Glioma represents the most prevalent and aggressive primary intracranial malignancy in adults, characterized by infiltrative growth and a dismal prognosis despite a rigorous multimodal standard of care comprising maximal safe resection, radiotherapy, and temozolomide chemotherapy [1], [2]. A fundamental driver of this therapeutic failure is the highly heterogeneous and profoundly immunosuppressive tumor microenvironment (TME) [3], [4]. Within this complex ecosystem, tumor-associated macrophages (TAM) constitute the predominant immune population. While TAM possess the plasticity to exert anti-tumor functions, their polarization toward a specialized, pro-tumorigenic phenotype known as Hypoxia-TAM [5], [6], [7].

Photodynamic therapy (PDT) has shown promise as a locoregional treatment for glioma recurrence by inducing cytotoxicity via reactive oxygen species (ROS) and enhancing anti-tumor immunity [8], [9]. However, the clinical utility of PDT is often compromised by a fundamental physiological paradox: the therapy consumes molecular oxygen to generate ROS, thereby acutely exacerbating intra-tumoral hypoxia. This effect is particularly prominent when the initial tumor oxygenation level is sub-optimal, a frequent condition in gliomas that contributes to PDT resistance [10], [11]. This therapy-induced hypoxic stress acts as a potent evolutionary pressure, selecting for resistant phenotypes and entrenching the immunosuppressive Hypoxia-TAM population. While the enrichment of these cells in recurrent gliomas is well-documented, the precise molecular drivers that orchestrate their metabolic adaptation and sustained immunosuppressive function following PDT [12], [13], [14]. Consequently, there is an urgent need to identify more precise and actionable metabolic nodes within Hypoxia-TAM that can be targeted to disrupt their pro-tumorigenic functions.

To address this critical gap, we employed single-cell RNA sequencing (scRNA-seq) on paired Pre-PDT and recurrent post-PDT (RP-PDT) clinical specimens. Through differential expression analysis and metabolic pathway enrichment, we identified phosphoglycerate kinase 1 (PGK1, a key glycolytic enzyme) as a top candidate gene significantly upregulated in the resistant myeloid population. Building upon this screening result and integrating it with existing literature regarding the non-canonical functions of glycolytic enzymes, we hypothesized that PGK1 serves as a critical metabolic switch, phosphorylating pyruvate dehydrogenase kinase 1 (PDHK1, a regulator of mitochondrial metabolism) to reinforce glycolysis and sustain the Hypoxia-TAM state [15], [16].

Furthermore, we explored the functional impact, showing that PGK1-driven glycolysis promotes secretion of osteopontin (SPP1, a pro-tumorigenic cytokine), which activates CD44 (a cell surface receptor mediating cell-matrix interactions) on glioma cells to facilitate mesenchymal transition and recurrence [17], [18], [19]. By combining the discovery power of single-cell transcriptomics with rigorous experimental validation in orthotopic murine models, this study establishes the PGK1-pPDHK1 (phosphorylated PDHK1) interaction as a novel, druggable target. Ultimately, our findings provide a mechanistic framework for targeting this specific metabolic vulnerability, offering a translational strategy to convert the hypoxic barrier induced by PDT into a therapeutic opportunity to sensitize glioma to treatment.