Dysregulation of zebrin-II cell subtypes in the cerebellum is a shared feature across polyglutamine ataxia mouse models and patients

Abstract

Spinocerebellar ataxia type 7 (SCA7) is a genetic neurodegenerative disorder caused by a CAG-polyglutamine repeat expansion. Purkinje cells (PCs) are central to the pathology of ataxias, but their low abundance in the cerebellum underrepresents their transcriptomes in sequencing assays. To address this issue, we developed a PC enrichment protocol and sequenced individual nuclei from mice and patients with SCA7. Single-nucleus RNA sequencing in SCA7-266Q mice revealed dysregulation of cell identity genes affecting glia and PCs. Specifically, genes marking zebrin-II PC subtypes accounted for the highest proportion of DEGs in symptomatic SCA7-266Q mice. These transcriptomic changes in SCA7-266Q mice were associated with increased numbers of inhibitory synapses as quantified by immunohistochemistry and reduced spiking of PCs in acute brain slices. Dysregulation of zebrin-II cell subtypes was the predominant signal in PCs of SCA7-266Q mice and was associated with the loss of zebrin-II striping in the cerebellum at motor symptom onset. We furthermore demonstrated zebrin-II stripe degradation in additional mouse models of polyglutamine ataxia and observed decreased zebrin-II expression in the cerebella of patients with SCA7. Our results suggest that a breakdown of zebrin subtype regulation is a shared pathological feature of polyglutamine ataxias.

Figure 1:. Purkinje-enriched snRNA-seq yields a high-resolution map of SCA7 cerebellum

(A) Relative cell type abundance ratios captured by snRNA-seq with standard nuclear extraction and Purkinje-enriched nuclear extraction. Ratios for standard nuclear extraction were obtained from the transcriptomic mouse cerebellar atlas (26). (B) Relative cell type abundance ratios quantified from snRNA-seq WT and SCA7 samples. Mean cell type proportion from n = 4 animals plotted with error bars representing Standard Deviation. (C) Uniform Manifold Approximation and Projection (UMAP) dimensional reduction of Purkinje-enriched snRNA-seq data obtained from 5-week-old pre-symptomatic SCA7–266Q mice and wild-type controls. Points colored by genotype of origin (coral = SCA7–266Q; teal = WT). (D) UMAP dimensional reduction of Purkinje-enriched snRNA-seq data obtained from 8.5-week-old symptomatic SCA7–266Q mice and wild-type controls. Points colored by genotype of origin (coral = SCA7–266Q; teal = WT). (E) Differential gene expression in Purkinje cell nuclei of 5-week-old SCA7–266Q and WT mice using replicate-aware pseudobulk DEseq2. Each point represents a significant DEG (cutoffs: adjusted P < 0.05; |log2-fold change| > 0.5). Up and Downregulated genes are quantified and shown in black numbers. (F) Differential gene expression results per cell type for 8.5-week-old symptomatic SCA7–266Q mice and WT controls. Each point represents a significant DEG (cutoffs: adjusted P < 0.05; |log2-fold change| > 0.5). Up and Downregulated genes are quantified and shown in black numbers.

Figure 2:. Signatures of glial identity loss identified in SCA7–266Q mice

(A) UMAP dimensional reduction for Astrocytes, Bergmann Glia, and Oligodendrocytes of 8.5-week-old SCA7–266Q mice and WT controls. Points colored for genotype of origin (coral = SCA7–266Q; teal = WT). (B) Volcano Plots showing differentially expressed genes between SCA7–266Q and WT nuclei of Astrocytes, Bergmann Glia, and Oligodendrocytes. Log2FC and –log₁₀(p-value) for each gene is shown with significant DEGs in red (cutoffs: adjusted P < 0.05; |log2-fold change| > 0.5). (C) Venn-Diagrams showing the overlap of significant DEGs from each glial cell type with previously identified CLIP-seq targets of the QK RNA-binding protein (46). Hypergeometric test used to calculate overlap enrichment. Bonferroni adjusted p-value for overlaps shown along with Gene IDs.

Figure 3:. Progressive transcriptional dysregulation in Purkinje cells of SCA7–266Q mice highlights changes in synaptic signaling.

(A) UMAP dimensional reduction for Purkinje cell nuclei of 5-week-old SCA7–266Q mice and WT controls. (B) Volcano plot showing differential gene expression between presymptomatic 5-week-old SCA7–266Q and WT Purkinje cell nuclei with pseudobulk DEseq2. Significant DEGs shown in red (cutoffs: adjusted P < 0.05; |log2-fold change| > 0.5). (C) UMAP dimensional reduction for Purkinje cell nuclei of 8.5-week-old SCA7–266Q mice and WT controls. (D) Volcano plot showing differential gene expression between 8.5-week-old symptomatic SCA7–266Q and WT Purkinje cell nuclei with pseudobulk DEseq2. Significant DEGs in red (cutoffs: adjusted P < 0.05; |log2-fold change| > 0.5). (E) Gene Ontology categories for DEGs (320 DEGs) in 8.5-week-old SCA7–266Q mice calculated with ShinyGO v0.77. Biological Process categories are sorted by -Log10(FDR) with vertical cutoff lines drawn at equivalent adjusted p-values of (dotted: P = 0.01, dashed: P = 0.001, solid: P = 0.0001).

Figure 4:. PCs of symptomatic SCA7–266Q mice contain increased numbers of inhibitory synapses

(A) Immunohistochemistry performed on 30μm cerebellar sections imaged in the molecular layer of lobule VII in 5-week-old WT and SCA7–266Q mice using antibodies against the vesicular GABA transporter (VGAT, green) and Gephyrin (red) to label pre- and postsynaptic GABAergic terminals respectively and Calbindin-1 (blue) to label PC dendrites. Image insets contain arrows pointing to yellow co-localized signal of intact inhibitory synapses on Calbindin-1 positive PC dendrites. Orange arrow demarcates a non-PC synapse. (B) Quantification of inhibitory synapses by intact synapse numbers (VGAT and Gephyrin co-labeling) in 15 μm z-scans. Inhibitory synapses overlapping Calbindin-1 were classified as PC-synapses. n = 6 mice/genotype; two images per animal; statistics calculated with two-tailed t-test: PC Synapses, P = 0.114 (n.s)., non-PC Synapses, P = 0.0537 (n.s.). (C) IHC as in (A) performed in the molecular layer of lobule VII in 8.5-week-old WT and SCA7–266Q mice (green = VGAT, red = Gephyrin, blue = Calbindin-1). Image insets contain arrows pointing to yellow co-localized signal of intact inhibitory synapses on Calbindin-1 positive PC dendrites. Orange arrow demarcates a non-PC synapse. (D) Quantification of inhibitory synapses in 15 μm z-scans. Inhibitory synapses overlapping Calbindin-1 were classified as PC-synapses. n=6 mice/genotype; two images per animal; statistics calculated with two-tailed t-test: PC Synapses * P = 0.0195, non-PC Synapses P = 0.102 (n.s.). (E) IHC as in (A) performed in the molecular layer of Lobule II in 8.5-week-old WT and SCA7–266Q mice (green = VGAT, red = Gephyrin, blue = Calbindin-1). Image insets contain arrows pointing to yellow co-localized signal of intact inhibitory synapses on Calbindin-1 positive PC dendrites. Orange arrow demarcates a non-PC synapse. (F) Quantification of inhibitory synapses in 15 μm z-scans. Inhibitory synapses overlapping Calbindin-1 were classified as PC-synapses. WT n = 7 mice, SCA7 n = 5 mice; two images per animal; statistics calculated with two-tailed t-test: PC Synapses P = 0.0781 (n.s.), non-PC synapses P = 0.789 (n.s.).

Figure 5:. Increased inhibitory post-synaptic currents in SCA7 Purkinje cells are associated with reduced spiking rates.

(A) Representative traces of spontaneous inhibitory post-synaptic currents (sIPSC). Purkinje cells were recorded in acute cerebellar slices from 8.5-week-old SCA7–266Q mice and WT littermates. Whole-cell voltage-clamped sIPSCs were measured at +10 mV. (B and C) sIPSC amplitudes and inter-event-intervals were measured for each sIPSC event before randomly sampling an equal number of events from each cell, shown plotted in a cumulative histogram. Wilcoxson Rank Sum test for IPSC amplitudes, *** P = 2.48×10^-13. A total of n = 6 WT and n = 15 SCA7 Purkinje Cells were recorded across 3 animals of each genotype. (D) Representative traces of spontaneous excitatory post-synaptic currents (sEPSC). Purkinje cells were recorded in acute cerebellar slices from 8.5-week-old SCA7–266Q mice and matched WT littermates. Whole cell voltage-clamped sEPSCs were measured at −75 mV. (E and F) sEPSC cumulative amplitudes and inter-event-intervals were measured for each sEPSC and randomly sampled, shown plotted in a cumulative histogram. Wilcoxson Rank Sum test for EPSC inter-event-intervals, ***p=2.20×10^-16. A total of n=18 WT and n=17 SCA7 Purkinje Cells were recorded across 3 animals of each genotype. (G) Representative traces showing cell-attached recordings of action potentials from a Purkinje cell of a WT mouse and a SCA7–266Q mouse. (H) Action potentials were analyzed to find inter-event-intervals and randomly sampled from each cell to produce a cumulative histogram. Wilcoxon Rank Sum test for Action Potential Inter-Event-Intervals, ***p=2.53×10^-10. A total of n=14 WT and n=14 SCA7 Purkinje Cells were recorded across 3 animals of each genotype. (I) The coefficient of variation of action potential inter-event-intervals was calculated for each cell, as well as the CV2 statistic. WT and SCA7–266Q groups were compared with a two-tailed t-test, P = n.s.

Figure 6:. Zebrin-II subtype dysregulation explains highest proportion of SCA7 Purkinje cell DEGs

(A) UMAP dimensional reduction of Purkinje-enriched snRNA-seq data obtained from 8.5-week-old symptomatic SCA7–266Q mice and wild-type controls (as in Fig. 3A), showing normalized Aldoc (zebrin-II) expression for each single-nucleus. (B) UMAP dimensional reduction of Purkinje-enriched snRNA-seq data obtained from 8.5-week-old WT mice. Points colored by UMAP cluster. (C) UMAP dimensional reduction of snRNA-seq data from 8.5-week-old WT mice (as in B), showing normalized Aldoc (zebrin-II) expression counts for each single-nucleus. (D) Volcano Plot showing DEGs between WT zebrin-positive and WT zebrin-negative clusters (from Fig. 6B). DEGs calculated with Wilcoxon Rank Sum test with significant genes shown in red (cutoffs: adjusted P < 0.05; |log2FC| > 0.25). (E) Venn-Diagram showing overlap of DEGs in PCs of 8.5-week-old SCA7–266Q mice (as in Fig. 3D) with WT zebrin-axis genes quantified in Fig. 6D. Statistical enrichment calculated with Hypergeometric test and adjusted with Bonferroni correction of 13,078 representing number of gene ontology biological process categories (adjusted P = 5.15e-31***). (F) Gene Ontology (GO) biological processes from Figure 3B re-plotted with Zebrin-Axis gene category added. (G) Significant DEGs between Zebrin-positive PCs of 5-week-old WT and SCA7–266Q mice. Genes were categorized according to their expression in the 8.5-week WT Zebrin-Axis (see volcano plot in D) as non-Zebrin (black) if evenly expressed, Zebrin-positive (green) if higher expressed in Zebrin-positive PCs, or Zebrin-negative (yellow) if higher expressed in Zebrin-negative PCs. Total numbers of up and downregulated genes are reported in respective colored numbers. DEGs were calculated with Pseudobulk DEseq2 (cutoffs: adjusted P < 0.05; |log2FC| > 0.5). (H) Significant DEGs between Zebrin-positive PCs of 8.5-week-old WT and SCA7–266Q mice. (I) Significant DEGs between Zebrin-negative PCs of 5-week-old WT and SCA7–266Q mice. (J) Significant DEGs between Zebrin-negative PCs of 8.5-week-old WT and SCA7–266Q mice. (K) Zebrin-axis DEGs calculated from PCs of 5-week-old WT mice (as in D) plotted against Zebrin-axis DEGs from 5-week-old SCA7–266Q mice by log2FC value. Linear regression was applied to find correlation of log2FC values from WT and SCA7 (slope = 0.778). Points are colored red if log2FC difference between WT and SCA7 is greater than 0.25. (L) Zebrin-axis DEGs calculated from PCs of 8.5-week-old WT mice plotted against Zebrin-axis DEGs from 8.5 week-old SCA7–266Q PCs by log2FC value. Linear regression (slope = 0.819).

Figure 7:. Zebrin-II parasagittal stripes are lost in SCA7–266Q mice and models of related polyglutamine ataxias

(A) Immunohistochemistry (IHC) performed on coronal cerebellar slices from 8.5-week-old WT and SCA7–266Q mice stained for Zebrin-II/Aldoc (green) and Calbindin (blue). Scale bars = 1,000 μm. (B and C) Representative images of IHC against Zebrin-II/Aldoc across the central 3 stripes of Cerebellar Lobule VIII (left). Stripe Amplitudes (right) were calculated with a loess polynomial regression (blue line) on median pixel intensity values for each x-coordinate of the yellow boxed areas in the images on left. Disease models and ages tested: WT - 5 weeks (n=3), SCA7 −5 weeks (n=3), WT - 8.5 weeks (n=3), SCA7 – 8.5 weeks (n=3), SCA1–175Q - 12 weeks (n=3), SCA2–127Q - 12 weeks (n=2), SCA3-YAC (Q84/Q84) - 8 weeks (n=4), and HD-N171–82Q - 13 weeks (n=3). Note that age-matched WT littermates for SCA1, SCA2 and HD models were also analyzed for Zebrin stripe intensity. Scale bars = 500 μm. (D) Representative images for IHC against Plcb4 in 8.5-week WT and SCA7 animals paired with measured stripe amplitudes; n=3 mice / genotype. Scale bars = 500 μm. (E) Mean Stripe Amplitudes for each disease model compared to respective WT controls with a two-tailed t-test; n=3 mice / genotype. (SCA7 8.5 weeks, *p = 0.013; SCA1–175Q 12 weeks, *p= 0.016; SCA2–127Q 12 weeks, * P = 0.032; SCA3-YAC-84Q 8 weeks, * P = 0.049). Error bars = s.e.m.

Figure 8:. Cerebellar tissue from patients with SCA7 shows disruption of zebrin-related subtypes in molecular layer interneurons

(A) qRT-PCR for selected PC zebrin-subtype genes on bulk RNA extracted from patients with SCA7 (n=4) and unaffected control (n=3) cerebella. Fold change shown relative to mean Ct value for control samples; statistics calculated with one-tailed t-test. (ALDOC, * P = 0.022; PLCB4, P = 0.115; DGKH, P = 0.110; TRABD2B, P = 0.146; FAM107B, * P = 0.0498; CALB1, * P= 0.045). (B) UMAP clusters for snRNA-seq performed on human post-mortem cerebellar tissues from 5 patients with SCA7 and 4 unaffected controls colored by genotype (coral = SCA7; teal = Healthy). (C) Molecular Layer Interneuron (MLI) sub-clustering results showing strong co-clustering of SCA7 nuclei. Nuclei are colored by disease status (coral = SCA7; teal = Healthy). (D) Human MLI UMAP from (C) showing ALDOC expression. (E) UMAP sub-clustering of MLI-1 cells from the mouse cerebellar atlas colored by UMAP cluster. (F) Mouse MLI-1 UMAP from (E) showing Aldoa expression. (G) Volcano plot showing differentially expressed genes between Aldoc/Aldoa-high and Aldoc/Aldoa-low clusters from the mouse cerebellar atlas. (Wilcoxon Rank Sum test; cutoffs: adjusted P < 0.05; |log2-fold change| > 0.25). (H) Venn-Diagram showing overlap of SCA7 Human MLI-1 DEGs with Mouse MLI-1 Aldoc/Aldoa Gradient DEGs. Statistical enrichment of overlap computed with hypergeometric test against a background set of all genes expressed in greater than 5% of MLI-1 cells. Bonferroni corrected p-value (P = 3.59 × 10⁻³⁴).