Mesenchymal Stem Cells and Fibroblasts Contribute to Microvascular Proliferation in Glioblastoma and are Correlated with Immunosuppression and Poor Outcome

Abstract

Microvascular proliferation (MVP) is a disease-defining hallmark of glioblastoma and other World Health Organization grade 4 gliomas. MVP also serves as a poor prognostic marker in various solid tumors. Despite its clinical significance, the mechanisms and biological consequences of MVP are controversial and remain unclear. In this study, we performed single-cell RNA sequencing on paired CD45-CD105+ vascular/perivascular stromal cells (PVSC) and CD45+CD105± immune cells from 16 primary glioma patient samples, both with and without MVP. This analysis revealed the presence of developmentally related mesenchymal stem cells alongside cancer-associated fibroblasts, pericytes, fibromyocytes, and smooth muscle cells within the CD45-CD105+ compartment. RNA velocity analysis identified PDGFRB as a putative driver gene guiding mesenchymal stem cells toward more mature PVSCs in the context of MVP. Signaling network analysis and digital spatial profiling uncovered interactions between PDGFRB+ PVSCs and immunosuppressive myeloid cell subsets enriched in the perivascular niche, suggesting targetable receptor-ligand interactions. Additionally, a gene signature of MVP-associated PVSCs from gliomas predicted worse prognosis in multiple other solid tumors. This study provides a transcriptomic cell atlas of PVSCs and immune cells in glioma, helping to refine the biological model of MVP which has traditionally focused on endothelial cells.

Fig. 1.. PDGFRβ⁺ PVSCs are a cellular component of MVP.

a. Experimental schematic of human tissue analyses. scRNA-seq was performed on 16 human glioma specimens. Libraries of CD45⁻CD105⁺ and CD45⁺ cells were generated separately, then bioinformatically integrated. Digital spatial profiling and multiplex immunofluorescence was performed on representative patient samples. b. UMAP visualization and annotation of cell clusters. Two MSC populations were identified, one of which clustered with fibroblastic cells and pericytes. MSC markers (NT5E, ENG, THY1, and PDGFRA) are highlighted in yellow, while PDGFRB is highlighted in red. T_naive/CM = naïve and central memory T cells; HLA-DRA_lo = HLA-DRA-low; IFN = interferon; T_reg = regulatory T cells; and cDC2 = classical type 2 dendritic cells. c. Stacked bar plot showing the proportion of cells contributed by individual patient tumors for each cell population in (b). d. Dot plot of the signature genes of PVSCs derived from the top 50 differentially expressed markers in PVSCs compared to all other cells. Note PDGFRB is one of the most specific and sensitive standalone markers. e. Representative H & E and multiplex immunofluorescence images of the perivascular niche in gliomas with (n = 3) and without (n = 4) MVP. Classical MSC markers (CD73, CD90, CD105), a pan-microglia/monocyte/macrophage marker (AIF1), an endothelial cell marker (CD31), a glioma stem cell/oligodendrocyte/oligodendrocyte precursor marker (OLIG2), as well as PDGFRβ, a marker derived from (a), were used to characterize the perivascular niche. The morphology of MVP corresponded mostly with PDGFRβ, not CD31. PDGFRβ was present, but not expanded morphologically in gliomas without MVP (Supplementary Fig. 5). An arrow indicates a PDGFRβ⁺CD73⁺CD90⁺CD105⁺ cell. A PDGFRβ⁻CD73⁺CD90⁺CD105⁺ cell is shown in Supplementary Fig. 5. f. Automated quantitative analysis of classical MSC markers (CD73, CD90, and CD105) and PDGFRβ in regions with MVP, regions without MVP, and regions adjacent to MVP that were normal-appearing was performed. Two populations of MSCs (CD73⁺CD90⁺CD105⁺) can be found in the perivascular niche, but only those also expressing PDGFRβ are associated with MVP (p = 0.009**, unpaired t-test with Welch’s correction).

Figure 2:. PDGFRB is a marker and potential driver gene of PVSCs.

a. UMAP generated from re-clustering the MSC cluster and contiguous cell clusters. b. Velocyto(37) Dynamical modeling. c. Velocyto(37) Latent time analysis. d. Scatter plot of spliced and unspliced mRNA in cells. PDGFRB is a putative driver gene that is induced transcriptionally during the MSC, fibroblastic subtype stage (green, as corresponds to the original UMAP in (a)), through the intermediate fibroblastic, pericytic and smooth muscle cell stage (red and purple) until it is downregulated in the fibroblastic, pericytic and fibromyocyte cell stage (blue). e. UMAP visualization of PDGFRB velocity. PDGFRB velocity is highest in the MSC, fibroblastic subtype cluster. f. UMAP visualization of PDGFRB expression. g. Schematic of proposed differentiation and potential de-differentation of MSCs into different perivascular cell states.

Fig. 3:. MVP is associated with changes in immune cell infiltration and immunosuppression.

a.UMAP visualization of the cellular composition of gliomas with (33,295 cells) and without MVP (42,846 cells). b. Stacked bar plot showing the proportion of cells contributed by “MVP” and “No MVP” patient tumors for each cell population in a. The dashed lines represent the 25.0%/75.0% marks. c. Top enriched biological processes in PVSCs in gliomas with and without MVP which include extracellular matrix/structure organization, external encapsulating structure organization and negative regulation of immune system processes. Legend to the right of the figure. d. Signal transduction network of brain-resident perivascular macrophages associated with MVP. Blue arrows represent receptors of brain-resident perivascular macrophages, red arrows represent ligands of PVSCs. The “edge cost” represents the probability of a functional interaction between two genes and the “node prize” represents cell-specific gene activity. Legend below. e. Signal transduction network of HLA-DRA_lo suppressive monocytes associated with MVP. Blue arrows represent receptors of brain-resident perivascular macrophages, red arrows represent ligands of PVSCs. The “edge cost” represents the probability of a functional interaction between two genes and the “node prize” represents cell-specific gene activity.

Fig. 4.. Immunosuppressive myeloid cell subsets are upregulated in the perivascular niches associated with MVP, along with PDGF/TGF-β signaling and extracellular matrix changes.

a. Representative example of CD45, PDGFRβ, and CD45⁻PDGFRβ⁻ (other) masks for the digital spatial profiling of a MVP-associated perivascular niche. CD31⁺ areas were subtracted from the PDGFRβ mask to maximize the accuracy of enrichment. b. The immune-related proteins of the CD45 mask show an upregulation of CD14 (associated with monocytes and myeloid-derived suppressor-like cells) and CD163 (associated with perivascular macrophages) along with immunosuppressive markers such as ARG1 and B7-H3. Immunostimulatory proteins such as STING, ICOS, 4–1BB, and GZMB are also upregulated. A log₂ fold change of 0.6 and adjusted p value of 0.05 were used as cutoffs. c. The pathway analysis obtained using digital spatial profiling shows upregulation of PDGF signaling, the TGF-β, matrix reorganization and collagen-associated axes in gliomas with MVP compared to those without. Top differentially downregulated processes are also shown. An adjusted p value cutoff of 0.05 was used.

Fig. 5.. The signature of MVP-associated PVSCs correlates with overall survival in LGG as well as other solid tumors.

a. Volcano plot showing the top differentially upregulated and downregulated genes in PVSCs with and without MVP. b. Expression levels of the top 5 differentially expressed genes in the TCGA-GBM, -LGG, and GTEx databases obtained using GEPIA2. Generally, COL3A1, PCOLCE, TIMP1, NNMT, and COL1A1 are more highly expressed in GBM than LGG, and the least expressed in normal tissues. c. The gene signature created from the top five differentially expressed genes in a. showed a significant difference in overall survival in the TCGA-LGG cohort (p = 3.9×10⁻⁹, log-rank test). d. The gene signature created from the top five differentially expressed genes in a. showed a significant difference in disease-free survival in the TCGA-LGG cohort (p = 0.00029, log-rank test). e. The gene signature created from the top 5 differentially expressed genes in a. showed a significant difference in overall survival in the TCGA-clear cell renal cell carcinoma cohort (p = 0.0095, log-rank test). f. The gene signature created from the top 5 differentially expressed genes in a. showed a significant difference in overall survival in the TCGA-muscle-invasive bladder cancer cohort (p = 0.044, log-rank test). g. The gene signature created from the top 5 differentially expressed genes in a. showed a significant difference in overall survival in the TCGA-mesothelioma cohort (p = 0.033). h. The gene signature created from the top 5 differentially expressed genes in a. showed a significant difference in overall survival in the TCGA-squamous cell carcinoma of the lung cohort (p = 0.043, log-rank test). i. The gene signature created from the top 5 differentially expressed genes in a. showed a significant difference in overall survival in the TCGA-uveal melanoma cohort (p = 0.0029, log-rank test).