Valosin-Containing Protein (VCP/p97) Is an Activator of Wild-Type Ataxin-3

Alterations in the ubiquitin-proteasome system (UPS) have been reported in several neurodegenerative disorders characterized by protein misfolding and aggregation, including the polylgutamine diseases. Machado-Joseph disease (MJD) or Spinocerebellar Ataxia type 3 is caused by a polyglutamine-encoding CAG expansion in the ATXN3 gene, which encodes a 42 kDa deubiquitinating enzyme (DUB), ataxin-3. We investigated ataxin-3 deubiquitinating activity and the functional relevance of ataxin-3 interactions with two proteins previously described to interact with ataxin-3, hHR23A and valosin-containing protein (VCP/p97). We confirmed ataxin-3 affinity for both hHR23A and VCP/p97. hHR23A and ataxin-3 were shown to co-localize in discrete nuclear foci, while VCP/p97 was primarily cytoplasmic. hHR23A and VCP/p97 recombinant proteins were added, separately or together, to normal and expanded ataxin-3 in in vitro deubiquitination assays to evaluate their influence on ataxin-3 activity. VCP/p97 was shown to be an activator specifically of wild-type ataxin-3, exhibiting no effect on expanded ataxin-3, In contrast, we observed no significant alterations in ataxin-3 enzyme kinetics or substrate preference in the presence of hHR23A alone or in combination with VCP. Based on our results we propose a model where ataxin-3 normally functions with its interactors to specify the cellular fate of ubiquitinated proteins.


Introduction
Ataxin-3 is a 42 kDa intracellular protein capable of binding and hydrolysing ubiquitin chains [1][2][3]. Ataxin-3 was first recognized as the protein implicated in Machado-Joseph disease (MJD), also known as spinocerebellar ataxia type 3 [4]. MJD is a polyglutamine disorder, one of a group of nine neurodegenerative diseases that share a common genetic cause: an expansion of a CAG trinucleotide repeat in the coding region of the respective disease genes [5][6][7]. In the case of MJD, the pathogenic expansion occurs in the ATXN3 gene, which encodes a deubiquitinating enzyme (DUB), ataxin-3. When expanded the CAG repeat encodes an abnormally long polyglutamine (polyQ) track near the C-terminus of ataxin-3 [8][9]. The polyQ expansion destabilizes ataxin-3 structure, increasing its propensity to misfold and form large intracellular aggregates [10][11]. Through its aggregation, expanded ataxin-3 sequesters house-keeping proteins that are essential for cell homeostasis inside intracellular aggregates [12][13][14]. Although these aggregates were initially speculated to cause neurodegeneration, more recent studies suggest that their formation represents a cellular defense mechanism against smaller toxic oligomers of the mutant protein [15][16].
The polyQ expansion alone can not account for all the specific characteristics and selective neuronal loss in MJD. The protein context in which the polyQ is expressed is a crucial determinant of the pathological mechanisms triggered by expanded ataxin-3 [16][17][18][19][20]. Many recent studies have focused on wild-type ataxin-3 function. Ataxin-3 has shown to be a DUB with an N-terminal catalytic ''Josephin'' domain with ubiquitin hydrolase activity, and two or three ubiquitin-interacting motifs (UIM) depending on the splice isoform, responsible for the interaction with ubiquitin chains [1][2]. Ataxin-3 possesses higher affinity for longer ubiquitin chains, containing at least four molecules of ubiquitin, and preferentially cleaves linkages between ubiquitin molecules established through lysine 63 (K63) [3]. These characteristics suggest that ataxin-3 is an ubiquitin chain editing enzyme.
As an enzyme, ataxin-3 is regulated at multiple levels in the cell. Mono-ubiquitination of ataxin-3, which is enhanced by proteotoxic stress, increases its ubiquitin hydrolase activity [21]. Ataxin-3's translocation between the nucleus and the cytosol is regulated by nuclear import and export signals in association with specific phosphorylation and dephosphorylation events [22][23]. The high degree of regulation sustains a central role for ataxin-3 in the protein quality control of the cell. Ataxin-3 has been implicated in ubiquitin-proteasome pathways [24], endoplasmic reticulum associated degradation (ERAD) [25] and the cytoprotective response to heat shock stress [26].
The study of ataxin-3 enzymatic activity, its nuclear-cytoplasm translocation, and the colocalization of other cellular proteins to intracellular aggregates formed by expanded ataxin-3 have spurred the search for ataxin-3 interactors. A growing number of interactors have been identified that are diverse in structure and function, reflecting the wide range of biological activities associated with ataxin-3 [27][28][29][30]. Some of these protein interac-tions occur independently of the polyQ expansion and likely relate to normal ataxin-3 function. In the current study, we have focused on two of these interactors, hHR23A and valosin-containing protein (VCP/p97). hHR23A and hHR23B, the human homologs of RAD23 yeast protein, are involved in DNA repair pathways and the delivery of ubiquitinated substrates to the proteasome for degradation [31]. hHR23A has two ubiquitin-associated (UBA) domains which recognize and bind ubiquitin motifs, and an ubiquitin-like domain which interacts with the globular catalytic domain of ataxin-3 [32]. VCP/p97 protein is a 97 kDa AAA ATPase involved in several cellular pathways, including the extraction of ubiquitinated proteins from the endoplasmic reticulum for degradation by ERAD [25]. VCP/p97 is an abundant cellular protein that has numerous interactions with other proteins [33]. An extensive study has uncovered the region in ataxin-3 crucial to the interaction with VCP/p97, which is situated after the second UIM, just prior to the polyQ domain [28].
In previous reports, ataxin-3 was shown to be a DUB with rather slow kinetics [3]. However, a post-translational modification of ataxin-3, ubiquitination, was highly effective at enhancing ubiquitin hydrolase activity of ataxin-3 [21]. Here, we sought to investigate whether VCP/p97 or hHR23A, both of which are involved in protein degradation pathways, also enhanced ataxin-3 activity through direct interactions. Modulation of ataxin-3 activity through protein-protein interactions might represent an additional level of regulation over its DUB activity and cellular roles. Moreover, a better understanding of how ataxin-3 is regulated may lead to the identification of pathological targets underlying neurodegeneration in MJD.

Immunoprecipitation
For the immunoprecipitation of endogenous proteins, COS-7 cells (ATCCH, UK) were lysed in FLAG lysis buffer (50 mM Tris, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, pH 7.4) supplemented with 1 mg/ml protease inhibitor cocktail (chymostatin, pepstatin A, leupeptin, and antipain). For the immunoprecipitation from in vitro samples, 100 nM of both recombinant proteins were incubated in deubiquitination buffer (50 mM HEPES, 0.5 mM EDTA, 0.1 mg/ml ovalbumine , 1 mM DTT, 1 mg/ml protease inhibitor cocktail, at 37u C for 5 hours. Cell lysates and in vitro samples were washed with Protein A Sepharose beads (GE Healthcare, UK) to reduce nonspecific binding to the beads, before the incubation with mouse monoclonal (1H9; 1:1000) or rabbit polyclonal (MJD; 1:1000) anti-ataxin-3 antibodies overnight, at 4uC with gentle agitation. Beads were then precipitated by centrifugation at 5006 g, 4uC for 5 min, washed six times with FLAG lysis buffer, and the protein was eluted with VCP/p97 Is an Activator of Wild-Type Ataxin-3 Laemmli buffer at room temperature. The eluted immunoprecipitates were collected in Spin-X centrifuge tube filter columns by centrifugation and stored at 280uC and resolved by SDS-PAGE followed by western blotting.

Pull-down experiments
Recombinant GST-VCP/p97 fusion protein or GST protein alone were expressed in BL-21 cells (GE Healthcare, UK) as described in the recombinant protein purification section. The cells were sonicated and centrifuged to discard cell debris. The lysates were poured into columns containing glutathione-Sepharose beads (GE Healthcare, UK) and incubated for 30 min on ice with occasional agitation. Beads were washed 6 times with FLAG lysis buffer before being added to COS-7 total extracts in FLAG Lysis buffer, which had been previously incubated with glutathione beads for 1 hour to reduce unspecific interactions with the beads. COS-7 total extracts were in contact with the beads for 2 hours, at 4uC in constant agitation. Beads were washed another 4 times with FLAG lysis buffer supplemented with protease inhibitors and resuspended in Laemmli buffer for 15 min, at room temperature. Beads were centrifuged a last time for 5 min, at 5006 g and the elutants were stored at 280uC.

Protease assays
100 nM human recombinant wild-type (Q22) or expanded (Q80) ataxin-3 was incubated with 250 nM K63 or K48 linked hexa-ubiquitin chains (Boston Biochem, USA) in the absence or presence of 100 nM recombinant VCP, 100 nM recombinant hHR23A, or both. The protease assay was performed in in vitro deubiquitination buffer, at 37uC for up to 20 hours. Samples from the reaction mixture were collected at 0, 2, 5 and 20 hours. Reactions were stopped with loading buffer at room temperature.

Statistical analysis
Data were expressed as mean 6 SEM of the number of experiments indicated in the figure legends. Comparisons between two conditions were performed by two-tailed unpaired t-test. Significance was defined at p,0.05.

VCP/p97 and hHR23A interact directly with ataxin-3
It was previously shown that ataxin-3 can bind and cleave ubiquitin chains [1][2][3]. Since protein-protein interactions are known to modulate enzymatic activity, we hypothesized that hHR23A and/or VCP/p97 could modify the ubiquitin hydrolase activity of ataxin-3 through direct interaction. To test this hypothesis we investigated the interaction between these three proteins in vitro and in cells.
Immunostaining for hHR23A and ataxin-3 in COS-7 cells revealed a strong spatial correlation ( Figure 1A). Endogenous hHR23A and ataxin-3 were both more prevalent in the nucleus. Within this organelle, hHR23A and ataxin-3 exhibited a punctate pattern with additional diffuse distribution throughout the nucleoplasm, excluding the nucleolus. Ataxin-3 and hHR23A were often present in the same nuclear puncta, although the proteins did not perfectly co-localized through the entire cell. The presence of both proteins in specific subnuclear regions may indicate the compartmentalization of nuclear areas dedicated to protein degradation. Immunoprecipitation of endogenous ataxin-3 from COS-7 cell extracts co-precipitated endogenous hHR23A, suggesting that these co-localized proteins also interact in the cell ( Figure 1B). As previously reported [27,35], immunoprecipitation of recombinant ataxin-3 co-precipitated recombinant hHR23A, demonstrating that hHR23A and ataxin-3 directly interact in vitro ( Figure 1C).
VCP/p97 is an abundant cellular protein that is homogeneous distributed throughout the cytosol of immunostained COS-7 cells, as shown in Figure 1D. Unlike ataxin-3 and hHR23A, VCP/p97 was not enriched in the nucleus. Nevertheless, recombinant GST-VCP/p97 fusion protein bound to GSH-beads was able to pulldown endogenous ataxin-3 from COS-7 cell extracts ( Figure 1E), indicating a significant interaction between the two proteins. Recombinant human wild-type (Q22) ataxin-3 also co-immunoprecipitated recombinant VCP/p97 in vitro ( Figure 1F), supporting a direct interaction between ataxin-3 and VCP/p97.

VCP/p97 enhances wild-type ataxin-3 activity in vitro
Ataxin-3 interacts with VCP/p97 through a region near the polyQ domain and far from the Josephin domain [28]. The interaction with VCP/p97 may position ataxin-3 in close proximity to polyubiquitinated protein substrates in the cell, as implied by its regulatory effect of the ERAD pathway [25]. To determine whether VCP/p97 can alter ataxin-3 activity, we performed in vitro deubiquitination experiments in which recombinant human VCP/p97 was added to the reaction. Ataxin-3 enzymatic activity was assessed through in vitro deubiquitination assays as in Burnett et al. (2003) and Winborn et al. (2008). Ataxin-3 acts on longer ubiquitin chains and exhibits preference for ubiquitin chains with ubiquitin molecules linked through K63 [3]. Thus, an ubiquitin chain containing six ubiquitin molecules linked through K63 was used as an optimal substrate to evaluate the influence of hHR23A in ataxin-3 deubiquitinating activity. Ataxin-3 (Q22) incubated with K63 linked hexa-ubiquitin chains generated lower molecular weight reaction products, resulting from the cleavage of ubiquitin linkages (Figure 2A). When VCP/ p97 was added to the reaction employing K63-linked chain, an increase in the accumulation of reaction products was observed (Figures 2A, 2B). In contrast, VCP/p97 did not enhance significantly the cleavage of K48-linked chains by ataxin-3 ( Figures 2C, 2D). Increased cleavage of K63-linked chains in the presence of VCP was not due to ataxin-3 stabilization, because there were no differences in ataxin-3 levels through the reaction course, in the presence or absence of VCP (Figure 2A). To rule out the possibility that the enhanced K63-linked chain cleavage reflected a direct activity of VCP/p97 on ubiquitin chains, we incubated catalytically inactive ataxin-3 mutant (C14A) with VCP/p97 under the same reaction conditions. As shown in Figure 2E, no reaction products were detected, demonstrating that VCP/p97 Is an Activator of Wild-Type Ataxin-3

VCP/p97 does not enhance ubiquitin hydrolase activity of expanded ataxin-3
Previous in vitro studies of ataxin-3 activity have not identified any differences in substrate preference or reaction kinetics between wild-type (Q22) and expanded (Q80) ataxin-3 [3]. Nevertheless, we sought to investigate whether VCP/p97 similarly stimulated the protease activity of expanded ataxin-3. To test this, we expressed and purified human recombinant ataxin-3 with 80 glutamines (Q80). We assayed recombinant human expanded ataxin-3 (Q80) in the same reaction conditions used for nonpathogenic ataxin-3, in the absence or presence of VCP/ p97. Unexpectedly, VCP/p97 did not alter expanded ataxin-3 activity towards K63-linked ubiquitin chains (Figures 3A, 3B). This result validates the direct VCP stimulation of wild-type ataxin-3. An indirect effect of VCP/p97 over ataxin-3 activity (e.g. increased accessibility to the ubiquitin chains) would produce a similar effect in both wild-type and expanded proteins.

Blockade of VCP/p97-ataxin-3 protein interaction abolishes VCP/p97 stimulation of ataxin-3 protease activity
The region of ataxin-3 that mediates the interaction with VCP/ p97 is located between the first two ubiquitin-interacting motifs (UIM)s, near the polyQ track [28]. A sequence of four amino acids in ataxin-3 is crucial for this interaction, the basic amino acids 282 RKRR (arginine/lysine-rich motif), which also constitute part of a nuclear localization signal [22,[36][37]. A four amino acid substitution ( 282 RKRR to HNHH) has been shown to block the interaction between ataxin-3 and VCP/p97 in vitro. We substituted these four amino acids in wild-type (Q22) and expanded (Q80) ataxin-3 constructs in order to block the interaction between ataxin-3 and VCP/p97. Recombinant ataxin-3 mutants were incubated with human recombinant VCP/p97 under the same experimental conditions of in vitro deubiquitination assays, followed by immunoprecipitation of ataxin-3. As expected, the 282 RKRR -HNHH amino acids substitution effectively reduced the interaction of ataxin-3 mutants with VCP/p97 in vitro ( Figure 4A). Whereas VCP/p97 was co-immunoprecipitated with wild-type (Q22) and expanded (Q80) ataxin-3, almost no VCP/p97 coimmunoprecipitated with the wild-type (Q22) and expanded (Q80) ( 282 RKRR-HNHH ataxin-3 variants). We then tested the ubiquitin hydrolase activity of the 282 RKRR-HNHH ataxin-3 mutants in vitro. Wild-type (Q22) and expanded (Q80) 282 RKRR-HNHH mutants were both catalytically active, however the protease activity of these mutants was slightly reduced compared to normal Q22 and Q80 ataxin-3 ( Figure S1), registering a decrease in the total amount and in the rate of production of low molecular weight ubiquitin chains (Figures 4B, 4D). Moreover, adding VCP/p97 to the reaction system did not increase the ubiquitin hydrolase activity of either 282 RKRR-HNHH mutant (Figures 4C, 4E). These results demonstrate that VCP/p97 stimulation of wild-type (Q22) ataxin-3 protease activity is mediated through a direct interaction with the enzyme.
hHR23A does not change the kinetics or the substrate preference of ataxin-3 hHR23A can simultaneously bind ubiquitin chains and ataxin-3 [27,32,38]. In our co-localization studies, hHR23A and ataxin-3 revealed a high homology in their nuclear distributions. For these reasons, hHR23A emerged as a strong candidate for a direct modulator of the ubiquitin hydrolase activity of ataxin-3. To test this model, we quantified the enzymatic activity of recombinant human wild-type ataxin-3 (Q22) in the absence or presence of recombinant hHR23A ( Figure 5). As reported previously, ataxin-3 (Q22) deubiquitinates K63-linked ubiquitin chains and promotes the appearance of smaller ubiquitin chains over time. No enhancement in the rate of accumulation of reaction products was observed in the presence of hHR23A (Figures 5A, 5B); rather, there was a trend towards decreased accumulation of reaction products. hHR23A did not alter the substrate preference of ataxin-3 either, as ataxin-3 continued to cleave K63-linked chains more robustly than K48-linked ubiquitin chains ( Figures 5C, 5D). Hence, hHR23A does not alter ataxin-3 activity, at least in these in vitro assays. Our data is in accordance with a recent report also showing no alteration in the ability of ataxin-3 to cleave ubiquitin chains in the presence of hHR23A [39]. As expected, catalytically inactive ataxin-3 mutant (C14A) failed to generate reaction products ( Figure 5E). hHR23A and VCP/p97 are both involved in protein quality control through their connections with protein degradation and the ubiquitin-proteasome system [31][32][33]. Recent studies have established ataxin-3 as an important contributor to ubiquitindependent protein quality control [25][26]. Caenorhabditis elegans homologues for hHR23A, VCP/p97 and ataxin-3 have been previously found to assemble in a trimolecular complex [40]. To assess if these proteins could establish a similar trimolecular complex in our cellular system, endogenous hHR23A was immunoprecipitated from COS-7 cells and co-immunoprecipitation of endogenous VCP/p97 and endogenous ataxin-3 was assessed through western blotting. Accordingly, both endogenous VCP/p97 and endogenous ataxin-3 were co-immunoprecipitated with hHR23A ( Figure 6A). Since ataxin-3, hHR23A and VCP/ p97 are linked to the same cellular pathways, we assessed whether ataxin-3 enzymatic activity is altered when exposed to two protein interactions with different effects regarding ataxin-3 stimulation. In vitro deubiquitination assays were performed with wild-type (Q22) ataxin-3 in the absence or presence of hHR23A and VCP/p97 simultaneously, using K63-linked hexa-ubiquitin chains in the reaction (Figures 6B, 6C). In both conditions wild-type ataxin-3 was able to cleave ubiquitin chains. However, when hHR23A and VCP/p97 were both added to the reaction solution, we did not observe an increase in wild-type ataxin-3 activity (Figures 6B, 6C), as we have previously observed with VCP/p97 alone. Interestingly, ataxin-3 in conjunction with hHR23A and VCP/p97 had similar activity to ataxin-3 alone, displaying neither the increase in activity caused by VCP/p97, nor the trend to reduced activity observed with hHR23A. Taken together, these results suggest that,
Ataxin-3 binds and cleaves poly-ubiquitin chains [1][2]. With respect to ubiquitin hydrolase activity, ataxin-3 displays clear substrate preference: ataxin-3 preferentially cleaves K63-linked over K48-linked ubiquitin chains, and longer over shorter chains [3]. Even with its preferred substrates, the reaction kinetics of ataxin-3 in vitro are slower than those of other well-studied deubiquitinating enzymes [39]. Recent studies have shown that mono-ubiquitination of ataxin-3 itself is a strong activator of ataxin-3 activity [21], but in cells under basal conditions the amount of mono-ubiquitinated ataxin-3 is minor. Therefore, we investigated whether two well known ataxin-3 interactors could modulate ataxin-3 protease activity. VCP/p97 is an abundant protein, comprising approximately one percent of total protein in some cells [41][42]. Though it is distributed throughout the nucleus and cytoplasm of COS-7 cells, VCP/p97 is more abundant in the cytosol, in contrast with ataxin-3, which is predominantly nuclear in these cells. Despite this difference in localization, we established both a cellular and a direct in vitro interaction between ataxin-3 and VCP/p97 in our experiments, confirming previously published results [25,28,43]. Importantly, VCP/p97 proved to be an activator of wild-type ataxin-3 ubiquitin hydrolase activity in vitro, nearly doubling the accumulation of lower molecular weight ubiquitin chain reaction products. Activation has also been reported for other deubiquitinating enzymes [44][45] and most recently, a synergistic cooperation between ataxin-3 and CDC-48 (C. elegans orthologue of VCP/p97) was found to have a role in ageing regulation and ubiquitin-mediated proteolysis in C. elegans [46]. VCP/p97 did not appear to alter ataxin-3 substrate preference, as K63-linked hexaubiquitin chains were still cleaved more extensively than K48linked hexa-ubiquitin chains. VCP/p97 also had no effect on the activity of mutant (i.e. expanded) ataxin-3. The region of ataxin-3 essential for VCP/p97 interaction is situated just amino-terminal to the polyQ region, so it is conceivable that polyQ repeat length may affect the interaction between the two proteins. Previous studies have reported an increased interaction between VCP/p97 and polyQ-expanded ataxin-3 [28] and our own in vitro co- immunoprecipitation studies are consistent with this observation. Even so, this interaction is apparently not sufficient to stimulate the activity of expanded ataxin-3. Although the changes in ataxin-3 structure induced by the expanded polyQ domain seem to promote a stronger association with VCP/p97, the same conformational alterations might abrogate any allosteric effect in favor of catalytic activation. It is important to recall that VCP/p97 normally is organized as a ring-like hexamer rather than as a monomer [47]. A single VCP/p97 hexamer might engage differentially with normal and expanded ataxin-3 polypeptides in a manner that permits activation only of normal ataxin-3 molecules. Additional conformational and structural studies of both proteins in this interaction will be needed to define the mechanism.
Our results confirm that the 282 RKRR sequence of ataxin-3 mediates binding to VCP/p97 [28] and establish that this region is essential for the activation of ataxin-3 activity by VCP/p97. The 282 RKRR-HNHH substitution disrupts the interaction of VCP/ p97 with wild-type and expanded ataxin-3, and blocks VCP/p97activation of ataxin-3 DUB activity. We thus conclude that VCP/ p97-mediated enhancement of ataxin-3 DUB activity is dependent on direct VCP/p97 interaction with ataxin-3. A competent protein-protein interaction with VCP is mandatory to induce a significant increase in wild-type ataxin-3 activity. Therefore, a more restrained protein-protein interaction might also justify the ineffectiveness of VCP/p97 in increasing expanded ataxin-3 ubiquitin hydrolase activity.
Ataxin-3 resides in the nucleus, cytosol and possibly mitochondria [36,[48][49]. In some types of cells, ataxin-3 is predominantly nuclear while in other cells it is largely cytoplasmic. In COS-7 cells, we observed a primarily nuclear distribution of ataxin-3, similar to what we have observed in human fibroblasts and in mouse cerebellar granule cells (Laço et al., unpublished data). Ataxin-3 exhibits the capacity to shuttle in and out of the nucleus, presumably responding to specific stimuli [22,26,50]. In the nuclear compartment, ataxin-3 exhibits a punctate pattern with a superimposed, lighter homogeneous distribution throughout the nucleoplasm that spares the nucleolus. This nuclear pattern is very similar to that of hHR23A; indeed, we observed a high degree of co-localization of these two proteins, to the subnuclear foci. Interestingly, the 20S and 19S subunits of the proteasome also exhibit a punctate nuclear distribution in COS-7 cells and several other tested cell types [51][52]. These proteasome-enriched nuclear foci are in close association with the ataxin-3 containingnuclear foci (not shown). The discrete distribution of ataxin-3 and hHR23A into specific nuclear regions, might represent specialized nuclear compartments dedicated to protein quality control and protein degradation.
Interestingly, and despite the tight relation in nuclear distribution and a clear interaction in cellular assays and in vitro between ataxin-3 and hHR23A, we found that hHR23A does not influence the ubiquitin hydrolase activity of ataxin-3. hHR23A did not increase ataxin-3 DUB activity or change its substrate preference, in accordance with recent data by Nicastro and colleagues (2010). On the contrary, ataxin-3 actually displayed a non-significant trend towards decreased activity when hHR23A was added to the reaction system. Both ataxin-3 and hHR23A are able to bind ubiquitin chains [53], therefore the decrease in ataxin-3 activity might reflect competition for the ubiquitin chains by the two proteins. Ataxin-3 would be left with less free non-hHR23A bound ubiquitin chains to work on, reducing the quantity of end products of deubiquitination. hHR23A was first proposed to be part of the DNA repair machinery [54]; while other studies have shown an influence of ataxin-3 in nuclear gene expression [29,[55][56]. The close association to proteasome-enriched nuclear areas and the influence in expression of specific factors and nuclear proteins levels may link various physiological roles of ataxin-3 and hHR23A to protein degradation.
Recent studies have expanded the number of proteins known to interact with ataxin-3 [29][30][57][58][59][60]. In a cellular context, ataxin-3 likely does not function alone [25,61], since protein-protein interactions are established continuously and dynamically in the cell. New protein-protein interactions involving ataxin-3 may be established along with its deubiquitination activity over proteinsubstrates. Indeed, distinct protein interactions may differently modulate ataxin-3 ubiquitin hydrolase activity, as observed in our study regarding the role of hHR23A and VCP/p97. We exposed ataxin-3 to both hHR23A and VCP/p97 in an attempt to better understand how ataxin-3 activity may be regulated in the cell. Interestingly, the presence of hHR23A negated the activation induced by VCP/p97. Competition between hHR23A and VCP/ p97 for a common binding site on ataxin-3 seems unlikely since the described areas of interaction for these two proteins are in different regions of ataxin-3 [28,38]. Nevertheless, interaction with hHR23A could change the affinity of ataxin-3 for VCP/p97 and thus impair VCP/p97-induced activation of ataxin-3. Also, competition for hexa-ubiquitin chains between hHR23A and ataxin-3 might reduce substrate availability for ataxin-3 when hHR23A and VCP/p97 are added, thereby reducing ataxin-3 DUB activity. On the other hand, ataxin-3 is probably deubiquitinating faster the available ubiquitin chains due to VCP/p97 activation. VCP stimulation of ataxin-3 activity might compensate the lack of substrate and the slight inhibitory effect of hHR23A, originating similar total amounts of reaction products. These interesting data demonstrate that ataxin-3 activity in the cell might be highly regulated and the repercussions of ataxin-3 enzymatic action might derive from the equilibrium of all the various inputs coming from the different interactions ataxin-3 is establishing at a specific moment.
Taken together with previously published studies, our results favor the view that ataxin-3 engages in multiple protein-protein interactions that influence its DUB activity within the cell [25,30,40,59]. Such interactions might influence the selection of ubiquitinated proteins on which ataxin-3 would act. Ataxin-3 is an interesting DUB in part because it can discriminate between different ubiquitin chain conformations [3]. Resembling the phosphorylation and dephosphorylation procedures, different ubiquitin signals made of different ubiquitin chains with different arrangements and topologies will almost certainly trigger different cellular outcomes. The selective activity of ataxin-3 over some ubiquitin linkages may alter the ubiquitin signal on the ubiquitinated substrates and consequently determine their future cellular role. In this way, ataxin-3 may represent an important checkpoint on the future fate of ubiquitinated proteins in the cell and its interactions may play a major role in modulating this event [46].
Our results and previous published data on VCP/p97 and hHR23A lead us to propose a model in which these proteins interact with ataxin-3 at different points of the processing of ubiquitinated protein-substrates (Figure 7). VCP/p97 and ataxin-3 have been described to work together in the retrotranslocation of ubiquitinated substrates from the endoplasmic reticulum in the ERAD [25]. Recently, it has been shown that ataxin-3 deubiquitinates C-terminus of Hsp70-interacting protein (CHIP) and parkin [3,59]. VCP/p97 may bind to ubiquitinated CHIP, parkin or other protein-substrates, present them to ataxin-3 and activate the enzyme. Ataxin-3 would cleave preferentially K63 linkages out of the ubiquitin chains, releasing a K48-linkage  ubiquitin chain. The interaction with VCP would facilitate the ataxin-3 editing of ubiquitin chains on its substrates. hHR23A would easily recognize the K48-linkage enriched ubiquitin chain left attached to the protein-substrates, since hHR23A has higher affinity for K48 linkages [32,62]. Finally, hHR23A would take the ubiquitinated substrate out of the catalytic pocket of ataxin-3 and would transfer it to the proteasome or other alternative cellular destination, as depicted in Figure 7.
Taking this model into account, the absence of stimulation of expanded ataxin-3 by VCP/p97 may have strong implications in MJD pathology. We may hypothesize that without the activation of VCP/p97, expanded ataxin-3 would work less efficiently as a DUB, leading to an accumulation of ubiquitinated proteins in the cell. Such a reduction in ubiquitin chain processing by expanded ataxin-3 not only would favor the accumulation, and possibly aggregation, of ubiquitinated proteins in the cell, but also might perturb the delivery of ubiquitinated substrates to their appropriate cellular destination. Thus, the disturbance of normal cellular pathways of ubiquitin signaling would add up to other cellular stresses promoted by the expression of a pathogenic polyQ fragment. The presence of a polyQ expansion in ataxin-3 may impair protein homeostasis through direct inhibition of the proteasome [63][64] and thus promote the formation of expanded ataxin-3 aggregates. In addition, the polyQ expansion in ataxin-3 protein may compromise its ubiquitin chain editing activity over ubiquitinated substrates. The buildup of protein malfunction and misleading signals in the cell together with expanded ataxin-3's intrinsic propensity to aggregate, may initiate a chronic cascade of events, culminating in neuronal dysfunction and cell death. Figure S1 ( 282 RKRR-HNHH) ataxin-3 VCP-binding mutants exhibit a small reduction in their protease activity. Ubiquitin protease assay for (A) wild-type (Q22) normal and ( 282 RKRR-HNHH) ataxin-3 mutant and (B) expanded (Q80) and ( 282 RKRR-HNHH) ataxin-3 mutant, using K63-linked hexaubiquitin chains as substrate. Samples were collected at times 0, 2, 5 and 20 hours and analyzed by western blotting using an antiubiquitin antibody. (TIF)