Background: The incidence of early-onset colorectal cancer (CRC) has been rising since the 1990s in many parts of the world.  Although the exact reason for this phenomenon remains uncertain, it is hypothesized that increased risk factor exposures in young individuals since the mid-20th century might play a substantial etiological role.  Also, in early-onset CRC, the intestinal microbiota is considered to play interrelated pathogenic roles.  Among intestinal bacteria, Escherichia coli strains harboring the polyketide synthase (pks) genomic island (pks+ E. coli) produce the genotoxin colibactin, which induces DNA damage and generates a distinct mutational signature.  A large international study identified an enrichment of colibactin-induced mutational signatures in early-onset CRC.  Therefore, linking risk factors with tumor tissue pks+ E. coli status and colibactin-induced mutational signatures can provide valuable insights into the pathogenesis of early-onset CRC. 
 
Objectives: To test our hypotheses that the associations of early-life and long-term exposures with CRC incidence might differ by tumor tissue pks+ E. coli status and colibactin-induced mutational signatures (early-onset CRC-associated tumor characteristics).
 
Methods: We used the resources of two U.S.-wide prospective cohort studies (Nurses’ Health Study and Health Professionals Follow-Up Study) with data on tissue pks+ E. coli and colibactin-induced mutational signatures.  We examined the differential associations of a history of being breastfed and empirical lifestyle index for insulin resistance (ELIR) with CRC incidence, subclassified by tumor tissue pks+ E. coli and the colibactin-induced mutational signature, using an inverse probability weighted duplication-method Cox proportional hazards regression model.
 
Results: During follow-up of 118,746 participants (2,011,959 person-years), we documented 4,606 incident CRC cases, including 1,012 cases with available pks+ E. coli data and 689 cases with tumor exome-sequencing data.  Compared to non-breastfed individuals, breastfed individuals experienced higher incidence of pks? E. coli-positive CRC (multivariable-adjusted hazard ratio [HR], 2.07, 95% confidence interval [CI]: 1.10-3.89) but not microbe-negative CRC (HR 0.93, 95% CI: 0.75-1.17; P heterogeneity = 0.051).  Similarly, compared to non-breastfed individuals, breastfed individuals were consistently associated with increased incidence of CRC harboring colibactin-induced mutations (HR 2.34, 95% CI: 1.59-3.44).  For ELIR and CRC incidence, multivariable-adjusted HRs (with 95% CIs) in individuals with the highest (vs. lowest) ELIR tertile were 2.26 (1.30-3.93; P trend = 0.0016) for pks+ E. coli-positive CRC incidence and 1.02 (0.86-1.20) for pks+ E. coli-negative CRC incidence (P heterogeneity = 0.0017).  Similar differential associations were observed in the analyses of the colibactin-induced mutational signatures.
 
Conclusions: Our findings indicate that a history of being breastfed and long-term insulin resistance-promoting lifestyle patterns facilitate pks+ E. coli-induced colorectal carcinogenesis, which is implicated in the pathogenesis of early-onset CRC.  Considering that pks? E. coli may be acquired early in life and that colibactin-induced mutations are enriched in early-onset CRC, our findings suggest that exposures in early-life, such as being breastfed and insulin resistance-promoting lifestyle patterns, may facilitate pks? E. coli-induced colorectal carcinogenesis.  Our integrative molecular epidemiologic studies highlight a need for further integrative population studies to investigate potential biological, environmental, and lifestyle factors contributing to the rising burden of early-onset cancers.