PCR Followed by Electrospray Ionization Mass Spectrometry for Broad-Range Identification of Fungal Pathogens

Abstract

Invasive fungal infections are a significant cause of morbidity and mortality among immunocompromised patients. Early and accurate identification of these pathogens is central to direct therapy and to improve overall outcome. PCR coupled with electrospray ionization mass spectrometry (PCR/ESI-MS) was evaluated as a novel means for identification of fungal pathogens. Using a database grounded by 60 ATCC reference strains, a total of 394 clinical fungal isolates (264 molds and 130 yeasts) were analyzed by PCR/ESI-MS; results were compared to phenotypic identification, and discrepant results were sequence confirmed. PCR/ESI-MS identified 81.4% of molds to either the genus or species level, with concordance rates of 89.7% and 87.4%, respectively, to phenotypic identification. Likewise, PCR/ESI-MS was able to identify 98.4% of yeasts to either the genus or species level, agreeing with 100% of phenotypic results at both the genus and species level. PCR/ESI-MS performed best with Aspergillus and Candida isolates, generating species-level identification in 94.4% and 99.2% of isolates, respectively. PCR/ESI-MS is a promising new technology for broad-range detection and identification of medically important fungal pathogens that cause invasive mycoses.

INTRODUCTION

Invasive fungal infections (IFIs) are responsible for significant morbidity and mortality, particularly among immunocompromised patients. Early identification of fungal pathogens is important for optimal treatment of IFIs (1); however, traditional detection methods, including culture, direct exam, and histopathology, all have limitations of sensitivity, specificity, and ability to provide rapid identification (2–4). Molecular techniques, such as PCR, have shown enhanced analytic sensitivity and specificity (5). There have been several publications showing the value of PCR in the detection and characterization of those molds involved in IFIs, but with limited range of coverage. Most published assays target only a single organism or organism group, most typically Aspergillus or Candida species (6–8). Given the increasing variety of yeasts and molds now commonly seen in IFIs (9, 10), a broader PCR-based method could prove to be valuable for diagnosis. PCR coupled with electrospray ionization mass spectrometry (PCR/ESI-MS) offers the possibility for broad-range identification while maintaining the sensitivity of PCR.

Over the past 10 years, PCR/ESI-MS has been developed and used in a wide variety of diagnostic settings, including biodefense, forensics, and clinical diagnostics (11). The PCR/ESI-MS system relies on multiple pairs of broad-range primers to cover a group of organisms, which may be as narrow as a single genus or as broad as the pan-fungal approach explored in the present study. The “molecular fingerprint” obtained from PCR/ESI-MS can be coupled with a computer-based algorithm to determine the identification of the organism by comparison with reference database signatures (11, 12). PCR/ESI-MS has been used in clinical diagnostics for the detection of bacterial and viral pathogens (13–19). More-limited work with fungi has focused primarily on identification of Candida species (6). In this study, a novel PCR/ESI-MS broad-range fungal assay was developed to identify the most common agents of IFI, including Ascomycota (e.g., Aspergillus spp., Candida spp., and Pneumocystis spp.), Basidiomycota (e.g., Cryptococcus spp.), and Mucorales (e.g., Mucor spp. and Rhizopus spp.). Herein we describe the evaluation of this method using cultured reference strains and clinical isolates encompassing a wide variety of medically important yeasts and molds.

MATERIALS AND METHODS

Study design.

A 16-primer pair PCR/ESI-MS broad fungal assay was developed by Ibis Biosciences (Carlsbad, CA) to provide universal fungal detection, as well as differentiation to the species level of common fungal pathogens known to cause IFI. The performance characteristics of this broad-range assay were assessed measuring analytical sensitivity and conducting a feasibility study to look at sampling stability and assay interference. The assay was then tested against a set of phenotypically identified clinical isolates, including both yeasts and molds.

Reference strains and clinical isolates.

Sixty reference strains used in this study were obtained from the American Type Culture Collection (ATCC; Manassas, VA) and are listed in Table S1 in the supplemental material. All permits required for transport were obtained prior to shipping. In the case of Fusarium oxysporum, Ibis Biosciences obtained the USDA/Plant Protection and Quarantine permit for receipt of lyophilized F. oxysporum from the ATCC and only purified DNA (not requiring a permit) was sent to St. Jude Children's Research Hospital (SJCRH). For Coccidioides immitis, purified DNA was sent from Viracell only to Ibis Biosciences, and a USDA Veterinary Services 16-3 permit was obtained prior to shipment. Twelve Mucorales, 41 Ascomycetes, and 7 Basidiomycetes reference strains were used in this study. Within the Ascomycetes and Basidiomycetes were a total of 16 yeasts. Clinical isolates recovered from patient specimens submitted for fungal diagnosis were collected between March 1999 and October 2011 at two institutions, SJCRH and Johns Hopkins Hospital (JHH). Patient identifiers were removed from all isolates selected for study, and PCR/ESI-MS testing was performed blinded to original culture data collected at the time of diagnosis from either SJCRH or JHH. A total of 401 clinical isolates were evaluated by Ibis Biosciences using PCR/ESI-MS, 134 isolates collected retrospectively from SJCRH and 267 collected prospectively at JHH. Of those initial 401 isolates, 4 were excluded due to contamination during shipment and 3 were excluded due to contaminated or mixed culture. Results for PCR/ESI-MS and phenotypic identification were directly compared to each other. Discrepancies were resolved using ITS and D1/D2 sequencing, which was completed at SJCRH. Phenotypic identification was considered to be the reference standard when stratified taxonomically. This study was designated exempt by both SJCRH and Johns Hopkins University institutional review boards (IRBs).

Fungal culture.

Rehydrated ATCC material from freeze-dried stocks and clinical isolates were each inoculated to Sabouraud dextrose agar (Fisher Scientific, Suwanee, GA) and held at 29°C until a confluent mycelium (mold) or visible isolated colonies (yeast) were observed, typically after 2 to 7 days of incubation. Retrospectively collected isolates (SJCRH) were stored in water culture at room temperature prior to use in the study.

Nucleic acid extraction.

Nucleic acid was extracted using the PCR/ESI-MS DNA preparation kit (CE Mark) following the standard extraction protocol, with modifications of 1- or 3-mm biopsy punches (Integra Miltex, York, PA) used to obtain samples from agar plates. The punch samples, which included either mycelial or yeast colonies and underlying agar, were placed in Negative Matrix tubes (Abbott Molecular, Des Plaines, IL) that contained 1,100 μl of filtered 10% bovine serum albumin (Negative Matrix solution) and zirconium-yttrium beads. Before extraction, 25 μl proteinase K (600 units/ml) and 325 μl lysis buffer (Abbott Molecular) were added, and the PCR/ESI-MS extraction protocol was performed.

Genomic DNA for sequencing was extracted from reference strains and discordant clinical strains using the QIAamp UCP pathogen minikit and pathogen lysis tubes L (Qiagen, Valencia, CA). The extraction steps followed the manufacturer's protocols for pretreatment of microbial DNA from biological fluids or cultures using mechanical prelysis and sample preparation spin protocol, with slight modifications. Conidia and sporangiospores were collected by rubbing a sterile polyester-tipped applicator over the mycelium mat; the swab was then transferred to a collection tube and washed in 1 ml PCR grade water, and the resulting pellet was resuspended in 500 μl ATL buffer containing the optional DX reagent, transferred to the pathogen lysis tube L, and vortexed at maximum speed for 10 min. Samples were eluted in 35 μl of PCR grade water. Quantity and purity of DNA were determined using the NanoDrop 200c spectrophotometer (Thermo Scientific, Waltham, MA).

PCR amplification and sequencing.

Ribosomal internal transcribed spacer (ITS) and D1/D2 regions were amplified using HotStar HiFidelity Taq (Qiagen) and the primer pairs ITS-1/ITS-4 and NL-1/NL-4, respectively (20, 21). Fifty-microliter reactions were carried out according to the manufacturer's specifications. A touchdown cycling protocol was used with the following parameters: initial activation at 95°C for 10 min, followed by 25 cycles of denaturation at 95°C for 30 s, annealing at 65°C for 30 s minus 1°C per cycle, and extension at 68°C for 1 min. Amplification was carried out over 25 cycles, consisting of 95°C for 30 s, 50°C for 30 s, and 72°C for 1 min. A final extension period was included (72°C for 5 min), and the reaction was held at 4°C. PCR products were purified using the Qiagen PCR purification kit (Qiagen) according to the manufacturer's directions and quantitated using the NanoDrop 200c spectrophotometer (Thermo Scientific).

Sequencing was carried out on the ITS and D1/D2 regions using the protocol for BigDye Terminator version 3.1 chemistry (Applied Biosystems, Carlsbad, CA). Sequencing primers included ITS-1, ITS-2, ITS-3, and ITS-4 (ITS region) (21) and NL-1, NL-2, NL-3, and NL-4 (D1/D2 region) (20). DNA sequencing data were configured using the software CLC Main Workbench version 6.0 (CLCbio, Cambridge, MA), and the sequence identity was determined by using the NCBI BLAST program and the nucleotide collection database (nr/nt; NCBI, Bethesda, MD).

Phenotypic identification.

For both SJCRH and JHH mold isolates, phenotypic identifications were determined by visual observation of macroscopic colony appearance and lactophenol cotton blue microscopic slide mount features. Growth rate and temperature tolerance were also determined. For JHH yeast isolates, phenotypic identification was based on colony morphology including the use of CHROMagar (BD, Franklin, NJ), microscopic examination, germ tube formation, and biochemical reactions using API panels (bioMérieux, Durham, NC).

PCR/ESI-MS assay.

All samples were run on the PLEX-ID platform at Ibis Biosciences using PCR/ESI-MS methodology as described previously (11, 12). Sixteen primer pairs were assembled to incorporate various priming ranges that provided a balance between universal fungal detection and fine characterization of the most common fungi involved in IFIs (see Results) (Table 1 and Fig. 1). A synthetic DNA construct was also included within each reaction (20 copies per reaction) as an internal positive amplification control and as a quantitative calibrant. This synthetic construct (approximately 1,800 nucleotides) contained all 16 primer-binding regions; however, these regions each contained either insertions or deletions, causing a mass shift to allow product differentiation from the actual sample. By comparing the signals obtained from the synthetic construct amplicons and the target amplicon, the approximate template copy number could be determined (11).

Table 1.

PCR/ESI-MS broad fungal assay primer pairs

Ibis primer pair no.	Primer sequence (5′-3′)a	Molecular targetb	GenBank accession no.	Amplicon size (bp)	Fungal detection range
5186	TCATTTTTGGTAAGCAGAACTGGCGA	LSU rRNA	M26190.1	107–108	Broad
	TGTCTAGATGAACTAACACCTTTTGTGGT
5187	TGCCGAAGTTTCCCTCAGGATAGCA	LSU rRNA	M26190.1	72–78	Broad
	TCAAGGCCTCTAATCATTCGCTTTACCTC
5178	TGCAGGCCTATGCTCGAATACATTAGCAT	SSU rRNA	EF397944	99–101	Ascomycetes and Basidiomycetes
	TGTCCCTATTAATCATTACGGCGGTCCTA
3030	TGTGAAGCGGCAAAAGCTCAAATTT	LSU rRNA	X70659.1	116–130	Ascomycetes and Basidiomycetes
	TTCTCACCCTCTGTGACGGCCTGTTCC
5185	TGGAGTCTAACATCTATGCGAGTGTT	LSU rRNA	X70659.1	137–153	Ascomycetes and some Basidiomycetes
	TCCCACAGCTATGCTCTTACTCAAATCC
3766	TTGTGTAGAATAGGTGGGAGCTTCGGC	LSU rRNA	X70659.1	148–154	Ascomycetes and Basidiomycetes
	TCTGACAATGTCTTCAACCCGGATC
5181	TCCAAATTACGTGCCAGCAGTCG	mtSSU rRNA	DQ415396.1	90	Mucorales and Ascomycetes (except Candida)
	TCCCTACCGTCTAGGTACCCTTTAAACC
4836	TCTTGGATTGACCGAAGACAAACTACTG	SSU rRNA	AF113430.1	136–136	Most Ascomycetes, Basidiomycetes, and Mucorales
	TCTCTAGTCGGCATCGTTTGTGGTTAAG
4837	TGCCGCGGTGCTCACTCTTTC	tub	AM261879.1	109	Most Ascomycetes (except Candida)
	TGTATCGGCCGTTGCGGAAGTC
5172	TCCCCGTAACAATTTTCTTATTCTTCTTAGTATTAG	mt cytB	AB025730.1	106	Mucorales
	TGCATTGGGTTAGCTGGAATATAATTATCAGGATG
5174	TCGAAGACGATTAGATACCGTCGTAGTC	SSU rRNA	AF113430	113–116	Mucorales
	TCAGAGCCCAAAAACTTTACTTTCGCTAAG
3862	TCCAAGTACTTGACACATGCTAATCG	mtSSU rRNA	NC_007445.1	102–133	Eurotiales (specifically Aspergillus and Penicillium)
	TCCCCTTACTTTAAGGTAGCCAAATTATC
3865	TGGTACAGTGGAGTATGCTGTTTAATTGGA	mtSSU rRNA	AF285261.1	93–131	Saccharomycetales (specifically Candida)
	TCTGACGACAACAATGTAACGCCTG
3867	TTCGATACCCGTGTAGTTCTAGTAGTAAAC	mtSSU rRNA	NC_004691.1	113–127	Saccharomycetales (specifically Candida)
	TCATTATTGCTAACGTACTCTTCAGGTGG
4145	TACCAAGCCAATGACGAGTAGGG	mtSSU rRNA	NC_004336.1	87	Tremellales (specifically Cryptococcus)
	TGCTGCACATAATCTATGCTCTGGAC
4437	TGACGAGTTCATGAGGGCAGGC	hpr	U01067	77	Extraction control (pumpkin DNA)
	TCTGGCCTTTCAGCAAGTTTCCAAC

^aEach row of the primer sequence column contains the forward and reverse primers for the corresponding primer pair number.

^bLSU, large subunit; SSU, small subunit; mt, mitochondria.

Fig 1.

Schematic showing the coverage of fungal species provided by the assay. Colored boxes indicate the coverage attained by each primer pair whose name is reported within square brackets. Each species mentioned has been uniquely characterized using culture samples. Species reported within the same box show distinct signatures that use the same combination of primer pairs.

Molecular signature database and reporting by PCR/ESI-MS.

The PCR/ESI-MS fungal molecular signature database, consisting of the base composition signatures of individual fungi, was initially constructed from an in silico survey of existing fungal entries in the GenBank nr database and later completed with actual spectral signatures of 60 diverse reference strains (see Table S1 in the supplemental material). After the addition of the 60 reference strains, database version S1.4.0.2-A6.336.379.302-IVD01_6.20.21.15.1 and software version SPS.1.2.6993/REL-GenX-V07R004-8281 were not further modified and consisted of approximately 400 complete fungal signatures. This database was then used to assign identities to the clinical isolates. Among the reference strains used in the database, where readily attainable, a type strain was selected. Identities of most other reference strains were either verified by ITS and/or D1/D2 sequencing or by independent references in the peer-reviewed literature. PCR/ESI-MS fungal signatures were typically defined across 6 to 10 loci depending on phylogeny. Aspergillus and Candida isolates were identified by PCR/ESI-MS to the species or species complex level when possible. All other organisms were reported to the genus level when possible. A molecular identification was given if 60% of the retrieved loci were an exact match with database signatures. The cutoffs were at least a 60% exact match for genus- or species-level identification, 50 to 60% exact match for “fungi detected—no identification can be provided,” and below a 50% exact match for “not detected.” The 60% threshold was set after reviewing the diversity of existing database signatures. All signatures were highly redundant due to the inclusion of at least one uniquely discriminating base count, thus retaining the ability to uniquely associate with a single fungus even when 40% of the expected loci were missing. Although the system has the potential to recognize single nucleotide polymorphism (SNP) variants, this feature was turned off for the broad fungal assay since closely related base counts could result from naturally occurring variation in a given species or from closely related but distinct species.

PCR/ESI-MS broad fungal assay development. (i) Analytical sensitivity.

Amplicon detection was carried out on 13 different fungal genomic preparations, challenging all primer pairs. Detection was carried out in the presence and absence of 3 μg of human DNA. Original fungal DNA stocks, either of known concentration or empirically determined based on the quantitative calibrant (described above), were tested over a 2-fold serial dilution range from 64,000 copies/ml (320 copies/reaction [rxn]) to 200 copies/ml (1 copy/rxn) and run in triplicate to establish limits of detection (LOD).

(ii) Interference.

A core panel of five fungi (Aspergillus fumigatus [AF], Candida albicans [CA], Candida glabrata [CG], Cryptococcus neoformans [CN], and Mucor racemosus [MR]), representing the minimum number needed to test all assay primers, was used in the interference study. To assess specimen type interference, 1-mm fungal punches for each member of the panel were each mixed with an additional 3-mm Sabouraud agar punch that came from plates that had been overlaid for 16 h with either blood, wound fluid, cerebrospinal fluid (CSF) or bronchoalveolar lavage (BAL) fluid, obtained from Bio-Med Supply, LLC (Carlsbad, CA) and processed for fungal detection. To assess interference of commonly used fungal growth media (BD, Franklin Lakes, NJ) the 1-mm core panel punches were mixed with 3-mm punches from inhibitory mold agar with gentamicin, brain heart infusion agar with blood, chloramphenicol and gentamicin, potato dextrose agar with tartaric acid, Mycosel agar, and Sabouraud dextrose agar with chloramphenicol. These mixtures were then processed for fungal detection. To assess bacterial interference, Citrobacter freundii, Haemophilus influenzae, Lactobacillus casei, Neisseria mucosa, Staphylococcus aureus, and Staphylococcus epidermidis were each spiked into the Negative Matrix tubes at 10⁵ CFU/ml, together with a 1-mm punch from each of the core panel fungi. The tubes were then processed for fungal detection by PCR/ESI-MS. To assess potential interference between mixed fungal isolates, the five members of the core panel (AF, CA, CG, CN, and MR) were run using PCR/ESI-MS in all possible pairwise combinations at ratios of 1:1, 1:9, and 9:1. All interference reactions were run in triplicate.

(iii) Stability study.

Biopsy punches of each of the core panel fungi used in the interference studies (see above) were stored in Negative Matrix tubes at room temperature, 4°C, or −20°C for 1 day, 3 days, 1 week, 1 month, or 2 months. After incubation, proteinase K and lysis buffer were added and samples were processed. Four replicates of each fungus were run at each corresponding time point.

(iv) Data analysis.

PCR/ESI-MS reference strain identifications were compared to ATCC designations and sequencing results in a pairwise manner. Species-level comparisons using sequence data were considered concordant if designations matched at either the species or species complex level. PCR/ESI-MS clinical isolate identifications were compared to phenotypic identification results. Samples were classified as having either a species- or genus-level match. Mismatched identifications of clinical samples were resolved by sequencing.

RESULTS

Assay development.

A 16-primer pair panel for the characterization of fungal species by PCR/ESI-MS was assembled (Table 1 and Fig. 1). The primer pairs developed and used in this assay, particularly those that targeted large-subunit (LSU) rRNA, were chosen to cover the broad phylogenetic diversity seen in clinically relevant fungi. Two primer pairs (5186 and 5187) were designed to amplify the LSU rRNA for broad-range fungal detection; those two primer pairs offer limited resolution but yield amplicons with all fungi tested to date, including the Mucorales. The Ascomycetes and Basidiomycetes are primarily resolved with the next ribosomal primer pairs, 5178, 3030, 5185, and 3766. A more robust discrimination of Ascomycetes molds was achieved using primer pairs 5181, 4836, and 4837, yielding a 7- to 9-locus signature for most Ascomycetes species. Primers 5181 and 4836 also provided coverage of some of the Mucorales, but further resolution was provided using dedicated primer pairs 5172 and 5174. The panel was completed with primer pairs targeting the more rapidly evolving mitochondrial rRNA, providing narrow-scope but highly discriminant signatures of Candida spp. (3865 and 3867), Aspergillus spp. (3862), and Cryptococcus spp. (4145). The last primer pair (4437) targeted a pumpkin DNA extraction control.

Analytic sensitivity.

The PCR/ESI-MS limit of detection (LOD) for all four tested Aspergillus spp. was near 300 copies/ml (approximately 15 copies/rxn). The remainder of the tested strains, except Rhizopus oryzae and R. microsporus, showed LODs ranging from 20 to 180 copies/ml (1 to 9 copies/rxn). R. oryzae and R. microsporus gave LODs of 900 copies/ml (45 copies/rxn) and 700 copies/ml (35 copies/rxn), respectively (Table 2). LOD did not change for any of the fungi when human DNA was present (data not shown).

Table 2.

Limit of detection (LOD) achieved by PCR/ESI-MS broad fungal assay for selected fungal species

Fungus	LOD measured ina:
	gen/ml	gen/rxn
Aspergillus fumigatus	300	15
Aspergillus flavus	320	16
Aspergillus terreus	280	14
Aspergillus niger	280	14
Fusarium solani	180	9
Candida albicans	80	4
Candida glabrata	40	2
Cryptococcus albidus	100	5
Sporothrix schenckii	140	7
Scedosporium prolificans	20	1
Scedosporium apiospermum	100	5
Rhizopus oryzae	900	45
Rhizopus microsporus	700	35

^aCopy number based on an internal calibrant or quantified DNA. gen, genomes.

Interference study.

No inhibition was seen in the presence of blood, wound fluid, CSF, or BAL fluid (see Table S2 in the supplemental material). All fungi were identified irrespective of the agar type used (see Table S2). However, Saccharomyces cerevisiae was detected in the negative controls for inhibitory mold agar with gentamicin, brain heart infusion agar with blood, chloramphenicol, and gentamicin, as well as Mycosel agar. The Saccharomyces detection likely derived from yeast-based ingredients in the media. The detection of fungal panel members was not inhibited by the presence of any of the tested bacteria. To simulate mixed fungal culture, biopsy punches from different fungi were mixed prior to analysis. In most cases, no inhibition was seen in fungal detection. Detection of A. fumigatus was inhibited when mixed with either Candida albicans or Candida glabrata (data not shown). In all other mixtures, both fungi were successfully identified, and there were no misidentifications due to superimposition of multiple fungal signatures (data not shown).

Stability testing.

Punches could be stored in negative matrix for 72 h at ambient temperatures or 2 months at 4°C or −20°C without a decrease in sensitivity.

Identification of clinical isolates. (i) Molds.

A total of 264 filamentous fungi were evaluated. PCR/ESI-MS was able to identify 48.1% (127/264) to the species level and an additional 33.3% (88/264) to the genus level. Mold was detected without a genus- or species-level identification 14.4% (38/264) of the time. No fungus was detected in 3.8% (10/264) of samples; the assay had an overall run failure rate of 0.38% (1/264).

Of the 127 isolates that were given species-level identifications by the PCR/ESI-MS broad fungal assay, 87.4% showed full concordance to either the species level or the species complex level (Aspergillus spp.) between PCR/ESI-MS and phenotypic results. Of the remaining isolates, 5.5% were not given species-level identification by phenotype, 1.6% were mismatched at the species level, and 5.5% were complete mismatches at both the genus and species levels when comparing the two methods. Among those 88 isolates that were identified only to the genus level by PCR/ESI-MS, there was 89.7% concordance with phenotypic identification. Of those 18 isolates mismatched at either the species or genus level, sequence data confirmed that the PCR/ESI-MS assay was correct in 2 instances while phenotypic identification was correct in the remaining 16 instances.

Results were stratified taxonomically to determine the performance of PCR/ESI-MS and the correlation with genus and species (Table 3; see also supplemental material). Of the 129 isolates identified as Aspergillus by phenotypic identification, PCR/ESI-MS identified 97.7% (126/129) of the samples to the correct genus level, while three isolates were detected by PCR/ESI-MS but could not be identified. Of those 126 isolates, PCR/ESI-MS provided species-level identification for 94.4% (119/126), compared to 89.9% (116/129) with phenotypic methods. Comparing only the isolates with a species-level identification by both methods, concordance between PCR/ESI-MS and phenotype was 99.1% (112/113), including the 3 isolates phenotypically identified as Aspergillus sydowii but matched by PCR/ESI-MS to Aspergillus versicolor (both within the Aspergillus versicolor species complex). The single discordant isolate was identified phenotypically as Aspergillus flavus, but Aspergillus niger was the closest match by PCR/ESI-MS. Based on sequence data, this isolate was identified as Aspergillus nomius (data not shown), which belongs to the Aspergillus flavus complex. Exserohilum spp., Fusarium spp., and Scedosporium spp. showed a 100% genus-level match between the two methods; however, only 2/5 (40%), 4/11 (36.4%), and 7/9 (77.8%) isolates, respectively, were given species-level designations by phenotypic methods, and none of the isolates were given species-level designations by PCR/ESI-MS. All isolates of Alternaria spp., Bipolaris spp., and Curvularia spp. were identified to the genus level but none were given species designations by either method. Among genuine Cladosporium species (none were Cladophialophora species), 8/10 (80%) matched at the genus level, and the remaining two isolates were detected but not identified by PCR/ESI-MS. Similarly, PCR/ESI-MS identified 5/6 (83.3%) of the Rhizopus isolates to the genus level; the remaining isolate was not detected by PCR/ESI-MS. Nine isolates were identified as Mucor spp. by phenotype; PCR/ESI-MS matched the phenotypic identification for 6/9 isolates (66.7%) but neither method gave species-level identifications. Two isolates were identified as Mucor by phenotype and as Rhizomucor and Scedosporium by PCR/ESI-MS. Based on sequencing data, these two isolates were Rhizomucor pusillus and Scedosporium apiospermum, confirming PCR/ESI-MS identifications (data not shown). The remaining sample was detected but could not be identified by PCR/ESI-MS. Among Penicillium spp. (none were Penicillium marneffei), all but one sample (18/19) either were misidentified as Aspergillus or Fusarium, were not detected, or had DNA detected but no identification could be determined. Among the isolates that were misidentified, sequence data confirmed that all but one (sequence identified as Paecilomyces lilacinus) were Penicillium spp., agreeing with phenotypic identifications. In the case of Paecilomyces spp., two of five isolates had DNA detected without identification, while two isolates were misidentified by PCR/ESI-MS as Fusarium, and in one sample, the PCR/ESI-MS run failed (Table 3). Based on sequence data, both misidentified isolates were identified as Paecilomyces lilacinus (data not shown).

Table 3.

Summary of PCR/ESI-MS broad fungal assay results from clinical isolates

Genus	No. of samples	PCR/ESI-MS result (no. of samples identified/total no. of samples)a
Molds
Alternaria spp.	10	Genus-level identification (10/10)
Aspergillus spp.	129	Genus-level identification (126/129)
		Speciated (119/126)
		Species match (109/113)b
		Species complex match (3/113)b
		Fungi detected but not identified (3/129)
Bipolaris and Curvularia spp.	15	Genus-level identification (15/15)
Cladosporium spp.	10	Genus-level identification (8/10)
		Fungi detected but not identified (2/10)
Cunninghamella spp.	2	Genus-level identification (2/2)
Exserohilum spp.	6	Genus-level identification (5/6)
		Not detected (1/6)
Fusarium spp.	11	Genus-level identification (11/11)
Mucor spp.	9	Genus-level identification (6/9)
		Genus-level misidentification (2/9)
		Fungi detected but not identified (1/9)
Paecilomyces spp.	4c	Genus-level misidentification (2/4)
		Fungi detected but not identified (2/4)
Penicillium spp.	19	Genus-level identification (1/19)
		Genus-level misidentification (5/19)
		Fungi detected but not identified (11/19)
		Not detected (2/19)
Rhizopus and Rhizomucor spp.	7	Genus-level identification (6/7)
		Not detected (1/7)
Scedosporium spp.	9	Genus-level identification (9/9)
Other characterized species (18 genera)	26	Genus-level misidentification (7/26)
		Fungi detected but not identified (15/26)
		Not detected (4/26)
Nonsporulating molds and unknown spp.	6	Fungus detected but not identified (4/6)
		Not detected (2/6)
Yeasts
Saccharomycetes (Candida, Pichia, and Clavispora spp.)	127	Genus-level identification (127/127)
		Speciated (126/127)
		Species-level identification (126/126)
Cryptococcus spp.	2	Not detected (2/2)
Rhodotorula spp.	1	Genus-level identification (1/1)

^aPCR/ESI-MS results compared to the phenotypic reference standard.

^bIsolates in which PCR/ESI-MS and phenotypic methods both produced species-level identifications.

^cExcludes one failed run.

(ii) Yeasts.

A total of 130 isolates identified as yeast by culture and phenotypic characterization were analyzed. PCR/ESI-MS identified 96.9% (126/130) of isolates to the species level and 1.5% (2/130) to the genus level. Although 98.5% (128/130) of yeast isolates were detected with the PCR/ESI-MS, the only two clinical Cryptococcus isolates (C. albidus and C. laurentii) could not be detected (Table 3). Of the isolates that were identified to the species level by both methods, there was 100% concordance. There was a 100% match between the PCR/ESI-MS and phenotypic methods for isolates that were identified only to the genus level.

DISCUSSION

These results demonstrate the potential value of PCR/ESI-MS for the identification of clinically relevant fungal isolates and set the stage for future studies examining its use in direct detection from clinical specimens. Molecular methods have become a crucial tool for the detection and characterization of infectious diseases. Their application to the diagnosis of fungal infections has been somewhat more limited, with commercial systems mostly limited to single agent assays (6, 7). DNA sequencing-based methods targeting the ITS and D1/D2 regions have been shown to be useful for the accurate identification of fungal isolates (5, 20–26). The PCR/ESI-MS system effectively combines broad phylogenetic coverage of sequence-based methods with the sensitivity of PCR. It does so in the context of a commercially available, high-throughput system whose value has been previously demonstrated for several other applications related to infectious disease diagnostics (13–19).

The assay requires a small amount of input material (1- to 3-mm colonies); therefore, identification can be achieved as soon as a fungal colony becomes visible. Harvesting fungi using the punch biopsy technique shown here is much easier than by loop, as some fungi can be very difficult to collect from solid culture media. This study also showed that clinical isolates can be used after prolonged storage, in the presence of different agar medium types, mixed bacterial or fungal populations, without apparent inhibition.

When used for organism identification of clinical culture isolates, the system demonstrated a high degree of concordance with phenotypic identification. PCR/ESI-MS identified 98.4% of the total yeast isolates to either the species (96.9%) or genus (1.5%) level, of which 100% were concordant to the species level with phenotypic methods. Similar findings were recently reported in a study comparing PCR/ESI-MS with repetitive-sequence-based PCR (rep-PCR) and sequencing for the detection of Candida species (6), in which 98.1% accuracy was observed. In the current study, the signatures of 17 reference strains of yeast (9/9 Candida species) were determined, of which 9 were repeatedly identified to the species level among the tested clinical isolates. These results indicate a high level of accuracy for the identification of clinically important yeasts. This confirms and expands upon previous results in this area (6) and suggests that PCR/ESI-MS may be useful, particularly for reliable identification of uncommon Candida isolates which may be misidentified using other systems (27).

The PCR/ESI-MS system identified 81.4% of clinical mold isolates to either the species (48.1%) or genus (33.3%) level. Concordance of PCR/ESI-MS with phenotypic results was 87.4% and 89.7%, respectively. Sequence data from clinical strains with discordant phenotypic and PCR/ESI-MS results showed that in the majority of the cases (16/18 or 88.9%), sequencing agreed with phenotypic identification (data not shown). For these isolates, the discrepancy arose from the incomplete matching of the signature retrieved by PCR/ESI-MS with an existing database entry. Thirty-eight clinical mold isolates (14.4%) were detected but could not be identified by PCR/ESI-MS. Of these, the majority were of the Penicillium, Pithomyces, Paecilomyces, and Trichophyton genera (per phenotypic methods). Such inconclusive identifications revealed local gaps in the phylogenetic distribution of the reference isolates initially represented in the database (see Table S3 in the supplemental material) and may be overcome by future expansion of the PCR/ESI-MS fungal database or by further expanding primer coverage to other possible genetic targets, such as elongation factor, β-tubulin, and calmodulin (28–30). Similarly, species-level identification could be improved by the incorporation of additional reference strains to the PCR/ESI-MS database.

Identification to the genus level only may be very helpful for some clinical decisions. For example, distinction between Cryptococcus spp. and Candida spp. from bloodstream isolates can provide critical information; echinocandins are not active against the former but are active against Candida spp. Differentiation between Aspergillus spp. and Rhizopus spp. is important, as voriconazole is active against organisms within the former but not against the latter.

There are some genera whose misidentification may have important clinical implications. For example, non-marneffei Penicillium spp. were identified in 18/19 cases as Aspergillus or Fusarium spp. This misidentification of non-marneffei Penicillium spp., which are considered usually to be contaminants, may trigger a therapeutic intervention for Aspergillus spp. or Fusarium spp. Trichophyton mentagrophytes misidentified as Aspergillus flavus may suggest a cutaneous lesion caused by invasive aspergillosis versus dermatophytosis. The latter organism would require systemic antifungal therapy, while the former organism would warrant topical therapy if recovered from a cutaneous lesion. However, cutaneous lesions caused by these two genera are clearly distinctive, with aspergillosis being characterized by necrosis (31). Further adjustments in primer design may be warranted in order to establish clearer distinction of the less-virulent genera of non-marneffei Penicillium spp. versus Aspergillus spp. and Fusarium spp.

Notwithstanding these limitations, the data here suggest not only a high degree of accuracy for identification of yeasts and molds but also a high degree of specificity, apparently unaffected by common sample matrices, interfering substances, mixed cultures, or prolonged storage. The PCR/ESI-MS system has been used for direct detection of bacterial and viral pathogens; similar application to broad-range fungal detection could have tremendous clinical utility. The potential for a broad-range fungal assay has consistently eluded investigators. However, earlier detection is a fundamental challenge in improving clinical outcome in infections that continue to have a high degree of morbidity and mortality. Future, prospective clinical studies will explore the application of PCR/ESI-MS for direct detection and identification of fungal pathogens in a variety of clinical sample types.

The PCR/ESI-MS system used here lists for approximately $500,000, with costs of $30 per sample and a turnaround time of 24 h. Fungal identification by phenotypic methods costs $5 per sample. These represent material costs only, beginning with cultured isolates, and exclude the cost of initial culture, common to both systems. Identification by morphology may be done immediately if sporulation is present on the initial isolation media; however, some cases require subculture and days to weeks for definitive identification. This demonstrates a potential advantage of the experimental system in providing rapid identification in some cases. However, it should be noted that direct sequencing of such isolates, as well as the application of other technologies, may provide similar benefits. Most notably, matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) has recently been shown to be effective for the identification of both yeasts and molds (32–35). A recent study comparing PCR/ESI-MS and MALDI-TOF MS in the identification of yeast and bacteria showed that both have similar performances for routine use (33). Given that MALDI-TOF uses proteomic information from intact cells, limitations may include variability in protein expression and the integrity of intact cells (34). PCR/ESI-MS uses genetic information, potentially expanding its utility for epidemiology and infection control, as well as for the identification of uncultivable organisms (33). Another potential advantage of PCR/ESI-MS lies in its sensitivity and the ability (as previously demonstrated with other infectious agents) to directly detect pathogens without prior culture. The present work is the first step in establishing such capability. Having shown effective identification and analytic sensitivity in the present study, future studies will address direct detection from clinical samples. The present study was designed primarily to assess accuracy of identification compared to reference methods (sequencing) and concordance with phenotypic methods, and so limited conclusions can be reached regarding direct detection capabilities. Spectrum and number of individual genospecies of fungi tested were also limitations, as was the use of phenotypic testing as a reference method. The number of clinical isolates tested here was large, but the vast spectrum of phylogenetic heterogeneity seen among even clinically relevant fungi is such that no one study can hope to completely cover it with any degree of redundancy that might be desired when assessing a clinical assay. Studies such as ours will need to be reinforced with others over time, particularly as the PCR/ESI-MS database is expanded. Finally, in order to encompass the breadth of organisms that were included here, multiple centers were necessarily involved. This likely resulted in some lack of uniformity in phenotypic identification, which was unavoidable but must be recognized when evaluating results.

PCR/ESI-MS is capable of identifying a broad range of medically important fungi, including clinically relevant yeasts and molds. The availability of a commercially available system that incorporates both the sensitivity of PCR and the specificity of ESI-MS in a broad-range system could have a significant impact on the care of immunocompromised patients. Future studies will be required to confirm findings related to the identification of culture isolates and to explore the use of this system as a means of detection and characterization of fungal pathogens directly from clinical specimens.