UC Berkeley

Epigenetic mechanisms, particularly DNA methylation, are a possible link between environmental and biological determinants of health. As the DNA methylome undergoes rearrangement in utero and is susceptible to environmental insults, it may be a mechanism explaining the developmental origins of human disease with public health importance. However, the epidemiologic studies needed to identify the role DNA methylation plays mediating environmental exposure disease risk still face several obstacles.

This dissertation addresses knowledge gaps impeding the rigorous adoption of genome-scale measures of site-specific DNA methylation, like the Illumina Infinium HumanMethylation450 (450K) BeadChip®, into epidemiologic study designs. We then investigate the impact of prenatal exposure to polybrominated diphenyl ethers (PBDEs) on DNA methylation of children at birth. PBDEs are a class of flame retardant chemicals widely used in U.S. consumer products over the last 40 years that have previously been associated with adverse neurobehavioral outcomes, obesity, and other effects.

Specifically, we aimed to:

1) Identify and minimize sources of technical variation for site-specific DNA methylation measured by 450K BeadChip assay in epidemiologic studies

2) Characterize sources of biological variation due to host factors (e.g. blood cell composition and sex) in measures of whole blood DNA methylation at birth

3) Determine whether prenatal exposure to PBDEs is associated with differential methylation patterns of CpG sites in umbilical cord blood

Our results identified that the newly proposed All Sample Mean Normalization (ASMN) procedure performed consistently well, both at reducing batch effects and improving replicate comparability compared to several other leading normalization methods. It can be successfully implemented in epidemiologic studies to enhance 450K DNA methylation data preprocessing.

In our examination biological variation, we found that a standard approach in epigenome-wide analysis – minfi white blood cell composition estimation – did not correlate well with white cell counts from newborns (ρ = -0.05 for granulocytes; ρ = -0.03 for lymphocytes), but improved substantially (ρ = 0.77 for granulocytes; ρ = 0.75 for lymphocytes) in older children likely due to increasing similarity with minfi’s adult reference data as children aged. This suggests that minfi may not currently be appropriate for analysis involving newborns or young children.

Additionally, results on DNA methylation differences by sex identified 3,031 differentially methylated positions (DMPs) and 3,604 sex-associated differentially methylated regions (DMRs) on autosomes that were mostly hypermethylated in girls compared to boys. Our hits were significantly enriched for gene ontology terms related to nervous system development and behavior.

Finally, we investigated the impact of exposure to PBDEs during the highly susceptible prenatal period on DNA methylation of Mexican-American children enrolled in the Center for Health Assessment of Mothers and Children of Salinas (CHAMACOS) at birth. We identified between 6 and 48 DMRs in umbilical cord blood associated with different measures of prenatal PBDE exposure. BDE’s-47, -99 and Σ4BDE had fewer (from 6 to 9), mostly hypomethylated DMRs. Prenatal BDE-100 and -153 levels were associated with more DMRs (11 and 48 respectively) and the majority hypermethylated. The PBDE-DMRs we found were located in genes (e.g. NRBP1, CDH9, NTN1, S100A13) involved in biologically relevant functions (including axon guidance and tumor suppression) given the health effects observed in association with BDE exposure to date.