To determine if neocentromere formation occurs in the C. deuterogattii reference strain R265, either centromere 9 or 10 was deleted. Centromere 9 (CEN9) was deleted by CRISPR-Cas9-mediated transformation. Two guide RNAs flanking the centromere were used and CEN9 was replaced with a NAT dominant drug-resistance gene by homologous recombination. Biolistic transformation was used to replace centromere 10 (CEN10) with either the NAT or NEO dominant drug-resistance gene via homologous recombination. Viable transformants with the correct integration and deletion were obtained and confirmed by 5’ junction, 3’ junction, loss of deleted regions, and spanning PCRs as well as Southern blot analysis for cen10∆ (Figure 1—figure supplement 1, Figure 1—figure supplement 2). Multiple independent cen9∆ and cen10∆ deletion mutants (cen9∆-A to -F and cen10∆-A to -G) were obtained from independent transformations. Pulsed-field gel electrophoresis (PFGE) confirmed that cenΔ mutants had a wild-type karyotype and that chromosome 9 and 10 remained linear, because a circular chromosome would not have entered the gel (Figure 1—figure supplement 3).

The formation of neocentromeres on chromosome 10 in C. deuterogattii was infrequent. A total of 99 independent biolistic transformations resulted in only seven confirmed cen10∆ mutants (7/21 total candidate transformants, 33% homologous integration), suggesting that CEN10 deletion is lethal in most circumstances. In comparison, deletion of nonessential genes by homologous recombination in the C. deuterogattii R265 strain typically results in ~100 colonies with a high success rate (~80–90% homologous integration). We estimate that the likelihood of deleting a centromere and recovering a viable colony is at least 1000-fold lower than would be expected from the deletion of a non-essential gene. The deletion of CEN9 was more efficient as this was mediated by CRISPR-Cas9 cleavage with two guide RNAs and a repair allele.

Chromatin immunoprecipitation of mCherry-CENP-A followed by high-throughput sequencing (ChIP-seq) for six cen9∆ (-A to -F) and seven cen10∆ mutants (-A to -G) was performed (Figure 1). Prior to the ChIP-seq experiment, all of the centromere deletion mutants were streak purified from single colonies. The sequence reads were mapped to a complete whole-genome assembly, followed by the normalization of the reads by subtraction of the input from the ChIPed sample (Yadav et al., 2018). To quantify the ChIP-seq data, the CENP-A-enriched regions were compared with the centromeres previously identified based on CENP-C enrichment. Both the CENP-A- and CENP-C-enriched peaks were congruent for all of the native centromeres (Yadav et al., 2018). This analysis identified 13 of the 14 native centromeres (CEN1-8, CEN11-14 and depending on the centromere mutant either CEN9 or CEN10), indicating that, as expected, the native centromere of chromosome 9 or 10 was missing in all of the cen9∆ and cen10∆ deletion mutants respectively (Figure 1). Instead, neocentromeres were observed.

Except for the neocentromere of isolate cen10∆-E, the neocentromeres formed in close proximity to the native centromere (CEN9 and CEN10). Almost all neocentromeres were shorter than the native centromere, with the exception of cen10∆-G which was larger than native centromere 10 (Table 1).

In three of the independent cen9∆ mutants (cen9∆-B, -C and -E), neocentromeres formed at the same chromosomal location (Figure 1A). Interestingly, two independent cen10∆ mutants (cen10∆-A and cen10∆-C) contained two CENP-A-enriched regions on chromosome 10, with a primary peak and a smaller secondary peak with reduced levels of CENP-A (1.3- to 1.75-fold lower) compared to the primary CENP-A peak (Figure 1C). The chromosomal location of the secondary peak was similar to the neocentromere of cen10∆-B (which had only one neocentromere) (Figure 1B).

The two CENP-A-enriched regions suggest four possible models: 1) aneuploidy in which cells harbor two chromosomes, 2) a dicentric chromosome with two neocentromeres (neodicentric), 3) instability between two different neocentromere states (neocentromere switching), 4) or only one CENP-A-enriched region functions as a centromere and the second CENP-A-enriched region is not bound by the kinetochore (Figure 1).

The neocentromeres were located in unique, nonrepetitive sequences and were not flanked by repetitive regions. The GC content of neocentromeres is similar to the overall GC content of chromosome 9 and 10, whereas the native centromere has a lower GC content (Table 1). Comparing the reference genome with de novo genome assemblies of cen10∆-A, cen10∆-B, and cen10∆-E confirmed that transposable elements did not enter these genomic regions during neocentromere formation (Supplementary file 1). Instead of spanning repeats and transposable elements like the native centromeres, neocentromeres span genes.

All of the neocentromeres of chromosome 9 formed in a region within 26 kb of the chromosomal location of the native centromere 9 (Figure 1A). Interestingly, the neocentromeres of the independent cen9∆-B, cen9∆-C, and cen9∆-E mutants all formed at the same chromosomal location and had the same length (4.41 kb); these neocentromeres spanned three genes. One gene was completely covered by CENP-A and this gene encodes a transglycosylase SLT domain-containing protein. The two other genes (a gene encoding a xylosylphosphotransferase and a gene encoding glutamate synthase (NADPH/NADH)) were partially covered with CENP-A. Mutant cen9∆-A had a 3.87 kb long neocentromere located 26 kb 3’ to the native centromere and spanned two genes. The first gene was completely spanned by CENP-A and encodes an ESCRT-II complex subunit (Vps25) protein. The second gene was only partially covered and encodes an iron regulator protein. The neocentromere of cen9∆-D was located directly to the left of the native centromere and was 4.37 kb in length. This neocentromere spanned two genes, coding for a hypothetical protein (92% covered by CENP-A) and a gene encoding for Derlin-2/3 that was completely covered by CENP-A. Lastly, the neocentromere of mutant cen9∆-F was 3.83 kb in length and spanned one gene (encoding a xylosylphosphotransferase, XPT1), which was completely covered by CENP-A. This neocentromere was located 12 kb away (3’) from the native centromere 9.

Like the neocentromeres of cen9∆ mutants, the neocentromeres of cen10∆ mutants also spanned genes and interestingly, the kinetochore protein CENP-C was located inside the neocentromere of cen10∆-B and in the secondary peak of cen10∆-A and -C (Table 1, Figure 1B). The neocentromere in cen10∆-B spanned 4.46 kb, was located 242 kb away from the 3’ region of the native CEN10, and was located 115 kb from the telomere (Figure 1B). In addition to the gene encoding CENP-C, the CENP-A-enriched region spanned a hypothetical protein (Table 1). The primary CENP-A-enriched region of cen10∆-A and cen10∆-C spanned a gene encoding a serine/threonine-protein phosphatase 2A activator 2 (RRD2) and a hypothetical protein (Figure 1D). This neocentromere spanned 2.85 kb and was located closer to the native CEN10 (21 kb from the native centromere) than the neocentromere of cen10∆-B and the secondary CENP-A peak of cen10∆-A and cen10∆-C. The neocentromere of cen10∆-D was the smallest neocentromere (2.5 kb) and partially (88.4%) spanned a gene encoding a Ser/Thr protein kinase and formed 7.4 kb from the location of the native CEN10 (Figure 1D). The neocentromere of cen10∆-E spanned two hypothetical proteins, was 4.38 kb in length and was located directly adjacent to the right telomere (Figure 1E). Mutant cen10∆-F had a neocentromere of 2.64 kb, which spanned one hypothetical gene completely and two genes (hypothetical and a hexokinase (HXK1)) partially; the neocentromere formed at a chromosomal location 20 kb 5’ of the native centromere (Figure 1D). The neocentromere of cen10∆-G was the largest neocentromere with a CENP-A-enriched region of 7.97 kb, and was in fact larger than the native CEN10. This neocentromere spanned four genes, including a gene coding for a high osmolarity protein (Sho1) and three genes coding for hypothetical proteins (Figure 1D).

To test if the kinetochore was binding to the CENP-A-enriched regions of chromosomes 9 and 10, and to validate if the neocentromeres were fully functional as centromeres, two additional kinetochore proteins were epitope-tagged with GFP. cen9∆ mutants were transformed with an overlap PCR product expressing CENPC-GFP. As the neocentromeres of three cen10∆ mutants spanned the gene encoding CENP-C, all cen10∆ mutants were transformed with an overlap PCR product, expressing Mis12-GFP. In addition to the cen9∆ and cen10∆ mutants, the wild type was transformed with constructs expressing Mis12-GFP and CENP-C-GFP, and these served as controls. ChIP-qPCRs for cen9∆ mutants, cen10∆ mutants, and wild-type strains with Mis12-GFP or CENP-C-GFP were performed (Figure 1—figure supplement 4). Because Mis12 is an outer kinetochore protein, the formaldehyde cross-linking was extended to 45 min (15 min was used for CENP-A and CENP-C) for this protein. For all qPCR analyses, the native centromere 6 CEN6) was used as an internal control and for each neocentromere specific primer pairs were designed. For cen9∆ and cen10∆ mutants a similar level of Mis12 or CENP-C enrichment at the neocentromeres and (CEN6) was observed. This suggested that the CENP-A-enriched regions of chromosome 9 of the cen9∆ mutants and chromosome 10 of cen10∆ mutants identified by ChIP-seq were functional centromeres and indeed neocentromeres (Figure 1—figure supplement 4).

Previously generated RNA sequence data were remapped to the C. deuterogattii reference strain R265 and analyzed to determine if the regions where neocentromeres formed in the cenΔ mutants were transcribed in the wild type (Figure 1; Supplementary file 2; Schneider et al., 2012). In the wild-type strain, all genes spanned by neocentromeres in the cenΔ mutants were expressed (Figure 1). However, the expression levels of the neocentromeric genes were lower than their neighboring genes. For example, the expression level of the gene (CNBG_5686) flanking native centromere 9 was three times higher than the genes spanned by neocentromeres in cen9∆ mutants (Figure 1A). The same trend was observed in the cen10∆ mutants. Here, the expression level of the gene (CNBG_4365) 3’ flanking the neocentromere (primary CENP-A peak) of cen10∆-A and cen10∆-C was more than six times higher than the genes spanned by the neocentromere (Figure 1D). Also, the neocentromere of cen10∆-D is flanked by genes whose expression was two times higher than the genes spanned by the neocentromere (Figure 1D). This suggests that neocentromeres are formed in chromosomal regions with lower gene expression in C. deuterogattii. The majority of the genes (24/28) flanking native centromeres are transcribed in the direction towards the native centromere. All of the neocentromeres observed span one or more genes and most of the flanking genes are transcribed in the direction away from the neocentromere.

The expression levels of the neocentromeric genes in cen9∆ and cen10∆ mutants were assayed by qPCR (Figure 2). The neocentromeric genes of chromosome 9 were normalized to actin. To compensate for the ploidy levels of chromosome 10 in cen10∆ mutants, a housekeeping gene located on chromosome 10 was used to normalize the expression of genes spanned by neocentromeres located on chromosome 10. The expression levels of the CENP-A-associated neocentromeric genes were all found to be similar to the wild-type strain (Figure 2).

 cen10∆ mutants were noted to grow more slowly than wild type. To investigate this, the growth of cen10∆ and wild-type strains was measured during the course of a 22-hour cell growth experiment (Figure 3A). The majority of cen10∆ mutants exhibited slower growth rates compared to the wild-type parental strain R265. Six of seven cen10∆ mutants exhibited significant fitness defects compared to the wild-type strain, with doubling times ranging from 101 to 111 min compared to 81 min for the wild type (Figure 3B). In contrast, one mutant, which has a telocentric neocentromere (cen10∆-E), grew similarly to the wild type and had a similar doubling time (84 min for the mutant vs 81 min for the wild-type strain). Compared to the wild type, cen10∆ mutants with increased doubling times produced smaller colonies during growth on non-selective media (Figure 1—figure supplement 5C).

To compare fitness, a competition assay was performed with 1:1 mixtures of wild-type and cen9∆ or cen10∆ mutants grown in liquid YPD medium (Figure 3C). With no growth defect, the expectation was that the wild-type strain and centromere deletion mutants would grow at the same growth rate, resulting in a 1:1 ratio. In fact, fewer cen10∆ cells were found in the population after growth in competition with the wild-type strain, and this observation is consistent with the slower doubling time of cen10∆ mutants resulting in reduced fitness compared to wild type (Figure 3). Compared to the wild-type cells, there were fewer cen9∆ mutant cells in the population. However, the number was closer to a 1:1 ratio (Figure 3C). The ratio of the cen9∆ mutants in the population was similar to the ratio of the cen10∆-E mutant, which had a wild-type growth rate. Due to this observation, we hypothesize that the growth rate of the cen9∆ mutants is similar to wild type.

Because deletion of a centromere could lead to defects in chromosome segregation, cenΔ mutants were assessed for aneuploidy (Figure 4). Overall, cen10∆ mutants exhibited a mixture of large and small colony sizes during growth on YPD medium at 37°C, while cen9∆ mutants exhibited a uniform, wild-type like, colony size (Figure 1—figure supplement 5).

Aneuploidy in C. neoformans often leads to a similar mixed colony size phenotype as that observed in the cen10∆ mutants (Sun et al., 2014). To exacerbate the aneuploidy-associated slow growth phenotype, four cen10∆ mutants were grown at elevated temperature (37°C), causing these isolates to produce smaller, growth-impaired and larger, growth-improved colonies (Figure 3—figure supplement 1). Three small and two large colonies were selected from each isolate and whole-genome analysis was performed based on Illumina sequencing. Sequences were mapped to the reference R265 genome, revealing that the small colonies were indeed aneuploid (Figure 4A). The small colonies of cen10∆-B and cen10∆-C had ploidy levels for chromosome 10 in the range of 1.25- to 1.36-fold higher compared to the other 13 chromosomes, which suggested that only a proportion of the cells (25% to 36%), were aneuploid (Figure 4B). The remainder of the genome was euploid. Chromosome 10 of the small colonies derived from isolate cen10∆-A and cen10∆-E exhibited ploidy levels ranging from 1.1- to 1.14-fold, reflecting less aneuploidy. Importantly, for all of the large colonies derived from isolates cen10∆-A, cen10∆-B, cen10∆-C, and cen10∆-E the fold coverage of chromosome 10 was restored to the wild-type euploid level (1.0 fold compared to wild type). The ploidy levels of chromosome 10 were 1-fold for all of the large colonies compared to wild type, indicating that the ploidy level of chromosome 10 of the large colonies was restored to euploid.

Based on whole-genome sequencing and PFGE analysis, fusion of cen10∆ chromosome 10 to other chromosomes was a common event in the large colonies (Figure 5, Figure 6). Whole-genome sequence analysis revealed that sequences corresponding to the 3’ subtelomeric region of chromosome 10 (including one gene) were absent in the sequences obtained from all of the large colonies analyzed (Figure 4—figure supplement 1A). In addition, the large colonies of cen10∆-A were missing sequences for two genes in the 5’ subtelomeric region of chromosome 4 (Figure 4—figure supplement 1B). Large colonies of cen10∆-B were missing 18.5 kb at the 5’ subtelomere of chromosome 7 (including eight genes) (Figure 4—figure supplement 1C). The large colonies of cen10∆-E lacked a small part of one gene in the 3’ subtelomeric region of chromosome 1. In total, of the 14 subtelomeric genes that were lost in these three chromosome-fusion isolates, ten encoded hypothetical proteins and four encoded proteins with predicted functions. BlastN analysis in the de novo genome assemblies of the large colonies confirmed that the subtelomeric regions were not located on minichromosomes or inserted in other chromosomes. Seven genes have homologs in C. neoformans and are present in C. neoformans deletion libraries (Liu et al., 2008; Supplementary file 3). This observation suggested that either subtelomeric deletions occurred, or that chromosomal fusions led to the loss of subtelomeric regions. Notably, sequences from the small colonies spanned the entire genome with no evidence of these subtelomeric deletions (Figure 4—figure supplement 1F).

We hypothesized that the subtelomeric gene loss was due to chromosomal fusion and tested this hypothesis with de novo genome assemblies and PFGE (Figure 5 and Figure 6). Based on de novo genome assemblies for the large colonies of cen10∆-A, cen10∆-B, and cen10∆-E, chromosome 10 fused with chromosome 4, 7, or 1, respectively (Figure 5 and Supplementary file 3). In the large colony of cen10∆-A (cen10∆-A-L1), the fusion occurred between chromosome 10 and chromosome 4 (Figure 5A). Chromosomal fusion led to the loss of the CNBG_10211 gene (on chromosome 4), and the fusion junction was within the CNBG_6174 gene of chromosome 4. For cen10∆-B-L1, the chromosomal fusion occurred between chromosomes 10 and 7 (Figure 5B). Seven genes of chromosome 7 were lost in the fused chromosome. The chromosome fusion junction was intergenic on chromosome 7. cen10∆-E-L1 was due to a chromosomal fusion between chromosomes 10 and 1 (Figure 5C). The fusion was intragenic for both chromosomes. The fusion point occurred in CNBG_6141 on chromosome 10 and CNBG_10308 on chromosome 1.

Because all of the large cen10∆ colonies had chromosome 10 fusions, we examined the fusion location on chromosome 10 in detail. The fusions occurred 1.7, 0.3, and 3.6 kb from the chromosome 10 gene CNBG_6142, respectively (Figure 5—figure supplement 1). The fusion occurred in unique DNA sequences and was not flanked by repetitive regions. The overlapping region between chromosome 10 and the fused chromosome was at most 6 bp, suggesting that these fusions occurred via microhomology-mediated end joining (MMEJ) (also known as alternative nonhomologous end-joining [Alt-NHEJ]).

Chromosome fusion may result in loss of the neocentromeres, and the kinetochore may bind to the native centromere of the fused chromosome and function as the active centromere. This hypothesis was tested by performing ChIP-qPCR for CENP-A binding (Figure 6A). For each neocentromere (either of the cen10∆-A or cen10∆-B mutant), CENP-A enrichment was tested with four primer sets located in the neocentromere. The CENP-A enrichment for these four locations was tested in 1) the initial cen10∆ mutant, 2) a large colony derived from a specific cen10∆ mutant and 3) wild type (Figure 6A). As expected, the ChIP-qPCR analysis showed CENP-A enrichment for the neocentromeres of the initial cen10∆-A and cen10∆-B mutants. The neocentromeric regions of cen10∆-A and cen10∆-B were not enriched with CENP-A in the wild-type strain, showing there was no occupancy by CENP-A prior to neocentromere formation. For all analyzed cen10∆ chromosome 10 fusion isolates, the neocentromeres were not CENP-A-associated, and were similar to the wild-type background levels. Therefore, the neocentromeres were no longer active in the chromosome fusion strains (Figure 6A). This suggests that the native centromere of the fusion partner of chromosome 10 (i.e., chromosome 1, 4, or 7) was the active centromere of the Chr10-Chr1, Chr10-Chr4, and Chr10-Chr7 fusions.

In addition to cen10∆-A, -B, and -E mutants, whole-genome sequencing was performed for two large colonies of cen10∆-C. Although it was not possible to identify the chromosome fusion based on whole-genome sequencing data for either of the large colonies of the cen10∆-C mutant, PFGE analysis showed that cen10∆-C-L1 had a fusion between chromosomes 10 and 13 (Figure 6b). cen10∆-C-L2 had read coverage of 1.99-fold for a region of ~200 kb of chromosome 10 (Figure 4B). The rest of chromosome 10 was euploid, suggesting that the ~200 kb region was duplicated and was either a single chromosome or fused to another chromosome in this isolate. PFGE analysis suggested that this fragment was duplicated on chromosome 10, resulting in a larger chromosome (Figure 6B). In contrast to the other fused chromosomes, this chromosomal fragment did not fuse to a chromosome with a native centromere, and the fact that the mutant still exhibited a fitness defect was consistent with this interpretation. The larger chromosome was euploid, suggesting that the unstable neocentromere(s), rather than causing aneuploidy, resulted in a fitness cost in this isolate.

The native centromeres of C. deuterogattii are found in repetitive regions and are flanked by, but do not contain, protein-encoding genes (Yadav et al., 2018). By contrast, neocentromeres of C. deuterogattii span genes, lack repetitive elements, and like the native centromeres, are flanked by genes. In general (with one exception), the neocentromeres of C. deuterogattii are significantly shorter than the native centromeres, whereas most neocentromeres in other species have similar lengths as the native centromeres.

Native centromeres of S. pombe have a central core that is enriched with CENP-A and flanked by repetitive pericentric regions (Ishii et al., 2008). While neocentromere formation in S. pombe favors repeats in the pericentric regions, neocentromere formation is possible without the repetitive pericentric regions (Ishii et al., 2008). The majority of the neocentromeres in C. albicans and chickens are formed close to native centromeres due to seeding of CENP-A that is located near the native centromere (the so-called CENP-A cloud) (Ketel et al., 2009; Shang et al., 2013). Our results and the earlier reports discussed, suggest that the chromosomal location of the native centromere is the main determinant of neocentromere formation. One exception was the neocentromere of cen10∆-E, which directly flanked the left telomere. Interestingly, this was the only cen10∆ mutant that had a growth rate similar to wild type.

Several C. deuterogattii neocentromeres formed in the same location; however, there is no apparent consensus between the different regions occupied by different neocentromeres. A similar trend has been observed in neocentromere formation in C. albicans (Ketel et al., 2009). Evolutionary new centromeres (ECNs) in the largest crucifer tribe Arabideae originated several times independently and are located in the same chromosomal location (Mandáková et al., 2020). Our results suggest that neocentromeres form by mechanisms that do not rely on nearby transposable elements/repeats to initiate de novo centromere assembly.

Neocentromeres induced in several species can span genes, resulting in silencing or reduced gene expression. For example, all genes within five independent neocentromeres in C. albicans that spanned nine genes were suppressed (Burrack et al., 2016). In S. pombe, neocentromeres span genes that are only expressed in response to nitrogen starvation in the wild-type strain, and neocentromere formation silences these genes during nitrogen starvation (Ishii et al., 2008). The native centromere 8 of rice contains an approximately 750 kb CENP-A-enriched region with four genes that are expressed in both leaf and root tissues of three closely related species (Fan et al., 2011; Nagaki et al., 2004). Neocentromeres of rice span genes that are expressed at similar levels as in the wild type (Zhang et al., 2013). Chicken neocentromeres have been induced on chromosome Z or 5 (Shang et al., 2013). Chromosome Z neocentromeres span eight genes, but in wild-type cells only MAMDC2 is expressed during normal growth. The other seven genes were either not expressed at any detectable level in all tested developmental stages or were only expressed during early embryonic stages (Shang et al., 2013). When a neocentromere formed, expression of the MAMDC2-encoding gene was reduced 20- to 100-fold. Chromosome 5 of chickens is diploid, and neocentromeres on this chromosome span genes that are expressed. The hypothesis behind this phenomenon is that one allele functions as a centromere, while the other allele codes for the genes.

Because the cen10∆ mutants of C. deuterogattii were aneuploid, the expression of genes spanned by chromosome 10 neocentromeres was normalized to expression levels of a housekeeping gene located on chromosome 10. The expression of genes enriched for CENP-A chromatin is similar to that of wild type, and if the allelic hypothesis, like in chickens, were valid, the expectation would be a 60% reduction in expression levels. The genes spanned by neocentromeres of cen9∆ mutants are also expressed at wild-type levels. As the cen9∆ mutants have uniform, wild-type colony sizes, the ploidy levels of these mutants were not tested and we hypothesize that these mutants are haploid/euploid.

Genes contained in regions in which C. deuterogattii neocentromeres formed in cenΔ mutants were actively expressed in the wild-type strain, and this is similar to human neocentromeres that can form in regions with or without gene expression (Alonso et al., 2010; Marshall et al., 2008). However, we have identified chromosomal regions that lack gene expression on chromosomes 9 and 10, although these regions were not close to the native centromere.

Of the C. deuterogattii genes spanned by the neocentromere region, one encodes the kinetochore component CENP-C. Several independent biolistic transformations were performed to delete the gene encoding CENP-C, but all attempts were unsuccessful. This suggests that CENPC is an essential gene. In fission yeast, deletion of the gene encoding the CENP-C homolog Cnp3 was lethal at 36°C, but mutants were still viable at 30°C (Suma et al., 2018). However, CENP-A was mislocalized in the cnp3∆ mutants. Another gene partially located inside a C. deuterogattii neocentromere encodes the serine/threonine-protein phosphatase 2A activator 2 (RRD2). The RRD2 homolog is not essential in S. cerevisiae (Higgs and Peterson, 2005).

Compared with other haploid fungi, the neocentromeric genes of C. deuterogattii are similar to the native centromeric genes of the haploid plant pathogenic fungus Zymoseptoria tritici. Z. tritici has short regional centromeres with an average size of 10.3 kb, and 18 out of 21 native centromeres have a total of 39 expressed genes (Schotanus et al., 2015).

The appearance of two CENP-A-enriched regions of C. deuterogattii cen10∆ mutants could be explained in a few ways. First, neocentromere formation could lead to a dicentric chromosome 10 in which the centromeres may differ in functional capacity. Dicentric chromosomes are not by definition unstable, for example the dominant-negative mutation of the mammalian telomere protein TRF2 results in chromosome fusions, leading to the formation of dicentric chromosomes (Stimpson et al., 2010). The formation of dicentric chromosomes occurred in 97% of the fused mammalian chromosomes, which were stable for at least 180 cell divisions (Stimpson et al., 2010). Several microscopic studies showed that chromosomes with two regions of centromere-protein enrichment are stable (Higgins et al., 2005; Stimpson et al., 2012; Stimpson et al., 2010; Sullivan and Willard, 1998). This suggests that a dineocentric chromosome 10 could be stable in the population. Second, the two CENP-A-enriched peaks could be the result of a mixed population and either due to an unstable primary neocentromere and/or aneuploidy. The primary neocentromere could be associated with the majority of the cells, whereas the secondary CENP-A peak would be only found in a small number of cells (and the primary neocentromere is lost in these isolates). This is reflected by lower CENP-A enrichment for the secondary peak, and the hypothesis of putative dicentrics is due to a mixture of alleles in the population. Third, the neocentromeres could be unstable, which could lead to the formation of two CENP-A-enriched regions with centromere function switching between the regions. However, our data would argue against this latter model. Prior to the ChIP-seq analysis of the cen10∆ mutants, colonies were isolated by streak purification (eight times), suggesting that the presence of two distinct CENP-A peaks occurs continuously.

Neocentromere formation in chickens results in a low number of aneuploid cells (Shang et al., 2013). Based on whole-genome sequencing of a population of cells, the C. deuterogattii cen10∆ isolates are partially aneuploid for chromosome 10. For fully aneuploid isolates, the coverage of Illumina reads is expected to be 2-fold; the cen10∆ isolates with two CENP-A peaks showed aneuploidy levels up to 1.28-fold or were even euploid. This suggests that, like the chicken neocentromeric isolates, only a small number of cells in a population of C. deuterogattii cen10∆ isolates are aneuploid.

In C. albicans, deletion of centromere 5 results in neocentromere formation, and these isolates have fitness similar to the wild-type strain (Ketel et al., 2009). Similar results were reported for neocentromeres in chicken and S. pombe, in which strains with neocentromeres or chromosome fusion have a growth rate similar to the wild-type strain (Ishii et al., 2008; Shang et al., 2013).

If centromere deletions occurred in nature, we hypothesize that the wild type would outcompete all of the cen∆ isolates. The virulence of the cen∆ mutants was not assayed. Based on reduced fitness of the cen∆ mutants, we hypothesize that pathogenicity of the cen∆ mutants would be lower than the wild type. However, when chromosome fusion occurs the growth rate is restored to a near wild-type level and we hypothesize that the isolates with 13 chromosomes could have virulence similar to the wild type. Several genes were lost due to the fusion events in the cen∆ mutants; to our knowledge these lost genes have not been associated with pathogenicity of C. deuterogattii.

Deletion of a centromere in S. pombe leads to either neocentromere formation or chromosome fusion due to a noncanonical homologous recombination pathway (Ishii et al., 2008; Ohno et al., 2016). This is in contrast to neocentromere formation in C. deuterogattii, which results in 100% neocentromere formation. Based on PFGE analysis, the karyotype of the cenΔ isolates is wild type at 30°C, but chromosome fusion can occur at 37°C within the cen10∆ mutants and lead to improved growth at 30°C.

The location of the cen10∆ neocentromere had no apparent influence on the ability to undergo chromosome fusion as we have shown with a telocentric neocentromere, a dicentric mutant, and a neocentromere located 118 kb away from the telomere.

The fused chromosomes have no or only short homology at the breakpoints that is insufficient for homologous recombination, suggesting that the chromosome fusions arise via MMEJ. Future experiments to test this hypothesis could involve deleting genes involved in the MMEJ pathway, such as CDC9 and DNL4 (Sinha et al., 2016).

A prominent chromosome fusion occurred during the speciation of humans. Compared to other great apes, humans have a reduced karyotype, which is due to the fusion of two ancestral chromosomes that resulted in chromosome 2 in modern humans, Denisovans, and Neanderthals (Miga, 2017). Human chromosome2 still harbors signatures of telomeric repeats at the fusion point (interstitial telomeric sequences [ITS]), suggesting that this chromosome is derived from a telomere-telomere fusion. By synteny analysis, the inactive centromere of chimpanzee chromosome 2b can be identified on human chromosome 2, and there are relics of α satellite DNA at this now extinct centromere (Miga, 2017). Moreover, a dominant-negative mutation of the human telomeric protein TRF2 leads to telomere-telomere fusions, mainly between acrocentric chromosomes (Stimpson et al., 2010; van Steensel et al., 1998). In the fungal species Malassezia, chromosome breakage followed by chromosome fusion has led to speciation (Sankaranarayanan et al., 2020). The short regional centromeres (3–5 kb) are fragile and this led most likely to chromosome reduction. By contrast in C. deuterogattii, the chromosomes involved in chromosomal fusion of the cen10∆ mutants were all metacentric, and fusion occurred in non-telomeric sequences.

Another example of telomeric fusions is the presence of ITS regions in several genomes. In budding yeast, the experimental introduction of an ITS into an intron of the URA3 gene resulted in four classes of chromosome rearrangements, including: 1) inversion, 2) gene conversion, 3) mini-chromosome formation due to deletion or duplication, and 4) mini-chromosome formation due to translocation (Aksenova et al., 2013). Based on our de novo genome assemblies of the C. deuterogattii large-colony cen10∆ mutants, chromosome fusions occurred with no signs of chromosome rearrangements. Thus, these chromosome fusions did not produce ITS regions, which would otherwise destabilize the genome.

Our work shows that, like in other model systems, neocentromeres can be induced in C. deuterogattii. However, C. deuterogattii neocentromeres have several unique characteristics, such as spanning genes whose expression is unaffected by centromere assembly. In some instances, deletion of CEN10 led to chromosome fusion, resulting in enhanced fitness and leading to inactivation of the neocentromere. Presumably, deletion of other centromeres could be carried out, leading to a C. deuterogattii strain with only one or a few chromosomes, as was recently reported in S. cerevisiae (Luo et al., 2018; Shao et al., 2018).

Primers are listed in Supplementary file 4. Strains used in this study are listed in Supplementary file 5. All strains were stored in glycerol at −80°C, inoculated on solid YPD (yeast extract, peptone, and dextrose) media, and grown for two days at 30°C. Liquid YPD cultures were inoculated from single colonies of solid media and grown, while shaking, at 30°C overnight.

DNA sequences (1 to 1.5 kb) of the CEN10-flanking regions were PCR-amplified with Phusion High-Fidelity DNA Polymerase (NEB, Ipswich MA, USA). Flanking regions were fused on both sides of either the NEO or NAT dominant selectable marker via overlap PCR, conferring G418 or nourseothricin resistance, respectively. Deletion of C. deuterogattii CEN10 was achieved through homologous recombination via biolistic introduction of an overlap-PCR product as previously described (Billmyre et al., 2017; Davidson et al., 2002). Deletion of CEN9 was performed by CRISPR-CAS9 mediated transformation with two guide RNAs flanking CEN9 and homologous recombination was mediated by the introduction of an overlap PCR product as previously described (Fan and Lin, 2018). Transformants were selected on YPD medium containing G418 (200 μg/mL) or nourseothricin (100 μg/mL).

Subsequently, the 5’ junction, 3’ junction, and spanning PCR and Southern blot analyses were performed to confirm the correct replacement of CEN10 by the appropriate drug resistance marker. To identify centromeres, the gene CNBG_0491, which encodes CENP-A, was N-terminally fused to the gene encoding the fluorescent mCherry protein by overlap PCR, and C. deuterogattii strains were biolistically transformed as previously described (Billmyre et al., 2017). A subset of cen9∆ mutants were biolistically transformed with an overlap PCR product containing CENPC C-terminally fused with GFP. As three cen10∆ mutants have a neocentromere that spans the gene encoding CENPC, a subset of cen10∆ mutants were transformed instead with an overlap PCR product containing MIS12 C-terminally fused with GFP. Both PCR products encoding CENPC-GFP and MIS12-GFP were randomly integrated in the genome and confirmed by a PCR spanning either CENPC-GFP or MIS12-GFP.

Three replicate cultures for seven independent cen10∆ deletion mutants and the wild-type strain were grown in liquid YPD at 30°C overnight. Cells were diluted to an OD₆₀₀ of 0.01 and grown in 50 mL YPD at 30°C. The OD₆₀₀ of the triplicate cultures was measured every two hours with a SmartSpec 3000 (BioRad) until stationary phase was reached (T = 22 hr).

For competition assays, three independent replicate cultures (cen9∆, cen10∆, control, and wild type) were grown overnight in 8 mL YPD. Subsequently, the cell density of the cultures was determined using a hemocytometer. For each independent cenΔ deletion mutant, 500,000 cells were co-cultured in a 1:1 ratio with wild-type cells. After 24 hr, the cultures were inoculated on 1) a YPD plate to determine the total colony-forming units (CFUs) and 2) a YPD plate containing G418 or nourseothricin to calculate the proportion of cen10∆ mutant CFUs compared to the wild-type CFUs. Plates were incubated at 30°C and the colonies were counted after 4 days. The cell morphology of >1000 cells of the wild type and of five cen10∆ mutant strains was analyzed, and the number of elongated cells was quantified (Figure 1—figure supplement 5).

Genomic DNA was isolated using the CTAB protocol and sent to the Duke University Sequencing and Genomic Technologies Shared Resource facility for library preparation and Illumina sequencing. Sequencing was performed with a HiSeq 4000 sequencer, and 150 bp paired-end reads were generated. The resulting DNA sequence reads were trimmed, quality-filtered, and subsequently mapped with Bowtie2 to a complete PacBio, Nanopore-based, Illumina Pilon-corrected, whole-genome assembly of the C. deuterogattii R265 reference genome (version R265_fin_nuclear). Reads were visualized with IGV (Langmead, 2010; Quinlan and Hall, 2010; Thorvaldsdóttir et al., 2013; Yadav et al., 2018). Previously generated RNA sequencing reads (NCBI, SRA: SRR5209627) were remapped to the C. deuterogattii R265 reference genome by HISAT2 according to the default settings (Schneider et al., 2012; Pertea et al., 2016).

Genomes were de novo assembled with Spades using the default conditions (Bankevich et al., 2012). Genome assemblies were confirmed with PCRs using primers flanking the chromosome fusions and the PCR products obtained span the chromosomal fusions (Figure 5—figure supplement 1). The read coverage at chromosome fusions was analyzed and compared to the average read coverage of the contig (Figure 5—figure supplement 1).

ChIP analyses were performed as previously described with minor modifications (Schotanus et al., 2015; Soyer et al., 2015). In short, 500 mL YPD cultures (1000 ml YPD for Mis12 ChIPs) were grown overnight at 30°C, after which 37% formaldehyde was added to a final concentration of 0.5% for crosslinking. The cultures were then incubated for 15 min, formaldehyde was quenched with 2.5 M glycine (1/20 vol), and cells were washed with cold PBS. The crosslinking time of Mis12-GFP tagged isolates was extended to 45 min. Cells were resuspended in chromatin buffer (50 mM HEPES-NaOH, pH 7.5; 20 mM NaCl; 1 mM Na-EDTA, pH 8.0; 1% [v/v] Triton X-100; 0.1% [w/v] sodium deoxycholate [DOC]) containing protease inhibitors (cOmplet Tablets, mini EDTA-free EASYpack, Roche), followed by homogenization by bead beating with a miniBead beater (BioSpec products) using 18 cycles of 1.5 min on and 1.5 min off. The supernatant containing chromatin was sheared by sonication (24 cycles of 15 s on, 15 s off, burst at high level) (Bioruptor UCD-200, Diagenode). Chromatin was isolated by centrifugation, and the supernatant was divided into a sample fraction and a sonication control. The sample fraction was precleared with protein-A beads (1 to 3 hr) and subsequently divided into two aliquots. One tube served as the input control, and a mCherry or GFP antibody (MCherry: ab183628, Abcam, GFP: ab290, Abcam) was added to the remaining half of the sample. The samples were incubated overnight at 4°C and then processed according to a previously published protocol (Soyer et al., 2015). After completing the ChIP experiment, the samples were analyzed by ChIP-qPCR or sent to the Duke University Sequencing and Genomic Technologies Shared Resource facility for library preparation and Illumina sequencing. Samples cen10∆-A, cen10∆-B and cen10∆-C were sequenced with a HiSeq 2500 sequencer, and single reads of 50 bp were obtained. All other ChIP-seq samples were sequenced with a NovaSeq 600 sequencer and 50 bp PE reads were obtained. For each centromere mutant and the wild type, a ChIPed and input sample were sequenced. Reads were mapped to the reference genome, similar to the whole-genome sequencing reads. To analyze the ChIP-seq data the ChIPed sample was normalized with the input sample and visualized with the IGV viewer.

Cells were grown in an overnight culture of 25 mL YPD at 30°C. RNA was isolated with TRIzol LS (Thermo Fisher Scientific) according to the manufacturer’s instructions. Subsequently, cDNA was synthesized with the SuperScriptFirst-Strand Synthesis System (Thermo Fisher Scientific) according to the manufacturer’s instructions. qPCRs were performed in triplicate with Brilliant III Ultra-Fast SYBR Green qPCR Master Mix (Agilent Technologies) on an ABI 1900HT qPCR machine.

Isolation of whole chromosomes and conditions for PFGE and chromoblot analysis were performed as previously described (Findley et al., 2012).