Multi-omics profiling reveals the intratumoral microbiota-immune-metabolism axis associated with preoperative serum CEA status in colorectal cancer

Abstract Background and objectives Carcinoembryonic antigen (CEA) is one of the most widely used serum markers in the clinical management of colorectal cancer (CRC), but the intratumoral ecological basis of the differences in CEA status between patients remains unclear. The aim of this study was to integrate intratumoral microbiome, transcriptome, and metabolome data to resolve the multi-omics features associated with preoperative serum CEA status. Methods 63 patients with primary CRC without preoperative antitumor therapy and undergoing radical resection were included in this study and divided into a high CEA group (H-CEA, n = 27) and a low CEA group (L-CEA, n = 36) according to the preoperative serum CEA 5 ng/mL cut-off value. 16 S rRNA gene sequencing, RNA-seq and non-targeted metabolomics analyses were carried out on the tumor tissues to compare the intratumoral microbiota structure, host transcriptional profiles, immune cell infiltration and metabolite differences between the two groups, and further cross-omics correlation, mediational analyses and machine learning modeling. Results Compared with the L-CEA group, the intratumoral microbiota Observed, Chao1 and ACE richness indices were elevated in the H-CEA group, whereas the Shannon, Simpson and Coverage indices did not show significant differences; PERMANOVA based on Bray-Curtis, Jaccard and weighted/unweighted UniFrac distances neither showed significant overall community structure separation. Linear Discriminant Analysis Effect Size (LEfSe) analysis identified 39 CEA-related differential taxonomic units, 33 of which were enriched in the H-CEA group, including Selenomonas sputigena, Campylobacter gracilis, Peptococcus niger, and Pyramidobacter piscolens. Transcriptomic and immune-infiltration analyses revealed that H-CEA tumors were accompanied by increased activated mast cell and neutrophil infiltration and showed enhanced transcriptional features related to inflammation, chemotaxis, endothelial activation, and immunomodulation. Pyramidobacter -related taxa were associated with neutrophil infiltration, PTGS1, S100A8, SELE, and multiple chemokines. Metabolomics screened five candidate differential metabolites; exploratory mediation analyses suggested that magnoline may partially mediate the association between Synergistaceae/Pyramidobacter and Oscillospiraceae -related taxa and CEA status, with mediation ratios of 57.4% and 53.2%, respectively. A random forest classifier based on the differential microbial taxa showed exploratory discriminative performance in internal cross-validation, although these estimates are internally derived and may be optimistic and require validation in independent cohorts. Conclusion This study depicts the intratumoral microbiota-immunity-metabolism covariance network associated with preoperative serum CEA status and proposes the Pyramidobacter -myeloid inflammatory immunity axis and the microbiota-magnoline metabolism axis as clues to candidate mechanisms. Given that this study is a single-center observational cohort, correlations, mediating effects, and model performance still need to be validated in independent cohorts, spatial localization, and mechanistic experiments.