Abstract
Brainstem gliomas (BSG) is a highly malignant central nervous system childhood tumors with 5-year survival rate <10%. Metabolism during radiotherapy is a dynamic and precisely programmed process, improving clinical outcomes and guiding therapy decisions of BSG. Here we construct diagnostic and prognostic assays of BSG via circulating metabolites based on both cross-sectional study and longitudinal cohort study with 106 BSG patients. We employ nanoparticle enhanced laser desorption/ionization mass spectrometry to characterize static and dynamic snapshots of metabolites during BSG radiotherapy. We show that this serological tool reaches the area under the curve of 0.933 for BSG diagnosis in an independent blind test and predicts risk of patients with significant differences (p < 0.05) in prognostic outcomes. We further identify eight distinct temporal patterns of metabolite regulation associated with radiotherapy responses and tracked the metabolic trajectory via dynamic metabolic snapshots throughout radiotherapy process. If further validated, this framework could be extended to derive comprehensive metabolic pictures for cancers including but not limited to BSG.
Introduction
Brainstem gliomas (BSG) represent a significant clinical challenge in pediatric oncology. The brainstem's critical role in regulating vital functions makes traditional tissue biopsies extremely hazardous. Furthermore, current gold-standard imaging techniques like MRI often lack the sensitivity (65–73% accuracy) to detect minimal lesions or provide deep molecular insights. While liquid biopsies targeting circulating tumor DNA (ctDNA) have shown promise in other cancers, the blood-brain barrier (BBB) often prevents these large molecules from entering the systemic circulation.
Small metabolites, however, can traverse the BBB more readily, particularly when the barrier is compromised by tumor growth. This study explores the use of serum metabolic profiling as a non-invasive "liquid biopsy" to overcome these diagnostic and monitoring hurdles.
Methodology
The researchers developed a high-throughput platform using Nanoparticle-Enhanced Laser Desorption/Ionization Mass Spectrometry (NPELDI-MS). Unlike conventional mass spectrometry, which can be slow and requires significant sample volumes, NPELDI-MS utilizes ferric nanoparticles to enrich metabolites, allowing for the analysis of just 0.1 µL of serum in approximately 2 seconds per sample.
The workflow involves capturing "metabolic snapshots"—static snapshots for initial diagnosis and prognosis, and dynamic snapshots through longitudinal sampling to monitor treatment response. Machine learning algorithms, including LASSO and Random Forest, were employed to identify specific metabolite panels from a pool of over 400 features.
Study Design and Datasets
The study utilized a large cohort of BSG patients and healthy donors (HDs) collected between 2017 and 2023. The discovery and validation phases involved 174 BSG patients and 125 HDs, followed by an external validation with 161 independent samples.
A longitudinal cohort of 62 patients provided serum samples at three critical time points: before, during, and after radiotherapy. To ensure specificity, the metabolic signatures were also tested against patients with other neurological conditions like encephalitis and epilepsy.
Key Insights and Results
Diagnostic and Prognostic Performance
The optimized diagnostic model achieved an AUC of 0.933 in independent blind tests, significantly outperforming traditional MRI and ctDNA-based methods. The prognostic panel successfully stratified patients into high-risk and low-risk groups, with the low-risk group showing significantly prolonged median overall survival (p < 0.05).
Metabolic Rewiring and Treatment Monitoring
The identified biomarkers, such as lactic acid, valine, and threonine, are closely linked to the "Warburg effect" and specific amino acid pathways utilized by tumors for growth. The longitudinal analysis revealed eight distinct temporal patterns of metabolite regulation during radiotherapy. This suggests that metabolic response is a programmed process that can be tracked to provide a molecular layer of information beyond what is visible on MRI. Notably, the metabolic signature distinguished between volume-decreasing and volume-increasing tumors with an AUC of 0.818.
Clinical Scalability
The NPELDI-MS platform demonstrated high throughput, processing 384 samples in roughly 3 hours. Biological validation through RNA sequencing and in vitro studies confirmed that the identified metabolites directly influence tumor cell proliferation.
Discussion and Future Directions
While promising, the study's primary limitation is its focus on a single ethnic population (Chinese). Future work must validate these metabolic signatures across diverse global cohorts. Additionally, longer-term follow-up is required to confirm the durability of prognostic markers. The authors envision the eventual transition of this technology to miniaturized mass spectrometers or dried blood spot testing, which would enhance accessibility in low-resource clinical settings. This framework provides a foundation for comprehensive metabolic monitoring that could be adapted for a wide range of cancers.
Summary updated