Use of integrated multi-omics analysis to analyze prognostic molecular subtype in collecting duct carcinoma.

Abstract

585 Background: Collecting duct carcinoma (CDC) originates from the principal cells of the collecting ducts in the renal medulla and accounts for less than 2% of all renal cell carcinomas (RCCs). Due to its low incidence, the transcriptomic features and molecular subtypes of CDC have only been described in small-sample studies. This study employs a multi-omics approach to investigate CDC, aiming to deepen our understanding of CDC and provide some references for future treatment options. Methods: A total of 103 tumor samples and 15 paired adjacent non-tumor samples underwent whole transcriptome sequencing. Clinical pathological characteristics, treatments, and prognostic information were collected for all patients. Molecular typing of CDC was performed using Non-negative Matrix Factorization. The clinical and biological characteristics of each subtype were analyzed. Tissues from different NMF classifications were collected for single-cell sequencing to compare the biological characteristics. To enhance clinical accessibility, plasma was analyzed by LC-MS/MS and correlated with RNAseq data to identify representative plasma biomarkers for different NMF subtypes. Results: Compared to normal tissue, CDC exhibited significant differences in various biological functions including mechanisms and regulation of ion and substance transport across cellular membranes, as well as pathways such as proximal tubule bicarbonate reclamation, aldosterone-regulated sodium reabsorption and protein digestion. Tumor samples were classified into two transcriptional subtypes through machine learning: NMF1 and NMF2. NMF1 subtype had a significantly longer overall survival compared to those with the NMF2 subtype (median OS of 7.06 months vs 2.04 months, P =0.014). The infiltration proportions of T cells, CD8+ T cells, cytotoxic lymphocytes, endothelial cells, myeloid dendritic cells, NK cells, and neutrophils were all higher in NMF1 tumor tissues compared to NMF2, while the proportions of fibroblasts and monocytic lineage were lower in NMF1. These immune infiltration results were further validated by single-cell sequencing analysis. Using Lasso regression combined with survival data, we identified characteristic gene sets for each NMF subtype. Further association analysis with plasma LC-MS/MS revealed distinct protein expression profiles for each NMF subtype. These findings suggest that the NMF1 subtype has a relatively better prognosis and a 'hotter' tumor microenvironment, potentially benefiting from immunotherapy and/or anti-angiogenic therapy. Plasma biopsies could serve as a basis for molecular typing. Conclusions: This study is the largest to date to integrate clinical prognostic information in a multi-omics investigation of CDC. It provides critical insights for clinical research on targeted therapies for specific molecular subtypes of CDC.