A SIR-independent role for cohesin in subtelomeric silencing and organization

To test whether cohesin affects telomere-proximal silencing, a temperature-sensitive (ts) allele of the MCD1 gene, mcd1-1, was introduced into a silencing reporter strain, CCFY101 (a kind gift from Kurt Runge, Cleveland Clinic, Cleveland), having the URA3 reporter inserted adjacent to the right telomere of chromosome V (Fig. 1A). The mcd1-1 mutant is viable at 23 °C but displays a strong ts phenotype resulting in inviability at 37 °C (7). Only a very slight growth defect is observed at 31 °C (Fig. 1A, Top), but at this semipermissive temperature almost complete abrogation of silencing of the telomere-proximal reporter URA3 was observed, as evidenced by failure of the mcd1-1 silencing reporter strains to grow on plates containing 5-fluoroorotic acid (5-FOA) that selects against Ura3⁺ cells (Fig. 1A, Bottom). We further confirmed this result in two other telomere silencing reporter strains, ROY783, harboring a synthetic telomere on chromosome VR having URA3 inserted close to the telomere (21) (SI Appendix, Fig. S1A), and YEF452 (38), having URA3 near the telomere of the left arm of chromosome XI (SI Appendix, Fig. S1B). Once again, the telomere VR-proximal URA3 was strongly derepressed at 31 °C but not at 23 °C, while the chromosome XIL–proximal URA3 was only mildly derepressed in mcd1-1 at 31 °C when assayed on 5-FOA plates (SI Appendix, Fig. S1 A and B, respectively). We also assayed RNA levels of the telomere-proximal URA3 reporter in derivatives of CCFY101 that were modified such that the endogenous URA3 locus is replaced by KanMX4 and these strains hence completely lack sequences homologous to URA3 elsewhere in the genome. Real-time PCR analysis of RNA isolated from wild-type or mcd1-1 derivatives of these strains showed an increase (3.58-fold, P = 0.001) in the levels of URA3 transcripts in the mutant (Fig. 1A, Bottom), confirming that enhanced 5-FOA sensitivity detected in the growth assays may indeed be due to an increase in URA3 transcripts resulting from derepression of transcription of the reporter and not due to differential sensitization of the mutant to 5-FOA due to altered nucleotide metabolism (39).

We also examined the effect of the mcd1-1 mutation on silencing of naturally occurring subtelomeric genes. We compared the transcript levels of one such subtelomeric gene located near a SIR-binding peak in the genome (YFR057W), whose silencing was previously reported to be abrogated significantly in sir mutants (36). Real-time qPCR analysis of RNA from wild-type or mcd1-1 cells grown at 31 °C showed a substantial increase in transcripts corresponding to YFR057W in the mutant (Fig. 1B, Top) (59.4-fold, P = 0.0007), while only a modest increase (2.43-fold) was observed at the permissive temperature (23 °C) (SI Appendix, Fig. S1C). Likewise, significant derepression of YFR057W transcripts was also observed in smc3-1 (Fig. 1B, Bottom) (11.3-fold, P = 0.0028), a mutant defective in another subunit (Smc3) of cohesin, indicating that the observed silencing defect is not limited to dysfunctional Mcd1 in mcd1-1 but represents a more general role of the cohesin complex.

While mapping the transcriptional landscape of budding yeast chromosome ends, it was reported that genes present within ∼20 kb of chromosome ends are expressed at a lower level compared with the rest of the genome, but only a small subset (∼6%) of these subtelomeric genes are repressed in a SIR-dependent manner (36, 37). We examined expression levels of two other subtelomeric genes, SEO1 and YHL044w, that were also derepressed in mcd1-1 (SI Appendix, Fig. S1D) (24.10-fold, P = 0.0067; 7.43-fold, P = 0.0004) and smc3-1 (SI Appendix, Fig. S1E) (3.84-fold, P = 0.0304; 3.06-fold, P = 0.0005). The expression of these genes was unaffected in sir2Δ here (SI Appendix, Fig. S1F) and in the earlier study (36). We also found that asynchronous mcd1-1 cells showed a cell-cycle distribution profile indistinguishable from wild type at 31 °C (SI Appendix, Fig. S2A) and derepression of subtelomeric genes was also observed in nocodazole-arrested mcd1-1 cells (SI Appendix, Fig. S2B), confirming that the changes in gene expression were not an indirect consequence of cell-cycle arrest in the mutant. G1-arrested wild-type and mutant cells had comparable YFR057w transcripts (SI Appendix, Fig. S2C), indicating that cohesin does not repress subtelomeric gene expression in G1, possibly because Mcd1 is limiting in G1.

We extended the differential gene expression analysis comparing mcd1-1 with wild type at a genome-wide scale by RNA sequencing. Approximately 42% of genes located within 20 kb of telomeres exhibited significantly increased levels of transcripts (≥2-fold; Dataset S1), whereas only ∼10% of genes located elsewhere in the genome were similarly derepressed (Fig. 1C and SI Appendix, Table S1). In contrast, only 0.69% of subtelomeric genes were down-regulated (Dataset S1), whereas 5.21% of telomere-distal genes were down-regulated (Fig. 1C and SI Appendix, Table S1); a majority of the telomere-distal genes were unaffected.

We also compared the observed fraction of derepressed subtelomeric genes in the mutant with the genome-wide average using the χ² test. Notably, genes repressed by cohesin are significantly (at significance level α = 0.0001) enriched within subtelomeric regions between 4 and 24 kb from the chromosome ends (Fig. 1C, Bottom and SI Appendix, Table S2). Repression of genes within 4 kb from the end is also impaired (18.57% relative to the genome-wide average of 11.44%, α = 0.05; SI Appendix, Table S2) in mcd1-1, albeit to a lesser extent. When we analyzed the effects on gene expression with respect to position across individual chromosomes, we observed enrichment of clusters of derepressed genes in the subtelomeric regions in mcd1-1, as shown for several representative chromosomes (Fig. 1D) and all chromosomes (SI Appendix, Fig. S3). Notably, down-regulated genes in mcd1-1 were scarce near chromosome ends; more strikingly, clusters of repressed genes were absent near ends (Fig. 1D and SI Appendix, Fig. S3).

Formation of silent chromatin at telomeres depends on the SIR proteins Sir2, Sir3, and Sir4 (29). To rule out the possibility that expression of SIR proteins may be limiting in mcd1-1, resulting in derepression of silencing, we compared their steady-state levels in cell lysates by western blotting and found comparable levels of Sir2, Sir3-6HA, and Sir4-9myc in wild-type and mutant cells (SI Appendix, Fig. S4A). We then assayed the association of SIR proteins with YFR057w, one of the native subtelomeric genes that was strongly derepressed in the mutants (Fig. 1B), by chromatin immunoprecipitation (Fig. 2A). Surprisingly, there was no significant reduction in binding of Sir2, Sir3, or Sir4 to the YFR057w ORF (Fig. 2A) in mcd1-1. The binding of Sir2 to a site upstream (1,456 bp) and downstream (870 bp) of YFR057w (or 2.5 kb and 6 bp from the end, respectively) was unaltered (SI Appendix, Fig. S4B, Left graph). Binding and spreading of Sir3 1,456 and 3,126 bp upstream of YFR057w (or 2.5 kb and 4 kb from the end, respectively) was also not significantly altered (SI Appendix, Fig. S4B, Right graph). In addition, no defect in Sir2 binding to YFR057w was observed in smc3-1 (SI Appendix, Fig. S4C).

We also examined telomere silencing-associated histone marks at YFR057w in mcd1-1. Not much increase in H4K16ac, a hallmark of euchromatin, was observed in mcd1-1 (1.69-fold), whereas sir2Δ (defective in deacetylation of H4K16) showed a substantial increase in H4K16ac relative to wild type (5.55-fold) (Fig. 2B, Middle and SI Appendix, Fig. S4D, respectively). In addition, mcd1-1 showed no change in H3K4me3, an active chromatin mark (Fig. 2B, Left), and not much increase in H3K79me2 (1.45-fold) (Fig. 2B, Right). However, association of cohesin with the CAR (25) about 0.5 kb from the end of telomere VIR, near YFR057w, was decreased by 38% in the mutant (SI Appendix, Fig. S4E). Interestingly, this was accompanied by enhanced binding of Cet1 (by 43%) (SI Appendix, Fig. S4F), a component of the active RNA Pol II complex to YFR057w, suggesting that enhanced transcript levels of these genes may arise in part from enhanced transcription as a result of increased RNA Pol II occupancy to some extent.

Since we did not observe any difference in association of Sir proteins with derepressed subtelomeric genes, and Sir proteins were silencing-competent in mcd1-1 (SI Appendix, Fig. S5 and Text), we wondered whether the role of cohesin in subtelomeric silencing could be Sir-independent. To test whether the silencing defect in mcd1-1 was indeed SIR-independent, we compared the transcript levels of YFR057w in mcd1-1 sir2Δ and mcd1-1 sir4Δ double mutants with mcd1-1, sir2Δ or sir4Δ, and wild-type cells. We found that YFR057w transcripts were significantly more abundant in the double mutants relative to either of the sir mutants, mcd1-1, or wild-type cells (Fig. 2 C and D). Likewise, comparison of YFR057w transcript levels in the smc3-1 sir2Δ double mutant with smc3-1, sir2Δ, and wild-type cells also yielded similar results (Fig. 2E), confirming that cohesin can indeed contribute to subtelomeric silencing independent of Sir proteins. Similar results were obtained with yku70, a mutant defective in Yku70 that participates in the SIR-mediated silencing pathway by recruiting silencing proteins (SI Appendix, Fig. S6 and Text).

Our interesting yet confounding observation of a disruption of subtelomeric silencing uncoupled from defects in SIR recruitment and formation of histone activation marks in the cohesin mutant resembles earlier observations made for subtelomeric derepression during the stress response (40, 41). A hallmark of rapamycin (that induces stress by targeting target of rapamycin kinases)-induced derepression of telomeric silencing is Mpk1-mediated Sir3 hyperphosphorylation (41). To test whether a similar stress response may be activated in mcd1-1, we examined Sir3 phosphorylation. Indeed, mcd1-1 cells had enhanced slow-migrating electrophoretic forms of Sir3 similar to the rapamycin-induced hyperphosphorylated forms (Fig. 3A) (41) whose accumulation was Mpk1-dependent (Fig. 3B). In addition, the phosphorylated form of Mpk1 was more abundant in the mutant, indicating its activation in mcd1-1 (Fig. 3C). Mutant cells exhibited other properties of Mpk1-mediated stress signaling, such as derepression of PAU3 (Fig. 3D), a member of the subtelomeric PAU gene family, encoding cell-wall mannoproteins. Deletion of MPK1 in mcd1-1 only slightly diminished YFR057w transcripts relative to mcd1-1 (Fig. 3E). These observations indicate that while Mpk1-mediated signaling was activated in mcd1-1, it only contributes to derepression to a very limited extent; there was a substantial additional contribution of cohesin to subtelomeric silencing, independent of the stress pathway-induced Sir3 hyperphosphorylation.

Given our failure to detect significant changes in molecular composition of subtelomeric heterochromatin, we examined the effect of cohesin dysfunction on telomere tethering. The sister-chromatid tethering function of cohesin near a telomere and at an arm site was indeed impaired in mcd1-1 at 31 °C (SI Appendix, Fig. S7 A and B and Text). To explore additional potential mechanisms for derepression of telomere-proximal genes, we examined telomere tethering to the nuclear envelope. The mcd1-1 mutation was integrated in a strain having a Tel-VR–proximal GFP-based chromosomal tag (42) that also expresses Nup49-GFP that illuminates the nuclear envelope, facilitating precise delineation of the nuclear boundary (Fig. 4A). The percentage of nuclei having tethered versus untethered configuration was enumerated in G2/M-arrested cells at 31 °C; G1 cells and anaphase cells were excluded from the analysis. We found a modest but statistically significant increase in untethering of Tel-VR from the nuclear envelope in mcd1-1 (from 26% in the wild type to 41% in the mutant) (Fig. 4A, Right). Hence, our findings indicate that cohesin contributes to telomere cohesion and nuclear membrane tethering of telomeres.

Cohesin also has a role in chromosome condensation (2, 7). Earlier studies demonstrated that telomeres have a special compact organization particularly refractory to DNA-modifying enzymes (30). We compared compaction in wild-type and mcd1-1 cells by quantifying Dam-methylase (an enzyme that methylates adenine within a GATC sequence) accessibility of subtelomeric chromatin near Tel-VIR (30). Subtelomeric chromatin (at YFR057w and SEO1) in the mutant appeared to be more accessible as it was more susceptible to cleavage by Dpn1, a methylation-dependent restriction enzyme, indicating a relatively open organization; the same region was equally sensitive to Sau3A1, a methylation-insensitive restriction endonuclease (Fig. 4B and SI Appendix, Fig. S7C). These findings indicate that subtelomeric chromatin in mcd1-1 assumes a more open conformation. Interestingly, the same region was even more accessible in the mcd1-1 sir2 double mutant, indicating their independent contributions to closed chromatin configuration near telomeres (Fig. 4B). A telomere-distal euchromatin region (at YFL015c) was also more accessible in mcd1-1 (SI Appendix, Fig. S7D), indicating that open configuration in the mutant is not limited to subtelomeric regions.

Assembly of heterochromatin domains occurs by a conserved mechanism that involves action of histone-modifying silencing proteins and the association of silencing proteins with the modified histones (43). In budding yeast, telomere silencing depends upon the SIR complex that associates with telomeres and silenced subtelomeric regions and forms a heterochromatin domain (29). Interestingly, two genome-wide studies revealed that while the zone of subtelomeric silencing extends to ∼20 kb from the end, only a very small fraction (∼6%) of telomere-proximal repressed genes were silenced in a SIR-dependent manner (36, 37). The mechanism for repression of genes in the zone not overlapping with Sir-binding regions is poorly understood. Here, we found that repression of subtelomeric genes (both SIR-dependent and -independent) is decreased in cohesin mutants, indicating a requirement for cohesin in subtelomeric silencing. Silencing at the mating loci HMR and HML, as well as ribosomal DNA, was also impaired in the cohesin mutant, albeit to a lesser extent (SI Appendix, Fig. S8 and Text). This contrasts with an earlier observation on the role of cohesin in silencing in budding yeast at the silent mating locus HMR, where it was inferred that cohesin hinders the establishment of silencing at HMR (24). Our findings establish cohesin as an important determinant of repression in the extensive zone of subtelomeric heterochromatin that is SIR-independent and extends beyond SIR complex-binding regions.

Derepression of subtelomeric genes in the cohesin mutant was observed in SIR-binding as well as nonbinding regions. Interestingly, based on genetic interactions with sir mutants, we found that the contribution of cohesin to telomere silencing is largely SIR-independent. SIR proteins retain their association with desilenced subtelomeric loci (e.g., YFR057w) in the cohesin mutants. This contrasts with observations in Schizosaccharomyces pombe, where loss of silencing was observed in a region located between telomere heterochromatin and arm euchromatin (that has combined eu/heterochromatin histone marks) in the rad21 mutant (44), but in this case the fission yeast silencing protein Swi6 and a silencing-associated histone mark, H3K9me, were also observed to be depleted in this region. Unexpectedly, in our study, an active chromatin-associated histone mark, H3K4me3, is unaltered in the cohesin mutant in the derepressed subtelomeric regions relative to wild-type cells, and H4K16ac and H3K79me2 are also only marginally increased in the mutant. The requirement of cohesin for subtelomeric repression in both budding and fission yeasts suggests that this may be a general role of cohesin in many eukaryotes.

Enhanced transcription of subtelomeric genes in the cohesin mutant without a change in binding of silencing proteins or histone modifications is intriguing. Interestingly, there is some precedence for activation of subtelomeric genes without a decrease in Sir protein association or alterations in histone modifications. For example, a silencing defect in the absence of depletion of Sir proteins or silencing-associated histone modifications was noted in the case of activation of a heat shock-inducible transgene system (40) where >200-fold activation of transcription occurred without much change in Sir3 binding. In the same study, transcription activation (150-fold) of the natural subtelomeric gene YFR057w by the drug cycloheximide occurred without substantial changes in histone occupancy or modification. Other instances of activation of a subtelomeric gene unaccompanied by changes in binding of silencing proteins have been reported in the case of transcriptional activation of a subtelomeric URA3 reporter by growth of cells in medium lacking uracil (45), where there was an increase in H4 acetylation while Sir3 and Rap1 binding was retained. Interestingly, Kitada et al. (46) reported that the epigenetically variegated ON vs. OFF states of the reporter do not differ with respect to Rap1 or Sir3 binding or deacetylation of H4K16.

Since Sir protein binding at derepressed telomeres was unaltered, we wondered whether there may be a change in a posttranslationally modified form of the silencing proteins. It has been reported that Mpk1-mediated Sir3 phosphorylation is induced upon stress or rapamycin treatment, which correlates with reduced subtelomeric silencing (41). Indeed, Sir3 was hyperphosphorylated in the cohesin mutant in an Mpk1-dependent manner, indicating that the Mpk1-mediated stress pathway was activated in this mutant. However, our findings indicate that while elevated Sir3 phosphorylation contributes to the silencing defect in the cohesin mutant, this is to a limited extent.

Cohesin’s role in multiple chromosomal processes depends on its association with chromosomes and its ability to tether DNA molecules. Telomere cohesion was indeed impaired in our study under conditions where telomeric silencing was reduced in the cohesin mutant. In this study, we also examined cohesin’s role in telomere localization to the perinuclear compartment that favors silencing. Yeast telomeres cluster at the nuclear periphery and are anchored to the nuclear envelope (32, 33). The nuclear membrane-adjacent peripheral nuclear space has been described as a subcompartment of the nucleus that favors silencing of genes localized in this region (34, 35). Given our failure to detect changes in SIR binding and gene activation-specific histone marks, we examined the effect of cohesin dysfunction on telomere tethering to the nuclear envelope to explore additional potential mechanisms of derepression. We found that telomere tethering to the nuclear envelope was moderately impaired in the cohesin mutant, indicating a role for cohesin in telomere tethering that may influence telomere-proximal gene repression.

In budding yeast, cohesin also helps in chromosome arm compaction or condensation. Cohesin’s role in mitotic chromosome condensation was first revealed by characterization of the mcd1 mutant (7) that showed defects in both cohesion and condensation. Recent genome-wide studies analyzing the conformation of mitotic chromosomes in yeast (2) have also revealed the role of cohesin in compaction via formation of intra-arm loops. In this study (2), the scc1-73 mutant displayed fewer loops and diminished short-distance (<100-kb) contacts indicative of defective mitotic compaction at the restrictive temperature. Our current study reveals that cohesin dysfunction is also associated with a less compact subtelomeric chromatin organization. Interestingly, the effect of cohesin on compaction of subtelomeric domains was also independent of SIR-mediated compaction in the same region. These defects in organization of repressed subtelomeric chromatin domains are likely to contribute to the transcriptional derepression of subtelomeric genes observed in the cohesin mutants (as depicted in the model; Fig. 4C).

The proposed model for cohesin-mediated subtelomeric gene repression based on our findings is shown in Fig. 4C. In wild-type cells, cohesin-mediated compaction may not only shield subtelomeric genes, rendering them inaccessible to the transcriptional machinery, but may also place them within the peripheral nuclear subcompartment that favors silencing. In mutant cells, the cohesin dysfunction-associated condensation defect may open up chromatin organization, making genes more accessible to the transcriptional machinery, and may also result in their relocation outside the peripheral silent compartment. In the case of chromosome ends that remain tethered to the nuclear envelope, telomeric genes may remain silent while subtelomeric genes may become derepressed (shown for the chromosome on the left in the mutant; Fig. 4C). Our results also reveal a modest impairment of telomere tethering to the nuclear envelope in the cohesin mutant (Fig. 4A). Untethering of the chromosome end from the envelope coupled with decompaction may derepress telomeric genes as well as subtelomeric genes (shown for the untethered chromosome on the right in the mutant; Fig. 4C). This is consistent with our results summarized in the graph in Fig. 1C and SI Appendix, Table S2 that show only a modest impairment of repression of telomeric genes that are within 4 kb from the end (18.57% for telomeric genes compared with the genome-wide average of 11.44%, at significance level α = 0.05; SI Appendix, Table S2) and a substantial derepression of subtelomeric genes (ranging from 30 to 60% for genes within 4 to 24 kb from the end, at significance level α = 0.0001; SI Appendix, Table S2) in the mutant. The effects on expression of nontelomeric regions are not depicted here, although cohesin’s role in compaction is not limited to subtelomeric regions.

In conclusion, our work reveals an alternative way by which cohesin can modulate gene expression. The requirement for cohesin in subtelomeric silencing suggests another mechanism by which the expression of multiple (subtelomeric) genes can be altered in a cohesin mutant. We propose that cohesin contributes to repression of subtelomeric genes by chromatin compaction that renders them inaccessible to the transcriptional machinery, and positions these genes that are adjacent to tethered chromosome ends within the peripheral silent compartment. Our findings establish cohesin as an important determinant of extended subtelomeric gene repression that extends up to 20 kb from budding yeast chromosome ends, well beyond the zone of Sir binding, and provide evidence for a Sir-independent mode of repression mediated by cohesin over extended chromosomal domains.