Although conventional transgenic mice can cause severe and early disease onset, accelerating experimentation, such models are vulnerable to expression pattern artifacts [23]. We therefore employed a knock-in mouse model. The mouse model carrying a repeat of approximately 200 CAGs in the endogenous gene (HdhQ200) was initially established in a C57BL/6J background [24, 25]. Since approximately 20% of mice in this background demonstrate nocturnal hyperactivity [26], which could create gene expression noise, we backcrossed the mutant allele into the 129S4 background. To characterize the disease phenotype in this new genetic background, we performed behavioral and histological analyses on S4-HdhQ200 heterozygous mice and littermate controls (hereafter HD and control respectively) at different ages (Fig. S1).

HD mice in the S4 background are mildly affected

To assess behavioral changes during disease progression, we evaluated animals between the ages of 3–18 months using standard paradigms. At each experimental time point, mice performed a session on the accelerating rotarod, followed by walking on a balance beam, 1 h of burrowing and a second balance beam task. Body weight was monitored over the course of the behavioral assessment. Neither males nor females showed a significant difference in body weight between HD and controls (Fig. 1A, B), although female HD mice tended to weigh less. Rotarod performance declined with aging for both control and HD mice, but significant differences were only observed at 15 months (p < 0.05; Fig. 1C). Some HD mice performed very poorly on the balance beam task beginning at 12 months, but the majority performed as well as controls and the group average was similar to controls (Fig. 1D, F). Burrowing is an innate, instinctive, and rewarding behavior in mice, and burrowing paradigms are used as an assessment of overall well-being and the effects of disease on the performance of spontaneous behavior [27]. HD mice burrowed less at 9 and 12 months (p < 0.05). This trend continued at 15 and 18 months, but since several HD mice performed as well as controls, the differences were not significant (Fig. 1E). A larger cohort of mice may have produced more statistically significant differences, but it would not have changed the overall conclusion that S4-HdhQ200 mice showed only a very mild behavioral phenotype, even at an advanced age, similar to when this mutation is in the C57Bl/6 background [24, 25].

Fig. 1

S4-HdhQ200 mice have a mild behavioral phenotype. A, B Body weight is not significantly reduced in male (A), or female (B) HD mice compared to littermate controls. C HD mice remain on the rotarod for less time than control mice at 15 months. D, F HD mice show no significantly different performance in the balance beam task. For each time point, two repeat measures were performed, with a one-hour burrowing period between the first (D) and second (F) test. E HD mice burrow less than littermate controls at 9 months and 12 months. HD n = 12 (green), control n = 7 (black), unpaired t-test without corrections for multiple testing, (*p < 0.05)

In contrast to the mild behavioral phenotype, the histopathological phenotype revealed changes consistent with HD. NIIs are common in polyglutamine diseases, such as HD and several spinocerebellar ataxias, and their extent can be indicative of disease stage. To improve detection of Htt protein aggregates, soluble Htt was partially removed by applying a mild proteinase K treatment, resulting in bright puncta on a background of diffuse staining, only in HD mice (Fig. 2A, B). The proportion of cells with proteinase K-resistant Htt aggregates was measured at 4, 9 and 18 months in eight brain regions (Fig. 2C). By 9 months of age, aggregates were detected in all brain regions, however hippocampus, striatum, and cerebellum had the highest percentage (Fig. 2D, Fig. S2). Immunofluorescent co-staining of Htt with the nuclear stain DAPI revealed that nearly all aggregates in these brain regions were NIIs (Fig. 2E, Fig S2). In contrast, aggregates in the thalamus, brain stem, midbrain, and hypothalamus were much less frequent and presented mostly as cytosolic aggregation foci, appearing like NIIs but not within a nucleus (Fig. 2D, E; Fig. S2). To determine if regions with high NII load expressed Htt the highest, in situ hybridization (ISH) for Htt mRNA was performed. Surprisingly, the striatum and cerebellum were among the lowest expressing regions, whereas the thalamus was the second highest expressing region, indicating something other than expression levels determines NII load (Fig. 2F). We also observed a significant, age-dependent downregulation of Htt mRNA levels in HD mice in all brain regions except the hypothalamus by 9 months (Fig. 2F; Fig. S2).

Fig. 2

S4-HdhQ200 mice show HD-typical neuropathology by 9 months of age. A Htt aggregates (2 examples marked with arrowheads) present in 9 months (mos) old HD striatum (A), were absent from 9 mos old control striatum (B). C Overview of brain regions analyzed. D–F The proportion of aggregates per cell (D), the percent of aggregates that are NIIs (E), and the number of RNA molecules detected per cell (F). All charts share the key in (F), using the same color scheme as in (C); statistical analyses are presented in Fig. S2. G–H Representative images of RNA ISH for Ppp1r1b in 9 mos old brain sections (G), and quantification and analysis in (H). Unpaired t-test without corrections for multiple testing (***p < 0.001). The scale bar in (B) represents 25 µm, and the scale bars in (G) represent 100 µm

Previous analyses of similar HD knock-in mouse models revealed downregulation of common HD-associated genes in striatal cell types early in the disease [10, 14], suggesting loss of striatal identity may be one of the earliest features of HD pathology. We therefore performed ISH staining for one such marker, Ppp1r1b (DARPP32), which is abundant throughout the striatum and in Purkinje neurons of the cerebellum. Ppp1r1b was reduced by 38% in the striatum at 9 months, and 51% by 18 months but was unchanged in the cerebellum at all time points (Fig. 2G, H). Staining for Gfap and Iba1 showed no obvious signs of astrogliosis or microgliosis at any disease stage (Fig. S3). Finally, no differences were detected between HD and control mice at 4 months of age, indicating the abnormalities detected at 9 and 18 months were not caused by a developmental defect. In summary, like other HD knock-in mice, S4-HdhQ200 mice have very mild behavioral and neuropathological changes, with abundant NII accumulation in specific brain regions and Ppp1r1b reduction in the striatum at 9 months. Therefore, here we considered 9 months as an ideal early disease stage for assessment of cell type-specific translatome changes.

Capture of cell type-specific translatomes

To study cell type-specific translatomes we employed the RiboTag method [15]. It utilizes a knock-in mouse line in which the endogenous gene encoding large subunit ribosomal protein 22 (Rpl22) has been engineered such that activation by Cre recombinase results in a hemagglutinin (HA) antibody epitope being encoded on the C-terminus of Rpl22, and consequently HA-tagged ribosomes (RiboTag). From these samples, one can immunoprecipitate (IP) HA-tagged ribosomes from cell types of interest, and sequence the attached translating mRNAs, representing the translatome. To this end, we crossed HdhQ7/Q200-Rpl22HAflox/flox mice with homozygous Cre-driver lines to generate HD and control mice expressing RiboTag in targeted cell types. We targeted general populations of either GABAergic (Gad2+) or glutamatergic (vGluT2+) neurons in the cerebellum and the cerebrum, where the cerebrum is the remaining part after removal of the cerebellum and olfactory bulb from the brain. In the cerebrum, we also targeted the subset of GABAergic neurons expressing the neuropeptide parvalbumin (PV) (Fig. 3A). We have previously shown that these Cre lines direct RiboTag expression in the desired cell types [28]. Since the PV Cre line targets the same cerebellar cell types as the Gad2 line, and were thus redundant, the cerebellar PV samples were not processed.

Fig. 3

Translatome analysis with RiboTag reveals cell type-specific responses in HD. A HdhQ7/Q200 mice homozygous for RiboTag were crossed with homozygous Cre-driver lines to facilitate cell type-specific expression of Rpl22-HA in glutamatergic (vGluT2+) and GABAergic (Gad2+) neurons in the cerebellum and cerebrum, and parvalbumin (PV+)-expressing neurons in the cerebrum. Cerebellum and cerebrum (without olfactory bulb) were separated and flash frozen at 9 months. For RiboTag IPs, cell type-specific HA-tagged ribosomes were immunoprecipitated from tissue homogenate using anti-HA antibody bound to magnetic beads. Total RNA was prepared from an aliquot of the same tissue homogenate. B Expression levels of Htt mRNA were slightly reduced in all cell types, significantly in cerebral vGluT2+ and PV+ neurons (* FDR ≤ 0.05; ** FDR ≤ 0.01; *** FDR ≤ 0.001). C HD-typical genes showed downregulation in cerebral cell types. D Differentially expressed genes (DEGs) (FDR ≤ 0.05) for each cell type. E Top five significantly enriched terms (FDR ≤ 0.05) among DEGs reveal cell type-specific responses to mHtt

Upon sequencing RiboTag-captured mRNAs, we found Htt mRNA levels were significantly reduced in cerebral PV+ and vGluT2+ neurons in HD mice, with the same trend in the other cell types (Fig. 3B), consistent with the histological analysis (Fig. 2F). Principal component analysis (PCA) of IP and total RNA samples showed a clear separation of RiboTag IP samples by targeted cell type, in contrast to total RNA samples (Fig. S4A, B), indicating that we successfully obtained cell type-specific translatomes from RiboTag IPs. This was confirmed by analyzing the expression of cell type-specific marker genes in comparison to total RNA samples (Fig. S4C, D). General GABAergic markers Gad2, Slc32a1, Dlx2, and Lhx6 were enriched in cerebral Gad2+ and PV+, but not vGluT2+ neurons (Fig. S4C). PV+ samples showed enrichment for genes important for the development and function of cortical PV+ interneurons (Syt2, Cplx1, Nek7, Ank1, Kcnc2, Gpr176, Pthlh) [29,30,31,32], while marker genes for Gad2+/PV− neurons, including genes with high striatal expression levels such as Adora2a, Drd1, Drd2, Penk, and Ppp1r1b, were depleted in PV+ samples, as were markers of other PV− GABAergic interneuron subtypes (Sst, Vip and Htr3a). Cerebellar Gad2+ samples (Fig. S4D) showed enrichment of markers for Purkinje (Pcp, Grid2, Esrrb, Foxp2), Basket (Kcna2), Golgi (Lgi2, Gdj2, Sorcs3), and Stellate cells (Grik3), while cerebellar vGluT2+ samples showed enrichment for granule cell markers (Gabra6, Cntn2, Zic1, Etv1, Nfia) [33, 34]. As expected, astrocyte and microglia markers [35], were depleted in all IP samples compared to total RNA. Taken together, these results confirm the successful isolation of cell type-specific translatomes with RiboTag and the inclusion of expected neuronal populations in both targeted regions.

Translatome analysis reveals cell type-specific responses to mHtt

Next, we performed differential gene expression analysis to study the cell type-specific responses of targeted neurons to mHtt. For cerebral neurons, we observed downregulation of known HD-associated neuronal genes, such as Drd1, Drd2, Penk, Ppp1r1b, Pde10a, Arpp21, and Pcp4, all commonly reported to be downregulated in both human HD patients and various HD mouse models [36,37,38] (Fig. 3C). Several of these genes show high striatal expression in early vulnerable populations i.e., Gad2+ SPNs, so the observed downregulation of these genes in cerebral Gad2+ neurons further supports a striatal phenotype in our HD mice at 9 months. Comparison of our cell type-specific data with bulk RNAseq of striata from the zQ175DN HD knock-in model [10] showed a highly significant overlap between gene expression changes detected in cerebral Gad2+ samples and bulk striatal tissue at all observed time points (Fig. S5A), further indicating that our S4-HdhQ200 mice have HD-typical phenotypes. There was also a significant overlap of bulk RNAseq data with our cell type-specific results for cerebellar Gad2+ and vGluT2+ neurons (Fig. S5B). Notably, these HD-associated genes were also downregulated in cerebral vGluT2+ or PV+ neurons, which may be explained in at least two ways. First, downregulation of striatal marker genes may occur not only in highly vulnerable striatal projection neurons but in other cell types as well, as shown in a recent mouse study for glutamatergic corticostriatal projection neurons, astroglia, and cholinergic interneurons [14]. Second, the downregulation of these genes may occur in regions beyond the striatum, which would be detected when multiple regions are combined.

Surprisingly, cerebellar vGluT2+ neurons showed the overall highest number of differentially expressed genes (DEGs) with 626 (Fig. 3D). To discover functional associations, we analyzed the overrepresentation of gene ontology (GO) terms in the biological process classes among cell type-specific DEGs, which revealed cell type-specific responses (Fig. 3E). DEGs of cerebral Gad2+ neurons were associated with dopamine transport (FDR = 0.009, rich factor = 0.07) and secretion (FDR = 0.02, rich factor = 0.07), ion transport (FDR = 0.009, rich factor = 0.1), downregulation of neuropeptides vasopressin and proenkephalin (Vip, Penk), and downregulation of several genes encoding neurotransmitter receptor components for dopamine (Drd1), glutamate (Grm3), acetylcholine (Chrna6), cannabinoid (Cnr1), and serotonin (Htr4) (Fig. S6A). Cerebral vGluT2+ neurons were enriched for gene sets associated with cognition, memory (FDR = 0.05, rich factor = 0.03), and locomotory behavior (FDR = 0.04, rich factor = 0.02), while PV+ neurons were associated with myelination (FDR = 0.004, rich factor = 0.13) and neurotransmitter uptake (FDR < 0.001, rich factor = 0.13). Top enriched pathways in cerebellar vGluT2+ neurons were related to exocytosis (FDR = 0.03, rich factor = 0.13) and vesicle fusion (FDR = 0.002, rich factor = 0.55; “regulation of vesicle fusion” FDR = 0.01, rich factor = 0.28) (Fig. 3E), driven by downregulation of genes encoding the central components involved in calcium-dependent synaptic vesicle fusion and exocytosis, such as complexin-encoding genes Cplx2 and Cplx3, Snap25, and Vamp2, as well as differential expression of Ca2+ sensors Syt7, Doc2b and Otof (Fig. S6A). Cerebellar Gad2+ neurons, in contrast, indicated activation of mitotic cell cycle regulation, in particular G1/S transition (FDR = 0.004, rich factor = 0.017), with upregulation of cyclin D1 (Ccnd1), a regulator of G1/S progression, downregulation of antiproliferative p21/cyclin-dependent kinase inhibitor 1a (Cdkn1a) and polo-like kinase 5 (Plk5) (Fig. S6A). Ccnd1 was also among the strongest upregulated genes in cerebellar vGluT2+ neurons (log2 fold change = 1.8, FDR < 0.001). This was validated by ISH for Ccnd1 mRNA in tissue sections of 9 months old HdhQ200 mice (Fig. 4A), showing a 1.7-fold increase in Ccnd1 staining in the granule layer in HD mice (Wilcoxon rank sum test, p = 0.002) (Fig. 4B). Therefore, the cerebellum shows a robust, cell type-specific response to the HD mutation as early as 9 months with a focus on pathways involving neurotransmission, vesicles, or cell cycle reentry.

Fig. 4

Ccnd1 ISH staining is increased in the granular layer of 9-month-old HD mice. A In situ hybridization for Ccnd1 mRNA (red) shows increased signal in the anterior cerebellar lobules III-V in 9-month-old HD mice (bottom) compared to littermate controls (top). B Quantification of mean fluorescence intensity in the granule layer of lobule IV/V (white square). Analysis was performed on two measures obtained from five biological replicated per group. At 9 months, HD samples (n = 10, Shapiro-Wilks: W = 0.87, p = 0.11) show a significant 1.7-fold increase of Ccnd1 staining compared to controls (n = 10, Shapiro-Wilks test: W = 0.83, p = 0.03), Wilcoxon rank sum test, p = 0.002, n = 10, matched control and HD samples are indicated by dot color and connecting line

Cerebellar neurons show cell type-specific responses to mHtt

Given the high number of DEGs detected in cerebellar vGluT2+ neurons and that nearly all of the cerebellar Gad2+ DEGs were also changed in vGluT2+ cerebella, we conducted additional analyses of expression changes. Comparison of DEGs between cerebellar cell types revealed that the shared DEGs changed with the same directionality (Fig. S6B), again suggesting a similar response. We next performed gene set enrichment analysis (GSEA) on both cell types, using six different statistical methods to calculate enrichment using a consensus score (CS) to rank results [21]. GSEA is a useful tool to detect pathways that have a coordinated response even when individual genes of that pathway have small and statistically insignificant changes. Interestingly, the GSEA indicated that, despite sharing the majority of the DEGs with cerebellar vGluT2+ neurons, cerebellar Gad2+ neurons had a remarkably different response (Fig. 5). Genes associated with “translation and co-translational targeting of proteins to membrane” and endoplasmic reticulum (CS = 1) were enriched in both cerebellar cell types (Fig. S7). However, Gad2+ neurons uniquely showed upregulation of protein glycosylation-related gene sets (CS = 1.5) and ATP synthesis (CS = 1), oxidative phosphorylation (CS = 5.5), and Parkinson’s disease pathway (CS = 1), suggesting an increase of mitochondrial function in these neurons. In contrast, vGluT2+ cerebellar neurons made a cell type-specific response in the form of upregulation of cell differentiation-associated gene sets (CS = 1), cell cycle regulation (CS = 1) and chromosome organization (CS = 1), autophagy (CS = 1), metabolic processes, upregulation of PI3K-Akt signaling pathway genes (CS = 12) and apoptosis (CS = 12). This suggests that, despite superficial similarities, cerebellar cell types activate different pathways in response to mHtt. Additionally, the enrichment of genes associated with differentiation, PI3K-Akt signaling, and apoptosis, together with the upregulation of Ccnd1, may indicate that cell cycle regulation is affected in cerebellar vGluT2+ neurons. Interestingly, we found significant overlaps of DEGs identified in both vGluT2+ and Gad2+ cerebellar neurons with DEGs in the cerebellum of a knock-in mouse model of Spinocerebellar ataxia 1 (SCA1) [OMIM:164400], an autosomal dominant disease caused by expansion of a polyglutamine encoding CAG repeat in the Ataxin1 gene (Atxn1) (Fig. S8) [39]. This revealed that genes detected in both vGluT2+ and Gad2+ cerebellar cell types were also among the earliest DEGs in SCA1 (Fig. S8).

Fig. 5

Cerebellar neurons show early and cell type-specific responses. Gene set enrichment analysis shows a largely cell type-specific response in Gad2+ and vGlut2+ cerebellar neurons, shared terms are indicated in red. GO terms were reduced to parent terms based on semantic similarity. The ratio indicates the relative number of enriched genes to gene set size

Polycomb repressor complex 1 (PRC1) protein genes are upregulated in cerebellar vGluT2+ neurons

We next investigated if the differential expression patterns we observed in cerebellar vGluT2+ neurons could be driven by specific transcription factors using the ChIP Enrichment Analysis (ChEA) database [22] to perform overrepresentation analysis against the transcription factor targets determined by Chip-X. This analysis revealed that 259 out of 626 DEGs were among target genes of polycomb repressor complex 2 (PRC2) components SUZ12 or EZH2, PRC1 core component RING2B/RNF2, or PRC-associated factor MTF2 (Fig. 6A). Of these 259 genes, 218 were associated with PRC2 core proteins EZH2 (97 DEGs) and SUZ12 (207 DEGs), and 131 DEGs with PRC1 protein RING2B/RNF2 showing near equal distribution of up- and downregulated DEGs (Fig. 6B). Levels of PRC2 core components and associated factors were not affected, but we found increased expression of canonical PRC1 complex components chromobox proteins Cbx2, Cbx4, Cbx8 and polycomb group ring finger 2 (Pcgf2) in cerebellar vGluT2+ neurons (Fig. 6C). Although PRC2 regulation is impaired in HD [40, 41], little is known on the involvement of PRC1 in HD. However, given the proposed roles of CBX proteins in regulating cell cycle progression across various checkpoints [42], these findings further support the view of aberrant cell cycle regulation in cerebellar vGluT2+ neurons.

Fig. 6

Polycomb repressor complex 1 (PRC1) proteins are upregulated in cerebellar vGluT2+ neurons. A Overrepresentation analysis against ChEA ChIPseq gene sets from mouse embryonic stem cells (ESCs/MESCs) show DEGs detected in vGluT2+ neurons are overrepresented among targets associated with polycomb repressor complex (PRC) proteins (FDR ≤ 0.01). B PRC-associated DEGs show similar distribution between up and down-regulated DEGs. C PRC1 protein genes Cbx2, Cbx4, Cbx8 and Pcgf2 are significantly upregulated in cerebellar vGluT2+ neurons (green) in HD, whereas PRC2 core protein genes (Eed, Ezh1, Ezh2, Suz12) and associated factors (Aebp2, Jarid2, Rbbp4) are not differentially expressed. (* FDR ≤ 0.05; ** FDR ≤ 0.01; *** FDR ≤ 0.001)