Fig 2: VCAM1 controls the Aβ chemotaxis of chemotactic microglia and subsequent Aβ clearance upon IL-33 treatment.a–f, VCAM1 blockade inhibits the IL-33-induced Aβ chemotaxis of microglia. a, Schematic diagram showing the protocol for neutralizing antibody administration before IL-33 treatment in APP/PS1 mice. b,c, Representative images (b) and violin plot (c) showing the distances between chemotactic microglia and the nearest Aβ plaque 24 h after administration of a VCAM1-neutralizing antibody in IL-33-treated APP/PS1 mice (IgG Con: n = 61 microglia from 5 mice; IgG IL-33: n = 131 microglia from 5 mice; αVCAM1 Con: n = 48 microglia from 5 mice; αVCAM1 IL-33: n = 109 microglia from 5 mice; Kruskal–Wallis test with Dunn’s multiple comparisons test). Dotted circle indicates 10 μm from the perimeter of the Aβ plaque. Arrowheads indicate Vcam1-expressing microglia. Scale bar = 10 μm. d–f, Representative images (d) and bar plots showing the proportions of Aβ plaque-associated microglia (e) and phagocytic microglia (f) 24 h after administration of antibody against VCAM1 in IL-33-treated APP/PS1 mice (IgG Con: n = 9; IgG IL-33: n = 8 for panel e and n = 7 for panel f; αVCAM1 IL-33: n = 9 for panel e and n = 8 for panel f; one-way ANOVA with Dunnett’s multiple comparisons test). Scale bar = 20 μm. g–k, Genetic ablation of VCAM1 in microglia inhibits the Aβ chemotaxis of microglia and microglia-mediated Aβ clearance upon IL-33 treatment. g, Schematic diagram showing the protocol for tamoxifen and IL-33 administration in APP/PS1;VCAM1-icKO mice. h,i, Representative images (h) and bar plot (i) showing the proportions of phagocytic microglia 24 h after IL-33 treatment in APP/PS1;VCAM1-icKO mice (wild-type (WT) Con: n = 11; WT IL-33: n = 12; icKO Con: n = 11; icKO IL-33: n = 13; two-way ANOVA with Šidák’s multiple comparisons test). Scale bar = 20 μm. j,k, Representative images (j) and bar plot (k) showing the Aβ plaque areas in the cortex 48 h after IL-33 treatment in APP/PS1-icKO mice (WT Con: n = 7; WT IL-33: n = 12; icKO Con: n = 11; icKO IL-33: n = 12; two-way ANOVA with Šidák’s multiple comparisons test). Scale bar = 200 μm. All data are mean ± s.e.m.Source data

Fig 3: ApoE is a chemoattractant that directs VCAM1-dependent chemotaxis of microglia.a, STRINGdb protein–protein interaction between VCAM1 (green) and Aβ plaque-associated proteins (orange). b, Schematic diagram showing the protocol for injecting protein-coated beads followed by IL-33 treatment in APP/PS1 mice. c–e, Representative images (c) and bar plots showing the numbers of microglia (d) and MHC-II+ microglia (e) within the area of BSA-, ApoE-, CD44-, and ITGB2-coated beads after IL-33 treatment (Con: BSA: n = 3, ApoE: n = 6, CD44: n = 6, ITGB2: n = 4; IL-33: BSA: n = 5, ApoE: n = 6, CD44: n = 6, ITGB2: n = 5; two-way ANOVA with Šidák’s multiple comparisons test). Dotted line indicates bead area. Scale bar = 20 μm. f, Representative images showing the Vcam1-expressing microglia surrounding ApoE-coated beads after IL-33 treatment. Arrowheads indicate Vcam1-expressing microglia. Scale bars = 50 μm (left) and 20 μm (right). Experiment was repeated for three batches with similar results. g–i, Representative images (g) and bar plots showing the numbers of microglia (h) and MHC-II+ microglia (i) within the areas of beads coated with nonlipidated or lipidated ApoE in IL-33-treated APP/PS1 mice (nonlipidated: n = 7; lipidated: n = 7; two-tailed paired t-test). Dotted line indicates bead area. Scale bar = 20 μm. j–o, VCAM1 is essential for the ApoE chemotaxis of microglia after IL-33 treatment. j–l, Representative images (j) and bar plots showing the numbers of microglia (k) and MHC-II+ microglia (l) within the ApoE-coated bead areas after co-injection of rVCAM1 in IL-33-treated APP/PS1 mice (Fc: n = 6; rVCAM1: n = 6; two-tailed paired t-test). Dotted line indicates bead area. Scale bar = 20 μm. m–o, Representative images (m) and bar plots showing the numbers of microglia (n) and MHC-II+ microglia (o) within the ApoE-coated bead areas in IL-33-treated APP/PS1;VCAM1-icKO mice (wild-type [WT]: n = 6; icKO: n = 6; two-tailed unpaired t-test). Dotted line indicates bead area. Scale bar = 20 μm. All data are mean ± s.e.m.Source data

Fig 4: ST2-dependent induction of chemotactic microglia is required for Aβ clearance upon IL-33 treatment.a,b, Chemotactic microglia migrate toward Aβ plaques after IL-33 treatment. Representative images (a) and violin plot (b) showing the distance between chemotactic microglia (that is, Vcam1+ Cx3cr1+ cells) and the nearest Aβ plaque 3, 8 and 24 h after IL-33 treatment (3 h: n = 83 microglia from four mice; 8 h: n = 92 microglia from four mice; 24 h: n = 89 microglia from four mice; Kruskal–Wallis test with Dunn’s multiple comparisons test). Dotted circle indicates 10 μm from the perimeter of the Aβ plaque. Arrowheads indicate Vcam1-expressing microglia. Scale bar = 10 μm. c,d, Genetic ablation of ST2 inhibits the microglial chemotactic state 3 h after IL-33 treatment. Representative contour plots (c) and bar plot (d) showing the proportions of chemotactic microglia in each group (Con APP/PS1;ST2WT mice: n = 4; IL-33–treated APP/PS1;ST2WT mice: n = 5; Con APP/PS1;ST2KO mice: n = 4; IL-33–treated APP/PS1 mice; ST2KO: n = 5; two-way ANOVA with Šidák’s multiple comparisons test). e–g, Genetic ablation of ST2 inhibits the induction of MHC-II+ phagocytic microglia 24 h after IL-33 treatment. Representative images (e) and bar plots showing the proportions of Aβ plaque-associated microglia (f) and phagocytic microglia (g) in each group (Con APP/PS1;ST2WT mice: n = 9 for panel f and n = 10 for panel g; IL-33-treated APP/PS1;ST2WT mice: n = 10 for panel f and n = 8 for panel g; Con APP/PS1;ST2KO mice: n = 8; IL-33-treated APP/PS1;ST2KO mice: n = 9; two-way ANOVA with Šidák’s multiple comparisons test). Arrowheads indicate phagocytic microglia. Scale bar = 20 μm. h,i, Genetic ablation of ST2 attenuates Aβ clearance induced by IL-33 48 h treatment. Representative images (h) and bar plot (i) showing the Aβ plaque area in the cortex 48 h after IL-33 treatment in each group (Con APP/PS1;ST2WT mice: n = 10; IL-33-treated APP/PS1;ST2WT mice: n = 9; Con APP/PS1;ST2KO mice: n = 9; IL-33–treated APP/PS1;ST2KO mice: n = 7; two-way ANOVA with Šidák’s multiple comparisons test). Scale bar = 200 μm. All data are mean ± s.e.m.Source data

Fig 5: IL-33-responsive microglia undergo stepwise transcriptomic reprogramming.a–c, IL-33 induces prolonged expression of the microglial chemotactic gene signature. a, Heatmap showing the expression levels of 1,433 IL-33-induced genes in microglia 3, 8 and 24 h after IL-33 treatment (adjusted P value < 0.05). The bar on the far right indicates genes showing transient activation (red) and prolonged activation (orange). b, Bar plot showing the top GO pathways associated with the genes showing transient (red) and prolonged (orange) activation (as in panel a). FDR, false discovery rate. c, Protein–protein interaction analysis of genes exhibiting prolonged activation. d–g, IL-33 regulates microglial heterogeneity in a sequential manner. d, Uniform Manifold Approximation and Projection (UMAP) plot of three microglial subtypes showing unbiased clustering of 72,519 microglia from APP/PS1 mice treated with IL-33 or control for 3, 8, or 24 h (each condition corresponds to 3 biological independent samples). e, Dot plot showing the expression levels of the top signature genes of the 3 microglial subtypes. f,g, UMAP plots (f) and bar plots (g) showing the proportions of homeostatic, chemotactic and DAM in the four conditions (n = 3 mice in each condition; one-way ANOVA with Dunnett’s multiple comparisons test). Con, control. h–j, The developmental lineage of IL-33RM involves the sequential homeostatic–chemotactic–phagocytic state transition. UMAP plots showing the cell trajectory (h) and pseudotime ordering (i) of the IL-33RM developmental lineage. j, Heatmap visualizing the smoothed expression levels of the top homeostatic, chemotactic, DAM, and phagocytic signature genes (as described in the text) along the IL-33RM developmental lineage. All data are mean ± standard error of the mean (s.e.m.).Source data