TDP-43-dependent mis-splicing of KCNQ2 triggers intrinsic neuronal hyperexcitability in ALS/FTD

Abstract

Motor neuron hyperexcitability is a broadly observed yet poorly understood feature of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). Nuclear depletion and cytoplasmic aggregation of the RNA splicing protein TAR DNA-binding protein 43 (TDP-43) are observed in most ALS and FTD patients. Here we show that TDP-43 dysfunction causes mis-splicing of KCNQ2, which encodes a voltage-gated potassium channel (K_v7.2) that regulates neuronal excitability. Using iPSC-derived neurons and postmortem ALS/FTD brain and spinal cord tissue we find widespread, disease-specific and TDP-43-specific skipping of an exon encoding the KCNQ2 pore domain. The mis-spliced mRNA escapes degradation and is translated into a nonfunctional protein with severely reduced ion conductance that aggregates in the endoplasmic reticulum and causes intrinsic hyperexcitability in ALS neuronal models. This event, which correlates with higher phosphorylated TDP-43 levels and earlier age of disease onset in patients, can be rescued by splice-modulating antisense oligonucleotides that dampen hyperexcitability in induced pluripotent stem cell cortical neurons and spinal motor neurons with TDP-43 depletion. Our work reveals that nuclear TDP-43 maintains the fidelity of KCNQ2 expression and function and provides a mechanistic link between established excitability disruption in ALS/FTD patients and TDP-43 dysfunction.

Fig. 1. TDP-43 depletion in neurons leads to skipping of a constitutively expressed exon in the epilepsy gene, KCNQ2.

a, Differential splicing analysis of stem cell-derived MNs treated with TDP-43 (n = 6) or scrambled (n = 6) siRNAs reported previously. Annotated alternative splicing events are shown in green, de novo events are shown in orange and a de novo KCNQ2 event is shown in red. b, Mis-spliced genes after TDP-43 loss-of-function (LOF) involved in neuronal excitability. Left, integration of TDP-43 mis-spliced genes derived from purified MN, i³Neurons and TDP-43-low neuronal nuclei yields 522 genes; 47 are involved in the regulation of ‘membrane potential’ or ‘synaptic signaling’. Right, two genes (KCNQ2 and CACNA1E) are mis-spliced in all three datasets. c, Sashimi plot depicting HISAT2-mapped sequencing coverage along exons 4–6 of KCNQ2. The gene model indicates the locations of exons and introns. Junction spanning reads mapping to the canonical junctions (exons 4–5 and 5–6) are displayed with solid lines, and de novo junction (exons 4–6) reads are displayed with a dashed line. Read counts and percent spliced index (PSI) scores for each splice junction are denoted. KCNQ2 exon skipping is only detected under TDP-43 depletion. siTDP-43, siRNA against TDP-43; siSCR, scrambled control siRNA. d, RT–PCR assay shows KCNQ2 mis-splicing in cortical neurons depleted of TDP-43 by siRNA. e,f, RT–qPCR of mis-spliced KCNQ2^∆E5 that increases (e) and total KCNQ2 transcript that remains unaltered (f) after TDP-43 depletion. For e and f, circles represent n = 3 biological replicates, values represent mean expression; error bars, s.e.m. g–i, TDP-43 directly binds to KCNQ2 pre-mRNA. g, iCLIP sequencing coverage in the KCNQ2 locus suggests that TDP-43 binds KCNQ2. TDP-43 consensus motifs (UGNNUG) and (UG)_n simple repeat are shown below the gene model. RIP amplicon used in h and i is indicated. h, Cross-linking immunoprecipitation followed by RT–PCR analysis shows enrichment of KCNQ2 in the TDP-43-bound fraction compared to IgG immunoprecipitates (IP) using the same quantity of lysate for IPs. i, Quantification of results present in h. Circles represent n = 3 biological replicates; values represent mean relative enrichment; error bars, s.e.m. For e, f and i, P values are the result of an unpaired two-sided t-test. Source data

Fig. 2. Expression of KCNQ2^∆E5 is highly specific to ALS/FTD and TDP-43 pathology.

a, Target ALS/New York Genome Center RNA-seq datasets, including 118 non-neurological controls, 982 ALS and 24 other neurological disease samples from distinct CNS regions. b, Strategy to detect KCNQ2^∆E5 reads. c, Violin plots of CNS samples with KCNQ2^∆E5 reads in RNA-seq datasets stratified by disease status reveal specificity to ALS. d, Percentage of ALS samples across CNS regions with KCNQ2^∆E5 transcripts. Red shade intensity is relative to the proportion of samples with detectable KCNQ2^∆E5. e, Violin plots of ALS samples with KCNQ2^∆E5 reads in RNA-seq datasets stratified by genetics reveal specificity to TDP-43 pathologies. In c and e, each point represents a unique sample, and the y axis indicates the number of unique KCNQ2^∆E5 reads. P values are the result of an unpaired one-sided Wilcoxon rank sum test. f–i, RT–qPCR analysis in brain samples from patients with ALS/FTD (n = 89) or controls (n = 27) for KCNQ2^∆E5 (f) and wild type KCNQ2 (g). Boxplots represent interquartile range (IQR), whiskers indicate IQR limits ±1.5× IQR. h,i, KCNQ2^∆E5 abundance increases with pTDP-43 levels (h), and KCNQ2^∆E5 abundance is higher in individuals with early disease onset (i). For additional statistics, see Supplementary Table 2. For f–i, points indicate data obtained from unique postmortem samples. P values are the result of linear regression models adjusted for age at death, sex and RIN. j,k, Abundance of KCNQ2^∆E5 in TDP-43-high and TDP-43-low neuronal nuclei from patients with ALS/FTD with C9orf72 hyperexpansions. j, Top, schematic of neuronal nuclei sorted by flow cytometry into fractions with high and low TDP-43 levels. Bottom, quantification of KCNQ2 shows 100% wild type KCNQ2 transcripts in TDP-43-high nuclei and ~50% reduction in TDP-43-low nuclei and a concomitant ~50% increase in KCNQ2^∆E5 transcripts. Violin plots of KCNQ2 levels; each point represents a unique sample. k, Sashimi plot depicting HISAT2-mapped sequencing coverage along exons 4–6 of KCNQ2 in purified neuronal nuclei from individual patients. Canonical junctions (exons 4–5 and 5–6) are displayed with solid lines; KCNQ2 exon 5 skipping reads are displayed with red dashed lines; read counts are denoted. KCNQ2 exon skipping is only detected in TDP-43-low neurons. Source data

Fig. 3. KCNQ2^∆E5 is a loss-of-function allele exerting dominant negative activity.

a, Amino acid sequence (220–280) and illustration for KCNQ2^WT, KCNQ2^∆E5 and DEE variants KCNQ2^G279S and KCNQ2^T274M. Exon 5 deletion is shown by the dashed line. b, Schematic outlining heterologous expression model and SyncroPatch recordings. Left, KCNQ2 transgenes are transfected into CHO-K_v7.3 cells either alone (homozygous; top left) or with KCNQ2^WT (heterozygous; bottom left). Right, cells are plated into 384-well plates for automated voltage-clamp recordings. c, SyncroPatch results for homozygous expression of KCNQ2 transgenes. Summary data (means; error bars, s.e.m.) for peak outward current density as a function of command voltage. P value determined by one-way ANOVA, #P < 0.001 for all steps starting at −50 to +40 mV for KCNQ2^∆E5, G279S and T274M relative to WT. n represents cells denoted within the panel. d, Representative KCNQ2 current (XE991-sensitive) traces. No significant current is observed from KCNQ2^∆E5, KCNQ2^T274M or KCNQ2^G279S co-expressed with KCNQ3. Scale bar, 50 pA, 200 ms. e, Peak current density at +40 mV plotted as a percentage of KCNQ2^WT current (means; error bars, s.e.m.); n represents cells as in c. P value determined by one-way ANOVA, pairwise multiple comparisons by Holm–Sidak method. Circle represents the average of all cells, which are shown individually in Extended Data Fig. 4d,e. f, SyncroPatch results for heterozygous (×2) expression of KCNQ2 transgenes. Summary data (means; error bars, s.e.m.) for peak outward current density. One-way ANOVA, #P < 0.001 for all steps starting at −40 to +40 mV for KCNQ2^∆E5, G279S and T274M relative to WT; n represents cells denoted within the panel. g, Representative traces as in d. Scale bar, 50 pA, 200 ms. h, Peak current density at +40 mV plotted as a percentage of KCNQ2^WT (×2) current (means; error bars, s.e.m.); n represents cells as in f. P value determined by one-way ANOVA, pairwise multiple comparisons by Holm–Sidak method. Each circle represents the average of cells recorded from n = 3 independent experiments, as shown in Extended Data Fig. 4f,g. See Supplementary Table 3 for statistics for c and f. i, Top, schematic of correlation between clinical severity of KCNQ2 variants and M-current reduction. Variants are classified as PV (non-pathogenic population variants), SLFNE (mild cases of epilepsy) and DEE (severe developmental and epileptic encephalopathy). Bottom, SyncroPatch data from grouped KCNQ2 variants (heterozygous with KCNQ2^WT, previously reported) in comparison to KCNQ2^∆E5. Each circle represents the mean value recorded from all cells of a particular variant. PVs (n = 15) result in 101 ± 8% of WT, SFLNE (n = 15) in 74.1 ± 6% of WT and DEE variants (n = 40) in 56.6 ± 5% of WT current density. KCNQ2^∆E5-expressing cells are identical to the average of DEE variants (56.6 ± 5% of WT). No statistics were run; error bars, s.e.m. j, Top, schematic of cell surface biotinylation experiments. Bottom, western blot of input and cell surface fractions from WT/WT (GFP–KCNQ2^WT + mCh–KCNQ2^WT) and ∆E5/WT (GFP–KCNQ2^∆E5 + mCh–KCNQ2^WT). k,l, Quantification of surface levels of GFP–KCNQ2 (k) and mCh–KCNQ2 (l) normalized to transferrin receptor (TfR). Relative intensities are normalized to WT/WT in each experiment and plotted as fold change. Each shape represents the result from one of n = 3 independent experiments. P value determined by two-tailed paired t-test; error bars, s.e.m. Source data

Fig. 4. KCNQ2^∆E5 accumulates in the ER and causes hyperexcitability in edited iPS cell-derived neurons.

a, Schematic of CRISPR–Cas9 editing of embryonic stem cells (ESCs) to generate homozygous KCNQ2^∆E5/∆E5 lines and differentiation into cortical excitatory neurons. b, RT–PCR shows that KCNQ2^∆E5/∆E5 neurons express KCNQ2^∆E5 and KCNQ2^WT/WT express KCNQ2^WT. c, Representative immunocytochemical images of neurons stained with DAPI, MAP2, KCNQ2 and ANK-G. Arrowheads denote the beginning of AIS. Asterisk (*) denotes KCNQ2 localization in the AIS for WT (top) and accumulated in the soma for KCNQ2^∆E5/∆E5 neurons (bottom). Yellow dashed line outlines the cell body. Scale bar, 10 μm. d, Percentage of WT (n = 51, 0%) and KCNQ2^∆E5/∆E5 (n = 79, 100%) neurons with somatic accumulation of KCNQ2. e, Quantification of KCNQ2 signal intensity variation. Statistical significance determined by unpaired, two-tailed Student’s t-test. Data are shown as mean ± s.e.m.; each circle corresponds to one neuron. a.u., arbitrary units. f, Representative image of KCNQ2^∆E5/∆E5 neuron stained with DAPI, MAP2, calnexin and KCNQ2. Top, maximum Z-projection; middle, 3D views with neuron rotated forward; bottom, 3D views from below. g, Representative raster plot of neuronal activity recorded in a MEA well for control (top) and KCNQ2^∆E5/∆E5 (bottom). Rows depict individual electrodes; black lines represent single spikes; blue lines indicate ‘bursts’. h–o, Longitudinal analysis of neuronal MEA recordings for days 9–43 (h–k) or 12–43 (l–o). Data are presented as means from n = 3 independent experiments (n = 59 wells for WT and n = 64 for KCNQ2^∆E5/∆E5); circles represent means; shaded areas, s.e.m. Two-way repeated-measures ANOVA was used for h, i and l; mixed-effects model restricted maximum likelihood) for j, k, m, n and o. P values: in black indicate genotype effects and in pink reflect genotype × day interactions. MEA metrics are indicated within each panel. p, Experimental schematic (top) and representative raster plot from MEA wells (bottom) during treatment with the K_v7 agonist ICA-069673 (1 μM). For each metric, pre-ICA-069673 and post-ICA-069673 values are represented as the percent of baseline values (right). Each circle-pair represents the change in activity recorded from a well (total number of wells from two replicate MEA plates were combined for analysis: n = 20 for WT and n = 19 wells for KCNQ2^∆E5/∆E5. P value determined by unpaired, two-tailed Student’s t-test. Source data

Fig. 5. TDP-43-depleted iPS cell-derived spinal MNs exhibit hyperexcitability that can be rescued by KCNQ2 smASOs.

a, Top, KCNQ2 and smASO design strategy. Bottom, RT–qPCR of KCNQ2^∆E5 expression. P value determined by an unpaired, two-tailed t-test of each smASO relative to splice-modulating control ASO (smCTRL). b, Experimental schematic: iPS cell MNs are treated with a control ASO (CTRL) or an ASO to knock down TDP-43 (TDP-KD) and smASOs between days 20 and 30 for analysis; images of neurons analyzed by patch clamp. Three treatment groups: (1) CTRL + smCTRL, (2) TDP-KD + smCTRL and (3) TDP-KD + smKCNQ2. c, RT–qPCR of TARDBP expression. P value determined by a paired, two-tailed t-test; ns: not significant. d, RT–PCR shows KCNQ2 mis-splicing upon TDP-43 KD and partial rescue with smASO treatment. e, RT–qPCR of KCNQ2^∆E5. P value determined by a paired, two-tailed t-test. For a, c and e, each circle represents one of n = 4 biological replicates; data are presented as means; error bars, s.e.m. f, Representative traces of spontaneous APs counted while cells were at RMP for all three groups. Asterisks denote individual spontaneous APs; black dashed line denotes 0 mV; colored dashed line denotes RMP. Scale bar, 10 mV, 1 s. g, RMP recorded from all three groups; number of neurons: (1) n = 53, (2) n = 63, (3) n = 30. P value determined by one-way ANOVA (F_2,144 = 8.3, P = 0.0004) followed by Tukey’s multiple comparisons test. h, Spontaneous AP frequency recorded from all three groups; number of neurons: (1) n = 54, (2) n = 63, (3) n = 30. P value determined by one-way ANOVA (F_2,143 = 16.82, P < 0.0001) followed by Tukey’s multiple comparisons test. i, Representative traces of the post-burst mAHP. Scale bar, 5 mV, 250 ms. j, Post-burst mAHP recorded from all three groups; number of neurons: (1) n = 51, (2) n = 55, (3) n = 28. P value determined by one-way ANOVA (F_2,131 = 4.7, P = 0.0105) followed by Tukey’s multiple comparisons test. k, Post-burst sAHP recorded from all three groups; number of neurons: (1) n = 51, (2) n = 55, (3) n = 28. P value determined by one-way ANOVA (F_2,131 = 5.01, P = 0.008) followed by Tukey’s multiple comparisons test. For g, h, j and k, each circle represents neurons recorded from n = 4 biological replicates. Data are presented as means; error bars, s.e.m. Source data

Fig. 6. TDP-43-depleted iPS cell-derived cortical Halo-iNeurons exhibit hyperexcitability that can be rescued by KCNQ2 smASOs.

a, Experimental schematic: cortical neurons were differentiated from HaloTag TDP-43 iPS cells and treated with vehicle (CTRL) or PROTAC to knock down TDP-43. Some TDP-43 KD neurons were also treated with control smASO or smKCNQ2 to repress KCNQ2^∆E5. b, Left, RT–PCR assay shows KCNQ2 mis-splicing upon TDP-43 KD and partial rescue with smASO treatment. Right, RT–qPCR of KCNQ2^∆E5. P value determined by unpaired, two-tailed t-test. Data are means; error bars, s.e.m. Circles are one of n = 3 biological replicates. c, Left, M-current from CTRL (n = 29) and TDP-43 KD (n = 36) neurons. P value determined by unpaired, two-tailed t-test. Right, representative traces and measurements of individual neurons for pre-XE991 and post-XE991 treatment (n = 7). P value determined by paired, two-tailed t-test. d, Left, M-current from TDP-43 KD + smCTRL (n = 27) and TDP-43 KD + smKCNQ2 (n = 28) neurons. P value determined by unpaired, two-tailed t-test. Right, representative traces and measurements of individual neurons for pre-XE991 and post-XE991 treatment (n = 8). P value determined by paired, two-tailed t-test. For c and d, circles represent one cell; data are means; error bars, s.e.m. Scale bar, 100 pA, 500 ms. e,f, Representative traces of spontaneous APs. e, CTRL neurons were not firing at rest (CTRL, −XE991), while XE991 caused increasing spiking (CTRL, +XE991). TDP-43 KDs were active at rest (TDP-43 KD, −XE991), while XE991 did not cause significant firing changes (TDP-43 KD, +XE991). Scale bar, 20 mV, 500 ms. f, TDP-43 KD + smCTRL neurons were active at rest (TDP-43 KD + smCTRL, −XE991) and unresponsive to XE991 (TDP-43 KD + smCTRL, +XE991). TDP-43 KD + smKCNQ2 neurons were quiet at rest (TDP-43 KD + smKCNQ2, −XE991) and responsive to XE991 (TDP-43 KD + smKCNQ2, +XE991). Scale bar, 20 mV, 500 ms. g, Left, RMP from CTRL (n = 29) and TDP-43 KD (n = 35) neurons. Right, RMP from TDP-43 KD + smCTRL (n = 27) and TDP-43 KD + smKCNQ2 (n = 27) neurons. P values determined by unpaired, two-sided t-test; each circle represents one cell. Data are means; error bars, s.e.m. h, Left, spontaneous AP frequency from CTRL (n = 29) and TDP-43 KD (n = 37) neurons. P value determined by unpaired, two-sided t-test. Right, AP measurements of individual CTRL (top) and TDP-43 KD (bottom) neurons for pre-XE991 and post-XE991 treatment. P values determined by paired, two-sided t-test; each circle represents one cell. Data are means; error bars, s.e.m. i, Left, spontaneous AP frequency from TDP-43 KD + smCTRL (n = 27) and TDP-43 KD + smKCNQ2 (n = 27) neurons. P value determined by unpaired, two-sided t-test. Right, AP measurements of individual KD + smCTRL (top, n = 9) and KD + smKCNQ2 (bottom, n = 8) neurons for pre-XE991 and post-XE991 treatment. P values determined by paired, two-sided t-test; each circle represents one cell. Data are means; error bars, s.e.m. Source data

Fig. 7. KCNQ2 protein forms focal accumulations in postmortem spinal MN samples from patients with ALS.

a–c, Samples analyzed at Northwestern University (NU cohort). a, Immunohistochemistry of neurons from non-neurological control (n = 3) and sporadic ALS (sALS) (n = 3) spinal cord sections stained for DAPI (blue), MAP2 (white), KCNQ2 (green) and TDP-43 (red). b, KCNQ2 signal intensity variation was significantly higher in sALS cases: controls, (1) n = 25, (2) n = 21, (3) n = 14; sALS, (1) n = 19, (2) n = 20, (3) n = 30. Data are shown as coefficients of variation (mean-normalized standard deviation); each circle represents one neuron. c, Neurons stained for DAPI (blue), MAP2 (white), KCNQ2 (green) and ubiquitin (red). The square box denotes the magnified region shown at the bottom. Scale bar, 10 μm. d–i, Samples analyzed at Columbia University (CU cohort). d, Immunohistochemistry of MNs from non-neurological control and ALS lumbar spinal cord sections stained for DAPI (blue), KCNQ2 (green) and pTDP-43 (red). e, KCNQ2 signal intensity variation was significantly higher in ALS cases: controls, n = 29 neurons from three individuals; ALS, n = 142 neurons from 15 patients. f, KCNQ2 signal intensity variation was significantly higher in C9orf72 repeat expansion and sALS cases: controls, n = 29 neurons from three individuals; C9orf72, n = 40 from four patients; sALS, n = 102 from 11 patients. g, Neurons stained for DAPI (blue), KCNQ2 (green) and GRP78 (orange). h, sALS neurons from spinal cord sections stained for DAPI (blue), KCNQ2 (green) and pTDP-43 (red). Representative cases of localization of KCNQ2 in large globular pTDP-43 aggregates. i, KCNQ2 signal intensity variation was significantly higher in neurons from ALS patients, irrespective of pTDP-43 pathology: control, n = 29; ALS without TDP-43 pathology, n = 86; ALS with TDP-43 pathology, n = 56. P values in b and e determined by unpaired, two-tailed t-test with Welch’s correction; in f and i by Brown–Forsythe and Welch ANOVA test followed by Benjamini and Hochberg correction for multiple comparisons. Scale bars for NU, 10 μm; for CU, 15 μm. Detailed sample information is included in Supplementary Table 5. Source data

Extended Data Fig. 1. Identification and prioritization of TDP-43-dependent KCNQ2 mis-splicing.

(a) Schematics of distinct classes of alternative splicing events. (b) Heatmap and classification of splicing events that are altered with TDP-43 depletion. Percent Spliced Index (PSI) quantifies proportion of reads mapping to one splice variant relative to all splice variants from the same locus. de novo junctions are splice variants that are not normally expressed at detectable levels. (c) Overlap of 125 differential local splice variants (LSVs) detected in purified MNs after TDP-43 knockdown, with other datasets of TDP-43 depletion reported in Liu et al. and Brown et al.. (d) Related to Fig. 1b. Left: of the 522 genes with altered splicing in neuronal datasets, 47 are involved in the “regulation of membrane potential” and “synaptic signaling”. Right: list of 47 genes. Those with de novo events, such as cryptic exon de-repression are shown in red. (e-f) Genes involved in the regulation of membrane potential and synaptic signaling are not commonly differentially expressed (DE) in neurons depleted of TDP-43. (e) Total number of DE genes in each set of neuronal datasets. (f) Left: number of DE genes in each set, involved in the “regulation of membrane potential” and “synaptic signaling” and Right: Overlap of DE genes. (g) Sashimi plot depicting HISAT2-mapped sequencing coverage and junction spanning reads of alternatively spliced locus in CACNA1E. Inclusion of annotated cassette exon increases following TDP-43 loss. (h) Top: KCNQ2 gene model, including location of skipped exon 5. Bottom: exon 5 is deeply conserved. Graphic is generated using the UCSC genome browser, where each dot indicates conservation of the related amino acid. (i) Amino acid sequence alignment for WT and KCNQ2^∆E5. A fragment of the peptide is shown, and individual domains are boxed, shaded, and annotated. Exon 5 codes for transmembrane domain 5, extracellular components and a portion of the intramembrane pore-forming domain. (j) Confirmation of mis-spliced KCNQ2 by Sanger sequencing. Source data

Extended Data Fig. 2. Identification and prioritization of TDP-43-dependent KCNQ2 mis-splicing.

(a) Verification of mis-spliced KCNQ2 in day 40 human iPSC-MNs. Representative image of n > 3 biological replicates; marker in base pairs. (b) KCNQ2^∆E5 abundance increases with TDP-43 knockdown level. Comparison of KCNQ2^∆E5 and TDP-43 level in reported datasets^,. (c-e) Upregulation of KCNQ2^∆E5 and decrease of KCNQ2^WT in Dox-inducible TDP-43 aggregation model in SH-SY5Y cells. (c) Detection of KCNQ2^∆E5 upon Dox induction of TDP-43 containing 12 Q/N domain repeats; marker in base pairs. (d) Percent Spliced Index (PSI) of KCNQ2 (exon 5 inclusion or skipping) in control versus aggregation-prone TDP-43 cells. Points represent n = 3 independent biological replicates; values are mean +/- SEM; p-values are the result of unpaired two-tailed t-test. (e) Immunofluorescence of TDP-43 aggregates in SH-SY5Y 12Q/N cells; NT=not treated control. (f) Conservation of genomic locus of KCNQ2 exons 4-6 between human and mouse. Mouse conservation is reflected on a black to white color scale where black is highly conserved and white is not conserved. Human KCNQ2 exonic sequences are conserved, while intronic sequences are poorly conserved in mice. The 1-kb long UG repeat found upstream of human exon 5 is not present in the mouse genome. (g-h) Splicing of murine Kcnq2 is not regulated by TDP-43; data from Polymenidou et al.. (g) TDP-43 depletion in mouse brains treated with Tardbp targeting ASOs. (h) Sashimi plot depicting HISAT2-mapped sequencing coverage and junction spanning reads of syntenic mouse Kcnq2 locus. There is no exon skipping following TDP-43 reduction (TDP-43 antisense oligonucleotide; ASO). (i) Comparison of KCNQ2^∆E5 abundance in human and mouse stem-cell derived neurons with TARDBP knockdown. qRT-PCR-based relative expression of KCNQ2^∆E5 and TARDBP. Circles represent relative expression of KCNQ2^∆E5 or TARDBP from n = 3 independent biological replicates of TARDBP knockdowns in MNs. Data are presented as mean +/- SEM; p-values are the result of unpaired t-test. (j) Top: Sashimi plot depicting HISAT2-mapped sequencing coverage and junction spanning reads of the entire mouse Kcnq2 gene. Bottom: gene models. TDP-43 depletion did not induce alternative splicing in this gene; Data from Polymenidou et al.. Source data

Extended Data Fig. 3. Expression of KCNQ2^∆E5 is highly specific to ALS/FTD and TDP-43 pathology.

(a) Reads mapping to KCNQ2^∆E5 in RNA-seq datasets stratified by disease status and CNS tissue source. Violin plots shown, each point represents a unique sample, and the score indicates number of unique KCNQ2^∆E5 reads. P-values are the result of unpaired one-sided Wilcoxon Rank Sum tests. (b) Boxplots displaying abundance of KCNQ2, STMN2 and UNC13A in CNS regions. Points reflect gene expression from individual patient CNS samples from Target ALS/NYGC. Boxes show the interquartile range of gene expression in TPMs, whiskers reflect those beyond the interquartile range. Number of unique samples indicated within the panel. (c) Relative abundance of TDP-43 targets KCNQ2, STMN2 and UNC13A in human cortex and spinal cord using the same datasets employed for splicing analyses in Fig. 2a-e. Data shown as violin plots with underlying gene expression values from postmortem samples (datapoints). P-values are the result of unpaired, two-sided t-tests. (d) Proportion of regional CNS tissue samples with aberrant splicing events. Left: KCNQ2^∆E5; Middle: cryptic truncation of STMN2; Right: cryptic exon inclusion in UNC13A. Numbers of samples with and without detection of events are included (detected: not detected). (e) Expression of KCNQ2, STMN2 and UNC13A in different cell types in the human cortex. Data from Schirmer et al. and Velmeshev et al.. (f) Related to Fig. 2f-i. Correlation between KCNQ2^∆E5 abundance by qRT-PCR and disease duration. P-values are the result of linear regression models adjusted for age at death, sex, and RIN. For additional statistics see Supplementary Table 2. Source data

Extended Data Fig. 4. KCNQ2^∆E5 is a loss-of-function allele exerting dominant negative activity.

(a) Structural predictions generated by Alpha Fold for single KCNQ2 subunits, WT (top left) and KCNQ2^∆E5 (bottom left); and heterotetrametric channels consisting of WT (x2) + KCNQ3 (x2) (top right); or WT + KCNQ2^∆E5 + KCNQ3 (x2) (bottom right). (b) Peak current density recorded at -30mV to +40 mV steps (same cells and n as in Fig. 3f plotted as a percentage of WT (x2) current (mean ± SEM); p-values: one-way ANOVA, pairwise-multiple comparisons by Holm-Sidak method. (c) Voltage-dependence of activation measured in heterozygous experiments for WT (x2) (V½= -27.4 ± 0.8, k = 11.8 ± 0.3, n = 98), WT + KCNQ2^∆E5 (V½= -27.3 ± 0.8, k = 11.9 ± 0.3, n = 107), WT + KCNQ2^T274M (V½= -25.7 ± 0.7, k = 10.9 ± 0.3, n = 129), WT + KCNQ2^G279S (V½= -27.8 ± 0.9, k = 9.6 ± 0.2, n = 98). (d-g) Peak current density values plotted as a percentage of WT (x1) for (d-e) or WT (x2) for (f-g). Data separated into replicate experiments. Each circle is one cell, bars are mean ± SEM. P-value between WT and KCNQ2 variants determined by one-way ANOVA, multiple comparisons with Tukey’s test. (d-e) Data for homozygous KCNQ2-expressing cells was collected from one replicate experiment. Cells were held at voltage steps ranging from -30mV to +40 mV (same cells and n as shown in Fig. 3e). Data for (d) - 20 mV and (e) + 40 m voltage steps shown (1 value out of visual range for replicate experiment #1 at -20mV, KCNQ2^WT: 324.8%). (f-g) Data for heterozygous KCNQ2-expressing cells was collected from n = 3 replicate experiments. Cells were held at voltage steps ranging from -30mV to +40 mV (same cells and n as in Fig. 3h and Ext. Figure 4b-c). Data for (e) -20mV and (f) + 40 mV voltage steps shown (2 values out of visual range for replicate experiment #2 at -20mV, WT: 335.5% and at +40 mV, KCNQ2^∆E5: -51.6%). (h) Manual whole-cell voltage clamp recordings of fused (GFP or mCherry) hKCNQ2 transgenes used in cell surface biotinylation experiments. Current recorded from CHO-Kv7.3 cells expressing: GFP- KCNQ2^WT (n = 6), GFP-KCNQ2^∆E5 (n = 5), mCh-KCNQ2^WT (n = 5), KCNQ2^WT-IRES-GFP (n = 6). Summary data (mean ± SEM) for outward current density as a function of command voltage. Fusion does not cause significant alterations in current (KCNQ2^WT-IRES-GFP vs. GFP- KCNQ2^WT: p = 0.4695 and KCNQ2^WT-IRES-GFP vs. mCh-KCNQ2^WT: p = 0.0815; both determined by repeated measures ANOVA. (i-j) WB replicates #2 and #3 of Fig. 3j-l. Source data

Extended Data Fig. 5. KCNQ2^∆E5 accumulates in the ER and causes hyperexcitability in edited iPSC- derived neurons.

(a) Details of CRISPR mutagenesis of KCNQ2. The four gRNAs targeting KCNQ2 are presented along with details of deletions induced in and KCNQ2^∆E5/∆E5 cells. (b) KCNQ2 allele copy number assay for unrelated iPSC cell line, isogenic control and KCNQ2^∆E5/∆E5 ESCs. (c) Karyotype results for isogenic control and KCNQ2^∆E5/∆E5 ESCs. (d) Representative images of NGN2 cortical neurons stained with DAPI, GFP, and MAP2. (e) Representative images of NGN2 cortical neurons stained with DAPI, MAP2 and KCNQ2. Scale bar: 25 μm. Yellow dashed line outlines neuronal cell body. Letters signify individual neurons for which greyscale images of MAP2 and KCNQ2 signal are magnified. Scale bar: 10 μm. (f) Quantification of the mean KCNQ2 signal was significantly higher in KCNQ2^∆E5/∆E5 neurons (p < 0.0001). (g) Quantification of the max KCNQ2 signal was also significantly higher in KCNQ2^∆E5/∆E5 neurons (p = 0.0035). Statistical significance for (f-g) was determined by unpaired, two-tailed student’s t-test. Data are shown as mean ± SEM, each circle corresponds to one neuron (control: n = 51, KCNQ2^∆E5/∆E5: n = 79). Source data

Extended Data Fig. 6. TDP-43 depleted iPSC-derived spinal motor neurons exhibit hyperexcitability that can be rescued by splice-modulating KCNQ2 ASOs.

(a) Schematic illustrating splice modulating ASO (smASO) screen strategy in iPSC-MNs. (b) qRT-PCR of TARDBP expression in iPSC-MNs at day 32 of the screen; KD: knockdown. Data are presented as mean ± SEM; Statistical significance was determined by 2-tailed, unpaired t-test. Circles represent a biological replicate. (c) qRT-PCR of mis-spliced KCNQ2^∆E5 after treatment with smASO-KCNQ2 candidate #2 at three doses (1μM, 3μM, 10μM). Data were presented as mean ± SEM; p value is the result of an unpaired, 2-tailed t-test. Circles represent a biological replicate. (d) Representative ICC images of iPSC-derived MNs stained for DAPI, TDP-43 and MAP2. Scale bar: 10 μm. (e) Cell capacitance recorded from all 3 groups; number of neurons: TDP-KD+smCTRL n = 55, TDP-KD+smCTRL n = 60, TDP-KD+smKCNQ2 n = 39 and (f) Input resistance; number of neurons: TDP-KD+smCTRL n = 49, TDP-KD+smCTRL n = 53, TDP- KD+smKCNQ2 n = 27. Statistical significance determined by a one-way ANOVA; no significant differences between groups. Data represented as mean ± SEM. (g) Representative traces for spontaneous APs recorded from neurons for all 3 groups. Scale bar: 10 mV/5 ms. (h-n) Spontaneous AP properties from neurons for all 3 groups; number of neurons: CTRL+smCTRL n = 51, TDP-KD+smCTRL n = 57, TDP-KD+smKCNQ2 n = 34. Each circle represents the average of all spontaneous APs recorded from one cell over the first 15 seconds of whole cell configuration. Cells recorded from n = 4 independent biological replicates. Statistical significance determined by a one-way ANOVA followed by a Tukey’s multiple comparisons test when applicable. Data presented as the mean ± SEM. Source data

Extended Data Fig. 7. TDP-43 depleted iPSC-derived cortical Halo-iNeurons exhibit hyperexcitability that can be rescued by splice-modulating KCNQ2 ASOs.

(a) Experimental schematic of HaloTag TDP-43 iPSC line and degradation via ubiquitin-proteasomal pathway. (b) WB analysis of TDP-43 in KD experiments. (c) Top: schematic of treatment groups. Bottom: RT-PCR assay reveals some mis-splicing of KCNQ2 because of tagging TDP-43 and significant enhancement upon TDP-43 KD. Right: qRT-PCR analysis of mis-spliced KCNQ2^∆E5. Data are mean ± SEM; p-value: one way ANOVA (F(2,9) = 77.97), 13 control vs. Halo+DMSO, I3 vs. Halo+PROTAC, Halo+DMSO vs. Halo+PROTAC. Circles represent 1 of n = 4 biological replicates. (d) Cell capacitance and (e) Input resistance was not significantly different between groups, CTRL (n = 29, n = 28), TDP-43 KD (n = 37, n = 34), TDP-43 KD+smCTRL (n = 21, n = 19) and TDP43 KD+smKCNQ2 (n = 23, n = 22). Each circle represents one cell. Data are mean ± SEM; p-value: unpaired, two-tailed t-test (ns: not significant, p > 0.05 in all cases). (f) Left: mAHP measured from single APs recorded from CTRL (n = 34) and TDP-43 KD neurons (n = 27). Each circle represents one cell. Data are mean ± SEM; p-value: unpaired, two-tailed t-test. Right: Representative traces and quantification of mAHPs of XE991 treatment. Data presented as pre-and post- XE991 for individual cells from CTRL (n = 7) and TDP-43 KD neurons (n = 7); p-value: paired two-tailed t-test. (g) Left: mAHP measured from single APs recorded from TDP-43 KD+smCTRL (n = 25) and TDP-43 KD+smKCNQ2 (n = 27). Each circle represents one cell. Data are mean ± SEM; p-value: unpaired, two-tailed t-test. Right: Representative traces and quantification of mAHPs of XE991 treatment for TDP-43 KD+smCTRL experiments. Data presented as pre-and post-XE991 for individual cells from TDP-43 KD+smCTRL (n = 8) and TDP-43 KD+smKCNQ2 (n = 8); p-value: paired, two-tailed t-test. Source data

Extended Data Fig. 8. KCNQ2 protein forms focal accumulations in postmortem ALS patient spinal MNs.

(a-d) Samples analyzed at Northwestern University (NU cohort). (a-b) Related to Fig. 7a; representative images from control and sALS spinal cord sections stained for DAPI (blue), MAP2 (white), KCNQ2 (green), TDP-43 (red). (c-d) Related to Fig. 7c; representative images from control and sALS spinal cord sections stained for DAPI (blue), MAP2 (white), KCNQ2 (green), Ubiquitin (red). (e-h) Samples analyzed at Columbia University (CU cohort). (e-f) Related to Fig. 7d; representative images from healthy control and ALS spinal cord sections stained for DAPI (blue), KCNQ2 (green) and phospho-TDP-43 (red). (f) Note selective signal of phospho-TDP-43 in ALS samples. (g) Higher normalized KCNQ2 intensity in ALS samples relative to control samples. Neuron counts; control n = 29 from 3 individuals, ALS n = 142 from 15 patients; p-value determined by unpaired, two-tailed t-test with Welch’s correction. (h) Representative image from ALS spinal cord section stained for DAPI (blue), KCNQ2 (green) and Ubiquitin (orange). Note cytoplasmic overlap between KCNQ2 accumulations and Ubiquitin; n = 3 sALS and control cases were assessed. Scale bars for NU: 20 μm and CU: 15 μm. Source data