Investigating the Immune Response to Glioblastoma in Unique Central Nervous System Niches

Abstract

Glioblastoma (GBM) is a highly aggressive and the most common primary tumor of the central nervous system (CNS). Despite recent improvements in anti-PD-1 therapy, there remains to be seen any significant benefits of such therapies in patients with GBM. One significant barrier to developing effective brain tumor therapies is our incomplete understanding of how antigen uptake and presentation are orchestrated within the CNS, a challenge compounded by its unique anatomy, including the dura and choroid plexus. Studies have highlighted the importance of type 1 conventional dendritic cells (cDC1s) in mounting neoantigen-specific immune responses and mediating immune checkpoint blockade benefits in preclinical brain tumor models. Yet despite their established importance, the transcriptional programs and T cell priming functions of cDC1s within the brain tumor microenvironment remain poorly characterized. Similarly, type 2 conventional dendritic cells (cDC2s) have begun to be studied across various tumor types, including their role in fostering immunosuppression following high-dose MHC II-neoantigen vaccination in sarcoma models, but their function in brain tumors has received comparatively less attention. A deeper understanding of both DC subsets therefore represents a largely untapped therapeutic opportunity. Clinically, DC-Vax, the most advanced DC-based therapy employed monocyte-derived DCs pulsed with tumor lysate as a therapeutic modality for patients with GBM. However, they demonstrated only modest efficacy, highlighting the need for a more mechanistic understanding of how endogenous DCs function within the tumor microenvironment. Beyond tumor-infiltrating DC biology, there is an equally nascent appreciation of how CNS-specific anatomy shapes the broader anti-tumor immune responses. The dura mater, long regarded as a passive structural barrier, has recently been recognized as an immunologically active compartment. The discovery of meningeal lymphatic vessels and dural-associated lymphoid tissues (DALT) has shown the dura can serve as a site of immune activation in infectious and autoimmune settings, yet its contribution to anti-tumor immunity remains largely unexplored. Modulating the dural immune compartment may therefore represent a novel and underappreciated avenue for GBM therapeutic intervention. In this thesis, we aimed at addressing both gaps. I characterized the functional and transcriptional properties of cDCs in both human and murine brain tumors and investigated the immunological remodeling of the dura in response to intracranial tumor burden in preclinical models. In my first aim, I addressed two central questions, 1) whether tumor-infiltrating dendritic cells that have acquired tumor-derived antigen retain the functional capacity to prime tumor-infiltrating lymphocytes (TILs) despite chronic exposure to the highly immunosuppressive GBM microenvironment, and 2) whether antigen-loaded DCs harbor distinct transcriptional signatures relative to their antigen-negative counterparts. To trace tumor antigen acquisition by infiltrating APCs, we utilized a "transfer of fluorescence" strategy in which fluorescent proteins serve as a surrogate marker of tumor-derived materials. In patient samples, we leveraged 5-aminolevulinic acid (5-ALA), which is selectively converted to fluorescent protoporphyrin IX (PPIX) within tumor cells and often used in tumor resection surgeries to trace tumor margins. Using flow cytometry, we demonstrated that tumor-infiltrating DCs, but not TILs or peripheral blood DCs, accumulate detectable PPIX signal, indicating preferential phagocytosis of tumor-derived material by DCs. We then assessed whether these cells could directly prime autologous TILs ex vivo, measuring T cell activation by IFN-γ ELISpot assay. To complement our human data, we used the CT2A-zsGreen-minOVA preclinical model, in which zsGreen fluorescent protein is fused to a minOVA construct carrying the SIINFEKL epitope. This system enabled simultaneous tracking of tumor antigen uptake by cDCs and functional assessment of antigen cross-presentation through OT-I T cell priming. Notably, in both the human and preclinical setting, tumor-infiltrating DCs successfully acquired tumor-derived materials, PPIX and zsGreen, respectively and were able to activate T cells ex vivo. These findings represent among the first direct evidence that tumor-infiltrating cDCs retain sufficient antigen load and cross-presentation capacity to activate T cells. Finally, single cell sequencing of patient tumor samples revealed that antigen uptake by tumor-infiltrating APCs is associated with enrichment of specific gene expression programs, notably IFN- γ response pathways and oxidative phosphorylation. Furthermore, comparison of tumor-infiltrating APCs and peripheral APCs show an enrichment of MHC class I and II signatures in addition to cell motility and various immune-related pathways in the former. In my second aim, I investigated the role of cDC2s in GBM using the Δ1+2+3 mouse strain, which lacks both cDC2s and monocytes. Strikingly, Δ1+2+3 mice survived significantly longer than wild-type B6 and cDC1-deficient Δ32 mice in both GL261 and CT2A tumor models, pointing to a tumor-promoting rather than protective role of cDC2s and/or monocytes in the preclinical GBM setting. In GL261, this survival benefit was accompanied by a remodeled myeloid compartment with fewer tumor-associated macrophages (TAMs), and a compensatory increase in cross-presenting cDC1s within the tumor. Furthermore, there was an observed increase in CD8+ T cell infiltration that were differentiated, activated, and functional, with increased expression of granzyme B, TNFα, and IFN- γ. In CT2A, survival was similarly extended despite no clear increase in T cell infiltration. Hence, a larger cohort with comprehensive phenotypic and functional analysis will be needed to fully characterize the antitumor T response in this model. Together, these findings suggest that cDC2s and monocytes impede effective antitumor immunity in GBM, and that their absence shifts the tumor microenvironment toward a more pro-inflammatory state with increased cDC1 and CD8+ T cell infiltration. Finally, in my third aim, I established the dura mater as an immunologically active site in preclinical GBM models. Compared to sham control, the dura of tumor-bearing mice showed greater immune infiltration, detectable tumor-derived materials, and tumor-specific T cells, directly implicating the dura in mediating immune response to intracranial tumors. Dural CD8⁺ T cells exhibited elevated CD44, PD-1, and granzyme B expression relative to sham controls, with an activation state comparable to TIL and markedly higher than that observed in the ipsilateral superficial cervical lymph nodes (iSCLN). Importantly, FACS-sorted dural cDCs from CT2A-zsGreen-minOVA tumor-bearing mice primed naïve OT-I T cells more potently than splenic cDCs ex vivo, providing the first direct evidence that dural cDCs carry sufficient antigen to activate T cells without additional stimulation. Consistent with this priming capacity, SIINFEKL-specific CD8⁺ T cells accumulated in the dura at frequencies comparable to the tumor and far exceeding the iSCLN, displaying activated, granzyme B⁺ cytotoxic profiles. Adoptive transfer of cell trace violet-labeled OT-I T cells showed that proliferated tumor-specific T cells accumulate within the dura. Finally, micro-dissection of distinct dural regions revealed spatially heterogeneous immune infiltration, with the DALT and superior sagittal sinus showing the greatest tumor-driven enrichment of T cells and myeloid cells. Together, these findings position the dura as a spatially organized, previously underappreciated site of antigen presentation in preclinical GBM models. Ultimately, these findings have direct relevance to how we think about CNS immune surveillance and the design of future immunotherapeutic strategies for patients.