Single-cell profiling of brain pericyte heterogeneity following ischemic stroke unveils distinct pericyte subtype-targeted neural reprogramming potential and its underlying mechanisms

Abstract

Rationale: Brain pericytes can acquire multipotency to produce multi-lineage cells following injury. However, pericytes are a heterogenous population and it remains unknown whether there are different potencies from different subsets of pericytes in response to injury. Methods: We used an ischemic stroke model combined with pericyte lineage tracing animal models to investigate brain pericyte heterogeneity under both naïve and brain injury conditions via single-cell RNA-sequencing and immunohistochemistry analysis. In addition, we developed an NG2⁺ pericyte neural reprogramming culture model from both murine and humans to unveil the role of energy sensor, AMP-dependent kinase (AMPK), activity in modulating the reprogramming/differentiation process to convert pericytes to functional neurons by targeting a Ser 436 phosphorylation on CREB-binding protein (CBP), a histone acetyltransferase. Results: We showed that two distinct pericyte subpopulations, marked by NG2⁺ and Tbx18⁺, had different potency following brain injury. NG2⁺ pericytes expressed dominant neural reprogramming potential to produce newborn neurons, while Tbx18⁺ pericytes displayed dominant multipotency to produce endothelial cells, fibroblasts, and microglia following ischemic stroke. In addition, we discovered that AMPK modulators facilitated pericyte-to-neuron conversion by modulating Ser436 phosphorylation status of CBP, to coordinate an acetylation shift between Sox2 and histone H2B, and to regulate Sox2 nuclear-cytoplasmic trafficking during the reprogramming/differentiation process. Finally, we showed that sequential treatment of compound C (CpdC) and metformin, AMPK inhibitor and activator respectively, robustly facilitated the conversion of human pericytes into functional neurons. Conclusion: We revealed that two distinct subtypes of pericytes possess different reprogramming potencies in response to physical and ischemic injuries. We also developed a genomic integration-free methodology to reprogram human pericytes into functional neurons by targeting NG2⁺ pericytes.

Keywords: CBP S436 phosphorylation; Sox2; acetylation; cellular reprogramming; focal ischemic stroke; histone 2B; induced neural stem cells; neuronal differentiation; pericytes.

Figure 1

Naïve NG2⁺ pericytes and Tbx18⁺ pericytes are two distinct pericyte populations. (A) Schematic of experimental timeline, created with BioRender.com. 10-7 days before injury both NG2-CreER^T2/Ai14-flx and Tbx18-CreER^T2/Ai14-flx mice received tamoxifen injections daily for 4 days. Intracerebral injections of ET-1/L-NAME (or saline) were performed 3 days prior to scRNA-seq. Three groups of Tbx18-Ai14⁺ and NG2-Ai14⁺ cells from 1) no injury (naïve), 2) physical injury (saline), and 3) ischemic injury (ET-1/L-NAME) were FAC sorted for tdT (Ai14)⁺/DAPI^- and scRNA-seq was performed. (B) Number of cells obtained and mean number of genes per cell for both naïve NG2-tdT⁺ and naïve Tbx18-tdT⁺. (C) Visualization of cells from naïve NG2-tdT⁺ and naïve Tbx18-tdT⁺ groups after PCA and UMAP, coloured by Seurat clustering and annotated by cell type. (D) Proportion of cells in each cluster for naïve NG2-tdT⁺ and naïve Tbx18-tdT⁺. (E) UMAP visualization of NG2-tdT⁺ (orange) and Tbx18-tdT⁺ (purple) groups. (F) Visualization of the total cell population after PCA and UMAP, coloured by expression of key marker genes (Jun, Atp13a5, Pdgfrβ, Vtn, Acta2, Cldn5, Ly6c1, Nnmt, Pdgfrα, and Mog). (G) Visualization of the total cell population colored by pseudotime using Monocle3. (H) Flowchart of naïve NG2-CreER^T2 X YFP-flx mice receiving tamoxifen treatment 10 days prior to sacrifice for immunohistochemistry. (I-J) Image and quantitative analysis of the proportion of Tbx18⁺/NG2-YFP⁺ and NG2-YFP⁺/Tbx18⁺ cells in cortical sections from naïve mice, immunostained for NG2-YFP (green) and Tbx18 (red) and counterstained for Hoechst (blue). Scale bar: 50 µm.

Figure 2

NG2⁺ pericytes show strong neurogenic potential following ischemic stroke by reprogramming into radial glial cells. (A) Number of cells obtained and mean number of genes per cell for naïve, physical injury, and ischemic stroke conditions obtained from NG2-tdT⁺ mice. (B-C) Visualization of cells from NG2-tdT⁺ naïve, physical injury, and ischemic stroke after PCA and UMAP, coloured by Seurat clustering and annotated by cell type. (D) The proportion of cells in each cluster for naïve, physical injury, and ischemic stroke obtained from NG2-tdT⁺ mice. (E) Visualization of the total NG2-tdT⁺ cell population after PCA and UMAP, coloured by expression of key marker genes (Jun, Abcc9, Pdgfrβ, Vtn, Acta2, Mog, Cldn5, Pdgfrα, Fabp7, Sox2, Nnmt, and C1qa). (F) Visualization of the total NG2-tdT⁺ cell population colored by pseudotime using Monocle3. (G) Velocity vectors for the total NG2-tdT⁺ cell population, visualized and calculated from RNA velocity using the dynamic model, projected onto UMAP visualizations of clusters 1, 6, and 7. (H) GO enrichment (biological process) results of upregulated differentially expressed genes in cluster 7 (RGPs) compared to cluster 1 (canonical pericytes). Log2 fold-change > 0.25 and p-value adjusted < 0.05.

Figure 3

NG2⁺ pericytes exhibit strong neural reprogramming potential following brain injury. (A) Flowchart of brain injury induced by intracerebral injections of ET-1/L-NAME (or saline) into the sensory-motor cortex of NG2-CreER^T2/Ai14-flx mice receiving tamoxifen treatment 7 days prior to injury and sacrificed at 3 days after injury for immunohistochemistry. Cresyl violet image of a brain section at 3 days post-stroke. The red box shows where representative immunohistochemical images were taken. (B-C) Images and quantitative analysis of the proportion of Sox2⁺/tdT⁺ i-NSCs in the cortex sections from mice receiving ET-1/L-NAME (stroke) or saline (physical injury) injections, immunostained for Sox2 (green) and tdT (red), and counterstained for Hoechst (blue). Scale bar: 100 µm. (D-E) Images and quantitative analysis of the percentage of DCX⁺/tdT⁺ neuroblasts in the cortex sections, immunostained for DCX (green) and tdT (red) and counterstained for Hoechst (blue). Scale bar: 100 µm (left panel); 50 µm (right panel). White boxes in the left panels were enlarged in the right panels. (F-G) Images and quantitative analysis of the percentage of Olig1⁺/tdT⁺ OL lineage cells in the cortex sections, immunostained for Olig1 (green) and tdT (red) and counterstained for Hoechst (blue). Scale bar: 100 µm. (H-I) Images and quantitative analysis of the proportion of Iba1⁺/tdT⁺ microglia in the cortex sections, immunostained for Iba1 (green) and tdT (red) and counterstained for Hoechst (blue). Scale bar: 100 µm. White boxes in the upper panels were enlarged in the bottom panels. (J-K) Images and quantitative analysis of the percentage of Col1a1⁺/tdT⁺ fibroblasts in the cortex sections, immunostained for Col1a1 (green) and tdT (red) and counterstained for Hoechst (blue). Scale bar: 100 µm (left panel); 50 µm (right panel). White boxes in the left panels were enlarged in the right panels. (L-M) Images and quantitative analysis of the proportion of NG2-tdT⁺ cells that were adjacent to CD31⁺ micro-vessels, immunostained for CD31 (green) and tdT⁺ (red) and counterstained for Hoechst (blue). Scale bar: 100 µm. (N) Quantitative analysis of the percentage of CD31⁺/tdT⁺ micro-vessels, as shown in (L) in the cortex sections. Arrows denote co-labelled cells. PM: Pia Mater. (O-P) Images and quantitative analysis of the proportion of NeuN⁺/tdT⁺ in the cortex 28 days post-stroke, immunostained for NeuN (green) and tdT (red), and counterstained for Hoechst (blue). Scale bar: 50 µm. n=3-4 animals/group. Student t-test, *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 4

Tbx18⁺ pericytes exhibit strong vascular-genic potential following ischemic stroke. (A) Number of cells obtained and mean number of genes per cell for naïve, physical injury, and ischemic stroke conditions obtained from Tbx18-tdT⁺ mice. (B-C) Visualization of cells from Tbx18-tdT⁺ naïve, physical injury, and ischemic stroke after PCA and UMAP, coloured by Seurat clustering and annotated by cell type. (D) The proportion of cells in each cluster for naïve, physical injury, and ischemic injury was obtained from Tbx18-tdT⁺ mice. (E) Spp1 expression visualized in total Tbx18-tdT⁺ population and in integrated naïve NG2-tdT⁺ and naïve Tbx18-tdT⁺ population after PCA and UMAP. (F) Visualization of the total Tbx18-tdT⁺ cell population after PCA and UMAP, coloured by expression of key marker genes (Cldn5, Cd34, Acta2, Abcc9, Vtn, Pdgfrβ, C1qa, Col1a1, Col15a1, and Sox2). (G) Visualization of the total Tbx18-tdT⁺ cell population, coloured by pseudotime using Monocle3. (H-I) Visualization of cells subset from Tbx18-tdT⁺ vascular-genic pericytes in naïve, physical injury, and ischemic stroke after PCA and UMAP, coloured by Seurat clustering and annotated by cell type. (J) The proportion of cells in each subcluster for cells subset from Tbx18-tdT⁺ vascular-genic pericytes in naïve, physical injury, and ischemic stroke. (K) Visualization of the cell subsets from Tbx18-tdT⁺ vascular-genic pericytes after PCA and UMAP, coloured by expression of key marker genes (Cldn5, Acta2, Vtn, C1qa, and Col1a1).

Figure 5

Tbx18⁺ pericytes predominantly produce micro-vessels following brain injury. (A) Flowchart of brain injury induced by intracerebral injections of ET-1/L-NAME (or saline) into the sensory-motor cortex of Tbx18-CreER^T2/Ai14-flx mice receiving tamoxifen treatment 7 days prior to injury and sacrificed at 3 days after injury for immunohistochemistry. Cresyl violet image of a brain section at 3 days post-stroke. The red box shows where representative immunohistochemical images were taken. (B-C) Images and quantitative analysis of the proportion of Sox2⁺/tdT⁺ i-NSCs in the cortex sections from mice receiving ET-1/L-NAME (stroke) or saline (physical injury) injections, immunostained for Sox2 (green) and tdT⁺ (red) and counterstained for Hoechst (blue). (D-E) Images and quantitative analysis of the percentage of DCX⁺/tdT⁺ neuroblasts in the cortex sections, immunostained for DCX (green) and tdT⁺ (red) and counterstained for Hoechst (blue). (F-G) Images and quantitative analysis of the percentage of Olig1⁺/tdT⁺ OL lineage cells in the cortex sections, immunostained for Olig1 (green) and tdT (red) and counterstained for Hoechst (blue). (H-I) Images and quantitative analysis of the proportion of Iba1⁺/tdT⁺ microglia in the cortex sections, immunostained for Iba1 (green) and tdT⁺ (red) and counterstained for Hoechst (blue). (J-K) Images and quantitative analysis of the percentage of Col1a1⁺/tdT⁺ fibroblasts in the cortex sections, immunostained for Col1a1 (green) and tdT⁺ (red) and counterstained for Hoechst (blue). (L-M) Images and quantitative analysis of the proportion of Tbx18-tdT⁺ cells that were adjacent to CD31⁺ micro-vessels, immunostained for CD31 (green) and tdT⁺ (red) and counterstained for Hoechst (blue). (N) Quantitative analysis of the percentage of CD31⁺/tdT⁺ micro-vessels, as shown in (L) in the cortex sections. Arrows denote co-labelled cells. Scale bar: 100 µm. n=4 animals/group. Student t-test, *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 6

Inactivation of the aPKC-CBP pathway induces an acetylation shift from Sox2 to H2B and Sox2 nuclear import during i-NSC reprogramming from a-pericytes. (A) a-pericytes expanded on an uncoated plastic dish were immunostained with pericyte markers, NG2 (red) and Pdgfrβ (green), counterstained with Hoechst (blue). (B-C) i-neurospheres were produced from cultured a-pericytes isolated from stroke-injured Sox2-GFP reporter mouse cortical tissues when treated with neural conditioned medium (NCM) and expressed GFP signal (B); cytospin of these i-neurospheres and immunocytochemistry for Sox2 (red), GFP (green), counterstained for Hoechst (blue) at 2 and 4 weeks upon reprogramming (NCM treatment, C). (D-G) Confocal images and quantitative analyses of Sox2 nucleus/cytosol intensity ratio in WT and CbpS436A i-NSCs (D-E), in the absence and presence of CpdC (F-G) at 3 weeks upon reprogramming. (H) Immunoprecipitation analysis of Sox2 acetylation in WT and CbpS436A i-NSCs (left panels) or in the absence and presence of CpdC (right panels). i-NSC lysates were immunoprecipitated with a Sox2 antibody, washed and then blotted with pan-acetyl and Sox2 antibodies. Western blot analysis for H2BK5 acetylation in i-NSCs from WT and CbpS436A mice. Blots were probed for acetyl-H2BK5 and total H2B. (I) Graphs show relative levels of acetylation of Sox2 and H2BK5 over total Sox2 and H2B, respectively, normalized to control samples (WT and WT-DMSO, respectively). Scale bar: 20 µm. ***P < 0.05; n=3-5 animals/group.

Figure 7

Activation of the aPKC-CBP pathway induces an acetylation shift from H2B to Sox2 and Sox2 nuclear export during neuronal differentiation of SVZ NPC and i-NSC. (A) Schematic of experimental design for the differentiation of SVZ NPCs using a neurosphere culture model. (B) Immunoprecipitation analysis of Sox2 acetylation in WT and CbpS436A differentiating SVZ NPCs in the absence and presence of metformin (left panels). SVZ NPCs lysates were immunoprecipitated with a Sox2 antibody, washed and then blotted with pan-acetylated, CBP and Sox2 antibodies. Western blot analysis for H2BK5 acetylation in SVZ NPCs from WT and CbpS436A mice (right panels). Blots were probed for acetyl-H2BK5 and total H2B. (C) Graphs show relative levels of acetylation of Sox2 and H2B over total Sox2 and H2B, respectively, normalized to controls for WT (top panel) and CbpS436A (bottom panel) SVZ NPCs without metformin. (D) Graphs show relative levels of CBP-IB over total pulled-down Sox2, normalized to controls for WT (top panel) and CbpS436A (bottom panel) SVZ NPCs in the absence of metformin. (E-F) Confocal images and quantitative analyses of Sox2 nucleus/cytosol intensity ratio in WT and CbpS436A SVZ NPCs 6 days upon neuronal differentiation in the absence and presence of metformin and treated with MG132 (1 µM) 16 h prior to fixation. (G-H) Confocal images and quantitative analyses of Sox2 nucleus/cytosol intensity ratio in WT and CbpS436A i-NSCs 2 days upon neuronal differentiation in the absence and presence of metformin. (I-J) Photographs and quantitative analyses of the percentage of βIII tubulin-positive newly born neurons from WT and CbpS436A i-NSCs 7 days upon neuronal differentiation in the absence and presence of metformin. Scale bar: 20 µm. **P < 0.01; *P < 0.05; n=3-5 animals/group.

Figure 8

Sequential treatment of CpdC and metformin facilitates reprogramming /differentiation of NG2⁺ pericytes into functional neurons in culture. (A-B) Representative image and quantification of cultured human pericytes isolated from human leptomeningeal tissues, immunostained for NG2 (red) and Pdgfrβ (green), counterstained with Hoechst (blue). (C) Experimental timeline of cultured human pericytes undergoing different conditions for neural reprogramming. (D-E) Representative images and quantification of the proportion of nuclear Sox2⁺ (red) i-NSCs over total live cells from either oxygen-glucose deprivation (OGD) or oxygen deprivation (O2 Dep) conditions. Experimental groups treated with CpdC were denoted with +C. Arrows denote Sox2⁺ i-NSCs, counterstained with Hoechst. (F-G) Representative images and quantification of the proportion of βIII tubulin⁺ (green) neurons over total live cells upon receiving NDM treatment in the absence and presence of EGF (E), FGF2 (F), and metformin (M) for 1 week, analysed with One-way ANOVA. (H) Quantification of the proportion of βIII tubulin⁺ neurons over total live cells upon receiving NDM + E + F + M treatment for two weeks, analysed with Student t-test. (I-K) Representative images and quantification of the proportion of Vglut2⁺ (green) and NeuN⁺ (red) neurons over total live cells upon receiving NDM + E + F + M treatment for two weeks, analysed with Student t-test, . Arrows denote Vglut2⁺/NeuN⁺ double-labeled neurons. n = 3-4 donor tissues per group. (L) Representative fluorescence traces of Fluo 4 AM before and after glutamate (10 µM) addition at 60 s from control pericytes and i-neurons derived reprogrammed pericytes. (M) Representative time-lapse fluorescent images of a single i-neuron labeled with Fluo 4 AM (glutamate addition, red). (N) Quantification of the amplitude of spikes in response to glutamate from control pericytes and i-neurons differentiated from reprogrammed pericytes. n=14 cells for control pericytes, and n=24 i-neurons from 3 donor tissues. Scale bar: 100 µm; ** P < 0.01, *** P < 0.001, **** P < 0.0001.