Microglial control of astrocytes in response to microbial metabolites

Abstract

Microglia and astrocytes modulate inflammation and neurodegeneration in the central nervous system (CNS)^1-3. Microglia modulate pro-inflammatory and neurotoxic activities in astrocytes, but the mechanisms involved are not completely understood^4,5. Here we report that TGFα and VEGF-B produced by microglia regulate the pathogenic activities of astrocytes in the experimental autoimmune encephalomyelitis (EAE) mouse model of multiple sclerosis. Microglia-derived TGFα acts via the ErbB1 receptor in astrocytes to limit their pathogenic activities and EAE development. Conversely, microglial VEGF-B triggers FLT-1 signalling in astrocytes and worsens EAE. VEGF-B and TGFα also participate in the microglial control of human astrocytes. Furthermore, expression of TGFα and VEGF-B in CD14⁺ cells correlates with the multiple sclerosis lesion stage. Finally, metabolites of dietary tryptophan produced by the commensal flora control microglial activation and TGFα and VEGF-B production, modulating the transcriptional program of astrocytes and CNS inflammation through a mechanism mediated by the aryl hydrocarbon receptor. In summary, we identified positive and negative regulators that mediate the microglial control of astrocytes. Moreover, these findings define a pathway through which microbial metabolites limit pathogenic activities of microglia and astrocytes, and suppress CNS inflammation. This pathway may guide new therapies for multiple sclerosis and other neurological disorders.

Extended Data 1.. Contribution of AHR in CNS resident and infiltrating immune cells during EAE.

(a) qPCR of indicated genes from microglia, splenic macrophages, and astrocytes from control and CX3CR1-AHR mice on day 28 after EAE induction. Data are mean ± s.e.m. of n = 8 independent samples per group. P values were determined by two-sided Student’s t-test. (b) Flow cytometry analysis of AHR expression in microglia, monocytes and astrocytes from Control and CX3CR1-AHR mice 21 days after EAE induction. Thin line depicts isotype control, thick line AHR staining, and numbers indicate percentage of AHR positive cells. Representative of stainings of n=3 mice per group. (c) Spinal cord samples from naïve Control and CX3CR1-AHR mice were stained for Iba-1 and DAPI and Iba-1⁺ microglia/mm² were determined. n=5 mice per group. Data are mean ± s.e.m. and P value was determined by two-sided Student’s t-test. n.s. not significant. (d) TUNEL staining in Iba-1⁺ microglia in spinal cord sections of Control and CX3CR1-AHR mice as in (c). For the positive control, slides were cooked at 98°C in Citrate buffer during 60 minutes using a vapor cooker. Solid arrows show TUNEL positive microglia. Representative of n = 5 independent experiments. (e) Number of CNS-infiltrating (top) and splenic T-cells (bottom), and splenic pro-inflammatory monocytes (bottom) as determined by flow cytometry. n = 5 samples per group for CNS, n = 4 samples per group for spleen. Data are mean ± s.e.m. and P values were determined by two-sided Student’s t-test. (f) Proliferation assay from splenocytes isolated on day 28 of the experiment (Data are mean ± s.e.m. of n=4 biologically independent samples per group, representative of two independent experiments). (g) Bone marrow chimera were generated using WT mice irradiated as recipients, reconstituted with Control or CX3CR1-AHR bone marrow. Recipients of bone marrow were then rested for 3 weeks and thereafter treated with weekly tamoxifen gavages (4 mg) for another 3 weeks; after a total of 6 weeks, EAE was induced and tamoxifen administration continued weekly during EAE. Left, flow cytometry analysis of AHR expression in microglia and monocytes 21 days after EAE induction. Thin line depicts isotype control, thick line AHR staining, and numbers indicate percentage of AHR positive cells. Representative of stainings of n=3 independent mice per group. Right, EAE clinical course in bone marrow chimera mice. Data are mean ± s.e.m. and P values were determined by two-way ANOVA of n = 4 mice per group. (h) Control and CX3CR1-AHR mice were treated with with oral tamoxifen weekly starting from 5 weeks of age. EAE was induced at 8 weeks under continuation of weekly tamoxifen administration. Left, intracellular FACS staining for AHR in microglia and monocytes from at day 21 of EAE. Representative of stainings of n=3 independent mice per group. Right, clinical course of control and CX3CR1-AHR bone marrow chimera mice. Data are mean ± s.e.m. and P values were determined by two-way ANOVA of n = 4 mice per group.

Extended Data 2.. Topical and molecular regulation of TGF-α and VEGF-B.

(a) Ingenuity pathway analysis of differentially regulated pathways in astrocytes from n = 3 control vs CX3CR1-AHR mice per group during EAE. (b)Tgfa and Vegfb expression determined by qPCR in microglia from brain, cerebellum and spinal cord 21 days after EAE induction (left). Data are mean ± s.e.m. and P values were determined by one-way ANOVA followed by Tukey’s post-hoc test of n = 8 mice per group. (c) Predicted NF-κB and AHR responsive sites (NREs and XREs, respectively) in Vegfb and Tgfa promoters. (d) Microglia were isolated by FACS sorting from control and CX3CR1-AHR mice during EAE. Ex vivo ChIP assay of NF-kB p65 or AHR binding to predicted binding sites in Vegfb promoter. Data are mean ± s.e.m. and P values were determined by one-way ANOVA followed by Tukey’s post-hoc test of n = 3 mice per group. Representative of 2 independent experiments. (e) Reporter assay using a construct in which the Vegfb promoter controls luciferase expression (pVegfb-Luc); luciferase activity in HEK293 cells 24 hrs after transfection with pVegfb-Luc, pTK-Renilla, and plasmids expressing AHR or NF-κB p65. Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 4 biological replicates. (f) Ex vivo ChIP assay as in (d) for AHR binding in Tgfa promoter. Data are mean ± s.e.m. and P values were determined by one-way ANOVA followed by Tukey’s post-hoc test of n = 3 mice per group. Representative of 2 independent experiments. (g) Reporter assays using a construct in which the Tgfa promoter controls luciferase expression (pTgfa-Luc); luciferase activity in HEK293 cells 24 hrs after transfection with pTgfa-Luc, pTK-Renilla, and plasmids expressing AHR or Control. Data are mean ± s.e.m. and P values were determined by two-sided Student’s t-test. Representative of 2 independent experiments with n = 3 biological replicates.

Extended Data 3.. TGF-α and VEGF-B are regulated by AHR in highly purified astrocytes and microglia.

(a,b) Murine microglia were activated with LPS in the presence or absence of the AHR inhibitor CH223191. 24 hours later, activation medium was removed and substituted with fresh medium after extensive washes. 48 hours later, microglia conditioned medium (MCM) was harvested and applied to cultures of primary astrocytes. (a) Gene expression in microglia 24 hours after activation in the presence or absence of CH223191. (b) Gene expression in astrocytes after 24 hours exposure to MCM. Data are mean ± s.e.m. and P values were determined by two-sided Student’s t-test. Representative of 2 independent experiments with n = 3 biological replicates. (c) Representative FACS stainings for CD11b and CD45 in primary astrocyte and microglia cultures. Numbers indicate percentages in respective gate. Representative of n = 3 independent experiments. (d) Representative FACS stainings for GFAP and GLAST in astrocyte cultures as in (b). Representative of n = 3 independent experiments. (e,f) qPCR analysis of mRNA expression in astrocyte and microglia cultures. n=4 independent cultures, Data are mean ± s.e.m. and P values were determined by two-sided Student’s t-test. Representative of 2 independent experiments with n = 4 biological replicates. (g) Effect of TGF-α and VEGF-B on gene expression in primary astrocytes activated with TNF-α and IL-1β, determined by pPCR after 24 hrs. Representative of three independent experiments with n = 3 biological replicates. Data are mean ± s.e.m., P values determined by one-way ANOVA followed by Tukey’s post-hoc test. (h,i) Primary murine astrocytes were activated with TNF-α and IL-1β and treated with TGF-α or VEGF-B. 24 hours later, culture medium was substituted by fresh medium after extensive washes. 48 hours later, ACM was added to mouse neurons (h) and oligodendrocytes (i) in culture, and cytotoxicity was determined by quantifying lactate dehydrogenase (LDH) release after 24 hrs. n=3 biological replicates. Data are mean + s.e.m. representative of two independent experiments. P values determined by one-way ANOVA followed by Tukey’s post-hoc test. n.s. not significant. (j) CD11b⁺Ly6C^hi monocyte migration assay performed using ACM from astrocytes activated in the presence of TGF-α or VEGF-B. n=4 biological replicates. Data are mean + s.e.m. representative of two independent experiments. P values determined by one-way ANOVA followed by Tukey’s post-hoc test. n.s. not significant. (k) qPCR analysis for Nos2 expression in microglia co-cultured with astrocytes activated in the presence of TGF-α or VEGF-B. n=3 biological replicates. Data are mean + s.e.m. representative of two independent experiments. P values determined by one-way ANOVA followed by Tukey’s post-hoc test. n.s. not significant.

Extended Data 4.. Phenotypical and functional effects of knock-down of microglial TGF-α and VEGF-B.

(a) Quantification of astrocyte numbers in spinal cord sections of knock-down mice. Sox9 positive astrocytes per mm² were quantified in spinal cord sections of n=4 mice per group. n.s. (not significant) as determined by one-way ANOVA followed by Tukey’s post-hoc test. (b) IMARIS reconstruction of GFAP+ astrocytes in spinal cord sections as in (a) and quantification of dendrite length, branches, volume, terminal points, and segments of n=4 mice per group. n.s. (not significant) as determined by one-way ANOVA followed by Tukey’s post-hoc test. (c,d) qPCR analysis of Tgfa and Vegfb expression in sorted CNS-infiltrating inflammatory monocytes (c) and microglia (d) from mice injected with pCD11b-shControl, pCD11b-shTgfa, and pCD11b-shVegfb 7 days after EAE induction. Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 3 biological replicates. (e) qPCR analysis of Erbb1 and Flt1 expression in mice injected with pGFAP-shControl, pGFAP-shErbb1, and pCD11b-shFlt1 7 days after EAE induction. Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 3 biological replicates. (f) Flow cytometry analysis of VEGF-B and TGF-α expression in microglia from mice injected with pCD11b-shControl, pCD11b-shTgfa, and pCD11b-shVegfb 7 days after EAE induction (Left). Quantification of VEGF-B and TGF-α positive microglia in n=5 mice per group (Right). Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 5 biological replicates. (g) Flow cytometry analysis of FLT-1 and Erb-B1 expression in astrocytes from mice injected with pGFAP-shControl, pGFAP-shErbb1, and pCD11b-shFlt1 7 days after EAE induction (Left). Quantification of FLT-1 and Erb-B1 positive microglia in n=5 mice per group (Right). Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 5 biological replicates. (h) Naïve mice were injected with Lysolecithin, VEGF-B, or PBS into the corpus callosum by stereotaxic injection and 6 days later, brains were analyzed by myelin staining. Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 5 biological replicates.

Extended Data 5.. Directionality of TGF-α and VEGF-B signaling during EAE.

(a,b) EAE development in wild type mice injected with (a) pGFAP-shControl, pGFAP-shErbb1 and pCD11b-shFlt1, or (b) pCD11b-shControl, pCD11b-shTgfa and pCD11b-shVegfb. Clinical course n=5 mice per group. Representative of 2 independent experiments with n=5 mice per group. (c) Flow cytometry analysis of TGF-α and VEGF-B expression in astrocytes as in (a) (Left). Quantification of cytokine positive astrocytes (Right). Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 4 biological replicates. (d) Flow cytometry analysis of FLT-1 and Erb-B1 expression in microglia as in (b) (Left). Quantification of surface receptor expression of microglia (Right). Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 4 biological replicates.

Extended Data 6.. Regulation and transcriptional effects of TGF-α and VEGF-B during EAE.

(a,b) NanoString analysis of mRNA expression in astrocytes from EAE mice injected with pCD11b-shVegfb or pCD11b-shTgfa (a) and pGFAP-shFlt1 or pGFAP-shErbb1 (b, see also Fig. 2k,l). Fold change in relative expression relative to control as determined by log₂(shKD/shControl)). Representative of 2 independent experiments with pooled RNA isolated from n=3 mice per group. (c) Principal component analysis of gene expression in astrocytes isolated as in (a,b). Representative of 2 independent experiments with pooled RNA isolated from n=3 mice per group. (d) Ingenuity pathway analysis of significantly regulated pathways from astrocytes as in (a,b). Representative of 2 independent experiments with pooled RNA isolated from n=3 mice per group. (e) Left, representative flow cytometry plots depicting NF-κB p65 phosphorylation in WT astrocytes stimulated for 15 mins with Vehicle (top) or TNF-α/IL-1β (bottom) in the presence of TGF-α, VEGF-B, or their combination. Numbers indicate percentage of FITC⁺ cells. Bar graphs, quantification of FITC+ cells. Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 4 biological replicates. (f) Primary murine astrocytes were exposed to VEGF-B or Vehicle and pharmacological blocker of NF-κB activation. RNA was harvested after 18 hours and subjected to qPCR analyses for the indicated genes. Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 3 biological replicates. (g) Primary murine astrocytes were activated with TNF-α/IL-1β in the presence of VEGF-B or Vehicle and a pharmacological blocker of NF-κB activation. RNA was harvested after 18 hours and subjected to qPCR analyses for the indicated genes. Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 3 biological replicates.

Extended Data 7.. Role of AHR in astrocytes and microglia during EAE.

(a-c) EAE was induced in Control (WT), GFAP-Cr AHR^flox/flox (GFAP-AHR), or CX3CR1-AHR EAE mice. Starting from day 7, mice were injected daily intraperitoneally with Indoxyl-3-sulfate (I3S), given a tryptophan depleted diet (TDD), or kept on a control diet (Control). Clinical course of EAE mice under treatment conditions as indicated. Representative of 2 independent experiments with n = 4 mice per group. Data are mean ± s.e.m. and P values were determined by two-way ANOVA. (d-f) EAE was induced in WT mice, which were treated with lentiviruses to knock-down AHR in astrocytes (pGFAP-shAhr) or microglia (pCD11b-shAhr). A noncoding RNA was used as control. Flow cytometry quantification of AHR expression in astrocytes and microglia by FACS. (d) Representative histograms of n=4 mice per group. Numbers indicate percentage of AHR positive cells, thin lines isotype control, thick lines AHR staining. (e) Quantification of AHR positive astrocytes and microglia as in (d). Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 4 biological replicates. (f) EAE mice with knock down of AHR in astrocytes, microglia, or both as in (d) were subjected to daily I3S injections, TDD, or Control diet conditions starting on day 14 after disease induction. Clinical course of n=4 mice per group. Representative of 2 independent experiments with n = 4 mice per group. Data are mean ± s.e.m. and P values were determined by two-way ANOVA. (g) Quantification of CNS infiltrating pro-inflammatory monocytes as determined by FACS at day 28 of EAE. Data are mean ± s.e.m. and P values were determined by one way ANOVA followed by Tukey’s post-hoc test. Representative of 2 independent experiments with n = 3 biological replicates.

Extended Data 8.. Dietary factors influence mouse and human TGF-α and VEGF-B expression.

(a) Ingenuity pathway analysis of NF-κB signaling comparing TDD to TDD+Trp diet in control animals. Colors code for up- and down-regulation of individual members in red (up) and blue (down). Normalized reads of n = 2 independent samples per group. (b) mRNA expression determined by qPCR in from EAE mice as in Fig. 3a. Data are mean + s.e.m. and representative of two independent experiments with n = 3 replicates. P values determined by one-way ANOVA followed by Tukey’s post-hoc test. (c) Quantification of co-expression of AHR and CD14, VEGF-B and CD14, TGF-α and CD14 in immunofluorescence stainings of human white matter brain tissue of NAWM, active, or chronic MS lesions for AHR (left), VEGF-B (middle), or TGF-α (right), CD14 (green), and DAPI (blue). Data shown are representative of n = 12 fields from three distinct MS brains. (d) Ratio of VEGF-B to TGF-α intensities. Ratio of means from (g) + s.e.m of n = 25 fields. P values derived by one-way ANOVA followed by Tukey’s post-hoc test.

Figure 1.. AHR limits microglial pro-inflammatory transcriptional responses during EAE.

(a) EAE clinical scores in control and CX3CR1-AHR mice (n=10 mice per group). Data are mean ± s.e.m. and representative of two independent experiments. P derived by two-way ANOVA. (b) Representative spinal cord sections from control and CX3CR1-AHR EAE mice stained for Luxol Fast blue (LFB) for demyelination (top), and MAC3 for macrophage infiltration (bottom). Representative of 3 sections of n=3 mice. Right, quantification of demyelination and macrophage infiltration. Data are mean ± s.e.m.; P values determined by two-sided Student’s t-test. (c) Pro-inflammatory monocytes in the CNS of Control and CX3CR1-AHR EAE mice. Data are mean ± s.e.m. and representative of two independent experiments with n = 5 mice per group. P value was determined by two-sided Student’s t-test. (d) Microglial mRNA expression determined by qPCR in control (n = 8) and CX3CR1-AHR (n = 8) EAE mice; P value determined by two-sided Student’s t-test. (e) Lys310-acetyl p65 in Iba-1 positive cells in control and CX3CR1-AHR EAE mice (left). Quantification of Iba-1/p65 double positive cells (right). Data are mean ± s.e.m. representative of two independent experiments with n = 9 mice per group; P value determined by two-sided Student’s t-test. (f) Heat map of 9.957 genes expressed in microglia from control and CX3CR1-AHR mice (n = 3 mice per group). Gene expression is row-centered and log₂-transformed, and saturated at levels −0.5 and +0.5 for visualization satisfying a false discovery rate (FDR)<0.1. (g) Micloglial mRNA expression determined by qPCR in control (n = 5) and CX3CR1-AHR (n = 5) EAE mice. Data are mean ± s.e.m. and P values were determined by two-sided Student’s t-test.

Figure 2.. AHR-regulated microglial TGF-α and VEGF-B control astrocytes during EAE.

(a) Heat map of 14.823 genes (detected at level 0.1 in at least 2 of 3 samples) expressed in astrocytes from control and CX3CR1-AHR mice. Gene expression is row-centered log₂-transformed and saturated at −0.5 and +0.5 levels for visualization satisfying an FDR<0.1; n = 3 independent biological samples per group. (b) mRNA expression determined by qPCR in control and CX3CR1-AHR EAE mice (n = 5 per group). Data are mean ± s.e.m., P values determined by two-sided Student’s t-test. (c) Network diagram of differentially regulated genes in astrocytes and their predicted upstream regulators in microglia (n = 3 independent samples per group). (d) Microglial mRNA expression determined by qPCR in control and CX3CR1-AHR EAE mice. Data are mean ± s.e.m. P values determined by two-sided Student’s t-test; representative of two independent experiments with n = 6 control and n = 8 (Vegfb) or n = 6 (Tgfa) CX3CR1-AHR mice per group. (e) Effect of MCM and blocking antibodies to TGF-α and VEGF-B on gene expression in primary astrocytes determined by pPCR after 24 hrs. Representative of three independent experiments with n = 3 biological replicates. Data are mean ± s.e.m., P values determined by one-way ANOVA followed by Tukey’s post-hoc test. n.s. not significant. (f,g) EAE in C57Bl/6J mice injected with lentiviral knock-down constructs targeting Tgfa, Vegfb or control in microglia (f), or Erbb1, Flt1 or control in astrocytes (g). Data are mean ± s.e.m. and representative of two independent experiments with n=5 mice per group, P values determined by two-way ANOVA.

Figure 3.. Trp metabolites control microglia/astrocyte interactions and CNS inflammation.

(a) Clinical scores in control and CX3CR1-AHR mice treated with TDD, TDD+Trp or TDD+I3S from day 21 after EAE induction (n=10 mice per group). Data are mean ± s.e.m. representative of two independent experiments. P values were derived by two-way ANOVA. n.s. not significant. (b) Microglia were isolated and subjected to RNA-sequencing. Heatmap of expressed genes of normalized reads of n = 2 independent samples per group. (c) t-distributed stochastic neighbor embedding (tSNE) plot of RNA-Sequencing data isolated from microglia of mice as in (b). (d) microglial mRNA expression determined by qPCR in EAE mice as in (a). Data are mean + s.e.m. representative of two independent experiments with n = 3 replicates. P values determined by one-way ANOVA followed by Tukey’s post-hoc test. (e) mRNA expression determined by qPCR in microglia from EAE mice as in (a). Data are mean + s.e.m. and representative of two independent experiments with n = 3 replicates. P values determined by one-way ANOVA followed by Tukey’s post-hoc test. (f) Heatmap depicting mRNA expression in astrocytes from EAE mice as in (a), as determined by RNA-seq of normalized reads of n = 2 independent samples per group.

Figure 4.. VEGF-B and TGF-α control human astrocytes and are expressed by CD14⁺ cells in MS lesions.

(a-c) mRNA expression determined by qPCR in human microglia activated in the presence of I3S or the AHR antagonist CH223191 24 hours after activation. n=4 biological replicates. Data are mean + s.e.m. representative of three independent experiments. P values derived by one-way ANOVA followed by Tukey’s post-hoc test. (d) mRNA expression determined by qPCR in primary human astrocytes activated in the presence of TGF-α or VEGF-B. n=4 biological replicates. Data are mean + s.e.m. representative of three independent experiments. P values derived by one-way ANOVA followed by Tukey’s post-hoc test. (e) Immunofluorescence staining for AHR (A, red), VEGF-B (B, red), or TGF-α (C, red), CD14 (green), and DAPI (blue) in human brain samples corresponding to normal appearing white matter (NAWM), active, and chronic MS lesions Data are representative of n = 12 fields from three distinct MS brains. Inserts highlight co-expression of AHR and CD14, VEGF-B and CD14, and TGF-α and CD14. (f) Myelin oligodendrocyte glycoprotein (MOG, green) staining in MS tissues. Nuclear staining was done using Hoescht (blue). Representative sections of NAWM, active and chronic lesions from MS patients (n=3). (g) Quantification of AHR, VEGF-B and TGF-α expression in NAWM, active, or chronic lesions in MS tissue. Data shown are representative of n = 25 fields from three distinct MS brains. Data are mean + s.e.m. P values derived by one-way ANOVA followed by Tukey’s post-hoc test.