Human microglia maturation is underpinned by specific gene regulatory networks

Abstract

Microglia phenotypes are highly regulated by the brain environment, but the transcriptional networks that specify the maturation of human microglia are poorly understood. Here, we characterized stage-specific transcriptomes and epigenetic landscapes of fetal and postnatal human microglia and acquired corresponding data in induced pluripotent stem cell (iPSC)-derived microglia, in cerebral organoids, and following engraftment into humanized mice. Parallel development of computational approaches that considered transcription factor (TF) co-occurrence and enhancer activity allowed prediction of shared and state-specific gene regulatory networks associated with fetal and postnatal microglia. Additionally, many features of the human fetal-to-postnatal transition were recapitulated in a time-dependent manner following the engraftment of iPSC cells into humanized mice. These data and accompanying computational approaches will facilitate further efforts to elucidate mechanisms by which human microglia acquire stage- and disease-specific phenotypes.

Keywords: epigenomics; humanized; microglia; neurodevelopment; neurological disorders; stem cells; transcription factors; transcriptomics.

Figure 1.. Changing transcriptome profile of human microglia during fetal development.

A. Heatmap expression z-scores of the top 100 genes from each significantly correlated module identified by WGCNA, ranked by Kleinberg’s hub centrality scores. PN=postnatal, CTX=cortex B. MA plot of gene expression between human fetal (DEG, blue) and postnatal microglia (DEG, red). C. Bar charts of the expression of genes represented in GO term analysis of DEGs between fetal and postnatal microglia. Only genes with p-adj < 0.001 and FC > 2 are shown. D. Heatmap expression z-scores of DEGs between first and second trimester fetal microglia. E. MA plot of gene expression differences between male (DEG, blue) and female (DEG, orange) first trimester fetal microglia (left). Bar charts of expression of sex-biased autosomal genes in fetal microglia with male-bias (top right) and female bias (bottom right). F. Bar charts of expression of NDD-associated genes in microglia and brain cortex. Genes with p-adj < 0.001 and FC > 2 are shown. EE=epileptic encephalopathy, Scz=Schizophrenia. Please also see Figure S1-2.

Figure 2.. Putative ligand-receptor interactions influencing microglia development.

A. Circos plot indicating NicheNet prediction of ligand to target genes and predicted receptors in fetal and postnatal microglia. Please also see Figure S3. B. Heatmaps depicting ligand activity score (left) and ligand-receptor interaction score (middle) represented in 2A and of RNA expression of ligands (right) and receptors (bottom). C. Heatmap of gene expression of ligand (bottom)-receptor (top) pairs between human fetal bulk cortex, postnatal bulk cortex, fetal microglia and postnatal microglia. D. Representative images of 3 independent experiments of multi-fluorescent RNAscope of human postnatal brains probed for P2RY12 (red), SORL1 (green), APOE (white); and DAPI. Arrowheads indicate microglia co-expressing P2RY12 and SORL1. E. Representative images of 3 independent experiments multi-fluorescent RNAscope of human postnatal brains probed for P2RY12 (red) or IHC for IBA1 (red), CCL3 (green), CCR5 (white); and DAPI. Arrowheads indicate microglia co-expressing P2RY12 or IBA1, CCL3, and CCR5.

Figure 3.. Maturation of human microglia remodels the active enhancer landscape.

A. (left) Scatter plot of H3K27ac ChIP-seq signal around distal (>1000bp from TSS) ATAC-seq peaks. Differentially acetylated regions (FC > 2, p-adj < 0.05) enriched in fetal (blue) and postnatal (red) microglia are colored. (right) De novo motifs analysis of differentially acetylated regions in fetal (top) or postnatal (bottom) microglia. B. Bar chart showing log2FC of candidate TFs known to bind motifs identified in A, with TFs in blue higher expressed in fetal microglia and red higher expressed in postnatal microglia. *** indicates FC > 2, p-adj < 0.001. Grey, no significant differences. C. (left) UMAP projection and color clustering of 27,041 scATAC-seq profiles of fetal and postnatal microglia. Each dot represents one cell. (right) Bar chart indicating the relative fetal and postnatal contribution to each cluster. D. Heatmap of average chromVAR score per motif and per cluster. Scores are averaged for all cells within each cluster and z-score normalized. E. UMAP visualization of enrichment for indicated motifs using ChromVar. F. Heatmap of gene expression of TFs best matching motifs identified in Figure 2E in fetal and postnatal microglia. Stars indicate genes that are differentially expressed (FC > 2, p-adj <0.05) between fetal and postnatal microglia.

Figure 4.. Human microglia have gene regulatory networks unique to development-stage.

A. Schematic of Transcription factor Interaction inference from Motif co-Occurrence Networks (TIMON). Height of TF is directly correlated with motif score. B. TIMON analysis of all microglia (fetal and postnatal). Node represents a TF motif, edge represents significant (p<0.001) co-occurrence between two transcription factors. Node sizes are proportional to the node degree. Nodes with degree >10 are labeled. C. Fetal microglia specific enhancer motif co-occurrence networks, nodes with degree >3 are labeled. D. Postnatal microglia specific enhancer motif co-occurrence networks, nodes with degree >3 are labeled. E. Bubble plots of significant co-occurring transcription factor pairs in fetal and postnatal microglia. Node size represents the co-occurrence frequency, opacity is proportional to the significance level. F. MITF-PU.1-ELF2 (top) is a TF clique of the fetal microglia network. (middle) GO terms of genes linked to enhancers containing MITF-PU.1-ELF2. (bottom) UCSC browser tracks showing expression of LGALS3 in fetal (blue) and postnatal microglia (red) and corresponding ATAC-seq and H3K27ac ChIP-seq peaks associated with LGALS3. Enhancer regions with MITF-PU.1-ELF2 combinatorial motifs are highlighted yellow. Star indicates the enhancer region in which binding of indicated TFs were validated in Supplemental Figure 4G. G. NR2C2-IRF2-MAFB-PRDM1 (top) is a clique of the postnatal microglia network. (middle) GO terms of genes linked to enhancers containing NR2C2-IRF2-MAFB-PRDM1. (bottom) UCSC browser tracks showing expression of CD74 in fetal (blue) and postnatal microglia (red) and corresponding ATAC-seq and H3K27ac ChIP-seq peaks associated with CD163. Combinatorial NR2C2-IRF2-MAFB-PRDM1 motifs are highlighted yellow. Star indicates the enhancer region in which binding of indicated TFs were validated in Supplemental Figure 4H.

Figure 5.. Transcriptional heterogeneity of primary human microglia during development.

A. Annotation of scRNA-seq clusters with inset depicting RNA velocity analysis of cluster 0. B. (left) tSNE projection of scRNA-seq analysis of fetal and postnatal microglia. Each dot represents one cell with coloring indicating age contribution. Bar graphs illustrating age contribution (middle) and sample contribution (right) to each cluster. C. Split violin plots showing the distribution of gene expression per cluster for fetal (blue) and PN (red) microglia. D. Fetal microglia specific TIMON analysis for cluster 1. E. Circle plot representing TFAct scores per cluster with height of bars indicative of TF activity scores. Please also see Figure S6. F. tSNE projections of TF activity scores for MEF2C (left) and ATF4 (right).

Figure 6.. iPSC and xenotransplantation systems capture distinct stages of human microglia maturation.

A. Schematic demonstrating derivations of HPCs, iMGs, oMGs, and xMGs from iPSCs. B. Immunohistochemistry depicting oMGs (IBA1, red; PU1, white) in proximity to neurons (MAP2, green) in organoids (left) and xMGs (right) (huCD45, red; human nuclei-Ku80, green; DAPI, blue) 3 weeks after engraftment into humanized mouse brains. C. Ratio-ratio plot of genes comparing FC in gene expression between oMGs relative to cerebral organoids versus FC in gene expression between fetal microglia relative to fetal cortex. Pearson’s correlation coefficient is indicated in bottom right. D. TROM correspondence map of the transcriptomes of fetal and postnatal microglia and cortex, iMG, oMGs, and xMGs. Values are TROM scores, 6 being best match. E. Violin plots of gene expression (TPM) that are associated with indicated neurodevelopmental disorders. F. PCA of distal ATAC-seq peaks (> 1000bp from TSS) of primary human microglia, in vitro iPSC models and in vivo xMGs. G. (left) Scatterplot of distal ATAC-seq peaks showing differentially accessible regions in oMGs (blue) and xMGs-8weeks post-engraftment (purple). (right) De novo motif analysis of differentially accessible regions in oMGs (top) and xMGs (bottom). H. Bar chart of expression of TFs known to bind to DNA motifs identified in (G). *p<0.05, ** p<0.01, *** p <0.001, **** p<0.0001. I. Heatmap of LDSC analysis of enrichment of genetic variants associated with listed conditions displayed as -log10(q) value for significance of enrichment for promoter-proximal (light grey) and distal elements (dark grey) in indicated cell types.

Figure 7.. Transcription factor networks underlying environment-dependent genes in microglia.

A. Heatmap of gene expression changes of environmentally dependent genes across primary human microglia and iPSC-systems. B. Mean TPM (LOESS fit) of microglia genes that are increased in in vitro microglia as compared to ex vivo microglia. C. Mean TPM (LOESS fit) of environmentally dependent microglia genes that regain expression in xMGs. D. Transcription factor network, derived from TIMON, of in vitro human microglia. E. Transcription factor network, derived from TIMON, of ex vivo human microglia. F. JUN-C/EBPα-SREBP2 (top) is a clique of the in vitro microglia network. (middle) GO terms of genes linked to enhancers containing JUN-C/EBPα-SREBP2. (bottom) UCSC browser tracks showing RNA expression of CD163 in in vitro (brown) and ex vivo microglia (red) and corresponding ATAC-seq and H3K27ac ChIP-seq peaks associated with CD163. Enhancer regions with JUN-C/EBPα-SREBP2 combination motifs are highlighted in yellow. G. IRF8-EGR1-STAT2 (top) is a clique of the ex vivo microglia network. (middle) GO terms of genes linked to enhancers containing IRF8-EGR1-STAT2. (bottom) UCSC browser tracks showing RNA expression of MEF2A in in vitro (brown) and ex vivo microglia (red) with corresponding ATAC-seq and H3K27ac ChIP-seq peaks. Enhancer regions in which combinations of IRF8-EGR1-STAT2 motifs are detected highlighted in yellow. H. Bar chart of expression of select brain environment regulated TFs. Genes are p <0.001 by one-way ANOVA.