Precise description of metabolomic states using NGS uncover new potential biomarkers of response to TKIs in ccRCC

Renal cell carcinomas (RCC) undergo extensive metabolic reprogramming, which support tumor progression and therapy resistance. Understanding these changes is essential for identifying resistance mechanisms and therapeutic targets. While RNA sequencing enables estimation of metabolic changes, current gene expression-based metabolomic signatures often lack specificity and include unrelated or conflicting genes. Here, we combine transcriptomic and metabolomic data to create refined metabolic signatures for glycolysis (Glyc), the kynurenine pathway of tryptophan catabolism (Trp), and the urea cycle (UC), which may inform prediction of tyrosine kinase inhibitors (TKIs) response in patients with clear cell RCC (ccRCC). To develop metabolic signatures, an initial gene pool was assembled by merging unique genes from existing metabolomic signatures in MSigDB (v2024.1) database. This gene list was refined using part of BostonGene ccRCC metacohort (701 samples), by filtering them based on technical and biological criteria: (i) median expression ≥ 2 TPM; (ii) positive Spearman cross-correlation among genes in the signature; and (iii) confirmed biological relevance through correlation with paired metabolomic and NGS data, retaining only genes with r > 0.2 to target metabolites (like L-lactic acid, kynurenine, or urea). Clinical significance of signatures, focusing on associations with TKI response, was assessed using data from the metacohort with available therapy response (853 samples). Overall, TME and survival analysis was assessed on the whole metacohort (n = 4,583). All metabolic signatures demonstrated statistically significant differences between tumor and normal samples (P < 0.001), with higher scores for Glyc and Trp signatures in tumors and lower scores for the UC signature, consistent with previously reported findings. Notably, our signatures showed the strongest differentiation between tumor and normal samples compared to publicly available signatures. The metabolic signatures were also significantly associated with tumor microenvironment (TME) subtypes (P < 0.001) (Bagaev et al., 2021, Cancer Cell ). Higher Trp signature scores were observed in immune-enriched subtype, whereas lower Trp and higher UC signature scores were found in the immune-depleted subtype. Survival analysis revealed that low UC scores correlated with worse overall survival (P < 0.001, MW U-test). In TKI-treated patients, complete responders exhibited higher Trp scores, while those with progressive disease had lower UC scores, which were associated with significantly poorer overall survival. We developed metabolism-based gene signatures and demonstrated that the Trp and UC signatures strongly correlate with TKI response and patient survival. The UC and Trp pathways represent novel candidate biomarkers for future patient stratification aimed at overcoming TKI resistance in RCC.