One of cancer's defining features is metabolic reprogramming - the way tumor cells change how they generate energy and process nutrients to support rapid growth, invasion, and survival in hostile environments. In clear cell renal cell carcinoma (ccRCC), the most common form of kidney cancer, metabolic changes are particularly prominent. Most ccRCC tumors have mutations in the VHL gene, which triggers a cascade of changes that mimic low-oxygen conditions and dramatically alter metabolism.
These metabolic changes affect not just the tumor cells themselves but also the surrounding immune environment. Abnormal lactate metabolism and mitochondrial dysfunction can impair immune cell function, making it harder for the body's defenses to attack the tumor and reducing the effectiveness of immunotherapy.
This study aimed to identify specific metabolism-related genes in ccRCC whose expression patterns could predict patient survival - creating what the researchers called a Metabolism-Related Prognostic Signature (MRPS).
The researchers began with single-cell RNA sequencing (scRNA-seq) of 26 ccRCC samples, analyzing over 81,000 individual cells. This allowed them to identify a specific population of tumor cells they called META_active cells - malignant kidney cancer cells with notably higher metabolic activity than other tumor cells, characterized by active glycolysis, amino acid metabolism, glutathione metabolism, and other metabolic pathways.
Using a technique called BayesPrism deconvolution, they estimated how much of each patient's bulk tumor sample consisted of these META_active cells. Patients with higher proportions of META_active cells had worse survival, establishing this metabolically active subpopulation as clinically relevant.
To build the MRPS, they tested 118 combinations of 10 different machine learning algorithms - including Random Survival Forest, LASSO, GBM, and others - to find the most accurate and stable combination. The best model used RSF for feature selection and GBM for prediction, identifying 17 metabolism-related genes that collectively predict survival across multiple patient cohorts.
The MRPS model was validated across four patient cohorts - the TCGA training cohort, TCGA testing cohort, and two independent external cohorts (E-MTAB-1980 and Meta). In the training cohort, the model's 1-, 3-, and 5-year survival prediction AUC values were 0.874, 0.808, and 0.823 respectively - strong predictive performance.
When compared to 51 previously published prognostic signatures for ccRCC, MRPS achieved the highest C-index across most cohorts. This is a meaningful benchmark: many existing signatures work well in their discovery dataset but fail to generalize. MRPS showed more consistent performance across diverse patient populations.
High MRPS risk scores were associated with more advanced tumor stage, metastatic disease, and higher tumor grade. The MRPS also predicted immunotherapy response: patients in the low-risk group had significantly better survival when treated with nivolumab (anti-PD-1) immunotherapy, suggesting the signature could help identify who will benefit from immunotherapy.
Among the 17 genes in MRPS, the researchers focused special attention on GGT6 (Gamma-Glutamyltransferase 6), a gene involved in glutathione metabolism that had not previously been studied in ccRCC. GGT6 was expressed at lower levels in tumors compared to normal kidney tissue, and patients with higher GGT6 expression had better survival.
In laboratory experiments with ccRCC cell lines, when GGT6 was knocked down (silenced using siRNA), the cancer cells proliferated faster and invaded more aggressively. This suggests GGT6 normally acts as a brake on tumor aggressiveness - when its expression is lost, the cancer becomes more dangerous.
These wet-lab experiments transformed GGT6 from a computational finding into a biologically validated target, providing mechanistic evidence for why GGT6 expression matters for patient outcomes.
The MRPS has several potential clinical applications. As a prognostic tool, it could help identify high-risk patients who need more intensive surveillance after surgery or who should be prioritized for adjuvant therapy clinical trials. As a predictive biomarker for immunotherapy, it could help oncologists decide whether nivolumab or similar drugs are likely to benefit a specific patient.
The researchers also developed a dynamic nomogram - an interactive tool that combines MRPS with standard clinical factors like tumor stage and grade - to generate individualized survival probability estimates for each patient. This kind of integrated risk calculator is the type of clinical decision support tool that could realistically be used in oncology consultations.
Drug sensitivity analysis using the MRPS identified several candidate therapeutic agents that appeared to be more effective in high-risk patients, including sepantronium bromide and niclosamide, which have been tested in clinical trials for other cancers and could potentially be repurposed for high-risk ccRCC.
MRPS represents a rigorous, methodologically strong approach to building a prognostic signature: it used single-cell analysis to identify biologically meaningful cell populations, validated findings across four cohorts, benchmarked against 51 competitor signatures, and confirmed key findings with laboratory experiments. This multi-layered validation gives more confidence in the results than many published signatures.
The broader message is that metabolic reprogramming is not just a hallmark of ccRCC biology - it is a measurable predictor of outcomes and treatment response. Understanding and targeting tumor metabolism may offer new avenues for improving care for the many patients whose cancer does not respond adequately to current treatments.