Studying meiotic cohesin in somatic cells reveals that Rec8-containing cohesin requires Stag3 to function and is regulated by Wapl and sororin

Cohesin is a multiprotein complex that provides sister chromatid cohesion by embracing both DNA copies in its middle. In mitotically dividing cells, an Smc1α–Smc3 heterodimer associates with the kleisin subunit Scc1 (also known as RAD21) forming a tripartite ring. Human somatic cells express two variants of a fourth subunit, Stag1 and Stag2, which bind Scc1 in a mutually exclusive manner (Nasmyth, 2011; Nasmyth and Haering, 2009).

Vertebrate cohesin is loaded in telophase but is constantly removed from chromatin owing to the action of Wapl (Gandhi et al., 2006; Kueng et al., 2006), until S phase when a pool of cohesin is stabilised by the Wapl antagonist sororin (also known as CDCA5) (Nishiyama et al., 2010; Rankin et al., 2005). In early mitosis, Wapl-dependent removal of cohesin from chromosome arms is boosted when phosphorylation inactivates sororin (Dreier et al., 2011; Hauf et al., 2005; Liu et al., 2013). A small pool of cohesin, however, is resistant to this so-called prophase pathway signalling. At centromeres, shugoshin 1 (Sgo1) recruits the protein phosphatase 2A (PP2A) to keep sororin active by means of dephosphorylation (Liu et al., 2013; McGuinness et al., 2005; Tang et al., 2006). Centromeric cohesion is resolved in anaphase when separase (also known as ESPL1) cleaves Scc1 from the remaining chromosomal cohesin (Hauf et al., 2001).

In a prototypic meiosis, recombination events between non-sister chromatids create bivalents, in which the homologous chromosomes are held together by cohesin on chromosome arms located distally to crossovers. Segregation of homologous chromosomes in meiosis I is achieved by selective loss of cohesin at chromosome arms. In the second meiotic division, cohesion is resolved at the centromeres allowing the disjunction of sister chromatids. Interestingly, both waves of cohesin removal depend on separase activity. However, centromeric cohesion is maintained during meiosis I because (1) Scc1 is functionally replaced by Rec8, which is different from its mitotic counterpart in that it must be phosphorylated in order to be cleaved by separase (Kudo et al., 2009; Riedel et al., 2006) and (2) Sgo2, which replaces Sgo1 in mammalian meiosis, recruits PP2A to keep centromeric Rec8 in a dephosphorylated state (Lee et al., 2008; Llano et al., 2008).

An unresolved question is whether Sgo1 is only competent to protect Scc1 cohesin from the prophase pathway and whether, conversely, Sgo2 might only shield Rec8 cohesin from separase.

In addition to Stag1 and Stag2, germ cells contain a third variant named Stag3 (Prieto et al., 2001). Furthermore, vertebrate meiocytes express two variants of Smc1, Smc1α and the meiosis-specific Smc1β (Revenkova et al., 2001; Revenkova et al., 2004). Which of the possible combinations of cohesin subunits actually exist in germ cells is controversial (Ishiguro et al., 2011; Lee and Hirano, 2011; Revenkova et al., 2004). Moreover, while it is clear that Rec8 is crucial for cohesion in meiocytes (Tachibana-Konwalski et al., 2010), whether and which other meiosis-specific cohesin subunits are essential for cohesion still needs clarification. Wapl is expressed during meiosis, and mutant spermatocytes with dysregulated Wapl exhibit increased chromosomal Smc3 levels (Brieño-Enríquez et al., 2016; Kuroda et al., 2005). However, whether the Rec8 cohesin is responsive to Wapl-dependent ring opening remains enigmatic.

Here, we exploit somatic cells to study human meiotic cohesin subunits and reveal the following: (1) Rec8 does not interact with Stag1 or Stag2 but robustly associates with Stag3 and becomes functional (i.e. loaded onto chromatin and cohesive) only in presence of this meiosis-specific cohesin subunit; (2) Rec8 cohesin physically interacts with and is regulated by sororin and its antagonist Wapl; (3) Sgo1 and Sgo2 are indeed specific for distinct cohesin complexes as specified by the kleisin subunit.

We generated transgenic Hek293 cell lines inducibly expressing C-terminally GFP-tagged Rec8 or, as a control, Scc1 and compared the cellular localisation of both fusion proteins. Notably, in all clones analysed the levels of Scc1–GFP were lower than Rec8–GFP (data not shown) consistent with previous experiments indicating that cells try to keep a constant level of Scc1 (Schöckel et al., 2011).

Consistent with a recent report (Rong et al., 2017), we found that Rec8, in contrast to Scc1, was excluded from the nucleus (Fig. 1A). We reasoned that another meiosis-specific cohesin subunit might be necessary for the nuclear targeting of Rec8. First, we asked whether Rec8 is competent to associate with all non-kleisin subunits of the somatic cohesin complex. Immunoprecipitation (IP) experiments revealed a robust interaction of Rec8 with Smc3 and Smc1α (Fig. 1B). Interestingly, studies conducted in spermatocytes revealed an association of Rec8 only with Smc1β and not with Smc1α (Ishiguro et al., 2011; Lee and Hirano, 2011). Importantly, endogenous Stag1 and Stag2, while readily co-purifying with Scc1, failed to co-IP with Rec8 (Fig. 1C; Fig. S1A). We directly compared transiently expressed Flag-tagged Stag2 and Stag3 in their ability to associate with Rec8 and confirmed that only Stag3, and not Stag2, exhibited a strong interaction with the kleisin (Fig. 2A). Moreover, Stag3 changed the localisation of Rec8 from the cytoplasm to the nucleus, while ectopic expression of Stag2 had no such effect (Fig. 2B–D; Fig. S2A,B). In addition, the nuclear Rec8–GFP signal became resistant to preextraction in presence of Stag3, indicating that Rec8 is also loaded onto chromatin in a Stag3-dependent manner. Stag3 triggered this effect irrespectively of its expression from a transiently transfected plasmid (Fig. S2A,B) or a genomically integrated transgene (Fig. 2B–D). Rec8 behaved the same in these kinds of experiments with both a C-terminal GFP tag and an N-terminal Myc₅ tag indicating that the tag did not affect kleisin localisation (Fig. S2C,D). Taken together, our data suggest that Rec8 might only be functional when assembled in a cohesin complex that also contains Stag3. Although some somatic cell types express Smc1β (Mannini et al., 2015), we could not detect any Smc1β transcript as judged by means of a PCR on isolated cDNA (Fig. S1B).

Rigorous testing of whether Rec8-containing cohesin can mediate cohesion in somatic cells requires abrogation of Scc1 cohesin function. In our hands, the knockdown of SCC1 was incomplete and did not result in a penetrant phenotype. Therefore, we used overexpression of hyperactive separase (Boos et al., 2008; Holland and Taylor, 2006) or depletion of Sgo1 (McGuinness et al., 2005; see further details below) as two independent and established methods to induce premature sister chromatid separation (SCS) in prometaphase cells. We then asked whether the cohesion defect could be rescued by presence of Stag3 and Rec8.

First, parental Hek293 cells, cells expressing Rec8 with Stag2, and Rec8 with Stag3 were transfected to express hyperactive separase and arrested with nocodazole treatment. Subsequent chromosome spreading revealed untimely SCS in an average of 31% of the cells without the transgene or with REC8 and STAG2 (Fig. 3A; lanes 2 and 3; Fig. S3A), which represents a fivefold increase relative to control cells (lane 1). Importantly, precocious SCS occurred in merely 10% of cells that expressed Rec8 and Stag3 in addition to hyperactive separase (lane 5). Thus, Rec8–Stag3-containing cohesin alleviates separase-dependent cohesion failure. Given that a surplus of chromosomal Scc1 cohesin leaves anaphase largely unaffected (Hauf et al., 2005), the above result also argues that Rec8–Stag3 cohesin represents a poor substrate for separase in prometaphase. This could be explained by the fact that somatic cells express Sgo2, whose canonical function is to protect Rec8 from separase-dependent cleavage throughout meiosis I. Indeed, the percentage of Rec8–Stag3-expressing cells that suffered from premature SCS tripled when expression of hyperactive separase was combined with depletion of Sgo2, while siRNA against SGO1 had no effect (Fig. 3A, lanes 6 and 7). However, this rescue effect does not prove that Sgo2 protects Rec8 from separase because, as shown below, it might also be explained by Sgo2-dependent protection of Rec8 cohesin from the prophase pathway.

To independently confirm the functionality of Rec8–Stag3 cohesin, we next asked whether it could suppress the effect of a Sgo1 depletion, which on its own leaves centromeric Scc1 cohesin unprotected against the prophase pathway, hence resulting in untimely SCS. To this end, cells were arrested through treatment with nocodazole and analysed by chromosome spreading, as above, but only after transfection of siRNA against SGO1 (or, as control, GL2 siRNA targeting luciferase) (Fig. 3B). As expected, removal of Sgo1 caused loss of cohesion in 77% of parental Hek293 cells and Rec8–Stag2-expressing cells (lanes 2 and 4). Remarkably, expression of Rec8 and Stag3 almost fully suppressed the premature loss of cohesion in the absence of Sgo1 to background level (10% versus 7%, respectively; lanes 1 and 6). Thus, Rec8–Stag3 cohesin renders Sgo1 dispensable, which provides strong experimental proof that meiotic cohesin can indeed mediate sister chromatid pairing in somatic cells. Our data are in accordance with studies reporting Stag3 to be a component of all meiosis-specific cohesin complexes in mice and its loss to result in strong cohesion defects in germ cells (Fukuda et al., 2014; Hopkins et al., 2014; Ishiguro et al., 2011). Our results demonstrating the functionality of the Smc1α–Rec8–Stag3 complex are also in line with the finding that Smc1β-deficient meiocytes retain some cohesion throughout prophase I (Biswas et al., 2013).

The rescue of the Sgo1-depletion phenotype by Rec8–Stag3 cohesin demonstrates that it is not removed by prophase pathway signalling, at least under the conditions of a nocodazole arrest. This could mean that meiotic cohesin is resistant to the prophase pathway altogether. Alternatively, it might be susceptible to Wapl-dependent ring opening, in principle, but protected by Sgo2. To clarify this issue, we first asked whether the concomitant depletion of both Sgo1 and Sgo2 would cause premature SCS in Rec8–Stag3-expressing cells (Fig. 3B). Indeed, relative to the single depletion of Sgo1, depletion of both shugoshins increased precocious dissolution of sisters in Rec8–Stag3-expressing cells almost threefold (lanes 6 and 7). This likely represents an underestimation because (for unknown reasons) knockdown of Sgo2 in parental Hek293 and Rec8–Stag2-expressing cells had an opposing effect, where it slightly (but reproducibly) suppressed the Sgo1 depletion phenotype (lanes 2 to 5). Taken together, our results suggest that Sgo1 can only shield Scc1 cohesin, while Sgo2 is solely capable of Rec8–Stag3 cohesin protection. Accordingly, Scc1 does preferentially co-IP with Sgo1 whereas Rec8 shows a robust interaction only with Sgo2 (Fig. S3B,C). Our findings rationalise why Sgo2 is dispensable for mitosis (Llano et al., 2008; Orth et al., 2011). However, they contradict previous reports showing that Sgo1 was able to shield Scc1 cohesin from separase and that Sgo2 contributed to cohesion protection in somatic cells (Huang et al., 2007; Kitajima et al., 2006; Lee et al., 2008).

To address the putative susceptibility of Rec8–Stag3 cohesin to the prophase pathway more directly, we conducted a co-IP experiment demonstrating that Rec8 interacts not only with Pds5 (the Pds5A or Pds5B form) but also with Wapl (Fig. 4A; Fig. S4A). We found no GFP signal on mitotic chromatin in Rec8–Stag3-expressing cells, which is in line with the suspected Wapl-dependent removal of meiotic cohesin from chromosome arms in early mitosis (Fig. 4B). An siRNA-mediated depletion of Wapl extends the residence time of Scc1 cohesin on chromosome arms such that it becomes detectable on pro(meta)phase chromatin (Buheitel and Stemmann, 2013). We expected the same for meiotic cohesin and, hence, were surprised that we were unable to detect GFP-positive mitotic chromosomes in Rec8–Stag3-expressing cells upon depletion of WAPL with siRNA (Fig. 4B; Fig. S4B). This could be explained if, in the absence of Wapl, the increased level of chromatin-associated endogenous Scc1 out-competed Rec8–GFP, thus masking the effect of Wapl depletion. Hence, we partially depleted Scc1 to reduce the competition between endogenous Scc1 and Rec8–GFP without causing a cohesion defect. Indeed, co-depletion of Wapl and Scc1 resulted in a significantly increased Rec8–GFP signal on mitotic chromosomes of nocodazole-arrested Rec8–Stag3-expressing cells, while depletion of Scc1 alone had no such effect (Fig. 4B; Fig. S4B). We, thus, conclude that Wapl binds Rec8–Stag3 cohesin and is able to remove these meiotic ring complexes from chromatin. Consistent with this, recent studies have demonstrated an important function of Wapl during meiosis in A. thaliana, S. cerevisiae and C. elegans (Challa et al., 2016; Crawley et al., 2016; De et al., 2014).

Sororin antagonises the anti-cohesion activity of Wapl in somatic metazoan cells (Nishiyama et al., 2010; Rankin et al., 2005). Hence, we asked whether sororin protects also meiotic cohesin. First, we tested for physical interaction between sororin and Rec8. Indeed, sororin specifically co-purified not only with Scc1 but also with Rec8 from nuclease-treated lysates of corresponding G2-arrested transgenic cells (Fig. 4C; Fig. S4A). A reciprocal experiment where sororin was precipitated from cell lysates confirmed the interaction (Fig. S4C). Next, we assessed the functional relevance of this interaction. Depletion of sororin leads to untimely SCS in mitosis due to failure in protection of centromeric cohesin from Wapl (Nishiyama et al., 2010; Rankin et al., 2005). We hypothesised that the presence of Rec8 and Stag3 should fail to rescue this cohesion defect if Rec8-containing cohesin was protected from Wapl in a sororin-dependent manner. We transfected parental Hek293 and Rec8–Stag3-expressing cells with a sororin (CDCA5)-targeting siRNA (or GL2 as a control), arrested the cells in prometaphase and performed chromosome spreads. As expected, control cells showed severe premature SCS upon knockdown of sororin (Fig. 4D). Importantly, Rec8–Stag3-expressing cells also suffered from loss of cohesion when sororin was limiting, albeit slightly less than control cells. In summary, we propose that Wapl can remove meiotic cohesin rings from chromatin and that sororin antagonises this process.

In S. pombe, forced co-expression of Rec8 and the meiotic shugoshin in mitotically dividing cells is toxic (Kitajima et al., 2004). Therefore, it is surprising that expression of functional Sgo2-protected Rec8 cohesin goes ‘unpunished’ in human cells. This can only be explained if inactivation of shugoshin and separase-dependent cleavage of Rec8 fail in yeast mitosis but are properly mimicked by human somatic cells. How shugoshin is inactivated in meiosis has recently been worked out for S. cerevisiae but likely occurs by a different, as yet speculative, mechanism in mammals (Argüello-Miranda et al., 2017; Chambon et al., 2013; Gómez et al., 2007; Jonak et al., 2017; Krishnan et al., 2017; Moshkin et al., 2013). Our work now offers the opportunity to address this important issue in a human model system.

Here, we extend previous studies (Gutiérrez-Caballero et al., 2011; Rong et al., 2017) and establish that certain aspects of human meiosis-specific cohesin subunits can be functionally assessed in somatic cell culture. This system avoids ethical concerns regarding human meiosis research, is amenable to rapid experimental manipulation, and offers unlimited material for cell biological and biochemical analyses. Although it cannot replace the investigations of meiocytes it can quickly generate testable predictions. Beginning to exploit this system, we show (1) that Rec8-containing chromosomal cohesin is susceptible to removal by Wapl and is protected from this by sororin; (2) that Sgo2 can only shield Rec8 cohesin but not Scc1 cohesin; and (3) that phosphorylation and cleavage of Rec8, as well as inactivation of Sgo2, can be studied in somatic cells.

FlpIn T-Rex Hek293 cells (Thermo Fisher Scientific) were cultured in high-glucose Dulbecco's modified Eagle's medium (Biowest) plus 10% fetal bovine serum (Biowest) at 37°C in a humidified 5% CO₂ incubator. The cell line was authenticated by selection with Zeocin selection reagent (Thermo Fisher Scientific) and routinely tested for mycoplasma contamination. Transfection was performed using a calcium phosphate-based method. cDNAs encoding SCC1-EGFP and REC8-EGFP were cloned into the pcDNAL/FRT/TO vector (Buheitel and Stemmann, 2013). Flp-recombinase-mediated integration of this construct into the unique FRT site of the parental FlpIn T-Rex Hek293 cell genome allowed for a second round of site-specific integration by means of Cre recombinase. To this end, cDNAs encoding FLAG₃-STAG2 and FLAG₃-STAG3 were cloned into pcDNA5/loxP/TO (Buheitel and Stemmann, 2013) and transfected together with pIC-Cre (Gu et al., 2013). At 48 h after transfection of the plasmids, cells were selected with hygromycin (90 μg/ml) or geneticin (120 μg/ml). Individual clones resulting from selection for 1 to 2 weeks were analysed by western blotting for Dox-induced expression of the transgene(s). To induce prometaphase arrest, the culture medium was supplemented with 0.2 µg/ml nocodazole (Sigma). S-phase arrest was induced by addition of 2 mM thymidine (Applichem) in the culture medium.

Myc was detected in western blots with a mouse monoclonal antibody from the Developmental Studies Hybridoma Bank (9E10, affinity purified over protein G-sepharose, 0.2 μg/ml). In immunofluorescence microscopy, Myc was stained with a mouse monoclonal antibody from Millipore (4A6, 1:1500). Scc1 probing in western blots was performed using a commercial mouse monoclonal antibody from Millipore (05-908, 1:1000). A mouse monoclonal antibody against Wapl was raised against its N-terminal 88 amino acids, affinity-purified using the antigenic peptide, and used in immunoblots at 2 μg/ml. A rabbit polyclonal antibody against sororin was raised against bacterially expressed full-length protein, affinity-purified using the antigenic protein, and used in immunoblots at 1.7 μg/ml. This antibody was also coupled to protein-A–Sepharose (GE Healthcare) and used for sororin IP. Susannah Rankin (Department of Cell Biology, University of Oklahoma, USA) generously provided polyclonal rabbit antibodies against Xenopus Stag1, Smc3 and Pds5A (used at 3.5 µg/ml, 1 µg/ml and 1.5 µg/ml, respectively), which are known to cross-react with human proteins. The polyclonal rabbit antibodies against human Stag2 and Pds5B (used at 3.3 µg/ml and 1.4 µg/ml, respectively), were kind gifts from Jan-Michael Peters (Mitosis and Chromosome Biology, Research Institute of Molecular Pathology, Austria). Two commercial polyclonal rabbit antibodies were used for detection of Sgo1 (Abcam, ab21633, 1:300), and for IP and detection of Sgo2 (Bethyl, A301-261A, 1:600). GFP was detected in immunofluorescence microscopy by means of a polyclonal rabbit antibody (Herzog et al., 2013), and in immunoblots by a mouse monoclonal antibody (Hellmuth et al., 2015) affinity-purified over protein G–Sepharose and used at 4 μg/ml. For α-tubulin staining in western blots, we used a mouse monoclonal antibody from the Developmental Studies Hybridoma Bank (12G10, hybridoma supernatant at 1:200). Hec1 was stained in immunofluorescence microscopy experiments by using a mouse monoclonal antibody from Genetex (GTX70268, 1:800). For immunofluorescence staining of the centromeric region, we used a human autoantibody (CREST) from ImmunoVision (hct-0100, 1:1000). Unspecific rabbit IgG were purchased from Bethyl.

Secondary antibodies for western blot were horseradish peroxidase (HRP)-conjugated goat anti-rabbit and anti-mouse IgGs (Sigma Aldrich, both used at 1:15,000). Secondary antibodies for immunofluorescence microscopy (all 1:500) were: goat anti-rabbit-IgG conjugated to Alexa Fluor 488, goat anti-rabbit-IgG conjugated to Cy3 and anti-mouse-IgG conjugated to Cy3 (all from Life Technologies) and goat anti-human-IgG Cy3 (Bethyl).

For depletion of SCC1 and WAPL mRNAs, two siRNAs each were used. In case of SCC1 both siRNAs were used at a concentration of 50 nM. WAPL siRNAs were used at a concentration of 70 nM. These siRNAs and the siRNA targeting luciferase (GL2) were as described previously (Schöckel et al., 2011). The SGO2-specific siRNA sequence and concentration was also previously published (Orth et al., 2011). The siRNA for depletion of sororin (CDCA5) targeted the 5′UTR, had the sequence 5′-UGGAGGAGCUCGAGACGGATT-3′, and was used at a concentration of 120 nM. The siRNA against SGO1 had the sequence 5′-GAUGACAGCUCCAGAAAUUTT-3′ and was used at a concentration of 50 nM.

Immunofluorescence microscopy was performed essentially as described previously (Mohr et al., 2015). Briefly, cells were grown on poly-L-lysine-coated coverslips and washed once with PBS and fixed with 3.7% (w/v) paraformaldehyde in PBS for 10 min. Where indicated, the cells were pre-extracted with 0.2% (v/v) Triton X-100 in PBS prior to fixation. Afterwards, cells were treated with 100 mM glycine in PBS to quench residual fixative followed by a PBS wash. All fixed cells were permeabilised by incubation in 0.5% (v/v) Triton X-100 in PBS for 5 min. After washing once with PBS, samples were incubated in 1% (w/v) bovine serum albumin (BSA) in PBS overnight at 4°C. Coverslips were transferred onto parafilm, placed in a wet chamber and incubated at room temperature with the corresponding primary antibody diluted in 1% (w/v) BSA in PBS. After four washes with 0.1% (v/v) Triton X-100 in PBS (WS) coverslips were incubated with a dilution of fluorescently labelled secondary antibodies for 40 min. The samples were washed once with WS, once with 1 μg/ml Hoechst 33342 in WS and again four times with WS. Finally, the coverslips were mounted onto glass slides in 2.33% (w/v) diazabicyclo-[2,2,2]-octane, 20 mM Tris-HCl pH 8.0 in 78% (v/v) glycerol and imaged using a Zeiss Axioplan 2 fluorescence microscope with a Plan-Apochromat 100×1.40 NA oil DIC objective, an AxioCam MRm CCD camera and AxioVision software version 4.8.2.0. Note that the images within an experiment showing cells with or without preextraction were recorded with different exposure times.

Chromosome spreads were prepared as described previously (Hellmuth et al., 2015).

IPs were performed essentially as described (Hellmuth et al., 2015). Briefly, cells were harvested and lysed in lysis buffer [20 mM Tris-HCl (pH 7.7), 100 mM NaCl, 5 mM MgCl₂, 0.1% (v/v) Triton X-100, 5% (v/v) glycerol, 10 mM NaF, 20 mM β-glycerophosphate, 1 tablet/50 ml Complete protease inhibitor, EDTA-free (Roche)]. The lysates were centrifuged at 16,000 g and 4°C for 30 min. Where indicated, the chromatin in the lysates was digested with benzonase nuclease (ChemCruz) prior to centrifugation. The corresponding beads were added to the lysates and after a 3 h incubation at 4°C the beads were washed three times with lysis buffer (200 g, 1.5 min, 4°C) and eluted by boiling in SDS sample buffer. Coomassie staining was performed as described (Candiano et al., 2004).

No statistical calculations were used to predetermine sample sizes. Our sample sizes represent those generally used in this field of research. All western blotting and IP experiments were independently repeated a minimum of two times with similar results. Immunofluorescence microscopy experiments without quantification were independently repeated at least three times and each time a minimum of 50 cells were assessed.

For chromosome spread analysis the investigators were blinded to sample allocation.

We thank Monika Ohlraun, Jutta Hübner, and Markus Hermann for technical assistance and Katja Wassmann, Stefan Heidmann and Philip Kahlen for critical reading of the manuscript. Please note that some data in the paper form part of the PhD thesis of P.G.W. (University of Bayreuth, 2017).