Resilience to Endoplasmic Reticulum Stress Mitigates Membrane Hyperexcitability Underlying Late Disease Onset in a Murine Model of SCA6

Abstract

Objective

An enduring puzzle in many inherited neurological disorders is the late onset of symptoms despite expression of function‐impairing mutant protein early in life. We examined the basis for onset of impairment in spinocerebellar ataxia type 6 (SCA6), a canonical late‐onset neurodegenerative ataxia which results from a polyglutamine expansion in the voltage gated calcium channel, Cav2.1.

Methods

We performed serial transcriptome analysis with weighted gene correlation network analysis to investigate mechanisms for resilience in SCA6 mice that prevent onset of symptoms. We examined changes in membrane excitability that result in cerebellar Purkinje neuron spiking abnormalities through patch‐clamp recordings of Purkinje neurons in acute brain slices.

Results

Using unbiased transcriptome analysis, we identified endoplasmic reticulum (ER) stress as a driver of disease. Using spatial transcriptome analysis, we identified Purkinje neuron specific changes in unfolded protein response (UPR) related pathways. Novel activation of a store‐operated calcium current due to ER stress is the cause for Purkinje neuron spiking abnormalities in SCA6 mice. The impairments in Purkinje neuron spiking are unrelated to Cav2.1 ion‐flux function. Redundant pathways of the UPR act through a HSP90‐dependent mechanism to mitigate this ER stress.

Interpretation

Our studies support a model whereby proteotoxicity from misfolded mutant Cav2.1 is mitigated by a HSP90‐dependent UPR, and age‐related breakdown of this response causes motor dysfunction and aberrant Purkinje neuron spiking. These studies elucidate a mechanism of resilience connecting aberrant proteostasis and calcium‐dependent intrinsic membrane hyperexcitability to explain delayed disease onset more widely in age‐dependent neurodegenerative disease. ANN NEUROL 2026;99:502–522

Spinocerebellar ataxia type 6 (SCA6), a dominantly inherited cerebellar ataxia, is a canonical late‐onset neurodegenerative disorder with symptom onset in mid‐ to late‐life. Median age of disease onset in SCA6 is in the early 50s with some individuals who do not manifest symptoms until their 70s. ¹ SCA6 results from the expansion of a glutamine‐encoding CAG repeat expansion in the CACNA1A gene, which encodes the P/Q‐type voltage‐gated calcium channel, Cav2.1. ² Cav2.1 is vital for neurotransmission at the neuromuscular junction and for cerebellar Purkinje neuron function. Loss‐of‐function mutations in the channel cause early onset episodic ataxia or a congenital ataxia in humans. ³ , ⁴ , ⁵ Cacna1a knockout mice exhibit dystonia and ataxia apparent by postnatal day 10 and die by 3 to 4 weeks of age, ⁶ demonstrating the importance of the channel for cerebellar function early in life. It is therefore puzzling that symptom onset in SCA6 is late in life, in spite of the importance of Cav2.1 in early postnatal life.

SCA6^84Q/+ mice with a hyperexpanded CAG repeat knocked into one of the endogenous mouse Cacna1a alleles have been generated to model human SCA6. ⁷ Heterozygous SCA6^84Q/+ mice, the most precise genetic model currently available, develop motor impairment at 19 months of age, in a manner similar to late symptom‐onset in individuals with SCA6. Motor impairment in SCA6^84Q/+ mice is concurrent with abnormally irregular spiking of cerebellar Purkinje neurons. The irregular spiking is the likely driver of motor impairment, as improving spiking regularity pharmacologically, also improves motor dysfunction. ⁸ The relationship between the CAG repeat expansion in Cacna1a and alterations in spiking is unclear.

Purkinje neurons, the sole output from the cerebellar cortex, are autonomous pacemakers that fire action potentials at a high frequency even in the absence of synaptic input. ⁹ , ¹⁰ Purkinje neuron pacemaking is possible through the interplay of a specific group of ion channels, whose biophysical properties allow for fine tuning of the frequency and regularity of spiking. ¹⁰ In a number of mouse models of spinocerebellar ataxia (SCA) that result from a CAG repeat expansion in the respective genes, a reduction in firing frequency or spiking regularity are seen concurrent with onset of motor impairment. ⁸ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ A shared reduction in ion channel transcript levels of Kcnma1, Cacna1g, Trpc3, and Itpr1 is seen in SCA1, SCA2, and SCA7. ¹³ , ¹⁵ , ¹⁶ , ¹⁸ Although SCA6^84Q/+ mice display alterations in spiking similar to that seen in models of SCA1 and SCA7, the basis for the changes in spiking remains unexplored.

The endoplasmic reticulum (ER) is vital for proper protein folding. Protein folding in the ER is mediated through a number of resident chaperone proteins. The ER is also the major site of calcium storage in the cell. Perturbation of ER‐luminal calcium concentrations inhibits chaperone function. ¹⁹ A number of conditions, including protein misfolding and ER calcium depletion, result in ER stress. Cells activate a signaling pathway known as the unfolded protein response (UPR) to alleviate ER stress. There are 3 canonical pathways that have been identified, including XBP1, ATF6, and PERK‐eIF2α. ¹⁹ These pathways are thought to separately mediate ER stress responses under conditions of abnormal proteostasis. Xbp1 transcripts undergo unconventional splicing by IRE1α, which leads to generation of a transcription factor sXBP1. ²⁰ Increased function of XBP1 has been linked to enhanced chaperone‐mediated protein folding among other functions related to the UPR. ²¹ , ²² Upon ER stress, ATF6 translocates to the Golgi apparatus where it is cleaved to generate ATF6f. ATF6f subsequently induces expression of Xbp1 and regulates the UPR together with XBP1. ²³ , ²⁴ PERK activation induces phosphorylation of eIF2α (peIF2α), which inhibits protein synthesis to prevent ER protein overload. ²⁵ Under conditions of unmitigated ER stress, eIF2α phosphorylation tips cells toward apoptosis. ²⁶ Glutamine encoding CAG repeat (polyQ) disorders are postulated to result in misfolded proteins and activate ER stress responses. ²⁷ However, whether and how ER stress leads to disease in polyQ ataxia remains unknown.

We examined the basis for the late‐onset disease phenotype in SCA6^84Q/+ mice, where motor impairment is only evident at 19 months. Utilizing unbiased transcriptome analysis in presymptomatic SCA6^84Q/+ mice, we discovered a homeostatic response that engages the UPR at 6 months, long preceding onset of motor impairment. This resilience to ER stress is associated with alterations in all 3 canonical UPR pathways and mediated through the HSP90 chaperone machinery. ER stress‐mediated activation of a calcium‐release activated calcium current drives the increased membrane excitability and aberrant spiking in SCA6^84Q/+ Purkinje neurons. We have identified a mechanism for preservation of Purkinje neuron health in SCA6 that explains the late onset of the disease phenotype, and the mechanism for cerebellar impairment in SCA6.

Methods

Animals

All animal studies were reviewed and approved by the University of Texas Southwestern Medical Center. The SCA6^84Q/+ mice were generated by the laboratory of Dr Huda Y. Zoghbi ⁷ and deposited at the Jackson Laboratory. The SCA6^84Q/+ mice used in the current study were obtained from the Jackson Laboratory through recovery of frozen embryos (RRID: IMSR_JAX:008683). The XBP1^flox/flox mice were a gift from Dr Laurie Glimcher, who generated these mice. ²⁸ PCP2Cre^+/− mice (RRID:IMSR_JAX:010536), expressing cre recombinase solely in cerebellar Purkinje neurons ²⁹ were obtained from Jackson Research Labs. PCP2Cre^+/− and SCA6^84Q/+ mice in an XBP1^flox/flox background were first generated. PCP2Cre^+/− XBP1^flox/flox and SCA6^84Q/+ XBP1^flox/flox were crossed to subsequently generate PC XBP1^+/+SCA6^+/+ (XBP1^flox/floxSCA6^+/+ PCP2Cre^−/−), PC XBP1^−/−SCA6^+/+ (XBP1^flox/floxSCA6^+/+ PCP2Cre^+/−), PC XBP1^+/+SCA6^84Q/+ (XBP1^flox/floxSCA6^{84Q /+} PCP2Cre^−/−), and PC XBP1^−/−SCA6^84Q/+ (XBP1^flox/floxSCA6^84Q–/+ PCP2Cre^+/−) mice.

Acute Cerebellar Slice Preparation

Artificial cerebellar slice preparation (aCSF) was prepared as follows (in mmol/L): 125 NaCl, 3.8 KCl, 26 NaHCO₃, 1.25 NaH₂PO₄, 2 CaCl₂, 1 MgCl₂, and 10 glucose. Mice were anesthetized with isoflurane inhalation before decapitation. Brains were rapidly removed for acute slices preparation. Slices were prepared at 300 μM on a VT1200 vibratome (Leica Biosystems, Buffalo Grove, IL) using the “hot cut” technique. The aCSF were maintained between 33°C and 36°C during tissue sectioning. Following sectioning, the slices were incubated in pre‐warmed, carbogen‐bubbled (95% O₂, 5% CO₂) aCSF at 36°C for 45 minutes before recording. Slices were maintained in carbogen‐bubbled aCSF at room temperature for the rest of the experiment.

Patch Clamp Electrophysiology

Borosilicate glass patch pipettes of 3 to 5 MΩ were filled with an internal pipette solution containing (in mmol/L): 119 K‐gluconate, 2 Na‐gluconate, 6 NaCl, 2 MgCl₂, 0.9 EGTA, 10 HEPES, 14 tris‐phosphocreatine, 4 MgATP, and 0.3 tris‐GTP, at pH 7.3 and osmolarity 290 mOsm. Patch‐clamp recordings were performed at 33°C. Pre‐warmed, carbogen‐bubbled aCSF was perfused at a rate of 150 ml/hour for all recordings. Recordings were acquired using an Axopatch 200B amplifier, Digidata 1440A interface, and pClamp‐10 software (Molecular Devices, San Jose, CA). Current clamp data were acquired at 100 kHz in the fast current‐clamp mode with bridge balance compensation, and filtered at 2 kHz. Cells were included only if the series resistance did not exceed 15 MΩ at any point during the recording, and if the series resistance did not change by more than 20% during the course of the recording. Voltage data have been corrected for a 10 mV liquid junction potential. A prior study in SCA6^84Q/+ mice did not mention changes in spiking in relation to region of the cerebellum examined. ⁸ Recordings were, therefore, performed without regard to the anterior versus the posterior cerebellum. The majority of the recordings were performed from the anterior vermis and all cells were pooled for analysis.

The inhibitory post‐synaptic current (IPSC) recordings were performed with an internal pipette solution containing (in mmol/l) 150 CsCl, 10 HEPES, 0.5 CaCl₂, and 5 EGTA, at pH 7.2. Perforated patch clamp recordings used this same internal solution. ³⁰ Prior to perforated patch clamp IPSC recordings, gramicidin was dissolved in DMSO at a concentration of 2 mg/ml. The solution was then diluted in the electrode internal solution to obtain a final concentration of 5 μg/ml. Gramicidin containing electrode solution was made fresh every hour to ensure perforating potency. The solution was not filtered so as to not trap the antibiotic out of solution. The tip of the electrode was filled with regular internal solution to enhance seal formation. Input resistance was monitored through recording to ensure stable seal formation without significant leak. After obtaining baseline characterization in perforated patch mode, picrotoxin (50 μM) was perfused in the bath solution. The patch was then ruptured to generate conventional whole‐cell recordings.

Pharmacology

For some recordings, aCSF contained the following reagents (specified in the Results section for individual experiments): YM 58483 (Tocris, cat#: 3939; 20 μM), 17‐AAG (Cayman Chemicals, cat#: 11039; 1 μM), 6,7‐dinitroquinoxaline‐2,3‐dione (DNQX; Sigma Aldrich, cat#: D0540; 10 μM); and picrotoxin (Sigma Aldrich, cat#: P1675; 50 μM). Purkinje neurons were recorded under baseline aCSF condition for at least 5 minutes before switching to reagent‐contained aCSF. For 17‐AAG in particular, slices were pre‐incubated with 17‐AAG containing aCSF for 1 hour at 36°C before recording. Slices were perfused with reagent‐contained aCSF for at least 5 minutes and up to 10 minutes. All reported data were analyzed from the final 1 minute of the baseline recording, and the final 1 minute of the recording in the presence of the perfused compounds.

Cells for pharmacology were serially identified as long as they met the inclusion criteria for cell health and baseline firing properties. A predetermined number of 7 to 10 cells were used, based on our empirical studies indicating that with paired statistics this number is sufficient to identify a biologically meaningful change in firing frequency or coefficient of variation of the interspike interval (CV). The smaller number of cells for each set of pharmacology experiments will, therefore, not represent the mean of the larger total population of cells (typically greater than 25 cells), because there is considerable heterogeneity in the baseline firing properties. In order to correct for inherent biological variability and slice health, the investigator performing the studies was consistent for each set of studies, and the experimental and control groups were performed either concurrently or on subsequent days.

RNA Extraction and Quantitative Real Time‐Polymerase Chain Reaction

Mice were deeply euthanized with isoflurane inhalation before decapitation. The cerebella were rapidly removed, flash‐frozen in liquid nitrogen, and stored at −80°C for long‐term use. RNA was isolated from the whole cerebellum using TRIzol (Life Technologies, cat#: 11596‐026) and purified with RNeasy Mini Kit (Qiagen, cat#: 74104). Reverse transcription was performed with iScript cDNA Synthesis Kit (Bio‐Rad, cat#: 1708891) using a C1000 Touch Thermal Cycler (Bio‐Rad). Quantitative real‐time polymerase chain reaction (RT‐PCR) was performed with the iQ SYBR Green Supermix (Bio‐Rad, cat#: 1708887) using a CFX96 Touch Real‐Time PCR qPCR system (Bio‐Rad). Gene expression was normalized to Actin (Actb) levels. Cycles of threshold (Ct) values for each gene were obtained in triplicate and averaged for statistical comparisons. Delta Ct values were calculated as C_t ^{target gene} – C_t ^Actb. Relative fold changes in gene expression were calculated using the 2^−∆∆Ct method. ³¹ The primers (5′‐3′) used for quantitative RT‐PCR are listed below.

Cacna1g Forward: GTCGCTTTGGGTATCTTTGG.

Cacna1g Reverse: TACTCCAGCATCCCAGCAAT.

Cacna1a Forward: CCAGGAGCTCACCAAGGATG.

Cacna1a Reverse: GCAGACAGGGGACTCACTTC.

Kcnma1 Forward: AGCCAACGATAAGCTGTGGT.

Kcnma1 Reverse: AATCTCAAGCCAAGCCAACT.

Itpr1 Forward: GGCAGAGATGATCAGGGAAA.

Itpr1 Reverse: AGCTCGTTCTGTTCCCCTTC.

Trpc3 Forward: GAGGTGAATGAAGGTGAACTGA.

Trpc3 Reverse: CGTCGCTTGGCTCTTATCTT.

Kcnn2 Forward: ACCCTAGTGGATCTGGCAAA.

Kcnn2 Reverse: GGGTG ACGATCCTTTTCTCA.

Orai2 Forward: ACAGTCAGGCCTGGTCC.

Orai2 Reverse: TGGTGGTTAGACGTGACGAG.

Stim1 Forward: CCTTGTCCATGCAGTCCCC.

Stim1 Reverse: GGGTCAAATCCCTCTGAGATCC.

sXbp1 Forward: GAGTCCGCAGCAGGTG.

sXbp1 Reverse: GTGTCAGAGTCCATGGGA.

Actb Forward: CGGTTCCGATGCCCTGAGGCTCTT.

Actb Reverse: CGTCACACTTCATGATGGAATTGA.

RNA Sequencing and Weighted Gene Co‐Expression Network Analysis

The mRNA sequencing was performed on a NovaSeq instrument (Illumina, Inc.) with approximately 155 million reads and 150 bp paired‐end reads. Five cerebellar samples from each genotype from each age were used. Samples were prepared individually and pooled together to reduce batch effects. The samples were then demultiplexed into fastq files and statistics were collected using FastQC (Babraham Bioinformatics). The fastq files were trimmed using Trimmomatic. ³² One hundred fifty bp reads were aligned using STAR, ³³ and the percentages of unique reads were calculated. Raw read counts per gene per sample were calculated using HTSeq. ³⁴ Outliers were removed based on principal component analysis and hierarchical clustering. Differential gene expression analysis was performed using DESeq2. ³⁵

Weighted Gene Co‐Expression Network Analysis (WGCNA) was performed as described previously. ³⁶ Briefly, WGCNA was performed on a total of 34 RNA‐seq samples across 2 genotypes (WT and SCA6^84Q/+) and 4 age groups (3 months, 6 months, 12 months, and 19 months). The R software package for WGCNA ³⁷ was used to build a gene co‐expression network using filtered counts per million (CPM) data (CPM > = 1 across all replicates of one or more of each of the conditions). A signed network was constructed using the blockwiseModules function of the WGCNA R software package. A value of 16 was chosen as Beta with highest scale‐free R square (R² = 0.814). For other parameters, we used corType = pearson, maxBlockSize = 10,000, reassignThreshold = 1 × 10 ⁻² , mergeCutHeight = 0.05, deepSplit = 4, detectCutHeight = 0.995, and minModuleSize = 50. Visualizations of network plots were created using igraph R software package, ³⁸ representing the top 250 edges based on ranked weights.

Spatial Transcriptome Analysis

We designed a gene panel containing 354 genes using the Vizgen MERSCOPE platform. The gene panel included 254 differentially expressed genes (DEGs) that were identified at 6 months in SCA6^84Q/+ mice using bulk cerebellar RNAseq, in addition to cell markers, and some genes that were previously identified to be important for cerebellar function in this model of SCA6. ³⁹ , ⁴⁰ Ten μm sagittal cerebellar sections were prepared on a cryostat and imaging was performed using the manufacturer's recommended protocol. Cell segmentation was performed using a trained model to identify Purkinje neurons. Segmentation utilized the Vizgen Post‐processing Tool (version 1.3.0); and GitHub (https://github.com/Vizgen/vizgen-postprocessing). After cell segmentation MERFISH data quality check, cell clustering was performed using a standard workflow (https://satijalab.org/seurat/articles/) provided in Seurat Vignettes using RStudio (version 2024.12.1). Using a nonlinear dimensionality reduction algorithm, t‐distributed Stochastic Neighbor Embedding (tSNE), we performed clustering for each region from 2 independent experiments. We merged the dataset from 8 mice (total = 2 wild‐type [WT] and 2 SCA6^84Q/+ male mice and 2 WT and 2 SCA6^84Q/+ female mice) and after preprocessing, hierarchical clustering was performed on the merged dataset to identify groups of cells with similar gene expression profiles. Cells were initially grouped into 21 clusters. Based on cell marker expression, ⁴¹ we identified 11 major cell clusters (cell types) and the cell classes were annotated. Purkinje neurons were identified based on expression of marker genes, ⁴² Calb1, Skor2 and Ryr1, Itpr1, and Rora. Verification of accuracy of the number of Purkinje neurons in this cluster was confirmed by the observation that the number of Purkinje neurons in this cluster was similar to the number of Purkinje neurons that were manually counted across all the samples. Further, the categorical and tSNE coordinate data for the identified Purkinje neuron cluster was mapped back to the MERSCOPE visualizer to confirm that the spatial location corresponded to the cerebellar Purkinje neuron layer. Other marker genes used for annotation of cell clusters include Aldh1l1, Gfap, and Slc1a3 for astrocytes, Gabra6 and Grin2c for granule neurons, Slc32a1 for Golgi neurons, Fl11 for vascular cells, Fos, Junb, and Hdc for ependymal cells, Pax2, Npy, and Slc32a1 for molecular layer interneurons, Plcb4 for unipolar brush cells, Olig1, Olig2, Cldn11, Cspg4, and Mal for oligodendrocyte lineage cells, and Cx3cr1 for microglia. Differential gene expression analysis was performed using DESeq2.

Western Blot Analysis

Mice were deeply euthanized with isoflurane inhalation before decapitation. Cerebella were rapidly removed, flash‐frozen in liquid nitrogen, and stored at −80°C for long‐term use. Cerebellar tissue was homogenized using the Teflon tissue homogenizer tube in Igepal lysis buffer (50 mM Tris–HCl, 150 mM NaCl, 5 mM EDTA, 1 mM EGTA, pH 8.0, and 1% Igepal CA‐640; Sigma‐Aldrich, cat#: 4906845001) with protease (Sigma‐Aldrich, cat#: 4693132001) and phosphatase (Roche, cat# 04906837) inhibitor added. The whole cell lysates (WCLs) were then centrifuged at 14,000 rpm for 20 minutes at 4°C in temperature. The supernatant was then aliquoted and stored at −80°C. Lysates were subsequently thawed on ice, and protein concentration was determined using a Bicinchoninic acid (BCA) protein assay kit (Pierce BCA Protein Assay Kit, cat#: 23227). Samples were prepared for loading with 30 μg of protein denatured in 2X Laemmli Sample Buffer (Bio‐Rad, cat#: 1610737) with 5% of 2‐Mercaptoethanol (Sigma, cat#: M6250) at 95°C for 5 minutes. Protein samples were loaded on 10% precast gel (Bio‐Rad, cat#: 4561034). The gel was run using Tris‐Glycine‐SDS buffer (Bio‐Rad, cat#: 1610732) for 1.5 hours at a constant voltage of 100 V and the resolved proteins were transferred on a polyvinylidene difluoride (PVDF) membrane, using Tris‐Glycine buffer (Bio‐Rad, cat#: 1610734) in 20% methanol for 1.5 to 2 hours at a constant current of voltage of 100 V at 4°C. After transfer, the membrane was blocked with 5% skimmed milk in Tris‐buffered saline (TBS) and with tween 20 (T; TBS‐T; [0.1% Tween 20], for 1 hour at room temperature (RT). The membrane was then incubated with primary antibody overnight at 4°C according to the following dilutions in 5% BSA in TBS‐T for immunoblot analysis: sXBP1 (Protein tech, cat#: 24868‐1‐AP) at 1:1000, ATF6 (Cell Signaling Technology, cat#: 65880) at 1:1000, phosphorylated eIF2α (Cell Signaling Technology, cat#: 3398) at 1:1000, eIF2α (Cell Signaling Technology, cat#: 5324) at 1:1000, and GAPDH (Protein tech, cat#: 60004‐I‐Ig) at 1:5000 dilutions. After primary antibody incubation, the membrane was washed 3 × 10 minutes each with TBS‐T and then incubated with secondary antibodies in 5% skimmed milk for 1 hour on rocker at RT. The secondary antibody was removed, and the membrane was washed 3 × 10 minutes each with TBS‐T. The blots were developed using the Western blotting substrate (Thermo Scientific, cat#: 32209) using the Chemi‐Doc imaging system (Bio‐Rad) and densitometry was performed using Image Lab software from Bio‐Rad.

Tissue Immunofluorescence and Microscopy

Mice were deeply euthanized with isoflurane inhalation before decapitation. Brains were placed in 1% paraformaldehyde for 1 hour at RT before transferring to a solution of 30% sucrose in phosphate‐buffered saline (PBS) for 48 hours at 4°C. Brains were preserved in a 1:1 mixture of 30% sucrose in PBS:optimal cutting temperature (OCT) compound (Fisher Scientific, cat#: 23–730‐571) and stored at −80°C. Brains were sectioned to 20‐μm thickness using a CM1850 cryostat (Leica). Tissue was permeabilized with 0.4% triton in PBS and blocked with 5% goat serum in 0.1% triton in PBS. Tissue was incubated with primary antibodies in PBS containing 0.1% triton and 2% normal goat serum at 4°C overnight. The following primary antibodies were used for immunostaining: rabbit anti‐Cav2.1 (1:1000; Abcam, cat#: ab32642‐200), mouse anti‐Cav3.1 (1:150; NeuroMab, cat#: 75–206), sXBP1 (1:400; Cell Signaling Technology, cat#: 40435), mouse anti‐Calbindin (1:400; Sigma‐Aldrich, cat#: C9848), rabbit anti‐Calbindin (1:400; Sigma‐Aldrich, cat#: C2724), rabbit anti‐ORAI2 (1:200; Proteintech, cat#: 20592‐I‐AP), mouse anti‐STIM1 (1:200; Invitrogen, cat#: MA1‐19451), goat anti‐Calbindin (1:40; Invitrogen, cat#: PA5‐46936), and mouse anti‐ATF6alpha (1:500; Santa Cruz Biotechnology, cat#: sc‐166659). Secondary antibody in PBS was applied to tissue for an hour at RT. Secondary antibodies used were as follows: Alexa Fluor 488 goat anti‐mouse IgG (H + L; 1:200 or 1:500; Invitrogen, cat#: A11001), Alexa Fluor 488 goat anti‐rabbit IgG (H + L; 1:200 or 1:500; Invitrogen, cat#: A11034), Alexa Fluor 594 goat anti‐rabbit IgG (H + L; 1:200 or 1:500; Invitrogen, cat#: A11012), Alexa Fluor 594 goat anti‐mouse IgG (H + L; 1:200; Invitrogen, cat#: A11032), and Alexa Fluor 647 donkey anti‐goat IgG (H + L; 1:500, Invitrogen, cat#: A32849TR). Sections were imaged using an Axioskop 2 plus microscope (Zeiss). Cav2.1 and Cav3.1 staining intensity was quantified using ImageJ software by measuring the mean pixel intensity within a box covering the molecular layer of cerebellar lobule 5. The sXBP1 intensity was quantified using ImageJ software by measuring the mean pixel intensity within a circle covering the nuclear region of Purkinje neurons. Representative images were acquired on a C2+ confocal microscope (Nikon) at 60× magnification.

Statistical Analysis

Data were compiled in Excel and analyzed using SigmaPlot. Baseline data for firing was confirmed to be normally distributed using a Shapiro–Wilk test. Electrophysiology data for pharmacological studies were analyzed using either unpaired or paired Student's t‐test as appropriate. Statistical significance was determined with an α‐level of 0.05 for all studies. For RNAseq and spatial transcriptome analysis, criteria used for filtering genes were an absolute log2 fold change ≥0.3 and an adjusted p value ≤ 0.05.

Results

Irregular Purkinje Neuron Spiking in SCA6 is Not Due to Ion Channel Transcript Dysregulation

SCA6^84Q/+ mice, where a hyperexpanded CAG encoding polyglutamine (polyQ) repeat is knocked into one of the endogenous mouse Cacna1a alleles, represent a genetically precise model of the dominantly inherited human neurological disorder. Heterozygous SCA6^84Q/+ mice do not display motor impairment until 19 months of age. ⁷ , ⁸ In other models of SCA, motor dysfunction is associated with irregular cerebellar Purkinje neuron spiking. ¹³ , ⁴³ Prior work has identified Purkinje neuron spiking irregularity in 19‐month SCA6^84Q/+ mice. ⁸ It is unclear, however, whether asynchronous onset of irregular Purkinje neurons spiking reaches a threshold at 19 months to cause motor dysfunction. To examine Purkinje neuron spiking regularity prior to the onset of motor impairment, we performed noninvasive cell‐attached patch‐clamp recordings in acute cerebellar slices in SCA6^84Q/+ mice at near physiological temperatures at various ages. Purkinje neurons showed no change in spiking regularity or firing frequency in SCA6^84Q/+ mice at 3 months (Fig 1A–C and Supplementary Fig S1A), 6 months (Fig 1D–F and Supplementary Fig S1B), or 12 months (Fig 1G–I and Supplementary Fig S1C) of age. Significant spiking irregularly was observed only in 19‐month SCA6^84Q/+ mice compared to WT mice (Fig 1J–L), with no alteration in firing frequency (Fig S1D). Both male and female 19‐month SCA6^84Q/+ mice exhibited similar Purkinje neuron spiking irregularity (data not shown), and, therefore, subsequent studies were done without regard to sex. These data suggest that in SCA6^84Q/+ mice, spiking irregularity is likely the driver of motor impairment. Alterations in ion channel transcript levels are associated with spiking impairment in SCA1 ¹⁸ , ⁴³ , ⁴⁴ and SCA7 ¹³ mouse models. To examine whether changes in the expression of ion channels may account for irregular spiking, we used quantitative RT‐PCR and compared several key ion channel transcript levels in the cerebella of WT and SCA6^84Q/+ mice at different ages (Supplementary Fig S1E–H). The key ion channel transcripts examined included calcium channels (including Cacna1g, Cacna1a, Trpc3, and Itpr1), and calcium‐activated potassium channels (including Kcnma1 and Kcnn2). No significant changes in transcript levels of these ion channels were identified at 3 months of age (Supplementary Fig S1E). Although small but significant cerebellar changes were noted in transcript levels of several ion channels in 6‐month (Supplementary Fig S1F), 12‐month (Supplementary Fig S1G), and 19‐month SCA6^84Q/+ mice (Supplementary Fig S1H), the changes were inconsistent across different ages. We further compared protein levels of Cav2.1 (encoded by Cacna1a) and Cav3.1 (encoded by Cacna1g) in WT and SCA6^84Q/+ mice at 6 months and 19 months of age using immunostaining. The intensity of Cav2.1 and Cav3.1 staining in the cerebellar molecular layer were comparable between WT and SCA6^84Q/+ mice at both ages (Supplementary Fig S1I–L), indicating the expression of Cav2.1 and Cav3.1 are unchanged despite small changes in ion channel transcript levels. Taken together, these data suggest that age‐dependent Purkinje neuron dysfunction is not caused by changes in expression of ion channels.

FIGURE 1.

Irregular Purkinje neuron spiking in SCA6 is age dependent. (A) Representative trace of spontaneous Purkinje neuron spiking from acute cerebellar slices of 3‐month WT mice. (B) Representative trace of spontaneous Purkinje neuron spiking from acute cerebellar slices of 3‐month SCA6^84Q/+ mice. (C) Purkinje neuron spiking regularity, indicated by the coefficient of variation of the interspike interval, is comparable between 3‐month WT (N = 3) and SCA6^84Q/+ mice (N = 3). WT cells = 37 and SCA6^84Q/+ cells = 42. Student's t test, n.s. = not significant. (D) Representative trace of spontaneous Purkinje neuron spiking from acute cerebellar slices of 6‐month WT mice. (E) Representative trace of spontaneous Purkinje neuron spiking from acute cerebellar slices of 6‐month SCA6^84Q/+ mice. (F) Purkinje neuron spiking regularity, indicated by the coefficient of variation of the interspike interval, is comparable between 6‐month WT (N = 3) and SCA6^84Q/+ mice (N = 3). WT cells = 28, and SCA6^84Q/+ cells = 30. Student's t test, n.s. = not significant. (G) Representative trace of spontaneous Purkinje neuron spiking from acute cerebellar slices of 12‐month WT mice. (H) Representative trace of spontaneous Purkinje neuron spiking from acute cerebellar slices of 12‐month SCA6^84Q/+ mice. (I) Purkinje neuron spiking regularity, indicated by the coefficient of variation of the interspike interval, is comparable between 12‐month WT (N = 4) and SCA6^84Q/+ mice (N = 3). WT cells = 20, and SCA6^84Q/+ cells = 31. Student's t test, n.s. = not significant. (J) Representative trace of spontaneous Purkinje neuron spiking from acute cerebellar slices of 19‐month WT mice. (K) Representative trace of spontaneous Purkinje neuron spiking from acute cerebellar slices of 19‐month SCA6^84Q/+ mice showing spiking irregularity. (l) The coefficient of variation of the interspike interval, representing spiking regularity, is significantly abnormal in 19‐month SCA6^84Q/+ mice (N = 11) compared to WT mice (N = 4). WT cells = 35, and SCA6^84Q/+ cells = 64. Student's t test, ****p < 0.0001. WT = wild‐type. [Color figure can be viewed at www.annalsofneurology.org]

Irregular Purkinje Neuron Spiking in SCA6 Results from Altered Intrinsic Membrane Excitability

The pacemaker firing of Purkinje neurons is modulated by synaptic input. In order to examine whether the irregular spiking observed in 19‐month SCA6^84Q/+ mice is driven by a synaptic or intrinsic mechanism, we evaluated Purkinje neuron spiking in the presence of inhibitors of fast synaptic transmission. In cerebellar slices from 19‐month SCA6^84Q/+ mice, perfusion of a combination of picrotoxin, a GABA_A receptor antagonist, and 6,7‐dinitroquinoxaline‐2,3‐dione (DNQX), a non‐NMDA receptor antagonist, significantly improved Purkinje neuron spiking irregularity (Supplementary Fig S2A) without affecting firing frequency (Supplementary Fig S2B). Picrotoxin alone improved Purkinje neuron spiking irregularity (see Supplementary Fig S2A), whereas DNQX alone had no effect on modulating Purkinje neuron spiking irregularity (Supplementary Fig S2C) in 19‐month SCA6^84Q/+ mice. Neither picrotoxin alone nor DNQX alone altered Purkinje neuron firing frequency (Fig 2B and Supplementary Fig S2D). As expected from prior studies suggesting that tonic inhibition influences Purkinje neuron spiking regularity, ⁹ picrotoxin improved spiking regularity without affecting firing frequency in 19‐month WT Purkinje neurons (Supplementary Fig S2E, S2F). To further determine whether alterations in inhibitory neurotransmission in Purkinje neurons in SCA6^84Q/+ mice are responsible for irregular Purkinje neuron spiking, we examined spontaneous IPSCs. We measured the picrotoxin‐sensitive (Supplementary Fig S2G) IPSC frequency and regularity in Purkinje neurons of 19‐month WT and SCA6^84Q/+ mice. There was no difference in IPSC frequency (Fig 2C) or regularity (Fig 2D). These data suggest that although intact inhibitory synaptic transmission influences Purkinje neuron spiking regularity in general, it does not explain the difference in spiking regularity between WT and SCA6^84Q/+ Purkinje neurons.

FIGURE 2.

Irregular Purkinje neuron spiking in SCA6^84Q/+ mice is due to changes in intrinsic membrane excitability. (A) Picrotoxin improves Purkinje neuron spiking irregularity in 19‐month SCA6^84Q/+ mice (N = 2). n = 9 cells. Paired t test, *p < 0.05. (B) Picrotoxin has no effect on Purkinje neuron firing frequency in 19‐month SCA6^84Q/+ mice (N = 2). n = 9 cells. Paired t test, n.s. = not significant. (C) Frequency of IPSCs in Purkinje neuron is comparable between 19‐month WT (N = 2) and SCA6^84Q/+ mice (N = 2). WT cells = 5, and SCA6^84Q/+ cells = 11. Student's t test, n.s. = not significant. (D) Regularity of IPSCs in Purkinje neuron is comparable between 19‐month WT (N = 2) and SCA6^84Q/+ mice (N = 2). WT cells = 5, and SCA6^84Q/+ cells = 11. Student's t test, n.s. = not significant. (E) Representative trace of Purkinje neuron spiking in the cell‐attached configuration from 19‐month SCA6^84Q/+ mice. (F) Representative trace of Purkinje neuron spiking in the whole‐cell configuration from the same cell in E. (G) Patch break‐in in Purkinje cells whose spiking was monitored noninvasively in the cell‐attached configuration improves Purkinje neuron spiking irregularity in 19‐month SCA6^84Q/+ mice (N = 6). n = 20 cells. Paired t test, ****p < 0.0001. (H) Patch break‐in in Purkinje cells whose spiking was monitored noninvasively in the cell‐attached configuration increases Purkinje neuron firing frequency in 19‐month SCA6^84Q/+ mice (N = 6). n = 20 cells. Paired t test, ****p < 0.0001. (I) Purkinje neuron spiking regularity in the whole‐cell patch clamp configuration is comparable between 19‐month WT (N = 4) and SCA6^84Q/+ mice (N = 9). WT cells = 33, and SCA6^84Q/+ cells = 24. Student's t test, n.s. = not significant. (J) Characterization of the AHP in 19‐month WT and SCA6^84Q/+ Purkinje neurons. (K) The amplitude of the AHP as defined as the most hyperpolarized membrane potential, is similar between WT and SCA6^84Q/+ Purkinje neuron at 19 months. WT mice = 4, WT cells = 33, and SCA6^84Q/+ mice = 7, SCA6^84Q/+ cells = 26. Student's t test, n.s. = not significant. (L) The chloride reversal potential is comparable between 19‐month WT (n = 4 cells, N = 4 mice) and SCA6^84Q/+ Purkinje neurons (n = 4 cells, N = 4 mice). Student's t test, n.s. = not significant. (M) Left: Representative traces of IPSCs using gramicidin perforated‐patch recordings showing a chloride reversal potential of −90 mV in 19‐month SCA6^84Q/+ Purkinje neurons. Please note that current is inward at −100 mV and outward at −80 mV. Rs = 28 MΩ. Right: Representative traces showing IPSCs in 19‐month SCA6^84Q/+ Purkinje neuron in the whole‐cell patch clamp configuration. Please note that the IPSCs are inward at all the voltages shown. Rs = 10 MΩ. AHP = afterhyperpolarization; IPSC = inhibitory post‐synaptic current; WT = wild‐type. [Color figure can be viewed at www.annalsofneurology.org]

To examine whether altered intrinsic excitability may in fact be responsible for Purkinje neuron irregular spiking, we recorded spiking from the same 19‐month SCA6^84Q/+ Purkinje neurons in the cell‐attached configuration and subsequently the whole‐cell patch clamp configuration. If irregular Purkinje neuron spiking in SCA6^84Q/+ mice is due to a synaptic mechanism, the spiking irregularity seen in the cell‐attached configuration would be preserved in the whole‐cell patch‐clamp configuration in 19‐month SCA6^84Q/+ mice. We performed noninvasive cell‐attached recordings and following patch break‐in, whole‐cell voltage recordings in the same cells (Fig 2E, F). Spiking irregularity was significantly improved in the whole‐cell configuration compared to the cell‐attached configuration in 19‐month SCA6^84Q/+ Purkinje neurons (Fig 2G). Purkinje neuron firing frequency also increased significantly in the whole‐cell configuration (Fig 2H). Further, spiking regularity and firing frequency between 19‐month WT and SCA6^84Q/+ mice in the whole‐cell configuration were comparable (Fig 2I and Supplementary Fig S2H), unlike in the cell‐attached configuration where spiking was irregular in SCA6^84Q/+ mice compared to WT controls (see Fig 1L). These results suggest that an abnormality in intrinsic Purkinje neuron membrane excitability explains the irregular spiking in 19‐month SCA6^84Q/+ Purkinje neurons, and that dialysis/buffering of intracellular contents ameliorates the cause of the irregular spiking in SCA6^84Q/+ Purkinje neurons.

An intrinsic mechanism that results in irregular Purkinje neuron spiking is abnormalities in the afterhyperpolarization (AHP) phase of the action potential, ¹³ , ⁴³ , ⁴⁵ which is sensitive to intracellular calcium buffering. In addition, the AHP is strongly influenced by calcium flux through Cav2.1 and a reduction in Cav2.1 currents produces irregular Purkinje neuron spiking. ⁴⁵ To determine whether irregular Purkinje neuron spiking in 19‐month SCA6^84Q/+ mice results from changes in the AHP, we compared the amplitude of the AHP during spontaneous spiking, between 19‐month WT and SCA6^84Q/+ mice using whole‐cell patch‐clamp recordings with intracellular calcium buffering that preserves normal Purkinje neuron AHP amplitude. ¹³ , ¹⁶ , ⁴³ No significant changes in AHP amplitude were observed (Fig 2J, K). Other spike parameters analyzed, including spike peak, depolarization slope, repolarization slope, spike half‐width, and input resistance, also failed to reveal biologically meaningful differences between WT and SCA6^84Q/+ Purkinje neurons (Supplementary Table S1) that would explain the spiking irregularity. Taken together, these results suggest that the irregular Purkinje neuron spiking observed in 19‐month SCA6^84Q/+ mice results from an intrinsic mechanism different from alterations in expression or function of ion channels, including Cav2.1, previously implicated in producing this aberrant spiking phenotype.

Although GABA_A neurotransmission did not differentially affect spiking regularity between SCA6^84Q/+ and WT Purkinje neurons, during development, GABA_A‐neurotransmission causes membrane depolarization. This in turn is mediated by elevated cytosolic chloride concentrations and the resulting more depolarized chloride reversal potential. ⁴⁶ , ⁴⁷ To exclude the possibility that GABA_A transmission was depolarizing in SCA6^84Q/+ Purkinje neurons, we measured the chloride reversal potential in 19‐month WT and SCA6^84Q/+ Purkinje neurons using gramicidin perforated‐patch recordings. Gramicidin channels are impermeable to divalent cations and all anions, ³⁰ allowing us to measure the chloride reversal potential of spontaneous Purkinje neuron IPSCs. Perforated‐patch recordings revealed a chloride reversal potential of −90 mV in 19‐month SCA6^84Q/+ Purkinje neurons, which is comparable to the chloride reversal potential in WT Purkinje neurons (Fig 2L, M). Taken together, we conclude that an increase in intrinsic membrane excitability leads to irregular spiking in SCA6 Purkinje neurons.

Identification of ER Stress as a Potential Driver of Disease

Prior studies have demonstrated the utility of unbiased transcriptome analysis to identify the molecular basis for changes in Purkinje neuron intrinsic membrane excitability in cerebellar ataxia. ¹³ , ¹⁸ We performed bulk RNAseq from cerebella of WT and SCA6^84Q/+ mice at 3, 6, 12, and 19 months of age and used unbiased WGCNA to identify co‐expressed gene networks. We then examined which gene networks were significantly enriched in DEGs between WT and SCA6^84Q/+ mice. We determined whether DEGs between WT and SCA6^84Q/+ mice at 3, 6, 12, and 19 months of age could provide insight into mechanisms underlying resilience to spiking abnormalities at presymptomatic ages, and thereby also provide insight into mechanism for spiking irregularities in symptomatic 19‐month SCA6^84Q/+ mice. WGCNA identified correlated gene expression modules as a function of age (Fig 3A, top, and Supplementary Table S2). None of these gene expression modules displayed correlated changes as a function of genotype (SCA6^84Q/+ vs WT). Gene expression modules containing ion channel genes implicated in inherited ataxia are contained in the white and yellow modules. The yellow module had no significant enrichment of DEGs between WT and SCA6^84Q/+ mice at any age (Fig 3A, middle, and see Supplementary Table S2). No ion channel gene transcripts in the white module were differentially expressed between WT and SCA6^84Q/+ mice (see Fig 3A, middle, and Supplementary Table S2). The gene expression module that showed the most significant enrichment of genes differentially expressed between WT and SCA6^84Q/+ mice was the lightcyan module (highlighted in Fig 3A, bottom). The lightcyan module showed a highly significant enrichment of DEGs that were upregulated in SCA6^84Q/+ compared with WT mice at 6 months of age. Examination of the genes within the lightcyan module revealed ER‐resident UPR‐related genes, with transcripts for upregulated ER chaperones (Hsp90b1, Hyou1, Hspa5, Dnajb9, Dnajb11, Pdia3, and Dnajc3) and ER calcium buffers (Calr; Fig 3B) at 6 months of age. Among the transcripts that were significantly reduced at 12 months of age in the lightcyan module, Creld2 is also a UPR‐related gene (see Fig 3B). The UPR is a protective pathway that is activated by ER stress. ²⁵ UPR‐related genes are regulated by 3 major pathways, namely, XBP1, ATF6, and PERK‐eIF2α. The DEGs in the lightcyan module are known transcriptional targets of XBP1 and ATF6. ⁴⁸

FIGURE 3.

WGCNA identifies ER stress as a potential driver of disease in SCA6. (A) Top: WGCNA identifies correlated gene modules as a function of age but not as a function of genotype (SCA6^84Q/+ vs WT). Middle: Ion channel genes implicated in inherited ataxia are enriched in white and yellow modules. The yellow module exhibits no enrichment of DEGs between WT and SCA6^84Q/+ mice at any age. The ion channels gene transcripts in the white module are not differentially expressed in SCA6^84Q/+ mice. Bottom: WGCNA identifies genes within the lightcyan module (highlighted) as significantly upregulated in 6‐month SCA6^84Q/+ compared to WT mice. (B) Network plot showing the genes within the lightcyan module and the corresponding changes in expression at different ages between WT and SCA6^84Q/+ mice. There is upregulation of ER resident chaperones (Hsp90b1, Hyou1, Hspa5, and Dnajb11), folding enzymes (Pdia3, Pdia4, and Pdia6) and ER calcium buffers (Calr) in 6‐month SCA6^84Q/+ mice. Downregulation of the UPR‐related gene, Creld2, is noted in 12‐month SCA6^84Q/+ mice. The size of the circle represents the weight of the correlation with other module genes. Note that genes that are not outlined exhibit correlated expression with other module genes but the expression is not differentially altered in SCA6^84Q/+ mice (upregulated DEGs outlined in green; and downregulated DEGs outlined in blue). (C) Cells were grouped into transcriptionally distinct clusters using unsupervised clustering of MERFISH data. Eleven cell clusters, annotated by different colors, were identified based on cell marker expression. (D) Volcano plot of MERFISH data showing differentially expressed transcripts in SCA6^84Q/+ Purkinje neurons compared to WT littermate controls. Almost all the significantly altered DEGs were higher in SCA6^84Q/+ Purkinje neurons compared to WT controls. Note that the transcripts of the UPR‐related genes, Dnaja1, Hsp90aa1, Pdia3, and Hspa8, were significantly higher in SCA6^84Q/+ Purkinje neurons. Bdnf transcripts were also higher in SCA6^84Q/+ Purkinje neurons. Xbp1 (not sXbp1, which could not be assessed by MERFISH) transcripts were also higher in SCA6^84Q/+ Purkinje neurons, although not statistically significant (N = 4 WT and 4 SCA6^84Q/+ mice, which were age and sex matched, and pooled for analysis). DEGs = differentially expressed genes; ER = endoplasmic reticulum; UPR = unfolded protein response; WGCNA = Weighted Gene Correlation Network Analysis; WT = wild‐type.

In order to examine the specificity of the changes in the UPR for cerebellar Purkinje neurons, we performed spatial transcriptome analysis using multiplexed error‐robust fluorescence in situ hybridization (MERFISH) on a MERSCOPE platform. Probes were directed against the DEGs between 6‐month WT and SCA6^84Q/+ cerebella identified with bulk cerebellar RNAseq, other genes in the lightcyan module, and genes previously implicated in SCA6 pathogenesis. Following unbiased cell clustering, Purkinje neuron somata were identified based on their location and the expression of transcripts characteristic of Purkinje neurons (Fig 3C, and Supplementary Fig S3A–E). Ninety‐one gene transcripts were differentially expressed in SCA6^84Q/+ Purkinje neurons compared with WT littermate controls. Transcripts of the UPR related genes, Dnaja1, Hsp90aa1, Pdia3, and Hspa8, were significantly elevated in SCA6^84Q/+ Purkinje neurons (Fig 3D, Supplementary Table S3). This suggests that enhancement of UPR in response to ER stress may play a homeostatic role in SCA6^84Q/+ Purkinje neurons at presymptomatic ages.

Alterations in all Three Canonical UPR Pathways are Present in SCA6

Because we identified UPR‐related genes as the most significantly enriched DEGs between WT and SCA6^84Q/+ mice, we sought to examine whether there are changes in proteostasis in SCA6^84Q/+ mice. Three canonical UPR pathways are identified: XBP1, ATF6, and PERK‐eIF2α. ¹⁹ , ²⁵ Of these, XBP1 is the most evolutionarily conserved UPR pathway. ²⁵ We observed that Xbp1 is co‐expressed with other UPR‐related genes in the lightcyan module (see Fig 3B). Xbp1 undergoes non‐canonical splicing to generate the active transcription factor sXBP1. ²⁰ To investigate the role of XBP1 in proteostasis in pre‐symptomatic SCA6^84Q/+ mice, we examined spliced Xbp1 (sXbp1) transcript levels in the cerebella of 6‐month SCA6^84Q/+ mice. We examined activation of UPR pathways separately in male and female mice. A significant elevation of sXbp1 transcript levels was seen in both male and female SCA6^84Q/+ mice at 6 months of age, although the degree of sXbp1 elevation was higher in female mice (Fig 4A). We also examined the protein levels of sXBP1 in the whole cerebellum of 6‐month male and female SCA6^84Q/+ mice (Fig 4B). Paradoxically, sXBP1 protein levels were reduced in both male and female SCA6^84Q/+ mice, although the reduction was only significant in male mice (Fig 4C). We further examined protein expression of sXBP1 using immunostaining to explore the subcellular localization of sXBP1 in Purkinje neurons (Fig 4D). The sXBP1 predominantly localized to nuclei of Purkinje neurons. The intensity of sXBP1 was significantly reduced in the nuclei of 6‐month SCA6^84Q/+ Purkinje neurons (Supplementary Fig S4A). Interestingly, despite decreased intensity of sXBP1 staining in SCA6^84Q/+ Purkinje neuron nuclei (Fig 4D), we observed greater focal nuclear localization of sXBP1, suggesting active DNA binding, ⁴⁹ in 6‐month SCA6^84Q/+ Purkinje neurons, although this was not statistically significant (see Fig 4D, inset, and Supplementary Fig S4B). Together with transcriptome data (see Fig 3) showing upregulation of sXbp1 targets in cerebella and Purkinje neurons of 6‐month SCA6^84Q/+ mice, these data suggest enhanced function of sXBP1 in SCA6^84Q/+ Purkinje neurons. The UPR is canonically mediated by 2 other pathways in addition to XBP1; namely ATF6 and PERK‐eIF2α ²⁵ , ⁵⁰ Cerebellar ATF6 levels were significantly reduced in 6‐month male but not female SCA6^84Q/+ mice (Fig 4E, F). Cerebellar peIF2α relative to total eIF2α (peIF2α/eIF2α) levels were unchanged in both 6‐month male and female SCA6^84Q/+ mice (Fig 4G, H). Consistent with the immunoblots from the whole cerebellum, immunostaining for ATF6 revealed reduced ATF6 in Purkinje neurons (Supplementary Fig S4C).

FIGURE 4.

Alterations in all three canonical UPR pathways are present in SCA6. (A) Quantitative RT‐PCR of sXbp1 in 6‐month WT (N = 16) and SCA6^84Q/+ mice (N = 15). The transcript levels of sXbp1 are significantly increased in both 6‐month male and female SCA6^84Q/+ mice. Student's t test, *p < 0.05, **p < 0.01. (B) Representative Western blot showing reduced cerebellar sXBP1 levels in 6‐month male and female SCA6^84Q/+ mice and age‐ and sex‐matched WT mice. (C) Quantification of cerebellar sXBP1 normalized to cerebellar GAPDH levels in 6‐month male (N = 7) and female (N = 6) SCA6^84Q/+ mice compared with WT male (N = 7) and female (N = 6) mice. Cerebellar sXBP1 levels are reduced in both male and female SCA6^84Q/+ mice, although the reduction is only significant in male mice. Student's t test, **p < 0.01. (D) Immunostaining showing sXBP1 (green) and Calbindin (red) in 6‐month WT and SCA6^84Q/+ mice. Note overall reduced staining of sXBP1 in Purkinje neuron nuclei in SCA6^84Q/+ mice. Inset of WT Purkinje neuron (cell #3 from right) and SCA6^84Q/+ Purkinje neuron (cell #1 from right) showing that in some SCA6^84Q/+ Purkinje neurons, sXBP1 appears to form distinct puncta. (E) Representative Western blot showing reduced cerebellar ATF6 levels in 6‐month male SCA6^84Q/+ mice compared to 6‐month WT male mice. (F) Quantification of cerebellar ATF6 levels normalized to cerebellar GAPDH levels in 6‐month male (N = 7) and female (N = 6) SCA6^84Q/+ mice compared to WT male (N = 7) and female (N = 6) mice. Cerebellar ATF6 levels are reduced only in 6‐month male but not female SCA6^84Q/+ mice. Student's t test, **p < 0.01, n.s. = not significant. (G) Representative western blot showing comparable cerebellar levels of peIF2α/eIF2α in 6‐month female WT mice and female SCA6^84Q/+ mice. (H) Quantification of the ratio of cerebellar peIF2α to eIF2α levels, each separately normalized to cerebellar GAPDH levels, in 6‐month male (N = 4) and female (N = 3) SCA6^84Q/+ mice compared to WT male (N = 4) and female (N = 3) mice. Cerebellar peIF2α/eIF2α levels are unchanged in both 6‐month male and female SCA6^84Q/+ mice. Student's t test, n.s. = not significant. (I) Representative western blot showing reduced cerebellar peIF2α/eIF2α levels in 6‐month female PC XBP1^−/−SCA6^84Q/+ mice compared to 6‐month female PC XBP1^+/+SCA6^+/+ mice. (J) Quantification of cerebellar peIF2α/eIF2α levels in 6‐month female PC XBP1^−/−SCA6^84Q/+ mice (N = 3) compared to 6‐month female PC XBP1^+/+SCA6^+/+ mice (N = 3). Cerebellar peIF2α/eIF2α levels are significantly reduced in female PC XBP1^−/−SCA6^84Q/+ mice. Student's t test, *p < 0.05. (K) Quantitative RT‐PCR of sXbp1 in 19‐month WT (N = 8) and SCA6^84Q/+ mice (N = 8). The transcript levels of sXbp1 are comparable between WT and SCA6^84Q/+ mice. Student's t test, n.s. = not significant. (L) Quantification of cerebellar sXBP1 levels in 19‐month WT (N = 6) and SCA6^84Q/+ (N = 6) mice. Cerebellar sXBP1 levels are similar between 19‐month WT and SCA6^84Q/+ mice. Student's t test, n.s. = not significant. (M) Quantification of cerebellar ATF6 levels in 19‐month WT (N = 8) and SCA6^84Q/+ (N = 9) mice. Cerebellar ATF6 levels are significantly reduced in 19‐month SCA6^84Q/+ mice. Student's t test, ****p < 0.0001. (N) Quantification of cerebellar peIF2α/eIF2α levels in 19‐month WT (N = 6) and SCA6^84Q/+ (N = 6) mice. Cerebellar peIF2α/eIF2α levels are significantly reduced in 19‐month SCA6^84Q/+ mice. Student's t test, *p < 0.05. (O) Representative immunostaining images showing ATF6 (green) and Calbindin (red) in 19‐month WT mice. (P) Representative immunostaining images showing ATF6 (green) and Calbindin (red) in 19‐month SCA6^84Q/+ mice. (Q) Quantification of ATF6 intensity in 19‐month WT (N = 5 mice, 202 cells) and SCA6^84Q/+ (N = 5 mice, 182 cells) Purkinje neurons. The ATF6 intensity is significantly higher in SCA6^84Q/+ Purkinje neurons. Student's t test, **p < 0.01. RT‐PCR = real‐time polymerase chain reaction; UPR = unfolded protein response; WT = wild‐type.

To examine the role for XBP1 in SCA6^84Q/+ Purkinje neurons further, we generated SCA6^84Q/+ mice where XBP1 is deleted in cerebellar Purkinje neurons (Supplementary Fig S4D). We obtained WT mice with XBP1 in Purkinje neurons (PC XBP1^+/+SCA6^+/+), WT mice without XBP1 in Purkinje neurons (PC XBP1^−/−SCA6^+/+), SCA6^84Q/+ mice with XBP1 in Purkinje neurons (PC XBP1^+/+SCA6^84Q/+), and SCA6^84Q/+ mice without XBP1 in Purkinje neurons (PC XBP1^−/−SCA6^84Q/+). We first evaluated motor performance in all 4 genotypes using the rotarod assay at 1‐month and 6‐months of age. We expected PC XBP1^−/−SCA6^84Q/+ mice to show motor impairment similar to 19‐month SCA6^84Q/+ mice (Supplementary Fig S4E). We did not, however, observe motor impairment in PC XBP1^−/−SCA6^84Q/+ mice either at 1‐month or 6‐months of age (Supplementary Fig S4F). We then examined Purkinje neuron spiking regularity and firing frequency in all 4 genotypes. Consistent with the lack of a motor phenotype, Purkinje neuron spiking regularity and firing frequency were comparable across all genotypes (Supplementary Fig S4G, H). These results suggest that XBP1 alone does not mediate the resilience to ER stress in SCA6^84Q/+ Purkinje neurons and that redundancy in UPR pathways is present. Because there was a more prominent increase in sXbp1 transcript levels in female SCA6^84Q/+ mice, we examined protein levels of XBP1, ATF6, and peIF2α/eIF2α in 6‐month female PC XBP1^−/−SCA6^84Q/+ mice. Interestingly, cerebellar peIF2α/eIF2α levels were significantly downregulated in 6‐month female PC XBP1^−/−SCA6^84Q/+ mice (Fig 4I, J). Additionally, cerebellar sXBP1 and ATF6 levels were reduced, although the reduction was not statistically significant (Supplementary Fig S4I, J).

Our transcriptome data indicated a lack of upregulation of UPR‐related gene transcripts in 19‐month SCA6^84Q/+ mice. We therefore investigated changes in XBP1, ATF6, and peIF2α/eIF2α in 19‐month SCA6^84Q/+ mice. sXbp1 transcript levels were comparable between 19‐month WT and SCA6^84Q/+ mice (Fig 4K). We separately examined UPR pathways in 19‐month male and female SCA6^84Q/+ mice. Because no sex differences were observed, we pooled 19‐month male and female SCA6^84Q/+ cerebellar samples together for analysis. Cerebellar sXBP1 levels were similar between 19‐month WT and SCA6^84Q/+ mice (Fig 4L). Cerebellar levels of both ATF6 and peIF2α/eIF2α were reduced in 19‐month SCA6^84Q/+ mice (Fig 4M, N). Immunostaining for ATF6, however, displayed a small but significantly greater level of ATF6 in 19‐month SCA6^84Q/+ Purkinje neurons (Fig 4O–Q). These results suggest that UPR‐related pathways are engaged at all ages in SCA6^84Q/+ mice, indicating that ER stress is present across the lifespan of SCA6^84Q/+ mice.

ER Stress Induces Irregular Purkinje Neuron Spiking in SCA6

We wished to examine the relationship between the observed changes in Purkinje neuron spiking and upregulation of ER stress pathways in the cerebella of SCA6^84Q/+ mice. The UPR is triggered by ER stress, and a potent cause of ER stress is ER calcium depletion. ⁵¹ , ⁵² Depletion of ER calcium activates plasma membrane calcium‐release activated calcium (CRAC) channels. ⁵³ We therefore reasoned that ER calcium dyshomeostasis may be responsible for irregular Purkinje neuron spiking through the activation of CRAC channels in 19‐month SCA6^84Q/+ mice. We performed noninvasive cell‐attached recordings in 19‐month SCA6^84Q/+ mice, and examined changes in spike regularity following perfusion of 20 μM YM 58483 (also known as BTP2), a canonical CRAC channel inhibitor, ⁵³ in acute cerebellar slices. Spiking regularity and firing frequency were examined before YM 58483 perfusion (Fig 5A) and 5 minutes following YM 58483 perfusion (Fig 5B). Inhibiting the CRAC current in Purkinje neurons from 19‐month SCA6^84Q/+ mice significantly improved spiking irregularity (Fig 5C) and increased firing frequency (Fig 5D). To examine whether the effects were SCA6^84Q/+ specific, we perfused YM 58483 on acute cerebellar slices from 19‐month WT mice. YM 58483 did not alter the spiking regularity (Fig 5E) or firing frequency (Fig 5F) in WT mice, indicating that an increase in CRAC current is specific to Purkinje neurons in 19‐month SCA6^84Q/+ mice. The CRAC channel subunit enriched in Purkinje neurons is the Orai2/STIM1 subunit containing channel complex. ⁵⁴ Quantitative RT‐PCR revealed no changes in transcript levels of Orai2 or Stim1 in 3‐month (Supplementary Fig S5A), 12‐month (Supplementary Fig S5C), or 19‐month (Supplementary Fig S5D) SCA6^84Q/+ mice, although a small increase of both transcripts was noted in 6‐month SCA6^84Q/+ mice (Supplementary Fig S5B). Immunostaining for Orai2 and STIM1 revealed a modest but significant reduction in both these CRAC channel subunits in 19‐month SCA6^84Q/+ Purkinje neurons (Supplementary Fig S5E–H). No significant changes in transcripts of Orai1/2/3 or Stim1/2 were observed at any age in SCA6^84Q/+ mice in the RNAseq dataset (see Supplementary Table S2). YM 58483 inhibits TRPC3 channels in addition to CRAC channels, and TRPC3 channels have been reported to contribute to store‐operated calcium entry in Purkinje neurons. ⁵⁵ We, therefore, examined whether Pyr10, a TRPC3 inhibitor ⁵⁶ could improve spiking irregularity in 19‐month SCA6^84Q/+ mice. Pyr10 failed to improve spiking regularity (Fig 5G) or change firing frequency (Fig 5H) in 19‐month SCA6^84Q/+ Purkinje neurons, indicating that the observed increase in Purkinje neuron spiking irregularity is likely largely due to an increased CRAC current.

FIGURE 5.

Irregular Purkinje neuron spiking in 19‐month SCA6^84Q/+ mice results from age‐dependent activation of CRAC channels. (A) Representative trace of baseline spontaneous Purkinje neuron irregular spiking in 19‐month SCA6^84Q/+ mice before 20 μM YM 58483 perfusion. (B) Following perfusion of the CRAC channel inhibitor in the cell shown in panel A, spontaneous Purkinje neuron spiking irregularity improves in 19‐month SCA6^84Q/+ mice. (C) YM 58483 significantly improves Purkinje neuron spiking irregularity in 19‐month SCA6^84Q/+ mice (N = 4). n = 9 cells. Paired t test, *p < 0.05. (D) YM 58483 significantly increases Purkinje neuron firing frequency in 19‐month SCA6^84Q/+ mice (N = 4). n = 9 cells. Paired t test, ***p < 0.001. (E) YM 58483 has no effect on Purkinje neuron spiking regularity in 19‐month WT mice (N = 3). n = 8 cells. Paired t test, n.s. = not significant. (F) YM 58483 has no effect on Purkinje neuron firing frequency in 19‐month WT mice (N = 3). n = 8 cells. Paired t test, n.s. = not significant. (G) Pyr10 has no effect on Purkinje neuron spiking regularity in 19‐month SCA6^84Q/+ mice (N = 2). n = 8 cells. Paired t test, n.s. = not significant. (H) Pyr10 has no effect on Purkinje neuron firing frequency in 19‐month SCA6^84Q/+ mice (N = 2). n = 8 cells. Paired t test, n.s. = not significant. (I) Representative trace of baseline spontaneous irregular Purkinje neuron spiking in 14‐week Atxn1^154Q/2Q mice before YM 58483 perfusion. (J) Following perfusion of YM 58483 perfusion in the cell in I, there is no change in spiking irregularity in 14‐week Atxn1^154Q/2Q mice. (K) YM 58483 does not improve Purkinje neuron spiking irregularity in 14‐week Atxn1^154Q/2Q mice (N = 3). n = 10 cells. Paired t test, n.s. = not significant. (L) YM 58483 has no effect on Purkinje neuron firing frequency in 14‐week Atxn1^154Q/2Q mice (N = 3). n = 10 cells. Paired t test, n.s. = not significant. CRAC = calcium‐release activated calcium. [Color figure can be viewed at www.annalsofneurology.org]

To investigate whether irregular Purkinje neuron spiking as a result of an increased CRAC current is unique to SCA6^84Q/+ mice, we perfused YM 58483 on acute cerebellar slices from 14‐week SCA1 mice (Atxn1^154Q/2Q mice), where Purkinje neuron spiking irregularity is similar to SCA6^84Q/+ mice. ⁴³ Inhibiting CRAC had no effect on Purkinje neuron spiking regularity or firing frequency (Fig 5I–L), which is consistent with our prior studies showing that the irregular spiking in SCA1 is due to a reduced AHP and potassium channel transcripts. ⁴³ Taken together, our results indicate that an increase in CRAC current is the cause of the spiking irregularity in Purkinje neurons of symptomatic SCA6^84Q/+ mice.

HSP90 Mitigates ER Stress‐Induced Membrane Hyperexcitability

We wished to examine the mechanism for UPR‐mediated compensation that prevents CRAC channel activation in SCA6^84Q/+ Purkinje neurons. Of the many molecular chaperones whose transcripts are upregulated, we identified HSP90 as particularly important. HSP90 has been previously implicated in other late‐onset neurodegenerative diseases characterized by protein aggregation, ⁵⁷ and upregulation of the ER‐resident HSP90 isoform, GRP94 (encoded by Hsp90b1) is considered a hallmark of the UPR. ⁵⁸ We observed upregulation of transcripts of Hsp90b1, in addition to upregulated transcripts of Hsp90aa1, Hsp90ab1 (cytosolic isoforms of HSP90), and the HSP90 co‐chaperone Ahsa1 (AHA1) in 6‐month SCA6^84Q/+ cerebella (see the Table 1 and Supplementary Table S2).

Table 1.

Transcript Levels of Hsp90‐Related Genes are Upregulated in 6‐Month SCA6^84Q/+ Mice

SCA684Q/+ × WT
6 Months
Gene	Base Mean	Log2 Fold Change	LfcSE	Stat	p	p adj
Hsp90ab1	86001.45736	0.404713493	0.10649	3.80051	0.000144	0.003722
Hsp90b1	36971.0823	0.350081462	0.04772	7.33671	2.19E‐13	1.10E‐10
Hsp90aa1	74180.79037	0.331443086	0.04502	7.36153	1.82E‐13	1.06E‐10
Ahsa1	7838.91479	0.485988819	0.08152	5.96145	2.50E‐09	4.55E‐07

p adj = adjusted p value; WT = wild‐type.

We investigated whether inhibiting HSP90 in 6‐month presymptomatic SCA6^84Q/+ mice could activate CRAC channels to produce the irregular Purkinje neuron spiking seen in symptomatic 19‐month SCA6^84Q/+ mice. We performed recordings in the presence of 1 μM 17‐AAG, a pan‐HSP90‐specific reversible inhibitor, ⁵⁹ to examine the role for HSP90 in maintaining regular spiking in 6‐month SCA6^84Q/+ Purkinje neurons. The 17‐AAG significantly increased Purkinje neuron spiking irregularity in 6‐month SCA6^84Q/+ mice (Fig 6A, B) but not in WT mice (Fig 6B, and Supplementary Fig S6A). The 17‐AAG did not alter Purkinje neuron firing frequency in 6‐month SCA6^84Q/+ mice (Supplementary Fig S6B). Importantly, perfusion of the CRAC current inhibitor, YM 58483, significantly improved the 17‐AAG‐induced irregular Purkinje neuron spiking in 6‐month SCA6^84Q/+ mice (Fig 6C) without altering firing frequency (Supplementary Fig S6C). Additionally, YM 58483 did not produce a significant effect on Purkinje neuron spiking regularity (Fig 6D) or firing frequency (Supplementary Fig S6D) in 17‐AAG‐treated 6‐month WT cerebellar slices. ER stress is often associated with ER calcium depletion. ER stress can, however, activate CRAC channels independently of ER calcium depletion. ⁶⁰ We examined whether ER calcium depletion alone is sufficient to explain the irregular spiking in SCA6^84Q/+ Purkinje neurons. We used thapsigargin, an agent that depletes ER calcium stores in cerebellar slices from 6‐month WT and SCA6^84Q/+ mice. Purkinje neuron firing regularity (Supplementary Fig S6E) was similar between thapsigargin treated WT, SCA6^84Q/+ Purkinje neurons and DMSO (the diluent) treated WT Purkinje neurons. In addition, a CRAC channel inhibitor was unable to significantly improve spiking regularity (Supplementary Fig S6F) in thapsigargin treated SCA6^84Q/+ Purkinje neurons. These results suggest that the ER stress observed in SCA6^84Q/+ Purkinje neurons does not act solely through ER calcium depletion to activate CRAC channels.

FIGURE 6.

HSP90 mitigates ER stress‐induced membrane hyperexcitability in pre‐symptomatic SCA6^84Q/+ mice. (A) Representative trace of irregular Purkinje neuron spiking in 6‐month SCA6^84Q/+ cerebellar slices incubated with 1 μM 17‐AAG, a pan‐HSP90 reversible inhibitor. (B) Incubation with 1 μM 17‐AAG has no effect on Purkinje neuron spiking regularity in 6‐month WT mice (N = 3). In 6‐month SCA6^84Q/+ (N = 5) cerebellar slices incubated with 1 μM 17‐AAG, Purkinje neuron spiking is irregular compared to WT mice. Dashed lines represent baseline Purkinje neuron spiking regularity from Figure 1F. WT cells = 29, SCA6^84Q/+ cells = 30. Student's t test, *p < 0.05. (C) YM 58483 improves 17‐AAG‐induced Purkinje neuron spiking irregularity in 6‐month SCA6^84Q/+ mice (N = 2). n = 10 cells. Paired t test, **p < 0.01. (D) YM 58483 has no effect on Purkinje neuron spiking in 17‐AAG‐pretreated cerebellar slices from 6‐month WT mice (N = 2). n = 8 cells. Paired t test, n.s. = not significant. (E) 17‐AAG has no effect on Purkinje neuron spiking regularity in 19‐month WT mice. DMSO‐treated mice = 2, DMSO‐treated cells = 29; 17‐AAG‐treated mice = 3, 17‐AAG‐treated cells: n = 29. Student's t test, n.s. = not significant. (F) 17‐AAG activates a CRAC current in 19‐month WT Purkinje neurons as indicated by YM 58483 improving Purkinje neuron spiking regularity in 17‐AAG‐pretreated 19‐month WT cerebellar slices (N = 2). n = 10 cells. Paired t test, **p < 0.01. CRAC = calcium‐release activated calcium; ER = endoplasmic reticulum; WT = wild‐type. [Color figure can be viewed at www.annalsofneurology.org]

Next, we sought to understand whether ER‐stress related activation of CRAC channels occurs normally as a function of age. Inhibition of HSP90 with 17‐AAG did not alter Purkinje neuron spiking regularity (Fig 6E) or firing frequency (Supplementary Fig S6G) in 19‐month WT mice. Unlike Purkinje neurons from 6‐month WT mice, inhibition of the CRAC current significantly improved Purkinje neuron spiking regularity (Fig 6F) without changing firing frequency (Supplementary Fig S6H) in 17‐AAG treated 19‐month WT cerebellar slices, indicating that 17‐AAG was able to activate a CRAC current in 19‐month WT Purkinje neurons. These results suggest that there is an age‐related compromise of the ability of HSP90 to mitigate ER‐stress‐related activation of CRAC channels, even though it does not reach the threshold, at least at 19 months of age, to cause abnormal Purkinje neuron spiking in WT mice. Taken together, our data indicate that activation of the UPR serves as a compensatory mechanism to mitigate ER‐stress induced by polyQ Cav2.1 in SCA6 and that this resilience mechanism is compromised as a function of age, resulting in irregular Purkinje neuron spiking and disease onset.

Discussion

A number of neurodegenerative disorders are associated with protein misfolding. It is, however, unclear whether aberrant proteostasis‐induced ER stress is a driver of disease, or a target for therapeutic intervention. Our unbiased identification of a co‐expressed gene network that is associated with resilience to ER‐stress allows an interrogation of genes that are protective in late‐onset neurodegenerative disease beyond SCA6. For example, examination of the network of genes in the lightcyan module identifies Nrn1 expression as co‐expressed with other ER‐resilience genes regulated by XBP1, although Nrn1 is not differentially expressed in SCA6^84Q/+ mice. Nrn1, which encodes Neuritin, is a neurotrophic factor implicated in cognitive resilience in Alzheimer's disease, ⁶¹ but the mechanism through which it confers dendrite resilience to degeneration is unclear. Our data suggest that Nrn1 may confer resilience to age‐related cognitive decline through mitigating ER stress. These results also suggest that mitigation of ER stress may not be uniformly mediated by the same set of genes, and that resilience is conferred by genes that are unique to specific age‐related neurodegenerative disorders.

The mechanism of disease in SCA6 is uncertain. When the CAG‐repeat expansion in CACNA1A was initially identified as the cause of SCA6, and because SCA6 is allelic with episodic ataxia type 2, a childhood onset episodic disorder resulting from CACNA1A haploinsufficiency and without progressive neurodegeneration, ⁴ , ⁵ it was suggested that a loss of channel function could be responsible for disease. This was supported by in vitro studies in heterologous overexpression systems indicating that the biophysical properties of the channel were altered by the polyQ expansion in Cav2.1. ⁶² , ⁶³ , ⁶⁴ In Purkinje neurons of SCA6^84Q/+ mice ⁷ and in human iPSC derived cortical neurons from individuals with SCA6, ⁶⁵ however, the biophysical properties and current density of Cav2.1 are unchanged. Recent work has suggested that an alternative transcript of CACNA1A, α1ACT, residing in the C‐terminus that contains the polyQ region, may contribute to disease in SCA6. ⁶⁶ The α1ACT acts as a transcription factor regulating genes involved in neurite outgrowth in early development. ⁶⁶ Viral‐mediated expression of polyQ‐α1ACT at postnatal day 1 results in motor impairment and Purkinje neuron degeneration by 4‐postnatal weeks in mice. ⁶⁷ The early onset phenotype of polyQ‐α1ACT transgenic mice is inconsistent with the late onset phenotype in human SCA6. SCA6^84Q/+ mice do not appear to have polyQ‐α1ACT expression ⁷ and also more closely recapitulate the late onset of symptoms that is seen in human disease. A recent study suggests that in neural cells overexpressing either polyQ‐α1ACT or polyQ‐Cav2.1, mutant protein localizes to the ER, induces ER stress, and mediates apoptosis in association with activation of the PERK‐eIF2α pathway. ⁶⁸ This suggests that ER stress is the primary mediator of disease in SCA6 rather than transcriptional dysregulation. The 84‐repeat polyQ pathogenic expansion in SCA6^84Q/+ mice, however, significantly exceeds the modest 21‐30‐polyQ pathogenic expansion in humans. ² The relevance of our identification of ER stress in SCA6^84Q/+ mice should be interpreted cautiously in human SCA6 where the polyQ expansion is much more modest. A recent study revealed changes in mitochondrial structure and function following disease onset and during disease progression in SCA6^84Q/84Q mice. ⁶⁹ Homozygosity for the hyperexpanded polyQ repeat in Cav2.1 advances the age of onset to 7 months as compared with 19 months in heterozygous SCA6^84Q/+ mice. ⁷ , ⁷⁰ It is possible that the observed mitochondrial phenotypes are secondary to uncompensated ER stress due to the increased gene dosage of mutant Cacna1a in these mice. Future studies could examine the presence of ER stress in human iPSC derived neurons from individuals with SCA6.

Changes in Purkinje neuron spiking, with changes in either firing regularity or frequency, or both, are a consistent feature of disease in mouse models of spinocerebellar ataxia, ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁶ , ¹⁷ , ⁴³ including SCA6. ⁸ The changes in firing are likely the cause of the motor phenotype in SCA6 as in SCA6^84Q/84Q mice, pharmacologically improving firing properties is associated with improvement in motor dysfunction. ⁸ In addition, the improvement in motor function by exercise and increasing BDNF signaling, is associated with correction of spiking abnormalities in SCA6^84Q/84Q mice. ⁴⁰ Our identification of elevated Bdnf transcripts specifically in Purkinje neurons in presymptomatic mice suggests that BDNF plays a role in preserving function in SCA6. The identification of spiking changes mediated by increased CRAC current could suggest that inhibition of CRAC channels may be a therapeutic strategy to improve motor impairment in SCA6. CRAC channels, however, play an important role in cerebellar learning and STIM1, the most abundant STIM subunit of CRAC channels in Purkinje neurons, plays an essential role in mGluR1 dependent synaptic transmission and motor behavior. ⁵⁵ Selective loss of STIM1 in Purkinje neurons results in motor impairment, with even haploinsufficiency of STIM1 resulting in motor deficits. ⁷¹ Targeting CRAC channels as a strategy to improve motor impairment in SCA6 would, therefore, likely not be a viable therapeutic strategy. Our identification of a relationship between ER stress and Purkinje neuron firing dysfunction suggests that the activation of specific HSP90 pathways has potential to improve motor impairment. Identification of chemical chaperones that promote folding of polyQ Cav2.1 is also a potential therapeutic strategy for SCA6. Promoting channel folding is a well‐established strategy in a different channelopathy, cystic fibrosis, where CFTR correctors that promote channel folding ⁷² are also able to relieve ER stress. ⁷³

Three canonical pathways are recognized to mitigate the deleterious consequences of ER stress. These are mediated by the XBP1, ATF6, and PERK‐eIF2α pathways. UPR targets differ depending on cell type and context. ²⁵ Remarkably, the gene transcripts in the lightcyan module mediating the protective response in 6‐month SCA6^84Q/+ mice, including Pdia4, Hsp90b1, Hspa5, Hyou1, Pdia6, Calr, and Pdia3, overlap with gene transcripts that have been previously observed to be upregulated from enforced dimerization of sXBP1 and ATF6f.⁴⁸ This enforced dimer of ATF6f/sXBP1 reduced the abnormal aggregation of mutant huntingtin and α‐synuclein in vitro and demonstrated neuroprotection in preclinical models of Huntington's and Parkinson's disease. Our findings in SCA6 suggest that endogenous homeostatic mechanisms too can harness responses from more than one arm of the 3 canonical UPR pathways. We also identified redundancy the UPR pathways mitigating ER stress in SCA6. Deleting XBP1 specifically in cerebellar Purkinje neurons in SCA6^84Q/+ mice resulted in downregulation of both ATF6 and peIF2α. Increased peIF2α has been linked to ER stress‐induced apoptosis. ²⁶ Our results suggest that when XBP1‐mediated upregulation of chaperone responses is abrogated, Purkinje neurons utilize alternative UPR pathways of downregulating peIF2α and ATF6. Additionally, our results suggest enhanced function of XBP1 in presymptomatic SCA6^84Q/+ cerebella despite decreased protein levels. XBP1 has been shown to play competing roles in ER stress. XBP1 deficiency has been linked to enhanced macroautophagy and protects against Huntington's disease. ⁷⁴ Enhanced XBP1 activity can promote protein folding, reduce protein aggregation, and confer protection in other Huntington's disease models. ⁷⁵ It is possible that XBP1 plays both roles simultaneously in SCA6^84Q/+ mice to mediate resilience to ER stress. On the one hand, decreased levels of cerebellar XBP1 may promote degradation of polyQ‐Cav2.1. On the other hand, enhanced function of XBP1 may simultaneously promote proper polyQ‐Cav2.1 folding via upregulation of HSP90 and other chaperones. Further studies are needed to more fully understand the role of individual UPR pathways in SCA6.

Protein misfolding and changes in the UPR have been identified in a number of age‐related neurodegenerative disorders, including in Parkinson's, Alzheimer's, Huntington's and Creutzfeldt–Jakob disease. ²⁷ Postmortem brains from patients with these diseases has identified UPR activation, and induction of ER chaperones and ER stress markers. ⁷⁶ Calcium dyshomeostasis has also separately been implicated in these disorders. ⁷⁷ , ⁷⁸ , ⁷⁹ The identification of a mechanism of disease in SCA6 that connects abnormal proteostasis and calcium‐dependent membrane excitability that also explains delayed onset of disease likely applies to a variety of other age‐dependent neurodegenerative disorders. Examination of autopsy material from individuals with SCA6 for ER stress and UPR‐related markers could be a focus for future studies. Exploring the relationship among aberrant proteostasis, ER stress, and calcium dyshomeostasis further may be fruitful in late onset neurodegenerative disorders as enhancers of specific components of the UPR that confer protection from calcium dyshomeostasis could lead to novel treatment for SCA6 and other age‐related neurodegenerative disorders.

Author contributions

HH, TLC, A‐HP, GK, and VGS contributed to the conception of the study; HH, TLC, MF, MD, DMJ, SI, PK, and VGS contributed to the acquisition and analysis of data; HH, DMJ, PK and VGS contributed to drafting the text or preparing the figures.

Potential Conflicts of Interest

Nothing to report.

Supporting information

Figure S1. Purkinje neuron firing frequency and changes in ion channel transcripts/protein in SCA6^84Q/+ mice.

Figure S2. Irregular Purkinje neuron spiking in SCA6^84Q/+ mice is due to changes in intrinsic membrane excitability.

Figure S3. Spatial transcriptome analysis of Purkinje neurons.

Figure S4. Knockout of XBP1 in Purkinje neurons has no effect on the motor phenotype, Purkinje neuron spiking regularity and firing frequency but alters other UPR pathways in SCA6^84Q/+ mice.

Figure S5. The increase in CRAC current in 19‐month SCA6^84Q/+ mice is not caused by increased expression of Orai2/Stim1 subunits in cerebella.

Figure S6. 17‐AAG has no effect on Purkinje neuron firing frequency in 6‐month SCA6^84Q/+ mice or 19‐month wild‐type mice.

Table S1. Electrophysiological parameters of Purkinje neurons from 6‐month and 19‐month wild‐type and SCA6^84Q/+ mice.

Table S2. DEGs between wild‐type and SCA6^84Q/+ cerebella at differenet ages using bulk cerebellum RNAseq.

Table S3. DEGs between wild‐type and SCA6^84Q/+ Purkinje neurons at 6 months using MERSCOPE spatial transcriptome analaysis.

Data Availability

The transcriptome dataset obtained from RNAseq analysis of cerebellar RNA isolated from 3, 6, 12, and 19‐month WT and SCA6^84Q/+ mice is available at the NCBI GEO website, accession GEO: GSE264100 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE264100). Code information can be found on GitHub (https://github.com/konopkalab/understanding_resilience_in_sca6).