Abstract
Neuronal hyperexcitability is a hallmark of amyotrophic lateral sclerosis (ALS), but its relationship with the TDP-43 aggregates that comprise the predominant pathology in over 90% of ALS cases remains unclear. Emerging evidence indicates that TDP-43 pathology induces neuronal hyperexcitability, which may contribute to excitotoxic neuronal death. To characterize TDP-43 mediated network excitability changes in a disease-relevant model, we performed in vivo continuous electroencephalography monitoring and ex vivo acute hippocampal slice electrophysiology in rNLS8 mice (males and females), which express human TDP-43 with a defective nuclear localization signal (hTDP-43ΔNLS). Surprisingly, we identified the presence of seizures in ∼64% of rNLS8 mice beginning ∼2.5 weeks after transgene induction (off-DOX). More broadly, we observed longitudinal changes in cortical EEG patterns and circuit hyperexcitability preceding neurodegeneration of vulnerable hippocampal subfields. Consistent with previous reports, we have observed broad dysregulation of AMPA subunit expression in mice expressing hTDP-43ΔNLS. These changes were most pronounced in the hippocampus, where we hypothesized they promote hyperexcitability and ultimately, excitotoxic cell death. Interestingly, hippocampal injection of AAV encoding inhibitory DREADDs (hM4Di) and daily activation with CNO ligand rescued anxiety deficits on the elevated zero maze but did not reduce neurodegeneration. Moreover, therapeutic doses of the antiseizure medications, valproic acid and levetiracetam, did not improve behavior or prevent neurodegeneration. These results highlight the complex relationship between TDP-43-mediated neuronal hyperexcitability and neurodegeneration. Although targeting hyperexcitability may ameliorate some behavioral deficits, our study suggests it may not be sufficient to halt or slow neurodegeneration in TDP-43-related proteinopathies.
Significance Statement
Cytoplasmic aggregates of TDP-43 are the predominant pathology in over 90% of ALS and 50% of frontotemporal lobar degeneration cases. Understanding how TDP-43 pathology promotes neurodegeneration may lead to therapeutic strategies to slow disease progression in humans. In this study, we identified hippocampal network hyperexcitability and generalized seizures that preceded neurodegeneration in the inducible rNLS8 mouse model. Local suppression of hippocampal hyperexcitability with chemogenetics (hM4Di) improved behavioral function but did not reduce neuron loss. Systemic antiseizure medications had no beneficial effects. These results highlight the complexity of TDP-43-induced excitability changes but ultimately suggest that directly targeting hyperexcitability may not be therapeutically effective.
Introduction
Neuronal hyperexcitability is a physiological hallmark of amyotrophic lateral sclerosis (ALS) yet its relationship to the underlying TAR DNA-binding protein 43 (TDP-43) pathology and neurodegeneration remains poorly understood. Electrophysiological studies using transcranial magnetic stimulation (TMS) have consistently identified neuronal hyperexcitability in ALS patient cohorts, both in the brain (Vucic et al., 2011) and in the periphery (Vucic and Kiernan, 2006a). Cortical TMS studies have shown ALS patients to present with increased motor evoked potential amplitude, a reduced cortical silent period, increased intracortical facilitation, and reduced short-interval intracortical inhibition (Vucic et al., 2011). Hyperexcitability appears early in disease (Vucic and Kiernan, 2006b) and correlates with cognitive impairment (Agarwal et al., 2021; Higashihara et al., 2021). Indeed, cortical hyperexcitability may be a key distinguishing feature of ALS versus mimicking neuromuscular disorders (Vucic et al., 2011; Menon et al., 2015). Early studies have implicated glutamate excitotoxicity as a proximate cause of neuron death (Roisen et al., 1982; Plaitakis and Caroscio, 1987; Rothstein et al., 1992; Rothstein, 1995) and the antiglutamatergic agent riluzole became the first approved ALS therapeutic (Lacomblez et al., 1996). Excitotoxicity remains central to the so-called “Dying Forward” hypothesis, which posits that excitotoxic input from upper motor neurons in cortical layer V promotes cell death of lower motor neurons in the spinal cord (Odierna et al., 2024).
TDP-43 is predominantly nuclear, while phosphorylated cytoplasmic aggregates, with accompanying nuclear depletion, is the primary pathology found in over 90% of ALS cases (Neumann et al., 2006). Toxicity resulting from TDP-43 pathology is believed to arise from a combination of gain-of-toxic function and loss-of-normal function effects (Chen-Plotkin et al., 2010). Recent reports in mouse and cell models suggest TDP-43 dysfunction induces hyperexcitability. Patch-clamp experiments in acute slice cultures have demonstrated intrinsic increases in excitability among layer V pyramidal neurons in mutant TDP-43 (Q331K; Fogarty et al., 2016) and CamKIIa-ΔNLS4 (Dyer et al., 2021a) mice. Additionally, in separate studies, layer V motor neurons from both the CamKIIa-ΔNLS4 bigenic (Dyer et al., 2021b) and TDP-43 (A315T) transgenic (Handley et al., 2017) mice experience widespread loss of dendritic spines consistent with sustained hyperactivity (Jiang et al., 1998).
Although evidence indicates TDP-43 dysfunction can affect neuronal activity, these alterations are complex, and it is unclear whether they lead directly to neurodegeneration. In a 2013 paper, astrocyte-specific expression of mutant TDP-43 (M337V) lead to downregulation of the glutamate transporter, GLT-1/EAAT2, that in turn promoted excitotoxicity in neurons (Tong et al., 2013). Additionally, in CamKIIa-ΔNLS4 mice, it has been reported that layer V motor neuron hyperexcitability precedes synaptic remodeling and cell death of lower motor neurons (Reale et al., 2023). However, these changes were accompanied by potentially compensatory increases in signaling from inhibitory circuits. Additionally, patch-clamp recordings in those same layer V neurons showed decreases in excitatory output, even as intrinsic excitability increased (Dyer et al., 2023). Thus, while multiple reports suggest TDP-43 can induce hyperexcitability, this may not necessarily drive excitotoxic neurodegeneration.
Following a behavioral study to characterize early cognitive changes in rNLS8 mice, we performed continuous electroencephalography (EEG) with epidural screw electrodes. We observed complex changes in EEG activity throughout the disease course and identified the presence of repetitive generalized seizures. Western blot analysis revealed broad dysregulation of AMPAR subunit expression in the off-DOX rNLS8 hippocampus. Using multielectrode electrophysiology in acute hippocampal slices, we found that mossy fiber hyperexcitability preceded neurodegeneration of the CA3 subfield. To determine whether hyperexcitability promoted neurodegeneration, we treated rNLS8 mice with chemogenetics and antiseizure medications (ASMs). Interestingly, we observed a rescue of deficits on the elevated zero maze (EZM) with AAV encoded inhibitory Designer Receptors Exclusively Activated by Designer Drugs (DREADDs). However, no reduction in neurodegeneration was observed with either ASMs or chemogenetics.
Materials and Methods
Animal husbandry and breeding
rNLS8 bigenic mice were generated by breeding tetO-hTDP-43ΔNLS4 (+/−) and tTA-NEFH8 (+/− or +/+) transgenic mice as previously described (Walker et al., 2015) and maintained on doxycycline containing chow (“on-DOX”; 200 mg/kg, Bio-Serv) until ∼2–6 months of age, when hTDP-43ΔNLS expression was induced by replacing doxycycline feed with standard rodent chow (“off-DOX”). Their diet was supplemented with moistened chow pellets and DietGel76A (ClearH2O) as the motor phenotype progressed. Nonbigenic littermates and on-DOX rNLS8 were used as controls, as indicated. The motor dysfunction phenotype and time course of progression were consistent with reports from when the line was first generated (Walker et al., 2015), which suggests a stable phenotype. All animal procedures were conducted with approval by the University of Pennsylvania Institutional Animal Care and Use Committee. Approximately equal number males and females were used for all experiments. A complete list of animals used in these studies is included in Table S1.
Epidural electrode implantation
On-DOX rNLS8 mice were anesthetized with isoflurane (4% induction, 1–2% maintenance) and subcutaneously injected with meloxicam-ER (6 mg/kg) and bupivacaine (3.75 mg/kg) to provide analgesia. Sterile instruments and aseptic technique were used throughout the procedure. Mice were attached to a stereotaxic instrument and the scalp was resected extending from approximately the coronal to just behind the lambdoid skull sutures. Burr holes for recording electrodes were made with a handheld microdrill on the skull overlaying the hippocampus bilaterally or over the right hemisphere only (−0.14 cm anterior-posterior and ±0.12 cm medial-laterally from bregma). Two additional burr holes were created bilaterally behind the lambdoid sutures for ground and reference electrodes. Screw electrodes (00–96 × 1.6 mm, Protech International E363/96/1.6/SPC) were implanted at a depth of one-third to one-half the shaft of the screw and connected to a plastic pedestal cap (Protech International MS363). Ortho-Jet acrylic resin (Lang Dental) was applied to close the surgical site and cement the pedestal cap to the skull. The total weight of the electrodes, cement, and cap was <2 g. A subcutaneous meloxicam-ER (6 mg/kg) injection was administered once per day for an additional 3 d to provide postprocedure analgesia.
Electroencephalography and seizure analysis
rNLS8 mice, after at least 3 d recovery from electrode implantation, were connected to a tethered three-channel EEG/EMG system (Pinnacle Technology). EEG was sampled at 1 kHz with 16 bit resolution and a 100 Hz low-pass filter with Sirenia Pro EEG acquisition software (Pinnacle Technology). Synchronous video was captured at 10 frames per second. EEG recording was suspended weekly during cage changes but was otherwise continuous. Automated seizure detection was performed with Sirenia Pro seizure analysis software (Pinnacle Technology, v1.8) using a line-length algorithm. All EEG traces and corresponding videos of seizure-candidate events were reviewed to rule out motion artifacts and other false positives.
Power spectral density estimation
Select 24 h EEG epochs from rNLS8 mice that were monitored for at least 4 weeks off-DOX were chosen for power spectral density analysis (n = 11 seizing, n = 7 seizure-free mice). Only seizure-free epochs were included. A separate cohort of (n = 5) on-DOX rNLS8 mice were recorded continuously for over 4 weeks to provide a cross-sectional control to rule out iatrogenic effects on EEG from electrode implantation and chronic recording. For analysis, EEG files were converted to European data format (.edf) and analyzed in R (v4.4.1) using the Welch's power spectral density estimate function in the “gsignal” package (v0.3.7) with a Hamming window and 0.5 segment overlap. Area under the curve calculations were performed using the AUC function in the “DescTools” package (v0.99.57). Power in EEG frequency bands was calculated by summing the area under the curve calculations for the bands of interest: delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–80 Hz). Relative power was calculated by dividing the band power by the total power from 0 to 100 Hz.
Accelerating rotarod
To test motor coordination, mice were transferred to rotating arms of an accelerating rotarod (Model 7650, Ugo Basile) for three runs. (1) Habituation: 300 s at constant 4 rpm, mice that fell during this phase were returned to the rotarod. (2 and 3) Timed trial: latency to fall (seconds) was assessed during 4–40 rpm acceleration over 300 s. Mice that did not fall were assigned a latency to fall of 300 s. Mice were given 3 min rest between each trial. The longest latency to fall from the two timed trials was used for analysis.
Wire hang
In some studies, wire hang was used as a complement to the accelerating rotarod as a general measure of strength and motor function. Mice were transferred to a wire cage lid and suspended upside down above a clean cage (∼30–60 cm). The longest latency to fall from two trials was used for analysis. Mice that remained suspended for the maximum time of 180 s during the first trial were not retested.
Open field test
Gross motor function and exploratory behavior were assessed with an open field test. Individual mice were transferred to an arena (40 × 40 cm) for 15 min of free exploration (Gould et al., 2009). Movement (distance traveled, meters) was automatically tracked by overhead camera controlled by EthoVision software (Noldus).
Elevated zero maze
Fear and anxiety behaviors were examined with the EZM, which consists of an elevated circular platform with two walled (enclosed) quadrants on opposite sides adjacent to two open quadrants (Shepherd et al., 1994). Mice were transferred to an enclosed quadrant for 5 min of free exploration while being recorded by a video camera from the above. Transitions were recorded when both the front and hindlimbs crossed into a new quadrant. Percent time in open quadrants was defined as the total number of seconds in the open quadrants divided by the total test time (300 s). Mice that fell or jumped off of the apparatus during testing were excluded from analysis.
Y maze
Spatial working memory was assessed with the Y maze, an apparatus consisting of three walled arms (“A,” “B,” “C”) that form a “Y” shape (Kraeuter et al., 2019). Mice were transferred to the distal portion of an arm, facing the center and allowed to explore for 5 min while behavior was recorded from an overhead camera. Both front and hindlimbs needed to cross the arm boundary for an arm entry to be recorded. Alternations were recorded when the mouse entered all three arms (e.g., “ABC,” “ACB,” “BCA,” “BAC,” “CAB,” “CBA”; % Spontaneous Alternations = Number of Alternations / (Number of Arm Entries − 2) × 100).
Social interaction test
Mice were transferred to the center chamber of a three-chambered apparatus while recorded by a video camera from above. Before the start of testing, identical plexiglass cylinders with numerous holes for air exchange were placed into the center of each of the end chambers. During habituation (Phase I, 10 min), mice explored the apparatus with empty cylinders. In the social choice phase (Phase II, 10 min), one plexiglass cylinder was loaded with a novel same-sex stimulus mouse, while the plexiglass cylinder in the opposite end-chamber was loaded an inanimate object (e.g., paperweight). For the final, direct interaction phase (Phase III, 5 min), the plexiglass cylinders were removed, allowing free interaction between the mice. Sniffing time was defined as the number of seconds the test mouse sniffed the plexiglass cylinders containing the stimulus mouse (Phase II) or sniffed/groomed the stimulus mouse directly (Phase 3). Only interactions initiated by the test mouse were scored.
Intracerebral AAV injection
Adult mice were anaesthetized with an intraperitoneal (i.p.) injection of ketamine (60–100 mg/kg)–xylazine (8–12 mg/kg)–acepromazine (0.5–2 mg/kg) and immobilized on a robotic stereotaxic frame (NeuroStar). AAV9 encoding hM4D(Gi)-mCherry (Addgene, 50475) or mCherry control (Addgene, 114472) under the human synapsin 1 promoter (hSyn) was injected into the unilateral hippocampus (−1.94 mm anterior-posterior, +1.80 mm medial-lateral, −2.14 mm dorsal-ventral) with a 33 gauge Hamilton syringe (2.5 µl volume, ∼1.15 × 1010 genome copies total). After AAV injection, the surgical area was closed with sutures and mice were monitored for recovery.
DREADD agonist eyedrop administration
Two weeks after AAV injection, mice were taken off-DOX and began receiving daily administration of the DREADD agonist, clozapine-N-oxide (CNO) dihydrochloride (Hello Bio) for 6 weeks. CNO was diluted in sterile water to 20 mM and applied via bilateral eyedrops (Keenan et al., 2017), using a P10 pipette, to achieve a total dose of 1 mg/kg (∼1–2 µl per eye). Eyedrops were administered by touching eye with a droplet formed at the end of the pipette tip. The pipette tip itself did not contact the eye. If the droplet was dislodged before reaching the eye, the eyedrop was reapplied. Mice were weighed weekly to maintain proper dosage.
Valproic acid administration
As a pharmacological alternative to the DREADDs, a subgroup of rNLS8 mice were taken off-DOX and simultaneously administered valproic acid in their drinking water for a duration of 6 weeks. Valproic acid sodium salt (Hello Bio, HB0867) was diluted in drinking water (∼1–2 mg/ml) to achieve an effective dose of ∼300 mg·kg−1 [mean weight of the mice in the cage (kg) × 300 mg·kg−1/assumed 5 ml daily water consumption]. Mice were weighed weekly to maintain the appropriate dose. The drinking water–valproic solution was freshly prepared daily. Daily dosage was selected based on a previous study to achieve effective anticonvulsant activity (Ohdo et al., 1989).
Levetiracetam administration
rNLS8 mice were administered ∼100 mg/kg/day levetiracetam (Selleck Chemicals, S1356) in 0.5% methylcellulose and 0.1% Tween-80 by oral gavage (200 µl males; 130–150 µl females; calculated from the mean weight of males and females) from 1 to 6 weeks after transgene induction (off-DOX). Gavage solutions were prepared weekly from a 400 mg/ml levetiracetam stock solution (stored in aliquots at −20°C). Control animals were administered with an equal volume of gavage vehicle alone. Daily dosage was based on a previous report to achieve anticonvulsant activity (Smucny et al., 2015).
Mouse tissue collection
Mice were lethally anaesthetized with a cocktail of ketamine–xylazine–acepromazine and transcardially perfused with PBS and 10% neutral buffered formalin (NBF) or PBS only. Brains were immersion fixed in 10% NBF overnight at 4°C and then rinsed in Tris leaching buffer (50 mM Tris, 150 mM NaCl, pH 8.0). For Western blot analysis, the mice were perfused with PBS only and the brains dissected fresh; one hemisphere was processed for protein extractions while the other hemisphere was immersion fixed in 10% NBF overnight. Postfixed brains were grossly sectioned in the coronal plane with a mouse brain matrix and processed for paraffin embedding. Paraffin-embedded tissue blocks were cut on a rotary microtome into 6-µm-thick sections, mounted on StarFrost adhesive slides, and stored at room temperature until use.
Protein extraction and Western blot analysis
In a subset of mice, the cortex and intact hippocampus from one brain hemisphere were dissected, weighed, and snap frozen on dry ice. The tissue was Dounce homogenized and sonicated in freshly prepared radioimmunoprecipitation assay (RIPA) buffer with 1 mM PMSF and phosphatase and protease inhibitors. Samples were centrifuged at ∼100,000 × g for 30 min. The resulting supernatant was collected as the RIPA soluble fraction. A commercial bicinchoninic acid (BCA; Pierce) was performed to determine protein concentration. RIPA soluble lysates were diluted to 1 mg/ml in Laemmli buffer with DTT and boiled for 10 min. Precision Plus Protein Dual Color Standards (Bio-Rad) and protein samples (20 µg) were loaded into handcast polyacrylamide gels and separated by electrophoresis (∼80 V for 30 min followed by 120 V for 1.5 h) in Tris-glycine-SDS buffer. Afterward, proteins were wet transferred onto nitrocellulose membranes with 0.45 µm pores (∼1.15 h at 100 V or overnight at 20 V at 4°C). Total protein content was visualized with Ponceau S solution. Membranes were washed thoroughly with Tris-buffered saline (TBS) with Tween 20 (TBS-T) followed by TBS and incubated with Rockland blocking buffer (MB-070) diluted 1:1 with TBS for 1 h. Primary antibodies were diluted in blocking buffer and added to membranes for overnight incubation at 4°C. The antibodies used in this study include the following: mouse anti-hTDP-43 (clone 241, in house) 1:5,000 + rabbit anti-GAPDH (10494-1-AP, Proteintech) 1:20,000; mouse anti-GluA1 (67642-1-Ig, Proteintech) 1:1,000 + rabbit anti-GAPDH 1:20,000; rabbit anti-GluA2 (AB1768, Millipore) 1:2,000 + mouse anti-GAPDH (clone 6C5, AM4300, Fisher) 1:20,000; rabbit anti-GluA3 (29588-1-AP, Proteintech) 1:1,000 + mouse anti-GAPDH 1:20,000; rabbit anti-GluA4 (23350-1-AP, Proteintech) 1:1,000 + mouse anti-GAPDH 1:20,000. The following day, membranes were washed thoroughly with TBS-T and TBS and incubated with blocking buffer for 1 h before incubation with secondary near-IR fluorescent protein conjugated antibodies (LI-COR, 1:20,000 in blocking buffer) in the dark for an additional 1 h at room temperature (RT). After secondary antibody incubation, membranes underwent additional washes and imaging on a LI-COR Odyssey scanner. Immunoreactivity was quantified with LI-COR Image Studio software.
Immunohistochemistry
Formalin-fixed paraffin-embedded tissue sections were deparaffinized in xylene, rehydrated in graded alcohols, and washed in 0.1 M Tris buffer, pH 7.6. Endogenous peroxides were quenched by incubating sections in 5% H2O2 in methanol for 30 min at room temperature, and antigen retrieval was performed by microwaving in citrate buffer (95°C for 15 min, room temperature for 50 min). Sections were blocked in 2% fetal bovine serum (FBS) in 0.1 M Tris buffer for 5 min and then incubated with primary antibody overnight at 4°C. On Day 2, sections were incubated at room temperature with biotinylated secondary antibody for 2 h, avidin and biotinylated horse radish peroxidase reagent (VectorLabs) for 1 h, and ImmPACT DAB (VectorLabs) for 5–10 min. Finally, sections were dehydrated in graded alcohols, cleared with xylene, and cover glass mounted with CytoSeal 60. For immunofluorescence, peroxidase quenching was excluded, and sections were mounted with Fluoromount-G with DAPI (SouthernBiotech) after incubation with a fluorescently tagged secondary antibody. Primary antibodies used in this study include the following: rabbit anti-NeuN (ABN78, Millipore) 1:1,000, mouse anti-NeuN (A60, Millipore) 1:1,000, rat anti-p(409/410)-TDP-43 (clone 1D3, CNDR) 1:300, rat anti-GFAP (clone 2.2.B10, CNDR) 1:2,000, rabbit anti-mCherry (26765-1-AP, Proteintech) 1:1,000, and rabbit anti-Iba1 (Wako, 019-19741) 1:1,000.
Brain slice preparation
Adult mice were deeply anesthetized with pentobarbital or a cocktail of ketamine, acepromazine, and xylazine and transcardially perfused with ice-cold dissection buffer (in mM: 11 sucrose, 1 MgCl2, 0.5 CaCl2, 26 NaHCO3, 10 glucose, 3 KCl, 1.25 NaH2PO4; 290–310 mmol/kg). Brains were quickly dissected in chilled oxygenated dissection buffer and one brain hemisphere was transferred to a vibratome (Leica VT100 s) to generate horizontal brain sections (300 µm). Slices were incubated in oxygenated artificial cerebrospinal fluid (ACSF; in mM: 1.2 MgSO4, 2 CaCl2, 26 NaHCO3, 10 glucose, 124 NaCl, 5 KCl, 1.25 NaH2PO4; 290–310 mmol/kg) for 50 min before recording (∼32°C for 30 min, then room temperature for 20 min).
Extracellular recordings
A multielectrode array (MEA) recording system (MED64, Alpha MED Scientific) was used to perform evoked field excitatory postsynaptic potential recordings (fEPSPs). The MEA probes had 64 planar electrodes in an 8 × 8 pattern with 150 µm interelectrode spacing (MED-PD5155) and were continuously perfused at room temperature with oxygenated ACSF containing 60 µM picrotoxin to block inhibitory GABAergic currents. Slices were transferred to the MEA chamber, and the hippocampus was aligned to the electrode array. Slices were anchored with a harp and allowed to recover for 20 min before recording. Hippocampal subfields were identified by anatomical landmarks under bright-field microscopy. Stimulation and analysis channels were selected by assessing maximal fEPSP response to single electrical pulse among electrodes in the region of interest. Input–output curves were captured for the mossy fiber→CA3 and CA3→CA1 circuits (5–100 µA, 5 µA steps, 30 s interstimulus interval). In some recordings, slices needed to be repositioned after CA3 recordings to align the CA1 region to the electrode array. Traces were analyzed with custom code in R to identify max fEPSP amplitude per pulse stimulus. Analysis was performed on the mean fEPSP amplitudes recorded from one to three tissue slices per mouse.
Experimental design and statistical analysis
Statistical analysis was performed with GraphPad Prism 10 using two-tailed t tests (α = 0.05), one-way ANOVA, two-way ANOVA, or mixed-effects models with post hoc multiple-comparisons test as indicated. The figures were prepared using Adobe Photoshop (ver. 25.11). The diagram in Figure 2a was prepared with Biorender.com.
Code accessibility
R code will be provided upon request.
Results
Cognitive and social behavioral deficits precede neurodegeneration in off-DOX rNLS8 mice
rNLS8 transgenic mice express full-length human TDP-43 with a defective nuclear localization signal (hTDP-43ΔNLS) under the control of the neurofilament heavy subunit promoter (Walker et al., 2015). Transgene expression is induced by replacing doxycycline chow with normal feed, resulting in cytoplasmic TDP-43 accumulation and subsequent depletion of endogenous nuclear TDP-43. Motor deficits begin ∼2 weeks off-DOX, and neurodegeneration, evidenced by decreased cortical thickness and neuromuscular junction denervation, begins ∼4 weeks off-DOX (Walker et al., 2015).
While previous work has done much to characterize motor dysfunction in the rNLS8 model, less is known about cognitive deficits. To characterize changes in cognitive function, particularly early in the disease before the onset of neurodegeneration, we subjected a cohort of rNLS8 mice (n = 8) to a weekly behavioral battery from an on-DOX baseline to 4 weeks off-DOX. As expected, off-DOX rNLS8 experienced weight loss (Fig. 1a) and progressive motor impairments (Fig. 1b) consistent with previous reports (Walker et al., 2015). At 4 weeks, off-DOX mice experienced a 68% reduction in performance on the rotarod assay versus on-DOX controls (75 vs 236 s mean latency to fall) and lost ∼22% of their baseline weight (vs +10% weight gain for on-DOX controls). Nevertheless, motor deficits did not grossly impair ambulation or negatively impact exploratory behavior in the open field assay (Fig. 1c), ruling out potential secondary effects on behavioral measures. Indeed, we observed a transient increase in exploratory behavior at the 3 week timepoint in off-DOX mice. Hyperlocomotor activity is a hallmark characteristic of CamKIIa-ΔNLS4 mice (Alfieri et al., 2014). These mice express the same mutant hTDP-43ΔNLS construct as the rNLS8 but under the brain-specific, Ca2+/calmodulin-dependent protein kinase IIa promoter, which largely spares them the severe motor deficits observed in the rNLS8 mouse line. It is possible that a hyperlocomotor phenotype in rNLS8 mice is masked by progressive, spinal cord-mediated, motor dysfunction in later disease stages.
Early motor and cognitive deficits in off-DOX rNLS8 mice. a, Weight change from on-DOX baseline (“Week 0”). b, Latency to fall in accelerating rotarod. c, Total exploration distance in a 15 min open field test. rNLS8 mice experienced progressive loss of motor coordination but retained the ability to ambulate to 4 weeks off-DOX. d, Percent time in open quadrants and total number of quadrant transitions on EZM. rNLS8 mice experienced profound deficits in anxiety–fear behaviors beginning at 3 weeks off-DOX. e, Percent spontaneous alternations and total number of arm entries on Y maze. A mild transient deficit was observed in spatial memory at 3 weeks off-DOX. Impaired sociability was observed on three-chambered social interaction test at 4 weeks as assessed by time sniffing during (f) social choice (restrained stimulus mouse) and (g) direct interaction (unrestrained stimulus mouse) phases. n = 8 each on-DOX control and off-DOX rNLS8 mice at 0, 1, 2, and 3 weeks; n = 6 off-DOX rNLS8 at 4 weeks and for social interaction test. Mean ± error bars (SEM). Mixed-effects model with Šidák multiple-comparisons test (time-matched on-DOX vs off-DOX) (a–e) and two-tailed t test (f, g). *p < 0.05; ***p < 0.01; ****p < 0.0001.
Additionally, we observed substantial deficits in anxiety-like behaviors on the EZM beginning at 3 weeks off-DOX (Fig. 1d), which suggested functional impairment in limbic structures including the amygdala and hippocampus (Silveira et al., 1993; File and Gonzalez, 1996; Walf and Frye, 2007). However, we observed only a minimal, transient, deficit on Y maze (Fig. 1e), a more straightforward assessment of spatial memory and hippocampal function.
Finally, as sociability deficits have been reported in CamKIIa-ΔNLS4 mice (Alfieri et al., 2014), we performed a three-chambered social interaction test at 4 weeks off-DOX. Similar to CamKIIa-NLS4 mice, off-DOX rNLS8 mice displayed reduced sociability with a novel same-sex mouse (Fig. 1f,g). As these early cognitive deficits substantially preceded disease timepoints associated with neurodegeneration, we hypothesized that TDP-43-induced alterations in network activity may be leading to impaired limbic function.
Longitudinal changes in EEG patterns and development of spontaneous seizures in off-DOX rNLS8 mice
To characterize global changes in neuron activity, we performed continuous EEG recording in 34 adult (3–8 months of age) rNLS8 mice, approximately equal numbers of males and females. Recording electrodes were placed either bilaterally (two-channel recording) or only over the right hemisphere (one-channel recording; Fig. 2a). Notably, we observed multiple changes in EEG measures of neural network function over the disease course (Fig. 2b–d). This includes an approximate 59% decrease in total EEG power at 4 weeks off-DOX compared with baseline accompanied by changes in the relative power across the frequency spectrum (0–80 Hz). At 4 weeks off-DOX, rNLS8 mice showed an increase in the relative EEG power of the delta (32%) and gamma (61%) frequency bands along with a decrease in power of the alpha (29%) and beta (32%) bands (Fig. 2d). To control for potential iatrogenic effects of electrode implantation and continuous recording, we implanted and continuously a separate cohort of rNLS8 mice (n = 5) maintained on-DOX (Fig. 2b,c). Notably, the power spectra after a median of 4.4 weeks (range, 2–5 weeks) continuous recoding resembled the on-DOX baseline of the longitudinal cohort, suggesting that the changes in EEG spectral density and power resulted from off-DOX hTDP-43ΔNLS expression rather than potential signal degradation from continuous recording.
Seizures in off-DOX rNLS8 mice. rNLS8 mice underwent continuous EEG monitoring including an on-DOX baseline and extending into the off-DOX disease time course (n = 34). a, Diagram highlighting screw electrode placement in rNLS8 mice. Recording electrodes were implanted above the hippocampus bilaterally or unilaterally on the right hemisphere. Reference electrodes were implanted bilaterally behind the lambdoid suture. b, Welch's power spectral density estimation and (c) total EEG power area under the curve of a 24 h EEG epoch from mice that were recorded for at least 4 weeks off-DOX. Epochs include an on-dox baseline, ∼2 (10–15 d) and 4 weeks (25–29 d) off-DOX (n = 18 mice). At 4 weeks off dox, there was a significant reduction in total EEG power. Power spectra from a separate cohort of continuously recorded on-DOX mice (“4wk on-DOX”; n = 5) resembled the on-DOX baseline from the longitudinal off-DOX cohort. d, Relative power of EEG frequency bands indicated a time-dependent modest increase in delta and gamma bands accompanied by decreased power in the alpha and beta bands. e, Representative seizure trace (calibration, 5 s, 300 µV), f, survival curve showing latency to seizure onset among all mice recorded (n = 34), and g, seizure frequency among mice that developed seizures and were monitored for at least 4 weeks off-DOX (n = 12). Mean ± error bars (SEM). One-way (c) and two-way (d) ANOVA with Dunnett's multiple-comparisons test (vs on-DOX Baseline). *p < 0.05; ***p < 0.01; ****p < 0.0001.
More strikingly, we observed repetitive generalized seizures beginning 2.5 weeks off-DOX (17.6 d median, 10.1 d min, 25.8 d max) in ∼64% of rNLS8 mice (Fig. 2e–g). All detected seizures featured a motor behavioral component, often consisting of limb myoclonus (Racine stage 3) or rearing, falling, and wild running (Racine stage 4–5; Movie 1; Racine, 1972). In two-channel recordings, seizures were equally apparent on both channels. Seizures lasted typically less than 1 min, and mice returned to normal behavior at the conclusion. Seizure frequency varied but, among mice that underwent at least 4 weeks of off-DOX EEG monitoring, typically consisted of multiple seizures per day (0.66–7.8 per day). Importantly, no seizures were detected during the on-DOX baseline recording period (2–21 d; mean, 8 d) prior to hTDP-43ΔNLS induction in the longitudinal cohort or at any time in the separate control cohort maintained on-DOX. Consequently, it is unlikely that seizures are of iatrogenic origin related to electrode implantation.
Seizure in an off-DOX rNLS8 mouse. [View online]
Interestingly, ∼36% of rNLS8 mice remained seizure-free. At 4 weeks off-DOX, EEG characteristics including total and relative band power were largely indistinguishable between seizing and nonseizing mice, except for a modest, nonstatistically significant increase in relative delta band power in rNLS8 mice with seizures (n = 10) compared with those that were seizure-free (n = 8; mean, 29.8% increase; p = 0.27; Fig. 3a–d). Off-DOX weight loss, hippocampal neuronal survival, motor impairment, and elevated zero and Y maze performance were comparable between seizure and seizure-free groups (Fig. 3e–j). Age was comparable between seizing (mean, 4.2 months) and nonseizing (mean, 4.9) mice. While the study was not powered to detect sex differences in epileptogenesis, ∼67% of females and 40% of males developed seizures within 4 weeks of transgene expression (p = 0.39 on Fisher's exact test).
Seizure-free rNLS8 mice are phenotypically similar to mice with seizures. a, Representative interictal traces from a mouse with seizures and a seizure-free rNLS8 mouse (calibration, 5 s, 300 µV). b, Welch's power spectral density estimation; c, total EEG power area under the curve; and d, relative power of EEG frequency bands of a 24 h EEG epoch at 4 weeks (25–29 d) off-DOX were similar between seizing and nonseizing mice (n = 9 nonseizing and n = 10 seizing mice). Additionally, no significant differences were observed in (e) dorsal hippocampus CA3 neuron survival or (f–j) behavioral deficits between seizing and nonseizing groups. Mean ± error bars (SEM). Mixed-effects model with Šidák multiple-comparisons test (seizing vs nonseizing) (d) and two-tailed t test (c, e–j). All comparisons were nonsignificant.
A previous report suggested TDP-43 dysfunction may alter α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) subunit composition in off-DOX rNLS8 mice (Wright et al., 2021). As these changes may promote hyperexcitability, we performed Western blot analysis of cortex and hippocampal RIPA soluble lysates from on-DOX and 4 week off-DOX (seizing and nonseizing) rNLS8 mice to evaluate AMPAR subunit expression, including GluA2, which regulates AMPAR calcium permeability (Geiger et al., 1995). We suspected that AMPAR subunit composition in seizure-free rNLS8 may be more normal (i.e., similar to on-DOX controls) than mice with seizures. In the cortex, AMPAR GluA1–3 subunits appeared similar among groups (Fig. 4a–d). However, we did find an approximate sevenfold increase in GluA4 expression among all 4 week off-DOX mice relative to on-DOX controls (Fig. 4e,f). In addition, we observed no difference in soluble hTDP-43 levels between seizure-free and seizing off-DOX rNLS8 mice (Fig. 4g), ruling out heterogeneous transgene expression as the culprit. Unlike the cortex, we observe broad changes in AMPAR subunit expression in the hippocampus (Fig. 5a–f). Notably, we identified an approximate 46% decrease in GluA1, a 49% decrease in GluA2, and 47% decrease in GluA3 (Fig. 5b–d). Interestingly, as in cortex, GluA4 levels were markedly increased (Fig. 5e,f). No differences in hippocampal AMPAR subunit expression were observed between seizing and nonseizing mice. Nevertheless, altered AMPAR subunit expression within the hippocampus of off-DOX mice may play a pivotal role in driving seizure generation. To that point, it is possible that the nonseizing off-DOX mice developed localized, epileptiform discharges, or other excitability changes that evaded detection by epidural electrodes.
Cortical AMPAR subunit and hTDP-43 expression in off-DOX rNLS8 mice. RIPA soluble mouse cortex lysates from on-DOX (n = 3) and 4 week off-DOX rNLS8 mice (n = 8 seizing and n = 4 nonseizing mice). 20 µg protein was loaded per lane. a, Immunoblots and (b–g) quantification of signal intensity, normalized to GAPDH loading control for AMPAR subunits GluA1, GluA2, GluA3, and GluA4 along with hTDP-43 transgene and the GluA4:GluA2 ratio. Mean values for on-DOX mice were set at 1. Mean ± error bars (S.E.M.). One-way ANOVA with Tukey's multiple-comparisons test (b–g). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Comparisons between seizing and nonseizing mice were nonsignificant.
Hippocampal AMPAR subunit expression in off-DOX rNLS8 mice. RIPA soluble mouse hippocampus lysates from on-DOX (n = 3) and 4 week off-DOX rNLS8 mice (n = 8 seizing and n = 4 nonseizing mice). 20 µg protein was loaded per lane. Unlike the cortex, complex changes in AMPAR subunit expression were observed. a, Immunoblots and (b–f) quantification of signal intensity, normalized to GAPDH loading control for AMPAR subunits GluA1, GluA2, GluA3, and GluA4 along with the GluA4:GluA2 ratio. Mean values for on-DOX mice were set at 1. Mean ± error bars (SEM). One-way ANOVA with Tukey's multiple-comparisons test (b–f).*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Comparisons between seizing and nonseizing mice were nonsignificant.
Network hyperexcitability precedes hippocampal CA3 region-specific neurodegeneration
Given the observed in vivo excitability changes and broad changes in AMPAR subunit expression within the hippocampus, we preformed MEA electrophysiology in acute hippocampal slices to examine circuit-level excitability in off-DOX rNLS8 mice (Fig. 6). We evoked excitatory postsynaptic potentials (fEPSPs; 0–100 µA stimulation, in 5 µA steps) from nontransgenic littermate controls and 3.1–4.1 week off-DOX rNLS8 mice (n = 4) in CA3 and CA1 hippocampal subfields. To limit confounding effects by inhibitory GABAergic circuits, all experiments were performed in the presence of 60 µM picrotoxin. Interestingly, we observed a dramatic increase in excitability in the mossy fiber–CA3 circuit, while the Schaffer collateral–CA1 circuit appeared unaffected. Notably, we did not observe obvious differences in hTDP-43ΔNLS expression between the CA3 and CA1 subfields (Fig. 7a), suggesting that regional heterogeneity in transgene expression was not the cause of the difference in excitability.
Hyperexcitability in CA3 hippocampus of off-DOX rNLS8 mice. a–c, Evoked fEPSPs in CA1 from nTg littermates and 3.5–4.5 week off-DOX rNLS8 mice (n = 4 mice, 1–3 slices per mouse) showed similar amplitudes. d–f, However, off-DOX rNLS8 CA3 hippocampus displayed much greater excitability than nonbigenic (nTg) littermate controls. a, d, Representative fEPSP traces (nTg, blue; rNLS8, red) and micrographs of hippocampal slices (150 µm distance between electrodes; red square, stim electrode, green square, recording electrode). b, e, Input–output curves of fEPSP amplitude (5–100 µA stimulation, in 5 µA steps, 30 s interstim interval). c, f, Area under the curve analysis. Mean ± error bars (SEM). Two-tailed t test. ****p < 0.0001.
Selective death of CA3 neurons in the dorsal hippocampus between 4 and 6 weeks off-DOX in rNLS8 mice. a, Immunofluorescence of cytoplasmic hTDP-43 (antibody 5104, green; DAPI, blue) in a 2 week off-DOX rNLS8 mouse. Representative NeuN immunohistochemistry in (b) nTg control and (c) 6 week off-DOX rNLS8 hippocampus (arrows show CA3 thinning). NeuN-positive cell counts per section in the (d) dentate gyrus (DG), (e) CA3, and (f) CA1 subfields (n = 11 control, n = 2 on-DOX rNLS8, n = 9 nTg littermates); n = 9, ≤2wk off-DOX; n = 7, 4wk off-DOX; n = 8, 6wk off-DOX; n = 9, ≥8wk off-DOX). Mean ± error bars (SEM). Scale bars, 200 µm. One-way ANOVA with Dunnett's multiple-comparisons test (vs Ctrl). **p < 0.01; ****p < 0.0001. Between 4 and 6 weeks, ∼47–56% of dorsal hippocampus CA3 neurons are lost in off-DOX rNLS8 mice. Mild degeneration is seen in the dentate gyrus at late timepoints while CA1 appears unaffected.
We next examined whether the observed hyperexcitability in the CA3 circuit early in disease may lead to enhanced neurodegeneration in later disease stages. To address this question, we performed detailed NeuN-positive cell counts of the dorsal hippocampus in coronal sections of off-DOX rNLS8 mice (Fig. 7b,c). Consistent with our hypothesis, we identified substantial neurodegeneration specific to the CA3/CA2 subfields between 6–8 weeks off-DOX (∼50% cell loss). Mild neuron loss was also detected in the dentate gyrus (DG) beginning at 8 weeks off-DOX. Interestingly, no cell loss was observed in the CA1 region, despite high expression of hTDP-43ΔNLS. These differences in neurodegeneration led us to suspect that excitotoxic mechanisms may be a major driver of cell death in this model.
Suppressing excitability with AAV encoded DREADDs reduces anxiety–fear behavioral deficits but not neurodegeneration in the rNLS8 hippocampus
To directly address whether hyperexcitability contributed to neurodegeneration in rNLS8 mice, we performed separate studies treating rNLS8 mice with inhibitory DREADDs (hM4Di-mCherry) chemogenetics localized to the hippocampus or systemically with ASMs. As a previous report in macaques indicated that that AAV encoded hM4Di receptors could attenuate bicuculline-induced discharges (Miyakawa et al., 2023), we hypothesized that DREADDs could likewise attenuate TDP-43-induced hyperexcitability in the hippocampus, which might rescue CA3 neuron death. To address this question, we performed unilateral hippocampal injection of AAV9 encoding hM4Di-mCherry or mCherry only under the human synapsin promoter in a cohort of on-DOX rNLS8 mice (Fig. 8). hTDP-43ΔNLS was induced 2 weeks after injection, while on-DOX mice served as controls. DREADDs receptors were activated daily with administration of the DREADD agonist CNO beginning at day of off-DOX and continuing for 6 weeks. Although we hypothesized that suppressing hippocampal excitability unilaterally may rescue local neurodegeneration, we did not expect to observe gross phenotypic changes. Consequently, we treated an additional cohort of rNLS8 mice with the ASM, valproic acid (VPA), dissolved in the drinking water. In SOD1 (G93A) mice, VPA administration in drinking water was shown to prolong survival (Sugai et al., 2004); thus, we expected VPA to provide a therapeutic benefit in rNLS8 mice.
Hippocampal injected AAV-DREADDs (hM4Di-mCherry) improved fear and anxiety deficits without affecting neurodegeneration. Representative immunohistochemistry for mCherry in (a) mCherry vector controls and (b) inhibitory DREADD injected rNLS8 hippocampus at 6 week off-DOX (mCherry, red; DAPI, blue). c, DREADDs injection resulted in robust expression along the mossy fiber pathway in the dorsal hippocampus (MFI, mean fluorescence intensity). d, No reduction in neurodegeneration was observed in counts of NeuN-positive neurons in CA3 of the dorsal hippocampus (n = 5, ON-DOX; n = 8 mCherry; n = 9 hM4Di; n = 7 VPA). e, However, DREADD injected mice showed dramatic improvements in the EZM (n = 5, ON-DOX; n = 4 mCherry; n = 6 hM4Di; n = 5 VPA). Motor performance in (f) EZM quadrant transitions, (g) accelerating rotarod or (h) wire hang, as well as (i) weight loss were similar across all off-DOX groups. Mean ± error bars (SEM). One-way (c–h) and two-way (i) ANOVA with Tukey's multiple-comparisons test. **p < 0.01; ****p < 0.0001. Each group was compared with every other group; nonsignificant comparisons are not shown.
Contrary to our hypothesis, histological analysis at 6 weeks off-DOX revealed no improvement in neuron survival in the vulnerable hippocampal CA3 field in the DREADDs or VPA treated groups (Fig. 8d). However, at 5.8 weeks off-DOX, we observed a dramatic rescue of anxiety–fear behaviors on the EZM in the DREADDs cohort but not in VPA treated mice (Fig. 8e). Given the dramatic behavioral rescue with DREADDs, we suspected that insufficient suppression of hyperexcitability may have been responsible for the lack of improvements in neuron survival.
Consequently, we performed a second study using an ASM with a different mechanism of action than VPA. Here, we administered levetiracetam (LEV) or methylcellulose vehicle by oral gavage to rNLS8 mice daily from 1 to 6 weeks off-DOX (Fig. 9). Dosing was started at 1 week off-DOX rather than at the day of off-DOX to reduce the total number of gavages, which may stress the mice. Behavioral tests were performed at 4 weeks off-DOX and histology at 6 weeks. Previously, LEV was shown to reduce cognitive dysfunction in APP23/MAPT mice (Zheng et al., 2022) and reduce aberrant network activity in hAPPJ20 mice (Sanchez et al., 2012) when administered at comparable dosing. To enhance our ability to detect therapeutic effects in the rNLS8 disease course, we included a social interaction test, where we showed previously that rNLS8 mice have deficits, and additional immunostaining for markers of astrogliosis and microgliosis. However, similar to VPA treatment, we observed no reduction in neurodegeneration in the CA3 subfield of the dorsal hippocampus or amelioration of behavioral deficits in LEV-treated mice (Fig. 9a–c,h–m). Weight loss and secondary immunohistochemical markers of pathology and glial reactivity in the hippocampus including phosphorylated TDP-43 (409/410), GFAP, and Iba1 were likewise similar between vehicle and LEV mice (Fig. 9d–g). Unexpectedly, we observed a mild reduction in sociability in LEV mice in the three-chambered social interaction test (Fig. 9j,k). Although, to our knowledge, impairments in sociability have not been shown in mice administered LEV, mood alterations including irritability and fatigue are common side effects of LEV treatment in humans, which could have contributed to reduced sociability in our study.
ASM therapy in rNLS8 mice with levetiracetam did not reduce neurodegeneration at 6 weeks off-DOX or improve behavioral deficits at 4 weeks off-DOX. Representative NeuN immunohistochemistry in 6wk off-DOX (a) vehicle control (VEH; n = 8) and (b) levetiracetam (LEV; n = 7) treated mice. Arrows point to CA3 subfield neuron loss. VEH- and LEV-treated mice did not differ in (c) CA3 cell counts, (d–f) area occupied analysis of secondary histological markers in the dorsal hippocampus, (g) off-DOX weight loss, and (h–m) 4 week off-DOX behavior assays. Unexpectedly we observed mild reduced sociability in LEV-treated mice in both (j) social choice (restrained stimulus mouse) and (k) direct interaction (unrestrained mouse) phases. No other differences among groups were observed. Mean ± error bars (SEM). Two-tailed t tests (c–i, k–m) and (j) mixed-effects model with Šidák multiple-comparisons test (VEH vs LEV). *p < 0.05.
Discussion
Hyperexcitability is a defining feature of ALS but its relationship to TDP-43 pathology and neurodegeneration are unclear. Here, we found that ∼64% of rNLS8 mice develop generalized seizures after hTDP-43ΔNLS transgene induction. While our initial behavioral study revealed impairments in fear and anxiety behaviors in off-DOX versus on-DOX rNLS8 mice, we did not identify any significant phenotypic difference between mice with seizures and those that remained seizure-free.
The molecular basis for hyperexcitability in ALS is not fully understood but likely involves a complex dysregulation of glutamatergic neurotransmission that increases susceptibility to excitotoxicity. Postmortem studies with ALS patients have identified depletion of the astrocytic glutamate transporter, EAAT2 (Rothstein et al., 1995), in spinal cord and motor cortex, along with decreased neuronal expression of the glutamate ionotropic receptor AMPA type subunit 2 (GluA2), which regulates AMPA receptor calcium permeability, in spinal cord motoneurons (Kawahara et al., 2004). Interestingly, EAAT2 and AMPAR dysregulation have been identified in SOD-1 (Howland et al., 2002; Van Damme et al., 2005; Tortarolo et al., 2006), c9ORF72 (Selvaraj et al., 2018), FUS (Kia et al., 2018), and TDP-43 (Polymenidou et al., 2011; Tong et al., 2013; Jiang et al., 2019; Wright et al., 2021; Dyer et al., 2021a) ALS experimental models, which suggest excitotoxicity, may represent a common disease mechanism. Evidence that TDP-43 can alter AMPAR function is strengthened by previous studies indicating TDP-43 regulates the RNAs encoding the AMPAR subunits: GluA2, GluA3, and GluA4 (Sephton et al., 2011). In our study we observed broad reductions in GluA2, GluA3, and GluA1 in the off-DOX rNLS8 hippocampus but not in the cortex. These changes were accompanied by a selective increase in GluA4 expression in both hippocampus and cortex. AMPAR subunit expression was similar between seizing and nonseizing mice. Nevertheless, it is possible AMPAR dysregulation contributed to seizure generation. Notably, the power spectra from both nonseizing and seizing mice were similar and distinct from on-DOX mice, suggesting nonseizing mice were not spared global alterations in network activity. Additionally, nonseizing mice may have developed localized hippocampal epileptiform discharges that were not detected by epidural electrodes.
At the circuit level, MEA electrophysiology in acute hippocampal sections revealed substantial hyperexcitability of the mossy fiber pathway preceding profound neurodegeneration in the CA3 subfield in off-DOX rNLS8 mice. Despite high cytoplasmic expression of hTDP-43ΔNLS, neurons in CA1 appeared relatively unaffected. Unlike the CamKIIa-NLS4 mice, we did not observe more than mild degeneration of the dentate gyrus. As the dentate gyrus was relatively preserved, we hypothesized that excessive excitatory input from mossy fibers may promote CA3 neurodegeneration via glutamate excitotoxicity, similar to what has been in reported in CamKIIa-ΔNLS4 mice, where corticomotor neuron hyperexcitability preceded degeneration of lower motor neurons (Dyer et al., 2021a).
To test whether hyperexcitability promoted neurodegeneration of CA3 pyramidal neurons, we used three approaches to suppress hippocampal hyperexcitability. First, we injected AAV encoded DREADDs into the hippocampus, which we activated with daily eyedrop administration of CNO. Interestingly, we observed a rescue of anxiety deficits on the EZM but no corresponding reduction in neuron loss. In a separate cohort, we administered the commonly used ASM, valproic acid, in the drinking water. The valproic acid had no measurable effect on neurodegeneration or behavioral phenotype. As disease progression may alter water consumption in rNLS8 mice, it is possible that the dose the mice received varied unexpectedly. To address this issue, we performed a final study where we administered the ASM, levetiracetam, by oral gavage. Like valproic acid treatment, this had no therapeutic effect on neurodegeneration or behavioral deficits.
Ultimately, results from this study suggest that directly targeting excess excitability with pharmacology may not be an effective therapeutic approach in TDP-43 proteinopathies. However, our behavioral rescue with the inhibitory DREADDs suggests that it may have the potential to address symptoms. While we cannot rule out that more effective suppression of hyperexcitability would provide neuroprotection, our results are consistent with a previous report showing that the glutamate release inhibitor, riluzole (8 mg/kg/day), also did not reduce disease burden in rNLS8 mice (Wright et al., 2021). More importantly, despite the Kv7 activator, ezogabine, providing some improvement in physiological readouts of hyperexcitability (Wainger et al., 2021), ASMs have largely failed to demonstrate therapeutic benefit for ALS patients (Eisen et al., 1993; Miller et al., 2001; Cudkowicz et al., 2003; Piepers et al., 2009; Aizawa et al., 2022; Boll et al., 2025). This is not completely surprising as emerging reports suggest TDP-43-induced excitability changes are highly complex and involve potential intrinsic and extrinsic compensatory mechanisms including alterations in excitatory output (Dyer et al., 2023), increased inhibitory signaling (Reale et al., 2023), and potentially novel protective effects of microglia (Xie et al., 2024) or other non-neuronal cell types. Our findings add to this body of literature, highlighting the need for a better understanding of the cellular and molecular processes underlying TDP-43 proteinopathies, including ALS. This includes whether changes in neuronal excitability contribute to TDP-43 disease progression and how best to address those changes therapeutically.
Footnotes
This work was supported by National Institute on Aging, 5T32AG000255, R01AG077692; National Institute of Neurological Disorders and Stroke, 5R01NS110688, R01NS101156, R37NS115439; and Seed Grant 2023-2024 from the American Epilepsy Society.
The authors declare no competing financial interests.
This paper contains supplemental material available at: https://doi.org/10.1523/JNEUROSCI.2297-24.2025
- Correspondence should be addressed to Virginia M. Y. Lee at vmylee{at}upenn.edu.
