Integrated single-cell and spatial transcriptomic profiling in ALS uncovers peripheral-to-central immune infiltration and reprogramming

Ziyang Zhang; Lynn van Olst; Francesco Alessandrini; Matthew Wright; Alex J Edwards; Jake Boles; Anait Nalbandian; Anne V Forsyth; Nate Shepard; Thomas Watson; Evan Kaspi; Angeli Mittal; Joshua Kuruvilla; Natalie Piehl; Abhirami Ramakrishnan; Stanley Appel; Evangelos Kiskinis; David Gate

doi:10.1038/s41593-026-02300-5

. 2026 May 14;29(7):1735–1748. doi: 10.1038/s41593-026-02300-5

Integrated single-cell and spatial transcriptomic profiling in ALS uncovers peripheral-to-central immune infiltration and reprogramming

Ziyang Zhang ^1,^2,^#, Lynn van Olst ^1,^2,^#, Francesco Alessandrini ^2,^#, Matthew Wright ^2,^#, Alex J Edwards ^1,², Jake Boles ^1,², Anait Nalbandian ^1,², Anne V Forsyth ^1,², Nate Shepard ², Thomas Watson ^1,², Evan Kaspi ², Angeli Mittal ², Joshua Kuruvilla ^1,², Natalie Piehl ^1,², Abhirami Ramakrishnan ^1,², Stanley Appel ³, Evangelos Kiskinis ^2,^✉, David Gate ^1,^2,^✉

PMCID: PMC13337491 PMID: 42135512

Abstract

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder marked by progressive motor neuron (MN) degeneration in the brain and spinal cord. Although neuroinflammation is increasingly recognized as a hallmark of ALS, the precise molecular programs linking immune responses to MN pathology remain poorly defined. Using an integrated approach that combines single-cell and bulk RNA sequencing with spatial proteogenomics, we characterized both shared and distinct immune dynamics in peripheral blood and spinal cord tissues from patients with sporadic ALS and those carrying C9orf72 repeat expansions. Our analysis revealed broad immune remodeling in C9orf72 ALS, ALS subtype-specific and progression-associated differences in monocyte activation and antigen-experienced CD8 effector memory T cells with clonal features consistent with antigen-driven responses. Spatial mapping revealed complement activation and lipid-programmed myeloid states converging at sites of MN loss and TDP-43 pathology. Together, these findings connect peripheral and central immune alterations to ALS heterogeneity and highlight stratified immunomodulation as a potential therapeutic strategy.

Subject terms: Neuroimmunology, Amyotrophic lateral sclerosis

This study reveals that the immune system has a role in driving ALS. These findings link blood immune changes to spinal cord damage and suggest personalized treatments targeting specific immune pathways.

Main

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by the progressive degeneration of motor neurons (MNs) in the spinal cord, brainstem and motor cortex¹. This degeneration leads to muscle weakness, spasticity and paralysis, ultimately resulting in respiratory failure¹. Despite extensive research, the precise mechanisms underlying ALS pathogenesis remain incompletely understood. However, an accumulating body of evidence suggests that neuroinflammation is a central feature of ALS pathoetiology².

Neuroinflammatory responses in ALS are marked by microglial activation³, astrocytic dysfunction^4,5 and the infiltration of peripheral immune cells⁶ into the central nervous system (CNS). Together, these immune-driven processes are thought to exacerbate MN damage through a complex interplay between the CNS and peripheral immunity. Previous studies have identified key inflammatory pathways and immune-cell subsets involved in ALS^3–8. Yet, the spatial relationship between immune activation and MN pathology has been poorly resolved, particularly across sporadic ALS (sALS) and familial ALS (fALS) forms of the disease.

To address these gaps, we used single-cell RNA sequencing (scRNA-seq), bulk RNA-seq and spatial proteogenomics (SP) to profile the immune landscape of patients with ALS. By analyzing both peripheral blood and CNS tissues, we elucidate the molecular and cellular underpinnings of inflammation across patients with sALS and fALS harboring C9orf72 repeat expansions (C9-ALS). Our findings reveal significant activation of innate immune pathways in the peripheral blood of sALS and patients with C9-ALS and a connection between peripheral immune dysregulation and MN pathology in the ALS spinal cord.

Altogether, we present an integrated transcriptomic resource of peripheral and CNS immune-cell phenotypes and their spatial localization in the spinal cord of sALS and fALS. These insights advance our understanding of the inflammatory mechanisms underlying disease pathophysiology and open avenues for therapeutic strategies to counteract neuroinflammation in ALS.

Results

Peripheral immune dysregulation in sALS and C9-ALS

We began by profiling peripheral blood mononuclear cells (PBMCs) from 22 patients with ALS and 18 healthy controls (HCs) using scRNA-seq (Fig. 1a). The ALS cohort comprised 18 patients with sALS and 4 with C9-ALS (Fig. 1a,b). To account for disease heterogeneity, patients with sALS were further stratified into slow-progressing (sALS-slow) and fast-progressing (sALS-fast) subgroups based on ALS Functional Rating Scale scores (Supplementary Table 1). Cohorts were well balanced for age and sex (Fig. 1c and Extended Data Fig. 1a). Because all patients with C9-ALS were female, comparisons involving the C9-ALS group were limited to female HCs to control for sex-specific immune differences.

Extended Data Fig. 1 — a, Bar plot showing sex distribution. b,c, Box plots comparing mitochondrial (b) and ribosomal (c) transcript content in 10x Genomics and CRISPR-cleaned single-cell RNA sequencing datasets (10x Genomics: n = 236,539 cells; CRISPR-clean: n = 279,856 cells). Each observation represents one cell. Centerline, median; box limits, 25th and 75th percentiles; whiskers extend to the most extreme values within 1.5× the interquartile range. d,e, UMI count (d) and gene count (e) by sample across ALS subtypes. f,g, Mitochondrial (f) and ribosomal (g) expression levels by sample across ALS subtypes. h, Scaled expression of marker genes across annotated cell types. i, Stacked bar plots showing the percentage of cell types identified across HC and ALS groups. j, UpSet plot summarizing intersecting differentially expressed gene sets in ALS vs. HC comparisons across cell types. k, Pseudobulk expression analysis of C1QA and C1QB in CD16⁺ monocytes from C9-ALS patients and healthy controls. l, GSEA pathway enrichment analysis identifying pathways significantly enriched in CD14⁺ monocytes across the entire ALS cohort and individual ALS diagnostic subgroups. m, GSEA pathway enrichment analysis of adaptive immune cells comparing ALS versus healthy controls across at least two cell populations. Raw P values for panels l and m were computed by fgsea and adjusted for multiple testing using the Benjamini–Hochberg method. HC, healthy control; ALS, amyotrophic lateral sclerosis; sALS, sporadic ALS; C9-ALS, C9orf72 ALS; Mono, monocytes; MAIT, mucosal-associated invariant T cells; TCM, T central memory; T_EM, T effector memory; IFN, interferon; GSEA, gene set enrichment analysis; PBMCs, peripheral blood mononuclear cells.

To enhance the detection of biologically relevant transcripts, we used clustered regularly interspaced short palindromic repeats (CRISPR) technology for targeted transcript enrichment in our scRNA-seq libraries. This depletion method redirected sequencing reads from abundant mitochondrial and ribosomal molecules to lower-expressing, high-value transcripts in an unbiased manner. This approach led to a pronounced reduction in mitochondrial and ribosomal reads (Extended Data Fig. 1b,c). Notably, no significant differences were observed in unique molecular identifiers or gene counts, nor in mitochondrial or ribosomal gene content across samples, ensuring consistent data quality (Extended Data Fig. 1d–g).

Uniform manifold approximation and projection (UMAP) revealed distinct clusters of immune-cell types across ALS and HC cohorts (Fig. 1d and Extended Data Fig. 1h). However, analysis of cell-type distributions did not reveal significant changes to immune populations in patients with ALS compared to HCs (Fig. 1e and Extended Data Fig. 1i).

Differential expression analysis across all immune-cell types revealed the most unique transcriptional alterations in patients with C9-ALS, followed by fast-progressing sALS cases (Fig. 1f). Notably, C9orf72 transcript levels were reduced in PBMC subsets from individuals with C9orf72-associated ALS compared to HCs (Fig. 1g). Differential expression analysis revealed pronounced downregulation of C9orf72 in CD16⁺ monocytes, memory B cells and intermediate B cells (Fig. 1h). These findings indicate that C9orf72 haploinsufficiency affects multiple circulating immune-cell subsets in ALS.

To further investigate immune dysregulation in ALS, we then examined cell-type-specific transcriptional changes across the PBMC compartment. Innate immune cells, including monocytes and mucosal-associated invariant T cells, demonstrated the highest enrichment of differentially expressed genes (DEGs; Fig. 1i,j, Extended Data Fig. 1j and Source data Fig. 1i,j). CD16⁺ monocytes were more transcriptionally dysregulated in sALS, whereas CD14⁺ monocytes exhibited greater alterations in C9-ALS.

Differential expression analysis directly comparing C9-ALS and sALS, as well as fast versus slow progression within sALS, revealed that the majority of DEGs were concentrated in CD16⁺ and CD14⁺ monocytes (Fig. 2a). We therefore focused on CD16⁺ monocytes to identify genes associated with disease progression in sALS, and on CD14⁺ monocytes to investigate subtype-specific changes in C9-ALS. Within sALS, DEG analysis of CD16⁺ monocytes revealed significant upregulation of pro-inflammatory genes in fast progressors. Among these, key components of the innate immune complement system, including complement C1q A chain (C1QA) and complement C1q B chain (C1QB), were specifically elevated in the ALS-fast subgroup (Fig. 2b and Source data Fig. 2). A direct comparison of fast versus slow progressors within sALS confirmed increased expression of both genes (Fig. 2c). Additional genes associated with fast progression included MSR1, which encodes macrophage scavenger receptor 1, a pattern-recognition receptor involved in the uptake of modified lipoproteins, apoptotic cells and cellular debris^9,10, and IFITM3, which encodes interferon-induced transmembrane protein 3, a downstream effector of type I interferon signaling known to modulate antiviral responses and amplify pro-inflammatory pathways^11,12. Pseudobulk analysis further confirmed increased expression of C1QA and C1QB in sALS-fast (Fig. 2d and Extended Data Fig. 1k). These findings align with prior studies demonstrating increased concentrations of complement activation products in plasma¹³ and cerebrospinal fluid (CSF)^14–16, as well as their deposition on innervated motor end plates in the intercostal muscle of patients with ALS¹⁷. Collectively, our results underscore complement activation in CD16⁺ monocytes as a key peripheral immune feature linked to sALS disease severity and progression.

Fig. 2 — a, Bar plot summarizing the number of shared and unique DEGs between comparisons of different ALS subtypes across immune-cell types. b, log₂(FC) plots for CD16⁺ monocytes (sALS-fast versus HC; sALS-slow versus HC). c, Volcano plot of DEGs between sALS-fast and sALS-slow in CD16⁺ monocytes. P values were calculated using MAST and adjusted for multiple comparisons using the Benjamini–Hochberg method. d, Per-donor pseudobulk normalized expression of *C1QA* and *C1QB* in CD16⁺ monocytes from HC (n = 18) and patients with sALS (n = 9 sALS-slow patients and n = 9 sALS-fast patients). Plots represent mean ± s.e.m. Outliers (>2 s.d. from the group mean) are shown as hollow points and were excluded from statistical testing. Two-sided Wilcoxon rank-sum test was used to compare groups. e, log₂(FC) plots for CD14⁺ monocytes (sALS-fast versus HC; C9-ALS versus female HC). f, Volcano plot of DEGs between C9-ALS and sALS in CD14⁺ monocytes. P values were calculated using MAST and adjusted for multiple comparisons using the Benjamini–Hochberg method. g, GSEA of hallmark pathways comparing all patients with ALS against all HCs; only pathways enriched in at least two populations of innate immune cells are shown. Raw P values were computed by fgsea and adjusted for multiple testing using the Benjamini–Hochberg method.

Source data

Comparison of sALS and C9-ALS to HCs in CD14⁺ monocytes revealed distinct gene expression profiles that differentiate C9-ALS. Notably, the pattern-recognition receptor gene C-type lectin domain containing 7A (CLEC7A); the innate immune regulator tribbles pseudokinase 1 (TRIB1) and early growth response (EGR) genes EGR1 and EGR3 were specifically upregulated in C9-ALS (Fig. 2e and Source data Fig. 2). In contrast, genes uniquely upregulated in sALS included C-C motif chemokine ligand 3 (CCL3) and C-C motif chemokine ligand 3 like 1 (CCL3L1), highlighting subtype-specific inflammatory signatures (Fig. 2e and Source data Fig. 2). Direct comparison of CD14⁺ monocytes between C9-ALS and sALS further highlighted these subtype-specific programs. C9-ALS showed increased expression of IFITM2 and IFITM3, interferon-stimulated genes (ISGs) involved in antiviral defense and immune modulation¹⁸, whereas sALS showed elevated CCL3L1 and IL1B expression, key immunomodulators that mediate immune-cell recruitment¹⁹ and represent a downstream product of inflammasome activation, respectively²⁰ (Fig. 2f). These findings demonstrate that unique innate immune activation pathways and transcriptional programs distinguish C9-ALS from sALS, suggesting distinct pathophysiological mechanisms underlying these ALS patient groups.

Gene set enrichment analysis (GSEA) of pathways shared by multiple innate immune cells revealed an increase in oxidative phosphorylation and a reduction in interferon-γ signaling in the combined ALS cohort (Fig. 2g). When focusing specifically on CD14⁺ monocytes, we observed significant enrichment of hallmark pathways related to reactive oxygen species production and hypoxia, which were increased only in sALS (Extended Data Fig. 1l). This suggests that the combined ALS signal may be primarily driven by the sALS subgroup. Tumor necrosis factor signaling through NF-κB was enriched in both patients with sALS and C9-ALS, while the inflammatory response signature in CD14⁺ monocytes was unique to C9-ALS (Extended Data Fig. 1l). Extending our analysis to adaptive immune cells, we identified pathways shared across multiple subsets (Extended Data Fig. 1m). Specifically, we found a downregulation of oxidative phosphorylation in CD4⁺ T cells and double-negative (dn) T cells, accompanied by an upregulation of tumor necrosis factor–NF-κB signaling and the p53 pathway across several adaptive immune-cell types. The contrasting effect of oxidative phosphorylation being higher in innate immune cells but lower in adaptive cells may reflect chronic innate activation compared to adaptive exhaustion. Notably, these findings prompted further examination of adaptive immune dysregulation in ALS.

To address this, we first examined DEGs in adaptive immune cells of C9-ALS and sALS compared to HCs. Notably, in each comparison, a high number of DEGs were identified in CD8 T effector memory (T_EM) cells, a subset known to be antigen experienced (Extended Data Fig. 2a). To investigate further, we analyzed DEGs in clonally expanded T cell populations using our established T-cell-receptor-sequencing (TCR-seq) methods^21–23. This analysis revealed numerous TCRβ chain genes that were uniquely dysregulated in C9-ALS or sALS compared to HCs, suggesting distinct CD8 T_EM antigen-specific responses in each ALS subtype (Extended Data Fig. 2b). To assess clonal diversity, we analyzed TCR sequences from our scRNA-seq dataset. Clonotypes were defined as expanded if the same TCRβ CDR3 amino acid sequence was observed at least twice within an individual’s repertoire, a threshold commonly used to identify antigen-driven expansion²⁴. While the number of expanded clonotypes did not differ significantly across groups, sALS-slow patients exhibited the highest proportion of unique clonotypes, suggestive of a more diverse, potentially polyclonal T cell response (Extended Data Fig. 2c). To investigate whether transcriptional changes observed in CD8⁺ T_EM cells reflected underlying TCR gene usage biases, we compared differentially expressed TCR genes from the scRNA-seq dataset with V gene usage patterns from paired scTCR-seq data (Extended Data Fig. 2d). Notably, TRBV6-5, which was upregulated in sALS clonal CD8⁺ T_EMs, showed the highest enrichment in sALS-fast patients, suggesting subtype-specific T cell activation.

Extended Data Fig. 2 — a, UpSet plot summarizing intersecting differentially expressed genes in clonally expanded adaptive immune cells; sALS vs. HC (left), C9-ALS vs. HC (right). b, Volcano plots showing DEGs in clonally expanded CD8 T_EM cells for C9-ALS (left) and sALS (right) against female healthy controls and all healthy controls, respectively. P values were calculated using MAST and adjusted for multiple comparisons using the Benjamini–Hochberg method. c, Mean and standard deviation of the unique clone proportion across diagnosis groups. Each data point represents an individual sample, scaled to its respective total number of contigs. Statistical significance was assessed using one-way ANOVA. n = 18 HC, n = 9 sALS-slow patients, n = 9 sALS-fast patients, n = 4 C9-ALS patients. d, Proportional usage of beta chain genes across diagnosis groups. HC, healthy control; ALS, amyotrophic lateral sclerosis; sALS, sporadic ALS; C9-ALS, C9orf72 ALS; T_EM, T effector memory.

Cumulatively, our findings reveal a striking alteration in immune-cell composition and activation states in patients with ALS, particularly among monocytes. The distinct inflammatory signatures observed in patients with sALS-fast and C9-ALS provide evidence for subtype-specific immune dysregulation. Furthermore, the identification of unique TCRβ chain gene expression patterns and a high number of DEGs in CD8 T_EM cells underscores the potential involvement of antigen-specific T cell responses in ALS. These results highlight the role of peripheral immune activation in ALS pathophysiology and suggest potential therapeutic targets to modulate immune responses. To facilitate additional investigation, we have made these scRNA-seq data available as an interactive online resource at https://gatelabnu.shinyapps.io/als_app/.

Gene expression changes and immune-cell dynamics in ALS CNS tissues

To investigate inflammatory responses in patients with ALS within the CNS, we analyzed bulk RNA-seq datasets generated by the New York Genome Center (NYGC) ALS Consortium and Target ALS²⁵ with a focus on immune signaling. We used a dataset of 1,464 postmortem samples derived from HCs (n = 92), ALS (n = 237), frontotemporal dementia (FTD; n = 50) and ALS/FTD (n = 26) patients, including tissues from the cerebellum, cortex (frontal, motor, occipital, temporal), hippocampus and spinal cord (cervical, thoracic, lumbar), collected from multiple clinical sites using standardized protocols (Fig. 3a, Extended Data Fig. 3a and Source data Fig. 3). We performed differential gene expression analysis between controls and ALS, ALS/FTD and FTD cases for each CNS tissue with at least three independent samples, while controlling for several variables (Methods; Fig. 3a and Extended Data Fig. 3b,c). In ALS cases, we observed the largest number of DEGs in the lumbar spine, motor cortex and cervical spine, respectively (Fig. 3a and Extended Data Fig. 3c). ALS/FTD samples displayed a similar pattern (Extended Data Fig. 3b,c), while FTD cases exhibited a substantially larger number of DEGs in the frontal cortex compared to ALS cases. This is consistent with the clinical presentation of these patients (Fig. 3a and Extended Data Fig. 3b,c). Intriguingly, the cerebellum, a region typically not thought to be involved in ALS/FTD^26,27, exhibited a considerable number of DEGs (Extended Data Fig. 3c). To determine whether the DEG effect size was influenced by sample size, we used all ALS tissues and conducted serial downsampling. This analysis showed that, even in cases with as few as ten samples per tissue, spinal cord sections were the most affected CNS region in patients with ALS, as measured by DEG number (Extended Data Fig. 3d).

Fig. 3 — a, Schematic representation of different CNS regions examined, color coded based on the number of DEGs when comparing ALS or FTD postmortem samples to HCs from the Target-ALS/NYGC databases. b, Volcano plot of DEG analysis in the motor cortex of patients with ALS compared to HCs. c, Lollipop plot displaying the top 15 most enriched terms as determined by GSEA for cell-type markers, in the motor cortex of patients with ALS, compared to HCs. d, Volcano plot of DEG analysis in the cervical spine of patients with ALS compared to HCs. e, Lollipop plot displaying the top 15 most enriched terms as determined by GSEA for cell-type markers, in the cervical spine of patients with ALS, compared to HCs. f, Schematic representation of the cellular deconvolution method used to infer population changes in bulk RNA-seq datasets of CNS tissue based on scRNA-seq datasets^30–32. g, Heatmap displaying inferred cell-type changes in ALS cortical tissues (frontal, motor, occipital, temporal cortices) based on the cellular deconvolution. h, Heatmap displaying cell-type changes in ALS spinal cord sections (lumbar, thoracic, cervical) based on the cellular deconvolution (top), and cell-type changes in ALS cervical spinal cord stratified by ALS-associated mutations (C9-ALS, sALS, *SOD1*-ALS; bottom). Statistical significance in g and h determined by Wilcoxon test; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. MC, motor cortex; OC, occipital cortex; TC, temporal cortex; FC, frontal cortex; CB, cerebellum; CS, cervical spinal cord; TS, thoracic spinal cord; LS, lumbar spinal cord; OPC, oligodendrocyte precursor cell; SMC, smooth muscle cell; Prolif., proliferating.

Source data

Extended Data Fig. 3 — a, Principal component analysis (PCA) of bulk RNA-seq tissue samples used in the analysis from the Target-ALS/NYGC databases. b, Schematic representation of different CNS regions examined, color coded based on the number of DEGs when comparing ALS–FTD postmortem samples to HCs. c, Number of DEGs in each CNS region when comparing ALS, ALS–FTD or FTD postmortem samples to HCs. d, Bar plot displaying the number of DEGs in each CNS region when comparing ALS postmortem samples to HCs (top panel). Box plots (bottom panels) showing the number of DEGs in each CNS region when comparing ALS postmortem samples to HCs in serial downsampling analysis, with as few as 10 samples per region. e, Schematic representation illustrating the enrichment of marker genes from different microglia clusters annotated in ref. ²⁹ within the microglia subclasses from the Molecular Signatures Database (MSigDB) gene set. f, Clustered dot plot displaying GSEA results for cell-type marker genes, based on DEG analysis comparing ALS patient tissue to HC tissue. Only statistically significant terms are shown. The size of the dots is proportional to the p-value, while the color represents the normalized enrichment score (NES) of gene set elements. Statistical significance determined by GSEA.

We then focused on ALS cases, which had the largest sample size, to investigate the nature of gene expression changes. In the ALS motor cortex, some of the most significantly upregulated genes included MGAT4A, a Golgi-enriched glycosyltransferase enzyme; NAA50, an acetyltransferase-encoding gene; and the ferroptosis-associated ACSL4 gene (Fig. 3b). GSEA against cortical cell-type markers²⁸ downloaded from the Molecular Signatures Database (MSigDB) revealed a strong upregulation of oligodendrocyte, microglia and inhibitory neuron signatures and a concordant downregulation of excitatory neurons and endothelial cells (Fig. 3c). In the ALS cervical spine, we observed upregulation of several immune-related genes including CHIT1, GPNMB, CD48, APOE, APOC1 and the complement genes C1QA, C1QB and C1QC (Fig. 3d). GSEA revealed a significant enrichment for microglia, brain immune cells, astrocytes and B cell and T cell signatures (Fig. 2e). In contrast, endothelial cells, excitatory neuron and oligodendrocyte terms were de-enriched (Fig. 3e). The upregulation of immune-cell expression signatures, particularly Microglia 3, was consistently found in both the motor cortex and cervical spinal cord of patients with ALS. To define the nature of this microglial subtype, we leveraged a recently reported annotation of human microglia transcriptional states²⁹ and found an enrichment of genes related to lipid processing, ribosome biogenesis, phagocytic and inflammatory microglia states (Extended Data Fig. 3e). Intriguingly, although the cerebellum exhibited a very high number of DEGs, there was no significant enrichment for Microglia 3, while brain immune-cell signatures were de-enriched in this tissue except for cycling microglia 1 (Extended Data Fig. 3f).

To determine whether the DEGs within the bulk RNA-seq of multicellular tissue samples were influenced by differential cell proportions, we performed bioinformatic cell-type deconvolution (Fig. 3f). To ensure accurate cell-type identification, we used multiple single-nucleus RNA-seq datasets from the human frontal cortex³⁰, motor cortex³⁰ and spinal cord^31,32 for tissue-specific cell deconvolution. We used a signature matrix generated from a frontal cortex single-cell dataset³⁰ to deconvolute the corresponding frontal cortex bulk RNA-seq data, a motor cortex dataset³⁰ to deconvolute motor cortex data and a combination of both regions to deconvolute the occipital and temporal cortex data (Extended Data Fig. 4a,b; Methods). Tissue-specific deconvolution is critical, as the inferred proportions of certain cell subtypes, such as inhibitory neurons, pericytes, T cells and oligodendrocyte precursor cells, are strongly affected by the source dataset (Extended Data Fig. 4c). Consistent with the GSEA analysis, we found a significant increase in inhibitory interneurons and a reduction in endothelial cells, in both the motor and the frontal cortex of patients with ALS (Fig. 3g). We also found an increase in the proportion of microglial cells specifically in the motor cortex, while other brain regions showed no substantial population changes, except for an increase in fibroblasts in the occipital cortex (Fig. 3g). More granular deconvolution of distinct neuronal subtypes revealed a complex interplay between excitatory and inhibitory neuron changes in the frontal and motor cortex of patients with ALS (Extended Data Fig. 4d,e). There were inferred increases in neurogliaform (SST-NPY-neg) and Rosehip (Rosehip-LAMP-PMEPA1) interneurons, as well as bidirectional changes in excitatory neurons (Extended Data Fig. 4e).

Extended Data Fig. 4 — a, UMAP representation of the clustering of human motor and frontal cortex cells from the single-cell dataset in ref. ³⁰, with broadly annotated cell types. b, Dot plot displaying markers of each cell type and their expression across single-cell clusters used for cellular deconvolution in the human motor and frontal cortex. c, Scatterplot showing the correlation between the proportions of inferred cell types in the motor cortex, using a cell-type-specific signature matrix generated from single-cell datasets of either the motor or frontal cortex. d, UMAP representation of clustering of human motor and frontal cortex cells from ref. ³⁰, with more granularly defined cell types. e, Heatmap displaying changes in inhibitory and excitatory subtypes in the ALS motor and frontal cortex compared to HCs, inferred by cellular deconvolution. The bar graph on the right shows the proportion of each cluster in healthy individuals in both regions. Statistical significance determined by Wilcoxon test; *p < 0.1, **p < 0.01, ***p < 0.001, ****p < 0.0001.

To deconvolute the bulk RNA-seq spinal cord data, we integrated single-nucleus RNA-seq data from a recently published spinal cord atlas³², with another spinal cord dataset generated using a neuronal enrichment strategy³¹ (Extended Data Fig. 5a–c; Methods). This analysis revealed significant proportional changes in several cell types. MNs and oligodendrocytes were reduced within the lumbar and cervical spinal cord sections of patients with ALS (Fig. 3h, top), while several types of immune cells, including macrophages, microglia and lymphocytes, were increased, along with astrocytes, Schwann cells and oligodendrocyte precursor cells (Fig. 3h, top). Notably, these changes were consistent across distinct ALS subtypes, including mutant C9-ALS, SOD1-ALS and sALS cases (Fig. 3h, bottom). To validate the shift from neurons to immune cells inferred by bioinformatic analysis, we performed immunohistochemistry (IHC) on cervical spinal cord sections from patients with ALS (n = 7) and HCs (n = 4–8). We observed a striking reduction in the number of large Microtubule-associated protein 2 (MAP2)⁺ neurons in the anterior horn of ALS cases (Extended Data Fig. 5d), along with a concomitant increase in the fraction of IBA1⁺ myeloid cells throughout the spinal cord (Fig. 4a,b and Extended Data Fig. 5e). Interestingly, these IBA1⁺ cells appeared larger and more morphologically complex (Fig. 4c and Extended Data Fig. 5f), suggesting an activated phenotype. Overall, the increase in the proportion of immune cells was more pronounced in the spinal cord compared to other CNS regions in patients with ALS (Fig. 4d).

Extended Data Fig. 5 — a, Schematic representation of the integration of the MN signature from the single-cell analysis in ref. ³¹, into the spinal cord single-cell atlas from ref. ³²; see Methods. b, UMAP representation of the clustering of different cell types in the human spinal cord from the single-cell dataset in ref. ³² including the MN population from ref. ³¹. c, Dot plot displaying well-characterized markers of each cell type and their expression across single-cell clusters used for cellular deconvolution in the human spinal cord. d, Representative fluorescence microscopy images of the anterior horn of the cervical spinal cord in sporadic sALS patients and HCs, immunolabeled for MAP2 (left). Scale bar = 100 µm. Box plot displaying the number of MAP2⁺ neurons detected in the anterior horn of a single cervical spinal cord section from ALS patients (n = 10) and HCs (n = 8). Plot represents mean ± s.e.m. Statistical significance determined by two-sided unpaired t-test; p < 0.0001. e, Representative fluorescence microscopy images of an entire cervical spinal cord section immunolabeled for IBA1, with a magnified inset. Orange arrows indicate IBA1/DAPI⁺ cells; 4 HC and 7 ALS sections have been analyzed. f, Superplot displaying the morphology of IBA1⁺ cells in the cervical spinal cord of ALS patients and HCs, represented as a roughness factor. Small dots represent individual cells, large dots the mean value per patient (used for statistical analysis), and colors denote individual patients. Statistical significance determined by two-sided Welch’s t-test; p < 0.01. g, Scatterplot showing the correlation between changes in the proportion of microglia cells in the cervical spinal cord and disease duration in ALS patients. P-value determined by linear regression; shaded area represents the 95% confidence interval. h, Scatterplot showing the correlation between changes in the proportion of immune cells in the cervical spinal cord and the age of onset in ALS patients. P-value determined by linear regression; shaded area represents the 95% confidence interval. i, Heatmap indicating the magnitude (r) and significance (p) of the correlation between cell-type proportions and disease duration in the cervical spine of ALS patients. Statistical significance in g–i determined by linear regression. j, Histogram showing distribution of disease duration among ALS patients in the Target ALS/NYGC cohort. Red dashed lines indicate the first and third quartiles (22 and 48 months, respectively). k, Representative fluorescence microscopy images of IBA1/LSP1/DAPI⁺ cells in ALS cervical spine. l, Bar plots quantifying IBA1/LSP1/DAPI⁺ cells detected by IHC in whole ALS (n = 9) and HC (n = 5) spinal cord sections, in the gray matter and in the white matter. Plots represent mean ± s.e.m. Statistical significance determined by two-sided unpaired t-test.

Finally, we examined whether gene expression and cell-type changes correlated with ALS disease duration. In terms of cell-type changes, there was a negative correlation between the proportion of immune cells and disease duration (Fig. 4e), as well as the proportion of microglia (Extended Data Fig. 5g). There was no significant correlation between immune cells and age of disease onset (Extended Data Fig. 5h), nor was the same association with disease duration seen in other non-neuronal cell types (Extended Data Fig. 5i). To examine the association with disease progression in another way, we selected all 226 patients with ALS from the cohort with available disease duration data and partitioned them into fast and slow progressors. Specifically, patients in the top 25% and bottom 25% percentile in terms of disease duration (<22 or >48 months, respectively) were classified as fast progressing and slow progressing (Extended Data Fig. 5j). Using our original spinal cord deconvolution, we compared immune proportions between these groups. Notably, patients with a short disease course exhibited significantly higher proportions of macrophages and microglia in their spinal cords based on cell deconvolution analysis (Fig. 4f). Together, these results suggest that immune-cell changes appear to be more related to disease progression rather than the time of initial onset of ALS.

An unbiased analysis highlighted numerous myeloid genes negatively correlated with disease duration (Fig. 4g). These included CHIT1, which is substantially upregulated in ALS spinal cord³³, as the top gene showing a strong negative correlation with disease duration in the ALS patient cohort (Fig. 4g,h and Source data Fig. 4g). leukocyte-specific protein 1 (LSP1) was the third top gene negatively associated with disease duration and is known to localize to nascent phagocytic cups during Fcγ receptor-mediated phagocytosis³⁴. Consistent with a role in phagocytic activation, we found that LSP1⁺/IBA1⁺ microglia were increased in both the gray and white matter regions of the ALS spinal cord by IHC analysis (Extended Data Fig. 5k,l). Additionally, APOC1, implicated in macrophage activation through its role in lipid metabolism and inflammation, and the complement genes C1QA, C1QB and C1QC, which we also found to be upregulated in peripheral blood of patients with ALS (Fig. 2b–d), all showed a negative correlation with disease duration (Fig. 4h). To investigate whether the genes with the strongest negative correlation to disease duration were enriched for a particular pathway, we performed GSEA, which revealed microglia-enriched and immune-related categories as strong negative predictors of disease duration (Fig. 4i). These findings highlight profound cell-type shifts and immune activation in the ALS spinal cord, with reductions in neurons and oligodendrocytes accompanied by increased immune-cell populations, including activated myeloid cells, and suggest that upregulation of CHIT1, LSP1, APOC1 and microglia cells may serve as molecular markers linked to disease progression.

Regional specificity of inflammatory changes in ALS spinal cord

To further investigate the molecular mechanisms underlying spinal cord inflammation in ALS, we used SP to concurrently assess RNA and protein expression in spinal cord sections from nine HC, five sALS and four C9-ALS cases (Fig. 5a and Supplementary Table 1). Our approach included a panel of 31 barcoded immune antibodies, which were codelivered to tissue slides along with RNA probes. This enabled a spatially resolved analysis of both immune-cell populations and gene expression profiles within the tissue (Methods). Each SP spot was annotated based on our tissue staining (Methods) to correspond to one of four regions—anterior/lateral white matter, posterior white matter, anterior or posterior horns (Fig. 5b). Differential gene expression analysis between ALS and HC spinal cords was conducted for each region, identifying DEGs in the anterior/lateral white matter and anterior horns (Fig. 5c). Consistent with findings in peripheral monocytes of ALS-fast patients (Fig. 2a), we noted upregulation of complement genes C1QA and C1QB, as well as C1QC in the anterior/lateral white matter of ALS spinal cords compared to HC tissues (Fig. 5d and Source data Fig. 5d). Consistent with the bulk RNA-seq analysis (Fig. 3d), APOC1 emerged as a shared DEG across both the anterior/lateral white matter and anterior horns (Fig. 5d). Additionally, in the anterior horn where MNs reside, we only found two downregulated genes in ALS tissues—neurofilament light chain (NEFL) and stathmin 2 (STMN2), likely reflecting MN loss (Fig. 5d and Source data Fig. 5d). The level of neurofilament light chain protein detected in CSF and blood has emerged as a reliable biomarker of neuroaxonal damage in ALS clinical trials³⁵. STMN2 has a crucial role in maintaining neurofilament-dependent axoplasmic organization, essential for preserving the caliber and conduction velocity of myelinated large-diameter axons³⁶. Critically, STMN2 mRNA is a target of TAR DNA-binding protein 43 (TDP-43), which binds it directly to maintain its splicing fidelity³⁷. In ALS, TDP-43 dysfunction leads to the mis-splicing and downregulation of STMN2 (refs. ^37–40). These findings highlight region-specific transcriptional changes in the ALS spinal cord, with complement activation and APOC1 upregulation emerging as key molecular features linked to neuroinflammation.

Fig. 5 — a, Schematic of SP workflow combining mRNA probes and a 31-plex immune-antibody panel to analyze cervical spinal cord regions in ALS (n = 9 HC, n = 5 sALS, n = 4 C9-ALS). b, Morphological annotation of cervical spinal cord regions showing anterior/lateral white matter, posterior white matter, anterior horns and posterior horns for HC, sALS and C9-ALS. c, UpSet plot summarizing unique and shared DEGs across spinal cord regions (anterior/lateral white matter, posterior white matter, anterior horns and posterior horns) in ALS versus HC. d, Volcano plots of DEGs in anterior/lateral white matter (left) and anterior horns (right) comparing ALS to HC. e, Representative spatial maps highlighting C2L cell-type-enriched ST spots in ALS spinal cord. f, Box plots of normalized C2L scores for oligodendrocytes, macrophages, lymphocytes and microglia in ALS and HC spinal cord regions. Each point represents one donor. Box plots compare HC (n = 9) versus pooled ALS (n = 9), with point colors indicating diagnosis subtype. Box plots show median (centerline), IQR (box) and 1.5× IQR whiskers. Statistical significance was assessed using a two-sided Wilcoxon rank-sum test. Oligodendrocytes comparison (leftmost), P = 8.2 × 10⁻⁵. g, Volcano plots of DEGs in macrophage-enriched and neuron-enriched ST spots in ALS versus HC. h, Volcano plots of DEPs in anterior/lateral white matter and anterior horns comparing sALS or C9-ALS to HC. Downregulated proteins are shown in blue, upregulated in red, shared DEG/DEP in green and proteins without overlap in purple. i, Confocal images of ALS spinal cord showing DAPI, CD8, IBA1 and MAP2 staining to illustrate spatial localization of immune and neuronal cells. For d, g and h, P values were calculated using MAST and adjusted for multiple comparisons using the Benjamini–Hochberg method. *APOC1*, apolipoprotein C1; *CHI3L1*, chitinase-3-like protein 1; *SPP1*, secreted phosphoprotein 1; *IBA1*, ionized calcium-binding adapter molecule 1; WM, white matter; Ant., anterior; Lat., lateral. Schematic in panel a created in BioRender; Fontanna, T. https://biorender.com/a23j248 (2024).

Source data

Using Cell2Location (C2L) cell-type deconvolution⁴¹, we incorporated a single-nucleus RNA-seq dataset³² to predict cell types within spatial spots (Fig. 5e and Extended Data Fig. 6a; Methods). Predicted cell types localized to their expected regions within the spinal cord (Extended Data Fig. 6b), with neurons aligning with regions of MAP2 positivity (Extended Data Fig. 6c). A comparison of predicted cell-type proportions between HC, sALS, and C9-ALS spinal cords revealed reduced oligodendrocyte and neuron proportions in both ALS groups relative to HCs (Extended Data Fig. 6d). C2L-estimated cell-type counts were normalized by expressing each as a percentage of the total inferred cells within the same spatial spot. Normalized C2L prediction scores, focused on the anterior/lateral white matter and anterior horns due to the presence of DEGs (Fig. 5d), showed significant reductions in C2L-predicted oligodendrocyte cell numbers alongside increased macrophage, lymphocyte, and microglia predictions in the anterior/lateral white matter (Fig. 5f). Similarly, the anterior horn had C2L-estimated increases in macrophage and microglia in ALS (Fig. 5f). We also detected a nonsignificant reduction in neuronal C2L counts and an increase in lymphocytes (Extended Data Fig. 6e). These results provide spatially resolved evidence of immune-cell infiltration and glial dysregulation in ALS-affected spinal cord regions, further emphasizing the contribution of neuroinflammation to disease pathology.

We then performed differential expression analysis on cell-type enriched spots in ALS (Extended Data Fig. 6f). In macrophage-enriched spots, APOC1 was upregulated, indicating macrophage activation within the ALS spinal cord (Fig. 5g and Source data Fig. 5g). In neuron-enriched spots, we identified increased expression of complement signaling genes, including C1QC and complement C3 (C3; Fig. 5g and Source data Fig. 5d). C3 is known to regulate synapse elimination⁴², to influence axonal growth⁴³ and is implicated in neurodegenerative disease⁴⁴. Additionally, neuron-enriched spots showed downregulation of NEFL and STMN2, both of which are core structural components of MNs (Fig. 5g and Source data Fig. 5d). Collectively, these findings highlight the intricate relationship between immune activation and neuronal dysfunction in the ALS spinal cord, emphasizing the potential roles of macrophage activation, complement signaling and impaired axonal maintenance in disease progression.

Differential protein expression (DEP) analysis confirmed distinct molecular changes specific to subgroups and spinal cord regions. Significant differences in protein expression between sALS and HC cohorts were identified in the anterior/lateral white matter and anterior horns. Notable DEPs included the monocyte adhesion molecule integrin subunit α X (ITGAX) and the macrophage motility factor syndecan-1 (SDC1), among others (Fig. 5h and Source data Fig. 5h). Within the anterior horns, comparisons between sALS and HC cohorts revealed DEPs such as the CD8 T cell protein cluster of differentiation 8 α (CD8A) and ITGAX (Fig. 5h and Source data Fig. 5h).

In the C9-ALS anterior/lateral white matter, enriched proteins included lysosomal protein cluster of differentiation 68 (CD68) and Fcγ receptor IIIa (FCGR3A), among others (Fig. 5h and Source data Fig. 5h). Similarly, comparisons of the anterior horns between C9-ALS and HC participants identified DEPs including CD68, ITGAX and FCGR3A (Fig. 5h and Source data Fig. 5h).

We confirmed the increased abundance of IBA1/CD68⁺ myeloid cells in both the white and gray matters of patients with ALS by IHC (Extended Data Fig. 6g). Intriguingly, CD8 T cells were observed interacting with CD68⁺ (Extended Data Fig. 6h) and IBA1⁺ (Fig. 5i) myeloid cells in the ALS spinal cord, and IBA1⁺ myeloid cells were found interacting with MAP2⁺ neurons (Fig. 5i). These findings emphasize the regional specificity of molecular and cellular alterations in ALS. Collectively, SP of ALS spinal cord tissue revealed substantial alterations in protein expression, cellular composition and intercellular interactions compared to HC participants.

Immune dysregulation associates with ALS spinal cord pathology

We then aimed to determine whether the peripheral and central immune dysregulation observed above was linked to MN degeneration. To address this, we developed a panel of 17 MN-enriched genes to evaluate MN loss in our SP data (Fig. 6a). Using this approach, we identified a significant reduction in MN signals in the anterior horn among ALS cases (Fig. 6b). To further explore transcriptomic changes around MNs, we annotated MN-enriched spatial spots (Fig. 6c). Differential expression analysis of MN-enriched spots revealed upregulation of complement genes C1QA, C1QB and C1QC, as well as APOC1 and CD163 (Fig. 6d and Source data Fig. 6d). In contrast, STMN2 and NEFL, typically highly expressed in MNs, were downregulated in these same spots (Fig. 6d and Source data Fig. 6d). Pseudobulk quantification confirmed significant upregulation of C1QC, APOC1 and CD163 in MN-enriched spots (Fig. 6e). Notably, APOC1 binds to CD163, a scavenger receptor on macrophages, influencing their polarization⁴⁵. We identified APOC1 and CD163 in anterior horn regions containing IBA1⁺ myeloid cells (Fig. 6f). Collectively, these findings highlight an association between innate immune inflammation and MN loss in the ALS spinal cord.

Fig. 6 — a, Schematic of the 17-gene MN enrichment panel used to evaluate MN loss in SP data. b, Box plots of MN-associated gene expression scores in anterior horn spatial spots from controls and ALS donors. For each donor, SCT-normalized expression was averaged across an MN gene set shown in a and across anterior horn spots. Each point represents one donor (HC, n = 9; sALS, n = 5; C9-ALS, n = 4); box plots compare HC (n = 9) versus pooled ALS (n = 9), with point colors indicating diagnosis subtype. Box plots show median (centerline), IQR (box) and 1.5× IQR whiskers. Statistical significance was assessed using a two-sided Wilcoxon rank-sum test. c, A representative spinal cord section indicating manually labeled anterior horn and MN-enriched spots. d, Differential gene expression comparing ALS to HC in MN-enriched spots identified upregulation of complement genes (*C1QA*, *C1QB*, *C1QC* and *APOC1*), and downregulation of *STMN2* and *NEFL*. e, Pseudobulk normalized expression of selected genes in MN-enriched spatial spots from the anterior horn. Each point represents one donor (HC, n = 9; sALS, n = 2; C9-ALS, n = 4 biological replicates; ALS total n = 6). Plots represent mean ± s.e.m. Outliers (>2 s.d. from the group mean) are shown as hollow points and were excluded from statistical testing. Statistical significance was assessed using two-sided Wilcoxon rank-sum test. f, Representative IHC images of APOC1 and CD163 in anterior horns of ALS and control participants. g, Representative images of wild-type TDP-43 localization, nuclear clearance and cytoplasmic aggregation in neurons. h, Schematic of TDP-43 pathology mapping in ALS cervical spinal cord, highlighting TDP-43 pathology and adjacent regions (top), and C2L-annotated neuron-enriched ST spots (bottom); regions outside of anterior horns are grayed out. i, Differential gene expression analysis comparing TDP-43-proximal and TDP-43-distal neuron-enriched spots. j, Expression levels of *APOC1*, *C1QA*, *C1QB* and *C1QC* in bulk RNA-seq spinal cord samples of patients with ALS in which cryptic *STMN2* was present (n = 118) or absent (n = 113). Box plots show median (centerline), IQR (box) and 1.5× IQR whiskers. Statistical significance determined by Wilcoxon rank-sum exact test; P values are shown within each panel in j. For d and i, P values were calculated using MAST and adjusted for multiple comparisons using the Benjamini–Hochberg method. Avg, average; wt, wild-type.

Source data

To investigate the spatial association between inflammation and ALS neuropathology, we performed IHC for TDP-43 and integrated the imaging with the SP data. Mislocalization and cytoplasmic aggregation of TDP-43 occur in affected neurons in more than 97% ALS cases^46,47. However, it remains unclear whether TDP-43 pathology contributes to MN loss through neuroinflammation. To address this, we categorized MNs localized in the anterior horn based on TDP-43 staining as having wild-type nuclear localization, pathological nuclear clearance or pathological cytoplasmic aggregation (Fig. 6g and Extended Data Fig. 6i). We then integrated these TDP-43 pathology annotations with the spatial transcriptomic (ST) assay, aligning them to C2L neuronal-enriched spots (Fig. 6h). Differential expression analysis of neuron-enriched anterior horn spots showed upregulation of inflammatory genes in regions proximal to TDP-43 pathology compared with distant neuron-enriched spots (Fig. 6i and Source data Fig. 6i). These included complement and innate immune signaling (C3, C1QA–C1QC, C1S), myeloid receptors and activation programs (TREM2, CD14, CD74, CEBPB), lipid handling and scavenger pathways (APOE, APOC1, CD163), lysosomal and antigen-processing functions (CD68, CTSS, IFI30, GPNMB), inflammatory and stress-associated mediators (S100A9, S100A6, CHI3L1) and general cytoskeletal or metabolic stress responses (ACTB, SOD2), together with secreted immune-associated factors such as SPP1 and FCGFBP. Intriguingly, CST3, which encodes cystatin C, a key endogenous inhibitor of lysosomal cysteine proteases, was downregulated, suggesting reduced inhibition of cathepsin activity in TDP-43-proximal regions. (Fig. 6i and Source data Fig. 6i). Permutation-based spatial controls showed that gene-level enrichment near TDP-43 pathology was modest at the resolution of individual spots, warranting a tempered interpretation of the transcriptional associations. We visualized the signal intensity of several of these markers in the anterior horn in relation to TDP-43 pathology (Extended Data Fig. 6j).

To further investigate the correlation between TDP-43 pathology and inflammation, we examined the NYGC bulk RNA-seq dataset (Fig. 3) and used mis-splicing of STMN2 (ref. ³⁷) as a molecular surrogate of TDP-43 dysfunction. We found a strong positive correlation across ALS cases that exhibited mis-spliced STMN2 and expression of APOC1, C1QA, C1QB and C1QC (Fig. 6j). Notably, the proportion of immune cells—specifically macrophages and microglial cells, but not lymphocytes—was significantly increased in the presence of cryptic STMN2 in ALS spinal cord samples (Extended Data Fig. 6k). Finally, the presence of cryptic STMN2 was significantly associated with shorter disease duration (Extended Data Fig. 6l). These findings suggest a potential link among TDP-43 pathology, complement activation and neuroinflammation in ALS.

Discussion

By integrating single-cell and bulk transcriptomics with SP, we mapped convergent and divergent immune signatures across blood and spinal cord tissues and linked them to MN pathology in ALS. Our data indicate that (1) C9orf72 haploinsufficiency is associated with broad immune remodeling in carriers, (2) innate immune activation in monocytes diverges by subtype and progression rate, (3) antigen-experienced CD8 T_EM cells display transcriptional and clonal features consistent with disease-relevant specificity and (4) complement-programmed and lipid-programmed myeloid responses spatially converge with MN loss and TDP-43 pathology in the spinal cord. These findings link peripheral and central immunity to ALS heterogeneity and nominate stratified immunomodulation strategies.

We observed broad downregulation of C9orf72 transcripts across circulating immune populations in C9-ALS, most pronounced in CD16⁺ monocytes and B cell subsets. This aligns with prior studies on humans showing reduced C9orf72 expression in expansion carriers^48,49. Genetic and experimental evidence demonstrates that C9orf72 regulates myeloid vesicle trafficking, autophagy and immune tone, and that its loss skews macrophages and microglia toward inflammatory states^7,8,50. Together, these data support a model in which C9orf72 haploinsufficiency establishes a permissive background for aberrant innate immune activation in carriers.

Stratification of monocytes revealed two distinct axes. In sALS, CD16⁺ monocytes were the most transcriptionally dysregulated subtype, with complement genes (C1QA/C1QB), the scavenger receptor MSR1 and ISGs (for example, IFITM3) upregulated and associated with faster progression. These changes indicate both classical-pathway priming and interferon-responsive remodeling. Increased complement gene expression dovetails with human biomarker studies reporting elevated complement activation in ALS plasma and CSF, as well as with immunohistochemical evidence of C1q at innervated motor end plates, consistent with early neuromuscular junction involvement^17,51,52. MSR1 supports uptake of modified lipoproteins, apoptotic bodies and debris⁵³, thereby linking the transcriptional programs observed in sALS-fast monocytes to heightened phagocytic and inflammatory activity. In contrast, in C9-ALS, CD14⁺ monocytes showed the greatest transcriptional changes, including upregulation of CLEC7A, TRIB1 and EGR1/EGR3 alongside ISGs (for example, IFITM2/3). This profile suggests a skewed pattern-recognition and interferon-responsive program, with IFITM3 acting as a membrane-modifying ISG that tunes innate signaling downstream of viral and toll-like receptor stimuli^54,55.

Adaptive immune alterations were prominent across patients with ALS, with CD8 T_EM cells showing a high DEG burden and clonal features consistent with prior antigen exposure. These findings align with human studies reporting activated T cell profiles in blood and CSF that stratify progression, as well as with evidence of clonal expansion in ALS^56–59. Although our cohort was not powered to define definitive Vβ usage biases, the convergence of transcriptional activation, clonal expansion and prognostic associations across datasets supports a role for antigen-driven T cell responses in ALS. This interpretation, additionally supported by bulk RNA-seq from a large ALS spinal cord cohort that revealed increased T cell signatures after deconvolution, is consistent with a recent single-cell study showing T cell and natural killer cell enrichment in CNS tissue⁶⁰. While prior studies have described shifts in immune-cell proportions in ALS, particularly among T cells and natural killer cells^58,61–64, we did not observe consistent changes in the relative abundance of major adaptive immune populations across ALS subtypes. This discrepancy likely reflects technical limitations of scRNA-seq in accurately quantifying cell types in modestly sized cohorts, rather than a lack of adaptive immune engagement. Bulk transcriptomics highlighted the spinal cord as the most transcriptionally perturbed site in ALS, showing the largest DEG burden and strongest immune-cell enrichment. In this tissue, reduced neuronal signatures co-occurred with increased myeloid and lymphoid cells and fewer oligodendrocytes, consistent with prior studies^25,65. Notably, spinal cord immune-cell activation correlated strongly with shorter disease duration but not with age of onset, suggesting that inflammation is a downstream rather than initiating event in ALS. Deconvolution stratified by disease duration further revealed stronger immune-cell signatures in fast progressors compared to slow progressors, consistent with our single-cell peripheral immune profiling, where fast cases showed more pronounced complement-associated and interferon-associated gene programs in monocytes.

SP localized immune signals to the anterior horn and adjacent white matter of ALS spinal cord tissue. In these regions, we observed upregulation of complement genes (C1QA/C1QB/C1QC; C3) and a macrophage-associated lipid signature (for example, APOC1), together with downregulation of MN-enriched transcripts (STMN2, NEFL). These spatial patterns align with evidence that microglial and macrophage states are tightly coupled to lipid handling and phagocytosis in neurodegeneration^66,67. APOC1 has been implicated in macrophage polarization programs and reported to interact with CD163/CD206 in other disease contexts⁴⁵. While largely derived from oncology, these data support an APOC1–myeloid axis relevant to inflamed, lipid-rich niches. Converging with these spatial observations, our bulk RNA-seq of a large ALS spinal cord cohort showed complement pathway enrichment and higher APOC1 expression, both associated with shorter disease duration. Together, these findings extend prior histopathological reports of complement activation⁶⁸ and point to a complement-driven immune response that may contribute to disease progression.

TDP-43 proteinopathy, present in the vast majority of ALS cases, disrupts RNA metabolism and induces mis-splicing events in several target mRNAs, including STMN2 (refs. ^37,40,69,70). Our SP map of the spinal cord revealed reduced NEFL and STMN2 in MN-enriched spots alongside upregulated complement and myeloid activation markers. These immune alterations, including complement activation (C1QB), microglial receptor signaling (TREM2), lipid and scavenger pathways (APOC1, CD163, ABCA1) and lysosomal/stress responses (IFI30, CTSZ, GPNMB, S100A9/S100A910, CHI3L1), were concentrated near regions of TDP-43 pathology. The spatial convergence of innate immune activation, NEFL/STMN2 loss and TDP-43 pathology within degenerating MNs suggests that TDP-43 dysfunction may mechanistically link RNA misprocessing to immune effector pathways, thereby amplifying MN degeneration and fueling local immune responses. One mechanistic hypothesis is that aberrant splicing generates neopeptides presented on major histocompatibility complex molecules as neoantigens, driving adaptive immune activation. Indeed, recent studies demonstrate that TDP-43-dependent cryptic neopeptides are detectable in ALS/FTD CSF^71,72 and may be recognized by CD8⁺ T cells in ALS and inclusion body myositis⁷³. Relatedly, cryptic antigens have recently been shown to be targets for T cell recognition in cancer⁷⁴. Although our deconvolution and mis-splicing analyses of bulk RNA did not reveal a direct association between TDP-43-associated mis-splicing and increased T cell frequency, larger studies will be required to clarify how RNA misprocessing intersects with adaptive immune activation in ALS. Consistent with this framework, recent studies have reported autoreactive T cell responses against C9orf72 in patients with ALS, supporting a potential autoimmune component in the disease⁷⁵.

Our data highlight bidirectional immune crosstalk at the neuroimmune interface in ALS. We found increased complement signaling molecules in both PBMCs and CNS tissues. Notably, randomized trials targeting the terminal complement pathway (C5; zilucoplan, ravulizumab)^76,77 have been negative, suggesting that earlier-stage or compartment-specific complement activity (for example, C1q at synapses or neuromuscular junctions) may represent a more tractable target. Alternatively, immune endophenotype-based patient selection may be required⁷⁸. A C1q inhibitor (ANX005) is under clinical evaluation in ALS⁷⁹, but peer-reviewed efficacy data are not yet available. Beyond complement, our stratified monocyte and T cell signatures suggest precise immunomodulation. For example, interferon-tuned monocyte programs in C9-ALS or antigen-driven interventions for clonally expanded CD8 T_EM cells, paired with disease biomarkers, may guide patient selection and pharmacodynamic endpoints. Immune-targeting strategies may also need to be tailored to genetic and immunological context—complement or interferon blockade may be most effective in fast-progressing sALS, whereas targeting pattern-recognition and vesicle trafficking pathways may hold greater promise in C9-ALS.

Limitations of this work include sample size and limited power for repertoire analyses, underscoring the need for larger, longitudinal cohorts. Additionally, we acknowledge that vaccination or recent infection history could potentially influence TCR repertoire expansions; however, such metadata were not available for this cohort. Finally, while mobility and activity levels have been reported to influence immune profiles⁸⁰, such data were also not available.

In summary, integration of multimodal data across blood and CNS demonstrates that ALS neuroinflammation is neither homogeneous nor incidental but stratified by genetic background and progression rate. We identify monocyte subset-specific programs, complement dysregulation and clonally expanded T cells as central features of ALS, converging on MN loss and TDP-43 pathology in the spinal cord. Together, these findings provide a framework for mechanistic stratification and highlight immune pathways with potential for biomarker development and therapeutic intervention.

Methods

Study participants

PBMCs were acquired through the Houston Methodist Neurological Institute. All participants were free from acute infectious diseases and were in good physical condition. Bulk RNA-seq data were acquired from the Target ALS Data Engine. Spinal cord sections were acquired from the Target ALS Multicenter Human Postmortem Tissue Core. The study of PBMCs and spinal cord tissues was approved by the Institutional Review Board of Northwestern University. No statistical methods were used to predetermine sample sizes but our sample sizes are similar to those reported in previous publications^81–84.

Ethics declarations for human tissues

Written consent was obtained from all participants who provided tissues for this study. The study of de-identified PBMCs and spinal cord tissues was approved by the Institutional Review Board of Northwestern University.

Human tissues

Blood was isolated from patients through venipuncture. PBMCs were isolated from blood by layering diluted blood (1:1 in PBS) on top of an equal volume of Ficoll, followed by centrifugation and isolation of the buffy coat. PBMCs were frozen in a mixture of FBS (90%) and dimethyl sulfoxide (10%) overnight at −80 °C in a Mr. Frosty freezing container (Thermo Fisher Scientific) and transferred the following day to liquid nitrogen for storage. No formal randomization was performed, as samples were collected based on predefined clinical and experimental criteria. Analysis for microglia and neuron quantifications (Fig. 4b and Extended Data Fig. 5d,l) was performed blindly to experimental conditions. For all other experiments, data collection and analysis were not performed blindly to the conditions of the experiments.

Immune profiling of PBMCs

PBMCs were sorted on Sytox Blue live/dead dye on the Sony SH800 in FACS buffer containing PBS with 0.4% BSA. A total of 100,000 cells were sorted into 750 µl FACS buffer, centrifuged at 350g for 7 min and resuspended in 85 µl FACS buffer to achieve ~1,000 cells per µl. Furthermore, 16.5 µl of cells were mixed with 23-µl water, then prepared according to the 10x Genomics Next GEM 5′ (v1.1) protocol with single indexes. cDNA was processed for whole RNA and TCR libraries. To avoid batch effects, all 40 samples were sorted and loaded into GEMs on the same day and all libraries were generated simultaneously using a 96-well PCR plate and a plate magnet.

scRNA-seq + TCR profiling

scRNA-seq and TCR-seq of PBMCs were performed using the 10x Genomics Chromium Next GEM Single Cell 5′ (v2) platform with the immune profiling kit. Library preparation followed the standard 10x Genomics protocol. Sequencing was conducted by Novogene on an Illumina NovaSeq 6000 system, with base calling performed using the Illumina RTA3 method. RNA reads were aligned to the hg38 reference genome, and gene expression matrices were generated using Cell Ranger (v6.0.0). TCR reads were also aligned to hg38, and clonotype/contig matrices were produced using Cell Ranger.

scRNA-seq preprocessing

Filtered scRNA feature barcode matrices are generated with Cell Ranger (v8.0.0) count module against reference GRCh38 and—expect-cells 10,000. We then corrected for contamination using SoupX (v1.6.2) with default settings. Doublets are removed from the corrected matrices using DoubletFinder (v2.0.4) using doubletFunder(), PCs = 1:12, pN = 0.25, pK set to optional value calculated per recommendation. Cells with more than 10% mitochondrial expression are removed from the matrices. Clonal expansion and receptor sequences for each sample are generated with Cell Ranger vdj module against Ensembl 7.1.0 reference with default settings. Nonimmune cells assigned with receptors are removed. Finally, only cells with at least 250 UMI and 500 features are retained. Cleaned counts are further processed per Seurat’s (v5.0.1) integration vignette with NormalizeData(), FindVariableFeatures(), ScaleData and RunPCA(). We then integrated the Seurat object using IntegrateLayers() with method = HarmonyIntegration. Finally, Harmony UMAP is generated using RunUMAP() with reduction = ‘harmony’, dims = 1:30. Cell-type annotation is based on a reference CITE-seq dataset of 162,00 PBMC cells measured with 228 antibodies acquired from Seurat. SCTransform() is called on all samples with vst.flavor = ‘v2’. Labels are then mapped from the reference dataset using Seurat’s FindTransferAnchors() and TransferData() functions with reference.reduction = ‘spca’ and dims = 1:50.

CRISPR-powered scRNA profiling

scRNA libraries were processed using the CRISPRclean Single Cell RNA Boost Kit (Jumpcode Genomics, KIT1018) to remove highly expressed ribosomal and mitochondrial transcripts, as well as nonvariable genes, before sequencing. After cDNA amplification and adaptor ligation with the 10x Genomics Chromium Next GEM Single Cell 3′ Reagent Kit (v3.1; 10x Genomics, 1000268), CRISPR ribonucleoprotein complexes were formed by incubating Cas9 nuclease (Jumpcode Genomics, REA1000) with guide RNAs (Jumpcode Genomics, REA1025-003, REA1120-001) targeting undesired sequences. CRISPR digestion was performed at 42 °C for 60 min, followed by AMPure XP bead-based size selection (Beckman Coulter, A63881) to remove cleaved fragments. The remaining library was eluted in nuclease-free water and carried forward to sample index PCR and double-sided size selection using SPRIselect beads (Beckman Coulter, B23318), as per the 10x Genomics Next GEM (v3.1) single-cell 3′ gene expression protocol. Libraries were quantified using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Q32854) on a Qubit Fluorometer (Thermo Fisher Scientific, Q33238), and fragment distributions were assessed on an Agilent 2100 Bioanalyzer (Agilent Technologies) before sequencing on an Illumina NovaSeq 6000 (Illumina).

scRNA-seq differential expression gene analysis

For our scRNA-seq PBMC dataset, we excluded samples containing fewer than three cells for any particular cell type. We also removed any features for which neither group exhibited an expression level above 10%, or both groups had expression levels below 1%. The expression level is calculated with log-normalized counts. Confound variables of age and sex are included for all analyses, except for comparing C9orf72 expansion carriers against female controls. In this case, only age is used as a covariate. The differential expression is performed with Seurat’s (v5.0.1) FindMarkers() function on the SCT assay and the test, after correcting for each sample’s SCT model with PrepSCTFindMarkers(). The MAST P values were adjusted with the Benjamini–Hochberg method. The significance cutoff for all PBMC scRNA-seq differential expression analyses is an adjusted P value of 0.01 and a log₂ fold change (FC) of ±0.585. To identify differential expression in clonally expanded cells, we similarly analyzed cells whose receptors were sequenced more than once, as described above. Notably, only cell types with more than five clonally expanded cells in both comparison groups are retained for differential expression analysis.

Bulk RNA-seq analysis of NYGC ALS Consortium and Target ALS datasets

Alignment

RNA-seq data from patients with ALS and controls were downloaded from the Target ALS website (www.targetals.org) and NYGC GEO repositories (GSE137810), respectively. Cutadapt (v3.4) was used to trim Illumina adaptors from raw reads. Low-quality bases were removed from the end of reads (-q 20,20) and reads that were less than 25 bp after trimming were removed (-m 25). Trimmed reads were then aligned to the GRCh38 genome with gene annotations from Gencode V38 using STAR (v2.7.5a). STAR was run in two pass modes (--twopassMode Basic), along with ENCODE options (--outFilterType BySJout, --outFilterMultimapNmax 20, --alignSJoverhangMin 8, --alignSJDBoverhangMin 1, --outFilterMismatchNmax 999, --outFilterMismatchNoverReadLmax 0.04, --alignIntronMin 20, --alignIntronMax 1000000, --alignMatesGapMax 1000000). After alignment, gene-level counts were determined by featureCounts (v2.0.1).

Differential expression

DESeq2 (v1.46.0) was used for differential expression analysis among healthy participants and either ALS, ALS–FTD or FTD disease groups in each tissue. If any disease group had fewer than three samples in a given tissue, all samples from that group were removed from the analysis. The known variables of sex, RNA integrity number and age at death were controlled for the design matrix. Vst-normalized expression values were supplied to the R package ‘sva’ (v3.54.0) to estimate additional covariates modeling latent variability in the data. A total of 21 surrogate variables were estimated for cerebellum samples; 20, 23, 7 and 12 surrogate variables were estimated for frontal, motor, occipital and temporal cortices, respectively; and 19, 19 and 12 surrogate variables were estimated for cervical, lumbar and thoracic spine, respectively. These surrogate variables were added to the design matrix, differential expression analysis was performed with the ‘DESeq2()’ command and results were extracted with the results() command, specifying either the ALS, ALS–FTD or FTD disease group as the coefficient of interest.

Cell-type enrichment

To determine whether differential expression results were driven by changes in underlying cell populations, we performed GSEA using markers of different cell types from the Broad Institute’s MSigDB. First, samples were ranked by their Wald statistic from the DESeq2 analysis, with the most upregulated genes ranked highest and the most downregulated ranked lowest. We then applied GSEA using the R package ‘fgsea’ against the FAN_EMBRYONIC_CTX cell-type markers from the cell-type markers/C8 terms from MSigDB.

Bulk RNA-seq downsampling analysis

To determine how robust and reproducible our DEG results were in each tissue, we applied a downsampling approach to our dataset. For each tissue, initial ten samples were drawn randomly without replacement from the control and ALS groups. Sex, RNA integrity number and age of death were used as covariables, and surrogate variables were estimated with sva and added to the design matrix as in the main analysis; then the number of DEGs between the ALS and healthy groups was determined. This process was repeated 1,000 times. The size of the downsampled was increased by 10 samples (for example, next 20 samples were drawn randomly), and this process was repeated. The size of the downsampled group was increased by ten samples each iteration until all control samples from a tissue were sampled. Distribution of the number of DEGs per iteration was visualized in ggplot2.

Quantification of truncated STMN2

Truncated STMN2 levels were quantified by counting the number of correctly spliced and mis-spliced STMN2 reads in each sample. Regtools junctions extract was first run on each sample to extract all exon–exon junction counts, then, for each sample, the number of reads spanning the STMN2 exon1–exon2 junction and the exon1–cryptic exon junction were determined. Proportion of mis-splicing was calculated as E1–CE reads/(E1–E2 reads + E1–CE reads). Samples were additionally binarized into those with evidence of TDP-43 dysfunction (that is, having at least one read supporting the STMN2 E1–CE junction) and those with no evidence of TDP-43 dysfunction.

scRNA-seq dataset processing and cellular deconvolution for bulk RNA-seq analysis

Cortex

Processed counts and metadata from ref. ³⁰ were downloaded from Synapse (syn53421586). Counts and metadata for each sample were compiled into individual Seurat objects, then combined into a single Seurat object. To control for batch effects, samples were integrated using Harmony. For downstream analysis, the Seurat object was split into frontal and motor cortex-derived cells, then subsampled based on cell identity.

Spinal cord

Aggregated counts and metadata for each cell were downloaded from the publicly available GEO accession (GSE190442) associated with the spinal cord atlas discussed in ref. ³². Counts and metadata were used to create a Seurat object. The author’s original ‘MN’ group identified cells with relatively few unique expressed genes compared to other cell types, and the group’s marker genes included nonspecific, ubiquitously expressed genes. As such, we removed these cells and defined our own MN population in this dataset. To this end, we used a dataset containing neurons isolated from cervical spine (GSE228778) associated with the details mentioned in ref. ³¹. We first determined genes in this dataset that were upregulated in the MN population compared to all other neurons present in the dataset. Next, we used the most significant genes from this analysis to apply an ‘MN’ score to each cell in the dataset discussed in ref. ³². Briefly, genes with FDR of <1 × 10⁻⁵ and log(FC) of >0 were used as highly specific MN markers. These genes were used in the AddModuleScore() command from Seurat to calculate a score of how highly these genes were expressed compared to a set of control genes. After applying this score to all cells, the five cells with the highest MN score were called ‘putative MNs’.

Cell-type deconvolution

To estimate the proportion of cell types in each tissue, CIBERSORTx was applied to cortical and spinal tissue bulk RNA-seq datasets. First, both the cortical and spinal cord single-nucleus datasets were subsampled, taking 100 cells at random from each cell type. To account for region-specific differences in cortical cell types, we also sampled cells from only the motor and frontal cortices to create motor-specific and frontal-specific signature matrices. Then, four separate signature matrices were created for each tissue from the subsampled datasets. For each of these, the fraction was set to 0.5 and replicates were set to 25. Next, TMM-normalized counts per million were generated for each set of samples, using the calcNormFactors() and cpm() commands from edgeR (v4.4.2). Normalized counts for each group were supplied to the CIBERSORTx fractions module to estimate the proportion of inferred cell types for each sample. Datasets derived from frontal, motor and other cortical tissues were deconvoluted with the motor, frontal and combined cortex signature matrices, respectively, while spinal cord datasets were deconvoluted with the spinal matrix.

Correlation of gene expression and cell-type changes in ALS with disease duration and age of onset

To determine the relationship between gene expression and clinical features such as disease duration, we examined Pearson correlation between log-transformed TMM-normalized expression and various clinical data. P values were determined using linear regression with the lm() function in R and then corrected for multiple testing using FDR.

Pseudobulk gene expression analysis

Selected DEGs are pseudobulked by averaging the log-normalized SCT counts within each sample. A rank-sum test is performed on the same groups as conducted with MAST. Outliers defined as two standard errors away from the mean are excluded from the P-value calculation.

GSEA for scRNA-seq

We used the hallmark gene sets from the human MSigDB. We then calculated the ranking of genes ordering in descending order, the product of multiplying the −1 × log₁₀(P_adj) with sign(log₂(FC)) for each gene. This ranking is then used to sort the genes in increasing order and passed to fgsea() from fgsea (v1.28.0) with minSize = 15, maxSize = 500 to identify enriched pathways.

Fluorescence immunostaining, imaging and quantification

Five-micrometer-thick FFPE spinal cord sections were deparaffinized, decrosslinked by boiling for 10 min in Antigen Decloaker buffer (Biocare Medical, CB910M), and stained with the appropriate antibodies depending on the target of interest and DAPI (Thermo Fisher Scientific, 62248). The following antibodies were used: rabbit anti-IBA1 (1:1,000; WAKO, 019-19741), chicken anti-MAP2 (1:2,500; Abcam, ab5392), anti-CD68 (1:400; Abcam, ab955) and mouse LSP1 (1:100; Novus Biologicals, NBP2-79725). TrueBlack Plus Lipofuscin Autofluorescence Quencher (Biotum, 23007) was used according to the manufacturer’s instructions. Whole spinal cord slide images were acquired using a Nikon Ti2 fluorescence imaging system with a ×10 objective, using the NIS-Elements JOBS module. The quantification of microglia/macrophage cells across the entire section was performed using the NIS-Elements General Analysis module. Background intensity was corrected using a rolling-ball algorithm, and cells were identified through fluorescence signal thresholding combined with colocalization with the DAPI signal, ensuring that only cells present within the observed section and stained with DAPI were counted. To exclude small or transversely sectioned microglial processes, a minimum cytoplasmic area constraint was applied during segmentation. Microglia cell area and shape were assessed through the same NIS-Elements General Analysis module. The shape evaluation was based on the roughness value, which ranges from 0 to 1, where a value of ‘1’ indicates minimal roughness (that is, a perfectly circular object). Digitalized images of microglia cell distribution were obtained by plotting the centroid of each identified cell as a single dot on the section. CD68 images were acquired using a Nikon W1 fluorescence imaging system with a ×40 objective. To evaluate the number of macrophages in the cervical spinal cord, IBA1⁺/CD68⁺/DAPI⁺ cells were counted per field of view in ten fields (white matter) or four fields (gray matter) and averaged for each individual sample. The analysis included two control and four ALS samples.

SP

FFPE spinal cord sections were cut at 5-µm thickness onto glass slides. Tissues were deparaffinized, decrosslinked and stained with a combination of DAPI (1:100; Thermo Fisher Scientific, 62248), goat anti-IBA1 (1:250; Abcam, 107159), chicken anti-MAP2 (1:2,500; Abcam, ab5392) and rabbit anti-TDP-43 (1:500; ProteinTech, 80001-1-RR). Notably, we used TrueBlack Plus Lipofuscin Autofluorescence Quencher (Biotum, 23014) according to the manufacturer’s instructions. Tissues were imaged on a Nikon A1 confocal imaging system using a ×40 objective. After imaging, spatial gene and protein expression libraries were immediately prepared according to the user guide for Visium CytAssist Spatial Gene and Protein Expression Reagent Kits (Rev B, CG000494; 10x Genomics). We used the Visium Human Transcriptome Probe Set (v2.0) for RNA transcript detection, along with the Human FFPE Immune Profiling Panel, which includes a 35-plex CytAssist panel of antibodies, both intracellular and extracellular, sourced from BioLegend and Abcam for protein detection. This panel also comprises four isotype controls. Final libraries were sequenced as detailed above for ST. The targeted sequencing depth was 25,000 reads per tissue-covered spot for gene expression libraries, and 5,000 reads per tissue-covered spot for protein expression libraries, as recommended.

SP data processing

Spatial alignment between IHC images and Visium data was performed using the recommended 10x Genomics Visium CytAssist workflow in Loupe Browser (v7). Briefly, manual, image-based registration was performed by selecting corresponding anatomical landmarks between the IHC and CytAssist images to generate an alignment file. Alignment was performed without prior knowledge of sample identity or comparison groups. After initial alignment, images were visually inspected and iteratively adjusted to minimize any discontinuities, consistent with developer guidelines. We removed spots outside of tissue manually using Loupe Browser and outlier spots with extreme RNA or feature counts on a per-sample basis. Furthermore, spots with more than 20% mitochondrial expression are removed. The finalized alignment file was then used in SpaceRanger (v2.1.1) to generate spatially aligned RNA and protein matrices. We then performed SCTransform on each sample with mitochondrial expression as a covariate and variable.features.n = 8000. We then manually annotated each slide. Annotation of anterior white matter and posterior white matter is based on the anatomy of cervical spinal and the MAP2 staining. Posterior horns and anterior horns are annotated based on a combination of transcriptomic signatures.

TDP-43 annotation and differential gene analysis

TDP-43 pathology was manually annotated based on immunostaining, and spatial spots containing TDP-43 signal, as well as their immediately adjacent spots on the Visium array, were classified as TDP-43+. Differential expression analysis was performed between TDP-43⁺ neuron-enriched spots and remaining neuron-enriched spots, restricted to the anterior horn region. Benjamini–Hochberg correction was used to correct for statistical significance. P values were adjusted with the Benjamini–Hochberg method. Adjusted P value of 0.01 and a log₂(FC) of ±0.585 are used as the significance threshold. Permutation of TDP-43 labels was performed to confirm that observed signals were not driven by spatial or technical artifacts.

SP differential expression gene analysis

We first subsetted our data to the desired morphological annotation region or enriched cell types. We then used the log-normalized SCT counts to calculate the FCs of compared groups. We again removed any features for which neither group exhibited an expression level above 10%, or both groups had expression levels below 1%. Ribosomal, mitochondrial and hemoglobin genes were removed from these comparisons. Cellular detection rate (CDR) was calculated based on the normalized RNA counts. We then used Seurat’s PrepSCTFinderMarkers() with default settings to correct SCTransform counts for each sample and exported the log-transformed SCT counts to create a sca object with MAST’s (v1.20.0) FromMatrix() with default settings. We then used the zlm function on this object with centered CDR and sample ID as covariates. The MAST P values were adjusted with the Benjamini–Hochberg method. Adjusted P value of 0.01 and a log₂(FC) of ±0.585 are used as the significance threshold.

SP DEP analysis

We subsetted our data to morphological annotation regions or enriched cell types. We used the isotype-normalized counts data output from SpaceRanger (v2.1.1). We used a negative binomial generalized linear model using Seurat’s (v5.0.1) FindMarkers(), min.pct = 0.01, logfc.threshold = -Inf, test.use = ‘negbinom’. We then used the Benjamini–Hochberg method to adjust the raw P values. Adjusted P value of 0.01 and a log₂(FC) of ±0.585 were used as the significance threshold.

Deconvolution of SP spots

We used a reference scRNA-seq dataset³² but modified the MN population using a separate dataset³¹. Mitochondrial and hemoglobin genes were removed from the training process. Additional filtering was performed with the filter_genes function using cell_count_cutoff = 5, cell_percentage_cutoff2 = 0.03, nonz_mean_cutoff = 1.12. Batch_key parameter was configured as sample ID. We trained the reference model using C2L’s (v0.1.3) mod.tran() with max_epochs = 500, batch_size = 2500, train_size = 1, lr = 0.002. Next, we calculated the proportion of genes expressed per spot (CDR) from raw counts and standardized this value. We then used c2l.models.Cell2location to map cell types to our SP data with N_cells_per_location = 7 and detection_alpha = 20 to account for technical variabilities. We then trained this model with centered CDR and UMI as covariates for 5,000 epochs.

Annotation of cell-type enrichment

We exported the 5% quantile of the posterior distribution from the C2L to represent the deconvoluted cell number at each spot. We then normalized each cell-type score by dividing the number of cells of that type by the total cell number at that spot. We defined a spot to be enriched for a cell type if it was four median absolute deviations away from the median score for that cell type within each sample. Furthermore, a spot was considered enriched for astrocytes and oligodendrocytes if the normalized cell-type score was above 30 and 60, respectively.

MN-enriched differential gene expression analysis

MN-enriched ST spots from the anterior horns were selected based on MN distance grouping, restricting analyses to samples with at least ten MN-enriched spots. Binarized MN segmentation masks, manually generated from histological images, were used to define spatial proximity to MN signals. These masks were iteratively dilated to generate concentric bands of 10 µm thickness, with decreasing intensity encoding increasing distance from the MN signal. Spots overlapping the innermost band were classified as MN enriched. Differential expression between ALS and control conditions was assessed using a MAST hurdle model on SCTransform-normalized data, with disease status as the primary effect and the centered cellular detection rate included as a covariate. Statistical significance was adjusted using the Benjamini–Hochberg method. DEGs were defined by an adjusted P value of <0.05 and an absolute log₂(FC) of >log₂(1.5).

Quantification and statistical analysis

Statistical analyses were conducted primarily using R (v4.2.3) and GraphPad Prism (v10.2.1). In GraphPad Prism, normality and variance equality were assessed using the Shapiro–Wilk and F tests, respectively, to guide test selection. For two-group comparisons, we applied unpaired two-tailed Student’s t tests (with or without Welch’s correction for unequal variances) or Mann–Whitney tests. For comparisons involving more than two groups, we used one-way analysis of variance with Tukey’s multiple comparisons test, Welch analysis of variance with Dunnett’s T3 multiple comparisons test or Kruskal–Wallis test with Dunn’s multiple comparisons test. Statistical significance was defined as P < 0.05 for all analyses. Data distribution was assumed to be normal when such assumptions are needed to select statistical tests but this was not formally tested.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41593-026-02300-5.

Supplementary information

Supplementary Information^{(406.6KB, pdf)}

Supplementary Table 1.

Reporting Summary^{(67.2KB, pdf)}

Source data

Source Data Fig. 1^{(5.2MB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 2^{(2.6MB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 3^{(44.3MB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 4^{(4.6MB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 5^{(374.6KB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 6^{(306.1KB, xlsx)}

Source analysis result for figure production.

Acknowledgements

We thank the NYGC ALS Consortium and Target ALS for providing the bulk RNA-seq datasets, and the Target ALS Foundation Biorepository Core Facility for providing the postmortem patient specimens used for SP and IHC analyses. We also thank Tony Wyss-Coray (Stanford University) for helpful discussion. We are grateful to the following funding sources: US National Institutes of Health (NIH), National Institute on Neurological Disorders and Stroke and NIH National Institute on Aging (R01AG078713 to D.G. and R01NS104219 and R01NS134166 to E. Kiskinis), the Les Turner ALS Foundation (to E. Kiskinis and D.G.), the New York Stem Cell Foundation (to E. Kiskinis), Target ALS Foundation (to E. Kiskinis and D.G.) and BrightFocus Foundation (to D.G.). E. Kiskinis is a New York Stem Cell Foundation—Robertson investigator. D.G. is a recipient of the New Academic Investigator Award from Target ALS.

Extended data

Author contributions

E. Kiskinis and D.G. conceived of and supervised the study and secured funding. D.G. generated the scRNA-seq datasets. L.v.O., F.A., A.J.E., J.B., A.N., N.S., N.P., J.K. and A.R. generated spatial datasets. S.A., D.G., E. Kiskinis, F.A. and M.W. coordinated patient recruitment and sample collection. Z.Z., A.V.F. and T.W. analyzed the single-cell and spatial datasets. M.W. analyzed the bulk RNA-seq data. E. Kaspi and A.M. assisted with data analysis. F.A., A.M., E. Kaspi and L.v.O. performed, acquired and quantified immunofluorescence stainings. Z.Z., L.v.O., F.A. and M.W. designed and led the analyses, and interpreted the data with input from all authors. Z.Z., L.v.O., F.A., M.W., E. Kiskinis and D.G. wrote the paper with input from all authors.

Peer review

Peer review information

Nature Neuroscience thanks Sean Bendall, Aaron Burberry, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

Spatial RNA and scRNA-seq data have been deposited at GEO under accessions GSE288365 and GSE268995, respectively. Any additional information required to reanalyze the data reported in this paper, including IHC images, is available from the lead contacts upon request. Data can be explored and requested through a central hub located at https://sites.google.com/view/alsshiny/home. Source data are provided with this paper.

Code availability

All code used to generate the figures in this study can be found at https://github.com/gatelabNW/als_immune_public and https://github.com/KiskinisLab/TargetALS-BulkAnalysis.

Competing interests

E. Kiskinis is an academic cofounder of NuCyRNA Therapeutics and SAB member of Axion Biosystems, ResQ Biotech, Synapticure, Live Like Lou Foundation and Packard Foundation. D.G. has been a consultant/advisor for Merck related to anti-inflammatory therapy for ALS. Named companies and foundations were not involved in this project. The other authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Ziyang Zhang, Lynn van Olst, Francesco Alessandrini, Matthew Wright.

Contributor Information

Evangelos Kiskinis, Email: evangelos.kiskinis@northwestern.edu.

David Gate, Email: dgate@northwestern.edu.

Extended data

is available for this paper at 10.1038/s41593-026-02300-5.

Supplementary information

The online version contains supplementary material available at 10.1038/s41593-026-02300-5.

References

1.Feldman, E. L. et al. Amyotrophic lateral sclerosis. Lancet400, 1363–1380 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Beers, D. R. & Appel, S. H. Immune dysregulation in amyotrophic lateral sclerosis: mechanisms and emerging therapies. Lancet Neurol.18, 211–220 (2019). [DOI] [PubMed] [Google Scholar]
3.Henkel, J. S. et al. Presence of dendritic cells, MCP-1, and activated microglia/macrophages in amyotrophic lateral sclerosis spinal cord tissue. Ann. Neurol.55, 221–235 (2004). [DOI] [PubMed] [Google Scholar]
4.Kushner, P. D., Stephenson, D. T. & Wright, S. Reactive astrogliosis is widespread in the subcortical white matter of amyotrophic lateral sclerosis brain. J. Neuropathol. Exp. Neurol.50, 263–277 (1991). [DOI] [PubMed] [Google Scholar]
5.Schiffer, D., Cordera, S., Cavalla, P. & Migheli, A. Reactive astrogliosis of the spinal cord in amyotrophic lateral sclerosis. J. Neurol. Sci.139, 27–33 (1996). [DOI] [PubMed] [Google Scholar]
6.Engelhardt, J. I., Tajti, J. & Appel, S. H. Lymphocytic infiltrates in the spinal cord in amyotrophic lateral sclerosis. Arch. Neurol.50, 30–36 (1993). [DOI] [PubMed] [Google Scholar]
7.O’Rourke, J. G. et al. C9orf72 is required for proper macrophage and microglial function in mice. Science351, 1324–1329 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Burberry, A. et al. Loss-of-function mutations in the C9ORF72 mouse ortholog cause fatal autoimmune disease. Sci. Transl. Med.8, 347ra393 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Sheng, W., Ji, G. & Zhang, L. Role of macrophage scavenger receptor MSR1 in the progression of non-alcoholic steatohepatitis. Front. Immunol.13, 1050984 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Gudgeon, J., Marin-Rubio, J. L. & Trost, M. The role of macrophage scavenger receptor 1 (MSR1) in inflammatory disorders and cancer. Front. Immunol.13, 1012002 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Harmon, E. et al. Increased expression of interferon-induced transmembrane 3 (IFITM3) in stroke and other inflammatory conditions in the brain. Int. J. Mol. Sci.23, 8885 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Diamond, M. S. & Farzan, M. The broad-spectrum antiviral functions of IFIT and IFITM proteins. Nat. Rev. Immunol.13, 46–57 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Goldknopf, I. L. et al. Complement C3c and related protein biomarkers in amyotrophic lateral sclerosis and Parkinson’s disease. Biochem. Biophys. Res. Commun.342, 1034–1039 (2006). [DOI] [PubMed] [Google Scholar]
14.Apostolski, S. et al. Serum and CSF immunological findings in ALS. Acta Neurol. Scand.83, 96–98 (1991). [DOI] [PubMed] [Google Scholar]
15.Tsuboi, Y. & Yamada, T. Increased concentration of C4d complement protein in CSF in amyotrophic lateral sclerosis. J. Neurol. Neurosurg. Psychiatry57, 859–861 (1994). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Annunziata, P. & Volpi, N. High levels of C3c in the cerebrospinal fluid from amyotrophic lateral sclerosis patients. Acta Neurol. Scand.72, 61–64 (1985). [DOI] [PubMed] [Google Scholar]
17.Bahia El Idrissi, N. et al. Complement activation at the motor end-plates in amyotrophic lateral sclerosis. J. Neuroinflammation13, 72 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Schoggins, J. W. & Rice, C. M. Interferon-stimulated genes and their antiviral effector functions. Curr. Opin. Virol.1, 519–525 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Yang, W. et al. Blocking CCL3-mediated neutrophil recruitment into the brain alleviates immunopathology following severe enterovirus 71 infection. iScience27, 111388 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Martin-Sanchez, F. et al. Inflammasome-dependent IL-1β release depends upon membrane permeabilisation. Cell Death Differ.23, 1219–1231 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Gate, D. et al. Clonally expanded CD8 T cells patrol the cerebrospinal fluid in Alzheimer’s disease. Nature577, 399–404 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Oh, H. et al. Methods to investigate intrathecal adaptive immunity in neurodegeneration. Mol. Neurodegener.16, 3 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Piehl, N. et al. Cerebrospinal fluid immune dysregulation during healthy brain aging and cognitive impairment. Cell185, 5028–5039 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Rosati, E. et al. Overview of methodologies for T-cell receptor repertoire analysis. BMC Biotechnol.17, 61 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Humphrey, J. et al. Integrative transcriptomic analysis of the amyotrophic lateral sclerosis spinal cord implicates glial activation and suggests new risk genes. Nat. Neurosci.26, 150–162 (2023). [DOI] [PubMed] [Google Scholar]
26.Kabiljo, R., Iacoangeli, A., Al-Chalabi, A. & Rosenzweig, I. Amyotrophic lateral sclerosis and cerebellum. Sci. Rep.12, 12586 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Prell, T. & Grosskreutz, J. The involvement of the cerebellum in amyotrophic lateral sclerosis. Amyotroph. Lateral Scler. Frontotemporal Degener.14, 507–515 (2013). [DOI] [PubMed] [Google Scholar]
28.Fan, X. et al. Spatial transcriptomic survey of human embryonic cerebral cortex by single-cell RNA-seq analysis. Cell Res.28, 730–745 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Sun, N. et al. Human microglial state dynamics in Alzheimer’s disease progression. Cell186, 4386–4403 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Pineda, S. S. et al. Single-cell dissection of the human motor and prefrontal cortices in ALS and FTLD. Cell187, 1971–1989 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Gautier, O. et al. Challenges of profiling motor neuron transcriptomes from human spinal cord. Neuron111, 3739–3741 (2023). [DOI] [PubMed] [Google Scholar]
32.Yadav, A. et al. A cellular taxonomy of the adult human spinal cord. Neuron111, 328–344 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Steinacker, P. et al. Chitotriosidase (CHIT1) is increased in microglia and macrophages in spinal cord of amyotrophic lateral sclerosis and cerebrospinal fluid levels correlate with disease severity and progression. J. Neurol. Neurosurg. Psychiatry89, 239–247 (2018). [DOI] [PubMed] [Google Scholar]
34.Maxeiner, S. et al. Crucial role for the LSP1-myosin1e bimolecular complex in the regulation of Fcγ receptor-driven phagocytosis. Mol. Biol. Cell26, 1652–1664 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Lu, C. H. et al. Neurofilament light chain: a prognostic biomarker in amyotrophic lateral sclerosis. Neurology84, 2247–2257 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Lopez-Erauskin, J. et al. Stathmin-2 loss leads to neurofilament-dependent axonal collapse driving motor and sensory denervation. Nat. Neurosci.27, 34–47 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Klim, J. R. et al. ALS-implicated protein TDP-43 sustains levels of STMN2, a mediator of motor neuron growth and repair. Nat. Neurosci.22, 167–179 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Baughn, M. W. et al. Mechanism of STMN2 cryptic splice-polyadenylation and its correction for TDP-43 proteinopathies. Science379, 1140–1149 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Prudencio, M. et al. Truncated stathmin-2 is a marker of TDP-43 pathology in frontotemporal dementia. J. Clin. Invest.130, 6080–6092 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Melamed, Z. et al. Premature polyadenylation-mediated loss of stathmin-2 is a hallmark of TDP-43-dependent neurodegeneration. Nat. Neurosci.22, 180–190 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Kleshchevnikov, V. et al. Cell2location maps fine-grained cell types in spatial transcriptomics. Nat. Biotechnol.40, 661–671 (2022). [DOI] [PubMed] [Google Scholar]
42.Stevens, B. et al. The classical complement cascade mediates CNS synapse elimination. Cell131, 1164–1178 (2007). [DOI] [PubMed] [Google Scholar]
43.Peterson, S. L., Nguyen, H. X., Mendez, O. A. & Anderson, A. J. Complement protein C3 suppresses axon growth and promotes neuron loss. Sci. Rep.7, 12904 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Walker, D. G. & McGeer, P. L. Complement gene expression in human brain: comparison between normal and Alzheimer disease cases. Brain Res. Mol. Brain Res.14, 109–116 (1992). [DOI] [PubMed] [Google Scholar]
45.Ren, L. et al. Systematic pan-cancer analysis identifies APOC1 as an immunological biomarker which regulates macrophage polarization and promotes tumor metastasis. Pharmacol. Res.183, 106376 (2022). [DOI] [PubMed] [Google Scholar]
46.Neumann, M. et al. Ubiquitinated TDP-43 in frontotemporal lobar degeneration and amyotrophic lateral sclerosis. Science314, 130–133 (2006). [DOI] [PubMed] [Google Scholar]
47.Arai, T. et al. TDP-43 is a component of ubiquitin-positive tau-negative inclusions in frontotemporal lobar degeneration and amyotrophic lateral sclerosis. Biochem. Biophys. Res. Commun.351, 602–611 (2006). [DOI] [PubMed] [Google Scholar]
48.Masrori, P., Beckers, J., Gossye, H. & Van Damme, P. The role of inflammation in neurodegeneration: novel insights into the role of the immune system in C9orf72 HRE-mediated ALS/FTD. Mol. Neurodegener.17, 22 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Belzil, V. V. et al. Reduced C9orf72 gene expression in c9FTD/ALS is caused by histone trimethylation, an epigenetic event detectable in blood. Acta Neuropathol.126, 895–905 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Atanasio, A. et al. C9orf72 ablation causes immune dysregulation characterized by leukocyte expansion, autoantibody production, and glomerulonephropathy in mice. Sci. Rep.6, 23204 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Kjaeldgaard, A. L. et al. Complement profiles in patients with amyotrophic lateral sclerosis: a prospective observational cohort study. J. Inflamm. Res.14, 1043–1053 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Mantovani, S. et al. Elevation of the terminal complement activation products C5a and C5b-9 in ALS patient blood. J. Neuroimmunol.276, 213–218 (2014). [DOI] [PubMed] [Google Scholar]
53.Govaere, O. et al. Macrophage scavenger receptor 1 mediates lipid-induced inflammation in non-alcoholic fatty liver disease. J. Hepatol.76, 1001–1012 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Clement, M. et al. IFITM3 restricts virus-induced inflammatory cytokine production by limiting Nogo-B mediated TLR responses. Nat. Commun.13, 5294 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Gomez-Herranz, M., Taylor, J. & Sloan, R. D. IFITM proteins: understanding their diverse roles in viral infection, cancer, and immunity. J. Biol. Chem.299, 102741 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Itou, T. et al. Th17 and effector CD8 T cells relate to disease progression in amyotrophic lateral sclerosis: a case control study. J. Neuroinflammation21, 331 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Rolfes, L. et al. Amyotrophic lateral sclerosis patients show increased peripheral and intrathecal T-cell activation. Brain Commun.3, fcab157 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Yazdani, S. et al. T cell responses at diagnosis of amyotrophic lateral sclerosis predict disease progression. Nat. Commun.13, 6733 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Kim, H. J. et al. Single-cell analysis reveals expanded CD8⁺GZMK^high T cells in CSF and shared peripheral clones in sporadic amyotrophic lateral sclerosis. Brain Commun.6, fcae428 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Tuddenham, J. F. et al. Single-cell transcriptomic landscape of the neuroimmune compartment in amyotrophic lateral sclerosis brain and spinal cord. Acta Neuropathol.150, 10 (2025). [DOI] [PMC free article] [PubMed]
61.Campisi, L. et al. Clonally expanded CD8 T cells characterize amyotrophic lateral sclerosis-4. Nature606, 945–952 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Piccoli, T. et al. Role of the immune system in amyotrophic lateral sclerosis. Analysis of the natural killer cells and other circulating lymphocytes in a cohort of ALS patients. BMC Neurol.23, 222 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
63.Garofalo, S. et al. Natural killer cells modulate motor neuron-immune cell cross talk in models of amyotrophic lateral sclerosis. Nat. Commun.11, 1773 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Murdock, B. J. et al. NK cells associate with ALS in a sex- and age-dependent manner. JCI Insight6, e147129 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.D’Erchia, A. M. et al. Massive transcriptome sequencing of human spinal cord tissues provides new insights into motor neuron degeneration in ALS. Sci. Rep.7, 10046 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Loving, B. A. & Bruce, K. D. Lipid and lipoprotein metabolism in microglia. Front. Physiol.11, 393 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Yamamoto, S. & Masuda, T. Lipid in microglial biology—from material to mediator. Inflamm. Regen.43, 38 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Sta, M. et al. Innate and adaptive immunity in amyotrophic lateral sclerosis: evidence of complement activation. Neurobiol. Dis.42, 211–220 (2011). [DOI] [PubMed] [Google Scholar]
69.Guerra San Juan, I. et al. Loss of mouse Stmn2 function causes motor neuropathy. Neuron110, 1671–1688 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Ma, X. R. et al. TDP-43 represses cryptic exon inclusion in the FTD-ALS gene UNC13A. Nature603, 124–130 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Irwin, K. E. et al. A fluid biomarker reveals loss of TDP-43 splicing repression in presymptomatic ALS–FTD. Nat. Med.30, 382–393 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Seddighi, S. et al. Mis-spliced transcripts generate de novo proteins in TDP-43-related ALS/FTD. Sci. Transl. Med.16, eadg7162 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Chizari, S., et al. TDP-43 pathology induces CD8⁺ T cell activation through cryptic epitope recognition. Preprint at bioRxiv10.1101/2025.06.22.660773 (2025).
74.Ely, Z. A. et al. Pancreatic cancer-restricted cryptic antigens are targets for T cell recognition. Science388, eadk3487 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Michaelis, T. et al. Autoimmune response to C9orf72 protein in amyotrophic lateral sclerosis. Nature647, 970–978 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
76.Genge, A. et al. Efficacy and safety of ravulizumab, a complement C5 inhibitor, in adults with amyotrophic lateral sclerosis: a randomized clinical trial. JAMA Neurol.80, 1089–1097 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
77.Paganoni, S. et al. Efficacy and safety of zilucoplan in amyotrophic lateral sclerosis: a randomized clinical trial. JAMA Netw. Open8, e2459058 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
78.Zelek, W. M. & Morgan, B. P. Targeting complement in neurodegeneration: challenges, risks, and strategies. Trends Pharmacol. Sci.43, 615–628 (2022). [DOI] [PubMed] [Google Scholar]
79.Mohammad, Q. D. et al. Results from a phase 1 study evaluating the safety, tolerability, pharmacokinetics, pharmacodynamics, and efficacy of ANX005, a C1q inhibitor, in patients with Guillain–Barré syndrome. J. Peripher. Nerv. Syst.30, e70009 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
80.Meyer-Lindemann, U., Moggio, A., Dutsch, A., Kessler, T. & Sager, H. B. The impact of exercise on immunity, metabolism, and atherosclerosis. Int. J. Mol. Sci.24, 3394 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
81.Wu, S. Z. et al. A single-cell and spatially resolved atlas of human breast cancers. Nat. Genet.53, 1334–1347 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
82.Kuppe, C. et al. Spatial multi-omic map of human myocardial infarction. Nature608, 766–777 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
83.Zhang, M. et al. Molecularly defined and spatially resolved cell atlas of the whole mouse brain. Nature624, 343–354 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
84.Bassiouni, R. et al. Spatial transcriptomic analysis of a diverse patient cohort reveals a conserved architecture in triple-negative breast cancer. Cancer Res.83, 34–48 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(406.6KB, pdf)}

Supplementary Table 1.

Reporting Summary^{(67.2KB, pdf)}

Source Data Fig. 1^{(5.2MB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 2^{(2.6MB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 3^{(44.3MB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 4^{(4.6MB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 5^{(374.6KB, xlsx)}

Source analysis result for figure production.

Source Data Fig. 6^{(306.1KB, xlsx)}

Source analysis result for figure production.

Data Availability Statement

All code used to generate the figures in this study can be found at https://github.com/gatelabNW/als_immune_public and https://github.com/KiskinisLab/TargetALS-BulkAnalysis.

[CR1] 1.Feldman, E. L. et al. Amyotrophic lateral sclerosis. Lancet400, 1363–1380 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR2] 2.Beers, D. R. & Appel, S. H. Immune dysregulation in amyotrophic lateral sclerosis: mechanisms and emerging therapies. Lancet Neurol.18, 211–220 (2019). [DOI] [PubMed] [Google Scholar]

[CR3] 3.Henkel, J. S. et al. Presence of dendritic cells, MCP-1, and activated microglia/macrophages in amyotrophic lateral sclerosis spinal cord tissue. Ann. Neurol.55, 221–235 (2004). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Kushner, P. D., Stephenson, D. T. & Wright, S. Reactive astrogliosis is widespread in the subcortical white matter of amyotrophic lateral sclerosis brain. J. Neuropathol. Exp. Neurol.50, 263–277 (1991). [DOI] [PubMed] [Google Scholar]

[CR5] 5.Schiffer, D., Cordera, S., Cavalla, P. & Migheli, A. Reactive astrogliosis of the spinal cord in amyotrophic lateral sclerosis. J. Neurol. Sci.139, 27–33 (1996). [DOI] [PubMed] [Google Scholar]

[CR6] 6.Engelhardt, J. I., Tajti, J. & Appel, S. H. Lymphocytic infiltrates in the spinal cord in amyotrophic lateral sclerosis. Arch. Neurol.50, 30–36 (1993). [DOI] [PubMed] [Google Scholar]

[CR7] 7.O’Rourke, J. G. et al. C9orf72 is required for proper macrophage and microglial function in mice. Science351, 1324–1329 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Burberry, A. et al. Loss-of-function mutations in the C9ORF72 mouse ortholog cause fatal autoimmune disease. Sci. Transl. Med.8, 347ra393 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Sheng, W., Ji, G. & Zhang, L. Role of macrophage scavenger receptor MSR1 in the progression of non-alcoholic steatohepatitis. Front. Immunol.13, 1050984 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Gudgeon, J., Marin-Rubio, J. L. & Trost, M. The role of macrophage scavenger receptor 1 (MSR1) in inflammatory disorders and cancer. Front. Immunol.13, 1012002 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Harmon, E. et al. Increased expression of interferon-induced transmembrane 3 (IFITM3) in stroke and other inflammatory conditions in the brain. Int. J. Mol. Sci.23, 8885 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Diamond, M. S. & Farzan, M. The broad-spectrum antiviral functions of IFIT and IFITM proteins. Nat. Rev. Immunol.13, 46–57 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Goldknopf, I. L. et al. Complement C3c and related protein biomarkers in amyotrophic lateral sclerosis and Parkinson’s disease. Biochem. Biophys. Res. Commun.342, 1034–1039 (2006). [DOI] [PubMed] [Google Scholar]

[CR14] 14.Apostolski, S. et al. Serum and CSF immunological findings in ALS. Acta Neurol. Scand.83, 96–98 (1991). [DOI] [PubMed] [Google Scholar]

[CR15] 15.Tsuboi, Y. & Yamada, T. Increased concentration of C4d complement protein in CSF in amyotrophic lateral sclerosis. J. Neurol. Neurosurg. Psychiatry57, 859–861 (1994). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Annunziata, P. & Volpi, N. High levels of C3c in the cerebrospinal fluid from amyotrophic lateral sclerosis patients. Acta Neurol. Scand.72, 61–64 (1985). [DOI] [PubMed] [Google Scholar]

[CR17] 17.Bahia El Idrissi, N. et al. Complement activation at the motor end-plates in amyotrophic lateral sclerosis. J. Neuroinflammation13, 72 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Schoggins, J. W. & Rice, C. M. Interferon-stimulated genes and their antiviral effector functions. Curr. Opin. Virol.1, 519–525 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR19] 19.Yang, W. et al. Blocking CCL3-mediated neutrophil recruitment into the brain alleviates immunopathology following severe enterovirus 71 infection. iScience27, 111388 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR20] 20.Martin-Sanchez, F. et al. Inflammasome-dependent IL-1β release depends upon membrane permeabilisation. Cell Death Differ.23, 1219–1231 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Gate, D. et al. Clonally expanded CD8 T cells patrol the cerebrospinal fluid in Alzheimer’s disease. Nature577, 399–404 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Oh, H. et al. Methods to investigate intrathecal adaptive immunity in neurodegeneration. Mol. Neurodegener.16, 3 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Piehl, N. et al. Cerebrospinal fluid immune dysregulation during healthy brain aging and cognitive impairment. Cell185, 5028–5039 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Rosati, E. et al. Overview of methodologies for T-cell receptor repertoire analysis. BMC Biotechnol.17, 61 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Humphrey, J. et al. Integrative transcriptomic analysis of the amyotrophic lateral sclerosis spinal cord implicates glial activation and suggests new risk genes. Nat. Neurosci.26, 150–162 (2023). [DOI] [PubMed] [Google Scholar]

[CR26] 26.Kabiljo, R., Iacoangeli, A., Al-Chalabi, A. & Rosenzweig, I. Amyotrophic lateral sclerosis and cerebellum. Sci. Rep.12, 12586 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Prell, T. & Grosskreutz, J. The involvement of the cerebellum in amyotrophic lateral sclerosis. Amyotroph. Lateral Scler. Frontotemporal Degener.14, 507–515 (2013). [DOI] [PubMed] [Google Scholar]

[CR28] 28.Fan, X. et al. Spatial transcriptomic survey of human embryonic cerebral cortex by single-cell RNA-seq analysis. Cell Res.28, 730–745 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR29] 29.Sun, N. et al. Human microglial state dynamics in Alzheimer’s disease progression. Cell186, 4386–4403 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR30] 30.Pineda, S. S. et al. Single-cell dissection of the human motor and prefrontal cortices in ALS and FTLD. Cell187, 1971–1989 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.Gautier, O. et al. Challenges of profiling motor neuron transcriptomes from human spinal cord. Neuron111, 3739–3741 (2023). [DOI] [PubMed] [Google Scholar]

[CR32] 32.Yadav, A. et al. A cellular taxonomy of the adult human spinal cord. Neuron111, 328–344 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR33] 33.Steinacker, P. et al. Chitotriosidase (CHIT1) is increased in microglia and macrophages in spinal cord of amyotrophic lateral sclerosis and cerebrospinal fluid levels correlate with disease severity and progression. J. Neurol. Neurosurg. Psychiatry89, 239–247 (2018). [DOI] [PubMed] [Google Scholar]

[CR34] 34.Maxeiner, S. et al. Crucial role for the LSP1-myosin1e bimolecular complex in the regulation of Fcγ receptor-driven phagocytosis. Mol. Biol. Cell26, 1652–1664 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Lu, C. H. et al. Neurofilament light chain: a prognostic biomarker in amyotrophic lateral sclerosis. Neurology84, 2247–2257 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR36] 36.Lopez-Erauskin, J. et al. Stathmin-2 loss leads to neurofilament-dependent axonal collapse driving motor and sensory denervation. Nat. Neurosci.27, 34–47 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Klim, J. R. et al. ALS-implicated protein TDP-43 sustains levels of STMN2, a mediator of motor neuron growth and repair. Nat. Neurosci.22, 167–179 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Baughn, M. W. et al. Mechanism of STMN2 cryptic splice-polyadenylation and its correction for TDP-43 proteinopathies. Science379, 1140–1149 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR39] 39.Prudencio, M. et al. Truncated stathmin-2 is a marker of TDP-43 pathology in frontotemporal dementia. J. Clin. Invest.130, 6080–6092 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.Melamed, Z. et al. Premature polyadenylation-mediated loss of stathmin-2 is a hallmark of TDP-43-dependent neurodegeneration. Nat. Neurosci.22, 180–190 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Kleshchevnikov, V. et al. Cell2location maps fine-grained cell types in spatial transcriptomics. Nat. Biotechnol.40, 661–671 (2022). [DOI] [PubMed] [Google Scholar]

[CR42] 42.Stevens, B. et al. The classical complement cascade mediates CNS synapse elimination. Cell131, 1164–1178 (2007). [DOI] [PubMed] [Google Scholar]

[CR43] 43.Peterson, S. L., Nguyen, H. X., Mendez, O. A. & Anderson, A. J. Complement protein C3 suppresses axon growth and promotes neuron loss. Sci. Rep.7, 12904 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR44] 44.Walker, D. G. & McGeer, P. L. Complement gene expression in human brain: comparison between normal and Alzheimer disease cases. Brain Res. Mol. Brain Res.14, 109–116 (1992). [DOI] [PubMed] [Google Scholar]

[CR45] 45.Ren, L. et al. Systematic pan-cancer analysis identifies APOC1 as an immunological biomarker which regulates macrophage polarization and promotes tumor metastasis. Pharmacol. Res.183, 106376 (2022). [DOI] [PubMed] [Google Scholar]

[CR46] 46.Neumann, M. et al. Ubiquitinated TDP-43 in frontotemporal lobar degeneration and amyotrophic lateral sclerosis. Science314, 130–133 (2006). [DOI] [PubMed] [Google Scholar]

[CR47] 47.Arai, T. et al. TDP-43 is a component of ubiquitin-positive tau-negative inclusions in frontotemporal lobar degeneration and amyotrophic lateral sclerosis. Biochem. Biophys. Res. Commun.351, 602–611 (2006). [DOI] [PubMed] [Google Scholar]

[CR48] 48.Masrori, P., Beckers, J., Gossye, H. & Van Damme, P. The role of inflammation in neurodegeneration: novel insights into the role of the immune system in C9orf72 HRE-mediated ALS/FTD. Mol. Neurodegener.17, 22 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Belzil, V. V. et al. Reduced C9orf72 gene expression in c9FTD/ALS is caused by histone trimethylation, an epigenetic event detectable in blood. Acta Neuropathol.126, 895–905 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR50] 50.Atanasio, A. et al. C9orf72 ablation causes immune dysregulation characterized by leukocyte expansion, autoantibody production, and glomerulonephropathy in mice. Sci. Rep.6, 23204 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR51] 51.Kjaeldgaard, A. L. et al. Complement profiles in patients with amyotrophic lateral sclerosis: a prospective observational cohort study. J. Inflamm. Res.14, 1043–1053 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR52] 52.Mantovani, S. et al. Elevation of the terminal complement activation products C5a and C5b-9 in ALS patient blood. J. Neuroimmunol.276, 213–218 (2014). [DOI] [PubMed] [Google Scholar]

[CR53] 53.Govaere, O. et al. Macrophage scavenger receptor 1 mediates lipid-induced inflammation in non-alcoholic fatty liver disease. J. Hepatol.76, 1001–1012 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR54] 54.Clement, M. et al. IFITM3 restricts virus-induced inflammatory cytokine production by limiting Nogo-B mediated TLR responses. Nat. Commun.13, 5294 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR55] 55.Gomez-Herranz, M., Taylor, J. & Sloan, R. D. IFITM proteins: understanding their diverse roles in viral infection, cancer, and immunity. J. Biol. Chem.299, 102741 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR56] 56.Itou, T. et al. Th17 and effector CD8 T cells relate to disease progression in amyotrophic lateral sclerosis: a case control study. J. Neuroinflammation21, 331 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Rolfes, L. et al. Amyotrophic lateral sclerosis patients show increased peripheral and intrathecal T-cell activation. Brain Commun.3, fcab157 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR58] 58.Yazdani, S. et al. T cell responses at diagnosis of amyotrophic lateral sclerosis predict disease progression. Nat. Commun.13, 6733 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR59] 59.Kim, H. J. et al. Single-cell analysis reveals expanded CD8⁺GZMK^high T cells in CSF and shared peripheral clones in sporadic amyotrophic lateral sclerosis. Brain Commun.6, fcae428 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR60] 60.Tuddenham, J. F. et al. Single-cell transcriptomic landscape of the neuroimmune compartment in amyotrophic lateral sclerosis brain and spinal cord. Acta Neuropathol.150, 10 (2025). [DOI] [PMC free article] [PubMed]

[CR61] 61.Campisi, L. et al. Clonally expanded CD8 T cells characterize amyotrophic lateral sclerosis-4. Nature606, 945–952 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR62] 62.Piccoli, T. et al. Role of the immune system in amyotrophic lateral sclerosis. Analysis of the natural killer cells and other circulating lymphocytes in a cohort of ALS patients. BMC Neurol.23, 222 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR63] 63.Garofalo, S. et al. Natural killer cells modulate motor neuron-immune cell cross talk in models of amyotrophic lateral sclerosis. Nat. Commun.11, 1773 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR64] 64.Murdock, B. J. et al. NK cells associate with ALS in a sex- and age-dependent manner. JCI Insight6, e147129 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR65] 65.D’Erchia, A. M. et al. Massive transcriptome sequencing of human spinal cord tissues provides new insights into motor neuron degeneration in ALS. Sci. Rep.7, 10046 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR66] 66.Loving, B. A. & Bruce, K. D. Lipid and lipoprotein metabolism in microglia. Front. Physiol.11, 393 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR67] 67.Yamamoto, S. & Masuda, T. Lipid in microglial biology—from material to mediator. Inflamm. Regen.43, 38 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR68] 68.Sta, M. et al. Innate and adaptive immunity in amyotrophic lateral sclerosis: evidence of complement activation. Neurobiol. Dis.42, 211–220 (2011). [DOI] [PubMed] [Google Scholar]

[CR69] 69.Guerra San Juan, I. et al. Loss of mouse Stmn2 function causes motor neuropathy. Neuron110, 1671–1688 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR70] 70.Ma, X. R. et al. TDP-43 represses cryptic exon inclusion in the FTD-ALS gene UNC13A. Nature603, 124–130 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR71] 71.Irwin, K. E. et al. A fluid biomarker reveals loss of TDP-43 splicing repression in presymptomatic ALS–FTD. Nat. Med.30, 382–393 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR72] 72.Seddighi, S. et al. Mis-spliced transcripts generate de novo proteins in TDP-43-related ALS/FTD. Sci. Transl. Med.16, eadg7162 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR73] 73.Chizari, S., et al. TDP-43 pathology induces CD8⁺ T cell activation through cryptic epitope recognition. Preprint at bioRxiv10.1101/2025.06.22.660773 (2025).

[CR74] 74.Ely, Z. A. et al. Pancreatic cancer-restricted cryptic antigens are targets for T cell recognition. Science388, eadk3487 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR75] 75.Michaelis, T. et al. Autoimmune response to C9orf72 protein in amyotrophic lateral sclerosis. Nature647, 970–978 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR76] 76.Genge, A. et al. Efficacy and safety of ravulizumab, a complement C5 inhibitor, in adults with amyotrophic lateral sclerosis: a randomized clinical trial. JAMA Neurol.80, 1089–1097 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR77] 77.Paganoni, S. et al. Efficacy and safety of zilucoplan in amyotrophic lateral sclerosis: a randomized clinical trial. JAMA Netw. Open8, e2459058 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR78] 78.Zelek, W. M. & Morgan, B. P. Targeting complement in neurodegeneration: challenges, risks, and strategies. Trends Pharmacol. Sci.43, 615–628 (2022). [DOI] [PubMed] [Google Scholar]

[CR79] 79.Mohammad, Q. D. et al. Results from a phase 1 study evaluating the safety, tolerability, pharmacokinetics, pharmacodynamics, and efficacy of ANX005, a C1q inhibitor, in patients with Guillain–Barré syndrome. J. Peripher. Nerv. Syst.30, e70009 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR80] 80.Meyer-Lindemann, U., Moggio, A., Dutsch, A., Kessler, T. & Sager, H. B. The impact of exercise on immunity, metabolism, and atherosclerosis. Int. J. Mol. Sci.24, 3394 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR81] 81.Wu, S. Z. et al. A single-cell and spatially resolved atlas of human breast cancers. Nat. Genet.53, 1334–1347 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR82] 82.Kuppe, C. et al. Spatial multi-omic map of human myocardial infarction. Nature608, 766–777 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR83] 83.Zhang, M. et al. Molecularly defined and spatially resolved cell atlas of the whole mouse brain. Nature624, 343–354 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR84] 84.Bassiouni, R. et al. Spatial transcriptomic analysis of a diverse patient cohort reveals a conserved architecture in triple-negative breast cancer. Cancer Res.83, 34–48 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Integrated single-cell and spatial transcriptomic profiling in ALS uncovers peripheral-to-central immune infiltration and reprogramming

Ziyang Zhang

Lynn van Olst

Francesco Alessandrini

Matthew Wright

Alex J Edwards

Jake Boles

Anait Nalbandian

Anne V Forsyth

Nate Shepard

Thomas Watson

Evan Kaspi

Angeli Mittal

Joshua Kuruvilla

Natalie Piehl

Abhirami Ramakrishnan

Stanley Appel

Evangelos Kiskinis

David Gate

Abstract

Main

Results

Peripheral immune dysregulation in sALS and C9-ALS

Fig. 1. Single-cell transcriptomic analysis of peripheral immune dysregulation in patients with ALS.

Extended Data Fig. 1. Quality control, cell composition, and differential gene expression analyses across PBMC single-cell RNA-seq datasets in ALS and healthy controls.

Fig. 2. Single-cell transcriptomic analysis of dysregulation in innate immune cells of patients with ALS.

Extended Data Fig. 2. Quality control, cell-type composition, and transcriptomic and pathway analyses of PBMC single-cell RNA-seq data across ALS and healthy controls.

Gene expression changes and immune-cell dynamics in ALS CNS tissues

Fig. 3. Bulk RNA analysis of ALS CNS tissues reveals widespread changes in spinal cord cell types.

Extended Data Fig. 3. Regional and cell-type-specific differential expression and enrichment analyses of bulk CNS transcriptomes across ALS and related neurodegenerative conditions.

Extended Data Fig. 4. Single-cell-informed cellular deconvolution reveals altered neuronal subtype composition in ALS motor and frontal cortex.

Extended Data Fig. 5. Single-cell reference mapping and cellular deconvolution identify neuronal subtype alterations in ALS motor and frontal cortex.

Fig. 4. Innate immune cells are dysregulated in the ALS spinal cord and associated with disease duration.

Regional specificity of inflammatory changes in ALS spinal cord

Fig. 5. SP reveals region-type-specific and cell-type-specific immune changes in ALS spinal cord.

Extended Data Fig. 6. Integrated single-nucleus, spatial, and histological analyses reveal immune activation and neuronal pathology in ALS spinal cord.

Immune dysregulation associates with ALS spinal cord pathology

Fig. 6. Inflammation localized to degenerating MNs and TDP-43 pathology within the anterior horn of the ALS spinal cord.

Discussion

Methods

Study participants

Ethics declarations for human tissues

Human tissues

Immune profiling of PBMCs

scRNA-seq + TCR profiling

scRNA-seq preprocessing

CRISPR-powered scRNA profiling

scRNA-seq differential expression gene analysis

Bulk RNA-seq analysis of NYGC ALS Consortium and Target ALS datasets

Alignment

Differential expression

Cell-type enrichment

Bulk RNA-seq downsampling analysis

Quantification of truncated STMN2

scRNA-seq dataset processing and cellular deconvolution for bulk RNA-seq analysis

Cortex

Spinal cord

Cell-type deconvolution

Correlation of gene expression and cell-type changes in ALS with disease duration and age of onset

Pseudobulk gene expression analysis

GSEA for scRNA-seq

Fluorescence immunostaining, imaging and quantification

SP

SP data processing

TDP-43 annotation and differential gene analysis

SP differential expression gene analysis

SP DEP analysis

Deconvolution of SP spots

Annotation of cell-type enrichment

MN-enriched differential gene expression analysis

Quantification and statistical analysis

Reporting summary

Online content

Supplementary information

Source data

Acknowledgements

Extended data

Author contributions

Peer review