Chromatin architecture at susceptible gene loci in cerebellar Purkinje cells characterizes DNA damage-induced neurodegeneration

Abstract

The pathogenesis of inherited genome instability neurodegenerative syndromes remains largely unknown. Here, we report new disease-relevant murine models of genome instability–driven neurodegeneration involving disabled ATM and APTX that develop debilitating ataxia. We show that neurodegeneration and ataxia result from transcriptional interference in the cerebellum via aberrant messenger RNA splicing. Unexpectedly, these splicing defects were restricted to only Purkinje cells, disrupting the expression of critical homeostatic regulators including ITPR1, GRID2, and CA8. Abundant genotoxic R loops were also found at these Purkinje cell gene loci, further exacerbating DNA damage and transcriptional disruption. Using ATAC-seq to profile global chromatin accessibility in the cerebellum, we found a notably unique chromatin conformation specifically in Purkinje chromatin at the affected gene loci, thereby promoting susceptibility to DNA damage. These data reveal the pathogenic basis of DNA damage in the nervous system and suggest chromatin conformation as a feature in directing genome instability–associated neuropathology.

Fig. 1.. Inactivation of DNA repair factors result in progressive ataxia and cerebellar dysfunction.

(A) Mice with dual inactivation of ATM and APTX (Atm^Nes-cre;Aptx^−/−) progressively develop a profound ataxia commencing from 6 to 7 months of age. Single-mutant (Atm^Nes-cre or Aptx^−/−) mice do not develop ataxia. Photo credit: Young Don Kwak, SJCRH. Schematic of in vivo extracellular recordings from PCs (B) and deep cerebellar neurons (C) in awake, head-restrained 12-month-old mice. Example raw traces are shown from control animals (black) and Atm^Nes-cre;Aptx^−/− mice (blue). Quantification of the average firing rate of PCs recorded from control mice and age-matched Atm^Nes-cre;Aptx^−/− mice shows a significant decrease in the double-mutant animals. The interspike interval (ISI) coefficient of variation (CV) of Atm^Nes-cre;Aptx^−/− animals was significantly increased compared with control cerebellum, indicating a highly irregular firing of PCs in double-mutant mice. The predominant firing rate (PFR) was not significantly different between control and mutants. sp/s, spikes/s; ns, not significant, **P < 0.01 and ****P < 0.0001.

Fig. 2.. Disrupted cerebellar homeostasis in the DNA repair–deficient cerebellum results from perturbation of gene expression.

(A) Heatmaps showing the top 1000 expressed genes from RNA-seq analysis from affected 10-month-old Atm^Nes-cre;Aptx^−/− cerebella (n = 6) show that a similar group of genes are decreased in expression. While there is a clear difference between gene expression in double-mutant cerebella compared to WT (n = 4), single Aptx^−/− (n = 2), or Atm^−/− (n = 2) mutants show a trend toward gene expression reflective of the double mutant, as indicated by asterisks. (B) A common set of 58 genes are decreased in double-mutant cerebella, while 25 genes are increased in expression. dKO, double knockout. (C) A volcano plot shows the change in gene expression compared to significance of expression change with representative gene names indicated. (D) An example of gene set enrichment analysis (GSEA) shows a significant decrease in genes required for synaptic function in the double-mutant cerebella. (E) Western blot analysis shows that ITPR1 levels decrease with age, and most ITPR1 is absent in the double-mutant cerebellum by 10 months of age. In contrast to the cerebellum, ITPR1 levels are not affected in the double mutant cortex. (F) Immunohistochemistry for ITPR1 shows a loss of PC expression in the 11-month-old double-mutant cerebellum compared to WT tissue. (G) In many PCs, expression of ITPR1 is markedly decreased/absent as indicated by arrows.

Fig. 3.. Aberrant RNA splicing in cerebellar PCs in the ataxia mutants.

(A) Splicing defects occur in the Atm^Nes-cre;Aptx^−/− cerebellum. rMATS analysis (32, 77) shows alterations in all classes of mRNA splicing in the double mutants. Additional identification of intron retention was performed using the splicing deficiency score algorithm (33). (B) An example of intron retention in Itpr1 shows all mutant Atm^Nes-cre;Aptx^−/− have a similar recurring splicing abnormality in the cerebellum, but not in the cortex (red dashed boxes). (C) Gene set enrichment using a PC-specific gene set (35) shows that gene expression in these neurons is affected in the double-mutant cerebellum. The adjacent heatmap shows the expression of the PC genes in the double-mutant cerebellum compared to WT. NES, normalized enrichment scores. (D) Gene expression (mRNA-seq) compared to intron retention (splicing deficiency score) shows that PC genes feature high-intron retention and lower gene expression. Representative Purkinje and granule neuron–specific gene names are listed. The significant threshold of P < 0.05 was applied as a differential cutoff. The graphs indicate a strong correlation between intron retention and gene expression changes in PCs, but not granule neuron gene expression in the double mutants.

Fig. 4.. RNA Pol II promoter occupancy is reduced in the double-mutant cerebellum.

(A) IGV visualization of comparative RNA Pol II binding after Pol II (pSer5) ChIP-seq of PC-specific loci from library size–normalized data using three independent 12-month-old WT and three Atm^Nes-cre;Aptx^−/− cerebella. Peaks found in Purkinje gene promoters show decreased RNA Pol II (pSer5) binding, and many including Cacnb4, Lynx1, and Pnpla3 show a significant decrease when log₂FC is adjusted P < 0.05. Gene promoters and exon/intron structure are indicated under the respective IGV image for each gene; arrows indicate the transcriptional start site and the direction of transcription. Each IGV track represents an independent tissue sample. (B) Plot of average signal of all samples versus log₂FC between WT and knockout (KO) and from promoter/transcription start site (TSS) peaks identified in RNA Pol II (pSer5) ChIP-seq. The red dashed line at 0 log₂FC indicates a mean representative of no change in promoter occupancy, but with overall RNA Pol II (pSer5) binding reduced in the double-mutant samples. The black line indicates the mean log₂FC of all peaks at x̅ = −0.54, while purple dots represent genes from the “Purkinje late” gene set (35), showing that their mean log₂FC is lower than the general mean, x̅ = −0.63, P = 0.13. (C) Western blots of cerebellar tissue extracts show similar levels of RNA Pol II and RNA Pol II (pSer5) in WT, Aptx^−/−, and two times independent 12-month-old Atm^Nes-cre;Aptx^−/− tissue. (D) Quantitative analysis of RNA Pol II binding using ChIP-qPCR shows a decrease in bound RNA Pol II and a further decrease in RNA Pol II (pSer5) at promoter sites for PC-specific gene loci (Cancnb4, ****P < 0.001; and Lynx1, ***P < 0.01), but not at Gapdh.

Fig. 5.. R loops occur in the double-mutant cerebellum.

(A) Immunoblotting of isolated genomic DNA shows a high level of R loops in the double-mutant neurospheres. Specificity of the S9.6 antibody is confirmed as signal is absent after RNase H treatment of samples. DNA loading is shown using SYBL Gold nucleic acid stain. (B) Quantification of R loops in Atm^Nes-cre;Aptx^−/− and Atm^Nes-cre;Parp1^−/− neurospheres show high levels of nuclear R loops compared to controls identified via immunocytochemistry using the S9.6 antibody. (C) R loop formation in the cerebellum is substantially increased in the double-mutant tissue (~8×), which is abolished by RNase H treatment showing specificity of the S9.6 antibody. (D) DRIP-PCR analysis shows that R loops are present at elevated levels in Itpr1 and Grid2 in the double-mutant cerebellum. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. (E) Immunohistochemistry of cerebellar sections from control and mutant, using S9.6 to identify R loops. RNase H treatment confirms the specificity of the S9.6 antibody. IGL, inner granule layer; PCL, PC layer; ML, molecular layer. (F) DRIP-qPCR shows R loop levels in Itpr1, Grid2, and Lynx1 in cerebellar tissues from the double-mutant compared to controls at 6 and 12 months of age. ActinB was used as a positive locus for R loops and AK132462 as a negative control. *P < 0.05, ***P < 0.001, and ****P < 0.0001.

Fig. 6.. R loops result from defective splicing, causing DNA damage and inhibiting gene expression.

(A) Splicing inhibition after isoginkgetin or PladB results in DNA damage as shown by γH2AX immunocytochemistry. Relative γH2AX levels are quantified in the adjacent plot. Unt, untreated control. (B) Inhibition of expression of ITPR1 and CACNB4 after PladB treatment in SH-SY5Y cells is relieved by RNase H1 but not a catalytic mutant (R57A) or RPA binding mutant (D210N). (C) Western blot analysis confirms equal expression of the WT, catalytic, and proliferating cell nuclear antigen–binding mutant versions of RNase H1. Ponceau staining shows equal protein loading. mks, Molecular Weight Marker. (D) Consistent with restoration of gene expression, RNA Pol II occupancy is also restored by RNase H1 (RNH) after PladB; relative Pol II enrichment after ChIP assessed by qPCR is shown for ITPR1, CACNB4, and LYNX1. (E) DNA damage after splicing inhibitor treatment (isoginkgetin) results from R loops, as γH2AX foci are abrogated by expression of WT RNase H1 but not the catalytic R57A mutant. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Fig. 7.. Global ATAC-seq profiling of Purkinje and granule neurons.

(A) Purkinje and granule neuron nuclei from four pooled WT cerebella were isolated via flow cytometry, using NeuN and ITPR1, respectively. (B) The Venn diagrams show the overlap between MACS2 peaks in Purkinje or granule nuclei and whole cerebellum as determined by positional overlap using UpSetR. (C) Scatterplot showing the relationship between gene length and the ATAC coverage in GC and PC, respectively. Gene lists represent the most differentially expressed genes in PCs and GCs, as determined by Rosenberg et al. (35). Gene length in base pairs is shown on the x axis. The sum of the base pair length of all MACS2 peaks annotated to the gene by HOMER is shown on the y axis. (D) Representative IGV image of the distribution of ATAC peak at the Grid2 locus in granule versus Purkinje neurons, showing the broad openness of the locus in PCs. (E) Whole-gene views of bigwig files and associated bed files of MACS2 peaks for granule and PCs. The comparison reveals the unique genomic open configuration across the entire gene body of Grid2, Itpr1, Ca8, and Ttll5 in PCs. (F) γH2AX ChIP-qPCR shows elevated levels of γH2AX at Itpr1 and Grid2 loci in the double-mutant cerebellum (red) at 3 months of age. In contrast, Ube2e2, the expression of which is not altered in the double mutant, does not show significant differences in binding of γH2AX between WT (blue) and the double mutant (red). Experiments were done in triplicate; primer positions and exon structure are shown for each genomic locus. *P < 0.05, **P < 0.01, and ***P < 0.001.

Fig. 8.. PC chromatin architecture dictates genotoxic stress susceptibility.

The cerebellum is particularly affected by genotoxic stress. (A) Despite nervous system–wide damage in the case of inactivation of ATM and APTX, it is the cerebellum (boxed) that shows an overt defect. Within the cerebellum, the PCs are primarily affected by DNA damage. (B) The laminar structure of the cerebellum is shown (an expanded version of the smaller inner box of the cerebellar region) and indicates the three main layers: the inner granule layer, which contains the granule neurons; the Purkinje layer, which contains PCs (shown in red) interspersed with Bergmann glia cells (gray); and the molecular layer. The molecular layer contains PC dendrites that synapse with parallel fibers emanating from granule neurons and also interneurons (blue). (C) Granule cells and PCs have distinct chromatin architecture at certain gene loci. The architecture of these loci can predispose to DNA breaks that affect transcription by affecting RNA Pol II kinetics, causing splicing defects that lead to coincident R loop formation that can exacerbate DNA damage. Collectively, this detrimental effect on transcription leads to the pathogenic reduction of expression of critical homeostatic genes and subsequent PC dysfunction, resulting in ataxia.