Clear cell renal cell carcinoma (ccRCC) is the most common subtype of kidney cancer, accounting for approximately 70% of cases. Despite improvements in early detection and localized treatment, about 17% of patients present with distant metastasis at diagnosis and an additional 30% of initially localized cases eventually progress, resulting in poor outcomes. Substantial tumor heterogeneity limits the benefit most patients receive from precision oncology approaches.
Tryptophan (Trp) is an essential amino acid that serves as a precursor to multiple signaling metabolites. Of its three major degradation pathways, the kynurenine (Kyn) pathway is dominant, processing over 95% of free tryptophan. In RCC, the kynurenine pathway is upregulated while alternative tryptophan degradation enzymes are downregulated, producing a characteristic metabolic signature. Kynurenine pathway metabolites suppress antitumor immune responses, promote cancer cell proliferation, and contribute to immune evasion.
Despite documented tryptophan metabolism disruption in RCC, no validated prognostic signature based on tryptophan metabolism had been developed for ccRCC. The study aimed to fill this gap by integrating single-cell RNA sequencing, bulk transcriptomics, multi-omics data, and a comprehensive machine learning framework to build and validate a tryptophan metabolism-related prognostic signature.
Single-cell RNA sequencing data from 45,427 cells across 12 ccRCC and adjacent normal tissue samples were analyzed using the Seurat pipeline. Tryptophan metabolism enrichment scores computed with the AUCell algorithm showed a progressive decline across normal, transitional, and malignant proximal tubule cells, with pseudotime analysis confirming that tryptophan metabolism decreases as cells acquire malignant characteristics during tumor progression.
hdWGCNA (high-dimensional weighted correlation network analysis) applied to single-cell data and WGCNA applied to bulk TCGA-KIRC data independently identified tryptophan metabolism-correlated gene modules. The intersection of both analyses yielded 142 overlapping tryptophan metabolism-related genes, which were then filtered to 35 candidates with significant prognostic associations across TCGA, CPTAC, and E-MTAB-1980 datasets by univariate Cox regression.
From these 35 candidates, 246 combinations of 10 machine learning algorithms were systematically evaluated on the TCGA training set. The combination of backward stepwise Cox regression and generalized boosted regression modeling (GBM) achieved the highest mean concordance index (C-index of 0.758) across three validation datasets, yielding a final 11-gene tryptophan metabolism-related signature (TMRS) comprising AGMAT, BBOX1, DDAH1, G6PC, SLC13A1, ALDH6A1, EPHX2, SMIM24, AQP1, ACAA2, and LGALS2.
The TMRS demonstrated strong prognostic performance across all validation datasets. In the TCGA dataset, 1-, 3-, and 5-year AUCs for overall survival prediction were 0.87, 0.85, and 0.83 respectively. In the independent E-MTAB-1980 dataset, AUCs were 0.83, 0.86, and 0.83, demonstrating reproducible performance. The CPTAC dataset showed lower 5-year AUC (0.59), attributed to its short follow-up duration with only 9.2% of patients monitored beyond five years.
High TMRS risk scores correlated significantly with advanced tumor stage and higher histological grade across all three datasets. Multivariate Cox regression confirmed that TMRS retained independent prognostic value after adjusting for age, grade, and disease stage. Subgroup analyses showed TMRS was especially effective for early-stage ccRCC patients, where conventional staging provides less prognostic discrimination.
Nomograms integrating the TMRS risk score with clinical characteristics achieved Harrell's C-index scores approaching 0.8 for overall survival prediction across all datasets, with calibration curves confirming close agreement between predicted and observed survival probabilities. Decision curve analysis further supported the clinical utility of the nomogram over standard staging-based models.
Gene set variation analysis on Hallmark signatures revealed that high TMRS scores correlated positively with malignancy-related pathways including epithelial-mesenchymal transition, G2M checkpoint activation, and E2F target expression. Metabolic pathways including fatty acid metabolism and oxidative phosphorylation showed negative correlation with high TMRS scores, consistent with the metabolic shift toward glycolysis and immune-tolerant states seen in aggressive ccRCC.
Immune microenvironment analysis using ESTIMATE, CIBERSORT, and ssGSEA showed that high-TMRS tumors exhibited an inflamed but dysfunctional immune phenotype, with elevated immune cell infiltration paired with markers of immune exhaustion and dysfunction. This pattern is consistent with the known immunosuppressive effects of kynurenine pathway metabolites, which accumulate when tryptophan degradation is dysregulated.
High TMRS tumors showed elevated tumor mutation burden and higher mutant-allele tumor heterogeneity scores, suggesting greater genomic instability. Patients with high TMRS scores also showed predicted resistance to immunotherapy based on the TIDE algorithm and differential predicted sensitivity to certain targeted therapies, opening the possibility of using the signature to stratify patients for specific treatment regimens.
All 11 TMRS genes showed reduced expression in ccRCC tumor samples compared to adjacent normal tissues at both transcript and protein levels, confirmed across three transcriptomic datasets, CPTAC proteomics, and immunohistochemical images from the Human Protein Atlas. Most signature genes also correlated significantly with disease stage and histological grade, supporting their biological relevance to tumor progression.
DDAH1 (dimethylarginine dimethylaminohydrolase 1) was selected for experimental validation. In ccRCC tissue specimens, DDAH1 expression was lower in tumor compared to normal adjacent tissue, confirmed by quantitative RT-PCR and Western blot. siRNA-mediated knockdown of DDAH1 in 786-O and Caki-1 ccRCC cell lines enhanced proliferation (measured by CCK-8 assay) and increased metastatic migration and invasion (measured by transwell assay), establishing a functional role for DDAH1 loss in promoting ccRCC aggressiveness.
Tryptophan and kynurenine levels were directly measured in DDAH1-knockdown cell supernatants, confirming that DDAH1 loss is associated with altered tryptophan metabolism at the cellular level. These experiments link the bioinformatic signature to a specific biological mechanism and validate DDAH1 as a candidate tumor suppressor gene in ccRCC whose downregulation promotes tumor progression.
The study is limited by the retrospective nature of all datasets used and the absence of prospective clinical trial data to validate TMRS-based treatment stratification. The relatively low 5-year AUC in the CPTAC cohort reflects limited follow-up rather than model failure, but longer follow-up data from proteomic datasets would be valuable for confirming long-term predictive accuracy.
The experimental validation was performed in two established cell lines and clinical specimens from a single hospital. Broader validation across patient-derived organoids, in vivo mouse models, and multi-institutional tissue cohorts would be necessary to confirm the clinical role of DDAH1 and other TMRS genes before they could be considered therapeutic targets.
The TMRS framework could potentially be extended to investigate tryptophan metabolism in other RCC subtypes and to identify pharmacological strategies that target the kynurenine pathway. Given the established role of IDO1 and TDO inhibitors as immunotherapy adjuvants in other cancer types, the TMRS signature may help predict which ccRCC patients would most benefit from such metabolic interventions.