Predicting response of gemcitabine plus cisplatin chemotherapy using RNA sequencing in patients with advanced urothelial carcinoma.

Abstract

66 Background: Despite of advancement using new agents including ICIs in treatment of urothelial carcinoma (UC), cisplatin-based chemotherapy is still standard of care in both 1st line palliative setting and perioperative setting. Previous studies regarding response prediction of cisplatin-based chemotherapy are limited in selecting target molecule or gene, there have been no comprehensive study for whole genome information. This study aimed to identify key genes to predict response of gemcitabine plus cisplatin (GC) in advanced UC by mRNA sequencing (RNA-seq). Methods: Between JAN 2016 and DEC 2019, advanced UC patients treated with GC and underwent response evaluation were enrolled. Patients were divided into the responder group (CR or PR) and non-responder group (SD or PD). RNA was extracted from the formalin fixed embedded tissue (FFED) and RNA-seq was performed on the Illumina RNA-seq platform. Differentially expressed genes (DEGs) were identified between responder group and non-responder group. Marker selection was performed as following steps: (1) DESeq2 (Adjusted P-value < 0.1) and Mann-Whitney U test (P-value < 0.05), (2) random forest (RF) and logistic regression, and (3) assessments for relevant markers by analyzing RNA/Protein expression level, survival information, and antibody staining results using ProteinAtlas. Results: Twenty-nine patients received GC chemotherapy during study period. Among them, 10 patients who had available FFED specimens and clear response evaluation were enrolled and performed RNA-seq. Median age was 65.5 years (range, 62-76), and all were male. They were grouped of 7 patients in responder and 3 patients in non-responder. Among 13,773 genes based on RNAseq, we identified 37 DEGs candidate (Step 1), followed by 22 DEGs selected (Step 2), and finally 9 DEGs were identified (Step 3). Among the 9 DEGs, 3 genes (UPK2, PCGF3, and POFUT1) were consistently expressed with survival data of TCGA, the expression of 3 (ITGA3, HSPA1B, GSTM3) was the opposite of that of TCGA, and the remaining 3 genes (S100A2, TDRP, and RPS6KA2) were not relevant. Conclusions: We found 3 key genes (UPK2, PCGF3, and POFUT1) that related with favorable response of GC in our advanced UC cohort which were confirmed in TCGA data. Our study results should be further investigated in an expansion cohort using immunohistochemistry staining matched with key genes for convenient practical application.