Background: Elevated plasma C-reactive protein (CRP) - a non-specific marker of systemic inflammation - is associated with an increased risk of cancer in observational studies. However, this relationship is unlikely to be causal and may instead reflect confounding by unknown or unmeasured factors upstream to CRP. Here, we propose a novel approach for identifying phenotypic confounders of two traits by employing cross-trait pleiotropy analysis to detect genetic loci that jointly affect both traits and multi-trait colocalisation to identify molecular phenotypes mediating these effects. We apply this approach to CRP and 10 cancers and demonstrate how it can provide insight into novel inflammation-related disease pathways and therapeutic targets for cancer prevention.
Methods: We used multivariable-adjusted proportional hazards models to investigate the association between pre-diagnostic CRP and risk of 11 cancers previously linked to chronic inflammation in the UK Biobank (N=35,797 cases, 378,234 controls). We then employed cis-focused and reverse Mendelian randomization (MR) analysis to examine bidirectional causal effects between CRP and risk of these cancers using summary genetic data on plasma CRP (N=575,531) and cancer susceptibility (N=388,173 cases, ≤1,663,971 controls). Genome-wide pleiotropy analysis to identify pleiotropic loci influencing CRP and cancer was performed using PLACO (PPLACO<5x10-8) and colocalisation to confirm shared causal variants at these loci (PPH4>0.50). We mapped lead SNPs at pleiotropic loci to the nearest protein-coding gene and used HyprColoc multi-trait colocalisation to identify proteomic and immune cell-specific transcriptomic mediators influencing CRP and cancer at each locus. Finally, we mapped candidate effector genes to targets of approved medications in ChEMBL/DrugBank to investigate drug repurposing opportunities.
Results: In multivariable-adjusted proportional hazard models, pre-diagnostic CRP was associated with increased risk of 10 of 11 cancers evaluated in the UK Biobank. In contrast, we found no evidence of a causal relationship between plasma CRP and risk of these cancers in bidirectional MR, suggesting that observational associations may be confounded. In genome-wide pleiotropy and colocalisation analysis we identified 90 loci with shared aetiological effects on CRP and cancer including 38 mapped to genes with established roles in inflammation and cancer including IL6 (lung cancer) and JAK1 (non-Hodgkin lymphoma), and 52 novel loci including CEBPB (breast cancer) and GCKR (colorectal cancer). Multi-trait colocalisation identified putative molecular mediators at 51 loci including a pleiotropic effect on CRP-breast cancer at TLR1 mediated by plasma TLR1 levels and CRP-lung cancer at FUBP1 mediated by memory B-cell FUBP1 expression. 11 candidate effector genes encode targets of approved or investigational medications, including PDE4D and CASP8, indicating potential opportunities for their repurposing as novel pharmacological agents for cancer prevention.
Discussion: Through analysis of CRP and 10 cancers, we demonstrate how the identification of shared genetic mechanisms underlying non-causal phenotypic associations can provide insight into potential molecular confounders driving these associations. Our findings both recapitulate known biological pathways in cancer development, providing proof-of-principle for our approach, and identify novel molecular mechanisms contributing to inflammation-related cancer risk. The proposed approach provides a generalisable framework for leveraging non-causal phenotypic relationships to yield novel insights into disease mechanisms and therapeutic targets for disease prevention.