Background

The placenta is a transient organ that facilitates the growth of the fetus by exchanging nutrients and metabolic biproducts with the maternal circulation. Responsible for the growth and wellbeing of the fetus, it is also a source of many pregnancy-related morbidities. Specifically, the placenta has been implicated in fetal growth restriction and hypertensive diseases of pregnancy [1,2,3]. Fetal sex is also known to impact pregnancy related morbidities [4, 5]. Male fetal sex (XY) has been associated with increased rates of preterm delivery, preterm prelabor rupture of membranes [4], as well as increased rates of gestational diabetes [6]. Male fetuses are also larger on ultrasound evaluations [7]. Given these differences in morbidity and growth profiles of pregnancies with male fetuses, further understanding is needed of gene expression and the upstream regulatory mechanisms responsible for sexually dimorphic placental phenotypes, particularly in early pregnancy.

The first trimester is a critical timepoint in placenta establishment (placentation). The invasion of trophoblast cells into the uterine wall and subsequent remodeling of spiral uterine arteries is thought to be involved in the progression of placental dysfunction including pregnancy loss, growth restriction, and hypertensive disease of pregnancy [8]. Further understanding the first trimester placenta is therefore essential in understanding, and some day preventing, pregnancy related morbidities.

We and others have found sexually dimorphic gene expression in human placentas using RNA-sequencing (RNA-seq) [9,10,11]. Upstream regulators of gene expression such as DNA methylation at CpG dinucleotide sites is also important to understand and can vary by organs and tissues [12]. Other studies have found sex differences in methylation patterns and gene expression in term placentas [13, 14]. Inkster et al. looked at 343 term placentas and found 162 CpG sites that were differentially methylated by sex [13]. Andrews et al. looked at 532 term placentas (overlapping Inkster’s cohort) and found 5,212 CpG sites differentially methylated by sex [14]. First trimester placenta in healthy and normal karyotype pregnancies has been more difficult to study, but it is essential to understand the methylation profile of normal placentation which impacts the rest of pregnancy. First trimester placenta tissue was collected from discarded sample after chorionic villus sampling (CVS), a prenatal diagnostic test performed in late first trimester most commonly for advanced maternal age. Our cohort is composed of singleton, normal karyotype pregnancies conceived without fertility treatments, continued until live births. The goal of the present study is to evaluate human first trimester placenta for sex differences in DNA methylation and to identify which methylation changes alter gene expression.

Methods

Subject recruitment and sample procurement

For this sex differences analysis, we used the 56 spontaneous subjects from the original n = 138 Spontaneous/Medically Assisted/Assisted Reproductive Technology (SMAART) DNA methylation cohort, where spontaneous refers to pregnancies conceived without fertility treatments [15]. Subjects who were already undergoing chorionic villus sampling (CVS) for medical reasons were recruited with informed consent on the day of their clinical visit. Research involvement did not affect the amount of tissue collected from the patient or other clinical decisions. Chorionic villi (first trimester placenta) tissue discarded after prenatal diagnostic testing was placed in 250 µL of RNAlater RNA stabilizing solution (QIAGEN) and stored at − 80 °C until further processing for research. Genomics DNA and total RNA were isolated from the same tissue samples with the AllPrep DNA/RNA Mini Kit (QIAGEN) and utilized for DNA methylation and total RNA-seq projects as previously described [15, 16]. All protocols were performed in accordance with the institutional review board’s guidelines at the Cedars-Sinai Medical Center under IRB Protocols Pro00006806 and Pro00008600. All pregnancies were singleton, normal karyotype, and resulted in live births. The gestational ages range at CVS was 71–99 days (mean 82 ± 6 days standard deviation). Pregnancies delivered at 39 ± 2 weeks gestation, all full term except two preterm births in the male cohort.

Analysis of demographic data

Demographics data collected included maternal/paternal age, race, ethnicity, maternal BMI, fetal race/ethnicity, maternal conditions such as hypertension, diabetes, thyroid disease, and other medical conditions. Pregnancy outcomes evaluated included gestational age at time of CVS, crown rump length (CRL) at time of CVS, gestational age at delivery, birthweight in grams, and mode of delivery. Pregnancy complications included hypertension, diabetes, coagulation disorders, placenta previa, placenta abruption, and other placental problems. Demographics analysis was performed to compare the cohorts by fetal sex. Means and standard deviations were reported for continuous variables and proportions were reported as percentages. The student’s T-test was used for normally distributed data and Chi-square test was used to compare categorical variables. Demographics were performed for the DNA methylation cohort (n = 56) and the RNA-seq cohort (n = 74).

DNA methylation data and filtering

The Infinium MethylationEPIC BeadChip array was filtered to 743,461 autosomal probes, including 741,145 CpG sites and 2316 CpH sites, with the following filtering steps. Unreliable or masked probes were dropped using Dr Wanding Zhou’s EPIC array manifest file “EPIC.hg38.manifest.tsv.gz” column MASK_general = TRUE [17], release 9/9/2018 available at (https://zwdzwd.github.io/InfiniumAnnotation). Non-autosomal probes were dropped (chrX, chrY, mitochondrial sites, and NA). Probes overlapping a single nucleotide polymorphism were dropped, identified using probeType = “rs”. Probe annotations such as gene context were added from the MethylationEPIC v1.0 B5 manifest (Illumina), available at (https://support.illumina.com/array/array_kits/infinium-methylationepic-beadchip-kit/downloads.html).

Principal components analysis

For DNA methylation, principal components analysis was performed with R package eigen v4.2.0 for spectral decomposition of a matrix.following methods for Eigenstrat software [18, 19]. The input was an M-values matrix of all EPIC array probes (for fetal sex verification only) or 743,461 autosomal probes (filtered as described above). For RNA-sequencing, principal components analysis was performed with R packages DESeq2 v1.4.0 (to normalize counts for sequencing depth and perform a variance stabilizing transformation), limma v3.56.1 (to batch correct for two sequencing runs with function removeBatchEffect), and base R function prcomp with the top 500 most variable genes [20, 21]. Scatter plots were created and combined with R packages ggplot2, cowplot, and pdftools.

Differentially methylated probes (DMPs), regions (DMRs), and gene ontology

The Infinium MethylationEPIC BeadChip array results for 56 placentas (25 female, 31 male) at 743,461 probes were analyzed in a generalized linear model to compare female versus male methylation (M-values) without covariates, using the limma v3.56.0 R package function lmFit [21]. The P-values were adjusted for multiple comparisons using Benjamini-Hochberg’s False Discovery Rate method to generate “FDR” values [22]. Probes were significantly sexually dimorphic at FDR < 0.05, equivalent to P < 1 × 10–5 in this study. Enriched gene ontology (GO) terms from differentially methylated probes (DMPs) at FDR < 0.05 were identified using R package missMethyl v1.34.0, which accounts for the number of probes per gene and the number of probes analyzed total [23]. Additionally, probes reached a stricter “genome-wide significance” threshold if they met unadjusted P-value < 9 × 10–8, the threshold recommended for the EPIC BeadChip array [24]. Quantile–quantile plots were created using R package qqman v0.1.9 [25]. Manhattan plots were created with the most significant 5000 probes by unadjusted P-value with R packages ggplot2, ggrepel, and cowplot. The x-axis uses the chromosome and GRCh38 start site from the EPIC manifest and the y-axis separates probes by either − log10(P-value) or limma output “logFC” describing direction of methylation. To identify differentially methylated regions (DMRs) between female and male placentas, methylation (M-values) for 56 placentas at 743,461 probes were input into DMRcate v2.14.0 functions cpg.annotate and dmrcate [26, 27]. We used the default values for Gaussian kernel bandwidth for smoothed-function estimation (lambda = 1000) and scaling factor for bandwidth (C = 2).

RNA-sequencing and sex differences analysis

We utilized our largest dataset with total RNA-seq of spontaneous and healthy first trimester human placentas, n = 74 (34 females and 40 males), available at The National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO) under accession GSE215421 [16]. The GSE215421 data includes re-sequenced samples from bulk total RNA-seq projects under accessions GSE109082 (n = 39) and GSE131874 (n = 4) [9, 28]. For each placenta sample, the RNA and DNA were isolated on the same day using the two-column kit, AllPrep DNA/RNA Mini Kit (QIAGEN), as previously described [9, 15, 16, 28, 29]. The total RNA and genomic DNA elutions were stored at − 80 °C until use for RNA-seq or the Illumina EPIC array. The counts matrix for GSE215421 was filtered to compensate for imperfect ribosomal RNA (rRNA) depletion during library preparation. Ensembl Gene IDs with gene biotype “rRNA” in the Ensembl release 91 annotations were dropped for all samples during bioinformatics analysis. Ensembl Gene IDs with zero counts across all n = 74 samples were also dropped. The final counts matrix contained 74 placentas and 55,319 genes. Differential expression analysis was performed using R package DESeq2 v1.40.0 to fit each gene into a negative binomial generalized linear model to compare females versus males, batch adjusted for two sequencing runs [20, 30]. The resulting Wald test P-values were adjusted for multiple comparisons using the Benjamini–Hochberg False Discovery Rate (FDR) procedure [22].

Integrated analysis comparing gene expression and DNA methylation

DMPs with sex differences at P < 0.01 were matched to corresponding genes using the gene symbols and Ensembl Transcript IDs in the EPIC manifest (GENCODE release 12), cross-referenced with the Ensembl BioMart server using R package biomaRt with GRCh37 and GRCh38 (Ensembl release 91) to output Ensembl Gene IDs corresponding to specific probeIDs. These Ensembl Gene IDs were used to merge RNA-seq results (n = 74 placentas) to DNA methylation results (n = 56 placentas), creating a dataset of 35,862 matches (12,560 unique probeIDs and 7737 unique genes). Using 51 placentas with data from both experiments, gene expression was correlated to DNA methylation using a generalized linear model with R version 4.3.3 function lm and model “gene expression ~ methylation M values”. Specifically, gene expression was defined as counts normalized for sequencing depth (“baseMeans” from DESeq2), transformed with log2(x + 1) and then corrected for batch using limma::removeBatchEffect.

Results

Demographics

Two largely overlapping placenta cohorts were used for first trimester sex differences analyses (Fig. 1A). Demographics analyses were performed to compare each cohort by fetal sex (Table 1, Additional file 1). The methylation cohort included 56 pregnancies (25 carrying a female fetus, 31 a male fetus). The methylation cohort subjects were mostly Caucasian non-Hispanic but had no differences in maternal/paternal age, race, ethnicity, BMI, maternal medical conditions, pregnancy outcomes, or pregnancy complications (Supplemental Table S1, Table 1). The RNA-seq cohort included 74 pregnancies (34 carrying a female fetus, 40 a male fetus). RNA-seq cohort subjects were mostly Caucasian, non-Hispanic. However paternal ethnicity was more often non-Hispanic (93%) in the male fetus cohort than in the female fetus cohort (85%) (p = 0.041). No differences were noted in maternal age, race, ethnicity, BMI, maternal medical conditions, pregnancy outcomes, or pregnancy complications (Supplemental Table S2, Table 1). Principal components analysis showed that when sex chromosomes were included, samples separated by fetal sex and not by other demographics variables (Additional file 2). When sex chromosomes were excluded, fetal Hispanic samples somewhat clustered in DNA methylation data, but were still located within the larger group (Additional file 2).

Fig. 1
figure 1

Sample cohorts and quality control. A Summary of the first trimester placenta cohorts and analyses. F = females, M = males. B Principal components analysis for the DNA methylation cohort. The pre-filtered dataset confirms fetal sex. The filtered dataset no longer shows clustering by sex. C Quantile–Quantile plot for the differentially methylated probes (DMP) analysis. Points are individual methylation sites (probes). Red line at Pobserved = Pexpected

Full size image
Table 1 Brief demographics and pregnancy outcomes
Full size table

DNA methylation analysis quality control

DNA methylation of first trimester human placentas was measured using the EPIC array. The fetal sex of processed placenta samples was verified experimentally using a principal components analysis starting with all sites on the EPIC array, including chromosomes X and Y which segregate samples by fetal sex (Fig. 1B, Additional file 2). However, sex chromosome DNA methylation cannot be accurately compared between female and male samples due to differences in copy numbers and phenomena such as X chromosome inactivation which involves wide spans of DNA and histone methylation across one copy of the female chromosome X, but not the other. Differentially methylated probes (DMPs) analysis was performed to identify sexually dimorphic sites using only autosomal chromosomes, without other sources of unreliability such as masked probes or probes at single nucleotide polymorphisms. Principal components analysis of the final sites shows that placental samples no longer cluster by fetal sex without the sex chromosomes, indicating that global placental DNA methylation is largely similar between sexes (Fig. 1B, Additional file 2). The Quantile–Quantile plot of P-values shows a lambda value of 1.1249 indicating low inflation (Fig. 1C).

Differentially methylated probes (DMPs) affected by fetal sex

We identified 151 autosomal differentially methylated probes (DMPs) significantly altered between female and male placentas at FDR < 0.05 (here equal to P < 1 × 10–5), with a slight majority of 59% (89 DMPs) hypermethylated in females, and 62 hypermethylated in males (Additional file 3, Additional file 4). Among these 151 DMPs, there were 29 CpG sites that reached genome-wide significance (P < 9 × 10–8), with 10 hypermethylated in females and 19 hypermethylated in males (Table 2). The most significantly female hypermethylated probe was cg00167275, located in the 5′ untranslated region (5′UTR) and first exon of glutamate dehydrogenase 1 (GLUD1), and before the transcription start site of double-strand break regulator FAM35A (also called SHLD2). The most significantly male-hypermethylated probe was cg17612569, located before the transcription start site of transcription factor subunit GABPA and at different locations of ATP synthase subunit ATP5J transcripts (5′UTR, first exon, and coding region). No CpH (CpA, CpC, or CpT) probes reached FDR < 0.05 significance.

Table 2 Differentially methylated probes (DMPs) in first trimester placenta that reach genome-wide significance (P < 9 × 10–8)
Full size table

Manhattan plots were created to visualize site significance across the genome (Fig. 2A). Chromosome 5 showed a cluster of 13 probes mapped to the promoter region of zinc finger protein 300 (ZNF300) gene, 200 bp or 1500 bp upstream of the ZNF300 transcription start site (Additional file 4). Twelve sequential ZNF300 probes were significantly hypermethylated in male placentas with strong trends, including 9 in a CpG island and 4 in a south shore (downstream of the CpG island), and only the final 13th probe (further downstream from the rest) showed female-hypermethylation. Additional clusters of methylation differences were evident across the genome, though not all reached genome-wide significance (Fig. 2B).

Fig. 2
figure 2

Manhattan plots of the top probes by P-values. Chromosomes alternate colors and magenta points highlight gene ZNF300. A The top 5000 probes by P-values separated by significance. Higher − log10(P) values means probes are more significantly sex different. Probes reach FDR < 0.05 (P < 1 × 10–5) significance above the solid green and genome-wide significance above the dashed green line at P = 9 × 10–8. B Only the significant probes (FDR < 0.05), plotted for direction of methylation: positive values indicate hypermethylation in female placentas, and negative values indicate hypermethylation in male placentas. C Top 10 enriched gene ontology terms for the DMPs, sorted by P-value

Full size image

Gene ontology analysis was performed using the 151 DMPs, resulting in 248 enriched gene ontology terms at P < 0.05 (Additional file 5). The five most significant terms were “endothelial cell–cell adhesion” (P = 0.000123), “presynaptic active zone” (P = 0.000694), “intracellular non-membrane-bounded organelle” (P = 0.00204), “non-membrane-bounded organelle” (P = 0.002047), and “S100 protein binding” (P = 0.002195) (Fig. 2C).

Differentially methylated regions

A detailed analysis of differentially methylated regions (DMRs) was performed to confirm findings on the Manhattan plot. We identified eleven DMRs on autosomal chromosomes, eight regions hypermethylated in female placentas and three regions hypermethylated in male placentas (Table 3). The most significant DMR was composed of 14 probes in ZNF300, hypermethylated in males. The longest DMR between female and male placentas was male-hypermethylated, located in genes GABPA/ATP5J, and composed of 23 probes across 1543 base pairs. The most significant female-hypermethylated DMR was located in ZNF311 with 16 probes. Another zinc finger domain-containing transcription factor, ZNF175, appeared to be among the most male-hypermethylated genes (Fig. 2B) and was confirmed as a significant male-hypermethylated DMR. Chromosome 14 had three DMR, all female-hypermethylated, including two forkhead box transcription factor genes (FOXG1 and FOXA1). Chromosome 6 had two DMRs (ZNF311 and C6orf47/C6orf47-AS1), also both female-hypermethylated. Other DMRs were located on separate chromosomes.

Table 3 Differentially methylated regions (DMRs) between female and male placentas at first trimester
Full size table

Comparison to RNA-seq sex differences

We previously found ZNF300 gene expression significantly upregulated in female placenta, compared to male, in bulk total RNA-seq of n = 39 [9]. We combined additional bulk total RNA-seq data to perform a new sex differences analysis on n = 74 first trimester human placenta samples. Differential expression analysis identified 152 differentially expressed genes (DEGs) between females and males at FDR < 0.05, with 48 female-upregulated and 104 male-upregulated (Additional file 6). The most female-upregulated gene was XIST (167-fold higher in females, FDR = 0), an X-linked gene involved in X-chromosome inactivation. The most male-upregulated genes were Y-linked protein coding genes RPS4Y1, DDX3Y, UTY, EIF1AY, ZFY, USP9Y, KDM5D, and non-coding TTTY15 (Fig. 3A). By fold change, the most sexually dimorphic autosomal genes were chemokine ligand 9 (CXCL9, 6.3-fold higher in males, FDR = 0.0021) and insulin-like growth factor binding protein 1 (IGFBP1, 6.27-fold higher in females, FDR = 0.0090) (Fig. 3B). By significance, the most sexually dimorphic gene was ZNF300 (2.12-fold higher in females, FDR = 2.98 × 10–8).

Fig. 3
figure 3

RNA-seq sex differences and correlation to DNA methylation. A, B Volcano plots for RNA-seq sex differences in placenta, with dashed lines at P = 0.05 and FDR = 0.05 (these lines overlap in A due to the y-axis range). C Violin plots comparing female and male gene expression: log2(baseMeans) adjusted for batch. Horizontal lines indicate the mean. Sex differences: *P < 0.05, **FDR < 0.05. D Manhattan plot of correlation results with the overlapping cohort (n = 51), comparing gene expression (“expr”) to DNA methylation (“methyl”). Each point is a probe. Probes located in the same gene are grouped with rectangles. Non-significant (“NS”, RNA-seq P ≥ 0.05) genes are semi-transparent. Magenta color highlights three genes with correlation FDR < 0.05 and RNA-seq FDR < 0.05 (FAM228A, ZNF300 and CSMD1)

Full size image

DMPs at FDR < 0.05 were matched to the RNA-seq results to identify overlapping genes of interest (Additional file 7). Of the 151 DMPs, 42 were located in genes that showed at least nominally significant sex differences in RNA-seq (P < 0.05), and 17 DMPs were located in two genes significant after adjustment for multiple comparisons (FDR < 0.05 in the RNA-seq): ZNF300 which was consistently male-hypermethylated, and tumor suppressor gene CUB and Sushi multiple domains 1 (CSMD1) which had gene expression 1.78-fold higher in males but inconsistent CpG hypermethylation directions in the gene body (none in the promoter region). Probe cg05994094 was genome-wide significant and hypermethylated in females and its corresponding gene, coiled-coil domain-containing protein 178 (CCDC178), had 1.46-fold higher gene expression in males at P < 0.05 (Fig. 3C), consistent with expectations. Three genes from the DMR region analysis were associated with sexually dimorphic gene expression: ZNF300, ZNF311, and CCDC178.

To directly access which DNA methylation sex differences affect gene expression, we integrated the datasets. DNA methylation probes with sex differences at P < 0.01 (12,560 probes) were matched to overlapping or nearby genes, resulting in 35,862 gene-probe matches for a correlation analysis using the 51 overlapping placentas (Additional file 8). Of these, 3083 (5.7%) and 396 (0.72%) gene-probe matches reached correlation significances of P < 0.05 or FDR < 0.05, respectively, indicating that methylation sex differences alter expression of a small subset of placental genes. When limited to sexually dimorphic genes identified earlier in the full cohort RNA-seq, 152 gene-probe matches (146 unique probes) show significant methylation correlation to gene expression in first trimester placenta (P < 0.01 DNA methylation, P < 0.05 RNA-seq, and FDR < 0.05 correlation). Of these, three genes met FDR < 0.05 in the RNA-seq: FAM228A, ZNF300, and CSMD1 (Fig. 3C), indicating strong sex differences in the placental transcriptome. FAM228A showed a significant inverse (negative) correlation, indicating that female-hypermethylation causes male-upregulated expression (Fig. 3D). ZNF300 showed an inverse correlation as well. CSMD1 showed a positive correlation for the most significant 12 probes (correlating at FDR < 0.05), suggesting that hypermethylation at these sites increases gene expression. The correlation analysis also highlighted proteasome subunit alpha type 8 (PSMA8) which showed a significant inverse correlation in eleven FDR < 0.05 probes. ZNF175, a sexually dimorphic DMR, showed strong inverse correlation between placental DNA methylation and gene expression (FDR < 0.05), though the gene was not significantly sex different in this RNA-seq.

Methylation sex differences across studies

The 151 DMPs were cross-referenced across different studies to identify which are uniquely sex different in first trimester placenta, which show placental sex differences later in gestation, and which also show sex differences in the whole blood of neonates or adults (Figs. 4, 5). All studies were compared at significance threshold FDR < 0.05, and non-significant sites were examined for directional trends.

Fig. 4
figure 4

Comparison with other sex differences studies, chromosomes 1–7. DMPs FDR < 0.05 in this study are compared to DMPs in Andrews et al., Santos et al., Inkster et al., and Grant [13, 14, 32, 33]. DMPs are sorted by chromosome position and labeled “genes_chromosome_probeID”. RNA-seq sex differences in this study are highlighted for significant (FDR < 0.05) and nominally significant (P < 0.05) genes using asterisks, with color bars for genes with > 2 DMPs

Full size image
Fig. 5
figure 5

Comparison with other sex differences studies, chromosomes 8–22. Continuation of Fig. 4, with similar labeling

Full size image

Our DMPs showed good consistency with Andrews et al. [14] sex differences analysis of human placenta at different gestational ages, including an independent cohort of 22 first trimester placentas at 8–14 weeks. We found novel results for 60 DMPs (40%) due to the higher number of methylation sites in the EPIC array, compared to the older 450 k array. Of 91 (60%) DMPs remaining, methylation directions were consistent across gestation: 84, 83, 66, and 86 DMPs matched in the first, second, third trimester (preterm), and full term placenta cohorts, respectively. At FDR < 0.05, we identified 3, 16, 6, and 71 DMPs with shared significance, respectively. The 3 DMPs significant in both first trimester cohorts were cg26605406 (ZNF175), cg20696478 (ZNF175), cg17612569 (GABPA/ATP5J), all male-hypermethylated genes significant in our region analysis (Table 2). We identified more significant matches with Andrews’ second trimester and full term cohorts, possibly because our cohort is at late first trimester (10–14 weeks). Inkster et al.’s [13] full term analysis includes some overlap with Andrews’ full term cohort and showed similar results, with 79 DMPs in the same direction (61 DMPs significant) and 66 DMPs with no available data. We also examined a sex-stratified comparison of second versus third trimester placenta [31], but no DMPs overlapped.

We also compared our 151 DMPs to two EPIC array studies of preterm placenta and whole blood. In Santos et al.’s [32] sex differences analysis of the Extremely Low Gestational Age Newborn (ELGAN) cohort (23–27 weeks, n = 1171), we found FDR < 0.05 matches for 51 DMPs in preterm placenta (34%) and 22 DMPs in neonatal peripheral blood (15%). In Grant et al.’s [33] adult blood analysis, only 8 DMPs (5%) matched, the lowest percent of all compared. One CpG methylation site was significantly male-hypermethylated at FDR < 0.05 in all studies including adult blood: cg17612569 (GABPA/ATP5J, chr21) (Fig. 5). Another near-unanimous methylation site was female-hypermethylated cg06710937 (chr13) near genes RFESDP1, LINC00621, BASP1P1, and IPMKP1, none of which were sex different in our placental RNA-seq.

Placental female-hypermethylation resulted in higher gene expression for males at several genes including: GABA-receptor subunit GABRB1 (chr4), transmembrane glycoprotein embigin (EMB, chr5), transcription factors ZNF311 (chr6) and GTF3C6 (chr6), methionine sulfoxide reductase A (MSRA, chr8), the cell cycle regulating transcription factor TFDP1 (chr13), read-through product NEDD8-MDP1 (chr14), fibroblast growth factor 7 (FGF7, chr15), CCDC178 (chr18). Of these transcriptome-altering DMPs, MSRA female-hypermethylation is uniquely significantly at first trimester placenta. TFDP1 and EMB methylation may also be unique to first trimester placenta, but information is lacking across gestation. CSMD1 DMPs (chr8) had inconsistent methylation directions, mostly male-hypermethylated except for female-hypermethylated cg12269374 located in the gene body. Two DMPs showed female-hypermethylation and female-upregulated gene expression: cg16446577 at peptidase inhibitor PI16 (chr6) which was consistently female-hypermethylated across gestation, and cg18685455 (chr6) which corresponds to AHI1 (involved in vesicle trafficking) as well as one of several regions encoding Y_RNA and is thus difficult to interpret.

Placental male-hypermethylation rarely resulted in higher gene expression for female placentas, only seen for our most significant result ZNF300 (chr5), and for long non-coding gene AP001065.15 also called LINC02575 (chr21). ZNF300 was consistently male-hypermethylated across gestation in placenta, but mostly not blood except for cg08580836 which was significant in neonatal blood [32]. AP001065.15 (cg20877594) was also male-hypermethylated in preterm placenta and matching neonatal blood [32]. Three other genes were male-hypermethylated in placenta but showed nominally higher gene expression in males: RAS oncogene family member RAB7A (chr3), CSMD1 (chr8) discussed above, and nuclear lamin B2 (LMNB2, chr19). RAB7A was male-hypermethylated across gestation in placenta, whereas LMNB2 may be unique to first trimester placenta.

Discussion

Global DNA methylation sex differences with associated gene expression differences are small in the first trimester placenta, however we identified 151 differentially methylated probes (DMPs) and 11 differentially methylated regions (DMRs) in first trimester placenta. Since we conducted DNA methylation and RNA-seq analysis on the same cohort, we were able to determine if changes in DNA methylation coincided with changes in gene expression in the same samples. Of interest, most significant DMPs and DMRs do not cause gene expression changes in the genes where they are localized, with only 42 DMPs and 3 DMRs associated with sexually dimorphic expression. This suggests that upstream mechanisms other than DNA methylation may play a role in sexually dimorphic gene regulation in the first trimester placenta during placentation.

Three of the 11 DMRs were associated with zinc finger protein genes (ZNF300, ZNF311, ZNF175), encoding DNA-binding proteins that often function as transcription factors. Our most significant result was male-hypermethylated ZNF300, found significantly sexually dimorphic in multiple placental DNA methylation and RNA-seq studies across gestation, including our previous RNA-seq study as well as other independent studies [9, 10, 13, 14, 34]. Placenta sex differences studies where it was not significant either could not evaluate (P-value = no value) or did not include it in the final analysis due to low gene expression in placenta [10, 11, 28]. ZNF300, though poorly expressed, may contribute to sexually dimorphic pregnancy outcomes. Hypermethylation of the ZNF300 gene in placenta is associated with intrauterine growth restriction in twin pregnancies at term [35], as well as preeclampsia at term [36]. ZNF300 is also involved in cellular differentiation, with knockdown of ZNF300 leading to increased proliferation but decreased differentiation of megakaryocytes and erythrocytes from a progenitor cell line [37]. ZNF300 therefore could be involved in adequate differentiation of placenta cells leading to proper establishment of the maternal-placental unit, though more research is needed. ZNF175 was also significantly male-hypermethylated at first trimester here and a previous smaller study, but not significant in the second trimester and results were mixed at term [13, 14]. Although gene expression trended weakly higher in females in our first trimester (this study) and Olney et al.’s third trimester total RNA-seq [31], ZNF175 only reached P < 0.1 and was not significant after adjustment for multiple comparisons in either trimester’s total RNA-seq. However, in our larger independent cohort analysis of messenger RNA (mRNA)-seq placental sex differences, we do find ZNF175 significantly 1.4-fold higher in female first trimester (FDR = 5.35E−6), but not significant in third trimester (P > 0.2) [34]. This suggests that ZNF175 sex differences in placental gene expression exist in early pregnancy but are more subtle, possibly only seen in polyadenylated transcripts or larger datasets. Differential methylation of ZNF175 may play a role in sexually dimorphic placental disease states. ZNF175 is hypomethylated and overexpressed in placenta of patients with preeclampsia, compared to controls [38,39,40]. Fetal sex appears to have a gestational age-dependent effect on preeclampsia risk, with female fetal sex more associated with preterm preeclampsia [6, 41], and otherwise conflicting results depending on study population [6, 42,43,44]. ZNF311 was the most significantly female-hypermethylated DMR and showed significantly higher gene expression in males at first trimester, but loses sexual dimorphism by term [13, 14]. There is limited information regarding the role of ZNF311. It is an oncogene associated with tumor infiltrating immune cells in gliomas [45], and thus may play a role in the immune milieu at the maternal–fetal interface.

Genes with differentially methylated CpG sites and corresponding differential expression, unique to our dataset in the first trimester placenta, include EMB, LMNB2, NEDD8-MDP1, and TFDP1. The genes EMB and TFDP1 contained female-hypermethylated sites and showed male upregulated gene expression in first trimester placenta. Both EMB, which encodes the Embigin transmembrane glycoprotein, and TFDP1 which encodes a member of a family of transcription factors which control cell cycle regulators, are involved in cell growth which may play a role in sexual dimorphic outcomes in fetal growth [46]. Additionally, knockout of TFDP1 with CRISPR/Cas9 reduces the invasion phenotype of HTR8 cells, a model of placental extravillous trophoblasts [47], and thus TFDP1 alterations may directly affect placentation. LMNB2 encodes lamin B2, a nuclear-localized protein involved in cell proliferation and apoptosis [48]. We found LAMNB2 gene to be male-hypermethylated with higher gene expression in females. Abnormal expression of LMNB2 has been identified in multiple types of cancers, likely due to the impact on cellular proliferation and DNA damage repair [48]. Differential gene expression among the sexes may lead to differences in nuclear membrane stability, particularly in syncytiotrophoblasts, and cellular proliferation leading to alterations in placental function.

In first trimester, CSMD1 gene was male-hypermethylated and showed a positive correlation with gene expression. CSMD1 coverage in other placental array studies is poor and inconsistent. However, consistent with our results in first trimester, Gong et al.  who used term placenta and bisulfite sequencing found male-hypermethylation of CSMD1, then confirmed higher gene expression in males [49]. Furthermore, they found placenta-specific CSMD1 gene regulation, with DNA methylation only sexually dimorphic in placenta but not eight other somatic tissues, and a placenta-specific transcript that starts downstream of the canonical CSMD1 transcription start site. The DMPs we identified in first trimester placenta are still within the gene body and not the promoter, indicating an alternative gene regulation method than the typical gene suppression by promoter methylation. CSMD1 gene expression has multiple tumor-suppressing functions through the SMAD pathway (promotes apoptosis, reduces cell proliferation and migration) [50], likely contributing to trophoblast cell regulation during placentation in first trimester. Notably, a polymorphism in CSMD1 (rs2924725) affects COVID-19 hospitalization risk differently in females and males [51]. This suggests that the function of CSMD1 is vulnerable to site-specific differences (either by DNA methylation or polymorphism) and that the downstream effect is further influenced by sex.

The most significant pathways of the DMPs included two cell–cell adhesion pathways. Cell–cell adhesion is an essential pathway for placentation which involves many categories of adhesion molecules (e.g. integrins) involved in within-placenta cell adhesion such as establishment of the syncytiotrophoblasts layer [52], as well as maternal–fetal interactions such as the development of the spiral arteries necessary for oxygen and nutrient transfer during pregnancy [53, 54]. Endothelial and epithelial cell–cell adhesion regulation specifically is essential for establishment of placental vascular networks, for example when trophoblasts (epithelial) invade uterine spiral uterine arteries (which begin with an endothelial lining) and restructure the cell environment [55, 56]. Defects or subtle alterations in cell–cell adhesion can contribute to the development of preeclampsia [54].

This placenta DNA methylation sex differences study has several strengths, including a larger single cohort and a primary focus on characterizing first trimester placenta. The study was designed to minimize confounding variables that affect methylation measurements. All placenta samples in this study were normal karyotype, collected within a short window of gestational age (10–14 weeks) in the same hospital, with DNA and RNA simultaneously co-isolated from the same tissue per subject, and DNA methylation measurements performed using the same array by the same facility for all samples. Similarly, RNA-seq was performed in the same facility for all samples. Previous placenta DNA methylation sex differences studies primarily focused on term placenta, with study designs and sample filtering most rigorous for term placenta [13, 14]. Due to challenges acquiring early pregnancy placenta, when first trimester samples were included in previous studies, they were analyzed from combined datasets of smaller studies performed on different methylation arrays (450k or EPIC), with wider allowances for gestational age (8–14 weeks) or no gestational age specified beyond trimester, and inconsistent tissue processing [14, 57, 58]. These allowances were necessary to combine enough first trimester samples for preliminary analysis, since first trimester samples are very limited, but this follow up study was necessary. This work provides improved datasets and analyses of placental sex differences at first trimester, specifically within the window of placentation which is a critical timepoint for pregnancy development, for both DNA methylation and total RNA-seq. Furthermore, all pregnancies resulted in live births.

We acknowledge study limitations such as a lack of third trimester samples, advanced parental ages, and limited racial diversity (most subjects were Caucasian). Placental DNA methylation sex differences are already well-characterized at later pregnancy and we benefited from prior work. Advanced parental ages are due to the source of first trimester placenta which was leftover tissue from prenatal diagnostic testing. Advanced maternal age does not significantly affect global DNA methylation in first trimester placenta, though individual regions are altered [59]. Parental age, race and ethnic diversity is limited by biobank availability, though we have shown here and in another mRNA-seq study that placenta samples did not separate by parental ages or race [60]. Fetal ethnicity (Hispanic versus non-Hispanic) may affect DNA methylation and future studies with larger cohorts are needed to explore this.

Perspective and significance

This study is the most comprehensive analysis of fetal sex and its effect on human placental DNA methylation and gene expression. Prior studies show that sex differences affect pregnancy development and outcomes, but placenta studies at early pregnancy are limited and often subject to batch effects. We characterized DNA methylation using a single unified cohort and the most current and comprehensive human DNA methylation array available (the Infinium MethylationEPIC BeadChip array, with twice as many sites as the Infinium HumanMethylation450k BeadChip array). We performed differential gene expression analysis with an overlapping cohort and identified genes with DNA methylation differences and corresponding gene expression changes. We compared our results to other sex differences studies and found that placental sex differences across gestation were largely consistent in direction of hypermethylation, though not always in significance. Several genes were also significant in term placenta studies, showing that fetal sex alters placental epigenetics and gene expression early and that some changes are sustained until delivery. Comparisons with blood methylation studies found that placental sex differences best matched neonatal blood, but not adult blood, further indicating that some sex differences are age-dependent. We also identified genes with DNA methylation sex differences uniquely present in first trimester placenta, not seen later in gestation or in blood. These unique differences may be important for placentation and other processes of early pregnancy. Future work is needed to understand the specific functions of identified genes within the context of early pregnancy and placenta, as most of these genes have been studied in other organs in non-pregnant subjects.