Background: Immune checkpoint inhibitors (ICIs) showed efficacy in metastatic urothelial carcinoma (UC) and promising activity in muscle-invasive and non-muscle invasive UC of the bladder. Recent studies revealed the immunomodulatory effect of the gut microbiota on ICIs efficacy across several malignancies, identifying microbial “signatures” associated with response to therapy and effective antitumoral T-cell activity. In our study, we aimed to study the stool microbiota in patients undergoing neoadjuvant immunotherapy (IO) for muscle-invasive UC. Methods: Pre-IO stools were available for analysis from 42 patients enrolled in the PURE-01 trial (NCT02736266), testing 3x200mg flat-dose pembrolizumab every 21 days before radical cystectomy (RC). All samples were collected using Stool Nucleic Acid Collection and Preservation Tubes (Norgen) and extracted using the Stool DNA Isolation Kit (Norgen), according to the manufacturer’s protocol. 16s sequencing was performed using standardized protocols at the internal facility, using mock communities and DNA standards (ZymoBIOMICS) to control for extraction and sequencing contaminations. A QIIME-based bioinformatic pipeline was used for microbiome analyses. Complete response (CR) to neoadjuvant IO was defined as ypT0N0 at pathologic examination on radical cystectomy specimens, while partial response (PR) was defined as < ypT2N0. Concomitant antibiotic therapy (ABT) was defined as any ATB between 30 days prior to the first pembrolizumab dose and the planned RC. Results: In our study sample, 23 patients responded to IO (21 CR + 2 PR). Overall median age was 68.5 years. Among responders, 20 (87%) patients had a smoking history (vs. 15 (79%) in non-responders) and 4 (17%) underwent concomitant ABT (vs. 6 (32%) in non-responders). Alpha-diversity assessed by richness (ACE index) was higher in responders vs. non-responders (p = 0.05), while no significant diversity was found. Beta-diversity did not show clear clustering of responders vs. non-responders. LEfSe identified 16 bacterial taxa with a linear discriminant analysis (LDA) score ≥2.5 that were differently enriched between responders and non-responders. Among them, we identified the genus Sutterella enriched in responders (p = 0.02), while the species Ruminococcus bromii was enriched in non-responders (p = 0.02). Conclusions: Our analyses showed an association between response to neoadjuvant-IO and microbiome composition in an intention-to-cure population with muscle invasive UC. We found bacterial taxa specifically enriched in responders or non-responders using pre-therapy stool specimens. The identified taxa may be tested in future studies as potential indicators of therapy outcomes, alone or in combination with other IO biomarkers. These results may also inspire new strategies of gut microbiota modulation to promote response in immunotherapy-refractory patients.