The Machado–Joseph disease-associated mutant form of ataxin-3 regulates parkin ubiquitination and stability

Abstract

Machado–Joseph disease (MJD), the most common dominantly inherited ataxia worldwide, is caused by a polyglutamine (polyQ) expansion in the deubiquitinating (DUB) enzyme ataxin-3. Interestingly, MJD can present clinically with features of Parkinsonism. In this study, we identify parkin, an E3 ubiquitin-ligase responsible for a common familial form of Parkinson's disease, as a novel ataxin-3 binding partner. The interaction between ataxin-3 and parkin is direct, involves multiple domains and is greatly enhanced by parkin self-ubiquitination. Moreover, ataxin-3 deubiquitinates parkin directly in vitro and in cells. Compared with wild-type ataxin-3, MJD-linked polyQ-expanded mutant ataxin-3 is more active, possibly owing to its greater efficiency at DUB K27- and K29-linked Ub conjugates on parkin. Remarkably, mutant but not wild-type ataxin-3 promotes the clearance of parkin via the autophagy pathway. The finding is consistent with the reduction in parkin levels observed in the brains of transgenic mice over-expressing polyQ-expanded but not wild-type ataxin-3, raising the intriguing possibility that increased turnover of parkin may contribute to the pathogenesis of MJD and help explain some of its parkinsonian features.

INTRODUCTION

Machado–Joseph disease (MJD) or spinocerebellar ataxia type 3 (SCA3), the most common dominantly inherited ataxia worldwide (1), is caused by an unstable CAG trinucleotide repeat expansion in the SCA3/MJD1 gene. The CAG expansion leads to elongation of the polyglutamine (polyQ) tract within the ataxin-3 protein (2), a feature shared with other polyQ diseases such as Huntington's disease and several SCAs (3). The polyQ repeat expansion is believed to confer a toxic gain-of-function to the various polyQ-encoding proteins, possibly promoting their misfolding and aggregation, but the precise mechanism of disease pathogenesis remains unknown. Although these shared features suggest similarities in disease mechanisms, many differences exist between the distinct disorders, both at a clinical and at a neuropathological level, which cannot be fully accounted for by the polyQ expansion alone. Indeed, as these proteins differ considerably outside of the polyQ region, the distinctive features of these proteins and their normal function may help account for disease specificity.

Two structural features of ataxin-3 implicate it in the ubiquitin (Ub) system. Ataxin-3 contains three Ub-interacting motifs (UIMs), flanking the polyQ region, which are involved in binding Ub and Ub-like domains (4,5), as well as an N-terminal Josephin domain, which defines a distinct class of deubiquitinating (DUB) enzymes (6–9). The Ub system is critical for a number of vital cellular functions, including protein degradation, trafficking and signaling. Moreover, defects in the Ub system have been implicated in several neurodegenerative diseases including SCAs, Parkinson's disease (PD) and Huntington's disease (10). Ubiquitination involves the attachment of Ub onto substrate proteins, requiring the coordinated activity of an E1 Ub-activating enzyme, an E2 Ub-conjugating enzyme and an E3 Ub-ligase (11). One of the seven lysine (K) residues within the attached Ub can, in turn, be modified by another Ub molecule resulting in the formation of a poly-Ub chain on substrates. Chains linked via K48 residues in Ub target substrates for degradation to the 26S proteasome, whereas the addition of a single Ub or Ub chains linked via other linkages, such as K27, K29 and K63, can target proteins into different signaling pathways and for degradation via the lysosomal and autophagy pathways (12–16). Regulation also occurs via deubiquitination, mediated by DUBs, which can remove Ub from substrates, thereby regulating their stability (17).

Interestingly, MJD can present with clinical and neuropathological features of PD and the affected brain regions can overlap (18). Mutations in the parkin gene cause one of the most common familial forms of PD (19,20) and the parkin protein also functions in the Ub system as an E3 Ub-ligase (21). Parkin is believed to play a neuroprotective role in both inherited and sporadic PD (22), and the inactivation of parkin, either genetically via PD-linked mutations or post-translationally (23–25), can interfere with its normal function. At the N-terminus, parkin contains a Ub-like (Ubl) domain, which interacts with Ub-binding domains such as the UIMs in Eps15 (26) and the SH3 domain in endophilin-A (27). At the C-terminus, parkin contains three RING motifs (RING0, RING1 and RING2) (28), with an in-between-RING (IR) domain separating the RING1 and RING2 domains, thereby defining parkin as a RING-type E3 Ub-ligases. Parkin has been demonstrated to mediate the assembly of K48-linked, K63-linked and K27-linked Ub chains as well as the attachment of monoubiquitin, supporting a role for parkin in both proteasome-dependent and -independent ubiquitination pathways (14,29–32). A number of DUBs have been shown to pair with specific E3 Ub-ligases. Typically, the DUB antagonizes ubiquitination of the cognate E3, thereby protecting it from proteasomal degradation (33–35). However, to date, no such DUB partner has been identified for parkin. Considering that MJD and PD exhibit overlapping clinical and neuropathological features, that parkin and ataxin-3 have opposing E3/DUB activities in the Ub system and that they encode UIM and Ubl domains with the potential to bind each other, we tested whether ataxin-3 interacts with parkin.

In this study, we identify a functional interaction between ataxin-3 and parkin. Both wild-type and polyQ-expanded mutant ataxin-3 can deubiquitinate parkin, regardless of the lysine residue used to assemble poly-Ub chains. However, polyQ-expanded mutant ataxin-3 deubiquitinates parkin more efficiently than wild-type ataxin-3, possibly owing to its enhanced activity toward K27- and K29-linked Ub on parkin. Remarkably, mutant but not wild-type ataxin-3 promotes the clearance of parkin via the autophagy pathway. The finding is consistent with the reduction in parkin levels observed in the brains of MJD transgenic mice. Thus, our findings implicate increased parkin turnover via autophagy in the pathogenesis of MJD, possibly explaining some of the parkinsonian features observed in the disease.

RESULTS

Ataxin-3 interacts directly with parkin

We showed previously that the parkin Ubl (Ubl^parkin) binds the UIM protein Eps15 (26). Considering the clinical and neuropathological overlap between MJD and PD and the fact that both parkin and ataxin-3 are involved in the Ub pathway, we speculated that the Ubl^parkin could also bind the UIMs in ataxin-3 (Fig. 1A). We found that myc-ataxin-3 (14Q) containing three functional UIMs, transfected in HEK293 cells, binds wild-type GST-Ubl^parkin (Fig. 1B), but not an R42P PD-linked mutant Ubl (GST-Ubl^R42P). A conserved serine residue within the UIMs is required for Ub and Ubl binding (36). Mutating the serine to alanine in all three ataxin-3 UIMs abolished the interaction with the Ubl^parkin. UIM2 and 3 are separated by a polyQ tract, which is expanded in MJD, suggesting that polyQ length could modulate parkin Ubl binding. However, we found that the binding of ataxin-3 to the Ubl^parkin was unaffected by an expansion in the polyQ tract (Fig. 1C). Moreover, co-immunoprecipitations (co-IPs) from transfected HEK293 cells show that full-length parkin interacts with both wild-type (14Q) and expanded (82Q) ataxin-3 in cells (Fig. 1D). Importantly, endogenous ataxin-3 from mouse brain could bind GST-parkin (Fig. 1E), and this interaction was demonstrated to be direct with bacterially expressed recombinant His-ataxin-3 binding GST-parkin in vitro (Fig. 2A). Interestingly, we found that parkin lacking the N-terminal Ubl (GST-ΔUbl^parkin) could also bind His-ataxin-3, suggesting the existence of a second interaction site (Fig. 2A). Further co-IPs and in vitro binding assays mapped this second site to the parkin IBR-RING2 domains and the ataxin-3 Josephin domain (Fig. 2B and C). Taken together, the data indicate a bimodal interaction between parkin and ataxin-3, involving Ubl:UIM and IBR-RING:Josephin binding.

Figure 1.

The Ubl^parkin binds ataxin-3 via the UIMs. (A) Schematic representation of the constructs used in this study. The ataxin-3 constructs are N-terminally tagged with Myc or His epitopes, whereas the parkin constructs are N-terminally tagged with GST, HA or FLAG epitopes. (B) The parkin Ubl interacts with ataxin-3 via its UIMs. GST, GST-Ubl^parkin and GST-Ubl^R42P were immobilized on glutathione–Sepharose beads and incubated with lysates from HEK293T cells expressing either myc-ataxin-3 (14Q) or myc-ataxin-3^{TRIPLE UIM} (S236A, S256A, S347A) and analyzed by ataxin-3 immunoblotting. (C) The Ubl^parkin binds both wild-type and MJD-associated polyQ-expanded ataxin-3. GST-Ubl^parkin pulldowns from HEK293T cell lysates expressing either wild-type (14Q) or polyQ-expanded (82Q) myc-ataxin-3 immunoblotted for ataxin-3. (D) Parkin interacts with both wild-type (14Q) and expanded (82Q) ataxin-3 in cells. HEK293T cells were transfected with the indicated combinations of myc-ataxin-3 (14Q) and HA-Parkin. Lysates were immunoprecipitated with anti-myc and immunoblotted for parkin and ataxin-3. (E) GST-parkin can interact with endogenous ataxin-3. GST and GST-parkin were incubated with rat brain lysates and stained for Ponceau S and immunoblotted for ataxin-3.

Figure 2.

Characterization of a second interaction site between ataxin-3 and parkin. (A) Ataxin-3 can interact with parkin lacking the Ubl domain. GST, GST-parkin, GST-ΔUbl^parkin and GST-Ubl^parkin were incubated with His-ataxin-3 (14Q) and immunoblotted for anti-His. (B) The ataxin-3 Josephin domain interacts directly with the parkin IR-RING2 domain. GST-Ubl^parkin, GST-IR^parkin, GST-RIR^parkin and GST-parkin were incubated with His-Josephin and immunoblotted for anti-His. (C) The ataxin-3 Josephin domain can interact with ΔUbl^parkin in cells. HEK293T cells were transiently transfected with myc-Josephin and either full-length HA-parkin or HA-ΔUbl^parkin. Lysates were immunoprecipitated with anti-myc and immunoblotted for parkin and myc. Asterisk denotes Ig heavy chain.

Parkin self-ubiquitination enhances ataxin-3 binding

Parkin, like other E3s, can ubiquitinate itself (26,30). Immunoblotting for Ub or parkin or staining with Coomassie shows that recombinant GST-parkin robustly ubiquitinates itself in vitro when incubated with an E1, E2, Ub and ATP (Fig. 3A). Interestingly, parkin self-ubiquitination markedly increased its binding to ataxin-3 (Fig. 3B). Parkin can conjugate both mono-Ub and poly-Ub chains onto itself (29,31,32). To determine the type of Ub conjugates involved in enhancing ataxin-3 binding, we carried out pulldowns with parkin ubiquitinated using either wild-type Ub or a Ub mutant (Ub^K0), which cannot support the assembly of poly-Ub chains because all seven lysines have been mutated to arginine. As shown in Figure 3B, comparable amounts of parkin ubiquitination were obtained using either wild-type Ub or Ub^K0 but only parkin conjugated with wild-type Ub enhanced ataxin-3 binding (Fig. 3B). The findings indicate that parkin polyubiquitination but not monoubiquitination or even extensive multi-monoubiquitination enhances the binding of ataxin-3.

Figure 3.

Ataxin-3 interacts with poly-Ub conjugates on parkin. (A) GST-parkin ubiquitinates itself in vitro. GST-Parkin, bound to glutathione–Sepharose, was incubated for 1 h at 37°C in reactions containing E1, E2 (His-Ubc7), Ub and ATP. Reactions were analyzed by Ub and parkin immunoblotting and by Coomassie staining. (B) Poly-Ub conjugates enhance the binding of ataxin-3 to GST-parkin. GST/GST-parkin bound to glutathione–Sepharose was ubiquitinated in vitro for indicated time intervals with Ub or for 1 h with lysine-free Ub (Ub^K0) and added into pulldown reactions containing His-ataxin-3 (14Q). Pulldowns were stained for Ponceau S and immunoblotted for ataxin-3. (C) Poly-Ub conjugates enhance the binding of ataxin-3 to parkin even in the absence of the Ubl. GST/GST-ΔUbl^parkin bound to glutathione-sepharose was ubiquitinated in vitro for indicated time intervals with Ub or for 1h with lysine free Ub (Ub^K0) and added into pulldown reactions containing His-ataxin-3 (14Q). Pulldowns were stained for Ponceau S and immunoblotted for ataxin-3. Note: Increased ataxin-3 binding in the –Ub lane is due to higher amounts of GST- ΔUbl^parkin as shown with the ponceau staining. (D) Expanded ataxin-3 binds ubiquitinated parkin. GST/GST-parkin bound to glutathione-sepharose was ubiquitinated for 1h followed by pulldowns in the presence of wild-type (14Q) or expanded (82Q) His-ataxin-3. Pulldowns were stained for Ponceau S and immunoblotted for ataxin-3.

These findings raise the possibility that parkin self-ubiquitination increases the accessibility of the parkin Ubl domain, leading to an increase in ataxin-3 binding. However, pulldowns with ubiquitinated GST-ΔUbl^parkin (Fig. 3C) demonstrate that ataxin-3 binds ubiquitinated GST-ΔUbl^parkin as strongly as ubiquitinated full-length GST-parkin. Thus, it is the poly-Ub conjugates on parkin per se rather than the accessibility of the Ubl that enhance the interaction between ataxin-3 and parkin. Neither ataxin-3 DUB activity nor the size of the polyQ repeat affected this enhancement, as polyubiquitinated parkin binds both the catalytically inactive (C14S) mutant ataxin-3 (37) (data not shown) and expanded (82Q) ataxin-3 (Fig. 3D) as strongly as wild-type ataxin-3. Together, these findings indicate that in addition to the direct Ubl-UIM and RING-Josephin interactions, ataxin-3 binding is enhanced by parkin ubiquitination.

Ataxin-3 deubiquitinates parkin

To date, no DUB partner has been identified for parkin. Given that ataxin-3 and parkin bind each other, we asked whether they could also interact functionally. FLAG-parkin, HA-Ub and either wild-type (14Q) myc-ataxin-3 or a catalytically inactive myc-ataxin-3^C14S were transfected into HEK293 cells and parkin was immunoprecipitated using an anti-FLAG resin (Fig. 4A). We found that levels of parkin-Ub conjugates, as measured by HA immunoblotting, were lower in cells transfected with wild-type (14Q) myc-ataxin-3 compared with cells transfected with catalytically inactive myc-ataxin-3^C14S or with FLAG-parkin alone (Fig. 4B). Conversely, we did not find evidence that parkin expression affected ataxin-3 ubiquitination (data not shown). Thus, ataxin-3 reduces the extent of parkin self-ubiquitination in cells and this effect requires ataxin-3 DUB activity.

Figure 4.

Ataxin-3 deubiquitinates parkin. (A and B) Ataxin-3 deubiquitinates FLAG-parkin in HEK293T cells. HEK293T cells were transfected with the indicated combination of FLAG-Parkin, myc-ataxin-3 (14Q)/myc-ataxin-3^C14S and HA-Ub^WT. Lysates were immunoprecipitated with FLAG resin and analyzed by HA immunoblotting. The optical densities of the Ub-parkin conjugates were quantified using NIH imageJ and the data represent the mean ± standard error of the mean (SEM) for three independent experiments (*P < 0.05). (C and D) Expanded ataxin-3 deubiquitinates FLAG-parkin in HEK293 cells with greater efficiency compared with wild-type ataxin-3. HEK293T cells were transfected with the indicated combination of FLAG-Parkin, myc-ataxin-3 (14Q)/myc-ataxin-3 (82Q) and either HA-Ub^WT or one of the indicated HA-Ub^Konly mutants. Lysates were immunoprecipitated with FLAG resin and analyzed by HA immunoblotting. The optical densities of the Ub-parkin conjugates were quantified using NIH imageJ and the data represents the mean ± standard error of the mean (SEM) for three independent experiments (*P < 0.05). (E) Wild-type and polyQ-expanded ataxin-3 (82Q) deubiquitinate parkin in vitro. GST-parkin, bound to glutathione–Sepharose, was incubated in an in vitro ubiquitination assay alone or with His-ataxin-3 (14Q)/His-ataxin-3 (82Q)/His-ataxin-3^C14S for 1h at 37°C. Reactions were immunoblotted for Ub, parkin and ataxin-3.

Ataxin-3 has been shown to deubiquitinate free Ub chains linked via K63 more efficiently than those linked via K48 (38). To determine whether ataxin-3 also preferentially targeted certain Ub–Ub linkages in chains conjugated to parkin, we used a panel of HA-tagged K-only Ub mutants. These mutants contain a single lysine, with all other lysines mutated to arginine. Thus, they can only form poly-Ub chains linked via the single remaining lysine. We found that FLAG-parkin, like other RING domain E3 Ub-ligases (39), can assemble poly-Ub chains on itself using several of the different K-only Ub mutants, including K6, K27, K29 and K63 (Fig. 4C, left panel). Interestingly, both wild-type (14Q) and expanded (82Q) myc-ataxin-3 reduced the levels of parkin-Ub conjugates, regardless of the K-only Ub mutant used (Fig. 4C, middle and right panels). However, compared with wild-type ataxin-3, expanded (82Q) ataxin-3 was more efficient at deubiquitinating wild-type, K27- and K29-linked HA-Ub (Fig. 4D). Thus, whereas both wild-type and polyQ-expanded ataxin-3 can reduce Ub conjugates on parkin in cells, the expanded ataxin-3 is more efficient, possibly due to enhanced deubiquitination of K27- and K29-linked Ub conjugates. Although little is known about the function of these atypical Ub chains, our findings are in line with recent work (14), identifying a role for K27-linked Ub conjugates in parkin-dependent autophagy of mitochondria (mitophagy).

Together, the findings above strongly suggest that ataxin-3 reduces Ub conjugates on parkin via its DUB activity. To test whether ataxin-3 can directly deubiquitinate parkin, in vitro GST-parkin self-ubiquitination assays were carried out as described above (Fig. 3A) in the presence of bacterially expressed and purified His-ataxin-3. Wild-type, but not catalytically inactive (C14S), His-ataxin-3 dramatically reduced the extent of parkin self-ubiquitination (Fig. 4E). Consistent with our findings above in cells, polyQ-expanded (82Q) His-ataxin-3 was also more active than wild-type ataxin-3 in vitro. Thus, ataxin-3 deubiquitinates parkin directly in vitro, which likely explains the reduction in parkin self-ubiquitination observed in cells. Such activity requires the active site cysteine in the Josephin domain and expansion of the polyQ tract enhanced the DUB activity of ataxin-3 toward parkin.

Ataxin-3 preferentially deubiquitinates parkin

Next, we asked whether the effect of ataxin-3 on parkin was specific or whether it reflected an intrinsic property of ataxin-3 to hydrolyze Ub linkages in general. Compared with similar amounts of the DUBs A20 and Usp2, ataxin-3 is relatively inefficient at disassembling free Ub chains (Fig. 5A). This may reflect a preference of ataxin-3 for substrate-bound Ub chains compared with free Ub chains. Yet, when incubated with brain lysates containing a broad range of Ub-protein conjugates, His-ataxin-3 (14Q) did not substantially decrease the levels of ubiquitinated proteins, in contrast to Usp2, which rapidly hydrolyzed the Ub-protein conjugates (Fig. 5B). As a further measure of specificity, we found that, compared with its effect on parkin, ataxin-3 had very little effect on the extent of self-ubiquitination of two other RING-type E3 ligases, SIAH and CIAP2 (Fig. 5C). Conversely, neither of the DUBs tested above could deubiquitinate parkin to the same extent as ataxin-3 (Fig. 5D), despite their comparatively stronger activity against free Ub chains (Fig. 5A) and Ub-protein conjugates (Fig. 5B). Taken together, these findings indicate that ataxin-3 preferentially deubiquitinates parkin.

Figure 5.

Ataxin-3-mediated deubiquitination of parkin is specific. (A) Ataxin-3 hydrolyzes free K48- or K63-linked free Ub chains inefficiently compared with other DUBs. K48- and K63-linked Ub chains were incubated with His-ataxin-3 (14Q) or the catalytic domain of His-A20 or Usp2 for 1 h at 37°C and analyzed by Ub immunoblotting. (B) Ataxin-3 does not exhibit broad DUB activity. Mouse brain lysates were incubated at 37°C for the indicated times with either buffer alone, His-ataxin-3 (14Q) or the core domain of USP2 and immunoblotted for Ub. (C) Ataxin-3 specifically deubiquitinates parkin but not other RING-type E3 Ub-ligases. In vitro ubiquitination reactions were assembled with GST-parkin, GST-SIAH or GST-CIAP2 in the presence or absence of His-ataxin-3 (14Q) for 1 h at 37°C and immunoblotted for Ub. (D) Parkin is specifically deubiquitinated by ataxin-3 but not other DUBs. In vitro ubiquitination reactions were assembled containing GST-parkin in the presence or absence of His-ataxin-3 (14Q), His-A20 or the core domain of USP2 and immunoblotted for Ub and His. Note that USP2 is not His-tagged.

Several E2 Ub-conjugating enzymes have been shown to support parkin-mediated ubiquitination (21,32,40). Therefore, it is possible that different E2s may dictate the conjugation of different types of Ub chains on parkin, which in turn may affect the ability of ataxin-3 to deubiquitinate parkin. However, we found that ataxin-3 efficiently deubiquitinated parkin in vitro irrespective of the E2 (Ubc4, UbcH7 or Ubc13/Uev1) used (Supplementary Material, Fig. S1A and B). Finally, His-ataxin-3 (14Q) could deubiquitinate MBP-parkin in vitro as efficiently as GST-parkin (Supplementary Material, Fig. S1C), demonstrating that ataxin-3 deubiquitinates parkin directly, irrespective of the N-terminal tag. Taken together, our results indicate that ataxin-3-mediated deubiquitination is preferential for parkin and independent of the E2 and N-terminal tag.

Expanded ataxin-3 promotes clearance of parkin via autophagy

Many E3-ligases are unstable as a result of their propensity to ubiquitinate and target themselves to the proteasome for degradation. For instance, the E3-ligase MDM2 ubiquitinates itself, leading to its rapid clearance by the proteasome (41). The presence of its DUB partner HAUSP mediates the removal of these Ub conjugates, thus protecting MDM2 from proteasomal degradation, a scenario observed with other E3-DUB pairs (42). We therefore asked whether ataxin-3-mediated deubiquitination of parkin could regulate parkin stability. Cycloheximide (CHX) pulse-chase experiments were carried out in HEK293 cells transfected with FLAG-parkin alone or with FLAG-parkin in the presence of either wild-type (14Q) or expanded (82Q) myc-ataxin-3. Consistent with previous studies (30), we found that parkin is inherently stable over a 24 h period in the presence of CHX (Fig. 6A and B). Remarkably, co-transfection of expanded (82Q), but not wild-type (14Q), ataxin-3 significantly reduced parkin levels over the course of the pulse-chase experiment (Fig. 6A and B). Thus, expanded (82Q) ataxin-3 directly promotes parkin degradation in cells. To explore the mechanisms involved, we inhibited the two main pathways for intracellular protein degradation, the Ub–proteasome system and autophagy. Surprisingly, neither MG132 nor lactacystin, two proteasome inhibitors, could suppress the effect of expanded ataxin-3 on parkin degradation (Fig. 6C and D). In contrast, 3-methyladenine (3-MA), a well-characterized inhibitor of autophagy, completely blocked the effect of expanded ataxin-3 on parkin clearance (Fig. 6C and D). Taken together, the findings suggest a novel mechanism whereby MJD-linked expanded ataxin-3 targets parkin for degradation via the autophagy pathway.

Figure 6.

Expanded ataxin-3 promotes clearance of parkin via the autophagy pathway. (A and B) Expanded ataxin-3 promotes the clearance of parkin. HEK293 cells were transfected with FLAG-parkin and either pcDNA, wild-type (14Q) or expanded (82Q) myc-ataxin-3. Transfected cells were harvested after treatment with CHX for 0, 18 or 24 h. Lysates were analyzed by immunoblotting for FLAG, actin (as a loading control) and ataxin-3. (B) Protein levels were quantified using NIH imageJ and normalized to actin. The data represent the mean ± SEM for three individual experiments (*P < 0.05, Student's t-test). (C and D) Expanded ataxin-3 promotes the clearance of parkin through the autophagy pathway. (C) HEK293 cells were co-transfected with FLAG-parkin and the expanded (82Q) myc-ataxin-3. Transfected cells were left untreated or treated with CHX for 24 h either alone or in the presence of the proteasomal inhibitor MG132 or lactacystin or the autophagy inhibitor 3-MA. Cells were harvested and analyzed by immunoblotting for FLAG, actin (as a loading control) and ataxin-3. (D) Protein levels were quantified using NIH imageJ and normalized to actin. The data represent the mean ± SEM for three individual experiments (*P < 0.05, Student's t-test).

PolyQ-expanded ataxin-3 reduces parkin levels in vivo

Given the effect of polyQ-expanded ataxin-3 in promoting clearance of parkin in HEK293 cells (Fig. 6), we next sought to determine whether ataxin-3 could also regulate parkin levels in mouse brain in vivo. Using brain homogenates prepared from wild-type and ataxin-3 knockout mice (43), we detected no differences in endogenous parkin levels (Fig. 7A). Similarly, overexpression of wild-type ataxin-3 (15Q) in the brains of transgenic YAC-ataxin-3 mice (44) did not affect parkin levels (Fig. 7B and D). Thus, consistent with our pulse-chase experiments in cells, manipulating the levels of wild-type ataxin-3 does not affect steady-state parkin levels in mouse brain. In contrast, we observed an ∼80% decrease in parkin levels in brain lysates from YAC-ataxin-3 (MJD84.2) transgenic mice, expressing expanded human ataxin-3 (84Q), compared with wild-type non-transgenic littermates (Fig. 7B and D). This decrease could not be explained by the co-aggregation of parkin with expanded ataxin-3 in insoluble inclusions, as no changes in parkin levels were observed in the insoluble brain fractions from MJD84.2 mice compared to non-transgenic mice (Fig. 7C). We also found no changes in the levels of ataxin-3 in brain lysates from parkin knockout mice, indicating that parkin does not reciprocally regulate steady-state ataxin-3 levels (Supplementary material, Supplementary Data). Taken together with the experiments above (Fig. 6), these findings suggest that MJD-associated expanded ataxin-3 promotes parkin degradation via autophagy both in cells and in vivo in mouse brain. Considering the neuroprotective functions of parkin, our findings point toward a novel mechanism for neurodegeneration in MJD.

Figure 7.

Reduced parkin levels in MJD mouse brain. (A) Deletion of ataxin-3 does not affect steady-state parkin levels in vivo. Brain homogenates from ataxin-3 knockout mice and wild-type littermates were analyzed by immunoblotting for parkin, ataxin-3 and GADPH (as a loading control). (B–E) Parkin levels are decreased in the brains of YAC-ataxin-3 (MJD84.2) transgenic mice. (B and C) Triton X-100-soluble and -insoluble fractions, prepared from the brains of 6-month-old YAC-transgenic mice expressing ataxin-3 with 84 glutamine repeats, were analyzed by parkin, ataxin-3 and actin (as a loading control) immunoblotting and compared with non-transgenic littermates. (D and E) Protein levels were quantified using NIH imageJ and normalized to actin. The data represent the mean ± SEM for the soluble fraction in three non-transgenic and three transgenic mice. **P < 0.01, Student's t-test. No differences were detected for the insoluble fraction.

DISCUSSION

The clinical and pathological overlap between PD and MJD is well established (18). However, it is less clear whether or how the pathogenic mechanisms underlying these disorders are related. In this study, we report a novel interaction between ataxin-3 and parkin, two proteins involved in the etiology of these disorders. We identify ataxin-3 as the first DUB partner for parkin and show that both wild-type and expanded ataxin-3 can deubiquitinate parkin. We find that mutant polyQ-expanded ataxin-3 exhibits increased DUB activity toward parkin, possibly by hydrolyzing K27- and K29-linked parkin-Ub conjugates with greater efficiency. Remarkably, we also find that MJD-associated mutant but not wild-type ataxin-3 promotes the clearance of parkin through the autophagy pathway. The findings raise the intriguing possibility that a decrease in parkin levels may contribute to the pathogenesis of MJD and account for the parkinsonian features observed in patients with the disease.

The role of ataxin-3 in the Ub pathway stems not only from its DUB activity, encoded by the Josephin domain, but also from its UIMs, which bind Ub and proteins with Ubl domains (5,45,46). Previous work has established the parkin Ubl as an important interaction module, linking parkin to SH3 domains within endocytic BAR-proteins such as endophilin-A (27) and to UIMs within Eps15 and S5a (26,47,48). As with Eps15, binding between parkin and ataxin-3 is multi-modal, involving an interaction between the IBR-RING2 and Josephin domain in addition to the Ubl:UIM interaction. We also observe an enhancement of ataxin-3 binding to parkin through an interaction between ataxin-3 and the poly-Ub conjugates attached to parkin. The increase in ataxin-3 binding requires the formation of poly-Ub chains on parkin, as no increase was observed with multi-mono- and/or monoubiquitinated parkin. Moreover, our time course experiments suggest that Ub conjugates on parkin need to attain a certain length in order to increase ataxin-3 binding, consistent with previous studies showing that ataxin-3 was unable to bind poly-Ub chains with less than four Ub moieties (7,49). Taken together, the findings suggest a complex multi-layered interaction involving direct Ubl: UIM and IR-RING2:Josephin binding between parkin and ataxin-3, as well as a second parkin activity-dependent interaction between Ub conjugates on parkin and ataxin-3.

Although many DUBs have been shown to interact with E3 Ub-ligases (41,50), to date, no DUB partner had been identified for parkin. Like other E3s, parkin can ubiquitinate itself (32,40). We show here that ataxin-3 can hydrolyze Ub conjugates on parkin both directly in vitro and in cells, establishing ataxin-3 as the first DUB partner of parkin. In contrast, ataxin-3 failed to deubiquitinate other E3 ligases and other DUBs failed to deubiquitinate parkin. Moreover, whereas a previous study found that parkin could ubiquitinate a small fragment of ataxin-3 containing the isolated polyQ tract (51), we found no evidence of parkin-mediated ubiquitination of full-length ataxin-3 in this study (data not shown). Together, these data indicate that ataxin-3-mediated deubiquitination of parkin is specific. By antagonizing E3 auto-ubiquitination, DUBs are generally believed to stablize their cognate E3 partner by pre-empting their targeting to the proteasome. In fact, overexpression of a DUB can stabilize its cognate substrate, as has been demonstrated for the DUB enzyme FAM (Fat Facets in Mouse), and its endogenous substrate, β-catenin (52). In contrast, the presence of the expanded ataxin-3 mediates a decrease in parkin levels (Figs 6 and 7), which would make the MJD-associated ataxin-3 the first known example of a DUB to downregulate, rather than upregulate levels of its protein substrate. Somewhat surprisingly, we found no evidence of wild-type ataxin-3 stabilizing parkin in cultured cells or in brain. In contrast to other E3s, we found parkin to be inherently long-lived and insensitive to proteasomal degradation in cells, which may at least partly explain the lack of any further stabilization by overexpressing ataxin-3. Indeed, neither transgenic overexpression of wild-type ataxin-3 nor knockout of endogenous ataxin-3 affected steady-state parkin levels in mouse brain. Thus, despite its capacity to efficiently deubiquitinate parkin, ataxin-3 does not regulate parkin stability or turnover in cells or in vivo.

Ub conjugates linked via K48 target substrates to the proteasome, whereas those linked via one of the six other lysines can alter the function of the modified protein without leading to degradation. Using HA-Ub^Konly mutants, we observe parkin to ubiquitinate itself predominantly via K6, K27, K29 and K63 rather than via K48. The finding may help explain the relative insensitivity to proteasomal degradation in our pulse-chase assays and is consistent with previous reports showing parkin to be long-lived (53) and to ubiquitinate itself via non-canonical linkages (29,32). Once these considerations are taken into account, the failure of ataxin-3 to regulate parkin stability appears much less surprising. What then is the function of ataxin-3-mediated deubiquitination of parkin? The K6, K27, K29 and K63 Ub linkages favored by parkin have been implicated in trafficking, signaling, DNA repair and autophagy (13–16,39,54). Whether deubiquitination by wild-type ataxin-3 regulates the targeting of parkin to one of these pathways will be a question of great interest for future work. Ubiquitination has also been shown to activate E3-ligases such as Bmi-RING1b (39), TRAF6 (55) and Smurf (56). As the effects of auto-ubiquitination on parkin activity are currently unknown, it is difficult to determine exactly how ataxin-3 might regulate parkin activity. Our finding that ataxin-3 binds poly-Ub chains on parkin suggests that parkin activity regulates the interaction with ataxin-3, possibly positioning it in a favorable orientation to deubiquitinate parkin or one of its substrates. Indeed, kinetic analysis of Ub chain assembly reveals that ataxin-3-mediated deubiquitination is functionally coupled to parkin ubiquitination, suggesting that ataxin-3 controls the abundance and edits the architecture of Ub chains on parkin concurrently with chain assembly (manuscript in preparation). Understanding the basis for such functional coupling will undoubtedly help elucidate the role of ataxin-3 in regulating parkin activity and targeting to appropriate cellular pathways.

Parkin is well known to be neuroprotective in a number of settings with a complete or partial loss of the parkin protein or a reduction in parkin function implicated in the pathogenesis of PD (22). A growing list of processes have been shown to regulate parkin function and hence influence neurodegeneration, including ubiquitination by the E3 Ub-ligase Nrdp1 (53), phosphorylation by Cdk5 (57) and Casein kinase II (58), nitrosylation (23,24) and attack by dopamine (25). We now add expansion of the ataxin-3 polyQ tract to this list by showing that MJD-linked mutant ataxin-3 promotes the autophagic degradation of parkin. We propose that destabilization of parkin in the context of MJD could account for some of the parkinsonian features observed in the disease. We do not claim that reducing parkin levels is required or fully accounts for MJD pathogenesis. It is well accepted that the polyQ expansion confers a toxic gain-of-function to the various polyQ-disease proteins, most plausibly by promoting their misfolding and aggregation (3). However, it is unlikely that the many clinical, neuropathological and cell biological differences between the distinct CAG expansion diseases can be fully explained by the polyQ expansion alone. Rather, it is likely that domains outside of the polyQ region, responsible for specific interactions such as the one we identify between ataxin-3 and parkin, play a crucial role in conferring disease specificity and pathogenesis.

Whereas the polyQ expansion had no apparent effect on the ability of ataxin-3 to bind parkin, it did increase its efficiency in DUB parkin. This effect is unlikely to stem from differences in the intrinsic hydrolase activities of wild-type and expanded ataxin-3, as previous work has found these to be similar (59). Studies with unanchored Ub chains suggest that ataxin-3 preferentially targets K63- compared with K48-linked Ub chains (38), raising the possibility that the difference we observe between wild-type and mutant ataxin-3 may reflect the preferential deubiquitination of certain linkages in parkin-Ub conjugates. Indeed, whereas both wild-type and mutant ataxin-3 could deubiquitinate parkin regardless of the linkages used in chain assembly, we found that MJD-associated expanded ataxin-3 deubiquitinated K27- and K29-linked Ub conjugates on parkin with greater efficiency. K27- and K29-linked Ub conjugates have been implicated in lysosomal and autophagic degradation pathways (13–15). However, the mechanisms involved remain poorly characterized. One possibility is that the presence of Ub conjugates on parkin linked in this manner may protect it from autophagic degradation. In this scenario, the preferential removal of K27- and K29-linked Ub conjugates from parkin by expanded ataxin-3 would oppose this protective effect and enhance parkin turnover by autophagy. Alternatively, the effect of expanded ataxin-3 on parkin turnover by autophagy may be independent of its catalytic DUB activity and involve a direct effect of the polyQ expansion on the rate of autophagy or on the previously described role of parkin in trafficking and autophagy (14,26,27,60).

Autophagy has been implicated in clearing mutant polyQ protein aggregates in several models of polyQ-expansion diseases, including MJD (61–63). Thus, our findings raise the possibility that this presumably beneficial effect of autophagy could be mitigated by the concomitant degradation of neuroprotective factors such as parkin. Irrespective of its own degradation, we cannot exclude that parkin is also involved in the autophagic process per se, as has been reported for the autophagic removal of damaged mitochondria (14,60). Interestingly, mitochondrial abnormalities have been observed in mouse models of polyQ diseases, including MJD (64,65), and a pool of ataxin-3 localizes to mitochondria (66). Thus, given the emerging role for mitochondria in both PD and MJD, it will be important to determine whether the relationship between ataxin-3 and parkin converge at the level of mitochondrial autophagy and quality control to influence the pathogenesis of these two disease.

MATERIALS AND METHODS

Antibodies, plasmids and reagents

Antibodies were used to detect ataxin-3 (Chemicon), parkin (Cell Signaling), Ub (Covance), actin (Chemicon), His (Novagen), HA (Roche), FLAG (Sigma), GADPH (Chemicon) and Myc (gift of Phil Barker). Myc- and His-tagged ataxin-3 (14Q) and ataxin-3 (82Q) were generated by sub-cloning GFP-ataxin-3 (14Q) and GFP-ataxin-3 (82Q) (gift of Guy Rouleau) into pCMV-Myc (Clontech) and pTrc-His (Invitrogen). Myc- and His-Josephin were sub-cloned into the same vectors by PCR using the primers: 5′-CCGCTCGAGCGGGGAGTCCATCTTCCACGAGAAACAA-3′ and 5′-CGGGGTACCCCGTTGTGCTAATTCTTCTCCAATAAG-3′.

Ataxin-3 C14S, UIM1 (S236A), UIM2 (S256A), UIM3 (S347A), UIM1 + 2, UIM1 + 3, UIM2 + 3 and triple UIM mutants were generated using QuickChange Site-Directed Mutagenesis (Stratagene). HA-Ub plasmids were a gift of Ted Dawson. Rat parkin plasmids were described previously (26). CHX and 3-MA were purchased from Sigma. MG132 and lactacystin were purchased from Boston Biochem.

Cell culture, transfections and pulse chase assays

HEK293T cells were maintained at 37°C and 5% CO₂ in DMEM with 10% fetal bovine serum, 2 mm glutamine, 100 U/ml of penicillin and 100 µg/ml of streptomycin. HEK293T cells were transfected using Lipofectamine 2000 (Invitrogen). For pulse chase assays, CHX (20 µg/ml) was added 24 h post-transfection for the indicated intervals either alone or in the presence of MG132 (5 µm), lactacystin (1 µm) or 3-MA (5 mm). Cells were lysed in RIPA buffer containing a cocktail of protease inhibitors.

Protein expression, binding assays and immunoprecipitations

GST and His fusion proteins were expressed in Escherichia coli strain BL21 (DE3) and affinity-purified using Glutathione–Sepharose 4B beads (Amersham) or Ni-NTA Agarose (Qiagen). Bead-immobilized GST proteins were incubated with 2 μg of purified His proteins at 4°C for 16 h in 1 ml of 20 mm HEPES-KOH pH 7.4, 100 mm NaCl, 10% glycerol, 0.1% Triton X-100, 0.5 mm DTT or with 500 μg of HEK293T cell lysate in 50 mm HEPES-KOH pH 7.4, 100 mm NaCl, 1% Triton X-100, 10% glycerol and protease inhibitors at 4°C for 3 h followed by extensive washing. For immunoprecipitations, HEK293T cells were lysed in 25 mm HEPES-KOH pH 7.4, 125 mm KOAc, 2.5 mm MgOAc, 5 mm EGTA and 1 mm DTT with protease inhibitors and incubated with primary antibody for 2 h at 4°C, followed by the addition of Protein G–Sepharose beads for 1 h at 4°C. For FLAG-parkin immunoprecipitations, cells were lysed in 2% SDS, 20 mm Tris–HCl pH 7.4, 150 mm NaCl with boiling followed by neutralization by dilution to a final concentration of 20 mm Tris–HCl pH 7.4, 150 mm NaCl, 1% Triton X-100, 0.5% SDS, 1 mm EDTA, 10 mmN-ethylmaleimide and protease inhibitors. Lysates (1 mg/ml) were incubated with 20 μl of Anti-FLAG resin (Sigma) for 2 h at 4°C. FLAG-parkin was eluted with 40 μg FLAG Peptide (Sigma) in 20 mm Tris–HCl pH 7.4, 150 mm NaCl at 4°C for 30 min.

In vitro ubiquitination and deubiquitination assays

In vitro ubiquitination reactions were carried out as described (30). Briefly, ∼0.25 µm maltose-eluted MBP-parkin or GST-parkin, GST-SIAH (gift of Dr Frédéric Luton) or GST-CIAP2 (gift of Phil Barker), bound to glutathione–Sepharose beads, were incubated in 40 μl of ubiquitination buffer (50 mm Tris–HCl pH 8.0, 5 mm MgCl₂, 4 mm ATP, 2 mm DTT) containing 90 nm E1 (Boston Biochem), 1 μm E2 [His-Ubc7 (gift of Ryosuke Takahashi), His-Ubc4 (67), UbcH7 and Ubc13/Uev1 (Boston Biochem)] and 0.2 mm Ub (Boston Biochem). The reactions were incubated at 37°C for 1 h unless otherwise indicated and analyzed by SDS–PAGE followed by immunoblotting or Coomassie staining. For deubiquitination, ∼0.25 µm His-ataxin-3 (14Q), His-ataxin-3 (82Q), His-ataxin-3 (14Q)^C14S or the catalytic domain of His-A20 (Boston Biochem) or Usp2 (also known as UBP^testis) (68) were incubated with the ubiquitination reactions above or with 1 μg of K48- or K63-linked free Ub chains (Boston Biochem) in 50 mm HEPES-KOH pH 7.4, 150 mm KCl, 10 mm DTT, 5% glycerol, 0.01% Triton X-100 for 1 h at 37°C. To perform GST pulldowns and to examine the cleavage of Ub-protein conjugates using rat brain lysates, brain fractions were prepared as described (30). Ten micrograms of His-ataxin-3 or the core domain of Usp2 were incubated with 250 μg of fraction S3 from brain in 10 mm HEPES, 7.4; 0.5 mm EDTA, 8.0; 1 mm DTT at 37°C for the indicated times and analyzed by Ub immunoblotting. Pulldowns with ubiquitinated proteins were performed as described above with GST or GST-parkin ubiquitination reactions incubated for 1 h before pulldowns were assembled containing 2 μg of His-ataxin-3 proteins.

Transgenic ataxin-3 mouse brain fractions

Ataxin-3 knockout, YAC-15Q and 84Q transgenic mice were generated as previously described (44). Brains from 6-month-old mice were homogenized in 50 mm Tris–HCl pH 7.4, 150 mm sodium chloride, 0.1% SDS, 0.5% sodium deoxycholate, 1% NP-40 with protease inhibitors. Soluble and insoluble fractions were separated by centrifugation (3000g, 15 min) and were analyzed by immunoblotting for parkin, ataxin-3, actin and GADPH.

FUNDING

This work was supported by a CIHR operating grant no. 84345. T.M.D. is supported by fellowships from the Parkinson Society of Canada and the National Ataxia Foundation. A.D. is supported by a grant from the Deutsche Forshungsgemeinschaft (DJ 65/4-1). E.A.F. is supported by the FRSQ Chercheur-Boursier senior.

ACKNOWLEDGEMENTS

We thank Dr Lenore Beitel and Dr Jean-Francois Trempe for critical reading of the manuscript. We thank Dr Randall Pittman (University of Pennsylvania) for scientific discussions. We thank Katrin Schulz and Ioana Medrea for technical assistance.

Conflict of Interest statement. None declared.

Author notes

†

These authors contributed equally to the work.