ALS-linked TDP-43 mutations produce aberrant RNA splicing and adult-onset motor neuron disease without aggregation or loss of nuclear TDP-43

Establishment of TDP-43 Transgenic Mice Expressing Wild-Type and Mutant TDP-43 Broadly in the Central Nervous System.

Transgenic mice were produced that express either wild-type or ALS-linked mutant TDP-43 broadly throughout the central nervous system, using the murine prion-promoter (47) previously reported to drive transgene expression most abundantly in the central nervous system, both in neurons and astrocytes (48). cDNAs encoding wild-type or either of two ALS-linked mutants of TDP-43 (6) [Q331K (glutamine to lysine substitution at amino acid position 331) and M337V (methionine to valine substitution at amino acid position 337)] were fused to an N-terminal myc-tag under control of the murine prion promoter (Fig. 1A). From >20 founders for each gene, three lines were selected and established for TDP-43^Wild-Type, TDP-43^Q331K, and TDP-43^M337V, respectively.

Analysis of whole-tissue lysates by immunoblotting with a polyclonal anti–TDP-43 antibody recognizing recombinant mouse and human TDP-43 protein with equal affinity confirmed that transgene expression was confined primarily to the brain and spinal cord, with very low to no expression in some other tissues (Fig. 1B). Immunoblotting of protein extracts or real-time quantitative RT-PCR (RT-qPCR) of RNAs, respectively, from whole brain identified levels of transgene expression for each of the nine lines (Fig. S1). Analysis of cortex and spinal cord identified lines expressing comparable levels of human TDP-43 mRNA, ranging from 1× to 1.5× the level of endogenous TDP-43 in nontransgenic mice, which was recapitulated in the accumulated protein levels from whole brain and spinal cord lysates (Fig. 1 C and D, and Fig. S1A). Immunofluorescent staining with myc antibody to detect human TDP-43 showed that it accumulated in neurons and glia of the spinal cord (Fig. 1E) in a pattern that mimicked the endogenous expression pattern of TDP-43.

Mice Expressing Mutant Human TDP-43 Develop Adult-Onset Motor Deficits and Hindlimb Weakness.

Four lines of mice were selected in which wild-type TDP-43, TDP-43^M337V, or TDP-43^Q331K were accumulated to levels comparable to endogenous TDP-43 in nontransgenic mice, as well as one line with a lower level (referred TDP-43^Q331K-low). At birth, mice expressing wild-type human TDP-43 or either TDP-43 mutant were indistinguishable from their nontransgenic littermates. However, in contrast to their nontransgenic littermates or TDP-43^Wild-Type mice, TDP-43^Q331K transgenic mice developed a tremor at as early as 3 mo of age (Fig. S2A). This tremor worsened with aging (Movie S1) and was accompanied by the development of hindlimb clasping indicative of spastic motor impairment (27, 49) (Fig. S2B).

Motor performance of TDP-43^Wild-Type, TDP-43^Q331K-Low, TDP-43^Q331K, and TDP-43^M337V mice was followed during aging. Motor performance (on an accelerating rotarod) was normal at the earliest time points analyzed in each of the mutant TDP-43–expressing lines (Fig. 2 A and B). In contrast with nontransgenic and TDP-43^Wild-Type mice, all three mutant TDP-43 lines developed significant age-dependent motor deficits by 10 mo of age, with the most severe deficit and earliest onset (3 mo) in the higher expressing TDP-43^Q331K line (Fig. 2B). A significant decrease in hindlimb grip strength by 10 mo of age accompanied age-dependent motor impairment in TDP-43^Q331K animals (Fig. 2C). Analysis of animals at later time points (e.g., after 17 mo of age) showed no further exacerbation of motor impairment (Fig. S3A), indicating a window of adult onset, active degeneration up to 10 mo of age, after which there is little further worsening of motor phenotype despite continued transgene accumulation in the remaining motor neurons and surrounding glia (see, for example, below and Fig. 5).

To determine whether the motor deficits and hindlimb weakness observed in TDP-43^Q331K transgenic mice were accompanied by neuromuscular abnormalities similar to those clinically observed in human ALS, we performed electrophysiological analyses on TDP-43^Wild-Type and TDP-43^Q331K transgenic mice (Fig. 3A). Resting electromyographic (EMG) recordings were obtained from the gastrocnemius muscle in the absence of any stimulus (in isoflurane-anesthetized animals). As expected, in SOD1^G37R symptomatic mice that will go on to develop fatal paralytic disease, high-frequency spontaneous firings of the motor units (fibrillations) were recorded (Fig. 3 B, v), consistent with the known widespread denervation. Recordings in TDP-43^Q331K mice (Fig. 3 B, iii) revealed the presence of muscle fibrillations similar to those observed in presymptomatic SOD1^G37R mice (Fig. 3 B, iv), indicating neuromuscular denervation and motor unit degeneration and regeneration. These spontaneous firings were found only in the presence of mutant TDP-43–dependent mice, as recordings in nontransgenic and expression matched TDP-43^Wild-Type mice (Fig. 3 B, i and ii) lacked such spontaneous EMG activity.

Because both upper and lower motor neuron deficits can develop in ALS patients, we tested if the spontaneous EMG activity associated with neuromuscular denervation observed in TDP-43^Q331K mice was accompanied by upper motor neuron deficits. To this end, myogenic motor evoked potentials (MMEPs), which report a measure of connectivity of the entire neuromuscular unit, including the motor cortex, interneurons, α-motorneurons, and the neuromuscular junction, were recorded from the gastrocnemius muscle after electrical stimulation of the motor cortex (Fig. 3C). Use of this measure in SOD1^G37R mice revealed that, as expected, MMEPs were decreased (from ∼2.4 mV in nontransgenic animals) to 1.5 mV at presymptomatic stages, and were almost completely lost in symptomatic SOD1^G37R mice (0.03 mV). TDP-43^Q331K transgenic mice also developed significant decreases (to 0.18 mV; n = 3) in MMEP amplitude (Fig. 3C) compared with nontransgenic (2.4 mV; n = 4) and TDP-43^Wild-Type transgenic mice (1.8 mV; n = 3), indicating a disruption of the neuromuscular unit.

To determine whether the decrease in MMEPs reported above involved disruption of the connectivity between the upper and lower motor neurons, spinal cord motor-evoked potentials (MEPs) were then recorded from the dorsal surface of an exposed thoracic (T12) segment after electrical stimulation of the motor cortex. MEPs consist of multiple waves, with the two earliest peaks (N1, N2) corresponding to the activation of the extrapyramidal system. No significant differences were found in the amplitude of the N1 wave when comparing nontransgenic (0.20 mV; n = 4), TDP-43^Wild-Type (0.21 mV; n = 5), and TDP-43^Q331K transgenic mice (0.20 mV; n = 4) (Fig. 3D). Quantification of immunofluorescently labeled Ctip2⁺ upper motor neurons in cortex layer V in TDP-43^Q331K transgenic animals showed no loss of upper motor neurons, consistent with the retention of upper motor neuron function indicated by the persistence of cortical MEPs (Fig. S4A). Therefore, spontaneous muscle firing recorded using EMGs, accompanied by decreased MMEP amplitudes, must reflect motor dysfunction associated primarily with lower motor neuron degeneration in TDP-43^Q331K mice.

TDP-43^Q331K Transgenic Mice Develop Age-Dependent Motor Neuron Loss and Motor Axon Degeneration.

Having observed the development of motor deficits and electrophysiological abnormalities, we examined the central nervous system for signs of neurodegeneration. Immunofluorescent staining for the astrocytic marker GFAP and the microglial marker Iba-1 revealed immunoreactive astrocytes and infiltrating microglia (Fig. S4B) in the ventral horn of the spinal cord in 10- to 12-mo-old TDP-43^Q331K transgenic mice. Consistent with the initial identification of the Q331K mutation in an ALS patient with classic lower motor-neuron involvement (6), quantification of choline acetyl-transferase (ChAT)-positive lower motor neurons in lumbar spinal cords from TDP-43^Q331K mice revealed a significant, age-dependent decrease (P = 0.04, n = 3) in motor neuron number. Loss of motor neurons initiated before 2 mo of age and continued until about 12 mo of age, yielding a reduction of ∼35% compared with age-matched nontransgenic or TDP-43^Wild-Type transgenic animals (Fig. 4A). Additionally, the milder motor deficits seen by 10–12 mo of age in both TDP-43^Q331K-Low and TDP-43^M337V mice were accompanied by a trend in age-dependent reduction in ChAT-positive motor neurons compared with nontransgenic or TDP-43^Wild-Type transgenic animals in the ventral lumbar horns at 10–12 mo of age (Fig. 4A).

Quantification of axons remaining in the fifth lumbar (L5) roots of TDP-43^Q331K transgenic mice revealed a significant reduction (623 vs. 923 in nontransgenic, ***P < 0.001) of total motor axons by 10–12 mo of age (Fig. 4B), with the most significant reduction in large caliber α-motor axons (>3.5 μm) that innervate muscle (50) (Fig. 4C). Morphological examination of the remaining motor axons revealed the presence of degenerating axons, characterized by the appearance of vacuolization and myelin defects (Fig. 4D). Similar to what was demonstrated behaviorally, there was no further loss of L5 motor axons at 20 mo of age (Fig. S3 B–D), indicating a period of active degeneration up to 10 mo of age.

We therefore focused our remaining analyses within this disease period. Quantification of postsynaptic neuromuscular junctions using α-bungarotoxin staining of the gastrocnemius muscle revealed a significant reduction (30%; P = 0.02) in the number of neuromuscular junction endplates in of TDP-43^Q331K mice (Fig. 4E). In contrast, TDP-43^Q331K-Lowand TDP-43^M337V transgenic mice displayed numbers of α-bungarotoxin–positive neuromuscular junction endplates that were not significantly different from nontransgenic or TDP-43^Wild-Type animals, consistent with functional deficits without overt structural changes in lower motor neuron synapses. Morphological examination of hematoxylin and eosin (H&E)-stained sections of gastrocnemius muscle from 10-mo-old TDP-43^Q331K animals revealed regions of damaged muscle fibers and regions of regeneration characterized by the appearance of centralized nuclei (Fig. 4G). TDP-43^Wild-Type, TDP-43^Q331K-low and TDP-43^M337V animals lacked similar morphological abnormalities in the muscle fibers. Additionally, neuromuscular junctions in TDP-43^Q331K mice appeared malformed with an abnormal bleb-like appearance (Fig. 4F). Examination of the descending corticospinal tracts in the dorsal and lateral columns of the spinal cords revealed minor degeneration in all mutant-expressing lines, including TDP-43^Q331K-Low and TDP-43^M337V, suggesting that degeneration of other regions of the motor circuit is responsible for the phenotypes observed in the low mutant-expressing lines.

Cytoplasmic Mislocalization of TDP-43 Is Not Required for the Development of Motor Neuron Disease.

A primary pathological feature in ALS patients is the accumulation of cytoplasmic, ubiquitinated and insoluble TDP-43–containing aggregates within the neurons of the nervous system (4). To test whether the development of motor neuron disease was accompanied by a similar alteration in TDP-43 localization, we first performed nuclear and cytoplasmic fractionation of whole spinal cord and brain from 10- to 12-mo-old TDP-43 transgenic mice (Fig. 5A). Immunoblotting for mouse and human TDP-43 in enriched fractions from either cytoplasm (Fig. 5 B, i and iii) or nuclei (Fig. 5 B, ii and iv) derived from cortex or spinal cord revealed that the majority of human TDP-43 remained in the nuclear fraction, just as it did in nontransgenic animals. Examination of longer exposures revealed a similarly small proportion of both endogenous mouse and human TDP-43 fractionated within the cytosol.

Immunofluorescent staining in lumbar spinal cord sections of TDP-43^Q331K mice from 2 to 10 mo of age revealed that the majority of mutant TDP-43 remained nuclear (Fig. 5C). Additionally, analysis of sequential biochemical extraction using high-salt and urea buffers revealed that the majority of both mouse and human TDP-43 was extracted in the high-salt soluble fraction (Fig. S5 B and C). Analysis for nontransgenic and all transgenic lines of a final extraction of the nuclear pellets with SDS demonstrated that only a small fraction (Fig. S5D) of human and endogenous mouse TDP-43 was in an initially insoluble form that could be solubilized with SDS.

Widespread Splicing Changes from Altered TDP-43 Levels Within the Central Nervous System.

To evaluate if normal patterns of alternative splicing were disrupted upon reduction of mouse TDP-43 and its replacement with wild-type or mutant human TDP-43 at or above the level of endogenous TDP-43 in normal mice, RNA extracted from cortices of 2-mo-old mice were examined using Affymetrix splicing-sensitive microarrays (Fig. 6A). Analysis of splicing changes in cortices of TDP-43^Wild-Type, TDP-43^Q331K, and TDP-43^Q331K-Low revealed that 824, 1,195, and 208 exons, respectively, were differentially spliced relative to mRNA from nontransgenic mice, using a statistical threshold that captured as many significant changes as possible (absolute separation score of <0.3) (51, 52) (Fig. 6A).

We focused on the largest represented class of alternative splicing events, namely cassette exons (i.e., exons that can be included or excluded in the final mRNA) (white slice of pie charts in Fig. 6A). For comparison, we reanalyzed our previous splicing array data (11) using a consistent threshold (absolute separation score of <0.3), identifying 737 cassette exons that were alternatively spliced upon depletion of endogenous mouse TDP-43 in the central nervous system (Fig. S6). These exons, the abundance of which is normally regulated by TDP-43 levels, were then compared with the 314, 533, and 97 cassette exons that were misregulated in human TDP-43^Wild-Type, TDP-43^Q331K, and TDP-43^Q331K-Low transgenic animals.

To determine if misregulated exons were associated with endogenous TDP-43 binding [as measured by cross-linking and immunoprecipitation followed by deep sequencing (11)], we separated the 737 exons into “direct” (175, which contained TDP-43 binding sites within 2 kb of the exon) and “indirect” (562 without TDP-43 binding sites near the exon) targets (Fig. 6B). Notably, when comparing the percentages of direct and indirect targets that overlapped with TDP-43^Q331K–dependent misregulated exons, we observed a statistically significant (P < 0.0005 by χ² analysis) 1.8-fold enrichment of direct targets [26% (45/175)] compared with indirect ones [14% (81/562)]. A similar (2.4-fold) enrichment was seen for altered splicing of direct targets in TDP-43^Q331K-Low [6% of direct targets (11/175) compared with 2.6% of indirect targets (15/562)]. No significant enrichment was seen in a similar comparison for TDP-43^Wild-Type [9% (16/175) to 7% (39/562)] (Fig. 6C). Thus, the human mutant transgenes preferentially affect direct, endogenous TDP-43–dependent alternative splicing events.

Mutant TDP-43 Retention or Enhancement of TDP-43 Function in Splicing of Most pre-mRNAs.

To test whether the splicing changes for cassette exons reflected a loss or gain of normal TDP-43 function, we compared the directionality of exon inclusion when TDP-43 was depleted or in response to expression in mice of the transgene-encoded human TDP-43 (Fig. 6D). Of the cassette exon-splicing changes observed in the TDP-43 transgenic mice, we found that 55 of 314 in TDP-43^Wild-Type (Fig. 6 D, i and ii), 126 of 533 in TDP-43^Q331K (Fig. 6 D, iii and iv), and 26 of 97 TDP-43^Q331K-Low (Fig. 6 D, v and vi) overlapped with cassette exon changes identified following TDP-43 depletion. For directly bound exon targets, the splicing of which was altered by expression of human TDP-43, the majority of the affected cassette exons overlapping with direct targets changed in a direction opposite to loss of TDP-43 [TDP-43^Wild-Type (13/16), TDP-43^Q331K (35/45), TDP-43^Q331K-Low (10/11)] (in gray, Fig. 6 D, i, iii, and v).

Furthermore, at higher TDP-43^Q331K levels (and the correspondingly stronger reduction in endogenous mouse TDP-43), splicing of 35 of 45 directly bound cassette exons (Fig. 6 D, iii) changed in a manner consistent with an elevated normal function of endogenous mouse TDP-43 (the opposite of TDP-43 depletion). In other words, these cassette exons, whose inclusion (or exclusion) was previously shown to be dependent on mouse TDP-43, were more included (or excluded) when human TDP-43^Q331K was expressed, consistent with this mutant retaining normal (or enhanced) TDP-43 activity for most pre-mRNAs.

Mutant TDP-43–Dependent Loss-of-Function for Splicing of Specific pre-mRNAs.

Ten direct exon targets in TDP-43^Q331K mice were altered in a manner indicative of a reduction in endogenous TDP-43 function in facilitating the splicing of these TDP-43 target exons (Fig. 6 D, iii). Exons whose splicing were reported altered by microarray analysis, as well as ones of biological interest but not represented on the array, were then analyzed by semiquantitative RT-PCR (Fig. 6 E and F, Fig. S7, and Table S1). To identify and validate splicing changes that are specifically affected by ALS-linked mutant TDP-43, we focused on changes occurring only in both TDP-43^Q331K and TDP-43^Q331K-Low-transgenic mice, but not in TDP-43^Wild-Type mice (the accumulated transgene level of which is similar to that of TDP-43^Q331K-Low transgenic mice). Within these changes, the presence of TDP-43 mutant produced splicing changes in Kcnip2, Abhd14a, Ctnnd1, and Atp2b1, which mimicked reduction of endogenous TDP-43 activity (Fig. 6E, Fig. S7A, and Table S1), demonstrating a loss-of-normal function of human TDP-43^Q331K, which even at a high level of expression did not replace the endogenous mouse TDP-43 for the splicing of these exons.

Dose-Dependent Enhanced Splicing of Some pre-mRNAs by Human TDP-43.

Within the exons that overlapped with those that changed upon TDP-43 depletion (Fig. 6 D, i, iii, and v), we found some that primarily displayed a dose-dependent TDP-43^Q331K or TDP-43^Wild-Type pattern of splicing in the opposite direction to what occurs after TDP-43 depletion (Fig. 6F, Sort1, and Fig. S7B, Ppm2c, AU040829, and Caly, the exons of which were more excluded, in contrast to more included in TDP-43 depletion). Added to this finding, for other pre-mRNA targets, enhancement of normal TDP-43 function in splicing was seen specifically for TDP-43^Q331K, as demonstrated by dose-dependent splicing changes in a direction opposite what is seen with TDP-43 depletion (e.g., Eif4h and Taf1b, the exons of which were more excluded, in contrast to increased inclusion in TDP-43 depletion) (Fig. 6E and Table S1).

These findings for individual RNAs were supported globally by a genome-wide analysis of direct TDP-43 targets, the exons of which were included or excluded upon human TDP-43 expression (Fig. 6G). On average, ∼20% of unaffected exons represented on the array were direct targets (green line in Fig. 6G). However, when focusing on alternatively changing cassette exons, we found an overall increase in the percentages of direct TDP-43 exon targets (∼43%, TDP-43^Q331K, and TDP-43^Wild-Type; up to 53% in TDP-43^Q331K-low) that were excluded in the presence of the human TDP-43 protein, compared with only ∼20% of the included exons (Fig. 6 G, ii). Surprisingly, 39% of excluded exons in the TDP-43^Q331K animals not previously demonstrated as regulated by endogenous TDP-43 depletion were also direct targets (Fig. 6 G, iii). That is, some exons that were unchanged following TDP-43 depletion displayed a dose-dependent pattern of splicing in the direction of enhanced exclusion in the presence of the human TDP-43 (Fig. 6F, Atxn2, Fig. S7C, Zfp414, and Kcnj3, the exons of which were more excluded, in contrast to no change upon TDP-43 depletion, and Table S1). This finding supports a hypothesis that certain exons, although bound by TDP-43, are only misregulated upon elevation of wild-type or mutant TDP-43, but are not altered upon its loss. In other words, elevation or enhancement of TDP-43 activity through dose or mutation produces aberrant splicing events separate from those resulting from loss of TDP-43.

Mutant-Specific Splicing Alterations Accompany Motor Neuron Disease Development.

Having identified that the human transgenes cause both loss and enhancement of normal function, we focused on alternative splicing changes within the spinal cord, the tissue that develops the degenerative phenotype. RNA extracted from the spinal cords of 2-mo-old nontransgenic, TDP-43^Wild-Type, and TDP-43^Q331K animals before the onset of significant motor dysfunction were analyzed on the same splicing-sensitive microarrays (Fig. 7A). We identified 4,462 and 4,399 alternative splicing events in the TDP-43^Wild-Type and TDP-43^Q331K spinal cord conditions, respectively (Fig. 7A). Approximately 38% of changes in TDP-43^Wild-Type cortex (118) and 42% of changes observed in TDP-43^Q331K cortex (226) were also present in the spinal cord (Fig. 7B). Notably, with the exception of a cassette exon within the Kctd9 gene, altered splicing of all 1,060 cassette exons that overlapped between spinal cord in TDP-43^Wild-Type and TDP-43^Q331K mice occurred in the same direction, demonstrating that at least for this subset of RNA targets, the mutant remains functionally similar to wild-type TDP-43 (Fig. 7C).

Use of semiquantitative RT-PCR validated a set of splicing changes that were consistent with those identified in the cortex (Fig. 7D and Table S1), again demonstrating that TDP-43^Q331K mutant confers both loss and retention of normal TDP-43 function. Furthermore, among the 419 splicing events unique in TDP-43^Q331K spinal cord were several RNAs, the encoded proteins of which are involved in neurological function and transmission (Table S2), including the synaptic cell-adhesion molecules neurexins 1 and 3 (Nrxn1 and Nrxn3), protein phosphatase 3 (also known as calcineurin, Ppp3ca), and glutamate receptor 2 (Gria2, encoding GluR2), which has been proposed to modify motor neuron vulnerability to excitotoxicity (53–55). In contrast, the 394 events unique to the TDP-43^Wild-Type spinal cord condition did not include a similar representation from genes involved in neurological function.

Generation of Transgenic Mice Expressing Floxed Wild-Type and Mutant Human TDP-43.

Flanking SalI digestion sites were inserted by PCR into cDNAs containing N-terminal myc-tagged full-length wild-type or mutant (Q331K or M337V) TDP-43 and cloned into the XhoI site of MoPrP.XhoI (ATCC #JHU-2). The resultant construct was digested upstream of the minimal PrP promoter and downstream of the final PrP exon 3, subcloned into a shuttle vector containing loxP flanking sites (Fig. 1A), and linearized using XhoI. The final construct was injected into the pronuclei of fertilized C57Bl6/C3H hybrid eggs and implanted into pseudopregnant female mice. Founder mice were then backcrossed to C57Bl6 to establish lines (mice used in this report were backcrossed to C57Bl6 for a minimum of four generations).

RT-qPCR and Immunoblotting.

Total RNA from half of the mouse spinal cord was isolated using TRIzol (Invitrogen) extraction and mRNA levels were determined by RT-qPCR using the iQSYBR Green supermix (Bio-Rad). See details in SI Materials and Methods. The protein fraction from each TRIzol sample was immunoblotted (using nitrocellulose membranes) and bound proteins detected using antibodies to mouse and human TDP-43 (ProteinTech 12892, 1:1,000) or mouse GAPDH (AbCam clone 6C5, 1:20,000).

Immunohistochemistry.

Tissue preparation for immunohistochemistry was performed as described previously (50). Next, 30-μm free-floating sections were stained using standard protocols previously established in our laboratory with the indicated antibodies. See SI Materials and Methods for details. Confocal images were acquired on a Nikon Eclipse laser scanning confocal microscope using the Nikon EZ-C1 software.

Morphometric Analysis and Quantification of Motor Axons.

Roots from lumbar level 5 of the spinal cord (L5) were dissected from three to four mice per genotype at 2 mo and 10–12 mo of age, and thin sections (0.75 μm) were cut, stained with Toluidine blue, and quantified as previously described (50). L5 axons are reported as mean ± SD.

Quantification of Upper and Lower Motor Neurons.

ChAT-positive ventral horn motor neurons were counted from at least 30 sections per animal (minimum of three animals per genotype) at 2 and 12 mo of age and are reported as average ± SD. Quantification of Ctip2⁺ neurons in cortex layer five from 10- to 12-mo-old animals was performed on six consecutive sections of frontal cortex (∼2.34 mm Bregma), in the motor M1 region in an area of 0.08 mm² and are reported as the average number of motor neurons per square millimeter ± SEM. Matching sections were chosen for each set of mice.

Quantification of Neuromuscular Junctions.

A total of ∼1,000 neuromuscular junctions were counted from at least 10 sections of gastrocnemius from 10- to 12-mo-old animals. To obtain the number of neuromuscular junctions per section, the average number of neuromuscular junctions was divided by the number of sections counted per animal, with three to four animals per genotype. The average number of neuromuscular junctions per section was reported as mean ± SEM. Statistical analysis was performed using one-way ANOVA with Bonferroni’s post hoc test.

Animal Behavior and Electrophysiology.

These studies were carried out under protocols approved by the Institutional Animal Care and Use Committee of the University of California at San Diego and were in compliance with Association for Assessment of Laboratory Animal Care guidelines for animal use. All studies were performed in such a manner as to minimize group size and animal suffering. See details in SI Materials and Methods.

Nuclear-Cytosolic Fractionation.

Nuclear-cytosolic fractionation of cortex and spinal cords from young and old mice was performed as described previously (26, 62). See SI Materials and Methods for details. Equivalent volumes were prepared for immunoblotting with the indicated antibodies.

Sequential Biochemical Fractionation.

Sequential biochemical fractionation on cortex and spinal cord was performed as described previously (63). For detailed protocols, see SI Materials and Methods. Equivalent volumes of samples were separated on 4–12% Bis-Tris gradient gels for immunoblotting with the indicated antibodies.

Microarray.

Microarray data analysis was performed as previously described (11). For each microarray condition, the log₂ ratio of skipping intensities to inclusion intensities was estimated using least-squares analysis. Significantly changing splicing events between TDP-43^Q331K, TDP-43^Q331K-low, TDP-43^Wild-Type, and control nontransgenic mice were identified using a q-value < 0.05 and an absolute separation score > 0.3.

RT-PCR Validation of Splicing Targets Identified by Microarray.

RT-PCR was performed on cDNA generated from the cortex and spinal cords of 2-mo-old transgenic animals. See details in SI Materials and Methods. Gel imaging and quantification of the isoforms was performed using the Bio-Rad Chemidoc software. The intensity ratios between products including the cassette exon and skipping the cassette exon were averaged for a minimum of three biological replicates per genotype. Splicing alterations significantly different from the nontransgenic RNAs were determined with Student’s t test (*P < 0.05, **P < 0.01, and ***P < 0.001).