Introduction

Amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) are neurodegenerative diseases that share multiple pathological features, particularly the presence of intracellular protein inclusions. Of proteins contributing to disease, ubiquitinated, phosphorylated, and biochemically-insoluble forms of TAR DNA binding protein 43 (TDP-43) lead to neurodegeneration in >95% of both sporadic and familial ALS cases and about 45% of FTD cases [1, 2]. TDP-43 is an important RNA-binding protein that typically shuttles between the nucleus and cytoplasm under physiological conditions, but TDP-43 accumulation in the cytoplasm of neurons results in neurodegeneration [3, 4]. The pathways leading to neuronal demise caused by cytoplasmic TDP-43 pathology remain to be fully defined [5], and understanding the mechanisms that cause neuron death may reveal new therapeutic opportunities.

Many pathogenic mechanisms have been proposed to contribute to neurodegeneration in ALS, including the integrated stress response (ISR) [6], protein synthesis dysregulation [7], oxidative stress [6], and neuroinflammation [8, 9]. Indeed, chronic activation of the ISR likely contributes to motor neuron death in ALS and FTD [10, 11]. The ISR is a master regulatory pathway that fine-tunes proteostasis in cells, controlling protein synthesis, folding, and decay processes, and is crucial for cell survival [11]. Under diverse cellular stress conditions, ISR signaling upregulates select downstream transcripts to restore protein homeostasis. Activating transcription factor 4 (ATF4) is a crucial ISR regulator that targets genes to promote protein folding and degradation, increasing cell survival [12]. Nevertheless, prolonged disturbance of proteostasis enhances ISR signaling and consequently results in ATF4-mediated upregulation of CCAAT/enhancer-binding homologous protein (CHOP), which can trigger a pro-apoptotic signal cascade [11, 13]. ISR activation, evidenced by activation of ISR sensor kinases, has been reported in postmortem samples of ALS cases [6]. Moreover, activation of ISR enhances TDP-43 aggregation and stress granule formation, which are hallmarks of TDP-43-linked disease [14,15,16], suggesting a direct association of TDP-43 dysfunction with ISR activation. In addition, TDP-43 toxicity increases levels of CHOP protein in cell lines and sporadic ALS spinal cord tissues [15, 17]. Pharmacological modulation for ISR using small-molecule drugs can modify neurodegeneration [7], however, the actions of these drugs are not restricted to the ISR pathways. Therefore, it remains unclear whether specific modulation of ISR effectors can modulate disease in mammalian models of cytoplasmic TDP-43 proteinopathy.

Transgenic mice with doxycycline (Dox)-suppressible expression of human TDP-43 containing a defective nuclear localization signal (hTDP-43∆NLS) under the control of the neurofilament heavy chain promoter (rNLS8 mice) are one of the most disease-relevant models of ALS, displaying progressive TDP-43 cytoplasmic accumulation coinciding with rapid ALS-like neurodegeneration and motor decline [18, 19]. In this study, we, therefore, screened for genes that may regulate and respond to cellular stress pathways in the cortex of rNLS8 mice prior to disease onset and throughout early disease stages. We thereby identified alterations in expression of several genes across multiple molecular pathways, including upregulation of genes involved in the ISR, DNA damage response, apoptosis, and neuroinflammation, and downregulation of genes involved in glycolysis and ion exchange, in the rNLS8 mouse cortex. Remarkably, several genes involved in the ISR, DNA damage response, and apoptosis were dramatically increased even prior to disease onset. Notably, anti- and pro-apoptotic genes were up-regulated prior to disease onset, with a switch towards solely pro-apoptotic signaling at later disease stages in rNLS8 mice. Our results also revealed a significant elevation of cleaved caspase-3 levels in late-disease rNLS8 mice, supporting ISR-mediated apoptosis as a critical contributor to neurodegeneration in disease. To investigate whether amelioration of ISR signaling could protect against disease, we, therefore, performed intracerebroventricular (ICV) injection of antisense oligonucleotides (ASOs) targeting Chop [20] to suppress Chop expression in rNLS8 mouse central nervous system. ASO treatment decreased Chop levels but did not affect overall TDP-43 pathology, neurodegeneration, gliosis, or motor deficits in rNLS8 mice. Taken together, we conclude that TDP-43 mislocalisation leads to early perturbation of multiple molecular pathways that crosstalk with ISR signals, with dysregulation of apoptosis signaling contributing to disease in rNLS8 mice.

Materials and methods

Animals

To generate samples for RT2 qPCR arrays, rNLS8 transgenic mice and littermate controls on a mixed B6/C3H F1 background were produced as previously described [18]. Founder monogenic B6;C3-Tg(NEFH-tTA)8Vle/J (NEFH-tTA line 8, stock #025397) mice and monogenic B6;C3-Tg(tetO-TARDBP*)4Vle/J (tetO-hTDP-43ΔNLS line 4, stock #014650) mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA) [19]. Intercross breeder and experimental mice were fed with chow containing 200 mg/kg doxycycline (Dox) (Specialty Feeds, Australia). Experiments were conducted with approval from the Animal Ethics Committee of Macquarie University (#2016-026). For all other experiments, rNLS8 mice were produced from the intercross of homozygous tetO-hTDP-43ΔNLS line 4 mice with hemizygous NEFH-tTA line 8 mice both on a pure C56BL/6JAusb background following >10 generations of backcrossing and were fed with Dox-containing chow (200 mg/kg, Specialty Feeds, Australia) [21]. Experiments were conducted with approval from the Animal Ethics Committee of The University of Queensland (#QBI/131/18). During all experiments, mice were switched to normal chow to induce expression of hTDP-43∆NLS. Male and female mice at approximately ten weeks of age were housed in temperature- and humidity-controlled conditions (21 ± 1 °C, 55 ± 5%) with a 12 h light/dark cycle (lights on at 06:00 h). Mice were randomly allocated to groups for all experiments of time point analyses and ASO administration groups, the size of which were calculated by G*Power (version 3.1). Both sexes were included and balanced between groups. Littermate non-transgenic and monogenic animals were used as controls. One mouse was excluded from all analyses due to non-neurological disease (malocclusion). Three mice in the ASO experiments were excluded from the qPCR and histological analyses since they reached the humane end-point prior to the predefined time point for tissue collection of 6 weeks off Dox. All experimental procedures were conducted under the guidelines of the National Health and Medical Research Council of Australia in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.

Surgery procedures and ASO administration

Two Chop ASOs (known as Chop-ASO #3 and #5 respectively) were synthesised and purified as previously described [20, 22], with the ASO sequences shown in Table 1. For intracerebroventricular (ICV) injection, mice were anaesthetised by isoflurane (Abbott Laboratories) using an isoflurane machine (Kent Scientific Co. Torrington, CT, USAs) at the rate of 2–4% according to body weights of mice. Bilateral ICV injections were conducted stereotaxically using a 30 G needle connected to a Hamilton syringe by polyethylene tubing into lateral ventricles: anteroposterior = 0; lateral = ±  0.9; depth = 2.3 (coordinates are in millimetres relative to the Bregma) [23]. A total of 10 μL of ASO solution (50 mg/mL prepared in sterile saline solution) or saline solution as vehicle were injected into each of the left and right ventricles over 2 min (5 μL in each ventricle). After injections, mice were housed in home cages for two weeks before the removal of the Dox diet to induce neuronal hTDP-43ΔNLS expression.

Table 1 Sequences and details of the Chop ASOs.

Mouse monitoring and behavioural tests

Mice were monitored and weighed three times per week after Dox feed was removed as previously described [18, 19]. Briefly, for observation of collapsing splay or clasping of hindlimbs, mice were suspended by the tail for >5 s. The failure to extend both hindlimbs was recorded as a positive response of collapsing splay and holding hindlimbs together was recorded as a presence of clasping splay. Mice were tested for rotarod performance of motor coordination and balance once per week, as previously described [19]. Briefly, mice were placed on a rotarod apparatus (Ugo Basile SRL, Gemonio, Italy) at a speed of 5 rpm with acceleration up to 40 rpm within 300 s. The time to fall was recorded. If mice were still running at the end of the testing session, their times were recorded as 300 s. Three training sessions were performed one week prior to time off Dox, and two test sessions were conducted weekly, with the final score being the highest time of the two test sessions. Experimenters were blinded to treatment group for hindlimb clasping and rotarod assessments.

RNA extraction

RNA was extracted from the homogenates of the rostral cortices (anteroposteriorly approximate 3.0 to −0.5 mm, Bregma) that contain the primary and secondary motor cortex or the lumbar spinal cord with Qiazol (Qiagen, #79306) using the Qiagen RNeasy Mini Kit (Qiagen, #74104) and Precellys tissue homogeniser (Bertin Instruments, Montigny-le-Bretonneux, France). On-column DNase I digestion was conducted using RNase-free DNase I (Qiagen, #79254). The concentration of extracted RNA was determined using a NanoRNA kit (Agilent, #5067-1511) and Bioanalyzer (Agilent 2100, Santa Clara, CS, USA). cDNA was synthesized from 1 μg total RNA using the SuperScript™ VILO™ Master Mix (Thermo Fisher, #11755050).

RT2 PCR arrays and data analysis

RT2 Profiler™ PCR Arrays (SABioscience Corporation, #PAMM-003Z) were used to detect differentially expressed cell stress and death genes of interest in the rostral cortices of control and rNLS8 mice at 2 and 4 weeks off Dox. The SYBR-Green based RT2 PCR arrays contained 84 genes plus 5 housekeeping genes and assay controls. cDNA was mixed with SYBR-green as per manufacturer instructions and loaded onto a 384 well plate, with each gene was assessed in quadruplicate. RT2 PCR arrays were performed on the Vii7 Real-Time PCR System (Applied Biosystems, Waltham, MA, USA). Fold-change analysis was performed using the ExpressionSuite Software (version 1.1), whereby each gene was determined by exponentiation of 2-ΔΔCt. A fold change of greater than 2 and p < 0.05 was deemed significant. The ß2 microglobulin (B2m) gene was excluded from the housekeeping genes in analysis as the expression of B2m gene was significantly increased in the cortex of rNLS8 mice off Dox compared to controls (Supplementary Fig. 1A), in line with previous findings [24].

For hierarchical clustering, fold change data of genes of interest was processed using the web-based Morpheus software (Broad Institute, Cambridge, MA, USA). Classification of differentially expressed genes was performed to indicate co-regulated and functionally related genes. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses was performed using Metascape (version 3.0) with default parameters (minimum overlap = 3, p-value cutoff = 0.01, and minimum enrichment = 1.5) [25]. Protein-protein interaction enrichment analysis was performed using the Molecular Complex Detection (MCODE) algorithm using Metascape to identify densely connected components from examined pathways with default parameters [26]. Venn diagrams were constructed using InteractiVenn [27]. Volcano plots were generated using Prism-GraphPad software (version 9).

Real-time quantitative PCR (qPCR) assay

Target gene real-time qPCR analyses were conducted using SYBR Green Master Mix (Bioline #BIO-98005) with the LightCycler® 480 System (Roche, Basel, Switzerland) to validate the RT2 PCR array results. The relative changes in gene expression were calculated based on the housekeeping gene ß-actin (Actb) according to the 2-∆∆CT method [28] and normalised to control or saline-treated control groups. The sequences of primers are available in Supplementary Table 1.

Protein extraction and immunoblotting (IB)

Mouse rostral cortices or lumbar spinal cords were collected for IB as described above and snap-frozen on dry ice. Tissues were then thawed on ice and homogenised in 5× v/w RIPA lysis buffer (50 mM Tris, 150 mM NaCl, 1% NP-40, 5 mM EDTA, 0.5% sodium deoxycholate, and 0.1% SDS, pH 8.0) containing 1 mM PMSF and protease and phosphatase inhibitor cocktails (Sigma, # 4906845001 and #11836170001) with three 1.4 mm Zirconium oxide beads (Bertin Instruments, #P000927-LYSK0-A) using Precellys tissue homogeniser (Bertin Instruments, #P000669-PR240-A). Samples were centrifuged at 4 °C, 100,000 g for 30 min, and the supernatant was taken as the RIPA-soluble fraction. The remaining pellet was washed with RIPA buffer as above, this supernatant was discarded, and the resulting pellet was dissolved in 5× v/w urea buffer (7 M urea, 2 M thiourea, 4% CHAPS, and 30 mM Tris, pH 8.5) using the Precellys homogeniser and centrifuged at 22 °C, 100,000 g for 30 min. This supernatant was taken as the RIPA-insoluble/urea-soluble fraction. Protein concentrations of the RIPA-soluble fractions were determined using the Pierce™ BCA Protein Assay Kit (ThermoFisher Scientific #23225).

Protein samples of RIPA-soluble or RIPA-insoluble/urea-soluble fractions were separated by electrophoresis (120 V for 90 min) on a polyacrylamide gel containing 12% acrylamide in the presence of reducing agent (2-mercaptoethanol, Sigma #63689). After SDS-PAGE, proteins were transferred to nitrocellulose membranes (LI-COR Biosciences #P/N926-31092) and incubated in blocking solution (5% (w/v) BSA, 0.05% (w/v) Tween-20 in TBS (TBST)) and then incubated overnight in primary antibody diluted in the blocking solution. Primary antibodies used for immunoblotting were rabbit anti-TDP-43 polyclonal antibody (PAb) for detecting both human and mouse TDP-43 (Proteintech #10782-2-AP, 1:2000), rat anti-phospho-S409/410 TDP-43 monoclonal antibody (MAb) (Biolegend # 829901, 1:1000), mouse anti-CHOP Mab (Invitrogen #MA1-250, 1:1000), a rabbit anti-caspase 3 for detection of cleaved forms of caspase-3 (Abcam #9664, 1:1000), mouse anti-GAPDH Mab (Proteintech #60004-1-Ig, 1:10,000), and rabbit anti-GAPDH PAb (Proteintech #10494-1-AP, 1:2000). The nitrocellulose membranes were washed with TBST and incubated with IRDye secondary antibodies (LI-COR Biosciences, 1:20,000) for 1 h. Protein bands were visualized using the Odyssey CLx Imaging System (LI-COR Biosciences, Lincoln, NE, USA) and quantified using Image Studio Lite software (LI-COR Biosciences, Lincoln, NE, USA). The relative changes of protein levels were calculated by quantifying the bands in RIPA soluble fractions relative to the band of the internal reference GAPDH protein, or by quantifying the bands in RIPA-insoluble/urea-soluble fractions relative to total protein and subsequently normalised to the control group or saline-treated control group.

Histology and immunofluorescence labeling (IF)

Mice were perfused using PBS followed by 4% paraformaldehyde (PFA). The rostral cortex and lumbar spinal cord tissues were rapidly dissected and post-fixed in 4% PFA for 1 h at room temperature and 4 °C overnight. Post-fixed tissues were rinsed in phosphate-buffered saline (PBS) pH 7.4, dehydrated with a series of increasing concentrations of ethanol, and then embedded in paraffin. Embedded tissues were sectioned at 10 μm thickness. For IF, paraffin-embedded sections were deparaffinised in xylene and rehydrated through a series of decreasing concentrations of ethanol. After rehydration, sections were subjected to antigen retrieval solution (10 mM sodium citrate, 0.05% Tween-20, pH 6.0) at 95 °C for 10 min and then were allowed to cool. Sections were then blocked in 1% BSA in PBS containing 0.25% (v/v) Triton X-100 (PBST) and incubated in primary antibodies overnight at 4 °C. Primary antibodies used for IF were rabbit anti-TDP-43 polyclonal antibody (PAb) for detecting pan-TDP-43 for both human and mouse TDP-43 (Proteintech #10782-2-AP, 1:1000), rat anti-glial fibrillary acidic protein (GFAP) Mab (ThermoFisher Scientific #13-0300, 1:500), and mouse anti-CHOP Mab (Invitrogen #MA1-250, 1:1000). Sections were then washed in PBST and incubated with fluorophore-conjugated secondary antibodies, including donkey anti-rat conjugated Alexa-488 (ThermoFisher Scientific #21208, 1:1000) and donkey anti-rabbit conjugated Alexa-647 (ThermoFisher Scientific #48272, 1:1000). Nuclei were stained using 4′,6-diamidino-2-phenylindole (DAPI, Sigma-Aldrich #D8417, 1:1000).

Fluorescence microscopy and Image analysis

Images for IF sections were acquired using an epifluorescence microscope Axio Imager (Zeiss) under x20 objective (0.8 NA / 0.55 mm WD), providing a pixel size of 0.323 μm using Zen imaging software (Zeiss). Images were background-corrected using ImageJ software with the rolling ball background subtraction algorithm and were then quantified for protein levels of TDP-43 using CellProfiler software (Broad Institute, version 4.2.1) as previously published [29].

Statistical analysis

For two-value data, statistical analyses were conducted using a two-tailed t-test. One-way ANOVA was used for analyses of datasets with three values with Bonferroni’s post hoc test using Prism-GraphPad software (version 9). Statistical significance is indicated as *p < 0.05, **p < 0.01, and ***p < 0.001. All data are presented as mean±standard error of the mean (SEM).

Results

Early activation of cell stress and death pathways in the cortex of rNLS8 mice

To determine how numerous cell stress and death pathways are regulated in the cortex of rNLS8 mice in early disease stages, we conducted RT2 PCR arrays to assess the expression of critical genes at time of approximate disease onset (2 weeks off Dox, 2 WOD) and an early disease stage (4 weeks off Dox, 4 WOD) (Fig. 1A, Supplementary Fig. 2). We assessed 84 mouse genes in ten functional groups, with four housekeeping genes (B2m excluded; Supplementary Fig. 1A) and three negative controls (the full gene expression dataset is available in Supplementary Table 2). Firstly, we confirmed the expected increases of human TARDBP (hTARDBP) mRNA levels in the cortex of rNLS8 mice from 1 WOD (Supplementary Fig. 1B), prior to assessment of cell stress gene changes, as well as at 2 and 4 WOD (Fig. 1B) relative to control mice, using real-time qPCR. Notably, the level of hTARDBP mRNA was apparently higher in the cortex (Fig. 1B, Supplementary Fig. 1D) than in the spinal cord (Supplementary Figs. 1C, 1E) of rNLS8 mice, consistent with our previous observation of higher hTDP-43ΔNLS protein levels and earlier accumulation of TDP-43 pathology in the cortex than in the spinal cord in this model [19]. In contrast, the mRNA levels of mouse Tardbp (mTardbp) were unchanged in rNLS8 mice compared to controls at 2 and 4 WOD (Fig. 1B), potentially due to the lack of sensitivity in this bulk tissue analysis to detect the expected auto-regulatory loss of mTardbp in neurons at these timepoints. However, as expected, mTardbp levels were decreased in the cortex of rNLS8 at 6 WOD (Supplementary Fig. 1D). These findings are in line with previous reports that overexpression of hTDP-43∆NLS decreases endogenous mTardbp mRNA and mTDP-43 protein levels via an auto-regulatory mechanism [19, 30, 31].

Fig. 1: Cellular stress and death signaling pathways are activated in the cortex of rNLS8 mice at early disease stages.
figure 1

A Experimental schema. B Real-time qPCR analysis of mouse Tardbp and human TARDBP gene expression in the cortex of rNLS8 mice at 2 and 4 weeks off Dox (WOD) relative to controls. n = 5. Mean ± SEM. **p < 0.01 and ***p < 0.001. C Total number of unique and shared statistically significantly upregulated or downregulated genes in rNLS8 mice at each time point, relative to the control mice. D Unsupervised hierarchical clustering heatmap for expression of all assayed genes in RT2 PCR array from rNLS8 and control mice at each timepoint, with values given as normalised gene expression levels (log2 fold changes). Statistically significantly (p < 0.05) upregulated and downregulated genes are indicated in red and blue, respectively. * indicates statistical significance for 2 WOD and # for 4 WOD. E Gene ontology enrichment analyses for all the genes of interest in RT2 PCR arrays. F MCODE network analyses for protein-protein interaction of identified differentially expressed genes in rNLS8 mice at 2 and 4 WOD, respectively. G Volcano plots indicate the magnitude (x axis, as log2 fold changes) and statistical significance (y axis, as -log10 p-values) of gene expression changes in rNLS8 mice relative to control mice at each timepoint. Upregulated genes (fold change >2) are shown in red and downregulated genes (fold change < -2) are shown in blue.

At 2 weeks off Dox, taken as disease onset, we identified a total of nine statistically significantly upregulated genes (fold change >2, p < 0.05) in the rNLS8 mice (Fig. 1C, the expression dataset of identified genes is available in Table 2). At 4 weeks off Dox, taken as an early disease stage after accumulation of TDP-43 pathology, we detected six upregulated genes and two downregulated genes (fold change < −2, p < 0.05) in the rNLS8 mice relative to controls (Fig. 1C). Among these dysregulated genes, we identified four genes that were upregulated at both 2-week and 4-week timepoints in the cortex of rNLS8 mice in the RT2 array, namely activating transcription factor 4 (Atf4), pro-apoptotic BH3 Interacting Domain Death Agonist (Bid), interleukin 6 (Il6), and tumor necrosis factor-alpha (Tnf). Consistent alteration in these pathways indicates involvement of these genes in disease over an extended period, suggesting that these pathways may be important for disease progression.

Table 2 Genes showing statistical significant difference with >2-fold change in cortex of rNLS8 mice at 2 weeks or 4 weeks off Dox compared to controls detected on RT2 PCR array.

The hierarchical heatmap showing the magnitude of fold change in expression between rNLS8 mice and controls in the RT2 array demonstrated a clear bi-clustering pattern of upregulated genes (in red) and downregulated genes (in blue) over the disease course (Fig. 1D). The GO enrichment analyses of all pre-selected genes revealed that the top three biological processes relate to ISR signaling: positive regulation of cell death (GO:0010942), cellular response to chemical stress (GO:0062197), and regulation of cellular response to stress (GO:0080135) (Fig. 1E). Furthermore, the molecule complex detection (MCODE) network for protein-protein interaction revealed significant interactions of identified genes among biological processes, notably the positive regulation of cell death via the ATF4-CHOP axis of the ISR in the disease-onset rNLS8 mice at 2 weeks off Dox (Fig. 1F).

To further interrogate the gene changes in rNLS8 mice, we grouped analysed genes based on biological pathway at both 2-week and 4-week timepoints. At 2 weeks off Dox, upregulated genes act within several pathways: the ISR, being Atf4 (~3.7 fold increased in rNLS8 mice) and C/EBP homologous protein (Chop, ~2.3 fold); apoptosis signaling, being Bid (~2.6 fold); oxidative stress response, being Heme Oxygenase 1 (Homx1, ~2.4 fold); DNA damage response, being DNA Damage Inducible Gamma (Gadd45g, ~2.7 fold) and transformation related protein 53 (Trp53, ~2.3 fold); neuroinflammation, being chemokine (CC motif) ligand 12 (Ccl12, ~15.5 fold) and Il6 (~6.7 fold), and; TNF family signaling, being Tnf (~7.4 fold) (Supplementary Fig. 2). The data suggest that TDP-43 cytoplasmic mislocalisation in neurons of rNLS8 mice results in disruption of cellular homeostasis including ISR activation, DNA damage, oxidative stress, and neuroinflammation as early as disease onset.

At 4 weeks off Dox, upregulated genes largely belong to the same biological pathways as at 2 weeks: the ISR, being Atf4 (~ 2.5 fold increased in rNLS8 mice); apoptosis signaling, being Bid (~2.5 fold); neuroinflammation, being the C-reactive protein gene (Crp, ~2.8 fold), Il6 (~ 5.5 fold), and the Toll-like receptor 4 gene (Tlr4, ~2.1 fold), and; TNF family ligand, being Tnf (~ 9.0 fold) (Supplementary Fig. 2). Furthermore, we also identified two downregulated genes in the cortex of rNLS8 mice at 4 weeks off Dox that were not altered at 2 weeks, involved in: cellular metabolism (glycolysis), being lactate dehydrogenase A (Ldha, ~ 0.5 fold), and; ion metabolism, being the solute carrier family 9 member A3 (Slc9a3, ~0.5 fold) (Supplementary Fig. 2). These results suggest there is a prolonged elevation of cell stress response during disease and later dysregulation in cellular metabolism in the rNLS8 mice.

To validate our findings in RT2 array analyses and determine whether changes occur even before disease onset, we conducted real-time qPCR for all 13 genes detected as significantly altered in at least one of the timepoints in the rNLS8 mice, in separate experiments using an independent set of mouse cortex samples. In addition to disease onset (2 weeks) and early disease (4 weeks) timepoints, we additionally analysed samples from mice prior to disease onset (1 week off Dox), a timepoint at which rNLS8 mice display increased cytoplasmic human TDP-43 protein levels but no overt motor phenotypes [19]. Intriguingly, five of the 11 upregulated genes originally identified as altered in array data were already upregulated at 1 week timepoint (Fig. 2A), including genes involved in: the ISR, being Atf4 (~1.5 fold) and Chop (~2.1 fold); apoptosis signaling, being Bid (~2.2 fold), and; DNA damage response, being Gadd45g (~1.4 fold) and Trp53 (~1.7 fold). Expression of these five genes were also consistently elevated at 2 weeks and 4 weeks off Dox relative to the control mice, consistent with the array findings. Moreover, qPCR results showed consistent up-regulation of selected genes in rNLS8 mice on both the original mixed B6/C3H background (Supplementary Fig. 3A), as employed in the RT2 array, and on pure C56BL/6JAusb background (Fig. 2A), as employed in the following experiments. Consistent upregulation of these genes suggests the early and prolonged activation of cellular stress pathways and pro-apoptosis signaling in the cortex of rNLS8 mice upon the induction of hTDP-43∆NLS expression, even before disease onset.

Fig. 2: rNLS8 mice display dysregulation of genes in multiple cellular stress pathways and activation of the ISR in the cortex even prior to disease onset.
figure 2

A Real-time qPCR verified upregulated RT2 array-identified genes (Atf4, Chop, Bid, Trp53 and Gadd45g) in the cortex of rNLS8 mice at 2 and 4 weeks off Dox (WOD), and revealed similar upregulation at 1 WOD. B Additional ISR-related genes (Ppp1r15a and spliced Xbp1) were also upregulated at 1 WOD. C The RT2 array-identified upregulated genes Ccl12, Il6, Tnf, Tlr4 and Homx1 were verified as increased at 2 and 4 WOD but were unaltered at 1 WOD. D Differential alterations were also identified in the RT2 array-identified downregulated genes (Lhda, Slc9a3 and Crp) at 1 and 4 WOD. Values were normalized to Actb as the housekeeping gene. E Immunoblotting for the proteins of eIF2α and p-eIF2α (S51) in the cortex of rNLS8 mice at 4 WOD. Approximate molecular weights (kDa) are indicated. Quantification of immunoblots showed increased eIF2α (F) and p-eIF2α (G) in the cortex of rNLS8 mice relative to control mice, respectively. H Immunoblotting for ATF4 protein in the cortex of rNLS8 mice at 4 WOD. I Quantification of immunoblots showed increased ATF4 protein in the cortex of rNLS8 mice relative to control mice. n = 5. Mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

Given the prominence of changes in the ISR pathway early in the rNLS8 mice, we also examined changes in additional ISR genes not included in the array data, including levels of protein phosphatase 1 regulatory subunit 15 A mRNA (Ppp1r15a) gene encoding GADD34 that is the direct target of ISR-ATF4-CHOP axis [11], and mRNA splicing of X-box binding protein 1 (Xbp1) that is a potent transcription factor under regulation of the inositol-requiring enzyme 1 (IRE1) branch of unfolded protein response (UPR) [12]. Notably, we identified upregulation of the Ppp1r15a gene (~ 1.4 fold, Fig. 2B) at 1-week timepoint, but not 2- or 4-week timepoint, suggesting impairment of GADD34-mediated negative feedback regulation of the ISR. As expected, the total and unspliced Xbp1 mRNAs showed no significant differences between groups (Fig. 2B), in line with the array data for total Xbp1 mRNAs (Supplementary Fig. 2A). However, there was marked elevation of spliced (active) Xbp1 mRNA (~ 2.2 fold) in rNLS8 mice at the 1-week and 2-week timepoints. Overall, these findings indicate the early activation of the ISR and broader UPR signaling from the beginning of induction of hTDP-43∆NLS expression in rNLS8 mice, with later loss of this potentially protective response during disease progression.

In line with the RT2 array data, we confirmed the mRNA levels for five further upregulated genes that were significantly increased in the cortex of rNLS8 mice at 2- and 4-week timepoints compared to control mice (Fig. 2C), including those involved in: neuroinflammation, being Ccl12, Il6, Tnf, and Trl4, and oxidative stress response, being Homx1. Notably, the expression of these five genes was not significantly different between rNLS8 and control mice at the 1-week timepoint, suggesting that activation of neuroinflammation and the oxidative stress response may be later events than ISR and apoptotic signaling in the cortex of rNLS8 mice.

Concerning the two genes identified as downregulated in the RT2 array data, we verified the downregulation of Ldha and Slc9a3 genes in the cortex of rNLS8 mice compared to the control mice at the 4-week timepoint (Fig. 2D). Unexpectedly, although both genes were not different between groups at the 2-week timepoint, we also detected a decrease of Ldha mRNA in the cortex of rNLS8 mice relative to control mice at the 1-week timepoint, suggesting early deficiency in metabolism may occur in rNLS8 mice. In contrast, we observed a significant increase of Slc9a3 mRNA at the 1-week timepoint compared to control mice, suggesting that effects of disruption of sodium absorption may change over time in rNLS8 mice. Surprisingly, we detected a decrease of neuroinflammatory Crp mRNA in the cortex of rNLS8 mice at the 4-week timepoint in qPCR results, in contrast to the increase of Crp in the RT2 array, suggesting that changes in this gene may not be a consistent feature of disease in the rNLS8 mice (Fig. 2D). Together, these findings reveal involvement of several important pathways, most notably activation of the ISR, beginning even prior to disease onset and persisting throughout early disease stages in rNLS8 mice.

To investigate whether the detected changes in the rNLS8 mouse cortex were also reflected in the spinal cord, we assessed the levels of a select set of seven of the genes that were increased in rNLS cortex, in 2-week spinal cord tissues using qPCR. Notably, at this timepoint, the levels of Chop, Atf4, Bid, Trp53 and Gadd45g, which were increased from 1-week in the cortex, and of Il6, which was increased from 2 weeks in the cortex, were unaltered in rNLS8 mouse spinal cord compared to controls at 2 weeks (Supplementary Fig. 3B). However, Ccl12 levels were dramatically increased in 2-week rNLS8 mouse spinal cord (Supplementary Fig. 3B). These findings suggest that the very early alterations in ISR and apoptotic signaling seen in the cortex are not necessarily reflected at the early disease stages in the spinal cord of rNLS8 mice, although neuroinflammatory signals including increased levels of Ccl12 are apparent.

To further biochemically validate activation of the ISR, we conducted immunoblotting (IB) to assess the phosphorylation of the eukaryotic initiation factor 2α (p-eIF2α) at serine 51 (S51), the core effector of the ISR, and the selective downstream target ATF4 protein [11] in rNLS8 mice at early disease stage (4 week timepoint). The IB results revealed slight but significant increases of eIF2α protein (~1.1 fold, Fig. 2E, F) and significant increases of p-eIF2α (S51) (~1.4 fold, Fig. 2E, G) in the cortex of rNLS8 mice compared to control mice at the 4 week timepoint. Notably, we also detected a dramatic increase of ATF4 protein (~3.8 fold, Fig. 2H, I) in rNLS8 mice relative to control mice. These data further confirm activation of the ISR caused by accumulation of TDP-43 pathology in rNLS8 mice at early disease stages.

Antisense oligonucleotide-mediated knockdown of Chop does not ameliorate motor deficits in rNLS8 mice

Apoptosis has been proposed as the final pathway leading to neurodegeneration in ALS and FTD [10]. Our array and qPCR results revealed consistent upregulation of Atf4, Chop, and also Bid (an apoptotic gene downstream of ISR-ATF4-CHOP signaling) in the cortex of rNLS8 mice as early as 1 week off Dox, suggesting that ISR-mediated apoptosis may be a pathogenic mechanism associated with TDP-43 pathology. CHOP protein not only converges upstream multiple stress signals such as ISR [11, 12], DNA damage [32] and inflammation [33], which we found to be dysregulated in rNLS8 mice (Fig. 2), but CHOP also acts as a critical transcription factor to directly regulate the expression of downstream apoptosis genes and to coordinate the initiation of the apoptosis cascade [13]. Importantly, previous evidence indicates that knockdown of Chop gene using shRNA can inhibit neuronal apoptosis [34]. Therefore, we selected Chop as the target gene to manipulate inhibition of apoptotic signaling in rNLS8 mice, aiming to protect neurons against cell death and thereby prevent motor deficits. Previously, several RNase H-gapmer ASOs were reported to specifically target the Chop gene, successfully decreasing levels of CHOP protein in cells and mice [20]. We thus examined the efficacy of these previously reported Chop ASOs [20], and selected the most effective Chop ASO (previously reported as ASO#3), which suppressed Chop gene expression in the cortex of rNLS8 mice relative to control levels (Supplementary Fig. 4), for subsequent experiments in rNLS8 mice.

To investigate the effect of knockdown of Chop in rNLS8 mice, we conducted bilateral ICV injection of Chop ASO or its vehicle (saline) to adult littermate control or rNLS8 mice (10 weeks old), two weeks prior to Dox removal to ensure that decreased Chop mRNA occurred from the beginning of induction of hTDP-43∆NLS expression (Fig. 3A). We examined neurological and behavioural phenotypes, TDP-43 pathology, and motor neurodegeneration in treated mice. Firstly, qPCR showed significantly higher (by ~1.5 fold) Chop mRNA levels in the cortex of saline-treated rNLS8 mice (referred as the rNLS-Sal group) at mid-disease stage of 6 weeks off Dox (eight weeks after ICV injection of ASOs), compared to the saline-treated control mice (the Ctrl-Sal group) (Fig. 3B). The data indicate that the early increases in Chop mRNA previously detected from 1 week off Dox continued until at least this later disease state at 6 weeks off Dox. The qPCR data also confirmed that Chop ASO successfully decreased the Chop mRNA in the cortex in rNLS8 mice (the rNLS-ASO group), to approximately the level of control saline-injected mice (Fig. 3B). Although there were no significant differences in the mRNA levels of Chop in the spinal cord between Ctrl-Sal and rNLS-Sal mice, Chop was decreased by ~40% in the spinal cord of the rNLS-ASO group compared to the Ctrl-Sal or rNLS-Sal groups (Fig. 3B), indicating that direct ICV injection of Chop ASO can effectively decrease both cortex and spinal cord Chop gene expression.

Fig. 3: CHOP protein levels are increased in rNLS8 mice at 6 weeks off Dox, but ASO-mediated knockdown of Chop prior to induction of hTDP-43ΔNLS expression does not alter disease phenotypes or TDP-43 levels or solubility.
figure 3

A Experimental schema. B Real-time qPCR assessment demonstrated Chop ASO decreased Chop mRNA levels in the cortex and spinal cord of treated rNLS8 mice at 6 weeks off Dox (WOD) (n = 4). Data are normalised to Actb as the housekeeping gene. C Immunoblotting for levels of CHOP protein in the cortex and spinal cord of treated rNLS8 mice at 6 WOD (n = 4). Approximate molecular weights (kDa) are indicated. D Quantification of immunoblots showed increased CHOP protein in the cortex and a trend of decrease of CHOP protein with ASO treatment in rNLS8 mice at 6 WOD. However, there were no significant differences between groups in CHOP protein levels in the lumbar spinal cord. E–G. Compared to saline-treated rNLS8 mice (rNLS-Sal), Chop ASO treatment led to no apparent beneficial effects on pathological phenotypes and motor deficits in clasping or rotarod test and did not alter the decline of body mass or affect the number of mice that had not reached humane endpoint by 6 WOD: with no mice reaching humane endpoint in Ctrl-Sal group (n = 12), three mice reaching humane endpoint in rNLS-Sal group (n = 14), and six mice reaching humane endpoint in rNLS-ASO group (n = 13) (H). IB assessed TDP-43 in the RIPA-soluble and RIPA-insoluble fractions and p-TDP-43 in the RIPA-insoluble fractions for the cortex (I–L) and spinal cord (M–P) of mice at 6 WOD (n = 4). Approximate molecular weights (kDa) are indicated on the left and target protein on the right. GAPDH was used as the loading control for RIPA-soluble IB and total protein for RIPA-insoluble IB. Quantification of IB for different molecular weight species of TDP-43 or p-TDP-43 in the RIPA-soluble and -insoluble fractions in the cortex (J–L) and the spinal cord (N–P) (n = 4). Representative IF images for pan-TDP-43 staining (human+mouse) (green) with the nuclear marker DAPI (blue) in the cortex (Q) and spinal cord (T) of treated mice at 6 WOD. Scale bars represent 50 µm. Quantification of IF for pan-TDP-43 protein in identified cell or cytoplasmic region in the cortex (R, S) and spinal cord (U, V) of treatment groups. Ctrl-Sal (n = 5), rNLS-Sal (n = 6) and rNLS-ASO (n = 6). Mean ± SEM. *p < 0.05, **p < 0.01 and ***p < 0.001.

Furthermore, we conducted immunoblotting to analyse CHOP protein levels in the cortex and spinal cord of rNLS8 mice after ASO treatment (Fig. 3C). Despite difficulties in detecting this protein via immunoblotting, we identified significant elevation of CHOP protein in the cortex of the rNLS8 mice (rNLS-Sal) of approximately 2.1-fold compared to the control (Ctrl-Sal), with a trend for a decrease in CHOP protein in rNLS-ASO mice (approximately 1.6-fold compared to rNLS-Sal mice), although this decrease did not reach statistical significance (Fig. 3D). CHOP protein levels were difficult to detect in the spinal cord and there were no apparent differences in the levels of CHOP protein in the spinal cord of treatment groups despite the decrease detected via qPCR, although there was substantial variation in detected protein levels between mice (Fig. 3D). Immunofluorescence of the motor cortex in mice at 6 weeks off Dox indicated that increased CHOP protein was primarily localized to neurons expressing hTDP-43ΔNLS, with some additional surrounding non-neuronal cells also showing increased CHOP levels in rNLS8 mice (Supplementary Fig. 5). Overall, consistent with previous evidence for a successful decrease of CHOP protein upon treatment with the ASO used here [20], our results confirmed successful knockdown of Chop in the nervous systems of rNLS8 mice.

We further examined the effects of knockdown of Chop gene on neurological phenotypes of rNLS8 mice until late disease stages. rNLS8 mice developed characteristic features of progressive neurological decline compared to control mice, including hind-limb clasping onset (Fig. 3E) and progressive motor dysfunction (Fig. 3F), and body weight loss (Fig. 3G), similar to results previously reported in mixed B6/C3H background [19] and recently in pure C56BL/6JAusb background [21]. Notably, Chop ASO treatment had no significant effects on any of the examined neurological phenotypes (Fig. 3E–G). Further, there was no difference in the number of rNLS-ASO mice remaining on-study at 6 weeks off Dox compared to the rNLS-Sal mice (Fig. 3H), although later disease stages were not assessed. Lastly, Chop ASO did not significantly affect the expression of genes identified by RT2 array in the cortex or spinal cord of treated mice comparing to the rNLS-Sal mice (Supplementary Fig. 6). Overall, these results demonstrate that knockdown of Chop gene did not affect the onset or early progression of motor phenotypes in rNLS8 mice.

Knockdown of Chop does not alter overall TDP-43 pathology in rNLS8 mice

As TDP-43 pathology, such as cytoplasmic aggregation, modification and phosphorylation, is the critical pathological feature of ALS and FTD [1, 5], we next examined the outcomes of knockdown of Chop on TDP-43 pathology in treated rNLS8 mice by assessing the levels of overall soluble and insoluble TDP-43 protein, TDP-43 fragments, and pTDP-43. We detected significant increases of total TDP-43 protein (human and mouse TDP-43) in the RIPA-soluble fractions of cortex (Figs. 3I, J) and spinal cord (Figs. 3M, N) of rNLS-Sal mice compared to the Ctrl-Sal group, as expected. However, Chop knockdown caused no significant changes to the levels of RIPA-soluble total TDP-43 protein in cortex (Figs. 3I, J) or spinal cord (Figs. 3M, N) of rNLS8 mice compared to the rNLS-Sal group. The levels of total TDP-43 protein were also dramatically increased in the RIPA-insoluble fraction of both cortex (Figs. 3I, K) and spinal cord (Figs. 3M, O) of the rNLS-Sal mice relative to Ctrl-Sal mice, as seen previously [19]. Notably, we detected accumulation of RIPA-soluble and -insoluble TDP-43 C-terminal fragments (CTFs) in the rNLS8 mice with the antibodies and conditions used in this study (Figs. 3I, M), in contrast to the original report of this model [19]. Chop knockdown had no effect on levels of total, CTF or high-molecular weight TDP-43 in the RIPA-insoluble fraction in the cortex of rNLS8 mice (Figs. 3I, K), and there was similarly no effect on total or the high-molecular weight TDP-43 species in the RIPA-insoluble fraction in the spinal cord of rNLS8 mice (Figs. 3M, O). However, interestingly, the spinal cord samples from rNLS-ASO mice showed a modest but statistically significant decrease in the low-molecular-weight TDP-43 CTFs (20–37 kDa) compared to rNLS-Sal mice (Fig. 3O). We further examined levels of pTDP-43 at serines 409/410 in the RIPA-insoluble protein fractions, demonstrating significant increases of pTDP-43 in both cortex (Fig. 3L) and spinal cord (Fig. 3P) of the rNLS-Sal mice compared the Ctrl-Sal mice, as expected. Nevertheless, knockdown of Chop did not change the levels of pTDP-43 in either cortex or spinal cord of rNLS8 mice.

To further analyse the effect of Chop knockdown on TDP-43 accumulation in the cytoplasm of neurons in rNLS8 mice, we conducted IF for TDP-43 and analysed neurons in the cortex (Fig. 3Q) and spinal cord (Fig. 3R), with TDP-43 primarily localised in the nuclei of neurons of the Ctrl-Sal mice. In contrast, TDP-43 was localised to cytoplasmic regions in the neurons in the layer V of the cortex (Fig. 3Q), and neurons in the ventral horn of the lumbar spinal cord in rNLS8 mice (Fig. 3T), as expected [19]. Nevertheless, Chop ASO treatment did not result in any changes in overall levels or cytoplasmic TDP-43 distribution in neurons of cortex or spinal cord in rNLS8 mice. Quantification of IF confirmed the elevation of TDP-43 in the rNLS-Sal and rNLS-ASO mice shown as fluorescence intensity measured per cell or within the cytosolic regions in cortical (Figs. 3R, S) and spinal cord neurons (Figs. 3U, V). Nevertheless, suppression of Chop by ASO led to no significant differences in total or cytosolic fluorescence intensity of TDP-43 between rNLS-Sal and rNLS-ASO groups. Overall, these data demonstrate that knockdown of Chop did not consistently affect the overall accumulation or mislocalisation of TDP-43 in rNLS8 mice.

rNLS8 mice display changes in apoptosis signaling over time and increased cleaved caspase-3 levels that are not affected by knockdown of Chop

It remains unclear how TDP-43 pathology affects the regulation of the downstream apoptotic mediators of CHOP activation. Thus, we first examined the expression of these apoptotic genes in rNLS8 mice at early disease stages. The qPCR results revealed significant elevation of anti-apoptotic Bcl2 gene (~1.3 fold, Fig. 4A), Bcl2 homology 3 (BH3)-only pro-apoptotic initiator Bim gene (~1.5 fold, Fig. 4B) and Noxa gene (~2.2 fold, Fig. 4C) in the cortex of rNLS8 mice before disease phenotypes at 1 week off Dox, compared to control mice. There were no significant changes in Bcl2 or Bim mRNAs at 2 or 4 weeks off Dox in rNLS8 mice, although upregulation of Noxa mRNA was detected at 2 weeks but not 4 weeks in rNLS8 mice (Fig. 4C). Notably, Puma mRNA level was not altered at 1 week off Dox but was significantly increased in the rNLS8 mice at 2- and 4-week timepoints, after disease onset (Fig. 4D). Dysregulation of these apoptosis genes suggests involvement of apoptosis signaling in neurodegeneration caused by TDP-43 pathology in rNLS8 mice, and suggests a differential switch from early competition between anti- and pro-apoptotic signals to later elevation of pro-apoptotic signaling at later disease phases.

Fig. 4: rNLS8 mice display early activation of apoptosis, and increased cleaved caspase-3 levels in the cortex that is not altered by knockdown of Chop.
figure 4

Real-time qPCR results revealed dysregulation of Chop target apoptosis genes in the cortex of rNLS8 mice at 1, 2 and 4 weeks off Dox (WOD) (n = 5): A Bcl2; B Bim; C Noxa; D Puma. Real-time qPCR results revealed expression changes of CHOP downstream genes in (E) and spinal cord (F) of mice at 6 WOD after ASO treatment (n = 4). Actb as the housekeeping gene. G IB for cleaved caspase-3 in the cortex and spinal cord of mice at 6 WOD (n = 4). Approximate molecular weight markers (kDa) are shown, and GAPDH is used as loading control. H Quantification of IB for cleaved caspase-3 (17 kDa) in the cortex and spinal cord of treated mice. Mean ± SEM. *p < 0.05, **p < 0.01 and ***p < 0.001.

Further, we assessed the expression of these apoptotic genes in rNLS8 mice at the mid-stage of disease (6 weeks off Dox) and with knockdown of Chop. Interestingly, we detected significantly decreased mRNA levels of the anti-apoptotic Bcl2 gene as well as pro-apoptotic Bim and Noxa genes, and no difference in Puma mRNAs, in the cortex of the rNLS-Sal mice compared to the Ctrl-Sal mice at 6 weeks off Dox (Fig. 4E). This is in contrast to the upregulation of these genes identified at earlier disease stages of rNLS8 mice, and may be due to extensive neurodegeneration that occurs at this stage of disease [19]. Notably, Bid expression, which was upregulated in rNLS8 mice at early disease stages (1, 2 and 4 weeks off Dox, Fig. 2A), was not significantly different between control mice (Ctrl-Sal) and rNLS8 mice at this later disease stage (6 weeks off Dox). rNLS8 mice also displayed no alteration of examined genes in the spinal cord relative to the control mice (Ctrl-Sal) at 6 weeks off Dox (Fig. 4F). Further, Chop knockdown had no effect on the mRNA levels of anti- or pro-apoptosis genes in either cortex or spinal cord of treated mice at 6 weeks off Dox (Figs. 4E, F).

Given that caspases play essential roles in orchestrating apoptosis and caspase-3 is a primary downstream executioner among these proteases, we also examined the activation of caspase-3 in the cortex of rNLS8 mice by quantifying the levels of a specific 17 kDa caspase-3 cleavage product (Fig. 4G), which is required for caspase-3 activity in apoptosis [35]. We detected a significant elevation of cleaved caspase-3 (~17 kDa band) in the cortex of the rNLS8 mice (rNLS-Sal) compared to the control mice (Ctrl-Sal) at the 6-week timepoint (Fig. 4H), suggesting that activation of apoptosis may drive neurodegeneration in the later stages of disease in rNLS8 mice. Similarly, there was a trend of increased cleaved caspase-3 levels in the spinal cord of rNLS8 mice (rNLS-Sal) relative to the control mice (Ctrl-Sal), although it did not reach statistical significance (Fig. 4H). Elevation of cleaved caspase-3 is consistent with the previous report that demonstrated dramatic motor neuron loss in the cortex of rNLS8 mice at 6 weeks off Dox [19]. Nevertheless, knockdown of Chop had no effect on the level of cleaved caspase-3 in either cortex or spinal cord of treated mice (rNLS-ASO) relative to the rNLS-Sal group (Fig. 4H). Overall, these data demonstrate that knockdown of Chop did not ameliorate the disease-associated activation of apoptosis signaling in rNLS8 mice, suggesting that alternative modulators or pathways may account for the upregulation of these apoptosis genes caused by TDP-43 pathology.

rNLS8 mice display dramatic astrogliosis and dysregulation of astrocytic genes, which are not ameliorated by knockdown of Chop

In addition to the cell-autonomous effect of ISR in neurons, dysfunctional astrocytes can also contribute to the pathogenesis of human neurodegenerative diseases [36]. Given that astrogliosis is an established feature of both ALS [37] and the rNLS8 mouse model [19], we assessed the activation of astrocytes that contribute to glutamate-induced motor neuron death in ALS [38] and which would be expected to be ameliorated by a beneficial treatment strategy. Firstly, we assessed astrocytes using the pan-reactive astrocytic marker GFAP by IF (Figs. 5A, D). The results indicated a dramatic increase in the number of GFAP-positive astrocytes with ramified cell processes in the cortex and spinal cord of rNLS8 mice (rNLS-Sal) compared to the Ctrl-Sal mice (Fig. 5B). Knockdown of Chop slightly but significantly increased the number of reactive astrocytes in treated mice (rNLS-ASO) relative to that of rNLS-Sal mice (Fig. 5B), with a slight elevation of GFAP intensity in the cortex of rNLS-ASO mice compared to the control mice (Ctrl-Sal) (Fig. 5C). However, we found no difference in either astrocyte number (Fig. 5E) or GFAP intensity (Fig. 5F) in the spinal cord among treatment groups despite a trend for increase in rNLS8 mice.

Fig. 5: rNLS8 mice develop reactive astrogliosis and altered astrocytic gene expression, which are not ameliorated by knockdown of Chop.
figure 5

Representative IF images for pan-TDP-43 (human+mouse) (green), reactive-astrocytic marker GFAP (red), and the nuclear marker DAPI (blue) in the layer V of cortex (A) or the ventral horn of spinal cord (D) of mice at 6 weeks off Dox (WOD). Scale bars represent 50 µm. Cell counting for GFAP positively labeled astrocytes in the cortex (B) and spinal cord (E) of treated mice (n = 3). Quantification of IF for GFAP intensity in the cortex (C) and spinal cord (F) of treated mice at 6 WOD (n = 3). qPCR analysis of astrocytic gene expression in the cortex (G) and spinal cord (H) of Chop ASO-treated mice at 6 WOD (n = 4). Data are normalised to Actb as the housekeeping gene. Mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001.

The ISR has been shown to impair the neurotrophic function and to modulate inflammatory states of astrocytes, by dysregulation of astrocytic gene expression [39]. Therefore, to determine whether knockdown of Chop affected the expression of genes important for astrocytic physiological function and reactivity, we examined levels of two representative astrocytic genes. The astrocytic solute carrier family 1 member 2 gene Slc1a2/Glt1 that encodes glutamate transporter GLT1/EAAT2 plays key roles in glutamate reuptake and modulation of homeostatic brain function, and is deficient in ALS [40]. The pan-reactive astrocytic gene Serpina3n encodes the serine protease inhibitor that is upregulated during neurodegeneration [36]. We detected a significant decrease of the Slc1a2/Glt1 mRNAs in the cortex (Fig. 5G) and spinal cord (Fig. 5H) of rNLS8 mice (rNLS-Sal) compared to the control mice (Ctrl-Sal). However, there were no significant differences in the Slc1a2/Glt1 mRNAs of rNLS-ASO mice relative to the rNLS-Sal group in the cortex or spinal cord. Similarly, rNLS8 mice (rNLS-Sal) displayed dramatic upregulation of Serpina3n in the spinal cord compared to Ctrl-Sal mice (Fig. 5H), although Serpina3n mRNAs was not significantly different in the cortex between Ctrl-Sal and rNLS-Sal mice (Fig. 5G). Nevertheless, knockdown of Chop did not affect Serpina3n mRNAs in treated mice (rNLS-ASO) compared to the rNLS-Sal group. Overall, these results not only confirmed previous reported astrogliosis in rNLS8 mice at the mid-disease stage [19] but also revealed dramatic dysregulation of disease-related astrocytic genes, suggesting disruption of astrocytic function may contribute to neurodegeneration in this model. Nevertheless, we did not identify any evidence for amelioration of disease-related astrogliosis or astrocytic gene expression alterations upon knockdown of Chop in rNLS8 mice.

Discussion

Here, we aimed to determine which cellular stress pathways contribute to disease onset and progression in ALS, using a well-characterized mouse model of cytoplasmic TDP-43 proteinopathy. Our data identified early dysregulation of multiple genes in several signaling pathways in the cortex of pre-onset rNLS8 mice, including Atf4, Chop, Bid, Gadd45g, and Trp53, extending throughout disease progression (summarized in Fig. 6A). Most notably, we identified early activation of the ISR and alterations in apoptosis pathways even in rNLS8 mice prior to onset of motor phenotypes and continuing through early disease stages, leading to later activation of caspase-3 (Fig. 6B). However, suppression of Chop, a key activator of the ISR, beginning prior to induction of hTDP-43ΔNLS expression, did not ameliorate disease pathology, onset, or progression in rNLS8 mice. Together, these studies demonstrate that cellular stress pathways, possibly including early anti-apoptotic signaling, are activated prior to disease onset and may contribute to neurodegeneration in rNLS8 mice. However, suppression of Chop throughout disease is not sufficient to prevent a switch to pro-apoptotic signaling associated with disease progression.

Fig. 6: TDP-43 pathology causes early activation of the ISR and apoptosis signaling in rNLS8 mice.
figure 6

A Summary of gene expression changes in the cortex of rNLS8 mice over time. B Mechanism schema. TDP-43 mislocalisation induces early activation of multiple cell stress signaling pathways in the cortex of rNLS8 mice even before disease onset, such as the ISR, UPR, DNA damage response, and apoptosis. Notably, the prolonged activation of these cell stress pathways caused by TDP-43 pathology, in particular the ISR, continues through early-disease stages where rNLS8 mice displayed dramatic elevation of p-eIF2α and ATF4 protein, and impairment of cellular metabolism. Dysregulation of stress pathways leads to reduction of anti-apoptotic Bcl2, continued increased expression of pro-apoptotic genes (Bid, Bim, Noxa, and Puma), and increases of cleaved caspase-3 and astrogliosis, likely contributing to neurodegeneration in rNLS8 mice.

In contrast to healthy neurons with tightly controlled protein synthesis, folding and degradation, in disease protein aggregation triggers the ISR which can ultimately lead to neuronal death [11, 12]. TDP-43 pathology is associated with activation of ISR in both neurons and non-neuron cells in nervous system postmortem tissues of people who have died with ALS or FTD [41], with increased CHOP levels identified in multiple types of cells in the postmortem tissues of human ALS [17] as well as in cell and mouse models for ALS and FTD [15, 20, 42, 43]. Here in this study, the IF results demonstrate CHOP protein primarily in hTDP-43ΔNLS-positive neurons, although CHOP was also detected in some surrounding non-neuronal cells in rNLS mice. This suggests potential involvement of both cell-autonomous as well as non-cell-autonomous mechanisms in ISR activation in disease. Interestingly, in vivo, TDP-43 pathology results in global protein translation inhibition, which is the main effect of the ISR [44]. Moreover, in cell culture, disease-associated mutant TDP-43 proteins trigger ISR activation and elevation of downstream gene expression including Chop, involving an amplifying cell stress response potentially leading to cell death [15]. Indeed, future experiments are required to assess the effects of modulation of other ISR genes such as Atf4, which has broader downstream effects [11, 12] and may be involved in the clearance of TDP-43 inclusions via autophagy [45]. Furthermore, CHOP was recently identified as one of the top upregulated genes in a human neuronal model of TDP-43 pathology, and was shown to be a direct target of TDP-43-mediated gene expression with increased CHOP levels in neurons with depleted nuclear TDP-43 in human FTLD-TDP-derived samples [46]. In addition, mice expressing ALS-associated mutant SOD1 [43, 47] or ALS and FTD-associated FUS [48] also display activation of ISR and/or impairment of ongoing protein synthesis, including upregulation of Chop [49]. The ISR is also activated in neurons expressing expanded C9orf72 hexanucleotide repeats, and is similarly seen in the cortices of people who have died with C9orf72-linked ALS and FTD [50]. Together, these studies indicate that ISR activation, and Chop in particular, is involved broadly in disease, although it remains unclear whether these pathways may be amenable to therapeutic targeting in TDP-43-related disease. Our data show that knockdown of Chop using ASO decreased Chop expression, but had no impact on motor deficits or overall TDP-43 pathology. These findings are consistent with previous reports that neither genetic knockout nor ASO-mediated knockdown of Chop led to beneficial outcomes in the mutant SOD1G93A mouse model of ALS [20, 51] or rTg4510 mice for Alzheimer’s disease [52]. Nevertheless, we did detect a slight albeit statistically significant decrease of insoluble TDP-43 CTFs in ASO-treated rNLS8 mice. Indeed, previous work has suggested that TDP-43 fragmentation may occur through activation of caspases such as caspase-3, a transcriptional target of CHOP [53, 54]. However, although TDP-43 CTFs are a pathological hallmark in the brains of people with ALS or FTD, current evidence suggests that TDP-43 CTFs may not act as a primary driver of disease pathogenesis [54]. Therefore, the disease relevance of a small change in TDP-43 CTFs in ASO-treated rNLS8 mice in our study remains unclear. Future cell culture experiments to address the biochemical alterations caused by CHOP ablation in neurons and the effects of decreased CHOP levels on TDP-43 pathology may be needed to reveal the mechanisms involved in these processes.

Prolonged ISR activation induces apoptosis via cell death downstream of the intrinsic pathway that is regulated by BCL2 family members (BIM, NOXA, PUMA), and the extrinsic pathway (BID) that can be triggered by various cell death signals [10]. Notably, our results revealed that induction of hTDP-43∆NLS expression rapidly induced both anti-apoptotic (Bcl2) and pro-apoptotic genes (Bim, Bid, Noxa) in rNLS8 mice even before disease onset. In contrast, from disease onset, Bcl2 mRNA in rNLS8 mice returned to the level similar to the control mice, coinciding with elevation of pro-apoptotic Bid, Noxa, and Puma mRNAs in rNLS8 mice, suggesting a switch from pro-survival to pro-death signalling over time. Importantly, our findings revealed that activation of these pro-apoptotic genes occurred differentially over the disease course (summarized in Fig. 6). For example, Bim mRNA was increased only prior to disease onset, in line with the previous report that Bim gene can be induced by overexpression of TDP-43 in neurons [42]. In contrast, p53-mediated transcription of Noxa and Puma was increased in rNLS8 mice during early disease stages, consistent with our data showing upregulation of Trp53 and Gadd45g. Notably, Bid, a gene that can be induced by a variety of upstream signals from the ISR, DNA damage, oxidative stress, and death receptors (for example, TNF and its receptor), was upregulated prior to disease onset and into early disease stages. This differential regulation of apoptotic genes indicates that multiple stress signaling pathways may co-contribute to and even enhance pro-apoptotic signaling in rNLS8 mice at different disease stages. Moreover, we speculate that significant neuron loss occurs in the motor cortex of rNLS8 mice by 6 weeks off Dox, which may partly account for the late decrease of Bcl2, Bim, and Noxa expression in rNLS8 mice, consistent with previous findings that show decreases of Bcl2 mRNA in postmortem tissues in ALS and other human neurodegenerative diseases [35]. Importantly, rNLS8 mice displayed elevation of pro-apoptotic executioner caspase-3 at the mid-stage of disease, providing direct evidence that supports the previous findings of increased caspase-3 in brain and spinal cord autopsy tissues of people with ALS and FTD [54]. Overall, our data, therefore, supports the hypothesis that TDP-43 pathology induces apoptosis contributing to neurodegeneration, but reveals complex regulation between pro-survival and pro-death signaling that begins early in the disease course. Thus, targeting apoptosis may be a beneficial therapeutic strategy for treatment of ALS and FTD. Notably, we did not detect any changes in expression of necroptosis genes Ripk1 and Ripk3 at 2 or 4 weeks off Dox in rNLS8 mice, despite the detected alterations of apoptosis signalling even at these early timepoints, aligning with recent work suggesting that necroptosis is not a primary driver of neurodegeneration in ALS and FTD [55].

Our results also demonstrated dramatic astrogliosis in the cortex and spinal cord of rNLS8 mice, with significant upregulation of pan-reactive astrocytic marker Serpina3n in rNLS-ASO mice, which can be induced upon neuronal damage and contributes to neuroinflammation [36, 37]. Our finding is consistent with previous studies showing upregulation of Serpina3n in a conditional TDP-43 knockout mouse model for FTD [56], suggesting that the loss of endogenous nuclear TDP-43 in neurons of the rNLS8 mice [19] may account for this phenotype. Notably, our results also indicated the downregulation of the Slc1a2 gene encoding glutamate transporter GLT1 in rNLS8 mice. This is in line with the previous observation that GLT1 expression is selectively decreased in astrocytes of ALS postmortem tissues [40] and also in TDP-43-associated FTD cases [57]. Hence, our results further support the hypothesis that disease-associated astrocyte reactivity and deficiency of astrocytic reuptake of glutamate at the synaptic cleft contribute to neurotoxicity in ALS and FTD. Interestingly, suppression of Chop via ASO slightly increased the number of reactive astrocytes, potentially supporting the previous findings that CHOP is involved in astrocyte survival [58] or astrocyte reactivation [39] under stress conditions.

In addition to ISR, other biological processes also appear active early in the rNLS8 mice, such as DNA damage response, UPR and dysfunction in cellular metabolic functions. Our data revealed dramatic elevation of Trp53 and Gadd45g genes as early as 1 week off Dox and continued to early disease stages. Previous research showed depletion of TDP-43 or ALS-linked mutant TDP-43 induces DNA damage and consequently p53-dependent apoptosis of motor neurons [59, 60], while GADD45G acts as a stress sensor for p53-mediated apoptosis [61, 62]. Moreover, it is proposed that oxidative stress is a pathogenic mechanism in disease, in parallel with or downstream of ISR [4, 6]. Our results reveal elevation of the oxidative stress-responsive gene Homx1 in the cortex of rNLS8 mice from the 2-week time point, in line with evidence from the motor cortex of ALS or FTD-linked mutant TDP-43 transgenic mice [6, 63, 64] and in the lumbar spinal cord of SOD1G93A mice for ALS [65]. However, most oxidative stress genes remained unchanged in our study, suggesting that further studies are required to identify the roles of oxidative stress genes in TDP-43-related disease. Additionally, we detected significant increases of the spliced (active) form of Xbp1, a key component of the UPR mediated by activation of IRE1, in rNLS8 mice at 1 and 2 weeks off Dox. These results indicate involvement of early activation of UPR signaling in rNLS8 mice in response to accumulation of TDP-43 pathology, in addition to the ISR. Further, our data demonstrate early deficiency of cellular metabolism in rNLS8 mice at early disease stage (4 weeks off Dox). LDHA, encoded by the Ldha gene, is specifically expressed in cortical neurons (including Layer V neurons) in human and rodent brains, and plays a vital role in neuronal health, with knockdown of Ldha inducing intrinsic apoptotic signalling [66,67,68]. This suggests loss of LDHA as a mechanism by which apoptosis may be triggered in rNLS8 mice. Since many of these pathways are also regulated at translational and post-translational levels, it is worth further investigation using high-throughput protein screening, for instance, quantitative proteomics [69], to characterise the interplay among pathways involved in ALS and FTD.

Of relevance for therapeutic development, ASOs have become an increasingly promising class of drugs to target conventionally undruggable molecules or pathways in a broad range of human neurological and psychiatric diseases. For instance, the FDA-approved Nusinersen targeting SMN2 to treat spinal muscular atrophy has shown remarkable efficacy in the clinic [70]. Recently, there have been newly developed ASO drugs for ALS and FTD in clinical trials, such as Tofersen to decrease SOD1 protein synthesis [71] and an ASO to inhibit translation of C9orf72 mRNA (WVE-004, Wave Life Sciences). Moreover, preclinical trials have also been undertaken to ameliorate TDP-43 pathology via modulation of TDP-43-regulating proteins such as Ataxin-2 [53, 72], which are now in Phase I clinical trials (ClinicalTrials.gov, Identifier# NCT04494256). Further identification of disease-relevant targets that are able to protect against neurodegeneration caused by TDP-43 dysfunction offers hope for development of additional improved therapeutics, potentially including antisense approaches, for ALS and FTD [73].

Limitations

Among the limitations of our work, we analysed a single timepoint of administration of the Chop ASO, namely two weeks prior to the removal of Dox from rNLS8 mice. This allowed us to assess the effects of Chop suppression from before induction of hTDP-43ΔNLS expression and throughout the entirety of the study. Nevertheless, short-term activation of ISR may actually be protective for cells under acute cellular stress conditions [11, 12, 16], and so it is possible that the activation of Chop in rNLS8 mice is a beneficial response, at least early in the disease course. Indeed, our findings of shifts in activation or inhibition of different signalling pathways prior to onset and throughout disease indicate that the effects of ISR activation may be modulated by other contributing factors, such as changes in apoptosis signalling, at different disease points. Knockdown of Chop from the beginning of induction of hTDP-43∆NLS may have thereby ablated any neuroprotective effects of an initial ISR activation in neurons. Therefore, future studies could determine whether Chop suppression beginning only later in the disease course is beneficial in extending survival. Indeed, in this study we did not examine whether knockdown of Chop can modify the disease phenotypes or lifespan of rNLS8 mice beyond 6 weeks off Dox, which may be necessary to identify any beneficial longer-term effects. Furthermore, recent work has highlighted the importance of the level of cellular stress for effective responses to small molecule modulators of the ISR. For example, the small molecular ISRIB, which overcomes eIF2a phosphorylation-mediated inhibition of translation by stimulating eIF2B activity, effectively rescues protein translation inhibition under conditions of acute viral infection with low levels of p-eIF2a but is ineffective in chronic infection with high levels of p-eIF2a present [74]. These results suggest that a critical threshold of ISR activation exists in disease conditions above which normally protective therapeutic approaches become ineffective [74]. Additional studies are also required to investigate changes at a single-cell level. Lastly, ISR activation is also involved in glial activation [39]. It will therefore be important to further investigate using cell-specific approaches the effects of ISR activation on hTDP-43∆NLS-expressing neurons as well as surrounding glial cells, in particular reactive astrocytes [75] and disease-associated microglia [76] that are involved in disease development of ALS and FTD. Regardless, our work also highlights other likely mediators of neurotoxicity which may be more promising targets for further development, including a switch between the balance or anti- and pro-apoptotic signalling. Further work is required to investigate therapeutically relevant targets and to define the timing of therapeutic intervention to successfully ameliorate disease.

Conclusion

Taken altogether, we identified early activation of several distinct cellular stress pathways, notably the ISR and pro-apoptotic signalling, in the cortex of rNLS8 mice even prior to disease onset. These pathways are essential for neuronal survival-and-death decisions, indicating that the biological pathway towards neurodegeneration begins very early after accumulation of cytoplasmic TDP-43 in disease. Our data revealed that cytoplasmic TDP-43 accumulation induces pro-apoptotic caspase-3 in parallel with decreases of anti-apoptotic Bcl2 in the central nervous system of TDP-43 mice. We conclude that multiple cellular stress pathways are active early in rNLS8 mice, notably the pro-apoptotic signalling that is activated even prior to disease onset, which likely trigger eventual motor neuron death via apoptosis following loss of competing anti-apoptotic signals. Although knockdown of the key ISR mediator Chop had no effect on motor deficits and did not ameliorate overall TDP-43 pathology in rNLS8 mice, our findings indicate that targeting of cell stress and death signalling may be a potential promising avenue for treatment of TDP-43-associated ALS and FTD.