Neuronal STING activation in amyotrophic lateral sclerosis and frontotemporal dementia

The stimulator of interferon genes (STING) pathway has been implicated in neurodegenerative diseases, including Parkinson’s disease and amyotrophic lateral sclerosis (ALS). While prior studies have focused on STING within immune cells, little is known about STING within neurons. Here, we document neuronal activation of the STING pathway in human postmortem cortical and spinal motor neurons from individuals affected by familial or sporadic ALS. This process takes place selectively in the most vulnerable cortical and spinal motor neurons but not in neurons that are less affected by the disease. Concordant STING activation in layer V cortical motor neurons occurs in a mouse model of C9orf72 repeat-associated ALS and frontotemporal dementia (FTD). To establish that STING activation occurs in a neuron-autonomous manner, we demonstrate the integrity of the STING signaling pathway, including both upstream activators and downstream innate immune response effectors, in dissociated mouse cortical neurons and neurons derived from control human induced pluripotent stem cells (iPSCs). Human iPSC-derived neurons harboring different familial ALS-causing mutations exhibit increased STING signaling with DNA damage as a main driver. The elevated downstream inflammatory markers present in ALS iPSC-derived neurons can be suppressed with a STING inhibitor. Our results reveal an immunophenotype that consists of innate immune signaling driven by the STING pathway and occurs specifically within vulnerable neurons in ALS/FTD. Supplementary Information The online version contains supplementary material available at 10.1007/s00401-024-02688-z.


Introduction
Neuroinflammation, characterized largely by the initiation and propagation of innate immune signaling, leads to the activation of microglia and astrocytes within the central nervous system and plays an increasingly appreciated role in neurodegenerative diseases [36].Current understanding posits that microglia and astrocytes orchestrate a neuroinflammatory response, either in direct response to exogenous neuronal damage or through an autonomous inflammatory capacity.Recent studies have implicated the STING pathway, a primary link between foreign or damaged DNA and innate immune activation [11], in neuroinflammation and neurodegenerative diseases [76].Cytoplasmic double-stranded DNA, recognized by the DNA sensor cyclic GMP-AMP synthase (cGAS), leads to 2′3′-cyclic GMP-AMP (cGAMP) synthesis, activation of the canonical STING-TBK-1-IRF3 pathway, and downstream type-I interferon signaling.Damaged nuclear DNA can also trigger the non-canonical STING pathway, which operates independently of cGAS and results in the predominant activation of NF-κB [20,39].Increased STING signaling occurs in Parkinson's disease [89] and amyotrophic lateral sclerosis (ALS) [62,102], for which global STING knockout reduced disease progression in mouse models [62,89,102].Such benefits have been thought to result from STING elimination in glial and myeloid cells [76], consistent with the reduction in interferon markers observed after blocking STING in myeloid cells from individuals with ALS [62].Because the accumulation of DNA damage within neurons is a primary feature of ALS, frontotemporal dementia (FTD), and other dementias [14,21,60], we hypothesized that STING may initiate an innate immune response within neurons themselves.We focused on the spectrum of ALS/FTD on account of the central role of DNA damage at the convergence of disease mechanisms.
ALS is a devastating and rapidly fatal degenerative disease of the motor nervous system [10], and it shares pathological features with about 50% of FTD cases [5].While approximately 90% of ALS cases are sporadic (sALS), 10% are familial (fALS) and result from a mutation in one of more than 30 genes that span a broad range of biological functions.These variants include C9orf72 intronic repeat expansion, the most common familial cause of both ALS and FTD and a major contributor to sporadic disease as well [24].The roles of many ALS genes implicate DNA damage as an overarching mechanism, particularly C9orf72, FUS, and TARDBP [46], and postmortem analysis of laser-captured cortical and spinal motor neurons from sALS subjects reveals upregulation of DNA damage and response pathways [42][43][44].
In this study, we established that STING activation occurs in vulnerable cortical and spinal motor neurons in both fALS and sALS but not in less affected neurons.Leveraging parallel approaches across postmortem tissue from multiple fALS variants and sALS, mouse models of C9orf72 repeat expansion disease, and neurons derived from human iPSCs harboring fALS mutations, we observed a consistent activation of STING and its downstream effectors.The parallel results in iPSC-derived neurons suggest that STING activation in neurons occurs in a neuron-autonomous manner, with DNA damage serving as a key mechanism.Supporting the importance of STING signaling in neuronal inflammation in ALS, STING inhibition blocks inherent innate immune markers present in fALS iPSC-derived neurons.

Materials and methods
All experiments were conducted in accordance with relevant guidelines and followed protocols approved by the Massachusetts General Hospital Institutional Review Board, Institutional Animal Care and Use Committee, and Partners Institutional Biosafety Committee.Studies performed in postmortem human tissue were approved by the Massachusetts Alzheimer's Disease Research Center (ADRC) and the Veterans Affairs Biorepository Brain Bank (VABBB).

Postmortem human and mouse tissue immunohistochemistry and immunostaining
Postmortem motor cortex tissue included sections from C9orf72 repeat expansion with ALS (five ALS alone and three ALS/FTD), p-TDP-43-negative AD (8), and non-neurological disease controls (6) provided by the ADRC, as well as TDP-43, FUS, PFN1, KIF5A, NEK1 fALS, and four non-neurological controls provided by the VABBB.Information on human samples is provided in Supplementary Table 1.Briefly, formalin-fixed, paraffin-embedded (FFPE) slides from postmortem human primary motor cortex and occipital cortex were de-paraffinized with xylene followed by a descending ethanol series.Peroxidase quenching was performed by incubating slides in 3% hydrogen peroxide solution for 30 min.Sections were rinsed with PBS, and antigen retrieval was then performed by microwaving sections in 10 mM citrate buffer, pH 6.0, and allowing them to cool on ice for 30 min.The sections were permeabilized using 0.4% Triton X-100 (Millipore Sigma, 9400) for 8 min at room temperature (RT) and incubated with a blocking solution containing 2.5% normal horse serum (Vectastain ABC kit) for 90 min at RT.The sections were then incubated with primary antibody anti-STING (1/50, R&D systems, AF6516) overnight at 4 °C.Immunolabeling was then visualized using the biotinylated secondary antibody sheep IgG horseradish peroxidase (HRP) for 90 min at RT and incubated with the avidin-biotin enzyme complex for 30 min.Stains were visualized by incubation with the Vector peroxidase substrate kit DAB (Vector-Labs).Slides were then dehydrated in ascending ethanol/ xylene series and coverslipped with Cytoseal X. Immunohistochemical negative controls included incubation of sections with no primary or secondary antibody but with all other immunoperoxidase-DAB steps unchanged.Investigators were blinded throughout the processing and analysis of IHC staining.Imaging of immunoperoxidase sections was done using a 40 × objective on a Nanozoomer.
For C9orf72 mouse tissue, FFPE sagittal brains from four AAV (G 4 C 2 ) 2 (two males and two females) and five AAV (G 4 C 2 ) 149 (three males and two females) 1-year-old mice were sliced into 6 μm sections using a microtome.For STING antibody validation, 1-year-old wild-type (WT) and goldenticket (gt) STING gt/gt mice were transcardially perfused with cold 0.01 M PBS, followed by cold 4% PFA in 0.01 M PBS.Their colon and spleen were collected and cut on a vibratome into coronal sections of 40 μm.
For individual experiments, NGN2 neurons were premixed with Matrigel and plated in poly-D-lysine-coated plates, at either 10,000 cells/well (384-well plate, immunostaining) or 250,000 cells per well (24-well plate, RNA) in induction media supplemented with CET.After 24 h, the neurons were maintained at 37 °C with 5% CO 2 in Neurobasal supplemented with 2% B27-supplement, NaCl (50 mM), 1% GlutaMAX, NT-3 (10 μg/mL, PeproTech, 450-03), BDNF (10 μg/mL), and 1% penicillin/streptomycin.The following day, the neurons were cultured in long-term media supplemented with aphidicolin (5uM, Cell Signaling Technology, 32,774).After 2 days, media was exchanged with fresh long-term media without aphidicolin.Every 3 days, 50% of the media was replaced with fresh media and maintained up to day 30 in culture.The neurons were grown for 10 to 30 days for time course analyses and for 20 days before receiving drug treatments.

Lentiviral shRNA neuronal transduction
Two separate lentiviral shRNAs directed against TDP-43 and a scrambled shRNA control were obtained from Origene (TL308946V).For shRNA lentiviral particle transduction, isoTDP-43 +/+ NGN2 neurons were incubated for 4 h with media containing lentivirus particles for GFP-shScramble or GFP-shTDP-43 and cultured for 48 h.Transduction efficiency was assessed by GFP expression.Neuronal cultures were analyzed for STING and γH2AX levels 48 h post-transduction.

RT-qPCR
RNA was extracted from cells using Trizol reagent (Life Technologies, 15,596-018) and purified according to the manufacturer's instructions.Purified RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, 4,368,814) according to the manufacturer's instructions.RT-qPCR was carried out using iQ™ SYBR Green Supermix (Biorad, 1,708,882) on a Biorad CFX96 instrument with the following program: initial denaturation at 95 °C for 360 s; 40 cycles of 95 °C for 30 s, 55 °C for 60 s, and a melt curve step.The quantification cycle for the mRNAs of interest was normalized to TATA-box binding protein (TBP), hypoxanthine phosphoribosyltransferase (HPRT), or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) reference mRNA, and data were expressed as fold change or log2 fold change over vehicle treatment.Gene-specific primer sequences are listed in Supplementary Table 3.

Immunocytochemistry, image acquisition, and analysis
iPSC-derived neurons and primary cortical neuron cultures were fixed for 20 min with 4% paraformaldehyde (Thermo Fisher Scientific, 28,908) at RT. Three gentle washes with PBS (Thermo Fisher Scientific, 10,010,049) were followed by cell membrane permeabilization using 0.25% Triton X-100 (Millipore Sigma, 9400) for 8 min at RT. Cells were incubated with a blocking solution containing 5% bovine serum albumin (BSA) and 10% Normal Donkey Serum for 45 min at RT.The blocking solution was removed, and cells were incubated with primary antibodies diluted in fresh blocking solution overnight at 4 °C.Primary antibodies included anti-TUJ1 (1/250, Aves Lab, TUJ), anti-CTIP2 (1/100, abcam, ab18465), anti-p-IRF3 (1/50, Cell Signaling Technology, 29047S), anti-p-NF-κB (1/50, Santa Cruz Biotechnology, sc-136548), anti-phospho-histone H2A.X (γH2AX, 1/50, Cell Signaling Technology, 9718 T), anticalnexin (1/50, Novus Biological, NB300-518), anti-STING (1/100, Proteintech, 19,851-1-AP; mouse cells), and anti-STING (1/50, R&D systems, AF6516; human cells).Antibodies are listed in Supplementary Table 2.After three washes with PBS, cells were incubated with species-specific Alexa Fluor (− 488, − 647, − 568; Thermo Fisher Scientific) secondary antibodies at a 1/500 dilution for 1 h at RT. Finally, the cells were incubated for 5 min with DAPI solution for nuclear staining and washed three times in PBS.Confocal images were acquired with an Image X-Press Micro Confocal (Molecular Devices) or LSM900 confocal microscope (Zeiss).Four z-stacks were acquired per field (1.6 μm step size), 15 to 25 fields per well using a 40 × objective for iPSC-derived neurons and a 20 × objective for primary mouse cortical neurons.Within each experiment, all groups were imaged with the same acquisition settings.Automatic quantifications were performed using a custom Fiji/ImageJ-based plugin (National Institutes of Health, version 1.53c) [86] for all cellular analyses except TDP-43 depletion, which was performed by blinded manual quantifications of γH2AX and STING-positive neurons performed on 80 to 100 cells per well.For area and integrated intensity measurements of stained cytoplasmic STING and nuclear p-IRF3 and p-NF-kB, and γH2AX, human iPSCneurons and primary mouse cortical neurons were analyzed using a custom Fiji/ImageJ-based plugin (National Institutes of Health).To investigate ER-activated STING, the nuclear mask was dilated (10 iterations), and the perinuclear stained STING area and integrated intensity were analyzed within the dilated nuclear mask.

Quantification and statistical analysis
All data were plotted and statistically analyzed on R-studio software (version 4.2.1).Boxplots and bars represented the mean and standard error of the mean.Unpaired two-tailed Student's t tests (R-studio) were used to compare between groups.Differences were considered statistically significant for p values < 0.05.p values were depicted in graphs as *, **, ***, **** representing p < 0.05, p < 0.01, p < 0.001, and p < 0.0001, respectively.

Data and code availability
Custom scripts used to analyze immunofluorescence and RNA-seq of published datasets are available at https:// github.com/ waing erlab/ STING.Any additional information reported in this paper will be shared by the lead contact upon request.

Unbiased analysis of expression datasets suggests neuronal activation of the STING pathway in ALS
We first wanted to investigate whether STING signaling may be present in central nervous system neurons and contribute to neurodegenerative disease, beyond its well-recognized role in immune cells [6].We selected three publicly available high-quality transcriptional datasets of spinal motor neurons (SMNs): pure populations of SMNs derived from human iPSCs containing pathologic C9orf72 repeat expansions and isogenic control lines [1] or other fALS mutations (TARDBP G298S/+ , PFN1 G118V/+ , and SOD1 G85R/+ ) and isogenic control lines [35], as well as laser-captured postmortem SMNs from sALS cases and controls [49].We analyzed the enrichment of 25 core genes of the cGAS-STING pathway, including 12 from a prior study [62] and 13 from the literature [52,62,68].Gene set enrichment analysis (GSEA) demonstrated differential activation of the STING pathway in both ALS iPSC-derived SMN datasets, despite markedly different techniques for generating the SMNs from iPSCs, and the postmortem laser-captured ALS SMNs (Supplementary Fig. 1a).In contrast, a similar analysis using 30 randomly selected genes (the same number of genes analyzed in the STING pathway) did not show significant enrichment in ALS versus control samples for any dataset (Supplementary Fig. 1b).Compared to the well-controlled iPSC SMN datasets, the higher variability observed in the sALS dataset may reflect greater mechanistic heterogeneity present in postmortem sALS tissue as well as greater impact of background variation [94].
Using the same datasets, we then examined the downstream effectors of STING.Cytosolic double-stranded DNA preferentially activates the canonical STING-TBK1-IRF3 pathway [11].In contrast, nuclear DNA damage can elicit the non-canonical STING pathway, which does not require the presence of cytoplasmic DNA and culminates in strong NF-κB activation [20,39].Synergistic STING amplification via these two pathways may occur [95] and induce an exacerbated neuroinflammatory response [30].While selective activation of the canonical IRF3 signaling was previously reported in myeloid cells from C9orf72 patients [62], our unbiased analysis of vulnerable motor neurons from the three RNA-seq datasets showed enrichment of genes from both the canonical (IRF3) and non-canonical (NF-κB) pathways (Supplementary Fig. 1c).Consistent with the activation of STING by both canonical and non-canonical pathways, GSEA also highlighted an enrichment of genes related to DNA damage response (Supplementary Fig. 1d).Thus, we hypothesized that this dual pathway activation of both IRF3 and NF-κB components may contribute to the role of STING in neurons and their vulnerability to neurodegenerative diseases.

Human postmortem fALS and sALS brains and spinal cords show STING activation in vulnerable ALS neurons
To assess STING activation and its specificity for diseaserelevant neurons, we first obtained postmortem motor cortex tissue sections from eight C9orf72 repeat expansion with ALS (five ALS alone and three ALS/FTD), eight phospho-TDP-43 (p-TDP-43)-negative Alzheimer's disease (AD), and six non-neurological disease control brains.The tissues were matched for age, sex, and postmortem interval (Supplementary Table 1).The specificity of the STING antibody was validated using colon and spleen sections (high-level expression of STING) from wild-type (WT) versus STING goldenticket (gt) mice (STING gt/gt ), which do not produce the protein [85](Supplementary Fig. 2a).Using 3,3'-diaminobenzidine (DAB) immunohistochemistry for STING, we identified only limited staining, which appeared to be restricted to vessels, in control sections (Fig. 1a).Glial STING activation was observed in AD and ALS cases (Supplementary Fig. 2b, left panels); cellular identities were confirmed by immunostaining for microglia (IBA1) and astrocytes (GFAP) (Supplementary Fig. 2b, right panels).Layer V cortical motor neurons, Betz cells in humans, are among the most affected neurons in ALS [32,72,84].In contrast to AD and non-neurological disease control brains, all C9orf72 ALS brains examined showed punctate signal within the neurons of typical layer V Betz morphology (Fig. 1a, lower panels), including large cell body and apical dendrite.Quantification showed a five-fold increase in layer V STING-positive neurons in C9orf72 sections compared to AD and non-neurological disease controls (Fig. 1b, left panel), whereas no increase was observed in the superficial layer II/III neurons (Fig. 1b, right panel).We found similar results in the brains of individuals with other fALS gene mutations, including TDP-43, PFN1, FUS, KIF5a, and NEK1, again with increased STING signal in deep layer V pyramidal neurons but not layer II/III neurons (Fig. 1a, c).
To determine whether STING activation was also observed in sALS, we measured STING levels in postmortem motor cortex tissue sections from 12 sALS and 6 nonneurological disease-matched control brains.We observed a markedly increased number of STING-positive layer V neurons in sALS compared to control sections, with few STING-positive superficial layer II/III neurons in sALS or control sections (Fig. 1d).To confirm our ability to identify layer V pyramidal cortical neurons based on DAB staining morphology and to show definitively that the increase in STING expression in the C9orf72 brains was neuronal and specific for layer V neurons, we took advantage of the mu-crystallin (CRYM) marker, which labels layer V cortical motor neurons [4,66] and detects their degeneration in ALS mouse models [61].Staining for CRYM protein in the human motor cortex sections showed selectivity for deep layer V pyramidal neurons (Supplementary Fig. 2c).Co-staining with the neuronal marker HUC/HUD, CRYM, and STING confirmed the presence of increased neuronal STING activation within layer V cortical motor neurons in C9orf72 ALS brains (Fig. 1e).Consistent with the DAB results, the activation of STING had a remarkable layer V specificity, as STING activation was strong in layer V but nearly absent in layers II/III (Fig. 1e, upper vs lower).Within the STING-positive layer V neurons, we also detected activation of the downstream effector p-IRF3 in the nucleus of CRYM-positive neurons (Supplementary Fig. 2d).
We next evaluated whether STING activation may also occur in SMNs, the other focus of greatest ALS vulnerability in addition to layer V cortical motor neurons.Spinal cord  1f, left panel) compared to non-neurological disease controls.A quantitative analysis demonstrated a fourfold increase in STING-positive SMNs in the ventral horn of sALS cases (n = 13), a five-fold increase in fALS cases with C9orf72 mutation (n = 6), and a three-fold increase in fALS cases with other mutations (n = 4), compared to nonneurological disease controls (Fig. 1g, right panel).
To investigate whether neuronal STING activation is limited to the most disease-relevant regions, we analyzed a brain region largely spared in ALS, the occipital cortex, from sALS and C9orf72 cases compared to non-neurological disease controls.In contrast to our results in the primary motor cortex, no sALS or C9orf72 ALS brains showed STING activation in either layer V or layer II/III neurons (Supplementary Fig. 2e).These findings collectively indicate neuronal activation of the STING pathway specifically in vulnerable cortical motor neurons and SMNs of sALS and a broad range of fALS cases.

STING signaling is increased in a C9orf72 mouse model
To test whether the observed increase of STING activation in postmortem tissue may also be present during earlier  [13,16].For mouse studies, we identified layer V cortical motor neurons using the COUP-TF-interacting protein 2 (CTIP2), a nuclear marker that specifically labels layer V neurons [4,66], including cortical motor neurons [23,61].Indeed, when we first quantified CTIP2-positive cortical motor neurons in motor cortices of 1-year-old animals, we found a significant reduction (36.4%) in AAV (G 4 C 2 ) 149 compared to control AAV (G 4 C 2 ) 2 motor cortices (Fig. 2a, b).In contrast, the number of BRN2positive layer II/III neurons was unchanged between AAV (G 4 C 2 ) 149 and AAV (G 4 C 2 ) 2 motor cortices (Supplementary Fig. 3a).
We then investigated whether STING was activated in motor cortices of AAV (G 4 C 2 ) 149 mice and whether this activation occurred selectively in vulnerable deep layer V cortical motor neurons.We did not detect neuronal STING accumulation in either AAV (G 4 C 2 ) 149 or AAV (G 4 C 2 ) 2 brains in superficial cortical layers (Fig. 2c upper panels).However, co-immunostaining of STING with CTIP2 revealed a significant accumulation of STING within layer V neurons of AAV (G 4 C 2 ) 149 but not AAV (G 4 C 2 ) 2 sections (Fig. 2c; lower panels of Fig. 2d, Supplementary Fig. 3b).We observed increased STING staining immediately adjacent to CTIP2-positive nuclei (18.9% of CTIP2 + neurons), consistent with the typical perinuclear spatial pattern of activated STING and the results in the human postmortem tissue (Fig. 2d).Consistent with the differential STING activation, p-IRF3 was increased in AAV (G 4 C 2 ) 149 mice but not in AAV (G 4 C 2 ) 2 control mice.Furthermore, within the AAV (G 4 C 2 ) 149 cortices, p-IRF3 was restricted to deep layers and almost exclusively co-localized within STING-positive neurons (Fig. 2e).Increased p-NF-κB was also observed in the cortices of AAV (G 4 C 2 ) 149 compared to AAV (G 4 C 2 ) 2 animals (Supplementary Fig. 3c).Thus, the mouse model studies confirmed the specificity for vulnerable neurons revealed in the human postmortem tissue: STING activation in the motor cortex occurred specifically within susceptible layer V CTIP2-positive cortical motor neurons of (G 4 C 2 ) 149 but not (G 4 C 2 ) 2 mice.
The STING pathway is present and functional in primary mouse cortical neurons and healthy control human iPSC-derived neurons STING signaling has been well characterized in microglia and immune cells [11,41,74] but has not been previously documented in central nervous system neurons.To interrogate the functional capacity of STING signaling in neurons, we treated primary mouse cortical neurons and cortical-like neurons derived from human healthy control iPSCs via NGN2 overexpression [104] with modulators of the STING pathway (Fig. 3, primary mouse neurons; Supplementary Fig. 4, iPSC-derived neurons).Treating neurons with a STING agonist (DMXAA and cGAMP for mouse and human neurons, respectively) increased STING pathway activation, indicated by phosphorylation of the downstream nuclear regulators IRF3 (p-IRF3) and NF-κB (p-NF-κB) (Fig. 3a upper and middle panels) and increased expression of canonical IRF3 (Cxcl10, Ifnb1, Ifna) and non-canonical NF-κB (Il6, Il1b, Tnfa) downstream target inflammatory markers as measured by RT-qPCR (Fig. 3b upper and middle panels; Supplementary Fig. 4a) [2,15,20,38,90].Next, we assessed the effects of inhibiting any potential basal STING activity using inhibitors of STING [31] and cGAS, H-151 [31] and RU.521 [96], respectively.These inhibitors yielded reduction in the intrinsic levels of the downstream cytokines Tnfa, Il6, and Cxcl10, indicating the presence of some baseline level of STING signaling within the neurons (Fig. 3b).Notably, we also observed modulation of the STING protein and RNA, which were both increased by STING agonists and decreased by STING inhibitors, consistent with literature documenting a positive feedback loop on STING transcription [33,58] (Fig. 3a, b lower panels; Supplementary Fig. 4b).Thus, both STING and its established regulatory and downstream signaling pathways are present, functional, and modifiable within central nervous system neurons.

SMNs and NGN2 neurons derived from a range of human fALS iPSCs show neuron-autonomous STING signaling activation
To determine whether STING is activated in ALS iPSCderived neurons without the influence of other cell types, we analyzed levels of STING and its downstream mediators in SMNs and NGN2 neurons derived from multiple ALS and control iPSC lines (Fig. 4a), including an isogenic pair of TDP-43 lines (isogenic unedited control (isoTDP-43 +/+ ) and an edited line into which an fALS mutation was introduced Fig. 4 STING pathway is activated in a range of human ALS iPSCderived neurons in the absence of exogenous stressors.a Schematic for iPSC-derived small-molecule SMN differentiation and NGN2 transcription factor (TF)-based induction.b Left, immunofluorescence staining for STING (red), TUJ1 (white), and DAPI (blue) in iPSC-derived small-molecule SMNs at day 35 of differentiation (D35) from isoTDP-43 +/+ control compared to isoTDP-43 +/ G298S (upper panels) and healthy control 11a compared to C9orf72 19f SMNs (lower panels).Right, quantification of time-dependent changes in cytoplasmic STING area and STING intensity in isoTDP-43 +/+ (gray) compared to isoTDP-43 +/G298S (blue) (above) and healthy control (11a, dark gray) compared to C9orf72 (19f, orange) (below) SMNs.c Left, staining for STING (red), TUJ1 (white), and DAPI (blue) in iPSC-derived TF NGN2 neurons at day 10 of differentiation (D10) from isoTDP-43 +/+ compared to isoTDP-43 +/G298S (upper panels) and healthy control (FA10) compared to C9orf72 (ND74) (lower panels) NGN2 neurons.Right, quantification of time-dependent changes in cytoplasmic STING area and STING intensity in isoTDP-43 +/+ (gray) compared to isoTDP-43 +/ G298S (blue) (above) and healthy controls (gradient of gray: light, FA10, medium, Kolf2.1;dark, 11a) compared to C9orf72 (yellow, ND74; orange, ND76; red, ND80) (below) NGN2 neurons.d RT-qPCR analysis showing increase of canonical IRF3 (IFNA, IFNB1, ISG54) and non-canonical NF-κB (TNFA, IL1B response genes in isoTDP-43 +/+ (gray) compared to isoTDP-43 +/G298S (blue) (above) and healthy control (11a, dark gray) compared to C9orf72 (19f, orange) (below) SMNs at day 35.e RT-qPCR analysis of downstream canonical IRF3 and non-canonical NF-κB response genes in isoTDP-43 +/+ (gray) compared to isoTDP-43 +/G298S (blue) at day 30 (above) and healthy control (gray, same three indicated lines) compared to C9orf72 (orange, same three indicated lines) (below) NGN2 neurons at day 10.b, c Each object represents a well, and each symbol represents an independent differentiation (triangles, squares, and circles).)) (Fig. 4) [77], an independent TDP-43 +/G298S ALS subject line (TDP-43 +/G298S ) (Supplementary Fig. 5a), a range of C9orf72 repeat expansion lines, and separate control lines (Fig. 4).In both iPSC-derived SMNs and NGN2 neurons, we observed a time-dependent increase in STING activation within ALS compared to control neurons (Fig. 4b, c; Supplementary Fig. 5a) and a parallel increased expression level of both canonical and non-canonical downstream genes (Fig. 4d, e; Supplementary Fig. 5b,c), together suggesting a hyperactive innate immune response in these ALS neurons.Thus, STING activation in ALS iPSC neurons occurs in a neuron-autonomous and time-dependent manner.It begins after only brief culture periods, does not require additional stressors, and includes both transcriptional and translational components.Given that STING is present within and adjacent to the endoplasmic reticulum and has a predominant perinuclear localization when activated [18,27,29], we next characterized the subcellular location of STING within ALS neurons [40].STING partially co-localized with the ER marker calnexin but also included substantial perinuclear staining (Supplementary Fig. 5d).To quantify the perinuclear component, we measured STING levels adjacent to the nucleus by dilating a mask generated from the nuclear DAPI signal and observed increased amounts in ALS compared to control neurons (Supplementary Fig. 5e).

Activation of STING in neurons by DNA damage
A primary role of STING is to activate an innate immune response to DNA damage, and neuronal DNA damage accumulation is a key early feature [88] and a potential mechanistic driver in neurodegenerative diseases, including AD, ALS, and FTD [14,21,47,60,69,78].To investigate whether DNA damage elicits STING activation in neurons, we treated primary mouse cortical neurons with etoposide, which damages nuclear DNA by inhibiting type II topoisomerase [101], and observed both increased nuclear γH2AX, a marker of nuclear DNA damage and repair (Fig. 5a,b), and cytoplasmic STING levels (Fig. 5a,c).Increased nuclear translocation of both p-IRF3 and p-NF-κB and elevated downstream inflammatory gene expression indicated activation of both canonical and noncanonical STING pathways (Fig. 5a,d,e).One potential mechanism of DNA damage in ALS relates to cortical hyperexcitability, which occurs in both ALS and FTD [8,63,97], and may be related to increased glutamatergic signaling [83].Notably, glutamate receptor agonists can cause DNA damage in parallel with increased neuronal activity [59,91].We extended this reasoning to investigate whether DNA damage resulting from increased neuronal excitability would also induce STING pathway activation.Short pulses of glutamate (1 h) yielded increases in nuclear γH2AX and cytoplasmic STING levels (Fig. 5a-c).STING was elevated at both the protein and RNA levels (Fig. 5c, f).We again observed increased levels of both nuclear p-IRF3 and p-NF-κB and their respective downstream target genes by RT-qPCR (Fig. 5d, e).

Both TDP-43 depletion and C9orf72 dipeptide repeat proteins lead to DNA damage and STING pathway activation in iPSC-derived neurons
Two additional ALS-and FTD-related processes that have been implicated in DNA damage include TDP-43 dysregulation [25,28,37,48,50,64,100,106] and C9orf72 dipeptide repeat (DPR) production [3,22,56,73,98,105].We considered whether these features also elicit STING activation.First, we performed TDP-43 knockdown by using GFP-labeled lentiviral particles containing shRNA against TARDBP in control iPSC-derived NGN2 neurons and analyzed the number of STING-and γH2AX-positive neurons 48 h after transduction.ShRNA-mediated knockdown of TDP-43 via either of the two tested shRNA TDP-43 lentiviral particles resulted in significantly increased DNA damage and STING activation compared to a shRNA scramble control (Fig. 6a, b).Next, we modeled C9orf2 DPR toxicity by applying exogenously synthesized GR 20 [22] in iPSC-derived NGN2 neurons for 24 h and analyzed the number of STING and γH2AX positive neurons [67].We found that GR 20 dipeptide repeats caused not only DNA damage but also STING activation compared to either a control (GAPR) 10 peptide [67] or vehicle control (Figs.6c, d).These findings demonstrate that both TDP-43 depletion and exposure to GR dipeptides lead to DNA damage and STING activation in iPSC-derived NGN2 neurons.The results provide diseaserelevant mechanisms that link DNA damage to STING and innate immune activation across different ALS variants.

STING pathway inhibition abrogates the enhanced inflammatory signal in ALS iPSC-derived neurons
We have now identified potential disease-relevant activators of the STING pathway in ALS iPSC-derived neurons and demonstrated that such mechanisms are sufficient to yield STING activation and stimulation of downstream inflammatory cytokines.To show that STING is indeed necessary for the observed elevated innate immune signaling present in ALS iPSC-derived neurons, we treated cells with STING pathway blockers and assessed the effects on downstream cytokines.We observed that both the STING inhibitor H151 and the cGAS inhibitor RU.521 suppressed the increased interferon response gene expression for both canonical and non-canonical STING pathways in NGN2 neurons derived from isoTDP-43 +/G298S compared to isoTDP-43 +/+ iPSC lines (Fig. 7a) and several C9orf72 repeat expansion compared to control iPSC lines (Fig. 7b).These findings establish the contribution of STING signaling to neuron-intrinsic inflammatory signaling in ALS.

Discussion
While the timing, drivers, and consequences of neuroinflammation in neurodegenerative diseases remain contested, the prevailing conceptual framework involves an initial neuronal insult after which inflammatory cells initiate an innate immune response that leads to neuroinflammation and eventually neurodegeneration [79].Here, we demonstrated that STING activation occurred within vulnerable neurons in ALS/FTD through a neuron-autonomous mechanism, and the activation included both the canonical IRF3 and noncanonical NF-κB pathways.We provided concordant evidence from human postmortem cortices and spinal cords across a broad range of fALS mutations and sALS, a mouse model of C9orf72 repeat expansion ALS/FTD, and iPSCderived neurons harboring fALS mutations.Each of these approaches confirmed important features of the STING neuronal process in ALS/FTD: the human postmortem studies showed the relevance to both sALS and fALS and the specificity for the most disease-susceptible cell types, the layer V cortical motor neurons and SMNs; the C9orf72 mouse model suggested that STING activation occurred at relatively early time periods in the disease course; finally, the iPSC modeling studies supported the neuron-autonomous capacity of the STING signaling process.
Intrinsic neuronal innate immune responses have been recently identified in neurons as potential contributors to Huntington's disease and C9orf72 ALS [53,82,105], although the proposed mechanisms have been RNA rather than DNA-based.Among the pathogen-associated molecular pattern (PAMP) and damage-associated molecular pattern (DAMP) receptors available to produce host-initiated neuroinflammation [26,51], the STING pathway appears to be well positioned to respond to DNA damage, which lies at the crossroads of pathophysiological mechanisms in ALS/FTD.DNA damage results from mutations in numerous fALS genes [71,80,99], C9orf72 repeat expansion [3,22,56,73,98,105], hyperexcitability [59,91], and nuclear TDP-43 depletion [25,28,48,50,64,100], and we now extend those studies to demonstrate that DNA damage from each leads to STING activation in neurons.Beyond these mechanisms, de-repression of endogenous retroviruses [54] may serve as another source of DNA damage and STING activation [94].In addition to nuclear DNA damage, STING may be activated by mitochondrial damage and subsequent mitochondrial DNA release [11,17], a process that occurs following TDP-43 overexpression [102].Other ALS pathological mechanisms may amplify STING activation.For example, reduced clearance of STING by autophagy inhibition can yield increased STING signaling, as shown after C9orf72 depletion in myeloid cells [62].Deficits of autophagy in ALS iPSC neuronal modeling studies share similar implications [7], although interpretation is complicated by the role of STING in promoting autophagy at least in some cell types (2024) 147: 56 56 Page 16 of 20 [29].Links between autophagy and the distinct mechanisms of nuclear DNA damage and DNA release from damaged mitochondria may form additional mechanistic layers [70,81,98].
In the ALS iPSC-derived neurons, upregulation of both STING and its downstream signaling pathways occurred at surprisingly early times--day 10 in NGN2 neurons and day 35 in small-molecule-derived SMNs.At such times, robust disease-relevant phenotypes have proved hard to detect [34].Notably, the activation of STING and its downstream effectors occurred without the addition of pharmacological stressors, which have been necessary to model many of the core neuropathological features within iPSC-derived neurons [34].In contrast with STING activation in myeloid cells following C9orf72 depletion, activation included both the canonical IRF3 pathway and the non-canonical NF-κB pathway, the latter occurring in response to nuclear DNA damage without requiring cytoplasmic DNA release and detection [20].Moreover, interactions driven by STING activation between neurons and neighboring glial cells may potentiate the intrinsic neuronal inflammatory response.
What determines the selective vulnerability of specific neuronal subtypes among different neurodegenerative diseases remains a central unanswered question.The use of cortical layer-specific markers, including CRYM and CTIP2 in human and mouse studies, respectively, allowed us to confirm increased STING activation in layer V neurons within the motor cortex.In contrast, such activation was absent in the superficial layer II/III neurons within the motor cortex and throughout the occipital cortex, both largely spared in ALS.Based on our findings of STING activation in postmortem ALS SMNs, we expected and obtained concordant results with iPSC-derived SMNs.We did not necessarily expect that iPSC-derived NGN2 neurons would exhibit the same phenotype, given that they have been described as general models of excitatory cortical neurons.However, these neurons do not map well onto the subtypes of bona fide human cortical neurons [55,104], and their expression of deep layer markers, such as FEZF2 and PCP4 [35], may be sufficient to drive some deep-layer phenotypes, in this case, STING activation in the presence of fALS mutations.Further investigation will be necessary to determine the specific pathways that elicit STING activation and the connections to selective vulnerability in ALS/FTD.
The functional integrity of STING signaling, observed within both primary mouse and human iPSC-derived neurons, and the robust STING activation in ALS iPSC-derived neurons establish a neuron-autonomous aspect of STING signaling.Blockage of STING within such neurons abrogates the immunophenotype of increased interferon signaling observed in ALS iPSC-derived neurons.Understanding how STING activity within neurons contributes to the broader contexts of neuroinflammation and neurodegeneration will require cell-type selective modulation of STING and other implicated innate immune pathways [82,105].Through independent and joint modulation of these pathways in co-culture and mouse models, novel aspects of these pathways in neurons may be dissected from their more traditional roles within microglia and astrocytes.

Fig. 1
Fig. 1 STING accumulates in layer V CRYM-positive cortical pyramidal neurons in the postmortem motor cortex and SMNs from multiple fALS and sALS variants.a Immunoperoxidase staining showing STING protein in layer II/III (upper panels) and layer V (lower panels) pyramidal neurons from matched non-neurological control, AD, C9orf72, as well as several additional fALS (mutations in TDP-43, PFN1, FUS, KIF5A, and NEK1) motor cortex brain sections.Bottom left inset in each picture shows high-magnification view of STING signal including in cell body and apical dendrite of layer V pyramidal neurons (bottom row).b, c Boxplots depicting the number of layer II/III and layer V STING-positive cortical pyramidal neurons in the motor cortex of C9orf72 (3 FTD/ALS dark blue dots and 5 ALS light blue dots) compared to AD (n = 8, purple) and matched non-neurological control (n = 5, gray) brains (b) and other fALS mutations (n = 6, colored as indicated) compared to separate additional matched non-neurological control brains (n = 4, gray) (c).Each dot represents an individual.d Left, immunoperoxidase staining showing STING protein in layer II/III and layer V pyramidal neurons from the motor cortex of sALS and separate additional matched nonneurological control brains.Right, boxplots depicting the number of layer II/III and layer V STING-positive cortical pyramidal neurons in the motor cortex of sALS (n = 12, purple) compared to non-neurological control (n = 6, gray) brains.(e) Representative immunofluorescence images from C9orf72 brains for neurons (HUC/HUD, white), CRYM (green), and STING (red).f Immunoperoxidase staining showing STING protein in ventral spinal cord SMNs from C9orf72, sALS, and additional fALS cases (TDP-43, FUS, NEK1) and matched non-neurological controls.g Boxplots depicting the number of STING-positive SMNs in C9orf72 (three FTD/ALS dark blue dots and three ALS light blue dots) and sALS (n = 12, purple) compared to matched non-neurological controls (n = 6, gray) and other fALS mutations (n = 4, colored as indicated) compared to separate additional matched non-neurological controls (n = 3).Scale bars = 100 μm and 40 μm for insets.All data are shown as mean ± s.e.m (boxplots), unpaired two-tailed Student's t test: p < 0.05*; p < 0.01 **; p < 0.001 *** ◂

Fig. 2
Fig. 2 STING signaling is increased in deep layer V neurons in C9orf72 mouse model.a Representative images of layer V CTIP2positive cortical pyramidal neurons (white) in the motor cortex of 1-year-old mice expressing 149 G 4 C 2 -repeats ((G 4 C 2 ) 149 , three males, two females) compared to 2 G 4 C 2 -repeats control ((G 4 C 2 ) 2 , two males, two females).b Boxplot depicting the number of layer V CTIP2-positive neurons in (G 4 C 2 ) 2 (gray) and (G 4 C 2 ) 149 (orange) mice.c Immunostaining for STING (green), CTIP2 (magenta), and DAPI (blue) in layer II/III (above) and layer V (below) in the motor cortex of (G 4 C 2 ) 2 control and (G 4 C 2 ) 149 mice.Representative images