Hemorrhagic stroke-induced subtype of inflammatory reactive astrocytes disrupts blood-brain barrier

Abstract

Astrocytes undergo disease-specific transcriptomic changes upon brain injury. However, phenotypic changes of astrocytes and their functions remain unclear after hemorrhagic stroke. Here we reported hemorrhagic stroke induced a group of inflammatory reactive astrocytes with high expression of Gfap and Vimentin, as well as inflammation-related genes lipocalin-2 (Lcn2), Complement component 3 (C3), and Serpina3n. In addition, we demonstrated that depletion of microglia but not macrophages inhibited the expression of inflammation-related genes in inflammatory reactive astrocytes. RNA sequencing showed that blood-brain barrier (BBB) disruption-related gene matrix metalloproteinase-3 (MMP3) was highly upregulated in inflammatory reactive astrocytes. Pharmacological inhibition of MMP3 in astrocytes or specific deletion of astrocytic MMP3 reduced BBB disruption and improved neurological outcomes of hemorrhagic stroke mice. Our study demonstrated that hemorrhagic stroke induced a group of inflammatory reactive astrocytes that were actively involved in disrupting BBB through MMP3, highlighting a specific group of inflammatory reactive astrocytes as a critical driver for BBB disruption in neurological diseases.

Keywords: Single-cell RNA sequencing; blood-brain barrier; hemorrhagic stroke; matrix metalloproteinase-3; reactive astrocytes.

Figure 1.

scRNA-seq reveals heterogeneity of astrocytes in the striatum of hemorrhagic stroke mice. (a) A UMAP plot of 33,048 single cells from the striatum in control (11,308) and 3 days post hemorrhagic (21,740) mice. Colors denote different cell clusters. n = 8 mice per group. (b) A UMAP plot of 3,262 astrocytes from the striatum in control and 3 days post hemorrhagic mice. Colors denote different astrocyte phenotypes. (c) Expression of reactive and inflammatory genes in astrocytes. (d) Volcano plots comparing DEGs among cluster 2 versus cluster 0 and 1. (e) Feature plot of Lcn2, Serpina3n, Vim, Gfap, C3, S100a10, S100a6 genes expressed in astrocytes and (f) Representative immunostaining images of astrocytes (GFAP, green) colocalized with Lcn2, Serpina3n, Vim, C3, S100a10, and S100a6 (red) in the peri-lesion area at 3 days after hemorrhagic stroke. Scale bars, 100 μm.

Figure 2.

Inflammation-related genes in reactive astrocytes were induced by residential microglia but not infiltrated macrophages. (a–f) Representative immunostaining images of astrocytes (GFAP, green) colocalized with Lcn2 (a), Serpina3n (b), Vim (c), C3d (d), S100a10 (e), and S100a6 (f) (red) in the hemorrhagic perifocal area of mice treated with vehicle, PLX5622, F70101C-A, and PLX5622+F70101C-A at day 3 of hemorrhagic stroke. Scale bars, 100 μm. Percentage of Lcn2⁺/GFAP⁺ (a), Serpina3n⁺/GFAP⁺ (b), Vim⁺/GFAP⁺ (c), C3d⁺/GFAP⁺ (d), S100a10⁺/GFAP⁺ (e), and S100a6⁺/GFAP⁺ (f) cells in the hemorrhagic perifocal area at day 3 of hemorrhagic stroke. n = 6 mice per group. One-way ANOVA followed by Dunnett’s test. *p < 0.05, ***p < 0.001, compared with the Vehicle group. Data are mean ± SD.

Figure 3.

RNA-seq analysis of astrocytes treated with conditioned medium from LPS-stimulated microglia (LPS-MCM) or hemin-stimulated microglia (Hemin-MCM). (a) Volcano plot showed the upregulated (right) and downregulated (left) genes between astrocytes treated with LPS-MCM and astrocytes treated with medium derived from resting microglia (Resting-MCM). The horizontal axis is log₂ fold change, and the vertical axis is -log₁₀ p value, p < 0.05. (b) Heatmap showed the overall distribution of differentially expressed genes between LPS-MCM treated astrocytes and Resting-MCM treated astrocytes. Gene expression data was colored in orange for high expression and blue for low expression. (c) KEGG pathway enrichment analysis of differentially expressed genes in LPS-MCM treated astrocytes versus Resting-MCM treated astrocytes. (d) Volcano plot showed the upregulated (right) and downregulated (left) genes between Hemin-stimulated microglia-conditioned medium (Hemin-MCM) treated astrocytes and Resting-MCM treated astrocytes. (e) Heatmap showed the overall distribution of differentially expressed genes between Hemin-MCM treated astrocytes and Resting-MCM treated astrocytes. (f) KEGG pathway enrichment analysis of differentially expressed genes in Hemin-MCM treated astrocytes versus Resting-MCM treated astrocytes. (g) Real-time PCR analysis of MMP3 in Resting-MCM treated, LPS-MCM treated, and Hemin-MCM treated astrocytes. n = 3 biologically independent primary astrocytes cultures. GAPDH was used as an internal control. One-way ANOVA followed by Tukey’s test. **p < 0.01 compared with the Resting-MCM group. (h) Representative immunostaining images of MMP3 (red) and GFAP (green) in the hemorrhagic perifocal area at day 3 of hemorrhagic stroke. Scale bars, 50 μm. (i) Representative immunoblots of MMP3 in the hemorrhagic perifocal area at day 3 of hemorrhagic stroke. (j) Quantification of MMP3 protein levels. n = 3–4 mice per group. β-actin was used as an internal control. Two-sided, unpaired Student’s test. ***p < 0.001 and (k) Quantification of MMP3 activity in the hemorrhagic perifocal area at day 3 of hemorrhagic stroke. n = 3–4 mice per group. Two-sided, unpaired Student’s test. ***p < 0.001. All data are mean ± SD.

Figure 4.

Conditioned medium derived from LPS-MCM or Hemin-MCM treated astrocytes reduced endothelial tight junctions, which reversed by inhibition of MMP3 in vitro. (a) Schematic diagram showed endothelial cells were treated with medium derived from astrocytes that were treated with LPS-MCM (LPS-MCM-ACM) or Hemin-MCM (Hemin-MCM-ACM) for 24 h. (b) Representative immunostaining images showed tight junction proteins ZO-1 (red) and Claudin-5 (red) expressed in CD31⁺ endothelial cells (green) that were treated by Resting-MCM-ACM, LPS-MCM-ACM, and Hemin-MCM-ACM. Scale bars, 25 μm. (c) Representative immunoblots of ZO-1 and Claudin-5 in endothelial cells that were treated by Resting-MCM-ACM, LPS-MCM-ACM, and Hemin-MCM-ACM. (d) Quantification of ZO-1 and Claudin-5 protein levels in endothelial cells. n = 3 biologically independent endothelial cell cultures. β-actin was used as an internal control. One-way ANOVA followed by Tukey’s test. *p < 0.05, **p < 0.01. (e) Schematic diagram showed treatment of endothelial cells with Hemin-MCM-ACM or medium derived from astrocytes that were treated with Hemin-MCM and MMP3 inhibitor (Hemin-MCM+MMP3 inhibitor-ACM) for 24 h. (f) Representative immunostaining images showed tight junction proteins ZO-1 (red) and Claudin-5 (red) expression in CD31⁺ endothelial cells (green) that were treated by Resting-MCM-ACM, Hemin-MCM-ACM, Hemin+MCM-MMP3 inhibitor-ACM. Scale bars, 25 μm. (g) Representative immunoblots of ZO-1 and Claudin-5 in endothelial cells that were treated by Resting-MCM-ACM, Hemin-MCM-ACM and Hemin-MCM+MMP3 inhibitor-ACM and (h) Quantification of ZO-1 and Claudin-5 protein levels in endothelial cells. n = 3 biologically independent endothelial cell cultures. β-actin was used as an internal control. One-way ANOVA followed by Tukey’s test. *p < 0.05, **p < 0.01. All data are mean ± SD.

Figure 5.

Conditional knockout of MMP3 in astrocytes attenuated BBB disruption and neurobehavioral deficits after hemorrhagic stroke. (a) Representative perfused whole brains of Oil treated sham mice (Oil Sham), tamoxifen treated Aldh1l1^Cre-ERT2; MMP3^flox/flox sham mice (Tam Sham), Oil treated hemorrhagic stroke mice (Oil HS) and tamoxifen treated Aldh1l1^Cre-ERT2; MMP3^flox/flox Continued.hemorrhagic stroke mice (Tam HS) after Evans blue injection at day 3 after hemorrhagic stroke. (b) Quantification of extravasated Evans blue in brains at day 3 after hemorrhagic stroke. n = 4–6 mice per group. One-way ANOVA followed by Dunnett’s test. ***p < 0.001. (c) Representative brain sections and immunostaining images of IgG leakage at day 3 after hemorrhagic stroke. Scale bars, 50 μm. (d) Semi-quantification of IgG intensity in the hemorrhagic perifocal area at day 3 after hemorrhagic stroke. Statistics are derived from 16 slices, n = 6 mice per group. One-way ANOVA followed by Tukey’s test. ***p < 0.001. (e) Representative immunoblots of ZO-1 and Claudin-5 in the hemorrhagic perifocal area at day 3 after hemorrhagic stroke. (f) Quantification of ZO-1 and Claudin-5 protein levels. n = 6 mice per group. β-actin was used as an internal control. One-way ANOVA followed by Tukey’s test. ***p < 0.001. (g) TEM images showed representative capillaries in the hemorrhagic perifocal area at day 3 after hemorrhagic stroke. Scale bars, 0.5 μm. BM, basement membrane. Endothelial BM and parenchymal BM is lightly shaded blue and yellow, respectively. (h, i) Quantification of endothelial BM and parenchymal BM thickness in the hemorrhagic perifocal area at day 3 after hemorrhagic stroke. Statistics are derived from 15 capillaries, n = 3 mice per group. One-way ANOVA followed by Tukey’s test. ***p < 0.001. (j) Representative gelatin zymogram showing MMP9 activity in the hemorrhagic perifocal area at day 3 after hemorrhagic stroke. (k) Quantification of pro-MMP9 and active MMP9 activities in the hemorrhagic perifocal area at day 3 of hemorrhagic stroke. n = 4 mice per group. One-way ANOVA followed by Tukey’s test. ***p < 0.001. (l–o) Neurobehavioral outcomes were assessed by four neurobehavioral tests including the mNSS (l), EBST (m), rotarod test (n), and grid walking test (o). n = 10–17 mice per group. Two-way ANOVA followed by Bonferroni’s test. *p < 0.05, **p < 0.01, compared with the Oil HS group. All data are mean ± SD.