Southeastern Brain Tumor Foundation (SBTF) Award - A Single Cell Atlas of Glioblastoma Under Bevacizumab Treatment Reveals Mechanisms of Antiangiogenic Therapy Resistance

Introduction: Anti-angiogenic therapies such as bevacizumab have not improved overall survival in glioblastoma despite tumor hypervascularity. Hypothesized resistance mechanisms include co-option of existing vessels, bone marrow-derived endothelial progenitor recruitment, and vessel intussusception, but transcriptional changes in tumoral endothelium during progression on bevacizumab remain unknown.

Methods: Single-nucleus RNA sequencing of frozen GBM tissue samples of patients with progression on bevacizumab (n=6) undergoing surgery within four weeks of last bevacizumab dose and treatment-naïve GBM patients (n=4).

Results: We analyzed 662 endothelial cells (494 bevacizumab-naïve vs 168 bevacizumab-treated) and 420 mural cells (pericytes and vascular smooth muscle cells) (249 vs 171). CNV analysis showed tumor stem-cell transdifferentiation rates into endothelium (16.1% vs 3.75) and mural cells (19.8% vs 6.25%) decreased after bevacizumab resistance. Bevacizumab-treated endothelial and mural cells downregulated inflammatory genes (endothelial, mural: CCL3, log2FC=-3.73, -2.53; B2M, log2FC=-1.94, -0.95; HLA-A, log2FC=-2.13, -1.54; IL1B, log2FC=-2.8, -1.04) but upregulated long non-coding RNAs (lncRNAs) (NEAT1, log2FC=3.07, 1.44). While treated endothelium increased angiopoietin-2 (log2FC=2.2), mural cells upregulated IL23RA (log2FC=4.93, all p< 0.001). Gene ontology showed enrichment of antigen processing and presentation in naïve (p=4.04e-5) but hypoxia response (p=0.004) in treated endothelium, with further clustering identifying naïve proliferating (GO: mitotic nuclear division, p=2.9e-4) and treated RNA splicing (p=0.017) subpopulations. Naïve mural cells had enriched immune cell activation (p=3.5e-12) but facilitated IL-10/12 production (p=0.012) after treatment. GSEA of 6,871 treatment-naïve and 20,952 bevacizumab-treated tumor cells showed decreased angiogenesis (p=0.0008).

Conclusion : Tumoral endothelial and vascular mural cells in bevacizumab-treated glioblastoma demonstrate decreased inflammation and shift towards fibrosis. Fewer proliferating endothelium and negative enrichment for angiogenesis among bevacizumab-treated tumor cells suggest successful devascularization. Tumor progression in this context may proceed via perivascular invasion, supported by increased levels of vessel co-option marker angiopoietin-2. Future work will explore the role of lncRNAs such as NEAT1 and whether cancer-associated fibroblasts among mural cells may promote fibrosis in bevacizumab-resistant GBM.