INVOLVEMENT OF SUMO MODIFICATION IN MBD1- AND MCAF1-MEDIATED HETEROCHROMATIN FORMATION

Small ubiquitin-related modifiers, SUMO-2/3 and SUMO-1, are involved in gene regulation and nuclear structures. However, little is known about the roles of SUMO, in heterochromatin formation of mammalian cells. Here we demonstrate that SUMOs directly interact with human MCAF1, which forms complexes with either the methyl-CpG-binding protein MBD1 or SETDB1, which trimethylates histone H3 at lysine 9 (H3-K9) in the presence of MCAF1. Modification of MBD1 with either SUMO-2/3 or SUMO-1 facilitated the interaction between MBD1 and MCAF1, suggesting that SUMOylation links the methylation of DNA and histones. In a cultured human cell line, SUMOs were localized in MBD1- and MCAF1-containing heterochromatin regions that were enriched in trimethyl-H3-K9 and the heterochromatin proteins HP1beta and HP1gamma. Specific knockdown of either SUMO-2/3 or SUMO-1 induced dissociation of MCAF1, trimethyl-H3-K9, and the HP1 proteins from the MBD1-containing heterochromatin foci, suggesting a requirement for SUMOs for heterochromatin assembly. These findings provide insights into the roles of SUMOylation in the regulation of heterochromatin formation and gene silencing.

SUMOs are small ubiquitin-related modifiers that covalently attach to their target proteins, a process referred to as SUMOylation. This process involves an enzymatic pathway, which is similar to the pathway employed in the ubiquitin conjugation cascade (1)(2)(3). In contrast to ubiquitin, however, there are at least three SUMO paralogs in human cells, namely SUMO-2/SMT3A, SUMO-3/SMT3B and SUMO-1/SMT3C (4). SUMO-2 and -3 are more closely related to each other (95% amino acid identity) than to SUMO-1 (~50% identity). It has become increasingly clear that a wide variety of cellular proteins can be modified by either SUMO-2/3, SUMO-1 or both, leading to alterations in many signaling pathways associated with their target proteins. Most, but not all, of the SUMOylation target proteins appear to be involved in maintenance of nuclear integrity, regulation of nuclear transport or control of chromatin functions, such as transcription, replication, repair, recombination and chromatin modification (5,6).
Currently, there are numerous examples of co-repressor proteins that can be SUMOylated and transcription factors whose activities are down-regulated by SUMOylation, implying the involvement of a SUMO modification system in the regulation of transcriptional repression and maintenance of silenced heterochromatin (7)(8)(9). Although the exact mechanisms by which SUMO modification contributes to such anti-activation activities in chromatin remain poorly defined, it has been speculated that SUMOylation mediates changes in gene expression and chromatin assembly, at least, by serving as a binding platform for recruiting other chromatin proteins that may transmit or amplify the SUMO signal and maintain 2 the silenced state of chromatin (7)(8)(9). More recently, SUMO-mediated recruitment of histone deacetylases has been reported to play an important role (10)(11)(12)(13)(14), implying the possibility that histone-modifying enzymes may act as effectors for SUMO modification signals.
Covalent addition of a methyl group to the 5 position of cytosine (5mC) is the major modification of DNA in vertebrate genomes. This modification predominantly occurs within CpG dinucleotides, and is involved in a wide range of biological phenomena, including genomic imprinting, X-chromosome inactivation and tissue-specific gene expressions (15,16). Furthermore, it is well established that sites of DNA methylation are recognized by a family of proteins designated methyl-CpG-binding proteins (16,17). These proteins utilize transcriptional co-repressors or mediators to silence transcription and also modify the surrounding chromatin, thereby providing a link between DNA methylation and chromatin remodeling and modifications. Among these proteins, MBD1 is relatively well characterized, and has been implicated in regulating chromatin structure and gene silencing through a currently unknown mechanism that probably involves histone modifications, such as deacetylation and H3-K9 methylation (16)(17)(18)(19)(20)(21).
Considering the biological consequences of the effect of MBD1 on DNA methylation, the roles of an MBD1-interacting protein, designated MBD1-containing chromatin-associated factor 1 (MCAF1; also known as ATFa-associated modulator (AM)), are intriguing. Specifically, this protein has been reported to interact with MBD1 via the transcriptional repression domain (TRD) at the carboxyl-terminal region of MBD1 (21), and it also forms a tight complex with a histone H3-K9 methyltransferase, SETDB1/ESET, that appears to facilitate SETDB1-dependent conversion of dimethyl-H3-K9 to the trimethyl state (22). In addition to MBD1 and SETDB1, MCAF1 also interacts with several transcriptional factors, including ATFa (23), Sp1 (24) and the homeobox-containing zinc finger protein ZHX1 (25), suggesting that it acts as an recruiter for a wide range of proteins that can modulate gene regulation and chromatin formation. However, the mechanisms involved in regulating the assembly of macromolecular complexes containing MCAF1 and how such chromatin complexes contribute to the regulation of gene silencing and heterochromatin formation remain largely uncharacterized.
Here, we report that MCAF1 interacts with SUMO-2/3 and SUMO-1, with a preference for SUMO-2/3, via short peptide sequences similar to the SUMO-binding motif. Using a human cultured cell line, we demonstrate SUMOylation of MBD1 both in vitro and in vivo, as well as facilitated anchorage of MCAF1 to SUMOylated MBD1. Furthermore, RNA interference (RNAi) experiments directed against SUMO-2/3 or SUMO-1 reveal that depletion of the SUMO pathways perturbs the assembly of MCAF1, trimethyl-H3-K9, HP1 and HP1 at MBD1-containing heterochromatin. Taken together, our results indicate that SUMOs function as epigenetic modulators for heterochromatin formation, at least in part, by regulating the MCAF1-MBD1 interaction.

RESULTS
MCAF1 binds to SUMO-1 and SUMO-2/3 with a preference for SUMO-2/3. To identify possible chromatin proteins that may specifically recognize the SUMO moiety and be recruited to sites of SUMOylation, we performed a yeast two-hybrid screening using the Gal4-SUMO-3G-SUMO-3G fragment, which was incapable of conjugation, as bait. After multiple rounds of screening of a mouse 11-day embryo cDNA library, we isolated a clone encoding amino acids 787-1306 of the ATFa-associated factor AM, a mouse homolog of human MCAF1 (Fig. 1A).
As a step toward understanding the role of MCAF1 binding to SUMO, we first performed in vitro GST pull-down assays to investigate whether the binding was direct and whether the binding affinities varied among the SUMO paralogs. When beads bound to GST-SUMO-1, GST-SUMO-2, GST-SUMO-3 or GST alone were incubated with a lysate from C-33A human cervical cancer cells, MCAF1 associated more efficiently with GST-SUMO-2/3-beads than with GST-SUMO-1-beads (Fig. 1B). MCAF1 binding to SUMO-1 and SUMO-2/3 was direct, since bacterially expressed recombinant full-length (His) 6 -MCAF1 protein was retained on GST-SUMO-conjugated beads (Fig. 1C).

MCAF1 contains a conserved amino acid sequence similar to the SUMO-binding motif.
Mouse and human MCAF1 proteins have been demonstrated to belong to an evolutionarily conserved family of MCAF/AM proteins (24). A comparison of MCAF family members revealed two highly conserved domains in both mouse and human MCAF1 proteins, referred to as Domain 1 and Domain 2. To determine the region responsible for binding to SUMOs, we generated a series of deletion mutants of human MCAF1, and performed GST pull-down assays. The results revealed that 11 amino acid residues (amino acids 965-975) of MCAF1 (MCAF1 965-975 ) located between Domain 1 and Domain 2 were sufficient for a specific interaction with (His) 6 -SUMO-3 (Fig.  1A). A semi-quantitative binding analysis revealed that the affinity of MCAF1 965-975 for (His) 6 -SUMO-2/3 was approximately 5-fold higher than that for (His) 6 -SUMO-1 (Fig. 1D).
A database search for MCAF1 965-975 using the basic local alignment search tool algorithm BLAST detected several peptide sequences encoded in mouse, rat, dog and chicken MCAF1 (Supplemental Fig. 1A), suggesting evolutionary conservation of this amino acid sequence. We also found that the N-terminal half of this sequence partially overlapped with a recently described hydrophobic amino acid cluster in the SUMO-binding motif present in PIAS1, PIASx, SAE2, PML IV, RanBP2/Nup358 and thymine DNA glycosylase (Fig. 1E) (6,28,(30)(31)(32). It should be mentioned that our alanine scanning analysis revealed that valine (V) at residue 966, isoleucine (I) at 967, aspartic acid (D) at 968, leucine (L) at 969 and threonine (T) at 970 in MCAF1 965-975 were essential for the binding (Supplemental Fig. 1B). In addition, we found that peptides in which D at 973 5 or glutamic acid (E) at 975 was substituted for asparagine (N) or alanine (A), respectively, showed dramatically reduced affinities for SUMO-2/3 (Supplemental Fig. 1B) as well as SUMO-1 (data not shown), suggesting an important role of this acidic amino acid cluster for the binding. Although several previously described SUMO-interacting peptides, including PIAS1, PIASx, SAE2 and PML IV, also contain an acidic cluster at their C-terminus ( Fig. 1E) (6,30), the roles of the negatively-charged residues in the SUMO binding have not been investigated in detail.

The SUMO-MCAF1 interaction occurs in vivo.
To examine the importance of the amino acid residues in MCAF1 965-975 for the binding to SUMO-2/3, two MCAF1 point-mutants in which D968 and L969 were substituted for alanine (A) were generated and designated (His) 6 -MCAF1-D968A and (His) 6 -MCAF1-L969A, respectively. GST pull-down assays revealed that neither of the mutants showed stable binding to GST-SUMO-2/3 ( Fig. 2A), despite the observed association between wild-type (His) 6  Given that Myc-SUMO-3G is a SUMO-3 mutant incapable of conjugation, these data provide evidence that MCAF1 interacts non-covalently with SUMO-2/3 in vivo, and further confirm that D968 and L969 are necessary components for the binding to SUMO-2/3. Similar results were obtained when Myc-SUMO-1G was used (data not shown).
Endogenous SUMO-2/3, SUMO-1 and MCAF1 are colocalized. Thus far, our results had implied that MCAF1 acts as a SUMO-binding protein.
Therefore, we next attempted to assess the physiological relevance of the endogenous SUMO-MCAF1 interaction using mammalian cultured cells. As shown in Fig. 3A, indirect immunofluorescence experiments revealed a large number of C-33A cells containing discrete nuclear foci in which SUMO-2/3 and MCAF1 were colocalized, although some SUMO-2/3 and MCAF1 may also be distributed throughout the nucleoplasm, except for the nucleolus. Both the sizes and numbers of the SUMO-2/3-MCAF1 foci varied among individual cells with an average of 2-6 clear and large foci per cell. Double-staining with anti-SUMO-2/3 and anti-SUMO-1 antibodies (Fig. 3B, upper panel) or anti-MCAF1 and anti-SUMO-1 antibodies (data not shown) revealed that the merged signals of the nuclear foci were almost indistinguishable, suggesting enrichment of SUMO-1 in SUMO-2/3-MCAF1 foci as well. Therefore, these data indicate the relevance of not only SUMO-2/3-MCAF1, but also SUMO-1-MCAF1, interactions in vivo. Although it is possible that a larger amount of SUMO-2/3 may be enriched with MCAF1 via the preferential association of SUMO-2/3 with MCAF1, our present indirect immunofluorescence technique could not quantitatively compare the levels of SUMO-2/3 versus SUMO-1 accumulation in MCAF1 foci.
SUMOylated forms of MCAF1 are barely detectable. Since MCAF1 and SUMOs interacted in vitro and colocalized in vivo, one could argue that MCAF1 itself could be an efficient substrate for SUMOylation. To elucidate whether MCAF1 was SUMOylated in C-33A cells, we performed an immunoprecipitation analysis of endogenous MCAF1 (Supplemental Fig. 2). When the proteins immunoprecipitated with the anti-MCAF1 antibody were probed with either anti-MCAF1, anti-SUMO-2/3 or anti-SUMO-1 antibodies, none of the antibodies detected any SUMOylated bands, which were expected to migrate more slowly than the 240-kDa band of non-modified full-length MCAF1 (Supplemental Fig. 2, arrowhead). Since more than 95% of the total cellular pool of MCAF1 could be extracted under our experimental conditions (data not shown), these results indicate that either most of the endogenous MCAF1 is modified very poorly, if at all, by SUMOs or SUMOylated MCAF1 is unstable in C-33A cells.

MBD1
is an efficient substrate for SUMOylation. We therefore hypothesized that there may be SUMOylated proteins in MBD1-containing heterochromatin, and that SUMOylation of such proteins may provide sufficient anchoring of MCAF1 to heterochromatin regions. A straightforward test of this idea was to determine the SUMOylated proteins residing in heterochromatin, and to investigate whether SUMOylation of such proteins provided a sufficient binding platform for MCAF1. For these purposes, we undertook a "candidate screening" approach to identify proteins that could be SUMOylated and should be present in heterochromatin.
Among the proteins tested, the methyl-CpG-binding protein MBD1 was identified as a good candidate. As shown in Fig. 4A and B, the bacterial SUMOylation system (27) revealed that a GST protein fused to the full-length form of MBD1 (GST-MBD1) was efficiently SUMOylated, further demonstrating that the C-terminal region of MBD1 (MBD1 373-605 ), which contained the TRD, was responsible for SUMOylation. Moreover, we revealed SUMOylation of MBD1 using an ectopic expression system in mammalian cultured cells (Fig. 4C). When we expressed FLAG-tagged MBD1 in HeLa cells and performed an immunoprecipitation analysis with an anti-FLAG antibody, multiple high-molecular mass bands were detected by the anti-SUMO-2/3 antibodies, suggesting that FLAG-MBD1 was efficiently modified by endogenous SUMO-2/3 (lanes 2 and 5). Multiple SUMOylated bands were also detected when FLAG-MBD1 was transfected along with Myc-SUMO-3 (lanes 3 and 6). Of note, FLAG-MBD1 proteins modified by Myc-SUMO-3 migrated slightly more slowly than FLAG-MBD1 proteins modified by endogenous SUMO-2/3, due to the addition of Myc-tag moieties. We also found that ectopically expressed FLAG-MBD1 was efficiently modified by either endogenous SUMO-1 or ectopically expressed Myc-SUMO-1 (data not shown). Taken together, these results demonstrate that MBD1 is efficiently modified by either SUMO-2/3, SUMO-1 or both in vitro as well as in vivo.
Endogenous MBD1 is SUMOylated. To formally confirm that MBD1 is indeed a physiological SUMOylation substrate, we immunoprecipitated endogenous MBD1, and investigated the existence of the SUMOylated form of MBD1. As shown in Fig. 4D, in addition to the 85-and 90-kDa bands corresponding to non-modified forms of the endogenous spliced variants of MBD1 (26,33), an anti-MBD1 antibody precipitated multiple high-molecular mass bands, particularly around 140-and 180-kDa (white circles). Furthermore, these bands could be superimposed with bands detected by the anti-SUMO-2/3 and anti-SUMO-1 antibodies, demonstrating SUMOylation of endogenous MBD1. It should be noted that a single molecule of SUMO is expected to migrate at around approximately ~15-kDa. Therefore, the band shifts from 85~90-kDa (non-modified forms) to 140~180-kDa (modified forms) suggested that 4-6 molecules of either SUMO-2/3, SUMO-1 or both were appended to a single molecule of MBD1. Taken together, these results indicate that a significant proportion of endogenous MBD1 is modified by either SUMO-2/3, SUMO-1 or both in 7 vivo.
Depletion of either the SUMO-2/3 or SUMO-1 pathway perturbs the assembly of MCAF1, trimethy-H3-K9 and HP1 foci. Our finding that SUMOylation of MBD1 facilitates its association with MCAF1 supports the idea that MCAF1 is enriched in heterochromatin regions, at least in part, via anchoring to SUMO-modified MBD1. Therefore, we used RNAi directed against SUMO-2/3 to investigate whether SUMO-2/3 is critical for the assembly of MCAF1 at MBD1-containing heterochromatin regions in vivo.
To demonstrate the specificity of the RNAi technique, we used two different siRNA duplexes for SUMO-1 (SUMO-1#1 and #2 siRNAs) and 2 independent combinations of SUMO-2 and -3 siRNA duplexes (SUMO-2/3 #1 and #2 siRNAs). For control experiments, a siRNA against firefly luciferase GL3 was used. In each siRNA experiment, judging from immunofluorescence analysis using anti-SUMO-2/3 or anti-SUMO-1 antibodies, ~60% of the cells showed no detectable levels of either SUMO-1 or SUMO-2/3 at 60 h post-transfection with either SUMO-2/3 or SUMO-1 siRNAs, respectively, whereas no reduction in the signals was observed in the control experiments ( Fig. 6A-E, and data not shown).
A comparison of the SUMO knockdown cells and control cells revealed no significant differences in terms of the localization of MBD1 or the number of MBD1 foci (data not shown), implying that depletion of either the SUMO-2/3 or SUMO-1 pathway may not completely impair the ability of MBD1 to localize at specific DNA regions, possibly via interactions between the methyl-CpG-binding domain of MBD1 and methylated DNA. In contrast, 52% and 61% of SUMO-2/3 and SUMO-1 knockdown cells, respectively, showed no significant accumulation of MCAF1 at MBD1-containing foci ( Fig. 6A and  F). Moreover, 75% and 67% of SUMO-2/3 and SUMO-1 knockdown cells, respectively, showed decreased signals for trimethyl-H3-K9 around the MBD1-containing regions ( Fig. 6B and F). In agreement with these observations, 57% and 43% of SUMO-2/3 and SUMO-1 knockdown cells, respectively, showed HP1 delocalization from the MBD1 foci ( Fig. 6C and F), implying that the formation of MBD1-containing heterochromatin was perturbed. We further tested the subcellular 8 localizations of other HP1 proteins, HP1 and HP1 , and found delocalization of HP1 , but not HP1 , in SUMO-2/3 and SUMO-1 siRNA-treated cells ( Fig. 6D and E, and data not shown). Taken together, we conclude that the SUMO-2/3 and SUMO-1 pathways are involved in the anchorage of MCAF1, and are also critical for methylation of histone H3-K9 and targeting of HP1 and HP1 to MBD1-containing heterochromatin. (DsRed-MCAF1 965-975 ) was transiently overexpressed in C-33A cells, the signals for both endogenous SUMO-2/3 (Fig. 7A) and SUMO-1 (data not shown) were remarkably reduced, suggesting that overproduction of monomeric DsRed-MCAF1 965-975 effectively perturbed both SUMO pathways, possibly via sequestration of the SUMOs and/or competition with pre-existing SUMO-binding proteins. Significantly, in these DsRed-MCAF1 965-975 -expressing cells, the numbers of cells in which both MCAF1 and HP1 were delocalized from MBD1-containing foci increased ( Fig. 7B and C), reminiscent of the depletion of the SUMO pathway by siRNAs described above. Since overexpression of monomeric DsRed alone had no apparent effect on the localizations of MCAF1 and HP1 at MBD1 foci ( Fig. 7A-C, upper panels), these results indicate the importance of the SUMO-binding region, MCAF1 965-975 , for regulating the anchorage of MCAF1 to MBD1-containing heterochromatin and further support the idea that SUMO-MCAF1 interactions are required for the proper assembly of MBD1-containing heterochromatin.

MCAF1 is a SUMO-binding protein.
In the present paper, we have identified a previously unknown SUMO-binding region in MCAF1 (MCAF1 965-975 ) that facilitates its interaction with SUMOs. Significantly, we have revealed that MCAF1 965-975 shows similarity to a previously characterized SUMO-binding motif (6,28,30-32). Moreover, we found that MCAF1 preferentially interacts with SUMO-2/3 rather than SUMO-1 in vitro. These results imply a previously unappreciated function of MCAF1 as a modulator for SUMO modification signals with the potential to preferentially transmit signals derived from SUMO-2/3-modified proteins. In addition, we found that following multisite-modification and/or poly-chain formation of SUMOs on MBD1 373-605 , MCAF1 appeared to further increase its affinity for MBD1. This finding further supports the idea that MCAF1 functions as a modulator of SUMO signaling and suggests that MCAF1 may be able to amplify SUMO signals to downstream events. Thus, our results define MCAF1 as a novel class of SUMO regulator and predict the existence of a previously undescribed SUMO-MCAF1-based regulatory network.

MBD1 is a SUMOylation substrate.
In the present study, we have shown for the first time that the methyl-CpG-binding protein MBD1 is SUMOylated in mammalian cells, suggesting direct linkage of the SUMO modification pathway with a wide variety of important epigenetic cellular phenomena regulated by DNA methylation, including gene silencing and heterochromatin formation. It is feasible that other methyl DNA-binding proteins, besides MBD1, that reside in heterochromatin may also be SUMOylated. We are currently investigating whether other methyl DNA-binding proteins, including other MBD family proteins (16), MeCP2(34) and kaiso (35), are SUMOylated using the bacterial SUMOylation system.
It is currently unclear how SUMOylation of MBD1 is regulated and whether a SUMO E3 ligase that upregulates SUMOylation of MBD1 is present. Since previous studies have shown direct interactions of MBD1 with MCAF1 (21,24), we first suspected that MCAF1 may act as a SUMO E3 ligase toward MBD1. However, neither binding of a SUMO E2 enzyme (Ubc9) to MCAF1 nor the ability of MCAF1 to enhance SUMOylation of MBD1 in vitro was observed (Supplemental Fig.  3A and B). In addition, while many of the SUMO E3 ligases reported to date have the ability to 9 become auto-SUMOylated (36-39), MCAF1 appears to be inefficiently auto-SUMOylated. Thus, there is still poor evidence to support a role for MCAF1 as a SUMO E3 ligase. It will be interesting in the near future to identify the SUMO E3 ligases for MBD1 and to elucidate how such E3 ligases regulate the SUMOylation of MBD1 during cell cycle progression and cellular differentiation.
The SUMO modification pathway cross-talks with the DNA and histone methylation pathways. We have shown that SUMOylation of MBD1 provides sufficient anchoring of MCAF1 at heterochromatin regions in C-33A cells and demonstrated that depletion of either the SUMO-2/3 or SUMO-1 pathway results in delocalization of trimethyl-H3-K9, HP1 and HP1 from MBD1 foci, suggesting the possibility that destabilization of the interaction between MCAF1 and MBD1 perturbs the histone methylation pathway and the proper assembly of heterochromatin proteins at MBD1-containing DNA regions.
The molecular mechanism for how the SUMO pathway is linked to histone methylation and heterochromatin formation in mammalian cells currently remains to be fully elucidated, although two possibilities appear feasible. First, augmented recruitment of MCAF1 via an interaction with SUMOylated MBD1 may contribute to the maintenance of a stable assembly of H3-K9 methyltransferases at MBD1-containing heterochromatin regions, thereby stabilizing the association of HP1 with MBD1-containing heterochromatin. In the present study, we have not identified any such putative H3-K9 methyltransferases that can be recruited by MCAF1 complexed with SUMOylated MBD1. However, we suggest that SETDB1 is a likely candidate for this scenario, since previous reports have described that MCAF1 is a regulatory subunit of the SETDB1 histone H3-K9 methyltransferase complex (22,24) and that MCAF1 complexed with SETDB1 modulates the histone methylase activity of SETDB1, converting it from an H3-K9 dimethylase to a trimethylase (22,24). Second, anchorage of MCAF1 to SUMOylated MBD1 may enhance the recruitment of chromatin remodeling activity and/or histone chaperone activity that somehow preferentially incorporate trimethyl-H3-K9. The chromatin assembly factor CAF1, which interacts with MBD1, represents a likely candidate for this scenario (18,40), although neither facilitated association of CAF1 with MCAF1-MBD1 complexes nor accumulation of CAF1 at SUMO-MCAF1 foci in C-33A cells have yet been demonstrated.
Regardless of the mechanism, our results implies a direct link between the SUMO pathway and the methylation of DNA and histone methylation. In addition, the dramatic effects on the nuclear localizations of several heterochromatin proteins, including trimethyl-H3-K9 and HP1 proteins, in either SUMO-2/3-or SUMO-1-depleted cells reveal highly dynamic features of the SUMO modification pathway in the context of heterochromatin formation. It is intriguing that HP1 , a well-described constitutive heterochromatin protein in mammalian cells (41), seems to exhibit a different role in terms of the regulation of the SUMO-enriched heterochromatin regions in C-33A cells compared with the two other members of the HP1 protein family examined. Although these results imply that the various HP1 members have only overlapping functions with respect to the regulation of SUMO-enriched heterochromatin, the physiological relevance of this phenomenon remains to be elucidated in future studies.
In conclusion, our results have provided direct evidence that modification of MBD1 by either SUMO-2/3, SUMO-1 or both, and facilitated anchorage of MCAF1 to SUMOylated MBD1 participate in the formation of heterochromatin in C-33A cells. Considering the increasing numbers of SUMO substrates involved in chromatin modification, remodeling and epigenetic control, our findings lay the foundation for future exploration of currently undiscovered SUMO signaling events that regulate gene silencing and heterochromatin formation in both physiological and pathological situations.

FOOTNOTES
We would like to thank Dr. N. Saitoh and Dr. S. Watanabe for critically reading this manuscript, and all our laboratory staff members for their helpful discussions. H.S. dedicates this work to the late S. Mizuno, who was an excellent mentor during his graduate studies at Tohoku University, Sendai, Japan, and the late A.P. Wolffe, who was a good advisor during his post-doctoral training at NIH, Bethesda, USA. Their enthusiasm and curiosity were the motivation behind this study to find a link between the SUMO modification pathway and the mechanisms behind chromatin dynamics and epigenetic control.   6 and 10), and -L969A (lanes 7 and 11) mutants were incubated with beads containing 3 μg of GST-SUMO-2 (lanes 5-7) and GST-SUMO-3 (lanes 9-11), respectively. For control experiments, PBS buffer was mixed with GST-SUMO-2-or GST-SUMO-3-conjugated beads (lanes 4 and 8, respectively). Following incubation, GST pull-down assays were carried out, and the proteins associated with the beads were subjected to immunoblot analysis using an anti-T7 antibody. A total of 5% of the input of each lysate was loaded in    3 and 6). As a control, the vector plasmid only was transfected (lanes 1 and 4). Nuclear extracts prepared from the transfected cells were subjected to immunoprecipitation using an anti-FLAG M2 antibody, followed by immunoblotting with anti-FLAG M5 (lanes 1-3) or anti-SUMO-2/3 (lanes 4-6) antibodies. Immunoprecipitation was carried out with RIPA buffer using highly stringent conditions that do not allow the precipitation of non-covalently associated proteins. Arrows indicate the position of the expressed FLAG-MBD1 (non-modified form). Circles represent the positions of the expressed FLAG-MBD1 modified with endogenous SUMO-2/3. Arrowheads show the positions of the expressed FLAG-MBD1 modified with Myc-SUMO-3. (D) Endogenous MBD1 is SUMOylated in C-33A cells. Nuclear extracts were prepared from exponentially growing C-33A cells, followed by immunoprecipitation analysis using an anti-MBD1 antibody (lanes 3, 6 and 9) or IgG as a control (lanes 2, 5 and 8). Immunoprecipitation was carried out with RIPA buffer using highly stringent conditions. Immunoblot analysis of the immunoprecipitated proteins was carried out using anti-MBD1 (lanes 1-3), anti-SUMO-2/3 (lanes 4-6) or anti-SUMO-1 (lanes 7-9) antibodies. A total of 1% of each nuclear extract (input) was applied in lanes 1, 4 and 7, respectively. Arrows indicate the positions of the 85-kDa and 90-kDa proteins that may represent the two major splicing variants of MDB1 in C-33A cells.
Circles indicate several bands reactive to anti-MBD1, anti-SUMO-2/3 and anti-SUMO-1 antibodies. Black circles represent non-specific bands.   Single-point-mutants of MCAF1 965-975 were generated by individually changing appropriate amino acids to alanine (A), asparagine (N) or glutamine (Q). A GST fusion protein with each mutated peptide (5 μg) was mixed with a fixed amount of (His) 6 -SUMO-3 (5 μg) and analyzed by a GST pull-down assay. The proteins associated with GST-beads were separated by 15% SDS-PAGE, and visualized by Coomassie Brilliant Blue staining. The intensity of each (His) 6 -SUMO-3 band was measured, and the relative intensity to that of the SUMO-3 binding of the wild-type peptide was calculated (black bars). a.u., arbitrary relative band density units.