Novel metabolomic predictors of incident colorectal cancer in men and women

Abstract

Background: Metabolomic profiles may influence colorectal cancer (CRC) development. Few studies have performed prediagnostic metabolome-wide analyses with CRC risk.

Methods: We conducted a nested case-control study among women (Nurses' Health Study) and men (Health Professionals Follow-Up Study) who provided blood between 1989 and 1995. Over 22.9 years, 684 (409 Nurses' Health Study, 275 Health Professionals Follow-Up Study) individuals developed CRC and were matched 1:1 to unaffected participants. Liquid chromatography-mass spectrometry identified 255 plasma metabolites after quality control. Cohort-specific and combined metabolite association analyses were performed using conditional logistic regression. Metabolite set enrichment analysis was used to identify differential abundance in metabolite classes. The R Weighted Correlation Network Analysis package provided modules of covarying metabolites, which were tested for CRC association.

Results: Metabolite set enrichment analysis identified specific acylcarnitines associated with higher CRC risk and triacylglycerols with lower CRC risk among women and men. Further, phosphatidylcholines were associated with a higher risk of CRC among men. In an analysis restricted to CRC diagnosed 2 years after blood draw, myristoleic acid (odds ratio = 1.37 [95% CI = 1.15 to 1.62]; false discovery rate = 0.072) and C60:12 triacylglycerol (odds ratio = 0.75 [95% CI = 0.64 to 0.88]; false discovery rate = 0.072) were associated with CRC risk in women. Weighted correlation network analysis identified amino acids associated with CRC in men, fatty acid esters (carnitines) with distal CRC in men, and triradylcglycerols inversely associated with CRC in women.

Conclusions: We identified prediagnostic CRC-associated metabolites with distinct sex-specific profiles. These results provide insight into CRC etiopathogenesis and have implications for risk prediction strategies.

Figure 1.

Gene set enrichment analysis and the Weighted Correlation Network Analysis package identify metabolic signatures associated with CRC risk. A) Enriched metabolite set enrichment analysis terms, by cohort, with an false discovery rate <0.10. Positive normalized enrichment scores indicate enrichment among positively associated CRC metabolites and the opposite for negative normalized enrichment scores. B) Miami plot showing the ‒log₁₀(P value) association of all metabolites, grouped into modules, with CRC risk by nested case-control cohort. The strongest NHS modular association was the Light Salmon module, which was enriched for triradylcglycerols, with C60:12 TAG being the most strongly associated CRC metabolite within that module. The HPFS Dark Salmon module, enriched for peptides, had the strongest association with CRC, with N2, N2-dimethyguanosine being the most CRC-associated metabolite within that module. CRC = colorectal cancer; HPFS = Health Professionals Follow-Up Study; NHS = Nurses’ Health Study; PC = phosphatidylcholine; PE = phosphatidylethanolamine; TAG = triacylglycerol.