ABSTRACT
Rapid identification of yeast species isolates from clinical samples is particularly important given their innately variable antifungal susceptibility profiles. Here, we have evaluated the utility of pyrosequencing analysis of a portion of the internal transcribed spacer 2 region (ITS2) for identification of pathogenic yeasts. A total of 477 clinical isolates encompassing 43 different fungal species were subjected to pyrosequencing analysis in a strictly blinded study. The molecular identifications produced by pyrosequencing were compared with those obtained using conventional biochemical tests (AUXACOLOR2) and following PCR amplification and sequencing of the D1-D2 portion of the nuclear 28S large rRNA gene. More than 98% (469/477) of isolates encompassing 40 of the 43 fungal species tested were correctly identified by pyrosequencing of only 35 bp of ITS2. Moreover, BLAST searches of the public synchronized databases with the ITS2 pyrosequencing signature sequences revealed that there was only minimal sequence redundancy in the ITS2 under analysis. In all cases, the pyrosequencing signature sequences were unique to the yeast species (or species complex) under investigation. Finally, when pyrosequencing was combined with the Whatman FTA paper technology for the rapid extraction of fungal genomic DNA, molecular identification could be accomplished within 6 h from the time of starting from pure cultures.
Invasive fungal infections caused by Candida spp. remain a significant cause of morbidity and mortality in immunocompromised patients and those undergoing invasive procedures (19, 22, 26). While most studies agree that Candida albicans is the principal agent of nosocomial yeast infections, more than 150 yeast species from Candida and other genera have now been reported from mammalian infections (9), with C. glabrata, C. parapsilosis, C. tropicalis, C. lusitaniae, and C. krusei emerging over recent years to be significant opportunistic pathogens (16, 18, 23, 25). It is now well established that the patterns of antifungal susceptibilities vary substantially between different Candida species (20, 21). In addition, the widespread use of antifungal agents has been postulated to have contributed to a shift in species distributions via the emergence of inherently resistant species as significant pathogens (12, 13). Thus, informed therapeutic decisions frequently require the correct and rapid identification of an ever increasing number of potential pathogens.
The commercial systems available for the conventional identification of pathogenic yeasts rely upon a combination of morphological and biochemical features, employ tests that are time-consuming, and are designed to identify only the more common pathogens. Indeed, previous studies have demonstrated that such methods frequently fail to identify the less common pathogens and those common organisms that are misbehaving (1, 15) or to discriminate between closely related species (15, 27). Molecular identification, employing PCR amplification and sequencing of genomic regions that evolve slowly and that show high degrees of conservation, represents a rapid and sensitive alternative to conventional identification for yeasts and is also useful for the establishment of phylogenetic relationships. Among the genomic regions targeted, the nuclear rRNA gene cassette (and, notably, the internal transcribed regions [ITSs] and the D1-D2 portion of the 28S large rRNA gene) have previously proved sufficient to discriminate between the majority of species of clinically important yeasts (7, 8, 15). However, such approaches are still costly and cannot easily be performed more quickly than in vitro antifungal susceptibility testing.
Pyrosequencing technology (originally created by Pyrosequencing AB, now Biotage) is an extremely rapid DNA sequencing approach that employs novel chemistry to robustly sequence relatively short (<70-bp) target regions. Several reports have suggested that pyrosequencing could allow the identification of medically important yeasts (5, 10, 17), although for the most part only a few isolates representing the more common species were examined. In addition, we have previously shown that pyrosequencing of a region within ITS2 could identify some rare and closely related yeast species and was able to distinguish C. glabrata from its close genetic relative, C. nivariensis (4), and also to discriminate between C. parapsilosis, C. orthopsilosis, and C. metapsilosis (3).
In the current study, we have assessed the possibility of using pyrosequencing of a short fragment of ITS2 for identification of a wide spectrum of clinically important yeast species. In excess of 450 yeast isolates referred to the United Kingdom National Mycology Reference Laboratory (MRL), Bristol, United Kingdom, for identification were submitted to conventional identification, molecular identification based on 28S rRNA gene sequencing, and also pyrosequencing. Pyrosequencing successfully allowed the rapid and unambiguous identification of >98% of isolates, which encompassed 40 different species of pathogenic yeasts.
MATERIALS AND METHODS
Test isolates.A total of 477 isolates were tested. These included 88 reference isolates and type strains that had been stored in the National Collection of Pathogenic Fungi (NCPF) (Table 1) and 389 additional clinical isolates that had been submitted to MRL for identification. Isolates were subcultured twice on plates of Oxoid Sabouraud dextrose agar containing 0.5% (wt/vol) chloramphenicol (Unipath Limited, Basingstoke, England). Cultures were incubated for 24 h at 35°C prior to testing. All procedures used coded isolates, and their identities were revealed only after completion of conventional and molecular analyses.
Selected NCPF reference and type strains included in the study
Conventional identification methods.All the clinical isolates included in the present study were subjected to the conventional yeast identification strategy currently employed by the MRL: following initial germ tube testing, isolates were subjected to testing with the AUXACOLOR2 identification kit (Bio-Rad, Marnes-la-Coquette, France) exactly as described previously (6). Isolates were also subjected to PCR amplification-sequencing of the D1-D2 portion of the 28S rRNA gene exactly as described previously (15), using fungal genomic DNA that had been prepared using prepunched Whatman FTA papers as described previously (2).
Pyrosequencing of ITS2.Pyrosequencing analysis of a portion of ITS2 was performed essentially as described previously (3, 4), using genomic DNA prepared using prepunched Whatman FTA filters (2). Briefly, PCR amplification of a fragment of ITS2 was performed in 50-μl reaction volumes in the presence of 500 nM primers PyroF2B (5′-biotin-CTTTGAACGCACATTGCGCCCTCTGGTATTCC-3′) and PyroR2W (5′-TCCYCCGCTTATTGATATGCTTAAGTTCAGC-3′), 200 μM each deoxynucleoside triphosphate, 1.25 U of HotStar Taq polymerase (Qiagen, Valencia, CA), and a single FTA filter punch. Following enzyme activation at 94°C for 15 min, the reaction mixtures were subjected to 40 thermal cycles on a GeneAmp PCR systems 9700 thermocycler (Applied Biosystems, Foster City, CA) with the following parameters: 94°C (15 s), 50°C (15 s), and 72°C (90 s). Amplification success was evaluated by electrophoresis of a fraction of total amplification products in 1.2% (wt/vol) agarose gels run for 45 min at 120 V in Tris-borate buffer. ITS2 amplification products were subjected to pyrosequencing analysis using the primer PyroYS (5′-GCGGGTAGTCCTACCTGATTTGAG-3′) and the reagents supplied with the Pyrogold SQA kit using a PyroMark ID pyrosequencing instrument (Biotage AB, Sweden). Sequencing was performed on 40 μl of each successful PCR product, using 15 cycles of introduction of each the 4 nucleotides (60 nucleotide introductions in total). Analysis of the resulting sequences was performed using IdentiFire software (Biotage) with an extended sequence database generated at MRL using reference and type species. IdentiFire default parameters were employed for sequence analysis, with the exception of “sequence to be analyzed,” which was set to “user-defined, nucleotides 1 to 35.” The extended sequence database is available from MRL upon request (enquiries to Andy.Borman@uhBristol.nhs.uk).
RESULTS
Table 2 shows the pyrosequencing profiles generated for the various yeast species analyzed. Unique pyrosequencing profiles were obtained for the majority of the species tested and ranged from 35 nucleotides to in excess of 60 nucleotides in length (data not shown). In all cases there was 100% agreement between the pyrosequencing identification and the identification obtained by PCR amplification and conventional sequencing of some 300 to 400 bp of the D1-D2 region of each organism (Table 2 and data not shown). While correct identification of some species could be achieved by analysis of only 20 to 25 bp of sequence generated by pyrosequencing analysis, analysis of the first 35 bp allowed correct identification for 469 of the 477 isolates examined (Table 2). As reported previously, pyrosequencing analysis was able to discriminate between very closely related organisms and species complexes. Pyrosequencing separated C. albicans from C. dubliniensis and C. glabrata from C. nivariensis and C. bracarensis and also correctly identified the members of the C. parapsilosis species complex (C. parapsilosis, C. orthopsilosis, and C. metapsilosis [24]) (Table 2) (3, 4).
Manual alignments of the ITS2 pyrosequencing signature sequencesa
For the 469 isolates that were correctly identified by pyrosequencing, the similarities between the pyrosequencing signature sequence and the reference ITS2 sequences present in our database ranged from 80 to 100%. As previously reported, all mismatches resulted from machine errors in incorrectly discerning the numbers of nucleotides in homopolymeric stretches (17). These errors never resulted in the erroneous identification of one species as another and were easily corrected manually after close inspection of the pyrograms (data not shown). Eight isolates, corresponding to three species (C. lambica, n = 2; C. lipolytica, n = 2; C. pararugosa, n = 4) could not be identified by pyrosequencing analysis. Despite successful amplification of the appropriate portion of ITS2 with our PCR primers (Table 2; data not shown), no signature sequence could be obtained after pyrosequencing. A previous study (17) using pyrosequencing primers developed commercially by Biotage to target the same ITS2 reported similar failures with these species, which are likely due to sequence variations in the ITS2 target that prevent annealing of the pyrosequencing sequencing primer.
To assess whether the individual signature sequences generated by pyrosequencing were unique to the species under study, each of the signature sequences depicted in Table 2 was subjected to BLAST searches against the fungal sequences present in the public synchronized databases (Table 3; data not shown). For 30 of the 40 signature sequences, searches returned only the species that had been identified. For the other 10 species (denoted by asterisks in Table 2), BLAST searches returned several alternative species identifications, all with 100% identity over the ITS2 analyzed by pyrosequencing. However, for the most part, the alternative species identifications suggested by these database searches corresponded to rare yeast species that have been proposed to comprise cryptic/closely related species in species complexes containing the principal organism of interest (Table 3). For all such database searches, sequence matches that were returned were scrutinized for their likely validity by assessing whether the database entries were for type strains of the respective organisms and whether multiple examples of each sequence had been deposited from several independent laboratories. However, since we did not attempt to pyrosequence the type strains of the organisms returned in such searches, it is possible that some of the database returns represent organisms that had been erroneously identified by the depositor of the sequence. In summary, to date potential pyrosequencing signature sequence redundancy has not been observed between yeast species from unrelated genera or species complexes.
Organisms demonstrating potential sequence redundancy in the ITS2 targeta
DISCUSSION
The current study has evaluated the utility of pyrosequencing analysis in the rapid identification of yeast isolates cultured from clinical specimens. A total of 477 test isolates, encompassing 43 yeast species and including both reference and type strains and strains from clinical samples, were subjected to pyrosequencing analysis of a portion of ITS2. Species were included in the present study due to their occurrence in clinical specimens. Test organisms were selected to include as many disparate species as possible, while ensuring that multiple isolates of the most clinically relevant species were included. The organisms included therefore do not in any way reflect an epidemiological analysis of Candida infections. All isolates had previously been unambiguously identified using combinations of biochemical/phenotypic methods and conventional sequencing of the D1-D2 portion of the 28S rRNA gene but were submitted for pyrosequencing in a rigorously blinded manner.
Pyrosequencing of a 35-nucleotide portion of ITS2 allowed the successful identification of 469/477 isolates and 40/43 species. The 8 isolates that failed to be identified comprised three species (C. lambica, C. pararugosa, C. lipolytica) that are only rarely encountered in the clinical setting. Indeed, these three species account for less than 0.5% of all yeast isolates that are submitted to the MRL for identification (see, for example, reference 14). The ITS2 pyrosequencing signature sequences generated in the current study were unique to the species under study for 30 of the 40 species that were successfully identified. For the remaining 10 species, BLAST analyses of signature sequences against the public DNA databases returned several potential alternative species identifications (Table 3). In all cases, the alternative species were close genetic and phenotypic relatives of the test organism and can reasonably be considered members of single species complexes (Table 3). Currently, the clinical need to distinguish between such very close genetic relatives is uncertain (see, for example, reference 11), although it is somewhat unfortunate that the 35-nucleotide sequence analyzed by pyrosequencing did not allow separation of members of the Cryptococcus neoformans complex. Further studies will be required to determine if pyrosequencing of a second independent locus might allow identification of the individual species in these complexes and also the three species that to date have failed to be identified by pyrosequencing altogether.
Although previous studies have demonstrated that conventional sequencing of the entire ITS2 allowed the identification of greater than 98% of clinical yeast isolates from 86 species examined (14), the discriminatory power of pyrosequencing analysis of the 35-nucleotide region analyzed here is impressive. Pyrosequencing, when coupled with the Whatman FTA paper technology for the rapid extraction of fungal genomic DNA, could be accomplished within 6 h from the time of starting pure cultures and robustly discriminated between closely related species in several species complexes, including the C. parapsilosis species complex (C. parapsilosis, C. orthopsilosis, C. metapsilosis) and distinguished C. glabrata from C. nivariensis and C. bracarensis. This high discriminatory power is likely to stem at least in part from the extremely variable nature of the portion of ITS2 examined (Table 2). No single nucleotide position in the 35-nucleotide C. albicans signature sequence was conserved among all 40 species that yielded sequences. Moreover, sequence variability increased steadily from 5′ to 3′ across the ITS2 target, such that 16 of the 35 nucleotides in the C. albicans sequence were conserved in less than 50% of the study species (Table 2). Thus, overall sequence identity with C. albicans ranged from 25.7% (C. rugosa) to 68.6% (Lodderomyces elongisporus) over the target region, and 21 of the 40 species shared less than 50% identity with C. albicans in this region (Table 2). The extremely variable nature of this portion of ITS2 is probably the reason why our pyrosequencing discriminated more yeast species than pyrosequencing targeting a portion of the more conserved 18S rRNA gene reported by Gharizadeh et al. (10) and also explains why occasional errors of resolution of homopolymer runs in the ITS2 target sequence do not affect the final identification obtained by pyrosequencing.
Once errors in the resolution of homopolymeric runs were corrected, no intraspecific sequence variations were detected in the relatively short region of ITS2 examined by pyrosequencing analysis for all but one of the species examined. This is perhaps unsurprising, given that the variability in highly conserved genes such as those encoding ribosomal DNA is more limited than that in many other genes. However, three related but distinct pyrosequencing signature sequences were generated from isolates of Candida rugosa tested in a previous study (17) and in the current study (see the aligned sequences in Table 4). When each signature sequence was submitted to BLAST searches, all returned matches with 100% identities with reference strains of C. rugosa in the public synchronized databases (Table 4), but only one of the sequences matched the type strain for this organism (strain Montero 2; Table 4). Although it remains plausible that C. rugosa exhibits exceptional sequence variability in this short region of ITS2, we believe that these data might indicate that C. rugosa is in fact a species complex of several closely related but distinct organisms. Future studies will address this possibility.
Sequence variation detected by pyrosequencing of putative isolates of C. rugosa a
In conclusion, we have demonstrated that pyrosequencing analysis of a small portion of ITS2 is sufficient to reliably identify the vast majority of pathogenic yeast species. The pyrosequencing profiles produced for each species were reproducible and conserved across multiple isolates and are to date unique to the species (or species complex) of pathogenic yeast under investigation. Due to the extreme rapidity and facility of pyrosequencing analysis and the relatively low cost per sample analyzed, we believe that this approach is ideally suited to the accurate identification of yeast isolates. The high-throughput nature of the pyrosequencing technology also makes it ideally suited to evaluate the prevalence of cryptic minority species. Given the increasing evidence indicating species-specific variations in antifungal susceptibility profiles, we believe that rapid and accurate yeast identification will become increasingly important for informed therapeutic decisions.
ACKNOWLEDGMENTS
We are grateful to the other members of the Mycology Reference Laboratory for their interest and advice, to Frank Odds for sharing isolates of Candida metapsilosis and C. orthopsilosis, and to Whatman International and Biotage for the gift of reagents.
A.M.B. was funded in part by travel grants from Whatman International and Biotage AB.
FOOTNOTES
- Received 27 May 2010.
- Returned for modification 1 July 2010.
- Accepted 2 August 2010.
- Copyright © 2010 American Society for Microbiology
