AN EMPIRICAL DIETARY PATTERN ASSOCIATED WITH GUT MICROBIAL FEATURES IN RELATION TO COLORECTAL CANCER RISK

BACKGROUND AND AIMS: Chemoresistance is a major cause of colorectal cancer (CRC) recurrence and death. The critical role of gut microbiome in the efficacy of CRC chemotherapy remains unclear. We aim to characterize the gut microbiota signatures in responders and non-responders to chemotherapy and identify chemoresistance-associated pathogens in CRC.

METHODS: Stool samples were collected from CRC patients before chemotherapy (n=64) and analyzed by shotgun metagenomic sequencing. The effect of enterotoxigenic Bacteroides fragilis (ETBF) on chemoresistance was assessed in CRC cells and CRC mouse models of xenografts, Apc^Min/+ or azoxymethane (AOM)/dextran sodium sulfate (DSS) induced CRC tumorigenesis. Phage VA7 was used for targeting ETBF in vitro and in vivo. Bacterial adhesin-host receptor interaction was investigated by far western, biotin pull-down, and GST pull-down assays followed by mass spectrometry. RNA-seq was performed to evaluate involved pathways.
RESULTS: Bacteroides fragilis was identified as the top enriched bacterium in stool samples of CRC patients who were resistant to chemotherapy, and its abundance was associated with poor prognosis (P=0.019). Consistent with this observation, we demonstrated that ETBF co-culture with CRC cells (HT29 and HCT116) inhibited 5-FU and oxaliplatin (OXA)-induced cell apoptosis (both P<0.001). ETBF promoted chemoresistance in xenografts of nude mice (P<0.001). In particular, oral gavage of ETBF to Apc^Min/+ mice significantly attenuated the antitumor efficacy of 5-FU and OXA. Such effect was further confirmed in the second CRC tumorigenesis mouse model induced by AOM/DSS. Mechanistically, we revealed that ETBF activated Notch signaling pathway genes (Notch1, NICD, Hes1, MAML1, and c-Myc) identified by RNA-seq and confirmed by Western blot. ETBF-mediated chemoresistance critically depends on its direct interaction with CRC cells, as neither ETBF supernatant nor non-contact transwell co-culture significantly affected the efficacy of 5-FU and OXA. Concordantly, fluorescence in situ hybridization (FISH) and qPCR demonstrated the significant enrichment of ETBF in colonic tumors compared to adjacent normal tissues (P = 0.002) in mice. Scanning electron microscopy (SEM) visualized the attachment of ETBF to colon epithelial cells. ETBF surface protein SusD/RagB was identified to mediate its attachment to CRC cells. Finally, we demonstrated that ETBF-targeting phage VA7 selectively inhibited ETBF in vitro and in mice (P<0.05), and that VA7 co-administration abolished ETBF-induced chemoresistance (P<0.001).
CONCLUSION: We identified for the first time that ETBF provokes chemoresistance in CRC cells through its direct interaction with colon epithelial cells and activating Notch1 signaling. Targeting ETBF by phage VA7 is a promising therapeutic strategy to overcome ETBF-mediated chemoresistance in CRC.

Background
Diet interacts intimately with the gut microbiome and is also a major risk factor for colorectal cancer (CRC). However, epidemiologic evidence on how diet may influence CRC through alteration of the gut microbiome remains limited.

Methods
We leveraged two cohorts of 307 healthy men and 212 healthy women with serial stool metagenomes and detailed dietary data. We empirically characterized a dietary pattern associated with the CRC gut microbiome signature (labeled CRC Microbial Dietary Score [CRC-MDS]), which sums the abundances of 30 species enriched and 13 depleted in CRC based on their loadings in a recent meta-analysis of metagenomic studies (Thomas et al. Nat Med. 2019). We then applied the CRC-MDS to three prospective cohorts – the Health Professionals Follow-up Study (1986-2014), the Nurses’ Health Study (1984-2014), and the Nurses’ Health Study II (1991-2015), in which we assessed diet by semi-quantitative food frequency questionnaires every four years, confirmed self-reported CRC cases through medical record review, and collected tumor tissues from a subset of CRC patients. We evaluated the CRC-MDS association with CRC risk using Cox proportional hazards model, and further assessed CRC risk according to subsites and tumor tissue presence of Fusobacterium nucleatum (F. nucleatum).

Results
The CRC-MDS was characterized by high consumption of foods with greater industrial processing (e.g., processed meat, butter, and low/high-energy drinks) and low consumption of foods rich in fiber (e.g., fruits, vegetables, nuts, and whole grains) (Figure). The score was positively correlated with Western diet (Spearman ρ =0.44) and empirical dietary index for hyperinsulinemia score (ρ =0.54), and negatively correlated with the prudent dietary pattern score (ρ =-0.53). In the three cohorts of 259,190 participants, we documented 3,819 incident cases of CRC over 3,223,696 person-years of follow-up. Compared with participants in the lowest quintile of the CRC-MDS, those in the highest quintile had an increased risk of CRC (hazard ratio [HR], 1.25; 95% confidence interval [CI], 1.12-1.39; P_trend <0.001) after adjusting for putative CRC risk factors. The association was attenuated but remained statistically significant after adjusting for other dietary patterns (see details in Table). The association was similar across CRC subsites but differed according to the presence of F. nucleatum in tumor tissue. The CRC-MDS was strongly associated with a higher risk of F. nucleatum-positive CRC (HR_{Q5 vs Q1}, 2.52; 95% CI, 1.68-3.76; P_trend <0.001) but not F. nucleatum-negative CRC (HR_{Q5 vs Q1}, 1.09; 95% CI, 0.95-1.25; P_trend =0.19) (P_{heterogeneity} = 0.03).

Conclusions
The CRC-MDS was positively associated with CRC risk, especially F. nucleatum-positive CRC. Our findings support the gut microbiome as a mediator of diet in CRC development.