Disrupted structure and aberrant function of CHIP mediates the loss of motor and cognitive function in preclinical models of cerebellar CHIPopathy

CHIP (carboxyl terminus of heat shock 70-interacting protein) has long been recognized as an active member of the cellular protein quality control system given the ability of CHIP to function as both a co-chaperone and ubiquitin ligase. Mutations in CHIP are the driver of spinocerebellar autosomal recessive 16 (SCAR16), or cerebellar CHIPopathy, as we initially discovered this disease was caused by a loss of CHIP ubiquitin ligase function. The initial mutation describing SCAR16 was a missense mutation in the ubiquitin ligase domain of CHIP (p.T246M). Using multiple biophysical and cellular approaches, we demonstrate that T246M mutation results in structural disorganization and misfolding of the CHIP U-box domain, promoting oligomerization, and increased proteasome-dependent turnover. CHIP-T246M has no ligase activity, but maintains interactions with chaperones and alters the co-chaperone function of CHIP. To establish preclinical models of SCAR16, we engineered T246M at the endogenous locus in both mice and rats. Animals homozygous for T246M had both cognitive and motor cerebellar dysfunction distinct from those observed in the CHIP null animal model, as well as deficits in learning and memory, reflective of the cognitive deficits reported in SCAR16 patients. We conclude that the T246M mutation is not equivalent to the total loss of CHIP, supporting the concept that disease-causing CHIP mutations have different biophysical and functional repercussions on CHIP function that may directly correlate to the spectrum of clinical phenotypes observed in SCAR16 patients. Our findings both further expand our basic understanding of CHIP biology and provide meaningful mechanistic insight underlying the molecular drivers of SCAR16 disease pathology, which may be used to inform the development of novel therapeutics for this devastating disease.


Introduction
Protein quality control (PCQ) involves a specialized cellular surveillance system that monitors protein integrity, identifies unfolded or damaged proteins, and then either repairs or targets them for degradation. CHIP is abundantly expressed in most tissues and plays a central role in maintaining protein quality control [1]. CHIP is uniquely suited as a regulator of protein quality control due to its dual functions as both a co-chaperone protein and ubiquitin ligase enzyme. As a co-chaperone, CHIP interacts with heat shock protein (HSP)-bound proteins to aid in substrate stabilization and refolding [2]. Conversely, as a ubiquitin ligase CHIP ubiquitinates terminally-defective proteins and prepares them for degradation by the Ubiquitin Proteasome System (UPS) [3]. Since the discovery of CHIP in 1999 [1], numerous reports detailing CHIP's co-chaperone and ubiquitin ligase activities in both the brain and heart have been published [4][5][6][7]. However, recent reports describing surprising new roles for CHIP have emerged. These roles include autonomous chaperone activity [8,9], the regulation of cardiac metabolic homeostasis via the metabolic sensor AMPK (AMP-activated protein kinase) [9], and DNA damage repair [10]. Most recently, CHIP was implicated in the pathophysiology associated with cerebellar CHIPopathy, or spinocerelbellar autosomal recessive 16 (SCAR16) [11,12], representing the first direct association between a CHIP polymorphism and a human disease. Cerebellar CHIPopathy (MIM 615768) is a form of autosomal recessive spinocerebellar ataxia that can also be accompanied with hypogonadism, similar to the clinical phenotype of Gordon Holmes Syndrome (GHS) [13]. Despite over 100 years of clinical recognition, only recently have causal mutations for GHS been identified. These mutations include the ubiquitin ligase RNF216 and deubiquitinase OTUD4 [14], which suggest that faulty ubiquitination plays an essential role in the pathophysiology of ataxia. Using exome sequencing, we identified a mutation in STUB1 (the gene encoding CHIP) in two patients initially diagnosed with GHS [11]. We found that this STUB1 mutation (p.T246M) resulted in a loss in the ubiquitin ligase function of CHIP [11].
Combined with our studies demonstrating that mice lacking the expression of CHIP display motor deficiencies and some aspects of the hypogonadism observed in patients with STUB1 mutations, our CHIP knockout mouse represented the first animal model of CHIPopathy [11].
The amino acid substitutions reported in cases of cerebellar CHIPopathy result in nonsense, missense, frameshift, or splicing mutations; the majority of which are predicted to significantly alter protein function [12]. Given the clinical heterogeneity of neuroendocrine phenotypes in STUB1 patients, specific CHIP mutations likely have varying biophysical and functional consequences to CHIP function. In this context, an animal model with a total loss of CHIP may not adequately represent the spectrum of human disease represented by SCAR16. For these reasons, we complemented the biophysical and cellular repercussions of CHIP-T246M with two rodent models engineered with CRISPR/Cas9 to mimic this human mutation. Additionally, we performed in-depth behavioral assessments to determine the effects of T246M mutation at a whole-animal level to establish a suitable preclinical model for SCAR16. Studying CHIP mutations both in vitro and in vivo allows us to delineate the contribution of co-chaperone, ubiquitin ligase, and other emerging CHIP activities to specific deficits observed in a diseaserelevant context in vivo. These results provide insights that are valuable for the development of effective therapies for this devastating degenerative disease.

BIOCHEMICAL AND CELLULAR ANALYSIS OF CHIP-T246M
The T246M mutation destabilizes the structure of the U-box and CHIP-T246M forms decamers and dodecamers in vitro and in cells. Asymmetric homodimerization of CHIP as well as conformational flexibility are required for CHIP ubiquitin ligase activity. Critical to both the dimerization and conformational flexibility is the U-box domain [3,25] where T246 is located and is a highly conserved residue across CHIP homologs [11]. Furthermore, T246 is located in the core of a conserved beta hairpin turn that lies at the interface between the two molecules of the CHIP dimer ( Figure 1A). Modeling of the T246M amino acid substitution predicts that this mutation would impact the tertiary structure of the U-box and hence disrupt the formation of functional dimers, consequently reducing or abolishing CHIP's ubiquitin ligase function towards both chaperone and non-chaperone substrates. To test the effects of T246M substitution on Ubox structure, we performed solution NMR on purified WT and T246M CHIP U-box domains.
Whereas WT CHIP U-box exhibit distinct peaks spread throughout the 2D 15 N-1 H HSQC spectrum characteristic of a stable, structured domain (Figure 1B), the T246M spectrum is collapsed and has broad resonances, suggesting the loss of stable globular structure, multiple conformations and aggregation ( Figure 1B). Additionally, circular dichroism spectra were acquired for the isolated WT and T246M CHIP U-box domains. Compared to the secondary structure content of the WT U-box, the signal from T246M U-box was consistent with loss or regular secondary structure and a shift to random coil conformation ( Figure 1C). Further, we monitored the protein melting temperature (Tm) at 222 nm, the characteristic wavelength for αhelices. The WT U-box exhibited a sigmoidal curve consistent with unfolding of a globular, folded domain with a Tm of 30 °C, whereas T246M appeared to be unfolded at room temperature and did not undergo any change with temperature ( Figure 1C). Together these data suggest the T246M mutation destabilizes the U-box domain, resulting in a loss of secondary and tertiary structure.
Previous size exclusion chromatography (SEC) analysis suggested that full-length CHIP harboring the T246M formed large aggregates greater than 670 kDa [23,24], however, SEC alone cannot account for differences in protein conformation or account for protein-column interactions [26]. Given the known dynamics of CHIP conformation in solution, we used an analytical approach, SEC coupled with multi-angle light scattering, to determine true molar mass and radius in solution and compared these data to the composition of native CHIP-T246M when expressed in cells using blue native PAGE and immunoblot analysis. As expected, both the WT protein and the K30A control mutant were predominantly dimers, 72 kDa ( Figure 1D, Table 1). However, the U-box domain mutants T246M and H260Q were detected as higher-order oligomers, suggesting CHIP-T246M exist in cells predominantly as 10-12mers ( Figure 1D). Native gels of WT CHIP and the same point mutants expressed in COS7 cells also confirmed that CHIP-WT and CHIP-K30A form dimers whereas CHIP-T246M and CHIP-H260Q form higher-order oligomers ( Figure 1E). Next, we performed indirect immunofluorescence to observe CHIP localization and expression patterns in the same cell model. Not surprisingly, CHIP-WT protein is detected as diffuse staining throughout the cytoplasm and within the nucleus, while CHIP-T246M protein appears as punctate staining in the cytoplasm and perinuclear regions, perhaps reflective of CHIP-T246M oligomers ( Figure   1E). Taken together, these data suggest that the mutation destabilizes the U-box domain resulting in the loss of ubiquitin ligase activity of T246M CHIP previously observed [11] and promotion of oligomerization.
The T246M mutation increases co-chaperone activity. When cells are exposed to heat, the cochaperone activity of CHIP initiates the heat shock response through the activation of heat shock factor 1 (HSF1). At the same time, heat induces the oligomerization of CHIP both in vitro and in cells, enhancing the chaperone activity of the protein while not affecting ubiquitin ligase activity [8]. Remarkably, the oligomerization of CHIP-WT that occurs with heat resembles the oligomerization of CHIP-T246M and -H260Q [8]. Given T246M increased the interaction between CHIP and HSC70/HSP70 [11] and the oligomeric status of CHIP-T246M ( Figure 1E), we hypothesized that CHIP-T246M would still have the ability to co-chaperone HSF1. To test whether CHIP's regulation of HSF1 remains intact with T246M, we transiently expressed CHIP constructs in COS7 cells and measured the nuclear translocation of HSF1 and changes in HSF1 transcriptional activity. As previously reported [27] we observed that expression of WT CHIP promotes the nuclear translocation of HSF1 (Figure 2A) and stimulates HSF1 activity ( Figure   2B). Remarkably, expression of CHIP-T246M also promoted HSF1 nuclear translocation ( Figure 2A) and enhanced HSF1 transcriptional activity ( Figure 2B). In contrast, abolishing chaperone interactions via the K30A mutation of CHIP did not affect HSF1 dynamics. The ability of T246M to affect HSF1 was dependent on a functional TPR domain, as the effect of T246M on HSF1 activity is lost with the K30A-T246M double mutant ( Figure 2B).

Co-expression of CHIP-WT and CHIP-T246M does not alter oligomerization or localization.
The recessive basis of SCAR16 suggests that the wild-type allele is sufficient to overcome the disease-bearing mutated allele. Given the U-box is important to dimerization of CHIP we hypothesized that CHIP-WT would not interact with CHIP-T246M. Interestingly, using extracts from cells expressing both CHIP-WT and CHIP-T246M, these two proteins co-immunoprecipitated each other ( Figure 2C) suggesting the two forms of CHIP can interact with each other. However, when we directly separated the lysates via BN-PAGE immunoblot analysis, we observed only subtle differences in oligomerization patterns when CHIP-WT and CHIP-T246M are co-expressed compared to the single protein expression conditions ( Figure   2D). Given that BN-PAGE can limit antibody detection of proteins given the native conditions, possibly obscuring changes in visualizing heterocomplexes, we also analyzed expression via indirect immunofluorescence, and we observed that the localization of CHIP-WT was unaffected by co-expression of CHIP-T246M ( Figure 2E) despite extensive overlap in the localization of CHIP expression within cells ( Figure 2E). Together, these data suggest that while WT and T246M CHIP interact in crude extracts, the localization and dimerization status of CHIP-WT is largely unaffected by the presence of CHIP-T246M. Therefore, CHIP function likely also remains intact in the heterozygous condition and the presence of WT CHIP does not appear to alter the oligomerization or distribution of CHIP-T246M, consistent with the recessive nature of this disease-causing mutation.
Change in solubility and increased proteasome-dependent turnover of CHIP-T246. We consistently observed lower levels of soluble CHIP-T246M, CHIP-H260Q, and CHIP-K30A/T246M protein relative to CHIP-WT when transiently transfecting equal amounts of vector DNA ( Figure 2B). We hypothesized that this decrease in soluble CHIP-T246M expression could be due to changes in solubility and/or stability. To test for changes in solubility, we performed SDS-PAGE and immunoblot analysis of CHIP in both the soluble and insoluble fraction of whole cell lysates. We observed CHIP-T246M, -H260Q, and -K30A-T246M CHIP in both the soluble and insoluble fraction; in contrast, CHIP-WT or -K30A was found only in the soluble fraction ( Figure 2F). In addition to the change in solubility, we also observed a shorter half-life of soluble CHIP-T246M compared to CHIP-WT, an effect that that was dependent on proteasome activity ( Figure 2G). Taken together, our data suggest that the T246M mutation results in altered cellular distribution, solubility, and stability and likely contributes to the loss of CHIP function.
Protein levels of CHIP-T246M are reduced in fibroblasts from SCAR16 patients and mice engineered with the equivalent mutation. Prior to this report, analyses of CHIP-T246M by others were limited to in vitro analyses or exogenous expression of CHIP in cells that robustly express endogenous CHIP [23,24]. Thus, we utilized primary embryonic fibroblasts (MEFs) isolated from mice engineered with the corresponding murine amino acid substitution at the endogenous CHIP locus (T247M) [28]. Consistent with our exogenous expression models, soluble CHIP protein levels were robustly reduced in MEFs isolated from different M246/M246 mouse embryos relative to either T246/T246 or T246/M246 littermates ( Figure 3A). The decrease in steady-state CHIP in M246/M246 MEFs appeared to be at the post-translational level, as mRNA levels of Stub1 were equal across all three genotypes ( Figure 3B). We isolated protein extracts from patients homozygous for the CHIP-T246M mutation, and again, we measured a decrease in steady-state soluble CHIP-T246M protein ( Figure 3C). Together these data suggest that the reduction in CHIP-T246M protein is likely a result of post-translational regulation.
Endogenous CHIP-T246M is degraded by the proteasome. Given the difference in steady-state soluble protein levels of CHIP-T246M in primary cells from patients and our preclinical models, we evaluated the turnover rate of CHIP in MEFs. Similar to exogenously expressed CHIP proteins ( Figure 2G), the turnover of soluble CHIP-T246M protein was rapid (t1/2 = 1.2 h) compared to the stable, CHIP-WT protein. In fact, CHIP-T246M protein was undetectable after 6 hours of cycloheximide chase compared to approximately 75% of CHIP-WT protein that remained after 6 hours ( Figure 3D). To evaluate the solubility distribution of CHIP-T246M and whether this distribution was effected by proteasome inhibition, we treated MEFs with 20 µM MG132 or 0.05% DMSO control for 4 hours. We collected soluble, insoluble and total protein fractions from samples containing equal cell numbers and analyzed protein levels using immunoblot analysis. Total T246M protein is dramatically reduced compared to WT protein ( Figure 3E). Secondly, the change in total protein levels is dramatically higher with proteasome inhibition for T246M (4-fold) protein compared to WT, suggesting a robust proteasomedependent turnover of endogenous CHIP-T246M ( Figure 3E).

CHIP-T246M maintains protein-protein interactions and the response to heat shock.
We previously demonstrated that exogenous CHIP-T246M immunoprecipitates more chaperone clients, such as HSP70 and HSC70, compared to CHIP-WT [11]. However, given the strong decrease in steady-state CHIP-T246M protein expression seen in primary cells, it is possible that these interactions are lost, resulting in a CHIP null phenotype. Remarkably, in primary cells we found that CHIP-T246M maintains interactions with its chaperone substrate AMPK ( Figure   3F), as well as HSC70 (Figure 3G), at levels similar to CHIP-WT, suggesting that CHIP-T246M may still exhibit activity towards these proteins. To test this concept, we exposed primary cells to heat and to determine if cells expressing CHIP-T246M still respond to heat shock, as CHIP plays an important role in inducing the heat shock response via activation of HSP70. Unexpectedly, M246/M246 cells still maintained activation of HSP70 during the recovery period following heat shock ( Figure 3H), whereas cells completely lacking CHIP expression had an attenuated response as previously described [27]. These data demonstrate that CHIP-T246M may retain some chaperone function, despite the increase in turnover.

PRECLINICAL MODELS OF CEREBELLAR CHIPOPATHY
Generating in vivo models of CHIP-T246M. Given the advantages of both mouse and rat models to study neurological disease, we developed a rat model also harboring the same endogenous CHIP-T246M mutation. Similar to the decrease seen in fibroblasts isolated from patients with T246M mutations, steady-state levels of CHIP-T246M expression are 40% lower in protein extracts isolated from the cerebellums, whole brain, and testes of T246M rats ( Figure 4A) and further decrease with age ( Figure 4B). CHIP protein levels in mouse tissues isolated from M246 mice were more dramatically decreased (90% decreased) compared to CHIP levels in tissues from wild-type mice ( Figure 4C). The decrease in tissue expression was not due to differences in levels of the mRNA that encodes CHIP ( Figure 4D). The effect of the T246M mutation on CHIP was also evident via immunohistochemical analysis in both the mouse and rat models ( Figure 4E) where we also identified Purkinje cell degeneration, evidenced by a loss in calbindin staining ( Figure 4E). In both mouse and rat models, we observed lower body weights in the M246/M246 mice over time ( Figure 4F) as well as an increase in mortality in rats ( Figure 4G). In addition to lower body weights, there was selective tissue atrophy in the brain and testes of M246/M246 mice, known tissues that play a role in SCAR16 pathology, whereas the heart was not affected ( Figure 4H). Together, these data demonstrate that our two preclinical models of T246M recapitulate features of cerebellar CHIPopathy in patients including decreased protein expression of CHIP, neurodegeneration, as well as neuro-and gonadal-atrophy.
M246 results in progressive ataxia and alterations in gait. One measure of ataxia in rodent models is measured by performance on a rotating rod, known as the rotarod test. In both mouse and rat models, we found a decrease in rotarod performance. In mice, we found initial learning to be blunted ( Figure 5A) and an age-dependent decrease in performance starting around 30 weeks of age ( Figure 5B). Given our previous observation of a near complete lack of learning of rotarod behavior in mice completely lacking CHIP expression [11], we utilized a second metric of ataxia progression. This composite test is comprised of hind limb clasping, ledge test, gait, and kyphosis [29] and was used to measured ataxia onset and progression in both the CHIP-T246M and CHIP(-/-) mouse lines ( Figure 5C). Remarkably, we found that CHIP-/-mice were already ataxic at weaning (score = 4.4, Video 1), and the rate of ataxia progression was equivalent to suggesting the T246M mouse may reflect a more suitable preclinical model for SCAR16 compared to the CHIP-/-mouse. Likewise, in rats, we found robust decreases in rotarod performance at ages as young as 12 weeks of age that rapidly decreased with age ( Figure 5D and Video 2B). The loss in rotatrod performance was also accompanied by changes in gait ( Figure 5E -5H), and much like the effect on motor performance, the effect on gait was exacerbated by age with notable differences occurring at 32 weeks of age. Thus both the mouse and the rat model of CHIP-T246M demonstrate progressive ataxia that worsens with age and recapitulates the clinical phenotype observed in patients with cerebellar CHIPopathy that suffer from an early adult-onset progressive ataxia [12].
Behavioral repercussions of CHIP-T246M include loss of prepulse inhibition, hyperactivity, and cognitive dysfunction. We utilized our rodent models to determine effects of the T246M mutation on the additional clinical hallmarks of SCAR16, decreased sensorimotor reflexes and cognitive dysfunction. Given the onset of ataxia phenotype in both the mouse and rat models were observed around 30 weeks of age ( Figure 5B, 5C), we tested mice at ages both before and after this time point, 8-12 and 33 weeks, respectively. Using the acoustic startle and prepulse inhibition test we found no effects of M246 on the amplitude of the acoustic startle response in either young or adult mice ( Figure 6A). Similarly, all young mice had comparable levels of prepulse inhibition; however, older M246/M246 mice exhibited robust decreases in percent inhibition, indicating the emergence of sensorimotor deficits by age 33 weeks ( Figure 6B).
These degenerative deficits in prepulse inhibition are consistent with other mouse models of cerebellar degeneration with profound Purkinje cell loss [30].
Mice were tested in the elevated plus maze at eight weeks of age and there was no difference between genotypes in any of the parameters measured ( Table 2). However, we observed age- Our data suggest CHIP-T246M results in impulsive and risky exploration, as observed in mouse models for mania and impulsivity and overt hyperactivity [31,32]. Interestingly, both impulsivity and hyperactivity have been attributed to cognitive cerebellar dysfunction in humans [33][34][35]; therefore, we evaluated cognitive function in mice using the fear response test and in rats using the Morris water maze. The M246/M246 mice had impairments in the conditioned fear procedure, both in contextual and cue-dependent learning ( Figure 6F, 6G).
On the initial training day, all genotypes had similar, low levels of freezing before any exposure to the aversive foot shock. On the next day, or two weeks later, the lack of learning in M246/M246 mice was apparent as they did not increase their freezing time to either a contextual ( Figure 6F) or cue-dependent ( Figure 6G) stimulus. This lack of response could not be attributed to hearing impairment, since the mutant mice had normal performance in the acoustic startle test ( Figure 6A). Finally, to determine if the M246 mutation has similar effects on cognition in the rat model, we measured the performance of young and adult rats in the Morris water maze. Consistent with the conditioned fear responses in mice, learning was also impaired in M246/M246 animals, as measured by the escape latency ( Figure 6I). Moreover, young M246/M246 rats had a 42% decrease platform zone occupancy that worsened in older animals ( Figure 6H), consistent with a profound deficit in memory recall. Together these data suggest significant impairment in learning and memory as a result of CHIP-T246M and are consistent with the clinical phenotype of patients with T246M [11].
Additional testing in the 3-chamber choice test found that the M246/M246 genotype associated with altered social behavior ( Figure 6J-6M), such that M246/M246 mice had increased preference for social novelty towards the newly-introduced stranger 2 ( Figure 6L, 6M).
Further behavioral testing revealed the M246/M246 mice had reduced marble burying, indicating a decrease in exploratory digging ( Table 2); moreover, no effects of genotype were observed for olfactory ability in a buried food test ( Table 2). Overall, the results of the battery of behavioral assessments performed suggest that homozygous CHIP-T246M leads to the dysregulation of inhibitory processes governing activity, exploration, and sensorimotor gating, as well as impaired learning and memory in tests for conditioned fear and cognition.
Interestingly, impaired conditioned fear and decreased marble-burying were reported in mice with deletion of maternal E3 ubiquitin ligase Ube3a, a model for Angelman syndrome [36], again highlighting the critical role of protein ubiquitination in cerebellar homeostasis.
Alterations in the T246M proteome identified both known and potentially novel substrates of CHIP-dependent regulation. Given the role of CHIP as a chaperone, co-chaperone, and ubiquitin ligase it is likely that disruption in CHIP function alters the regulation of proteins critical to cerebellar function and protein homeostasis. To identify candidate proteins that may mediate the pathogenesis of SCAR16, we performed unbiased proteomics via mass spectroscopy to identify differentially expressed proteins in cerebellar lysates prepared from either T246/T246 or M246/M246 rats. We identified 63 and 80 unique proteins that were either less or more abundant in M246/M246 relative to T246/T246 cerebellums ( Figure 7A, Supplementary Table S1). As expected, we identified a 50% decrease in CHIP protein in M246/M246 brains as measured via MS (mean ratio = 0.5, p = 3.47E-08). We confirmed the differential levels of selected proteins that were either increased in M246/M246 cerebellums including Pde9a and Tau or decreased in M246/M246 cerebellums, such as Phlpp1, alpha-synuclein, CHIP, and Pink1 ( Figure 7A). Next, we analyzed the dataset using known and predicted protein-protein association data combined with functional enrichment analysis via the STRING database [37] to identify proteins or pathways that may play a role in the cerebral pathophysiology of SCAR16. Of the 143 initial proteins, 60 proteins were identified in a total of 21 protein-protein interaction clusters (Supplementary Table S2) including a family of six proteins that contained CHIP and other regulators of protein quality control, such as BAG3 and several F-box proteins ( Figure 7B). Additionally, we identified a cluster of proteins known to be affected by CHIP expression including tau (MAPT), alphasynuclein (SNCA), and PINK1 [5,38,39]. Of note, seven member cluster that was also functionally enriched for coagulation processes (wound healing) and extracellular matrix function was identified ( Figure 7B), and in total, 16 proteins involved in coagulation were affected by CHIP-T246M. Coagulation was recently identified as a significant component of spinocerebellar ataxia disease signatures (without Friedreich ataxia) from a large-scale analysis of 80,000 samples [40] and our analysis of CHIP-T246M cerebellums support the notion that coagulation may represent new targets for therapeutic interventions.

Discussion
CHIP plays a vital role in cerebellar homeostasis, however, how do the various mutations that cause cerebellar CHIPopathy drive the disease pathogenesis? The simplest explanation is that the CHIP mutations destabilize protein stability that lead to an effective CHIP null condition.
However, a varying degree of CHIP protein levels is found in fibroblasts isolated from patients with CHIP mutations. For example, the N65 mutation results in an 80% loss of CHIP protein [21], whereas other mutations, such as T246M ( Figure 3C) and the compound mutation M211I/E238* have more modest effects, approximately at the level of haploinsufficiency [16].
Moreover, the R119*/I294F and K145Q/P243K compound mutations do not appear to affect CHIP turnover [20]. Given the recessive nature of this disease in patients [12], and the lack of phenotype observed in our three rodent models of CHIP haploinsufficiency ( Figure 5C, 5D), these data suggest that either the loss or change in specific activities of CHIP drive the pathophysiologies associated with cerebellar CHIPopathy.
We identified that the T246M leads to disruption of the U-box structure ( Figure 1A, 1B, 1C), and the effect of the T246M conformational change on CHIP function appears to be three-fold.
First, the destabilization of the U-box results in no appreciable ligase activity as demonstrated previously by our group [11] and others [23,24]. Secondly, CHIP-T246M is degraded in a proteasome-dependent manner at a higher rate than CHIP-WT, both when expressed exogenously ( Figure 2G) and under native genomic conditions ( Figure 3D, 3E), resulting in decreased steady-state levels that we observed both in both T246M patients ( Figure 3C) and in our rodent models ( Figure 4A, 4B, 4C). Lastly, and more surprisingly, the T246M mutation promotes the formation of soluble oligomeric forms of CHIP comprised of 10-12 mers. CHIP functions as a dimer [41,42], however under heat stress, CHIP forms higher-order oligomers that appear to have increased chaperone activity [8]. Therefore, we tested the effect of T246M on the chaperone-mediated activation of HSF1 and found that T246M results in increased nuclear localization of HSF1 (Figure 2A) and increased HSF1 activity ( Figure 2B). We initially considered that CHIP-T246M was activating HSF1 simply because the cells were responding to a misfolded protein insult, and not a specific effect of CHIP-T246M on HSF1. To account for this, we engineered a CHIP-T246M protein that also contained the K30A mutation that abolishes the interaction between CHIP and its chaperone binding partners. Unlike T246M, the K30a-T246M double mutant had little effect on HSF1 activity ( Figure 2B) despite similar levels of insoluble CHIP protein ( Figure 2F). Therefore, while the T246M may abolish ligase activity, it appears that one consequence of these mutations may be altering chaperone function by failing to process substrates usually triaged by the CHIP-HSC(P)70 complex in a manner distinct from the CHIP null condition. We previously observed increased pulldown of HSP70 and HSC70 with CHIP-T246M [11], and even in our rodent model with reduced steady-state levels of CHIP-T246M, this interaction is maintained ( Figure 3G). In fact, comparable amounts of HSC70 as well as the chaperone substrate of CHIP, AMPK, immunoprecipitate with CHIP or CHIP-T246M ( Figure 3F, 3G) suggesting that even though the expression of CHIP-T246M is reduced, it is still engaging with known client proteins at similar levels. In cell culture models, the synthetic CHIP mutation, H260Q, acted in similar to T246M regarding HSF1 activation (Figure 2B), confirming a previous report that also identified that a U-box mutation led to changes in BAG3 protein levels and suppression of the macro autophagy pathway [43]. Likewise, it was proposed that enhancing either macro or chaperone mediated autophagy may be beneficial in CHIPopathies [16]. Interestingly, our proteomics analysis of the T246M rodent cerebellum revealed a decrease in Bag3 protein levels (Supplementary Figure S1, Supplementary Table S1), so one distinct possibility for the pathology associated with cerebellar CHIPopathy may be a disruption in either the macroautophagy or chaperone-mediated autophagy pathways, rendering cells susceptible to proteinopathies. Alteration of these autophagy pathways is implicated in other ataxias, such as spinocerebellar ataxia type 1, 3 7, and 14, [44][45][46][47].
Central to disease-based research is the establishment of preclinical models that can be used as a platform to both better understand the mechanism of disease and to test therapies. Therefore, we created both a mouse and rat model that genocopies the human T246M. We found that these models recapitulate the key features of SCAR16, notably: decreased steady-state protein expression due to proteasome-dependent turnover ( Figure 3D, 3E Figure 5C, 5D, and 5E); and impaired cognitive function ( Figure 6F, 6G, 6H, 6I). Given these phenotypes were exacerbated with age, these models provide us with the ability to interrogate the pathophysiology at difference time points in disease progression. These data further support our previous findings that CHIP plays a critical role in cerebellar maintenance [11]. Interestingly, while some behavioral deficits in M246/M246 rodents overlap with those observed in CHIP-/-mice [11], the majority were unique to T246M mutation, further supporting our hypothesis that disease-causing mutations in CHIP and the total loss of CHIP are not functionally equivalent. For example, M246/M246 mice exhibited overt hyperactivity ( Figure   6C) risky behavior (Figure 6D), deficits in prepulse inhibition (Figure 6B), and increased preference for social novelty (Figure 6K, 6M), phenotypes not observed in CHIP-/-mice [11].
We hypothesize that the phenotypic differences observed between CHIP-/-mice and T246M mice are likely reflective of our cell-based and in vitro findings that while T246M CHIP no longer functions as an E3 ubiquitin ligase, other CHIP functions remain intact despite this mutation.
Perhaps even more intriguingly, the T246M mutation may modify the co-chaperone activity of CHIP in a deleterious manner. For example, mutant CHIP may be unable to either ubiquitinate chaperone-engaged proteins or promote the refolding of proteins that are usually degraded by the proteasome. Similarly, mutant CHIP may promote the activation of compensatory pathways involved in proteins degradation or clearing protein aggregates such as autophagy [43] that could sensitize the cells to additional proteotoxic stress. Moreover, the formation of soluble CHIP oligomers that alter the chaperone or co-chaperone functions of CHIP may also contribute to the pathophysiology. Remarkably, the oligomerization of CHIP-WT with heat is thought to increase its intrinsic chaperone activity by enhancing the binding activity of CHIP to substrates [8], thereby providing an additional mechanism through which coding mutations could affect CHIP activity. Alternatively, CHIP could function as a sink for other components of the

BIOCHEMICAL AND CELLULAR STUDIES
Expression plasmids and recombinant proteins. Mammalian expression plasmids pcDNA3myc-CHIP, pcDNA3-myc-CHIP-K30A, pcDNA3-myc-CHIP-H260Q, HA-Ubiquitin, FLAG-SIRT6, FLAG-HSP70, β-galactosidase, and GFP were used as described previously [3,8,43,48,49]. Nuclear magnetic resonance spectroscopy. Human WT and T246M U-box (amino acid residues 212-303) recombinant proteins were expressed and purified as previously described for WT CHIP U-box [50]. NMR spectra were recorded using a Bruker Avance 600 MHz (1H) spectrometer at 20 °C in buffer containing 20 mM HEPES (pH 7.5), 50 mM NaCl, and 1 mM DTT as previously described [50]. NMR data were processed with NMRPipe     The test session consisted of three 10-min phases: a habituation period, a test for sociability, and a test for social novelty preference. For the sociability assay, mice were given a choice between being in the proximity of an unfamiliar conspecific ("stranger 1"), versus being alone. In the social novelty phase, mice were given a choice between the already-investigated stranger 1, versus a new, unfamiliar mouse ("stranger 2"). The social testing apparatus was a rectangular, 3- Training. On Day 1, each mouse was placed in the test chamber, contained in a soundattenuating box, and allowed to explore for 2 min. The mice were then exposed to a 30 s tone (80 dB), followed by a 2 s scrambled foot shock (0.4 mA). Mice received two additional shocktone pairings, with 80 s between each pairing.
Context-and cue-dependent learning. On Day 2, mice were placed back into the original conditioning chamber for a test of contextual learning. Levels of freezing (immobility) were determined across a 5 min session. On Day 3, mice were evaluated for associative learning to the auditory cue in another 5 min session. The conditioning chambers were modified using a Plexiglas insert to change the wall and floor surface, and a novel odor (dilute vanilla flavoring) was added to the sound-attenuating box. Mice were placed in the modified chamber and allowed to explore. After 2 min, the acoustic stimulus (an 80 dB tone) was presented for a 3 min period.
Levels of freezing before and during the stimulus were obtained by the image tracking system.
Two weeks following each test, mice were given second tests to evaluate retention of contextand cue-dependent learning.
Gait analysis. The CatWalk XT (Noldus information Technology, Wageningen, Netherlands) was used to analyze gait of unforced moving rats. CatWalk XT consists a hardware system of a long glass walkway plate, illuminated with green light that is reflected within the glass at points be touched, a high-speed video camera, and a software package for quantitative assessment of animal footprints. The parameters we observed included stride length: the distance between successive placements of the same paw; base of support: the average width between either the front paws or the hind paw; step cycle: the time in seconds between two consecutive initial contact of the same paw. All rats were trained to cross the runway in consistently at least six times a day for a week before any experimentation. A successful run was defined as an animal finishing the run down the tracks without any interruption or hesitation. Rats that failed the CatWalk training were excluded from the study. An average number of five replicate crossings by each rat was recorded. Rats were subjected to computer-assisted CatWalk monthly after 8 weeks-of-age.
Morris water maze. The Morris water maze task was used to assess learning and memory. The task was conducted in a round tank, 160 cm in diameter and 54 cm deep, filled with water. The wall was colored with non-toxic black paint to ensure opaqueness. Throughout testing, the water temperature was monitored and maintained at 21 ℃. The tank was divided into four equally sized quadrants, and a circular acrylic escape platform was placed in one of the quadrants. The escape platform was submerged in water by 2 cm so that it was not visible to the animals. A camera mounted above the tank recorded the movement of the animals in each trial. The Sunny Instruments Morris water maze Tracking Software was used to record the latency to reach the escape platform and the time spent in the target quadrant. The water maze task consisted of four training days with four trials on each day. In each trial, the animals were placed in the water facing the tank wall and had to locate the escape platform. The initial position of the animal was the vertices of one of the four quadrants and was different for each trial. It was assigned randomly and counterbalanced for the genotypes. Animals could utilize external visual cues on the walls surrounding the tank to locate the platform. The trial was completed when the animal either found the escape platform or 60 s had passed. If the animal was unable to locate the platform in 60 s, it was gently led to the platform. Animals were allowed to remain on the escape platform for 15 s before being removed and dried for the next trial. In addition, the platform position was kept constant between trials and days. The four trials of the first four training days were used as an indicator of spatial working memory. On day 5, the animals performed a 60 s probe trial without the platform. During this trial, the time spent in the target quadrant was recorded for each animal. FDR threshold of 1% was used, which is based on Picked protein FDR strategy, will also be estimated after protein inference (protein-level FDR <= 0.01). The protein quantification process includes the following steps: protein identification, tag impurity correction, data normalization, missing value imputation, protein ratio calculation, statistical analysis, results presentation. Protein identification was supported by all peptide matches with 95% confidence.
Comparisons were made between genotypes and proteins were considered to be differentially that met the criteria of p < 0.05 and absolute fold change > 1.2.
We filter excluded any proteins lacking interactions and used the Markov cluster algorithm (MCL) with an inflation parameter of three to identify clusters of related proteins based on their interaction network. Simultaneously, we used functional enrichment to identify biological processes, cellular components, and pathways that were over-represented in our protein list using a false discovery rate of less than 5%.

Acknowledgments
We