectrum antibacterial agents, prolonged hospitalization of critically ill patients, and the increased number of immunocompromised patients. Candida species comprise the fourth most common cause of nosocomial bloodstream infections, and Cryptococcus neoformans is the most common cause of fungus-related mortality in HIV-infected patients (15, 19). While Candida albicans is still involved in more than half of all Candida-related bloodstream infections, an increase in recovery of non-C. albicans Candida spp., Rhodotorula spp., Trichosporon spp., and Malassezia spp. has occurred (2, 29). Treatment with amphotericin B may be useful for these organisms and inefficient for those belonging to other genera (5, 8). While many Candida species remain susceptible to fluconazole, it is important to differentiate the more resistant organisms, namely, Candida glabrata, Candida krusei, Rhodotorula spp., and some members of the genus Trichosporon. Additionally, Rhodotorula spp. have an innate resistance to voriconazole, and Trichosporon, Cryptococcus, and Rhodotorula are intrinsically resistant to the echinocandins (1, 15). These organisms present new challenges not only to treatment but also to standard identification methods used in the clinical laboratory (4, 8, 28).
Commercially available biochemical test systems identify most of the commonly isolated species of yeast accurately but may result in no identification or misidentification of more-unusual isolates (4, 21, 28). Additionally, samples for these tests must be incubated for 1 to 3 days before results are obtained. To overcome the inaccuracies of biochemical identification methods, nucleic acid-based tests have been developed. These tests amplify and then sequence a target gene, such as the rRNA genes or the internal transcribed spacer (ITS) region (9, 10, 14, 17). While these assays are highly accurate, they require considerable processing time and costly reagents.
As an alternative to biochemical and genome-based identification schemes, proteomic profiling by mass spectral analysis was recently evaluated for use in species differentiation of a variety of microorganisms. Matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF) is emerging as a rapid and accurate tool for identifying pathogens, including Gram-positive and Gram-negative bacteria, mycobacteria, molds, and yeast species (3, 6, 11-13, 16, 18, 22, 23, 27). The technique can be performed rapidly, with minimal consumable expenses, and produces reproducible, species-specific spectral patterns that are not dependent upon the age of culture, growth conditions, or medium selection (7, 13, 20, 26).
In this work, we present the development of a yeast database library consisting of 109 type and reference strains (44 species in 8 genera), and we tested the robustness and accuracy of this library by using 194 well-characterized clinical isolates (23 species in 6 genera).
MATERIALS AND METHODS
Type and reference strains and clinical challenge isolates.In this study, type and reference strains from the American Type Culture Collection (ATCC), the University of Alberta Microfungus Collection and Herbarium (UAMH), and the National Institutes of Health (NIH) were used to construct the yeast database library. The accuracy of this test library for MALDI-TOF identification of clinical isolates of yeast was then evaluated with 194 isolates collected at the NIH Clinical Center (initially identified using the API 20C AUX yeast identification system [bioMerieux, Durham, NC]) or provided by M. Pfaller and D. Diekema (identified using Vitek 2 yeast cards [bioMerieux, Durham, NC]). For all organisms, isolates were cultured from frozen stock onto Sabouraud dextrose agar and incubated for 2 to 3 days at 30°C before being tested. Discrepancies in identification of reference strains and clinical isolates were resolved by Sanger sequencing of the ITS region (ITS1-5.8S-ITS2) or the secondary alcohol dehydrogenase gene (14, 24, 25).
Sample preparation.A single colony or multiple small colonies (enough to fill one-third of a 10-μl inoculating loop) were suspended in 1 ml of 70% ethanol (EtOH), briefly vortexed, and centrifuged in an Eppendorf 5415D centrifuge at 13,000 rpm for 2 min. The liquid was removed, and the pellet was quickly spun again, followed by removal of residual EtOH. The pellet was then resuspended in 50 μl 70% formic acid, 50 μl of acetonitrile was added, the sample was vortexed briefly, and the mixture was centrifuged for 2 min at 3,000 rpm. The supernatant was either tested immediately or stored at −20°C until ready for MALDI-TOF analysis. A portion of each extracted sample was subcultured onto Sabouraud dextrose agar and incubated at 30°C to ensure that no viable organisms remained.
MALDI-TOF MS.The yeast sample to be analyzed was warmed to room temperature, 1 μl was spotted onto a steel target plate (Bruker Daltonics Inc., Billerica, MA) and allowed to evaporate to dryness on a 37°C plate warmer (model XH-2002 Premiere slide warmer; Daigger, Vernon Hills, IL), and then 2 μl matrix solution was applied to the spot and dried. The MALDI matrix, α-cyano-4-hydroxycinnamic acid, was prepared daily as a saturated solution in 50% acetonitrile and 2.5% trifluoroacetic acid. After complete drying, the mixture was exposed to laser pulses, resulting in energy transfer from the matrix to the nonvolatile analyte molecules, with desorption of analyte into the gas phase. The ionized molecules were accelerated by electrical potentials through a flight tube to the mass spectrometer, with separation of the biomarkers determined by their mass/charge ratio (m/z; z typically is 1). The profiles of biomarkers were then compared with profiles of a collection of well-characterized organisms. Each sample was assayed in quadruplicate.
Samples were evaluated on an UltraFlex I TOF-TOF mass spectrometer (Bruker Daltonics Inc.) in linear positive-ion mode across the m/z range of 2,000 to 20,000, with gating of ions below m/z 400 and a delayed extraction time of 450 ns. For these studies, each spot was measured using 1,000 laser shots at 25 Hz in groups of 50 shots per sampling area of the spot. The data sampling rate was 0.5 GHz. Each plate was calibrated before being run, using a mixture of protein standards I and peptide standards II from Bruker Daltonics. Total processing time and spectral analysis took approximately 15 min for a single isolate, and additional samples increased this time by 1 to 2 min per sample.
Spectral analysis by Bruker MALDI BioTyper.Acceptable spectra for reference strains were defined by Bruker Daltonics as having a minimum of 25 peaks with a resolution of >400, among which 20 peaks must have a resolution of >500, where resolution is defined as the peak m/z divided by the peak width at 50% of the peak maximum. Spectra of clinical challenge isolates were analyzed using the reference strain database and MALDI BioTyper software (BioTyper Library v 2.0.4; Bruker Daltonics), a proprietary algorithm for spectral pattern matching resulting in a logarithmic score of 0 to 3. As part of this study, we assessed the minimum spectral score that accurately predicted species identification of all yeast isolates.
RESULTS
A total of 124 type and reference strains were evaluated for construction of the yeast database library. Acceptable spectra were not obtained with 5 reference organisms, that is, insufficient numbers of quality peaks were observed for inclusion of the reference organisms in the database library. Additionally, identifications of 10 organisms by MALDI-TOF and gene sequencing were concordant but differed from the identifications assigned by ATCC or UAMH (Table 1). Therefore, these 15 organisms were eliminated, and the final yeast database library consisted of 109 strains representing 44 species in 8 genera (Table 2).
Reference strains removed from the database library due to discordant identification results or unacceptable spectra
Reference strains and gene sequence-confirmed NIH clinical isolates evaluated for inclusion in the MALDI-TOF MS fungal database library
A set of 197 yeast clinical isolates, previously identified by biochemical analysis, was used to challenge the spectral library. A limited number of commonly isolated yeast and a broad spectrum of less common yeast were evaluated. Three organisms identified as Candida famata by biochemical tests were identified by gene sequencing as Pichia caribbica (2 isolates) or Pichia mississippiensis (1 isolate). These organisms are not in the database and gave no acceptable MALDI-TOF spectra, so the organisms were excluded from this analysis. Table 3 depicts the results of tests with the remaining 194 yeast isolates (23 species in 6 genera). Two organisms had MALDI-TOF spectral scores of <1.8 (for Candida rugosa, 1.5; and for C. neoformans, 1.3); all other organisms had scores of ≥1.8. Of the 192 organisms with scores of ≥1.8, the MALDI-TOF identification and initial identification were in agreement for 165 isolates, with 27 discordant identifications. Sequencing of the ITS region was used to resolve the discrepancies, with MALDI-TOF and sequencing results being in agreement for all 27 isolates. Thus, MALDI-TOF correctly identified 192 (99.0%) isolates, and all spectral scores of ≥1.8 accurately predicted the species identity.
Identification of clinical isolates by MALDI-TOF MS
During this study, the clinical isolates were assayed in quadruplicate, with each protein extract applied to 4 spots on the steel target plate. This allowed us to determine the number of assays per isolate that needed to be performed for acceptable identification. Of the 192 correctly identified species, the spectral score was ≥1.8 with the first spot for 172 (89.6%) isolates, with the first two spots for 187 (97.4%) isolates, and with the first three spots for 189 (98.4%) isolates. Low MALDI-TOF scores (i.e., <1.8) requiring repeat testing were most commonly observed with Candida guilliermondii, C. krusei, Candida lusitaniae, and C. neoformans.
DISCUSSION
In this work, we evaluated the use of MALDI-TOF analysis for the identification of clinically important yeast species. Traditional biochemical testing for yeast identification is inexpensive but time-consuming and sometimes inaccurate (4, 21, 28). Gene sequencing is highly accurate but is expensive, time-consuming, and technically demanding (17). The aim of this study was to determine if MALDI-TOF would be suitable for rapid yeast identification. Working with 109 reference and type strains representing 44 species in 8 genera, we created a MALDI-TOF spectral library and then challenged the library with 197 clinical isolates. Three clinical isolates gave no spectral score and could not be identified because they were not in the spectral library. Two organisms (C. rugosa and C. neoformans) gave consistently low spectral scores and could not be identified. Thus, of the 194 isolates (23 species in 6 genera) included in the spectral library, 192 (99.0%) were identified correctly at the species level by MALDI-TOF. These results compare favorably with the studies of Marklein et al. (12) and van Veen et al. (27). Marklein et al. (12) evaluated 18 reference strains representing 8 species of yeast and 267 clinical isolates (28 species in 7 genera). All reference organisms and 92.5% of their clinical isolates were identified correctly. The organisms that were not identified were not in the database library. When appropriate reference organisms were added to the database, all clinical isolates were identified correctly. van Veen et al. (27) evaluated the accuracy of MALDI-TOF for the identification of 80 yeast isolates (14 species in 7 genera). A total of 97.5% of isolates were identified correctly at the genus level, with 87.5% identified correctly at the species level, with most of the cases of unidentified organisms being due to insufficient entries in the database and to species-level identification being accepted only if the MALDI-TOF spectral score was >2.0. If we applied the same scoring criterion to the isolates tested in the current study, then the same proportion of clinical isolates (87.1%) would be identified correctly at the species level (Table 3); however, as stated above, spectral scores between 1.8 and 2.0 were consistently accurate for identification of yeast species in the present study. In the current study and earlier reports (12, 27), organisms not identified by MALDI-TOF resulted from (i) a low spectral score due to an insufficient number of peaks or (ii) the absence of the organism in the database. In contrast to the case for biochemical identification tests, no yeast isolate in any of these studies was misidentified as another species.
For studies validating a new identification system, it is important that the database be constructed with a comprehensive collection of accurately identified reference strains. Although we selected type and reference strains from the American Type Culture Collection and the University of Alberta Microfungus Collection, 11 ATCC reference organisms were identified incorrectly, as determined by gene sequencing (Table 1). Ten of these organisms were identified correctly by MALDI-TOF, with no acceptable spectra obtained for one isolate, Cryptococcus luteolus (Cryptococcus podzolicus) ATCC 42279. No acceptable MALDI-TOF spectra were obtained for four additional organisms (C. luteolus ATCC 42279, Geotrichum candidum ATCC 10834 and ATCC 74169, and Malassezia sympodialis ATCC 96803). This was most likely due to the rigorous standards used for inclusion of a reference strain in the MALDI-TOF database (refer to Materials and Methods), because both clinical isolates of G. candidum were identified correctly.
Spectra of clinical isolates were analyzed using the reference strain database and MALDI BioTyper software (BioTyper Library v 2.0.4; Bruker Daltonics), a proprietary algorithm for spectral pattern matching resulting in a logarithmic score of 0 to 3. The results reported in this study demonstrate that all organisms identified at the species level had a spectral score of 1.8 or greater. Five clinical isolates were not identified in this study: two organisms had unacceptably low spectral scores, 1.3 and 1.5, and three organisms had spectral scores of 0 because the species were not in the reference library database.
All isolates were tested in quadruplicate, that is, each sample preparation was spotted four times on the steel target plate. Because 89.6% and 97.4% of the isolates were identified with the first spot and the first two spots, respectively, we believe that testing should routinely be performed in duplicate, with retesting in duplicate if the spectral scores are <1.8. This is a practical approach to integrating this technology into the routine laboratory because there are no additional reagent costs for duplicate testing, the analysis of individual spots requires less than 5 min, and the steel plates are washed and reused.
In summary, use of MALDI-TOF mass spectrometry for the identification of clinically important yeast is rapid and accurate. Although the instrument cost is significant (a Bruker Daltonics Microflex LT mass spectrometer presently costs approximately $110,000), the reagent cost per test is pennies, and there are no consumable costs. Additionally, we have found that this technology can be used to identify a broad spectrum of organisms, including virtually all Gram-positive and Gram-negative bacteria, Nocardia and related Actinomycetes bacteria, mycobacteria, and fungi. Integration of this technology into the routine workflow of the laboratory allows us to identify most bacterial and fungal organisms within a few hours of initial detection and to eliminate the need for biochemical testing or to perform gene sequencing for the identification of most problem organisms.
ACKNOWLEDGMENTS
This research was supported by the Intramural Research Program of the NIH Clinical Center, Department of Laboratory Medicine.
FOOTNOTES
- Received 3 April 2009.
- Returned for modification 26 May 2009.
- Accepted 30 June 2010.
- Copyright © 2010 American Society for Microbiology