Because Purkinje cells constitute less than 0.1% of cerebellar cells, we used laser capture microdissection (LCM) to enrich for these neurons so as to identify misregulated genes (Fig. 1a; Supplementary Fig. 1a). We used tissue from the SCA1 knock-in mice (henceforth SCA1 mice) that express an expanded version of ATXN1 with 154 glutamines, a model that closely mirrors human SCA1⁶.

Since the yield of RNA from LCM material was insufficient for microarray screening, we employed a PCR-based approach to test for expression of candidate genes involved in key neurodegenerative pathways. One of the genes that we found down-regulated encodes for vascular endothelial growth factor-A (VEGF), an angiogenic and trophic factor implicated in motor neuron disorders^7–9(Fig 1b). Importantly, VEGF levels were decreased as early as postnatal day 30 (Supplementary Fig. 1b), well before any behavioral or pathological signs. In the cerebellum, VEGF is widely expressed in neurons, glia and endothelial cells^10,11, with strong expression in Purkinje neurons (Fig. 1d, Supplementary Fig. 1c, d).

Low Vegf mRNA was accompanied by decreased VEGF protein in SCA1 cerebella (Fig. 1c, d). The decrease in VEGF was most prominent in Purkinje neurons (granule cells also demonstrated a trend toward lower Vegf mRNA levels and a significant decrease of VEGF protein; Supplementary Fig. 1e–h). We also found that Vegf mRNA levels were lower in a transgenic SCA1 mouse model that expresses mutant ATXN1 only in Purkinje neurons¹² (Supplementary Fig. 1i).

We next asked whether ATXN1 is directly responsible for down-regulation of Vegf mRNA. To test whether ATXN1 modulates Vegf promoter activity, we used a Vegf luciferase reporter assay¹³. Both expanded (ATXN1-84Q) and wild-type (ATXN1-2Q) ATXN1 repressed Vegf in a dose-dependent manner (Fig. 1e). The ability of even wild-type ataxin to cause repression is in keeping with the observation that overexpression of wild-type ATXN1 can induce pathology in animal models, and the notion that SCA1 can in part be caused by a gain of ATXN1 normal function¹⁴. Mutating a phosphorylation site critical for ATXN1 toxicity (serine 776 to alanine; S776A)¹⁵ disrupted VEGF repression (Fig. 1f). Thus ATXN1 repression of the Vegf promoter correlates with its toxicity in vivo. Consistent with these reporter experiments, VEGF levels were reduced in primary cerebellar neurons from SCA1 mice (Supplementary Fig. 2a). We next performed chromatin immunoprecipitation (ChIP) using an ATXN1-specific antibody to test for ATXN1 occupancy of the Vegf promoter. We found that the Vegf promoter (but not the promoter of a closely related family member Vegfc⁸) was indeed enriched in ChIP lysates (Fig. 1g, Supplementary Fig. 2b). The low expression of Vegf in SCA1 mice was associated with hypoacetylation at the Vegf promoter (Fig. 1h); furthermore, inhibitors of histone deacetylases relieved ATXN-1 84Q-induced repression at the Vegf promoter (Supplementary Fig. 2c). These results suggest a role for histone acetylation in ATXN1-induced repression.

Since VEGF is an angiogenic factor, diminished VEGF could contribute to cerebellar dysfunction by limiting angiogenesis. We observed a significant decrease in cerebellar microvessel density and total vessel length in SCA1 mice (Fig. 1i, Supplementary Fig. 3a–d). We also detected evidence for hypoxia in SCA1 cerebella using pimonidazole (a chemical that can detect the effects of hypoxia) (Supplementary Fig. 3e, f).

In addition to its role in angiogenesis, VEGF is also a neurotrophic factor. Thus, inadequate VEGF levels could also be deleterious by limiting neurotrophic support to cerebellar neurons. We tested the effects of reduced VEGF levels on mixed cerebellar neuronal cultures that express VEGF and its receptor VEGF R2 (the predominant VEGF receptor in neurons) (Supplementary Figs. 4, 5 and 6). We demonstrated that cerebellar neurons are susceptible to a decrease in VEGF signaling using VEGF R2 tyrosine kinase inhibitors and a neutralizing VEGF antibody. These experiments suggest that low VEGF levels compromise the growth and survival of not only Purkinje neurons, but also other cerebellar neurons through a reduction in neurotrophic support.

We then examined whether replenishing VEGF improves the SCA1 phenotype in mice. We mated the SCA1 mice to a transgenic VEGF line that overexpresses human VEGF in neurons starting at embryonic day 14 (Supplementary Fig. 7a, b and¹⁶). Genetically increasing VEGF enhanced the Rotarod performance of mice at 13 weeks and 6 months, and improved SCA1 cerebellar pathology as measured by calbindin staining intensity (calbindin specifically labels Purkinje cells) and molecular layer thickness (Fig. 2a–e). In addition, VEGF overexpression increased cerebellar microvessel density (Fig. 2f).

We next tested whether pharmacological delivery of recombinant VEGF is beneficial after disease onset. Since VEGF cannot cross the blood brain barrier, we used an intracerebroventricular route to deliver VEGF (Supplementary Fig. 7c, d). Mirroring our genetic rescue experiments, we observed improved Rotarod performance along with amelioration in pathology (Supplementary Fig. 8).

Our findings demonstrate an important role for VEGF in SCA1 pathogenesis and suggest that reversal of low VEGF levels offers the potential for meaningful treatment in patients with SCA1. In addition, these findings advance our understanding of SCA1 neurodegeneration in several ways. First, they suggest a cross-talk between the degenerating nervous system and the vascular system, an unexplored area in SCA1 research. Second, our results may explain the alterations in energy metabolism and decreased oxygen consumption identifiable by functional imaging in SCA1¹⁷. Third, our work on VEGF provides clues to develop biomarkers to monitor disease progression (such as VEGF itself, or sequelae of VEGF signaling in blood or cerebrospinal fluid of patients). Finally, since ataxia pathways interact in pathogenic hubs¹⁸ our results could prove relevant to other ataxias and possibly even other neurodegenerative syndromes.

1