Abstract

The gut microbiota has emerged as an important modifier of colorectal cancer (CRC), yet how tumor genotype influences host-microbiota interactions and whether microbial signals are required throughout tumor progression remain unclear. Here, we combined genetically engineered mouse models, microbial multi-omics and a germ-free-compatible orthotopic transplantation system to define the functional contribution of the microbiota across distinct stages of CRC evolution. Across multiple CRC genotypes, we identified tumor-associated microbial ecosystem states characterized by distinct taxonomic, functional and metabolic configurations. To directly test their contribution to tumor progression, we established the first orthotopic CRC transplantation platform compatible with long-term experimentation in germ-free mice, enabling side-by-side comparison of genetically identical tumors in the presence or absence of microbiota. Using organoids spanning low-grade adenoma, high-grade adenoma and adenocarcinoma states, we found that the dependence on the presence of microbiota progressively decreases during malignant evolution. Whereas adenoma-derived organoids exhibited profound dependence on microbial exposure and failed or were markedly impaired in establishing tumors under germ-free conditions, adenocarcinoma organoids engrafted and metastasized in both germ-free and specific pathogen-free (SPF) hosts. Unexpectedly, comprehensive histological, immunological and transcriptomic analyses revealed highly similar tumor ecosystem states in advanced tumors arising under both microbial conditions, arguing against broad immune or epithelial defects as primary explanation for the observed phenotype. Together, our findings demonstrate that distinct oncogenic drivers establish specific microbial ecosystem states and reveal a stage-dependent role of the microbiota during colorectal tumorigenesis. Whereas microbial signals are critical during early stages of tumor progression and may promote malignant transformation, advanced tumors progressively acquire microbiota-independent growth programs and increasingly impose genotype-specific ecological signatures on the surrounding microbial ecosystem. More broadly, we establish a versatile framework for the causal dissection of tumor-microbiota interactions in cancer.