Maternal leukocyte composition, not placental pathology, underlies pre- eclampsia-associated differences in plasma cell-free DNA methylation: an integrative cell-free reduced-representation bisulfite sequencing study with tissue, trajectory, and genetic validation

Abstract Background. Plasma cell-free DNA (cfDNA) methylation is an attractive non-invasive readout in pregnancy, but the cellular origin of pre-eclampsia (PE)-associated methylation differences—placental trophoblast or maternal leukocytes—remains unresolved, and the answer determines what a cfDNA methylation biomarker would actually measure. Methods. We analysed cell-free reduced-representation bisulfite sequencing (cfRRBS) of 369 plasma samples (GSE282512), identifying differentially methylated regions (DMRs) with coverage-weighted region-level methylation and limma, adjusted for collection tube and batch. Origin was tested in five independent layers: cross-cohort placental expression meta-analysis; direct DMR testing in placental tissue (three cohorts, 105 samples); regulatory and transcription factor motif annotation; longitudinal trajectories with change-based deconvolution in two external cohorts; and genetic analyses (GoDMC cis-mQTL coverage and two-sample Mendelian randomization (MR) against pre-eclampsia and 16 blood cell and organ traits). Results. We identified 166 DMRs (105 hyper- and 61 hypomethylated), robust to collection tube and enriched in early-onset pre-eclampsia; an elastic net classifier reached a leave-one-out AUC of 0.932, reported as an internally validated candidate signature pending external validation. DMR genes showed no differential expression in placenta, and DMR CpGs in placental tissue carried smaller pre-eclampsia effects than genomic background (mean |z| 0.346 vs 0.444; permutation p = 1.0) against a positive control at |z| = 3.99. In pregnancy trajectories, placental-marker methylation rose with gestation while blood-cell markers fell, with postpartum reversal of the T-cell decline; in a small pre-eclampsia/hypertension subgroup, the neutrophil fraction was elevated at delivery and already in the first trimester (hypothesis-generating). DMR CpGs carried fewer cis-mQTLs than background (28.8% vs 37.7%, p = 2.1 × 10⁻⁹), and MR identified cg19490609 (SLC17A1) and cg15445000 as pre-eclampsia-associated loci with lineage-specific effects on blood cell counts. Conclusions. Across expression, tissue, trajectory, and genetic layers, pre-eclampsia-associated cfDNA methylation differences behave as dynamic fingerprints of maternal leukocyte composition rather than placental trophoblast lesions. cfDNA methylation signatures for pre-eclampsia should therefore be interpreted as readouts of maternal inflammatory and haematological state. These conclusions rest on a small, predominantly European-ancestry discovery cohort; clinical application requires large, independent, and ancestrally diverse validation cohorts.