Abstract
Amyotrophic lateral sclerosis (ALS) is a debilitating neurodegenerative disorder marked by progressive motor neuron degeneration and muscle denervation. A recent transcriptomic study integrating a wide range of human ALS samples revealed that the upregulation of p53, a downstream target of inflammatory stress, is commonly detected in familial and sporadic ALS cases by a mechanism linked to a transactive response DNA-binding protein 43 (TDP-43) dysfunction. In this study, we show that prolonged interferon-gamma (IFNγ) treatment of human induced pluripotent stem cell–derived spinal motor neurons results in a severe cytoplasmic aggregation of TDP-43. TDP-43 dysfunction resulting from either IFNγ exposure or an ALS-associated TDP-43 mutation was associated with the activation of the p53 pathway. This was accompanied by the hyperactivation of neuronal firing, followed by the complete loss of their electrophysiological function. Through a comparative single-cell transcriptome analysis, we have identified significant alterations in ALS-associated genes in motor neurons exposed to IFNγ, implicating their direct involvement in ALS pathology. Interestingly, IFNγ was found to induce significant levels of programmed death-ligand 1 (PD-L1) expression in motor neurons without affecting the levels of any other immune checkpoint proteins. This finding suggests a potential role of excessive PD-L1 expression in ALS development, given that PD-L1 was recently reported to impair neuronal firing ability in mice. Our findings suggest that exposing motor neurons to IFNγ could directly derive ALS pathogenesis, even without the presence of the inherent genetic mutation or functional glia component. Furthermore, this study provides a comprehensive list of potential candidate genes for future immunotherapeutic targets with which to treat sporadic forms of ALS, which account for 90% of all reported cases.
- amyotrophic lateral sclerosis
- interferon-gamma
- iPSC-derived spinal motor neuron
- neuroinflammation
- p53 pathway
- TDP-43 proteinopathy
Significance Statement
ALS is currently an incurable neurodegenerative disease that primarily damages the motor function. Ninety percent of the reported cases have unknown reasons, but their progression is extremely fast once triggered. A pathologic hallmark of ALS is an aggregation of RNA-/DNA-binding protein, transactive response DNA-binding protein 43 (TDP-43), but its pathologic link to the disease is yet to be elucidated. In this study, we found that interferon-gamma (IFNγ), an immune-derived cytokine, activates the p53 pathway which appears commonly in postmortem ALS tissue. Importantly, IFNγ triggered a significant cytoplasmic TDP-43 aggregation and impaired the electrophysiological function of human motor neurons. Furthermore, we found that ALS risk genes related to mitochondrial dysfunction are aberrantly expressed under the IFNγ exposure, which constitutes potential therapeutic targets of immune-dysregulated neurodegeneration in ALS.
Introduction
Amyotrophic lateral sclerosis (ALS) is an etiologically heterogeneous disease that primarily attacks motor neurons, leading to a progressive and ultimately fatal denervation of the skeletal muscle. Familial mutations are implicated as the cause for only 10% of ALS cases, while the pathologic origin of the other 90% remains unclear. The disease progresses rapidly once initiated so that the average life expectancy of patients is <5 years after the onset of symptoms (Wales et al., 2011; Hardiman et al., 2017).
The RNA-/DNA-binding protein, transactive response DNA-binding protein 43 (TDP-43), plays a central, but poorly understood, role in ALS etiology. Mutations of the gene encoding TDP-43, TARDBP, segregate with disease in some cases of familial ALS, and the loss of nuclear TDP-43, associated with the aberrant cytoplasmic accumulation of TDP-43 aggregates (TDP-434 proteinopathy), is present in 97% of both familial and nonfamilial forms of ALS. Mechanisms that initiate the TDP-43 proteinopathy and tie the TDP-43 dysfunction to motor neuron degeneration remain inadequately elucidated (Kabashi et al., 2008; Yuan et al., 2008; Barmada et al., 2010; Cohen et al., 2011; Vogt et al., 2018).
A recent collective analysis of transcriptomic studies using postmortem ALS tissues and induced pluripotent stem cell (iPSC)–derived motor neurons harboring diverse ALS-causing mutations highlighted the p53 pathway, one of the downstream targets of inflammatory stress, as playing a central role in the majority of both familial and nonfamilial ALS pathologies. Importantly, the TDP-43 proteinopathy was shown to be responsible for the p53 pathway activation. Postmortem ALS tissues with familial SOD1 and FUS mutations, which are the minority of ALS subtypes that do not develop TDP-43 proteinopathy, were the only exceptions where a significant p53 pathway activation was not observed (Maor-Nof et al., 2021; Ziff et al., 2023). Moreover, Vogt et al. (2018) found the increased expression of even normal TDP-43 in neural stem cells induces the p53 activation, implying a direct role of TDP-43 in regulating the p53 signaling in neurons. These findings suggest a significant contribution of the p53 pathway activator(s) in the development of ALS pathology.
Inflammatory cytokines are known to be elevated in ALS patients’ body fluids, and previous studies have shown that excessive levels of inflammatory cytokine exposure can trigger acute motor neuron death mainly through abnormal glial activation (Liu et al., 2015; Hu et al., 2017). However, it is still unclear whether such exposure causes motor neuron degeneration by driving the ALS-specific phenotype expression in these neurons. Given that IFNγ promotes the activation of the p53 pathway in cancer (Kim et al., 2009; Thiem et al., 2019), we hypothesized that the excessive exposure of IFNγ to motor neurons triggers the neuronal p53 activation, which in turn promotes the development of a characteristic ALS phenotype in motor neurons. Furthermore, we hypothesized that the studies of motor neurons exposed to IFNγ would reveal novel biomarker genes that could constitute potential immunotherapeutic targets for future ALS treatment.
In this study, we exposed the human iPSC–derived motor neurons to IFNγ for prolonged culture periods and examined the downstream impact on TDP-43 proteinopathy and electrophysiological impairment in these cells. We used both normal (WT) and TDP-43 mutant (TARDBPQ331K+/−) iPSC–derived motor neurons to compare the pathologic contributions of both the IFNγ treatment and ALS-causing mutation to the resulting phenotype. We also examined whether IFNγ induces the upregulation of p21 (CDKN1A), a direct transcriptional target of p53, in motor neurons (Abbas and Dutta, 2009). We used a single-cell transcriptomic analysis to investigate whether IFNγ exposure induces the mis-regulation of ALS-associated genes reported in other studies, including genes related to protein degradation, mitochondrial function, and neuromuscular junction (NMJ) maintenance (Maor-Nof et al., 2021; Namboori et al., 2021; Sturmey and Malaspina, 2022). We also investigated the role played by IFNγ in upregulating the programmed death-ligand 1 (PD-L1) expression in cultured motor neurons. Given that PD-L1 has been reported to significantly compromise the firing ability of sensory neurons in mice, our results imply a potential pathologic role of aberrant neuronal PD-L1 expression in neurodegenerative processes related to ALS. Collectively, our findings suggest that excessive IFNγ infiltration into the spinal cord may trigger a pathologic cascade of motor degeneration observed in the majority of ALS patients. These results provide potential immunotherapeutic targets to treat not only familial but sporadic forms of ALS, which account for roughly 90% of all reported cases.
Materials and Methods
Maintenance of human iPSC lines
WTC11 (WT, Q331K+/−) iPSCs were frozen in an mFreSR medium (StemCell Technologies) and stored under cryogenic conditions in liquid nitrogen. Ten centimeter dishes were treated with Matrigel (Thermo Fisher Scientific) diluted 1–60 in DMEM/F12 medium (Thermo Fisher Scientific) and incubated at 37°C/5% CO2 overnight. On the day of plating, vials of cells were removed from the liquid nitrogen storage and incubated in a 37°C water bath for 3 min. The content of the vial (1 ml) was then transferred to a 15 ml centrifuge tube and centrifuged for 3 min at 300 g. The supernatant was aspirated, and cells were resuspended in fresh mTeSR (StemCell Technologies) supplemented with 10 µM Y-27632 (Thermo Fisher Scientific), a specific inhibitor of the Rho kinase activity. Prepared Matrigel plates were then aspirated and washed with PBS containing Ca2+ and Mg2+ (Thermo Fisher Scientific), and the cell suspension was then plated evenly over the prepared culture surface. Y-27632 was removed from the culture medium the first day after plating, and cells were fed daily with fresh mTeSR from then on. The cells were incubated at 37°C/5% CO2 until the iPSC colonies filled the field of view when visualized using an Eclipse TS100 Microscope (Nikon) fitted with a 10× lens. At this point, the medium was aspirated and replaced with a TrypLE 1× solution (Invitrogen), incubated at 37°C/5% CO2 for 5 min. The detached cell colonies were dissociated with gentle trituration with a P1000 pipette, followed by adding the same volume of PBS with calcium and magnesium. The cell suspension was spun down and resuspended in fresh mTeSR with mild trituration to gently break up cell clusters. The cell suspension was then split across the desired number of Matrigel-coated plates and returned to the incubator. During the continued culture, any iPSC colonies displaying irregular boundaries, significant space between cells, or low nuclear-to-cytoplasmic ratios were carefully marked and removed from the culture using a fire-polished, sterile glass pipette.
Differentiation of iPSCs into spinal motor neurons
Human iPSCs were passaged onto Matrigel-coated six-well plates as described above and incubated at 37°C/5% CO2 in mTeSR until they reached ∼50% confluency. At this point, the cultures were differentiated into regionally unspecified neural progenitor cells (NPCs) using a monolayer differentiation method. Briefly, the undifferentiated Day 0 cells were treated with dual-SMAD inhibitors, SB431532 and LDN193189, as well as ascorbic acid (AA). CHIR99021 was added to the medium for the purpose of activating the wingless-related integration site pathway, which has been proven to enhance neuroepithelial differentiation and proliferation. On Day 5, the cultures were treated with all-trans retinoic acid (RA) for caudalization and the sonic hedgehog agonist purmorphamine, to give ventralization cues to the cells. On Day 11, the cells were passaged onto fresh Matrigel-coated six-well plates and treated with RA and AA until Day 18 to expand the neural progenitor population. These cells were then passaged at 100,000 cells/cm2 onto 0.01% poly-ʟ-ornithine (Sigma-Aldrich)- and 5 µg/ml laminin (Sigma-Aldrich)-coated surfaces and exposed to culture conditions promoting motor neuron differentiation. Specifically, the cells were fed with 10 ng/ml of neurotrophic factors (BDNF, GDNF, IGF-1 from R&D Systems, NT3 from StemCell Technologies) and AA (2.27 µM). During all stages of differentiation, a basal medium consisting of a 1:1 mix of Neurobasal medium and DMEM/F12, supplemented with B27, GlutaMAX, N2, nonessential amino acids, and penicillin–streptomycin, was used. All cells used in the described experiments were differentiated from WTC11 colonies between passages 45 and 55.
Cytokine treatment
WT and Q331K+/− mutant iPSC–derived neurons were exposed to seven different cytokines and treated from Day 23 to Day 25 of the ventrospinal neuron differentiation method described above. Recombinant human and mouse IFNγ, IFNα, IL-4, IL-6, IL-1β, IL-17, and TNFα were used at concentrations depicted in Supplementary Figure S3A. All of the cytokines were purchased from R&D Systems.
Coculture of CD8+ T-cells and iPSC-derived motor neurons
Whole blood was collected into 10 ml BD Vacutainer Plastic Collection Tubes with Sodium Heparin and inverted multiple times. To isolate peripheral blood mononuclear cells (PBMCs), we gently layered whole blood over a double volume of Ficoll in a Falcon tube and centrifuged for 30–40 min at 400 × g without brake. Four layers formed, each containing different cell types—the uppermost layer contained plasma, which was removed by pipetting. The second layer contained PBMCs, and these cells were gently removed using a pipette and added to PBS to wash off any remaining platelets. The pelleted cells were then counted, and the percentage viability was estimated using trypan blue staining. A 96-well cell culture plate was coated with a CD3-specific Ab (OKT3, eBioscience for human T-cells) and 5 μg/ml concentrations of PD-L1-Ig for control experiments. For the coculture of CD8+ T-cells and iPS-derived motor neurons, the culture plate was coated with poly-ʟ-ornithine, laminin, and a CD3-specific Ab sequentially. Human CD8+ T-cells were purified from PBMCs freshly isolated from whole blood using CD8a+ T Cell Isolation Kit II, human (Miltenyi Biotec). Purity was confirmed to be over 90% by flow cytometry. The T-cells were labeled with 1 μM CFSE, quenched by cold FBS, and incubated in plates coated with CD3-specific Ab. Human iPSC–derived neurons were cocultured with CD8+ T-cells purified from PBMCs at a 1:1 ratio (neurons:CD8+ T-cells) for 5 d in a culture medium with CD3 stimulation. On Day 5 after stimulation, the cells were stained with an Ab against CD8-APC and then analyzed for CFSE dilution by flow cytometry.
Flow cytometry
The following antibodies were used for the flow cytometry analysis: PD-L1 (29E.2A3, BioLegend), CD44 (C44Mab-5, BioLegend), and CD8 (SK1, BioLegend). The stained cells were analyzed using a BD FACSCalibur flow cytometer. For each sample, 10,000 events were recorded, and histograms were plotted. The percentages of cells were calculated using the FlowJo software (Treestar).
Immunocytochemistry
The cells were fixed in 4% paraformaldehyde for 15 min and permeabilized in 0.2% Triton X-100 solution, followed by blocking with 5% goat serum in PBS for 1 h at room temperature. The cells were then incubated with primary antibodies diluted in 0.5% BSA in PBS overnight at 4°C. The next day, the cells were washed three times with PBS. They were then incubated in a secondary antibody solution containing secondary antibodies diluted in 0.5% BSA in PBS overnight at 4°C. Counterstaining was performed with Vectashield containing DAPI (Vector Laboratories). Images were taken at the Garvey Imaging Core at the University of Washington's Institute for Stem Cell and Regenerative Medicine using a Leica SP8 Confocal System on an inverted microscope platform. Twelve bit 2,048 × 2,048 pixel images were acquired with the Leica LAS X software. The antibodies used in this study were as follows: mouse anti-Islet1 (1 in 200, Developmental Studies Hybridoma Bank), rabbit anti-β III tubulin (1 in 500, Sigma-Aldrich), rabbit anti-TDP43 (1 in 200, Invitrogen), Alexa Fluor 488–conjugated goat and anti-mouse secondary antibody (1:500, Invitrogen), Alexa Fluor 594–conjugated goat and anti-rabbit secondary antibody (1:500, Invitrogen), and Alexa Fluor 647–conjugated donkey and anti-goat secondary antibody (1:500, Invitrogen).
Electrophysiology
The population level function in motor neuron cultures was assessed using 48-well multielectrode arrays (MEAs) in conjunction with the Maestro MEA system (Axion Biosystems). Day 21 cells were plated on CytoView 48-well plates (Axion Biosystems) at the density of 100,000 cells per well. Plated cells were then subjected to a daily cytokine and blocking antibody treatment using a fresh medium from the next day onward. Electrophysiological recordings were taken every day from Day 22 to Day 40 for all experimental groups. During data acquisition, the standard recording settings for spontaneous neuronal spikes were used (Axis Software, version 2.5), and the cells were maintained at 37°C/5% CO2 throughout the 2 min recording period. The standard settings have 130× gain and record from 1 to 25,000 Hz, with a low-pass digital filter of 2 kHz for noise reduction. In all experiments, the spike detection was set at 5× the standard deviation of the noise, and the network burst detection was recorded if at least 25% of the electrodes in a given well showed synchronous activity. The reported results were calculated by averaging all of the electrodes in each well and then averaging data from duplicate wells. For the whole-cell patch-clamp experiments, the neurons were cultured with and without IFNγ treatment until Day 32. Recordings were performed on an inverted differential interference contrast microscope (Nikon) connected to an EPC10 patch-clamp amplifier and a computer running the Patchmaster software (HEKA). The cells on glass-bottom single-well dishes were loaded onto the microscope stage and bathed in a Tyrode's solution containing 140 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl2, 1 mM MgCl2, 10 mM glucose, and 10 mM HEPES. The intracellular recording solution (120 mM ʟ-aspartic acid, 20 mM KCl, 5 mM NaCl, 1 mM MgCl2, 3 mM Mg2+ATP, 5 mM EGTA, and 10 mM HEPES) was loaded into borosilicate glass patch pipettes (World Precision Instruments) with a resistance in the range of 2–6 MΩ. Offset potentials were nulled before the formation of a gigaΩ seal. Suction was then applied to disrupt the cell membrane in contact with the pipette tip. This process allowed electrical and molecular access to the intracellular space, and membrane potentials were then corrected by the subtraction of the liquid junction potential, calculated by Patchmaster. The current-clamp mode was used for recording the action potential behavior of neurons. Specifically, a 2 nA depolarizing single pulse was applied for 5 ms to induce single action potentials, and stepwise current injections of 10 pA from −30 to +70 pA for 500 ms were applied to trigger repetitive action potential firing. The action potential duration at 90% repolarization (APD90), depolarization speed, and repolarization speed were recorded for each neuron analyzed by the Patchmaster software. Inward and outward currents were evoked in voltage-clamp mode, by providing 500 ms depolarizing steps from −120 to +30 mV in 10 mV increments. To measure the resting membrane potential (RMP) of neurons in each group, we performed gap-free recordings of spontaneous activity in patched neurons in current-clamp mode with 0 pA current injection. The analyses of action potential waveforms and currents were performed using the Patchmaster software suite. All reagents used in this protocol were obtained from Sigma-Aldrich.
Quantitative real-time PCR
We extracted total RNA from cells using TRIzol (Invitrogen). After checking RNA purity and concentration, RNA was reverse transcribed to cDNA using the iScript SuperMix reagent (Bio-Rad Laboratories). Primers (Integrated DNA Technologies) were diluted in nuclease-free water with PowerUp SYBR Green Master Mix (Applied Biosystems), and quantitative PCR was performed on the Applied Biosystems 7300 machine. Relative RNA levels were calculated from cycle threshold values. The primer sequence for each gene is listed in Table 1.
Primer sequence for each gene
Single-cell RNA-seq
Single cells differentiated from WT and Q331K+/− mutant iPSC–derived neurons with or without IFNγ treatment were collected on Day 32 of differentiation, to generate a total of four independent single-cell libraries. Following a single-cell dissociation using TrypLE (Thermo Fisher Scientific), the cells were filtered through a 70 µm filter and spun down at 1,200 rpm for 5 min. The single-cell pellet was resuspended in neuron maintenance medium at a maximum concentration of 2,000,000 cells/ml, followed by measuring the cell viability using the Countess II automated cell counter (Invitrogen), which confirmed over 91% viability of cells in all groups. The cells were spun down at 1,200 rpm for 5 min again and resuspended in PBS + 0.04% BSA at a concentration of 15,000 cells/16.5 µl to target harvesting 10,000 recovered cells per library. cDNAs of single cells from each group were barcoded with 10x Genomics Chromium Next GEM Chip G Single Cell Kit and Chromium Next GEM Single Cell 3′ Kit v3.1, by following the manufacturer's protocol. Constructed libraries were loaded to the TapeStation (Agilent Technologies) to quantify the cDNA amount and confirm the fragmentation status, followed by the sequencing with the Illumina NextSeq 2000 at the Genomics Core with the University of Washington's Institute for Stem Cell and Regenerative Medicine. The sequencing was performed at an estimated read depth of 10,000 reads/cell. Sequenced FASTQ reads were initially processed using the CellRanger v3 software with settings recommended by the manufacturer (10x Genomics).
Transcriptomic data analysis
Read alignment, filtering, barcode counting, and unique molecular identifier (UMI) counting were performed using the CellRanger data processing software provided by 10x Genomics. The feature-barcoded matrix was obtained for further gene expression analysis with Seurat v4.0. The quantitative summary of sequencing was obtained from the CellRanger as follows: (1) The sequencing of single cells from the WT culture without IFNγ stimulation detected 1,911 genes per cell, 4,105 UMI per cell, and 21,312 mean reads per cell. (2) The sequencing of single cells from the WT culture with IFNγ stimulation detected 1,833 genes per cell, 3,726 UMI per cell, and 16,427 mean reads per cell. (3) The sequencing of single cells from the Q331K+/− mutant culture without IFNγ treatment detected 1,651 genes per cell, 3,443 UMI per cell, and 18,452 mean reads per cell. (4) The sequencing of single cells from the Q331K+/− mutant culture with IFNγ treatment detected 1,858 genes per cell, 3,803 UMI per cell, and 20,319 mean reads per cell. For the comparative gene expression analysis, the integration of single-cell data described by Stuart et al. was applied on the basis of the Seurat algorithm. Before applying the integration, cells with low UMI counts, doublets, and relatively high mitochondrial DNA content were removed. The cleared datasets were subsequently normalized in order to correct for differences in read depth and library size, using Seurat's “LogNormalize” function, which divides feature counts of the cell by its total counts, followed by multiplying the scale factor. Then, the integration method included in Seurat v4 aligned the shared cell populations across multiple datasets to identify cells that are in a matched biological state in each library, which allows comparative gene expression analysis across the selected libraries to be performed. The principal component analysis was performed with the integrated data post normalization, followed by running the “RunUMAP” function to generate two-dimensional Uniform Manifold Approximation and Projection (UMAP) using the top principal components detected in the dataset. The “FindConveredMarkers” function in Seurat identified conserved cell type markers for generating the feature plots with the “FeaturePlot” function. The clusters were annotated by analyzing the list of conserved marker genes in each cluster with the gene ontology (GO) database, the Human Protein Atlas, and the PanglaoDB. Differential gene analysis was performed by using the “FindMarkers” function in Seurat, and the result was displayed in volcano plots, heatmaps, and violin plots to visually highlight the contrast in gene expression across the selected transcriptomic datasets.
Statistical data analysis
All experiments were performed at least in triplicate and repeated using 2–3 independent differentiation runs for WTC11 neurons except for the single-cell RNA-seq. Significant differences between groups were evaluated using unpaired t tests for two conditions, or one-way ANOVA, with post hoc tests for multiple comparisons, for experiments with three or more groups. Mann–Whitney U tests and ANOVA on ranks were used to analyze the statistical significance of differences arising between sets of non-normally distributed data. In all experiments, a p value of <0.05 was considered significant. All statistical tests were performed using the GraphPad Prism statistics software.
Results
Motor neurons differentiate successfully from both normal and TARDBP mutant iPSCs
We established a small-molecule–based differentiation protocol using the WTC11 iPSC line (WT) to generate human ventrospinal motor neurons, essentially as described by Du et al. (2015) and Smith et al. (2021, 2022). Furthermore, we previously employed CRISPR-Cas9–mediated gene editing techniques to introduce a heterozygous (+/−), pathogenic Q331K mutation into the TARDBP gene locus of the WTC11 iPSC line, creating an ALS (TDP-43 mutant) line and an isogenic control pair (Smith et al., 2021). These iPSCs were induced to generate human ventrospinal motor neurons with WT and TARDBPQ331K+/− (Q331K+/−) genotypes using the protocol summarized in Figure 1A. We confirmed that both cultures exhibited high levels of CD271 expression, a positive marker for NPCs of the subventricular zone, on Day 21. These cells were also found to express negligible amounts of CD44, a marker for non-neuronal cells in this differentiation lineage, at the same time point (Fig. 1B). By Day 27, the differentiated cells from each iPSC group expressed both pan-neuronal (TuJ1) and spinal motor neuron–specific (ISL-1) markers, confirming a successful motor neuron differentiation from both genotypes (Fig. 1C).
WT and TARDBPQ331K+/− iPSC are differentiated into motor neurons expressing pan-neuronal and motor neuron–specific marker(s). A, Differentiation schematic illustrating the time-course treatment of small-molecule cocktails for promoting motor neuron differentiation from iPSCs. B, Flow cytometry data of Day 21 cultured NPCs from each group. Cells were stained with antibodies against human nerve growth factor receptors CD271 and CD44. C, Immunocytochemistry data of Day 27 cultures from each group, showing the expression of motor neuron–specific (ISL-1) and pan-neuronal (TuJ1) markers.
IFNγ induces p21 expression and TDP-43 mis-localization in ventrospinal motor neurons
Given that the increased expression of p53 downstream targets, such as p21, constitutes a common characteristic of various familial and sporadic ALS cases (Ziff et al., 2023), we assessed whether the exposure of motor neurons to IFNγ induced p53 pathway gene upregulation. Interestingly, we observed that IFNγ exposure significantly increased p21 expression in WT motor neurons, but did not further promote p21 expression in Q331K+/− mutant motor neurons. Instead, we confirmed that the expression of p21 was constitutively increased during the later stages of Q331K+/− motor neuron culture, which aligns with the previous findings that abnormal expression of TDP-43 potentiates p53 pathway activation (Vogt et al., 2018; Ziff et al., 2023; Fig. 2A). Next, we assessed whether the expression of mutant TDP-43 and/or continuous IFNγ treatment can cause TDP-43 proteinopathy, a defining ALS phenotypic characteristic observed in ∼97% of ALS patient's spinal cord tissues. At all timepoints tested, expressing mutant TDP-43 did not induce noticeable cytoplasmic translocation of TDP-43 protein in motor neurons per se. However, nucleic depletion and cytoplasmic aggregation of TDP-43 were strongly induced by prolonged IFNγ treatment (Fig. 2B), and this effect was more pronounced in TARDBP mutant motor neurons than in WT controls (Fig. 2C).
IFNγ triggers the p21 upregulation and TDP-43 proteinopathy in iPSC-derived motor neurons. A, qRT-PCR analysis for each experimental group quantifying the time-dependent expression of p21, the early transcriptional target of p53 signaling. UT, untreated control group; IFNγ, 5 ng/ml IFNγ-treated group. Flow of hypothesis for this study (right panel). B, Immunocytochemistry assay of WT and Q331K+/− mutant cultures at Days 25, 32, and 40, showing a significant cytoplasmic translocation and aggregation of the TDP-43 protein in response to the IFNγ stimulation at the later phases of both cultures. C, Quantitative analysis of the cytoplasmic TDP-43 translocation for each experimental condition (error bars indicate SD, n = 5 biological replicates; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant).
Exposure to IFNγ results in hyperactivation of motor neurons, followed by a complete loss of firing function
Next, we investigated whether the IFNγ exposure affected the electrophysiological function of motor neurons. We assessed the firing frequency of both WT and Q331K+/− mutant motor neurons using MEAs, which enable the recording of spontaneous field potential activity from populations of cultured neurons in a nondestructive and noninvasive manner. We treated neurons daily with three different concentrations of IFNγ (0.5, 5, 50 ng/ml) from Days 23 to 40 post neuronal induction while recording the population's firing activity across the same culture period. The time-course analysis of the mean firing rate shows a distinct difference in the firing pattern between untreated and IFNγ-treated neurons. Untreated neurons showed a gradual increase in the firing rate as they matured for both the normal and mutant groups and maintained firing rates above 5 Hz throughout the culture period from Day 27 onward. However, neurons treated with IFNγ above 5 ng/ml showed a drastic increase in the firing rate from Day 27, reaching an apex higher than 30 Hz by Day 32, followed by a sharp decline in the firing rate subsequently. Consequently, both the WT and Q331K+/− mutant cells exposed to IFNγ exhibited a complete loss of firing ability by Day 38, before the end of the assay period (Fig. 3A,B). While the IFNγ treatment was the major driver of this aberrant firing behavior, we also observed a significant difference in firing patterns between IFNγ-untreated WT and mutant neurons. Notably, the firing frequency of the mutant neurons was significantly lower than that of the WT neurons at the later phase of the recording (Fig. 3C).
IFNγ treatment induces pathological electrophysiological changes in iPSC-derived motor neurons. A, Plots illustrating an average weighted mean firing frequency of cultured neurons over time. High-concentration IFNγ-treated neurons show a distinct increase in firing rate during the midphase, followed by an acute silencing in the late phase of culture. B, Representative field potential traces from each experimental condition, illustrating an abnormal electrophysiological progression in response to IFNγ treatment. C, Collective summary of phase-dependent neuronal electrophysiology for weighted mean firing rate. D, Whole-cell patch-clamp data showing the RMP of Day 32 neurons from each group. E, % cells showing none, single, adaptive, and repetitive firing behavior in response to 500 ms depolarizing current injection. Single, nonrecurring firing; adaptive, repetitive firing with diminishing action potential amplitude in 500 ms recording period; repetitive, repeated firing without showing compromised action potential amplitude in 500 ms recording period (error bars indicate SD, n = 3 biological replicates for MEA, n = 12 cells collected across three biological replicates per condition for whole-cell patch-clamp experiments; *p < 0.05, **p < 0.01, ****p < 0.0001; ns, not significant).
Next, we compared the intrinsic neuronal properties of the four different experimental groups, as well as their responses to electrical stimulation, via the whole-cell patch clamp. The data revealed that both the TARDBP mutation and IFNγ exposure induced an abnormal depolarization of the motor neuron's RMP, although the effect of IFNγ was not significant in the mutant group, which already showed compromised RMP (Fig. 3D). Additionally, we found that the expression of the TARDBP mutation highly decreased the percentage of cells capable of firing repetitively in response to depolarizing current injection, whereas the IFNγ treatment did not cause a comparable level of impairment (Fig. 3E). No significant change in action potential duration, amplitude, depolarization velocity, or repolarization velocity was observed in response to either the TARDBP mutation or IFNγ treatment (data not shown).
Taken together, both TARDBP mutation and IFNγ treatment affected the electrophysiological function of cultured motor neurons. The data suggest that the influence of IFNγ treatment on iPSC-derived motor neuron function was more pronounced compared with that caused by mutation of the TARDBP gene, in terms of the strength of its effect on firing patterns, whereas the mutation led to a more unstable resting state in cultured cells. We found that WT motor neurons were similarly vulnerable to continuous IFNγ-mediated inflammatory stress, strongly suggesting that IFNγ could be a critical pathogenic driver of motor deficits in sporadic forms of ALS.
Exposure to IFNγ induces significant changes in the transcriptomic profile of iPSC-derived motor neurons, potentially contributing to the development of ALS
To gain insight into the potential mechanisms responsible for the IFNγ-mediated functional impairment and TDP-43 proteinopathy, we characterized the transcriptional changes induced by the exposure of motor neurons to IFNγ. WT and Q331K+/− mutant neurons were treated with IFNγ (5 ng/ml) for 10 d (Days 22–32 post neuronal induction), directly following the neurogenic phase of differentiation, and subjected to single-cell RNA-seq on the last day of stimulation. We generated four single-cell libraries defined by the mutation status and IFNγ treatment (Fig. 4A). A population analysis with cell type-specific markers showed that motor neurons constituted 37.7 and 40.0% of WT and Q331K+/− mutant populations, respectively. The motor neuron clusters expressed well-established markers of motor neurons, and the majority of non-neuronal cells in both groups were early glial progenitors (Fig. 4B,C). Overlaid UMAP with spinal motor neuron populations from each pair of comparisons showed that either IFNγ treatment or TARDBP mutation significantly altered the overall transcriptomic profile of neurons (Fig. 4D).
Population analysis and validation of motor neuron differentiation using single-cell RNA-seq. A, Schematic of the single-cell RNA-seq library construction. B, UMAP of WT and Q331K+/− culture on Day 32 shows bifurcation of population into a neuronal group and an early glial population. C, Expression of motor neuron markers in each group. D, Overlayed UMAP of motor neuron populations showing the transcriptomic profile differences in each pairwise comparison.
For more detailed gene expression analysis, we generated subgroups that isolated cells with a spinal motor neuron phenotype from each library and listed the average expression values of all transcripts detected in these cells, specifically. We then performed a comparative analysis of differential gene expression between groups by pairing the spinal motor neuron populations in each library, with a particular focus on investigating their transcriptomic change under the IFNγ exposure (Fig. 5A). Motor neuron stimulation by IFNγ generated 828 (WT) and 506 (Q331K+/−) differentially expressed genes (DEGs) with log2FC values >0.25 compared with each of the untreated controls (Fig. 5B). We sought to ascertain whether the exposure to IFNγ promoted the expression of ALS-related transcripts and, if so, how the expression trend differed between WT and Q331K+/− motor neurons. We compared the identified DEGs induced by the IFNγ treatment individually with the list of ALS risk genes included in the GO term “ALS” in the KEGG pathway analysis. We found 54 DEGs and 15 DEGs associated with ALS pathology in the IFNγ-stressed WT and Q331K+/− motor neurons, respectively. The result indicates that the IFNγ exposure may drive both WT and Q331K+/− mutant motor neurons toward an ALS pathological state (Fig. 5C). Functional GO analysis on the DEGs identified in each group showed a strong focus on metabolic function, suggesting the IFNγ exposure may exert its pathological impact through driving metabolic malfunction in motor neurons. The IFNγ-induced DEGs in WT motor neurons showed an additional functional relationship to ATP synthesis and endocytosis, whereas distinct DEGs only derived from Q331K+/− mutant motor neurons with the IFNγ treatment did not fall into those functional GO terms (Fig. 5D). We provided a list of all of the DEGs derived by the IFNγ treatment in each group of spinal motor neurons (Fig. 6).
Transcriptomic profiles of both WT and Q331K+/− mutant motor neurons are significantly altered by IFNγ treatment. A, Schematic of the single-cell RNA-seq libraries for comparison. B, Volcano plots showing the DEGs detected after IFNγ treatment. The diagram below the volcano plots shows the number of upregulated and downregulated genes for each group after IFNγ treatment. C, A Venn diagram showing the number of ALS-associated DEGs in each group. (D) The GO terms show the functional categories to which the ALS-associated DEGs are targeted. The numbers indicate the number of DEGs included in each GO term.
List of ALS-associated DEGs generated by 10 d of IFNγ treatment (5 ng/ml) to WT and Q331K+/− motor neurons. The bold horizontal line separates the DEGs upregulated by the IFNγ treatment (top) and downregulated by the IFNγ treatment (bottom). Common genes found from both WT and Q331K+/− mutant motor neuron populations are highlighted in either red or cyan as indicated.
Although quantitatively less impactful, we found the Q331K+/− mutation mis-regulated the expression of 325 genes in motor neurons, implying the mutation per se may still impact the motor neuron physiology (Fig. 7A,B). Among those DEGs, we found 19 ALS-associated genes mis-regulated (Fig. 7C). There was a relatively little difference in the gene expression between the IFNγ-treated WT and IFNγ-treated Q331K+/− mutant motor neurons (a total of 13 DEGs), and none of them were previously reported to be associated with ALS pathology (data not shown).
Differential gene expression analysis between WT and Q331K+/− mutant motor neurons without IFNγ treatment. A, Schematic showing the groups being compared. B, A volcano plot and diagram showing the total number of DEGs derived by the Q331K+/− mutation. C, List of DEGs associated with ALS pathology obtained from the same comparative dataset. D, GO analysis result with the upregulated and downregulated DEGs in total due to the Q331K+/− mutation. E, GO terms related to ALS-associated genes downregulated in the mutant group. F, Proteasome genes showing decreased expression with the Q331K+/− mutation.
We also investigated whether the DEGs induced solely by the Q331K+/− mutation have functional implications related to neurodegeneration. The DEGs upregulated in the mutant group were mainly related to cytoplasmic translation, response to muscle stretch, and glycolytic processes, whereas the genes downregulated by the mutation were related to the negative regulation of mitochondrial protein localization, cytoskeletal assembly, and axonal transport of the mitochondria (Fig. 7D). The ALS-associated DEGs are thought to have main roles in amino acid metabolic processing and proteasomal activity. The genes encoding proteasomal proteins, such as PSMB6, PSMB7, and PSMC5, were significantly downregulated in the mutant group, implying the mutation may impair motor neurons’ amino acid metabolic processing and proteolytic activities, which may lead to the accumulation of damaged or misfolded protein species (Fig. 7E,F).
In addition to the IFNγ-induced DEGs listed in the GO term “ALS” in the KEGG pathway analysis, we identified critical genes that could have a strong impact on motor neuron degeneration that were explicitly mis-regulated by IFNγ treatment. Specifically, we found that IFNγ triggered a significant increase in BST2, PRPH, and PTN expression and a decrease in CRABP1 (Fig. 8A). The similar trend of those gene mis-regulations was reported to be associated with neuroinflammation, protein aggregation, and NMJ breakdown in animal models, which we will discuss more in the following section.
The expression of selected DEGs associated with neurodegeneration and immune checkpoint proteins under IFNγ exposure. A, A set of DEG expression profiles for genes related to neurodegeneration, including data from single-cell RNA-seq and qRT-PCR for verification. B, Expression of immune checkpoint proteins in motor neurons exposed to IFNγ and/or in the presence of a TARDBP mutation. C, Time-course validation of PD-L1 mRNA expression in each group of motor neurons by qRT-PCR (error bars indicate SD, n = 3 biological replicates for qRT-PCR; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant).
Motor neurons exposed to INFγ exhibit aberrant increases in PD-L1 expression but no other immune checkpoint proteins
Our transcriptomic analysis showed an aberrant increase in only one immune checkpoint protein, PD-L1. Immune checkpoint proteins have been intensively studied in the cancer immunotherapy field in order to understand how tumors regulate the immune cell response by binding to their cognate receptors on immune cell surfaces. While many cancers under immune attack usually coexpress immune checkpoint proteins such as PD-L1, B7-H3, GAL9, B7-H4, PD-L2, CTLA4, and TNFRSF14 on their surfaces, motor neurons exposed to INFγ exclusively expressed PD-L1 (CD274). Under our experimental conditions, other immune checkpoint proteins were not responsive to IFNγ stimulation (Fig. 8B). We confirmed the dramatic increase of PD-L1 expression after the IFNγ stimulation through the time-course qRT-PCR assay at Days 25, 32, and 40 of neurons in both genotypes (Fig. 8C). Moreover, IFNγ was found to be particularly potent for eliciting neuronal PD-L1 expression in motor neurons, whereas other neuroinflammatory cytokines, such as IFN-α, IL-4, IL-6, IL-1b, IL-17, and TNF-α, did not affect neuronal PD-L1 expression (Fig. 9A), despite the neurons expressing receptors of those cytokines except for interleukins (Fig. 9B). Does PD-L1 in motor neurons have a canonical function as a suppressor of cytotoxic T-cell attack as it does in many cancer cell types? To address this question, we isolated T-cells from a normal donor's blood sample and labeled them with a carboxyfluorescein succinimidyl ester, which covalently attaches a fluorescent label to the amines of cellular proteins. IFNγ-treated neurons were cocultured with CFSE-labeled CD8+ T-cells for 5 d in the presence of anti-CD3, a marker of T-cell proliferation. Labeled T-cells without neurons in the presence of anti-CD3 or in the presence of both anti-CD3 and PD-L1-Fc proteins were used as controls (Fig. 10A). Exogenous treatment of PD-L1-Fc inhibited the proliferation of CD8+ T-cells by 37%. However, IFNγ-treated neurons reduced CD8+ T-cell proliferation by just 8.8% (WT) and 15% (Q331K+/−). The result suggests that neuronal PD-L1 is only modestly immunosuppressive, implying that neuronal PD-L1 may have another physiological role under the action of IFNγ (Fig. 10B).
Expression levels of PD-L1 in response to various inflammatory cytokines and the mRNA expression level for each cytokine's receptors in WT and Q331K+/− motor neurons. A, Flow cytometry analysis shows the level of PD-L1 expression in response to various types of proinflammatory cytokine treatment. The histogram shows a representative raw dataset, and the bar graph summarizes the PD-L1 expression level from repeated experiments. B, Single-cell RNA-seq data showing the relative expression level of each cytokine's receptor on motor neurons. Receptor and cytokine pair: IFNAR1, IFNAR2 for IFNα, IFNGR1, IFNGR2 for IFNγ, TNFRSF1A for TNFα, IL4R for IL4, IL6R, GP130 for IL6, IL17RA, and IL17RB for IL17 (error bars indicate SD, n = 4 biological replicates for flow cytometry; ****p < 0.0001).
Suppression of CD8+ T-cell proliferation by coculturing with IFNγ-treated neurons. A, Schematic of the cells and antibodies used (left panel). CFSE-labeled CD8+ T-cells were stimulated by incubating them in anti-CD3-coated plates for 5 d, in the absence or presence of IFNγ-treated iPSC-derived neurons. Quantification of CFSE-labeled CD8+ T-cell proliferation is depicted in each histogram (right panels). B, The result summarized in the bar graph showing % of proliferating T-cells.
Discussion
Prior studies have revealed that neuroinflammation, TDP-43 proteinopathy, and p53 pathway activation are important hallmarks of ALS pathology (Yuan et al., 2008; Maor-Nof et al., 2021; Ziff et al., 2023). In this study, we sought to explore how these features might be mechanistically linked, hypothesizing that the effects of inflammatory cytokines associated with neuroinflammation might lie upstream of TDP-43 dysfunction and p53 pathway activation. Our results demonstrate that an ALS-associated TDP-43 mutation enhances the expression of the p53 pathway effector, p21(CDKN1A), confirming a prior study that placed TDP-43 dysfunction upstream of p53 pathway activation. More importantly, we have demonstrated that prolonged exposure of motor neurons without TDP-43 mutations to IFNγ induced TDP-43 proteinopathy as well as p53 pathway activation as indicated by enhanced p21 expression. Thus, our findings suggest that neuroinflammatory cytokine exposure, TDP-43 dysfunction, and p53 pathway activation are sequentially linked in ALS etiology.
How might such an excessive exposure of motor neurons to IFNγ occur? CD8+ T-cells are a major source of IFNγ secretion, although NK cells and resident immune cells in the CNS, such as astrocytes and microglia, can also produce IFNγ (Murray et al., 2002; Coque et al., 2019; Kann et al., 2022). A T-cell population resides constitutively associated with the choroid plexus, where it may secrete IFNγ into the cerebrospinal fluid (CSF; Goverman, 2009; Kann et al., 2022). Importantly, the CSF of ALS patients exhibits an expansion of a T-cell subpopulation expressing eomesodermin, which drives the IFNγ expression (Yazdani et al., 2022). IFNγ is permeable to the blood–spinal barrier, particularly in the cervical region, implying that T-cells impacted by peripheral inflammation might expose motor neurons to IFNγ (Pan et al., 1997; Bartanusz et al., 2011; Kunis et al., 2013). Moreover, Bonney et al. (2019) showed that IFNγ even promotes its own permeability to the blood–brain barrier, implying that uncontrolled IFNγ secretion may broadly affect the physiology of the entire CNS. Finally, peripherally projecting motor axons lie outside the blood–brain barrier. The exposure of sympathetic axons to IFNγ mediates axonal STAT1 phosphorylation and the transport of these activated transcription factors to the somatic nucleus (Song et al., 2016). If a similar retrograde IFNγ axonal signaling pathway exists in motor neurons, T-cell–generated IFNγ circulating peripherally might affect motor neurons.
We showed that IFNγ exposure in iPSC-derived motor neurons triggers the onset of a characteristic ALS phenotype, including TDP-43 proteinopathy and electrophysiological impairment, even in WT neurons. This implies that IFNγ-induced stress may play a critical role in promoting the development of sporadic ALS following the potentially multiple bouts of upstream immunological events. To compare the IFNγ-induced phenotype with that caused by TDP-43 mutation, which also regulates the p53 pathway, we used TDP-43 mutant (Q331K+/−) iPSC–derived motor neurons in parallel with WT motor neurons for every assay. Motor neurons harboring the Q331K+/− mutation did not exhibit the characteristics of symptomatic ALS patients, such as cytoplasmic TDP-43 mis-localization or acute motor neuron death. Although mutant neurons at the late culture stage exhibited compromised neuronal firing function and depolarized RMP, this limited phenotype suggests that the TARDBP mutation per se may not be sufficient to drive ALS pathology in iPSC-derived motor neurons. Interestingly, neither the relative immaturity of iPSC-derived motor neurons nor the absence of familial TARDBP mutation prevented the initiation of TDP-43 proteinopathy once the cells were subjected to prolonged IFNγ treatment. IFNγ-treated motor neurons, regardless of the presence or absence of a TARDBP mutation, showed a dramatic increase in TDP-43 translocation from nuclei to cytoplasm. However, the phenotype was more severe in the mutant motor neurons, implying the potential role of TARDBP mutations in exacerbating cytoplasmic TDP-43 aggregation in response to IFNγ stimulation. This result implies that mis-localization of TDP-43 may require an upstream trigger beyond the mutation of the TARDBP gene, but does not discount the possibility that mutations in the TARDBP contribute to the development of a pathologic phenotype. Given that the absence of nuclear TDP-43 promotes p53 activation, this acute cytoplasmic TDP-43 translocation suggests that signaling downstream of IFNγ mediates TDP-43 proteinopathy and may constitute an attractive upstream target for ALS treatment.
We also showed that IFNγ significantly damages the electrophysiological function of neurons. IFNγ treatment triggered an explosive activation of motor neurons starting at the midphase of culture in both WT and Q331K+/− mutant groups, followed by their complete loss of function. However, we did not see evidence of excitotoxic behavior or noticeable cell death after IFNγ treatment in either WT or Q331K+/− mutant neurons, which is one of the early hallmarks of ALS pathology, suggesting that signaling pathways downstream of IFNγ did not immediately activate cell death signals in our cultured cells. This unique impairment pattern was not observed in mutant neurons without the presence of IFNγ, although the mutant neurons still exhibited a compromised firing rate. Together, these results suggest that IFNγ may induce electrophysiological damage in neurons by a mechanism different from that initiated by TARDBP mutations. Furthermore, we found that the TARDBP mutation has a more profound effect on increasing neurons’ RMP than the IFNγ treatment, supporting the idea of a distinct pattern of electrophysiological damage caused by the two different factors. Although mechanistic understanding is short, this observation raises the need for a different approach to understanding electrophysiological loss in ALS patients based on the type of upstream pathologic drivers.
Our single-cell RNA-seq analysis revealed that IFNγ triggers a significant transcriptomic alteration for both WT and Q331K+/− mutant neurons that trend toward the development of ALS-like and neurodegenerative phenotypes. Among the DEGs arising from IFNγ treatment (828 for WT, 506 for Q331K+/−), we identified 54 (WT) and 15 (Q331K+/−) ALS-associated genes, based on the KEGG pathway GO analysis. The functional GO analysis of those ALS-associated DEGs revealed that the aberrantly expressed genes are mostly related to either metabolic function or the host–virus interaction of neurons. The majority of abnormally expressed genes in IFNγ-stressed motor neurons encode proteins expressed in the inner membrane of the mitochondria that contribute to the respiratory chain and electron transport. Since the mitochondrial respiratory chain drives ATP synthesis, the dysregulation of this process may impair energy production and ROS regulation in motor neurons. Moreover, there is mounting evidence that mitochondrial dysfunction in ALS tissue contributes to disease progression through impaired respiration, protein import, mitochondrial transport, and neuronal homeostasis (Mehta et al., 2019; Choi et al., 2020; Zhao et al., 2022). It is notable that WT motor neurons under IFNγ stress recapitulated the mis-regulation of mitochondrial genes found in those mutant ALS models. The mitochondrial damage in response to the IFNγ treatment was evidenced by the increase in the expression of the OPTN in both types of motor neurons, which encodes a primary mitophagy receptor, optineurin, that plays a central role in degrading damaged mitochondria (Evans and Holzbaur, 2020). An important class of mitochondrial dysfunction that is unique to neurons concerns the axonal transport of mitochondria. To support axon function, the mitochondria continuously undergo anterograde and retrograde transport to supply energy, control ROS production, maintain calcium dynamics, and regulate synaptic function at distal sites (Hung, 2021; Zhao et al., 2022). As motor neurons have exceptionally long axons, they are particularly vulnerable to defects in mitochondrial transport in axons. Thus, the importance of maintaining mitochondrial homeostasis for motor neuron survival and function suggests that continuous exposure to IFNγ may trigger a pathologic cascade of neurodegeneration, initiated by mitochondrial dysfunction.
Although a smaller number of genes were mis-regulated in Q331K+/− mutant motor neurons without IFNγ treatment, we still found 19 altered ALS-associated DEGs (out of 325 DEGs) in these cells, and the majority of these 19 were downregulated. Interestingly, the expressions of pivotal genes that contribute to the proteasomal function of neurons were significantly reduced in mutant neurons. PSMB6, PSMB7, and PSMC5 are major subunits that contribute to the complete assembly of the proteasome complex (Gomes, 2013). Since proteasomes are an essential component of intracellular protein homeostasis mechanisms, compromised proteasome function may lead to reduced proteolytic activities, which could make cells susceptible to cytotoxic protein accumulation. This result suggests that Q331K+/− mutation may disrupt the homeostatic balance between neuronal protein production and degradation that could lead to significant damage in postmitotic motor neurons, which cannot dilute out accumulated protein aggregates during cell division. Given that more severe TDP-43 proteinopathy was observed in the IFNγ-stressed mutant neurons than the WT counterparts, the reduced expression of those proteasomal genes in mutant neurons may help explain the exacerbated cytoplasmic TDP-43 accumulation in response to the IFNγ exposure.
In addition to the ALS-associated DEGs mainly related to the metabolic function, IFNγ induced an aberrant expression of neurodegenerative genes that play critical roles in neuroinflammation (BST2, PTN), protein aggregation (PRPH), and NMJ breakdown (CRABP1).
It was of particular interest that IFNγ induces an abnormal expression of a gene responsible for regulating NMJ development and maintenance. CRABP1 is an RA-binding protein that was recently found to regulate motor function via the maintenance of the NMJ through the CRABP-CaMKII-Agrin axis. Lin et al. (2022) showed that CRABP1 knock-out mice exhibited adult-onset ALS-like phenotypes as characterized by behavioral, electrophysiological, and histological assays. The CRABP1-deficient mice showed significantly impaired NMJs, accompanied by progressive motor axon degeneration, which was rescued after the re-expression of the gene. It is notable that, in our hands, CRABP1 was one of the most downregulated genes in both genotypes of motor neurons following exposure to IFNγ. Given the importance of CRABP1 expression in maintaining NMJ and motor function, our result indicates that excessive immunological activity triggered by IFNγ exposure might contribute to the development of sporadic forms of ALS via direct targeting of the neuromuscular synapse.
Another interesting phenomenon observed in motor neurons exposed to IFNγ was a significant increase in PD-L1 expression (at both the RNA and protein level) but no other immune checkpoint proteins. PD-L1 is a well-preserved transmembrane protein across different types of cells and a verified immune checkpoint target for cancer immunotherapy. The increased expression of PD-L1 on several types of cancer cells conditionally suppresses immunity by binding to its cognate receptor, PD-1, on immune cells (Patel and Kurzrock, 2015; Doroshow et al., 2021). Although the expression level of immune checkpoint proteins in many other cell types typically increases upon immune attack, a selective increase of PD-L1 in neurons may implicate its potential involvement in ALS-related neurodegenerative mechanisms. Furthermore, the degree of PD-L1 protein increase in response to IFNγ stimulation (4–5-fold increase) greatly exceeds that which typically occurs in cancer cells (30–60% increase than normal cells), which tend to coexpress with other immune checkpoint proteins under IFNγ exposure (Patel and Kurzrock, 2015; Doroshow et al., 2021). The results collectively suggest a potential pathogenic action of excessive PD-L1 expression in motor neurons upon inflammatory stress. Although the downstream signals of PD-L1 in neurons are very understudied, this theory may be supported by recent discoveries showing that exogenous PD-L1 treatment in normal mice significantly impairs neuronal firing behaviors. Chen et al. (2017) showed that PD-L1 treatment to peripheral sensory neurons damages their firing ability in a dose-dependent manner through interaction with its cognate receptor, PD-1 on the same cells, implying a potential role for PD-L1 in regulating neuronal function.
Although current immunotherapeutic approaches employing immune checkpoint proteins have been highly focused on rejuvenating immunity for cancer treatment, mounting evidence suggests that uncontrolled immunity could be responsible for the rapid neurodegeneration in the course of ALS progression (Coque et al., 2019; Campisi et al., 2022; Yazdani et al., 2022). Thus, further investigation of the pathologic role of PD-L1 overexpression on the neuronal membrane is warranted to better understand how its neuron-specific downstream signaling may alter neurophysiology in response to the IFNγ-mediated inflammatory stress.
We believe these results raise important questions, the answers to which could lead to a better understanding of the pathogenic role of uncontrolled immunity-mediated neurodegeneration, which could be a key component driving rapid disease progression in ALS. First, it is currently unclear whether the inflammatory reaction occurs at the axon terminal and/or the NMJ level, the periphery of the patient's body, before spreading the pathologic effect to neuronal cell bodies in a retrograde manner. Second, speculative mechanisms mentioned here linking particular DEGs to IFNγ-mediated protein aggregation and abnormal firing patterns need to be experimentally tested. Third, it is still unclear whether p53 pathway activation, leading to p21 upregulation, is directly involved in ALS phenotype expression induced by IFNγ. Finally, it would be valuable to investigate the neuron-autonomous pathologic role of PD-L1, besides its immunomodulatory function.
Conclusion
Our findings suggest that IFNγ is a potent driver of p21 expression in motor neurons, which also triggers TDP-43 proteinopathy and electrophysiological impairment, as well as neurodegenerative transcript expression in both normal and TARDBP mutant iPSC–derived human motor neurons. We also confirmed that the Q331K+/− mutation in the TARDBP gene directly affects neuronal firing ability and induces p53 pathway expression, especially at the late stages of culture. However, the pathologic outcome triggered by IFNγ treatment was much more severe and extensive in terms of driving the adoption of an ALS-like phenotype. Both WT and Q331K+/− mutant iPSC–derived motor neurons exhibited dramatic decays in firing frequency, coupled with an increase in cytoplasmic TDP-43 aggregation, following IFNγ treatment. Importantly, IFNγ treatment not only induced a significant increase in inflammatory gene expression but also altered the motor neuron's transcriptomic profile toward an ALS-like state, presumably via damaging motor neuron's mitochondrial function. Additionally, we found that continuous exposure of motor neurons to IFNγ downregulated an essential gene for NMJ maintenance, which implies a new role for IFNγ in driving the deterioration of the synaptic connections between motor axons and skeletal muscle. Finally, we highlighted an exclusive increase of PD-L1 among immune checkpoint proteins following IFNγ treatment, implying an unknown role for PD-L1 in neurophysiology and ALS development. In all, our study provides strong evidence for a pathogenic link between IFNγ and subsequent spinal motor neuron degeneration that mirrors ALS pathology and a comprehensive list of mis-regulated genes in motor neurons in response to IFNγ exposure. These results suggest that immunomodulatory compounds may constitute strong targets for ameliorating symptoms, especially in sporadic forms of ALS.
Footnotes
This work was funded by NIH R03 TR004009 (A.S.T.S.) and supported by a Predoctoral Fellowship (C.C.) from the Institute for Stem Cell and Regenerative Medicine (ISCRM) at the University of Washington. Additional funding was provided by a Sponsored Research Agreement from Curi Bio, Inc. and a philanthropic gift from the Eileen & Larry Tietze Foundation (both awarded to D.L.M.) and by an Elsa U Pardee Foundation and MCC Patient Gift Fund (J.H.L.) and NIH P01 CA225517 (P.T.N). The single-cell RNA sequencing (RNA-seq) work performed in this study was supported by the ISCRM Genomics Core. We thank Dr. Mary C. Regier for her assistance with the single-cell library preparation for the sequencing experiment.
↵*C.C. and J.H.L. contributed equally to this work.
The authors declare no competing financial interests.
- Correspondence should be addressed to David L. Mack at dmack21{at}uw.edu or Alec S. T. Smith at astsmith{at}uw.edu.
