Integrated Molecular Epidemiology of Colorectal Cancer: An Exploration of Chronic Inflammation and the Gut Microbiome

Abstract

Background: Colorectal cancer (CRC) is a heterogeneous disease, arising from a complex environment of host inflammation and gut microbes. We investigated whether the novel circulating inflammatory molecule, Macrophage Inhibitory Cytokine-1 (MIC-1), was associated with risk of CRC and CRC-associated mortality. In addition, to gain insights into the role of the gut microbiome in CRC development, we explored how a particular microbial species, Fusobacterium nucleatum, may mediate the association between diet and CRC incidence. Finally, to evaluate the fecal microbiome as a biomarker for CRC, we measured the temporal dynamics and stability of gut microbiome communities, metagenomes, and metatranscriptomes in a large prospective cohort.

Methods: Data for these prospective analyses were collected from two ongoing cohorts, the Nurses’ Health Study (NHS) and the Health Professionals Follow-up Study (HPFS). In both cohorts, participants have returned questionnaires every 2 years to provide information about lifestyle and dietary factors, medication use, and diagnoses of CRC and other diseases. We measured plasma MIC-1 in pre-diagnostic blood samples (1990 in the NHS; 1994 in the HPFS) drawn from 618 participants with incident CRC and 950 matched controls. We also assessed the relationship between levels of MIC1 and prostaglandin-endoperoxide synthase 2 expression (PTGS2 or cyclooxygenase-2), measured in 245 tumor samples by immunohistochemistry. In addition, we calculated Western & Prudent dietary pattern scores among NHS and HPFS participants and classified 990 CRC cases into either F. nucleatum-positive tumors [Fuso(+)] or F. nucleatum-negative tumors [Fuso(-)] tumors according to levels of the nusG gene sequence unique to F nucleatum. Finally, we performed taxonomic and metagenomic profiling on 929 samples collected from 308 participants and metatranscriptomic profiling on 382 samples collected from 96 participants. Unconditional logistic regression models and Cox proportional hazards models were used to examine the association between MIC-1 and risk of CRC as well as CRC-specific mortality. Cox proportional hazards models were also used to examine the association between diet and CRC according to F. nucleatum status. Finally, we calculated Bray Curtis scores to estimate stability of the gut microbiome, and we used GLMs to associate putative factors with microbial stability.

Results: Plasma MIC-1 was associated with a higher risk of incident CRC (Ptrend = .004). Comparing extreme quintiles, the multivariate relative risk (RR) for incident CRC was 1.93 (95% confidence interval [CI] = 1.27 to 2.94). Among individuals with established CRC, plasma MIC-1 was associated with CRC-specific death (Ptrend =0.009). Comparing extreme quartiles, the multivariate hazards ratio (HR) for CRC-specific death was 2.40 (95% confidence interval: 1.33- 4.34). In addition, we found that compared to those in the lowest quartile of the Western dietary pattern, those in the highest quartile had a multivariable-adjusted HR for CRC of 1.30 (95% CI, 1.14-1.48; ptrend, 0.0001). Conversely, those in the highest quartile of Prudent dietary pattern had a RR of 0.87 (95% CI, 0.78-0.97; ptrend = 0.03). The association between Prudent diet and risk of CRC differed significantly according to F. nucleatum enrichment (Phet = 0.005). Comparing extreme quartiles, Prudent diet was associated with a multivariate RR of 0.41 (95% CI = 0.24 to 0.70, P trend = .003) for Fuso(+) tumors. In contrast, the corresponding RR was 0.98, 95% CI = 0.79 to 1.21, P trend = .73) for Fuso(-) tumors. This differential association appeared to be driven by greater intakes of vegetables (Phet = 0.02) and/or dietary fiber (Phet=0.02). Finally, we found evidence that human gut microbiome organismal composition and metagenomic profiles, but not metatranscriptomic data, remain highly personalized over time. Metagenomic stability accounted for 70% of metatranscriptomic stability. Congruently, we found that 80% of the pathways identified as differentially regulated were consistently over- or underexpressed over three or more time points.

Conclusions: Our results support an association of higher levels of circulating MIC-1 (GDF15) with CRC incidence and CRC-associated mortality. In addition, we found that higher Prudent dietary pattern scores as well as greater intakes of vegetables and dietary fiber are associated with lower risk for Fuso(+) CRC but not Fuso(-) tumors, suggesting a potential role of the microbiome in mediating the association of diet with CRC. Finally, we found that the vast majority of gut microbial species and metagenomes can potentially be used as indicators of long-term exposure for future biomarker studies of CRC.