INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The incidence of infections caused by non-Candida albicans Candida species is steadily increasing in AIDS and cancer patients (10, 14). Among these non-C. albicans species that cause infections, Candida glabrata is important because it is primarily resistant to fluconazole and therefore has recently been the focus of intensive research (15, 24, 25). The epidemiology of C. glabrata infections needs to be studied by obtaining measurements of the genotypic variation because intraspecific phenotypic differentiation is negligible. Although typing of C. glabrata intraspecific variants can be done by randomly amplified polymorphic DNA (RAPD) analysis, restriction fragment length polymorphism (RFLP) analysis, and molecular karyotyping, it is challenging to find adequate primers, sites, and probes that will enable discrimination of these closely related isolates (18, 26). Also, most differences in band patterns are difficult to interpret because band intensity and reproducibility are not absolutely consistent (23, 26). Direct comparison of DNA sequences is preferable for typing of these pathogenic yeasts because methods based on comparison of patterns (e.g., RAPD analysis, RFLP analysis, pulsed-field gel electrophoresis, and multilocus isozyme analysis) are indirect measurements of genetic divergence and can be affected by paralogy and nonindependence of characters (31). Differences in band patterns are much more difficult to analyze under probabilistic models of divergence, which are routinely used for direct sequence comparisons, such as substitution matrices used in maximum likelihood analysis (31). This limitation is due to the difficulty in attributing state-transition probabilities of pattern data as opposed to nucleotide data. The probabilistic approach is required to put the epidemiological transmission framework into a divergence-time perspective, because divergence times can be estimated from phylogenies inferred from DNA sequences. Accordingly, to build the phylogeny that reconstructs the path of orthologous steps that led to the observed divergence, the observed substitutions must be corrected by probabilistic models because of reverse, parallel, and convergent substitutions (17). Comparative analysis of molecular sequences has been used for a variety of taxonomic groups to determine relatedness (34). The rRNA subunit genes have extensively been used as macroevolutionary markers of microorganism phylogeny and taxonomy, including those of pathogenic yeasts, because of functional equivalence, size, and universal distribution (1, 33). However, intraspecific sequence comparisons of nuclear genes in eukaryotes, especially in those genes for which divergence was very recent or the evolutionary rate is very small, exhibit few or no substitutions which preclude the use of such sequences for typing. Mitochondrion-encoded genes evolve approximately 10-fold faster than nuclear genes due to low-fidelity replication, defective repair, and high concentrations of mutagens in mitochondria and therefore are more suitable for use for the resolution of close phylogenetic relationships (2, 3).

Here we tested whether sequences of the mitochondrion-encoded cytochrome c oxidase subunit 2 gene (COX2) could be used to discriminate intraspecific variants of the pathogenic yeast C. glabrata. We typed 23 different strains by comparison of COX2 sequences and found 13 haplotypes. Our data enabled the differentiation of clinical variants on the basis of 2.1% variant positions of 756 positions compared and suggest that at least two positions contain synonymous substitutions that reflect regional differences.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Clinical isolates. All Brazilian strains were obtained from the fungus collection of the Laboratório Especial de Micologia (LEMI), Disciplina de Doenças Infecciosas e Parasitárias (DIPA), Departamento de Medicina, Universidade Federal de São Paulo, and were collected from cancer, diabetes, and intensive care unit patients (abdominal surgery) (labeled C, D, and I in Table 1) from 29 December 1994 to 28 May 1997. Isolates from the United States were provided by David Perlin, Public Health Research Institute, New York University. The U.S. isolates were obtained from cancer patients at the Memorial Sloan Kettering Cancer Center in New York City over a 6-month period from 1 July 1998 to 31 December 1998 (Table 1).

Isolation of total DNA and mtRNA. Total genomic DNA was isolated by the fast miniprep protocol as described previously (32). Mitochondrial RNA (mtRNA) was isolated by a modification of the method described by Defontaine and collaborators (6). Briefly, yeasts were grown in YPD (1% yeast extract, 2% peptone, 2% dextrose) overnight at 30°C and were harvested by centrifugation at 500 × g for 5 min. Pellets were washed twice with water and once with SEM (1.2 M sorbitol, 50 mM EDTA, 2% mercaptoethanol), resuspended in 5 ml of Sol A (0.5 M Sorbitol, 10 mM EDTA, 50 mM Tris [pH 7.5]) containing 2% mercaptoethanol and 0.2 µg of Zymolyase 20T (ICN) per ml, and incubated at 37°C for 45 min with gentle agitation to digest the cell walls. The resulting spheroplast suspension was sonicated at 19 kHz with pulses of 300 ms/s for 1 min. The lysate was centrifuged at 1,000 × g for 10 min and the supernatant containing the mitochondria was centrifuged at 15,000 × g for 15 min. After the crude mitochondrial pellet was washed four times with Sol A it was resuspended in 1 ml of Trizol (GIBCO) and vortexed, and 200 µl of chloroform was added. After incubation for 2 min at room temperature the suspension was centrifuged at 12,000 × g at 4°C for 15 min. After precipitation with 500 µl of isopropanol and washing with 70% ethanol, the mtRNA was resuspended in 40 µl diethyl pyrocarbonate (DEPC)-treated water and quantitated by spectrophotometry as described previously (29).

Amplification (PCR) and sequencing. The complete coding region of COX2 was amplified in 50 µl of the PCR mixture with Taq DNA polymerase buffer (50 mM KCl, 1.5 mM MgCl₂, 10 mM Tris-HCl [pH 9.0]), 0.4 mM deoxynucleoside triphosphates, 2 mM MgCl₂, 120 pmol of each primer, 1.5 U of Taq DNA polymerase (Pharmacia), and 1.5 µg of total cellular DNA. Cycling conditions were 94°C for 7 min and 45 cycles of 94°C for 1 min, 42°C for 1 min, and 72°C for 1 min, followed by a final extension at 72°C for 7 min in a Perkin-Elmer 9600 thermocycler.

Comparative sequence analysis and phylogenetic inference. The sequences of both strands from each cloned amplicon were obtained in duplicate from different PCRs and were assembled into single contigs corresponding to each individual strain by using SeqMan from Lasergene-DNASTAR package (DNASTAR Inc., Madison, Wis.). Alignments were done by using the Clustal algorithm in the DNASTAR package (MegAlign) (16), and manual corrections were done by using the Seaview sequence editor for UNIX (12). Phylogenetic analysis was done by using Phylo_Win for UNIX by using the maximum likelihood and neighbor-joining algorithms (9, 12, 27). The substitution model used was F84 with transversion/transition ratios of 2.0 and 1.13, as inferred from the data. Bootstrap analysis (8) was done with 500 replications by using Phylo_Win. The small-subunit rRNA sequences were downloaded from GenBank and were aligned by using as a guide the general alignment of the Ribosomal Database Project (http://www.cme.msu.edu/rdp/) (21).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

COX2 primer design and amplification. The C. glabrata COX2-specific primers COF (5'-ATGTTAAATTTATTATATAA-3') and COR (5'-TTATTGTTCGTTTAATCATTC-3') were designed to amplify the entire coding region, and their sequences were determined from the published sequence (GenBank accession no. X69430) (4). For all 10 C. glabrata strains the 750-bp fragment was amplified from total cellular DNA (Fig. 1). Also, as expected, these primers did not amplify S. cerevisiae COX2 (Fig. 1A, lane 11) or C. albicans (data not shown) because of highly divergent sequences in the N terminus of the COX2 protein, which corresponds to primer COF. Because nuclear pseudogenes can be coamplified with the expressed mitochondrial copies, we also amplified COX2 from isolated mtRNA and converted it to cDNA with oligo(dT) due to polyadenylation of mitochondrial transcripts (Fig. 1B). As shown in Fig. 1B (lane 5), a single fragment of 750 bp was amplified from oligo(dT), and no fragments were amplified from the negative control (Fig. 1B, lane 4), which consisted of PCR with COF and COR primers from a mock RT reaction containing no reverse transcriptase. This result shows that the amplified fragment of 750 bp from oligo(dT) cDNAs was reverse transcriptase dependent and, therefore, was not derived from contaminating DNA in the mtRNA preparation.

View larger version (35K): [in this window] [in a new window]   FIG. 1.   PCR and RT-PCR of COX2 genes from total cellular DNA and mtRNA. (A) COX2 was amplified as a 756-bp product from total cellular DNAs of 10 strains of C. glabrata (lanes 1 to 10, respectively) but not from S. cerevisiae total cellular DNA (lane 11). C. glabrata DNA samples correspond to strains 05, 06, 12, 13, 36, 37, 38, 39, 47 and ATCC 90030, respectively (TABLE 1). (B) COX2 was amplified from purified mtRNA of C. glabrata 06 (Table 1) (lanes 5 and 7). The cDNAs were synthesized either from oligo(dT) primers (lane 4) or from the COR primer (lane 7). Negative controls consisted of mock RT reactions with oligo(dT) (lane 4) or COR primer (lane 6) lacking reverse transcriptase to demonstrate that the COX2 product was reverse transcriptase dependent. Amplification of actin cDNA from oligo(dT) was used to control for RNA integrity, RT efficiency, and contamination with nonmitochondrial mRNA (lane 3) with the corresponding mock RT reaction-negative control to exclude the possibility of DNA contamination (lane 2). Amplification of COX2 from total cellular DNA was used as a positive control for PCR (lane 1). The identities of the COX2 and actin cDNA amplicons were checked by sequencing. Lane M, molecular size markers (phage lambda DNA cut with HindIII for panel A and a 100-bp ladder [GIBCO BRL] for panel B).

Intraspecific sequence comparison of COX2. The amplicons shown in Fig. 1A (lanes 1 to 10) and Fig. 1B (lane 5) were cloned and sequenced. Alignment and comparison of these sequences shows that the COX2 fragments amplified from total cellular DNA correspond to expressed copies and not nuclear pseudogenes because the sequence of ATCC 90030 is identical in both mitochondrial cDNA and mitochondrial DNA (Fig. 2, second and third sequences in the alignment). A frameshift mutation at position 673 is present in all sequences, including that of the ATCC 90030 control, except strain CBS 138 and S. cerevisiae. This frameshift mutant in C. glabrata sequences (C insertion in position 673) is present in the mRNA, which suggests that frameshift mutation suppressor mechanisms, such as a suppressor tRNA and RNA polymerase slippage, might exist in the mitochondria of C. glabrata to yield a functional COX2 peptide. Substitutions between the different strains are summarized in Fig. 3 and were confirmed by sequencing both strands in duplicate and from independent PCRs to exclude Taq DNA polymerase artifacts and heteroplasmic effects. Also, identical sequences were obtained by direct sequencing of PCR products.

View larger version (114113110K): [in this window] [in a new window]   FIG. 2.   Alignment of COX2 sequences from C. glabrata clinical isolates and those from previously published strain CBS 138 and S. cerevisiae (S.c). Sequence names correspond to the names in Table 1 (C.g, C. glabrata). The cDNA sequence (second from top to bottom) was determined from strain ATCC 90030. The frameshift mutation (C insertion) is located at position 673. Boxed residues indicate differences from the consensus sequence.

View larger version (58K): [in this window] [in a new window]   FIG. 3.   Summary of alignment of variant positions of C. glabrata COX2. Open boxes indicated the synonymous substitutions, and underscores indicate the nonsynonymous substitutions. The asterisks at the bottom indicate the substitutions at positions 51 and 519 that separate type 1 strains from type 2 strains.

Evolutionary analysis of COX2. Phylogenetic analysis was used to estimate the rate of evolution of COX2 in yeasts and, consequently, the divergence times between different species and isolates. This is important for epidemiological investigations because it can provide estimates on how likely it is that different patients have acquired C. glabrata from single or different sources. The sequences of 18S ribosomal DNA (small-subunit rRNA gene) of five representative yeast species, K. thermotolerans, K. lactis, S. cerevisiae, S. exiguus, and C. glabrata, were aligned by using as a guide the general small-subunit rRNA alignment from the Ribosomal Database Project (21). The phylogenetic distances were inferred for these species by using the Kimura-2-parameter model for nucleotide substitution and the divergence times calculated by using the expression T = K_SSUrRNA/2r_SSUrRNA (20) (where T is the divergence time, K_SSUrRNA is the corrected phylogenetic distance for the small-subunit rRNA genes, and r_SSUrRNA is the evolutionary rate of the small-subunit rRNA gene, which is 0.85%/10⁸ years [7]). The evolutionary rate of COX2 genes in the yeast group (11.4%/10⁸ years) was estimated by using the alignments of the COX2 sequences. The phylogenetic distances of COX2 genes (Kimura-2-parameter model) between K. thermotolerans, K. lactis, S. cerevisiae, S. exiguus, and C. glabrata were used in the expression r_COX2 = K_COX2/2T (where r is the evolutionary rate, K is the phylogenetic distance, and T is the divergence time estimated from the 18S rRNA comparisons). Because r_COX2 is 11.4% sequence divergence/10⁸ years, we estimate that the comparison of COX2 genes would be able to detect mutations that occurred at least 570,000 years ago. One mutation in 756 bp represents a distance of 0.013% sequence divergence, and therefore, use of the expression T = K_COX2/2r_COX2, with r_COX2 equal to 11.4%, gives a value of T of 0.57 × 10⁶ years.

View larger version (18K): [in this window] [in a new window]   FIG. 4.   Correlation between percent COX2 sequence distances and taxonomic distances. Points in the plot indicate the average distance for the pairwise comparison for a given category, and error bars indicate the associated standard deviation.

View larger version (26K): [in this window] [in a new window]   FIG. 5.   Interspecific and intraspecific phylogenetic analysis of COX2 genes of Candida glabrata and closely related taxa. (A) Maximum likelihood tree for several fungi, rooted for S. pombe and N. crassa, to compare the average intraspecific distances of C. glabrata strains from their closest relatives. (B) Maximum likelihood phylogeny of C. glabrata strains supporting the existence of two types with a bootstrap value of 89% in 500 replications. The same topologies for trees in panels A and B were observed when neighbor-joining algorithms with HKY distances were used. The scale bars below the trees indicate percent divergence, and the numbers above the tree nodes indicate the percentage of that particular branch cluster in 500 bootstrap replications. Species abbreviations are as follows: S-c, S. cerevisiae; S.e, S. exiguus; W.su, W. saturnus var. suaveolens; W.sa, W. saturnus var. makrii; D.b, D. bruxellensis; K.l, K. lactis; K.t, K. thermotolerans; P.j, P. jadinii; C.g, C. glabrata; B.ci, B. custersii; N.c, N. crassa; Sc.p, S. pombe.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Typing of pathogenic yeasts is relevant for epidemiology because it helps to identify sources of contamination, detect outbreaks in hospital environments, and identify strains that are either more virulent or more resistant to antifungal drugs. Typing is dependent on the intraspecific variation in a species and is usually done by methods that use comparative analyses of profiles, such as RAPD and RFLP analyses. Despite their easy applicability, these methods are subjected to several systematic errors due to the nonindependence of characters (31). In the present work we tackled the problem of typing intraspecific variants of C. glabrata by comparison of COX2 sequences. A total of 16 of 756 (2.1%) positions were variable in our sample, and 13 different haplotypes were identified. Phylogenetic analysis of these data suggests that the 13 haplotypes can be grouped into two types, types 1 and 2, which is supported by 89% bootstrap replications. These two types are basically identified by synonymous substitutions at positions 51 and 519. These polymorphic positions could be correlated with the geographical origins of the strains because 82% of Brazilian isolates belong to type 2 and 90% of the U.S. isolates belong to type 1. The U.S. strains are more closely related to the American Type Culture Collection strain used as a reference and an internal control for sequencing. One explanation for the two Brazilian strains that have a type 1 COX2 gene and the U.S. strain that has a type 2 COX2 gene might be relatively recent migratory events. However, it would be ideal for more sequences to be studied to confirm or refute the expected hypothesis suggested by our data regarding the association of type 1 COX2 with U.S. strains and type 2 COX2 with Brazilian strains. None of the nonsynonymous changes affect the amino acid residues that are essential for the correct structure and function of the Cox2 peptide (Fig. 3) (13).

Strain CBS 138, whose sequence has been published previously, represents the most distantly related isolate in the data set. It has a C residue at position 51 and an A residue at position 52. This strain also does not have the frameshift mutation at position 673 that we observed in all other strains, including the American Type Culture Collection control strain. The lack of this frameshift mutation is observed in S. cerevisiae as well, which suggests that among C. glabrata strains there might exist a group of strains that are more closely related to S. cerevisiae and a group of strains that are more distantly related and that contain the frameshift mutation at position 673. Because we observed no alterations in the phenotypes of our isolates growing in culture under aerobic conditions, we conclude that the COX2 gene product of strains that contain the frameshift mutation must be a functional peptide. This frameshift mutation without suppressor mechanisms would originate a truncated protein without the copper binding domain which encompasses two histidines and two cysteines in the carboxy-terminal end (13). In fact, mitochondrial gene frameshift mutations and corresponding suppressor mechanisms have already been described in yeast mitochondria (11, 19, 28).

The amplification from total cellular DNA yields a fragment that corresponds to the expressed copy of COX2 and not a nuclear pseudogene. Sequence divergence analyses that use mitochondrial DNA are subject to this type of artifact, particularly when the mitochondrial gene recently moved to the nucleus. To verify whether pseudogenes are coamplified, it has been proposed that mitochondrial genes should be amplified from poly(A)⁺ mRNA (5). Although the level of DNA contamination in our mtRNA preparation is below the level of detection, some nuclear gene-encoded mRNA is present, as seen for the actin control (Fig. 1B, lane 3). Since nuclear pseudogenes do not have corresponding transcripts, this assay readily enabled us to control for this type of artifact.

We believe that the comparative sequence analysis of COX2 can be used to discriminate major subgroups of strains over broad geographic locations. Considering the evolutionary rate for COX2 that we calculated on the basis of calibrations for 18S ribosomal DNA, we estimate that the most closely related strains in our sample diverged approximately 5 × 10⁵ years ago and that the divergence between type 1 and type 2 strains occurred 1 × 10⁶ years ago. Strain CBS 138 would have diverged 2 × 10⁶ years ago. For typing of more closely related isolates, we need to use even faster markers such as the intergenic regions in the mitochondrial DNA. Nevertheless, the present work represents the first example of the typing or the intraspecific differentiation of pathogenic yeasts by direct comparison of a DNA sequence, with additional assessment of results of polymorphism analysis under a phylogenetically consistent probabilistic model.