Long pre-mRNA depletion and RNA missplicing contribute to neuronal vulnerability from loss of TDP-43

Abstract

We used cross-linking and immunoprecipitation coupled with high-throughput sequencing to identify binding sites in 6,304 genes as the brain RNA targets for TDP-43, an RNA binding protein that, when mutated, causes amyotrophic lateral sclerosis. Massively parallel sequencing and splicing-sensitive junction arrays revealed that levels of 601 mRNAs were changed (including Fus (Tls), progranulin and other transcripts encoding neurodegenerative disease-associated proteins) and 965 altered splicing events were detected (including in sortilin, the receptor for progranulin) following depletion of TDP-43 from mouse adult brain with antisense oligonucleotides. RNAs whose levels were most depleted by reduction in TDP-43 were derived from genes with very long introns and that encode proteins involved in synaptic activity. Lastly, we found that TDP-43 autoregulates its synthesis, in part by directly binding and enhancing splicing of an intron in the 3' untranslated region of its own transcript, thereby triggering nonsense-mediated RNA degradation.

Figure 1. TDP-43 binds distal introns of pre-mRNA transcripts through UG-rich sites in vivo

(a) Autoradiograph of TDP-43-RNA complexes trimmed by different concentrations of micrococcal nuclease (MNase) (left panel). Complexes within red box were used for library preparation and sequencing. Immunoblot showing TDP-43 in ~46kD and higher molecular weight complexes dependent on UV-crosslinking (UV) (right panel). (b) Example of a TDP-43 binding site (CLIP-cluster) on Semaphorin 3F defined by overlapping reads from 2 independent experiments surpassing a gene-specific threshold. (c) UCSC browser screenshot of neurexin 3 intron 8 (mm8; chr12:89842000-89847000), displaying four examples of TDP-43 binding modes. The right-most CLIP cluster represents a ‘canonical’ binding site coinciding GU-rich sequence motifs while the left-most cluster lacks GU-rich sequences and a region containing multiple GU-repeats shows no evidence of TDP-43 binding. The second CLIP cluster (middle purple-outlined box) with weak binding was found only when relaxing cluster-finding algorithm parameters. (d) Flow-chart illustrating the number of reads analyzed from both CLIP-seq experiments. (e) Histogram of Z-scores indicating the enrichment of GU-rich hexamers in CLIP-seq clusters compared to equally sized clusters, randomly distributed in the same pre-mRNAs. Sequences and Z-scores of the top 8 hexamers are indicated. Pie-charts enumerate clusters containing increasing counts of (GU)₂ compared to randomly distributed clusters (p≈0, χ²=21,662). (f) Pre-mRNAs were divided into annotated regions (upper panel). Distribution of TDP-43 (left panel), or previously published Argonaute binding sites as a control (right panel), showed preferential binding of TDP-43 in distal introns.

Figure 2. In vivo depletion of TDP-43 in mouse brain with antisense oligonucleotides (ASOs)

(a) Strategy for depletion of TDP-43 in mouse striatum. TDP-43-specific or control ASOs were injected into the striatum of adult mice. TDP-43 mRNA was degraded via endogenous RNase H digestion which specifically recognizes ASO-pre-mRNA (DNA/RNA) hybrids. Mice were sacrificed after 2 weeks and striata were dissected for RNA and protein extraction. (b) Semi-quantitative immunoblot demonstrating significant depletion of TDP-43 protein to 20% of controls in TDP-43 ASO-treated compared to saline and control ASO-treated animals (three upper panels). Quantitative RT-PCR showed similar TDP-43 depletion at the mRNA levels. (c) Flow-chart illustrating the number of reads sequenced and aligned from the RNA-seq experiments. (d) Differentially regulated genes identified by RNA-seq analysis. Scatter-plot revealed the presence of 362 and 239 genes (diamonds) that were significantly up- (red) or down-(green) regulated upon TDP-43 depletion. RNA-seq analysis confirmed that TDP-43 levels were reduced to 20% of control animals. (e) Normalized expression (based on RPKM values from RNA-seq) of four non-coding RNAs that are TDP-43 targets and are downregulated upon TDP-43 depletion. (f) Quantitative RT-PCR validation of non-coding RNA Meg3/Gtl2.

Figure 3. Binding of TDP-43 on long transcripts enriched in brain sustains their normal mRNA levels

(a) Correlation between RNA-seq and CLIP-seq data. Genes were ranked upon their degree of regulation after TDP-43 depletion (x-axis) and the mean number of intronic CLIP clusters found in the next 100 genes from the ranked list were plotted (y-axis; green line). Similarly, the mean total intron length for the next 100 genes was plotted (y-axis; red line). The cluster count for each up-regulated gene (red dots) and each down-regulated gene (green dots) was plotted using the same ordering (inset). (b) Quantitative RT-PCR for selected down-regulated genes with long introns (except for Chat) revealed significant reduction of all transcripts when compared to controls (p < 8×10⁻³). Standard deviation was calculated within each group for 3–5 biological replicates. (c) Cumulative distribution plot comparing exon length (left panel) or intron length (middle panel) across mouse brain tissue enriched genes (388 genes) and non-brain tissue enriched genes (15,153 genes). Genes enriched in brain have significantly longer median intron length compared to genes not enriched in brain (solid red line and black lines, p-value < 6.2×10⁻⁶ by two-sample Kolmogorov Smirnov goodness-of-fit hypothesis test) while a random subset of 388 genes shows no difference in intron length (dashed lines). Similar analysis across human brain tissue enriched genes (387 genes) and non-brain tissue enriched genes (17,985 genes) also showed significantly longer introns in brain enriched genes (solid red and black lines, p < 5.3×10⁻⁶) while a random subset of 387 genes shows no difference in intron length (dashed lines) (right panel).

Figure 4. TDP-43-mediates alternative splicing regulation of its RNA targets

(a) Schematic representation of different exon classes defined by EST and mRNA libraries, RNA-seq or splicing-sensitive microarray data as indicated (left panel). Bar plot displaying the percentage of exons that contain TDP-43 clusters within 2kb upstream and downstream of the exon-intron junctions (right panel). (b) Example of alternative splicing change on exon 18 of sortilin 1 analyzed using RNA-seq reads mapping to the exon body (left panel). Arrows depict increased density of reads in the TDP-43 knockdown samples compared to controls. 76% of the spliced-junction reads in the TDP-43 knockdown samples supported inclusion versus only 19% in the control oligo treated samples (right panel). (c) Comparison of mouse cassette exons detected by splicing-sensitive microarrays to conserved exons in human orthologous genes. 85% and 57% of human exons corresponding to the excluded and included mouse exons, upon TDP-43 depletion, respectively (left and middle pie charts), contained EST and mRNA evidence for alternative splicing. As a control, percentage of human exons orthologous to all mouse exons represented on the splicing-sensitive array, and that have alternative splicing evidence are shown at the right pie chart. (d) Semi-quantitative RT-PCR analyses of selected targets confirmed alternative splicing changes in TDP-43 knockdown samples compared to controls. Representative acrylamide gel pictures of RT-PCR products from control or knockdown adult brain samples (right panel). Quantification of splicing changes from three biological replicates per group (left panels); error bars represent standard deviation.

Figure 5. Autoregulation of TDP-43 through binding on the 3′UTR of its own transcript

(a) CLIP-seq reads and clusters on TDP-43 transcript showing binding mainly within an alternatively spliced part of the 3′ untranslated region (3′UTR), lacking long uninterrupted UG-repeats. (b) Quantitative RT-PCR showing ~50% reduction of endogenous TDP-43 mRNA in transgenic mice overexpressing human myc-TDP-43 not containing introns and 3′UTR. (c) Immunoblots confirming the reduction of endogenous TDP-43 protein (upper panel) to 50% of control levels (by densitometry, lower panel) in response to human myc-TDP-43 overexpression (d) Immunoblot showing reduction of endogenous TDP-43 protein in HeLa cells upon tetracycline induction of a transgene encoding GFP-myc-TDP-43-HA (annotated GFP-TDP-43). The ~30kD product accumulating after 48h (red arrow) was immunoreactive with four TDP-43-specific antibodies. (e) Quantitative RT-PCR using primers spanning the junctions of TDP-43 isoform 3 (upper panel), showed ~100-fold increase in response to tetracycline induction of GFP-TDP-43. TDP-43 isoform 3 was present at very low levels before tetracycline induction. (f) Luciferase assays showing significant reduction of Relative Fluorescence Units (RFUs) in cells expressing renilla luciferase under the control of long TDP-43 3′UTR when co-transfected with myc-TDP-43-HA. Cells treated with siRNA against UPF1 showed a significant increase of RFUs when expressing Luciferase with the long TDP-43 3′UTR, but not in controls. (g) Quantitative RT-PCR scoring levels of endogenous TDP-43 isoform 3 in HeLa cells transiently transfected with myc-TDP-43-HA, UPF1 siRNA or both simultaneously. TDP-43 isoform 3 was increased in response to elevated TDP-43 protein levels, blocking of NMD (by UPF1 siRNA) and there was a synergistic effect in the combined condition.

Figure 6. TDP-43 regulates expression of FUS/TLS and Progranulin

(a) CLIP-seq reads and clusters on Fus/Tls transcript showing TDP-43 binding within introns 6 and 7 (also annotated as an alternative 3′UTR) and the canonical 3′UTR. RNA-seq reads from control or TDP-43 knockdown (KD) samples (equal scales) show a slight reduction in Fus/Tls mRNA in the TDP-43 knockdown group and expression values from RNA-seq confirmed that Fus/Tls mRNA was reduced to 70% of control upon TDP-43 reduction. (b) Quantitative RT-PCR confirmed Fus/Tls mRNA downregulation upon TDP-43 depletion. Standard deviation was calculated within each group for 3 biological replicas. (c) Semi-quantitative immunoblot (upper panel) demonstrated a slight but consistent reduction of FUS/TLS protein in ASO-treated mice to ~70% of control levels as quantified by densitometric analysis (lower panel). (d) CLIP-seq reads and clusters on progranulin transcript (Grn) showing a sharp binding within the 3′UTR. RNA-seq reads from control or TDP-43 knockdown (KD) samples (equal scales) show a significant increase in progranulin mRNA in the TDP-43 knockdown group compared to controls. (E) Quantitative RT-PCR confirmed the statistically significant increase (p < 3×10⁻⁴) in progranulin mRNA in samples with reduced TDP-43 levels when compared to controls. Standard deviation was calculated within each group for 3 biological replicas.