Gut microbiota as a biomarker for the efficacy of pembrolizumab in patients with advanced urothelial carcinoma treated in a prospective multicenter study.

825 Background: Pembrolizumab (Pb) is a treatment (trt) for advanced urothelial carcinoma (aUC). In this study, we analyzed the fecal metaproteome of aUC patients (pts) to investigate the differences in the taxonomic and functional composition of the gut microbiome in responders (R) vs non-responders (NR) to pembrolizumab monotherapy as second-line trt. Methods: We designed a prospective multicenter study in France. Stool samples were collected before and nine weeks after the start of Pb, with a maximum follow-up of 36 weeks. Trt efficacy was evaluated using the Response Evaluation Criteria in Solid Tumors v1.1. Proteins were extracted from each sample and analyzed using high-resolution tandem mass spectrometry combined with liquid reverse phase chromatography. Raw data were interpreted without a priori assumptions. The R package Metacoder was used to analyze taxonomic diversity, while the mixOmics package was used to perform partial least squares discriminant analysis (sPLSDA) to reveal microbiota changes between R and NR pts. A Benjamini-Hochberg (BH)-corrected Wilcoxon test identified microbial functions with different frequencies in R and NR pts. A multivariate logistic regression analysis with Bonferroni correction was conducted to create a response-prediction classifier based on clinical variables and identified relevant microbial genera. The significance threshold was set at 5%. Results: From 2019 to 2022, 53 samples were taken from 34 different pts. Of these, 3 (8.8%) pts with complete response, 5 (14.7%) with partial response and 2 (5.9%) with stable disease were classified as R. Four (11.8%) deceased pts before response evaluation were excluded from the analysis. NR corresponded to disease progression in 20 pts (58.8%). No dysbiosis was detected. There were no statistically significant differences in alpha diversity over time or between R and NR. Beta diversity index showed microbiome stability at individual level during follow-up. sPLSDA at the genus level pooling all samples (before and after trt) showed a distinctive separation between R and NR (p ≤ 0.05). Of the genera contributing most to this separation, 4 genera were identified using logistic regression: Anaerostipes, Sutterella, Escherichia, Evtepia (adj. p ≤ 0.05). Functional composition analysis identified 31 signaling pathways that differ between R and NR. These signaling pathways were clustered highlighting 7 functions that differed between R and NR when modulated by Pb. Host function analysis identified the “response to external biotic stimuli” signaling pathway significantly associated to R (BH adj p-value ≤0.05). Conclusions: The composition of the gut microbiome and metabolic functions differed between R and NR pts with aUC under Pb. The genera Anaerostipes , Sutterella, Escherichia and Evtepia could be biomarkers for the efficacy of pembrolizumab. Clinical trial information: NCT03584659 .