Therapy of autoimmune inflammation in sporadic amyotrophic lateral sclerosis: Dimethyl fumarate and H‐151 downregulate inflammatory cytokines in the cGAS‐STING pathway

In sporadic amyotrophic lateral sclerosis (sALS), IL‐17A‐ and granzyme‐positive cytotoxic T lymphocytes (CTL), IL‐17A‐positive mast cells, and inflammatory macrophages invade the brain and spinal cord. In some patients, the disease starts following a trauma or a severe infection. We examined cytokines and cytokine regulators over the disease course and found that, since the early stages, peripheral blood mononuclear cells (PBMC) exhibit increased expression of inflammatory cytokines IL‐12A, IFN‐γ, and TNF‐α, as well as granzymes and the transcription factors STAT3 and STAT4. In later stages, PBMCs upregulated the autoimmunity‐associated cytokines IL‐23A and IL‐17B, and the chemokines CXCL9 and CXCL10, which attract CTL and monocytes into the central nervous system. The inflammation is fueled by the downregulation of IL‐10, TGFβ, and the inhibitory T‐cell co‐receptors CTLA4, LAG3, and PD‐1, and, in vitro, by stimulation with the ligand PD‐L1. We investigated in two sALS patients the regulation of the macrophage transcriptome by dimethyl fumarate (DMF), a drug approved against multiple sclerosis and psoriasis, and the cyclic GMP‐AMP synthase/stimulator of interferon genes (cGAS/STING) pathway inhibitor H‐151. Both DMF and H‐151 downregulated the expression of granzymes and the pro‐inflammatory cytokines IL‐1β, IL‐6, IL‐15, IL‐23A, and IFN‐γ, and induced a pro‐resolution macrophage phenotype. The eicosanoid epoxyeicosatrienoic acids (EET) from arachidonic acid was anti‐inflammatory in synergy with DMF. H‐151 and DMF are thus candidate drugs targeting the inflammation and autoimmunity in sALS via modulation of the NFκB and cGAS/STING pathways.


| INTRODUCTION
Amyotrophic lateral sclerosis (ALS) is a chronic, progressive neurodegenerative disorder of upper and lower motor neurons associated with an inflammatory attack in the brain and spinal cord ( Figure 1). ALS has familial (fALS) and sporadic (sALS) forms, and lacks effective therapy. The cause of fALS is attributed to the mutations of the genes for protein homeostasis, and RNA metabolism, involving the genes C9orf72, superoxide dismutase (SOD1), TAR DNA-binding protein 43 (TDP-43), and FUS-TLS, and altered cytoskeletal dynamics. 1 The cause of sALS is unknown, but epidemiological data show an association of sALS with prior trauma 2 and microbial infections. 3 Damage-associated molecular pattern (DAMP) molecules, including autologous nucleic acids released by trauma, and microbial agents with pathogen-associated molecular patterns (PAMPs) are potential (auto)inflammatory stimuli. The role of microbial infection was noted in the study of a genetically identical twin pair with both twins possessing common ALS mutations, but only the twin suffering a cat bite and a severe systemic Pasteurella infection later developed ALS, whereas the other twin remains healthy 11 years later. 4 The ALS twin displayed spontaneous production of the cytokines IL-6 and TNFα, and epigenetic modification through differential methylation of immune-related regions near EGFR and TNFRSF11A genes.
TDP-43, a nuclear DNA/RNA-binding protein, accumulates and aggregates in neurons of sALS patients, triggers the release of mitochondrial DNA, and activates cyclic GMP-AMP synthase (cGAS), which synthesizes 2′, 3′-cGMP-AMP (cGAMP), and activates the stimulator of interferon genes (cGAS-STING) pathway. 5 In this pathway, the endoplasmic reticulum (ER)-membrane adaptor stimulator of interferon genes (STING) is released from ER and traffics to Golgi, is palmitoylated and phosphorylated by TANK-binding kinase 1 (TBK1), and recruits the transcription factor interferon regulatory factor 3 (IRF3), which is phosphorylated. Phospho-IRF3 pro-resolution macrophage phenotype. The eicosanoid epoxyeicosatrienoic acids (EET) from arachidonic acid was anti-inflammatory in synergy with DMF. H-151 and DMF are thus candidate drugs targeting the inflammation and autoimmunity in sALS via modulation of the NFκB and cGAS/STING pathways.
The rationale for an anti-inflammatory therapy of sALS is supported by the increasing inflammatory and autoimmune transcripts in peripheral blood mononuclear cells (PBMC) over the disease course, and the sALS neuropathology. We initially speculated that the programmed cell death ligand 1 (PD-L1) would engage the inhibitory receptor PD1 and downregulate inflammation but, in ALS patients, PD-L1 engagement was aberrantly pro-inflammatory. Motivated by the increased evidence of autoimmunity and inflammation, we tested in ALS macrophages (a) dimethyl fumarate (DMF), the drug approved for the therapy of two autoimmune diseases, multiple sclerosis and psoriasis, (b) the inhibitor of the cGAS-STING pathway H-151, 5,6 and (c) the eicosanoid epoxyeicosatrienoic acids (EET). We demonstrate that DMF with epoxyeicosatrienoic acids (EET) and H-151 have anti-inflammatory and anti-autoimmune effects in immune cells, which suggest their therapeutic usefulness in ALS. 81-year-old; Ctr #2 F 51-year-old; Ctr #3 F 67-year-old; Ctr #4 F, 56-y/o) ( Figure 2). The diagnosis of sALS was established by primary neurologists according to upper and lower motor neuron signs, electromyography, and exclusion of other disorders and familial ALS. We evaluated the clinical stage according to the ALS functional rating scale (ALSFRS), rated on a scale from 0 to 48 with a score of 48 corresponding to control patients (Table 1).

| sALS and control patients
We investigated in vitro immune regulation of differentiated macrophages from sALS patient #9 (M, 40-year-old, ALS-FRS 16, loss of 1.33 points per month, duration 24 months) with a typical ALS presentation, and sALS patient #10 (F, 65-year-old, ALS-FRS score 39, loss of 0.37 points per month, duration 24 months) with dysarthria and weakness of facial muscles followed by sensory changes, characteristic of the ALS-related disorder facial onset sensory and motor neuronopathy (FOSMN). The latter patient had a history of severe impetigo infection (a source of microbial nucleic acids) in childhood. We investigated protein abundances in natural killer (NK) cells of an identical but ALS-discordant twin pair, the sALS twin #1 (M, 77-year-old, loss 0.57 points per month), and the identical healthy twin #2 ( Figure 2).

| Treatment of sALS macrophages by dimethyl fumarate (DMF) with epoxyeicosatrienoic acids, and the cGAS-STING inhibitor H-151
We investigated immune regulation in macrophages differentiated in vitro from PBMCs of the rapidly progressing younger sALS patient #9 and the slowly progressing older sALS patient #10. We tested possible therapeutic molecules dissolved in dimethyl sulfoxide (DMSO) and IMDM medium (Fisher): the soluble epoxide hydrolase inhibitor (sEHI) TPPU (1 μM), 13

| Immunohistochemistry of spinal cord tissues
The immunochemical staining of sALS paraffinembedded spinal cord tissues, after antigen retrieval using pH 6-buffer steam and blockade with goat serum, was performed using DAKO Envision Doublestain system with primary antibodies to the macrophage marker CD68, (DAKO), granzyme B, and IL-17A ( Figure 1).

| RNA-Seq analysis
We isolated RNA using the Quick-RNA miniprep kit (Zymo Research) and prepared RNA libraries using the Illumina TruSeq Stranded mRNA kit. We quantified final libraries using the Qubit BR dsDNA Assay (Life Technologies) and performed QC with the D1000 Assay on a 2200 Tape Station (Agilent Technologies). We sequenced PBMC libraries using an Illumina HiSeq 4000 to obtain 50 bp reads and evaluated the quality of the RNA-seq data using FastQC. We aligned the reads to the human genome (hg19) using STAR v2.5.0a. 14 We quantified raw read counts using HTSeq v0.6.1 15 and transformed them to Reads Per Million (RPM) for each sample. We analyzed the time course of the expression of cytokines, chemokines, granzymes, and transcription factors after disease onset.
For analyses of patient macrophages treated with candidate drugs, we aligned reads to hg38 using the RNAseq processing workflow TOIL. 16 Briefly, TOIL employs STAR for alignment and quantifies reads using both RSEM (RNAseq by Expectation Maximization) and Kallisto. We used RSEM upper quartile normalized counts for subsequent analyses. For each set of samples, we calculated expression log fold changes by comparing each patient's treated cells to the DMSO-treated control. Heatmaps were generated with the package pheatmap in R, and principal component analysis (PCA) was performed using the R function prcomp.

| Global proteomic analysis of NK cells
NK cells were isolated using a CD56+ selection kit (Stem Cells Technologies, Vancouver, BC, Canada) from the blood of a twin-pair discordant in the ALS diagnosis. The protein abundance ratio of NK cells was measured using a native hybrid quadrupole-Orbitrap mass spectrometer and high pH reverse-phase chromatography fractionation mass spectroscopy. We analyzed the protein abundance ratio of the NK immune-and cell adhesion proteins in the affected over the healthy twin.

| Inflammatory cytokines are upregulated in early disease and autoimmune cytokines and granzymes in late disease
The key sALS immunopathology of the ALS spinal cord gray matter involves the invasion by IL-17A-positive and granzyme B-positive cytotoxic CD8 cells, which engulf the neurons (Figure 1). Since this inflammatory neuropathology is derived from the circulating PBMCs, we analyzed by RNAseq the transcriptome of prospectively collected PBMCs over the ALS disease course since the disease onset. Two sALS patients provided prospective samples and eight patients and four control subjects (shown at baseline) provided one sample ( Figure 2).
The cytokines tumor necrosis factorα (TNFα) and interferonγ (IFNγ) were upregulated since disease onset in relation to controls shown at baseline and were downregulated 25-and 15-months post-diagnosis, respectively. Conversely, the anti-inflammatory cytokines IL-10 and transforming growth factor beta 1 (TGFβ1) transcripts decreased over the disease course. The autoimmunity cytokine IL-17 B increased in late disease, together with the expression of the pro-inflammatory chemokines CXCL9 and CXCL10, which attract immune cells into the CNS ( Figure 3A).
Increased expression of inflammatory cytokines, especially from the IL-23 and IL12 cytokine families, signal to the downstream transcription factors STAT1, STAT3, and STAT4. 17 Target genes of STAT 3 and STAT4 include the cytokine IFNG and granzymes. Granzymes are serine proteases toxic to neurons, which are released from natural killer (NK) cells and cytotoxic T cells. 18 Granzymes were induced following the induction of STAT4, which serves as the transcription factor for the granzyme B promoter. 19 Correspondingly, the expression of granzymes decreased late in the disease course after 30 months, following a decrease in STAT4 expression at 25 months ( Figure 3B).

| Systemic inflammation over the ALS disease course is fueled by the down-regulation of T-cell inhibitory co-receptors and upregulation of stimulatory T-cell co-receptors
T-cell inhibitory co-receptors such as cytotoxic T lymphocyte-associated protein 4 (CTLA4) and programmed cell death protein-1 (PD-1) are negative regulators of T-cell immune responses, while the stimulatory co-receptors show the opposite effect. 20 In ALS patients' T A B L E 1 sALS patients demographics, clinical data, functional rating score loss, and list of medications. PBMCs, CTLA4 showed gradual downregulation over the disease course. Another inhibitory co-receptor LAG3 showed initial upregulation with a sharp decline at 40 months post-onset. In contrast, the stimulatory coreceptors OX40 and GITR showed dramatic upregulation at 40 months post-onset. As expected, 21 in the whole cohort, late-surviving patients had a relative increase in the transcription factor FOXP3, a master regulator of the regulatory pathway in the development and function of immunoregulatory T cells (Tregs) ( Figure 3C).

| Proteomic analysis indicates overexpression of granzymes, kinases, cell adhesion, and apoptotic proteins in the natural killer cells of the ALS twin in comparison to the healthy twin
To identify the changes in natural killer (NK) T-cell populations of an ALS discordant and genetically identical twin-ship, we performed proteomics and evaluated proteins with relevant functions according to the ratio of the ALS twin/the healthy twin ( Figure 4). As expected, granzyme B, a protease toxic to neurons, was increased in the ALS twin. In addition, the surface proteins CD44 (cell surface glycoprotein involved in cell-cell interactions), CD8 cytotoxic T-cell antigen, and the histocompatibility antigen HLA-DRA were also increased, suggesting increased T-cell activity in the ALS twin. We performed proteomic analysis of immune regulation and cell death. Signaling by GTPases RAC1, RAC2, and RHOA stimulates cytoskeletal reorganization and inflammatory pathways. 22 Signaling by these GTPases is important for regulating cell proliferation, differentiation, and survival. These GTPases, as well as MAPK proteins and SRC kinase, were more abundant in NK cells of the ALS twin than in the healthy twin. Notably, SRC kinase inhibitors have been shown in a large phenotypic screen to protect from motor neuron degeneration, and MAPKp38 inhibition has also been proposed as a therapeutic strategy for ALS. 23,24 The cell junction and cell adhesion proteins actin (ACTB), talin1 (TLN1), integrin β3, and VIM were also more abundant in the ALS twin than in the healthy twin. Certain cell surface proteins are thought to serve as potential biomarkers for ALS and may be the target for autoantibodies. In addition, anti-ACTB antibodies may be a biomarker for ALS disease severity.
Finally, the immunophilin proteins, which bind to the immunosuppressive molecules FK506 and rapamycin, were higher in the ALS twin, as were caspase 3, death agonist protein BID, and superoxide dismutase-1 (SOD-1), which promote apoptosis (Figure 4). Taken together, these results suggested multiple dysregulated pathway functions in the NK cells of an ALS patient compared to his healthy identical twin.

PDL-1, PD-1 antibody in ALS macrophages
Given the pro-inflammatory signatures seen in the ALS time-course and the concomitant downregulation of T-cell inhibitory co-receptors, we hypothesized that manipulating immune checkpoint receptor-ligand interactions may reduce inflammatory cytokine production in the ALS macrophage transcriptome. PDL-1 (CD274) is a ligand on cancer cells and antigen-presenting cells that stimulates the inhibitory co-receptor programmed cell death protein 1 (PD-1, aka CD279) to block inflammatory activation of CTLs. We investigated the effects of PD-L1 (8 μg/mL) (Biolegend, San Diego, CA) and PD-1 antibody (2.3 μg/mL) (Abclonal Technology, Woburn, Mass) (in comparison to 1% dimethyl sulfoxide [DMSO]) on induction of the macrophage inflammatory cytokines, transcription factors, and regulatory co-receptors (CTLA4, LAG3, PDCD1). Contrary to the established function of PD-L1 to activate PD-1 and inhibit immune responses, treatment with PDL-1 generated a pro-inflammatory effect in sALS patients' macrophages by upregulating the expression of the inflammatory cytokines. Treatment with the PD-1 antibody, which blocks PD-1/PD-L1 interactions, was expected to be pro-inflammatory, but instead tended to be anti-inflammatory, decreasing the expression of IL1B and IL6 ( Figure 5). In contrast, treatment of the same patient's F I G U R E 3 (Continued) cells with DMF resulted in strong downregulation of all examined cytokines, granzymes, and NFκB target genes.

| Inhibition of inflammatory cytokines, granzymes, and NFκB in ALS macrophages by therapeutic molecules DMF and H-151
Given the autoimmune features of sALS, we next tested dimethyl fumarate (DMF) (Figure 6), a drug approved for the therapy of two autoimmune diseases, multiple sclerosis and psoriasis, without or with epoxyeicosatrienoic acids (EET) and the cGAS-STING inhibitor H-151 (Figure 7). We investigated these effects in the macrophage cultures of two ALS patients: the younger patient #9 (M, 40-year-old, ALS-FRS 16, severe loss of 1.33 ALS FRS points per month, duration 24 months, with typical ALS features) and the older patient #10 (F, 65-year-old, ALS-FRS score 39, less severe loss of 0.37 points per month, duration 24 months, with FOSMN). After 24-h treatment with either drug or DMSO, we collected total RNA and performed RNA sequencing. We calculated the log fold change of gene expression in DMF or H151-treated cells compared to DMSO. A negative log2 fold change indicates decreased expression of the gene upon drug treatment. a. Inhibition by dimethyl fumarate (DMF) with or without epoxyeicosatrienoic acids (EET) and TPPU We investigated the effect of DMF and H-151 on the expression of several categories of genes: inflammatory cytokines, granzymes, NFκB, and IFN target genes ( Figure 6). In both patients, we observed that DMF downregulated the inflammatory cytokines IL1β, IL15, and IFNγ. The downregulation of the expression of pro-inflammatory cytokines as well as granzymes and NFκB target genes, such as IL6 and TNFAIP3, by DMF appeared to be potentiated by the addition of EET, an epoxy fatty acid derived from the metabolism of arachidonic acid, known to have antiinflammatory effects, and, in some transcripts, by the soluble epoxide hydrolase inhibitor (sEHI) TPPU.

b. Inhibition by cGAS/STING inhibitor H-151
To investigate the role of the cGAS/STING pathway, we next examined the effect of the STING inhibitor H-151 on cytokines, NFκB target genes, and interferon signaling (e.g., IRF3 and IRF7) target genes in macrophages with or without PBMC (Figure 7). H-151 in a dose-responsive fashion (1 and 10 μg/mL) downregulated a subset of proinflammatory cytokines, including IL1B and TNF, and the effects were potentiated by the presence of PBMC in macrophage cultures. The direct NFκB target gene NFKBIA was also downregulated, suggesting decreased NFκB signaling activity. Importantly, downstream targets of cGAS/STING signaling such as ISG15, MX1, and IFIT1 were downregulated, supporting effective STING inhibition by H-151.

| Comparison of the effects of DMF versus H-151 on ALS patient macrophages
We first evaluated the reproducibility of biological replicates in the same cultures (R = 0.94) and in different cultures of the same patient (R = 0.98), in comparison to the cultures of different patients (R = 0.87) and different treatments (R = 0.83). Thus, we observed high concordance (R > 0.9) of biological replicates of the same patient ( Figure 8).
To examine the differences between DMF and H-151 treatment, we performed principal component analysis (PCA) on more recently collected samples from patient #10, which appear to cluster together. PCA on this subset showed that DMF versus H-151-treated samples separated along Principal Component 2 ( Figure 9A). This analysis identified a subset of inflammatory genes that most strongly drove this distinction: IFNG, CCL15, CXCL9, GZMB, and IGFBP15 were all downregulated in H151-treated samples more than in DMF-treated samples ( Figure 9B). In contrast, H-151 samples exhibited an increased expression of genes important in oncogenesis, such WNT10A, WNT7A, PLEKHB1, and SLC22A17 (solute carrier transporter), which have been implicated in or suggested as therapeutic targets for ALS ( Figure 9B). 25,26 Serendipitously, DMF+/− EET+/− TPPU inhibited these WNT and PLEKHB targets. Future investigation and eventually clinical trials need to determine whether the combination of H-151 with DMF would be synergistic against autoimmune inflammation in sALS patients.

| Clues about the origin of sporadic ALS
In a previous study of a genetically identical, ALS discordant twin pair, both twins had common ALS mutations, but only the ALS twin suffered a cat bite leading to a severe systemic Pasteurella infection and later developed ALS, whereas the other twin remains healthy 11 years later. 4 The ALS twin displayed spontaneous production of the cytokines IL-6 and TNF-α, and epigenetic modification through differential methylation of immune-related regions near EGFR and TNFRSF11A genes.

| Progressive inflammation over the course of sporadic ALS: An initial inflammation with interferon-γ and TNF-α and a later autoimmune inflammation with IL-17B
We examined in a cohort of eight sALS patients over the disease course the transcripts of inflammatory and autoimmune cytokines (Figure 3). The prospective results show an early upregulation of the inflammatory cytokines IFNγ and TNFα, which are signaling through STAT3 and STAT4, followed by a later increase of CD4 T cells and upregulation of IL-17B on Th17 cytotoxic T cells and the chemokines associated with STAT1 signaling. The inflammatory cytokines stimulating IL-17 expression in Th17 cells, that is, IL-1, IL-6, and IL-23A, are upregulated in response to aggregated superoxide dismutase-1(SOD-1) in vitro. 11 Over the sALS disease course, the stimulatory co-receptor OX-40 and GITR increased, whereas IL-10, TGFβ1, the inhibitory coreceptors LAG3 and CTLA4, and regulatory T cells (Tregs) decreased. We also analyzed the relation of inflammation to autoimmunity in PBMC and NK cells from a pair of identical, ALS-discordant twins. The increase of inflammation was also observed in this twin-ship proteomic study with an overall increase in the sALS twin's NK cells of inflammatory proteins, including granzymes, GTPases, and cell adhesion proteins (Figure 4).

| Anti-inflammatory therapies of ALS
The heterogeneity of inflammatory mechanisms in sALS have suggested that diverse immunotherapeutic strategies may be necessary. The IL-6 receptor antibody tocilizumab 12 showed temporary effects in a subset of sALS patients with an inflammatory Th1/Th17 signature but no effects in a second subset with a B-cell signature. 4 An in vitro study showed that induced pluripotent stem cell (iPSC)-derived M2 macrophages from sALS patients suppress activation of M1 macrophages and boost ALS Tregs. 27 In a mouse model, grafting protective macrophages at disease onset extended survival and decreased inflammation. 28 In patients with chronic autoimmune inflammation, macrophages have both beneficial immunosuppressive and pathogenic tissue-destructive roles. 29 We initially speculated that activation of the inhibitory co-receptor PD-1 by recombinant PD-L1 would be anti-inflammatory, that is, therapeutic. However, both recombinant PD-L1 ligand and recombinant PD-1 receptor were strongly pro-inflammatory, as observed in systemic sclerosis 30 and ALS, 31 whereas PD-1 antibody was antiinflammatory, demonstrating aberrant regulation of the ALS patients' immune system ( Figure 5).

| DMF and H-151 are anti-inflammatory drugs in sporadic ALS patients
We abandoned the PD-L1 strategy and chose a strategy with dimethyl fumarate (DMF), an approved drug against two autoimmune diseases, multiple sclerosis and psoriasis, and the inhibitor of the cGAS-STING pathway H-151 implicated in autoimmunity. Both drugs, DMF (in combination with EET) ( Figure 6) and H-151 (Figure 7) inhibited the inflammatory cytokines (IFNγ, IL-1β, IL-15, IL-23A, and TNF), the granzymes (B, M, and H), and the transcription factors NFkB1 and NFkB2 in macrophages of ALS patients. According to the literature, only DMF has been used in a clinical setting in a combination with riluzole in a randomized multisite Australian clinical trial in 107 sporadic ALS patients. 32 In that study, the DMF therapy (240 mg twice daily P.O.) was safe but, in comparison to placebo, not significantly active against ALS progression according to ALS FRS. Our in vitro results using the EET/DMF combination in ALS macrophages suggest that DMF may be effective when combined with a diet rich in certain polyunsaturated fatty acids (PUFAs).

| CONCLUSIONS
Both DMF (in synergy with EET and TPPU) and H-151 inhibit in immune cells of sALS patients inflammatory and autoimmune signaling, which support their investigation in a clinical trial.