Radial glia integrin avb8 regulates cell autonomous microglial TGFβ1 signaling that is necessary for microglial identity

Abstract

Microglial diversity arises from the interplay between inherent genetic programs and external environmental signals. However, the mechanisms by which these processes develop and interact within the growing brain are not yet fully understood. Here, we show that radial glia-expressed integrin beta 8 (ITGB8) activates microglia-expressed TGFβ1 to drive microglial development. Domain-restricted deletion of Itgb8 in these progenitors results in regionally restricted and developmentally arrested microglia that persist into adulthood. In the absence of autocrine TGFβ1 signaling, microglia adopt a similar phenotype, leading to neuromotor symptoms almost identical to Itgb8 mutant mice. In contrast, microglia lacking the canonical TGFβ signal transducers Smad2 and Smad3 have a less polarized dysmature phenotype and correspondingly less severe neuromotor dysfunction. Our study describes the spatio-temporal regulation of TGFβ activation and signaling in the brain necessary to promote microglial development, and provides evidence for the adoption of microglial developmental signaling pathways in brain injury or disease.

Fig. 1. Deletion of Itgb8 in early embryonic radial glia disrupts microglial maturation.

A Analysis of Itgb8 expression in the E14.5 mouse embryo in neural progenitor cells (NPCs) and radial glia, microglia, endothelial cells, and mural cells. B Itgb8^tdT reporter expression confirms strong Itgb8 expression in SOX9⁺, NESTIN⁺ radial progenitors at E14.5. Open arrowhead marks radial glia fibers; closed arrowhead marks ramified radial glia endfeet at the surface of the neuroepithelium. C Model describing developmental expression of Itgb8 in neuroepithelium and radial glia, and correlation with sequential timing of Cre recombination in Emx1^Cre, Nestin^Cre and hGFAP^Cre lines. D–F Deletion of Itgb8 from neuroepithelial and radial progenitors using indicated Cre lines. Coronal brain sections stained for tdT (Cre recombination, red), vascular endothelium (CD31, cyan), and macrophages/microglia (IBA1, yellow); hemorrhage (red blood cells marked by TER119, yellow) observed outside of vascular lumen (CD31, cyan); microglia precursors (CD206, yellow) and committed/homeostatic microglia (P2RY12, cyan). G Quantification of brain hemorrhage in Itgb8^fl/fl;Emx1^Cre and Itgb8^fl/fl;hGFAP^Cre E14.5 cerebral cortex and underlying striatum. p = 9.3*10⁻⁵. H Quantification of microglial P2RY12 expression in embryonic IBA1⁺ microglia in Itgb8^fl/fl;Emx1^Cre and Itgb8^fl/fl;hGFAP^Cre E14.5 cerebral cortex and underlying striatum. p = 5.1*10⁻⁵. I E14.5 brain section from Emx1^Cre;RG-brainbow mouse stained for membranous GFP (individual recombined radial glia; endfeet, green), microglia precursors (CD206, magenta), and committed/homeostatic microglia (P2RY12, yellow). Arrowheads indicate foot process of radial glia contacting pial-associated CD206⁺ presumptive microglia precursor (model in J to right). n = 3 for B, D–I. Error bars = mean +/- SEM. Scale bar in B = 500μm, D–F = 200μm and inset scale bar=30μm, I = 25 μm. ***p < 0.001, two-tailed t-test. Source data are provided as a Source Data file.

Fig. 2. Domain-specific microglial dysmaturity persists in Itgb8^fl/fl;Emx1^Cre mice.

A Comparison of the transcriptional properties of adult Itgb8^fl/fl; Emx1^Cre mutant and control Itgb8^fl/fl microglia using stage specific developmental markers. Itgb8^fl/fl;Emx1^Cre mutant dysmature microglia retain the gene expression profiles of early embryonic microglia, and lack expression of mature homeostatic microglial genes (teal). Heatmap shows z-score of normalized expression. Many of the early embryonic marker genes enriched in dysmature microglia in Itgb8^fl/fl; Emx1^Cre mutants are also MgND (orange) and BAM (gray) markers. B Analysis of developmental gene cluster expression in Itgb8^fl/fl; Emx1^Cre mutant microglia reveals enrichment for progenitor (cluster 1) and early embryonic phase (clusters 3-5) enriched gene sets. Graphs indicate mean ± SD of the average normalized expression in each cluster. *p < 0.05, ***p < 0.001; two-tailed t-test. C Whole brain sagittal immunostaining of adult Itgb8^fl/fl; Emx1^Cre mice revealed anatomically restricted maintenance of the microglial precursor marker CD206 in the cortex and hippocampus (asterisk and box 1), vs striatum (box 2). D Increased MGnD marker LGALS3 expression in cortical and hippocampal microglia in Itgb8^fl/fl; Emx1^Cre mice (asterisk and blow-up in box 1), and no change in the underlying striatum (blow-up in box 2). E Downregulation of the homeostatic marker P2RY12 in the cortex of a Itgb8^fl/fl; Emx1^Cre mouse, and maintenance in the underlying striatum. F Cortex-restricted upregulation of the reactive marker APOE in IBA1⁺ cells of the cerebral cortex, and lack of APOE expression in microglia in the striatum. Closed and open arrowheads in (E, F) mark dysmature cortical and phenotypically normal striatal microglia respectively. G Quantification of homeostatic (P2RY12) (***p = 2.9*10⁻¹⁰) and reactive (APOE) (***p = 9.68*10⁻⁵) markers in the cerebral cortex and striatum of the adult Itgb8^fl/fl; Emx1^Cre brain. H Expression of microglial status markers in isolated microglia. Heatmap shows z-score of normalized expression. Genes with an asterisk are differentially expressed (*p < 0.05; DESeq2, negative binomial generalized linear model and Wald test for significance testing). Error bars= mean +/- SEM. n = 4 for control and n = 5 for Itgb8^fl/fl; Emx1^Cre mutants in (A and B). Cx= cerebral cortex; Cc= corpus callosum; Str= striatum; Dashed line= cortical/striatal boundary. Scale bar in C, D = 2 mm and inset scale bar=200μm, E, F = 150μm, and inset scale bar=20μm. n = 3 for 2C-G. ***p < 0.001, two-tailed t-test. Source data are provided as a Source Data file.

Fig. 3. Dysmature microglia show pervasive epigenetic changes associated with MGnD and BAM gene upregulation.

A Tracks of ATAC-seq and H3K9ac ChIP-seq from Itgb8^fl/fl; Emx1^Cre mutant and Itgb8^fl/fl control microglia illustrating key homeostatic, border associated macrophage (BAM) and MgND marker gene bodies. B Comparison between differentially enriched genes (DEGs) (RNA-seq) and genes associated with differentially accessible peaks (DAPs) (ATAC-seq) in cortical and hippocampal microglia from Itgb8^fl/fl; Emx1^Cre vs Itgb8^fl/fl control brains (Padj < 0.05). Linear correlation between overlapping DAPs and DEGs, Log2FC. C Comparison between DEGs (RNA-seq) and genes associated with differentially enriched H3K9 ChIP-Seq peaks (DEPs) in Itgb8^fl/fl; Emx1^Cre (Padj < 0.05) vs control brains. Linear correlation between overlapping DEPs and DEGs Log2FC. D Heatmap of ATAC-seq showing top 100 DAPs comparing Itgb8^fl/fl; Emx1^Cre to control microglia. n = 7 control, 7 mutant. E Heatmap of H3K9ac ChIP-seq showing top 100 DEPs comparing Itgb8^fl/fl; Emx1^Cre to WT microglia. n = 4 control, 3 mutant; each biological replicate was derived from pooled microglia from 3 mice. F Motif enrichment (HOMER) for positive or negative enriched peaks. For epigenetic analysis, differential enrichment was determined with DESeq2 (negative binomial generalized linear model and Wald test for significance testing). Differentially accessible peaks were determined with adjustment for false discovery using benjamini hochberg method; Padj < 0.05).

Fig. 4. Microglia provide their own TGFβ1 to promote and maintain homeostasis.

A Sorted bulk-Seq analysis of embryonic microglia and BAMs reveals that Tgfb1 is expressed in both microglia and BAMs during embryonic development, whereas (B) P2ry12 and Pf4 are specific markers of these two respective populations. C, D Analysis of control (C) and conditional Cx3cr1^CreER mediated deletion (D) of Tgfb1 deletion in the E14.5 forebrain following E11.5, 12.5 and 13.5 tamoxifen induction. Analysis revealed no hemorrhage (CD31 in green, TER119 in magenta), no change in macrophage/blood vessel association (IBA1 in magenta, CD31 in green), loss of the homeostatic marker P2RY12 (in magenta, IBA1 in green), and loss of Tgfb1 (cyan) in Isolectin B4 (green) and tdT (red) labeled microglia, but not in IB4⁺ labeled blood vessels. E Analysis of conditional P2ry12^CreER mediated deletion of Tgfb1 deletion in E14.5 microglia following E11.5, 12.5 and 13.5 tamoxifen induction. Analysis revealed no brain hemorrhage (CD31 in green, TER119 in magenta), no change in macrophage/blood vessel association (IBA1 in magenta, CD31 in green), and loss of the homeostatic marker P2RY12 (magenta), in IBA1⁺ (green) microglia. F P2ry12^CreER recombination, as shown by ROSA-tdTomato (Ai14) Cre reporter expression, was restricted to microglia (closed arrowheads), and was not seen in the overlying meninges (open arrowheads). G Analysis of Pf4^Cre mediated deletion of Tgfb1 in the E14.5 forebrain. Analysis revealed no hemorrhage (CD31 in green, TER119 in magenta), no change in macrophage/blood vessel association (IBA1 in magenta, CD31 in green), and no loss of the homeostatic marker P2RY12 (magenta), in IBA1⁺ (green) microglia. H Pf4^Cre recombination was restricted to the embryonic meninges (open arrowheads), and was not seen in microglia (closed arrowheads). I, J Quantification of hemorrhage (I) and microglial P2RY12 expression (J) in different Tgfb1 conditional mutant models. n = 3 for histological analysis. In 2B, n = 2 (E14.5), n = 3 (E11.5, 12.5, 16.5) and n = 4 (E10.5, E18.5). ***p < 0.001, two-tailed t-test. In J, p = 4.3*10⁻⁶ and 2.28*10⁻⁷ for Cx3cr1^CreER and P2ry12^CreER comparisons to control, respectively. Error bars= mean +/- SEM. Gene expression in B = normalized RPKM. Scale bar in B–E and G = 150μm, inset scale bar=20μm. Scale bar in F and H = 50μm. Source data are provided as a Source Data file.

Fig. 5. Vascular Tgfb1 is not required for microglial development.

E14.5 coronal brain sections from (A) control (Tgfb^+/−) embryos, (B) embryos with global (Tgfb1^−/−) or cell-lineage specific deletion of Tgfb1 (Tgfb1^fl/fl) in (C) endothelial cells (Cdh5^CreER) (D, H) vascular mural cells (Pdgfrb^Cre), or (E, I) endothelial cells and microglia/macrophages (Tie2^Cre). Sections were stained for hemorrhage (TER119, magenta) and vasculature (CD31, green) or for committed/hemostatic microglia (IBA1, magenta and P2RY12, green), or to study P2RY12 expression in the context of hemorrhage (P2RY12, cyan and TER119, yellow). Only Tgfb1^−/− mutants have consistent evidence of vascular dysplasia (marked by X) and hemorrhage (asterisk), whereas mice with microglia/macrophage deletion of Tgfb1 (Tgfb^−/−, and Tgfb1^fl/fl;Tie2^Cre mutants) have presence of dysmature microglia (open arrowheads, blowups to right). Quantification of hemorrhage (F) and microglial P2RY12 expression (G) in different Tgfb1 conditional mutant models. ***p = 0.0004 in G. Panels in H and I show ROSA-tdTomato (Ai14) recombination pattern of Pdgfrb^Cre and Tie2^Cre mouse lines respectively, and P2RY12 expression (or lack thereof) in the sporadic hemorrhage (arrowheads) seen when Tgfb1 is deleted with these lines. n = 3. ***p < 0.001, two-tailed t-test. Error bars= mean +/- SEM. Cx=Cortex, Str=Striatum. Scale bar in (A–E, H and I) = 150μm, inset scale bar=30μm. Source data are provided as a Source Data file.

Fig. 6. Tgfb1 is required postnatally for microglial homeostasis.

A Bulk-seq analysis of Tgfb1 expression in the adult mouse brain. Analysis revealed enrichment for Tgfb1 expression primarily in microglia and vascular cells. B–E Analysis of conditional Cx3cr1^CreER mediated deletion of Tgfb1 deletion in the P30 mouse cerebral cortex following neonatal tamoxifen induction at P4,5 and 6. Analysis revealed a patchy distribution of dysmature microglia characterized by altered morphology and (B) Tgfb1 loss, (C) upregulation of CD206 and loss of pSMAD3 staining, (D) loss of the homeostatic marker TMEM119 and (E) loss of the homeostatic marker P2RY12. Open arrowheads in B-G mark dysmature microglia, closed arrowheads mark dysmature microglia. Postnatal deletion of Tgfb1 (P30-P60) using P2ry12^CreER resulted in isolated production of microglia with altered morphology (F), loss of pSMAD3 staining (F), downregulation of the homeostatic marker TMEM119 (G), and upregulation of CD206 (F). H Cartoon depiction of the role of TGFβ1 in the cell-autonomous control of microglial homeostasis. Microglial cartoon created in BioRender. McKinsey, G. (2025) https://BioRender.com/r69o433. Scale bar in (B–G) = 100μm, inset scale bar=20μm.

Fig. 7. Disruption of non-canonical TGFβ signaling in microglia drives disease-associated gene expression.

A Schematic of TGFβ signaling, with analyzed mutant mouse models analyzed by bulk and microglial flow cytometry noted by color (Itgb8=red; Tgfb1=cyan; Lrrc33=orange; Tgfb2=blue, Smad2/3=green). B Cumulative incidence of motor dysfunction seen in different conditional TGFβ pathway mutants as a function of age. Incidence included any sign of detriment to gait, appearance, or tremor, based on a 0,1,2 rating scale (see for details). p < 0.0001, two-tailed t-test. C Pearson´s correlation coefficient of bulk-seq normalized expression of differentially expressed genes across TGFβ mutant models. D Compensatory transcriptional changes of key TGFβ signaling genes in different TGFβ mutant models. Heatmap shows log₂FC. E Bulk-seq analysis of microglial homeostatic and disease associated (MGnD/DAM) microglial markers across TGFβ mutant models. Heatmap shows z-score of normalized expression. Differentially expressed genes in all mutants (p < 0.05; DESeq2) are indicated with dotted rectangles. F–H Comparison of control (F) Tgfb1^fl/+;Cx3cr1^Cre, (G) Tgfb1^fl/fl;Cx3cr1^Cre and (H) Smad2/3^fl/fl;Cx3cr1^Cre adult mice. Analysis revealed loss of the homeostatic marker P2RY12 in both conditional Tgfb1 and Smad2/3 mutants, and upregulation of the MGnD-associated microglial marker LGALS3 (see arrowheads). LGALS3 upregulation in Tgfb1 conditional mutants was significantly higher in white matter (asterisks in G and (H) and was only seen in the white matter of Smad2/3 conditional mutants (H). Scale bar in F–H = 50μm, inset scale bar=10μm.