Chromosome-level assemblies and updated phylogenetic relationships of Cryptococcus and Kwoniella

To enable detailed analyses of MAT locus structure and evolution, we expanded our previous genomic dataset (29 assemblies; [49]) by incorporating genome sequences for 12 additional Cryptococcus and 7 Kwoniella strains, including opposite mating types where available. Among the Cryptococcus strains, four represent three currently undescribed species (Cryptococcus sp. 3, sp. 4, and sp. 5; Fig 1A), while the remaining eight are opposite mating types of pathogenic Cryptococcus species previously sequenced [49]. For Kwoniella, three strains belong to the recently described species K. ovata, K. endophytica, and K. fici [48,70,71], while the remaining four strains are different mating types of K. mangrovensis, K. europaea, K. botswanensis, and K. heveanensis, as previously determined through PCR and mating tests [52,66,67]. All but three genomes were assembled from long-read sequencing data (PacBio or Oxford Nanopore) and polished with short-read (Illumina) data, yielding chromosome-level genome assemblies. The exceptions, K. europaea PYCC6162, K. botswanensis CBS12717, and K. heveanensis BCC8398 were sequenced solely with Illumina, resulting in more fragmented assemblies. Analyzed strains were inferred to be haploid based on genome sizes consistent with those previously documented for haploid strains [49]. In total, the expanded dataset comprises 27 Cryptococcus strains (spanning 17 species) and 22 Kwoniella strains (spanning 18 species) (S1 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Phylogeny, breeding systems, and MAT loci organization in Cryptococcus and Kwoniella.

(A) Maximum likelihood phylogeny of strains analyzed in this study, inferred from a concatenated data matrix comprising protein alignments of 3,086 single-copy genes. Clades are labeled as A–I, with clade E shown as a nested subclade within clade F for easier cross-referencing. Branch lengths are given as substitutions per site (scale bar). All branches are supported >95% by SH-aLTR and UFBoot tests unless otherwise indicated. Gene concordance factor (gCF) and site concordance factor (sCF) were assessed to evaluate genealogical concordance, revealing unresolved branching order of clades A, B, and C (dashed branches; see S1 Fig for details). A black asterisk denotes that VGV has not yet been associated with human infections, and a white asterisk marks strains sequenced solely with Illumina, yielding more fragmented assemblies. (B) Predicted breeding systems, MAT locus organization, and mating-type identity of the P/R locus. Breeding systems are classified as heterothallic or homothallic for species with defined sexual cycles, and as likely heterothallic for species inferred to exhibit heterothallism based on MAT gene content. (C) Heatmap of gene presence/absence and pseudogene annotations within inferred MAT loci. Genes are ordered by their frequency of presence across species, with the most conserved genes on the left. Genes marked as “unclear” indicate cases where presence within MAT is uncertain because locus boundaries could not be precisely determined. BSP3 and IKS1, evicted from the MAT locus in C. deneoformans [34], are listed as absent in this species. Gene essentiality, as predicted or experimentally validated in strain H99 in prior studies rather than direct assessments in each species, is displayed at the bottom. Genes scored as absent from MAT may still be present elsewhere in the genome. (D) Gene presence frequency plot summarizing panel C. Pseudogenes were treated as present for frequency calculations, while unclear cases were treated as absent. The identification of either SXI1 or SXI2 was scored as “present” and combined for frequency quantification. (E) MAT loci locations in Cryptococcus sp. 3 (CWM60451), K. fici, and K. europaea, illustrating bipolar and pseudobipolar arrangements, respectively. Genes typically found either within or near the HD and P/R loci in tetrapolar species are designated as HD- and P/R-associated genes, and their respective chromosomal locations are indicated by gold or teal vertical bars. The data underlying this figure can be found in S1 Appendix and at https://doi.org/10.5281/zenodo.16987438.

https://doi.org/10.1371/journal.pbio.3003417.g001

To establish evolutionary relationships across species and determine the placement of the newly sequenced taxa, we identified 3,086 single-copy orthologous genes shared across Cryptococcus and Kwoniella, along with three outgroup species (Tremella mesenterica, Saitozyma podzolica, and Bullera alba), and reconstructed phylogenies employing both Maximum Likelihood (ML) and coalescent-based methods. Both approaches produced largely congruent results, except for the variable placement of Cryptococcus clades B and C relative to clade A (Figs 1A and S1). This inconsistency, noted in earlier studies [23,45,47], remains unresolved with the current dataset. Despite this, phylogenetic placement of the newly identified Cryptococcus and Kwoniella species was resolved. Cryptococcus sp. 3, known from a single strain (CMW60451) isolated from a bark beetle (Lanurgus sp.) infesting twigs of the endangered conifer Widdringtonia cedarbergensis in the Cederberg Mountains of South Africa, was placed as the closest relative of C. depauperatus, albeit on a long branch. Cryptococcus sp. 4 [72] and Cryptococcus sp. 5 clustered with Cryptococcus sp. 6 (OR918) [49,73], together forming a sister clade to all other known Cryptococcus. Among Kwoniella, K. ovata and K. endophytica were identified as sister species closely related to K. dendrophila, while K. fici grouped with Kwoniella sp. 4. The new Cryptococcus and Kwoniella species will be formally described elsewhere.

Sexual reproduction, breeding systems, and MAT gene identification and organization across Cryptococcus and Kwoniella

Sexual cycles have been described for eight Cryptococcus species: seven primarily heterothallic (C. neoformans, C. deneoformans, C. gattii, C. bacillisporus, C. deuterogattii, C. amylolentus, C. floricola), and one homothallic (C. depauperatus) (Fig 1A) [23,47,74–78]. In Kwoniella, sexual reproduction has been confirmed only in K. mangrovensis and K. heveanensis, both heterothallic [45,52,66]. We sought to identify sexual cycles in species where none had been described, restricting tests to cases where opposite mating types were available. Crosses were performed under conditions known to induce mating in Cryptococcus [79] and examined for the presence of hyphal growth or other sexual structures (basidia and basidiospores).

In C. decagattii, sexual reproduction was confirmed by crossing strains 7685027 (MATα) and CBS11687 (MATa), which produced sexual structures (S2A Fig). In contrast, crosses between C. tetragattii IND107 and CBS11718 did not yield sexual structures (S2B Fig). For Cryptococcus sp. 3, sister to C. depauperatus, we observed key differences: while C. depauperatus is homothallic and exhibits continuous hyphal growth [23], Cryptococcus sp. 3 grew mostly as yeast and did not produce sexual structures alone, suggesting it is unlikely to be homothallic (S2C Fig). Within clade D species, we tested Cryptococcus sp. 4, where two strains of opposite mating types are available. Although some hyphal growth was observed, it was inconsistent, and no discernible sexual structures formed even after prolonged incubation (S2D Fig). These negative results may reflect low mating efficiency under laboratory conditions, and future genetic approaches such as CRG1 deletion [141,142] could help uncover latent sexual potential in these lineages. Given the absence of confirmed sexual reproduction in most of the species tested, we resorted to genomic data to examine MAT gene content and organization, inferring breeding systems (heterothallic or homothallic) and classifying MAT loci configurations as tetrapolar, pseudobipolar, or bipolar, as summarized in Fig 1B.

Consistent with previous studies [34,78], we confirmed that all pathogenic clade A species, including C. decagattii and C. tetragattii, have a bipolar mating configuration supported by genome assemblies from both mating types. In contrast, most of the nonpathogenic Cryptococcus and Kwoniella species appeared to be heterothallic and tetrapolar, consistent with the proposed ancestral state of basidiomycetes [9,80,81]. Three exceptions were identified: Cryptococcus sp. 3 (clade C), K. europaea (clade E), and K. fici (clade H) (Fig 1B and 1E). In Cryptococcus sp. 3 and K. fici, BLAST searches revealed that the key MAT genes (HD and the pheromone receptor STE3) are close together on the same chromosome, consistent with bipolarity (Fig 1E). In contrast, K. europaea exhibited a pseudobipolar configuration, with HD and STE3 on the same chromosome but separated by ~8.95 Mb (Fig 1E). These findings prompted detailed characterization of MAT loci across species and reconstruction of the chromosomal changes underlying bipolar configurations.

Structure and evolution of MAT loci in Cryptococcus and Kwoniella

Identification and general organization of MAT loci.

The structure and gene composition of MAT loci have been described for several Cryptococcus species within clades A, B, and C [16,23,34,35,47,78,82], but in Kwoniella analyses have been limited to K. mangrovensis and K. heveanensis, based solely on fosmid libraries from a single mating type, leaving the precise structure of MAT loci unresolved [66,67]. To expand on these analyses, we combined genome-wide synteny comparisons with gene genealogies to delineate MAT loci across both genera. The HD locus was defined based on the presence of the HD1 and HD2 genes, while the P/R locus boundaries were delineated by identifying regions of disrupted synteny between mating types (see Materials and methods).

In tetrapolar Cryptococcus and Kwoniella species, the HD locus is compact, consisting only of the divergently transcribed HD1/SXI1 and HD2/SXI2 genes (S3 Fig). As in other basidiomycetes, mating-type differences are mainly confined to the N-terminal regions of the encoded proteins, and this two-gene structure is strongly conserved. This organization persists in Kwoniella species, even though the HD locus predominantly resides in subtelomeric regions, inherently more prone to rearrangements (S3J–S3O Fig). By contrast, the P/R locus spans a larger chromosomal region and is more variable across species and between mating types of the same species (Figs 2 and S4). It is significantly expanded in tetrapolar Cryptococcus species, averaging 91.7 kb compared to 29.3 kb in Kwoniella (Figs 2 and S4; S2 Appendix; P < 0.0001, Mann–Whitney U test). Within Cryptococcus, sizes range from 84.5 kb in Cryptococcus sp. 5 (P/R a2) to 101.5 kb in C. floricola (P/R a1), while in Kwoniella, they range from 24.3 kb in K. newhampshirensis (P/R a1) to 34.4 kb in K. heveanensis (P/R a2) (S2 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Synteny analysis of the P/R locus in tetrapolar Cryptococcus and Kwoniella species.

(A, B) Detailed views of the P/R locus of representative Cryptococcus species from clades B and D. In both clades, the P/R locus exhibits lower GC content compared to other genomic regions. In clade D species, the predicted P/R locus contains only a subset of the genes found within the P/R locus of clade B species or the MAT locus of pathogenic Cryptococcus species. The additional genes in clade B or pathogenic species are instead located in the immediate downstream region, suggesting independent expansion of this genomic segment in clade D species. (C, D) P/R locus region in representative Kwoniella species of clades E and G. In a2 strains, the STE12 gene appears truncated. Remnants of the NDC80 gene (in clade E/F species) or the BSP3 gene (in clade G species) are observed at the right edge of the P/R a2 allele, likely resulting from inversion events. In all panels, P/R-associated genes are colored teal if their orthologs in Cryptococcus pathogens are located within the MAT locus, with darker teal indicating genes positioned in the flanking regions, and bright green for genes present within the P/R locus of C. amylolentus, but absent in most Cryptococcus pathogens. The P/R allele of each strain (a1 or a2) is indicated on the left. Chromosomes inverted relative to their original assembly orientations are marked with asterisks. GC content is depicted as the deviation from the genome average, calculated in 0.5 kb nonoverlapping windows. Gene essentiality within and in the immediate vicinity of the P/R locus is inferred from predictions or experimental validations in C. neoformans H99 rather than direct assessments in these species. Other features are annotated as shown in the key. See S4 Fig for additional comparisons. The data underlying this figure can be found at https://doi.org/10.5281/zenodo.16987438.

https://doi.org/10.1371/journal.pbio.3003417.g002

Evolution of the P/R locus in Kwoniella.

Gene content analysis of the Kwoniella P/R locus revealed nine genes shared across all species (Figs 1C, 2C, and 2D; S3 Appendix). These include a single pheromone gene (MFα or MFa) containing the canonical CAAX motif at its C-terminus, a pheromone receptor (STE3α or STE3a), and two additional genes (STE20 and STE12) with established mating-related roles and found within MAT in Cryptococcus pathogens [34,83,84]. Other conserved genes include DBP10 and REI1, both implicated in ribosome biogenesis in Saccharomyces cerevisiae [85,86]; ATG7, an autophagy-related gene [87]; a C2 domain-containing protein (C2prot), whose ortholog in S. pombe (Ync13) coordinates exocytosis, endocytosis, and cell-wall integrity during cytokinesis [88]; and NDC80, encoding a component of the NDC80 kinetochore complex, essential for chromosome segregation and spindle checkpoint activity [89]. In K. heveanensis and other clade G species, two additional genes are present within the P/R locus: BPS3, encoding a protein of unknown function with BTP/PZ and CAP-Gly domains (also present in the MAT locus of Cryptococcus pathogens), and UPB6, a predicted deubiquitinating enzyme associated with the 26S proteasome [90]. Of these, DBP10, REI1, and NDC80 are predicted to be essential in S. cerevisiae and in C. neoformans, based on previous studies [85,86,89,91,92].

Across Kwoniella, the P/R a2 allele is generally larger than the a1 (Figs 2C, 2D, and S4; S2 Appendix), mainly due to the presence of an additional gene of unknown function. This size difference is also accompanied by truncation of STE12 within the a2 allele, with fragments flanking STE3 on both sides in several species (e.g., K. mangrovensis and K. pini; Figs 2C and S4). Fragments of the 3′ end of NDC80 and BSP3 are likewise found at the right edge of the P/R a2 allele in certain clade E/F and clade G species, respectively, in addition to apparently intact copies (Figs 2C and S4). These fragments seem to be remnants of structural rearrangements, potentially associated with duplication tracks at inversion breakpoints. Reconstruction of the likely rearrangements leading to the extant P/R configuration in Kwoniella suggests three distinct inversions have shaped the locus in clades E/F: one involving STE3 and STE12, truncating STE12 in the a2 allele; a second relocating DBP10 and REI1 from the edge of the locus to its center; and a third moving NDC80 from the edge to the middle of the locus (S5A Fig). The first two inversions are conserved in clade G, indicating they are ancestral in Kwoniella, while clade G additionally underwent rearrangements repositioning BSP3-NDC80 and UBP6 to the center of the locus (S5B Fig). This suggests that the P/R a2 allele has experienced more modifications than the P/R a1 allele.

Additional complexity is seen in K. ovata and K. endophytica, where the P/R locus contains STE3a and MFa alleles alongside truncated versions of STE3α, with K. ovata additionally harboring a truncated MFα gene (Fig 3A). This mosaic configuration likely arose through intra-P/R recombination, as synteny analyses showed that the left side of the locus resembles that of P/R a2 strains, whereas the right side aligns more closely with the P/R a1 (Fig 3A). Gene genealogies further supported this scenario. While deep trans-specific polymorphism has been well-documented for STE3, STE12, and MF genes, reflecting their ancestral integration into the P/R locus across Cryptococcus and Kwoniella [23,47,67,78], shallower patterns were detected for NDC80, BSP3, ATG7, and UBP6. Specifically, NDC80 exhibited trans-specific polymorphism within clades E/F and G (Fig 3B), while ATG7, BSP3, and UBP6 displayed this pattern only within clade G (S5C Fig). Notably, the NDC80 allele at the right edge of the P/R locus in K. ovata and K. endophytica clustered more closely with a1 alleles from other species (Fig 3B), and a similar pattern was found in K. dendrophila, together pointing to a shared evolutionary history shaped by past intra-P/R recombination. This recombination event may have initially produced strains with intercompatible pheromone and receptor pairs, potentially enabling self-filamentous growth. However, subsequent loss of STE3α and MFα alleles appears to have eliminated self-filamentation. Consistently, solo cultures of K. ovata and K. endophytica on V8 pH 5 media did not form hyphae, indicating that extant species retain only relics of this ancestral recombination event. Collectively, these findings unveil a dynamic evolutionary history of the Kwoniella P/R locus, shaped by structural rearrangements, recombination, and small-scale lineage-specific expansions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The P/R locus in K. ovata, K. endophytica, and K. dendrophila likely arose through intra-P/R recombination.

(A) Synteny analysis comparing P/R a1 and P/R a2 alleles in select Kwoniella species from clade F. The P/R loci of K. ovata and K. endophytica contain STE3a and MFa alleles, along with truncated (ψ) versions of STE3α. Additionally, K. ovata harbors a truncated MFα gene. Synteny analysis indicates that the left side of the P/R locus in K. ovata, K. endophytica, and K. dendrophila is structurally more similar to P/R a2 alleles of K. pini and K. shivajii, while the right side aligns more closely with the P/R a1 alleles of K. bestiolae and K. dejecticola. (B) Gene genealogy of the NDC80 gene inferred with IQ-TREE2 (model JTT + G4) showing trans-specific polymorphism across clade E/F species (labeled as 1) and within clade G species (labeled as 2). The NDC80 alleles of K. ovata, K. endophytica, and K. dendrophila cluster more closely with a1 alleles than with other a2 alleles. The data underlying this figure can be found at https://doi.org/10.5281/zenodo.16987438.

https://doi.org/10.1371/journal.pbio.3003417.g003

Independent origins and evolutionary outcomes of MAT locus fusion

Post-fusion structural dynamics of MAT in Cryptococcus pathogens.

The fusion of the P/R and HD loci into a single nonrecombining MAT locus is well established in clade A, which encompasses all human-pathogenic Cryptococcus species. We extended MAT locus analysis to all recognized species in this clade (Figs 1, 4, S6, and S7; S1 Appendix). For C. deneoformans, our dataset included congenic strains JEC21α and JEC20a, their progenitor strains NIH12α and NIH433a, and an additional MATa strain (NIH430) (S6 Fig and S1 Appendix). For C. neoformans, we examined multiple strains of both mating types across the four main lineages (VNI, VNII, VNBI, and VNBII), including eight newly sequenced genomes assembled telomere-to-telomere (T2T) (Fig 4A and S1 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Synteny analysis of the mating-type locus (MATα and MATa) across different Cryptococcus neoformans lineages.

(A) Phylogenetic tree of representative C. neoformans strains from different VN lineages, inferred using maximum likelihood analysis of a concatenated matrix of protein alignments from 5,439 single-copy genes. Branch support is based on SH-aLTR and UFBoot tests. Mating types were determined based on the allelic version of the pheromone receptor gene (STE3). (B) Synteny analysis of chr. 5, where the MAT locus resides, showing significant conservation across strains, except within the MAT and centromere regions, which exhibit rearrangements. Strains are organized by mating type and lineage, as in panel D. (C) Box plot comparing MAT locus size between α and a strains of C. neoformans (C.n.) and C. gattii species complex (C.g.c.). The red dashed line, blue line, and boxes denote the mean value, median value, and interquartile range, respectively. Outliers are labeled with the corresponding strain name. Statistical significance (P-values) was determined by the Mann–Whitney U test. (D) Detailed synteny of the MAT locus, highlighting strain- and mating-type-specific rearrangements. The MATa configuration is more variable both across and within VN lineages, existing in three distinct configurations, whereas the MATα structure is more conserved, with limited rearrangements between ancestrally linked homeodomain (HD) and pheromone receptor (P/R) genes. Observed rearrangements appear to result from inversions mediated by identical copies of pheromone genes. KDZ transposons are detected within the MAT locus of certain strains. Genes typically associated with the P/R and HD loci in tetrapolar species are colored green and gold, respectively, with darker shades denoting genes flanking the MAT locus in Cryptococcus pathogens. Gene essentiality within the MAT locus, as predicted or experimentally validated in strain H99, is indicated by color-coded circles as given in the key. The data underlying this figure can be found at https://doi.org/10.5281/zenodo.16987438.

https://doi.org/10.1371/journal.pbio.3003417.g004

Synteny comparisons revealed that the MATa allele is generally longer than the MATα allele in both C. neoformans (P < 0.0001, Mann–Whitney U test) and the C. gattii species complex (P = 0.04142, Mann–Whitney U test; Fig 4C and S2 Appendix), with a similar trend in C. deneoformans, although limited sampling precluded statistical support. The MATa allele also exhibits greater structural diversity than MATα, as seen in the C. gattii complex (S7A Fig) and in C. neoformans, where frequent inversions were detected even among MATa strains of the same VN lineage (e.g., VNI strains Bt130 versus IUM96-2828, or strain Bt206 compared to other VNBII strains; Fig 4D). Inversion breakpoints were often associated with pheromone genes, consistent with the hypothesis that multiple pheromone copies may act as mediators of structural changes. In contrast, MATα is structurally more conserved with only limited inversions involving the pheromone–gene pairs flanking the PRT1 and ZNF1 genes (Figs 4D and S7). Comparison of the entire MAT chromosome across C. neoformans strains further showed that most of the intraspecific structural variation is confined to the MAT locus, with variation outside MAT largely restricted to the inherently dynamic centromeric and telomeric regions (Fig 4B).

As part of this analysis, we also detected significant structural divergence between the MATa alleles of C. neoformans VNBI (Ftc555-1 and Bt63) and VNI strains (Fig 4D). The VNI lineage has been hypothesized to have diverged initially with only the MATα allele, possibly due to a bottleneck associated with a small founding population, and later acquired the MATa allele via introgression from a VNBI MATa donor [95]. To date, only five VNI MATa strains have been identified [95–100], three of which are now represented by high-quality genome assemblies (125.91 and Bt130, sequenced in this study, and IUM96-2828, sequenced previously [101] and annotated here; Fig 4D). This structural divergence did not contradict the introgression model but raised questions about how such differences could be reconciled with an introgressed origin.

To revisit this question with improved data, we reanalyzed SNP data leveraging newly generated reference genomes. Genome-wide SNP analysis, with VNBI MATa strain Ftc555-1 as the reference, revealed markedly lower SNP densities within the MATa locus in VNI strains compared to VNBII strains (Fig 5A and 5B). Conversely, across the rest of the genome, VNI MATa strains diverged more strongly from the VNBI reference, while VNBII strains showed lower SNP densities (Fig 5A and 5B). Notably, the typically high SNP density between VNI and VNBI strains was restored immediately downstream of MAT and ~5 genes upstream, consistent with recombination hotspots flanking the locus [102] and previously mapped introgression boundaries [95]. Phylogenetic analyses reinforced these patterns: VNBI strains clustered more closely with VNBII strains in trees inferred from genome-wide and chr. 5-specific SNPs (Fig 5C and 5D), whereas MAT-specific SNP trees placed VNI and VNBI MATa strains together, consistent with a more recent shared ancestry for the MATa allele (Fig 5E). Together, these findings refine the VNBI-to-VNI introgression model (Fig 5F) and indicate that lineage-specific rearrangements following introgression account for the structural differences observed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. SNP distribution, phylogenetic analysis, and revised introgression model of the MATa allele from VNBI into VNI.

(A) Genome-wide distribution of SNPs across chromosome 5 in C. neoformans MATa strains from different VN lineages relative to the C. neoformans Ftc555-1 VNBI genome reference. Empty spaces in the plots represent either sites identical to the reference genome (monomorphic sites) or regions excluded due to low coverage or repetitive/duplicated sequences. Strains on the y-axis are ordered by the total number of SNPs across the genome, with a color gradient (right) indicating SNP counts: darker colors represent higher SNP densities relative to the reference (strains positioned at the top), while lighter colors indicate fewer SNPs (strains at the bottom). Reads from the reference strain were also mapped as a control. The x-axis shows genomic coordinates. (B) Zoomed-in view of the MAT locus region plus 50 kb flanking regions, with genes and other features represented as in Fig 4. Notably, VNI strains exhibit a markedly lower number of SNPs at the MATa locus relative to the VNBI reference, compared to VNBII strains. (C) Midpoint-rooted tree based on genome-wide SNPs across all chromosomes, illustrating the evolutionary relationships of the strains. (D) Midpoint-rooted tree constructed using SNPs from chromosome 5 only, showing similar relationships to the genome-wide analysis. (E) Midpoint-rooted tree based on SNPs restricted to the MAT locus, highlighting a closer relationship between VNI and VNBI strains. In all trees, branches are shown as the number of substitutions per site (scale bars) and have >95% support based on SH-aLTR and UFBoot tests, unless otherwise indicated. “Ref” represents the reference genome assembly employed for read mapping and is thus identical to Ftc555-1 reads, serving as a control to validate the SNP calling pipeline. (F) Proposed model for the introgression of the VNBI MATa allele into VNI, shown alongside the key stages of the C. neoformans sexual cycle (right). Step 1: possible loss of the MATa allele in the VNI lineage, likely due to a bottleneck; step 2: mating and cell-cell fusion between a VNBI (MATa) and a VNI (MATα) strain; step 3: formation of dikaryotic hyphae, nuclear fusion, meiosis, recombination, and sporulation; step 4: some of the resulting F1 progeny emerged with a ~50:50 VNBI:VNI genome composition, retaining the VNBI MATa locus; step 5: repeated backcrossing of the MATa F1 progeny to VNI MATα strains progressively purging VNBI genomic regions while retaining the introgressed MATa allele; step 6: after ~10 backcrosses, the genome became ~99.95% VNI, with only the VNBI MATa locus and possibly a small fraction (~0.05%) of the VNBI genome remaining; step 7: over time, the introgressed VNBI MATa locus diverged and underwent rearrangements within the VNI background, reflecting strain-specific changes. The data underlying this figure can be found at https://doi.org/10.5281/zenodo.16987438.

https://doi.org/10.1371/journal.pbio.3003417.g005

In addition to introgression, other forces also shaped MAT architecture in C. neoformans. Species-specific gene losses (e.g., NCP1 and NCP2, which are absent in most species) and TE insertions contributed further to allele differentiation (S7 Fig). TE activity appears especially important, as the C. deneoformans MAT locus contains over five times more TEs than genome averages outside centromeres [16,34,103]. In line with this, we identified novel insertions of two distinct KDZ transposons, recently characterized in C. neoformans as large mobile elements (~11 kb) with terminal inverted repeats and target site duplications [104]. KDZx was present in strains 125.91, C45, PMHc1023, and T4, while KDZ1 was found in strain Ze90-1 (Fig 4D). Altogether, these findings underscore the dynamic evolution of the MAT locus in Cryptococcus pathogens, shaped by an interplay of structural rearrangements, gene loss, TE activity, and introgression.

Independent fusion of P/R and HD loci in Cryptococcus sp. 3.

Our initial screening of MAT gene content revealed that the P/R- and HD-associated genes are adjacent on the same chromosome in Cryptococcus sp. 3, indicating a fusion of these two regions (Fig 1E). While only a single strain (CMW60451) is available, limiting precise delineation of MAT boundaries, the minimum estimated size is ~90 kb based on the distance between the two most distal pheromone genes (Fig 6C). If boundaries are extended to include additional genes that are part of the MAT locus in pathogenic Cryptococcus, the locus could span ~120 kb, from RPL22 on the left to STE12 on the right (Fig 6C). Under this broader definition, 15 genes are shared between Cryptococcus sp. 3 and clade A species (Figs 1C and 6C; S3 Appendix), including four MFα pheromone gene copies and three genes typically associated with the HD locus (RPL22, CAP1, and SPO14). Additionally, two regions within the MAT locus contain TEs: one with a complete KDZ-like transposon and another with only remnants of this element (Fig 6C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Transition to linked P/R-HD loci in Cryptococcus sp. 3 (CMW60451) and Cryptococcus pathogens.

(A) The karyotype of K. shandongensis (with 14 chrs.) served as the reference for reconstructing synteny blocks in pairwise comparisons. For simplicity, only synteny blocks corresponding to P/R- and HD-containing chromosomes (chrs. 12 and 5, colored teal and gold, respectively) were plotted in representative Cryptococcus species of clades A (C. neo. H99), B (C. amy. CBS6039), C (C. sp. CMW60451), and D (C. sp. DSM108351). Besides the MAT-containing chromosomes, synteny blocks of two additional chromosomes in K. shandongensis (chrs. 10 and 14), which correspond to a large portion of P/R and HD chromosomes in C. amylolentus were also plotted (color-coded light blue and green, respectively). Other synteny blocks can be visualized in S9A Fig. A red arrowhead pinpoints the predicted location of an inactivated centromere (ic) in Cryptococcus sp. CMW60451. (B) Linear chromosome plots depicting gene synteny conservation across species with zoomed-in views depicted in (C). In panel B, the MAT chromosome of bipolar species (C. neoformans and C. sp. 3, CMW60451) is shown twice: the left column depicts synteny relative to the P/R-containing chromosomes of tetrapolar species, and the right column depicts synteny relative to the HD-containing chromosomes.

https://doi.org/10.1371/journal.pbio.3003417.g006

Further analysis uncovered an unusual combination of MAT alleles. Unlike the bipolar Cryptococcus pathogens, where MATα strains carry HD1 (SXI1) linked to STE3α and MATa strains have HD2 (SXI2) linked to STE3a, Cryptococcus sp. 3 retains only SXI2 linked to STE3α. Although this distinct configuration suggested an independent fusion event at the origin of the MAT locus in Cryptococcus sp. 3, the unresolved phylogenetic placement of clade C relative to clade A left open the possibility that the two clades share a more recent common ancestor. In this scenario, the fusion of the P/R and HD loci could have occurred in their shared ancestor rather than independently in both lineages.

To test this hypothesis, we analyzed the phylogenetic clustering patterns of genes within the predicted MAT region shared with pathogenic Cryptococcus. Genes ancestrally present in the P/R locus, such as STE3, STE12, and MYO2, were excluded from the analysis because they exhibit deep trans-specific polymorphism predating the divergence of Cryptococcus species [23,47,67,78], making them uninformative. Instead, we focused on four genes (BSP3, PRT1, SPO14, and CAP1), which are part of MAT in pathogenic species and display mating-type-specific clustering within them [34]. The analysis revealed that Cryptococcus sp. 3 sequences do not group with either the a or α allele-specific clusters from pathogenic species; rather, they are distinct and fall outside these clusters, which is more consistent with an independent fusion event than one inherited from a shared ancestor (S8 Fig).

Having established from gene phylogenies that the MAT fusion in Cryptococcus sp. 3 likely arose independently, we next examined chromosome-scale structural changes underlying the P/R-HD linkage by reconstructing synteny blocks with SynChro [105]. K. shandongensis was selected as reference because (i) it is tetrapolar, with P/R and HD loci on separate chromosomes representing the ancestral state (Figs 1, S3N, and S4N); (ii) Kwoniella species show significantly fewer interchromosomal rearrangements than Cryptococcus, likely due to their simpler, smaller centromeres and lower TE density, as both high TE content and large, complex centromeres promote genomic rearrangements in Cryptococcus [49,106–108]; and (iii) it retains a 14-chromosome karyotype, which is the predicted ancestral state for both genera [49] (Figs 6 and S9A).

The nuclear genome of Cryptococcus sp. 3 comprises 13 chromosomes (S1 Appendix), representing a reduction from the ancestral 14-chromosome karyotype. While we hypothesized that this reduction could explain the P/R-HD fusion, examination of the lost centromere showed it did not involve the MAT-containing chromosome (detailed in S2 Text and S9B Fig). Thus, physical linkage of the P/R and HD loci in Cryptococcus sp. 3 arose independently of the karyotypic reduction. We therefore compared the MAT-containing chromosome of Cryptococcus sp. 3 to the P/R- and HD-bearing chromosomes of K. shandongensis. Synteny analysis revealed significant alignment between chr. 11 of Cryptococcus sp. 3 and large regions of K. shandongensis chrs. 12 and 5, which carry P/R and HD, respectively (Fig 7A). Additionally, a small segment from the P/R chromosome is embedded towards the left end of chr. 11 within a region otherwise derived from the HD chromosome. These patterns suggest that the physical linkage likely began with a translocation that brought the two loci onto the same chromosome, initially far apart, followed by an inversion relocating the HD locus closer to the P/R locus, resulting in their fusion (events 1–3 in Fig 7A). Notably, the centromere-flanking regions of chr. 11 in Cryptococcus sp. 3 show considerable synteny to the CEN12-flanking region in the K. shandongensis P/R chromosome, indicating a shared evolutionary origin and suggesting that the translocation did not involve centromeric recombination but occurred elsewhere (Fig 7A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Hypothesized mechanisms underlying the tetrapolar-to-bipolar transitions in Cryptococcus sp. 3 and in Cryptococcus pathogens.

Comparative synteny analysis between K. shandongensis and (A) Cryptococcus sp. 3 or (B) C. neoformans, illustrating the inferred sequence of chromosomal rearrangements leading to this transition. In both panels, the left side displays pairwise alignments, with syntenic regions connected by red (collinear) and gray (inverted) links, while the right side depicts a stepwise model of the proposed structural changes. The transition likely began with (1) an initial chromosomal break and reciprocal translocation (red dashed line), leading to (2) the repositioning of the P/R locus onto the same chromosome as the HD locus. (3) In Cryptococcus sp. 3, a subsequent inversion relocated the HD locus closer to the presumed ancestral position of the P/R locus, whereas in Cryptococcus pathogens, an inversion instead shifted the P/R locus toward the original location of the HD locus. (4) Additional structural modifications further refined the newly linked configuration, ultimately resulting in (5) the extant MAT locus organization observed in each species. These models suggest that despite independent evolutionary trajectories, both lineages underwent convergent genomic rearrangements that facilitated the transition to bipolar mating systems.

https://doi.org/10.1371/journal.pbio.3003417.g007

Tetrapolar to bipolar and pseudobipolar transitions in Kwoniella.

We recently reported that karyotype reduction within Kwoniella occurred progressively and independently, often with formation of “giant” chromosomes (up to 18 Mb) through repeated fusions [49]. Although such events could potentially lead to P/R-HD physical linkage, only two Kwoniella species harbor both loci on the same chromosome (Fig 1B). In K. europaea (clade E), the HD and P/R loci are located on the same chromosome but remain far apart (~8.95 Mb) (Figs 1E and S4G), representing a pseudobipolar configuration. This arrangement likely arose from a lineage-specific rearrangement, in which the progenitor “giant” chr. 1 of clade E species underwent a translocation with chr. 2 [49]. Given their physical distance, the two MAT loci are still expected to recombine.

In K. fici, by contrast, the HD locus lies less than 13 kb from the P/R gene cluster, both located near the end of chr. 1 (Fig 8D). This “giant” chromosome likely arose independently through fusion events, similar to those documented in Kwoniella clades E and G [49]. After divergence from clade I (K. shandongensis/K. newhampshirensis), the chromosomes corresponding to chrs. 11 and 12 of K. shandongensis fused in the descendant lineage, producing the ancestral P/R chromosome, which persists as a single chromosome in most Kwoniella species (e.g., chr. 2 in Kwoniella sp. 4 DSM27419; Fig 8A) [49]. In K. fici, however, this ancestral fusion product fragmented further: one block now resides at the terminal region of K. fici chr. 1, while another segment lies on chr. 4 (Fig 8C). Similarly, most of the K. shandongensis/Kwoniella sp. 4 HD chromosome (chr. 5), aligns with the opposite end of K. fici chr. 1, while a small portion is on chr. 4, adjacent to the P/R-derived block. Only a few key HD-associated genes (SXI1, SXI2, and RPL22) are positioned near the P/R locus gene cluster at the right end of chr. 1 (Fig 8D). Notably, this derived bipolar arrangement retains both HD genes, unlike bipolar Cryptococcus species where opposite mating types carry only one. Overall, these observations indicate that complex translocations united blocks from ancestral P/R and HD chromosomes, producing the unique fused configuration in K. fici.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Linkage of P/R and HD loci in Kwoniella fici and MAT locus structure.

(A) The karyotype of K. shandongensis (with 14 chrs.) served as a reference for reconstructing synteny blocks in pairwise comparisons with Kwoniella sp. 4 and K. fici. (B) Phylogenetic relationships among selected Kwoniella species, highlighting their MAT locus organization and inferred breeding systems. (C) Synteny comparison between K. fici and its closest relative Kwoniella sp. 4, illustrating chromosomal rearrangements underlying the linkage of P/R and HD loci in K. fici. A red double-headed arrow indicates that the centromere-proximal regions of chr. 2 in Kwoniella sp. 4 correspond to a region near the fusion point on the K. fici “giant” chromosome (chr. 1), whereas the telomere-proximal regions of Kwoniella sp. 4 chr. 2 align with more internalized regions. This suggests that a large pericentric inversion targeting the centromere-adjacent region is associated with this fusion event, as previously reported for other Kwoniella species [49]. (D) Detailed gene-level organization of the MAT locus in K. fici compared to Kwoniella sp. 4. Each track (labeled a–d in dark circles) corresponds to a specific region marked by a black bar in panel C. Chromosomes inverted relative to their original assembly orientations are marked with asterisks.

https://doi.org/10.1371/journal.pbio.3003417.g008

Analysis of sexual reproduction in Kwoniella mangrovensis

Kwoniella mangrovensis (clade G) can reproduce sexually in the lab [52,66] and has only three chromosomes. To better understand its reproductive biology, we analyzed recombination, ploidy variation, and the potential for self-fertility in the progeny. The two originally sequenced strains, CBS8507 (P/R a1, HD b1) and CBS8886 (P/R a2, HD b1), share identical HD b1 alleles, making them incompatible for mating (Fig 9A). However, Guerreiro and colleagues [66] reported a third strain, CBS10435, as haploid and compatible with CBS8507. We sequenced CBS10435 with Illumina and confirmed it carries the P/R a2 allele and a distinct HD b2 allele, rendering it compatible with CBS8507 (Fig 9A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Sexual reproduction and recombination in Kwoniella mangrovensis.

(A) Synteny analysis of the P/R and HD loci among three K. mangrovensis strains of different mating types (CBS8507, a1b1; CBS8886, a2b1; and CBS10435, a2b2). Note the pronounced divergence in the N-terminal domains of both HD1 (SXI1) and HD2 (SXI2) gene products, which are known to mediate mating-type specificity in other basidiomycetes. (B) Micrographs showing hyphal filaments extending from colony peripheries in a sexual cross between CBS8507 and CBS10435 after 2 weeks of incubation on CMA and V8 pH 5 media, in the dark, at room temperature. Insets depict two types of basidia: globose (a–c) and lageniform (b–d). Black arrows indicate clusters of cells emerging near basidia (either at the surface or embedded), scored as potential meiotic progeny (basidiospores). Single cells from these clusters were isolated using a micromanipulator, cultured into colonies, and genotyped by PCR-RFLP (see S5 Appendix). Putative recombinants were further analyzed using Illumina sequencing. Scale bars = 100 μm (25 μm in insets). (C) Genotypic analysis of selected meiotic progeny. SNP density between the parental strains was calculated as the number of SNPs per 10 kb (top). The genotypes of 9 segregants, derived from four distinct cell clusters, were inferred from SNP data and are depicted as follows: orange for regions inherited from CBS8507, purple for regions inherited from CBS10435, and brown for heterozygous regions (i.e., inherited from both parents). Instances of recombination or loss of heterozygosity (LOH) are detected by changes in genotype along the chromosomes. Discrepancies in ploidy, as inferred from FACS and sequencing read coverage, suggest potential genomic instability (marked by asterisks). Mating-type identity was inferred through sequencing, coverage analysis, and FACS. (D) Sequencing coverage plot for progeny MP59 with color-coded contributions from each parent. Haplotype blocks inferred from SNP data are overlaid for each chromosome for comparison (additional progeny data is presented in S8 Fig). (E) Self-filamentation phenotype of K. mangrovensis progeny. Parental strains CBS8507 and CBS10435 (grown individually and in co-culture) and their recovered progeny were cultivated on CMA at room temperature. Self-filamentation was assessed after 2 weeks of incubation. While neither parental strain exhibited self-filamentation in solo culture, their co-culture produced hyphal filaments and basidia. Progeny MP16, MP18, and MP54 also exhibited robust self-filamentation on both CMA and V8 pH 5 (see S9 Fig), whereas progeny MP49, MP51, and MP59 displayed weaker filamentous growth, detectable only on V8 pH 5 (S9 Fig). Scale bars = 200 μm. (F) Summary of the growth phenotype of the recovered progeny. The production of mycelium and basidia was classified as: extensive (+++), when observed across the entire periphery of the mating patch; moderate (++), when restricted to specific areas of the mating patch; poor (+), when limited to a single spot of the mating patch and slower to develop; and negative (−), when no filamentous growth was observed. Results represent observations from two independent tests.

https://doi.org/10.1371/journal.pbio.3003417.g009

Crosses between CBS8507 and CBS10435 on CMA and V8 pH 5 media produced basidia with characteristic globose and lageniform morphologies, consistent with prior descriptions (Fig 9B) [52]. Progeny were recovered by dissecting eight clusters of cells germinating from basidia along the hyphae (S5 Appendix). In total, 51 F1 progeny were obtained from both culture media. To avoid analyzing clonally derived genotypes, we screened for recombination with a PCR-RFLP approach using two markers per chromosome, positioned on either side of the centromere (S5 Appendix). Evidence of allele exchange and heterozygosity was detected, and nine F1 progeny with distinct genotypes were selected for whole-genome sequencing. SNPs between the parental strains served as markers to infer inheritance patterns associated with recombination and loss of heterozygosity (LOH), while read coverage analysis and FACS were combined to assess ploidy.

These analyses revealed that most F1 progeny were nonhaploid (Figs 9C, 9D, S10, and S11). Genotyping and FACS results were largely concordant, though some isolates (e.g., MP11 and MP18) showed discrepancies, suggesting variability within cell populations, possibly stemming from intrinsic genomic instability. Accordingly, mapping reads to a combined reference genome (CBS8507 and CBS10435) revealed uneven coverage for certain chromosomes. For example, CBS10435 chr. 2 exhibited low coverage in progeny MP59 (Fig 9D). Although CBS8507 and CBS10435 were previously reported as haploid [66], our FACS analysis showed that CBS10435 is predominantly diploid with a haploid subpopulation (Figs 9C and S11). This difference in parental ploidy likely underlies the aneuploidy observed in the progeny, as haploid–diploid matings are prone to chromosome mis-segregation, and the reduced karyotype of K. mangrovensis (3 chromosomes) may further amplify these effects.

Analysis of the MAT gene content revealed that some progeny carried fully compatible P/R and HD alleles (a1a2b1b2), potentially enabling self-fertility (Fig 9C). To investigate this, we cultured the F1 progeny under mating conditions and compared their phenotypes to a parental cross (Figs 9E, 9F, and S12). On CMA and V8 pH 5 media, progeny MP16, MP18, and MP54 with fully compatible MAT alleles produced hyphae and basidia similar to the parental cross. Some hyphal growth was also observed in solo culture of strains MP49, MP51, and MP59, exclusively on V8 pH 5, but the filaments were shorter and had no structures resembling basidia (S12 Fig). This morphology resembled that of previously described self-filamentous K. mangrovensis strains, which form irregular, highly branched hyphae without clamp cells [66]. Taken together, these analyses demonstrate that sexual reproduction in K. mangrovensis involves recombination and can generate nonhaploid progeny with potential for self-fertility.

Multiple independent P/R locus expansions in Cryptococcus and Kwoniella

The archetypal basidiomycete MAT loci are thought to consist of a compact P/R region with one pheromone receptor and one pheromone gene, and an unlinked HD region containing two divergently transcribed homeodomain transcription factors (HD1 and HD2). This configuration is retained in some lineages, such as in Malassezia species [20], but in most basidiomycetes the P/R locus has undergone expansions and structural changes, while the HD locus remains comparatively conserved (except in some rust fungi [109] and in mushroom-forming species with HD gene duplications [9,28]). For example, in Ustilago maydis, allele-specific differences in P/R locus size arise from two additional genes (lga2 and rga2) in the a2 allele that direct mitochondrial uniparental DNA inheritance [110,111]. Similarly, several Microbotryomycetes yeasts (Pucciniomycotina) harbor expanded P/R loci [26,112,113], as do tetrapolar Cryptococcus species [35,47,78]. Recent fully phased, chromosome-scale assemblies in rust fungi also uncovered multiple pheromone genes and STE3 receptor duplications on single haplotypes [109,114], though the full structure of the P/R locus in these fungi remains incompletely resolved.

Our analyses across tetrapolar Cryptococcus and Kwoniella support independent, lineage-specific P/R locus expansions. The inferred ancestral P/R locus likely comprised a conserved four-gene core (MF, STE3, STE12, and STE20) (Fig 10, event 1). In Kwoniella, this region has remained relatively compact, with 9 core conserved genes (Fig 10, event 2), though lineage-specific modifications were detected, including mating-type-specific changes (e.g., STE12 truncation in a2 alleles) and small-scale expansions through the inclusion of genes likely entrapped by local inversions. We note that these additions more likely reflect neutral consequences of inversions expanding recombination-suppressed regions, rather than adaptive recruitment, consistent with stepwise extension models of sex chromosome and MAT locus evolution [39,115]. In K. ovata, K. endophytica, and K. dendrophila, mosaic P/R configurations are better explained by past intra-locus recombination. This may have transiently enabled self-compatibility at the P/R locus, but the absence of self-filamentation in extant strains suggests a reversion to heterothallism. A full transition to homothallism would also require compatibility at the HD locus, either through mutations enabling pairing between otherwise nonself-compatible HD1/HD2 alleles or through bypass of HD function, as seen in C. depauperatus [23].

In contrast to Kwoniella, tetrapolar Cryptococcus species from clades B and D exhibit pronounced P/R locus expansion, but with limited gene-content overlap, indicating that these expansions occurred independently after divergence from a common ancestor (Fig 10, event 5). A distinctive feature of the Cryptococcus P/R locus is the duplication of pheromone (MF) genes, with each mating type typically carrying multiple copies arranged in divergent orientations. Similar patterns occur in Microbotryomycetes yeasts [26,112,113,116,117], and have also been reported in rust fungi [109,114]. However, in rusts, some of the duplicated MF genes encode distinct pheromones, possibly reflecting adaptation to different receptor specificities, whereas in Cryptococcus and Microbotryomycetes the MF copies encode identical mature peptides, suggesting functional redundancy rather than diversification.

Experimental evidence in C. deneoformans supports a dosage-based model, in which all three MFα gene copies are expressed and contribute additively to mating efficiency and filamentation [118]. Triple mutants show severe mating defects, while overexpression enhances differentiation, indicating that pheromone dosage promotes mating responsiveness [118]. Although direct functional evidence in other Cryptococcus species is lacking, these duplications may similarly enhance partner recognition. This interpretation aligns with sexual selection frameworks in fungi, where pheromone production can act as a costly, honest signal of mating quality (the handicap principle), which posits that only high-quality individuals can afford the cost of producing strong signals and are therefore more likely to be chosen as mates [119,120]. Studies in model yeasts have shown preferential mating toward high pheromone producers [121,122], but equivalent tests in basidiomycetes are limited, and the ecological context for such selection in Cryptococcus remains unclear. Why pheromone duplications recur in Cryptococcus but not in Kwoniella remains uncertain. Their absence in Kwoniella may reflect lineage-specific constraints or ecological differences—for instance, Kwoniella species produce embedded rather than aerial spores, which could reduce the need for long-range mate signaling—although the lack of ecological and population-level data prevents firm conclusions.

Notably, MF gene duplications can also introduce structural instability by facilitating the formation of inverted repeats, which can promote inversion loops and allele-specific structural rearrangements, potentially expanding recombination-suppressed regions between P/R alleles. One way to mitigate this instability is by clustering MF gene pairs more closely, thereby restricting the extent of rearrangements. Consistent with this, we observed reduced structural variation in C. neoformans MATα strains and in the C. gattii species complex, where the two MF gene pairs are positioned in close proximity, separated by only two genes. In contrast, MATa strains exhibit a more dispersed arrangement of MF genes, which appears to contribute to greater structural diversity across strains, including the C. neoformans VNI-specific rearrangements that arose following introgression of the MATa allele from the VNBI lineage. Once established, multiple MF gene copies may be maintained as templates for intrallelic repair via gene conversion, a nonreciprocal exchange that can occur during meiosis between homologous sequences but also at lower frequencies between sister chromatids, or within duplicated regions on a single chromosome during mitosis [123,124]. If confirmed, this mechanism would resemble the role of palindromic sequences in preserving gene integrity on the human Y chromosome [125–127]. Introgression events, such as the transfer of MATa alleles between C. neoformans lineages, may further buffer against degeneration by resetting mutational load [128], thereby helping maintain complex MAT configurations.

Integrated with these structural features, our analyses revealed that the P/R locus in clades B and D Cryptococcus species has consistently lower GC% relative to genome-wide averages and that this depletion is best explained by reduced gBGC in recombination-suppressed regions rather than by relaxed selection on codon usage. gBGC normally favors the fixation of G/C alleles in regions of frequent recombination, and its reduction allows the AT-biased mutational spectrum to dominate over time, driving local base composition shifts. Similar GC depletion has been reported in the Trichosporonales MAT locus [18]. Interestingly, this trend is not observed in the MAT locus of pathogenic Cryptococcus, suggesting that genomic GC background also influences the outcome. In high-GC genomes (e.g., clades B and D), recombination suppression accentuates AT accumulation, whereas in lower-GC genomes (pathogenic Cryptococcus; [49]), the absence of a shift suggests they may already be near equilibrium under mutation pressure, masking localized effects.

Independent transitions to bipolar and pseudobipolar via chromosomal translocations and inversions

Beyond the previously documented tetrapolar-to-bipolar transition in pathogenic Cryptococcus lineages (clade A), we identified three additional transitions: two leading to bipolarity (Cryptococcus sp. 3 and K. fici; Fig 10, events 6 and 4), and one resulting in a pseudobipolar arrangement (K. europaea; Fig 10, event 3). Our findings suggest that the transition to bipolarity in Cryptococcus sp. 3 likely occurred independently of that in pathogenic species. First, fusion of the P/R and HD loci in Cryptococcus sp. 3 was unrelated to the reduction from 14 to 13 chromosomes and instead originated from a translocation followed by an inversion. Second, the MAT locus in this species retains only SXI2 (HD2) linked to the STE3α, unlike in pathogenic species where SXI2 is associated with STE3a. Third, phylogenetic analyses revealed that many genes within the Cryptococcus sp. 3 MAT locus do not group with a or α allele-specific clusters from pathogenic species. If the fusion had predated the divergence of these clades, trans-specific polymorphism would be expected, but this pattern was not observed, which is more consistent with an independent origin.

However, whether the initial translocation event that brought the P/R and HD loci onto the same chromosome occurred independently in Cryptococcus sp. 3 or in a common ancestor of clades A and C, remains unresolved. This uncertainty stems from three main factors. First, phylogenetic ties among clades A (bipolar), B (tetrapolar), and C (which includes Cryptococcus sp. 3 and C. depauperatus) are still ambiguous. Second, although our synteny analyses indicate that the translocation breakpoints at the origin of the physical linkage between MAT loci differ between Cryptococcus sp. 3 and pathogenic species (supporting independent events), these differences might also reflect subsequent rearrangements that accumulated along the two lineages. Third, we have previously reported that C. depauperatus also shows a partial merging of HD- and P/R-derived genes into the same genomic region, but with more extensive gene reshuffling and HD gene loss likely associated with a transition to homothallism [23]. Further taxon sampling, combined with phylogenomic approaches tailored to deep divergences, will be required to conclusively resolve these relationships and clarify the broader implications of these transitions.

The tetrapolar-to-bipolar transition in Cryptococcus pathogens has been hypothesized to result from a chromosomal translocation mediated by ectopic recombination between repetitive elements in the centromeric regions of the P/R- and HD-containing chromosomes [35]. This model had support from several observations: (i) Cryptococcus centromeres are enriched with long-terminal-repeat retrotransposons (LTRs) from the Ty3 and Ty1 superfamilies, which are shared among different centromeres and may facilitate recombination [23,47,49,129,130]; (ii) chromosomal arm exchanges occurred during the divergence of Cryptococcus lineages from their common ancestor [35,108]; and (iii) the repair of experimentally induced double-strand breaks at centromere-specific transposons in C. neoformans frequently results in chromosomal translocations [108]. However, our broader comparisons indicate that the centromere of the C. neoformans MAT chromosome, along with extended flanking regions, is conserved across most Cryptococcus species and in K. shandongensis, suggesting an ancestral organization.

Indeed, our analyses suggest that translocations outside centromeric regions may have played a more prominent role in repositioning the P/R and HD loci onto the same chromosome (Fig 7). Such noncentromeric translocations also appear to underlie the transition to bipolarity in several Ustilaginales species, including Sporisorium scitamineum [30] and in the common ancestor of Ustilago hordei, U. nuda, and U. bromivora [31]. Noncentromeric translocation events are also common across other genomic regions when comparing genomes of different Cryptococcus species [49] and even among conspecific strains of C. neoformans [101,131]. Whether such translocations are influenced by TEs has not yet been systematically analyzed, but studies in Saccharomyces cerevisiae have shown they can occur via homologous recombination between retrotransposons on different chromosomes [132]. Given the abundance of TEs in Cryptococcus genomes [49], they may likewise have facilitated the chromosomal relocation of MAT loci and the subsequent transition to bipolarity. Our findings offer an alternative to models invoking intercentromeric recombination as the primary event initiating the tetrapolar-to-bipolar transition in Cryptococcus pathogens.

Genetically linked P/R-HD loci can also arise from the fusion of entire ancestral P/R and HD chromosomes, as reported in Microbotryum violaceum on Silene paradoxa [27]. In Kwoniella, repeated and independent chromosome fusions have drastically reduced chromosome numbers, with some extant species retaining only 3 (clade E) or 5 (clades G and H) chromosomes from a 14-chromosome ancestral karyotype [49]. Yet, aside from the bipolar arrangement in K. fici (clade H) and the pseudobipolar arrangement in K. europaea (clade E), no other instances of P/R-HD linkage or colocation were detected in Kwoniella, and even in these species the transition reflects chromosomal translocations rather than whole-chromosome fusions.

In K. fici, the transition to fused MAT loci did not result in loss of SXI1 (HD1) or SXI2 (HD2), possibly suggesting a relatively recent fusion where insufficient time has passed for one HD gene to decay in each mating type, as observed in Cryptococcus sp. 3, Cryptococcus pathogens [133], and Trichosporonales [18]. Other known instances of HD gene loss in bipolar species, albeit distinct, include: (i) Microbotryum violaceum on Silene caroliniana, where HD1 is absent in a2 strains, while a1 retains both HD genes [27]; (ii) Microbotryum superbum, where loss of HD function in mating compatibility may have relaxed constraints on HD2 maintenance in a1 strains, leading to its disruption while a2 strains retain both HD genes [117]; and (iii) xerotolerant Wallemia species, where HD1 is retained in MAT a1, while the opposite mating type appears to lack any HD gene [134]. Because HD compatibility is no longer an independent determinant of mating success in bipolar species, maintaining both HD genes in each MAT allele may become dispensable, with loss potentially reflecting genomic streamlining once MAT loci are fused and recombination suppressed. Yet, as in K. fici, the retention of both HD genes in each allele remains the predominant organization in other bipolar basidiomycetes with fused MAT loci, including several Microbotryum [27], Sporisorium [30], and Ustilago species [31,135,136], possibly reflecting differences in mating strategies or genomic constraints. For instance, if both HD genes have additional roles beyond mating-type determination that do not require heterodimerization, their retention could be selectively maintained, although this possibility remains largely unexplored in basidiomycetes.

In contrast to K. fici, where the MAT loci are fused, K. europaea has the two loci ~8.5 Mb apart on the same chromosome. This distance makes tight genetic linkage unlikely, though extended recombination suppression over megabase-long spans has been reported in Neurospora tetrasperma and Microbotryum [39,137]. Because crosses in K. europaea have not yielded sexual reproduction [66], it remains unclear whether this configuration represents an incipient step toward bipolarity or a neutral byproduct of chromosomal rearrangements. If recombination persists, this system would functionally resemble tetrapolarity, as the two loci would still reassort in most meioses, making MAT loci colocation unlikely to meaningfully increase sibling compatibility. In this case, the arrangement may simply reflect a stochastic translocation rather than a selectively maintained feature.

Together, these observations reinforce the view that chromosomal rearrangements provide the mechanistic basis for tetrapolar-to-bipolar transitions but do not inherently drive shifts in breeding systems. Instead, whether a lineage remains tetrapolar or transitions to bipolarity likely depends on ecological and evolutionary pressures such as opportunities for outcrossing versus inbreeding, the frequency of sibling or clonal mating after colonization, dispersal limitations, and the long-term stability of chromosomal rearrangements that maintain linkage between the loci. Accordingly, species with linked MAT loci are often associated with animals or plants as commensals or pathogens, where linkage facilitates mating among siblings and provides an advantage during host colonization [9,28,81,138,139]. A similar scenario has been proposed for the Trichosporonales, where MAT locus fusion may have originated in a host-associated ancestor and persisted in saprobic descendants [18]. Our results, however, show that bipolarity is not restricted to pathogens: Cryptococcus sp. 3 (isolated from bark beetles) and Kwoniella fici (isolated from fig) harbor fused MAT loci with no evidence of pathogenicity to date. This suggests that restricted partner availability—whether in insect-associated microhabitats or free-living clonal populations—may favor MAT loci fusion independently of pathogenicity. Once established, bipolar systems may be difficult to revert, and their stability could in turn facilitate the persistence or emergence of pathogenic lifestyles by ensuring mating success under the constrained conditions typical of infection.

Outlook

Our characterization of MAT loci gene content and organization allowed us to infer putative breeding systems (heterothallism versus homothallism) across species. However, determining how often sexual reproduction and recombination occur in nature, and whether mating involves primarily outcrossing or inbreeding, will require population-level analyses such as linkage disequilibrium decay [140]. Unlike basidiomycete macrofungi, where sexual structures can be directly observed in the field, or smut fungi, where disease symptoms visibly reflect the production of sexual spores, most Tremellomycetes are inconspicuous environmental isolates or parasites of lichens and other fungi, complicating the isolation of compatible mating partners. Their cryptic nature has led to many species being described from a single isolate [48], hindering mating assays. Progress will require expanded taxon and population-level sampling [48,50], particularly for nonpathogenic species, to clarify reproductive strategies, transitions in mating systems, and potentially capture changes in MAT locus organization as they occur. Uncovering the hidden diversity of Cryptococcus species—particularly in their likely African origin [50]—will also be crucial for linking ecology, speciation, and adaptation with the evolutionary forces shaping MAT locus evolution and reproductive transitions, with possible implications for pathogenic emergence.

Strains and media

Strains were routinely grown on YPD medium (10 g/L yeast extract, 20 g/L Bacto Peptone, 20 g/L dextrose, and 20 g/L agar) unless indicated otherwise. C. neoformans and C. deneoformans strains were incubated at 30 °C, while other Cryptococcus and Kwoniella species strains were grown at room temperature (20–23 °C). A complete list of strains is provided in S1 Appendix. Strain NCYC1536 was obtained from the National Collection of Yeast Cultures (Norwich, UK), and strain CMW60451 was isolated in October 2022 from a bark beetle (Lanurgus sp.) infesting twigs of Widdringtonia cedarbergensis in the Cederberg Mountains, South Africa.

Genomic DNA extraction

High-molecular-weight DNA was extracted with a cetyltrimethylammonium bromide extraction as previously described [108], minimizing DNA shearing during sample preparation. DNA quality was evaluated by determining the A260/A280 and A260/A230 ratios on NanoDrop spectrophotometer (Thermo). Integrity and fragment size were analyzed using clamped homogeneous electric field electrophoresis, and gDNA for short-read whole-genome sequencing (Illumina) was extracted with a phenol:chloroform-based protocol, both as previously described [49]. DNA concentration was measured with Qubit dsDNA Assay Kits (Invitrogen) on the Qubit fluorometer.

Genome sequencing

Whole-genome sequencing was performed with PacBio, Nanopore, and Illumina technologies. For K. heveanensis BCC8398 and K. mangrovensis CBS10435, Illumina sequencing at the Broad Institute Genomics Platform utilized “fragment” and “jumping” libraries, prepared and sequenced as previously described [49]. All other Illumina sequencing was conducted at the Duke University Sequencing and Genomic Technologies (DUSGT) Core, with libraries prepared with Kapa HyperPlus kit and sequenced as paired-end 2 × 150 bp reads on various Illumina platforms. For PacBio sequencing, 15–20 kb insert-size libraries were prepared and run on a PacBio RS II or Sequel (2.0 chemistry) system at the DUSGT. Nanopore sequencing was carried out in-house. Single-strain libraries were prepared with either the SQK-LSK108 or SQK-LSK110 kit, whereas up to three DNA samples were barcoded with the SQK-LSK109 and EXP-NBD103/EXP-NBD104, or SQK-NBD114.24 kits. Libraries were prepared following the manufacturer’s protocols, either individually or pooled, and sequenced on R9 (FLO-MN106) or R10 (FLO-MIN114) flow cells for 48–72 h at default voltage on a MinION Mk1B or Mk1C system. The MinION software version current at the time of each run was applied. Further details on sequencing platforms, basecalling, and demultiplexing are provided in S1 Appendix.

Genome assembly

Genomes were assembled with Canu [141] using Nanopore or PacBio sequencing data and default parameters. Assembly accuracy was improved through initial error correction with Medaka (https://github.com/nanoporetech/medaka) for Nanopore-based assemblies, followed by up to five rounds of iterative polishing with Pilon v1.22 [142] (--fix all) using Illumina reads aligned to the first pass-polished assembly with BWA-MEM v0.7.17-r1188 [143]. Pilon iterations were stopped early if no additional changes were detected, based on the absence of reported corrections in each round. Contigs containing only rDNA sequences, detected by Barrnap (https://github.com/tseemann/barrnap) (--kingdom euk), or those classified as mitochondrial DNA, were excluded from the final nuclear genome assembly. To confirm assembly completeness and evaluate telomeric regions, Nanopore/PacBio and Illumina reads were realigned to the Canu-corrected assembly using minimap2 v2.9-r720 [144] and BWA-MEM, respectively, and read coverage profiles were examined in the Integrative Genomics Viewer [145]. Draft assemblies for K. europaea PYCC6162 and K. botswanensis CBS12717 were generated with SPAdes v3.15.3 using default settings, while assemblies for K. heveanensis BCC8398 and K. mangrovensis CBS10435 were constructed with Allpaths [146]. Genome assemblies and raw sequencing data were deposited in DDBJ/EMBL/GenBank under the BioProject numbers listed in S1 Appendix, which also provides details on assembly parameters for each genome.

Gene prediction and annotation

Gene models were predicted using either BRAKER2 v2.1.5 [147] or Funannotate v1.8.9 (https://github.com/nextgenusfs/funannotate), as previously described [20,23,49]. Prior to gene prediction, assemblies were masked with RepeatMasker v4.1.4 in sensitive mode, applying a custom library that combined Cryptococcus–Kwoniella repeat families generated with RepeatModeler2 v2.0.4 [49], centromere-associated TEs, and curated Cryptococcus elements compiled from published sources [148] and represented in Dfam. Manual inspection and correction were performed only for mating-type gene models, guided by ortholog alignments, and included adjustments to exon–intron boundaries or addition of small ORFs missed by automated pipelines (e.g., MF pheromone genes). These curated annotations were incorporated into the final GFF files and are part of the genome submissions deposited in NCBI. Functional annotation was added to the final gene models using the Funannotate “annotate” module, which integrated data for PFAM and InterPro domains, Gene Ontology terms, fungal transcription factors, Cluster of Orthologous Genes (COGs), secondary metabolites, Carbohydrate-Active Enzymes (CAZYmes), secreted proteins, proteases (MEROPS), and Benchmarking Universal Single-Copy Orthologs (BUSCO) groups. InterPro domain data, COG annotations, and secondary metabolite predictions were generated with InterProScan v5.55-88.0, eggNOGmapper v.2.1.7 (eggNOG DB version: 5.0.2) [149], and AntiSMASH v6.1.0 [150], and then passed to Funannotate annotate with the options --iprscan, --eggnog, and --antismash. Specific parameters are provided in S1 Appendix.

Ortholog identification and sequence alignment

To construct the phylogenomic data matrix, single-copy orthologs (SC-OGs) were identified across Cryptococcus and Kwoniella species and three outgroups (Tremella mesenterica ATCC28783, GCA_004117975.1; Saitozyma podzolica DSM27192, GCA_003942215.1; and Bullera alba JCM2954, GCA_001600095.1) using OrthoFinder v3.0.1b1 with options: -M msa -S diamond_ultra_sens -I 1.5 -M msa -A mafft -T fasttree -t 48 -a 6. This identified 3,086 SC-OGs shared across species. Their amino acid sequences were individually aligned with MAFFT v7.310 [151] (--localpair --maxiterate 1000) and trimmed with TrimAl v1.4.rev22 [152] (-gappyout -keepheader). The same approach was applied to construct the phylogenomic data matrix for C. neoformans strains representing different VN groups, with three C. deneoformans outgroups, yielding 5,439 shared SC-OGs, which were then aligned and trimmed as above.

Species phylogeny, estimation of topological support, and gene genealogies

The updated phylogeny of Cryptococcus and Kwoniella species (Figs 1A and S1) was inferred using two complementary strategies: (i) concatenation-based partitioned ML phylogeny reconstruction in IQ-TREE v2.1.6 [153], and (ii) gene-based coalescence analysis in ASTRAL v5.7.8 [154]. For the concatenation approach, amino acid alignments of 3,086 SC-OGs were combined into a partitioned supermatrix (52 taxa, 3,086 partitions, 1,686,691 sites) using the “-p” option in IQ-TREE. The edge-linked proportional partition model [155] was applied to account for differences in evolutionary rates across partitions, and the best-fit substitution model for each partition was identified with ModelFinder [156] under the Bayesian information criterion. The ML tree was inferred with the parameters --seed 54321 -m MFP -msub nuclear -B 1000 -alrt 1000 -T 14, incorporating 1,000 ultrafast bootstrap (UFBoot) replicates [157] and Shimodaira–Hasegawa approximate likelihood ratio tests (SH-aLRT) for branch support assessment. For the coalescence-based approach, ML gene trees were constructed for each SC-OG alignment using IQ-TREE’s “-S” option, which performs both model selection and tree inference for individual alignments. These trees were then input to ASTRAL under default settings to reconstruct the species phylogeny. Quartet support values were computed with the -t 2 option, providing quartet support metrics for the primary topology (q1) and alternative topologies (q2, q3), and local posterior probability (LPP) support. Genealogical concordance and topological support for branches in the concatenated ML tree were evaluated with gene concordance factor (gCF) and site concordance factor (sCF) metrics in IQ-TREE. This analysis utilized both the best-scoring concatenated ML tree (concat.treefile) and the set of gene trees (loci.treefile), with options -t --gcf -p --scf 100. The concatenation-based approach was also applied to resolve relationships among C. neoformans VN lineages (Fig 4A). For gene genealogies, amino acid sequences of individual genes of interest were extracted from OGs identified by OrthoFinder, manually inspected, and reannotated as needed. Alignment, trimming, and ML reconstruction were performed as described above. Trees were visualized and annotated with iTOL v7. Detailed model parameters are provided in the tree files accessible at https://doi.org/10.5281/zenodo.16987438.

Synteny analyses, MAT loci delineation, and centromere identification

Conserved synteny blocks in pairwise genome comparisons were identified with SynChro [105], using a delta parameter of three for high stringency. Comparisons in Figs 6A, 8A and S9A employed the K. shandongensis genome as reference. Detailed linear synteny plots of chromosomes and specific genomic regions, including MAT loci, were generated with EasyFig v2.2.2 [158] using BLASTN. Centromeres positions were determined in silico as previously described [49], combining detection of centromere-associated LTR elements and synteny analysis. Centromere lengths were defined as the intergenic regions between flanking centromeric genes (S1 Appendix). MAT loci were initially identified by BLAST searches, with C. neoformans-derived MAT genes and their flanking proteins as queries. The HD loci were defined as the regions spanning the HD1 and HD2 genes. The P/R loci were delineated based on structural comparisons between mating types. For species with available strains of opposite mating types, the P/R locus was defined as the region where synteny between the mating types is disrupted, with boundaries set at the points where synteny is restored. For species with only a single strain, the P/R locus length was inferred from the distance between P/R-flanking genes, as determined from comparisons involving opposite mating types.

To infer the origin of structural rearrangements observed in the P/R locus in Kwoniella (S5A Fig), we manually inspected the orientation and synteny of P/R-associated genes across mating types and species. Rearrangement breakpoints were approximated by identifying changes in gene order and orientation between alleles, supported by the presence of gene fragments (e.g., partial NDC80, STE12, BSP3) and conserved synteny in the opposite mating type. These rearrangements were reconstructed by inverting or repositioning affected gene blocks in the more derived a2 allele to match the putative ancestral configuration in the more conserved a1 allele.

Statistical analyses of MAT loci lengths across species, clades, or mating types were conducted with Python3 with Pandas, Seaborn, Matplotlib, and SciPy libraries. Differences between the two groups were assessed using the two-sided Mann–Whitney U test. Scripts and raw data are available at https://doi.org/10.5281/zenodo.16987438. Plots were refined for clarity by adjusting color schemes, labels, and adding features in Adobe Illustrator.

Analysis of gene content in MAT loci, frequency across species, and gene essentiality classification

A curated gene presence/absence matrix was constructed after delineating MAT loci across species. Genes were categorized as “present” (1), “absent” (0), “unclear” (?), or “pseudogene” (pseudo). Unclear cases indicated instances where gene presence could not be conclusively determined due to uncertain locus boundaries. For frequency calculations, pseudogenes were treated as present and unclear cases as absent. The presence of either SXI1 or SXI2 was scored as present and combined for frequency estimates. Information on gene essentiality was inferred from C. neoformans H99, based on experimentally validated studies [34,159,160] or predictions from a high-throughput transposon mutagenesis (TN-seq) study [92]. Gene essentiality in other species was extrapolated solely from H99, and genes without H99 orthologs were categorized as “unknown” due to the lack of supporting data. A custom R script (0_calculate_gene_presence_frequency.R) analyzed and plotted gene presence frequencies. The matrix was also visualized as a heatmap (1_plot_gene_matrix_based_on_frequency.R), with genes ordered by frequency. SXI1 and SXI2 are shown as separate columns in the heatmap, but their ranking reflects their combined frequency. Analyses and visualizations were conducted in R with dplyr, ggplot2, ComplexHeatmap, and circlize packages. Gene presence/absence was further analyzed with a Python script (2_compare_MAT_gene_content_across_clades.py), which calculated clade-specific and pairwise presence, identified genes shared across specified clades, and highlighted clade-specific genes. Input matrices are provided in S2 and S3 Appendices and scripts are available at https://doi.org/10.5281/zenodo.16987438.

Base composition and codon-bias analyses

Genome-wide GC content and deviations from the mean GC% were analyzed with a custom Python script (00_gc_content_analysis_and_plots.py), employing a nonoverlapping sliding window (1 kb for genome scans; 0.25 kb for zoomed-in views). Regions of interest (e.g., P/R loci) were highlighted from BED file coordinates. Plots were further refined in Adobe Illustrator for producing Fig 2. To assess whether the P/R locus exhibited lower GC content than genome-wide averages, a second script (01_gc_content_analysis.py) compared P/R locus GC% to 1,000 randomly sampled genome-wide regions of equal size (excluding the P/R locus). A one-sample t test assessed whether the P/R locus GC content differed from the sampled mean, while a permutation test (10,000 iterations) generated a null distribution to calculate P-values and Z-scores. The analysis was automated with a shell script (01a_run_script_vs_wg.sh), consolidating results into a summary table (S4 Appendix).

To assess whether GC depletion within the P/R locus reflects reduced gBGC or relaxed selection on codon usage, we performed a four-step Python3-based analysis (NumPy, pandas, SciPy, Biopython): (i) neutral noncoding GC comparison, (ii) coding third-position composition (AT3 versus GC3), (iii) codon-bias indices (Fop, Nc, ENC, ΔENC), and (iv) CAI analyses controlling for GC3 and CDS length (details in S1 Text). As preparation for CAI-related analyses, species-specific CAI reference sets were generated with a custom script (02_build_cai_reference.py) in “rp_only” mode, selecting ribosomal-protein genes as proxies for highly expressed, codon-optimized, genes [161,162], since RNA-seq data for these species was unavailable. The script counted codons and generated a CAI weight table in which, within each synonymous family, relative adaptiveness weights (w) were normalized so that the most frequent codon had w = 1 and others 0 < w ≤ 1. Per-gene CAI values were then computed with 03_compute_cai_per_gene.py, which reads each species’ CAI weight table and CDS annotations and returns gene-level CAI (geometric mean of w across codons; start/stop excluded; nonACGT ignored; Met/Trp treated as neutral set to w = 1).

To test for GC depletion in the P/R region, noncoding GC within P/R (intergenic plus introns) was compared to same-chromosome, length-matched nulls with script 04_neutral_non_coding_GC.py using 10,000 P/R-excluded windows (--n_null 10000 --no_overlap --min_gap 1000). The script reports ΔGC (gc_intergenic – null_mean_intergenic), a standardized z, and a one-sided empirical permutation p-value (PR ≤ null).

Coding third-position composition (AT3 versus GC3) was analyzed with script 05_coding_composition_and_codon_bias.py by counting AT- versus GC-ending third positions inside P/R and background regions (same chromosome, outside P/R) and testing the 2 × 2 table by χ² with Fisher’s exact fallback. Using the same script and the CAI weights above, codon-bias indices were calculate: Fop (frequency of optimal codons based on the CAI reference); Nc (effective number of codons, representing the observed codon usage bias in a gene); ENC (the expected Nc based solely on the gene’s GC3, calculated from the GC3–Nc relationship described by Wright (1990) [163]); and ΔENC = Nc − ENC (representing codon bias beyond that expected from GC3 composition alone). To improve the stability of Nc estimates, particularly for short CDS, we applied a small pseudocount (0.5 per codon), excluded very small synonymous families, bounded intermediate calculations to avoid division by near-zero values, and constrained Nc values to the theoretical range of 20 (maximum bias) to 61 (no bias) before computing ΔENC. P/R versus background contrasts were evaluated as median differences using Mann–Whitney tests with 10,000 permutations.

Finally, CAI differences independent of nucleotide composition were tested using three complementary analyses implemented in script 06_cai_vs_gc3_controls.py (detailed in S1 Text), using the following parameters: --n_perm 10000, --gc3_tol 0.03, --len_tol 0.10, and --k_match 5. Briefly, we (i) fit CAI ~ inside_pr +  Δ(exon-intron GC) + log(length_cds) to all genes using ordinary-least-squares (OLS) with heteroskedasticity-consistent (HC1) robust standard errors; inside_pr = 1 for CDS fully within the P/R window and 0 for same-chromosome genes outside (boundary overlaps excluded); (ii) we trained CAI ~ Δ(exon-intron GC) + log(length_cds) on background genes only, computed residuals for all genes, and tested median(PR residual) − median(BG residual) with a one-sided permutation test (10,000 shuffles; alternative PR < BG); and (iii) we matched each P/R locus gene to same-chromosome controls within ±0.03 GC3 and ±10% CDS length (dropping unmatched P/R genes), computed CAI_match_median = median[CAI_PR − mean(CAI_BG matches)], and used a sign-flip permutation test (10,000 flips; alternative PR < BG). Fop was analyzed identically to step (iii) with the same matching tolerances and permutation settings. All scripts are available at https://doi.org/10.5281/zenodo.16987438, and per-species summaries and statistical tests are given in S4 Appendix.

Variant analysis and SNP distribution in C. neoformans MATa strains

Multi-genome variant analysis was performed with the Snippy pipeline v4.6.0 (https://github.com/tseemann/snippy) using the newly assembled genome of C. neoformans Ftc555-1 (VNBI) as the reference, with parameters: --cpus 10, --unmapped, --mincov 10, and --minfrac 0.9. Paired-end reads were obtained from the NCBI SRA database (S1 Appendix). Three datasets were independently used: (i) the entire genome, (ii) only the MAT locus region (chr_5:153,773-277,230), and (iii) only chr. 5. The MAT locus and chr. 5 were extracted from the reference genome to generate region-specific GenBank files for Snippy to identify SNPs and construct region-specific phylogenies. Visualization of SNV distributions was based on the whole-genome dataset and included both chromosome-wide and zoomed-in views of specific regions. A custom Python script processed merged SNV datasets, computed total SNP counts for each strain, and generated gradient color-coded SNP density plots. Final composite figures were refined using Adobe Illustrator. Core SNP alignments (core.aln) from each dataset were used for phylogenetic reconstruction with IQ-TREE2, incorporating 10,000 UFBoot replicates and SH-aLRT tests for branch support. Scripts for SNV analysis and visualization are available at https://doi.org/10.5281/zenodo.16987438.

Chromosome composition analysis of K. mangrovensis progeny

To characterize chromosome composition in K. mangrovensis progeny, a combined nuclear reference genome was constructed from assemblies of the two parental strains, CBS8507 and CBS10435. For CBS8507, the original assembly consisted of four contigs, two of which represented the same chromosome fragmented at the rDNA array. To streamline downstream analyses and facilitate interpretation, the two contigs were merged into a single scaffold: orientation was first established from rDNA gene direction; the corresponding ends were aligned and trimmed; and a 5-N gap was inserted to join them, yielding a finalized three-chromosome assembly. For CBS10435, the draft assembly of 37 contigs was reordered and reoriented to match the CBS8507 3-chromosome assembly using D-GENIES [164], which employs minimap2 for genome alignment. Raw Illumina paired-end reads from selected progeny were then mapped to a combined reference with the sppIDer pipeline [165], which sequentially mapped reads, applied stringent quality filtering (MQ > 3), and generated depth-of-coverage plots. Chromosome number and ploidy were estimated by integrating spider coverage data with flow cytometry results.

Mating assays, phenotyping on mating-inducing media, and microscopy

Mating assays with K. mangrovensis strains CBS8507 and CBS10435 were performed as previously described [66,67]. Equal amounts of cells from each strain were mixed on V8 agar (10 g/L yeast extract, 20 g/L Bacto Peptone, 20 g/L dextrose, 20 g/L agar; pH 5) or corn meal agar (CMA; 15 g/L corn meal agar, 5 g/L agar) and incubated in the dark at room temperature (20–23 °C) for up to 1 month. Plates were regularly monitored for the development of mating structures. Random F1 progeny were recovered by microdissecting individual cells from cell clusters [79] embedded in or on the surface of the mating media (V8 and CMA) at the edges of mating patches. A total of 51 progeny were collected, grown axenically on YPD plates, and stored as glycerol stocks at −80 °C. Selected F1 progeny (MP11, MP15, MP16, MP18, MP49, MP50, MP51, MP54, and MP59) were assessed for hyphal formation and mating structures on V8 pH5 and CMA, compared to the parental cross (CBS8507 × CBS10432) and the solo culture of each parental strain. Potentially compatible strain pairs from Cryptococcus species lacking documented sexual reproduction (IND107 × CBS11718; CBS11687 × 7685027; and DSM108351 × NCYC1536) were tested for mating ability under conditions known to induce mating in C. neoformans and C. deneoformans (V8 media, dark, room temperature) for up to a month, with regular monitoring [79]. Solo cultures of each strain were also evaluated for their ability to reproduce sexually without a compatible partner. Similar assays were performed for strains of newly described species lacking known compatible partners. All mating assays were repeated three separate times to ensure consistent phenotypic observations. To assess hyphal growth, basidia, and spores, the edges of yeast colonies and mating patches were examined and photographed with a Zeiss Axio Scope.A1 microscope equipped with an Axiocam Color camera and ZEN Lite v3.4 software.

Analysis of K. mangrovensis progeny by PCR-restriction fragment length polymorphism (RFLP) analysis

Kwoniella mangrovensis has only three chromosomes. To analyze the F1 progeny from the CBS8507 × CBS10432 crosses, six primer pairs (one per chromosomal arm) were designed for the initial screening. Primers were manually designed to produce distinct parental patterns after PCR product digestion with specific restriction enzymes. As a heterozygosity control, PCR products from a 1:1 DNA mixture of both parental strains were amplified for each primer pair, followed by digestion with the corresponding restriction enzyme. Results and primer sequences are provided in S5 Appendix.

Fluorescence-activated cell-sorting (FACS)

FACS was performed as previously described [166] to determine ploidy of K. mangrovensis CBS8507, CBS10432, and nine selected progeny. Strains were grown overnight at room temperature (21–23 °C) on YPD medium, harvested, and washed with PBS. Cells were fixed in 2 ml of 70% ethanol at 4 °C overnight, washed with 1 ml of NS buffer (10 mM Tris-HCl pH 7.2, 250 mM sucrose, 1 mM EDTA pH 8.0, 1 mM MgCl₂, 0.1mM CaCl₂, 0.1 mM ZnCl₂, 0.4 mM phenylmethylsulfonyl fluoride, and 7 mM β-mercaptoethanol), and then stained with 5 μl of propidium iodide (0.5 mg/ml) in 180 μl NS buffer with 20 μl of RNase (10 mg/ml) at 4 °C overnight. Lastly, 50 μl of stained cells were diluted in 2 ml of 50 mM Tris-HCl pH 8.0 and sonicated for 1 min before analysis at the Duke Cancer Institute Flow Cytometry Shared Resource. Data were collected from 10,000 cells using the FL1 channel on a Becton-Dickinson FACScan and analyzed with FlowJo software. Strains JEC21 and XL143 served as haploid and diploid controls, respectively [167].