The application of epiphenotyping approaches to DNA methylation array studies of the human placenta

204 placentas from three cohorts were processed for DNAme arrays in Vancouver, Canada. The three cohorts consisted of: (i) V-NORM, a normative population of term pregnancies recruited at BC Women’s Hospital (BCWH) in Vancouver, Canada (n=35), as part of a study on Epigenetics in Pregnancy Complications (EPIC); ii) V-SSRI, a population of pregnant individuals recruited in Vancouver, Canada in the 20th week of gestation (n=64), with/without clinical depression, and with/without the use of selective serotonin reuptake inhibitors (SSRIs); and (iii) QF-2011, a population of term pregnancies that were exposed to sudden-onset stress during gestation due to catastrophic flooding in the Australian state of Queensland in early January 2011 (n=105). DNA samples from all three cohorts were run on Illumina Infinium MethylationEPIC arrays in Vancouver, BC, Canada. Processing included DNA purification after extraction using the DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA), bisulfite conversion using the EZ DNAme Kit (Zymo Research, Orange, CA, USA), and hybridization to and processing of the Illumina Infinium MethylationEPIC BeadChip arrays according to the manufacturer’s protocol (Illumina, San Diego, CA, USA). Samples from the three cohorts were randomized and run in 3 array batches across 44 eight-sample chips as illustrated in Supp Figure 5. In the case of the V-SSRI and QF-2011 cohorts, sample exposure groups (SSRI exposed/non-exposed and objective flood stress high/low) and infant sex were carefully distributed across array chips (1-44) and rows (1-8) to minimize potential batch effects. DNAme data from raw IDAT files were read into R v4.2.2(54) and annotated with the Illumina Infinium MethylationEPIC v1.0 B4 Manifest. Several data quality control checks were undertaken using the R packages minfi(55,56), wateRmelon(57,58), and ewastools(59). First, each sample was assessed at 17 Illumina control probes to evaluate bisulfite conversion efficiency and array run quality; all samples passed the manufacturer-recommended thresholds at the control probes. Next, average total (methylated + unmethylated) fluorescence intensity was assessed between samples, and between array batches. All samples had similar total fluorescence, though samples run on the EPIC array in Batch 3 had slightly higher average intensities than those in Batch 1 and 2. Sample sex was assessed with the ewastools package(59), using the mean total fluorescence intensity (methylated + unmethylated) of the X and Y chromosome probes, normalized to the per-sample mean autosomal total fluorescence intensity, and was confirmed to match the clinically-reported sex of the infant in all cases. Sample genetic identity was assessed using the 59 SNP (‘rs’) probes on the EPIC array with the call_genotypes() and enumerate_sample_donors() functions (ewastools)(59). Finally, DNAme beta value density plots of all samples were visually assessed to determine overall similarity of the beta value distributions between samples, with no outliers identified. 204 placentas and 19 technical replicates were analyzed with the Illumina Infinium Methylation EPIC array. These samples came from 3 distinct collection cohorts but were processed for arrays at a single facility in Vancouver, Canada. The tools included in the PlaNET R package (https://www.bioconductor.org/packages/release/bioc/html/planet.html) were used to estimate epiphenotype variables of ancestry, gestational age, and cell type composition in these samples, and these estimates were evaluated for relationships with each other, and with technical and biological variables associated with these cohorts.

本研究共纳入来自3个独立队列的204份胎盘样本，所有样本均在加拿大温哥华完成DNA甲基化（DNA methylation, DNAme）芯片检测。3个队列分别为：
（i）V-NORM队列：为加拿大温哥华不列颠哥伦比亚省妇女医院（BC Women’s Hospital, BCWH）招募的足月妊娠正常人群，作为「妊娠并发症表观遗传学（Epigenetics in Pregnancy Complications, EPIC）」研究的一部分，样本量n=35；
（ii）V-SSRI队列：于加拿大温哥华招募的妊娠20周孕妇人群（n=64），部分受试者伴有临床抑郁症且/或使用过选择性5-羟色胺再摄取抑制剂（selective serotonin reuptake inhibitors, SSRIs）；
（iii）QF-2011队列：2011年1月初澳大利亚昆士兰州发生特大洪涝灾害后，宫内暴露于突发应激的足月妊娠人群（n=105）。

所有3个队列的DNA样本均在加拿大不列颠哥伦比亚省温哥华的实验室完成Illumina Infinium MethylationEPIC芯片杂交检测。样本处理流程包括：使用DNeasy血液与组织试剂盒（Qiagen，美国加利福尼亚州瓦伦西亚）完成提取后的DNA纯化，使用EZ DNA甲基化（DNAme）试剂盒（Zymo Research，美国加利福尼亚州奥兰治）完成亚硫酸氢盐转化，并依照制造商说明书（Illumina，美国加利福尼亚州圣迭戈）完成Illumina Infinium MethylationEPIC BeadChip芯片的杂交与后续处理。3个队列的样本经随机化处理后，分为3批芯片检测，共使用44张8样本芯片，具体实验设计详见补充图5（Supp Figure 5）。针对V-SSRI与QF-2011队列，研究将样本暴露组（SSRI暴露/非暴露、客观洪涝应激高/低）及胎儿性别均匀分配至各芯片（1-44号）及芯片孔行（1-8行），以尽可能降低潜在的批次效应。

原始IDAT格式的DNA甲基化（DNAme）数据被导入R v4.2.2(54)软件，并使用Illumina Infinium MethylationEPIC v1.0 B4注释文件完成注释。研究使用R包minfi(55,56)、wateRmelon(57,58)及ewastools(59)完成多项数据质量控制流程：首先，通过17个Illumina对照探针对每份样本进行评估，以检测亚硫酸氢盐转化效率与芯片运行质量，所有样本均达到制造商推荐的对照探针阈值标准；其次，评估样本间及芯片批次间的总荧光强度均值（甲基化信号+非甲基化信号），所有样本的总荧光强度水平基本一致，仅第3批EPIC芯片的样本平均荧光强度略高于第1、2批；随后，使用ewastools包(59)评估样本性别：以X、Y染色体探针的总荧光强度均值（甲基化信号+非甲基化信号），并以样本自身的常染色体总荧光强度均值进行标准化，最终所有样本的基因性别均与临床报告的胎儿性别一致；接着，使用ewastools包(59)中的call_genotypes()与enumerate_sample_donors()函数，依托EPIC芯片上的59个单核苷酸多态性（single nucleotide polymorphism, SNP，‘rs’位点）探针评估样本遗传同一性；最后，对所有样本的DNA甲基化β值密度图进行可视化评估，以检查样本间β值分布的整体相似性，未发现异常样本。

本研究共对204份胎盘样本及19份技术重复样本完成Illumina Infinium MethylationEPIC芯片检测。上述样本来自3个独立的采集队列，但均在加拿大温哥华的同一实验室完成芯片实验处理。研究使用PlaNET R包（https://www.bioconductor.org/packages/release/bioc/html/planet.html）中的工具，对样本的祖先血统、胎龄及细胞类型组成等表观表型变量进行估算，并评估这些估算值之间的相关性，以及它们与各队列相关的技术及生物学变量之间的关联。