4.1. Methylation Analysis

Maternal T cell leukocytes were isolated from blood samples by negative selection using RosetteSep kits (Stem Cell Technologies, Vancouver, BC, Canada). To prepare specimens for methylation analysis, the 500 ng of genomic DNA was treated with bisulfite reagents using the EZ-96 DNA methylation kit (Zymo Research, Irvine, CA, USA) according to the manufacturer’s protocol. Bisulfite-converted DNA samples were then analyzed using the Illumina Infinium EPIC Methylation BeadChip array V1 [Catalog #WG-317-1003, which included the arrays and the reagents used (available from Illumina (Illumina site link https://support.illumina.com/array/array_kits/infinium-methylationepic-beadchip-kit.html) accessed on 22 June 2022). This array allows for examination of more than 850,000 CpG sites at a single nucleotide resolution. EPIC analyses were conducted at the University of Minnesota Genome Center. Data pre-processing and quality control analyses were performed in R with Bioconductor package minfi 1.40.0 and ewastools 1.7. Samples that failed 2 of the 17 metrics described in the BeadArray Controls Reporter Software Guide were excluded [42]. Further quality control and functional normalization were conducted as others have done [43,44]. We also removed far outliers of median methylation values and wrong sex prediction samples. No samples were removed due to wrong sex prediction. Two samples did not pass quality control, therefore we performed normal out of band background (Noob) within-sample correction, and removed probes with low intensities (detected in the minfi 1.40.0 package) using the threshold of p > 0.01. Next, probe type adjustment was done using Rcp in the EnMix 1.30.0 package, and batch effects were adjusted using combat function. Beta (β) values for each CpG site were calculated using the minfi 1.40.0 package. β values were defined as M/(M + U + α), where M is the total methylated signal and U is the total unmethylated signal, ranging from 0.0 to 1.0. We removed probes that were cross-hybridising using the rmSNPandCH module in the DMRcate 2.8.0 package. We estimated the relative proportion of each cell type in our heterogenous peripheral blood samples [45]. We also further excluded CpG sites that were outside the 5% and 95% threshold for very low or high methylation for all samples [46]. The CpG sites were selected based on targeted genes using the EPIC array annotation file, resulting in a total of 424 CpG sites. Preprocessing excluded 15 sites leading to a final total of 409 CpG sites that were used for differential methylation analysis. Since methylation values at CpG sites can be cell-type specific [47], we conducted cell composition analysis using a modified version of Houseman and colleagues’ method [48] and the cell type proportions were adjusted using linear regression. Visualization of genomic regions with differentially methylated sites was done using the coMET package [49]. For preprocessing and analysis, we used R version v4.0.3 and various packages including minfi v1.40.0, ENmix 1.30.0, sva v3.42.0, DMRcate v2.8.0, and Ewastools v1.7. Also, the preprocessing script is available as the monklab.methyl package which is available in https://github.com/seonjoo/monklab.methyl accessed on 22 June 2022.