Pharmacological induction of heat shock proteins ameliorates toxicity of mutant PKCγ in spinocerebellar ataxia type 14

Amyloid and amyloid-like protein aggregations are hallmarks of multiple, varied neurodegenerative disorders, including Alzheimer's and Parkinson's diseases. We previously reported that spinocerebellar ataxia type 14 (SCA14), a dominant-inherited neurodegenerative disease that affects cerebellar Purkinje cells, is characterized by the intracellular formation of neurotoxic amyloid-like aggregates of genetic variants of protein kinase Cγ (PKCγ). A number of protein chaperones, including heat shock protein 70 (Hsp70), promote the degradation and/or refolding of misfolded proteins and thereby prevent their aggregation. Here, we report that, in various SCA14-associated, aggregating PKCγ variants, endogenous Hsp70 is incorporated into aggregates and that expression of these PKCγ mutants up-regulates Hsp70 expression. We observed that PKCγ binds Hsp70 and that this interaction is enhanced in the SCA14-associated variants, mediated by the kinase domain that is involved in amyloid-like fibril formation as well as the C2 domain of PKCγ. Pharmacological up-regulation of Hsp70 by the Hsp90 inhibitors celastrol and herbimycin A attenuated the aggregation of mutant PKCγ in primary cultured Purkinje cells. Up-regulation of Hsp70 diminished net PKCγ aggregation by preventing aggregate formation, resulting in decreased levels of apoptotic cell death among primary cultured Purkinje cells expressing the PKCγ variant. Of note, herbimycin A also ameliorated abnormal dendritic development. Extending our in vitro observations, administration of celastrol to mice up-regulated cerebellar Hsp70. Our findings identify heat shock proteins as important endogenous regulators of pathophysiological PKCγ aggregation and point to Hsp90 inhibition as a potential therapeutic strategy in the treatment of SCA14.

Many neurodegenerative disorders, including Alzheimer's disease, Parkinson's disease, and polyglutamine disease, are characterized by the accumulation of intracellular or extracellular protein aggregates. These aggregates, despite the heterogeneity of their component proteins, typically contain fibrillar amyloid-like structures with common characteristics, such as cross-␤-structure and the ability to bind lipophilic dyes such as Congo red (1). Whether fibrillar aggregates are pathogenic, inert, or protective in neurodegenerative diseases still remains controversial. Yet, previous studies have revealed that these aggregates are commonly immunoreactive for molecular chaperones, such as heat shock proteins (Hsps), 4 that facilitate protein folding, assembly, and degradation. Molecular chaperones are potent suppressors of neurodegeneration when overexpressed or induced pharmacologically and therefore represent promising therapeutic targets for amyloid-related neurodegenerative disorders (2).
Autosomal dominant spinocerebellar ataxias (SCAs) are a heterogeneous group of neurodegenerative disorders that are characterized by cerebellar atrophy and progressive ataxia. These disorders can be classified into at least 43 subtypes based on different genetic loci of the causal gene (3). Specifically, SCA14 is caused by a missense mutation or deletion in the PRKCG gene, which encodes protein kinase C␥ (PKC␥), a serine/threonine kinase that is highly expressed in the central nervous system and is especially abundant in cerebellar Purkinje cells (PCs) (4 -6). More than 30 different PKC␥ mutations have been identified in patients with SCA14. Yet, the mechanisms by which mutant PKC␥s cause cerebellar atrophy and PC degeneration in SCA14 are unclear. It is known that SCA14associated mutations increase aggregation of PKC␥ in mouse cerebellar PCs in vivo and in primary cultured PCs as well as cultured cell lines in vitro (7)(8)(9)(10)(11). Additionally, PKC␥ aggregates form amyloid-like fibrils in vitro (12,13). Thus, it has been hypothesized that the amyloid-like fibril formation of mutant PKC␥ is involved in the pathogenesis of SCA14, similar to that of other neurodegenerative disorders.
We recently found that mutant PKC␥ up-regulated Hsp70 and that knockdown of endogenous Hsp70 exacerbated mutant PKC␥ aggregate formation in primary cultured PCs and cultured neuronal cell lines (14). Given the known role of Hsp70 in neurodegenerative diseases, we hypothesized that Hsp70 might be a therapeutic target against SCA14. In this study, we sought to extend our previous observations by evaluating the effects of Hsp induction in primary cultured PCs expressing mutant PKC␥.

Incorporation and up-regulation of Hsp70 caused by aggregation of mutant PKC␥
To date, 32 point or deletion mutations of PKC␥ have been associated with SCA14; of these, 6 and 19 are located in the C1A and C1B diacylglycerol-binding domains, respectively. Additionally, several mutations in the C2 Ca 2ϩ -binding domain and kinase domain of PKC␥ have been reported (Fig. 1A) (12,(15)(16)(17)(18)(19)(20). Our previous study showed that Hsp70 (HSPA1A) was incorporated into aggregates of the C1B domain mutants S119P and G128D and that the knockdown of endogenous Hsp70 exacerbated mutant PKC␥-induced cytotoxicity (14). To build on this initial finding, we examined whether the aggregates of various mutants, with mutations located in different domains of PKC␥, were also immunoreactive for Hsp70, using SH-SY5Y cells. This neuroblastoma cell line has been used to study the aggregation of mutant PKC␥ in previous studies, and similar results have been obtained in this cell line and in primary cultured Purkinje cells (7,13,21). The immunofluorescence analysis showed that all the mutant (R41P, Q127R, V138E, and F643L) PKC␥green fluorescent proteins (GFPs) tested formed aggregates that co-localized with endogenous Hsp70, whereas most wildtype (WT) PKC␥-GFPs showed diffuse cytoplasmic expression (Fig. 1B). Consistent with a previous study (14), the up-regulation of endogenous Hsp70 in cells expressing mutant PKC␥s was also observed, and Hsp70 was more highly expressed in cells containing aggregates than in cells without aggregates (Fig. 1C). In contrast, Hsc70 (HSPA8), which is a member of the Hsp70 family and is known as a constitutively expressed chaperone, was also incorporated into the PKC␥ aggregates, although its endogenous expression levels were not altered by the PKC␥ aggregates (Fig. 1D). Up-regulation of Hsp70 was evident at 24 h after overexpression of V138E PKC␥-GFP and was significantly enhanced at 48 h (187 Ϯ 44%, compared with 0 h, n ϭ 4, p Ͻ 0.01) in SH-SY5Y cells, although the expression level of Hsc70 was unaffected (Fig. 1E). Thus, endogenous Hsp70 chaperones were incorporated into aggregates of various SCA14-associated PKC␥ mutants, and Hsp70, but not Hsc70, expression was significantly up-regulated by mutant protein expression.

Interaction of PKC␥ with Hsp70 via the C2 and kinase domains is facilitated by a SCA14 mutation
We next tested whether WT and mutant PKC␥ physically interact with Hsp70. Co-immunoprecipitation (IP) experi-ments using COS-7 cells overexpressing FLAG-WT PKC␥, FLAG-V138E PKC␥, and Myc-Hsp70 showed that the Myc-Hsp70 was co-immunoprecipitated with FLAG-WT PKC␥ and that the co-IP was significantly enhanced by the SCA14-associated V138E mutation (Fig. 1, F and G). It is known that Hsp70 preferentially binds to extended polypeptides with exposed hydrophobic residues (22). Interestingly, disorder propensity calculated by IUPred2A (https://iupred2a.elte.hu/) (23,24), based on the amino acid sequence of PKC␥, indicated a relatively high disorder propensity around C2 and the C terminus of KD compared with C1A and C1B (Fig. 1H). These results suggested that C2 and KD may have high tendencies to unfolded for transient periods so that Hsp70 can readily bind to PKC␥ through these domains. In fact, co-IP experiments using FLAG-tagged isolated domains (C1A, C1B, C2, and KD) of PKC␥ showed that Myc-Hsp70 was significantly co-immunoprecipitated with FLAGtagged C2 and KD but not FLAG-C1A or -C1B (Fig. 1I). These results suggested that PKC␥ binds Hsp70 and that the interaction is promoted by a SCA14-associated mutation and mediated by the C2 domain of PKC␥ as well as the kinase domain, which is involved in amyloid-like fibril formation (12).

Overexpression of Hsp70 reduces the aggregation of mutant PKC␥
We previously identified WT PKC␥ as a novel amyloidogenic protein that forms amyloid-like fibrils, and we found that SCA14-associated mutations accelerated amyloid-like fibril formation in cultured cells (12). Therefore, we next examined whether Hsp70 reduces the aggregation of PKC␥. We assessed CHO and SH-SY5Y cells co-expressing WT or mutant PKC␥-GFPs and Hsp70 or its inactive form (Hsp70 -T13G (25, 26)) using fluorescence microscopy, and we counted the numbers of aggregate-containing cells. Overexpression of Hsp70 significantly reduced aggregates of both WT and V138E PKC␥-GFP in CHO (Fig. 2, A and B) and SH-SY5Y cells (Fig. 2C), whereas that of Hsp70 -T13G had no significant effects. Although the anti-aggregation effect of Hsp70 was observed in both cell lines, the percentage of SH-SY5Y cells with aggregates was notably lower than that of CHO cells.
To address this issue, we performed an immunoblot analysis and examined the protein expression levels of endogenous Hsps in each cell line. Interestingly, although all the Hsps we tested were expressed in SH-SY5Y cells, the immunoreactive bands of Hsp70 and Hsp27 were below the detectable levels in CHO cells under basal conditions (Fig. 2, D and E). Moreover, Hsp70 up-regulation by overexpression of V138E PKC␥-GFP was not observed in CHO cells (Fig. 2F). Thus, it is plausible that the aggregation levels of mutant PKC␥-GFP are inversely associated with the expression level of endogenous Hsps. We next examined the dominant-negative effect of Hsp70 using Hsp70 -K71E, which exhibits intact substrate association, but weakens ATP binding (27), on aggregate formation in SH-SY5Y cells. Consistent with the results from a previous study that included knockdown experiments (14), Hsp70 -K71E significantly exacerbated the aggregation of various mutants, but not WT PKC␥-GFP, in SH-SY5Y cells (Fig. 2G). Together, these results strongly suggested that Hsp70 is critical for preventing the aggregation of mutant PKC␥.

Hsp cytoprotection in SCA14
Overexpression of Hsp70 ameliorates mutant PKC␥ cytotoxicity We previously showed that mutant PKC␥ overexpression induces cell death (9). Therefore, in this study, we examined

Hsp cytoprotection in SCA14
were detached from the cultured dishes (12), floating cells in the medium as well as attached cells on the dishes were harvested for the analysis. The cells were stained with 7-aminoactinomycin D (7-AAD), a fluorescent DNA dye that selectively stains the DNA of dead cells (28). Stained cells were assessed using flow cytometry and classified in accordance with GFP and 7-AAD fluorescence. Because mutant PKC␥-GFP-induced cell death has been observed in parallel with its expression level (9), cells with relatively high GFP fluorescence were extracted to examine the effect of Hsp70 overexpression. Cells expressing Hsp70 exhibited significant reduction in cell death caused by various mutant PKC␥-GFPs (Fig. 2H), whereas the co-expression of Hsp70 -T13G, an inactive form of Hsp70, had no effect on mutant PKC␥ cytotoxicity (Fig. 2I), suggesting that the chaperon activity of Hsp70 is required to protect against cell death caused by SCA14-associated mutants. Taken together, we found that inhibition of PKC␥ aggregation by Hsp70 ameliorated mutant PKC␥-induced cell death in CHO cells.

Hsp cytoprotection in SCA14
PKC␥-GFP (Fig. 3B). To verify the effects of incorporated Hsps, Myc-Hsp27, Myc-Hsp40, or Myc-Hsp70 was co-expressed with various mutant PKC␥-GFPs in CHO cells, and percentages of cells containing aggregates were counted. Similar to Hsp70, overexpression of Hsp40 reduced the aggregation of several mutant PKC␥-GFPs (Fig. 3C). Conversely, Hsp27 had no effect. Therefore, we confirmed the incorporation of endogenous Hsp70 and Hsp40 into PKC␥ aggregates using primary cultured PCs. The PC-specific expression of G128D PKC␥-GFP (G128D-GFP) was achieved with adenoviral infection of Ad-L7-tTA (tetracycline-controlled transactivator), which expresses tTA in a PC-specific manner, and Ad-TetOP-PKC␥-GFP on days in vitro 21 (DIV 21) (29); G128D is a C1B domain mutant that forms aggregates with high frequency (9). As shown in Fig. 3D, both Hsp70 and Hsp40 were incorporated into the aggregates in calbindin-positive PCs. Although Myc-Hsp40 did not show direct interaction with FLAG-PKC␥ in a co-IP assay ( Fig. 3E), Hsp40 is known as a co-chaperone of Hsp70, and the interaction between Hsp40 and Hsp70 has been previously reported (30). Therefore, Hsp40 may have indirectly interacted with PKC␥ aggregates through Hsp70. We next examined whether Hsp70 and Hsp40 exhibit synergistic effects on the aggregation. As shown in Fig. 3F, the co-expression of Myc-Hsp40 and Myc-Hsp70 in CHO cells produced a dramatic suppression of mutant PKC␥ aggregation. Among Hsps, both Hsp70 and Hsp40 are key components for diminishing aggregation of mutant PKC␥.

Overexpression or pharmacological induction of Hsps increases the detergent solubility of mutant PKC␥
It has been reported that detergent-insoluble mutated proteins are generated in the cellular and animal models of various neurodegenerative diseases (31)(32)(33). We have also shown that SCA14-associated mutations in PKC␥ increase its insolubility in Triton X-100, a nonionic detergent (9). Thus, we next examined whether Hsps affect the detergent solubility of mutant PKC␥. Transfected CHO cells were separated into 1% Triton X-100 -soluble and -insoluble fractions, and both fractions were separated using SDS-PAGE under nonreducing conditions. Western blot analysis revealed the presence of highmolecular weight (HMW) species (Ͼ150 kDa) as well as SDSsoluble monomeric FLAG-Q127R PKC␥ in the Triton X-100insoluble fraction (Fig. 4A), suggesting amyloid-like fibril formation by mutant PKC␥, as reported previously (12). The amounts of whole-monomeric and HMW FLAG-Q127R PKC␥ in both fractions were quantified by immunoblotting with anti-FLAG antibody, and the percentage distributions of monomeric and HMW PKC␥ in the Triton X-100 -soluble and -insoluble fractions were determined (Fig. 4, B and C). Overexpression of Myc-Hsp70 or Myc-Hsp40 alone significantly reduced HMW PKC␥ immunoreactivity in the Triton X-100insoluble fraction. Moreover, co-expression of Myc-Hsp70 and Myc-Hsp40 dramatically reduced HMW PKC␥ in the Triton X-100 -insoluble fraction and increased monomeric and HMW PKC␥ in the Triton X-100 -soluble fraction, consistent with the results shown in Fig. 3E. These results indicated that co-expression of Hsp40 and Hsp70 synergistically improved the detergent solubility of mutant PKC␥ and reduced its aggregation.
Celastrol, a member of the triterpenoid compound family, extracted from the root of Tripterygium wilfordii Hook F (Thunder of God Vine), is a known Hsp90 inhibitor that activates heat shock transcription factor 1 (HSF1), a major transcription factor for Hsps (34,35). It has been reported that celastrol treatment induces Hsps and reduces aggregation and cell toxicity in a cell culture model of polyglutamine disease (36). Therefore, we asked whether the up-regulation of Hsps by celastrol could alleviate mutant PKC␥ aggregation in SH-SY5Y cells. Celastrol treatment (3 M, 3 h) up-regulated endogenous Hsp70 and Hsp40 expression and reduced monomeric and HMW PKC␥ in the Triton X-100 -insoluble fraction of Q127R PKC␥-GFP compared with dimethyl sulfoxide (DMSO) treatment (Fig. 4, D-G). These results indicated that Hsp70 and Hsp40 up-regulation by celastrol alleviated the aggregation of mutant PKC␥ in vitro.

Hsp90 inhibition induces Hsp70 expression in PCs in vitro, and celastrol up-regulates Hsp70 in the mouse cerebellum in vivo
Although previous studies have shown that treatment with celastrol and herbimycin A, a benzoquinone ansamycin inhibitor of Hsp90, induces Hsp expression in various in vitro and in vivo models (36 -38), it is not known whether the treatment also induces Hsps in PCs of the mouse cerebellum. Thus, we examined the effect of celastrol and herbimycin A treatment on Hsp expression in PCs in vitro. Immunoblot analyses showed that both celastrol (3 M for 1 h, after which the cells were cultured for 23 h without celastrol) and herbimycin A (2 M for 24 h) treatments successfully up-regulated Hsp70 in primary cultured cerebellar cells (Fig. 5, A and B). In contrast to the results from SH-SY5Y cells (Fig. 4, D and E), Hsp40 expression was not up-regulated by Hsp90 inhibitors. Moreover, Hsc70 expression was not altered by Hsp90 inhibitors. To confirm the up-regulation of Hsp70 in PCs, celastrol-treated (3 M, 1 h/day, for 4 days) or herbimycin A-treated (2 M for 4 days) PCs were assessed using fluorescence microscopy. Both Hsp90 inhibitors enhanced Hsp70 expression in calbindin-positive PCs (Fig. 5C). We next tested the brain permeability of intraperitoneally administered celastrol and herbimycin A in C57BL/6N mice. Notably, Hsp70, but not Hsp40, was upregulated in cerebellar lysates by celastrol treatment (1 mg/kg) (Fig. 5D), although herbimycin A (2 mg/kg) had no effect on Hsp expression at 2 weeks (Fig. 5E). These results indicate that Hsp90 inhibition by celastrol or herbimycin A induced Hsp70 expression in PCs in vitro, and celastrol penetrated the mouse cerebellum and up-regulated Hsp70 in vivo.

Pharmacological Hsp70 induction inhibits aggregation in PCs expressing mutant PKC␥
We investigated the ability of celastrol or herbimycin A treatment to diminish mutant PKC␥ aggregation in PCs. The PCspecific expression of G128D-GFP was achieved with adenoviral infection on DIV 21. Drug treatments were started on DIV 26 when most PCs expressing mutant PKC␥s already contained

Hsp cytoprotection in SCA14
aggregates. After treatment with celastrol (3 M, 1 h/d) or herbimycin A (2 M, 24 h) for 4 days, PCs were fixed and stained with antibody against calbindin and assessed using fluorescence microscopy (Fig. 6A). To assess the intensity of aggregates, GFP fluorescence was converted into rainbow pseudocolor. As shown in Fig. 6B, aggregates were dispersed, and the intensities were lowered after the Hsp90 inhibitor treatments.
To quantify the effects of drug treatment on aggregate formation, the numbers of cells containing high-intensity aggregates, which appear as white spots (intensity exceeding 250 arbitrary units (a.u.)) in calbindin-positive PCs, were counted, and the percentages of cells containing high-intensity aggregates were calculated. Both celastrol and herbimycin A treatments significantly diminished the high-intensity aggregates in PCs (Fig. 6C).

Hsp cytoprotection in SCA14
Because celastrol is known to induce autophagy (39,40), we examined whether Hsp90 inhibitors induce activation of autophagy to enhance the aggregate clearance in primary cultured cerebellar cells. Autophagic activation was assessed by the level of LC3-II, a membrane-bound form of LC3 as a marker of autophagic activity (41). As a positive control for LC3-II detection, chloroquine, which is an autophagy inhibitor, was used to accumulate LC3-II. In agreement with previous reports, immunoblot analysis showed that chloroquine as well as celastrol significantly increases LC3-II levels, whereas herbimycin A has no effect (Fig. 6D). Thus, although in the case of celastrol the autophagic activation may also contribute to the clearance of the PKC␥ aggregates, herbimycin A treatment does not appear to induce autophagy in primary cultured cerebellar cells.

Hsp cytoprotection in SCA14 Aggregation of mutant PKC␥ is a dynamic process regulated by Hsp70
To examine in more detail the diminishment of aggregated PKC␥ by pharmacological enhancement of Hsp70 expression, we performed long-term time-lapse imaging of primary cerebellar cultures in which G128D-GFP was overexpressed by adenoviral infection under the PC-specific L7 promoter. GFP fluorescence was assessed hourly for 51 h, beginning 5 days after infection. We imaged 7-10 fields in each of the cultures, which were treated with either DMSO alone (vehicle control) or herbimycin A. Herbimycin A treatment was used to avoid the induction of autophagic activity. Remarkably, we observed dynamic formation and dissolution of aggregates. In control cultures, over the course of observation, aggregates were observed in 78.1% of the cells, including cells in which aggregates formed (51.2%), and in which aggregates formed and disassembled (19.0%) or formed, disassembled, and re-formed (7.9%); aggregates were lost in some cells that initially included aggregates (i.e. at time 0 h), and aggregation in these cells could be restored over time (Fig. 6, E and F and supporting Movie S1). Thus, there is a dynamic equilibrium between disaggregated and aggregated mutant PKC␥, which may reflect, under these conditions, an influence of endogenous WT Hsp. In cultures treated with herbimycin A, the number of cells containing aggregates over the time course of 51 h was diminished (61.9% versus 78.1%; not significant), which included cells in which aggregates formed (19.0% versus 51.2%; significant) and in which aggregates formed and disassembled (28.9% versus 19%; not significant), and formed, disassembled, and re-formed (14.0% versus 7.9%, not significant), indicating that the induction of Hsp70 prevents the formation of aggregates (Fig. 6F). These results are consistent with the end-point measurements presented in Fig. 6, B and C.
To further demonstrate a role of Hsp70 in the dynamic process of aggregation, we examined the effect of a cell-permeable Hsp70 inhibitor, PES-Cl, on herbimycin A-treated PCs. In detail, to induce endogenous Hsp70 expression, PCs were treated with herbimycin A for 8 h before addition of PES-Cl (2 M) or DMSO. GFP fluorescence was then assessed hourly for 46 h. In cultures treated with herbimycin A and PES-Cl, the number of cells that formed aggregates was significantly enhanced (Fig. 6G). Thus, drug-induced up-regulation of Hsp70 is able to shift the dynamic equilibrium governing amyloid formation toward the cytosolic form of PKC␥.

Herbimycin A attenuates the impairment of dendritic development and inhibits apoptosis in PCs expressing mutant PKC␥-GFP
We investigated the ability of celastrol or herbimycin A treatment to ameliorate dysfunction induced by mutant PKC␥ expression in PCs. Consistent with a previous study, WT PKC␥-GFP exhibited ubiquitous GFP fluorescence in the cytoplasm and dendrites, whereas G128D-GFP formed aggregates in the soma and was associated with abnormal dendritic development (Fig. 7A) (7). We next examined dendritic development, as assessed by dendritic length, number of nodes, and number of terminals, in PCs. Herbimycin A treatment significantly improved the abnormal dendritic development induced by G128D-GFP in PCs but had no effect on dendritic parameters in PCs expressing WT PKC␥-GFP (Fig. 7, B-E). In contrast, celastrol treatment failed to improve dendritic development in PCs expressing G128D-GFP and elicited dendritic regression in PCs expressing WT PKC␥-GFP.
Finally, we investigated the effect of celastrol and herbimycin A treatments on PC apoptosis induced by mutant PKC␥-GFP expression. Apoptosis was assessed by chromatin condensation or fragmentation in calbindin-positive PCs (Fig. 7F) (7,13). The ratio of apoptotic PCs expressing G128D-GFP was significantly higher than that of PCs expressing WT PKC␥-GFP (Fig.  7G). Additionally, both celastrol and herbimycin A treatments significantly reduced the number of apoptotic PCs expressing G128D-GFP but had no effects on the PCs expressing WT PKC␥-GFP (Fig. 7G). Taken together, both celastrol and herbimycin A induced Hsp70 expression and diminished the aggregation of mutant PKC␥, which ameliorated the apoptosis induced by mutant PKC␥ in primary cultured PCs. Furthermore, herbimycin A restored abnormal dendritic development in these PCs.

Discussion
In this study, we examined strategies for alleviating the neurotoxic effects of mutant PKC␥ aggregation in an in vitro model of SCA14, and we found that pharmacological induction of Hsp70 ameliorated mutant PKC␥ toxicity. The induction of Hsp70 by the Hsp90 inhibitors celastrol and herbimycin A inhibited mutant PKC␥ aggregation and reduced apoptosis in PCs. Herbimycin A also improved abnormal dendrite development in PCs. Celastrol, but not herbimycin A, induced Hsp70 expression in the mouse cerebellum in vivo and thus represents a potential therapeutic strategy against SCA14.

Hsp cytoprotection in SCA14
Mechanisms alleviating cytotoxic mutant PKC␥ aggregation Exposed hydrophobic regions of misfolded proteins cause protein aggregation and are typical recognition surfaces for Hsps (22). Here, we showed that PKC␥ physically interacted with Hsp70, and this interaction was enhanced by a SCA14associated mutation. Although Hsp70 interacted with the isolated C2 as well as the kinase domain of PKC␥, the aggregation propensity of the kinase domain, calculated by the Zyggregator method, was much higher than that of the C2 domain (12). This discrepancy may be rationalized by the observation that the amino acid sequences in the C2 and kinase domains have higher intrinsic disorder propensity than those in the C1A and C1B domains. The data suggested that, when these domains are isolated, the C2 and kinase domains become unstable due to a lack of inter-domain interactions that are otherwise present in fulllength PKC␥. Moreover, these domains tend to assume an unfolded structure with exposed hydrophobic regions that are recognizable by Hsp70. Conversely, WT C1A and C1B may assume globular conformations in which the hydrophobic regions are buried. Under basal conditions, full-length PKC␥ displays an inactive, closed conformation; the kinase domain is masked by regulatory domains, including C1A, C1B, and C2, such that the hydrophobic regions of these domains remain inside the molecule. Upon cellular stimulation, Ca 2ϩ and diacylglycerol bind the regulatory domain and induce conformational change into an open structure with exposed kinase and (potentially) C2 domains for activation (42). Although the activation of WT PKC␥ is tightly regulated by these second messengers, most SCA14-associated mutants exhibit a constitutively active phenotype (10,43,44). Moreover, Jezierska et al. (11) used FRET-fluorescence lifetime imaging microscopy in living cells to verify that V138E PKC␥, a C1B domain mutant, has an exposed C terminus. These results indicate that SCA14 mutations destabilize the closed conformation of PKC␥, causing the exposure of each domain. The loss of intramolecular inter-domain interactions caused by the formation of an active open structure in SCA14-associated mutants may facilitate the binding of Hsp70 to unfolded C2 and kinase domains. Notably, Hsp70 has been previously reported to associate with the C terminus of PKC␤II, an isoform of the conventional PKC; specifically, Hsp70 binds to a leucine residue that is conserved across PKCs to protect PKC␤II against down-regulation (42,45). Because this leucine residue (Leu-654) is located in the kinase domain of PKC␥, exposure of this domain caused by a SCA14-associated mutation likely uncovers one of the primary recognition sites for Hsp70. Additionally, SCA14-associated mutations in C1A or C1B decrease domain stability. The transient unfolding of C1A or C1B in mutant PKC␥ increases the risk of amyloid-like aggregation, as suggested previously (12), but should also make the protein a good substrate of Hsp70. Such direct binding of Hsp70 to aggregation-prone regions in mutant C1A or C1B should effectively prevent the formation of amyloid-like aggregates.
The results show that Hsc70 was also incorporated into the PKC␥ aggregates. This result is consistent with our previous report that Hsc70 was pulled down together with mutant PKC␥ (46). Therefore, it is plausible that its chaperone activity could contribute to amelioration of mutant PKC␥ toxicity. Hsc70 was expressed in CHO, SH-SY5Y, and primary cultured PCs. However, unlike Hsp70, Hsc70 levels were not up-regulated by mutant PKC␥; therefore, it seems that the effect of Hsc70 on aggregation is limited.
A myriad of co-chaperones and cofactors facilitate Hsp70 chaperone activity (47). Of these, Hsp40, also known as DNAJB1, is a co-chaperone that regulates the ATPase activity of Hsp70 to enable the refolding and disaggregation of misfolded proteins (30). Here, we found that the overexpression of Hsp70, combined with Hsp40, dramatically reduced PKC␥ aggregation, whereas Hsp40 overexpression alone had limited effects. Although Hsp40 did not physically bind to PKC␥, it was localized to PKC␥ aggregates. The J domain of Hsp40 interacts with Hsp70, suggesting that the incorporation of Hsp40 into aggregates occurred via association with Hsp70 as a co-chaperone. In contrast, Hsp105 is highly expressed in PCs (48,49), and serves as a nucleotide exchange factor for Hsp70, but was not co-localized with aggregates in this study. Finally, neither Hsp60 nor Hsp90 co-localized with the aggregates. It is known that Hsp60 serves as a mitochondrial chaperonin (50). Although Hsp90 maintains proper protein folding in the cytoplasm (51), it did not target mutated PKC␥. Mutant PKC␥ aggregates were weakly immunoreactive for Hsp27, but overexpression of this chaperone had no significant effect on aggregation. It is known that Hsp27 protects against cellular toxicity, without reducing aggregation, in several in vitro and in vivo models, including models of SCA17 (52,53). Thus, Hsp27 may also serve a role in SCA14 pathogenesis that is independent of aggregation suppression.
The Hsps participate in protein folding, refolding of misfolded proteins, protein degradation, and disaggregation of amyloid fibrils (51,54). They may exert cytoprotective effects in cells that express mutant PKC␥s, via several pathways. We previously demonstrated that Hsp70 knockdown significantly reduced the rate of mutant PKC␥ degradation and increased PKC␥ protein deposition (14). Mutant PKC␥ is degraded by both proteasomal and lysosomal pathways, and Hsp70 participates in both of these pathways (55). In this study, overexpression or pharmacological induction of Hsp70 improved the detergent solubility of mutant PKC␥. In both cases, monomeric forms of mutant PKC␥ were significantly increased in the detergent-soluble fraction, indicating the chaperone activity of Hsps on misfolded mutant PKC␥.

Hsp90 inhibition as a potential therapeutic strategy for the treatment of SCA14
As a major transcription factor for Hsps, HSF1 is activated in response to proteotoxic stress. Recently, HSF1 was pharmacologically targeted to induce Hsps and to decrease protein aggregation and cytotoxicity in several neurodegenerative disease models (36,38,56,57). Celastrol is one such compound that inhibits Hsp90 and subsequently acts as an HSF1 activator. Although the neuroprotective effects of celastrol have been demonstrated in models of amyotrophic lateral sclerosis, Alzheimer's disease, Huntington's disease, and Parkinson's disease, its effect on the cerebellum, especially PCs, was previously unclear. To the best of our knowledge, this is the first evidence

Hsp cytoprotection in SCA14
that celastrol treatment induces Hsp70 expression in the cerebellum in vivo, and it prevents aggregation to ameliorate cytotoxicity caused by the mutant PKC␥ in vitro. It is possible that the celastrol-induced autophagic activity in primary cultured cerebellar cells also potentiates clearance of PKC␥ aggregates, because our previous report confirmed that pharmacological activation of autophagy enhances the clearance of PKC␥ aggregates (55). It has also been demonstrated that celastrol treatment reduces polyglutamine aggregation and toxicity by upregulating Hsp70 via the activation of HSF1 in a cell culture model of Huntington's disease (36). Thus, celastrol may exert neuroprotective effects in other SCAs, such as SCA1-3, -6, -7, -12, and -17.
Herbal extracts containing celastrol have been used in traditional Chinese medicine to treat rheumatoid arthritis for many years. Purified celastrol exerts anti-inflammatory and anticancer effects through multiple molecular mechanisms (58). Among them, the suppression of nuclear factor -light-chainenhancer of activated B cells (NF-B) signaling is well known, and it is noteworthy that NF-B activation is required for dendritic growth in PCs through transcriptional up-regulation of the microtubule-associated proteins, MAP2 and MAP1B (59). At the doses tested where celastrol induced Hsp70, and diminished aggregation and apoptosis in PCs expressing mutant PKC␥, the treatment failed to improve dendritic development, and it also elicited dendritic regression in PCs expressing WT PKC␥. We speculate that these effects may reflect celastrolinduced suppression of NF-B signaling and dendritic toxicity in PCs. Additionally, celastrol exerts anti-obesity effects by increasing energy expenditure through pharmacological activation of HSF1 in mice (60,61). We also observed similar effects of celastrol treatment on body weight; whereas the weights of DMSO-treated mice increased over the 2-week study period (13.7 Ϯ 3.8% (S.E. of the mean), n ϭ 8), the weights of celastroltreated mice decreased (0.9 Ϯ 2.8%, n ϭ 7). Moreover, it was reported that co-application of celastrol and arimoclomol, a co-activator of HSF1, improved the efficiency of celastrol-mediated Hsp induction (62). Thus, additional studies regarding the optimization of celastrol treatment are required for its therapeutic application against SCA.
Herbimycin A, a benzoquinone ansamycin, is also an inhibitor of Hsp90 that binds its nucleotide-binding site to activate HSF1 (63). In primary cultured PCs expressing mutant PKC␥, we found that the up-regulation of Hsp70 by herbimycin A suppressed de novo amyloid and amorphous aggregation of mutant PKC␥, ameliorated abnormal dendritic formation, and diminished apoptosis.
Our results indicate that aggregation of mutant PKC␥ reflects a dynamic equilibrium between aggregate formation and disassembly. In both control but especially in herbimycin A-treated cells, we observed disaggregation of already formed aggregates, including relatively large aggregates (5 m and larger). Inasmuch as herbimycin A did not potentiate autophagic activity, it seems plausible that the enhanced clearance of aggregates by herbimycin A involves Hsp70-dependent disaggregation. This role would be consistent with the observations of Gao et al. (54), who reported that human Hsp70 disaggregase-associated chaperone components can disassemble purified ␣-synuclein amyloid fibrils that are characteristic of Parkinson's disease.
As described above, it is unlikely that celastrol or herbimycin will advance as drugs for SCA14, due to their toxicity and poor blood-brain barrier permeability. Recently, several novel smallmolecule Hsp90 inhibitors have been developed, some of which have been demonstrated to induce brain Hsps in murine models of Alzheimer's disease and Parkinson's disease (66,67). Furthermore, although some of these compounds have failed in clinical trials for various malignancies or cancers due to limited clinical activity, the adverse effects of the drugs were manageable (68). These compounds may be of utility as therapeutic agents against SCA14.

Immunofluorescence staining, confocal imaging, and cell counting
The SH-SY5Y and CHO cells expressing the indicated recombinant proteins in glass-bottom dishes were fixed with 4% paraformaldehyde/phosphate-buffered saline (PBS) for 1 h at room temperature. After washing, the cells were incubated in 10% normal goat serum/Tris-buffered saline (TBS) containing 0.1% Tween 20 (TBS-T) for 30 min. After blocking, the cells were incubated with the appropriate primary antibodies for 2 h at room temperature or overnight at 4°C, followed by incubation with appropriate fluorescence-tagged secondary antibodies. Immunostaining was assessed by confocal immunofluorescence microscopy (LSM700, Carl Zeiss, Germany).
The CHO and SH-SY5Y cells transfected with Hsp70 and/or PKC␥-GFP were cultured for 1-2 days and fixed, as described above. After two washes with PBS, cells were observed by confocal microscopy. A total of 200 -300 GFP-positive and Hsp70positive cells were counted.

Cell death analysis
CHO cells transfected using FuGENE 6 (Promega, Madison, WI) were cultured on 6-cm diameter dishes for 2 days. Flow cytometric analyses were conducted using the FACSCalibur system (BD Biosciences) as described previously (9), with slight modifications. Briefly, we used 7-AAD (559925, BD Biosciences) as a marker for dead cells. For each sample, the fluorescence of 2 ϫ 10 4 cells was recorded and analyzed by Cell Quest TM software (BD Biosciences).

Triton X-100 solubility fractionation and immunoblotting
One day after transfection, CHO cells in 6-well plate dishes were washed, harvested with ice-cold PBS, and resuspended in 40 l of TBS-T (150 mM NaCl, 1% Triton X-100, and 50 mM Tris-HCl, pH 7.4) containing protease inhibitor mixture. After homogenization with a sonicator, 10 l of each cell lysate was preserved for analysis as inputs. The remaining homogenates were centrifuged for 12 min at 20,000 ϫ g at 4°C, and the resulting supernatants were collected as the Triton X-100 -soluble

Hsp cytoprotection in SCA14
fraction. Supernatant protein concentrations were determined using a bicinchoninic protein assay kit (Pierce, ThermoFisher Scientific). The inputs and Triton X-100 -soluble fractions were mixed with SDS sample buffer. Pellets were resuspended with 2ϫ SDS sample buffer, sonicated, and used as the Triton X-100 -insoluble fraction. The SDS sample buffer did not contain any reducing agents. All fractions were boiled for 10 min at 50°C, and equal amounts of protein were subjected to SDS-PAGE and Western blotting. Signals were detected using the ChemiDoc XRS system (Bio-Rad), and quantitative analyses were performed using Quantity One software (Bio-Rad). All comparisons were made between bands on single blots.
All drug treatments were performed 5 h after transfection with Q127R PKC␥-GFP in SH-SY5Y cells. Celastrol (3 M) treatment was performed for 3 h, and subsequently, the medium containing celastrol was replaced with fresh medium. After 16 h of cultivation, cells were harvested. Harvested cells were subjected to Triton X-100 solubility fractionation, as described above.

Treatment of primary cultured PCs for Western blotting
Approximately 6 ϫ 10 5 mouse cerebellar primary cells were cultured in poly-L-lysine-coated 24-well dishes. On DIV 21, celastrol (3 M, 1 h, followed by 23 h of incubation without celastrol), herbimycin A (2 M, 24 h), or DMSO-treated primary cultured PCs were harvested with ice-cold PBS and resuspended in TBS-T containing a protease inhibitor mixture. After homogenization with a sonicator, whole-cell lysates were mixed with SDS sample buffer and boiled for 5 min at 95°C. Quantities of 40 g (Hsp70) and 10 g (Hsc70/Hsp40) of protein were subjected to SDS-PAGE to quantify the expression levels of Hsps and normalized to ␣-tubulin. For LC3-II detection, cells were directly harvested with 60 l of SDS sample buffer, boiled for 5 min at 95°C, and sonicated and then 20 l of the lysates were subjected to SDS-PAGE. Chloroquine (50 M, 17 h; C6628, Sigma) was used to accumulate LC3-II. The LC3-II expression levels were normalized to ␣-tubulin.

Treatment of primary cultured PCs for fluorescence microscopy observation
To observe the pharmacological induction of endogenous Hsp70 expression in the primary cultured PCs, the cells were treated with celastrol (3 M, 1 h/day), herbimycin A (2 M), or DMSO for 4 days on DIV 21. The cells were fixed and stained with antibodies against Hsp70 and calbindin.
The PCs expressing WT or G128D PKC␥-GFP were treated with celastrol (3 M, 1 h/day, for 4 days), herbimycin A (2 M, 4 days), or DMSO. Cells were fixed and stained for calbindin and assessed by confocal microscopy. The intensity of GFP was measured, and the aggregates showing fluorescence intensity Ͼ250 a.u. were designated as high-intensity aggregates. At least 100 calbindin-and GFP-positive cells were counted per dish. Total dendrite length and total nodes or end numbers were measured using Neurolucida and Neurolucida Explorer software (MBF Bioscience, Tokyo, Japan). At least 30 calbindin-and GFP-positive cells were analyzed per dish.

Long-term time-lapse imaging
Long-term time-lapse imaging was performed using a computer-assisted fluorescent microscope, the LCV100 incubation imaging system (Olympus, Tokyo, Japan). This device possesses the ability to monitor fluorescence in living cells under normal culture conditions (37°C, 95% humidity, and 5% CO 2 ). Five days after adenoviral transduction, GFP fluorescence of DMSO-treated or 2 M herbimycin A-treated primary cultured PCs was observed for 51 h. For the experiments performed to observe the effects of Hsp70 inhibition, the cells were treated with 2 M herbimycin A for 8 h, followed by addition of DMSO or PES-Cl (10 M), and then the cells were observed for 46 h using the LCV100. Sequential GFP fluorescence images of PCs were obtained every 1 h and analyzed using MetaMorph software (Molecular Devices Corp.).

Pharmacological induction of Hsps in mice
All animal studies were approved by the Institutional Animal Care and Use Committee and conducted in accordance with the Kobe University Animal Experimentation Regulations. All efforts were made to minimize suffering and to reduce the number of animals used. Four-week-old male C57BL/6N mice were obtained from Japan SLC, Inc. (Shizuoka, Japan) and treated with celastrol (1 mg/kg, i.p.), herbimycin A (2 mg/kg, i.p.), or DMSO (10% DMSO in PBS, i.p.) once daily for the indicated durations. Animals were humanely euthanized, and the cerebellum was dissected out and homogenized by sonication in ice-cold TBS-T (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100, pH 7.4) containing a protease inhibitor mixture. Brain homogenates were centrifuged at 20,000 ϫ g for 15 min at 4°C. The resulting supernatants were collected, mixed with SDS sample buffer, and boiled for 5 min at 95°C. Quantities of 10 g (Hsp40) and 40 g (Hsp70) of protein were subjected SDS-PAGE to quantify the expression levels of Hsps and normalized to ␣-tubulin.

Experimental design and statistical analysis
Two-tailed unpaired t-tests (to compare two groups) and one-way analyses of variance with Tukey's post hoc tests (to compare three or more groups) were conducted using Graph Pad Prism 5 (La Jolla, CA; RRID: SCR_002798). All numbers and sample sizes for each experiment are indicated in the corresponding figure legends.