Population Structure and Properties of Candida albicans, as Determined by Multilocus Sequence Typing
http://www.100md.com
微生物临床杂志 2005年第11期
Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD
The Peter Medawar Building for Pathogen Research and Department of Zoology, University of Oxford, OX1 3SY Oxford, United Kingdom
ABSTRACT
We submitted a panel of 416 isolates of Candida albicans from separate sources to multilocus sequence typing (MLST). The data generated determined a population structure in which four major clades of closely related isolates were delineated, together with eight minor clades comprising five or more isolates. By Fisher's exact test, a statistically significant association was found between particular clades and the anatomical source, geographical source, ABC genotype, decade of isolation, and homozygosity versus heterozygosity at the mating type-like locus (MTL) of the isolates in the clade. However, these associations may have been influenced by confounding variables, since in a univariate analysis of variance, only the clade associations with ABC type and anatomical source emerged as statistically significant, providing the first indication of possible differences between C. albicans strain type clades and their propensity to infect or colonize different anatomical locations. There were no significant differences between clades with respect to distributions of isolates resistant to fluconazole, itraconazole, or flucytosine. However, the majority of flucytosine-resistant isolates belonged to clade 1, and these isolates, but not flucytosine-resistant isolates in other clades, bore a unique mutation in the FUR1 gene that probably accounts for their resistance. A significantly higher proportion of isolates resistant to fluconazole, itraconazole, and flucytosine were homozygous at the MTL, suggesting that antifungal pressure may trigger a common mechanism that leads both to resistance and to MTL homozygosity. The utility of MLST for determining clade assignments of clinical isolates will form the basis for strain selection for future research into C. albicans virulence.
INTRODUCTION
Differentiation of microbial isolates by sequencing a small sample of unrelated housekeeping genes has become established as a reliable and effective method for typing strains of many bacteria (9, 50). Such multilocus sequence typing (MLST) is highly reproducible, and data can be archived in Web-based databases accessible to all users. For the fungal pathogen Candida albicans, an MLST system based on seven DNA fragments was developed as an optimal consensus for typing strains within the species (6) following two earlier proposed systems (5, 48). Because C. albicans is a diploid organism, sequence data contain heterozygous as well as homozygous sites, adding an extra discriminatory feature to MLST for this species.
Among other approaches to strain typing of C. albicans, DNA fingerprinting based on the moderately repetitive sequence Ca3 has been widely used. By this method, C. albicans populations were shown to comprise five major clades of closely related strain types (46), including clades enriched in isolates from Europe (36) and South Africa (2, 3). Resistance of C. albicans to flucytosine in vitro was found to be a property restricted almost entirely to isolates from a single C. albicans clade as determined by Ca3 fingerprinting (37), and the sole mechanism of resistance to flucytosine within this clade was the mutation of a single nucleotide in the FUR1 gene (14).
Typing by a combination of restriction fragment polymorphisms and randomly isolated polymorphic DNA also showed geographical enrichment of groups of related isolates (10), and "ABC typing," based on the size and number of internally transcribed spacer regions (ITS1) in DNA encoding rRNA showed a higher proportion of flucytosine-resistant isolates among type A strains (26).
We have now typed more than 400 isolates of C. albicans by MLST. Our data show how MLST reveals a population structure with four major clades of related strain types plus at least eight minor clades. It confirms the predominant association of flucytosine resistance with one major clade and reveals associations between clades and their geographical, anatomic, and temporal origins.
MATERIALS AND METHODS
C. albicans isolates. We typed 416 isolates from a large range of sources. Each isolate was from a separate source. Table 1 summarizes the details of the panel of isolates tested. Data for isolates described by Bougnoux et al. in their MLST work (5) were included in the present study; for these isolates, only extracted DNA had been provided by Marie-Elisabeth Bougnoux and Christophe d'Enfert for MLST (6), so we could not conduct antifungal susceptibility tests with these isolates. A set of 25 isolates was provided by David R. Soll (Table 1); these comprised five isolates from each of the five major C. albicans clades as determined by Ca3 fingerprinting (46) and served as a reference for cross-validation of typing methodologies. Isolate WO1, the original "white-opaque switcher" now undergoing full-genome sequencing, was also originally supplied by D. R. Soll.
A set of 36 isolates was provided by Jan Schmid; 35 of these were a subset of isolates that had been typed by Ca3 fingerprinting (42, 43). Another set of 22 clinical isolates was provided by James B. Anderson from his group's study of population genomics of fluconazole resistance in C. albicans (12, 13). Other notable isolates included SC5314, which was used for whole-genome sequencing of C. albicans (21) and is the wild-type isolate used widely for gene disruption research (16); a representative from the set of 16 serial isolates with increasing fluconazole resistance from a patient infected with human immunodeficiency virus (HIV) (38, 51), provided by Theodore White; isolate NCPF3153, often referred to as 3153A and used in many laboratories; and isolate MYA-2669, the type strain of the putative species Candida africana (49), purchased from the American Type Culture Collection. A large set of isolates was provided from the collection of the Belgian Institute for Hygiene, Epidemiology, and Medicine (IHEM) through the collaboration of Nicole Nolard and Franoise Symoens; some of these were subcultures originally obtained from the senior author's collection. Three isolates that had been used to study mechanisms of fluconazole resistance (41) were originally contributed by Dominique Sanglard.
Recent clinical isolates of C. albicans were provided by Elizabeth M. Johnson (Mycology Reference Laboratory, Bristol, United Kingdom) and Christopher C. Kibbler (Royal Free Hospital, London, United Kingdom) from their diagnostic mycology laboratories. Fresh oral commensal isolates of C. albicans were obtained from healthy undergraduate student volunteers in Aberdeen, United Kingdom. The remainder of the isolates typed came from the senior author's collection of C. albicans isolates, which was started in the 1970s. This collection, originally stored under distilled water (29), was transferred twice in its entirety, once for an investigation of the prevalence of the species C. dubliniensis in the collection (32). Some of these isolates had been previously described in epidemiological studies of C. albicans based on phenotypic typing tests (30, 31). The quality of the collection was confirmed by the near-identity of MLST data for several isolates from the same source (details not shown).
Isolates were maintained on Sabouraud agar (Oxoid, Basingstoke, United Kingdom) for the duration of this study.
MLST. Sequences of bases in the gene fragments AAT1a, ACC1, ADP1, MPIb, SYA1, VPS13, and ZWF1b were determined as previously described (6, 48). Isolates were designated as genotype A, B, or C on the basis of PCR for rRNA genes (26), and the status of the mating-type-like locus (MTL) in each strain was assessed as heterozygous (a/), homozygous (a/a), or homozygous (/) by PCR as previously described (48).
Sequencing of other genes. Genes FCY22, FCA1, and FUR1 encode, respectively, cytosine permease, cytosine deaminase, and uracil phosphoribosyltransferase in the pathway for conversion of flucytosine to the active antifungal 5-fluorouracil in C. albicans (14). For selected C. albicans isolates, the DNA sequences of these genes were determined by the same method as that used for MLST (above). For FCY22, the forward primers used were FCY22-1 (5'AACATAATCTGTAGTTATATCTC), FCY22-2 (ATTATTGGAAGCATTCCAGTG3'), FCY22-3 (5'TCCAACTGAAGCTGGTAATG3'), and FCY22-4 (TGCCATTCAATGTATATGGAG), and the reverse primer was 5'TAACAGGGCAGAATGAATCG3'. For FCA1, the forward primers were FCA1-1 (5'AATTCTCTTTCAATTCTTCTTG3') and FCA1-2 (5'TCACCATGTAGTATGTGTAC3'), with the reverse primer being 5'TTGAACAAGATGATGATGTTG3'. For FUR1, the forward primers were FUR1-1 (5'ATAATGGAGCATCTCGCAAC3') and FUR1-2 (5'TAGCTGAAGATATCAGTGAAC3'), with the reverse primer being 5'TGTTGTACATATAATCCTATAAG3'.
Susceptibility testing. MICs of fluconazole, flucytosine, and itraconazole were determined by microdilution tests by a previously described method and spectrophotometric endpoints (33) that give MIC results compatible with those obtained by the NCCLS reference method (27). Isolates with fluconazole MICs of 16 μg/ml or higher, flucytosine MICs of 8 μg/ml or higher, or itraconazole MICs of 0.25 μg/ml or higher were classed as resistant to these agents (these breakpoints indicate full susceptibility to the agents concerned [27]).
Data analysis. Sequence data were analyzed by scrutiny of capillary sequencer output chromatograms with DNASTAR Seqman software. Each unique (diploid) sequence for each of the seven DNA fragments was given a separate identifying number. Each unique combination of seven sequences was assigned a separate diploid sequence type (DST) number. The sequence and DST identifiers were compiled on a central database (http://test1.mlst.net/).
The eBURST package (http://eburst.mlst.net/ [15]) was used to determine putative relationships between isolates. With C. albicans data, this software scans allele sequences from pairs of isolates and records isolates as related when six of the seven sequences are identical between a pair. The eBURST algorithm places all related isolates into clonal complexes and, where possible, predicts the founding, or ancestral DST of each complex. The output is a display of the most parsimonious patterns of descent of each DST from the ancestral type.
Phylogenetic analyses by unweighted pair group method with arithmetic averages (UPGMA) and neighbor-joining algorithms were conducted with MEGA version 2.1 (22) applied to modified sequence data. The analyses were based only on the results for polymorphic bases to maximize their power to discriminate between isolates. The data set of only the variable bases was suitable for pairwise-difference analysis, which has been used previously with C. albicans MLST (5, 48). To obtain sequences that could be handled by the MEGA software, which is not programmed to analyze heterozygous code data, the following procedure was used. The results for the variable sites from the seven gene fragments sequenced were concatenated into a single sequence. For any pair of isolates, each with a diploid genome, the base at each polymorphic site could be homozygous and identical between the isolates, heterozygous and identical, homozygous and different, or heterozygous in one isolate and homozygous in another. For example, the sequencing result (a IUPAC single-letter code) for a given base across a set of strains might appear as A, T, or W (= A + T). Data from the polymorphic sites from the seven C. albicans alleles were therefore conjoined into a single sequence, and then each base in the sequence was rewritten twice for a homozygous (A, C, G, or T) datum or once each for the two component bases for a heterozygous (K, M, R, S, W, or Y) datum. These revised sequences could then be analyzed to generate dendrograms in MEGA 2.1. In this form, the analysis was the functional equivalent of scoring a pair of results as 1 for homozygous or heterozygous identical data, 0 for homozygous different data, and 0.5 when one polymorphic site had a heterozygous result and the other a homozygous result and then creating a difference matrix. Use of the modified sequences, rather than of difference matrices, enabled calculation of statistical significance for the cluster nodes by bootstrapping with 1,000 replications.
Distributions of properties between C. albicans clades were determined statistically by Fisher's exact test. Results were considered significant for a P of <0.05. A univariate analysis of variance (ANOVA) was used to determine associations between clades and A, B, or C type (with 1, 2, and 3 corresponding to A, B, and C, respectively); MTL status (with 1, 2, and 3 corresponding to a/, a/a, and /, respectively); fluconazole, itraconazole, and flucytosine susceptibilities (with 1 indicating susceptibility and 2 indicating resistance); decades of isolation (1 being a year prior to 1990, 2 being a year in the 1990s, and 3 being a year in the 2000s); geographical origin (1 being the United Kingdom, 2 being another European country, 3 being North America, and 4 being another country); and anatomical origin (1 being blood or other sterile site, 2 being the oropharynx, 3 being the vagina, and 4 being another location). Isolates from animals were excluded from this analysis.
Discriminatory power was calculated according to Hunter (20).
RESULTS
Reproducibility and sequencing errors. A total of 13 different C. albicans isolates was submitted for MLST in duplicate on different occasions, and the person conducting the sequencing was unaware of their identities. The seven genes sequenced yielded a total of 109 variable sites. Of the 13 isolates tested in duplicate, three yielded different DSTs. In two cases, the difference resulted from a single-base homozygous/heterozygous difference in one of the seven genes; in the third there were homozygous/heterozygous discrepancies at two sites in a single gene. Thus, in a total of 1,417 variable sites tested in duplicate, a sequencing error arose on four (0.28%) occasions, indicating a reproducibility of 99.72%. Sequence differences greater than three homozygous/heterozygous discrepancies between isolates were therefore assumed to indicate true interstrain differences.
Statistics of C. albicans MLST. The 416 isolates yielded 351 distinct DSTs, with a discriminatory power index of 0.9996. Among the seven fragments sequenced, ZWF1 gave the highest discriminatory ratio, yielding 84 different sequences from just 10 polymorphic sites, followed by SYA1 (70 alleles from 14 variable sites), VPS13 (103 alleles from 21 variable sites), AAT1a (57 alleles from 13 variable sites), ACC1 (37 alleles from 10 variable sites), ADP1 (46 alleles from 18 variable sites), and MPIb (36 alleles from 21 variable sites).
Details of the alleles and DSTs for each isolate are shown in the full version of Table S1 in the supplemental material, together with the geographical and anatomical sources of the isolates. Insufficient DNA for sample BougnCP10 precluded determination of ABC and MTL type for this isolate. Among the remaining 415 isolates, 281 (67.7%) were type A, 96 (23.1%) were type B, and 38 (9.2%) were type C. The great majority of isolates (371, or 89.4%) were heterozygous at the MTL, while 27 isolates (6.5%) were of homozygous MTL type / and 17 isolates (4.1%) were of type a/a.
Population structure of 416 unrelated C. albicans isolates. MLST data allow for different approaches to phylogenetic analysis of various degrees of sophistication. We utilized several approaches, with an emphasis on methods based on pair differences at polymorphic nucleotide sites and on gross allele sequence differences in an effort to determine robust subspecific clades that will allow easy assignment of future isolates to the population structure.
eBURST analysis of the set of 416 C. albicans isolates from unique sources revealed 18 clonal clusters, with 5 being based on six or more DSTs (Fig. 1). Cluster 1, the largest, comprised 105 isolates (25.2% of all isolates), with DST 69 as the putative ancestral type. Clonal cluster 2, the second largest, comprised 38 isolates and 27 DSTs and no putative founding DST (type 155 was represented by 7 isolates but did not emerge as a likely ancestor for the cluster). Clonal cluster 3 comprised 28 isolates and 17 DSTs, with type 124 being the predicted founding type. Cluster 4 comprised eight isolates and seven DSTs, with type 305 being the likely ancestral DST, and cluster 5 comprised seven isolates and six DSTs, with no putative ancestral type. The remaining clonal clusters consisted maximally of six isolates.
Of the 416 isolates analyzed by the eBURST software, 228 (54.8%) were assigned to clonal clusters, thus leaving almost half of the isolates with singleton status (Fig. 1).
Unlike the eBURST analysis, a UPGMA pairwise-difference dendrogram for the 416 separate-source isolates (Fig. 2) took no account of the gene fragment in which sequence polymorphisms arose and therefore provided a different perspective on relationships between isolates: isolates with small numbers of differences in the concatenated sequences for variable sites across all fragments sequenced were regarded as similar by UPGMA analysis, even if the polymorphic sites were located on separate fragments. Comparison of the clustering in the UPGMA dendrogram with the eBURST clonal clusters (Fig. 2) showed clonal clusters 11 and 17 and several eBURST singletons dispersed within the broad range of isolates in clonal cluster 1. Similarly, isolates in clonal clusters 5, 10, and 13 were mixed with each other and with several eBURST singletons by UPGMA analysis.
Although the UPGMA dendrogram for the concatenated variable sequences from the 416 separate-source isolates clearly suggested patterns of clustering between the isolates, bootstrap values for cluster nodes were generally low except for a few pairs of very closely related isolates. The following procedure was therefore used to determine which isolates were most likely to belong to clusters of highly related strains. The information from Ca3 typing, particularly the results for isolates belonging to known Ca3 clades (46), was considered in conjunction with clonal cluster information to determine a cutoff point that could be used to define UPGMA clades and outlying singletons.
All five of the isolates from clade I as determined by Ca3 typing (46) clustered within a large group of isolates closely related by UPGMA (Fig. 2); four of the five belonged to clonal cluster 1 by eBURST analysis. Similarly, 19 of the 21 isolates designated types A1 and A2 by Ca3 fingerprinting and designated by neighbor-joining statistics (42, 43) appeared in the same large UPGMA cluster that was broadly delimited by eBURST clonal cluster 1. All 155 isolates from AM2003/0046 to AM2004/0049 in Fig. 2 were therefore provisionally assigned to clade 1 as demarcated by MLST. The five Ca3 clade II isolates and both Ca3 group B isolates correlated with a well-delineated UPGMA cluster (Fig. 2); four of these seven isolates appeared in eBURST clonal cluster 2. Isolates from AM2004/0030 to IHEM17110 in Fig. 2 were therefore provisionally assigned to MLST clade 2. Similarly, the five isolates from Ca3 clade SA clustered together in the UPGMA dendrogram, with four of the five in clonal cluster 3. However, the UPGMA cluster designated MLST clade 4 in Fig. 2 (isolates HUN68 to IHEM16731), though appearing reasonably homogeneous, contained a subcluster comprising two of the Ca3 clade E isolates, and all of the members of clonal cluster 15. The five isolates from Ca3 clade III emerged in clonal clusters 9 or 10 or as singletons by eBURST analysis, but they clustered together in the UPGMA dendrogram in a group of isolates with the same overall level of similarity as those in MLST clades 1, 2, and 4. Isolates AM2003/0082 to b30998/5 were therefore provisionally considered to belong to MLST clade 3. It was noted that isolates previously designated as group C by neighbor-joining analysis of Ca3 typing (42, 43) did not show obvious affinities with eBURST- or UPGMA-defined clusters.
The provisional assignment of isolates to clades 1 to 4 indicated a maximum dissimilarity of 10 by pairwise-nucleotide difference analysis as calculated with the MEGA software. The following definition of a clade was therefore used: a cluster of at least five isolates with a dissimilarity not exceeding 10 by pairwise-difference analysis. By this definition, 348 (83.7%) of the 416 isolates could be assigned to 12 clades, as specified in Fig. 2. Reanalysis of the MLST variable-site data only for the 348 isolates that matched the clade definition confirmed the validity of the clade assignments in Fig. 2. The bootstrap values for the nodes defining clades were 85% (clade 1), 88% (clade 2), 87% (clade 3), 86% (clade 4), and 98 or 99% for the minor clades. The exception was clade 5, for which the node bootstrap value was only 64%, while the values for the node defining the pair of isolates 81/078 and AM2003/0084 and the node for the remaining four isolates nominated for clade 5 were 97% and 99%, respectively. These results suggest that clade 5 is a less robust grouping of isolates than the other clades and may need to be revised in future as further C. albicans isolates are added to the MLST database.
To examine the population structure of C. albicans clades by an alternative statistical approach the MLST data for the 416 isolates of separate origins were reanalyzed by determination of p-distance with the unrooted neighbor-joining algorithm in the MEGA2 software package. The result (Fig. 3) confirmed most of the isolate clusters determined by pairwise-distance UPGMA analysis (Fig. 2) and reiterated that the majority of the isolates studied could be assigned to one of four major clades. However, the neighbor-joining algorithm also produced some differences in the structure from the UPGMA analysis.
In Fig. 3, the three isolates that included the type strain of the putative species C. africana, while still forming a well-separated branch in the tree, nevertheless appeared less distant from other isolates than in the UPGMA dendrogram (Fig. 2). In the neighbor-joining tree, isolates 81/078 and AM2003/0084 clustered separately from the remainder of isolates classed as clade 5 in Fig. 2 (the position of the pair is indicated as clade 5a in Fig. 3), reemphasizing the previous conclusion that clade 5 is less robust than most others. The population structure by neighbor joining also removed isolates C82, AM2003/0065, BougnCP11, 73/025, and J981315 as a group from the rest of clade 3 in Fig. 2 to a new position shown as clade 3a in Fig. 3. The clade assignments in Fig. 2 were used for subsequent analyses of clade properties; however, it is self-evident that the most robust definition of C. albicans clades must await a larger future database of MLST results and possibly further approaches to statistical analyses.
Isolates and clusters of notable provenance. Three isolates, AM2003/0025, AM2003/0032, and MYA-2669, the type strain of the putative species C. africana, were indistinguishable by MLST but clustered with very low similarity to the other isolates in Fig. 2, confirming that MYA-2669 differs substantially from other C. albicans isolates. All three of these indistinguishable isolates came from genital swabs.
Isolate SC5314, which was used for the first whole-genome sequencing project with C. albicans (21), fell within clade 1. Isolate WO-1, which has been extensively used in research into phenotypic switching and mating in C. albicans (45), appeared as a singleton between clades 8 and 4 in Fig. 2.
Within clade 1, a group of four isolates—IHEM20416, J942149, IHEM20414, and IHEM20417—that were extremely similar by MLST was notable because all four were resistant to fluconazole, and the first three were also resistant to itraconazole. These three were / at the MTL (Table 1). All four of the isolates in this subcluster came from the mouths of AIDS patients in Germany. A similar subcluster of four closely related isolates within cluster 5, already noted above by the high significance of the bootstrap value for the cluster node, included three isolates resistant to both azole antifungal agents and three isolates that were a/a at the MTL. All four isolates came from the mouths of AIDS patients in France.
C. albicans clades: properties and clinical relevance. Table 2 summarizes the properties of the 12 individual C. albicans clades and of singleton isolates. The numbers of isolates in each of clades 5 to 12 was too small to permit reliable statistical analysis; these isolates were therefore grouped together as "minor clades" for the calculations detailed below.
Clades 1 and 2 comprised almost entirely isolates of ITS1 type A, while clade 3 isolates were almost entirely type B. Clades 5 and 6 isolates were all type B, and clade 7 isolates were all type A. Only clade 4 had a preponderance of type C isolates, but this clade also contained type A and type B isolates. The distributions of A, B, and C types between clades obviously differed significantly (Fisher's exact test, P < 0.0001).
There were 44 (10.6%) of 415 typeable isolates with homozygous alleles at the MTL. Just over half of these isolates (29) were in clade 1. The distribution of isolates homozygous at the MTL differed significantly between clades (by Fisher's exact test, P was 0.016; the test was applied to numbers of isolates that were a/, a/a, or / [Table 1]). Minor clades 8 to 11 contained no examples of MTL homozygous isolates, while a high proportion of isolates in clades 5 to 7 were homozygous at the MTL (Table 2).
Except for clade 9, the majority of isolates in each clade came from sources within Europe (information on geographical origin was available for 415 of the 416 isolates in Fig. 2). The 6 isolates in clade 5 were all of European origin, and there were no North American isolates among the 28 that made up clades 10 to 12. Table 3 breaks down the geographical-origin data more finely, but with minor clades treated as a single group. Statistical analysis of the distribution of isolate sources by clade as shown in Table 3 indicated a highly significant association of clade with geographical source (Fisher's exact test, P < 0.001). Even when the data were analyzed with French and other EU isolates combined into a single group and African isolates included with "other" geographical sources, the differential distribution of sources within each clade remained statistically significant (P < 0.001). It should particularly be noted that, among the 45 isolates in clade 4, equivalent to the "South African" clade based on Ca3 fingerprinting (2), only 3 isolates came from the African continent. The majority of African isolates in the present study belonged to clade 2.
Information on the anatomical source of isolates was available for 406 of the 416 isolates in Fig. 2. There were 11 isolates from animals, so analysis of relations between anatomical sources and clades were based on 395 isolates (Table 4). Clades 5, 6, and 9 contained no isolates from blood cultures (Table 2), and the proportion of isolates from blood and other sterile sites in clade 1 was lower than in the other major clades. The distribution of isolates from blood, the oropharynx, the vagina, and other sites varied between clades at a level that reached statistical significance (Fisher's exact test, P = 0.045).
The clade distribution of the 11 isolates from animals differed considerably from that of human isolates. Five of the isolates were singletons, four were in clade 10, and one each belonged to clades 1 and 9. This differential distribution was highly significant (Fisher's exact test, P = 0.003). Since 10 of the 11 animal isolates originated in Belgium, their clade distribution was compared separately with that of the 11 human isolates that came from Belgian patients only (5 in clade 1, 2 in clade 11, and 1 each in clades 2, 3, 4, 6, 9, and 12). Even with clades 5 to 12 treated as a single group, the distribution of animal isolates differed significantly from that of Belgian human isolates (P = 0.02).
For most of the clades, between 10% and 20% of the isolates were first cultured before 1990 (Table 2). In clades 1 to 3, 6, 7, and 9, isolates first cultured before 1990 were represented and most of the isolates were first cultured before the year 2000, while clades 4, 5, 8 and 10 to 12 bore evidence of possibly more recent origins (Table 2). The distribution of isolates of different ages differed significantly between the clades (by Fisher's exact test, P was 0.003; analysis done with clades 5 to 12 treated isolates as a single group, and dates of isolation segregated as pre-1990, 1990s, and 2000s).
A high proportion of the five isolates in clade 5 was resistant to azole antifungal agents, while clades 7 to 11 contained no isolates resistant to any of the three agents tested (Table 2). However, the overall differences in clade distributions of the 378 isolates with susceptibility data available were not significant for those that were resistant to fluconazole (Fisher's exact test, P = 0.289), itraconazole (P = 0.94), or flucytosine (P = 0.105). Other significant associations were, however, apparent for the antifungal resistant isolates. Fifteen (65%) of the 23 flucytosine-resistant isolates were members of clade 1. In every one of these isolates, flucytosine resistance was associated with a homozygous allele pair encoding cysteine at position 101 in the FUR1 gene. In a random selection of 14 flucytosine-susceptible isolates from clade 1, the equivalent position was either homozygous for arginine or heterozygous (cysteine/arginine). The eight flucytosine-resistant isolates from other clades and seven randomly selected susceptible isolates from other clades were all homozygous, encoding arginine at position 101, except for one flucytosine-resistant isolate for which no FUR1 sequence was obtained despite three attempts. No common mutation was found in the FUR1, FCY22, or FCA1 gene sequences of these non-clade-1, flucytosine-resistant isolates that might account for their resistance to the drug. Among the 35 fluconazole-resistant isolates, 17 (48.6%) were homozygous at the MTL. The same was true for 10 (41.7%) of the 24 itraconazole-resistant isolates and for 7 (21.2%) of the 23 flucytosine-resistant isolates. The prevalence of MTL homozygosity among the corresponding susceptible isolates was 7.3, 9.0, and 9.9%, respectively (by Fisher's exact test, P was <0.001, <0.001, and equal to 0.004, respectively). There was no significant association between itraconazole or flucytosine resistance and ABC type, but an unusually high proportion (40%) of 35 isolates resistant to fluconazole were ABC type B (Fisher's exact test, P = 0.013).
Twenty-eight of the 35 fluconazole-resistant isolates came from the oropharynx; all of these were from AIDS patients. For 109 oropharyngeal isolates, the HIV status of the patient source was known. The distribution of oral isolates from HIV-positive versus HIV-negative patients differed significantly between clades, as follows. For 48 HIV-negative patients, 26.8% were in clade 1, 17.9% in clade 2, 0.0% in clade 3, 17.9% in clade 4, and 23.2% in minor clades, and 14.3% were singletons. For 61 HIV-positive patients, the corresponding data were 50.7%, 9.0%, 9.0%, 6.0%, 16.4%, and 9.0% (Fisher's exact test, P < 0.001).
The individual data comparisons made by Fisher's exact test can be vulnerable to biases arising from uneven distributions of contributing variables other than the variable under analysis. A univariate ANOVA was done to determine the associations between properties and sources of isolates and their clade distribution. Animal isolates were excluded. The ANOVA was repeated with individual clades and singletons as the dependent variable and also with the minor clades included as a single variable. With individual clades analyzed, ABC type emerged as the most significant variable (P < 0.001) and anatomical source as the next most significant variable (P = 0.027). With minor clades grouped, the corresponding P values were <0.001 for ABC type and 0.065 for anatomical source. MTL status, fluconazole, itraconazole, and flucytosine resistance, decade of isolation, and geographical source were not significant clade-related variables by ANOVA.
DISCUSSION
This study has generated a considerable amount of data on the population structure of isolates of C. albicans and the relationship between clades of isolates and their properties of clinical relevance. We have provided evidence that different clades of C. albicans may differ significantly in the proportions of isolates from blood, the oropharynx, the vagina, and other sites. While, on the one hand, it is beyond dispute that the immune status of the host is the major factor determining the ability of a C. albicans to invade human tissues (8, 28), our data provide the first-ever clear suggestion that particular strain type clusters within the species may differ in their proportions of isolates from disseminated or mucosal infections. It is especially notable that, while the proportions of several properties of the isolates studied varied significantly between clades in single-factor tests, only ABC type and anatomical origin emerged as significant factors relating to clade number by univariate ANOVA. This result lends particular emphasis to the likely importance of an association between clades and putative virulence properties within the species of C. albicans.
Assignation of population structures and clade boundaries to a set of isolates cannot be done entirely objectively. The clade assignments in Fig. 2 take account of three types of information: the UPGMA dendrogram, results of eBURST analysis, and the results of Ca3 oligonucleotide fingerprinting for several of the isolates. Use of neighbor-joining statistics in place of pairwise-difference analyses further redistributed some isolates between clades (Fig. 3). We intend to evaluate a wider range of statistical approaches to clade designation, including Bayesian statistics, when our database of MLST results is at least twice as large as that reported in this study. For the time being, we consider pairwise-difference analyses to be the simplest and most useful approach for clade assignments in the population of isolates at hand.
The eBURST results were helpful in demarcating potential clades, but not definitive: many examples occurred where isolates that coclustered by Ca3 typing or by MLST analyzed by pair differences were designated as singletons by the eBURST test. It may be that the nature of sequence variation in C. albicans and the diploid genome of the species make eBURST analyses less relevant than they are to haploid bacteria (50). In common with others (4), we found an excellent correlation between UPGMA clusters for MLST and Ca3 fingerprinting data. Clades I, II, III, and SA as currently delimited by Ca3 typing (46) matched clades 1 to 4, respectively, as defined by MLST in the present study. Only MLST clade E failed to emerge as a single group of isolates as determined by MLST; the 5 representatives of Ca3 clade E split into two clusters by UPGMA analysis of MLST data (Fig. 2).
The very clear demarcation of geographical origin of strains determined by Ca3 typing as described previously (46) was not confirmed in the present study. Our panel of isolates was dominated by isolates of European origin, particularly by isolates from the United Kingdom, while the panel of isolates studied by Ca3 typing (36, 46) was dominated by isolates from South Africa. In common with the Ca3 studies, we found significant interclade differences in proportions of various geographical origins, but the relationships between clade and origin were not the same, almost certainly because of the large differences in proportions of isolates from different sources. Most notably, MLST clade 4, which corresponded to Ca3 clade SA, contained only three isolates of African origin, while most of the African isolates in our study clustered in clade 2. We have therefore chosen to refer to clades only by numbers until the MLST database contains a sufficiently large representation of isolates from all parts of the globe to allow more objective determination of possible geographic evolutionary origins of clades. The statistical significance of geographical variations between MLST clades determined by Fisher's exact test was not confirmed in the ANOVA, perhaps showing the effect of biases from differences in anatomical origin, date of isolation, etc. A larger MLST database will allow for statistical analyses of geographical origin based on, for example, only blood isolates or only oral isolates. It is to be expected that individual C. albicans clades may be enriched with isolates from a particular geographical source, and several previous studies by other DNA typing methods have concluded that geographical enrichment of C. albicans clades is a real phenomenon (10, 23, 44), even to the extent of regional differences within the American continent (36, 46).
However, there seems also to have been sufficient mixing of isolates, possibly because of international migrations, to blur the geographical distinctions, and a microsatellite-based typing survey found genetic homogeneity of isolates in four different locations in the United States (24).
One unequivocal conclusion that we can draw is that C. albicans isolates from animals are distributed differently from human isolates in the MLST population structure. The differences in the present study, which are self-evident even without the high statistical support, confirm a similarly large difference in distribution of strains from birds compared with human isolates by MLST done with a slightly different gene set (4).
The species C. africana was erected to delineate C. albicans isolates with atypical carbohydrate assimilation patterns (49). All the isolates with the characteristics of C. africana were cultured from male or female genitalia, and most came from patients in Africa. The C. africana type strain MYA-2669 (from the penis of a German patient) clustered indistinguishably with two United Kingdom vaginal isolates by MLST, so this type continues to be associated exclusively with isolates of genital origin. Although the three isolates were highly distinct from the remainder of the panel we typed by MLST in the UPGMA dendrogram (Fig. 2), they were less distinct in an unrooted neighbor-joining tree (Fig. 3). The fact that all seven of the gene fragments that we used for MLST easily gave PCR products with MYA-2669 contrasts with the situation that we encountered when we attempted to type C. parapsilosis isolates by MLST and ultimately proposed two new species to account for isolates that gave PCR products with only a minority of the gene set tested (47). In common with others (17), we think the case for separate species status for C. africana is not confirmed by the results of the present study.
Our data concerning the antifungal susceptibility of our panel of isolates confirms and extends existing information. Like Pujol et al. (37), we found that a majority of flucytosine-resistant isolates belonged to clade 1. Since almost all clade 1 isolates are also ABC type A, this conclusion corroborates the work of McCullough et al. (26), who also found a significantly higher proportion of flucytosine-resistant isolates among type A isolates. All clade 1 isolates with flucytosine resistance encoded arginine at position 101 in the FUR1 gene product, as previously reported by others (14, 18). This mutation will presumably retard or prevent the conversion of flucytosine to fungicidal 5-fluorouracil within the fungal cells. Flucytosine-susceptible isolates in clade 1 did not encode arginine at position 101, nor did resistant and susceptible isolates from other clades. This finding of an apparent clade-specific molecular resistance mechanism to an antifungal agent opens the possibility for development of a molecular diagnostic test to predict flucytosine resistance and raises the expectation that further research may elucidate clade-consistent molecular bases for resistance to other agents. Although we did not uncover any self-evident associations between C. albicans clades and resistance to azole antifungal agents, the numbers of resistant isolates available to us was relatively small, and we have not investigated the molecular basis of resistance in these isolates. It should be noted that all of the isolates tested against fluconazole, itraconazole, and flucytosine in this study were also tested for caspofungin susceptibility, but none showed resistance to this agent, so this result was not detailed.
It is notable that most of the azole-resistant isolates of C. albicans in our panel were oral isolates from HIV-positive patients. A recent, very large, global surveillance study shows a very low prevalence of fluconazole resistance among C. albicans isolates since the introduction of highly active antiretroviral therapy (HAART) in 1996 (34). In the absence of a substantial future failure of efficacy of HAART, it is likely to prove difficult to obtain substantial numbers of azole-resistant clinical isolates of C. albicans for further investigation. The strong association that we saw between homozygosity at the C. albicans mating type locus and fluconazole resistance was described previously (40). Our data extend this association to itraconazole and even to flucytosine, which exerts its antifungal effects by a mechanism different from that of the azoles. Previous experimentation showed that engineering homozygosity at the C. albicans MTL does not lead per se to the development of fluconazole resistance (35). The association that we describe between MTL homozygosity and resistance to several antifungals therefore suggests that a common event, perhaps a change in expression of a transcription factor (11), is responsible both for resistance development and MTL homozygosity in wild-type isolates.
An important consideration to emerge from the determination of C. albicans population structure concerns the choice of strains used for experimental work with the fungus. To date, almost all molecular research with C. albicans has been done with the single clade 1 isolate SC5314 and its Ura- derivatives, in the absence of positive selectable markers for molecular genetic work. Now that at least one positive selectable marker has been developed for C. albicans (52), our delineation of the four major clades of the species, and a preliminary demonstration of potential clade-related differences in the propensity to cause infection, form a basis for rational selection of a wider diversity of strains for future research in the field.
ACKNOWLEDGMENTS
This study was supported by a grant from the Wellcome Trust.
We gratefully acknowledge the many colleagues who have supplied us over many years with the isolates we submitted to MLST. We are particularly grateful to E. M. Johnson, C. C. Kibbler, N. Nolard, F. Symoens (at the IHEM collection), J. B. Anderson, D. R. Soll, and J. Schmid for supplying us with well-pedigreed isolates for this MLST project and to Marie-Elisabeth Bougnoux and Christophe d'Enfert for DNA from their panel of isolates.
Supplemental material for this article can be found at http://jcm.asm.org/.
REFERENCES
Anderson, J. M., and D. R. Soll. 1987. The unique phenotype of opaque cells in the "white-opaque transition" in Candida albicans. J. Bacteriol. 169:5579-5588.
Blignaut, E., C. Pujol, S. Lockhart, S. Joly, and D. R. Soll. 2002. Ca3 fingerprinting of Candida albicans isolates from human immunodeficiency virus-positive and healthy individuals reveals a new clade in South Africa. J. Clin. Microbiol. 40:826-836.
Blignaut, E., C. Pujol, S. Lockhart, S. Joly, and D. R. Soll. 2002. A new clade of Candida albicans among South African oral yeast isolates. J. Dent. Res. 81:2231.
Bougnoux, M.-E., A. D. M., S. Morand, M. Theraud, B. G. Spratt, and C. d'Enfert. 2004. Multilocus sequence typing of Candida albicans: strategies, data exchange and applications. Infect. Genet. Evol. 4:243-252.
Bougnoux, M.-E., S. Morand, and C. d'Enfert. 2002. Usefulness of multilocus sequence typing for characterization of clinical isolates of Candida albicans. J. Clin. Microbiol. 40:1290-1297.
Bougnoux, M. E., A. Tavanti, C. Bouchier, N. A. R. Gow, A. Magnier, A. D. Davidson, M. C. J. Maiden, C. d'Enfert, and F. C. Odds. 2003. Collaborative consensus for optimized multilocus sequence typing of Candida albicans. J. Clin. Microbiol. 41:5265-5266.
Burnie, J. P., F. C. Odds, W. Lee, C. Webster, and J. D. Williams. 1985. Outbreak of systemic Candida albicans in intensive care unit caused by cross infection. Brit. Med. J. 290:746-748.
Calderone, R. A. 2002. Candida and candidiasis. ASM Press, Washington, D.C.
Chan, M. S., M. C. J. Maiden, and B. G. Spratt. 2001. Database-driven multi locus sequence typing (MLST) of bacterial pathogens. Bioinformatics 17:1077-1083.
Clemons, K. V., F. Feroze, K. Holmberg, and D. A. Stevens. 1997. Comparative analysis of genetic variability among Candida albicans isolates from different geographic locales by three genotypic methods. J. Clin. Microbiol. 35:1332-1336.
Coste, A. T., M. Karababa, F. Ischer, J. Bille, and D. Sanglard. 2004. TAC1, transcriptional activator of CDR genes, is a new transcription factor involved in the regulation of Candida albicans ABC transporters CDR1 and CDR2. Eukaryot. Cell 3:1639-1652.
Cowen, L. E., A. Nantel, M. S. Whiteway, D. Y. Thomas, D. C. Tessier, L. M. Kohn, and J. B. Anderson. 2002. Population genomics of drug resistance in Candida albicans. Proc. Natl. Acad. Sci. USA 99:9284-9289.
Cowen, L. E., D. Sanglard, D. Calabrese, C. Sirjusingh, J. B. Anderson, and L. M. Kohn. 2000. Evolution of drug resistance in experimental populations of Candida albicans. J. Bacteriol. 182:1515-1522.
Dodgson, A. R., K. J. Dodgson, C. Pujol, M. A. Pfaller, and D. R. Soll. 2004. Clade-specific flucytosine resistance is due to a single nucleotide change in the FUR1 gene of Candida albicans. Antimicrob. Agents Chemother. 48:2223-2227.
Feil, E. J., B. C. Li, D. M. Aanensen, W. P. Hanage, and B. G. Spratt. 2004. eBURST: inferring patterns of evolutionary descent among clusters of related bacterial genotypes from multilocus sequence typing data. J. Bacteriol. 186:1518-1530.
Fonzi, W., and M. Irwin. 1993. Isogenic strain construction and gene mapping in Candida albicans. Genetics 134:717-728.
Forche, A., G. Schnian, Y. Grser, R. Vilgalys, and T. G. Mitchell. 1999. Genetic structure of typical and atypical populations of Candida albicans from Africa. Fungal Genet. Biol. 28:107-125.
Hope, W. W., L. Tabernero, D. W. Denning, and M. J. Anderson. 2004. Molecular mechanisms of primary resistance to flucytosine in Candida albicans. Antimicrob. Agents Chemother. 78:4377-4386.
Horsburgh, C. R., and C. H. Kirkpatrick. 1983. Long-term therapy of chronic mucocutaneous candidiasis with ketoconazole: experience with twenty-one patients. Am. J. Med. 74:23-29.
Hunter, P. R. 1990. Reproducibility and indices of discriminatory power of microbial typing methods. J. Clin. Microbiol. 28:1903-1905.
Jones, T., N. A. Federspiel, H. Chibana, J. Dungan, S. Kalman, B. B. Magee, G. Newport, Y. R. Thorstenson, N. Agabian, P. T. Magee, R. W. Davis, and S. Scherer. 2004. The diploid genome sequence of Candida albicans. Proc. Natl. Acad. Sci. USA 101:7329-7334.
Kumar, S., K. Tamura, I. B. Jakobsen, and M. Nei. 2001. MEGA2: Molecular Evolutionary Genetics Analysis software. Bioinformatics 17:1244-1245.
Lott, T. J., and M. M. Effat. 2001. Evidence for a more recently evolved clade within a Candida albicans North American population. Microbiology 147:1687-1692.
Lott, T. J., R. E. Fundyga, M. E. Brandt, L. H. Harrison, A. N. Sofair, R. A. Hajjeh, and D. W. Warnock. 2003. Stability of allelic frequencies and distributions of Candida albicans microsatellite loci from US population-based surveillance isolates. J. Clin. Microbiol. 41:1316-1321.
Mackenzie, D. W. R., and F. C. Odds. 1991. Non-identity and authentication of two major reference strains of Candida albicans. J. Med. Vet. Mycol. 29:255-261.
McCullough, M. J., K. V. Clemons, and D. A. Stevens. 1999. Molecular and phenotypic characterization of genotypic Candida albicans subgroups and comparison with Candida dubliniensis and Candida stellatoidea. J. Clin. Microbiol. 37:417-421.
NCCLS. 2002. Reference method for broth dilution antifungal susceptibility testing of yeasts; approved standard, 2nd ed., vol. 17(9). NCCLS, Wayne, Pa.
Odds, F. C. 1988. Candida and candidosis, 2nd ed. Bailliere Tindall, London, United Kingdom.
Odds, F. C. 1991. Long-term preservation of pathogenic yeasts under distilled water. J. Med. Vet. Mycol. 29:413-415.
Odds, F. C., A. B. Abbott, R. L. Stiller, H. J. Scholer, A. Polak, and D. A. Stevens. 1983. Analysis of Candida albicans phenotypes from different geographical and anatomical sources. J. Clin. Microbiol. 18:849-857.
Odds, F. C., C. C. Kibbler, E. Walker, A. Bhamra, H. G. Prentice, and P. Noone. 1989. Carriage of Candida species and C. albicans biotypes in patients undergoing chemotherapy or bone marrow transplantation for haematological disease. J. Clin. Pathol. 42:1259-1266.
Odds, F. C., L. Van Nuffel, and G. Dams. 1998. Prevalence of Candida dubliniensis isolates in a yeast stock collection. J. Clin. Microbiol. 36:2869-2873.
Odds, F. C., L. Vranckx, and F. Woestenborghs. 1995. Antifungal susceptibility testing of yeasts: evaluation of technical variables for test automation. Antimicrob. Agents Chemother. 39:2051-2060.
Pfaller, M. A., and D. J. Diekema. 2004. Twelve years of fluconazole in clinical practice: global trends in species distribution and fluconazole susceptibility of bloodstream isolates of Candida. Clin. Microbiol. Infect. 10(Suppl. 1):11-23.
Pujol, C., S. A. Messer, M. Pfaller, and D. R. Soll. 2003. Drug resistance is not directly affected by mating type locus zygosity in Candida albicans. Antimicrob Agents Chemother. 47:1207-1212.
Pujol, C., M. Pfaller, and D. R. Soll. 2002. Ca3 fingerprinting of Candida albicans bloodstream isolates from the United States, Canada, South America, and Europe reveals a European clade. J. Clin. Microbiol. 40:2729-2740.
Pujol, C., M. A. Pfaller, and D. R. Soll. 2004. Flucytosine resistance is restricted to a single genetic clade of Candida albicans. Antimicrob. Agents Chemother. 48:262-266.
Redding, S., J. Smith, G. Farinacci, M. Rinaldi, A. Fothergill, J. Rhinechalberg, and M. Pfaller. 1994. Resistance of Candida albicans to fluconazole during treatment of oropharyngeal candidiasis in a patient with AIDS—documentation by in vitro susceptibility testing and DNA subtype analysis. Clin. Infect. Dis. 18:240-242.
Rex, J. H., C. R. Cooper, W. G. Merz, J. N. Galgiani, and E. J. Anaissie. 1995. Detection of amphotericin B-resistant Candida isolates in a broth-based system. Antimicrob. Agents Chemother. 39:906-909.
Rustad, T. R., D. A. Stevens, M. A. Pfaller, and T. C. White. 2002. Homozygosity at the Candida albicans MTL locus associated with azole resistance. Microbiology 148:1061-1072.
Sanglard, D., K. Kuchler, F. Ischer, J. L. Pagani, M. Monod, and J. Bille. 1995. Mechanisms of resistance to azole antifungal agents in Candida albicans isolates from AIDS patients involve specific multidrug transporters. Antimicrob. Agents Chemother. 39:2378-2386.
Schmid, J., S. Herd, P. R. Hunter, R. D. Cannon, M. S. M. Yasin, S. Samad, M. Carr, D. Parr, W. McKinney, M. Schousboe, B. Harris, R. Ikram, M. Harris, A. Restrepo, G. Hoyos, and K. P. Singh. 1999. Evidence for a general-purpose genotype in Candida albicans, highly prevalent in multiple geographical regions, patient types and types of infection. Microbiology 145:2405-2413.
Schmid, J., P. R. Hunter, G. C. White, A. K. Nand, and R. D. Cannon. 1995. Physiological traits associated with success of Candida albicans strains as commensal colonizers and pathogens. J. Clin. Microbiol. 33:2920-2926.
Schonian, G., A. Forche, H. J. Tietz, M. Muller, Y. Graser, R. Vilgalys, T. G. Mitchell, and W. Presber. 2000. Genetic structure of geographically different populations of Candida albicans. Mycoses 43:51-56.
Soll, D. R. 2004. Mating-type locus homozygosis, phenotypic switching and mating: a unique sequence of dependencies in Candida albicans. Bioessays 26:10-20.
Soll, D. R., and C. Pujol. 2003. Candida albicans clades. FEMS Immunol. Med. Microbiol. 39:1-7.
Tavanti, A., A. D. Davidson, N. A. R. Gow, M. C. J. Maiden, and F. C. Odds. 2005. Candida orthopsilosis and Candida metapsilosis spp. nov. to replace Candida parapsilosis groups II and III. J. Clin. Microbiol. 43:284-292.
Tavanti, A., N. A. R. Gow, S. Senesi, M. C. J. Maiden, and F. C. Odds. 2003. optimization and validation of multilocus sequence typing for Candida albicans. J. Clin. Microbiol. 41:3765-3776.
Tietz, H. J., M. Hopp, A. Schmalreck, W. Sterry, and V. Czaika. 2001. Candida africana sp. nov., a new human pathogen or a variant of Candida albicans Mycoses 44:437-445.
Urwin, R., and M. C. J. Maiden. 2003. Multi-locus sequence typing: a tool for global epidemiology. Trends Microbiol. 11:479-487.
White, T. C. 1997. Increased mRNA levels of ERG16, CDR, and MDR1 correlate with increases in azole resistance in Candida albicans isolates from a patient infected with human immunodeficiency virus. Antimicrob. Agents Chemother. 41:1482-1487.
Wirsching, S., S. Michel, and J. Morschhauser. 2000. Targeted gene disruption in Candida albicans wild-type strains: the role of the MDR1 gene in fluconazole resistance of clinical Candida albicans isolates. Mol. Microbiol. 36:856-865.(Arianna Tavanti, Amanda D)
The Peter Medawar Building for Pathogen Research and Department of Zoology, University of Oxford, OX1 3SY Oxford, United Kingdom
ABSTRACT
We submitted a panel of 416 isolates of Candida albicans from separate sources to multilocus sequence typing (MLST). The data generated determined a population structure in which four major clades of closely related isolates were delineated, together with eight minor clades comprising five or more isolates. By Fisher's exact test, a statistically significant association was found between particular clades and the anatomical source, geographical source, ABC genotype, decade of isolation, and homozygosity versus heterozygosity at the mating type-like locus (MTL) of the isolates in the clade. However, these associations may have been influenced by confounding variables, since in a univariate analysis of variance, only the clade associations with ABC type and anatomical source emerged as statistically significant, providing the first indication of possible differences between C. albicans strain type clades and their propensity to infect or colonize different anatomical locations. There were no significant differences between clades with respect to distributions of isolates resistant to fluconazole, itraconazole, or flucytosine. However, the majority of flucytosine-resistant isolates belonged to clade 1, and these isolates, but not flucytosine-resistant isolates in other clades, bore a unique mutation in the FUR1 gene that probably accounts for their resistance. A significantly higher proportion of isolates resistant to fluconazole, itraconazole, and flucytosine were homozygous at the MTL, suggesting that antifungal pressure may trigger a common mechanism that leads both to resistance and to MTL homozygosity. The utility of MLST for determining clade assignments of clinical isolates will form the basis for strain selection for future research into C. albicans virulence.
INTRODUCTION
Differentiation of microbial isolates by sequencing a small sample of unrelated housekeeping genes has become established as a reliable and effective method for typing strains of many bacteria (9, 50). Such multilocus sequence typing (MLST) is highly reproducible, and data can be archived in Web-based databases accessible to all users. For the fungal pathogen Candida albicans, an MLST system based on seven DNA fragments was developed as an optimal consensus for typing strains within the species (6) following two earlier proposed systems (5, 48). Because C. albicans is a diploid organism, sequence data contain heterozygous as well as homozygous sites, adding an extra discriminatory feature to MLST for this species.
Among other approaches to strain typing of C. albicans, DNA fingerprinting based on the moderately repetitive sequence Ca3 has been widely used. By this method, C. albicans populations were shown to comprise five major clades of closely related strain types (46), including clades enriched in isolates from Europe (36) and South Africa (2, 3). Resistance of C. albicans to flucytosine in vitro was found to be a property restricted almost entirely to isolates from a single C. albicans clade as determined by Ca3 fingerprinting (37), and the sole mechanism of resistance to flucytosine within this clade was the mutation of a single nucleotide in the FUR1 gene (14).
Typing by a combination of restriction fragment polymorphisms and randomly isolated polymorphic DNA also showed geographical enrichment of groups of related isolates (10), and "ABC typing," based on the size and number of internally transcribed spacer regions (ITS1) in DNA encoding rRNA showed a higher proportion of flucytosine-resistant isolates among type A strains (26).
We have now typed more than 400 isolates of C. albicans by MLST. Our data show how MLST reveals a population structure with four major clades of related strain types plus at least eight minor clades. It confirms the predominant association of flucytosine resistance with one major clade and reveals associations between clades and their geographical, anatomic, and temporal origins.
MATERIALS AND METHODS
C. albicans isolates. We typed 416 isolates from a large range of sources. Each isolate was from a separate source. Table 1 summarizes the details of the panel of isolates tested. Data for isolates described by Bougnoux et al. in their MLST work (5) were included in the present study; for these isolates, only extracted DNA had been provided by Marie-Elisabeth Bougnoux and Christophe d'Enfert for MLST (6), so we could not conduct antifungal susceptibility tests with these isolates. A set of 25 isolates was provided by David R. Soll (Table 1); these comprised five isolates from each of the five major C. albicans clades as determined by Ca3 fingerprinting (46) and served as a reference for cross-validation of typing methodologies. Isolate WO1, the original "white-opaque switcher" now undergoing full-genome sequencing, was also originally supplied by D. R. Soll.
A set of 36 isolates was provided by Jan Schmid; 35 of these were a subset of isolates that had been typed by Ca3 fingerprinting (42, 43). Another set of 22 clinical isolates was provided by James B. Anderson from his group's study of population genomics of fluconazole resistance in C. albicans (12, 13). Other notable isolates included SC5314, which was used for whole-genome sequencing of C. albicans (21) and is the wild-type isolate used widely for gene disruption research (16); a representative from the set of 16 serial isolates with increasing fluconazole resistance from a patient infected with human immunodeficiency virus (HIV) (38, 51), provided by Theodore White; isolate NCPF3153, often referred to as 3153A and used in many laboratories; and isolate MYA-2669, the type strain of the putative species Candida africana (49), purchased from the American Type Culture Collection. A large set of isolates was provided from the collection of the Belgian Institute for Hygiene, Epidemiology, and Medicine (IHEM) through the collaboration of Nicole Nolard and Franoise Symoens; some of these were subcultures originally obtained from the senior author's collection. Three isolates that had been used to study mechanisms of fluconazole resistance (41) were originally contributed by Dominique Sanglard.
Recent clinical isolates of C. albicans were provided by Elizabeth M. Johnson (Mycology Reference Laboratory, Bristol, United Kingdom) and Christopher C. Kibbler (Royal Free Hospital, London, United Kingdom) from their diagnostic mycology laboratories. Fresh oral commensal isolates of C. albicans were obtained from healthy undergraduate student volunteers in Aberdeen, United Kingdom. The remainder of the isolates typed came from the senior author's collection of C. albicans isolates, which was started in the 1970s. This collection, originally stored under distilled water (29), was transferred twice in its entirety, once for an investigation of the prevalence of the species C. dubliniensis in the collection (32). Some of these isolates had been previously described in epidemiological studies of C. albicans based on phenotypic typing tests (30, 31). The quality of the collection was confirmed by the near-identity of MLST data for several isolates from the same source (details not shown).
Isolates were maintained on Sabouraud agar (Oxoid, Basingstoke, United Kingdom) for the duration of this study.
MLST. Sequences of bases in the gene fragments AAT1a, ACC1, ADP1, MPIb, SYA1, VPS13, and ZWF1b were determined as previously described (6, 48). Isolates were designated as genotype A, B, or C on the basis of PCR for rRNA genes (26), and the status of the mating-type-like locus (MTL) in each strain was assessed as heterozygous (a/), homozygous (a/a), or homozygous (/) by PCR as previously described (48).
Sequencing of other genes. Genes FCY22, FCA1, and FUR1 encode, respectively, cytosine permease, cytosine deaminase, and uracil phosphoribosyltransferase in the pathway for conversion of flucytosine to the active antifungal 5-fluorouracil in C. albicans (14). For selected C. albicans isolates, the DNA sequences of these genes were determined by the same method as that used for MLST (above). For FCY22, the forward primers used were FCY22-1 (5'AACATAATCTGTAGTTATATCTC), FCY22-2 (ATTATTGGAAGCATTCCAGTG3'), FCY22-3 (5'TCCAACTGAAGCTGGTAATG3'), and FCY22-4 (TGCCATTCAATGTATATGGAG), and the reverse primer was 5'TAACAGGGCAGAATGAATCG3'. For FCA1, the forward primers were FCA1-1 (5'AATTCTCTTTCAATTCTTCTTG3') and FCA1-2 (5'TCACCATGTAGTATGTGTAC3'), with the reverse primer being 5'TTGAACAAGATGATGATGTTG3'. For FUR1, the forward primers were FUR1-1 (5'ATAATGGAGCATCTCGCAAC3') and FUR1-2 (5'TAGCTGAAGATATCAGTGAAC3'), with the reverse primer being 5'TGTTGTACATATAATCCTATAAG3'.
Susceptibility testing. MICs of fluconazole, flucytosine, and itraconazole were determined by microdilution tests by a previously described method and spectrophotometric endpoints (33) that give MIC results compatible with those obtained by the NCCLS reference method (27). Isolates with fluconazole MICs of 16 μg/ml or higher, flucytosine MICs of 8 μg/ml or higher, or itraconazole MICs of 0.25 μg/ml or higher were classed as resistant to these agents (these breakpoints indicate full susceptibility to the agents concerned [27]).
Data analysis. Sequence data were analyzed by scrutiny of capillary sequencer output chromatograms with DNASTAR Seqman software. Each unique (diploid) sequence for each of the seven DNA fragments was given a separate identifying number. Each unique combination of seven sequences was assigned a separate diploid sequence type (DST) number. The sequence and DST identifiers were compiled on a central database (http://test1.mlst.net/).
The eBURST package (http://eburst.mlst.net/ [15]) was used to determine putative relationships between isolates. With C. albicans data, this software scans allele sequences from pairs of isolates and records isolates as related when six of the seven sequences are identical between a pair. The eBURST algorithm places all related isolates into clonal complexes and, where possible, predicts the founding, or ancestral DST of each complex. The output is a display of the most parsimonious patterns of descent of each DST from the ancestral type.
Phylogenetic analyses by unweighted pair group method with arithmetic averages (UPGMA) and neighbor-joining algorithms were conducted with MEGA version 2.1 (22) applied to modified sequence data. The analyses were based only on the results for polymorphic bases to maximize their power to discriminate between isolates. The data set of only the variable bases was suitable for pairwise-difference analysis, which has been used previously with C. albicans MLST (5, 48). To obtain sequences that could be handled by the MEGA software, which is not programmed to analyze heterozygous code data, the following procedure was used. The results for the variable sites from the seven gene fragments sequenced were concatenated into a single sequence. For any pair of isolates, each with a diploid genome, the base at each polymorphic site could be homozygous and identical between the isolates, heterozygous and identical, homozygous and different, or heterozygous in one isolate and homozygous in another. For example, the sequencing result (a IUPAC single-letter code) for a given base across a set of strains might appear as A, T, or W (= A + T). Data from the polymorphic sites from the seven C. albicans alleles were therefore conjoined into a single sequence, and then each base in the sequence was rewritten twice for a homozygous (A, C, G, or T) datum or once each for the two component bases for a heterozygous (K, M, R, S, W, or Y) datum. These revised sequences could then be analyzed to generate dendrograms in MEGA 2.1. In this form, the analysis was the functional equivalent of scoring a pair of results as 1 for homozygous or heterozygous identical data, 0 for homozygous different data, and 0.5 when one polymorphic site had a heterozygous result and the other a homozygous result and then creating a difference matrix. Use of the modified sequences, rather than of difference matrices, enabled calculation of statistical significance for the cluster nodes by bootstrapping with 1,000 replications.
Distributions of properties between C. albicans clades were determined statistically by Fisher's exact test. Results were considered significant for a P of <0.05. A univariate analysis of variance (ANOVA) was used to determine associations between clades and A, B, or C type (with 1, 2, and 3 corresponding to A, B, and C, respectively); MTL status (with 1, 2, and 3 corresponding to a/, a/a, and /, respectively); fluconazole, itraconazole, and flucytosine susceptibilities (with 1 indicating susceptibility and 2 indicating resistance); decades of isolation (1 being a year prior to 1990, 2 being a year in the 1990s, and 3 being a year in the 2000s); geographical origin (1 being the United Kingdom, 2 being another European country, 3 being North America, and 4 being another country); and anatomical origin (1 being blood or other sterile site, 2 being the oropharynx, 3 being the vagina, and 4 being another location). Isolates from animals were excluded from this analysis.
Discriminatory power was calculated according to Hunter (20).
RESULTS
Reproducibility and sequencing errors. A total of 13 different C. albicans isolates was submitted for MLST in duplicate on different occasions, and the person conducting the sequencing was unaware of their identities. The seven genes sequenced yielded a total of 109 variable sites. Of the 13 isolates tested in duplicate, three yielded different DSTs. In two cases, the difference resulted from a single-base homozygous/heterozygous difference in one of the seven genes; in the third there were homozygous/heterozygous discrepancies at two sites in a single gene. Thus, in a total of 1,417 variable sites tested in duplicate, a sequencing error arose on four (0.28%) occasions, indicating a reproducibility of 99.72%. Sequence differences greater than three homozygous/heterozygous discrepancies between isolates were therefore assumed to indicate true interstrain differences.
Statistics of C. albicans MLST. The 416 isolates yielded 351 distinct DSTs, with a discriminatory power index of 0.9996. Among the seven fragments sequenced, ZWF1 gave the highest discriminatory ratio, yielding 84 different sequences from just 10 polymorphic sites, followed by SYA1 (70 alleles from 14 variable sites), VPS13 (103 alleles from 21 variable sites), AAT1a (57 alleles from 13 variable sites), ACC1 (37 alleles from 10 variable sites), ADP1 (46 alleles from 18 variable sites), and MPIb (36 alleles from 21 variable sites).
Details of the alleles and DSTs for each isolate are shown in the full version of Table S1 in the supplemental material, together with the geographical and anatomical sources of the isolates. Insufficient DNA for sample BougnCP10 precluded determination of ABC and MTL type for this isolate. Among the remaining 415 isolates, 281 (67.7%) were type A, 96 (23.1%) were type B, and 38 (9.2%) were type C. The great majority of isolates (371, or 89.4%) were heterozygous at the MTL, while 27 isolates (6.5%) were of homozygous MTL type / and 17 isolates (4.1%) were of type a/a.
Population structure of 416 unrelated C. albicans isolates. MLST data allow for different approaches to phylogenetic analysis of various degrees of sophistication. We utilized several approaches, with an emphasis on methods based on pair differences at polymorphic nucleotide sites and on gross allele sequence differences in an effort to determine robust subspecific clades that will allow easy assignment of future isolates to the population structure.
eBURST analysis of the set of 416 C. albicans isolates from unique sources revealed 18 clonal clusters, with 5 being based on six or more DSTs (Fig. 1). Cluster 1, the largest, comprised 105 isolates (25.2% of all isolates), with DST 69 as the putative ancestral type. Clonal cluster 2, the second largest, comprised 38 isolates and 27 DSTs and no putative founding DST (type 155 was represented by 7 isolates but did not emerge as a likely ancestor for the cluster). Clonal cluster 3 comprised 28 isolates and 17 DSTs, with type 124 being the predicted founding type. Cluster 4 comprised eight isolates and seven DSTs, with type 305 being the likely ancestral DST, and cluster 5 comprised seven isolates and six DSTs, with no putative ancestral type. The remaining clonal clusters consisted maximally of six isolates.
Of the 416 isolates analyzed by the eBURST software, 228 (54.8%) were assigned to clonal clusters, thus leaving almost half of the isolates with singleton status (Fig. 1).
Unlike the eBURST analysis, a UPGMA pairwise-difference dendrogram for the 416 separate-source isolates (Fig. 2) took no account of the gene fragment in which sequence polymorphisms arose and therefore provided a different perspective on relationships between isolates: isolates with small numbers of differences in the concatenated sequences for variable sites across all fragments sequenced were regarded as similar by UPGMA analysis, even if the polymorphic sites were located on separate fragments. Comparison of the clustering in the UPGMA dendrogram with the eBURST clonal clusters (Fig. 2) showed clonal clusters 11 and 17 and several eBURST singletons dispersed within the broad range of isolates in clonal cluster 1. Similarly, isolates in clonal clusters 5, 10, and 13 were mixed with each other and with several eBURST singletons by UPGMA analysis.
Although the UPGMA dendrogram for the concatenated variable sequences from the 416 separate-source isolates clearly suggested patterns of clustering between the isolates, bootstrap values for cluster nodes were generally low except for a few pairs of very closely related isolates. The following procedure was therefore used to determine which isolates were most likely to belong to clusters of highly related strains. The information from Ca3 typing, particularly the results for isolates belonging to known Ca3 clades (46), was considered in conjunction with clonal cluster information to determine a cutoff point that could be used to define UPGMA clades and outlying singletons.
All five of the isolates from clade I as determined by Ca3 typing (46) clustered within a large group of isolates closely related by UPGMA (Fig. 2); four of the five belonged to clonal cluster 1 by eBURST analysis. Similarly, 19 of the 21 isolates designated types A1 and A2 by Ca3 fingerprinting and designated by neighbor-joining statistics (42, 43) appeared in the same large UPGMA cluster that was broadly delimited by eBURST clonal cluster 1. All 155 isolates from AM2003/0046 to AM2004/0049 in Fig. 2 were therefore provisionally assigned to clade 1 as demarcated by MLST. The five Ca3 clade II isolates and both Ca3 group B isolates correlated with a well-delineated UPGMA cluster (Fig. 2); four of these seven isolates appeared in eBURST clonal cluster 2. Isolates from AM2004/0030 to IHEM17110 in Fig. 2 were therefore provisionally assigned to MLST clade 2. Similarly, the five isolates from Ca3 clade SA clustered together in the UPGMA dendrogram, with four of the five in clonal cluster 3. However, the UPGMA cluster designated MLST clade 4 in Fig. 2 (isolates HUN68 to IHEM16731), though appearing reasonably homogeneous, contained a subcluster comprising two of the Ca3 clade E isolates, and all of the members of clonal cluster 15. The five isolates from Ca3 clade III emerged in clonal clusters 9 or 10 or as singletons by eBURST analysis, but they clustered together in the UPGMA dendrogram in a group of isolates with the same overall level of similarity as those in MLST clades 1, 2, and 4. Isolates AM2003/0082 to b30998/5 were therefore provisionally considered to belong to MLST clade 3. It was noted that isolates previously designated as group C by neighbor-joining analysis of Ca3 typing (42, 43) did not show obvious affinities with eBURST- or UPGMA-defined clusters.
The provisional assignment of isolates to clades 1 to 4 indicated a maximum dissimilarity of 10 by pairwise-nucleotide difference analysis as calculated with the MEGA software. The following definition of a clade was therefore used: a cluster of at least five isolates with a dissimilarity not exceeding 10 by pairwise-difference analysis. By this definition, 348 (83.7%) of the 416 isolates could be assigned to 12 clades, as specified in Fig. 2. Reanalysis of the MLST variable-site data only for the 348 isolates that matched the clade definition confirmed the validity of the clade assignments in Fig. 2. The bootstrap values for the nodes defining clades were 85% (clade 1), 88% (clade 2), 87% (clade 3), 86% (clade 4), and 98 or 99% for the minor clades. The exception was clade 5, for which the node bootstrap value was only 64%, while the values for the node defining the pair of isolates 81/078 and AM2003/0084 and the node for the remaining four isolates nominated for clade 5 were 97% and 99%, respectively. These results suggest that clade 5 is a less robust grouping of isolates than the other clades and may need to be revised in future as further C. albicans isolates are added to the MLST database.
To examine the population structure of C. albicans clades by an alternative statistical approach the MLST data for the 416 isolates of separate origins were reanalyzed by determination of p-distance with the unrooted neighbor-joining algorithm in the MEGA2 software package. The result (Fig. 3) confirmed most of the isolate clusters determined by pairwise-distance UPGMA analysis (Fig. 2) and reiterated that the majority of the isolates studied could be assigned to one of four major clades. However, the neighbor-joining algorithm also produced some differences in the structure from the UPGMA analysis.
In Fig. 3, the three isolates that included the type strain of the putative species C. africana, while still forming a well-separated branch in the tree, nevertheless appeared less distant from other isolates than in the UPGMA dendrogram (Fig. 2). In the neighbor-joining tree, isolates 81/078 and AM2003/0084 clustered separately from the remainder of isolates classed as clade 5 in Fig. 2 (the position of the pair is indicated as clade 5a in Fig. 3), reemphasizing the previous conclusion that clade 5 is less robust than most others. The population structure by neighbor joining also removed isolates C82, AM2003/0065, BougnCP11, 73/025, and J981315 as a group from the rest of clade 3 in Fig. 2 to a new position shown as clade 3a in Fig. 3. The clade assignments in Fig. 2 were used for subsequent analyses of clade properties; however, it is self-evident that the most robust definition of C. albicans clades must await a larger future database of MLST results and possibly further approaches to statistical analyses.
Isolates and clusters of notable provenance. Three isolates, AM2003/0025, AM2003/0032, and MYA-2669, the type strain of the putative species C. africana, were indistinguishable by MLST but clustered with very low similarity to the other isolates in Fig. 2, confirming that MYA-2669 differs substantially from other C. albicans isolates. All three of these indistinguishable isolates came from genital swabs.
Isolate SC5314, which was used for the first whole-genome sequencing project with C. albicans (21), fell within clade 1. Isolate WO-1, which has been extensively used in research into phenotypic switching and mating in C. albicans (45), appeared as a singleton between clades 8 and 4 in Fig. 2.
Within clade 1, a group of four isolates—IHEM20416, J942149, IHEM20414, and IHEM20417—that were extremely similar by MLST was notable because all four were resistant to fluconazole, and the first three were also resistant to itraconazole. These three were / at the MTL (Table 1). All four of the isolates in this subcluster came from the mouths of AIDS patients in Germany. A similar subcluster of four closely related isolates within cluster 5, already noted above by the high significance of the bootstrap value for the cluster node, included three isolates resistant to both azole antifungal agents and three isolates that were a/a at the MTL. All four isolates came from the mouths of AIDS patients in France.
C. albicans clades: properties and clinical relevance. Table 2 summarizes the properties of the 12 individual C. albicans clades and of singleton isolates. The numbers of isolates in each of clades 5 to 12 was too small to permit reliable statistical analysis; these isolates were therefore grouped together as "minor clades" for the calculations detailed below.
Clades 1 and 2 comprised almost entirely isolates of ITS1 type A, while clade 3 isolates were almost entirely type B. Clades 5 and 6 isolates were all type B, and clade 7 isolates were all type A. Only clade 4 had a preponderance of type C isolates, but this clade also contained type A and type B isolates. The distributions of A, B, and C types between clades obviously differed significantly (Fisher's exact test, P < 0.0001).
There were 44 (10.6%) of 415 typeable isolates with homozygous alleles at the MTL. Just over half of these isolates (29) were in clade 1. The distribution of isolates homozygous at the MTL differed significantly between clades (by Fisher's exact test, P was 0.016; the test was applied to numbers of isolates that were a/, a/a, or / [Table 1]). Minor clades 8 to 11 contained no examples of MTL homozygous isolates, while a high proportion of isolates in clades 5 to 7 were homozygous at the MTL (Table 2).
Except for clade 9, the majority of isolates in each clade came from sources within Europe (information on geographical origin was available for 415 of the 416 isolates in Fig. 2). The 6 isolates in clade 5 were all of European origin, and there were no North American isolates among the 28 that made up clades 10 to 12. Table 3 breaks down the geographical-origin data more finely, but with minor clades treated as a single group. Statistical analysis of the distribution of isolate sources by clade as shown in Table 3 indicated a highly significant association of clade with geographical source (Fisher's exact test, P < 0.001). Even when the data were analyzed with French and other EU isolates combined into a single group and African isolates included with "other" geographical sources, the differential distribution of sources within each clade remained statistically significant (P < 0.001). It should particularly be noted that, among the 45 isolates in clade 4, equivalent to the "South African" clade based on Ca3 fingerprinting (2), only 3 isolates came from the African continent. The majority of African isolates in the present study belonged to clade 2.
Information on the anatomical source of isolates was available for 406 of the 416 isolates in Fig. 2. There were 11 isolates from animals, so analysis of relations between anatomical sources and clades were based on 395 isolates (Table 4). Clades 5, 6, and 9 contained no isolates from blood cultures (Table 2), and the proportion of isolates from blood and other sterile sites in clade 1 was lower than in the other major clades. The distribution of isolates from blood, the oropharynx, the vagina, and other sites varied between clades at a level that reached statistical significance (Fisher's exact test, P = 0.045).
The clade distribution of the 11 isolates from animals differed considerably from that of human isolates. Five of the isolates were singletons, four were in clade 10, and one each belonged to clades 1 and 9. This differential distribution was highly significant (Fisher's exact test, P = 0.003). Since 10 of the 11 animal isolates originated in Belgium, their clade distribution was compared separately with that of the 11 human isolates that came from Belgian patients only (5 in clade 1, 2 in clade 11, and 1 each in clades 2, 3, 4, 6, 9, and 12). Even with clades 5 to 12 treated as a single group, the distribution of animal isolates differed significantly from that of Belgian human isolates (P = 0.02).
For most of the clades, between 10% and 20% of the isolates were first cultured before 1990 (Table 2). In clades 1 to 3, 6, 7, and 9, isolates first cultured before 1990 were represented and most of the isolates were first cultured before the year 2000, while clades 4, 5, 8 and 10 to 12 bore evidence of possibly more recent origins (Table 2). The distribution of isolates of different ages differed significantly between the clades (by Fisher's exact test, P was 0.003; analysis done with clades 5 to 12 treated isolates as a single group, and dates of isolation segregated as pre-1990, 1990s, and 2000s).
A high proportion of the five isolates in clade 5 was resistant to azole antifungal agents, while clades 7 to 11 contained no isolates resistant to any of the three agents tested (Table 2). However, the overall differences in clade distributions of the 378 isolates with susceptibility data available were not significant for those that were resistant to fluconazole (Fisher's exact test, P = 0.289), itraconazole (P = 0.94), or flucytosine (P = 0.105). Other significant associations were, however, apparent for the antifungal resistant isolates. Fifteen (65%) of the 23 flucytosine-resistant isolates were members of clade 1. In every one of these isolates, flucytosine resistance was associated with a homozygous allele pair encoding cysteine at position 101 in the FUR1 gene. In a random selection of 14 flucytosine-susceptible isolates from clade 1, the equivalent position was either homozygous for arginine or heterozygous (cysteine/arginine). The eight flucytosine-resistant isolates from other clades and seven randomly selected susceptible isolates from other clades were all homozygous, encoding arginine at position 101, except for one flucytosine-resistant isolate for which no FUR1 sequence was obtained despite three attempts. No common mutation was found in the FUR1, FCY22, or FCA1 gene sequences of these non-clade-1, flucytosine-resistant isolates that might account for their resistance to the drug. Among the 35 fluconazole-resistant isolates, 17 (48.6%) were homozygous at the MTL. The same was true for 10 (41.7%) of the 24 itraconazole-resistant isolates and for 7 (21.2%) of the 23 flucytosine-resistant isolates. The prevalence of MTL homozygosity among the corresponding susceptible isolates was 7.3, 9.0, and 9.9%, respectively (by Fisher's exact test, P was <0.001, <0.001, and equal to 0.004, respectively). There was no significant association between itraconazole or flucytosine resistance and ABC type, but an unusually high proportion (40%) of 35 isolates resistant to fluconazole were ABC type B (Fisher's exact test, P = 0.013).
Twenty-eight of the 35 fluconazole-resistant isolates came from the oropharynx; all of these were from AIDS patients. For 109 oropharyngeal isolates, the HIV status of the patient source was known. The distribution of oral isolates from HIV-positive versus HIV-negative patients differed significantly between clades, as follows. For 48 HIV-negative patients, 26.8% were in clade 1, 17.9% in clade 2, 0.0% in clade 3, 17.9% in clade 4, and 23.2% in minor clades, and 14.3% were singletons. For 61 HIV-positive patients, the corresponding data were 50.7%, 9.0%, 9.0%, 6.0%, 16.4%, and 9.0% (Fisher's exact test, P < 0.001).
The individual data comparisons made by Fisher's exact test can be vulnerable to biases arising from uneven distributions of contributing variables other than the variable under analysis. A univariate ANOVA was done to determine the associations between properties and sources of isolates and their clade distribution. Animal isolates were excluded. The ANOVA was repeated with individual clades and singletons as the dependent variable and also with the minor clades included as a single variable. With individual clades analyzed, ABC type emerged as the most significant variable (P < 0.001) and anatomical source as the next most significant variable (P = 0.027). With minor clades grouped, the corresponding P values were <0.001 for ABC type and 0.065 for anatomical source. MTL status, fluconazole, itraconazole, and flucytosine resistance, decade of isolation, and geographical source were not significant clade-related variables by ANOVA.
DISCUSSION
This study has generated a considerable amount of data on the population structure of isolates of C. albicans and the relationship between clades of isolates and their properties of clinical relevance. We have provided evidence that different clades of C. albicans may differ significantly in the proportions of isolates from blood, the oropharynx, the vagina, and other sites. While, on the one hand, it is beyond dispute that the immune status of the host is the major factor determining the ability of a C. albicans to invade human tissues (8, 28), our data provide the first-ever clear suggestion that particular strain type clusters within the species may differ in their proportions of isolates from disseminated or mucosal infections. It is especially notable that, while the proportions of several properties of the isolates studied varied significantly between clades in single-factor tests, only ABC type and anatomical origin emerged as significant factors relating to clade number by univariate ANOVA. This result lends particular emphasis to the likely importance of an association between clades and putative virulence properties within the species of C. albicans.
Assignation of population structures and clade boundaries to a set of isolates cannot be done entirely objectively. The clade assignments in Fig. 2 take account of three types of information: the UPGMA dendrogram, results of eBURST analysis, and the results of Ca3 oligonucleotide fingerprinting for several of the isolates. Use of neighbor-joining statistics in place of pairwise-difference analyses further redistributed some isolates between clades (Fig. 3). We intend to evaluate a wider range of statistical approaches to clade designation, including Bayesian statistics, when our database of MLST results is at least twice as large as that reported in this study. For the time being, we consider pairwise-difference analyses to be the simplest and most useful approach for clade assignments in the population of isolates at hand.
The eBURST results were helpful in demarcating potential clades, but not definitive: many examples occurred where isolates that coclustered by Ca3 typing or by MLST analyzed by pair differences were designated as singletons by the eBURST test. It may be that the nature of sequence variation in C. albicans and the diploid genome of the species make eBURST analyses less relevant than they are to haploid bacteria (50). In common with others (4), we found an excellent correlation between UPGMA clusters for MLST and Ca3 fingerprinting data. Clades I, II, III, and SA as currently delimited by Ca3 typing (46) matched clades 1 to 4, respectively, as defined by MLST in the present study. Only MLST clade E failed to emerge as a single group of isolates as determined by MLST; the 5 representatives of Ca3 clade E split into two clusters by UPGMA analysis of MLST data (Fig. 2).
The very clear demarcation of geographical origin of strains determined by Ca3 typing as described previously (46) was not confirmed in the present study. Our panel of isolates was dominated by isolates of European origin, particularly by isolates from the United Kingdom, while the panel of isolates studied by Ca3 typing (36, 46) was dominated by isolates from South Africa. In common with the Ca3 studies, we found significant interclade differences in proportions of various geographical origins, but the relationships between clade and origin were not the same, almost certainly because of the large differences in proportions of isolates from different sources. Most notably, MLST clade 4, which corresponded to Ca3 clade SA, contained only three isolates of African origin, while most of the African isolates in our study clustered in clade 2. We have therefore chosen to refer to clades only by numbers until the MLST database contains a sufficiently large representation of isolates from all parts of the globe to allow more objective determination of possible geographic evolutionary origins of clades. The statistical significance of geographical variations between MLST clades determined by Fisher's exact test was not confirmed in the ANOVA, perhaps showing the effect of biases from differences in anatomical origin, date of isolation, etc. A larger MLST database will allow for statistical analyses of geographical origin based on, for example, only blood isolates or only oral isolates. It is to be expected that individual C. albicans clades may be enriched with isolates from a particular geographical source, and several previous studies by other DNA typing methods have concluded that geographical enrichment of C. albicans clades is a real phenomenon (10, 23, 44), even to the extent of regional differences within the American continent (36, 46).
However, there seems also to have been sufficient mixing of isolates, possibly because of international migrations, to blur the geographical distinctions, and a microsatellite-based typing survey found genetic homogeneity of isolates in four different locations in the United States (24).
One unequivocal conclusion that we can draw is that C. albicans isolates from animals are distributed differently from human isolates in the MLST population structure. The differences in the present study, which are self-evident even without the high statistical support, confirm a similarly large difference in distribution of strains from birds compared with human isolates by MLST done with a slightly different gene set (4).
The species C. africana was erected to delineate C. albicans isolates with atypical carbohydrate assimilation patterns (49). All the isolates with the characteristics of C. africana were cultured from male or female genitalia, and most came from patients in Africa. The C. africana type strain MYA-2669 (from the penis of a German patient) clustered indistinguishably with two United Kingdom vaginal isolates by MLST, so this type continues to be associated exclusively with isolates of genital origin. Although the three isolates were highly distinct from the remainder of the panel we typed by MLST in the UPGMA dendrogram (Fig. 2), they were less distinct in an unrooted neighbor-joining tree (Fig. 3). The fact that all seven of the gene fragments that we used for MLST easily gave PCR products with MYA-2669 contrasts with the situation that we encountered when we attempted to type C. parapsilosis isolates by MLST and ultimately proposed two new species to account for isolates that gave PCR products with only a minority of the gene set tested (47). In common with others (17), we think the case for separate species status for C. africana is not confirmed by the results of the present study.
Our data concerning the antifungal susceptibility of our panel of isolates confirms and extends existing information. Like Pujol et al. (37), we found that a majority of flucytosine-resistant isolates belonged to clade 1. Since almost all clade 1 isolates are also ABC type A, this conclusion corroborates the work of McCullough et al. (26), who also found a significantly higher proportion of flucytosine-resistant isolates among type A isolates. All clade 1 isolates with flucytosine resistance encoded arginine at position 101 in the FUR1 gene product, as previously reported by others (14, 18). This mutation will presumably retard or prevent the conversion of flucytosine to fungicidal 5-fluorouracil within the fungal cells. Flucytosine-susceptible isolates in clade 1 did not encode arginine at position 101, nor did resistant and susceptible isolates from other clades. This finding of an apparent clade-specific molecular resistance mechanism to an antifungal agent opens the possibility for development of a molecular diagnostic test to predict flucytosine resistance and raises the expectation that further research may elucidate clade-consistent molecular bases for resistance to other agents. Although we did not uncover any self-evident associations between C. albicans clades and resistance to azole antifungal agents, the numbers of resistant isolates available to us was relatively small, and we have not investigated the molecular basis of resistance in these isolates. It should be noted that all of the isolates tested against fluconazole, itraconazole, and flucytosine in this study were also tested for caspofungin susceptibility, but none showed resistance to this agent, so this result was not detailed.
It is notable that most of the azole-resistant isolates of C. albicans in our panel were oral isolates from HIV-positive patients. A recent, very large, global surveillance study shows a very low prevalence of fluconazole resistance among C. albicans isolates since the introduction of highly active antiretroviral therapy (HAART) in 1996 (34). In the absence of a substantial future failure of efficacy of HAART, it is likely to prove difficult to obtain substantial numbers of azole-resistant clinical isolates of C. albicans for further investigation. The strong association that we saw between homozygosity at the C. albicans mating type locus and fluconazole resistance was described previously (40). Our data extend this association to itraconazole and even to flucytosine, which exerts its antifungal effects by a mechanism different from that of the azoles. Previous experimentation showed that engineering homozygosity at the C. albicans MTL does not lead per se to the development of fluconazole resistance (35). The association that we describe between MTL homozygosity and resistance to several antifungals therefore suggests that a common event, perhaps a change in expression of a transcription factor (11), is responsible both for resistance development and MTL homozygosity in wild-type isolates.
An important consideration to emerge from the determination of C. albicans population structure concerns the choice of strains used for experimental work with the fungus. To date, almost all molecular research with C. albicans has been done with the single clade 1 isolate SC5314 and its Ura- derivatives, in the absence of positive selectable markers for molecular genetic work. Now that at least one positive selectable marker has been developed for C. albicans (52), our delineation of the four major clades of the species, and a preliminary demonstration of potential clade-related differences in the propensity to cause infection, form a basis for rational selection of a wider diversity of strains for future research in the field.
ACKNOWLEDGMENTS
This study was supported by a grant from the Wellcome Trust.
We gratefully acknowledge the many colleagues who have supplied us over many years with the isolates we submitted to MLST. We are particularly grateful to E. M. Johnson, C. C. Kibbler, N. Nolard, F. Symoens (at the IHEM collection), J. B. Anderson, D. R. Soll, and J. Schmid for supplying us with well-pedigreed isolates for this MLST project and to Marie-Elisabeth Bougnoux and Christophe d'Enfert for DNA from their panel of isolates.
Supplemental material for this article can be found at http://jcm.asm.org/.
REFERENCES
Anderson, J. M., and D. R. Soll. 1987. The unique phenotype of opaque cells in the "white-opaque transition" in Candida albicans. J. Bacteriol. 169:5579-5588.
Blignaut, E., C. Pujol, S. Lockhart, S. Joly, and D. R. Soll. 2002. Ca3 fingerprinting of Candida albicans isolates from human immunodeficiency virus-positive and healthy individuals reveals a new clade in South Africa. J. Clin. Microbiol. 40:826-836.
Blignaut, E., C. Pujol, S. Lockhart, S. Joly, and D. R. Soll. 2002. A new clade of Candida albicans among South African oral yeast isolates. J. Dent. Res. 81:2231.
Bougnoux, M.-E., A. D. M., S. Morand, M. Theraud, B. G. Spratt, and C. d'Enfert. 2004. Multilocus sequence typing of Candida albicans: strategies, data exchange and applications. Infect. Genet. Evol. 4:243-252.
Bougnoux, M.-E., S. Morand, and C. d'Enfert. 2002. Usefulness of multilocus sequence typing for characterization of clinical isolates of Candida albicans. J. Clin. Microbiol. 40:1290-1297.
Bougnoux, M. E., A. Tavanti, C. Bouchier, N. A. R. Gow, A. Magnier, A. D. Davidson, M. C. J. Maiden, C. d'Enfert, and F. C. Odds. 2003. Collaborative consensus for optimized multilocus sequence typing of Candida albicans. J. Clin. Microbiol. 41:5265-5266.
Burnie, J. P., F. C. Odds, W. Lee, C. Webster, and J. D. Williams. 1985. Outbreak of systemic Candida albicans in intensive care unit caused by cross infection. Brit. Med. J. 290:746-748.
Calderone, R. A. 2002. Candida and candidiasis. ASM Press, Washington, D.C.
Chan, M. S., M. C. J. Maiden, and B. G. Spratt. 2001. Database-driven multi locus sequence typing (MLST) of bacterial pathogens. Bioinformatics 17:1077-1083.
Clemons, K. V., F. Feroze, K. Holmberg, and D. A. Stevens. 1997. Comparative analysis of genetic variability among Candida albicans isolates from different geographic locales by three genotypic methods. J. Clin. Microbiol. 35:1332-1336.
Coste, A. T., M. Karababa, F. Ischer, J. Bille, and D. Sanglard. 2004. TAC1, transcriptional activator of CDR genes, is a new transcription factor involved in the regulation of Candida albicans ABC transporters CDR1 and CDR2. Eukaryot. Cell 3:1639-1652.
Cowen, L. E., A. Nantel, M. S. Whiteway, D. Y. Thomas, D. C. Tessier, L. M. Kohn, and J. B. Anderson. 2002. Population genomics of drug resistance in Candida albicans. Proc. Natl. Acad. Sci. USA 99:9284-9289.
Cowen, L. E., D. Sanglard, D. Calabrese, C. Sirjusingh, J. B. Anderson, and L. M. Kohn. 2000. Evolution of drug resistance in experimental populations of Candida albicans. J. Bacteriol. 182:1515-1522.
Dodgson, A. R., K. J. Dodgson, C. Pujol, M. A. Pfaller, and D. R. Soll. 2004. Clade-specific flucytosine resistance is due to a single nucleotide change in the FUR1 gene of Candida albicans. Antimicrob. Agents Chemother. 48:2223-2227.
Feil, E. J., B. C. Li, D. M. Aanensen, W. P. Hanage, and B. G. Spratt. 2004. eBURST: inferring patterns of evolutionary descent among clusters of related bacterial genotypes from multilocus sequence typing data. J. Bacteriol. 186:1518-1530.
Fonzi, W., and M. Irwin. 1993. Isogenic strain construction and gene mapping in Candida albicans. Genetics 134:717-728.
Forche, A., G. Schnian, Y. Grser, R. Vilgalys, and T. G. Mitchell. 1999. Genetic structure of typical and atypical populations of Candida albicans from Africa. Fungal Genet. Biol. 28:107-125.
Hope, W. W., L. Tabernero, D. W. Denning, and M. J. Anderson. 2004. Molecular mechanisms of primary resistance to flucytosine in Candida albicans. Antimicrob. Agents Chemother. 78:4377-4386.
Horsburgh, C. R., and C. H. Kirkpatrick. 1983. Long-term therapy of chronic mucocutaneous candidiasis with ketoconazole: experience with twenty-one patients. Am. J. Med. 74:23-29.
Hunter, P. R. 1990. Reproducibility and indices of discriminatory power of microbial typing methods. J. Clin. Microbiol. 28:1903-1905.
Jones, T., N. A. Federspiel, H. Chibana, J. Dungan, S. Kalman, B. B. Magee, G. Newport, Y. R. Thorstenson, N. Agabian, P. T. Magee, R. W. Davis, and S. Scherer. 2004. The diploid genome sequence of Candida albicans. Proc. Natl. Acad. Sci. USA 101:7329-7334.
Kumar, S., K. Tamura, I. B. Jakobsen, and M. Nei. 2001. MEGA2: Molecular Evolutionary Genetics Analysis software. Bioinformatics 17:1244-1245.
Lott, T. J., and M. M. Effat. 2001. Evidence for a more recently evolved clade within a Candida albicans North American population. Microbiology 147:1687-1692.
Lott, T. J., R. E. Fundyga, M. E. Brandt, L. H. Harrison, A. N. Sofair, R. A. Hajjeh, and D. W. Warnock. 2003. Stability of allelic frequencies and distributions of Candida albicans microsatellite loci from US population-based surveillance isolates. J. Clin. Microbiol. 41:1316-1321.
Mackenzie, D. W. R., and F. C. Odds. 1991. Non-identity and authentication of two major reference strains of Candida albicans. J. Med. Vet. Mycol. 29:255-261.
McCullough, M. J., K. V. Clemons, and D. A. Stevens. 1999. Molecular and phenotypic characterization of genotypic Candida albicans subgroups and comparison with Candida dubliniensis and Candida stellatoidea. J. Clin. Microbiol. 37:417-421.
NCCLS. 2002. Reference method for broth dilution antifungal susceptibility testing of yeasts; approved standard, 2nd ed., vol. 17(9). NCCLS, Wayne, Pa.
Odds, F. C. 1988. Candida and candidosis, 2nd ed. Bailliere Tindall, London, United Kingdom.
Odds, F. C. 1991. Long-term preservation of pathogenic yeasts under distilled water. J. Med. Vet. Mycol. 29:413-415.
Odds, F. C., A. B. Abbott, R. L. Stiller, H. J. Scholer, A. Polak, and D. A. Stevens. 1983. Analysis of Candida albicans phenotypes from different geographical and anatomical sources. J. Clin. Microbiol. 18:849-857.
Odds, F. C., C. C. Kibbler, E. Walker, A. Bhamra, H. G. Prentice, and P. Noone. 1989. Carriage of Candida species and C. albicans biotypes in patients undergoing chemotherapy or bone marrow transplantation for haematological disease. J. Clin. Pathol. 42:1259-1266.
Odds, F. C., L. Van Nuffel, and G. Dams. 1998. Prevalence of Candida dubliniensis isolates in a yeast stock collection. J. Clin. Microbiol. 36:2869-2873.
Odds, F. C., L. Vranckx, and F. Woestenborghs. 1995. Antifungal susceptibility testing of yeasts: evaluation of technical variables for test automation. Antimicrob. Agents Chemother. 39:2051-2060.
Pfaller, M. A., and D. J. Diekema. 2004. Twelve years of fluconazole in clinical practice: global trends in species distribution and fluconazole susceptibility of bloodstream isolates of Candida. Clin. Microbiol. Infect. 10(Suppl. 1):11-23.
Pujol, C., S. A. Messer, M. Pfaller, and D. R. Soll. 2003. Drug resistance is not directly affected by mating type locus zygosity in Candida albicans. Antimicrob Agents Chemother. 47:1207-1212.
Pujol, C., M. Pfaller, and D. R. Soll. 2002. Ca3 fingerprinting of Candida albicans bloodstream isolates from the United States, Canada, South America, and Europe reveals a European clade. J. Clin. Microbiol. 40:2729-2740.
Pujol, C., M. A. Pfaller, and D. R. Soll. 2004. Flucytosine resistance is restricted to a single genetic clade of Candida albicans. Antimicrob. Agents Chemother. 48:262-266.
Redding, S., J. Smith, G. Farinacci, M. Rinaldi, A. Fothergill, J. Rhinechalberg, and M. Pfaller. 1994. Resistance of Candida albicans to fluconazole during treatment of oropharyngeal candidiasis in a patient with AIDS—documentation by in vitro susceptibility testing and DNA subtype analysis. Clin. Infect. Dis. 18:240-242.
Rex, J. H., C. R. Cooper, W. G. Merz, J. N. Galgiani, and E. J. Anaissie. 1995. Detection of amphotericin B-resistant Candida isolates in a broth-based system. Antimicrob. Agents Chemother. 39:906-909.
Rustad, T. R., D. A. Stevens, M. A. Pfaller, and T. C. White. 2002. Homozygosity at the Candida albicans MTL locus associated with azole resistance. Microbiology 148:1061-1072.
Sanglard, D., K. Kuchler, F. Ischer, J. L. Pagani, M. Monod, and J. Bille. 1995. Mechanisms of resistance to azole antifungal agents in Candida albicans isolates from AIDS patients involve specific multidrug transporters. Antimicrob. Agents Chemother. 39:2378-2386.
Schmid, J., S. Herd, P. R. Hunter, R. D. Cannon, M. S. M. Yasin, S. Samad, M. Carr, D. Parr, W. McKinney, M. Schousboe, B. Harris, R. Ikram, M. Harris, A. Restrepo, G. Hoyos, and K. P. Singh. 1999. Evidence for a general-purpose genotype in Candida albicans, highly prevalent in multiple geographical regions, patient types and types of infection. Microbiology 145:2405-2413.
Schmid, J., P. R. Hunter, G. C. White, A. K. Nand, and R. D. Cannon. 1995. Physiological traits associated with success of Candida albicans strains as commensal colonizers and pathogens. J. Clin. Microbiol. 33:2920-2926.
Schonian, G., A. Forche, H. J. Tietz, M. Muller, Y. Graser, R. Vilgalys, T. G. Mitchell, and W. Presber. 2000. Genetic structure of geographically different populations of Candida albicans. Mycoses 43:51-56.
Soll, D. R. 2004. Mating-type locus homozygosis, phenotypic switching and mating: a unique sequence of dependencies in Candida albicans. Bioessays 26:10-20.
Soll, D. R., and C. Pujol. 2003. Candida albicans clades. FEMS Immunol. Med. Microbiol. 39:1-7.
Tavanti, A., A. D. Davidson, N. A. R. Gow, M. C. J. Maiden, and F. C. Odds. 2005. Candida orthopsilosis and Candida metapsilosis spp. nov. to replace Candida parapsilosis groups II and III. J. Clin. Microbiol. 43:284-292.
Tavanti, A., N. A. R. Gow, S. Senesi, M. C. J. Maiden, and F. C. Odds. 2003. optimization and validation of multilocus sequence typing for Candida albicans. J. Clin. Microbiol. 41:3765-3776.
Tietz, H. J., M. Hopp, A. Schmalreck, W. Sterry, and V. Czaika. 2001. Candida africana sp. nov., a new human pathogen or a variant of Candida albicans Mycoses 44:437-445.
Urwin, R., and M. C. J. Maiden. 2003. Multi-locus sequence typing: a tool for global epidemiology. Trends Microbiol. 11:479-487.
White, T. C. 1997. Increased mRNA levels of ERG16, CDR, and MDR1 correlate with increases in azole resistance in Candida albicans isolates from a patient infected with human immunodeficiency virus. Antimicrob. Agents Chemother. 41:1482-1487.
Wirsching, S., S. Michel, and J. Morschhauser. 2000. Targeted gene disruption in Candida albicans wild-type strains: the role of the MDR1 gene in fluconazole resistance of clinical Candida albicans isolates. Mol. Microbiol. 36:856-865.(Arianna Tavanti, Amanda D)