Mating experiments.

To test for sexual compatibility, isolates were grown on MYP agar plates at 25°C for 3 days. Pairs of cell suspensions were then inoculated and mixed together on plates containing CMA, YCB, or MEA medium at 25°C and examined with a phase-contrast microscope (Zeiss, Germany) for the presence of filaments and sexual structures, up to 4 weeks of incubation. CMA medium contained 1.5% (wt/vol) corn meal agar (Difco) and 0.5% (wt/vol) agar. YCB medium contained 1.17% (wt/vol) yeast carbon base (Difco) and 2% agar. MEA medium contained 2.5% (wt/vol) malt extract (Difco) and 2% (wt/vol) agar.

Fluorescence microscopy and staining.

Cells and mycelial structures from positive matings were stained with Calcofluor white (0.05% solution; BD) and Sytox green (5 mM; Invitrogen) to detect the cell wall and nuclei, respectively, using the protocols described by Metin et al. (9).

FACS analysis.

The fluorescence-activated cell-sorting (FACS) protocol was modified from Tanaka et al. (15). Cells were grown overnight at 25°C in yeast extract-peptone-dextrose (YPD) broth, collected by centrifugation, and washed with 1× phosphate-buffered saline (PBS). Cells were then fixed in 1 ml of 70% ethanol overnight at 4°C with mild agitation. Upon centrifugation, cells were resuspended and washed with 1 ml of NS buffer (10 mM Tris-HCl [pH 7.2], 0.25 M sucrose, 1 mM EDTA, 1 mM MgCl₂, 0.1 mM CaCl₂, 0.55 mM phenylmethylsulfonyl fluoride, 0.1 mM ZnCl₂, 0.049% 2-mercaptoethanol). Cells were then resuspended in 180 μl NS buffer with 14 ml RNase A (15 mg/μl, Qiagen) and 6 ml of propidium iodide (1.0 μg/μl; CALBIOCHEM) and incubated in the dark for 2 to 4 h at room temperature. After incubation, 50 μl of the cells were mixed with 500 ml of Tris-PI mix (482 μl 1 M Tris [pH 7.5] and 18 μl propidium iodide, 1 μg/μl). Flow cytometry was performed on 10,000 cells with a slow laser scan, on the FL1 channel with a Becton, Dickinson FACScan. Selected haploid and diploid C. neoformans reference strains were used as controls for ploidy determination.

Physiological characterization.

Physiological tests were performed on sterile 96-well microplates (NunclonΔ Surface, Denmark) according to Kurtzman et al. (16) and read with a StatFax 2100 microplate reader (Awareness Technology Inc., United States) using absorbance measures at 630 nm. Media were prepared according to a protocol available at the CBS website (http://www.cbs.knaw.nl/collections/DefaultInfo.aspx?Page=YeastMethods). The inoculum was grown overnight on liquid MYP medium at 25°C, in a Certomat U orbital shaker (Sartorius, Germany) at 48 rpm, and diluted in sterile distilled water for microplate inoculation. Each well contained 200 μl of medium and 5 μl of diluted inoculum (initial absorbance of 0.04). After inoculation, the microplates were sealed with a sealing pellicle (NunclonΔ Surface, Denmark) and were incubated at 25°C in a Denley Wellwarm 1 (Denley) microplate incubator with shaking, up to 3 weeks. Each strain was tested in independent duplicates. The ability to grow at 30°C, 35°C, and 37°C was determined by inoculating MYP slants with each culture and incubating in water baths at the appropriate temperature for up to 4 days.

DNA extraction.

Genomic DNA for MAT gene amplification and sequencing and for MLST was obtained using a simplified phenol-chloroform extraction method following cell disruption using glass beads. Genomic DNA for fosmid library preparation was obtained using the cetyltrimethylammonium bromide (CTAB) extraction method described by Metin et al. (9).

Degenerate PCR.

Identification of potential MAT genes in K. mangrovensis relied on the design of degenerate primers for moderately conserved MAT genes in C. neoformans and relatives. Sequences of homologs of several Cryptococcus MAT genes, CAP1, LPD1, RPL22, RPO41, STE3, STE12, and STE20 (pheromone cluster) and SXI1 and SXI2 (HD cluster), from C. neoformans, C. gattii, C. amylolentus, K. (C.) heveanensis, and T. wingfieldii were aligned with BioEdit 7.0.9.0 software (17), using the built-in ClustalW version 1.4 (18) with default parameters. The newly designed primers were tested for PCR amplification and sequencing of putative MAT gene fragments with K. mangrovensis opposite mating type strains CBS 8507 (=K01) and CBS 10435 (=K02). The newly obtained sequences were checked for mating-type-specific polymorphisms, which are expected only for genes within MAT. Those sequences were also used to design new, less, or nondegenerate primers, specific for K. mangrovensis, namely, for STE20 and SX1/SXI2. PCR products with the expected size were gel extracted or directly purified using the QIAquick gel extraction kit (Qiagen) or Illustra GFX PCR DNA and the gel band purification kit (GE Healthcare), according to the manufacturer's instructions. In some cases, an intermediate cloning step was necessary prior to sequencing and used the TOPO TA cloning kit (Invitrogen), as described by Metin et al. (9). Primers and PCR amplification conditions used in this study are presented as supplemental material (see Text S1 in the supplemental material). Primers were synthesized by Integrated DNA Technologies (United States) or STABVida (Portugal).

Fosmid library preparation and probing.

We employed the Copy Control Fosmid Library production kit (Epicentre, Madison, WI) to generate fosmid libraries from K. mangrovensis strains CBS 8507 (K01) and CBS 10435 (K02), using the procedure described by Findley et al. (8). Approximately 1,500 fosmid clones were picked into 96-well plates and transferred to 384-well plates for long-term storage at −80°C. The 384-well plates were replicated onto high-density filters for hybridizations using MAT gene probes prepared from the amplicons obtained using degenerate PCR, as described by Metin et al. (9). Positive fosmids were confirmed using colony PCR with specific primers for K. mangrovensis.

MLST analysis.

Multilocus sequence typing (MLST) of selected strains of K. mangrovensis was based on the following nuclear loci: D1/D2 domains of the LSU rRNA gene (LSU), ITS1-5.8S rRNA gene-ITS2 (ITS), a fragment of the gene encoding the largest subunit of RNA polymerase II (RPB1), a fragment of the gene encoding the second-largest subunit of RNA polymerase II (RPB2), a fragment of the gene encoding translation elongation factor 1 alpha (TEF1), and a fragment of the gene encoding a licensing factor required for DNA replication initiation and cell proliferation (MCM7). The first five loci have been previously used as phylogenetic markers for fungi (e.g., by the Assembling the Fungal Tree of Life [AFTOL] consortium [19]), and MCM7 was proposed more recently for the same purpose by Schmitt et al. (20). The RPB1 and TEF1 gene fragments contained two and one intron, respectively, whereas RPB2 and MCM7 did not include introns. Specific or degenerate primers were ordered or designed as needed and used to amplify and sequence the six loci. MLST primer information and PCR conditions are available in Text S1 in the supplemental material.

DNA sequencing.

Purified PCR amplification products were sequenced by STABVida (Portugal). Primers used for PCR amplification were also used for sequencing, except for ITS and LSU, in which case internal primers ITS1 (21) and NL4 (22), respectively, were used, and for RPB2 only the reverse primer was used for sequencing, yielding a <700-bp sequence. Fosmid DNA was extracted with JETstar, the novel plasmid purification system (Genomed), according to the manufacturer's instructions. Fosmid DNA was sequenced and assembled by STABVida (Portugal), using next-generation sequencing technology.

Bioinformatic and phylogenetic analyses.

Annotation of the different genomic loci was done manually by comparing the obtained sequences with the annotated sequences of C. neoformans var. neoformans JEC21 and/or C. heveanensis CBS 569. The comparison was done by pairwise alignment (BioEdit 7.0.9.0) and using BLASTn, BLASTp, BLASTx, or tBLASTx (2.2.26 version) (23) searches. Multiple sequence alignments were performed with the genomic sequences using either MUSCLE (24) or the ClustalW built-in version of MEGA version 5.05 (25) with the default parameters. Independent alignments and phylogenetic analyses were performed for each locus. For MLST, independent alignments were concatenated manually for each strain. Phylogenetic relationships were inferred by the maximum likelihood (ML) method based on the general time reversible (GTR) model (26) from 1,000 bootstrap replicates, using gamma distributed of 5 discrete gamma categories, all sites of the data set, all the codon positions and noncoding sites, and nearest-neighbor interchange (NNI) as the ML heuristic method in MEGA version 5.05. To detect recombination in the MLST data set, we used the Φ_w statistical test of Bruen et al. (27) implemented in the SplitsTree software (28). For phylogenetic analysis of MAT genes, the neighbor-joining method based on the Kimura 2-parameter model or the ML method as detailed above were employed, also using MEGA version 5.05. Sequences from GenBank were used when available. Accession numbers for sequences used for MLST and MAT analyses are listed in Table S1 in the supplemental material. Sequences from C. neoformans var. neoformans JEC20 (MATa) and JEC21 (MATα), C. neoformans var. grubii 125.91 (MATa) and H99 (MATα), C. gattii E566 (MATa) and WM276 (MATα), and C. heveanensis CBS 569 (A1B1) and BCC 8398 (A2B2) were all retrieved either from GenBank or DOE Joint Genome Institute (JGI), while all sequences from Tremella mesenterica ATCC 24925 were retrieved from JGI, and therefore no accession number is provided. Upon annotation of the fosmid insert sequence, synteny analysis was based on comparison of gene order and orientation in the homologous region of closely aligned species.

Phylogeny and species limits of K. mangrovensis.

A multilocus sequence typing (MLST) approach was applied to a set of 14 strains previously identified as K. mangrovensis on the basis of the sequences of the D1/D2 domains of the LSU rRNA gene (LSU). The set of strains comprised five strains included in the original study of Statzell-Tallman et al. (13), four seawater isolates from the Bahamas (K01, K02, K04, K06) and one cork isolate (CECT 11955 = K25), and nine additional strains: one clinical isolate from the United States (K09), four isolates from Mopane trees in Botswana, one seawater isolate from the British Virgin Islands (CBS 11279 = K27), one isolate from grape juice in Switzerland (CBS 7868 = K28), one isolate from grape berries in Slovenia (ZIM 605 = K29) obtained by Čadež et al. (29), and one isolate from a tree hollow in Brazil (K30) (see Table S1 in the supplemental material). The isolates from the Florida Everglades and the beach isolate reported by Statzell-Tallman et al. (13) were no longer viable (J. W. Fell, personal communication) and could thus not be studied. FACS analyses showed that strains K01, K02, K04, K06, K09, K10, K14, K17, K21, K25, K28, K29, and K30 are all haploid (data not shown). Preliminary results of PCR fingerprinting with all the strains previously identified as K. mangrovensis listed in Table S1 in the supplemental material (data not shown) demonstrated that some strains were indistinguishable based on fingerprints and therefore closely related (e.g., K04 and K05, K10 and K11, K25 and K26). We thus selected genetically distinct strains or those from different samples or geographic origins for the MLST analysis. We also included in the MLST analysis the taxa studied by Findley et al. (14) but added two recently described species in the Kwoniella clade: Cryptococcus pinus (30), including the type strain as well as a recent isolate from insect larvae in Belgium (MUCL 53261; see Table S1 in the supplemental material), and Cryptococcus shivajii (31). Tremella mesenterica served as an outgroup for the analysis. We performed a preliminary phylogenetic analysis of LSU and ITS with all sequences available in GenBank and found three other recently described species—Cryptococcus cuniculi, Cryptococcus tronadorensis, and Kwoniella shandongensis—to be loosely linked to the Cryptococcus and Kwoniella clades (see Fig. S1 in the supplemental material). However, since they formed a weakly supported lineage separate from the two main clades, we have not included them in the MLST analysis.

The following six loci were used for the MLST analysis: LSU, ITS, RPB1, RPB2, TEF1, and MCM7. We were able to amplify and sequence most of the loci for all taxa; accession numbers of the obtained sequences are given in Table S1 in the supplemental material. However, we were unable to amplify MCM7 from K. heveanensis, F. depauperata, or T. wingfieldii. Firstly, we performed phylogenetic analyses, using maximum likelihood (ML), of the individual loci (see Fig. S2 to S4 in the supplemental material). We assessed congruence of the different trees by checking possible conflicts in nodes with >70% bootstrap support as proposed by Hillis and Bull (32). We found only limited conflict in the nodes connecting Cryptococcus bestiolae, Cryptococcus dejecticola, and Bullera dendrophila in the TEF1 and MCM7 trees (see Fig. S4A and B in the supplemental material). Phylogenetic analyses of the concatenated sequences of five loci (Fig. 2) or of six loci (excluding the three above-mentioned species; see Fig. S5 in the supplemental material) generated trees in which most branches had strong bootstrap support and confirmed the two main lineages recognized by Findley et al. (14): the Cryptococcus clade, which comprises the pathogenic species complex, as well as C. amylolentus, T. wingfieldii, and F. depauperata; and the Kwoniella clade, which comprises K. mangrovensis, the remaining Cryptococcus species (including the two strains of C. pinus and C. shivajii), and Bullera dendrophila. Cryptococcus bestiolae was frequently the K. mangrovensis closest relative (except in the TEF1 tree; see Fig. S4A), and K. heveanensis appeared to be basal to the Kwoniella clade, but with moderate bootstrap support. Moreover, in phylogenetic trees of individual loci, namely, LSU, ITS, and TEF1 (see Fig. S2A and B and S4A), K. heveanensis was not or only weakly connected to the Kwoniella clade.

Open in a separate window

Fig 2

Phylogenetic relationships among K. mangrovensis and selected members of the Tremellales based on the analysis of concatenated sequences of five genomic loci (LSU, ITS, RPB1, RPB2, and TEF1). Analysis performed with 3,457 positions in the final data set. The tree was constructed using the maximum likelihood method implemented in the MEGA 5.05 software. Numbers on branches are bootstrap values (>50) from 1,000 replicates. T. mesenterica was used as the outgroup in the analysis. T, type strain. GenBank/EMBL accession numbers of the sequences used in the analysis are listed in Table S1 in the supplemental material; additional sequences were retrieved from JGI.

The phylogenetic analysis of the combined MLST loci (Fig. 2) further revealed significant genetic heterogeneity among K. mangrovensis strains that clustered into three strongly supported and closely related lineages: one includes the type strain (K01) as well as K02, K04, K06, and K27; another includes K10, K14, K17, and K22; and the third includes K25, K28, and K29. The three lineages are also apparent when the MCM7 locus was included in the analysis (see Fig. S5 in the supplemental material). Moreover, those three lineages are genetically distinct in each of the six loci, except LSU (see Fig. S2A in the supplemental material), and the strains of each lineage clustered consistently on each tree (see Fig. S2 to S4), thus suggesting an absence of genetic exchange between the three lineages. We have applied the Φ_w statistical test (27) to the 6-gene data set of the 12 K. mangrovensis strains mentioned above and found no significant evidence for recombination. However, inclusion of strains K09 and K30 in the same test resulted in a significant probability for recombination. The latter two strains occupied inconsistent or intermediate positions on the trees (Fig. 2; see also Fig. S2 to S4) and will be further discussed below. Nucleotide differences in the coding regions of the four protein-coding genes between strains of the three above-mentioned lineages were almost always synonymous, and in RPB1 and TEF1, most of the nucleotide differences were located in the introns (data not shown). The phylogenetic trees depicted in Fig. 2 and Fig. S2 to S5 suggest that K. mangrovensis should be confined to the strongly supported clade represented by the five strains isolated from seawater in tropical mangrove regions of the Atlantic, while the other two sister lineages, represented by the three European isolates and the four Botswana isolates, respectively, correspond to two additional sibling species, for which the names Kwoniella europaea and Kwoniella botswanensis are proposed (Fig. 2). This proposal is supported by the results of the mating experiments, because fertile crosses were observed only among the strains of K. mangrovensis but not between strains of K. mangrovensis and any of the two novel taxa. Crosses among strains of K. europaea or K. botswanensis were, however, infertile, and thus we are unable to confirm their conspecificity based only on mating assays. In spite of the apparent absence of sexual reproduction in either K. europaea or K. botswanensis, their phylogenetic placement in the Kwoniella clade (Fig. 2) argues for their inclusion in the genus Kwoniella in view of the recent implementation of the “one fungus, one name” rule in fungal nomenclature (33). Based on physiological test results, the three species are similar (see Table S2). However, K. mangrovensis can be distinguished from K. europaea and K. botswanensis by the inability of the former to grow at 35°C and to assimilate l-tartaric acid. K. europaea and K. botswanensis differ only in the assimilation of alpha-methyl-glucoside (fast in the latter and delayed in the former) and l-tartaric acid.

Evolutionary distances between the three sibling Kwoniella species calculated on the basis of the observed nucleotide differences in the six MLST loci are presented in Fig. 3 (distance bars). The distances between K. mangrovensis and K. europaea or K. botswanensis were much higher than the distance between C. amylolentus and T. wingfieldii and slightly higher than the distance between C. neoformans var. neoformans and C. neoformans var. grubii, which supports a separate taxonomic status for the former species. As for K. europaea and K. botswanensis, the calculated distance between them is lower than the distances of any of the two to K. mangrovensis and slightly higher than the distance between C. amylolentus and T. wingfieldii. In the absence of additional data, a varietal status could be assigned to the two sister taxa. However, the observed congruence of MLST gene genealogies and MAT gene data (see below) support separation of K. europaea and K. botswanensis at the species level. Formal descriptions of the newly proposed species are presented in the Taxonomy section below.

Open in a separate window

Fig 3

Jukes-Cantor distances based on concatenated sequences of five loci. Distances were computed among or between the following taxa: K. botswanensis (Kb), K. mangrovensis (Km), K. europaea (Ke), C. pinus (Cp), C. amylolentus (Ca), T. wingfieldii (Tw), C. neoformans var. neoformans (Cnn), C. neoformans var. grubii (Cng), and C. gattii (Cg). Values indicated above each bar represent the average distances; vertical range lines represent extreme values when different strains of the same taxon were compared.

The two remaining strains (K09 and K30) formed a separate branch that is positioned next to K. europaea and K. botswanensis in the combined MLST trees (Fig. 2; see also Fig. S5 in the supplemental material). Strain K30 had unique alleles for all MLST loci, except for RPB2, which had an identical sequence to that of K09; however, K09 shared the alleles of LSU, ITS, and MCM7 with K. europaea. FACS analyses demonstrated that both strains are haploid (data not shown). We have thus refrained from assigning a definite taxonomic status to either of these two anomalous strains until additional strains and respective genetic data are available to resolve if they are interspecies hybrids or novel taxa. As for the two strains of C. pinus studied and in spite of their different origins (see Table S1 in the supplemental material), they had only a few differences in the sequences of three of the six MLST loci and were thus considered conspecific (Fig. 2 and ​and3;3; see also Fig. S2 to S5 in the supplemental material).

Identification of MAT genes in K. mangrovensis and sibling species.

Our degenerate PCR-based strategy enabled us to amplify fragments of genes with homologs in the C. neoformans MAT locus from two compatible mating strains of K. mangrovensis, K01 and K02, for the following loci: LPD1, RPO41, STE20, and SXI. In the latter case, we actually amplified a fragment containing the 5′ regions of the SXI1 and SXI2 genes, as well as the intergenic spacer (this fragment will be henceforth designated SXI), because each primer was anchored in the conserved homeodomain region of each gene. The orientation of the single primer combination that produced the expected amplicon indicates that SXI1 and SXI2 are adjacent and divergently transcribed in K. mangrovensis, as found previously in the tremellaceous tetrapolar species C. amylolentus, K. heveanensis, and T. mesenterica (8, 9). In the case of STE20, we amplified two fragments of the gene: one containing the kinase domain, toward the 3′ end of the gene (~450 bp), and the other spanning the Cdc42/Rac interactive binding (CRIB) domain, upstream of the kinase domain and containing one intron (~1,000 bp). Possible linkage to MAT was verified by comparing the sequences of each gene fragment from the two compatible strains of K. mangrovensis, because only genes inside MAT are expected to exhibit marked mating-type-specific polymorphisms. Significantly different alleles were found for STE20 and SXI (e.g., 89% sequence similarity for the STE20 kinase 450-bp fragment; 74% similarity for the SXI2 673-bp fragment) but not for LPD1 or RPO41 (99% sequence similarity for the LPD1 388-bp fragment; 99.7% sequence similarity for the RPO41 909-bp fragment). In view of previous results on MAT structure of closely related species in the Cryptococcus and Kwoniella clades (12) and the observed mating-type-specific polymorphisms of STE20 and SXI, we predict that the former is located in the P/R region of MAT and the latter in the HD region. To confirm MAT specificity of those genes and to verify the mating system, we amplified and sequenced the STE20 (CRIB-containing region) and SXI gene fragments from the full set of strains previously identified as K. mangrovensis (see Table S1). We were also able to amplify and sequence the same loci for other species in the Kwoniella clade, B. dendrophila, C. bestiolae, C. dejecticola, C. pinus, and C. shivajii (for the latter two species and B. dendrophila, we could not amplify SXI). Accession numbers of the obtained sequences are given in Table S1 in the supplemental material. Phylogenetic analyses of the sequences of each locus are presented in Fig. 4 (STE20) and ​and55 (SXI). We were able to identify two STE20 alleles in K. mangrovensis (KmA1 in strains K01, K03, K06, K07, and K27; and KmA2 in strains K02, K04, K05, and K08), two additional alleles in K. botswanensis (KbA1 in strains K10, K11, and K22; and KbA2 in strains K14, K16, K17, K20, K23, and K24), and two more alleles in K. europaea (KeA1 in strains K25, K26, and K29; and KeA2 in strain K28) (see Table S1; Fig. 4). Strains K09 and K30 each had unique STE20 alleles. We also obtained an STE20 sequence from the mating type A2 strain of K. heveanensis and found that it is mating type specific (67% similarity to the sequence of the A1 allele), in contrast with the findings of Metin et al. (9), who predicted that STE20 was not located within MAT.

Open in a separate window

Fig 4

Phylogenetic tree based on STE20 coding sequences (CRIB-containing gene fragment) of K. mangrovensis and selected members of the Cryptococcus and Kwoniella clades. Analysis performed with 853 positions in the final data set. The trees were constructed using the maximum likelihood method implemented in the software MEGA 5.05. Numbers on branches are bootstrap values (>50) from 1,000 replicates. T. mesenterica was used as the outgroup in this analysis. T, type strain. GenBank/EMBL accession numbers of the sequences used in the analysis are listed in Table S1 in the supplemental material.

Open in a separate window

Fig 5

Phylogenetic tree based on SXI sequences of K. mangrovensis and selected members of the Cryptococcus and Kwoniella clades. Analysis performed with 1,280 positions in the final data set. Conserved blocks from the alignment were selected using Gblocks, and the trees were constructed using the maximum likelihood method implemented in the MEGA 5.05 software. Numbers on branches are bootstrap values (>50) from 1,000 replicates. T. mesenterica was used as the outgroup in this analysis. T, type strain. GenBank/EMBL accession numbers of the sequences used in the analysis are listed in Table S1 in the supplemental material.

Observed nucleotide differences in STE20 sequences between mating types of the same species were much lower in the three sibling Kwoniella species (ranging from 34 to 69) than in K. heveanensis (327) or the pathogenic Cryptococcus species (403 to 412). Nonetheless, nucleotide differences located in the coding sequence in the Kwoniella species were not always synonymous and led to 5 to 7 amino acid substitutions in the translated sequences between mating types (data not shown). The fact that mating type specificity of STE20 is less pronounced in the three Kwoniella species is also apparent in the phylogenetic tree depicted in Fig. 4. In fact, the topology of the Kwoniella clade is more similar to the species phylogeny as obtained from the MLST analysis (Fig. 2), whereas among the species of the Cryptococcus clade, STE20 follows a mating-type-specific phylogeny. Interestingly, whereas both STE20 alleles of K. mangrovensis formed a single, well-supported clade, those of K. botswanensis and K. europaea seemed to follow a mating-type-specific phylogeny, with the KbA1 allele clustering with the KeA1 allele and vice versa (Fig. 4).

For the SXI region, we identified four alleles in K. mangrovensis (KmB1 in strains K01, K03, K04, K05, and K08; KmB2 in strain K02; KmB3 in strains K06 and K07; and KmB4 in strain K27), a single allele in K. botswanensis (KbB1), and three different alleles in K. europaea (KeB1 in strains K25 and K26; KeB2 in strain K28; and KeB3 in strain K29) (Fig. 5; see also Table S1 in the supplemental material). The finding of distinct alleles of the two putative MAT genes (STE20 and SXI) in K. mangrovensis, K. botswanensis, and K. europaea provides further key support for their separation at the species level. For the two anomalous strains, K30 had a unique SXI allele, and K09 surprisingly shared one of the SXI alleles (KeB1) of K. europaea (Fig. 5), which hints to the possibility of this strain being a hybrid. The topology of the SXI tree (Fig. 5) reflects the high level of polymorphism of the 5′ regions of the SXI genes as observed in previous studies of related species (8, 9) that classified these genes as mating type unique, in contrast to the mating-type-specific or species-specific phylogenies displayed by other MAT genes.

Assuming that STE20 belongs to the P/R locus, the latter appears to be biallelic in each sibling species, whereas the HD locus defined by the SXI alleles appears to be multiallelic, at least for K. mangrovensis and K. europaea. These results are compatible with the hypothesis that the three sibling Kwoniella species are tetrapolar. This hypothesis is further corroborated by the fact that the alleles at both loci are not linked, as they are found in different combinations in the different strains of each species (Fig. 4 and ​and5;5; see also Table S1 in the supplemental material). We have in fact identified five different mating types with different STE20/SXI allele combinations in K. mangrovensis, three different mating types in K. europaea, and two in K. botswanensis (see Table S1). The molecular mating types are fully concordant with the results of mating experiments for K. mangrovensis (Fig. 1A; see also Table S1). On the other hand, in K. europaea, we would expect fertile matings between strains K25 and K28 or between K28 and K29. However, we did not observe any mating reactions with all four K. europaea strains. This result could be due to the fact that the tested strains are sterile or to nonoptimal mating conditions, but we were not able to confirm any of these hypotheses. Finally, in K. botswanensis, one likely explanation for the absence of observed sexual reproduction is that all studied strains have the same SXI allele (KbB1).

We tried to design degenerate primers to amplify also the STE3 and STE12 genes, which belong to the P/R locus in all heterothallic species of the Tremellales studied to date (12). This approach proved difficult due to the high sequence variability of those genes across species. However, we observed that in most taxa for which MAT structure has been elucidated, the two genes are almost always adjacent, although their relative orientation varies across different species. We designed one primer in the conserved domains of each gene and ordered also the reverse complement of each primer so as to perform PCR amplifications using all four possible primer combinations (see Text S1 in the supplemental material). We were able to obtain STE3 and STE12 gene fragments with only one primer combination from all KmA1 strains of K. mangrovensis tested but not from any of the KmA2 strains. The relative orientation of the successful primer pair suggests that the STE3 and STE12 genes are adjacent and divergently transcribed, at least in KmA1 strains of K. mangrovensis. The STE3/STE12 gene fragments had identical sequences in the KmA1 strains analyzed (K01, K06, K27). We were also able to amplify and sequence with the same primer pair STE3/STE12 gene fragments from two KeA1 strains of K. europaea (K25, K29) and one KbA1 strain (K10) of K. botswanensis (see Fig. S6 in the supplemental material) but not from the KeA2 or KbA2 strains. These results constitute further evidence that the STE20, STE3, and STE12 genes identified define the alleles of the P/R locus in the three sibling species.

Taxonomic status of taxa in the Cryptococcus and Kwoniella clades.

We included in our multilocus study several taxa in the Cryptococcus and Kwoniella clades, and the phylogenetic analyses of the combined data set (Fig. 2; see also Fig. S5 in the supplemental material) have confirmed that B. dendrophila, C. bestiolae, C. dejecticola, C. pinus, and C. shivajii all belong, together with K. mangrovensis, K. botswanensis, and K. europaea, to the Kwoniella clade. In view of the recent implementation of the “one fungus, one name” rule to fungal nomenclature (33), the above-mentioned Bullera species and four Cryptococcus species should all be transferred to the genus Kwoniella, but we refrain from formalizing this proposal here since a taxonomic revision of the genus is beyond the scope of the present paper. The phylogenetic position of K. heveanensis was not consistent in the individual gene trees (see Fig. S2 to S4 in the supplemental material), but the five-gene ML tree had this taxon at a basal position in the Kwoniella clade (Fig. 2) with 89% bootstrap support, thus justifying its inclusion in the genus, in accordance with the findings of Findley et al. (14). In the Cryptococcus clade, a single genus name should also be adopted, which should be Cryptococcus given that C. neoformans is the type species of the genus and has precedence over Filobasidiella. Tsuchiyaea wingfieldii was considered by Fonseca et al. (37) to be a synonym of C. amylolentus. However, the MAT gene analyses of Findley et al. (8) led the authors to propose that the two taxa should be kept separate at the species level. Our multilocus and MAT gene analyses support the hypothesis of the latter authors even though the evolutionary distance between the two taxa is less than half the value observed for the pair C. neoformans var. neoformans and C. neoformans var. grubii (Fig. 3). The ensuing transfer of T. wingfieldii to the genus Cryptococcus as C. wingfieldii should, however, await the study of more strains (only a single isolate is currently known) to obtain crucial supporting evidence.

Mating system of K. mangrovensis and sibling species.

In contrast to the original assumption of a bipolar mating system (13), our results have shown that K. mangrovensis instead has a tetrapolar system with two unlinked loci: a P/R locus defined by at least the STE20 gene (see below) and an HD locus defined by two divergently transcribed SXI genes (Fig. 3 and ​and4).4). Several lines of evidence support this conclusion. First, we found two alleles for STE20 and four alleles for SXI among nine strains of K. mangrovensis. Moreover, the association of the different alleles at both loci was found to vary in the different strains, defining five different mating types of eight possible combinations (Fig. 3 and ​and4;4; see also Table S1 in the supplemental material). Based on the results of previous studies on the MAT loci of closely aligned taxa in the Cryptococcus and Kwoniella clades for which tetrapolar systems have been demonstrated (8, 9), K. mangrovensis STE20 defines a biallelic P/R locus and SXI defines a multiallelic HD locus. Additionally, the adjacent position of the two divergently transcribed SXI1 and SXI2 homologs in K. mangrovensis, demonstrated by the fact that we amplified fragments of those two genes with a specific pair of primers anchored in the homeodomain regions of each gene, has been found only in tetrapolar species (C. amylolentus, K. heveanensis, and T. mesenterica), whereas in the bipolar pathogenic Cryptococcus species, only one gene is present in the MAT locus of each mating type (5). The results of mating experiments (Fig. 1) were concordant with the molecular mating types determined for the different strains and provide additional evidence supporting our hypothesis. The same line of reasoning is applicable to K. botswanensis and K. europaea, and the results obtained for the STE20 and SXI genes (Fig. 3 and ​and4)4) suggest that a tetrapolar mating system is also operating in this species. However, mating data were not conclusive. The possibility that K. mangrovensis would have a tetrapolar mating system had been suggested by Metin et al. (9) based on their study of the MAT locus of K. heveanensis, for which a tetrapolar system was demonstrated.

Work carried out at CREM was partly supported by Fundação para a Ciência e a Tecnologia (Portugal) (FCT), project PTDC/BIA-MIC/113051/2009. Work at Duke was supported by NIH/NIAID R37 grant AI39115; and R01 grant AI50113 to J.H.

We are indebted to Sheng Sun (Duke University) and Michael Weiss (University of Tübingen) for discussions and advice, Anna Floyd (Duke University) for technical assistance, and J. W. Fell (University of Miami), Neza Čadež (ZIM culture collection), Heide-Marie Daniel (MUCL culture collection), Kennio Ferreira-Paim (Universidade Federal do Triângulo Mineiro, Brazil), and Shawn Lockhart (CDC) for providing some of the isolates used in this study.