Establishing A Causal Effect Of Epigenetic Changes On Risk Of Cancer

Rebecca RICHMOND, University of Bristol, United Kingdom
SHIHAB H. 2 , HAYCOCK P. 1,2 , GAUNT T. 1,2 , DAVEY SMITH G. 1,2 , RELTON C. 1,2

1 CRUK Integrative Cancer Epidemiology Programme, University of Bristol, Bristol, UK
2 MRC Integrative Epidemiology Unit, University of Bristol, Bristol, UK

Demethylation at AHRR is a sensitive indicator of cigarette smoking. AHRR has also emerged as the site where DNA methylation is most strongly (and inversely) associated with lung cancer risk. We used Mendelian randomization to establish whether methylation at AHRR has a causal effect on lung cancer, which may mediate the carcinogenic effects of tobacco smoke. This approach uses the random assortment of alleles at conception to circumvent confounding and reverse causation.
We identified two independent genetic variants, rs2671915 and rs2672766, robustly associated with AHRR (cg05575921) methylation in ~1,000 individuals from the Accessible Resource for Integrative Epigenomic Studies (ARIES) which were then used to establish a causal effect of methylation on lung cancer in  ~75,000 individuals from the International Lung Cancer Consortium (ILCCO).  
For each variant, we calculated the odds ratio (OR) for disease per unit increase in AHRR methylation by the Wald ratio (βGD/βGP), where βGD is the OR for disease per copy of the effect allele in ILCCO and βGP is the unit change in methylation per copy of the effect allele in ARIES, which were then combined in fixed effects meta-analysis.
We found no strong evidence for a causal effect of AHRR DNA methylation on total lung cancer risk (OR = 1.05 (0.93, 1.18) per unit increase in methylation) nor for adenocarcinoma (OR = 0.94 (0.79, 1.13)), although a causal effect was seen for squamous cell carcinoma (OR = 1.35 (1.11, 1.63)).
Results suggest that previous findings implicating increased AHRR methylation with decreased cancer risk are likely to be confounded and rather increased AHRR methylation may increase squamous cell carcinoma risk, which is more in line with the action of AHRR as a tumour suppressor. This approach can be extended to establish the causal role of methylation in mediating associations between other risk factors and cancer.