To determine the effects of exercise in SCA1, we implemented a mild exercise regimen in the Atxn1^154Q knock-in mice, which bear 154 CAG repeats targeted into the endogenous mouse locus to create a model that recapitulates all aspects of SCA1 (6). From 4 to 8 weeks of age, wild-type (WT) or Atxn1^154Q mice were placed on a fixed speed rotarod apparatus 5 times/week, while control mice were placed on an immobile rotarod apparatus. At 10 weeks of age we found no significant improvements in motor performance (Figure S1A and B), but Atxn1^154Q mice that were exercised showed a remarkable and highly significant extension of lifespan of (Figure 1A).

To determine the molecular mechanism of this rescue, we measured the expression of several growth factors in vulnerable tissues (cerebellum and brainstem) one week after the last exercise session. We found a sustained increase in the level of epidermal growth factor (EGF) in the brainstem but not the cerebellum (Figure 1B and S1). Since Drosophila data indicate that Capicua (Cic) lies downstream of EGF signaling (7), and we previously showed that Cic interacts with Atxn1(8), we measured Cic levels after exercise and found a significant decrease in the brainstem (Figure 1C and D) but not the cerebellum (Figure S1B). Notably, exercise did not affect brainstem Atxn1^154Q protein levels (Figure S1F). Additionally, primary brainstem neuronal cultures treated with recombinant EGF for 72 hours showed a dose-dependent decrease in Cic levels (Figure 1E). Thus, EGF regulates Cic levels in vivo and in vitro, and reduction of Cic might modulate the survival of Atxn1^154Q mice.

To test formally the effect of reduced Cic levels on SCA1 phenotypes, we generated a loss-of-function allele of Cic and backcrossed these mice onto a C57BL/6 strain (Figure S2A). Two isoforms of Cic, long (Cic-L) and short (Cic-S), are transcribed from alternative promoters. Cic-L^-/- mice completely lack the Cic-L isoform with ~10% of Cic-S remaining (Figure S2B). Importantly, Cic-L^+/- mice had a ~50% reduction of both Cic-L and Cic-S. Cic protein is reduced in Atxn1^-/- cerebellum (8), and we also found a Cic dose-dependent reduction of Atxn1 and its paralog Ataxin-1 Like (Atxn1L) in WT, Cic-L^+/-, and Cic-L^-/- cerebellum (Figure S2B), confirming the interdependency of Cic and Atxn1 paralog proteins in vivo.

To determine whether reducing the level of Cic would affect disease course in SCA1, we bred Atxn1^154Q male mice to Cic-L^+/- female mice to generate WT, Cic-L^+/-, Atxn1^154Q, and Atxn1^154Q; Cic-L^+/- mice, all on a pure C57BL/6 background. Because of their motor impairments, Atxn1^154Q mice showed less total locomotor activity in the open field assay than WT, but Atxn1^154Q; Cic-L^+/- mice performed significantly better than their Atxn1^154Q littermates (Figure 2A). Reduced Cic levels also significantly improved the motor coordination defects normally seen in Atxn1^154Q mice (Figure 2B and C). Reduced Cic levels also improved the learning and memory deficits Atxn1^154Q mice normally exhibit in the conditioned fear assay (Figure 2D). These improved phenotypes were accompanied by reduction in neuropathology: Atxn1^154Q; Cic-L^+/- mice had significantly more Purkinje cells than Atxn1^154Q mice at 40 weeks of age (Figure 3A and B). Thus, a 50% reduction of Cic levels is enough to mitigate the behavioral defects and Purkinje cell loss in Atxn1^154Q mice.

Reduction of Cic levels also improved phenotypes that could be attributed to other brain regions. Following weaning, Atxn1^154Q mice lost weight progressively, but loss of one Cic allele was enough to significantly rescue this weight loss (Figure 3C). As with the exercised Atxn1^154Q mice, the median age of survival of the Atxn1^154Q; Cic-L^+/- mice was significantly older than that of Atxn1^154Q mice (Figure 3D). These data likely explain why exercise improved the longevity of Atxn1^154Q mice but not the impaired coordination: exercise reduced Cic levels only in the brainstem and not in the cerebellum.

To determine the mechanism by which constitutive reduction of Cic rescues SCA1, we focused on the cerebellum, the primary site of dysfunction. We examined how the Atxn1^154Q protein influenced the transcriptional repressor function of Cic by comparing microarrays of Cic-L^-/- cerebella (Table S1) with previous microarrays from Atxn1^154Q cerebellum (9). We identified many “hyper-repressed” genes (upregulated in Cic-L^-/- but downregulated in Atxn1^154Q), with >50% containing a Cic motif (10) (Table S2). We selected several of these hyper-repressed target genes and found that genes that were downregulated in Atxn1^154Q cerebellum were in fact restored to near WT levels in Atxn1^154Q; Cic-L^+/- cerebellum, with significantly more Cic bound to their promoters in Atxn1^154Q vs WT mice (Figure 4A). This suggests that the mechanism of disease rescue is relief of Cic hyper-repression conferred by polyglutamine-expanded Atxn1.

In Drosophila, overexpression of Atxn1^82Q is rescued by overexpression of Cic and is exacerbated by Cic reduction (8, 11). Although this is likely due to a titration of the endogenous Drosophila Cic away from its normal transcriptional targets, it suggests that polyglutamine-expanded Atxn1 could cause a loss of Cic transcriptional function (de-repression). We identified many “de-repressed” genes (upregulated in both Cic-L^-/- and Atxn1^154Q cerebella, Table S3), that were in fact even further upregulated in Atxn1^154Q; Cic-L^+/- cerebellum, with significantly less Cic bound to their promoters in Atxn1^154Q vs WT mice (Figure 4B). This suggests that polyglutamine-expanded Atxn1 does indeed cause Cic to lose some transcriptional repressor activity, but this could not explain the genetic rescue; if partial loss of Cic activity was the driving factor in pathogenesis, the disease phenotype would be exacerbated in Atxn1^154Q; Cic-L^+/- mice. We propose that the polyglutamine-expanded Atxn1 causes Cic to bind more tightly to certain transcriptional targets (hyper-repressing them) while concomitantly causing Cic to bind less to, and thus upregulate (de-repress), other transcriptional targets. Genetic or exercise-induced reduction of Cic relieves the toxic hyper-repressive activity despite the concomitant loss of normal repressive function (Figure S3), consistent with the fact that the gain-of-function mechanism is the driving mechanism for toxicity in SCA1. While other polyglutamine diseases are caused by a gain-of-function mechanism despite partial loss of function of the involved protein (9, 11–13), here we have demonstrated that polyglutamine-expanded protein can cause concomitant gain- and loss-of-function effects on the same native protein partner. The level of polyglutamine-expanded Atxn1 protein was reduced by ~9% in Atxn1^154Q; Cic-L^+/- mice compared to Atxn1^154Q mice (Figure S2C and D), and while this could possibly contribute to the phenotypic rescue, we suggest that the rescue is more likely caused by relief of Cic-dependent hyper-repression. Therapeutics aimed at lowering Cic levels or disrupting the Cic-Atxn1^154Q protein interaction could potentially ameliorate the disease.

The effect of exercise on lifespan was long lasting, well after its discontinuation, underscoring the potential value of exercise beyond motor improvements. Thus, SCA1 individuals might benefit from an exercise program early in disease course (14, 15). It is interesting to note that genetic reduction of Cic also increased the lifespan of Atxn1^154Q mice (by 3.5 weeks), but the magnitude of the survival effect was greater (~ six weeks) in exercised Atxn1^154Q mice. Thus, other pathways such as enhanced growth factor signaling are likely to contribute to the effect. We cannot rule out the possibility that more intense or longer duration exercise could cause a sustained EGF increase and Cic decrease in the cerebellum that could lead to motor improvements. The exercise regimen we chose was quite gentle; a more rigorous or sustained exercise paradigm that engages the cerebellum more intensely might reduce cerebellar Cic levels and improve cerebellar phenotypes. It is encouraging that exercise and the accompanying increase in neuronal activity and metabolic demands do not seem to exacerbate the disease process in vulnerable neuronal populations, which may be important in a variety of neurodegenerative disorders.

SOM