358P Analysis of Microbiome Signatures in Plasma for Early CRC Detection

Colorectal cancer (CRC) represents the 3rd leading cause of cancer-related deaths in the United States. Patients diagnosed with distant stage of the disease represent 22% of all CRC cases and have 15% 5-years survival rates, in contrast to 91% 5-years survival rate of patients with localized disease at diagnosis, indicating how CRC early detection represents an urgent clinical need. Tissue microbiome has been identified as a promising biomarker for the early identification of several cancer types, including CRC. In this study we aimed to investigate presence of selected bacteria in CRC tissue and confirm tissue-based microbiome signals in plasma derived cell-free DNA (cfDNA) in the context of early CRC detection. 19 tissue resident bacteria strains with 772 species known to play an important role in CRC initiation and growth were selected based on literature. Whole genome bisulfite sequencing (WGBS) data from 48 pairs of CRC and matching control tissue were analyzed by mapping non-human genome sequencing reads with Kraken2 tool and calculating for each bacterial genus and each sample the Z-scores of reads percentage that map to the genus. Plasma cfDNA confirmation was done using WGBS data from 8 CRC patients (Stage0=1, Stage1=3, Stage2=3, Stage3=1) and 10 colonoscopy verified control patients. Unbiased hierarchical clustering and Random Forest classifier with leave-one-out stratification was used to evaluate the potential of using microbial sequence information for early detection. In this initial tissue sample cohort, the presence of all 19 bacterial strains was confirmed in CRC samples, with Akkermansia, Porphyromonas, Fusobacterium, Parvimonas showing the highest Z-score on average in comparison to controls. Further cfDNA analysis and model building achieved a sensitivity of 88% (7/8) at a specificity of 90% (9/10), with Stage0/1 sensitivity being as high as 75% (3/4), and Stage2/3 100% (4/4). These results represent a proof of principle on the feasibility of analyzing the microbiome from cfDNA and identifying CRC-specific microbial biomarkers with high accuracy, paving the way for the development of novel approaches for CRC blood-based early detection. Results of this study are further assessed on a larger clinical cohort.