Whatʼs new in glioma: from molecular insights to clinical innovations

The landscape of glioma research has evolved over the past decade, bridging fundamental neurobiology with clinical applications. From spatial tumor architecture to AI-powered diagnostics and molecular therapies, these advances are redefining precision neuro-oncology and offering new hope for patients facing one of medicineʼs most challenging malignancies.

Introduction

Gliomas represent the most common primary brain malignancy in adults, arising from glial cells and encompassing a spectrum from lower grade tumors to highly aggressive and diffusely infiltrating glioblastoma.¹ Despite maximal treatment with surgery, radiotherapy, and alkylating agent chemotherapy, the latter remains universally fatal with median survival of 15 months despite maximum therapy.^2,3 A recent randomized trial demonstrated that the addition of tumor treating fields to maintenance temozolomide improved median overall survival (OS) to 20,9 months compared to 16,0 months with chemotherapy alone.⁴ Low-grade gliomas often affect younger patients and while initially less aggressive, often progress over years causing significant morbidity. The biological complexity of these tumors, characterized by extensive intratumoral heterogeneity, infiltrative growth patterns, and integration into functional brain tissue as well as the blood-brain barrier has long challenged therapeutic development.⁵

This review highlights some of the breakthrough discoveries in the last years that are reshaping our conceptual understanding and clinical management of glioma. We examine pre-clinical insights into tumor spatial organization, neuron-glioma interactions, and neuroactive vulnerabilities that reveal actionable therapeutic targets. We then discuss artificial intelligence (AI)-powered diagnostic platforms enabling molecular classification with much faster turn-around times than conventional array-based technologies. Finally, we review clinical advances in evidence-based surgical strategies and molecular therapies that are moving glioma care toward precision medicine. Together, these developments mark an inflection point where glioma’s neurobiological origins, long considered a therapeutic liability, are being exploited as vulnerabilities.

Pre-clinical discoveries: spatial architecture, neuron-tumor networks, and immune interactions

Recent technological advances have displayed glioblastoma as a multilayered biological system. Spatial transcriptomics and proteomics revealed a five-layer architectural organization driven by hypoxia, which coordinates long-range spatial patterns across all cancer cell states.⁶ Tumors contain local microenvironments dominated by specific cellular states, with consistent co-localization patterns. Critically, tumors lacking hypoxia, such as lower grade IDH-mutant gliomas, demonstrate significantly less organization, establishing hypoxia as a master organizer of tumor heterogeneity. Beyond spatial architecture, glioblastomas seem to functionally integrate into neural circuits throughout the brain. Retrograde rabies virus tracing revealed widespread neuron-to-glioma synaptic connections, with cholinergic neurons particularly driving invasion.⁷ Interestingly, Tetzlaff et al. showed that radiotherapy enhances neuron-to-glioma connectivity through increased neuronal activity, while combined neuronal inhibition with radiotherapy showed increased therapeutic effects. These findings may indicate a role for neuron-to-glioma connections in therapeutic resistance.

A third biological aspect emerged from the discovery that glioblastoma HLA molecules present bacterial peptides to tumor-infiltrating lymphocytes.⁸ Unbiased antigen discovery approaches demonstrated that CD4+ T cell clones derived from patient tumor-infiltrating lymphocytes recognize peptides from pathogenic bacteria, gut microbiota, and tumor antigens with striking cross-reactivity. This molecular mimicry suggests the microbiome primes anti-tumor immunity, potentially explaining individual variation in immunotherapy responses. Together, these three biological domains – spatial, neural, and microbial – reveal glioblastoma as deeply integrated into its host environment, with vulnerabilities extending beyond the cancer cell itself.

Intraoperative diagnostics: AI-powered rapid tumor classification

Glioma classification has evolved from purely histological assessment to molecular-integrated diagnostics, with the 2021 WHO classification incorporating moleculas markers as essential diagnostic criteria.⁹ While molecular profiling has dramatically improved diagnostic accuracy and opened new therapeutic opportunities – from targeted therapies to precision clinical trials – traditional molecular testing requires days to weeks, limiting its utility during surgery and delaying the initiation of adjuvant therapy. This temporal gap has driven intensive research into rapid intraoperative molecular diagnostics.¹⁰

To date, intraoperative diagnosis relies on histological assessment of frozen tumor sections, which often does not lead to a clear diagnosis. Results frequently have to be revised based on post-operative molecular testing, potentially leading to suboptimal resection strategies. Nanopore sequencing (Oxford Nanopore Technologies) has emerged as an ultrarapid diagnostic method with distinct advantages including low setup cost and instant local data availability. Together with artificial intelligence, nanopore sequencing has started to revolutionize intraoperative CNS tumor diagnosis. Sturgeon, a transfer-learned neural network, was recently introduced and enables molecular CNS tumor classification during surgery using nanopore sequencing of methylation profiles. Results of a prospective real-time application across 25 surgeries were published in Nature in 2023 and showed that accurate diagnosis was achieved in 72% of cases within 90 minutes.¹¹ This molecular subclassification during initial surgery could enable molecularly-informed resection strategies in the future in routine clinical practice. Building on this foundation, Deacon et al. demonstrated that integrating three methylation classifiers with nanopore technology could provide both rapid intraoperative classification and comprehensive next-day molecular profiling – including detection of single nucleotide variants, copy number variants, and structural variants. In 50 prospective cases, their approach displayed a 90% concordance with final integrated diagnosis while maintaining the speed necessary for surgical decision-making.¹⁰

Complementing molecular sequencing, stimulated Raman histology (SRH) is a label-free optical imaging technique that uses laser light to provide H&E-like images of unprocessed tissue without staining or prior chemical processing.¹² When combined with deep convolutional neural networks, SRH provides rapid intraoperative diagnosis. Trained on over 2,5 million images, this system was able to predict tumor types in under 2,5 minutes – significantly faster than frozen sections which often take around 30 minutes. A multicenter trial (n=278) demonstrated non-inferior accuracy compared to pathologist interpretation (94,6% vs 93,9%).¹³ Together, these AI-powered platforms create diagnostic pathways independent of traditional pathology laboratories, expanding access to rapid molecular diagnosis during surgery.

Clinical therapy: precision treatment strategies

Evidence-based surgical thresholds

The 2016 WHO molecular reclassification created controversy regarding surgical extent in lower grade gliomas, with debates whether aggressive resection benefits all molecular subtypes or primarily astrocytomas, given oligodendrogliomas’ favorable prognosis. A randomized trial remains infeasible due to lack of equipoise among physicians and patients. Hervey-Jumper’s landmark analysis of 757 IDH-mutant grade 2 glioma patients published in 2023 established quantitative extent of resection (EOR) thresholds.¹⁴ Using propensity score matching to simulate randomized trials, it was shown that an EOR ≥75% improved overall survival while an EOR ≥80% improved progression-free survival in both astrocytoma and oligodendroglioma. The authors also identified three risk groups based on molecular subtype, tumor volume, and residual disease.

Building on these findings, a 2025 international multicenter study of 1,391 patients addressed the inconsistent terminology regarding resection outcomes across studies by proposing a standardized RANO classification based on residual T2-FLAIR tumor volume.¹⁵ This four-tier system also addressed the new debate around supramarginal resection in glioma¹⁶ and demonstrated that supramaximal resection beyond T2-FLAIR borders (class 1) yielded superior outcomes with 98% 10-year survival and 83% 5-year progression-free survival, compared to maximal resection (class 2, 82% and 44%), submaximal resection (class 3, 75% and 25%). Notably, survival benefits already emerged after 3 years in patients with astrocytoma but required 6–8 years to manifest in oligodendrogliomas, confirming that both subtypes benefit from extensive resection albeit with different temporal trajectories. These evidence-based classification systems provide objective surgical targets while emphasizing that maximal safe resection with neurological preservation remains paramount.

Molecularly-matched and repurposed therapies

The therapeutic landscape in glioma has recently been transformed by precision medicine approaches targeting specific molecular alterations. The landmark INDIGO phase 3 trial demonstrated vorasidenib’s efficacy in IDH-mutant grade 2 glioma, with median progression-free survival improving from 11,1 to 27,7 months (HR: 0,39; P<0,001) and delayed time to next intervention (HR 0,26).¹⁷ This oral brain-penetrant IDH1/2 inhibitor represents the first FDA-approved targeted therapy specifically for low-grade glioma. Vorasidenib additional improved seizure control, particularly in the oligodendroglioma group, and showed no negative effects on health-related quality of life or neurocognition compared to the placebo.¹⁸

In line with this single-target success, the phase 1/2 NCT Neuro Master Match (N2M2) umbrella trial evaluated molecularly-matched therapies in 228 newly diagnosed glioblastoma patients without MGMT promoter methylation, a population with extremely limited treatment options.¹⁹ Trial specific tumor boards stratified patients across subtrials based on matching alterations. Temsirolimus in patients with activated mammalian target of rapamycin (mTOR) kinase signaling achieved superior median overall survival compared to standard-of-care controls, meeting the primary endpoint. While atezolizumab (PD-L1 checkpoint inhibitor), asunercept (CD95L inhibitor), and palbociclib (CDK4/6 inhibitor) did not demonstrate a positive impact on survival, these precision medicine approaches are now possible because of the rapid molecular diagnostics described above.

Complementing targeted therapy development, systematic drug repurposing has recently been explored using a high-throughput screening of 132 repurposable neuroactive drugs across 27 patient samples. Using an ex vivo platform on fresh surgical tumor tissue, the authors identified class-diverse compounds with potent anti-glioblastoma activity.²⁰ Interestingly, interpretable machine learning of drug-target networks revealed convergence on tumor suppression through calcium signaling, enabling expanded in silico screening of over 1 million compounds with high patient validation accuracy. The antidepressant vortioxetine emerged as particularly promising, demonstrating calcium-driven pathway induction and synergizing with standard chemotherapy in preclinical models. This actionable framework rooted in glioma’s neural etiology offers rapid clinical translation potential, as many neuroactive compounds are already FDA-approved with established safety profiles.

Conclusion and future directions

The convergence of spatial biology, advancements in long-read sequencing platforms, and precision medicine is reshaping glioma care. Pre-clinical discoveries revealing hypoxia-driven architecture, neuron-tumor networks, and neuroactive vulnerabilities are informing rational therapeutic designs and AI-powered diagnostics enable real-time molecular classification within hours to days after surgery.

Key priorities moving forward involve bringing neuron-tumor network inhibition strategies into clinical practice, refining how microbiome-immune interactions can be leveraged therapeutically, and broadening platform trials that evaluate combination molecular treatments. The success with vorasidenib in IDH-mutant gliomas and the encouraging results with other targeted therapies illustrate an important principle: aligning targeted therapies with specific molecular weaknesses could revolutionize patient outcomes.