Paired Tumor Biopsies Reveal Spatiotemporal Myeloid Remodeling After Local Chemotherapy in Glioblastoma

ABSTRACT Background Convection-enhanced delivery of topotecan enables sustained local chemotherapy for recurrent glioblastoma and was associated with reduced tumor proliferation in our previous phase 1B clinical trial. That trial incorporated a paired pre- and post-treatment biopsy design - rare in glioblastoma clinical research - enabling tissue-anchored assessment of drug effect without reliance on radiographic or survival endpoints, which are notoriously difficult to interpret in this disease. However, the cellular and molecular consequences of local chemotherapy within the treated tumor microenvironment remain incompletely defined. Methods We integrated paired, MRI-localized pre- and post-treatment biopsies from a first-in-human CED-topotecan trial (n=5), leveraging the paired biopsy architecture, in which each patient serves as their own control and post-treatment specimens are spatially annotated relative to the MRI-defined infusion zone, to generate tissue-based evidence of drug effect without requiring large patient numbers. These biopsies were integrated with complementary experimental models, including a time-resolved syngeneic murine glioma CED model, acute patient-derived glioblastoma slice cultures, and in vitro human microglial and glioma systems. Clinical biopsies were analyzed by bulk RNA-seq, cell-type deconvolution, and multiplex immunofluorescence. Murine tumors were analyzed by survival, immunofluorescence, and single-cell RNA-seq; patient-derived slice cultures were profiled by single-cell RNA-seq. Results In paired human biopsies, CED-topotecan induced spatially restricted transcriptional remodeling within the infusion zone, characterized by suppression of proliferative tumor programs and enrichment of inflammatory, interferon, hypoxia, and mesenchymal signatures. Cell-type deconvolution and immunofluorescence linked this response to myeloid remodeling, including enrichment of monocyte-derived tumor-associated macrophage states, increased MARCO-positive myeloid populations, and pH2AX-positive genotoxic stress within Iba1-positive myeloid cells. In the murine CED model, topotecan prolonged survival and reduced tumor cellularity, while also inducing inflammatory and DNA-damage programs in tumor-associated macrophages that evolved by 7-days toward hypoxia, angiogenesis, TGF-β signaling, and mesenchymal/tissue-remodeling programs. Human slice culture and in vitro microglial systems confirmed stress-coupled inflammatory and DNA-damage responses in human myeloid cells. Conclusions Local topotecan delivery produces spatially structured tumor cytotoxicity together with a genotoxic, stress-coupled inflammatory myeloid response that evolves toward mesenchymal macrophage remodeling. By integrating paired clinical biopsies with time-resolved and mechanistic experimental models, this study provides a framework for understanding how local chemotherapy reshapes the glioblastoma microenvironment and for future studies evaluating dose, schedule, treatment duration, and combination strategies. These findings demonstrate that paired, spatially annotated tissue sampling from small, precisely characterized clinical cohorts can yield mechanistic insight that conventional radiographic and survival endpoints cannot provide, and support tissue-based response assessment as the appropriate paradigm for evaluating novel locoregional therapies in glioblastoma.