DNA methylation patterns derived from fetal vulnerability to maternal smoking relate to future child outcomes

Jane WY Ng
Scientific Abstract Maternal smoking during pregnancy (MSP) has an independent and causal effect on fetal health outcomes. Through accumulating epidemiologic and experimental evidence, our understanding of the breadth and duration of health effects of this toxic exposure is expanding. MSP has been linked to the etiology of many non-fatal, non-communicable common complex diseases (CCDs) such as depression and anxiety, poor cognitive performance, asthma, cardiovascular disease, diabetes and
more » ... y. Despite the potency and prevalence of this exposure around the world, the mechanisms mediating these effects on human health are still unknown. While numerous studies hypothesize that MSP dysregulates fetal developmental programming through epigenetic modifications such as DNA methylation (DNAm), there is yet not a single clinically useful epigenetic marker for CCDs. This is despite an explosion of human cohorts and animal models studying DNAm since about the early-to mid-2000's. This failure is juxtaposed with the success and rapid advancement of epigenetic markers and therapies in cancers of multiple forms within a similar period. Perhaps one of the greatest barriers to clinical translation is the "gap" between genes, epigenetics and phenotype in complex traits. A frequent finding in association studies of CCDs is that many individuals may have shared phenotypic traits, but at best weakly share individual environmental risk factors or genetic/epigenetic markers. This gap further widens with factors such as varying intronic genetic mutations, phenotype heterogeneity and complex gene x environment (GxE) interactions. In this thesis, we consider that mis-assignment of the contribution of genetic and environmental factors relevant to any given individual can lead to false conclusions regarding its effect on observed disease and/or epigenetic manifestations. Moreover, we reason that epigenetic differences persistent and potent enough to underlie the pathogenesis of CCDs must shift the mechanics of regulation across the genome. Thus, we speculate that DNAm related to CCD must alter chromosomal activity by orchestrating changes in DNA interactions that are stably maintained and have regulatory consequences on multiple genes through modification of their physical contact with chromosomal and other nuclear structures. Based on these premises, we explore context-based mapping of two entities: 1) individual-level risk profiling based on vulnerability to exposure rather than exposure alone and 2) DNAm profiling based on genome-wide patterns rather than single feature differences. In this way, we seek the relevance of a given MSP-related signal by iii couching it within its clinical-level and genome-level context in order to visualize and adjudicate its relation to health. We apply this context-dependent analytic approach to population-based data from the UK pregnancy cohort, Avon Longitudinal Study of Parents and Children (ALSPAC). These children have wide-ranging vulnerability to MSP and heterogeneous physical and mental outcomes -an ideal situation to model common epigenetic pathways among diverse GxE contexts for complex traits. We recruit specific multidimensional data analysis methods to extract a small number of DNAm patterns found in cord blood that are representative of fetal vulnerability to MSP. The biological coherence of these patterns is supported by three main findings. First, patterns are enriched for sites of chromosomal regulation at both genic and intronic regions. Second, specific patterns are shared among phenotypically similar children throughout childhood suggesting a common epigenetic shift underlying their physical and mental developmental trajectories. Third, most patterns persisted in blood collected in middle childhood and adolescence despite random, technical and physiologic methylation changes expected over time. This supports a robust relation to stable phenotypic effects starting from fetushood. We observed replication of these vulnerability patterns in cord DNAm data in an independent cohort (Generation R, Netherlands). These patterns were derived directly from ALSPAC with no additional clinical data from Generation R. Yet, these "template-based" DNAm patterns related similarly to later childhood phenotype within Generation R as in ALSPAC. The novelty of this work lies in its use of context-based patterns of risk and epigenetic differences to provide a more detailed map of complex trait architecture. As envisioned by biologists like Conrad Waddington over a half century ago, the overlap of such maps -rather than unimodal data points -may provide deeper and potentially more accurate insights into the molecular underpinnings of complex diseases. iv
doi:10.7939/r3-p554-wm21 fatcat:74dbvmoyznbsvfidspyd3fy3oi