SUMO: A MULTIFACETED MODIFIER OF CHROMATIN STRUCTURE AND FUNCTION

Summary

A major challenge in nuclear organization is the packaging of DNA into dynamic chromatin structures that can respond to changes in the transcriptional requirements of the cell. Posttranslational protein modifications, of histones and other chromatin-associated factors, are essential regulators of chromatin dynamics. In this review, we summarize studies demonstrating that posttranslational modification of proteins by small ubiquitin-related modifiers (SUMOs) regulates chromatin structure and function at multiple levels and through a variety of mechanisms to influence gene expression and maintain genome integrity.

Introduction

The discovery of the nucleosome, the iconic “beads on a string,” and finally the realization that there are higher order chromatin packing structures have made it clear that DNA is intricately organized. Since this time, significant progress has been made in identifying the proteins responsible for higher order DNA packaging and in understanding how regulation of these proteins affects chromatin structure. A major theme that has emerged is the important role of posttranslational protein modifications in modulating the functional accessibility of DNA. Of particular interest, recent global proteomic and genetic studies have linked modification by the small ubiquitin-related modifier (SUMO) to many processes involving chromatin, including transcriptional activation and repression, DNA replication and repair, as well as chromosome segregation (Golebiowski et al., 2009; Makhnevych et al., 2009). Here, we review the current knowledge of how SUMO modification (sumoylation) of chromatin-associated proteins regulates chromatin structure and function and thereby controls these essential cellular processes.

After introducing the sumoylation pathway and general connections between SUMO and chromatin, we will discuss the complex role of sumoylation in both euchromatin and heterochromatin environments. First, the multiple mechanisms by which sumoylation modulates gene expression through effects on DNA methylation, histones and transcriptional regulators will be reviewed. Subsequently, the functional role of sumoylation in repetitive DNA structures, including rDNA, telomeres, and centromeres will be discussed. We will highlight the unique functions of sumoylation within each of these domains as well as its common role as a protector of genomic integrity.

Several emerging themes will be reiterated throughout the review. First, that sumoylation often functions as a signal to facilitate protein-protein interactions on chromatin. These interactions may be simple hetero-dimeric associations, but can also involve assembly of very large multi-protein complexes. Second, that sumoylation also specifies multiple other fates, including effects on enzyme activity and changes in protein sub-cellular localization. And lastly, that although in many cases sumoylation is linked to heterochromatin and gene inactivation, a growing number of studies indicate that sumoylation also plays important roles in enhancing chromatin accessibility and gene activation. Thus, the effects of sumoylation are dichotomous and often context dependent.

SUMO Modification and Function

Mechanistically, sumoylation occurs through an enzyme cascade very similar to ubiquitylation (Figure 1A). The SUMO paralogs are synthesized as precursor proteins that are cleaved by a family of SUMO isopeptidases referred to as SENPs (Mukhopadhyay and Dasso, 2007). Mature SUMO is subsequently activated by a heterodimeric E1 activating enzyme (Aos1/Uba2), transferred to an E2 conjugating enzyme (Ubc9), and finally transferred to lysine residues in target proteins. This last step may be facilitated by the action of E3 ligases, which in addition to enhancing rates of sumoylation, are also believed to contribute to specificity (Gareau and Lima, 2010; Johnson, 2004). Substrate specificity in the sumoylation pathway, however, still remains poorly understood as only a single E2 enzyme and relatively few E3 ligases have been identified. Sumoylation is, however, highly dynamic and can be reversed by the action of desumoylating enzymes. In vertebrates, these isopeptidases include a family of six SENPs defined by a conserved cysteine protease domain, distinct-subcellular localizations and non-redundant functions (Mukhopadhyay and Dasso, 2007). In addition, several unique desumoylating enzymes have more recently been identified, including the metalloprotease Wss1, the PPPDE–domain containing proteins DeSI-1 and DeSI-2, and the ubiquitin-specific protease-like protein 1 (USPL1) (Mullen et al., 2010; Schulz et al., 2012; Shin et al., 2012). Sumoylation of individual proteins is likely to be regulated by a fine-tuned balance between conjugation and deconjugation (Gareau and Lima, 2010). Consistent with this, and as outlined below, both SUMO conjugating and deconjugating enzymes are important effectors of chromatin structure.

Figure 1. The SUMO pathway and molecular consequences of sumoylation.

(A) SUMO is synthesized as a precursor, processed to a mature form by SUMO-specific isopeptidases and covalently conjugated to protein substrates via an E1, E2 and E3 enzyme cascade. Sumoylated protein substrates are demodified by SUMO-specific isopeptidases. (B) The molecular consequences of sumoylation (S) include protein targeting, alteration of protein or enzyme function, effects on protein stability, and effects on protein-protein interactions. Sumoylation can promote or antagonize protein stability by either blocking ubiquitylation of lysine residues or by promoting ubiquitylation (Ub) upon recognition by SUMO-targeted ubiquitin ligases (STUbL). Effects on protein-protein interactions may be modulated at multiple levels, including polymeric chain formation and intersection with other posttranslational modifications such as phosphorylation (P).

Sumoylation of proteins can affect protein stability, enzymatic activity, alter localization, or mediate novel protein-protein interactions with other proteins containing SUMO-interacting motifs (SIMs) (Figure 1B) (Geiss-Friedlander and Melchior, 2007; Kerscher, 2007). In many instances, sumoylation may play a role in facilitating the assembly of large multi-protein complexes between proteins that are either covalently modified by SUMO and/or contain SIMs, as exemplified by PML nuclear bodies. In these sub-nuclear structures, SUMO acts as a scaffold to mediate interactions between the PML protein and other associated factors (Matunis et al., 2006; Shen et al., 2006). Although multiple effects of sumoylation on proteins have been discovered, the ability of SUMO to promote the assembly of multi-protein complexes is an especially prominent theme.

The diverse effects of sumoylation may be explained in part through the generation of functionally distinct signals. Although invertebrates express only a single SUMO, vertebrates express four paralogs (SUMO-1, SUMO-2, SUMO-3 and SUMO-4), each with the potential to act as unique signals by interacting with distinct downstream factors (Kerscher, 2007). SUMO-2 and SUMO-3 share ~97% identity with each other and likely represent redundant signals and are thus referred to as SUMO-2/3. However, they share only ~50% identity with SUMO-1 (Gareau and Lima, 2010). SUMO-4 shares ~86% identity with SUMO-2/3, but questions exist about its ability to be conjugated to other proteins (Owerbach et al., 2005; Wei et al., 2008). The ability of SUMOs to form polymeric chains provides an additional opportunity for signal diversification (Figure 1B) (Kerscher, 2007). Currently the best-studied functional role for polymeric SUMO chains involves their recognition by SUMO-targeted ubiquitin E3 ligases containing tandem SIMs (Perry et al., 2008). Other functional distinctions between paralogs and polymers remain to be fully understood. Finally, the diverse effects of sumoylation can also be explained through intersections with other posttranslational modification pathways (Figure 1B). For example, both phosphorylation and acetylation affect interactions between SUMO and downstream SUMO-binding proteins (Chang et al., 2011; Ullmann et al., 2012).

General connections between sumoylation, chromatin and transcription

Associations between sumoylation and chromatin structure have been well documented through numerous immunofluorescence microscopy studies. All three SUMO paralogs, for example, are detected in the heterochromatin XY bodies of rat pachytene spermatocytes (La Salle et al., 2008; Rogers et al., 2004; Vigodner, 2009; Vigodner et al., 2006), and SUMO-1 is associated with long stretches of constitutive heterochromatin in human spermatocytes (Metzler-Guillemain et al., 2008). In mitotic cells, SUMO-2/3 has been observed at the inner centromere of chromosomes and also along the length of chromosome arms as cells progress from metaphase through telophase (Ayaydin and Dasso, 2004; Azuma et al., 2005; Zhang et al., 2008). Associations between SUMO and mitotic chromosomes are also detected in S. cerevisiae (Biggins et al., 2001) and in D. melanogaster using polytene chromosome spreads (Lehembre et al., 2000), suggesting that sumoylation of chromatin-associated proteins has a conserved and fundamentally important function.

Associations between SUMO and chromatin are further supported by biochemical studies, including chromatin immunoprecipitation experiments (ChIP). In S. pombe, for example, ChIP experiments revealed that the SUMO E2 conjugating enzyme Ubc9 is chromatin bound and specifically enriched in regions of heterochromatin (Shin et al., 2005). Similarly, fractionation of X. laevis egg extracts demonstrated interactions between PIAS E3 ligases and chromatin (Azuma et al., 2005). Surprisingly, a comprehensive genome-wide ChIP analysis to detect the precise association of SUMO or SUMO-modified proteins with chromatin has not yet been reported. However, more targeted studies link SUMO or SUMO pathway enzymes to distinct chromatin domains, including pericentric heterochromatin, PcG bodies, the nucleolus, telomeres, and centromeres, as reviewed in detail below.

Studies related to the involvement of sumoylation in controlling transcription regulation provide the strongest evidence for functional links between SUMO and chromatin. Genetic approaches have revealed a general causal relationship between sumoylation and gene repression. Inducing hyper-sumoylation by targeting SUMO and/or Ubc9 to specific gene promoters primarily induces gene repression (Chupreta et al., 2005; Shiio and Eisenman, 2003). Consistently, inducing hypo-sumoylation by overexpressing SUMO isopeptidases or by depleting cells of Ubc9 or SUMO enhances ectopic gene expression (Ouyang et al., 2009; Poulin et al., 2005; Spektor et al., 2011). These effects are mediated at multiple levels, including direct effects on transcription factor activities (Gill, 2005). Transcription factors and co-regulators make up one of the most abundant classes of SUMO-modified proteins. Although clearly able to mediate transcriptional repression, sumoylation is not simply a negative regulator of transcription. The dichotomous role of SUMO in gene regulation is demonstrated by the observations that sumoylation of certain transcription factors, including Ikaros, enhance their transcriptional activity (Figure 2E) (Gómez-del Arco et al., 2005).

Figure 2. Sumoylation functions as an activator and a repressor of gene expression.

(A) Sumoylation represses gene expression by promoting DNA methylation (yellow dots) through DNMT1 activation. (B) Sumoylation represses gene expression by facilitating assembly of repressive complexes on methylated DNA and at promoters. Sumoylation also inhibits the activities of transcription factors (TFs) and affects HDAC recruitment and function. (C) Sumoylation promotes the assembly of repressive PcG bodies. (D) Sumoylation promotes DNA demethylation and gene activation through mechanisms involving the SUMO-targeted ubiquitin E3 ligase activity of RNF4. (E) Sumoylation facilitates the assembly of complexes on chromatin that promote transcription. (F) Sumoylation positively influences RNA polymerase II (RNA Pol II) recruitment to constitutively active gene promoters.

Studies in yeast also provide a striking example of the complexities of sumoylation as both an activator and repressor of transcription. ChIP analysis in S. cerevisiae reveals the presence of sumoylated proteins at the promoters of constitutively active genes and the recruitment of Ubc9 and SUMO to promoters of inducible genes in response to activation (Rosonina et al., 2010). Surprisingly, sumoylation is not only required for optimal transcriptional activation of constitutive genes but also for repression and timely inactivation of inducible genes. At constitutively active genes, sumoylation enhances transcription by promoting RNA polymerase II recruitment (Figure 2F). However, at inducible promoters sumoylation functions downstream of transcription initiation. Specifically, sumoylation of the transcription factor Gcn4 promotes its removal from promoters and its degradation, thereby limiting transcription reinitiation (Rosonina et al., 2010, 2012).

Another elegant example of the subtle and complex effects of sumoylation on transcription is illustrated by phenotypes in mice expressing a mutant form of the SF-1 transcription factor that cannot be sumoylated. Although studies in cultured cells indicate that sumoylation negatively regulates SF-1 transcriptional activity, mice expressing non-sumoylatable SF-1 fail to phenocopy a constitutively active SF-1 (Lee et al., 2011; Lee et al., 2005). Thus, sumoylation does not function as a simple on-off switch, but rather enhances the functional diversity of SF-1, adding a layer of regulation for fine-tuning gene expression during development.

The utility of sumoylation as a mechanism to fine tune transcription can be explained in part on its broad effects on chromatin modifications and structure. This is illustrated by studies of a well-characterized SUMO-1 modified transcription factor, Sp3. In cells expressing specific mutant isoforms of Sp3 that cannot be sumoylated, transcription activation and chromatin modifications at Sp3-targeted promoters are dramatically different from those observed at the same promoters in cells expressing wild type Sp3. For instance, levels of both DNA and histone methylation are reduced at promoters in cells expressing mutant Sp3 and, concomitantly, levels of histone methyltransferases, heterochromatin protein 1 (HP1), and two ATP-dependent chromatin remodelers are also reduced (Ross et al., 2002; Stielow et al., 2010; Stielow et al., 2008a; Stielow et al., 2008b). These findings provide a relatively simple view of how sumoylation of just one transcription factor exerts multiple effects, some direct and others indirect, to alter chromatin structure. However, even in the case of Sp3, the situation is not so simple, as evidenced through additional studies demonstrating that the effects of sumoylation are unique for different Sp3 isoforms and for different gene promoters (Ellis et al., 2006; Sapetschnig et al., 2004). An emerging view is that sumoylation sits at the intersection of multiple pathways, affecting the activities of not only transcription factors, but also other chromatin-associated proteins and chromatin modifying enzymes. Thus, effects of sumoylation on gene expression and chromatin structure represent collective effects on multiple, context-dependent, levels (Figure 2). How sumoylation affects gene expression at the level of chromatin structure and accessibility, and within the context of distinct genomic subdomains, is the focus of the following sections.

DNA Methylation

DNA methylation of CpG dinucleotides restricts DNA accessibility by two mechanisms. Methylation either blocks the binding of sequence-specific DNA binding proteins and/or recruits chromatin modifying complexes that promote a restrictive chromatin structure (Bird, 2002). Multiple lines of evidence indicate that sumoylation plays important roles in regulating CpG methylation and demethylation, as well as the assembly and functions of downstream complexes recruited to methylated DNA.

First, sumoylation of DNA methyltransferases (Dnmts) may alter their enzymatic activity (Figure 2A). This has been demonstrated most clearly for the maintenance methyltransferase, Dnmt1, whose SUMO-1 modification increases its activity toward S phase hemi-methylated DNA substrates in vitro (Lee and Muller, 2009). In addition, the de novo DNA methyltransferases Dnmt3a and Dnmt3b are both sumoylated in vivo, although the functional consequences of their modifications remain to be fully elucidated (Kang et al., 2001; Li et al., 2007; Ling et al., 2004). Strikingly, nearly all of Dnmt3a is sumoylated in cells overexpressing SUMO-1, an effect that correlates with a disruption of Dnmt3a interactions with histone deacetylases 1 and 2 (HDAC1 and -2) and a loss of Dnmt3a-mediated repression (Ling et al., 2004). Further studies are needed to determine whether these effects are strictly related to Dnmt3a sumoylation.

In addition to regulating DNA methylation, sumoylation also promotes DNA demethylation through mechanisms mediated by RNF4, a ubiquitin E3 ligase that specifically recognizes and ubiquitylates sumoylated proteins (Figure 2D) (Hu et al., 2010; Perry et al., 2008). RNF4 deficiency is embryonic lethal in mice. RNF4^-/- mouse embryonic fibroblasts, however, are viable but exhibit hypermethylation of genomic DNA. Conversely, overexpression of wild type RNF4, but not SUMO-binding or ubiquitin ligase mutants, results in global DNA demethylation (Hu et al., 2010). Thus, SUMO and ubiquitylation are required for RNF4-mediated DNA demethylation, although the precise mechanisms of action remain unclear. Intriguingly, one favored model for DNA demethylation is based on deamination of methylcytosines to create T:G mismatches that are repaired by thymidine DNA glycosylase (TDG) and base-excision repair (BER) (Wu and Zhang, 2010). TDG is known to interact with RNF4 (Hu et al., 2010) and sumoylation has been proposed to play an important role in regulating TDG by enhancing its enzymatic turnover (Baba et al., 2005; Hardeland et al., 2002). Thus, ubiquitylation of sumoylated TDG or other interacting proteins could produce a signal required for DNA demethylation and possibly BER in general.

Sumoylation also functions downstream of DNA methylation, affecting the assembly of methyl-CpG binding domain (MBD) proteins and other factors with methylated DNA (Figure 2B) (Bogdanovic and Veenstra, 2009). SUMO-1 and SUMO-2/3 both localize to heterochromatin domains enriched in MBD1, as well as heterochromatin proteins HP1 and MCAF1 (Uchimura et al., 2006). Formation of these heterochromatin domains is SUMO-dependent, as knockdown of either SUMO-1 or SUMO-2/3 disrupts the co-localization of HP1 and MCAF1 with MBD1-containing foci (Uchimura et al., 2006). Intriguingly, MBD1 and HP1 are both sumoylated, whereas MCAF1 binds all three SUMO paralogs (Lyst et al., 2006; Maison et al., 2011; Uchimura et al., 2006). Thus, it is tempting to speculate that MBD1-containing heterochromatin domains are organized around covalent and non-covalent SUMO interactions in a fashion similar to PML nuclear bodies (Matunis et al., 2006). In contrast with these repressive functions, however, sumoylation of MBD1 also interferes with its interactions with the histone methyltransferase SETDB1 and might thereby limit gene inactivation (Lyst et al., 2006). Therefore, sumoylation underlies multiple mechanisms for fine-tuning the functional properties of methylated DNA through effects both positive and negative, again demonstrating the dichotomous effects of SUMO on gene expression.

Histones and HDACs

Posttranslational modification of histones also represents a central mechanism for controlling chromatin structure and gene expression, and not surprisingly, histones are sumoylated. All four histones as well as the H2A.Z variant are sumoylated in S. cerevisiae (Kalocsay et al., 2009; Nathan et al., 2006), whereas only H4 has been shown to be modified in mammalian cells (Shiio and Eisenman, 2003). The functional significance of histone sumoylation is surprisingly not well understood. ChIP experiments involving exogenously expressed SUMO-histone fusion proteins in yeast reveals enrichment at subtelomeric regions, an area of the genome where SUMO is generally thought to antagonize transcriptional repression (Nathan et al., 2006; Xhemalce et al., 2004; Zhao and Blobel, 2005). In contrast, expression of SUMO-histone fusion proteins represses transcriptional reporters in both mammalian cells and in yeast at least in part through recruitment of HDACs and HP1 (Nathan et al., 2006; Shiio and Eisenman, 2003). Such findings suggest that histone sumoylation functions as a signal to recruit proteins to chromatin (Figure 2B). Consistent with this general concept, recruitment of the transcription corepressor complex, LSD1/CoREST1/HDAC, to chromatin is dependent on a SIM in CoREST1 (Ouyang et al., 2009). Whether CoREST1 recognizes sumoylated histones and/or other sumoylated factors, however, remains to be determined.

In addition to histones, multiple studies have identified HDACs as another important effector of SUMO-mediated transcriptional repression. Most simply, HDACs themselves are sumoylated (Figure 2B). Sumoylation of HDAC1 and HDAC4 is required for the full transcriptional repression activities at defined promoters (Cheng et al., 2004; David et al., 2002; Kirsh et al., 2002). Whether sumoylation directly affects HDAC activity or acts as a signal for the recruitment of other chromatin repressors, however, is a question that remains to be fully addressed. In addition to being directly modified, HDACs are also recruited to gene promoters in response to sumoylation of other factors, including transcription factors and cofactors such as Elk-1 and p300 (Garcia-Dominguez and Reyes, 2009; Girdwood et al., 2003; Yang and Sharrocks, 2004). These findings suggest that HDAC recruitment may be mediated through non-covalent interactions with SUMO, a suggestion that has been confirmed for at least HDAC1 which contains a functionally important SIM (Ahn et al., 2009). A third level of association between HDACs and sumoylation has been made based on the observations that HDACs 4, 5, and 7 appear to function as SUMO E3 ligases for certain substrates (Gao et al., 2008; Yang et al., 2011; Zhao et al., 2005). These findings are based largely on effects of HDAC overexpression, where an alternative mechanism for enhanced sumoylation might involve substrate binding and protection from isopeptidases. In either case, HDAC interaction would provide a feed-forward mechanism for enhancing sumoylation-mediated histone deacetylation and repression.

PcG Bodies

Polycomb group (PcG) bodies are subnuclear structures that function as small hubs of transcriptional repression. To facilitate repression, PcG bodies cluster distant DNA promoter elements and recruit chromatin remodeling complexes called polycomb repressive complexes (Bantignies and Cavalli, 2011). Given its involvement in repression and organizing large protein complexes, it is not surprising that SUMO localizes to PcG bodies (Figure 2C) (Kagey et al., 2003). In addition to SUMO, Ubc9, the SUMO isopeptidase SENP2, and the SUMO E3 ligase Cbx4/Pc2 all localize to PcG bodies (Kagey et al., 2005; Kagey et al., 2003; Kang et al., 2010). Because Pc2 stimulates the sumoylation of many repressive proteins, including Dnmt3a, CTCF, and components of the polycomb repressive complex 2, it is attractive to speculate that sumoylation regulates the dynamic recruitment and assembly of these proteins within PcG bodies in a fashion similar to PML nuclear bodies (Li et al., 2007; MacPherson et al., 2009; Matunis et al., 2006; Riising et al., 2008).

Consistent with essential functions in PcG body-mediated repression, two independent studies have demonstrated links between sumoylation and expression of PcG-body regulated genes. In C. elegans, depletion of SUMO, E1 or E2 conjugating enzymes results in ectopic expression of Hox genes normally controlled by PcG body recruitment (Zhang et al., 2004). The appropriate repression of Hox genes is dependent at least in part on sumoylation of the PcG protein SOP-2, which is required for the association of SOP-2 with PcG bodies (Zhang et al., 2004). SUMO-dependent assembly of PcG bodies is also conserved in mammalian cells and is also critical for normal gene expression during embryonic development. In particular, assembly of the polycomb repressive complex 1 (PRC1) at the promoters of genes important for normal heart development is misregulated in mice deficient in the SENP2 isopeptidase (Kang et al., 2010). This misregulation is due in part to hypersumoylation of the Cbx4/Pc2 SUMO E3 ligase and enhanced assembly of PRC1 complexes on the promoters of PcG target genes. These findings illustrate the important balance between SUMO conjugating enzymes and isopeptidases, which is a common theme in ubiquitylation (Sowa et al., 2009). Further studies are required to understand how the activities of Cbx/Pc2 and SENP2 are normally regulated to affect proper PcG body function.

Finally, in another example of the dichotomous effects of sumoylation, assembly of the PcG protein Sex Comb on Midleg (Scm) into repressor complexes in D. melanogaster appears to be negatively regulated by its SUMO modification (Smith et al., 2011). Whether sumoylation has universally opposing effects on PcG body formation in D. melanogaster compared to other organisms remains to be determined. An alternative and more appealing scenario is that sumoylation both positively and negatively affects PcG body assembly, with the ultimate effects on individual protein recruitment and gene expression being influenced by multiple, context-dependent factors and interactions.

Chromatin insulators

Sumoylation also influences gene expression by affecting the activities of chromatin insulator complexes. This function was first revealed by studies in D. melanogaster, demonstrating that loss of the PIAS E3 ligase homolog results in ablation of heterochromatin-euchromatin barriers and normal polytene chromosome banding patterns (Hari et al., 2001). Consistent with a role in regulating insulator functions, SUMO was subsequently localized to insulator bodies in D. melanogaster and two of the major protein components, Mod(mdg4)2.2/67.2 and CP190, were found to be sumoylated (Capelson and Corces, 2006; Golovnin et al., 2012). However, the function of SUMO in organizing and regulating the function of insulators is still unclear. Enhancing sumoylation by Ubc9 overexpression leads to dispersal of insulator bodies, suggesting that sumoylation may negatively affect local and/or long-range interactions between insulator complexes (Capelson and Corces, 2006). In contrast, SUMO depletion or expression of a Mod(Mdg4)2.2/67.2 mutant that cannot be sumoylated inhibits insulator body formation, arguing for a positive role in insulator assembly (Golovnin et al., 2012). Such opposing findings indicate that insulator assembly and/or maintenance may rely on a finely tuned balance of sumoylation and desumoylation, as required for the association or HP1α with pericentric DNA (Maison et al., 2012). Further analysis is needed to understand the function of SUMO in insulator activity in Drosophila and particularly in other species. CTCF, a well characterize vertebrate insulator protein, is sumoylated in human cells, but how sumoylation affects its insulating activities remains unknown (MacPherson et al., 2009).

The nucleolus

The nucleolus is a specialized sub-nuclear domain for ribosomal RNA (rRNA) gene expression and pre-ribosomal particle assembly (Boisvert et al., 2007). Studies in both vertebrates and yeast indicate that sumoylation plays important roles in the nucleolus, including regulation of rRNA processing and pre-ribosomal particle assembly (Castle et al., 2012; Finkbeiner et al., 2011; Panse et al., 2006; Yun et al., 2008). Consistent with this, SUMO-1 and SUMO-2/3 are detected within the nucleolus in vertebrate cells (Ayaydin and Dasso, 2004; Takahashi et al., 2008), as are the isopeptidases SENP3 and SENP5 (Gong and Yeh, 2006). Sumoylation also appears to have important effects on nucleolar rDNA structure and function. Thus, hyposumoylation due to defects in the SUMO E3 ligase Mms21, a component of the Smc5/6 complex, lead to abnormal nucleolar morphology in S. cerevisiae (Zhao and Blobel, 2005). In addition, aberrant activation of silenced rDNA occurs in S. cerevisiae strains deficient in the Slx5/Slx8 SUMO-targeted ubiquitin E3 ligase (Darst et al., 2008). How sumoylation affects rDNA chromatin structure and silencing remains to be fully characterized. However, recent proteomic identification of the SUMO substrates within the nucleolus should enhance these efforts (Matafora et al., 2009; Westman et al., 2010).

Because of the repetitive nature of rDNA genes, specialized SUMO-dependent DNA repair mechanisms exist to maintain stability of nucleolar rDNA loci (Figure 3E). DNA double-strand breaks within rDNA loci are repaired at extranucleolar sites in a manner dependent on the Smc5/6 complex and sumoylation of Rad52. Specifically, yeast strains expressing Smc5/6 mutants or a Rad52 mutant that cannot be sumoylated form DNA repair foci within the nucleolus itself and these strains exhibit hyper-recombination within the rDNA locus (Altmannova et al., 2010; Torres-Rosell et al., 2007). Whether Mms21-mediated sumoylation regulates DNA repair in other repetitive sequences by a similar mechanism remains to be determined, but the role of SUMO in general repetitive DNA maintenance is reviewed in detail below.

Figure 3. Sumoylation maintains genome integrity at repetitive DNA domains.

(A) Sumoylation limits telomere elongation by regulating interactions between Cdc13 and Stn1. (B) Sumoylation of unknown proteins affects sub-telomeric chromatin structure. (C) Sumoylation promotes the assembly of ALT PML nuclear bodies (APBs) essential for telomere maintenance in telomerase-deficient cells. (D) Sumoylation of Rad52 promotes the movement of DNA double-strand breaks from intranucleolar domains to the nucleolar periphery for optimal repair. (E) Sumoylation of HP1α regulates its association with α-satellite RNA and recruitment to centromeres. (F) Sumoylation functions to resolve DNA replication and repair intermediates within repetitive DNA domains and to promote decatenation.

Telomeres

Telomeres are composed of DNA repeats coated by resident telomere-binding protein complexes, called shelterin complexes, that protect them from degradation and inappropriate recognition by DNA repair enzymes (Blasco, 2007). Because telomere shortening jeopardizes genome integrity, maintenance of telomere length is a tightly controlled process. Telomere maintenance is controlled through multiple mechanisms, including recruitment of telomerase (the enzyme that catalyzes the addition of telomere DNA repeats), modulation of the heterochromatin environment of the subtelomeric regions, and modulation of the alternate lengthening of telomeres (ALT) pathway (Blasco, 2007). In strains of S. cerevisiae and S. pombe defective in SUMO, Ubc9, or SUMO E3 ligases, telomeres are abnormally elongated, demonstrating a role for sumoylation in affecting one or more of these mechanisms (Chen et al., 2007; Hang et al., 2011; Tanaka et al., 1999; Xhemalce et al., 2007; Xhemalce et al., 2004; Zhao and Blobel, 2005).

The telomere elongation observed in yeast mutants defective in sumoylation is telomerase-dependent, suggesting that sumoylation normally limits the accessibility of telomeres to telomerase (Figure 3A) (Xhemalce et al., 2007). To date, there is no demonstration that telomerase itself is regulated through sumoylation. Studies have shown that sumoylation affects the activity of Cdc13, a single-stranded telomere binding protein and regulator of telomerase recruitment, by promoting Cdc13 association with the telomerase inhibitor, Stn1. Consequently, yeast strains expressing a Cdc13 mutant that cannot be sumoylated have lengthened telomeres, while shortened telomeres are observed in strains expressing a Cdc13-SUMO fusion (Hang et al., 2011). Furthermore, multiple components of the shelterin complex, which are known to limit telomerase recruitment to telomeres (de Lange, 2005), are sumoylated (Ferreira et al., 2011; Hang et al., 2011; Lu et al., 2010; Pebernard et al., 2008; Potts and Yu, 2007). Sumoylation of one or more of these factors is likely to contribute to the full inhibitory effect of SUMO on telomere lengthening. However, the underlying molecular mechanisms remain unknown.

Sumoylation may also regulate telomere length by modulating the heterochromatin environment of the subtelomere (Figure 3B). Mutations in yeast that disrupt telomeric heterochromatin structure and silencing also cause shortened telomeres, indicating that heterochromatin proteins positively regulate telomere length (Blasco, 2007). Because SUMO negatively regulates telomere length, this model would predict that sumoylation antagonizes silencing in the subtelomere. Remarkably, the data confirms this model despite the more general association of SUMO with enhanced repression. Reducing levels of sumoylation in yeast leads to increased telomeric silencing (Xhemalce et al., 2004; Zhao and Blobel, 2005), while increasing sumoylation relieves telomeric silencing (Darst et al., 2008; Nagesh et al., 2012). The molecular mechanisms underlying these effects are not fully understood, but are likely to be complex. For instance, sumoylation is required for the clustering and anchoring of telomeres to the nuclear periphery, a process that stabilizes telomeric heterochromatin and limits telomerase activity (Ferreira et al., 2011; Hari et al., 2001; Zhao and Blobel, 2005).

Finally, sumoylation is involved in telomere maintenance through effects on the ALT pathway (Figure 3C). In mammalian cells that utilize ALT, knockdown of the SUMO E3 ligase Mms21 results in reduced telomere length and increased senescence. Although this may appear to contradict phenotypes observed in yeast, this effect is unique to ALT and is not observed if telomerase is introduced into cells (Potts and Yu, 2007). The requirement of sumoylation in ALT is explained in part because telomere elongation is dependent on assembly of subnuclear structures formed around telomeres called ALT-associated PML nuclear bodies (APBs). Similar to PML nuclear bodies, assembly of APBs is SUMO dependent. Artificially tethering SUMO or Mms21 to telomeric regions is sufficient to promote APB formation, while Mms21 knockdown limits APB formation (Chung et al., 2011; Potts and Yu, 2007). Thus, requirements for sumoylation in the ALT pathway are due at least in part to an essential role in APB formation.

Centromeres

Centromeres are specialized chromatin structures that form the foundation for kinetochores and are therefore essential for proper chromosome segregation during cell division (Henikoff and Dalal, 2005). The gene encoding SUMO was first identified in yeast as a high copy suppressor of a mutant allele of the centromere-associated protein Mif2 (the vertebrate CENP-C homolog) (Meluh and Koshland, 1995), providing an early indication of a connection between sumoylation and centromeres. Since then, immunofluorescence microscopy has demonstrated that SUMO-2/3 localizes to centromeres on chromosomes formed in Xenopus egg extracts and on mammalian mitotic and meiotic chromosomes (Ayaydin and Dasso, 2004; Azuma et al., 2005; Brown et al., 2008; La Salle et al., 2008; Vigodner et al., 2006; Zhang et al., 2008). Furthermore, various SUMO E3 ligases, including PIASy, PIAS3, and Nup358/RanBP2, are also present at centromeres of mitotic chromosomes (Bantignies et al., 2011; Hari et al., 2001; Joseph et al., 2004; Ryu and Azuma, 2010). Consistent with essential roles in regulating centromere and kinetochore function in mitosis, chromosome segregation defects occur when the SUMO pathway is either up or down regulated (Biggins et al., 2001; Diaz-Martinez et al., 2006; Hari et al., 2001; Joseph et al., 2004; Mukhopadhyay et al., 2010; Seufert et al., 1995; Zhang et al., 2008). Segregation, cohesion, and other roles for sumoylation during mitosis have been characterized and are reviewed in more detail elsewhere (Dasso, 2008). Here, we focus more specifically on effects of sumoylation on centromeric heterochromatin.

In contrast to mitosis, few studies have addressed potential roles for sumoylation at centromeres or kinetochores during interphase. Of particular interest, it is not known whether SUMO remains associated with centromeres throughout the cell cycle or whether its association is specific to mitosis. A role for sumoylation in the maintenance of centromeric heterochromatin during interphase is, however, suggested by the findings that inhibiting sumoylation in both yeast and mammalian cells results in activation of genes within normally repressed centromeric regions (Figure 3D) (Marshall et al., 2010; Shin et al., 2005; Xhemalce et al., 2007). For example, S. pombe mutants lacking the Pli1 SUMO E3 ligase exhibit reduced silencing and enhanced conversion of genes inserted into core centromeric regions (Xhemalce et al., 2004). How Pli1-dependent sumoylation normally restricts transcription and recombination within this centromeric region is not fully understood. Inhibition could be mediated through effects on transcription and recombination factors, or through more direct effects on the formation and/or maintenance of centromeric heterochromatin.

One mechanism by which centromeric chromatin structure is regulated by sumoylation is through recruitment of the heterochromatin factor HP1α (Figure 3D). Sumoylation of HP1α regulates its interactions with major α-satellite RNA transcripts which in turn directs pericentric DNA targeting (Maison et al., 2011). Although sumoylation occurs within the hinge domain of HP1α thought to be involved in RNA binding, the exact molecular mechanisms underlying SUMO-dependent interactions with α-satellite RNAs remain to be determined. Intriguingly, further studies have revealed that depletion of SENP7, a SUMO protease that localizes to HP1-enriched pericentric domains, disrupts HP1α localization (Maison et al., 2012). This finding suggests that localization of HP1α is dependent on transient sumoylation of HP1α followed by desumoylation and may explain the common “SUMO enigma” (Hay, 2005), namely that steady state levels of sumoylated HP1α in the cell represent only a relatively minor fraction of total HP1α. A role for sumoylation in HP1 targeting and gene silencing has also been observed in S. pombe, where HP1 mutants that cannot be sumoylated are less efficiently recruited to heterochromatin domains and compromised in their ability to repress gene expression (Shin et al., 2005). Thus, sumoylation of HP1 represents an important and conserved regulatory point for controlling heterochromatin structure and gene expression at centromeres and other chromatin domains.

Maintenance of repetitive DNA

Repetitive DNA sequences, including those found at telomeres, centromeres and within the rDNA gene loci, represent especially fragile domains in the genome due to issues related to replication and recombination (Lovett, 2004). The maintenance of these and other repetitive DNA domains is highly dependent on the activity of the cohesion-like complex, Smc5/6, and the associated SUMO E3 ligase, Mms21 (Stephan et al., 2011). Similar to other SMC complexes, Smc5/6 activities are mediated at least in part through cohesion-related effects on higher-order chromatin structure, but also through the targeting of Mms21 to appropriate DNA targets. Smc5/6 and Mms21 are essential for the maintenance of genome integrity, with mutants exhibiting gross chromosomal rearrangements and chromosome segregation defects. These defects are due in part to the incomplete resolution of replication-associated homologous recombination intermediates, particularly within the rDNA locus and at telomeres (Figure 3F) (Behlke-Steinert et al., 2009; Bermudez-Lopez et al., 2010; Stephan et al., 2011; Torres-Rosell et al., 2005; Torres-Rosell et al., 2007).

While the exact molecular targets and functions of sumoylation in the maintenance of heterochromatic repetitive DNA are not fully understood, a growing body of evidence indicates that replication through these domains requires DNA repair factors, including BRCA1 and Rad51 (Nakamura et al., 2008; Pageau and Lawrence, 2006). Sumoylation is intimately linked to the control of these and a large number of other DNA repair factors (Bergink and Jentsch, 2009; Cremona et al., 2012; Morris et al., 2009), suggesting that Mms21-dependent sumoylation is required for proper resolution of DNA repair intermediates produced during replication. Consistent with this, recombination-dependent DNA repair intermediates accumulate during replication in Smc5/6, Ubc9 and Mms21-deficient cells (Branzei et al., 2006; Stephan et al., 2011).

Sumoylation also plays an essential role in regulating the resolution of DNA intermediates at centromeres during sister chromatid separation in mitosis by affecting the localization and activities of topoisomerase II (Figure 3F) (Azuma et al., 2005; Azuma et al., 2003; Dawlaty et al., 2008). In addition, immunofluorescence microscopy studies in mammalian cells indicate that centromeres transiently co-localize with PML nuclear bodies in G2, a phenomena that is enhanced by proteasome inhibition (Everett et al., 1999). Although the functional significance of this association remains unexplored, telomere association with APBs is required for their maintenance in telomerase-deficient cells. It is therefore tempting to speculate by analogy to APBs that interactions between centromeres and PML nuclear bodies in late G2 promotes SUMO-dependent reactions required for maintenance of centromeric DNA.

Future perspectives and conclusions

Sumoylation functions as a multifaceted regulator of chromatin structure, gene expression and genome integrity. Its utility resides in part in the ability of SUMO to elicit diverse downstream consequences following conjugation to different proteins. These consequences include affects on protein activity, localization, stability and interactions with a wide range of SIM-containing proteins. Understanding the rules that define the effects of sumoylation on specific chromatin-associated proteins, which are determined by the nature of the proteins themselves as well as any downstream interacting proteins, remains an important challenge for the field. In particular, understanding how sumoylation of different proteins mediates interactions with specific downstream SIM-containing proteins is critical. Specificity is likely to involve bivalent recognition of SUMO-modified proteins through downstream factors that contain both SIMs and motifs for recognizing the modified protein itself, as recently determined for Srs2 recognition of SUMO-modified PCNA (Armstrong et al., 2012). Specificity is also very likely to be determined and regulated by the intersection with other posttranslational modification pathways, including ubiquitylation, phosphorylation and acetylation. Understanding the crosstalk between sumoylation and other posttranslational modifications in greater detail is also an important challenge for the field.

Defining the role of sumoylation in controlling chromatin structure and function more specifically and at a molecular level will also require a more detailed characterization of relevant SUMO-modified proteins. While the identification of SUMO-modified proteins with roles in chromatin structure and function has expanded greatly in the past several years, the functional effects of sumoylation on the majority of these proteins remain unknown. Characterizing the effects of sumoylation on individual proteins is often challenging, due in part to the relatively small fraction of most proteins modified at steady state. Other challenges involve identifying approaches for specifically affecting sumoylation of individual proteins or pathways. While many important studies linking sumoylation to chromatin structure and gene expression have relied on global activation or suppression of sumoylation, more targeted approaches are needed. The identification of new and functionally unique E3 ligases, as exemplified by Mms21, represents one avenue for developing more specific approaches. The identification of separation-of-function alleles of SUMO or SUMO pathway enzymes in yeast or other genetically tractable organisms could also prove valuable.

Finally, a more detailed understanding of the genome-wide interactions between SUMO (and SUMO pathway enzymes) and chromatin is needed. ChIP experiments on a small scale provided the surprising finding the SUMO is associated with the promoters of active but not repressed genes (Rosonina et al., 2010). Whole chromosome ChIP experiments, revealing genome-wide associations of SUMO during different stages of the cell cycle cell or under different cell growth conditions, could be particularly insightful and provide even more surprises.

In summary, we have reviewed the role of SUMO as a regulator of chromatin structure and function. In one important capacity, sumoylation regulates the assembly of multi-protein complexes on chromatin, including repressive complexes organized around sites of DNA methylation and PcG bodies, as well as transcriptional regulatory complexes at gene promoters. A recurring theme from the reviewed studies is the dichotomous role of sumoylation. By facilitating the assembly of distinct complexes, sumoylation affects the chromatin environment in ways that can either activate or repress gene expression. In addition to facilitating protein complex assembly, sumoylation also affects proteins in multiple other ways by mediating changes in localization, stability or enzymatic activity. Thus, another recurring theme is the diverse and context-dependent effects of sumoylation on chromatin-associated proteins. Finally, we have reviewed the prominent role played by sumoylation in maintaining the integrity of repetitive heterochromatin domains, including telomeres, centromeres and rDNA loci. Collectively, the reviewed studies reveal the incredible versatility of sumoylation, which affects chromatin structure and function at multiple levels and through multiple mechanisms.