Epigenetic Profiling of Human Placenta Throughout Early Gestation

Abstract

The differentiation of the placenta, especially during early gestation is important for pregnancy success. Whilst emerging evidence has shown that DNA methylation (DNAm) in placenta varies over gestation, to date, most studies have compared DNAm in a relatively short gestational age range. Little is known about the dynamics of DNAm patterns across early pregnancy from as early as 6 weeks’, and as late as 23 weeks’ gestation. This comprehensive analysis of the placental methylome will help to elucidate the previously poorly understood relationship between DNA methylation, placental development, and complications of pregnancy. The overall aim of this thesis is to characterise and interpret this relationship through the analysis of DNA methylation profiles from strictly phenotyped samples of human placenta, and matched maternal leukocytes, across early to mid-gestation, through bioinformatics analyses. All data herein were obtained using the Illumina Infinium® MethylationEPIC BeadChips (EPIC arrays). In this study, we first compared three different bioinformatics methods implemented with different algorithms for detecting differentially methylated regions (DMRs) between sample groups. Subsequent to these analyses we aimed to establish an inhouse pipeline for the quality control and analysis of EPIC array methylation data obtained from both GEO database, and from this study. The three methods used for the discovery of DMRs were bumphunter, Probe Lasso and DMRcate. After comparison of these three methods we were able to demonstrate unique advantages and disadvantages of each. Overall, DMRcate was considered the most appropriate method for the identification of DMRs in EPIC array methylation data from our placenta samples, with a better sensitivity than Probe Lasso and bumphunter methods and less false positive regions than the Probe Lasso method. Next, the established in-house pipeline was used for array data analyses. Initial unsupervised clustering using a PCA analysis of methylation data revealed several outliers within our data. These 6 samples did not cluster as expected with other placenta samples of the same gestational age. To investigate whether these outliers were caused by complicated pregnancy or technical issues, we compared the data from samples in our study with publicly available data from samples of placenta and placenta-associated tissues. Given the otherwise strong gestational age clustering observed in the PCA analysis, and the unknown pregnancy outcomes of the tissue in question, we hypothesized that the samples which failed to cluster within their gestational age group would cluster with other like samples. After preprocessing we included the public data in a new PCA analysis with results indicating that the outliers we identified were not pure placenta villous tissue, but rather these samples were a mix of both placental and maternal tissue. After assessing the quality of all placenta samples, and removing samples identified as containing maternal tissue, an epigenome wide DNAm study of placenta (n = 125) across 6-23 weeks’ gestation was performed. Placental DNA methylation changed throughout gestation, with methylation differences also found between groups up to and after 10 weeks’ gestation. Since maternal blood starts to flow into placenta at approximately 10 weeks’ gestation, these DNA methylation changes could be associated with a change in oxygen tension in the placenta. Further to the DNA methylation changes identified across early gestation, DNA methylation levels at partially methylated domains and imprinting control regions were stable in placenta across early gestation, suggesting an association with these regions and the basic function and development of the human placenta. Finally, DNA methylation changes of maternal leukocytes from matched maternal blood were investigated. We identified DNA methylation changes in maternal leukocytes associated with maternal smoking and with maternal age, and to a lesser degree we were able to identify changes in DNA methylation of maternal leukocytes that were associated with gestational age. Changes of cell proportion for maternal leukocytes were identified and a potential accelerated aging was found in pregnant women compared with non-pregnant women. These findings provide more information for real time assessment of pregnancy health using DNA methylation in maternal circulating leukocytes. In summary, the research reported here provides an insight into performing bioinformatics analyses and quality control of placental DNA methylation data obtained from EPIC array analyses. Further, this thesis adds to our understanding of placental development, health and disease through the characterisation of the DNA methylome of placenta and matched maternal leukocytes across early gestation.