Single-cell transcriptomics reveals distinct inflammation-induced microglia signatures

Abstract

Microglia are specialized parenchymal-resident phagocytes of the central nervous system (CNS) that actively support, defend and modulate the neural environment. Dysfunctional microglial responses are thought to worsen CNS diseases; nevertheless, their impact during neuroinflammatory processes remains largely obscure. Here, using a combination of single-cell RNA sequencing and multicolour flow cytometry, we comprehensively profile microglia in the brain of lipopolysaccharide (LPS)-injected mice. By excluding the contribution of other immune CNS-resident and peripheral cells, we show that microglia isolated from LPS-injected mice display a global downregulation of their homeostatic signature together with an upregulation of inflammatory genes. Notably, we identify distinct microglial activated profiles under inflammatory conditions, which greatly differ from neurodegenerative disease-associated profiles. These results provide insights into microglial heterogeneity and establish a resource for the identification of specific phenotypes in CNS disorders, such as neuroinflammatory and neurodegenerative diseases.

Keywords: heterogeneity; lipopolysaccharide; microglia; neuroinflammation; single‐cell RNA‐seq.

Figure 1. Characterization of acutely isolated CD11b⁺ CD45^int cells

1. FACS gating strategy representative of five independent experiments adopted to sort CD11b⁺CD45^int microglia distinctly from CD11b⁺CD45^high resident macrophages and CD11b⁻CD45^high lymphocytes.

2. Analysis of relative transcript levels of CD11b⁺CD45^int FACS‐sorted microglia compared with whole brain tissue by qPCR. Gene expression levels of microglia (Olfml3, Fcrls, Tmem119, Siglech, Gpr34, P2ry12), astrocyte (Gfap, Gjb6, Ntsr2, Aldh1l1), oligodendrocyte (Mobp, Mog, Cldn1) and neuron (Tubb3, Vglut1, NeuN) markers. Bars represent mean (n = 4; pool of one female and one male per biological replicate) of relative expression (Gapdh as housekeeping gene) ± SEM (*P < 0.05; **P < 0.01 by two‐tailed Student's t‐test). N.D., not detected.

3. Representative quantification of CD206 expression in CD11b⁺CD45^int microglia and CD11b⁺CD45^high resident macrophages. Values denote the percentage of the mean ± SEM of five independent experiments.

4. Representative images of two independent experiments showing microglia, resident macrophages and lymphocytes acquired with ImageStream imaging cytometer (Amnis) based on CD45, CD11b and CD206 expression levels (scale bar represents 7 μm).

Figure EV1. Microglial isolation and characterization

1. FACS gating strategy representative of five independent experiments adopted to sort CD11b⁺CD45^int microglial. Cells were distinguished from debris using forward (FSC‐A) and side (SSC‐A) scatters, followed by cell doublet and aggregate elimination (SSC‐H/SSC‐A). Dead cells were gated out by their strong positivity for the dead cell discrimination marker Hoechst. Single viable microglial cells were gated as CD11b⁺CD45^int.

2. Analysis of microglial purity by qPCR. Gene expression levels of microglial‐specific genes (Itgb5, Sall1, Hexb, Tgfb, Aif1, Cx3cr1, Mertk, Ctss, Tyrobp, Trem2, Itgam, Itgax) in purified microglia compared to whole brain. Bars represent mean (n = 4; one female and one male per sample) of relative expression (Gapdh as housekeeping gene) ± SEM (*P < 0.05; **P < 0.01 by two‐tailed Student's t‐test).

3. Quantification of CD11c, Ly6C and CCR2 expression in CD11b⁺CD45^int microglia and CD11b⁺CD45^high resident macrophages representative of five independent experiments. Values denote the percentage of the mean ± SEM of five independent experiments.

Figure 2. LPS stimulation induces an intrinsic loss of the microglia homeostatic signature

1. A–D

   Three‐ to four‐month‐old C57BL/6N mice were treated with an acute dose of LPS (4 μg/g body) or vehicle (saline). Microglia (pool of two mice per group per replicate; one female and one male) were FACS‐sorted 24 h later. (A) Gene expression levels of microglial homeostatic (Olfml3, Fcrls, Tmem119, Siglech, Gpr34, P2ry12, Mef2c), phagocytic (Tyrobp, Trem2) and inflammatory (Il1b, Tnf, Ccl2, Mrc1, Arg1) markers were analysed by qPCR. Bars represent mean of relative expression (% of saline; Gapdh as housekeeping gene) ± SEM (*P < 0.05; **P < 0.01 by two‐tailed Student's t‐test; n = 4). (B) Representative multicolour flow cytometry analysis of five independent experiments showing CD11b‐ and CD45‐positive populations in single viable cells in saline or LPS‐injected mouse brains. (C) Representative multicolour flow cytometry analysis showing the percentage of the mean ± SEM of five independent experiments of Ly6C‐ and CD206‐expressing cells in CD11b⁺CD45^int cells from saline or LPS‐injected mice. (D) Gene expression levels of the monocytic markers Ly6c1 and Ccr2 in purified microglia (n = 4) and isolated bone marrow monocytes (n = 2) by qPCR. Bars represent mean of relative expression (Gapdh as housekeeping gene) ± SEM (**P < 0.01 by two‐tailed Student's t‐test).

2. E

   Primary adult microglia were cultivated in the presence of TGF‐β (50 μg/ml) and M‐CSF (10 ng/ml), while neonatal cells were stimulated for 24 h with TGF‐β (50 μg/ml) followed by 6 h of stimulation with LPS (1 ng/ml) or left untreated. Expression levels of microglial homeostatic (Olfml3, Tmem119, Gpr34) and inflammatory (Il1b, Tnf, Ccl2) genes were analysed by qPCR. Bars represent mean of relative expression (Gapdh as housekeeping gene) ± SEM (*P < 0.05; **P < 0.01 by two‐tailed Student's t‐test).

Figure EV2. Ex vivo and in vitro microglial characterization

1. A, B

   Three‐ to four‐month‐old mice were treated with an acute dose of LPS (4 μg/g body) or vehicle (saline). Microglial (pool of two mice per group; one female and one male) were isolated 24 h later. (A) Gene expression levels of microglial homeostatic genes (Itgb5, Sall1, Hexb, Tgfb, Mertk, Ctss, Itgam, Cx3cr1) were analysed by qPCR. Bars represent mean of relative expression (% of saline; Gapdh as housekeeping gene) ± SEM (**P < 0.01 by two‐tailed Student's t‐test; n = 4). (B) Representative multicolour flow cytometry analysis of five independent experiments showing CD45, CD11b, CD86 and CD11c expression levels in CD11b⁺CD45^int microglia of saline or LPS‐injected mouse brains.

2. C

   Representative results of two independent experiments showing the purity of MACS‐isolated bone marrow monocytes based on the expression levels of the monocytic marker Ly6C.

3. D

   Comparison of the homeostatic signature (Tmem119, Siglech, Gpr34, P2ry12) between primary and acutely isolated microglia. Primary adult microglia were cultivated in the presence of TGF‐β (50 μg/ml) and M‐CSF (10 ng/ml), while neonatal cells were treated for 24 h with TGF‐β or left untreated. Gene expression levels were analysed by qPCR and normalized using Gapdh as housekeeping gene. Bars represent mean ± SEM (**P < 0.01 by two‐tailed Student's t‐test; n = 3).

Figure 3. Characterization of microglial activation at the single‐cell level

1. Heatmap showing clustering analysis of 1,247 single cells, featuring 100 most variable genes (FDR < 0.05). Single‐cell RNA‐seq results are obtained from two mice per group (one female and one male each). Values denote a score based on gene expression rank.

2. 2D‐tSNE representation of all single cells included in the study (n = 1,247) depicting the separation of microglia isolated from LPS‐injected mice (770 cells in red) and steady state (477 cells in blue) in two main clusters.

3. Expression of specific homeostatic (Tmem119, P2ry12, Siglech) and inflammatory (Ccl2, Gpr84) genes overlaid on the 2D‐tSNE space. Bars represent log2 (Count + 1).

4. Representative multicolour flow cytometry analysis of two independent experiments showing TMEM119 and P2RY12 expression levels in CD11b⁺CD45^int microglia of saline or LPS‐injected mouse brains. For the unconjugated TMEM119 antibody, negative stands for primary antibody without secondary antibody. For P2RY12 antibody, negative represents isotype PE control.

Figure EV3. Inflammatory‐ versus disease‐associated signatures

1. A

   Scatterplot comparing the fold change of genes (log2 scale) between microglia isolated from LPS‐injected mice (x‐axis) versus DAM (y‐axis) compared to homeostatic microglia (FDR < 0.05).

2. B

   Venn diagrams showing the number of genes upregulated (left) or downregulated (right) uniquely under LPS treatment, exclusively by DAM and shared between LPS and DAM compared to their corresponding controls (FDR < 0.05).

3. C–E

   Top 10 biological processes identified by Database for Annotation, Visualization and Integrated Discovery (DAVID) resulting from (C) 960 uniquely upregulated genes under LPS treatment, (D) 597 increased genes specific for DAM and (E) 215 shared genes versus control conditions.

4. F

   Inflammatory‐ and disease‐associated signatures identified by distinct genes corresponding to specific microglial functions (homeostasis, phagocytosis, lipid metabolism, inflammation and lysosomal activity).

Figure 4. Identification of microglial subpopulations under inflammatory conditions

1. 2D‐tSNE representation of 1,247 single cells isolated from naïve (blue)‐ and LPS‐treated mice showing two distinct subpopulations among the 770 cells isolated from LPS‐injected mice (n = 703, red; n = 67, yellow).

2. Venn diagram showing 732 genes uniquely upregulated in the “main LPS” cluster (red) and 241 genes exclusively increased in the “subset LPS” (yellow) compared to their corresponding controls (blue) (FDR < 0.05). A total of 274 genes were shared between the two LPS populations.

3. Venn diagram showing 1,055 genes uniquely downregulated in the “main LPS” cluster (red) and 29 genes exclusively decreased in the “subset LPS” (yellow) compared to their corresponding controls (blue) (FDR < 0.05). A total of 87 genes were shared between the two LPS populations.

4. Heatmap showing examples of specific genes mainly upregulated in “main LPS” (Manf) or “subset LPS” (Ash1l) and downregulated in “main LPS” (Mef2c) or “subset LPS” (Lamp1) overlaid on the 2D‐tSNE space. Bars represent log2 (Count + 1).

5. Gene set enrichment analysis (GO, top 10 biological processes) of 99 downregulated and 397 upregulated genes distinguishing cells in “subset LPS” versus “main LPS” (FDR < 0.05).

Figure EV4. Representation of mainly top deregulated genes in “main LPS” and “subset LPS”

Expression of specific genes uniquely upregulated in “main LPS” (Cd52, Cd63, Ctsl, Fth1, C5ar1) or “subset LPS” (Golga4, Stab 1, Atrx) and downregulated in “main LPS” (Maf) or “subset LPS” (Cd68, C1qc, Cd81) overlaid on the 2D‐tSNE space. Bars represent log2 (Count + 1).

Figure EV5. FACS and immunohistochemistry analyses of “main LPS” and “subset LPS” upregulated genes and further comparison with DAM

1. A

   Quantification of CD44, CD274 and NOTCH4 expression in CD11b⁺CD45^int microglia under homeostatic (in blue) and inflammatory (in red) conditions. Values denote the percentage of cells obtained from one representative experiment.

2. B

   Immunohistochemistry for “subset LPS” marker (NOTCH4) in brain sections of tissue isolated from LPS‐injected mice after 24 h (IBA1 in red and NOTCH4 in yellow). Cell nuclei were counterstained with Hoechst (in blue). Scale bars: left panel 10 μm, right panel 20 μm. Arrows show NOTCH4 expression in IBA1+ cells.

3. C, D

   Venn diagrams showing unique and commonly (C) upregulated and (D) downregulated genes among “main LPS” cluster (red), “subset LPS” (yellow) and DAM (green) (FDR < 0.05).

Figure 5. Pseudotime analysis

1. Branching analysis of LPS‐activated microglia by Monocle 2 leads to nine distinct clusters in a two‐dimensional state space inferred by generalized regression modelling (see Materials and Methods) showing the major difference of “subset LPS” (in yellow) compared to the other clusters corresponding to “main LPS” (in red).

2. Monocle estimated a pseudotime for each cell along the inferred cell trajectory within the state space showing a delayed activation pattern of “subset LPS” compared to the other fractions.

3. Pseudotime dynamics of inflammatory (Ccl12, Ccl2, Gpr84) and homeostatic (Mef2c, P2ry12, Siglech) genes in dependence on inferred cell states.