Disease-associated astrocyte epigenetic memory promotes CNS pathology

Abstract

Disease-associated astrocyte subsets contribute to the pathology of neurologic diseases, including multiple sclerosis and experimental autoimmune encephalomyelitis^1-8 (EAE), an experimental model for multiple sclerosis. However, little is known about the stability of these astrocyte subsets and their ability to integrate past stimulation events. Here we report the identification of an epigenetically controlled memory astrocyte subset that exhibits exacerbated pro-inflammatory responses upon rechallenge. Specifically, using a combination of single-cell RNA sequencing, assay for transposase-accessible chromatin with sequencing, chromatin immunoprecipitation with sequencing, focused interrogation of cells by nucleic acid detection and sequencing, and cell-specific in vivo CRISPR-Cas9-based genetic perturbation studies we established that astrocyte memory is controlled by the metabolic enzyme ATP-citrate lyase (ACLY), which produces acetyl coenzyme A (acetyl-CoA) that is used by histone acetyltransferase p300 to control chromatin accessibility. The number of ACLY⁺p300⁺ memory astrocytes is increased in acute and chronic EAE models, and their genetic inactivation ameliorated EAE. We also detected the pro-inflammatory memory phenotype in human astrocytes in vitro; single-cell RNA sequencing and immunohistochemistry studies detected increased numbers of ACLY⁺p300⁺ astrocytes in chronic multiple sclerosis lesions. In summary, these studies define an epigenetically controlled memory astrocyte subset that promotes CNS pathology in EAE and, potentially, multiple sclerosis. These findings may guide novel therapeutic approaches for multiple sclerosis and other neurologic diseases.

Extended Data Fig. 1 |. Analysis of astrocyte epigenetic memory in vivo and in vitro.

(a) Fluorescence-activated cell sorting (FACS) sorting schematic for astrocytes, microglia, and monocytes. (b) Volcano plot of differential gene expression increased (red) or decreased (blue) analyzed by RNA-seq of sorted astrocytes. (c) IPA predicted upstream regulators in isolated astrocytes. Up-regulated (red) or down-regulated (blue) in astrocytes stimulated twice (2X IL-1β + TNF) versus once (1X IL-1β + TNF) are shown. (d) GO pathway analysis of ATAC-seq accessible peaks of isolated astrocytes comparing 2X IL-1β + TNF versus 1X IL-1β + TNF. (e) Genome browser snapshots showing the ATAC-seq sequencing tracks at the Tlr2 locus. Only regions showing a significant increase (p-value < 0.05) in accessibility in astrocytes from mice stimulated twice (2X) versus once (1X) are highlighted by yellow boxes. (f) Homer DNA-motif enrichment analyses of differentially accessible peaks. (g) Experimental design for (h) to (j). Primary astrocytes received IL-1β/TNF stimulation once (1X) or twice (2X). (h) qPCR of astrocytes after 30 min activation with IL-1β/TNF on day 7 (n = 5 per group). Unpaired two-sided t-test. (i) Neuronal viability assay (n = 4 control; n = 4 1X; n = 3 2X). Unpaired two-sided t-test. (j) Effect of IL-1β/TNF stimulation on lactate release (n = 6 per group). Unpaired two-sided t-test. (k) qPCR analysis of astrocyte response after the first IL-1β/TNF stimulation (n = 5 per group). Unpaired two-sided t-test. (l) FACS analysis of EGFP expression in IL-1β/TNF stimulated primary astrocytes isolated from p65^EGFP reporter mice (0 min; n = 5; Other time point; n = 6 per group). ***P < 0.0001. Unpaired two-sided t-test. (m) FACS sorting schematic of EGFP positive/negative astrocytes after PBS or IL-1β/TNF stimulation for 18–24 h. (n) Primary astrocytes isolated from p65^EGFP reporter mice received IL-1β/TNF stimulation. After 18–24 h, EGFP positive astrocytes were sorted and analyzed by FACS of EGFP expression (n = 3–5 per group). Unpaired two-sided t-test. (o) Primary astrocytes isolated from p65^EGFP reporter mice received PBS or IL-1β/TNF stimulation. After 18–24 h, EGFP positive/negative astrocytes were sorted and cultured for 6 days to perform qPCR (n = 4 per group). Unpaired two-sided t-test. (p) p65 activation in EGFP positive/negative astrocytes 1 h after IL-1β/TNF stimulation on day 7 (n = 5 per group). Unpaired two-sided t-test. Data shown as mean ± s.e.m.

Extended Data Fig. 2 |. HAT enzyme expression and regulation of astrocyte epigenetic memory.

(a) EP300, Tip60 (KAT5), and PCAF (KAT2B) signaling network in isolated brain astrocytes that received ICV administration of IL-1β/TNF twice (2X) compared to once (1X). (b) qPCR analysis of Ep300, Kat5, and Kat2b expression in primary astrocytes stimulated after 30 min IL-1β/TNF stimulation on day 7 (n = 4 per group). Unpaired two-sided t-test. (c) Primary astrocytes were stimulated with IL-1β/TNF once (1X), twice (2X), or three times (3X). qPCR analysis of astrocytes after 30 min activation with IL-1β/TNF on day 14 (n = 4 per group). Unpaired two-sided t-test. (d) Experimental design for (e). (e) qPCR of primary astrocytes in the presence with/without C646 (p300/CBP inhibitor), MB-3 (Gcn5/PCAF inhibitor), and MG149 (Tip60 inhibitor) after 30 min stimulation with IL-1β/TNF on day 7 (n = 5 per group). Unpaired two-sided t-test. (f) Experimental design for Fig. 1j. (g) Gene expression 30 min after IL-1β/TNF stimulation on day 7 of EGFP positive/negative astrocytes treated with C646 or vehicle (n = 8 per group). Unpaired two-sided t-test. Data shown as mean ± s.e.m.

Extended Data Fig. 3 |. Astrocyte epigenetic memory and Ep300 signaling in EAE.

(a) Generation of astrocyte epigenetic memory signature filtered based on adjusted p-value and fold change. Up-signature (red) or down-signature (blue) in astrocytes stimulated twice (2X) than once (1X) are shown. (b) Astrocyte epigenetic memory signature score applied to naive and EAE scRNA-seq astrocyte dataset (Priming, Peak, and Remission). (c) Experimental design for (d,e). (d) EAE score for (e) (n = 8 per group). Data shown as mean ± s.e.m. Naïve and EAE induced C57BL/6 mice received ICV administration of IL-1β/TNF (EAE peak, Day 22), and 18–24 h later sorted brain astrocytes were analyzed. (e) qPCR of IL-1β/TNF response of astrocytes (Naïve; n = 12; EAE; n = 8 per group). Representative data of two independent experiments. Unpaired two-sided t-test. (f) Immunostaining (left) and quantification (right) of H3K27ac⁺ and p300⁺ astrocytes from sgScrmbl- and sgEp300-transduced mice at 23 days after EAE induction (n = 6 spinal cord sections; n = 3 mice per group). Unpaired two-sided t-test. (g) FluoroMyelin dye staining and percentage of myelin loss in spinal cord from sgScrmbl- and sgEp300-treated mice (n = 9 spinal cord sections; n = 3 mice per group). Lesions indicated by arrowheads. Unpaired two-sided t-test. (h) NF-kB signaling network comparing sgEp300-transduced versus sgScrmbl-transduced astrocytes. (i) Quantification of CNS-resident cells from sgScrmbl- and sgEp300-transduced mice (n = 9 per group). Unpaired two-sided t-test. (j,k) Analysis of CNS T cells (up) and splenic T cells (bottom) from sgScrmbl- or sgEp300-transduced mice (n = 5 per group). Unpaired two-sided t-test. (l) Genome browser snapshots showing the Relb locus. Only regions showing a significant decrease (p-value < 0.05) in accessibility in sgScrmbl-transduced versus sgEp300-transduced astrocytes are highlighted by yellow boxes. (m) Chip-qPCR analysis of p300 recruitment to promoters in primary astrocytes 30 min after IL-1β/TNF stimulation on day 7 (n = 3 per group). Unpaired two-sided t-test. Data shown as mean ± s.e.m.

Extended Data Fig. 4 |. ACLY and ACSS2 signaling in EAE astrocytes.

(a) qPCR analysis of Acly and Acss2 expression in primary astrocytes stimulated for 30 min with IL-1β/TNF on day 7 (n = 4 per group). Unpaired two-sided t-test. (b) Primary astrocytes received IL-1β/TNF stimulation once (1X), twice (2X), or three times (3X). qPCR analysis of astrocytes after 30 min activation with IL-1β/TNF on day 14 (n = 4 per group). Unpaired two-sided t-test. (c) Immunostaining (left) and quantification (right) of ACLY⁺ astrocytes in mice with/without EAE (n = 8 spinal cord sections (naïve); n = 9 spinal cord sections (EAE); n = 3 mice per group). Unpaired two-sided t-test. (d) Immunostaining (left) and quantification (right) of ACSS2⁺ astrocytes in mice with/without EAE (n = 6 spinal cord sections (naïve); n = 9 spinal cord sections (EAE); n = 3 mice per group). Unpaired two-sided t-test. Data shown as mean ± s.e.m.

Extended Data Fig. 5 |. Astrocyte Acly signaling and ACLY + p300+ astrocytes in EAE.

(a) Immunostaining (left) and quantification (right) of H3K27ac⁺ and ACLY⁺ astrocytes from sgScrmbl- and sgAcly-transduced mice at 21days after EAE induction (n = 6 spinal cord sections; n = 3 mice per group). Unpaired two-sided t-test. (b) Staining with FluoroMyelin dye and percentage of myelin loss in spinal cord of sgScrmbl- and sgAcly-treated mice (n = 9 spinal cord sections; n = 3 mice per group). Lesions indicated by arrowheads. Unpaired two-sided t-test. (c) Quantification of CNS-resident cells from sgScrmbl- and sgAcly-transduced mice (n = 5 per group). Unpaired two-sided t-test. (d,e) Analysis of CNS T cells (up) and splenic T cells (bottom) from sgScrmbl- or sgAcly-transduced mice (n = 5 per group). Unpaired two-sided t-test. Data shown as mean ± s.e.m. (f) Immunostaining (left) and quantification (right) of ACLY⁺SOX9⁺, p300⁺SOX9⁺, and ACLY⁺p300⁺SOX9⁺ astrocytes in EAE and control mice (n = 6 spinal cord sections; n = 3 mice per group). Unpaired two-sided t-test. Data shown as mean ± s.e.m.

Extended Data Fig. 6 |. Astrocyte epigenetic memory in NOD EAE.

(a) Experimental design for (c). (b) NOD EAE score for (c) (n = 3 per group). (c) Naïve and EAE induced NOD mice received ICV administration of IL-1β/TNF (EAE progressive, Day 124). After 18–24 h, sorted brain astrocytes were analyzed by qPCR (n = 3 per group). Unpaired two-sided t-test. (d) Immunostaining (left) and quantification (right) of H3K27ac⁺ and p300⁺ astrocytes in mice with/without NOD EAE (n = 9 spinal cord sections; n = 3 mice per group). Astrocyte H3K27ac levels were calculated as the mean signal intensity (arbitrary units) per GFAP+ cells using automated unbiased quantification. Unpaired two-sided t-test. (e) Immunostaining (left) and quantification (right) of ACLY⁺, p300⁺, and ACLY⁺p300⁺ astrocytes in mice with/without NOD EAE (n = 9 spinal cord sections; n = 3 mice per group). Unpaired two-sided t-test. (f) NOD EAE curves (sgScrmbl; n = 7; sgEp300; n = 8; sgAcly; n = 7). Lentivirus were injected at day 40. Representative data of two independent experiments. Regression slope two-sided t-test compared with sgScrmbl. (g) Volcano plot of differential gene expression determined by RNA-seq in astrocytes isolated from sgScrmbl-, sgEp300-, and sgAcly-transduced mice 64 days after NOD EAE induction (n = 3 sgScrmbl, n = 3 sgEp300, n = 2 sgAcly). (h) GSEA analysis comparing sgScrmbl-, sgEp300-, and sgAcly-transduced astrocytes. (i) Staining with FluoroMyelin dye and percentage of myelin loss from sgScrmbl-, sgEp300-, and sgAcly-transduced mice spinal cord (n = 6 spinal cord sections (sgEp300); n = 9 spinal cord sections (sgScrmbl, sgAcly); n = 3 mice per group. Lesions indicated by arrowheads. Unpaired two-sided t-test. Data shown as mean ± s.e.m.

Extended Data Fig. 7 |. ACLY and p300 signaling in NOD EAE astrocytes.

(a) Immunostaining (left) and quantification (right) of H3K27ac⁺ and p300⁺ astrocytes from sgScrmbl- and sgEp300-transduced NOD mice 64 days after EAE induction (n = 6 spinal cord sections; n = 3 mice per group). Unpaired two-sided t-test. (b) Immunostaining (left) and quantification (right) of H3K27ac⁺ and ACLY⁺ astrocytes from sgScrmbl- and sgAcly-transduced NOD mice 64 days after EAE induction (n = 6 spinal cord sections; n = 3 mice per group). Unpaired two-sided t-test. (c) Quantification of CNS-resident cells from sgScrmbl-, sgEp300, and sgAcly-transduced mice (n = 7 per group). Unpaired two-sided t-test. (d,e) Analysis of CNS T cells (up) and splenic T cells (bottom) from sgScrmbl-, sgEp300, and sgAcly-transduced mice (n = 4 per group). Unpaired two-sided t-test. Data shown as mean ± s.e.m.

Extended Data Fig. 8 |. Analysis of Acly⁺Ep300⁺ astrocytes by FIND-seq.

(a) FACS sorting schematic for tdTomato^Gfap astrocytes. (b-d) Schematic illustration of microfluidic devices utilized in FIND-seq. (b) The bubble-triggered device has with five inlets: i) cell inlet, ii) Oligo dT primer-conjugated agarose inlet, iii) lysis buffer inlet, iv) oil inlet, and v) pressurized air inlet. (c) The re-injector device has three inlets: i) agarose bead inlet, ii) TaqMan PCR master mix inlet, and iii) oil inlet. (d) The droplet sorter device has seven inlets: i) emulsion inlet, ii) spacing oil inlet, iii) additional spacing oil inlet, iv) bias oil inlet, v) saltwater inlet (for the electrode), vi) saltwater inlet (for the moat), and vii) pressurized air inlet. (e) Astrocytes are encapsulated in an agarose bead along with the lysis buffer. The genome entrapped in the agarose bead is stained with SYBR Green I and visualized using a fluorescence microscope. (f) cDNA, produced on the agarose bead, is amplified via WTA and validated using a Bioanalyzer. (g) Principal component analysis (PCA) plot of Acly⁻Ep300⁻, Acly⁺Ep300⁻, Acly⁻Ep300⁺, Acly⁺Ep300⁺ EAE astrocytes. (h) Violin plot depicting Acly⁺Ep300⁺ EAE astrocyte signature expression in EAE astrocytes. (i) IPA pathway analysis up-regulated (red) in cluster 1 astrocytes. (j) EP300-NF-κB signaling network of cluster 1 astrocytes. (k) IPA pathway analysis up-regulated (red) in cluster 10 astrocytes. (l) EP300-NF-κB signaling network of cluster 10 astrocytes.

Extended Data Fig. 9 |. Analysis of human astrocyte epigenetic memory and MS astrocyte scRNA-seq.

(a) Primary human fetal astrocytes received IL-1β/TNF stimulation once (1X) or twice (2X). qPCR of astrocytes in the presence with/without C646 (p300/CBP inhibitor) after 2 h stimulation with IL-1β/TNF on day 7 (n = 4 per group). Unpaired two-sided t-test. (b) Gene scatterplots of astrocyte markers. (c) Unsupervised clustering UMAP plot of astrocytes from patients with MS and control individuals from Schirmer et al. and Absinta et al.. (n = 16,276 cells). WM, white matter; CI, chronic inactive; CA, chronic active. (d) Significantly enriched genes by astrocyte cell type cluster. (e) Cluster distribution of CNS cells. (f) Violin plot depicting Acly⁺Ep300⁺ EAE astrocyte signature expression in MS astrocytes. (g) Astrocyte epigenetic memory signature score in astrocyte clusters of control and MS patients.

Extended Data Fig. 10 |. ACLY + p300+ astrocytes in MS patients.

(a) Immunostaining and quantification of ACLY⁺SOX9⁺, p300⁺SOX9⁺, ACLY⁺p300⁺SOX9⁺ astrocytes in CNS samples from MS patients with MS (n = 9 sections (Lesion); n = 6 sections (NAWM); n = 3 per patient) and controls (n = 3 sections; n = 3 per patient). WM, white matter; NAWM, normally appearing white matter. Unpaired two-sided t-test. Data shown as mean ± s.e.m.

Fig. 1 |. Pro-inflammatory stimuli induce astrocyte epigenetic memory.

a, Experimental design for b–g. C57BL/6 mice received ICV administration of IL-1β and TNF or PBS once (1×) or twice (2×). After 18 h, sorted brain astrocytes were analysed. b, Differential gene expression determined by RNA-seq in sorted astrocytes (n = 3 per group). c, Gene set enrichment analysis (GSEA) comparing astrocytes stimulated twice (2×) versus once (1×). Gene Ontology (GO) terms are shown. FDR, false discovery rate; NES, normalized enrichment score. d,e, IPA (d) and predicted upstream regulators (e) in isolated astrocytes. Up-regulated (red) or down-regulated (blue) genes in astrocytes stimulated twice (2×) versus once (1×) are shown. f,g, ATAC-seq analysis of astrocytes comparing 2× IL-1β + TNF versus 1× IL-1β + TNF (n = 5 per group). f, Genes with increased (red) or decreased (blue) accessibility in astrocytes. g, Genome browser snapshots showing ATAC-seq sequencing tracks at the Il1b locus. Regions with a significant increase (P < 0.05) in accessibility in astrocytes from mice stimulated twice (2×) versus once (1×) are indicated by the yellow band. h, Experimental design for i. Primary astrocytes isolated from p65-eGFP reporter mice received PBS or IL-1β + TNF stimulation. After 18–24 h, eGFP-positive and negative astrocytes were sorted and received IL-1β + TNF stimulation once (1×) or twice (2×) after 6 days. i, Quantitative PCR (qPCR) of eGFP-positive and negative astrocytes after 30 min stimulation with IL-1β + TNF on day 7 (n = 8 per group). Unpaired two-sided t-test. j, Gene expression 30 min after IL-1β + TNF stimulation on day 7 of eGFP-positive and negative astrocytes treated with C646 or vehicle (n = 8 per group). Unpaired two-sided t-test. Data are mean ± s.e.m. Dotted lines indicate significance (P < 0.05).

Fig. 2 |. p300 promotes astrocyte epigenetic memory in EAE.

a, Immunostaining (left) and quantification (right) of H3K27ac⁺ and p300⁺ astrocytes in mice with or without EAE (n = 8 spinal cord sections; n = 3 mice per group). Astrocyte H3K27ac levels were calculated as the mean signal intensity (arbitrary units) per GFAP⁺ cell using automated unbiased quantification. Unpaired two-sided t-test. Scale bars, 25 μm. b, EAE curves in mice treated with non-targeting single guide RNA (sgScrmbl) (n = 17) or single guide RNA (sgRNA) targeting Ep300 (sgEp300) (n = 14) mice. Representative data of three independent experiments. Two-way repeated measures of ANOVA. c, Differential gene expression determined by RNA-seq in astrocytes from sgScrmbl- and sgEp300-transduced mice 23 days after EAE induction (n = 3 per group). d, GSEA comparing sgScrmbl- and sgEp300-transduced astrocytes. e,f, ChIP–seq analysis of sgScrmbl- and sgEp300-transduced astrocytes (n = 3 per group). e, Heat map showing dynamic H3K27 acetylation marks. f, Genome browser snapshots showing the Il1b locus. Regions showing a significant decrease (P < 0.05) in accessibility in sgScrmbl-transduced versus sgEp300-transduced astrocytes are highlighted by the yellow band. g, Reactome pathway analysis showing down-regulated ChIP–seq-accessible peaks of isolated astrocytes comparing sgScrmbl-transduced versus sgEp300-transduced astrocytes. Dotted line indicates significance (P < 0.05). Data are mean ± s.e.m. LTR, long terminal repeat; pA, polyA; U6, promoter; WRPE, woodchuck hepatitis virus post-transcriptional regulatory element.

Fig. 3 |. ACLY controls p300-derived astrocyte epigenetic memory in EAE.

a, EAE curves. Left: sgScrmbl, n = 15; Acly-targeting sgRNA (sgAcly), n = 12. Right, sgScrmbl, n = 14; sgAcss2, n = 9. Representative data of at least two independent experiments. Two-way repeated measures of ANOVA. b, Differential gene expression determined by RNA-seq in astrocytes from sgScrmbl- and sgAcly-transduced mice 21 days after EAE induction (n = 4 per group). c, GSEA analysis comparing sgScrmbl- and sgAcly-transduced astrocytes. d, IPA showing down-regulated genes in sgAcly-versus sgScrmbl-transduced astrocytes. e, Chromatin immunoprecipitation with qPCR analysis of the abundance of H3K27ac in pro-inflammatory gene promoters, comparing sgScrmbl- and sgAcly-transduced astrocytes (n = 3 per group). Unpaired two-sided t-test. Dotted line indicates significance (P < 0.05). f, Analysis of the EP300–NF-κB signalling network, comparing sgScrmbl- and sgAcly-transduced astrocytes. g, Immunostaining (left; scale bars, 20 μm, middle; scale bars, 10 μm) and quantification (right) of ACLY⁺p300⁺ astrocytes in mice with or without EAE (n = 9 spinal cord sections; n = 3 mice per group). Unpaired two-sided t-test. Data are mean ± s.e.m.

Fig. 4 |. Analysis of Acly⁺Ep300⁺ astrocytes with FIND-seq.

a, TdTomato^Gfap astrocytes from naive and EAE mice were pooled (n = 3 per group) and analysed by FIND-seq using Acly and Ep300 expression. EAE curve (n = 3 per group). Two-way repeated measures of ANOVA. b, Representative droplet cytometry plots. Numbers in each quadrant display the percentage of astrocytes in droplets. c, Astrocyte epigenetic memory signature score in Acly⁻Ep300⁻, Acly⁺Ep300⁻, Acly⁻Ep300⁺ and Acly⁺Ep300⁺ EAE astrocytes. d, Uniform manifold approximation and projection (UMAP) plot of TdTomato^Gfap astrocytes (n = 24,963 cells). e, Acly⁺Ep300⁺ EAE astrocyte signature expression in EAE astrocyte clusters. f, Astrocyte epigenetic memory signature score in EAE astrocyte clusters. g, Fraction of cells per cluster overall. h, IPA up-regulated pathways (red) in cluster 3 astrocytes. i, ACLY–EP300–NF-κB signalling network of cluster 3 astrocytes. j, Pseudotime trajectory in TdTomato^Gfap astrocytes. Each cell is coloured to indicate its pseudotime value. k, Pseudotime analysis of gene expression in TdTomato^Gfap astrocytes.

Fig. 5 |. Identification of astrocyte epigenetic memory in multiple sclerosis.

a, UMAP plot of astrocytes from patients with multiple sclerosis and control individuals from Schirmer et al. and Absinta et al.. n = 16,276 cells. b, Acly⁺ Ep300⁺ EAE astrocyte signature expression in multiple sclerosis astrocyte clusters. c, Cluster analysis of astrocytes based on percent composition in multiple sclerosis. WM, white matter; CI, chronic inactive; CA, chronic active. d, GSEA analysis of cluster 2 astrocytes. e, IPA showing up-regulated pathways in cluster 2 astrocytes. f, ACLY–EP300–NF-κB signalling network of cluster 2 astrocytes. g, Pseudotime trajectory in multiple sclerosis astrocytes. Each cell is coloured to indicate its pseudotime value. h, Pseudotime analysis of gene expression in multiple sclerosis astrocytes. i, Immunostaining and quantification of ACLY⁺, p300⁺ and ACLY⁺p300⁺ astrocytes in tissue samples from patients with multiple sclerosis (n = 9 sections (lesion); n = 6 sections (NAWM); n = 3 per patient) and control individuals (n = 3 sections; n = 3 per patient). NAWM, normally appearing white matter. Unpaired two-sided t-test. Data are mean ± s.e.m. Scale bars, 20 μm.