Aggregation of Polyglutamine-expanded Ataxin 7 Protein Specifically Sequesters Ubiquitin-specific Protease 22 and Deteriorates Its Deubiquitinating Function in the Spt-Ada-Gcn5-Acetyltransferase (SAGA) Complex*

Background: Expansion of the polyQ tract in ataxin 7 is the main pathology of SCA7. Results: The aggregates formed by polyQ-expanded ataxin 7 sequester USP22 through specific interactions. Conclusion: Sequestration of USP22 impairs its deubiquitinating function in the SAGA complex. Significance: This provides evidence for the hijacking model in which specific sequestration leads to cytotoxicity and neurodegeneration. Human ataxin 7 (Atx7) is a component of the deubiquitination module (DUBm) in the Spt-Ada-Gcn5-acetyltransferase (SAGA) complex for transcriptional regulation, and expansion of its polyglutamine (polyQ) tract leads to spinocerebellar ataxia type 7. However, how polyQ expansion of Atx7 affects DUBm function remains elusive. We investigated the effects of polyQ-expanded Atx7 on ubiquitin-specific protease (USP22), an interacting partner of Atx7 functioning in deubiquitination of histone H2B. The results showed that the inclusions or aggregates formed by polyQ-expanded Atx7 specifically sequester USP22 through their interactions mediated by the N-terminal zinc finger domain of Atx7. The mutation of the zinc finger domain in Atx7 that disrupts its interaction with USP22 dramatically abolishes sequestration of USP22. Moreover, polyQ expansion of Atx7 decreases the deubiquitinating activity of USP22 and, consequently, increases the level of monoubiquitinated H2B. Therefore, we propose that polyQ-expanded Atx7 forms insoluble aggregates that sequester USP22 into a catalytically inactive state, and then the impaired DUBm loses the function to deubiquitinate monoubiquitinated histone H2B or H2A. This may result in dysfunction of the SAGA complex and transcriptional dysregulation in spinocerebellar ataxia type 7 disease.


Human ataxin 7 (Atx7) is a component of the deubiquitination module (DUBm) in the Spt-Ada-Gcn5-acetyltransferase
(SAGA) complex for transcriptional regulation, and expansion of its polyglutamine (polyQ) tract leads to spinocerebellar ataxia type 7. However, how polyQ expansion of Atx7 affects DUBm function remains elusive. We investigated the effects of polyQexpanded Atx7 on ubiquitin-specific protease (USP22), an interacting partner of Atx7 functioning in deubiquitination of histone H2B. The results showed that the inclusions or aggregates formed by polyQ-expanded Atx7 specifically sequester USP22 through their interactions mediated by the N-terminal zinc finger domain of Atx7. The mutation of the zinc finger domain in Atx7 that disrupts its interaction with USP22 dramatically abolishes sequestration of USP22. Moreover, polyQ expansion of Atx7 decreases the deubiquitinating activity of USP22 and, consequently, increases the level of monoubiquitinated H2B. Therefore, we propose that polyQ-expanded Atx7 forms insoluble aggregates that sequester USP22 into a catalytically inactive state, and then the impaired DUBm loses the function to deubiquitinate monoubiquitinated histone H2B or H2A. This may result in dysfunction of the SAGA complex and transcriptional dysregulation in spinocerebellar ataxia type 7 disease.
There are nine inherited neurodegenerative diseases that have been identified to date to be caused by polyglutamine (polyQ) 2 expansion of the related proteins (1)(2)(3). Spinocerebellar ataxia type 7 (SCA7) is caused by expansion of the polyQ tract in ataxin 7 (Atx7) (4,5). It is the only polyQ disease that leads to visual failure and eventual blindness, except for neurodegeneration (6). Many studies have revealed that there are nuclear inclusions in neurons and photoreceptor cells of SCA7 patients (7)(8)(9). These inclusions contain not only polyQ-expanded Atx7 but also other proteins, such as Cbl-associated proteins (10,11); CREB-binding protein, HSP70 chaperones, and 19S subunit of proteasome (12,13); and GCN5 (14). Dysfunction of these proteins might be related to the pathogenesis of SCA7.
Transcriptional dysregulation may be central to the pathogenesis of SCA7, especially affecting a subset of genes involved in neuronal function, such as the genes specifically expressed in rod photoreceptors (25)(26)(27). However, it is controversial whether polyQ expansion of Atx7 affects the histone acetyltransferase activity of SAGA (27)(28)(29). Moreover, it is reported that polyQ expansion may alter Atx7 binding and the level of H2Bub at the RELN promoter and, therefore, reduce reelin expression (14). Recently, Mohan et al. (30) reported that loss of Atx7 in Drosophila reduces the level of H2Bub and causes neural and retinal degeneration similar to the phenotype when polyQ-expanded Atx7 is overexpressed in Drosophila (31). Although many efforts have been made to reveal the pathogen-esis of SCA7, how polyQ-expanded Atx7 causes transcription alterations still needs to be elucidated.
Our previous studies have focused on the neurodegenerative disease-related proteins that sequester their interacting partners into insoluble aggregates or inclusions (11,(32)(33)(34). Recently, we revealed that aggregation of polyQ-expanded ataxin 3 (Atx3) specifically sequesters ubiquitin chains and P97/ VCP into inclusions and impairs the normal function of P97/ VCP (35). We investigated the effects of polyQ expansion in Atx7 on USP22 and found that polyQ-expanded Atx7 specifically sequesters USP22 into aggregates or inclusions and reduces the deubiquitination level of histone H2Bub. Therefore, we propose that aggregation of polyQ-expanded Atx7 specifically sequesters USP22 and deteriorates its deubiquitinating function in the SAGA complex.

Experimental Procedures
Plasmids, Antibodies, and Reagents-For prokaryotic expression, the coding sequence for USP22-N (1-170) was PCR-amplified and cloned into a pGEX-4T3 vector. The cDNA for Atx7-ZnF (residues 75-172) was cloned into pET-MG (36). For eukaryotic expression, full-length USP22, USP22-N (1-170), and USP22-USPD (172-525) were cloned into the FLAG-pcDNA3.0 vector, and USP5 was cloned into the FLAG-pCMV-Tag2B vector. Atx7 10Q -N (1-172) and its ZnF mutants were PCR-amplified and cloned into the FLAG-pcDNA3.0 vector or the HA-pcDNA3.0 vector. To obtain Atx7 EPQ -N and its ZnF mutants, we digested the DNA of full-length Atx7 100Q (stored by our laboratory) with restriction enzymes and retrieved the small fragment that contains the polyQ tract. We PCR-amplified the posterior fragment of Atx7 and then ligated the whole cDNA into the FLAG-pcDNA3.0 vector. To obtain the mutants of Atx7 EPQ , we applied a similar method as for Atx7 EPQ -N but cloned into pEGFP-N1. The polyQ tracts are variable during cloning of the plasmids, but the lengths are around 100Q. All the constructs were verified by DNA sequencing. For Western blotting, antibodies against FLAG and USP22 were from Sigma, anti-HA antibody was from Santa Cruz Biotechnology, anti-H2Bub was from Cell Signaling Technology, anti-H2B was from Bioworld, and anti-GAPDH was from Zen BioScience. All secondary antibodies were purchased from Jackson ImmunoResearch Laboratories. PVDF membranes were from PerkinElmer Life Sciences, and proteins were visualized using an ECL detection kit (Thermo Scientific). Ubiquitin vinyl sulfone was purchased from Boston Biochem. The FLAG peptide was obtained from ChinaPeptides Co.
Cell Culture and Transfection-Human HEK 293T cells or HeLa cells were cultured in DMEM (HyClone) supplemented with 10% fetal bovine serum (Gibco) and penicillin-streptomycin at 37°C under a humidified atmosphere containing 5% CO 2 . All transfections of HEK 293T cells were performed by using PolyJet TM reagent (SignaGen). To test H2Bub, HeLa cells were transfected with the respective plasmids by using Lipofectamine 2000 (Invitrogen) and harvested about 48 h (for fulllength Atx7) or 72 h (for Atx7-N) after transfection.
Immunoprecipitation and Immunofluorescence Microscopy-For immunoprecipitation, HEK 293T cells were harvested and lysed in radioimmune precipitation assay (RIPA) buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, and protease inhibitor mixture) on ice for 30 min, and then the lysates were incubated with FLAG antibodyconjugated agarose (Abmart) for 2 h at 4°C. The beads were washed with RIPA buffer four times and subjected to Western blot analysis.
For immunofluorescence microscopy, HEK 293T cells grown on cover slides were transfected with PolyJet TM . For cotransfection, about 48 h later, the cells were fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.1% Triton X-100 for 1 min, and blocked with 5% BSA for 1 h. Then the cells were incubated with an antibody against FLAG (1:100, Sigma) for 1 h at room temperature. After washing with PBS, the cells were labeled with a TRITC-conjugated anti-mouse antibody (1:100, Jackson ImmunoResearch Laboratories) for 1 h, and the nuclei were stained with Hoechst (Sigma). The cells were visualized on a Leica Microsystems TCS SP8 confocal microscope. To detect endogenous USP22, immunocytochemistry experiments were performed about 96 h after transfection.
Supernatant/Pellet Fractionation for Cell Lysates-About 72 h after transfection, HEK 293T cells were lysed in 100 l of radioimmune precipitation assay buffer on ice for 30 min and centrifuged at 16,200 ϫ g for 15 min at 4°C. Then the supernatant was added to 100 l of 2ϫ loading buffer (2% SDS), and the pellet was washed sufficiently with RIPA buffer three times at 4°C and added to 40 l of 4ϫ loading buffer (4% SDS). Finally the samples were subjected to Western blotting. For time course experiments, the cultured cells were harvested 24, 36, 48, 60, or 72 h after transfection.
Ubiquitin Vinyl Sulfone (Ub-VS) Assay-HEK 293T cells were harvested and then subjected to immunoprecipitation for 2 h. The beads were incubated with FLAG peptide to competitively elute the immunoprecipitated material for 1 h at 4°C in Ub-VS reaction buffer (100 mM Tris-HCl, 100 mM KCl, 3 mM DTT, and 5% glycerol (pH 8.0)). The immunoprecipitated material was concentrated with 3-kDa centrifugal filter devices (Millipore), incubated with 5 M Ub-VS in Ub-VS reaction buffer for 30 min at 37°C, and then used for Western blotting.
The Interaction between Atx7 and USP22 Depends on the N-terminal ZnF Domain of Atx7-To better understand the mechanism underlying the sequestration of USP22 by polyQexpanded Atx7, we explored the specific interaction between Atx7 and USP22. Atx7 contains a polyQ tract (Qn), a prolinerich region, a ZnF domain, and a SCA7 domain (39), whereas USP22 is comprised of an N-terminal ZnF domain and a USP domain (40) (Fig. 2A). A previous study has indicated that the N-terminal ZnF domain of Atx7 is critical for assembly of the DUB module (20), so we used Atx7-N (residues 1-172) to investigate the interaction with USP22. Atx7 10Q -N or Atx7 EPQ -N was transfected into HEK 293T cells, and the cell lysates were subjected to coimmunoprecipitation (co-IP). The data showed that endogenous USP22 could be coimmunoprecipitated by both the normal and polyQ-expanded proteins (Fig. 2B). Furthermore, we identified the specific region of USP22 associating with Atx7-N. Although not strictly quantitative, the co-IP experiment showed that the N terminus of USP22 (USP22-N, residues 1-170) could immunoprecipitate a considerable amount of Atx7 10Q -N whereas the C terminus of USP22 (USP22-C, residues 172-525) could not (Fig. 2C). This interaction between Atx7-N and USP22-N was also confirmed by a GST pulldown experiment (Fig. 2D). As shown, the ZnF domain of Atx7 interacts directly with the N-terminal domain of USP22, which is mainly comprised of a ZnF domain ( Fig. 2A).
A Mutation in the ZnF Domain of Atx7 Disrupts Its Interaction with and Sequestration of USP22-Our previous study indicated that specific interaction is important for the sequestration of its interacting partners by polyQ proteins (33,35). We proposed that the sequestration of USP22 by Atx7 EPQ is also closely associated with their interaction. To test this hypothesis, we prepared the Atx7 mutants to disrupt the interaction with USP22 and re-examined the sequestration. The three ZnF mutants of Atx7 were C2* (C150S/C153S), H2* (H165N/ H170N), and C2H2* (C150S/C153S ϩ H165N/H170N) (Fig.  3A). Co-IP experiment showed that these mutations largely disrupted the interaction between Atx7 10Q -N and endogenous USP22 (Fig. 3B). Therefore, we performed confocal microscopy to detect the sequestration of USP22 by Atx7 EPQ or its ZnF mutants (C2*, H2*, and C2H2*). Compared with wild-type Atx7 EPQ , the three mutants lost the ability to sequester endogenous USP22 into inclusions (Fig. 3C), suggesting that polyQexpanded Atx7 sequesters USP22 into inclusions depending on the specific interaction between them.
PolyQ-expanded Atx7 Sequesters Endogenous USP22 into Insoluble Aggregates-To further examine whether Atx7 EPQ can redistribute endogenous USP22, we performed supernatant/pellet fractionation of the cell lysates. The data showed that overexpression of Atx7 EPQ -N could increase the amount of USP22 in the pellet fraction because of the presence of Atx7 EPQ -N in the pellet (Fig. 4A) but Atx7 10Q -N could not. This hijacking effect was also confirmed by the supernatant/pellet fractionation experiments in a dose-dependent manner (Fig. 4,  B and C). Moreover, we also performed time course experiments on the sequestration of USP22 into the pellet fraction by . Atx7 10Q -GFP or Atx7 EPQ -GFP (100Q) was cotransfected with FLAG-USP22 into HEK 293T cells, and the cells were visualized by confocal fluorescence microscopy. Atx3 71Q -GFP and FLAG-USP5 were set as controls. USP22 and USP5 were stained with anti-FLAG antibody (red), and nuclei were stained with Hoechst (blue). Scale bar ϭ10 m. B, sequestration of endogenous USP22. HEK 293T cells transfected with Atx7 10Q -GFP or Atx7 EPQ -GFP (100Q) were visualized by confocal fluorescence microscopy. USP22 was stained with anti-USP22 antibody (red), and nuclei were stained with Hoechst (blue). Scale bar ϭ 10 m.
Next, we investigated whether Atx7 EPQ -N sequestered USP22 into aggregates depending on the ZnF domain. The result showed that, in comparison with the wild type, the amount of USP22 sequestered by the ZnF mutants decreased considerably (Fig. 4F). Taken together, these results demonstrate that polyQ-expanded Atx7 sequesters USP22 into insoluble aggregates depending on the interaction through the ZnF domain of Atx7.
PolyQ Expansion in Atx7 Reduces the Catalytic Activity of USP22-To detect whether polyQ-expanded Atx7 affects the catalytic activity of USP22, we immunoprecipitated endogenous USP22 with FLAG-tagged Atx7-N with different polyQ lengths and examined its catalytic activity using Ub-VS. Ub-VS is a C-terminally modified vinyl sulfone derivative of ubiquitin that can irreversibly modify an active deubiquitinating enzyme with a covalent bond (20,41,42). Compared with Atx7 10Q -N, which led to the formation of a considerable amount of USP22 Ub-VS derivative, Atx7 EPQ -N resulted in less product (Fig.  5, A and B), indicating that polyQ-expanded Atx7 reduces the catalytic activity of USP22.
PolyQ-expanded Atx7 Impairs the Deubiquitination of Histone H2Bub-Our studies revealed that polyQ-expanded Atx7 could affect USP22 not only by sequestration but also by impairing its catalytic activity. We wondered whether polyQ expansion of Atx7 affects the normal function of USP22. It has been demonstrated that incorporation of USP22 into the DUBm endows the SAGA complex with the main deubiquitinating function for histone H2B (19,20,22,43). Therefore, we investigated the effect of polyQ expansion in Atx7 on the H2Bub level. Compared with the mock vector, overexpression of Atx7 10Q -N had no influence on the H2Bub level, whereas overexpression of Atx7 EPQ -N significantly increased the level  (n ϭ 3). F, comparison of the sequestration of USP22 by Atx7 EPQ -N and its ZnF mutants. FLAG-tagged Atx7 EPQ -N and its mutants were transfected into HEK 293T cells, respectively. The polyQ lengths of Atx7 EPQ -N used were as follows: WT, 115Q; C2*, 115Q; H2*, 115Q; C2H2*, 106Q. of H2Bub (Fig. 6, A and B). However, overexpression of the ZnF mutants of Atx7 EPQ -N caused a significant decrease of the level of H2Bub compared with the wild type (Fig. 6, C and D). To further verify the above results, we also applied full-length Atx7 and repeated the experiment. The data showed that overexpression of full-length Atx7 EPQ significantly increased the level of H2Bub, whereas its C2H2* mutant restored H2Bub to the mock level (Fig. 6, E and F). Together, these findings corroborate that polyQ expansion in Atx7 impairs the deubiquitinating function of USP22 by specific interaction mediated by the N-terminal ZnF domain of Atx7.

Discussion
The development of neurodegenerative diseases is a progressive process. Protein aggregation and inclusion formation are considered to be a hallmark of these diseases. Accumulating evidence has revealed that protein aggregates or inclusions lead to cellular dysfunction through the sequestration of essential proteins (33, 35, 44 -47). We investigated the effect of polyQexpanded Atx7 on its interacting partner, USP22, and found that the aggregates or inclusions sequester USP22, which leads to dysfunction of USP22 and, consequently, an altered level of monoubiquitinated histone H2B. Moreover, the sequestration is an ongoing process over time. With the aggregation of polyQexpanded Atx7, the amounts of USP22 in the insoluble aggregates increase gradually.
A number of studies demonstrate that the inclusions formed by polyQ-expanded proteins such as Atx7 contain various other cellular essential proteins (12,13,48). We discovered, for the first time, that the inclusions or aggregates of polyQ-expanded Atx7 specifically sequester USP22, as in the case of polyQ-expanded Atx3 aggregates, which specifically sequester P97/VCP and ubiquitin conjugates (35). It has also been reported that GCN5 is present in the nuclear inclusions of polyQ-expanded Atx7 in a SCA7 astrocyte model (14). Given these findings, we speculate that the inclusions or aggregates of polyQ-expanded Atx7 also sequester other subunits of the SAGA complex, especially the components of the DUBm, which may ultimately ruin the function of the SAGA complex. Specific Interaction Is Vital for the Sequestration of USP22 by polyQ-expanded Atx7-The N-terminal region of Atx7 (residues 75-172), together with USP22, Atx7L3, and ENY2, can assemble into a minimal recombinant DUBm (20). Our studies indicate that sequestration of USP22 by polyQ-expanded Atx7 relies on the specific interaction between their harboring domains. We confirmed that the N-terminal ZnF domain of Atx7, whether normal or polyQ-expanded, can interact with USP22 and that the N-terminal region of USP22, including its harbored ZnF domain, contributes to the interaction and sequestration of USP22 into the inclusions formed by polyQexpanded Atx7. Mutation of the ZnF domain in Atx7 abolishes the interaction with and sequestration of USP22 into inclusions or insoluble aggregates. All of these data demonstrate that specific interaction is critical for the sequestration of USP22 by polyQ-expanded Atx7. This is also consistent with previous studies of the sequestration of interacting partners by polyQ (33, 35, 49 -51) and other amyloidogenic proteins (34,45,52).
We focused on the sequestration that is mediated by domaindomain interactions, whereas the polyQ tract is just the source for protein aggregation. On the other hand, some studies have revealed that the sequestration is mediated by coaggregation of polyQ tracts (53)(54)(55). Cooperative polyQ-polyQ interactions also prompt aggregation of the polyQ proteins. In this case, a polyQ protein can sequester another polyQ-containing protein, and they coaggregate into inclusions or insolubilities (56,57). This nonspecific sequestration may sometimes also deteriorate the functions of other polyQ proteins.
Aggregation of polyQ-expanded Atx7 Impairs the Function of USP22-The best known function of USP22 is deubiquitinating monoubiquitinated histone H2B (19,20,22). We revealed that polyQ-expanded Atx7 can increase the level of H2Bub, depending on the specific interaction through its ZnF domain, suggesting that polyQ expansion of Atx7 interferes with the normal function of USP22 and, therefore, the DUB module. Recently, studies have reported that polyQ expansion of Atx7 increases the global level of H2Bub without changing the bulk levels of H3 Lys-9/14 acetylation (14,58). Our result corroborates the finding that expansion of the polyQ tract in Atx7 leads to formation of aggregates and that the aggregates specifically sequester USP22 and other components of DUBm into a dysfunctional state. This process leads to loss of DUBm function in deubiquitinating histone H2Bub (Fig. 7). Because H2Bub is important for both transcription initiation and elongation (23,24), dysfunction of the DUBm in the SAGA complex caused by polyQexpanded Atx7 may be a major source for transcriptional dysregulation in SCA7 disease.
The Generality and Individuality of polyQ Diseases-Our previous work has demonstrated that polyQ-expanded Atx3 specifically sequesters P97/VCP and influences its function in down-regulating neddylation and proposed a hijacking model for the cytotoxicity and neurodegeneration caused by polyQexpanded proteins (35). Here we report that polyQ-expanded Atx7 sequesters USP22 into aggregates or inclusions through a specific interaction that impairs the function of the DUBm in deubiquitinating H2Bub. Generally, for a polyQ-disease protein, its expanded polyQ-tract region leads to the formation of aggregates or inclusions, and its flanking domains or motifs are responsible for specific interactions with other proteins. During aggregation, its interacting partners are sequestered into insoluble aggregates and become dysfunctional, whereas the soluble fraction of the active partners are decreased consequently, which is probably the major source of cytotoxicity and neurodegeneration. On the other hand, it is also mentioned that some polyQ disease proteins are able to sequester the components involved in cell quality control through nonspecific sequestration (44,47,59,60). However, different polyQ proteins have different domains or motifs that mediate specific interactions with respective partners. A specific flanking domain or motif determines its interacting proteins to be sequestered and dysfunctional, which, in our opinion, leads to the individualities of polyQ diseases.