TDP-43 loss induces cryptic polyadenylation in ALS/FTD

Abstract

Nuclear depletion and cytoplasmic aggregation of the RNA-binding protein TDP-43 are cellular hallmarks of amyotrophic lateral sclerosis (ALS). TDP-43 nuclear loss causes de-repression of cryptic exons, yet cryptic alternative polyadenylation (APA) events have been largely overlooked. In this study, we developed a bioinformatic pipeline to reliably identify alternative last exons, 3' untranslated region (3'UTR) extensions and intronic polyadenylation APA event types, and we identified cryptic APA sites induced by TDP-43 loss in induced pluripotent stem cell (iPSC)-derived neurons. TDP-43 binding sites are enriched at sites of these cryptic events, and TDP-43 can both repress and enhance APA. All categories of cryptic APA were also identified in ALS and frontotemporal dementia (FTD) postmortem brain tissue. RNA sequencing (RNA-seq), thiol(SH)-linked alkylation for the metabolic sequencing of RNA (SLAM-seq) and ribosome profiling (Ribo-seq) revealed that distinct cryptic APA categories have different downstream effects on transcript levels and that cryptic 3'UTR extensions can increase RNA stability, leading to increased translation. In summary, we demonstrate that TDP-43 nuclear depletion induces cryptic APA, expanding the palette of known consequences of TDP-43.

Fig. 1. TDP-43 depletion induces cryptic APA in a compendium of in vitro TDP-43 datasets.

a, Computational pipeline inferring differential last exon (LE) usage from bulk RNA-seq. Putative novel last exons (orange) are identified by comparing StringTie assembled transcripts (condition mean TPM > 1) to reference transcripts (purple). Putative last exons with a PAS <100 nt from PolyASite PAS or containing a conserved poly(A) signal hexamer (final 100 nt) are quantified with annotated last exons using Salmon and assessed for differential usage using DEXSeq. b, APA upon TDP-43 knockdown (TDP43KD). Points: PAS with adjusted P < 0.05 in ≥1 dataset (median values when >1 dataset). Cryptic PAS (orange): adjusted P < 0.05, mean control (Ctrl) usage <10% and TDP43KD–CTRL usage >10%. c, Cryptic APA RNA-seq coverage traces in control (gray) and TDP-43 knockdown (gold) i3Neuron. ALE: ARHGAP32. IPA: ANKRD27. 3’Ext: TLX1. Dashed lines: landmarks assessed for TDP-43 binding (d,e). All events are visualized in sense orientation. d, TDP-43 binding around ALE boundaries. Exon start: first nucleotide of the last exon. Top, mean SH-SY5Y TDP-43 iCLIP peak coverage (n = 2) ±1 s.e.m. (shaded interval) of positions relative to landmarks in cryptic (orange, n = 92) versus background (black, n = 929) ALEs. Two-sided Fisher’s exact test in the plotting window (exon start P = 0.005, PAS P = 0.019). Bottom, mean YG-containing hexamer coverage (Supplementary Fig. 3a) ±1 s.e.m. (shaded interval). e, TDP-43 binding maps around 3’Ext alternative PAS. Top, as in d (top) for cryptic (orange, n = 86) and background (black, n = 798) 3’Exts. Proximal P = 0.031, distal P = 0.003. Bottom, as in d (bottom) for e (top). f, ELK1 fluorescent reporter. CDS: mGreenLantern coding sequence. ELK1 3’UTR, proximal 3’UTR and the first 800 bp of cryptic 3’Ext. SV40, SV40 PAS. g, Nanopore sequencing traces of the reporter in TDP-43 knockdown SK-N-BE(2) cells. h, Reporter distal PAS usage upon increasing TDP-43 knockdown (low: 30, medium: 60, high: 1,000 ng ml⁻¹ doxycycline). Bars denote mean PAS usage fold change versus controls. n = 3 per variant. −96%: four variants; −20% and −24%: two variants; remaining: one variant.

Fig. 2. Cryptic APAs are detected in postmortem ALS/FTD RNA-seq datasets.

a, Heatmap of cryptic last exon usage in postmortem FACS-seq data. Cells are colored according to the magnitude of sample-wise difference in usage between TDP-43-depleted (TDPnegative) and TDP-43-positive (TDPpositive) cells. Rows represent individual cryptic last exons from in vitro that passed enrichment criteria (median sample-wise difference in usage (TDPnegative − TDPpositive) > 5%) and are arranged in descending order of the difference in usage within each event type. Columns represent individual patients within the cohort. b, RT−qPCR analysis after 3’RACE for the indicated 3’UTRs in frontal cortex samples of control patients (n = 4) and FTD (FTD-TDP, n = 4) cases with TDP-43 pathology. The RNA expression levels were normalized against GAPDH mRNA and expressed as relative fold change with respect to one control sample set to a value of 1. PHF2 and SIX3 genes (shown in Supplementary Fig. 3) were excluded owing to unspecific amplification of the cryptic isoforms in tissues. Data are represented as box plots (lower, middle and upper quartiles), and error bars span from the minimum to the maximum value. Two-sided Studentʼs unpaired t-test (NS P > 0.05, *P < 0.05). l, long; s, short. STMN2 P = 0.330 (canonical), 0.033 (ALE). SYNJ2 P = 0.847 (canonical), 0.031 (ALE). ARHGAP32 P = 0.500 (canonical), 0.021 (ALE s), 0.035 (ALE l). ELK1 P = 0.056 (canonical), 0.013 (3’Ext). TLX1 P = 0.130 (canonical), 0.041 (3’Ext). All P values are to 3 decimal places (d.p.). c, Selectively expressed cryptic ALEs (orange) and splicing events (purple) in tissues and samples with TDP-43 proteinopathy in the NYGC ALS Consortium dataset. Events are considered detected if at least two junction reads were detected in a sample. d, Detection of spliced reads for the cryptic ALE in PHF2 across samples in the NYGC ALS Consortium dataset. Color indicates whether disease subtype and region is expected (orange) or not expected (green) to have TDP-43 pathology and cryptic spliced read expression. e, As in d but for cryptic ALE in SYNJ2. NS, not significant. Source data

Fig. 3. Cryptic 3’ UTR extensions in transcription factor RNAs lead to increased RNA and protein levels by increased RNA stability and cytoplasmic RNA levels.

a, RNA-seq differential expression volcano plot (TDP-43 knockdown versus control i3Neurons). Cryptic 3’Ext (orange), ALE (blue) and IPA (green) containing genes with increased translation (Fig. 3b) are colored and labeled. y axis 50, −log₁₀-adjusted P (Padj) ≥ 50. b, Ribo-seq differential expression volcano plot (TDP43KD versus CTRL i3Neurons). Colors: cryptic 3’Ext (orange), ALE (blue) or IPA (green) containing genes. y axis 10, −log₁₀-adjusted P ≥ 10. c, ELK1 protein levels in Halo-TDP-43 i3Neurons. Top, ELK1 western blot showing increased ELK1 protein expression upon TDP-43 knockdown (n = 4 independent differentiations). Bottom, tubulin-normalized ELK1 band intensities (c, top) in control and TDP43KD Halo-TDP-43 i3Neurons. d, ELK1 transcription factor activity. Top, ELK1 cryptic 3’Ext RNA-seq coverage traces in control (black) and TDP-43 knockout (KO) (gold) HeLa cells. Bottom, GSEA enrichment plot for ChIP–seq-defined ELK1 target genes in TDP-43 knockout HeLa cells. Green line denotes GSEA enrichment statistic; red lines denote maximum value in upregulated (left) and downregulated (right) genes; black lines denote ELK1 target genes (n = 353). NES is relative to mean score of identically sized, randomly sampled gene sets. e, Decay curve for RNA produced before 4SU labeling (old) in control (gray, 4 h n = 1, others n = 2) and knockdown (orange, all n = 2) i3Neurons. Curves denote fitted estimate of old RNA levels. Points denote old RNA abundance estimates. Error bars denote upper and lower 95% credible interval. Inset text shows the gene-level GrandR-estimated half-lives. f, Representative images for FISH probes targeting the annotated (ELK1 total, green) 3’UTR and cryptic 3’UTR-specific (ELK1 cryptic, magenta) ELK1 sequences in control (top row) and TDP-43 knockdown (bottom row) i3Neurons. Scale bars, 10 µm. g, Extranuclear FISH signals for the ELK1 total and cryptic probes. Points denote foci counts (n = 10 images). Blue bars denote mean count. Two-sided, one-sample t-test after within-replicate control normalization (n = 3, *P < 0.05, total P = 0.009, cryptic P = 0.012 (3 d.p)). h, ELK1 canonical and cryptic (3’Ext) isoform 3’RACE and RT−qPCR of the cytoplasmic fraction of TDP-43-depleted SH-SY5Y cells. Bars denote mean fold change versus control cells ± s.d. (n = 3 biological replicates). Two-sided Studentʼs unpaired t-test (**P = 0.009, ***P = 7.535 × 10⁻⁸). Source data

Extended Data Fig. 1. 3’ RACE validation of cryptic APAs in i3Neurons.

RT-qPCR analysis after 3’RACE for the indicated 3’UTRs upon TDP-43 depletion (“TDP-43 KD”) in i3Neurons. The RNA expression levels were normalized against GAPDH mRNA and expressed as relative fold change with respect to the control condition (“Control”) set to a value of 1. Data are represented as the mean of the fold change ± standard deviation. n=4 biological replicates. Statistical analyses were performed using two-sided, Student unpaired t-test (n.s. p>0.05, * p<0.05, ** p<0.01, **** p<0.0001). ALE: alternative last exon (s short, l long), 3’Ext: 3’UTR extension, IPA: intronic polyadenylation. Exact p-values are reported in the form (canonical, ALE/IPA/3’Ext/ALE short, ALE long). STMN2 (p = 2.804×10⁻⁸, 3.280×10⁻⁵). SYNJ2 (p = 7.577×10⁻⁵, 6.916×10⁻⁶). ARGHAP32 (p = 1.536×10⁻⁴, 6.094×10⁻¹⁰, 2.018×10⁻⁴). PHF2 (p = 6.727×10⁻⁴, 1.680×10⁻⁴, 9.785×10⁻⁵). ELK1 (p = 8.711×10⁻¹⁰, 2.295×10⁻⁶). TLX1 (p = 1.271×10⁻⁴, 6.779×10⁻⁶). SIX3 (p = 3.248×10⁻³, 3.711×10⁻³). SIN3B (p = 1.735×10⁻¹, 3.490×10⁻⁴). CNPY3 (p = 8.970×10⁻⁴, 5.032×10⁻⁸). Source data

Extended Data Fig. 2. ELK1 3’-UTR APA reporter library.

A). ELK1 Cryptic PAS usage in control (‘CTRL’) conditions in a series of reporters with varying changes in UG content (x-axis, %). Original reporter (0); reporters with increasing amounts of UG deletion (20, 24, 32, 48, 56, 60, 96); the reporter where UG content is increased (92). B). Cryptic PAS usage in control and severe TDP-43 knockdown conditions for the original reporter (0), the reporter with increased UG content (92) and the reporter with the most UG deletion (96).

Extended Data Fig. 3. Consistency of enriched/selective ALEs between FACS-seq and NYGC datasets.

A). Overlap between ALEs passing enrichment threshold in the ‘Liu’ FACS-seq data (Fig. 2a) and splice junctions of ALEs passing selective detection thresholds in the New York Genome Centre (NYGC) ALS Consortium dataset (Fig. 2b). Cryptic ALEs in each intersection group are labelled directly underneath the event count. B). Heatmap of PAS usage in post-mortem FACS-seq data for NYGC-specific ALEs. Cells are labelled with and coloured in proportion to the magnitude of the sample-wise difference in PAS usage between TDP-43 depleted (TDPnegative) and TDP-43 positive (TDPpositive) nuclei. Rows are arranged in descending order of the median sample-wise difference in usage (TDPnegative - TDPpositive). Columns represent individual patients within the cohort. C). Detection statistics for FACS-seq specific ALEs in the NYGC ALS Consortium. ALEs are sorted in descending order of the detection enrichment ratio and bars are coloured according to expected presence (gold, ‘True’) or absence (grey, ‘False’) of TDP-43 proteinopathy. ALEs are considered detected if at least 2 junction reads were present in a sample.

Extended Data Fig. 4. Subcellular fractionation of SH-SY5Y upon TDP-43 depletion.

A). Western blots to evaluate the decrease of TDP-43 protein upon its depletion in SH-SY5Y cell line; Tubulin was used as loading control. For each experimental condition, two technical replicates were loaded on the gel. n=3 biological replicates. B). Bar-plots showing the percentage in the nuclear and cytoplasmic fractions in SH-SY5Y cell line for selected targets in control condition (“Ctrl”) or upon TDP-43 depletion (“TDP-43 KD”) detected through qRT-PCR analysis. GAPDH and pre-GAPDH were used as cytoplasmic and nuclear controls, respectively, for cell fractionation. STMN2 Cryptic, a well-reported cryptic exon, shows predominant cytoplasmic localization. The relative RNA distribution in the bars is represented as mean ± standard deviation. n=3 biological replicates. Statistical analyses were performed using Student unpaired t-test (n.s. p>0.05, * p<0.05, ** p<0.01, **** p<0.0001). GAPDH p-value (3 d.p.): 0.865, pre-GAPDH p: 0.936, STMN2 Cryptic p: 0.516. C). RT-qPCR analysis after 3’RACE on the nuclear fraction of SH-SY5Y cell line upon TDP-43 depletion (“TDP-43 KD”). The levels of ELK1 canonical (“Canonical”) and cryptic (“3’Ext”) isoforms are expressed as relative fold change with respect to the control condition (“Control”) set to a value of 1. Data are represented as the mean of the fold change ± standard deviation. n=3 biological replicates. Statistical analyses were performed using Student unpaired t-test (* p<0.05). 3’Ext: 3’UTR extension. ELK1 Canonical p-value (3 d.p.): 0.023, ELK1 3’Ext p: 0.041. D). Bar-plots showing the percentage in the nuclear and cytoplasmic fractions in SH-SY5Y cell line upon TDP-43 depletion for ELK1 canonical (“Canonical”) and cryptic (“3’Ext”) isoforms, as detected through qRT-PCR analysis. The relative RNA distribution in the bars is represented as mean ± standard deviation. n=3 biological replicates. 3’Ext: 3’UTR extension. Source data