Abstract
BACKGROUND: Tumor-associated microbiota are implicated in colorectal cancer (CRC). Formalin-fixed paraffin-embedded (FFPE) tumor tissue is widely available yet seldom profiled for microbiota. We tested whether quantitative and presence/absence signals of selected taxa in FFPE tumors associate with clinicopathological features. METHODS: DNA from FFPE primary CRCs (n = 52) was assayed by a targeted PCR panel quantifying 30 bacterial taxa, Candida spp., and total bacterial load. Presence/absence combinations were selected by the Apriori algorithm with Fisher's exact testing and 10,000-permutation empirical p-values. Quantitative features were modeled by LASSO logistic regression; discrimination of single taxa and combinations was evaluated by ROC/AUC. RESULTS: In relative-abundance analyses, Fusobacterium nucleatum showed pro-metastatic value (AUC = 0.622). The best absolute-abundance model for metastasis combined F. nucleatum, Faecalibacterium prausnitzii, total bacterial load, and Akkermansia muciniphila (AUC = 0.739). Anti-metastatic directionality in relative-abundance models was driven by Acinetobacter spp.; the two-taxon set Eubacterium rectale + Acinetobacter spp. achieved AUC = 0.747. CONCLUSIONS: PCR-based profiling of FFPE CRC tumors is feasible and reveals hypothesis-generating patterns. Signals linking F. nucleatum to metastatic CRC and Acinetobacter spp. to non-metastatic disease merit validation in larger cohorts; tumor-tissue microbiome features may complement clinicopathological assessment.