SUMOylation of the polyglutamine repeat protein, ataxin-1, is dependent on a functional nuclear localization signal.

SUMO (small ubiquitin-like modifier) is a member of the ubiquitin family of proteins. SUMO targets include proteins involved in numerous roles including nuclear transport and transcriptional regulation. The previous finding that mutant ataxin-1[82Q] disrupted promyelocytic leukemia (PML) oncogenic domains prompted us to determine whether ataxin-1 disrupts another component of PML oncogenic domains, Sp100 (100-kDa Speckled protein). Similar to the PML protein, mutant ataxin-1[82Q] redistributed Sp100 to mutant ataxin-1[82Q] nuclear inclusions. Based on the ability of PML and Sp100 to be covalently modified by SUMO, we investigated the ability of ataxin-1 to be SUMOylated. SUMO-1 was found to covalently modify the polyglutamine repeat protein ataxin-1. There was a decrease in ataxin-1 SUMOylation in the presence of the expanded polyglutamine tract, ataxin-1[82Q]. The phospho-mutant, ataxin-1[82Q]-S776A, restored SUMO levels to those of wild-type ataxin-1[30Q]. SUMOylation of ataxin-1 was dependent on a functional nuclear localization signal. Ataxin-1 SUMOylation was mapped to at least five lysine residues. Lys(16), Lys(194) preceding the polyglutamine tract, Lys(610)/Lys(697) in the C-terminal ataxin high mobility group domain, and Lys(746) all contribute to ataxin-1 SUMOylation.

Protein modification by small polypeptides is an important mechanism for tightly controlling protein events. Ubiquitin is one such attachment, and it is clear that ubiquitylation of proteins regulates more than their cellular metabolism. The ubiquitin-like modifiers are a group of proteins having homology to ubiquitin. These proteins modify their target proteins in an enzymatic pathway similar to ubiquitin conjugation. SUMO 1 (small ubiquitin-like modifiers) is a member of this family of proteins, and SUMO targets include proteins involved in subcellular trafficking and gene regulation (1,2). In some cases such as IB␣, SUMOylation functions as an antagonist to ubiquitylation, generating proteins resistant to degradation (3). SUMO targets include diverse substrates such as promyelocytic leukemia (PML), 100-kDa Speckled protein (Sp100), RanGAP1, histone deacetylase 4, and p53 (4 -11). The list of proteins targeted by SUMO is rapidly increasing, and this modification has now been implicated in several neurodegenerative diseases (12,13).
SUMO was identified in a yeast two-hybrid screen using PML as bait. SUMO has been given numerous other names including PML-interacting protein (PIC1), Sentrin, and GAPmodifying protein 1 but now is commonly referred to as SUMO (5). In mammalian cells, there are three SUMO genes, SUMO-1, SUMO-2, and SUMO-3. The SUMO-2 and SUMO-3 genes share ϳ50% identity to SUMO-1. Unlike SUMO-1, SUMO-2 and SUMO-3 contain an internal SUMO consensus sequence and are capable of forming poly-SUMO chains (14). There is some sequence homology between ubiquitin and SUMO (18%), but structurally they adopt nearly identical structures (15). Similar to ubiquitin, SUMO conjugation occurs via covalent modification of target lysine residues (15,16). Functionally, SUMOylation mimics ubiquitylation in a parallel but non-overlapping E1/E2/E3 enzyme cascade (16). To date, only a single E1 heterodimer complex (Uba2/Aos1) and a single E2-conjugating enzyme (Ubc9) have been described for SUMO (16). Three classes of E3-conjugating enzymes have been reported for SUMOylation: 1) the PIAS (Protein Inhibitors of Activated STATs) of which p53 is a substrate; 2) the nucleoporin RanBP2, a member of the nuclear pore complex; and 3) the polycomb group protein Pc2 (1,17,18). Spinocerebellar ataxia type 1 (SCA1) is a dominantly inherited progressive neurodegenerative disease that results in atrophy of cerebellar Purkinje cells (19). SCA1 is caused by the expansion of a CAG trinucleotide repeat tract in the SCA1 gene, resulting in an abnormally long polyglutamine tract within the protein (19). Accumulation of mutant ataxin-1[82Q] into nuclear inclusions is a hallmark of disease. Ataxin-1 contains a functional nuclear localization signal (NLS), and mutation of this NLS results in cytoplasmic localization of ataxin-1 and a reduction in disease pathology (20).
Previously, it was shown that mutant ataxin-1[82Q] redistributed PML oncogenic domains (PODs) to large mutant ataxin-1[82Q] foci or nuclear inclusions (21). The effect of this redistribution is unknown; however, the cellular distribution of PODs in relationship to nuclear organization is tightly regulated (22)(23)(24). The presence of PML is critical for the proper formation of PODs (25). SUMO modification of PML is required for the recruitment of other proteins into the PODs, suggesting a dynamic balance between PML SUMOylation and protein localization to PODs (16,26). PML and PODs are thought to have a role in transcription (both as co-activators and co-repressors), translation, DNA repair/DNA replication, mRNA transport/stability, and proteasome-dependent degradation (27).
Although recent studies have focused on PML and its involvement with PODs, the first protein isolated from PODs was Sp100 (5, 28). Sp100 does not interact directly with PML (4). Sp100 is SUMOylated, although SUMOylation is not essential for POD formation or nuclear localization, suggesting that Sp100 SUMOylation is involved in mediating other events. Sp100 contains a high mobility group (HMG-1) domain and has been shown to interact with members of the heterochromatin family (HP1) that function as transcriptional co-repressors. The SUMOylation site of Sp100 overlaps with the HP1 binding domain, and SUMOylation could regulate the interaction of Sp100 with HP1 (4,16,29).
In addition to PML and Sp100, SUMOylation has been described for RanGAP1 where SUMOylation targets RanGAP1 to the nuclear pore complex. As a result of a SUMO-1 E3 ligase residing at the nuclear pore, RanBP2, it has been hypothesized that a large number of proteins are SUMOylated at the nuclear pore (1). Besides regulating subnuclear localization, SUMOylation is hypothesized to regulate import/export efficiency.
Here we show that ataxin-1 is SUMOylated and that SUMOylation is dependent on the length of the polyglutamine tract, the ability of ataxin-1 to be phosphorylated at serine 776, and the integrity of the NLS of ataxin-1. The dependence of ataxin-1 on phosphorylation and nuclear localization is reminiscent of other nuclear body proteins, namely PML, Sp100, and histone deacetylase 4, where SUMOylation controls their nucleocytoplasmic trafficking as well as their ability to function as transcriptional regulators.
Alternately, cells were washed as described above and then harvested using denaturing buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% Triton-X-100, 0.5% sodium deoxycholate, and 0.1% SDS containing protease/phosphatase inhibitors as described above) directly in a 60-mm plate. After the addition of denaturing buffer, the cells were lysed with rocking at 4°C for 15 min (31). Extracts were sheared and centrifuged as described above. Immunoprecipitations were performed with 500 g of cell lysate, 50 l of protein A, and 2 l of anti-ataxin-1 antibody 11750 with rocking overnight at 4°C (4, 5). The following day, the beads were pelleted by centrifugation for 20 s at 12,000 rpm and washed four times with denaturing buffer. The beads were heated for 10 min at 70°C in SDS-loading dye (4 times, 30 l, Invitrogen) plus sample reducing agent (3 l, Invitrogen). Immunoprecipitations were electrophoresed through 4 -12% Bis-Tris gels using MOPS buffer (Invitrogen) and transferred to nitrocellulose membranes (Protran, Schleicher & Schuell). Membranes were probed overnight at 4°C with anti-SUMO-1 antibody (1:1000, Zymed Laboratories Inc.) or with anti-HA antibody (1:1000, Santa Cruz Biotechnology). Protein was visualized with enhanced chemiluminescence (ECL) (Amersham Biosciences). Total ataxin-1 protein levels (10 -15 g) were electrophoresed as described above; however, membranes were probed for 1 h with anti-ataxin-1 antibody (11750, 1:2500) at room temperature followed by detection with Western lighting chemiluminescence (PerkinElmer Life Sciences).
SUMOylation of Ataxin-1 in Vivo-Sequence analyses of SUMO-1 target proteins led to identification of a loose consensus sequence for SUMOylation in which a target lysine is modified, namely ⌿-K-X-D/E, where ⌿ is a hydrophobic residue and D/E represent an acidic residue (35). The examination of the sequence of ataxin-1 revealed seventeen possible SUMOylation sites (Fig. 3A and Table I). The putative SUMO-1 consensus sequences are depicted in relation to previously characterized regions within ataxin-1 (Fig. 3A). Twelve of the 17 sites are clustered near the C terminus of ataxin-1. Intriguingly one of the predicted SUMOylation sites in the N terminus is at the start of the polyglutamine stretch. Within the C terminus of ataxin-1, there is a region with high homology to the HMG-box transcription factor HBP1 (heterochromatin-binding protein 1) and Sp100-HMG that is designated the AXH domain (ataxin HMG domain) (Fig. 3A) (36). This domain in HBP1 and Sp100 is important for chromatin remodeling and also functions as a transcriptional repressor domain (29,37). In ataxin-1, this domain has been shown to be involved in RNA binding (38) and self-association (20,39), interaction with the nucleolar protein p80 coilin (40), and the ubiquitin protease, USP7 (41).
SUMO-1 has a calculated mass of 11.5 kDa and migrates in SDS-PAGE at 17-22 kDa (6, 10). SUMO-1-conjugated proteins can be identified in SDS-containing cell lysates (9, 10, 33), presumably because of inhibition of rampant SUMO hydrolases and the necessity to solubilize subnuclear structures such as the nuclear matrix (3,9,33). Ataxin-1 resides in the nuclear matrix and interacts with chromatin (21,42). SUMO modification occurs only on a small fraction of total substrate, and as a result, SUMO conjugates can be difficult to detect in vivo (16). To enhance the ability of detecting SUMOylated ataxin-1, HA-tagged SUMO-1 and ataxin-1 were overexpressed in COS-1 cells. 48 h post-transfection, the cells were lysed in the presence of SDS to denature proteins followed by immunoprecipitations with anti-ataxin-1 antibody (11NQ). As shown in Fig. 3B (lane 1, asterisks), immunoblotting with anti-SUMO-1 antibody revealed a ladder of bands consistent with covalent modification of ataxin-1[30Q] by SUMO-1. As seen with many multi-ubiquitylated proteins, higher molecular weight species formed a smear near the top of the gel (Fig. 3B, lanes 1-3). There were no SUMO-1-immunoreactive bands detected from cells transfected with empty vector HA (Fig. 4A, lane 1). Immunoblotting with monoclonal HA antibody resulted in the same series of bands (data not shown), consistent with the attachment of SUMO-1 to ataxin-1.
Previous studies have reported that the dimerization of a substrate can be important for its SUMOylation (4,8). Ataxin-1 contains a functional self-association domain (⌬77) (Fig. 3A) (20,39). Deletion of this region results in a protein that does not form inclusion bodies and does not alter PML distribution but still retains the ability to bind the nuclear matrix (20). Consistent with oligomerization having a role in SUMOylation, ataxin-1[82Q]-⌬77 showed a reduction in SUMOylation (Fig.  4A, lane 5 versus 3).
As a first step toward characterizing the biological functions of SUMOylation of ataxin-1, a series of experiments were performed to identify SUMO-1-modified lysine residues in ataxin-1. Thus, a number of lysine to arginine mutations were made in the context of full-length wild-type ataxin-1[30Q] and their effect on SUMOylation was determined ( Fig. 4A and Table I). SUMOylation of each ataxin-1 variant was normalized to wild-type ataxin-1[30Q] (Fig. 4B). The mutation of Lys 421 and Lys 692 to arginine had no effect on SUMOylation. In contrast, the mutation of Lys 16 abrogated ataxin-1 modification by SUMO-1 to the greatest extent. The mutation of Lys 194 , Lys 610 , Lys 697 , and Lys 746 to arginine all decreased the level of SUMOylation. Table I compares the level of SUMOylation and percentage homology to the SUMO-1 consensus sequence. Consistent with the multiple bands of SUMOylated ataxin-1 seen in Fig. 3B, these data indicate there are 5-7 possible sites of SUMO-1 SUMOylation in ataxin-1.
To confirm that SUMOylation of Lys 16 , Lys 194 , Lys 610 , Lys 697 , and Lys 746 represents the major SUMOylation sites on ataxin-1, constructs were made with multiple lysine to arginine mutations (Fig. 5A). As shown in Fig. 5B, the mutants generated were expressed at similar levels. Moreover, treatment with proteasome inhibitor did not result in an increase in total levels of ataxin-1 or the ataxin-1 SUMOylation mutants (data not shown). Thus, ataxin-1 lacking these SUMOylation sites did not appear to be less stable. The mutation of all five SUMOylation sites did not completely abrogate ataxin-1 SUMOylation, suggesting the presence of more than five SUMOylation sites (Fig. 5B, lane 2 versus 4). As shown in Fig.  5B, lane 3, the NLS mutant, ataxin-1[30Q]-K772T, completely eliminated ataxin-1 SUMOylation.
Mutation of the SUMOylation sites in PML resulted in the complete loss of POD formation, whereas mutation of the SUMOylation site of Sp100 does not affect nuclear localization or POD targeting (4,43). To determine whether SUMOylation of ataxin-1 had an affect on ataxin-1 localization, the subcellular localization of ataxin-1 SUMO mutants was examined by immunofluorescence (Fig. 5C). There was a decrease in the ability of ataxin-1 to co-localize with SUMO-1 as the number of putative ataxin-1 SUMOylation sites were increasingly mutated (Fig. 5C, a-t). Similar to Sp100, the mutation of ataxin-1 SUMOylation sites did not affect ataxin-1 nuclear import or inclusion formation (Fig. 5C, a-t). DISCUSSION Ataxin-1 overexpressed in transfected cells was found to be SUMOylated at multiple lysine residues. SUMOylation of ataxin-1 was dependent on its nuclear localization, phosphorylation at Ser 776 , self-association region, and polyglutamine length, all of which have a role in the subcellular distribution of ataxin-1. Thus, an important factor controlling ataxin-1 SUMOylation seems to be its subcellular distribution. Besides PML (43) and GRIP1 (44), which have three SUMOylation sites, the majority of substrates reported to date are SUMOylated at a single lysine residue. Consequently, SUMOylated targets can be divided into two groups, those that are conjugated to one SUMO moiety and those that are SUMOylated at multiple residues. Ataxin-1 is SUMOylated on at least five residues and is therefore a member of the latter group.
Ataxin-1 with a mutated NLS showed a dramatic decrease in its ability to be SUMOylated. However, ataxin-1 nuclear localization and nuclear inclusions were indistinguishable between wild-type ataxin-1 and the ataxin-1 SUMOylation site mutants. Thus, ataxin-1 SUMOylation does not appear necessary for nuclear import. However, ataxin-1 SUMOylation could contribute to nuclear import/export efficiency of ataxin-1. Nucleocytoplasmic trafficking and subcellular localization are important themes for SUMOylated target proteins. The SUMO E3 ligase, RanBP2, and the SUMO isopeptidase Senp2 localize to the nuclear pore, indicating a role for SUMOylation in the dynamic import/export process (45). However, the question remains as follows. Does SUMOylation depend on nuclear import, or does import depend on SUMOylation (1)? Mutation of the NLS in PML and Sp100 disrupted their SUMOylation (4,43). Restoring nuclear import of Sp100 by fusing the large T-antigen NLS to Sp100 NLS mutants restored Sp100 SUMOylation (4). The same was true for PML but complicated by the presence of three SUMOylation sites on PML. In both cases, in vitro SUMOylation assays were used with NLS mutants and all of the constructs were SUMOylated, even though in vivo modification was only detected in the presence of an intact NLS (8,43,46). This finding suggests that SUMOylation does not always regulate specific subnuclear targeting but could regulate import/export efficiency. SUMOylation could also have a role in regulating the trafficking of ataxin-1 within the nucleus. Using fluorescence recovery after photobleaching, trafficking of wild-type ataxin-1[30Q] within the nucleus and between the nucleus and cytoplasm has been demonstrated (47,48). In contrast, once in the nucleus, mutant ataxin-1-[82] no longer traffics and does not shuttle between the nucleus and cytoplasm (47,48). The decrease in mutant ataxin-1-[82] nucleocytoplasmic shuttling could be due to the decreased SUMOylation of mutant ataxin-1-[82] compared with wild-type ataxin-1[30Q].
The role of PODs as either "nuclear closets" or "catalytic surfaces" remains controversial (27). Despite this controversy, an overall theme of PML and Sp100 is their involvement with events at the chromatin level (2). Specifically, Sp100 functions as a co-repressor when bound to DNA (29). Further, multiple isoforms of Sp100 have been described, including Sp100 with a C-terminal HMG (29). Sp100 interacts with HP1, a non-histone chromosomal protein (29), and SUMOylation of Sp100 increases interaction with HP1 (49). This is true for other SUMO targets. SUMOylation of PML increases interaction with Daxx, and SUMOylation of RanGAP increases interaction with RanBP2. Because ataxin-1 may also function as a co-repressor when tethered to DNA (42) and has a C-terminal AXH domain with sequence similarity to HMG proteins (36), it is reasonable to suggest that ataxin-1 has a significant function within PODs. Perhaps one such function could be regulating the SUMOylation of other POD components. Sp100 SUMOylation regulates its ability to interact with HP1 (49). An alteration in Sp100 SUMOylation could control the ability of Sp100 ability to interact with HP1. In this way ataxin-1 could be involved in trafficking within PODs.
SUMO substrates include proteins implicated in gene transcription, including both co-activators and co-repressors (50). Recent reports (42, 51, 52) suggest a defect in transcription underpins polyglutamine repeat diseases. Moreover, ataxin-1 was shown to interact with the transcriptional co-repressor SMRT (Silencing Mediator of Retinoid and Thyroid hormone receptors), histone deacetylase 3, and function as a transcriptional co-repressor when tethered to DNA (42). Thus, it is intriguing to speculate that functional roles of the multiple SUMOylation events on ataxin-1 are to: 1) facilitate/stabilize macromolecular complexes with promoter specific proteins and 2) regulate trafficking at gene specific promoters. As a result of a decrease in SUMOylation of mutant ataxin-1[82Q], the cooperative interactions between ataxin-1 and other proteins would be disrupted, leading to a decrease in transcription at target genes. and endogenous SUMO-1. Endogenous SUMO-1 staining was visualized using a monoclonal anti-SUMO-1 antibody followed by mouse anti-Cy3 (f, g, h, i, j, p, q, r, s, and t). The different forms of ataxin-1 are visualized using rabbit anti-ataxin-1 antibody 11750 followed by rabbit anti-Alexa 488 (a, b, c, d, e, k, l, m, n, and o).