Ever since the discovery that the Eco1 acetyltransferase is essential for creating sister chromatid cohesion but not for loading cohesin on to chromosomes (Tóth et al., 1999
), it has been clear that this enzyme (Ivanov et al., 2002
) is central to understanding how cohesin holds sister DNAs together. The enzyme exerts its effect by acetylating Smc3’s NBD at K112 and K113. Remarkably, this normally essential modification can be mimicked and the lethality of eco1
mutants suppressed either by null alleles of Wapl or by highly specific missense mutations within Smc3, Pds5, and Scc3. This has led to the notion that Eco1 counteracts an activity intrinsic to cohesin that hinders its ability to build stable cohesion. A critical issue is whether this “antiestablishment” activity prevents creation of cohesive structures in the first place or merely destroys them after their creation during S phase (Rowland et al., 2009
). Our finding that Wapl is able to destroy cohesion long after replication is complete in cells lacking Eco1 is consistent with the latter hypothesis.
How then does antiestablishment destroy cohesion? The answer is suggested by our finding that a large fraction of unacetylated cohesin complexes associated with pericentric chromatin turns over and that this process is greatly reduced by mutations that bypass the need for Eco1, implying that antiestablishment is synonymous with a feature of cohesin that enables it to disengage from chromatin in the absence of α-kleisin cleavage (Gerlich et al., 2006
). In other words, it is cohesin’s “releasing activity” that destroys cohesion built in the absence of Smc3 acetylation. In mammalian cells, Wapl depletion not only reduces cohesin’s turnover on chromosomes but also increases the fraction of cohesin associated with chromatin, especially as cells enter mitosis (Kueng et al., 2006
). Surprisingly, less not more cohesin is found associated with yeast chromosomes in wpl1Δ
cells. An explanation for this phenomenon is that Scc1 protein levels are reduced about two-fold in wpl1Δ
cells, possibly due to lower rates of synthesis in late G1. The cause of this phenomenon is currently under investigation.
If releasing activity is an inherent aspect of cohesin complexes, then it is equally vital that cells possess a mechanism to neutralize it in a subset of cohesin complexes that entrap sister DNAs during replication. We suggest that this is the function of acetylation of Smc3 NBDs by Eco1, a model similar to that proposed for animal cells where acetylation has been proposed to block Wapl activity by recruiting sororin (Lafont et al., 2010
; Nishiyama et al., 2010
). According to this notion, Smc3 acetylation is required to keep releasing activity continuously in check and is therefore responsible for maintaining cohesion after S phase. Our finding that partial deacetylation of Smc3 due to Hos1 overexpression is accompanied by reduced sister chromatid cohesion (Beckouët et al., 2010
) is consistent with this. We currently have no explanation for the finding that about 20% of pericentric cohesin fails to turn over even in the absence of Eco1. If inactivating cohesin’s releasing activity is Eco1’s sole function, then why does this stable pool of cohesin complexes not support viable chromosome segregation?
If as envisaged by the ring model, cohesin’s stable or even semistable chromosomal association is mediated by entrapment of chromatin fibers, then release must involve their escape. The ring must have a DNA exit gate. Our finding that cohesin containing an Smc3-kleisin fusion protein fails to turn over once it has associated with chromatin and moreover creates cohesion without Eco1 suggests that the exit gate is situated at the Smc3/kleisin interface. A corollary is that the primary function of Smc3’s NBD acetylation by Eco1 is to block dissociation of cohesin’s Smc3/kleisin interface, which ensures that DNAs remain entrapped by cohesin’s tripartite ring. This conclusion implies that DNA exit catalyzed by cohesin’s releasing activity is not simply a reversal of entry (Nishiyama et al., 2010
). The ring must have separate DNA entry and exit gates.
Because it is the only component of cohesin’s releasing activity that does not also have roles in establishing or maintaining cohesion, Wapl may be rather directly involved disengaging α-kleisin’s N-terminal domain from Smc3 NBDs. How might it perform this task? One possibility is that, with help from Pds5 and Scc3, Wapl binds to the Smc3 NBD in a manner that precludes α-kleisin binding. Thus, Wapl and α-kleisin might compete for binding to the same site on Smc3. Elucidating how Smc3’s NBD binds to α-kleisin will be vital to understanding this process. Crucially, we suggest that K112 and K113 within Smc3’s NBD have a key destabilizing influence on its interaction with α-kleisin’s N-terminal domain and that this effect (whether direct or indirect) is neutralized by acetylation. Unlike α-kleisin, which will be bound to the complex via its C-terminal domain throughout the disengagement cycle, Wapl is never stably associated with cohesin and its displacement of α-kleisin from Smc3 can therefore only be a temporary event.
It is important to stress that releasing activity is not intrinsic to the Smc3/kleisin interface or even to the influence of Wapl. It also needs residues within Pds5 and Scc3 that are not required for Smc3’s association with α-kleisin or for recruiting Wapl. How Scc3 and Pds5 regulate dissociation allosterically and whether the process also involves binding and/or hydrolysis of ATP to Smc NBDs are crucial issues for future research. Another is whether the rapid turnover of Wapl within cohesin, which is far more rapid than that of its partner Pds5, is an intrinsic aspect of cohesin’s releasing activity.
According to our version of the ring model (B), acetylation should be viewed as a key that locks cohesin rings shut once DNAs have been trapped inside. This process is coupled to DNA replication (Rolef Ben-Shahar et al., 2008
) and in yeast is not reversed until cohesin rings are cleaved by separase (Beckouët et al., 2010
). The locked acetylated state may need to be extraordinarily robust, especially in cells such as oocytes where the cohesion holding bivalent chromosomes together may need to last for several weeks if not decades (Tachibana-Konwalski et al., 2010
). In mammalian cells, acetylation is insufficient to neutralize cohesin’s releasing activity. The modification promotes recruitment of sororin, which is thought to displace Wapl from its binding site on Pds5 (Lafont et al., 2010
; Nishiyama et al., 2010
). If sororin does not exist in yeast, which is not known for sure, then acetylation must alter some other aspect of cohesin. Our finding that turnover of a fraction of Pds5 molecules within pericentric chromatin is greatly reduced by Eco1 activity suggests that Smc3 acetylation alters the way Pds5 interacts with cohesin, a change that might have a role in neutralizing releasing activity.
A Model: Acetylation of Smc3 NBDs by Eco1 Prevents Transient Dissociation of the Smc3/Kleisin Interface and Thereby Blocks Escape of DNAs
There are good reasons to believe that the mechanism by which cohesin dissociates from yeast chromosomes via opening the Smc3/α-kleisin interface will apply to the “prophase pathway” that removes most cohesin from chromosome arms as animal cells enter mitosis. Because Wapl in mammals is required for cohesin’s turnover on chromatin during interphase as well as during prophase (Kueng et al., 2006
), it is likely that cohesin’s depletion from chromosome arms during prophase is triggered by hyperactivation of the same releasing activity that merely induces turnover during interphase. Interestingly, the prophase pathway is also dependent on phosphorylation of SA (Scc3) proteins (Hauf et al., 2005
), emphasizing that this subunit is intimately involved in releasing activity in animal cells as well as yeast.
The finding that yeast cells lacking releasing activity are viable raises the question as to why it is such a conserved feature of eukaryotic cohesin complexes. If it did not exist, there would be less or possibly no necessity for Eco1. Indeed, some eukaryotic organisms appear to lack both proteins (Nasmyth and Schleiffer, 2004
). It may be relevant in this regard that cohesin has functions besides mediating sister chromatid cohesion (Kagey et al., 2010
; Parelho et al., 2008
; Wendt et al., 2008
) and is important in nonproliferating as well as proliferating cells (Pauli et al., 2008
; Seitan et al., 2011
). It is thought that cohesin has important roles in regulating transcription, presumably through modulating the topology of interphase chromatin. Due to the dynamic nature of the transcription process, it is inconceivable that such functions could be mediated by cohesin complexes lacking a capacity for turning over. We suggest that a dynamic entrapment of chromatins is important for achieving the appropriate distribution of cohesin complexes along the genome; remodeling intrachromatid loops; removing what could be topological barriers under certain conditions during transcription, repair, or replication; and dissolving inappropriate connections between nonsister chromatids. This is in addition to the value of protecting a large fraction of the cohesin pool from separase as a consequence of its prior dissociation from chromosomes during prophase (Kucej and Zou, 2010
). Given that releasing activity destroys a state catalyzed by kollerin, it is possible that defects caused by haploinsufficiency of kollerin’s Scc2/Nipbl subunit (Pehlivan et al., 2012
) that are characteristic of Cornelia de Lange syndrome might be caused at least in part by releasing activity. If so, partial inhibition could conceivably alleviate any nondevelopmental symptoms.
Our finding that cohesin’s dissociation from chromatin involves opening the Smc3/kleisin interface is an important endorsement of the ring model. It also provides a theoretical framework for exploring how release is regulated by acetylation during S phase and by phosphorylation during mitosis. It may also have important implications for other eukaryotic Smc/kleisin complexes. Because the N-terminal domains of kleisins are highly conserved, it is conceivable that their regulated association with and dissociation from their cognate Smc NBDs will prove to be a conserved feature of these chromosomal machines. In which case, the topological principles according to which cohesin functions may apply also to condensin and Smc5/6 complexes.