|Home | About | Journals | Submit | Contact Us | Français|
The fission yeast clade, comprising Schizosaccharomyces pombe, S. octosporus, S. cryophilus and S. japonicus, occupies the basal branch of Ascomycete fungi and is an important model of eukaryote biology. A comparative annotation of these genomes identified a near extinction of transposons and the associated innovation of transposon-free centromeres. Expression analysis established that meiotic genes are subject to antisense transcription during vegetative growth, suggesting a mechanism for their tight regulation. In addition, trans-acting regulators control new genes within the context of expanded functional modules for meiosis and stress response. Differences in gene content and regulation also explain why, unlike the Saccharomycotina, fission yeasts cannot use ethanol as a primary carbon source. These analyses elucidate the genome structure and gene regulation of fission yeast and provide tools for investigation across the Schizosaccharomyces clade.
The fission yeast genus Schizosaccharomyces forms a broad and ancient clade within the Ascomycete fungi (Fig. 1A) with a distinct life history from other yeasts (1). Fission yeast grow preferentially as haploids, divide by medial fission rather than asymmetric budding, and have evolved a single-celled lifestyle independently from the budding yeasts (Saccharomycotina). Fission yeasts share important biological processes with metazoans, including chromosome structure and metabolism (relatively large chromosomes, large repetitive centromeres, heterochromatic histone methylation, chromodomain heterochromatin proteins, siRNA-regulated heterochromatin and TRF-family telomere binding proteins), G2/M cell cycle control, cytokinesis, the mitochondrial translation code, the RNAi pathway, the signalosome pathway and spliceosome components with metazoans. These features are absent or highly diverged in budding yeast. In general, core orthologous genes in fission yeast more closely resemble those of metazoans than do those of other Ascomycetes (2). Fission yeasts have also evolved innovations in carbon metabolism, including aerobic fermentation of glucose to ethanol (3). This convergent evolution with the budding yeast Saccharomyces cerevisiae offers insight into the evolution of complex phenotypes.
S. pombe is widely used as a model for basic cell-biological processes and to study genes implicated in human disease. To better understand its evolution and natural history, we have compared the genomes and transcriptomes of S. pombe, S. japonicus, S. octosporus and S. cryophilus, which constitute all known fission yeasts.
We sequenced and assembled the genomes of S. octosporus, S. cryophilus and S. japonicus using clone-based and clone-free whole-genome shotgun (WGS) approaches (Table S1). Each genome is ~11.5 Mb in size. S. octosporus and S. cryophilus are 38% GC; S. japonicus is 44%. By comparison, the S. pombe genome is 12.5 Mb in size and 36% GC. We assembled the S. octosporus and S. japonicus scaffolds into 3 full-length chromosomes of similar quality to the finished S. pombe genome (Figs. 1B, S1, S2 and Tables S2, S3) and identified telomeric sequence using WGS data (4). Telomere-repeats in S. japonicus (GTCTTA), S. octosporus (GGGTTACTT) and S. cryophilus (GGGTTACTT) matched a one and a half repeat-unit sequence at the putative telomerase-RNA locus, similar to the configuration in S. pombe (GGTTAC) (5). Using these motifs, we extended the S. japonicus and S. octosporus chromosomes into subtelomeric and telomeric sequence (4).
We constructed a phylogeny of the Schizosaccharomycetes within Ascomycota (Fig. 1A, S3) from 440 single-copy core orthologs, placing the monophyletic Schizosaccharomyces species as a basal sister group to the clade including the filamentous fungi (Pezizomycotina) and budding yeast (Saccharomycotina). We found an average amino acid identity of 55% between all 1:1 orthologs between S. pombe and S. japonicus, similar to that between humans and the cephalochordate amphioxus (Table S4). For the most closely related species, S. cryophilus and S. octosporus, 1:1 orthologs share 85% identity on average, similar to humans and dogs. The genetic diversity within S. pombe is low. Comparing S. pombe 972 to WGS of S. pombe NCYC132 and S. pombe var kambucha, two phenotypically distinct strains, revealed less than 1% nucleotide difference between the three strains (Fig. S4, Table S5).
Transposons and other repetitive sequences are thought to be crucial for centromeric function through the maintenance of heterochromatin (6). These sequences evolve rapidly, but the evolutionary relationship between centromeres, transposons and heterochromatin is unclear, in part because fungal centromeres have not generally been included in genome assemblies. The S. japonicus genome harbors 10 families of gypsy-type retrotransposons (4) (Figure S5 and Table S6). Sequence divergence of their reverse transcriptases suggests that these transposon families predate the last common ancestor of the Ascomycetes. However, a dramatic loss of transposons occurred after the divergence of S. japonicus; S. pombe harbors two related retrotransposon, Tf1 and Tf2; S. cryophilus has a single related retrotransposon, Tcry1; S. octosporus contains no transposons, but contains sequences related to reverse transcriptase and integrase that may represent extinct transposons (Fig. S5, Table S6).
The disappearance of transposons in the post-S. japonicus fission yeast species correlates with the appearance of the cbp1 gene family, suggesting a transition in the control of centromere function. In S. pombe, Cbp1 proteins bind centromeric repeats and are required for transposon silencing and genome stability (7, 8). Although described as orthologs of CENP-B, a human centromere-binding protein, Cbp1 proteins apparently evolved independently within the Schizosaccharomyces lineage from a domesticated Pogo-like DNA transposase (9). The appearance of the cbp1 gene family also correlates with the switch from RNAi-mediated transposon silencing in S. japonicus (see below) to a Cbp1-based mechanism in S. pombe, suggesting that this shift to Cbp1-based transposon control allowed the eradication of most transposons from the fission yeast genomes, possibly by promoting recombinational deletion between LTRs (8). Furthermore, the cbp1 family is evolving rapidly (Fig. S6), suggesting that Cbp1-based transposon silencing is a Schizosaccharomyces-specific innovation that arose after the divergence of S. japonicus.
The loss of transposons was accompanied by a significant reorganization of chromosome architecture that conserves centromere function, suggesting evolution of novel centromere structures that compensate for the loss of transposons. In S. japonicus, transposons cluster next to telomeres and centromeres, as in metazoans (Fig. 1B,C). In the other Schizosaccharomycetes, the subtelomeres and pericentromeres are also repetitive, but lack transposons (Fig. 1C). However, like S. japonicus, the centromeric and subtelomeric repeats are confined to pericentromeric and subtelomeric regions, respectively, with one exception — a centromeric repeat involved in transcriptional silencing at the S. pombe mating-type locus (10). We confirmed that the centromeres are heterochromatic by histone H3 lysine-9 methylation mapping (Fig. S7), and by showing that the S. japonicus centromeres are functional by meiotic mapping (Table S2).
Although centromeric repeats evolve rapidly, differing even between related strains (11), individual repeat sequences tend to be similar within strains (Fig. 1C). No similarity was observed between the centromeric repeats of S. pombe, S. octosporus or S. cryophilus. However, both S. pombe and S. octosporus centromeres contain repeated elements, highly similar between chromosomes, that are arrayed in a larger inverted repeat structure around a unique core sequence (Fig. 1C), suggesting that they are homogenized by non-reciprocal recombination. This contrasts with a lack of symmetry in S. japonicus, and implies that transposition occurs more rapidly than homogenization by recombination. Thus, the suppression of transposition likely led both to the degeneration of transposon sequences and to the evolution of symmetric centromeric repeats.
Despite the divergence of centromere sequence and of gene order on the chromosome arms, karyotype and pericentromeric gene order are conserved between S. pombe and S. octosporus (Fig. S8). Thus, although gene conversion maintains the similarity of centromeric repeats between the different centromeres, crossover recombination between centromeres is suppressed. We observed neither centromeric translocations nor neo-centromere events within these lineages, despite the fact that centromeres can occur at novel locations in manipulated S. pombe strains. The retention of repetitive elements in the centromeres of (12) S. pombe, S. octosporus and S. cryophilus, even as they have lost their transposons, implies that centromeric repeats have an important function.
Since siRNAs are involved in both transposon silencing and centromere function (13), we investigated these roles in the Schizosaccharomyces lineage. In S. pombe, the centromeric repeats produce dicer-dependent siRNAs required for maintenance of centromeric structure, function and transcriptional silencing via Argonaute-dependent heterochromatin formation (14). However, transposons are silenced in S. pombe by RNAi-independent mechanisms and do not produce abundant siRNAs (Figs. 1B, S9) (7). To investigate whether centromere-directed siRNA production is conserved within the transposon-rich centromeres of S. japonicus, we sequenced small RNAs from log-phase S. japonicus cultures (which have a modal size of 23 nt) (4) and found that 94% map to transposons, both telomeric and centromeric (Figs. 1B, S9). The fact that siRNAs map to transposons in S. japonicus but not in S. pombe suggests that either the fission yeast RNAi pathway targets repetitive sequences instead of mobile elements per se, or that the pathway evolved away from an ancestral role in transposon control to a dedicated role in heterochromatin function.
The structure of the mating-type loci and the cis-acting elements that regulate mating-type switching is highly conserved across all four species (Fig. S10). The expressed mat1 locus can contain either the plus (P) or the minus (M) allele and switches between the two by epigenetically-programmed gene conversion (15–17) from two heterochromatically-silenced donor cassettes: mat2-P and mat3-M (Figs. S10, S11). cis-acting regulatory sequences required for epigenetic imprinting and recombinational switching (18–20)are conserved (4) (Fig. S11), as is the epigenetically-programmed genomic mark associated with mat1 (15).
In contrast, none of the cis-acting sequences involved in transcriptional repression of the silent cassettes in S. pombe are identifiable in the other species, although the donor cassettes are enriched for H3K9me heterochromatin (Fig. S7). In S. japonicus, the silent mat loci directly abut the centromere of Chromosome 3, suggesting that they may be silenced by a positional effect. In S. octosporus and S. cryophilus, the mat loci are distant from the centromeres, but each contains a conserved region of transposon remnants, which may be silencing triggers. Interestingly, they also contain inverted repeats, albeit shorter and less similar to each other than the inverted repeats that flank the mat2/3 locus in S. pombe (21). Thus, their silencing strategies may share elements from both S. pombe and S. japonicus. These results suggest that the mechanisms of imprinting and switching have been conserved, but that the strategies for establishing heterochromatin are plastic.
We annotated the three genomes using standard methods and compared them with S. pombe (4). We then deep-sequenced polyA-enriched, strand-specific cDNA (22–24) (RNA-Seq), and constructed de novo transcript models (Fig. S12) for log phase, glucose depletion, early stationary phase and heat shock from S. pombe, S. octosporus, and S. japonicus and log phase, glucose depletion and heat shock from S. cryophilus.
In S. pombe, we reconstructed 4277 out of 5064 previously annotated genes; of the remaining 788 genes, 60% were covered over at least 90% of their length. 400 of our transcript models change coding exon structure of the gene, 95% of which maintained or improved conserved coding capacity (Table S7, S8 and Figure S12) (25). In addition, we identified 253 UTR introns. Lastly, we found 89 new protein-coding genes in S. pombe, 53 of which are conserved (Table S7 and Figure S13). We found no evidence that intron-rich fission yeast genes engage in metazoan-like alternative splicing (26). We found evidence for 433 alternative splicing events in S. pombe in the form of intron retention and alternative splice-donor or -acceptor usage, but no evidence of exon skipping or alternative exons; we found similar levels of splice variants in the other species (Table S9). However, since many of these variants disrupt the coding capacity (Fig. S14, S15) and only a minority of intron-retentions (146/393) are conserved between two or more species, we suspect that much of alternative splicing in fission yeast represents nonproductive splicing variants. Interestingly, in some cases the non-spliced variant may be the protein-coding isoform (Figure S14C,D, S15, Table S10).
The majority of stable fission yeast transcripts originate from annotated protein coding genes. Most of the S. pombe genome is transcribed (22) with 91% of nucleotides covered by at least one RNA-Seq read. However, most transcription, as measured by steady-state polyA-enriched RNA levels, is associated with well-defined transcripts, most of which are protein coding. Specifically, 37% of intergenic nucleotides (between the UTRs of annotated protein coding transcripts) are not detectibly expressed and 90% of transcribed intergenic nucleotides account for only 0.16% of the polyA-enriched transcript signal. Moreover, the median expression level of exonic sequence (99.1% of which are detectably expressed) is 305-fold higher than that of intergenic sequence (Fig. 2, Table S11), with intergenic transcription enriched within origins of DNA replication (Fig. S16) – gene-free loci with nucleosome-free regions (27–29) that may provide permissive loci for ectopic transcriptional initiation (30).
Transcription of coding genes is heavily biased to the sense strand. Of coding genes, 73% have less than 5% of their RNA-Seq reads on the antisense strand. Genes with more than 5% antisense reads are enriched for convergent transcripts with intergenic distances of less than 200 bp (p < 10−8, hypergeometric test), but not with those of greater than 200 bp (p > 0.1), suggesting that much antisense transcription is due to read through of 3'-termination sites (31) (Fig. S17). Thus, stable transcripts in fission yeast genomes are primarily associated with known transcription units. We discuss notable exceptions below.
Despite the evolutionary breadth of the fission-yeast clade, as measured by amino-acid divergence, their gene content and structure are remarkably conserved. Of ~5000 coding genes in fission yeast species, 4218 are 1:1:1:1 orthologs across the clade, with the remainder of the orthologous groups containing genes that have been duplicated or deleted since their last common ancestor (Tables 3, S12). Protein kinases are even more conserved in gene content; 93% (102/110) of S. pombe protein kinases are 1:1:1:1 orthologs (4). Moreover, of 3601 S. pombe introns in 2616 spliced 1:1:1:1 orthologs, 2901 (81%) are identical across the four species (Table S13). The majority of changes are due to the gain of species- and clade-specific genes (Table S12, S14) (4). Overall, the conservation of gene content, gene order and gene structure within Schizosaccharomyces is higher than expected given the level of amino acid divergence. From amino acid divergence, we estimate that the fission yeast clade arose about 250 million years ago (Fig. S3). However, the conservation of gene content is significantly higher than that within Saccharomyces or Kluveromyces, both of which have much lower amino acid divergence (Table S15), suggesting that fission yeast amino acid sequences are evolving anomalously quickly, or that genome structures are unusually stable.
The majority of gene changes are due to the gain of species- and clade-specific genes (Table S12). We tested whether gene gain is due to rapid divergence of orthologous genes by looking for co-linearity in regions with species-specific genes, and examined these regions for signs of sequence similarity. We found that 94/317 S. pombe-specific genes are in the same position relative to neighboring genes as genes specific to other species (Table S16). Of these, 9 show greater than 15% identity to a cognate gene in another species, suggesting that they are rapidly diverged orthologs (4).
We also found 34 S. pombe candidates for horizontal gene transfer from bacteria, including two published examples (4, 32, 33) (Table S17), and similar numbers in the other species. Of these, 16 appear to have occurred before the radiation of the clade, and 9 appear to be specific to S. pombe.
We identified 1097 putative transcript models in S. pombe supported by strand-specific RNA-Seq data but containing no obvious coding capacity and having no correspondence to well-defined non-coding RNAs (22, 24, 34) (Fig. S18, Tables 2, S18, S19). Of these potential ncRNAs, 449 are intergenic and 648 are antisense, overlapping a coding gene on the other strand by at least 30%. 213 of the ncRNAs overlap an annotated UTR on the same strand, suggesting that they may be alternative UTRs. Nevertheless, the data support 338 of the intergenic and 546 of the antisense ncRNAs as distinct transcripts (4).
Of the 338 distinct intergenic ncRNAs in S. pombe, 138 are conserved in location in at least one other species (Table 41). Moreover, 26 of the intergenic ncRNAs are conserved in sequence and of these, 9 are conserved in both location and sequence, suggesting they represent potentially biologically important noncoding RNAs. The transcripts that are conserved in location but not in sequence may represent functional transcripts that have diverged beyond recognition. Of the antisense transcripts, 328 (51%) are conserved across two or more genomes (Table S20), suggesting that they are biologically significant (35).
Across fission yeast, the ~250 genes with greater antisense transcription than sense transcription (Table S21) are significantly enriched for meiotic genes (p = 10−10 for S. pombe, hypergeometric test) (Fig. 3, S19, Table S22, S23), consistent with observations in S. pombe and S. cerevisiae (24, 35). Several antisense-transcribed genes have been proposed to be regulated by intron retention (36, 37), however these studies did not use strand-specific approaches, making it impossible to distinguish unspliced sense transcripts from antisense transcripts. We find no evidence of alternative splicing of any of these genes.
Antisense transcription of meiotic genes does not uniformly decrease as cognate sense transcription increases during meiosis (Fig. S20). This observation suggests that antisense transcription does not inhibit sense transcription, in contrast to the anti-correlation observed in S. cerevisiae (30, 35). Furthermore, meiotic genes are not enriched among genes with greater than 5% antisense transcription but less than 100% antisense transcription (p = 0.47, hypergeometric test), consistent with a stoichiometric mechanism of regulation in which antisense transcripts directly bind to and inhibit the stability or translation of sense transcripts.
To identify conserved modules of co-expressed genes, we examined expression patterns across the four conditions and between the four fission yeast with phylogenic clustering (Fig. 4). We found that patterns of gene expression between species grown in similar conditions are generally conserved, with dominant patterns associated with growth (log and heat shock) and stress (glucose depletion and early stationary phase). Moreover, similar expression clusters are enriched for similar gene annotations across the species.
Fission yeast up regulate genes involved in mitosis, including those involved in the kinetocore, the spindle pole body and the anaphase-promoting complex, in response to glucose depletion (Table S24). In contrast, several classes of genes involved in growth are down regulated (4). None of these genes is significantly regulated in glucose depletion in S. cerevisiae (38).
Promoter motifs with conserved regulatory function across Ascomycota show new functionality among the Schizosaccharomyces. For example, the motif bound by Rtg3 in S. cerevisiae is associated with amino acid metabolism genes across the phylum. In fission yeast however, it is also enriched in genes responsive to various stress responses (Fig. 5A). Of the stress genes that have Rtg3 motifs in S. pombe, 36% are found only in the Schizosaccharomyces clade, and many are also associated with the Atf1 motif, a conserved regulator of the stress response (Fig. 5B). Rtg3 does not have a detectable ortholog in the Schizosaccharomyces clade (39), but the motif recognized by Rtg3 in S. cerevisiae is clearly identifiable in fission yeast, suggesting that these regulatory motifs are more conserved than their binding proteins. We also found a similar acquisition of Schizosaccharomyces-specific genes by the Fkh1- and MBF-associated motifs, which regulate meiotic transcription in S. pombe (4, 40, 41). In particular, these two motifs were enriched in genes with antisense transcripts (Figure 5C). Most of the Fkh1/Mei4 target genes with antisense transcripts (80%, 47 genes) are meiotic genes, the majority of which are specific to the Schizosaccharomyces clade (Figure 5C).
Fission yeast and budding yeast of the Saccharomyces clade independently evolved the ability to produce ethanol by aerobic fermentation (3, 42). In contrast to the convergent evolution of ethanol production, the utilization of ethanol has not converged; although budding yeast can efficiently catabolize ethanol, fission yeast cannot use ethanol as a primary carbon source. The evolution of aerobic fermentation in budding yeast involved changes in gene content, most notably following a whole genome duplication (WGD) event, and in regulatory mechanisms of glucose repression (3, 43).
Like budding yeast, fission yeast have duplicate copies of the pyruvate decarboxylase (pdc) gene, needed to funnel pyruvate to fermentation. They also have orthologs of several activators and repressors of respiratory genes, including Hap2/3/4/5 complex members, the Adr1, Tup and Mig transcriptional regulators, and the Snf1-Sip1/2 kinase (3). However, there are substantial distinctions in gene content between fission yeast and the post-WGD budding yeast (Fig. S21). We identified loss of the glyoxylate cycle, loss of the glycogen biosynthesis, fewer glycolytic paralogs, loss of the gluconeogenic enzyme phosphoenolpyruvate carboxykinase, lack of expanded adh genes, and lack of transcriptional regulators of glucose repression as differences that illuminate the distinct metabolic capacities of fission yeast (4). All of these adaptations are consistent with the inability of fission yeast to consume ethanol as a sole carbon source. The loss of conserved enzymes highlights how fission yeast came to depend solely on glucose.
In both fission yeast and budding yeast, as glucose is depleted the expression of respiratory genes (oxidative phosphorylation enzymes, TCA cycle) is induced. However, unlike S. cerevisiae (38), in fission yeast the expression of the genes encoding the pyruvate dehydrogenase complex and adh1 are reduced, preventing the efficient use of pyruvate for respiration. Instead, the expression of the ald genes is induced, which may provide an alternative mechanism for generating acetyl-coA in fission yeast.
Thus, the lack of efficient ethanol catabolism by fission yeast demonstrates that aerobic fermentation did not evolve to create a consumable by-product. Instead, ethanol is a waste product, possibly produced because it is toxic to competing micro-organisms. Interestingly, aerobic fermentation appears to have evolved as early as 200 million years ago in fission yeast (Fig. S3), long before the WGD and subsequent evolution of aerobic fermentation in budding yeast.
Our comparative analysis of genome structure and expression in the fission yeast, especially the analysis of centromere structure and evolution, demonstrates how chromosomal features can be rearranged while retaining function and maintaining stable positions across taxa. We also provide insight into centromeric biology and elucidate conserved antisense transcription that may play a systematic role in meiotic gene regulation. Lastly this study informs on the major evolutionary innovation of aerobic alcohol fermentation in microbial metabolism that arose in parallel in the fission yeast and budding yeast lineages. As these results demonstrate, comparative analyses improve the power of fission yeast as a model for eukaryotic biology.