If SUMOylation drives the stabilization or formation of protein complexes, networks or nuclear bodies, then the question arises: how do such interactions become eventually disentangled? The SUMO pathway seems to offer three possibilities (Fig. 34). The first option is disassembly catalyzed by recruited deSUMOylation enzymes ULPs/SENPs, which typically possess multiple SIMs (Hickey et al., 2012). In humans, this mechanism is employed at sites of DSB repair and pre-ribosome assembly (Dou et al., 2010; Finkbeiner et al., 2011a; Haindl et al., 2008). In S. cerevisiae, the deSUMOylation enzyme Ulp1 localizes primarily to the inner face of the nuclear pore, but redistributes to the nucleolus under stress conditions (Sydorskyy et al., 2010), and a fraction also to the septin ring during final stages of cell division (Li and Hochstrasser, 2003; Makhnevych et al., 2007). The second deSUMOylation enzyme of yeast, Ulp2, appears to be more specific for chromatin- related functions like DNA repair and SUMO-regulated sister-chromatid cohesion at centromeres (Felberbaum and Hochstrasser, 2008; Kroetz et al., 2009; Lee et al., 2011).
A second way to disrupt SUMO-SIM-stabilized interactions is selective proteasomal degradation promoted by SUMO-targeted ubiquitin ligases. These enzymes are RING-type ubiquitin ligases, which, due to the presence of SIMs, possess affinities for SUMO conjugates (Geoffroy and Hay, 2009; Perry et al., 2008; Praefcke et al., 2012). One of the known S. cerevisiae STUbLs, Slx5/Slx8, is a heterodimer formed by two RING-finger proteins, of which Slx5 harbors two strong SIMs. Slx5/Slx8 is genetically linked to DNA repair and localizes to nuclear DNA repair foci (Cook et al., 2009; Prudden et al., 2007), but also at nuclear pores (Nagai et al., 2008). Monomeric RNF4, a vertebrate homolog of Slx5/Slx8, possesses four putative SIMs through which the RING-finger ligase is potentially targeted to multiple SUMOylated proteins including DDR factors (Galanty et al., 2012; Yin et al., 2012). Intriguingly, RNF4 localizes also to PML bodies, where it attaches a polyubiquitin chain specifically on PML proteins that are modified by poly-SUMO chains. Polyubiquitylation of PML causes its proteasomal degradation, which in turn leads to the destabilization of PML bodies (Lallemand-Breitenbach et al., 2008; Tatham et al., 2008). However, the finding that Slx5/Slx8 mediates ubiquitylation and degradation of
DISCUSSION
the yeast MAT 2 transcription factor independent of SUMOylation (Xie et al., 2010) brings to attention that targeting to SUMO conjugates is just one option for substrate selection of enzymes originally classified as STUbLs. Moreover, STUbLs may play non-proteolytic roles as well. For example, S. cerevisiae Rad18 binds via its SIM to SUMO-modified PCNA, thereby stimulating non-proteolytic mono- or K63-linked polyubiquitylation of a different subunit of PCNA (Parker and Ulrich, 2012). Likewise, in human cells upon DNA damage, RNF4 is recruited to SUMOylated DNA-damage response factors where it promotes formation of non-proteolytic K63-linked polyubiquitin chains (Yin et al., 2012).
Figure 34. Pathways Used for Disentangling SUMO-SIM Stabilized Protein Assemblies
First, deSUMOylation (left) weakens physical interactions. Second, recruitment of STUbLs (middle) mediates polyubiquitylation (K48-linked) of SUMOylated factors, typically leading to proteasomal degradation of the conjugate. Third, the chaperone-like enzyme Cdc48 (p97), in conjunction with its heterodimeric co-factor Ufd1-Npl4, recognizes SUMO conjugates and mediates their disassembly by an ATP-driven mechanism (right). Because the Cdc48 complex also associates with ubiquitin conjugates and ubiquitin ligases (not shown), it may integrate ubiquitin and SUMO signaling and may function as a STUbL as well. Notably, all three pathways employ SIM-harboring factors for SUMO conjugate recognition.
DISCUSSION
The third mechanism used to disrupt SUMO-SIM-mediated interactions involves Cdc48 (p97 in mammals), a conserved chaperone-like AAA-type ATPase. This homohexameric segregase was previously known to bind and dislodge specifically ubiquitylated proteins from their protein environment (Dantuma and Hoppe, 2012; Jentsch and Rumpf, 2007; Rape et al., 2001). For example, in endoplasmic reticulum-associated degradation (ERAD), Cdc48, assisted by substrate-recruiting ubiquitin-binding cofactors (Ufd1-Npl4), is thought to mobilize ubiquitylated ERAD substrates from the ER for cytosolic proteasomal degradation. However, it is now known that Cdc48 acts on SUMOylated substrates as well. Interestingly, both Cdc48 and Ufd1 not only bind ubiquitin via special domains, but also SUMO via SIMs (Bergink et al., 2013; Nie et al., 2012). In yeast, Cdc48 associates with SUMOylated Rad52 in complex with Rad51, which possesses a SIM but is not SUMO modified itself. By acting on this SUMO-SIM-stabilized complex, Cdc48 displaces the two proteins from chromatin. Notably, inactivation of Cdc48 leads to increased spontaneous recombination and causes aberrant Rad51 foci formation in yeast and mammalian cells (Bergink et al., 2013). Inferring from this finding, Cdc48 might be generally competent to break up SUMO-SIM-assisted protein complexes or networks. Moreover, because Cdc48 also associates with a number of ubiquitin ligases (Bohm et al., 2011; Koegl et al., 1999; Verma et al., 2011), the Cdc48 complex may not only function as a segregase but also as a multisubunit STUbL. Additionally, the presence of both ubiquitin- and SUMO-binding motifs suggests that the segregase complex may also integrate SUMO- and ubiquitin-dependent signals similar to human RAP80. This protein, via tandem SIM/UIM motifs, preferentially binds RNF4-synthesized hybrid SUMO-ubiquitin chains, thereby promoting BRCA1 recruitment to DNA damage sites (Guzzo et al., 2012).
Taken together, protein group SUMOylation not only targets protein pools already engaged in their respective functions, possibly explaining why the SUMOylated fraction of a given substrate is typically very small, but it also simultaneously marks these pools for subsequent action by other SIM-containing factors that ensure the dynamic nature of SUMO modification and offer extensive possibilities for the regulation of the SUMO-SIM-assisted protein assemblies. It is attractive to speculate, that similar to ULPs or STUbLs, other enzymatic activities (e.g., phoshphorylation, methylation, acetylation) could be recruited by SIM-harboring proteins to the SUMOylated substrates, further expanding the versatility of the SUMO pathway.