Earlier models suggest that cohesin’s ATPase head domain is only involved in the initial DNA binding step, and that ATP hydrolysis releases the DNA from cohesin. These models predict that stably DNA-bound cohesin should not show ATPase activity. However, recent literature suggests that Eco1 might promote cohesion by regulating the cohesin ATPase after the stable DNA binding of cohesin. If ATPase activity is required to regulate this second step of cohesion establishment, we should be able to observe ATPase activity for purified cohesin-DNA complexes. To test this possibility, we purified Schizosaccharomyces pombe cohesin and loader complexes as described previously (Murayama and Uhlmann, 2014). Purified cohesin showed basal ATPase activity that was stimulated by the presence of the loader complex and DNA, and abolished by a Walker A mutation in Smc3 (Figure 1—figure supplement 1), similar to published results (Murayama and Uhlmann, 2014). Cohesin binding to chromosomes in vivo is enriched at centromeres and at cohesion-associated regions (CARs) along the chromosome arms (Laloraya et al., 2000; Onn et al., 2008), and is highly salt resistant (Ciosk et al., 2000). To purify stable cohesin-DNA complexes that are physiologically relevant, we assembled cohesin on DNA molecules that contained CARC1 sequence and were coupled by both ends to dynabeads. Cohesin and its loader were incubated with DNA-beads under low salt conditions (25 mM KCl, 25 mM NaCl). The cohesin-DNA bead mix was washed with high salt (500 mM KCl) to remove any free cohesin or cohesin not stably bound to DNA (Figure 1B). The cohesin that remained bound to the DNA-beads was then eluted and quantified by Coomassie staining or Western blots. In the presence of the loader, 20% of the input cohesin was bound to DNA-beads after the high salt wash (Figure 1C,D). In the absence of the loader, 2-fold less cohesin bound to DNA (Figure 1D). Cohesin did not bind to beads that lack DNA (Figure 1C). In addition, this stable population of cohesin on DNA-beads could be eluted from the beads by either a restriction enzyme digest or a DNase treatment (Figure 1—figure supplement 2). These results suggest that cohesin bound specifically to the DNA that was coupled to beads, and did so in a salt-resistant and loader-inducible manner. These DNA binding features recapitulated the properties of stable cohesin-DNA complexes assembled in vivo (Ciosk et al., 2000).

To address whether stably DNA-bound cohesin was an active ATPase, we next assessed whether cohesin in our purified, stable cohesin-DNA complexes (CD-B, Figure 1B) retained its ATPase activity. DNA-beads were incubated with cohesin and loader, then washed with high salt buffer to remove free and unstably bound cohesin, as described above. Cohesin-DNA beads (CD-B) were then resuspended in buffer containing ATP to measure ATPase activity of CD-B, compared to basal and loader/DNA-stimulated activities of cohesin. No ATPase activity was detected when cohesin was omitted from the reaction (Figure 1E and Figure 1—figure supplement 1). As shown before (Murayama and Uhlmann, 2014), cohesin’s basal ATPase activity (sample 2) was stimulated about 3-fold by the presence of loader and DNA (sample 4), but only subtly increased by DNA alone (sample 3, Figure 1E). Our purified cohesin-DNA complexes on beads (CD-B, sample 5) showed about 4-fold increased ATPase activity compared to the basal activity of equal amounts of cohesin (sample 2, Figure 1E,F). Furthermore, the ATPase activity of CD-B was at least as high, if not higher, as the activity of equal concentrations of loader/DNA-stimulated cohesin in solution, where additional rounds of cohesin loading were possible (sample 4 compared to sample 5, Figure 1E,F).

The ATPase activity in the fraction containing purified cohesin-DNA complexes might have been derived from DNA-free cohesin that had dissociated from DNA-beads. To test this possibility, a parallel sample of CD-B was treated similarly to the ATPase assay to assess the amount of cohesin that remained bound to DNA through the course of the ATPase experiment. DNA-beads were separated from the supernatant after the 30-minute incubation with ATP and the amount of cohesin in each fraction was visualized by SDS-PAGE. Very little, if any, cohesin was found in the supernatant at the end of the incubation, suggesting that most of the ATPase activity was derived from cohesin bound to the DNA-beads (Figure 1G). Even in the presence of excess competitor DNA in solution, cohesin remained stably bound to the DNA-beads (Figure 1—figure supplement 3). Thus in vitro, cohesin can have robust ATPase activity while remaining stably bound to DNA.

In our preparations of purified cohesin-DNA complexes, a small amount of the loader complex was also present, due to the ability of the loader complex to bind DNA. Given that the loader complex stimulates cohesin’s ATPase activity in the presence of DNA, it was possible that the ATPase activity in our purified cohesin-DNA complexes depended upon the presence of the loader complex. To test this possibility, we repeated these experiments but omitted the loader from the binding reaction. Under these conditions, a smaller percentage of cohesin could be assembled on DNA in a salt-resistant manner (Figure 1D and Figure 1—figure supplement 4). Cohesin in these purified cohesin-DNA complexes that have no co-purifying loader complex also retained ATPase activity equivalent to that seen in the mixture of cohesin and DNA in solution (Figure 1—figure supplement 4). Thus, the persistence of ATPase activity in stable cohesin-DNA complexes was not dependent upon the loader. The presence of cohesin’s ATPase activity in the stable cohesin-DNA complex is consistent with a regulatory role for cohesin ATPase in cohesion establishment in vivo after cohesin stably binds to DNA.

Cohesin exhibited robust ATPase activity after stably binding DNA, which encouraged us to re-examine the in vivo role of cohesin ATPases and Eco1 in cohesin function. Cohesin-DNA complexes in eco1∆ wpl1∆ cells are stably bound to DNA but fail to establish cohesion, indicating a post-DNA binding step is required for cohesion. We postulated that this failure to generate cohesion was due to a misregulation of the cohesin ATPase active sites. If so, a subset of mutations altering the Smc1 or Smc3 ATPase active sites might mimic the state of the ATPase upon Eco1 acetylation, and thereby might suppress the cohesion defect of eco1∆ wpl1∆ cells.

To identify such suppressors, we exploited the fact that eco1∆ wpl1∆ cells remain viable in the absence of cohesion because of an unusual feature of budding yeast cell cycle that gives rise to a cohesin-independent mechanism of sister chromatid segregation (Guacci and Koshland, 2012). The reliance of eco1∆ wpl1∆ cells on this cohesin-independent segregation makes them sensitive to the microtubule depolymerizing drug benomyl. A screen for suppressors of the benomyl sensitivity of eco1∆ wpl1∆ cells should identify a subset of mutations that restore cohesion and the cohesin-dependent benomyl-resistant mechanism of segregation. Indeed, we recently reported one such benomyl-resistant suppressor allele, smc3-D1189H (Guacci et al., 2015), that restored cohesion. smc3-D1189H is located in the ATPase head domain, near the Smc3 ATPase active site. Due to this restoration of cohesion to eco1∆ wpl1∆ cells, we termed this mutation a cohesion activator mutation.

Here, we present two new SMC1 alleles from our screen, smc1-D1164E and smc1-Y1128C. Re-introduction of these mutations into the parent eco1∆ wpl1∆background generated benomyl-resistance identical to the initial suppressor eco1∆ wpl1∆ isolates and close to the resistance of wild type cells (Figure 2A), demonstrating their causal role for this phenotype. The SMC1 suppressor mutations are even more proximal to the Smc3 ATPase active site than smc3-D1189H (Figure 2B). smc1-Y1128C is adjacent to the signature motif, which is thought to modulate ATP binding, but this residue is not conserved. smc1-D1164E alters the invariant aspartate that is part of the conserved D-loop motif found in ABC and SMC ATPases (Figure 2C). Studies from other ABC ATPases suggest that this aspartate interacts with the catalytic H loop and Walker A (Hohl et al., 2012; la Rosa and Nelson, 2011), and mediates communications between the two ATPase active sites within the ATPase head domain (Furman et al., 2013). Taken together, it is likely that these suppressor mutations modulate the cohesin ATPase to promote cohesion.

Figure 2

Download asset Open asset

To begin to test this hypothesis, we first asked whether smc1-D1164E or smc1-Y1128C were also cohesion activator mutations, i.e. suppressed the severe cohesion defect of eco1∆ wpl1∆ cells. Haploid wild type cells and eco1∆ wpl1∆ parent cells containing SMC1, smc1-D1164E or smc1-Y1128C alleles were arrested in G1, then released into media containing nocodazole to arrest them in M phase (Figure 3A). To assay for cohesion, these cells contained a LacO array integrated at a CEN-proximal locus (TRP1), and a GFP-LacI fusion protein that bound to the array. In this assay, sister chromatids that have cohesion generate a single GFP spot, whereas chromatids lacking cohesion generate 2 GFP spots (Figure 3A). Wild type cells had robust cohesion as shown by few M phase cells with two GFP spots (Figure 3B and Figure 3—figure supplement 1A). eco1∆ wpl1∆ cells exhibited a severe cohesion defect (more than 60% of cells had two GFP spots), which was reduced by half by the smc1-D1164E suppressor (Figure 3B and Figure 3—figure supplement 1A). The smc1-Y1128C suppressor also reduced the eco1∆ wpl1∆ cohesion defect, but less effectively (Figure 3—figure supplement 2). Thus, both of these new suppressor mutations, like smc3-D1189H, suppressed significantly but not completely the cohesion defect of eco1∆ wpl1∆ cells. The increased CEN-proximal cohesion generated by smc1-D1164E or smc1-Y1128C was likely responsible for the increased benomyl resistance in the eco1∆ wpl1∆ background. Subsequent studies focused on smc1-D1164E, because its alters a critical and absolutely conserved D-loop aspartate, and it more robustly suppresses the cohesion defect of eco1∆ wpl1∆ cells.

Figure 3 with 6 supplements see all

Download asset Open asset

To better understand the impact of the smc1-D1164E suppressor on cohesion, we monitored cohesion at the CEN-proximal TRP4 locus at 20 minute intervals as cells progressed from G1 to M (Figure 3—figure supplement 1B). As expected, wild type cells showed robust cohesion where very few cells had separated sisters, and eco1∆ wpl1∆ cells showed sister chromatid separation starting in S phase (Figure 3C), consistent with a cohesion establishment defect that was shown previously for this strain (Guacci and Koshland, 2012; Guacci et al., 2015). The small amount of cohesion loss in smc1-D1164E eco1∆ wpl1∆ cells also began during S phase, indicating that the residual cohesion defect in these cells resulted from a failure in cohesion establishment. However, these cells displayed fewer separated sisters throughout the time course compared to eco1∆ wpl1∆ cells. These results indicated that (1) smc1-D1164E promotes cohesion establishment in the eco1∆ wpl1∆, albeit incompletely, and (2) once cohesion is formed in smc1-D1164E eco1∆ wpl1∆ cells, it is maintained. smc1-D1164E also partially restored cohesion establishment in eco1∆ wpl1∆ cells at the CEN-distal locus, LYS4 (Figure 3—figure supplement 3). We conclude that smc1-D1164E promotes cohesion establishment in the eco1∆ wpl1∆, albeit incompletely.

Since cohesin is already stably bound to DNA in eco1∆ wpl1∆ cells, the restoration of cohesion in smc1-D1164E eco1∆ wpl1∆ cells should occur without altering the DNA binding of cohesin. To test this prediction, we performed chromosome immunoprecipitation (ChIP) to compare cohesin binding to chromosomes in eco1∆ wpl1∆ cells containing SMC1 or smc1-D1164E allele. Cohesin subunits colocalize on chromosomes, making the analysis of any cohesin subunit a surrogate for cohesin binding (Glynn et al., 2004; Heidinger-Pauli et al., 2010a; Lengronne et al., 2004). Here we used anti-Mcd1 antibodies as a means to assess cohesin binding to chromosomes. Cohesin binding at CARs has been shown to be reduced 2-3 fold in eco1∆ wpl1∆ cells as compared to wild type cells (Guacci et al., 2015; Sutani et al., 2009). In eco1∆ wpl1∆ cells containing either SMC1 or smc1-D1164E allele, there was little or no difference in Mcd1 binding at the centromere-proximal CARC1 or centromere-distal CARL1 (Figure 3—figure supplement 4). This similarity in chromosomal binding indicates that cohesion restoration in eco1∆ wpl1∆ cells by the smc1-D1164E allele was not due to an altered level of cohesin binding to chromosomes, but instead was due to a change in cohesin function at a post-DNA binding step.

The restoration of cohesion in smc1-D1164E eco1∆ wpl1∆ cells indicated that the smc1-D1164E could bypass one or more aspects of cohesion regulation normally imposed by Eco1 and Wpl1. To understand better how smc1-D1164E, and by inference the Smc3 ATPase active site, interfaced with cohesin regulators, we examined its impact on the phenotypes of wpl1∆ and eco1∆single mutants. In a wpl1∆ background, SMC1and smc1-D1164E alleles had similar viability and the same moderate cohesion defect characteristic of wpl1∆ mutants (Figure 3—figure supplement 5). The smc1-D1164E allele suppressed the essential ECO1 function, as smc1-D1164E eco1∆ cells showed good viability (Figure 3—figure supplement 6). smc1-D1164E eco1∆ cells showed a 35% cohesion defect (Figure 3—figure supplement 6), which was greatly reduced compared to 65-70% cohesion defect that was observed previously for eco1 mutant cells or cells lacking both Eco1 and Wpl1 (Guacci et al., 2015; Sutani et al., 2009). Taken together, these results suggest that smc1-D1164E restored cohesion by phenocopying both Eco1’s antagonism of Wpl1 and its ability to promote stably DNA-bound cohesin to tether sister chromatids.

The partial restoration of cohesion in smc1-D1164E eco1∆ wpl1∆ cells could reflect the inability of smc1-D1164E to fully compensate for the loss of Eco1 or Wpl1. Alternatively, it might reflect an inherent cost of uncoupling cohesin from its regulators. To differentiate between these possibilities, we characterized the smc1-D1164E allele when both ECO1 and WPL1 were present. In plasmid shuffle assays (Materials and methods), the viability of cells containing the smc1-D1164E or SMC1 alleles was indistinguishable (Figure 4—figure supplement 1). However, otherwise wild type cells carrying the smc1-D1164E allele had a moderate cohesion defect at both CEN-proximal and CEN-distal loci, which arose around the time of replication (Figure 4—figure supplement 2). This defect was similar to that observed for the smc1-D1164E eco1∆ wpl1∆ cells. Thus, cohesin in smc1-D1164E cells had an inherent moderate cohesion establishment defect independent of cohesin regulators. Taken together, our results suggest that the smc1-D1164E allele was constitutively cohesive as it uncoupled cohesin function from its regulators, which reduced the efficiency of cohesion generation.

Our two new cohesion activator mutations mapped to residues intimately involved in the Smc3 ATPase active site, and the previously characterized third cohesion activator was in close proximity (Guacci et al., 2015). This observation suggested that the Smc3 but not the Smc1 ATPase active site might play a specialized role in activating stably DNA-bound cohesin to tether sister chromatids.

To test more directly whether the two active sites played distinct roles in cohesin regulation, we compared the phenotypes of smc1-D1164E to smc3-D1161E, the analogous glutamate substitution of the D-loop aspartate of the Smc1 ATPase active site (Figure 4A). We used Auxin Inducible Degron (AID) system to obtain conditional alleles of SMC1 and SMC3, SMC1-AID and SMC3-AID, respectively, which can be degraded upon the addition of auxin (Figure 4—figure supplement 3). In strains containing either SMC1-AID or SMC3-AID alleles as sole source of Smc1 or Smc3, we integrated our test alleles, smc1-D1164E and smc3-D1161E, respectively. The smc1-D1164E allele robustly suppressed the inviability associated with depletion of Smc1-AID, whereas smc3-D1161E could not support viability when Smc3-AID was depleted (Figures 4B,C). Similar results in viability were seen when these alleles were tested by plasmid shuffle assays (Figure 4—figure supplement 1 & 4). This difference in viability between the smc1-D1164E and smc3-D1161E alleles suggests that Smc1 and Smc3 ATPase active sites have distinct functions.

Figure 4 with 6 supplements see all

Download asset Open asset

We next assessed cohesion at CEN-proximal and CEN-distal loci in smc1-D1164E SMC1-AID and smc3-D1161E SMC3-AID strains under conditions in which AID tagged subunits remain depleted from G1 through M phase (Figure 4D). As expected, SMC1-AID and SMC3-AID cells showed severe cohesion defects that could be rescued by SMC1 and SMC3, respectively (Figure 4E and Figure 4—figure supplement 5A). smc1-D1164E SMC1-AID cells had only a modest cohesion defect, in line with our previous results for the smc1-D1164E in an otherwise wild type background. In contrast, in the presence of auxin, the smc3-D1161E SMC3-AID cells were severely compromised for cohesion, at levels comparable to cells that only have the SMC3-AID allele (~90% separated sisters, Figure 4F and Figure 4—figure supplement 5B). Similar results were obtained when cohesion was assessed at the CEN-distal locus (Figure 4—figure supplement 6). Thus, smc1-D1164E and smc3-D1161E differ in their ability to promote viability and cohesion.

To understand the molecular basis for this difference in cohesion, we performed ChIP on this panel of strains to assess cohesin binding to chromosomes under conditions in which the AID-tagged proteins were depleted. The level of cohesin binding to chromosomes in smc1-D1164E SMC1-AID cells was reduced about twofold compared to SMC1 SMC1-AID (Figure 5A and Figure 5—figure supplement 1). In contrast, cohesin binding to chromosomes in smc3-D1161E SMC3-AID cells was essentially at the background levels observed in SMC3-AID-only cells (Figure 5B and Figure 5—figure supplement 1). Taken together, these data suggest that the Smc1 ATPase active site regulates cohesin ATPase to modulate cohesin’s DNA binding, whereas the Smc3 active site regulates cohesin ATPase to modulate cohesin-mediated DNA tethering after its stable binding to DNA.

Figure 5 with 2 supplements see all

Download asset Open asset

Precedent that the D-loop mutations might impact ATPase activity come from studies of the homodimeric SMC protein Rad50. The mutation of the D-loop aspartate in Rad50 to an alanine (Rad50^DA) dramatically reduced its ATPase activity (la Rosa and Nelson, 2011) (Figure 5—figure supplement 2). Furthermore, the substitution of the D-loop aspartate with a glutamate (Rad50^DE) led to a 3-fold reduction in ATPase activity (Figure 5—figure supplement 2).

To address how the cohesin ATPase activity would be affected by these substitutions, we purified the S. pombe cohesin complex with the analogous mutations (Murayama and Uhlmann, 2014) and assayed the ATPase activity (a sum of the ATP hydrolysis by Smc1 and Smc3 active sites) for equivalent amounts of wild type and mutant cohesin complexes. The mutation in the S. pombe homolog of cohesin analogous to smc1-D1164E (SpSmc1^DE), which was competent for cohesion and viability in S. cerevisiae, reduced the ATPase activity of cohesin to levels close to the Smc3 walker A mutant (SpSmc3^WA, Figure 5C). In contrast, the D-loop-E mutation in the Smc1 ATPase active site (SpSmc3^DE), analogous to the smc3-D1161E allele that abolished chromosomal association of cohesin, led to a subtler reduction in cohesin’s total ATPase activity. These results suggested that the substitution of the D-loop aspartate with a glutamate in Smc3 and Smc1 ATPase active sites affect cohesin in unique ways that uncouple the level of ATPase activity from DNA tethering and DNA binding.

The novel and distinct phenotypes of smc1-D1164E and smc3-D1161E might result from a subtle change that alters rather than abrogates D-loop function, given the chemical similarity of aspartate and glutamate. To test whether these unusual phenotypes persisted with more radical substitutions of the D-loop aspartate, we changed it to alanine in the Smc1 or Smc3 ATPase active sites. We introduced these smc1-D1164A or smc3-D1161A alleles into strains bearing SMC1-AID or SMC3-AID alleles, respectively (Figure 6A). We then characterized these alleles under conditions in which the AID-tagged proteins were degraded (Figure 6—figure supplement 1). Neither smc1-D1164A nor smc3-D1161A allele was able to sustain viability on auxin plates (Figures 6B,C). Similar results in viability were seen using plasmid shuffle assays (Figure 6—figure supplement 2). Moreover, the smc1-D1164A or smc3-D1161A cells were both severely compromised for cohesion, although the cohesion defect in smc3-D1161A was more severe (Figures 6D,E, Figure 6—figure supplements 3 and 4). ChIP using anti-Mcd1 antibodies showed that smc1-D1164A had 7-10 fold less cohesin binding than wild type, whereas smc3-D1161A reduced cohesin binding to chromosomes to background levels seen with the SMC3-AID alone (Figures 7A,B). Similar results were obtained when tagged smc1-D1164A and smc3-D1161A proteins were assessed by ChIP (Figure 7—figure supplement 1). Therefore, the glutamate substitution in the D-loop of the Smc3 ATPase active site is unique in its ability to support cohesion suggesting it alters rather than abolishes the in vivo function of the Smc3 ATPase active site.

Figure 6 with 4 supplements see all

Download asset Open asset

Figure 7 with 1 supplement see all

Download asset Open asset

Finally, we asked whether analogous aspartate-to-alanine substitution mutations altered the ATPase activity of purified S. pombe cohesin. Both mutations (SpSmc1^DA and SpSmc3^DA) reduced the ATPase activity of cohesin to levels comparable to the walker A mutation (SpSmc3^WA, Figure 7C). The similar severe defects in ATPase activities for the S. pombe Smc1^DE and Smc1^DA cohesin complexes was striking given the dramatic differences in the analogous complexes to promote viability, cohesion and cohesin binding to DNA in S. cerevisiae (Figures 4–7). This functional difference reinforces the conclusion that the cohesin’s DNA binding and DNA tethering activities can be uncoupled from the level of cohesin ATPase activity.

Until this study, there was no evidence for a role for the ATPase regulating post-DNA binding steps. Here, we show that cohesin retains its ATPase activity after stably binding to DNA in vitro, implying ATP hydrolysis plays an additional role(s) after cohesin binds DNA. We also show that cohesion activator mutations in key residues of the Smc3 ATPase active site, smc1-D1164E and smc1-Y1128C, suppress the severe cohesion defect of eco1∆ wpl1∆ cells, similar to the previously described smc3-D1189H allele (Guacci et al., 2015). These cohesion activators restore cohesion in eco1∆ wpl1∆ cells without increasing the amount of cohesin stably bound to DNA (this study, Guacci et al., 2015). Lastly, a mutation analogous to one of these activator mutations, smc1-D1164E, abolishes most, if not all, cohesin ATPase activity when introduced in purified S. pombe cohesin complex. Taken together, our results suggest the robust ATPase activity of DNA-bound cohesin is physiologically relevant, likely acting as an inhibitor of the conversion of stably bound cohesin to a form capable of tethering sister chromatids.

We also show that the cohesion activator phenotype of the smc1-D1164E allele is a unique feature of the D-loop of the Smc3 ATPase active site. The analogous substitution in the D-loop of the Smc1 ATPase active site, smc3-D1161E, not only fails to act as a cohesion activator but also is unable to support viability, fails to load cohesin to DNA, and cannot generate cohesion. These results suggest that the Smc1 and Smc3 ATPase active sites differentially affect cohesin ATPase function and that the Smc3 ATPase active site has a distinct function in the generation of cohesion after stable DNA binding of cohesin. This distinct function provides an explanation for the evolutionary diversion of the SMC subunits. Indeed, use of asymmetrical ATPases for regulation purposes is common to ABC transporters (Antony and Hingorani, 2004; Furman et al., 2013; Yang et al., 2003). We suspect the differential roles of cohesin ATPase active sites and a specialized role for the Smc3 active site in DNA tethering was missed previously, in part due to their composite nature and the particular mutations that have been analyzed to date. How the D-loop of the Smc3 ATPase active site executes its specialized function in regulating the cohesin ATPase remains unclear. The D-loop may impact the nucleotide state only in the Smc3 ATPase active site. Alternatively, it could alter the nucleotide state of both active sites. Support for the latter comes from studies on the ABC ATPase domains in ABC transporters, which suggest that ATP hydrolysis by the two sites is cooperative (Holland and Blight, 1999), and that D-loops are involved in mediating the communicaton between the two active sites within the ATPase (Furman et al., 2013; Hohl et al., 2012, 2014).

The phenotypes of the D-loop mutants in the Smc1 and Smc3 active sites did not correlate with the ATPase activities of the corresponding purified S. pombe cohesin complexes. For example, cohesin complexes with aspartate-to-glutamate substitution in the D-loop of the Smc1 active site had significant ATPase activity in vitro, but were severely perturbed for DNA binding in S. cerevisiae. In addition, cohesin complexes with aspartate-to-glutamate substitution in the D-loop of the Smc3 ATPase had low levels of ATPase activity indistinguishable from the walker A mutant, but could promote DNA binding and cohesion in S. cerevisiae. While it is possible that this lack of correlation could be explained by our use of the ATPase activity of the S. pombe wild type and mutant cohesins as an approximation of the S. cerevisiae cohesin, this explanaition seems unlikely. The aspartate of the D-loop is absolutely conserved in all SMC complexes and highly conserved among very diverse ABC ATPase modules. Our analysis of analogous substitutions in the homodimeric Rad50 protein shows similar reductions in the ATPase activity. Finally, many published biochemical and structural studies have used cohesin complexes from different organisms as a proxy for S. cerevisiae cohesin to provide physiologically relevant insights. Intriguingly, the uncoupling of ATPase activity and function by D-loop mutations has been observed for several other ABC ATPases (Furman et al., 2013; Grossmann et al., 2014).

Therefore, we favor a model in which cohesin’s functions and conformations could be coordinated by the different states in the hydrolysis cycle of ATPase active sites. For example, the tethering-competent state of cohesin could require that the Smc3 active site be bound with ADP+Pi. The smc1-D1164E mutation may mimic the ADP-Pi state, whereas the smc1-D1164A mutant could be blocked elsewhere in the ATPase cycle (ATP bound or nucleotide free). Both the smc1-D1164E and smc1-D1164A would dramatically reduce overall ATPase activity but only the smc1-D1164E would promote cohesin function. A similar case of trapping the ADP+Pi state is hypothesized for the MRP1 ABC transporter, for a glutamate substitution of an aspartate residue (Qin and Cao, 2008). Elucidating the exact relationship between the particular nucleotide states of the Smc1 and Smc3 ATPase active sites and cohesin functions will likely require in vitro reconstitution of cohesin’s various functions and detailed structural studies.

The phenotypes of smc1-D1164E cells suggest that Eco1 modulation of cohesin’s ATPase impacts two distinct modes of cohesin regulation. First, smc1-D1164E suppresses the inviability of eco1∆ cells, which normally die because they are unable to inhibit Wpl1 (Rolef Ben-Shahar et al., 2008; Rowland et al., 2009; Sutani et al., 2009). This result suggests that Eco1 acetylation alters the Smc3 ATPase active site to make cohesin resistant to Wpl1. Second, smc1-D1164E also suppresses the cohesion defect in eco1△ wpl1△ cells, suggesting that it phenocopies the second function of Eco1 acetylation, the activation of DNA-bound cohesin to tether sister chromatids.

A clue to how Eco1 modulates cohesin ATPase to promote DNA tethering comes from our in vitro and in vivo characterization of smc1-D1164E. Since smc1-D1164E suppresses the eco1∆ cohesion defect, we propose that smc1-D1164E mimics the Eco1 acetylation of Smc3 by blocking the cohesin ATPase at a particular stage in its hydrolysis cycle, promoting DNA tethering. In the absence of acetylation, continued cycles of ATP hydrolysis prevent cohesin from maintaining the conformation necessary for tethering (Figure 7D). The completion of the ATPase cycle may render cohesin in a conformation capable of being recognized by and acted upon Wpl1, which leads to cohesin dissociation from DNA.

This model predicts that Eco1-mediated acetylation of the Smc3 may have an impact on the ATPase activity of cohesin. However, previous studies on in vitro acetylated cohesin or acetyl-mimic mutants of cohesin suggest that Eco1 acetylation does not impact cohesin’s total ATPase activity (Ladurner et al., 2014). This could be due to (1) the inherent inefficiency of acetylation both in vitro and in vivo, resulting in only a small subset of cohesin being acetylated, (2) the inability of glutamine substitutions of K112 and K113 in Smc3 to mimic the acetylated state, (3) the failure to assay cohesin ATPase activity on DNA. Alternatively, Eco1-mediated acetylation may stabilize an intermediate state of cohesin in a way that makes subsequent cycles of ATP hydrolysis unable to alter the tethering conformation. Either way, the result is that acetylation traps cohesin in its tethering form to stabilize cohesion during or soon after establishment. Testing whether Eco1-mediated acetylation directly downregulates the cohesin ATPase cycle when bound to DNA and how the ATPase cycle affects cohesin conformation will require better reconstitution of cohesin function in vitro.

Our model of transient conformational changes that must be stabilized provides a plasticity to cohesin-mediated DNA tethering. Cohesin that is bound to a locus could form transient tethers with a number of different DNA loci and at different times. The spatial and temporal control of Eco1 would ensure that only the proper tethers are stabilized, allowing a very tight regulation for cohesion establishment in sister chromatid cohesion. This plasticity of cohesin tethering and its Eco1-dependent regulation may be critical for other aspects of cohesin function, such as condensation, DNA repair and transcriptional regulation.

S.pombe cohesin complex was over-expressed and purified from S. cerevisiae as described (Murayama and Uhlmann, 2014) with minor modifications. Briefly, over-expressing cells were harvested and lysed with a cryomill. For cohesin purifications, 50 mL of cell powder was thawed in the presence of 100 mL CLH300 buffer (50 mM HEPES, pH 7.5, 1 mM DTT, 300 mM NaCl, 20% glycerol, and protease inhibitor cocktail from Roche) and the clarified by centrifugation. The clarified supernatant was bound to 2 mL IgG beads (Invitrogen) overnight in the presence of RNase A (10 μg/mL). The beads were washed 300 mL in H300 (300 mM NaCl, 25 mM HEPES pH 7.5, 0.5 mM TCEP, 10% glycerol). To elute protein off the beads, 5 mL of elution buffer (H300 with 2 mM MgCl₂, 5 U/mL Prescission protease, 10 μg/mL RNase A) was added. Eluted protein was diluted to achieve 100 mM NaCl and then bound to heparin column. The column was washed in 4xColumn Volume (CV) H100M (25 mM HEPES pH 7.5, 0.5 mM TCEP, 10% glycerol, 100 mM NaCl and 2 mM MgCl₂), 4xCV H300 with 2 mM MgCl₂. Cohesin was eluted from the column at 600 mM NaCl. Cohesin containing fractions were pooled, concentrated and frozen in same buffer containing 200 mM NaCl. Mutants of cohesin were generated by site-directed mutagenesis and over-expression strains were generated as described (Murayama and Uhlmann, 2014). Mutant proteins were purified using the same protocol as described above. The loader complex from S.pombe was expressed and purified as described (Murayama and Uhlmann, 2014). Rad50 protein from T4 bacteriophage was purified as described (la Rosa and Nelson, 2011).

CARC1 DNA substrates were prepared as described (Onn and Koshland, 2011). Briefly, for each binding reaction, 500 ng biotin-labeled DNA was assembled on 20 μL streptavidin-conjugated dynabeads. Beads were washed 6x 30 μL in CL1 buffer (35 mM Tris pH 7.5, 1 mM TCEP, 25 mM NaCl, 25 mM KCl, 1 mM MgCl₂, 15% glycerol and 0.003% Tween 20) and resuspended in 60 μL CL1. 80 nM cohesin complex and 0.5 mM ATP (with or without 80 nM loader) was added to beads and incubated at 30°C for 1 hour in a reaction volume of 100 μL. Assembled cohesin-DNA complexes were washed in 100 μL buffer once in CL1, three times in CL1 with 500 mM KCl, and once more with CL1. Resuspending beads in 30 μL SDS sample buffer eluted bead-bound cohesin-DNA complexes. Samples were analyzed by SDS-PAGE followed by Coomassie staining or western blotting.

Stable cohesin-DNA complexes assembled as described above were resuspended in CL1 buffer in the presence of 2U DNase or 5U Mnl I at 30°C for 30 minutes. Eluted proteins (supernatant) and beads were subjected to SDS-PAGE and proteins were visualized by Coomassie staining or Western blotting.

For basal ATPase activity of cohesin, 150 nM cohesin was diluted in 100 μL ATPase reaction buffer 1 (25 mM HEPES pH 7.5, 100 mM NaCl, 10% glycerol, 1 mM TCEP, 5 mM MgCl₂). 10 μL of this sample was run on SDS-PAGE to visualize cohesin used in ATPase experiments. The ATPase reaction was started by the addition of 20 μL of 5x ATP hot mix (2.5 mM ATP, 1 μL of 10 μCi/ μL ATP-γ-P32 in 96.5 μL 5xATP buffer) to 80 μL cohesin mix. Thus, in our ATPase reactions, the final concentration of cohesin was 120 nM, and ATP was 0.5 mM. At appropriate time points, 20 μL of this reaction mix was taken out and mixed with 380 μL of 0.25 mg/mL BSA. 300 μL of STOP buffer (100 μL KPO₄ 1 M, 100 μL 1N HCl, 100 μL 20% Norit) was immediately added to the samples. Samples were then spun at 4°C 10000 rpm for 3 minutes. 500 μL of the supernatant was taken out in a new tube and spun again at 4°C 10000 rpm for 3 minutes. 350 μL of the supernatant of this second spin was counted in 5 mL of scintillation cocktail using a scintillation counter. Reads at time zero were counted as background and subtracted from later time points. To get% ATP hydrolysis, 10 μL of 1:5 diluted 5xATP buffer was counted to represent 100% ATP hydrolysis. For loader- and DNA-stimulated ATPase activity of cohesin, 80 nM cohesin, 80 nM loader, and 2.5 μg plasmid DNA containing the sequence for CARC1 (pIO2) were incubated in 50 μL ATPase reaction buffer 2 (25 mM HEPES pH 7.5, 25 mM NaCl, 25 mM KCl, 10% glycerol, 1 mM TCEP, 5 mM MgCl₂). To measure the activity of stably DNA-bound cohesin complexes (CD-B), 80 nM cohesin, 80 nM loader and 0.5 mM ATP was incubated with 100 μL dynabeads coupled to 2.5 μg DNA in 500 μL total volume of CL1 at 30°C for 1 hour. Unbound/unstably bound cohesin was washed off using the following washes: 1x500 μL CL1, 3 x 500 μL in CL1 with 500 mM KCl, then 1x500 μL CL1. CD-B were then resuspended in 50 μL ATPase buffer 2. This sample of CD-B contained about% 20 of the initial input cohesin, which was approximately the same amount of cohesin as the other samples in this experiment (Figure 1E), as could be judged by the protein gel. The reaction was started by the addition of 10 μL 5xATP hot mix to 40 μL protein/DNA mix for 2 hours at 30°C and samples were processed as described above. The ATPase activity of T4 Rad50 and mutants was measured in ATPase buffer 1 for 2 hours at 30°C as described above, except 1 μM Rad50 was used. Error bars represent standard error of data from at least two independent experiments.

Yeast strains used in this study are A364A background, and their genotypes are listed in Supplementary file 1. SC minimal and YPD media were prepared as described (Guacci et al., 1997). Benomyl (a gift from Dupont) and camptothecin (Sigma) plates used to assess drug sensitivity were prepared as previously described (Guacci and Koshland, 2012). Preparation of auxin (Sigma) containing media for depletion of AID tagged proteins was as previously described (Eng et al., 2014).

Cells were grown to saturation in YPD media at 23°C then plated in 10-fold serial dilutions. Cells were incubated on plates at relevant temperatures or containing drugs as described. For plasmid shuffle assays, cells were grown to saturation in YPD media to allow loss of covering plasmid, then plated in 10-fold serial dilutions on YPD or FOA media.

Cohesion was monitored using the LacO-LacI system where cells contained a GFP-LacI fusion and tandem LacO repreats integrated at one chromosomal locus, which recruits the GFP-LacI (Straight et al., 1996). CEN-distal cohesion was monitored by integrating LacO repeats at LYS4, located 470 kb from CEN4. CEN-proximal cohesion is monitored by integrating LacO at TRP1, located 10 kb from CEN4. Cells were fixed, and processed to allow the number of GFP signals in each cell to be scored and the percentage of cells with 2-GFP spots determined as previously described (Guacci and Koshland, 2012). Data is from 2 independent experiments and 200-300 cells scored for each data point in each experiment.

Site directed mutagenesis using the Stratagene Quick-change kit was employed to generate all mutants described. Mutations were confirmed by sequencing the entire ORF to ensure it was the only change.

Details about the Auxin mediated destruction of AID tagged proteins in yeast was previously described (Eng et al., 2014). Briefly, the TIR1 E3-ubiquiting ligase placed under control of the GPD promoter and marked by C. Albicans TRP1 replaced the TRP1 gene on chromosome IV. SMC1, SMC3 and ECO1 were internally tagged with 3V5-AID2 sequences and transformed into yeast strains bearing TIR1 to replace SMC1, SMC3 and ECO1 at the endogenous locus. PCR screening and auxin mediated sensitivity were used to identify clones containing AID tagged genes.

Genetic screen of eco1∆ wpl1∆ cells for cohesion activator suppressors was done as previously described (Guacci et al., 2015).

Chromatin Immunoprecipitation (ChIP) was performed as previously described (Eng et al., 2014; Wahba et al., 2013).

Microscopy. Images were acquired with a Zeiss Axioplan2 microscope (100X objective, NA=1.40) equipped with a Quantix CCD camera (Photometrics).

Flow cytometry analysis was performed as previously described (Eng et al., 2014).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank Hugo Tapia, Thomas Eng, Jeremy Amon, Michelle Bloom, Orna Cohen-Fix and Jill Heidinger-Pauli for critically reading the manuscript, and other members of the Koshland lab for fruitful discussions. We thank Dr. Frank Uhlmann for sharing cohesin and loader expression strains and plasmids, Dr. Scott Nelson for sharing Rad50 expression plasmids, Dr. Stuart Lynn for advice on ATPase assays, the laboratory of David Drubin and Georjana Barnes for use of equipment. This work was funded by the National Institute of Health (GM092813). GC was supported by a fellowship from the Damon Runyon Cancer Research Foundation (DRG-2137-12).

© 2015, Çamdere et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.