Prenatal exposure to smoking has been linked to various adverse outcomes that can affect a child at birth and/or in later life. Given that this exposure is modifiable, it represents a pivotal area in which to alleviate major social, economic and health burdens. There is immense interest in epigenetic processes and the role that they might play in mediating complex disease risk through gene-environment interactions. This project adopts a genetic epidemiology approach, Mendelian randomisation, and aims to further our understanding of the contribution made by in utero exposures on child outcomes via epigenetic mechanisms. We will use SNP and DNA methylation data currently available from the ALSPAC cohort-both gene-specific and genome-wide - and supplement with 27K methylation array data as funding becomes available.

We will explore whether maternal smoking is related to infant outcomes via intrauterine influences and if so, whether this is mediated by epigenetic changes, (i.e. DNA methylation.) Studies of exposure to maternal smoking are challenged by significant confounding of exposure measures. Due to social stigma, self-report of smoking during pregnancy may be particularly prone to bias. As well, differences in smoking topography limit the correlation of reported smoking quantity with dose of inhaled tobacco products. Most importantly, there is a well-known association of maternal smoking with measures of social class. The Mendelian randomisation approach can be used to overcome these design issues. It allows for the use of a specific maternal genetic variant as a proxy for a given exposure. There are a growing number of such variants. For example, findings from several genome-wide association studies indicate that genetic variation in the region of the CHRNA5- CHRNA3- CHRNAB4 gene (chromosome 15q25) is closely linked to smoking quantity. Given that the randomisation of gene variants occurs at meiosis and precedes the development of phenotype, variation at this site can be used as a relatively unconfounded proxy of smoking quantity in smokers. The CHRN variants will be used together with other potentially informative smoking related SNPs both as individual instruments and as multiple instruments.

For smoking-related outcomes, we will select child phenotypes having strong evidence of gene-environment interactions based on animal and human data, specifically birth parameters, respiratory health and neurodevelopment phenotypes at various and multiple points throughout the lifecourse. We will investigate these outcomes as well as relevant covariates such as gestational age, related family history of outcome phenotypes, maternal obstetric health, maternal grandparent smoking history and family social status. To determine intrauterine effect, history of paternal smoking will be used as a control.

Using data already generated by the ALSPAC cohort, we will explore potential relations between maternal smoke exposure, (determined via genetic proxy and self-report,) and offspring DNA methylation patterns. Currently, global and gene-specific DNA methylation data for 192 children is being generated. This analysis will serve as proof of principle to support future funding requests to obtain further DNA methylation data. To date, there is limited data comprehensively linking fetal exposure to maternal smoking, gene-specific epigenetic modifications, changes in gene expression and finally disease phenotype. However, studies connecting two or more nodes in this pathway offer several potential candidate genes. We will maximize our capacity to detect smoking-related epigenetic changes in candidate genes by comparing "typical" and "atypical" cases and controls. "Typical" refers to mother-infant pairs that demonstrate the expected relation between maternal smoking and adverse infant outcomes, (e.g. maternal smoking and lower birth weight.) Conversely, "atypical" refers to pairs that demonstrate the opposite relation. Thus, atypical pairs reveal mechanisms that are unlikely related to the exposure and can help target candidate genes.

If evidence of a smoking-related intrauterine effect on DNA methylation is found, a secondary objective will be to interrogate the interaction between prenatal smoke exposure and methylation of genes associated with the outcomes, (e.g. birthweight and BDNF gene.) Such analyses may shed light on potential biological pathways linking smoke exposure to pathophysiologic processes in utero. Our findings can gain further support through cross referencing with 46K Illumina expression chip data when it becomes available.

Proving causality poses unique statistical issues when performing correlations between multiple genetic proxies and phenotype. We will use specific data analysis techniques that address directionality and the use of multiple genetic mediators to strengthen causal inferences.

The study will make a significant contribution to the emerging field of epigenetic epidemiology and will serve as proof of concept for further studies linking smoke exposure, epigenetic modifications and child outcomes.