Previous reports that are based on pulsed-field gel electrophoresis (PFGE) have suggested that chromosome number varies within the Malassezia species complex. The early diverged species M. slooffiae of Clade C was reported to have nine chromosomes (Boekhout et al., 1998). Clade B Malassezia species are reported to have nine (M. globosa and M. restricta), eight (M. sympodialis), or six chromosomes (M. pachydermatis). Among the Clade A species, M. obtusa and M. furfur CBS14141 were both reported to have seven chromosomes (Boekhout and Bosboom, 1994; Zhu et al., 2017). A high-quality reference genome is a prerequisite to understanding the rearrangements associated with chromosome number variation. In addition, such a reference genome will also assist in resolving ambiguities in PFGE-based estimates of chromosome number when similar-sized chromosomes are present. Complete genome assemblies were not available for many of the species with reported numbers of chromosomes. To obtain better-assembled reference genomes, we sequenced the genomes of M. slooffiae and M. globosa as representatives of the nine-chromosome state, and of M. furfur as a representative of the seven-chromosome state, using PacBio SMRT sequencing technology (Figure 1, Figure 1—figure supplement 1).

The M. globosa genome was completely assembled into nine contigs with telomeres on both ends (BioSample accession SAMN10720087). We validated these contigs by matching each band on the pulsed-field gel with the contig sizes from the genome assembly, and further confirmed these by chromoblot analysis following PFGE. This analysis shows that chromosome 5 contains the rDNA locus and migrates further than the expected size of 902 kb, as a diffuse ensemble of different sizes along with chromosome 3 (Figure 1A). The assembled genome of M. slooffiae has 14 contigs of which nine contigs have telomeres on both ends, indicative of nine chromosomes (BioSample accession SAMN10720088). Each of the nine contigs could be assigned to the bands observed in the pulsed-field gel (Figure 1B). For M. furfur, the final genome assembly consisted of seven contigs with telomeres on both ends and matched the expected chromosome sizes obtained from an earlier PFGE analysis of CBS14141 (Figure 1C). The complete genome assembly of M. sympodialis reported earlier is distributed into eight chromosomes with telomere repeats on both ends (Figure 1D), and serves as a representative of an eight-chromosome state in this study.

Changes in chromosome number are always associated with the birth or loss of centromeres, which stabilizes the karyotype in organisms with monocentric chromosomes. To understand the transitions between these different karyotypic states observed in the Malassezia species complex, we sought to validate centromeres in species representative of each karyotype experimentally.

Organisms that have point centromeres possess Ndc10, Cep3, and Ctf13 of the CBF3 complex, a cognate protein complex that is specific to point centromeres. None of these point-centromere-specific proteins could be detected in M. sympodialis. However, we could detect homologs of CENP-A, CENP-C, and most of the outer kinetochore proteins in the genome of M. sympodialis (Figure 2A and Figure 2—figure supplement 1). We functionally expressed an N-terminally GFP-tagged Mtw1 protein (Protein ID: SHO76526.1) from its native promoter, and the expression of the fusion protein was confirmed by western blotting (Figure 2B). Upon staining with anti-GFP antibodies and DAPI (4′,6-diamidino-2-phenylindole), we were able to detect punctate localization of Mtw1 at the nuclear periphery (Figure 2C), consistent with the clustered kinetochore localization observed in other yeasts (Goshima et al., 1999; Euskirchen, 2002; Roy et al., 2011). Live-cell images of MSY001 (GFP-MTW1) cells revealed that the kinetochores (GFP-Mtw1) remained clustered throughout the cell cycle, starting from unbudded G₁ cells in interphase to large-budded cells in mitosis (Figure 2D).

Having identified Mtw1 as an authentic kinetochore protein, we performed ChIP-sequencing using the GFP-Mtw1 expressing strain of M. sympodialis (MSY001). Mapping the reads to the reference genome of M. sympodialis strain ATCC42132 (Zhu et al., 2017) identified one significantly enriched locus on each of the eight chromosomes (Figure 3A). The lengths of the Mtw1-enriched centromere regions identified from the ChIP-seq analysis ranged from 3167 bp to 5143 bp with an average length of 4165 bp (Table 1). However, the region of maximum Mtw1 enrichment on each chromosome (based on the number of sequenced reads aligned) mapped to the intergenic region harboring the GC trough (approximately 1 kb long), which was previously predicted to be the centromeres of M. sympodialis (Figure 3B) (Zhu et al., 2017). The regions of Mtw1 enrichment span beyond the core centromeres and include active genes located proximal to these troughs (Figure 3B, Figure 3—figure supplement 1A). However, these open reading frames (ORFs) do not show consensus features such as the orientation of transcription or functional classification. We validated this enrichment by ChIP-qPCR analysis with primers homologous to the predicted centromeres compared to those homologous to a control region distant from the centromere (Figure 3—figure supplement 1B).

The presence of CENP-A nucleosomes at the kinetochore should result in reduced histone H3 enrichment at the centromeres when compared to a non-centromeric locus. To test this, we performed ChIP with anti-histone H3 antibodies and analyzed the immunoprecipitated (IP) DNA by qPCR. As compared to a control ORF region that is unlinked to the centromere (190 kb away from CEN1), the pericentric regions flanking the core centromere showed a marginal reduction in histone H3 enrichment, which was further reduced at the core, that maps to the GC trough with the highest enrichment of the kinetochore protein Mtw1. That the core centromere region showing the maximum depletion of histone H3 coincided with the regions most enriched with Mtw1 further supports that histone H3, in these regions, is possibly replaced by its centromere-specific variant CENP-A (Figure 3C).

To understand the features of M. sympodialis centromeres, we analyzed the centromere DNA sequences for the presence of consensus motifs or structures such as inverted repeats. PhyloGibbs-MP (Siddharthan et al., 2005; Siddharthan, 2008) predicted a 12-bp-long AT-rich motif common to all of the centromere sequences of M. sympodialis (Figure 3D). We swept the PWM from the PhyloGibbs-MP output across each chromosome of M. sympodialis and counted the number of motif predictions in a sliding 500-bp window, sliding by 100 bp at a time. Sites with log-likelihood-ratio (LLR) of >7.5 were counted as motif predictions. The LLR is the natural logarithm of the ratio of the likelihood of the12-bp substring arising as a sample from the PWM to the likelihood of it being generic ‘background’. In each case, the global peak coincides with the centromere (Figure 3—figure supplement 2A). In each chromosome, the centromere region shows between 7 and 13 motif matches, whereas no other 500-bp window shows more than three matches. This suggests that the AT-rich motif is more enriched at the centromeres than at any other region in the M. sympodialis genome (Figure 3—figure supplement 2A). To ensure that this is not an artifact of the GC-poor nature of the centromere, we repeated the analysis with a synthetic shuffled PWM, created by scrambling the order of the columns of the original PWM (that is, scrambling the positions in the motif while keeping the corresponding weight vectors the same). This shuffled motif showed more matches in the centromeres than are seen in the non-centromeric genomic sequence, but significantly fewer than are seen in the authentic centromeric sequences of most chromosomes (Figure 3—figure supplement 2B). A dot-plot analysis was performed to detect the presence of any direct or inverted-repeat structure associated with the centromeres in M. sympodialis. Analysis of all of the centromere sequences and 5-kb flanking sequences using SyMap confirmed the lack of direct/inverted repeat structures (Figure 3—figure supplement 2C).

In the absence of any centromere-exclusive DNA sequence, the unique and distinguishing features of centromere regions in M. sympodialis are an AT-rich core region of <1 kb (average AT content of 78% as compared to the genome average of 41.5%) that is enriched with the 12-bp motif (Figure 3—figure supplement 3) within a kinetochore protein-bound region of 3–5 kb. As expected, the kinetochore-bound region contains a reduced level of histone H3.

Using the unique centromere features identified in M. sympodialis, we predicted one centromere locus on each of the seven M. furfur chromosomes, and these all map to chromosomal GC troughs (Figure 4—figure supplement 1A, Table 2). We also predicted the centromeres in M. slooffiae, M. globosa, and M. restricta, each of which contains nine chromosomes (Figure 4—figure supplement 1B–D, Table 2). Each of the predicted centromere regions is enriched with the 12-bp AT-rich motif identified in M. sympodialis centromere sequences as compared to other regions in the genomes (Figure 4—figure supplement 2).

To validate the centromeric loci in M. furfur experimentally, we functionally expressed the centromeric histone H3 variant CENP-A with a 3xFLAG tag at the C-terminus (Figure 4A). We performed ChIP in strain MF001 (CENP-A-3xFLAG) and analyzed immunoprecipitated DNA by qPCR using primers specific to each of the seven predicted centromeres and to a centromere-unlinked control locus 1.3 Mb away from CEN1. Enrichment of CENP-A at all seven centromeres over the control locus confirmed that the predicted regions are indeed centromeres in M. furfur CBS14141 (Figure 4B,C).

Given the lack of genetic manipulation methods for M. slooffiae and M. globosa, we tested the enrichment of histone H3 at the predicted centromeres in these two species. All of the nine centromeric loci in these two species contained a reduced histone H3 level when compared to a control locus that was unlinked to centromeres (Figure 4D,E). Furthermore, upon analyzing the enrichment profile at one centromere (CEN1) in M. slooffiae, we observed a reduction in the enrichment levels of histone H3 at the GC troughs as compared to the flanking regions (Figure 4F). In the case of M. globosa, the regions spanning a centromere (CEN2) also depicted a similar reduction in the histone H3 levels (Figure 4G). Taken together, the significant reduction in the histone H3 levels at the predicted centromeres, indicative of the presence of CENP-A, suggests that these putative centromere regions are indeed bona fide centromeres in these species.

Synteny of genes across centromeres is largely conserved in closely related species (Byrne and Wolfe, 2005; Padmanabhan et al., 2008; Yadav et al., 2018b). To understand the transition between different chromosome number states, we analyzed the conservation of gene synteny across centromeres in these species. By mapping gene synteny at the centromeres of M. globosa and M. slooffiae (each with nine chromosomes), compared with that of M. sympodialis (containing eight chromosomes), we found complete gene synteny conservation in eight of the nine centromeres (Figure 5A,B). Thus, syntenic regions of all eight M. sympodialis centromeres are present in the genomes of M. globosa and M. slooffiae. In the case of M. restricta, seven putative centromeres are completely syntenic with M. sympodialis centromeres and one centromere retained partial gene synteny (Table 3). However, no gene synteny conservation was observed at the centromeres of Chr2 in M. globosa, Chr5 in M. slooffiae, or Chr8 in M. restricta (Table 3), indicating the loss of a centromere during the transition from the nine-chromosome state to the eight-chromosome state.

The GC trough corresponding to MgCEN2/MslCEN5 is flanked by genes that map to MsyChr2 on one arm and MsyChr4 on the other (Figure 5C,D). The centromere region in each of MgChr2 and MslChr5 marks a synteny breakpoint showing no homologous region in the M. sympodialis genome, indicating a loss of this centromere DNA sequence. We also observed that the genes flanking the breakpoint are conserved in M. sympodialis, suggesting that the specific intergenic region was involved (Figure 5E,F). Evidence for the internalization of telomere-adjacent ORFs or for the presence of interstitial telomere repeats that are indicative of telomere–telomere fusions was not detected in the M. sympodialis genome. These observations strongly support our hypothesis that breakage of MgCEN2/MslCEN5 (or the orthologous ancestral CEN) and fusion of the two acentric arms to other chromosomes resulted in the reduction in chromosome number observed when comparing these species.

To understand the basis of the change in chromosome number from nine to seven in Malassezia species, we compared the synteny of ORFs flanking the M. slooffiae or M. globosa centromeres with that of M. furfur. Of the nine centromeres in M. slooffiae, three centromeres belonged to conserved gene synteny blocks and four others retained partial gene synteny conservation in M. furfur (Figure 6A, Table 3). A similar pattern of gene synteny conservation was observed between M. globosa and M. furfur (Figure 6B, Table 3). The genes flanking the remaining two centromeres (CEN8 and CEN9) in M. slooffiae were present in conserved gene synteny blocks in the two arms of MfChr3 (Figure 6C). However, the regions corresponding to CEN8 and CEN9 in M. slooffiae appear to have evolved to decreased AT-richness in M. furfur. A similar centromere inactivation mechanism was observed when CEN8 and CEN9 of M. globosa were compared to the corresponding syntenic regions in M. furfur (Figure 6D). These results suggest centromere inactivation by changes in the centromeric DNA sequence in this species (Figure 6—figure supplement 1).

To trace the ancestral karyotype in Malassezia, we predicted centromeres and inferred chromosome numbers in other species of clades A and B on the basis of GC troughs and gene synteny. We identified putative centromeres in M. dermatis and M. nana in Clade B because of their relatively better-assembled genomes distributed in 18 and 13 scaffolds, respectively. Of these, we could predict eight centromeric regions marked by GC troughs that were also enriched with the 12-bp motif in each species (Figure 7—figure supplement 1A–B, Figure 7—figure supplement 2A–B and Table 4). Furthermore, in both of these species, the eight putative centromeres shared complete gene synteny with the regions spanning M. sympodialis centromeres, indicating that their common ancestor had eight chromosomes (green circle in Figure 7A, Figure 7—figure supplement 3 and Table 3). To map the common ancestor in Clade B Malassezia species, we analyzed regions flanking centromeres of Chr2 of M. globosa and Chr8 of M. restricta, both of which mapped to the gene synteny breakpoint of the genome of the Malassezia species that have eight chromosomes suggesting that their common ancestor, named as Ancestor B (Anc. B), also had nine chromosomes (Figure 7A, Figure 7—figure supplement 3). On the basis of our centromere predictions in Clade B species and synteny analysis, we propose that centromere breakage would have occurred in the common ancestor of M. sympodialis, M. nana, and M. dermatis after divergence from the common ancestor of M. globosa and M. restricta, which retained a nine-chromosome configuration (Figure 7A,B).

As mentioned earlier, M. furfur and M. obtusa of Clade A contain seven chromosomes each (Boekhout and Bosboom, 1994; Boekhout et al., 1998). To further understand the karyotype variations within this clade, we predicted the chromosome number in Malassezia vespertilionis and Malassezia japonica because their genomes are relatively well assembled (Sugita et al., 2003; Lorch et al., 2018). In both of these species, we were able to predict nine GC troughs, indicative of the centromeres of nine chromosomes (Figure 7—figure supplement 1C–D, Table 4). In the case of M. vespertilionis, all of the predicted centromeres showed enrichment of the 12-bp motif (Figure 7—figure supplement 2C). However, the 12-bp motif was found to be enriched in all of the predicted centromeres except the centromere of scaffold 7 of M. japonica (Figure 7—figure supplement 2D). The presence of species with nine chromosomes in Clade A suggests that the ancestral state in this clade, Anc. A, also contained nine chromosomes (Figure 7A).

We identified nine centromeres in M. slooffiae, the only species in Clade C with a well-assembled genome. The presence of species with nine chromosomes in each of the three clades of Malassezia species, the conservation of gene synteny across orthologous centromeres, and the similar centromere features shared by all nine species analyzed in this study suggest that Malassezia species diverged from a common ancestor that had nine chromosomes with short regional centromeres enriched with the 12-bp AT-rich DNA sequence motif.

Malassezia strains were grown on modified Dixon's media (malt extract 36 g/L, desiccated ox-bile 20 g/L, Tween40 10 mL/L, peptone 6 g/L, glycerol 2 mL/L, oleic acid 2.89 mL/L). M. sympodialis, M. furfur strains were grown at 30°C. Cultures of M. globosa and M. slooffiae were grown at 32°C.

The allele for N-terminal tagging of Mtw1 with GFP was prepared by gap repair in the Saccharomyces cerevisiae BY4741 strain (Eckert-Boulet et al., 2012). Briefly, a 1.6-kb fragment consisting of the upstream and promoter sequence of the MTW1 gene and a 1.6-kb fragment having the MTW1 ORF (Protein ID: SHO76526) along with the downstream sequence were amplified from M. sympodialis genomic DNA. The GFP ORF (without the stop codon) and NAT were amplified from plasmids pVY7 and pAIM1, respectively. S. cerevisiae was transformed with all four fragments and the linearized plasmid pGI3 (digested with KpnI and BamHI) and the epitope-tagged allele were assembled in an ordered way by gap repair. Total DNA was isolated from S. cerevisiae and the E. coli DH5α strain was transformed. The pGFP-Mtw1 construct was screened by restriction digestion and further confirmed by sequencing. The pGFP-Mtw1 construct was used to transform M. sympodialis strain ATCC42132 by Agrobacterium tumefaciens-mediated transconjugation (Ianiri et al., 2016; Ianiri et al., 2017a).

The allele for C-terminal tagging of CENP-A with a 3xFLAG epitope tag was prepared by gap repair in the Saccharomyces cerevisiae BY4741 strain (Eckert-Boulet et al., 2012). Briefly, a 1-kb fragment consisting of the upstream and promoter sequence of the CENP-A gene of M. furfur including the ORF (CENP-A ORF coordinates in Chr1: 1,453,468–1,453,921), as well as a 1-kb fragment containing the sequence downstream of CENP-A ORF, were amplified from M. furfur genomic DNA. The 3xFLAG tag was introduced in the reverse primer annealing to the CENP-A ORF. The NAT marker was amplified from plasmid pAIM1 as above. S. cerevisiae was transformed with all three fragments and plasmid pGI3 (digested with KpnI and BamHI) and the epitope-tagged allele were assembled in an ordered way by gap repair. Total DNA was isolated from S. cerevisiae and the E. coli DH5α strain was transformed. The resulting pMF1 construct was screened by restriction digestion and further confirmed by sequencing. The pMF1 construct was used to transform M. furfur strain CBS14141 by Agrobacterium tumefaciens-mediated transconjugation (Ianiri et al., 2016; Ianiri et al., 2017a) to obtain the epitope-tagged strain MF001.

The GFP-Mtw1 strain was inoculated to 1% v/v from a saturated starter culture grown in mDixon medium. After growth for 6 hr at 30 °C/ 150 rpm, these cells were pelleted at 4,000 rpm and washed three times with 1x phosphate-buffered saline (PBS), and the cell suspension was placed on a clean glass slide. A coverslip was placed on the spot and sealed prior to imaging. The images were acquired at room temperature using a laser scanning inverted confocal microscope LSM 880-Airyscan (ZEISS, Plan Apochromat 63x, NA oil 1.4) equipped with highly sensitive photo-detectors. The filters used were GFP/FITC 488 excitation and GFP/FITC 500/550 band pass, long pass for emission. Z-stack images were taken every 0.3 μm and processed using ZEISS Zen software or ImageJ. All of the images were processed post-acquisition with minimal adjustments to levels and linear contrast until the signals were highlighted.

Cells were grown on mDixon’s medium and washed with water by centrifugation at 4000 rpm for 5 min. Cells were resuspended in 10 mL of 5% (v/v) 2-mercaptoethanol solution in water and incubated at 30 °C/ 150 rpm for 45 min. The cells were pelleted, washed, and resuspended in 3 mL spheroplasting buffer (40 mM citric acid, 120 mM Na₂HPO₄ and 1.2 M sorbitol) for every 1.5 × 10⁹ cells. Cell clumps were dissociated by mild sonication for 30 s using the medium-intensity setting in a Bioruptor (Diagenode). Lysing enzymes from Trichoderma harzianum (Sigma), chitosanase (Sigma), and zymolyase-20T (MP Biomedicals) were added at 20 mg/mL, 0.2 µg/mL, and 100 µg/mL, respectively. The spheroplasting suspension was incubated at 30°C/65 rpm for 6–8 hr. The suspension was examined under a microscope to estimate the proportion of spheroplasts. Spheroplasts were washed with ice-cold 1xPBS and used as per the experimental design (adapted from Boekhout, 2003).

The GFP-Mtw1 strain was inoculated to 1% (v/v) from a saturated starter culture grown in mDixon medium. After growth for 6 hr, the cells were fixed by the addition of formaldehyde to a final concentration of 3.7% for 1 hr. Post-fixing, the cells were washed with water and taken for preparation of spheroplasts (as described above). Spheroplasts were washed with ice-cold 1xPBS and resuspended in ice-cold 1xPBS to a cell density suitable for microscopy. Slides for microscopy were washed with water and coated with poly L-lysine (15 µL of 10 mg/mL solution per well) for 5 min at room temperature. The solution was aspirated and washed once with water. The cell suspension was added to each well (15–20 µL) and allowed to stand at room temperature for 5 min. The cell suspension was aspirated and the slides were washed once with water to remove unbound cells. The slides were fixed in ice-cold methanol for 6 min followed by treatment with ice-cold acetone for 30 s. Post fixing, blocking solution (2% non-fat skim milk in 1xPBS) was added to each well, and slides were incubated at room temperature for 30 min. After this, the blocking solution was aspirated and primary antibodies were added (mouse anti-GFP antibodies [Sigma] at 1:100 dilution). After incubation for 1 hr at room temperature, each slide was washed eight times with 1xPBS giving a 2 min incubation for every wash. Secondary antibody solution (goat anti-mouse-AlexaFluor488 [Invitrogen] at 1:500 dilution) was added to each well and incubated for 1 hr in the dark at room temperature. Post-incubation, slides were washed as described above. Mounting medium (DAPI at 100 ng/mL in 70% glycerol) was added, incubated for 5 min and aspirated out. Slides were sealed with a clean coverslip before imaging. The images were acquired at room temperature using an inverted fluorescence microscope (ZEISS Axio Observer, Plan Apochromat 100x, NA oil 1.4). Z- stack images were taken every 0.3 μm and processed using ZEISS Zen software/ImageJ.

The ChIP protocol was adapted from that implemented for C. neoformans (Yadav et al., 2018b). Logarithmically grown cells were fixed with formaldehyde at a final concentration of 1% for 30 min for Mtw1 ChIP and 15 min for CENP-A, histone H3, and histone H4 ChIP. The reaction was quenched by the addition of glycine to a final concentration of 0.135 M. Cells were pelleted and processed for spheroplasting as described above. Spheroplasts were washed once sequentially using 10 mL of the following ice-cold buffers: 1xPBS, Buffer-I (0.25% Triton X-100, 10 mM EDTA, 0.5 mM EGTA, 10 mM Na-HEPES [pH 6.5]), and Buffer-II (200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 10 mM Na-HEPES [pH 6.5]). The pellet after the final wash was resuspended in 1 mL lysis buffer (50 mM HEPES [pH 7.4], 1% Triton X-100, 140 mM NaCl, 0.1% Na-deoxycholate, 1 mM EDTA) for every 1.5 × 10⁹ cells. Protease inhibitor cocktail was added to 1x final concentration.

The resuspended spheroplasts were sonicated with a Bioruptor (Diagenode) using a 30 s ON/OFF pulse at high-intensity mode with intermittent incubation on ice to obtain chromatin fragments in the size range of 100–300 bp. The lysate was cleared after sonication by centrifugation at 13,000 rpm for 10 min at 4°C. The input DNA fraction was separated at this step (1/10^th volume of lysate) and processed for de-crosslinking by the addition of 400 µL elution buffer (0.1 M NaHCO₃, 1% SDS) per 100 µL lysate (processing for de-crosslinking is mentioned below). The remaining lysate was split equally and processed as IP and control samples. For GFP-Mtw1 ChIP, 20 µL GFP-trap beads and blocked agarose beads, respectively, were used for IP and control. For CENP-A-3xFLAG ChIP, 20 µL anti-FLAG affinity gel and blocked agarose beads, respectively, were used for IP and control. In the case of histone H3 or histone H4 ChIP, 5 µL of antibodies were used per IP fraction along with 20 µL Protein-A sepharose beads. Samples were rotated for 6 hr at 4°C. Post incubation, samples were sequentially washed as follows: twice with 1 mL low salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris [pH 8.0], 150 mM NaCl), twice with 1 mL high salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris [pH 8.0], 500 mM NaCl), once with 1 mL LiCl wash buffer (0.25 M LiCl, 1% NP-40, 1% Na-deoxycholate, 1 mM EDTA, 10 mM Tris [pH 8.0]) and twice with 1 mL 1xTE (10 mM Tris [pH 8.0], 1 mM EDTA). Samples were rotated in a rotaspin for 5 min at room temperature for every wash (15 min in case of histone H3 or histone H4 ChIP). After washing, DNA was eluted from the beads twice using 250 µL elution buffer. The samples for elution were incubated at 65°C for 5 min, rotated for 15 min and then collected by centrifugation.

Samples were decrosslinked by the addition of 20 µL 5 M NaCl and incubation at 65°C for 6 hr. Following this, samples were deproteinized by the addition of 10 µL 0.5 M EDTA, 20 µL 1 M Tris [pH6.8], 2 µL Proteinase K (20 mg/L) and incubation at 45°C for 2 hr. After incubation, samples were treated with an equal volume of phenol-chloroform-isoamyl alcohol (25:24:1) mix, and the aqueous phase was extracted by centrifugation. DNA was precipitated by the addition of 1/10^th volume of 3 M Na-acetate, 1 µL glycogen (20 mg/mL), 1 mL absolute ethanol and incubation at −20°C for at least 12 hr. Finally, the samples were harvested by centrifugation at 13,000 rpm for 45 min at 4°C followed by washing the pellet once with ice-cold 70% ethanol. Air-dried pellets were then resuspended in 20 µL sterile MilliQ water with 10 µg/mL RNAse. Samples were either processed for library preparation for ChIP-sequencing or analyzed by qPCR with the primers listed in Supplementary file 1.

GFP-Mtw1 ChIP sequencing was performed at the Clevergene Biocorp. Pvt. Ltd., Bengaluru, India. A total of 46,704,720 and 63,524,912 150-bp paired-end reads were obtained for IP and Input samples, respectively. The reads were mapped to the M. sympodialis ATCC42132 genome using Geneious 9.0 (http://www.geneious.com/) with default conditions. Each read was allowed to map only once randomly anywhere in the genome. The alignments were exported to BAM files, sorted, and visualized using the Integrative Genomics Viewer (IGV, Broad Institute). The images from IGV were imported into Adobe Photoshop (Adobe) and scaled for representation purposes. RNA-sequencing data (E-MTAB-4589) from a previous study was downloaded from the ArrayExpress website, sorted, and visualized using IGV. GC-content was calculated using Geneious 9.0 with a sliding window size of 250 bp. The data were exported as wig files and further visualized using IGV.

Protein lysates for western blot were prepared by the TCA method. Overnight grown cultures of 1 mL were harvested, washed, and resuspended in 400 μL of 12.5% ice-cold TCA solution. The suspension was vortexed briefly and stored at – 20°C for 4–6 hr. The suspension was thawed on ice, pelleted at 14,000 rpm for 10 min, and washed twice with 350 μL of 80% acetone (ice-cold). The washed pellets were air-dried completely and resuspended in the desired volume of lysis buffer (0.1 N NaOH+1% SDS). Samples were separated in 12% polyacrylamide gels and transferred onto nitrocellulose membranes. For probing, mouse anti-GFP antibody (Roche) and the HRP conjugated goat anti-mouse secondary antibody (Abcam) were used at 1:3000 and 1:5000 dilution, respectively, in 2.5% skim milk powder in 1xPBS. The blots were developed using Chemiluminescence Ultra substrate (BioRad) and imaged using the VersaDoc imaging system (BioRad).

For CHEF analysis of M. globosa (CBS7966) and M. slooffiae (CBS7956), the cells were grown on solid mDixon medium and then collected and resuspended in 1xPBS. CHEF plugs were prepared as described in previous studies (Sun et al., 2014; Sun et al., 2017). Chromosomes were separated in 1% Megabase certified agarose gels made with 0.5 × TBE, using a BioRad CHEF-DR II System running at 3.2 V/cm with linear ramping switching time from 90 to 360 s for 120 hr in 0.5xTBE at 14°C. The gel was stained with EtBr and visualized under UV.

For the chromoblot analyses of M. globosa (CBS7966), the gel from the CHEF analysis was first transferred to a membrane, and the resulting chromoblots were then hybridized sequentially with four probes from chromosomes 3, 4, 5, and 6 of the CBS7966 genome assembly, respectively (see the Supplementary file 1 for the primer information), as described in previous studies (Findley et al., 2012; Yadav et al., 2018b).

Sequence reads were assembled with HGAP3 included in SMRTPortal v2.3 (PacBio, Menlo Park, CA, USA) and default parameters, except for the genome size set to 9 Mb. Assembly completeness was evaluated by checking for telomeric repeats. Non-telomeric contig-ends were aligned to other contigs using BLAST and unique overlaps were used to build complete chromosomes. Short telomere ends were extended using uniquely mapped reads longer than 10 kb and repolishing of the assembly using the resequencing pipeline in SMRTPortal v2.3. The assembly resulted in 19 contigs, with a total length of 9.2 Mb. 17 long and one short telomere could be identified. Six contigs had telomeres on both the 5'- and the 3'-end, and thus represent full-length chromosomes (Chr1, 2, 3, 6, 7, and 8). Six contigs had only one telomere and seven contigs had no telomeric sequence. Two contigs without telomeres were from the mitochondrion and two were from the ribosomal repeats. Chromosome 5 was constructed from two contigs ending in ribosomal rDNA repeats. The assembly contains six copies of the repeat, but read coverage suggests a length of 30–40 repeat units that cannot be resolved with the available read-length. The remaining contigs were used to build chromosomes 4 and 9, which share highly similar 5'-ends. The two ends can be distinguished by two microsatellite expansions. Chromosome 4 had a very short 3'-telomere from the default assembly, but the raw data contained a uniquely mapping read that extended several repeat units past the assembly end. After polishing the reference, all nine chromosomes had clear 5′- and 3′- telomeres.

Sequence reads were assembled using HGAP3 included in SMRTPortal v2.3 (PacBio, Menlo Park, CA, USA) with default parameters. This resulted in an assembly with 14 scaffolds with telomeric repeats at both ends in nine contigs. Of the remaining five contigs, three of them could be assigned to mitochondrial DNA on the basis of BLAST analysis with M. globosa. The remaining two contigs, of size 5.8 kb and 2.3 kb, did not show BLAST hits against the M. globosa or M. sympodialis genomes.

Sequence reads generated from CBS14141 were assembled using HGAP3 included in SMRTPortal v2.3 (PacBio, Menlo Park, CA, USA) with default parameters. This resulted in an assembly with eight scaffolds, of which the nuclear genome was organized in seven scaffolds that had telomere repeats at both ends.

Analysis of gene synteny conservation across the centromeres was performed by BLAST as follows. The genomes for M. restricta, M. nana, and M. dermatis were downloaded from the NCBI genomes portal. The PacBio assembled genomes of M. globosa and M. slooffiae were used for synteny analysis. Synteny analysis was done in the context of ORFs flanking the centromeres of M. sympodialis. The protein sequences for each of these ORFs served as the query in BLAST analysis against the genome of other species. The local database for each genome was set up in the Geneious software for this analysis. The percentage identity values for each of the ORFs are mentioned in the boxes in Figure 7—figure supplement 3. In addition, synteny analyses between M. globosa and M. sympodialis were conducted with megablast (word size: 28) and plotted together with GC content (calculated as the deviation from the genomic mean, in non-overlapping 1-kb windows), using Circos (v0.69–6) (Krzywinski et al., 2009). Additional whole-genome alignments were conducted with Satsuma (Grabherr et al., 2010), with default parameters. The linear synteny comparisons shown in Figures 5 and 6 were generated with the Python application EasyFig (Sullivan et al., 2011).

To reconstruct the phylogenetic relationship among the nine Malassezia species selected, orthologs were identified using the bidirectional best-hit (BDBH), COGtriangles (v2.1), and OrthoMCL (v1.4) algorithms implemented in the GET_HOMOLOGUES software package (Contreras-Moreira and Vinuesa, 2013). The proteome of M. sympodialis ATCC42132 was used as a reference. A phylogeny was inferred from a set of 738 protein sequences as follows. Individual protein sequences were aligned using MAFFT v7.310 (L-INS-i strategy), and poorly aligned regions were trimmed with TrimAl (-gappyout). The resulting alignments were concatenated to obtain a final supermatrix consisting of a total of 441,200 amino acid sites (159,270 parsimony-informative). This sequence was input to IQ-TREE v1.6.5 (Nguyen et al., 2015) and a maximum likelihood phylogeny was estimated using the LG+F+R4 amino acid model of substitution. Branch support values were obtained from 10,000 replicates of both ultrafast bootstrap approximation (UFBoot) and the nonparametric variant of the approximate likelihood ratio test (SH-aLRT) implemented in IQ-TREE. The best likelihood tree was graphically visualized with iTOL v4.3.3 (Letunic and Bork, 2019).