Sialyltransferase gene signature as a predictor of prognosis and therapeutic response in kidney renal clear cell carcinoma.

Background: Sialyltransferases are enzymes involved in the addition of sialic acid to glycoproteins and glycolipids, influencing various physiological and pathological processes. The expression and function of sialyltransferases in tumors, particularly in kidney renal clear cell carcinoma (KIRC) remained underexplored. This study aimed to develop a prognostic model based on sialyltransferase-related genes (SRGs) to predict the prognosis and treatment response of patients with KIRC.

Methods: We utilized RNA-Seq data of KIRC from The Cancer Genome Atlas (TCGA) database, selecting samples with survival data and clinical outcomes. Somatic mutation and neoantigen data were analyzed using the "maftools" package, and genes involved in the sialylation process were identified through the Molecular Signatures Database. Validation cohorts of KIRC samples were obtained from the International Cancer Genome Consortium (ICGC) database. Single-cell RNA sequencing (scRNA-seq) data were downloaded from the Gene Expression Omnibus (GEO) platform, and preprocessing, normalization, and dimensionality reduction analyses were conducted using the "Seurat" package. Differentially expressed sialylation genes were identified using the "limma" package, and their functional enrichment was assessed via Gene Ontology GO and KEGG analyses. Consensus clustering analysis was performed to identify molecular subtypes of KIRC based on sialylation, and drug sensitivity of different subtypes was evaluated using the "pRRophetic" package. A risk signature model comprising 5 SRGs was constructed through univariate and multivariate Cox regression analyses and validated in both the TCGA and ICGC cohorts. The "estimate" package was utilized to calculate immune and stromal scores for each KIRC sample, assessing the tumor immune microenvironment characteristics of different subtypes.

Results: Analysis of scRNA-seq data identified 25 cell subtypes, categorized into 9 cell types. CD4 + memory cells exhibited the highest potential interactions with other cell subtypes. We identified 14 differentially expressed sialylation genes and confirmed their enrichment in various biological pathways through GO and KEGG analyses. Consensus clustering analysis based on sialylation identified 2 molecular subtypes: C1 and C2. The C2 subtype demonstrated higher sialylation scores and poorer prognosis. Drug sensitivity analysis indicated that the C1 subtype had better responses to Dasatinib and Lapatinib, whereas the C2 subtype was more sensitive to Epothilone B and Vinorelbine. The risk signature model, constructed with five distinct SRGs, exhibited strong predictive accuracy, as indicated by Area Under the Curve (AUC) values of 0.68, 0.69, and 0.70 for 1-, 3-, and 5-year survival, respectively, across both the TCGA and ICGC validation cohorts. Immune microenvironment analysis revealed that the C1 subtype exhibited higher immune and stromal scores, while the C2 subtype showed significantly enhanced expression of immune checkpoint genes.

Conclusion: This study successfully developed a prognostic model based on SRGs, effectively predicting the prognosis and drug response of KIRC patients. The model demonstrated significant predictive performance and potential clinical application value. Furthermore, the study highlighted the critical role of sialylation in KIRC, offering new insights into its underlying mechanisms in tumor biology. These findings could guide personalized treatment strategies for KIRC patients, emphasizing the importance of sialylation in cancer prognosis and therapy.