An acetyltransferase-independent function of Eso1 regulates centromere cohesion

The lethality of eso1 mutants is due to activation of the spindle assembly checkpoint and is triggered by premature centromere separation. This novel role for Eso1 at the centromeres is independent of Smc3 acetylation, and the defect arises in an Smc5/6-dependent manner.

Abstract

Eukaryotes contain three essential Structural Maintenance of Chromosomes (SMC) complexes: cohesin, condensin, and Smc5/6. Cohesin forms a ring-shaped structure that embraces sister chromatids to promote their cohesion. The cohesiveness of cohesin is promoted by acetylation of N-terminal lysines of the Smc3 subunit by the acetyltransferases Eco1 in Saccharomyces cerevisiae and the homologue, Eso1, in Schizosaccharomyces pombe. In both yeasts, these acetyltransferases are essential for cell viability. However, whereas nonacetylatable Smc3 mutants are lethal in S. cerevisiae, they are not in S. pombe. We show that the lethality of a temperature-sensitive allele of eso1 (eso1-H17) is due to activation of the spindle assembly checkpoint (SAC) and is associated with premature centromere separation. The lack of cohesion at the centromeres does not correlate with Psm3 acetylation or cohesin levels at the centromeres, but is associated ith significantly reduced recruitment of the cohesin regulator Pds5. The SAC activation in this context is dependent on Smc5/6 function, which is required to remove cohesin from chromosome arms but not centromeres. The mitotic defects caused by Smc5/6 and Eso1 dysfunction are cosuppressed in double mutants. This identifies a novel function (or functions) for Eso1 and Smc5/6 at centromeres and extends the functional relationships between these SMC complexes.

INTRODUCTION

During the course of the cell cycle, chromosomes undergo massive reengineering to allow gene expression, DNA repair, DNA replication, and finally their compaction and segregation at mitosis. Multiprotein Structural Maintenance of Chromosomes (SMC) complexes, together with DNA topoisomerases, modulate the dynamics of chromosomes as they confront these challenges.

Each SMC complex contains a heterodimer of Smc proteins. These are large coiled-coil proteins with N-terminal Walker A and C-terminal Walker B ATP-binding motifs (Harvey et al., 2002). Eukaryotic SMC complexes are made up of condensin (Smc2 and 4), cohesin (Smc1 and 3), and Smc5/6 (Smc5 and 6). The Smc proteins contain two coiled-coil domains separated by a flexible hinge, where the proteins fold back on themselves to enable N- and C-terminal interactions through ATP binding by the Walker A and B motifs. These folded proteins interact at the hinge domains to form a V-shaped heterodimer. The globular domains are joined by a kleisin subunit to form a ring-shaped structure. Each complex contains a number of unique non-Smc subunits that form an additional layer of specificity for the function(s) of each complex (Hirano, 2006).

Condensin, together with type II DNA topoisomerases, controls the decatenation and condensation of chromosomes that is essential for sister chromatid separation (Hirano, 2005). Cohesin embraces replicated chromatids and, in doing so, facilitates DNA repair by homologous recombination and ensures sister chromatids are equally segregated into daughter cells at anaphase. Cohesin also controls gene expression by locus insulation and modulates the progression of replication forks (Parelho et al., 2008; Rubio et al., 2008; Wendt et al., 2008; Nasmyth and Haering, 2009; Terret et al., 2009; Merkenschlager and Odom, 2013). The function(s) of the third complex, Smc5/6, have remained enigmatic, but most studies have focused on recombinational repair of double-stranded DNA breaks and collapsed replication forks (Murray and Carr, 2008). However, the core Smc5/6 genes are essential for viability, with null mutants initiating but unable to complete chromosome segregation (Verkade et al., 1999; Harvey et al., 2004; Tapia-Alveal et al., 2014a).

During interphase, cohesin is highly dynamic and becomes enriched at sites of DNA damage (Sjogren and Nasmyth, 2001). As most eukaryotic cells enter prophase, the majority of cohesin is stripped off chromosome arms by a poorly characterized mechanism, and cohesin complexes remain physically intact (Tomonaga et al., 2000; Waizenegger et al., 2000; Adachi et al., 2008). This removal, known as the prophase pathway, allows sister chromatids to separate along their arms (but not centromeres), giving mitotically arrested chromosomes their characteristic X shape. At anaphase, the small pool of cohesin that remains at centromeres is removed by proteolytic cleavage of the kleisin subunit by the protease separase, opening the ring and enabling sister chromatids to separate. Thereafter cohesin is immediately reloaded, and the cycle continues. A variation on the cohesin cycle is observed in the budding yeast Saccharomyces cerevisiae, in which the prophase pathway is absent and all cohesin removal depends on separase cleavage of Scc1. Consequently new Scc1 needs to be synthesized to enable reassembly of the complex, and it is loaded later in G1 and S phase (Nasmyth and Haering, 2005, 2009; Nasmyth, 2011).

The cohesiveness of cohesin is established coincident with DNA replication and is then maintained throughout interphase by the acetylation of the Smc3 subunit (known as Psm3 in Schizosaccharomyces pombe) on two N-terminal lysines by homologues of human Esco1 and Esco2 (known as Eso1 in S. pombe and Eco1 in S. cerevisiae). This modification antagonizes anticohesive functions for two cohesin regulators, Pds5 and Wapl (known as Wpl1 in S. pombe and Rad61 in S. cerevisiae; Ivanov et al., 2002; Hou and Zou, 2005; Rolef Ben-Shahar et al., 2008; Zhang et al., 2008; Rowland et al., 2009; Beckouet et al., 2010). Pds5 has complex functional relationships with Eso1: both overexpression and deletion of pds5 suppress the lethality of eso1∆ and eso1-H17 (a temperature-sensitive allele) in S. pombe (Tanaka et al., 2001). Further, Pds5 and Eso1 physically interact, and Pds5 promotes the acetylation of the Smc3 homologue Psm3 by Eso1, thus also promoting the cohesiveness of cohesin (Vaur et al., 2012). The nonlinear nature of these interactions makes genetic analyses of these interactions complex and needs to be interpreted in light of all available data, cell cycle position, and presence of DNA damage.

The mitotic defects conferred by Smc5/6 dysfunction in S. pombe are characterized by sister centromere separation at anaphase without sister chromatid arm separation. This occurs spontaneously in null mutants and in hypomorphs after DNA damage or replication stress (Verkade et al., 1999; Harvey et al., 2004). The mitotic failure is due to the postanaphase retention of cohesin on chromosome arms but not centromeres (Outwin et al., 2009; Tapia-Alveal et al., 2014a, b), that is, a failure of the prophase pathway. Although the mechanism leading to cohesin retention is not yet clear, the mitotic defects can be overcome by ectopic expression of separase (Outwin et al., 2009) and the loss of the histone variant H2A.Z (Tapia-Alveal et al., 2014b); both of these scenarios significantly lower arm cohesin levels, although H2A.Z levels have little to no effect on centromeric cohesin. A similar dependence on Smc5/6 for cohesin removal has been found in S. cerevisiae during meiosis (Copsey et al., 2013) but not during mitosis (Jeppsson et al., 2014), in which the prophase pathway does not function. Of note, the genomic locations of cohesin and Smc5/6 are largely overlapping (Pebernard et al., 2008; Jeppsson et al., 2014), and thus there is significant cross-talk between these related complexes (Tapia-Alveal et al., 2014a).

Coincident with the predominance of the separase pathway in S. cerevisiae, nonacetylatable (K-to-R) smc3 mutants are lethal, as is the deletion of PDS5. The lethality of mutations in this region of Smc3 is suppressed, however, upon deletion of ECO1 (Rowland et al., 2009). However, nonacetylatable psm3-RR mutants are viable in S. pombe (Feytout et al., 2011), and pds5 is not an essential gene (Wang et al., 2002). On the other hand, eso1 is an essential gene, but the lethality of eso1⁻ mutants is still suppressed by pds5∆ and wpl1∆ (Tanaka et al., 2001; Feytout et al., 2011; Vaur et al., 2012). Thus Eso1 either has another substrate and/or another function(s) in addition to Psm3 acetylation that is essential for cell viability, and Pds5 and Wpl1 remain antagonists of these.

Here we study the essential nature of eso1 in S. pombe. We show that the terminal phenotype of eso1-H17 is mitotic arrest due to activation of the spindle assembly checkpoint. This is induced by premature separation of centromeres, although is not related to Psm3 acetylation status. Instead, this is associated with a depletion of Pds5 from centromeres. This defect is bypassed in Smc5/6 mutants, which in turn has its cohesin-retention defects suppressed by Eso1 inactivation. This extends the functional interactions of cohesin and Smc5/6 and also identifies a novel noncatalytic role for Eso1 in centromere cohesion. Given the similarity in cohesin cycles between S. pombe and humans, this new function for Eso1 and Smc5/6 is likely to be conserved in organisms that use the separase-independent pathway to reset the cohesin cycle.

RESULTS

Eso1 inactivation activates a lethal spindle checkpoint arrest

To understand why eso1 is an essential gene whereas Psm3 acetylation is not required for cell viability, we first characterized the terminal phenotype of a temperature-sensitive lethal eso1 allele, eso1-H17. At 25°C, eso1-H17 grows normally and resembles wild-type cells. At restrictive temperatures (32–36°C), ∼50% of eso1-H17 cells arrest in mitosis with short bipolar mitotic spindles (Figure 1A).

FIGURE 1:

eso1-H17 lethality is due to spindle assembly checkpoint activation by premature centromere separation. (A) The indicated strains expressing GFP-tagged α-tubulin were grown at 25°C and then shifted to 36°C for 4 h. The percentage of cells with either short (black) or long (red) spindles were quantified in three counts of 100 cells. Data are mean ± SD. Inset, sample image of eso1-H17 at 36°C. (B, C) An extra copy of the pds5 and wpl1 genes was integrated into the genome of eso1-H17 cells, which were propagated at 32–36°C, whereupon rare (<10⁻⁹) spontaneous suppressors arise. Suppressing mutations were identified by whole-genome resequencing of both parental and suppressor strains. A frameshift mutation in bub1 resulting in a stop codon at codon 106 (bub1-106*) was identified as suppressing the lethality of eso1-H17 (B) and further confirmed with null alleles of both bub1 and mad2 (C). (D) The indicated strains with the temperature-sensitive APC/C mutant cut9-665, a LacO array adjacent to centromere 2 and expressing GFP-LacI, were grown at 25°C (black) or shifted to 36°C for 4 h to arrest cells in metaphase (red). The percentage of cells with two visibly separate GFP foci was quantified from three counts of 100 cells. Data are mean ± SD. Inset, examples of eso1-H17 cells with two foci. (E) Separation of the ade8 locus, located 2.5 Mb from the cen2 on the right arm of chromosome 2, was assayed as in D.

We then screened for extragenic suppressors of eso1-H17. To avoid reisolating suppressor mutations in pds5 and wpl1, we duplicated these genes in the eso1-H17 genome, selected spontaneous suppressors by plating cells at restrictive temperature, and identified the suppressing mutation by whole-genome resequencing. Among the suppressors was a frameshift mutation in the spindle assembly checkpoint (SAC) kinase gene bub1, truncating the protein at residue 106 (N-terminal to the kinase domain; Figure 1B). We confirmed suppression of both growth (Figure 1C) and spindle morphology (Figure 1A) with null alleles of both bub1 and mad2. Thus the lethality of eso1-H17 is SAC dependent.

Because the spindles were normal in appearance (Figure 1A), we tested whether the SAC was being activated by premature centromere separation. To this end, we used strains with a lacO array integrated immediately adjacent to the centromere on chromosome 2, with lacI–green fluorescent protein (GFP) expressed constitutively from the his7 promoter (Outwin et al., 2009). To distinguish premature separation from normal separation, this was done on a temperature-sensitive cut9-665 genetic background; Cut9 is a component of the anaphase-promoting complex, and inactivation of Cut9 induces a metaphase arrest with paired centromeres (Samejima and Yanagida, 1994). This experiment showed premature separation in ∼35% of eso1-H17 cells (Figure 1D). Although lower than the percentage of cells with short spindles (∼50%, Figure 1A), this is likely an underestimate of actually separated centromeres that are not far enough apart to distinguish optically. A similar experiment using a lacO array integrated at the ade8 locus, which also on chromosome 2 but ∼2.5 Mb from the centromere, was also performed. eso1-H17 had an increase in separation of this locus at all temperatures, but the effect was less than half that seen for the centromeres (Figure 1E). As expected, bub1∆ also suppressed this premature-centromere-separation phenotype (Figure 1D). Thus inactivation of Eso1 manifests in the cohesion of centromeres, leading to SAC activation. However, whatever the SAC activating defect is, it is not by itself a lethal event.

Spindle checkpoint arrest is not due to N-terminal Psm3 hypoacetylation

The best-characterized function for Eso1 is the enforcement of cohesin's cohesiveness via acetylation of K105 and K106 on the Smc3 homologue, Psm3 (Feytout et al., 2011). Using a strain with both these lysines mutated to arginines (psm3-RR), we assessed spindle morphology and frequency and observed no accumulation of short spindles (Figure 2A).

FIGURE 2:

Spindle checkpoint activation is not caused by Psm3 hypoacetylation. (A) Wild-type, nonacetylatable psm3-RR and eso1-H17 cells expressing GFP-tagged α-tubulin were grown at 25°C and then shifted to 36°C for 4 h. The percentage of cells with either short (black) or long (red) spindles was quantified in three counts of 100 cells. Data are mean ± SD. (B) Western blot of total (anti-HA) and K106-acetylated (K106^Ac) Psm3 in the indicated strains at 25, 30, and 36°C (4 h each). The K106^Ac blots are a short (top) and long (middle) exposure of the same Western. This membrane was stripped and reprobed for anti-HA (bottom).

This suggested that SAC activation upon Eso1-H17 inactivation should not correlate with Psm3's acetylation status. To assess this, we used an acetyl-specific antibody for K106 (K106^Ac; Feytout et al., 2011), and indeed, we could not detect Psm3 acetylation in eso1-H17 cells grown at 25, 30, or 36°C (Figure 2B). The same result was also reported for eso1-H17 grown at 25°C by Feytout et al. (2011), who kindly provided the K106^Ac antibody. In addition, we generated a second eso1 allele, eso1-1. This has two point mutations, R810G and K811G. The analogous mutations in ECO1 in S. cerevisiae were originally reported to lack activity in vitro when purified from bacteria (Ivanov et al., 2002) but were subsequently shown to retain some autoacetyltransferase activity when purified from S. cerevisiae extracts (Unal et al., 2008). In S. pombe, eso1-1 cells have significantly reduced K106 acetylation, which is further reduced with increasing temperature to barely detectable levels at 36°C (Figure 2B). Despite this, eso1-1 cells grow normally (see later discussion of Figure 6E) and do not display mitotic defects. Thus we conclude that the SAC activation upon Eso1 inactivation reflects a function other than Psm3-K106 acetylation—most likely a noncatalytic function.

FIGURE 6:

Cosuppression between smc6-74 and eso1-H17 restores viability and mitotic completion. (A) The experimental protocol. Cells growing asynchronously at 25°C were treated with 11 mM HU and then left at 25°C for 5.5 h or shifted to 36°C for 4 h (cell cycle equivalents). HU was then removed by filtration, and cultures were split to 25 or 36°C for 4 h ± 10 μM latrunculin B. For cultures treated at 25°C, an additional 1.5 h was added at 25°C to allow completion of DNA replication before temperature shift. (B) Aliquots were taken from the 10 μM latrunculin B cultures, fixed, and stained with DAPI to assess percentage of normal binucleated cells (data are mean ± SD of >100 cells, n = 3). Aliquots were taken from the 0 μM latrunculin B cultures, serially diluted, plated to YES plates to assess relative viability compared with no-HU controls, and normalized to wild type (mean ± SD, n = 3). Note that regardless of temperature protocol, HU induces a blockade to anaphase in smc6-74 and eso1-H17 cells, but this is cosuppressed in the double mutant. (C) Micrographs of DAPI-stained cells of the indicated strains after incubation at 36°C in the presence of latrunculin B. The aberrant mitoses seen in smc6-74 and eso1-H17 cells are characteristic of all temperature combinations. Bar, 10 μm. (D) The cosuppression does not extend to colony formation at 32°C over 4 d or in the presence of methyl methanesulfonate (MMS) or HU over 5 d at 25°C. (E) SAC inactivation (bub1∆) does not confer colony formation on smc6-74 eso1-H17 cells.

Eso1 inactivation reduces centromeric Pds5

Cohesin levels at the centromeres are higher than on chromosome arms (Outwin et al., 2009; Tapia-Alveal et al., 2014b). As a cohesin regulator, the most obvious reason for the premature separation of centromeres was the absence of cohesin at this locus. We assayed this by chromatin immunoprecipitation (ChIP) and found no effect of eso1-H17 on cohesin at the outer repeats (OTRs) of the centromeres (Figure 3A). However, the OTRs must be separated, as the lacO arrays are adjacent to them but further away from the innermost repeats of the centromeres. Thus, although present, the centromeric cohesin in eso1-H17 cells cannot be cohesive at 36°C. We also observed no significant change in cohesin levels in subtelomeric DNA (STE1) or at loci on each of the chromosome arms (1L–3R), although, presumably, cohesiveness is also affected at these loci (cf. Figures 3A and 1E).

FIGURE 3:

Eso1 inactivation depletes Pds5, but not cohesin, from centromeres. (A) Cohesin (Rad21-GFP) and (B) Pds5 (Pds5-HA₃) levels were assayed by ChIP in wild-type (black) and eso1-H17 (red) cells grown at 25°C or shifted to 36°C for 4 h. Loci are the outer repeats of the centromeres (OTR), subtelomeric DNA on chromosome 2R (STE1), and each chromosome arm (1L–3R). Data (qPCR) are mean ± SEM, n = 6. Two-sided t test comparing wild type to eso1-H17: ^#p < 0.001, *p < 0.05. Nonannotated histogram bars represent nonsignificant differences.

Another cohesin regulator is Pds5, with complex functional relationships with Eso1 (Tanaka et al., 2001; Wang et al., 2002). However, through a physical interaction between Pds5 and Eso1, Pds5 is believed to promote Psm3 acetylation and promote sister chromatid cohesion (Vaur et al., 2012). We assayed the localization of Pds5 to the same loci used for the cohesin ChIP experiments (Figure 3B). Indeed, at 36°C in eso1-H17 cells, Pds5 levels were reduced at the centromeres by ∼70%. It was also reduced at some arm loci but not to the same extent as at the centromeres. Thus a defect in Eso1-dependent recruitment of Pds5 is likely at least one aspect of the centromere defect in eso1-H17. However, this cannot alone explain eso1's essential nature, as pds5∆ cells are viable, albeit with mitotic defects (Tanaka et al., 2001; Wang et al., 2002).

Spindle checkpoint activation in eso1-H17 is bypassed by Smc5/6 dysfunction

Previously we showed that the cohesin-related Smc5/6 complex mediates cohesin dynamics in S. pombe. Smc5/6 is required for the separase-independent prophase pathway of cohesin removal from chromosome arms; Smc5/6 genes are essential, but this defect can be induced in hypomorphs such as smc6-74 by DNA damage and replication stress (Outwin et al., 2009; Tapia-Alveal et al., 2014a, b). Although Smc5/6 does not mediate centromeric cohesin removal, we believed it was possible that Smc5/6 (as a cohesin regulator) and Eso1 might interact in the regulation of centromeric separation. Indeed, smc6-74 completely suppressed the SAC-activation (Figure 4A) and premature-centromere-separation (Figure 4B) phenotypes of eso1-H17 but did not restore colony formation at nonpermissive temperatures (see later discussion of Figure 6D). Because smc6-74 cells have no spindle defect on their own (Verkade et al., 1999), there is no interaction with SAC mutants (unpublished results). Of interest, this suppression mechanism did not restore Pds5 recruitment to the centromeres (Figure 4C), and changing pds5 gene copy number does not alter the phenotype (unpublished results). Thus either this suppression is a bypass suppression mechanism or, despite the cohesion-promoting activity of Pds5, its absence from centromeres in eso1-H17 is not functionally significant in the premature centromere separation in eso1-H17.

FIGURE 4:

Spindle checkpoint activation in eso1-H17 requires Smc5/6 function. (A) The indicated strains expressing GFP-tagged α-tubulin were grown at 25°C and then shifted to 36°C for 4 h. The percentage of cells with either short (black) or long (red) spindles was quantified in three counts of 100 cells. Data are mean ± SD. Short spindle accumulation in eso1-H17 is suppressed by smc6-74. (B) Premature centromere separation was assayed in the indicated strains as performed in Figure 1D. The accumulation of cells with separated centromere 2 in eso1-H17 is suppressed by smc6-74. (C) The indicated strains were grown at 25°C or shifted to 36°C, and Pds5 levels at the centromeres (OTR) were determined as in Figure 3. Despite suppression, smc6-74 does not restore depleted Pds5 levels in eso1-H17.

Damage-induced cohesin retention after Smc5/6 dysfunction does not require additional cohesin loading

To further explore the relationship between Smc5/6 and Eso1, we next asked whether the damage-induced cohesin retention in irradiated smc6-74 cells is the preexisting cohesin complexes in G2 cells already acetylated by Eso1 or newly recruited complexes at sites of DNA damage. This was designed to ask whether the observed retention required loading of cohesin to sites of DNA damage or, alternatively, whether the regular existing complexes were the retained pool of cohesin. To this end, we devised a protocol in which Mis4 inactivation (using temperature-sensitive mis4-242) blocks additional cohesin loading in cells before irradiation, and then we assayed arm and centromere separation microscopically (Figure 5, A and D). Compared to wild-type controls (Figure 5B), smc6-74 cells showed the same chromosome arm segregation defects regardless of Mis4 status (Figure 5C), indicating that it is the existing pool of arm cohesin that is retained in smc6-74.

FIGURE 5:

Preexisting cohesin complexes are retained following DNA damage in smc6-74 cells. (A) The experimental protocol. mis4-242 cells expressing a GFP-lacI fusion and carrying a lacO array directly adjacent to centromere 2 were used on both wild-type and smc6-74 backgrounds. G2 cells were collected by elutriation at 25°C, split, and then either left to recover at 25°C for 30 min or shifted to 36°C to thermally inactivate mis4-242 and prevent further loading of cohesin. Cells were then filtered and irradiated with 0 or 30 J/m² ultraviolet-C and returned to growth at 25 or 36°C for a further 5 h in the presence of the actin poison latrunculin B (10 μM; to prevent lethal cytokinetic events). Cells were then scored for nuclear and GFP focus number by microscopy. (B) Data from mis4-242 cells. (C) Data from mis4-242 smc6-74 cells. Data are mean ± SD for three counts of >100 cells. Note that cen2 separates without chromosome segregation (one nucleus, two centromeres) regardless of Mis4 status in smc6-74 cells. (D) Micrographs of merged DAPI and GFP signals showing normal cells with one or two nuclei (top) and examples of the aberrant mitoses seen in irradiation smc6-74 cells. Bar, 10 μm.

Cosuppression of mitotic defects after Eso1 and Smc5/6 dysfunction

Because the presence of cohesin on chromosomes does not indicate its cohesive state (e.g., Figure 3A), we used Mis4 inactivation in synchronized cells to study the dynamics of cohesin that is regulated by Smc5/6. We attempted to extend these studies to ask how eso1 affects the dynamics regulated by Smc5/6 after DNA damage or replication stress, but double mutants of both eso1-H17 and eso1-1 with mis4-242 are close to synthetic lethal at 25°C (unpublished data) and too severely growth inhibited to make meaningful conclusions. We therefore devised a protocol using hydroxyurea (HU) block and release, which both induces cohesin retention in smc6-74 (Outwin et al., 2009) and allows analysis of chromosome segregation and viability as cells synchronously pass through a single mitosis (Figure 6A). We combined this with temperature shifts to modulate Eso1-H17 activity. Of interest, this protocol showed eso1-H17 to be extremely HU sensitive at all temperatures, even more so than smc6-74. However, smc6-74 eso1-H17 double mutants showed cosuppression; that is, the double mutants completed mitosis more efficiently than the parental single mutants and had better corresponding survival (Figure 6, B and C) in this HU block-and-release protocol. This was not seen in colony formation assays on HU-containing plates at 25°C, or at 32°C without HU, but this would require efficient suppression over ∼40 cell cycles (Figure 6D). On long incubations, in liquid cultures, the double mutants accumulate cells that are dying in mitosis, and this is in keeping with the substantial but incomplete suppression. Similarly, smc6-74 eso1-H17 bub1∆ triple mutants also fail to form colonies at nonpermissive temperatures (Figure 6E).

Cosuppression was not seen with the psm3-RR mutant (Figure 7A), nor does this modify the sensitivity of smc6-74 to chronic exposure of low-dose HU (Figure 7B). However, in keeping with the cohesin-retention defects in this strain, acetyl-mimic (cohesion-promoting) psm3-NN did increase the HU sensitivity of smc6-74 (Figure 7B). Suppression by both protocols was also not seen with eso1-1 (Figure 7C), despite significantly reduced Psm3-K106 acetylation (Figure 2B). Because Eso1 in S. pombe is a fusion protein in which the acetyltransferase domain is fused to the bypass polymerase Polη, we deleted that polymerase domain and confirmed no interaction with smc6-74 (Figure 7D; Sheedy et al., 2005; Lee et al., 2007). Finally, we assayed HU sensitivity in cells lacking bub1 or mad2 and observed efficient suppression of the HU sensitivity of eso1-H17. Therefore SAC activation is the source of HU sensitivity of eso1-H17 cells (Figure 7E).

FIGURE 7:

Blocking Psm3 acetylation does not phenocopy eso1-H17. (A) The same protocol used in Figure 6 was used to assess suppression of smc6-74 by the nonacetylatable psm3-RR mutant and the acetyl-low eso1-1 mutant. In no case was suppression observed. Unlike eso1-H17, eso1-1 shows neither HU-induced mitotic defects nor suppression of smc6-74 despite a significant reduction in Psm3 K106 acetylation (Figure 2). (B) Acetyl-mimetic (psm3-NN) enhances the HU and MMS sensitivity of smc6-74, but nonacetylatable (psm3-RR) does not. (C) eso1-1 does not suppress the chronic HU sensitivity smc6-74. (D) A strain in which the DNA polymerase η domain region of eso1 is deleted (eso1-η∆) does not suppress the HU or MMS sensitivity of smc6-74. (E) The same HU block-and-release protocol used in Figure 6A was used to assess the effects of SAC mutations on the HU sensitivity of eso1-H17, which was suppressed by both bub1∆ and mad2∆.

Thus we conclude that the observed cosuppression of HU-induced mitotic failure is not due only to Psm3 hypoacetylation, although this may contribute in part to the suppression of the cohesin retention leading to HU-induced mitotic failure in smc6-74. In turn, the suppression of SAC activation by smc6-74 reverses the HU sensitivity of eso1-H17, indicating that HU-induced replication stress potentiates the centromeric defects caused by Eso1 dysfunction.

DISCUSSION

Many aspects of the regulation of the cohesin cycle have been firmly established in several experimental systems. Most notable among these is the budding yeast S. cerevisiae. In this system, cohesin's cohesiveness appears to be largely regulated by Eco1-catalyzed Smc3 acetylation and the subsequent blocking of anticohesive factors, and this activity is required for cell viability. In turn, removal of cohesin at anaphase in this organism seems also to be via one molecular event: the cleavage of Scc1 by separase (Nasmyth and Haering, 2009).

Although these events are conserved across eukaryotes, they appear to control only a subset of cohesin complexes. At least in S. pombe, Psm3 acetylation is dispensable for cell viability (Feytout et al., 2011). Similarly, in human cells, induced expression of nonacetylatable Smc3 has relatively modest effects on sister chromatid cohesion (Zhang et al., 2008), although this is in the presence of wild-type Smc3. However, at least in human cells, Smc3 acetylation does promote the processivity of DNA replication fork processivity (Terret et al., 2009) and, coupled with the role of cohesin as a transcriptional insulator (Wendt et al., 2008), provides possible alternative explanations for phenotypes associated with Smc3 hypoacetylation.

Despite the viability of psm3-RR mutants, eso1 is an essential gene in S. pombe. Because SAC ablation suppresses the lethality of eso1-H17 (Figure 1), an important component of the essential function(s) of Eso1 resides in promoting the cohesion of sister centromeres in a manner independent of acetylation of Psm3, at least on K105 and K106. Although it is possible that Eso1-H17 retains acetyltransferase activity for other sites and/or other proteins, no other acetylation sites in cohesin can be detected with anti–acetyl-lysine antibodies (Feytout et al., 2011). We tagged eso1 and eso1-H17 with various epitopes, and whereas Eso1-H17 protein is expressed at all temperatures, epitope-tagged eso1-H17 is only viable on a pds5∆ background at 25°C (unpublished data), and so addition of the tags renders the protein completely nonfunctional. Nevertheless, it suggests that the temperature sensitivity of eso1-H17 is not associated with degradation of the protein.

Note that suppression of the growth defect of eso1-H17 by mad2∆ was previously observed (Tanaka et al., 2000), but evidence of SAC activation and premature centromere separation was not reported. Our results show that cohesin is present at centromeres in eso1-H17 (Figure 3A), but as centromeres prematurely separate, these cohesin complexes cannot be cohesive. What is depleted from the centromeres in eso1-H17 is Pds5 (Figure 3B), and because the interaction between Pds5 and Eso1 promotes cohesiveness (Vaur et al., 2012), this defect is a likely cause of the premature centromere separation. Analyzing pds5∆ cells cannot mimic this because they will also be defective in cohesion regulation on the arms.

We also observed strong functional interactions between smc6-74 and eso1-H17. The blockade to sister chromatid separation caused by the retention of existing arm cohesin complexes in smc6-74 cells after DNA damage or replication stress is also suppressed by eso1-H17 (Figures 5 and 6). Of note, centromeric cohesin and cohesion are unaffected in smc6-74 and other Smc5/6 mutants (Outwin et al., 2009; Tapia-Alveal et al., 2014a, b). Because acetyl-mimetic psm3-NN enhances the HU sensitivity of smc6-74 (Figure 7B), the suppressing effects of eso1-H17 are likely to be, at least in part, via Psm3 hypoacetylation. This was not seen with psm3-RR, which may not recapitulate the spatial control of Psm3 acetylation and/or may indicate an additional level of regulation that relieves cohesin retention in the face of Smc5/6 dysfunction. Chemical (thiabendazole) or genetic (nda3-KM311 at 18°C) activation of the SAC by tubulin depolymerization does not suppress smc6-74 (unpublished data), suggesting that this is not the source of smc6-74 suppression.

In turn, the suppression of the premature centromere separation of eso1-H17 must be either independent of Pds5 recruitment or a compensatory bypass of this defect. Lethal SAC activation is potentiated in eso1-H17 by replication stress, which may exacerbate defective DNA replication in the absence of Eso1 function. Of note, the resident heterochromatin and repetitive nature of the centromeres make these difficult-to-replicate regions of the genome (Hayashi et al., 2009; Li et al., 2013). Whatever the molecular centromeric defect in eso1-H17, the suppression by smc6-74 suggests that it includes the Smc5/6-dependent formation of toxic structures that are further promoted by extrinsic replication stress. With its well-defined role in controlling recombination at both stably stalled and collapsed replication forks (Ampatzidou et al., 2006; Murray and Carr, 2008; Irmisch et al., 2009), the combination of smc6-74 with eso1-H17 may prevent such structures from forming. Testing and refining such a hypothesis will first require a better understanding of the centromeric function for Eso1. Although this is a problem that will require substantial future work, it is at least an area of investigation that begins to bring together what we know about cohesin, Smc5/6, and their regulators.

In cells in which a genetically or chemically induced molecular defect activates a cell cycle checkpoint, the viability of such cells is usually dependent on an intact checkpoint response. This is not the case for eso1-H17, for which SAC activation is actually the lethal response. Thus the centromeric defect conferred by Eso1 dysfunction is tolerable, albeit checkpoint activating.

Many aspects of the cohesin cycle are conserved between S. pombe and humans. Thus it will be important to understand whether there is a centromeric function for human ESCO1/2, as well as the relationship between the spindle checkpoint and ESCO1/2 (and other cohesin and Smc5/6 components) in the developmental disorders and cancers in which these genes are mutated (Skibbens et al., 2013; Price et al., 2014; Cucco and Musio, 2016; van der Crabben et al., 2016). Indeed, SAC dysfunction may be an explanation as to why some tumor cells can survive loss-of-function mutations in cohesin, whereas normal cells cannot.

MATERIALS AND METHODS

S. pombe culture and genetic methods

All strains were constructed by tetrad analysis, with compound mutants selected from non-parental ditypes. In each case, multiple progeny were analyzed and backcrossed to exclude the presence of suppressor mutations. Cells were grown in supplemented yeast extract plus glucose (YES) or EMM2 (Moreno et al., 1991). Culture methods for individual experiments are described in the relevant figure legends. Drug sensitivity assays were performed on solid YES medium, with 5-μl drops of 10-fold serial dilutions from cultures containing ∼4 × 10⁶ cells/ml and plates photographed after either 4 d (30°C) or 5 d (25°C). Ultraviolet-C sensitivity was determined by triplicate plating of 100–10,000 cells, which, after drying, were irradiated in a Stratalinker (Stratagene, San Diego, CA). Viability assays were determined by triplicate plating of a culture dilution series to YES and normalized to untreated wild-type cells. eso1-1 was constructed by fusion PCR and integration into the eso1 locus and confirmed by sequencing and Southern blotting. G2-synchronized cells were obtained by centrifugal elutriation using a Beckman (Brea, CA) JE-5.0 elutriation system. Strains are listed in the Supplementary Material.

Screen for suppressors of eso1-H17

eso1-H17 cells were transformed with genomic clones of both pds5 and wpl1. Single-copy integrants (at leu1 and nmt1, respectively) were selected and confirmed to phenocopy untransformed eso1-H17 cells. Spontaneous suppressors were selected and identified by whole-genome resequencing as previously described (Tapia-Alveal et al., 2014b).

Microscopy

Nuclei were visualized in cells fixed in 70% ethanol by staining with 1 μg/ml 4′,6-diamidine-2′-phenylindole dihydrochloride (DAPI). Microtubules were visualized in live cells expressing GFP-tubulin. Spindles were scored as short when present as newly growing spindles through to those as long as the diameter of an undivided nucleus (∼0.5–3.5 μm). On anaphase, short spindles rapidly elongate to a length of ∼10 μm, which were scored as long spindles. Intermediates are rarely observed but were scored as long when nuclear division was advancing. Centromere 2 was visualized in either live cells immobilized on thin agarose pads by expression of GFP-LacI from the his7 promoter in strains with a lacO array integrated immediately adjacent to the outer repeat (OTR; Outwin et al., 2009) or cells fixed in 90% methanol, stained with DAPI, and imaged immediately. Arm separation was assayed with a lacO array at the ade8 locus (Outwin et al., 2009). Microscopy was carried out on a Zeiss (Thornwood, NY) Z1 Imager with a 100×/1.4 Pan-Apo objective.

Protein extracts and Western blotting

To detect total and acetylated Psm3, cells were grown to mid logarithmic phase at 25°C and shifted to either 30 or 36°C for 4 h. Cells were harvested by centrifugation and snap frozen in liquid nitrogen. Cells were then disrupted in a mini beadbeater (Biospec) using the following buffer: 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, pH 7.5, 100 mM KCl, 2.5 mM MgCl₂, 0.25% Triton X-100, 1 mM dithiothreitol, 0.1% SDS, 10 mM sodium butyrate, 10% glycerol, 1 mM sodium orthovanadate, 20 mM β-glycerophosphate, 1 mM phenylmethylsulfonyl fluoride, and 5× protease inhibitor cocktail (P8340; Sigma-Aldrich, St. Louis, MO). Protein levels were quantified (Dc Protein assay; Bio-Rad), and 30 μg was separated on a 6% SDS–PAGE gel. Proteins were transferred to nitrocellulose in 10 mM 3-(cyclohexylamino)-1-propanesulfonic acid and 10% methanol at 70 V and 45 min. The membrane was blocked in Tris-buffered saline containing 0.1% Tween 20 (TBST) and 5% skim milk powder for 30 min. K106-acetylated Psm3 was detected with a 1:2000 dilution of an acetyl-specific antibody overnight (Feytout et al., 2011), followed by a 1:2000 dilution of horseradish peroxidase HRP)–conjugated anti-rabbit immunoglobulin G (IgG; GE Healthcare). Immune complexes were detected with Clarity Reagent (Bio-Rad). Subsequently the membrane was stripped with 200 mM sodium hydroxide for 2 × 10 min at room temperature, rinsed in TBST, and reprobed with 0.04 μg/ml anti-hemagglutinin (HA; 12CA5; Roche) overnight, followed by 1:2000 HRP-conjugated anti-mouse IgG and detected with Clarity Reagent.

Chromatin immunoprecipitation

Methods for ChIP and primer sequences for quantitative PCR (qPCR) were as previously described (Outwin et al., 2009). qPCR data are expressed as fold enrichment over an untagged isogenic control. In each case, the input DNA is 1/80 of the IPed DNA, so percentage recovery can be calculated by dividing the fold enrichment by 80. Oligonucleotides for qPCR are listed in the Supplementary Material.

Abbreviations used:

GFP

green fluorescent protein

HU

hydroxyurea

MMS

methyl methanesulfonate

OTR

centromeric outer repeats

SAC

spindle assembly checkpoint

SMC

structural maintenance of chromosomes.