Background: Autism spectrum disorders (ASD) are neurodevelopmental disorders characterised by deficits in social interaction and repetitive and stereotyped interests and behaviours. The aetiology ot ASD comprises known genetic factors [1,2] and increasing evidence suggests a role for intrauterine environmental factors, such as exposure to maternal smoking during pregnancy [3]. Maternal smoking during pregnancy has been associated with other adverse perinatal outcome such as lower birth weight and higher risk of other neurodevelopmental disorders such as attention deficit hyperactivity disorder [5], therefore smoking during pregnancy is a biologically plausible hypothesis in the aetiology of autism but the evidence is still inconsistent. Epigenetics is one potential mechanism linking intrauterine exposures and postnatal outcomes. DNA methylation changes induced by the prenatal exposure could affect genes that are implicated in the aetiology of ASD and therefore DNA methylation could be a causal mediator for the increase risk of ASD. A recent study in the ALSPAC has shown that maternal smoking is associated to DNA methylation changes across the genome, including the replication of genes that were found in other cohorts and new targets [6]. Although an association between maternal smoking in pregnancy and autism might be plausible, there is much inconsistencies in studies to date. In particular, when investigating causal effects in observational studies there is the possibility of unobserved and residual confounding and a way to overcome this issue is by using genetic variants as proxies for the exposure as in a Mendelian Randomization framework [7]. GWAS studies have observed robust associations between smoking and a locus on chromosome 15 involving the CHRNA3 gene, as well as associations in other genes such as BDNF [8].

Aim: We aim to investigate the causal relationship between maternal smoking during pregnancy and childhood's autistic traits and whether this is mediated via DNA methylation changes at birth.