Abstract 2370: Microbiome alterations of colorectal cancer according to age of onset, stage, anatomic location, and mismatch repair deficiency

Abstract Colorectal cancer (CRC) is the third most common and lethal cancer type worldwide. Although CRC incidence in the older population has decreased in many countries, early-onset CRC cases (< 50 years age, EOCRC) have been rapidly increasing globally. The fecal microbiome has been reproducibly associated with CRC. However, microbiome differences according to age of onset, stage, anatomic location, and genomic instability remain only minimally explored. Here, we integrated data generated from the ONCOBIOME consortium, Cancer Grand Challenges PROSPECT project and the ARETHUSA study, providing seven new cohorts of gut microbiome samples from healthy, adenoma- and CRC individuals (n total = 1, 668) combined with 12 public studies, for a total of 3, 784 metagenomic samples, including 190 EOCRC and 270 age-matched control samples. Information on tumor stage, anatomic location, microsatellite stability/instability-high (MSS/MSI-H) has been curated and harmonized across datasets. Metagenomes were analyzed using the bioBakery suite for community-, species-, strain-level, and gene-level profiling, and downstream integrative analysis was performed via meta-analysis and leave-one-dataset-out learning tasks. We obtained increased microbiome-based CRC prediction capabilities (average leave-one-dataset-out AUC = 0.84) and accurate species- and strain-level characterization of CRC stages (e.g., Slackia exigua and Ruminococcus bicirculans, respectively, between stage 0-II and III/IV), while the microbiome in adenoma samples did not differ substantially from controls (average leave-one-dataset-out AUC < 0.6). The fecal microbiome of right-sided CRC cases presented lower species and higher oral commensals richness than left-sided cases. The gut microbiome composition of MSI-H CRC was significantly different than MSS, as reflected by machine learning discrimination performances (average cross-validation AUC = 0.71), and characterized by reduced alpha-diversity and fewer biomarkers than in MSS cases, such as Ligilactobacillus salivarius and Limosilactobacillus fermentum. We confirmed that EOCRC microbiome was significantly different from age-matched controls (average leave-one-dataset-out AUC = 0.84), harboring typical CRC-associated bacteria, such as F. nucleatum and D. pneumosintes. Overall, we defined cross-cohort reproducible microbiome associations across tumor stages, primary location, genetic instability, and EOCRC, providing improved resolution within the CRC microbiome and more refined targets for diagnosis and therapeutic stratification. Citation Format: Gianmarco Piccinno, Kelsey N. Thompson, Paolo Manghi, Luca Lazzari, Chiara Pozzi, Andrew R. Ghazi, Andrew M. Thomas, Aitor Blanco-Miguez, Francesco Asnicar, Katarina Mladenovic, Federica Pinto, Federica Armanini, Giulio Ferrero, Sonia Tarallo, Long H. Nguyen, Yan Yan, Mingyang Song, Veronika Vymetalkova, Mario Trompetto, Vaclav Liska, Tomas Hucl, Pavel Vodicka, Beatrix Bencsiková, Martina Čarnogurská, Vlad Popovici, Salvatore Siena, Andrea Sartore-Bianchi, Michela Lizier, Carmen Correale, Alessandra Raimondi, Filippo Pietrantonio, Federica Marmorino, Chiara Cremolini, Barbara Pardini, Francesca Cordero, Lisa Derosa, Silvia Marsoni, Alberto Bardelli, Maria Rescigno, Yin Cao, Andrew T. Chan, Laurence Zitvogel, Curtis Huttenhower, Alessio Naccarati, Eva Budinska, Nicola Segata. Microbiome alterations of colorectal cancer according to age of onset, stage, anatomic location, and mismatch repair deficiency. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2370.