Variability of genome-wide DNA methylation and mRNA expression profiles in reproductive and endocrine disease related tissues

Genome-wide association studies in the fields of reproductive medicine and endocrinology are yielding robust genetic variants associated with disease. Integrated genomic, transcriptomic, and epigenomic molecular profiling studies are common methodologies used to understand the biologic pathways perturbed by these variants. However, molecular profiling resources do not include the tissue most relevant to many female reproductive traits, the endometrium, while the parameters influencing variability of results from its molecular profiling are unclear. We investigated the sources of DNA methylation and RNA expression profile variability in endometrium (n = 135), endometriotic disease tissue (endometriosis), and subcutaneous abdominal fat samples from 24 women, quantifying between-individual, within-tissue (cellular heterogeneity), and technical variation. DNA samples (n = 96) were analyzed using Illumina HumanMethlylation450 BeadChip arrays; RNA samples (n = 39) were analyzed using H12-expression arrays. Variance-component analyses showed that, for the top 10-50% variable DNA methylation/RNA expression sites, between-individual variation far exceeded within-tissue and technical variation. Menstrual-phase accounted for most variability in methylation/ expression patterns in endometrium (P-m = 7.8 x 10(-3), Pe = 8.4 x 10(-5)) but not in fat and endometriotic tissue; age was significantly associated with DNA methylation profile of endometrium (P-m = 9 x 10(-5)) and endometriotic disease tissue (P-m = 2.4 x 10(-5)); and smoking was significantly associated with DNA methylation in adipose tissue (P-m = 1.8 x 10(-3)). Hierarchical cluster analysis showed significantly different methylation signatures between endometrium and endometriotic tissue enriched for WNT signaling, angiogenesis, cadherin signaling, and gonadotropin-releasing-hormone-receptor pathways. Differential DNA methylation/ expression analyses suggested detection of a limited number of sites with large fold changes (FC > 4), but power calculations accounting for different sources of variability showed that for robust detection > 500 tissue samples are required. These results enable appropriate study design for large-scale expression and methylation tissue-based profiling relevant to many reproductive and endocrine traits.