Cleavage of Atrophin-1 at Caspase Site Aspartic Acid 109 Modulates Cytotoxicity*

Dentatorubropallidoluysian atrophy (DRPLA) is one of eight autosomal dominant neurodegenerative disorders characterized by an abnormal CAG repeat expansion which results in the expression of a protein with a polyglutamine stretch of excessive length. We have reported recently that four of the gene products (huntingtin, atrophin-1 (DRPLA), ataxin-3, and androgen receptor) associated with these open reading frame triplet repeat expansions are substrates for the cysteine protease cell death executioners, the caspases. This led us to hypothesize that caspase cleavage of these proteins may represent a common step in the pathogenesis of each of these four neurodegenerative diseases. Here we present evidence that caspase cleavage of atrophin-1 modulates cytotoxicity and aggregate formation. Cleavage of atrophin-1 at Asp109 by caspases is critical for cytotoxicity because a mutant atrophin-1 that is resistant to caspase cleavage is associated with significantly decreased toxicity. Further, the altered cellular localization within the nucleus and aggregate formation associated with the expanded form of atrophin-1 are completely suppressed by mutation of the caspase cleavage site at Asp109. These results provide support for the toxic fragment hypothesis whereby cleavage of atrophin-1 by caspases may be an important step in the pathogenesis of DRPLA. Therefore, inhibiting caspase cleavage of the polyglutamine-containing proteins may be a feasible therapeutic strategy to prevent cell death.

Dentatorubropallidoluysian atrophy (DRPLA) is one of eight autosomal dominant neurodegenerative disorders characterized by an abnormal CAG repeat expansion which results in the expression of a protein with a polyglutamine stretch of excessive length. We have reported recently that four of the gene products (huntingtin, atrophin-1 (DRPLA), ataxin-3, and androgen receptor) associated with these open reading frame triplet repeat expansions are substrates for the cysteine protease cell death executioners, the caspases. This led us to hypothesize that caspase cleavage of these proteins may represent a common step in the pathogenesis of each of these four neurodegenerative diseases. Here we present evidence that caspase cleavage of atrophin-1 modulates cytotoxicity and aggregate formation. Cleavage of atrophin-1 at Asp 109 by caspases is critical for cytotoxicity because a mutant atrophin-1 that is resistant to caspase cleavage is associated with significantly decreased toxicity. Further, the altered cellular localization within the nucleus and aggregate formation associated with the expanded form of atrophin-1 are completely suppressed by mutation of the caspase cleavage site at Asp 109 . These results provide support for the toxic fragment hypothesis whereby cleavage of atrophin-1 by caspases may be an important step in the pathogenesis of DRPLA. Therefore, inhibiting caspase cleavage of the polyglutamine-containing proteins may be a feasible therapeutic strategy to prevent cell death.
To date, eight different dominantly inherited neurodegenerative diseases have been shown to be associated with polyglutamine tract expansions in their respective proteins (1)(2)(3).
Because all of these disease-associated proteins share a similar mutation, i.e. CAG expansion in the coding region causing expansion of a polyglutamine stretch, they may have a common pathological mechanism leading to neuronal cytotoxicity. Except for the polyglutamine tract, it is generally believed that the eight disease proteins are unrelated because their amino acid sequences bear no discernible sequence homology. However, recent evidence from our laboratories (13-15) 2 suggest that seven of the eight identified polyglutamine repeat proteins involved in CAG expansion diseases contain caspase consensus cleavage sites (i.e. DXXD). This would imply that a second common feature of at least seven of the polyglutamine expansion disease proteins may be their involvement in the apoptotic cell death pathway as cellular substrates for the caspases. This finding has important implications because studies in vitro and in vivo indicate that the truncated forms of these proteins lead to the formation of intracellular aggregates, and thus caspase cleavage of the full-length proteins could in part explain how these cytotoxic truncated proteins are formed (16 -23).
To date, we have characterized the caspase cleavage of four of the polyglutamine repeat disease proteins: huntingtin, the androgen receptor, atrophin-1 (DRPLA), and ataxin-3 (Machado-Joseph disease) (13)(14)(15). These initial studies suggested that a caspase-dependent apoptotic pathway may be a critical factor in the generation of truncated proteins in some of these polyglutamine repeat disease proteins and raised a number of questions that warrant further investigation: Is the proteolytic pathway involving caspases required to explain a common mechanism of cytotoxicity of these proteins with expanded polyglutamine stretches? How are these caspase substrates involved in the apoptotic process and neurodegeneration? Are any of the functional domains within these proteins activated or inactivated by caspase cleavage? Does the proteo-lytic processing by caspases form the basis for selective neuronal loss characteristic of these neurodegenerative diseases? Does the distinct cellular localization of each of these structurally and functionally unrelated proteins determine which caspase family member cleaves them? Do other proteolytic pathways contribute to the cytotoxicity of these proteins?
Given our recent findings, we wished to address the following specific questions: Is caspase cleavage required for the cytotoxicity exhibited by these proteins? In other words, is the formation of a truncated protein containing the polyglutamine stretch via a caspase-dependent pathway required for cytotoxicity? Additionally, we wished to address whether the caspase cleavage site is required for the formation of the protein aggregates and/or altered cellular localization characteristic of these diseases. Of the four proteins we reported recently as caspase substrates, atrophin-1 is a particularly attractive candidate for our initial investigation because in all likelihood it contains only one caspase cleavage site within the entire protein. This is in contrast to ataxin-3, huntingtin, and androgen receptor, which are cleaved at multiple sites within the protein. Here we provide in vitro evidence that caspase cleavage of atrophin-1 modulates cytotoxicity, formation of protein aggregates, and its subcellular localization.

EXPERIMENTAL PROCEDURES
Culture and Transfection of Cells-Cells from the human embryonic kidney cell line 293T were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum with 1% penicillin/streptomycin. Transient transfection was carried out with pcDNA3, pcDNA3-DRPLA26, pRc/CMV-LacZ, pcDNA3-DRPLA65, pcDNA3-DRPLA26D-109N, and pcDNA3-DRPLA65 D109N (24). Preparation of the atrophin-1 constructs has been described previously (14,25). Using pRc/CMV-LacZ, transfection efficiency was determined by staining for the expression of ␤-galactosidase. Cell death was measured by trypan blue exclusion, acridine orange/ethidium bromide, and LacZ reporter gene cotransfection. Death was established as apoptotic based on acridine orange/ethidium bromide staining and assessment of caspase-3 activation. Cell death was measured 36 -50 h after transfection. Cellular death in confluent cells was induced with tamoxifen citrate at a concentration of 35 M 36 -48 h after transfection (26). Data were collected for three to five experiments and then compared by Student's t test for statistical significance. Apoptosis was also monitored with the ApoAlert caspase assay kit according to the manufacturer's instructions with the Ac-DEVD-AFC substrate (CLONTECH).
Caspase Western Analysis-Western blots were carried out as described previously using anti-caspase-3 mouse monoclonal antibody (Transduction Laboratories) (26).
Atrophin-1 Western Blotting-293T cells were transiently transfected at 40% confluence using a modified calcium phosphate protocol by mixing Qiagen-prepared DNA (Qiagen, Chatsworth, CA) with 2.5 mM CaCl 2 and 2 ϫ BBS (50 mM BES, 280 mM NaCl, 1.5 mM Na 2 HPO 4 , pH 7.0) and adding the mixture to cells immediately. After a 3-h incubation, the media were removed and replaced with fresh growth media. At 24 h post-transfection, cells were either treated with 35 M tamoxifen (Sigma) for 4 h or left untreated. Cells were harvested by gentle scraping into the growth media and centrifugation at 4,000 ϫ g for 4 min. Samples were washed once with PBS, centrifuged as before, and suspended in lysis buffer (20 mM HEPES, 5 mM MgCl 2 , 0.5 mM EDTA, 0.01% (w/v) sucrose, 1 mM phenylmethylsulfonyl fluoride, 10 mg/ml leupeptin, and 10 mg/ml aprotinin). Equal amounts of total cellular protein were mixed with 5 ϫ Laemmli sample buffer, denatured at 95°C for 5 min, and separated on 7.5% SDS-polyacrylamide gels. Protein was transferred electrophoretically to polyvinylidene difluoride membrane, immunoblotted with anti-atrophin-1 antibody, and detected using enhanced chemoluminescence (Amersham Pharmacia Biotech).
Site-directed Mutagenesis and Plasmid Construction-Human DRPLA26D109N and DRPLA65D109N were created using the QuikChange site-directed mutagenesis system from Stratagene. pcDNA3-DRPLA constructs were used as templates with the following two synthetic primers according to manufacturer instructions: 5Ј-CC-GATCTGGATAGCTTGAACGGGCGGAGCCTTAATG-3Ј and 5Ј-CATT-AAGGCTCCGCCCGTTCAAGCTATCCAGATCGG -3Ј.
Mapping of Caspase Cleavage Sites by Radiosequencing-Radiosequencing was performed as described previously (27,30). Plasmid pcDNA3-DRPLA26 was transcribed and translated with T7 polymerase using the TNT system (Promega) with either [ 35 S]methionine or [ 3 H]leucine. The translation was treated with caspase-3, separated by SDS-polyacrylamide gel electrophoresis, and electroblotted onto a polyvinylidene difluoride membrane. After autoradiography, the position of the [ 35 S]methionine-labeled atrophin-1 fragments was used to cut out the [ 3 H]leucine atrophin-1 bands from the polyvinylidene difluoride membrane. The samples were subjected to automated sequencing using an Applied Biosystems 476A sequencer, and the anilinothiazolinone derivatives in each cycle were counted in a scintillation counter. Comparison of the known positions of leucines relative to the caspase cleavage site aspartate allowed identification of the atrophin-1 cleavage site.
Immunofluorescence Microscopy-293T cells were grown on glass coverslips and transiently transfected with the indicated DRPLA construct as described above. At 36 h post-transfection, the cells were treated with 35 M tamoxifen for 45 min. After fixation in 4% paraformaldehyde and PBS solution for 20 min, the cells were washed and then permeabilized in 0.5% Triton X-100 PBS for 5 min. The DRPLA antibody utilized in these studies has been described by Wood et al. (31) and was raised in rabbits against synthetic peptide DRPLA425 (residues 425-439 of atrophin-1). The cells were washed twice, incubated at room temperature with anti-DRPLA antibody (1:200) for 1 h, washed three times with PBS, and then incubated in Texas red-conjugated antirabbit antibody (1:1,000) for 20 min. Cells were washed three times with PBS and then mounted onto slides with DAPI (4Ј,6Ј-diamindino-2-phenylindole, Sigma, 0.05 g/ml) in 90% glycerol and PBS as a nuclear counterstain. Immunofluorescence was observed using a Zeiss confocal microscope. Control experiments were performed, including incubation with secondary antibody only, and immunofluorescence of cells transfected with control plasmids.

RESULTS AND DISCUSSION
Analysis of the Atrophin-1 Caspase Cleavage Site-DRPLA is one of eight autosomal dominant neurodegenerative diseases with expansion of CAG trinucleotide repeats encoding polyglutamine stretches (32). This neurodegenerative disorder is characterized by progressive dementia, myoclonic epilepsy, cerebellar ataxia, and choreoathetotic movements. Like many of these disease-associated proteins, atrophin-1 is expressed ubiquitously in the central nervous system (33), and thus its expression pattern offers little clue to the relative susceptibility or resistance of certain neuronal populations of cells to undergo neurodegeneration.
We and others have demonstrated recently that atrophin-1 is cleaved by caspases (14,34). Atrophin-1 is one of at least 40 cellular caspase substrates identified, and its function, as well as its contribution to the apoptotic process, is unknown. Atrophin-1 contains a consensus caspase-3 cleavage site (14,35) near the NH 2 terminus of the protein ( 106 DSLD 109 ) (Fig. 1A), and the polyglutamine tract is located in the middle of the protein (Fig. 1A). The cleavage products generated during caspase-3 cleavage of in vitro translated atrophin-1 migrated at 145 and 150 kDa for constructs with 26 and 65 glutamines, respectively (Fig. 1B, lanes 2 and 6). These COOH-terminal fragments contain the polyglutamine tract and would be expected to lack the predicted nuclear targeting sequence located at the NH 2 terminus of the protein at amino acids 16 -32 (see Fig. 1B). NH 2 -terminal sequencing of the DRPLA cleavage product confirmed that atrophin-1 was cleaved at Asp 109 .
To analyze the functional significance of caspase cleavage we prepared constructs of DRPLA without a caspase cleavage site. Mutation of the caspase P1 residue in atrophin-1 from Asp 109 to Asn abolished the processing of the in vitro translated atro-phin-1 (Fig. 1B, lanes 4 and 8) by caspases. Because we have shown previously that atrophin-1 can be cleaved by caspase-1, caspase-7, and caspase-8 in addition to caspase-3, we evaluated whether the Asp 109 mutation abolished cleavage by multiple caspases present in transfected cells by Western blotting (Fig.  1C). Caspase cleavage products were observed in 293T cells transiently transfected with DRPLA26 or DRPLA65 after tamoxifen treatment, showing that tamoxifen challenge induced caspase activation in cells transfected with these constructs. In contrast, no cleavage products were generated in cells transfected with DRPLA26D109N or DRPLA65D109N after tamoxifen treatment. These results show that we have prepared atrophin-1 proteins resistant to caspase cleavage in vitro and in transfected cells. Therefore, we utilized these constructs to test whether this site influences the cytotoxicity of atrophin-1 in cell culture.
Increased Cytotoxicity of DRPLA Mutant Protein-We have recently developed an in vitro tissue culture model to investigate the cellular toxicity of polyglutamine repeat expansion disease proteins (15,17). In this system, a sublethal stress is induced by tamoxifen in transiently transfected human embryonic kidney 293T cells. A sublethal stress in these studies is defined as a stress delivered by a concentration of a pro-apoptotic agent that does not result in the processing of caspase-3 in plasmid control transfected cells. As shown in Fig. 2A, treatment of 293T cells with tamoxifen at a concentration of 35 M does not lead to processing of caspase-3 over a 24-h period of time. Higher concentrations of tamoxifen (Ͼ40 M) result in apoptotic cell death based on acridine orange/ethidium bromide staining of the cells (data not shown) and the processing of caspase-3 (Fig. 2B).
To investigate the cytotoxicity of atrophin-1, we transiently Interestingly, the expression of DRPLA26 is pro-apoptotic when exposed to an apoptotic stress. Because the function of atrophin-1 protein is unknown, it is difficult to speculate how overexpression of normal atrophin-1 enhances cellular death. However, there is a growing body of literature suggesting that many of the caspase substrates can act to enhance or block apoptotic cell death upon cleavage by caspases. For example, cleavage of presenilin-2 results in the generation of an antiapoptotic cleavage product (36). In contrast, expression of the caspase substrate mitogen-activated protein kinase kinase results in the generation of a pro-apoptotic fragment that enhances caspase activation through a positive feedback loop (27). Furthermore, we have reported recently that the truncated fragment of Huntington disease containing the normal polyglutamine repeat is pro-apoptotic (17). Our results demonstrate that DRPLA65 is more pro-apoptotic than DRPLA26, which indicates that the gain of function related to CAG length may influence downstream events in apoptosis.
The Pro-apoptotic Effect of Atrophin-1 Requires Cleavage-Next, we assessed the effect of blocking caspase cleavage of DRPLA26 and DRPLA65 on the pro-apoptotic effects of these proteins in culture. As shown in Fig. 3  is crucial to the pro-apoptotic effect of atrophin-1, we investigated whether apoptosis induction with tamoxifen modulated the formation of aggregates. Intracellular neuronal inclusions may be a common property for glutamine repeat expansion diseases (37). Aggregates have been reported recently in the brains of patients with DRPLA and are similar to those ob-served in Huntington disease (16,20,38). Immunofluorescence analysis of transiently transfected 293T cells or COS-7 cells indicated that atrophin-1 is localized at the outer border of the nucleus, consistent with its putative nuclear localization signal (Fig. 4A). This result differs from earlier work that has suggested that atrophin-1 is localized to the cytoplasm (20, 39). Some of these earlier studies used epitope-tagged proteins in which the epitope may have interfered with the nuclear targeting signal. Expression of DRPLA65, but not DRPLA26, led to formation of densely stained nuclei (granular in appearance) with altered nuclear distribution upon apoptosis induction compared with the controls (atrophin-1-transfected cells not treated with tamoxifen), as determined by confocal microscopy (Fig. 4, A and C). Normal atrophin-1 protein is localized to the outer edge of the nuclei before and after apoptotic stimulation, whereas the disease-associated form of atrophin-1 leads to dense, particulate staining throughout the nuclei during apoptotic stimulation. Further analysis (Fig. 5) of the diseaseassociated form of atrophin-1 demonstrates that atrophin-1 colocalizes with DAPI-stained nuclei during apoptotic stimulation with tamoxifen (Fig. 5, A-C), and confocal images under high magnification show that the aggregates are nuclear (Fig.  5, D-F). Thus, the disease-associated form of atrophin-1 shows an altered pattern of nuclear distribution compared with the normal form of atrophin-1. Modulation of aggregate formation did not occur in the absence of apoptosis induction for DR-PLA65-transfected cells (data not shown) during the first 48 h after transfection. Interestingly, longer periods of time after transfection (Ͼ72 h) resulted in increased formation of cyto-plasmic aggregates. These aggregates were larger in the DR-PLA65-transfected cells and occurred with higher frequency (Fig. 6).
Because aggregate formation was modulated by the stimulation of apoptosis, we next assessed whether mutation of the caspase cleavage site in DRPLA65 blocks formation of aggregates. Cells transfected with pcDNA3-DRPLA26 expressed atrophin-1 protein that was localized to the outer edge of the nucleus with a homogeneous pattern during apoptotic cell death with tamoxifen (Fig. 4A). Cells transfected with pcDNA3-DRPLA65 showed granular dense straining throughout the nucleus (Fig. 4C) during apoptotic cell death. In sharp contrast, cells transfected with pcDNA3-DRPLA65D109N did not show altered nuclear distribution and granular dense staining, suggesting that, at least in this system, caspase cleavage of DRPLA is required for aggregation (Fig. 4, B and D). It is of relevance to compare our results with those found for SCA1, given that it is also a nuclear protein yet does not appear to be cleaved by caspases. Interestingly, recent studies on SCA1 show that the subcellular localization of wild-type ataxin-1 differs from the mutant ataxin-1 both in vitro and in vivo. Wild-type ataxin-1 localizes to the nucleus in COS-1 cells (40), whereas mutant ataxin-1 shows a specific redistribution or disruption of the nuclear structure. In these studies there was no evidence that apoptosis modulated the formation of aggregates, which is consistent with our finding that ataxin-1 is not a caspase substrate. 2 Ataxin-1 redistribution may be important for the pathogenic mechanism in this disease, and additional studies will determine whether this is similarly the case for atrophin-1.
In this study, we show that cells transfected with expression constructs encoding atrophin-1 undergo enhanced apoptotic cell death that is mediated by a pro-apoptotic caspase cleavage product. In vitro mutagenesis of caspase cleavage site Asp 109 blocks production of this pro-apoptotic fragment and reduces cellular toxicity dramatically. This is consistent with our recent findings that expression of truncated huntingtin fragments resulted in significantly more cell death than the full-length huntingtin (17,22). Our results suggest that caspase cleavage is required for modulation of aggregate formation but does not determine whether aggregation is required for cellular toxicity. Recent work on ataxin-1 indicates that nuclear localization is critical for pathogenesis but not aggregation (41).
The results described here, along with our recent work (13)(14)(15)17), suggest that one common feature shared among at least seven of the polyglutamine repeat disease proteins is that they are cleaved by caspases to produce pro-apoptotic fragments. In this model, initial cleavage by caspases or other proteases would produce a toxic fragment with a gain of toxic function, e.g. aggregation or altered protein-protein interactions. This generation of a toxic fragment would lead to increased activation of caspases through a feedback loop. In other words, the toxic fragments may function as caspase amplifiers. This amplification loop would be highly dependent upon the cellular context such as caspase/inhibitor distribution within the cell as well as protein-protein and/or protein-ligand interaction with each type of polyglutamine repeat protein (14). Further, amplification would also be dependent upon the ability of a particular cell type to evoke proteolytic pathways that remove this toxic caspase-amplifying fragment. These experiments have not addressed what is the initial trigger for caspase activation in the disease process but suggest that proteolysis is important for cytotoxicity. However, physiological stresses that are otherwise sublethal may in the presence of a caspase amplification mechanism lead to cell death. In addition, the results in this study do not exclude additional mechanisms for proteolytic FIG. 6. Immunofluorescence of atrophin-1 normal (DRPLA26) and disease causing (DRPLA65) 3 days after transfection in 293T cells without apoptotic stimulation. Cells expressing DR-PLA65 (bottom panel) have cytoplasmic aggregates (arrows) that are found at much higher frequency and are larger than those found in cells expressing DRPLA26 (top panel). Images were collected with a confocal microscope.
cleavage of atrophin-1 or the other CAG-containing gene products generating a smaller toxic fragment.
Furthermore, the results do not offer an explanation for the specific pattern of neuronal loss in CAG repeat diseases. It is possible that alterations in caspase expression, caspase inhibitors, partner proteins, or downstream targets may determine the selective vulnerability for each of the CAG repeat diseases. Further studies directed at identifying the specific caspase family members that process the CAG repeat disease proteins, as well as a study of the regional specificity of the caspases in the brain, should shed light on this question. Finally, because blocking the cleavage of atrophin-1 inhibits its pro-apoptotic effect, such a strategy may prove useful for the treatment of neurodegenerative diseases associated with polyglutamine repeat expansions. Extension of these results to animal models of polyglutamine expansion diseases should prove useful.