Aberrant splicing in Huntington's disease accompanies disrupted TDP-43 activity and altered m6A RNA modification

Abstract

Huntington's disease (HD) is caused by a CAG repeat expansion in the HTT gene, leading to altered gene expression. However, the mechanisms leading to disrupted RNA processing in HD remain unclear. Here we identify TDP-43 and the N6-methyladenosine (m6A) writer protein METTL3 to be upstream regulators of exon skipping in multiple HD systems. Disrupted nuclear localization of TDP-43 and cytoplasmic accumulation of phosphorylated TDP-43 occurs in HD mouse and human brains, with TDP-43 also co-localizing with HTT nuclear aggregate-like bodies distinct from mutant HTT inclusions. The binding of TDP-43 onto RNAs encoding HD-associated differentially expressed and aberrantly spliced genes is decreased. Finally, m6A RNA modification is reduced on RNAs abnormally expressed in the striatum of HD R6/2 mouse brain, including at clustered sites adjacent to TDP-43 binding sites. Our evidence supports TDP-43 loss of function coupled with altered m6A modification as a mechanism underlying alternative splicing in HD.

Fig. 1. Aberrant AS in HD mice.

a, Schematics of detection of alternatively spliced exons in HD mouse models with RASL-seq and rMATS. b, Altered exon (AE) events annotated by rMATS: inclusion (IN) or exclusion (EX) detected in 3mos R6/2 cortex (left) and using two published HD RNA-seq datasets from Labadorf et al. and Al-Dalahmah et al. (prefrontal cortex BA9, grade III/IV, 20 HD, 49 controls; anterior cingulate cortex, grade III/IV, six HD, six controls). Top-ranked HOMER enriched motif for each dataset is shown at the bottom. Splicing events were filtered for FDR < 0.05. c, Heatmap showing the detection of the short isoform (excluded exon) and long isoform (included exon) from RASL-seq in cortex from the R6/2, Q175 and Q150 HD mouse models. Gradient scale represents z-scores of normalized gene counts. NT, non-transgenic control for the R6/2; WT, wild type; HET, heterozygous (Q7/Q175, Q7/Q150); HOMO, homozygous (Q175/Q175, Q150/Q150). d, Left: single-nuclei sequencing validation; cell type classification of total nuclei sequenced from the striatum of 3mos R6/2. Right: bar graphs show nuclei counts from single-nuclei sequencing and number of significant (P < 0.05) splicing events detected in cell types from DESJ-detection. From top to bottom: Astro, astrocytes; Inhib, inhibitory neurons; microglia; Oligo, oligodendrocytes; OPC, oligodendrocyte progenitor cells; premyelin cells; and vascular cells. e, Pie chart showing the detection of significant excluded exons in PacBio Iso-seq long-read sequencing. UE, upstream exon; DE, downstream exon; IN, included exon; EX, excluded exon. f, Canonical binding motif for TDP-43 (UG rich) and METTL3 (DRACH).

Fig. 2. Unannotated splicing events in HD drive gene expression changes.

a, Bar plot showing rMATS splicing changes that are annotated (known) versus not annotated (novel). Bar plots showing MAJIQ and LeafCutter output for the detection of novel unannotated splicing events in the cortex and striatum from 3mos R6/2 and NT. b, Venn diagram showing MAJIQ junction overlap between TDP-43 KD changes from Šušnjar et al. versus R6/2 cortex and striatum samples. c, Heatmap showing exclusion and inclusion of novel unannotated splicing events in 3mos HD R6/2 versus NT (cortex and striatum). Gradient scale represents z-scores of normalized gene counts. d,e, Integrated Genomics Viewer sashimi plots of increased CE splicing in R6/2 (Cntn6; d) or loss of novel unannotated exon (UE) in NT (Ssbp2; e) in R6/2 and corresponding gene expression changes of the genes that contain the splicing event with statistical significance determined by unpaired two-tailed t-test (Cntn6: P < 0.0001, t = 7.111, degrees of freedom (d.f.) = 18, F = 1.612, 95% confidence interval (CI): 10.13–18.64; Ssbp2: P < 0.0001, t = 13.35, d.f. = 18, F = 6.684, 95% CI: −63.21 to −46.02). Data are presented as mean values ± s.e.m. n = 10 biological replicates per genotype (five males, five females). iCLIP-seq, individual-nucleotide resolution UV cross-linking and immunoprecipitation followed by high-throughput sequencing. f, Heatmap showing exclusion and inclusion of novel unannotated splicing events detected by MAJIQ in HD patient cortex compared to non-HD control from Al-Dalahmah et al. and corresponding gene expression changes of the genes that contain the splicing event. Gradient scale represents z-scores of normalized gene counts. CTRL, control.

Fig. 3. TDP-43 knockdown corresponds to R6/2 transcriptional changes.

a, Reproducible R6/2 HD and NT IDR TDP-43 peaks were centered and plotted on all TDP-43 binding sites, compared to Down DEGs, Up DEGs and cassette exons. b, TDP-43 mRNA level by qPCR normalized to cyclophilin as a percentage of PBS control. CTX, cortex; SC, spinal cord; STR, striatum. n = 5. Black bar represents the mean. c, Heatmap showing clustering of 3mos HD R6/2, NT, TDP-43 ASO treated and control PBS treated on TDP-43 KD-dependent DGE changes. Gradient scale represents z-scores of normalized gene counts. d, Schematic of iPSC differentiation into MSNs with TDP-43 KD by siRNA. D, day; NPC, neural progenitor cell; NSC, neural stem cell. e, Left: western blot for TDP-43 protein levels after treatment of MSNs with TDP-43 siRNA. n = 3 differentiation replicates per condition. Right: bar graph plots TDP-43 intensity normalized to Revert total protein stain. Data are presented as mean values ± s.e.m. Statistical significance was determined by two-way ANOVA with Sidak’s multiple comparisons test (18Q: P < 0.0001, 95% CI: 6.230–10.20; 50Q: P < 0.0001, 95% CI: 6.084–10.06). IB, immunoblot; M, marker. f, Venn diagram showing the overlap of DEGs between HTT-18Q MSNs scramble control versus TDP-43 siRNA and HTT-18Q MSNs versus mHTT-50Q. g, Example of key gene expression changes anticipated from TDP-43 KD. n = 3 differentiation replicates per condition. Statistical significance was determined by unpaired two-tailed t-test. Data are presented as mean values ± s.e.m. (TDP-43 KD versus Ctrl STMN2: P < 0.0001, t = 24.41, d.f. = 4, F = 2.639, 95% CI: −768.6 to −611.6; 18Q versus 50Q STMN2: P = 0.0001, t = 14.49, d.f. = 4, F = 2.344, 95% CI: −496.4 to −336.7; TDP-43 KD versus Ctrl UNC13B: P = 0.0076, t = 4.988, d.f. = 4, F = 2.598, 95% CI: −64.21 to −18.29; 18Q versus 50Q UNC13B: P = 0.0132, t = 4.242, d.f. = 4, F = 2.249, 95% CI: −36.78 to −7.682; TDP-43 KD versus Ctrl CAMK2B: P = 0.0006, t = 10.04, d.f. = 4, F = 5.482, 95% CI: −47.26 to −26.79; 18Q versus 50Q CAMK2B: P = 0.0287, t = 3.345, d.f. = 4, F = 4.953, 95% CI: −29.00 to −2.693). Ctrl, control.

Fig. 4. TDP-43 protein mislocalizes in HD patient brain.

a, Representative IF staining images of SFG from patients with HD compared to non-HD control individuals showing decreased TDP-43 (yellow) signal intensity. b, Left: quantification of decreased nuclear TDP-43 signal intensity; five representative images were taken at ×40 from five HD and two control individuals. A CellProfiler pipeline was created to identify larger nuclei (enriched for neurons) by DAPI staining. The average of TPD-43 nuclear signal was obtained by measuring the intensity signal within a mask defined by DAPI. Each cell’s mean nuclear TDP-43 intensity is plotted. One-way ANOVA was performed with multiple comparisons and resulted in significant changes between all HD versus control comparisons (data not shown). Numbers on top of each group indicate the number of cells plotted. Right: dot plot showing grouped data by genotype. Statistical significance was derived from unpaired two-tailed t-test between control versus HD (P < 0.0001, t = 31.41, d.f. = 1317, F = 10.33, 95% CI: −0.03850 to −0.03397). Red bar represents the median. c, Representative IF staining images of the motor cortex from a patient with ALS (positive control) compared to patients with HD, using antibodies against total TDP-43 (yellow), phosphorylated TDP-43 (purple) and nuclear stain DAPI. White arrowhead indicates pTDP-43 cytoplasmic aggregation; red arrowhead (HD1) indicates cytoplasmic aggregate verified by orthogonal view. d, IF images showing pTDP-43 AL bodies (yellow) within MAP2-positive neurons (white). e, Bar graph showing counts of DAPI, MAP2-positive and AL bodies. Data are presented as mean values ± s.e.m. Each bar is derived from five random ×20 images from each patient. C, control; HD, patient with Huntington’s disease. %AL bodies is the number of AL bodies per patient normalized to the total number of Map2-positive neurons. Statistical significance was determined by unpaired two-tailed t-test (number of neurons: P = 0.3655, t = 0.9593, d.f. = 8, F = 5.329, 95% CI: −48.51 to 20.01; AL bodies (%): P = 0.0146, t = 3.103, d.f. = 8, F = 44.539, 95% CI: 0.08172–0.5545). Experiments in a–e were repeated at least three times with similar results represented above.

Fig. 5. m6A is dysregulated in HD.

a, Representative images showing modest decrease in Mettl3 signal in the R6/2 cortex and striatum. b, CellProfiler analysis of Mettl3 nuclear intensity; statistical significance was determined by unpaired two-tailed t-test (cortex: P = 0.0184, t = 2.693, d.f. = 13, F = 1.572, 95% CI: −0.1158 to −0.01270; striatum: P = 0.0093, t = 3.052, d.f. = 13, F = 1.044, 95% CI: −0.0774 to −0.0324). n = 8 (four males, four females) per genotype were quantified, each averaging three cortical or striatal regions. Data are presented as mean values ± s.e.m. c, MS for m6A showed no significant changes by unpaired two-tailed t-test (P = 0.0567, t = 1.982, d.f. = 30) in global m6A levels. Data are presented as mean values ± s.e.m. n = 16 per genotype (five males, 11 females). d, Violin plots showing increased m6A sites per gene on genes that are dysregulated in HD R6/2 (DEGs), in particular in downregulated genes compared to upregulated genes. Statistical significance was determined by unpaired two-tailed t-test. DEGs versus non-DEGs: P < 0.0001, t = 15.03, d.f. = 3818, F = 3.099, 95% CI: −9.075 to −6.981; Up versus Down: P < 0.0001, t = 11.52, d.f. = 3384, F = 2.870, 95% CI: 5.688–8.021. e, m6A site deposition (localized levels) on 266 genes dysregulated in R6/2 striatum. f, Histogram plot showing TDP-43 eCLIP sites relative to an m6A site (from m6A eCLIP data) (0 position) compared to Hnrnp C and Hnrnp L binding motifs. g, m6A site deposition (localized levels) on 266 genes dysregulated with TDP-43 KD by ASO. h, Representative images showing significant increase in m6A IF intensity in HD patient caudate nucleus (CN) with corresponding METTL3 staining. Dot plot with each dot representing the average m6A intensity per neuronal cell spanning five ×20 confocal images per patient (three controls, six HD). Samples were grouped to either control or HD group with the unpaired Studentʼs t-test for significance (for separated dot plots, see Supplementary Fig. 7g) (P < 0.0001, t = 13.82, d.f. = 956, F = 3.313, 95% CI: 0.09684–0.1289). Bar represents the median. Experiments in a, b and h were repeated at least three times with similar results represented above.