Higher Maternal Body Mass Index Is Associated With Lower Placental Expression of EPYC: A Genome-Wide Transcriptomic Study

Abstract

Context

Elevated body mass index (BMI) in pregnancy is associated with adverse maternal and fetal outcomes. The placental transcriptome may elucidate molecular mechanisms underlying these associations.

Objective

We examined the association of first-trimester maternal BMI with the placental transcriptome in the Gen3G prospective cohort.

Methods

We enrolled participants at 5 to 16 weeks of gestation and measured height and weight. We collected placenta samples at delivery. We performed whole-genome RNA sequencing using Illumina HiSeq 4000 and aligned RNA sequences based on the GTEx v8 pipeline. We conducted differential gene expression analysis of over 15 000 genes from 450 placental samples and reported the change in normalized gene expression per 1-unit increase in log₂ BMI (kg/m²) as a continuous variable using Limma Voom. We adjusted models for maternal age, fetal sex, gestational age at delivery, gravidity, and surrogate variables accounting for technical variability. We compared participants with BMI of 18.5 to 24.9 mg/kg² (N = 257) vs those with obesity (BMI ≥30 kg/m², N = 82) in secondary analyses.

Results

Participants’ mean ± SD age was 28.2 ± 4.4 years and BMI was 25.4 ± 5.5 kg/m² in early pregnancy. Higher maternal BMI was associated with lower placental expression of EPYC (slope = −1.94, false discovery rate [FDR]-adjusted P = 7.3 × 10⁻⁶ for continuous BMI; log₂ fold change = −1.35, FDR-adjusted P = 3.4 × 10⁻³ for BMI ≥30 vs BMI 18.5-24.9 kg/m²) and with higher placental expression of IGFBP6, CHRDL1, and CXCL13 after adjustment for covariates and accounting for multiple testing (FDR < 0.05).

Conclusion

Our genome-wide transcriptomic study revealed novel genes potentially implicated in placental biologic response to higher maternal BMI in early pregnancy.

More than two-thirds of women of reproductive age in the United States are affected by overweight or obesity (1). Elevated body mass index (BMI) influences the risk of maternal health outcomes in pregnancy such as pre-eclampsia (PE) and gestational diabetes mellitus (GDM), as well as many adverse birth outcomes including cesarean delivery, macrosomia, and stillbirth (2). The mechanisms by which entering pregnancy with an elevated BMI leads to such adverse outcomes are still unclear, but emerging evidence points to critical involvement of the placenta (3, 4). Studies of placental tissue, which serves as the maternal-fetal interface, provide a unique opportunity to understand the pathophysiological mechanisms underlying the associations of maternal BMI with health outcomes in pregnant women and their offspring.

The placenta is a known regulator of many aspects of maternal physiology in pregnancy, as well as fetal health and programming of future health and diseases (5-8). Several studies have shown that high maternal BMI is associated with altered morphology, hormone production and inflammation in the placenta (9-11). Placental weight and thickness are higher in pregnancies with maternal overweight or obesity (12). Placental implantation and growth are affected by maternal obesity in pregnancy, and histological features of inflammation are more likely to be observed with elevated maternal BMI (13). In pregnancies affected by obesity, previous studies have observed placental alterations in interleukin-6 production, macrophage/leukocyte infiltration, and vessel density (14-17). Placental leptin production and regulation (18), as well as placental fatty acid oxidation, are also altered with maternal obesity (19).

While select biomarkers and placental morphological parameters can be informative outcomes in studies of maternal BMI and the placenta, genome-wide expression profiling can reveal novel mechanisms through the use of an agnostic analytical approach. Differential gene expression profiles in the placenta may help identify novel key molecular regulators of maternal and fetal health in the context of elevated maternal BMI. One of the challenges in genome-wide transcriptomics, as with all high dimensional data analysis, is the need for sufficiently large sample sizes to detect associations that are robust. Very few studies of maternal BMI and placental transcriptomics have been conducted (20-22). Furthermore, these studies were small (total N = 10 to 180, including fewer than 30 participants with obesity), relied on array-based expression profiling, and used a case-control design, which is prone to selection bias. In this study, we examined the association of maternal BMI with placental whole-genome RNA sequencing in Genetics of Glucose regulation in Gestation and Growth (Gen3G), a large population-based pregnancy cohort. We considered both continuous BMI and BMI categories to assess differential placental gene expression, using state-of-the-art RNA sequencing and robust analytic methods. We explored sex-specific associations and assessed the influence of pregnancy complications in sensitivity analyses.

Methods

Genetics of Glucose Regulation in Gestation and Growth Cohort

The Gen3G cohort, which has previously been described (23), is a prospective cohort study of pregnant women based in Sherbrooke, Quebec, Canada. Briefly, we invited women to participate if they received prenatal care directly at or in a health center affiliated with the Centre Hospitalier Universitaire de Sherbrooke (CHUS; Sherbrooke, Quebec, Canada) and planned to deliver at the CHUS, the only hospital in the Eastern Township region offering obstetric care for deliveries. We enrolled 1024 participants in the first trimester of pregnancy between January 2010 and June 2013. Exclusion criteria included a history of preexisting diabetes or laboratory evidence of overt diabetes, multiple pregnancies, substance and/or alcohol use disorder, and use of medications that affect glucose metabolism. We did not exclude women with thyroid disorders or anemia. All participants provided written consent prior to study enrollment according to the Declaration of Helsinki. All protocols were approved by the IRB at the CHUS Clinical Research Center (protocol #07-027-A1). Baseline and birth characteristics of Gen3G participants included in this analysis were similar (see Supplemental Table 1 (24)) to participants that could not be included (due to loss to follow-up during pregnancy, placenta samples not collected, excluded or inadequate RNA quality).

Maternal BMI and Covariates

Study staff ascertained maternal BMI at the first study visit (between 5 and 16 weeks of pregnancy) by dividing measured weight by squared height (kg/m²). We assessed weight using a calibrated electronic scale and height using a wall stadiometer (in light street clothes, without shoes). We considered BMI as a continuous variable for our primary analyses, but also performed secondary analyses using categorization of BMI as follows: underweight <18.5, normal ≥18.5 to <25.0, overweight ≥25.0 to <30.0, and obesity ≥30.0 kg/m².

In the late second trimester (median 26 weeks), we measured maternal weight using the same standardized procedures. During this second visit, participating women completed a standardized 75-g oral glucose tolerance test (75g-OGTT) to screen and diagnose for GDM, and they completed additional research components (questionnaires). We ascertained GDM based on the International Association of Diabetes and Pregnancy Study Groups criteria (at least one value above the following thresholds: fasting glucose ≥92 mg/dL, 1-hour glucose ≥180 mg/dL, or 2-hour glucose ≥153 mg/dL) (25).

At birth, we collected information on delivery mode, gestational age, and child sex from medical records, as well as information on complications that occurred during pregnancy (PE and gestational hypertension [GH]) and at delivery, as reported previously (26). Women were classified as having PE if they had hypertension (at least 2 blood pressure measurements ≥140/90, from the twentieth week of gestation to 6 weeks after delivery) with proteinuria (≥300 mg/day or protein/creatinine ratio ≥0.3) (27) Women classified as having GH had hypertension without proteinuria during pregnancy or up to 6 weeks after delivery. PE and GH will henceforth be referred to as hypertensive disorders of pregnancy (HDP).

Placental Tissue Collection

Trained study staff collected placentas within 30 minutes of delivery using a standardized protocol (23). In brief, a 1-cm³ placental tissue sample was collected from the maternal facing side (within 5 cm radius of corresponding location of cord insertion on the other side). Placental biopsies included decidual tissue. Each collected sample was immediately put in RNA Later for at least 24 hours before storage at −80 °C until DNA/RNA extraction. A participant flow diagram for deliveries with collected placenta samples included in the present analysis is shown in Supplemental Figure 1 (24).

RNA Extraction, Sequencing, and Quality Control

Laboratory personnel extracted total RNA (average = 19.7 ± 7.1 µg) and checked the quality of each sample using an Agilent Bioanalyzer to determine the RNA integrity number (RIN; average RIN = 6.7 ± 0.8). For samples with a RIN value ≥5, we shipped 3 µg of each sample to the Broad Institute (Cambridge, Massachusetts, USA) for sequencing. Samples were rechecked for quality at the Broad Institute using the Caliper Life Sciences LabChip GX system to determine an RNA Quality Score (RQS) for each sample, which ranged from 3.3 to 7.8 across all samples (average RQS = 5.9). All samples with an RQS value of 4 or higher were submitted for RNA sequencing (N = 468 samples), and 250 ng of each sample underwent library preparation, using an automated variant of the Illumina TruSeq Stranded mRNA Sample Preparation Kit (Illumina, cat #RS-122-2103). Flowcell cluster amplification and sequencing were performed according to the manufacturer's protocols using the Illumina HiSeq 4000, to generate 101-bp paired end reads, average of 113 M total reads (range 33 to 378 M) per sample.

The following steps were based on the GTEx v8 pipeline. Briefly, we applied STAR v2.5.3a (28) to align FASTQ/FASTA files to the human GRCh38 reference genome, using the parameters specified at https://github.com/broadinstitute/gtex-pipeline. Duplicates were marked using Picard MarkDuplicates, and expression was quantified with RNASeQC v2.3.6 using the GENCODE v30 annotation (29).

Following quantification, we processed our samples through additional quality control (QC) steps. Inspecting the overall distribution of the dataset, we excluded samples (n = 7) with >1% of outlier genes (as defined by >3 times the interquartile range (IQR) above Q3 or >3 IQR below Q1). We also excluded 3 samples for genotyping mismatch, and 8 samples with high maternal glucose at first trimester. Overall, our RNAseq analytic dataset included 450 placental samples (Supplemental Figure 1 (24)). For gene expression pre-analytic steps, we removed genes with low abundance by keeping only those genes with at least a count of 6 reads in a minimum of 20% of samples and a transcript per million (TPM) values >0.5, as well as average mappability >0.8. After QC, 15 221 genes remained for analysis. Prior to analysis, we computed normalizing factors using the R statistical software package edgeR (30), then normalized and transformed gene counts to log₂ counts per million reads (CPM) using Voom from the Limma R package (31).

Statistical Analysis

We adjusted models for maternal age, fetal sex, gestational age at delivery, and gravidity as biological covariates in addition to computed surrogate variables (SVs) to account for unmeasured sources of variability, including batch effects. We used the EstDimRMTfunction from the R package isva (32) to estimate the number of significant SVs given the residuals from the regression of BMI and biological covariates from the normalized counts. The R package SmartSVA was used to compute 37 SVs in our processed RNAseq dataset. We used Limma to identify differentially expressed genes with log₂ maternal BMI as a continuous independent variable for our primary analysis. For BMI as a continuous predictor variable, the effect size (or slope) is shown (increase in normalized gene expression per log₂-unit increase in BMI). We ran secondary analyses using categorical BMI, comparing participants in the normal BMI category [18.5 to 24.9 kg/m²] vs those with BMI ≥ 30 kg/m² to assess for potential nonlinear associations, and because most prior literature is based on case-control designs using these or similar categories. For models with BMI as a categorical predictor, we express effect sizes as log₂ fold change (FC) in expression for the BMI ≥ 30 kg/m² group compared to the normal BMI category (reference group). We report the unadjusted P values, and P values adjusted for Benjamini-Hochberg false discovery rates (FDR) for all analyses (33). We hypothesized that differential gene expression with larger effect sizes are more likely to have biologic impact, so we chose to report genes with effect sizes (slope for continuous BMI, fold change for BMI categories) at least >5 SD above the average absolute effect estimate of differential expression in relation to maternal BMI across all genes and with an unadjusted P value of <.001 in our main results.

We conducted multiple sensitivity analyses on our models using continuous maternal BMI as an independent variable: first, we removed participants with GDM; second, we removed participants with HDP from our analysis. We reported genes differentially expressed using the same thresholds as our main continuous BMI analyses (effect sizes >5 SD above the average effect estimate, and with an unadjusted P < .001). We conducted analyses using maternal BMI (as continuous variable) measured in late second trimester (median 26 weeks), to investigate whether BMI closer to time of delivery would influence our results. We also further adjusted our main model for mode of delivery (C-section or vaginal) as an additional sensitivity analysis.

To explore whether maternal BMI associations with placental gene expression differ by fetal sex, we performed sex-stratified analyses, and reported all genes by effect size (slope by 1 unit of log₂ BMI) at least >5 SDs above the average and with P values < .05 in both female and male offspring (we used a less stringent P value threshold given small sample size in each stratum and the exploratory nature of these analyses). We further tested interaction by sex for the top differentially expressed genes in both females and males and report the unadjusted P values. Finally, we used GSEA software to perform Gene Ontology pathway enrichments of genes and report results with FDR q-values < 0.05 from each sex-stratum (34, 35).

Results

Gen3G participant characteristics included in our placental transcriptome study are shown in Table 1. Mean (SD) maternal age and BMI were 28.8 (4.4) years and 25.4 (5.5) kg/m², respectively. Overall, 3% of women had underweight, 57% had normal weight, 22% had overweight, and 18% had obesity BMI categorization. Mean (SD) gestational age at delivery was 39.4 (1.3) weeks and 54% of the offspring were male. Participants included in our analysis were similar to those who were not included in our analysis (Supplemental Table 1 (24)).

Table 1.

Characteristics of Gen3G participants (N = 450)

Maternal characteristics	Mean ± SD/Frequency (%)
Age, years	28.8 ± 4.4
Gravidity (% primigravid)	160 (35.6%)
First-trimester BMI, kg/m2	25.4 ± 5.5
Underweight (<18.5)	12 (2.7%)
Normal weight (18.5-24.9)	257 (57.1%)
Overweight (25-30)	99 (22.0%)
Obese (≥30)	82 (18.2%)
Second trimester BMI, kg/m2	27.9 ± 5.3
Underweight (<18.5)	0
Normal weight (18.5-24.9)	152 (33.9%)
Overweight (25-30)	178 (39.7%)
Obese (≥30)	118 (26.3%)
Gestational diabetes mellitus (GDM)	36 (8.1%)
Gestational hypertension (GH)/Pre-eclampsia (PE)	41 (9.2%)
Child Characteristics
Sex	241 M/209 F
Gestational age, weeks	39.4 ± 1.3
C-section rate	77 (17.1%)
Newborn weight (g)	3394.2 ± 468.7
Male	3443.0 ± 462.6
Female	3338.0 ± 470.5
Newborn length (cm)	50.9 ± 2.1
Placental weight (g)	537.9 ± 128.4

Missing data: 2 individuals for maternal BMI at second trimester; 3 individuals for GDM and for GH/PE status; 2 individuals for newborn length; 5 individuals for placental weight. Abbreviation: Gen3G, Genetics of Glucose regulation in Gestation and Growth.

For our analysis of continuous maternal BMI and placental gene expression, a total of 80 genes met our >5 SD above the average effect size threshold, and 7 out of those genes also met our criteria for P < .001 as reported in Table 2 and illustrated in Fig. 1A. The strongest differential gene expression signal was for EPYC, a gene that regulates collagen formation/fibrillogenesis, which showed reduced expression (slope = −1.94) with higher maternal BMI (P = 4.8 × 10⁻¹⁰). This association was statistically significant after adjustment for multiple comparisons (FDR-adjusted P = 7.3 × 10⁻⁶). We also observed that higher maternal BMI was significantly associated with higher placental expression of IGFBP6, CHRDL1, and CXCL13 after accounting for multiple testing (slopes = 0.72 to 1.01; FDR-adjusted P < .05). Additional top differential genes (>5 SD above the average effect size) demonstrating greater expression with higher maternal BMI were MEDAG, OMD, and PTGFR but did not reach FDR statistical threshold (Table 2 and Fig. 1A).

Table 2.

Differentially expressed RNA transcripts in placenta associated with maternal BMI at the first trimester of pregnancy

Gene symbol	Gene name	Gene function	Avg. TMM	Slope	UnadjustedP value	FDR-adjusted P valuea
EPYC	Epiphycan	Regulates fibrillogenesis (collagen formation)	2.33	−1.94	4.8E-10	7.30E-06
IGFBP6	Insulin-like growth factor binding protein 6	Regulates insulin-like growth factor binding activity	3.73	0.72	1.40E-08	1.10E-04
CHRDL1	Chordin-like protein 1	Antagonizes the function of BMP4 by binding to it and preventing its interaction with receptors	10.65	0.87	2.90E-08	1.40E-04
CXCL13	C-X-C motif chemokine ligand 13	B lymphocyte chemoattractant	1.17	1.01	1.40E-05	3.80E-02
PTGFR	Prostaglandin F receptor	A receptor for prostaglandin F2-alpha, a potent luteolytic agent	5.17	0.53	1.00E-04	1.20E-01
MEDAG	Mesenteric Estrogen-Dependent Adipogenesis	Promote adipocyte differentiation, lipid accumulation, and glucose uptake in mature adipocytes	2.42	0.58	1.10E-04	1.20E-01
OMD	Osteomodulin	Involved in cell adhesion and regulation of bone mineralization	1.55	0.59	2.00E-04	1.30E-01

Differentially expressed genes with >5 SD above average slope (> |0.46|) and unadjusted P value <.001 are shown. (N = 450). BMI as a continuous measure (log₂-transformed); slope for 1 unit of log₂ BMI. Model adjusted for fetal sex + gestational age + maternal age + gravidity + 37 surrogate variables.

Abbreviations: BMI, body mass index; FDR, false discovery rate; TMM, trimmed mean m-value.

^a Benjamini-Hochberg false discovery rate adjusted P value.

Figure 1.

Differentially expressed placental genes associated with maternal BMI at the first trimester of pregnancy. (A) Maternal BMI (log₂-transformed) was treated as a continuous variable. (B) Maternal BMI was categorized as normal (BMI ≥ 18.5 to <25) or obesity (BMI ≥ 30). Black dots: Labeled genes with unadjusted P < .001 and differential expression >5SD above the average log₂ fold change in gene expression level in relation to maternal BMI.

Differential placental gene expression findings for BMI as a categorical variable (N = 257 with BMI 18.5 to 24.9 vs N = 82 with BMI ≥30 kg/m²) yielded 74 genes which met the 5 SD above the average log₂ fold change threshold (>|0.36|), and among these 8 had P < .001 (Fig. 1B and Supplementary Table S2 (24)). As with our results for continuous BMI, EPYC showed strongly reduced gene expression (log₂ fold change = −1.35) in placentas from pregnancies in the obesity category compared to the normal weight category; this association was statistically significant after adjusting for multiple comparisons (FDR-adjusted P = .003). Most of the genes found to be differentially expressed in the categorical BMI analyses were overlapping with those found in the continuous BMI analyses. Additional genes emerging specifically from our categorical BMI analyses included higher placental expression of PENK, and CYP1A1 in the obesity category (Supplemental Table 2 (24)).

As expected, higher maternal BMI was associated with higher placental weight (Pearson r = 0.16, P = .0007). Among the placental genes differentially expressed in relation to BMI as continuous variable (Table 2), we detected modest positive associations for placental expression of PTGFR with placental weight and newborn growth metrics (offspring weight, offspring weight z-score, and offspring weight-to-length ratio; see Supplemental Table 3 (24)). We did not detect correlations between birth anthropometrics and placental expression at any of the other differentially expressed genes listed in Table 2.

Sensitivity Analyses—Excluding GDM or HDP

Among Gen3G participants included in our analyses, 36 (8.1%) developed GDM, and 41 (9.2%) developed HDP (Table 1). Women with GDM had higher first-trimester BMI (mean ± SD: 27.8 ± 6.8 kg/m²) as compared to women without GDM (25.2 ± 5.4 kg/m²), P = .03 for comparison. Women with HDP also had elevated first-trimester BMI (28.5 ± 6.2 kg/m²) as compared to women without HDP (25.0 ± 5.3 kg/m²), P < .001. Results of sensitivity analysis after excluding GDM cases for differential gene expression with continuous maternal BMI are shown in Supplemental Table 4A (24) and Fig. 2A. Removal of GDM cases (N = 36) from our analysis produced similar results to those of our main analysis, with EPYC being the most differentially expressed (slope = −1.99; P = 6.6 × 10⁻¹⁰; FDR-adjusted P = 1.0 × 10⁻⁵). In sensitivity analysis performed while excluding cases with HDP (Supplemental Table 4B (24) and Fig. 2B), we consistently observed EPYC as the top differentially expressed gene (slope = −2.09; P = 3.4 × 10⁻¹⁰; FDR-adjusted P = 5.2 × 10⁻⁶) with continuous maternal BMI. Results for IGFBP6, CHRDL1, and CXCL13 differential expression in both sensitivity analyses were very similar to our main analyses.

Figure 2.

Differentially expressed RNA transcripts associated with maternal BMI with the exclusion of pregnancy-related complications. The volcano plots represent top transcript with the exclusion of participants who developed (A) GDM and (B) HDP (PE or GH). Black points indicate genes that reached our thresholds; all labeled genes have unadjusted P < .001 and >5 SD effect size. BMI was log2-transformed and treated as a continuous variable.

Additional Sensitivity Analyses

In our additional sensitivity analyses using maternal BMI measured during the research visit conducted in late second trimester, we found very similar results (see Supplemental Table 5 (24)) to those reported using first-trimester BMI (Table 2). Our results remained essentially the same after further adjusting our main models for mode of delivery (Supplemental Table 6 (24)).

Sex-Stratified Analyses

Analyses stratified by fetal sex are shown in Supplemental Table 7A (female (24)) and Supplemental Table 7B (male (24)), which list all genes that shown maternal BMI differential expression slope 5 SD above the average effect size and P < .05. For placentas from male fetuses, EPYC was the gene with the largest effect size for change in placental gene expression with increasing maternal BMI (slope = −2.56; P = 3.3 × 10⁻⁹), in line with our main analyses. In analyses using placentas from female fetuses, we did not observe the same top genes showing differential expression by maternal BMI and found a slightly more modest estimate of differential expression for EPYC (Slope = −1.44; P = 6.1 × 10⁻³) while OGN (Osteoglycin), a member of the small leucine-rich proteoglycan (SLRP) family of proteins, was the top hit in females (slope = −1.31; P = 2.2 ×10⁻⁴). The remainder of the genes listed were largely different for male vs female (slope 5 SD above the average effect size and P < .05; Supplemental Tables 7A and 7B (24)). In order to understand whether the top genes hold biological relevance dependent on fetal sex, we further explored pathway enrichment of the Gene Ontology (GO) molecular functions domain for the top genes in our sex-stratified analyses for females (Supplemental Table 8A (24)) and males (Supplemental Table 8B (24)). We observed signal receptor binding pathway enrichment for differentially expressed genes in both males and females. For males only, we also observed enrichment of growth factor binding and glycosaminoglycan binding pathways.

Discussion

Transcriptional changes in the placenta may have implications for both maternal and fetal health. In this large prospective pregnancy cohort, we examined the association of maternal BMI, objectively measured early in pregnancy, with genome-wide expression profiles from carefully collected placenta samples. Elevated maternal BMI was associated with lower placental expression of EPYC, a gene that regulates fibrillogenesis. This finding was robust across all sensitivity analyses. Women with a BMI > 30 kg/m² in the first trimester had more than a 2-fold lower placental expression of EPYC compared to women with BMI in the normal range, a much larger effect size than any other differential expression signal in our dataset.

EPYC encodes Epiphycan and is predicted to be a secreted protein. Immunohistochemistry studies show that EPYC is expressed in both maternal decidual cells and the trophoblastic cells of the placenta (36-38). A single cell RNAseq study also showed detectable EPYC mRNA expression levels in cells isolated from the decidua (decidual stromal cells, vascular endothelial cells, maternal natural killer cells, and maternal fibroblasts) as well as in cells from placental villi (Hofbauer cells [fetal macrophages] and fetal fibroblasts) as illustrated in Supplementary Figure 2 (39). In previous studies with limited sample size (N < 25 participants), placental EPYC expression appeared as part of a gene set distinguishing normal weight mothers from those with obesity (40) and was potentially downregulated in mothers with obesity (22), in line with the direction of effect we observed in our study. However, previous studies reporting differential placental expression in women with high BMI did not correct for multiple testing and their findings related to EPYC would not have reached statistical significance if this had been considered. Higher EPYC expression in the placenta has been associated with intra-uterine growth restriction (41), suggesting that it may have implications for fetal growth and development. However, very little is currently known about the potential role of EPYC during pregnancy or its function in the placenta. EPYC is hypothesized to be involved in ERK signaling and integrin pathways, including in GnRH signaling (38). Its role in collagen formation may also contribute to its function in the placenta, given collagen's importance for trophoblast adhesion at the maternal-fetal interface (42). Yet, this remains to be investigated in further physiological and functional studies.

We observed greater placental expression of IGFBP6, CHRDL1, and CXCL13 with higher maternal BMI. All 3 genes are predicted to be secreted proteins, which suggest they might be released in maternal or fetal circulation (36-38). Insulin-like growth factor binding protein (IGFBP)-6 is among a group of IGFBPs that bind to and regulate insulin-like growth factor (IGF) signaling. In the placenta single cell atlas, IGFBP6 is most highly expressed in the decidual stromal cells (in addition to other decidual cell types) and the fibroblast-like cells (Supplementary Figure 2). IGFBP-6 predominantly inhibits IGF-2 activity, a major fetal growth factor. Murine models have demonstrated that IGF2 is highly expressed in the fetal blood vessels of the placenta and IGFBP6 in the myometrium, suggesting that the binding protein plays a critical role in modulating trophoblast invasion (43). Additionally, IGFBP-6 can undergo nuclear localization where it can perform IGF-independent activities such as regulating angiogenesis (44). Outside of pregnancy, circulating IGFBP-6 levels were higher in patients with type 1 diabetes and correlated with diabetes-related complications (45).

Chordin-like 1 (CHRDL1) is a structural glycoprotein that plays a role in embryonic cell differentiation, osteogenesis, and formation of the central nervous system, and it serves as an antagonist of bone morphogenetic protein 4 (BMP4). CHRDL1 has been found to be differentially expressed in mesenchymal stromal cells derived from amniotic fluid, and overexpression of the gene significantly reduced cellular proliferation and migration (46). Further, CHRDL1 has been shown to influence adipocyte differentiation with greater expression in subcutaneous compared to visceral fat (47, 48). Notably, we also observed higher maternal BMI being associated with greater placental expression of MEDAG, another gene putatively involved in adipocyte differentiation, as well as greater expression of the well-known adipokine leptin (LEP) in our female strata (P = .02, FDR-adjusted P = .6).

Chemokine (C-X-C motif) ligand 13 (CXCL13) is a B lymphocyte chemoattractant that has been shown to be in much higher serum concentration in pregnancy (compared to nonpregnant women), present in amniotic fluid and cord blood, and may play a role in fetal-maternal immune tolerance (49). Outside of pregnancy, women with a BMI > 30 kg/m² had higher circulating levels of CXCL13, compared to women in the normal weight category, reflecting pro-inflammatory profile (50).

It is notable that most of our differentially expressed genes in relation to higher maternal BMI are predominantly expressed by maternal decidual cells. Normal human placentation requires delicate regulation of the immune system to allow trophoblasts (fetal origin) to interact with maternal endometrium leading to transformation into decidual tissue (51). Current knowledge about how maternal obesity affects the interaction between trophoblasts and decidual cells is limited, but recent studies suggest that high maternal BMI is associated with altered decidual immunity (52) and placental inflammation (20, 22), which can lead to placentally mediated pregnancy complications.

Placental adaptations to maternal obesity have been shown to vary by fetal sex, including changes in placental size and efficiency, differences in placental steroid hormone production, and alterations in placental fatty acid oxidation (19, 53, 54). In the present study, we explored whether associations of maternal BMI with placental transcriptomics varied by fetal sex in stratified analyses. Results in both males and females showed relatively consistent associations for EPYC, albeit the coefficient of association seemed larger in males (slope = −2.56) vs females (slope = −1.44; P = .0004 for interaction). Meanwhile, other top genes seemed to differ between males and females. Pathway enrichment analyses for the top hits per fetal sex yielded results suggesting signal receptor binding molecular processes in both males and females while other emerging pathways were sex-specific. We remain cautious about interpretation of these findings, given that we consider these pathways analyses exploratory and that we included a list of genes identified using unadjusted P < .05 from each sex-stratified differential expression analyses.

Our study has several strengths. First, our analysis of placentas from over 400 participants is significantly larger than prior studies investigating maternal BMI and placental gene expression. Our gene expression analysis is based on RNA sequencing, which compared to microarray profiling methods, is less biased and has higher resolution, a lower limit of detection, and a wider dynamic range for interrogating gene expression (55). Our large sample size, with a full range of maternal BMI, allowed us to examine associations between placental transcriptomics and continuous maternal BMI, in addition to categorical comparisons (ie, obesity vs normal weight). In our cohort, trained research staff measured height and weight in the first and late second trimesters to calculate an objective measure of maternal BMI, limiting misclassification of self-reported pre-pregnancy BMI. Our results for first and second trimester BMI showed very similar associations. However, given the correlation of maternal BMI across trimesters, it is difficult to determine whether it is BMI early or later in the pregnancy that is more related to placental gene expression levels. Our study also had additional limitations. First, our study is observational, so we cannot confirm any causal effects of maternal BMI on differential placental gene expression. We collected placentas at delivery, so we could not assess any differential expression that may have occurred in early or mid-pregnancy and that have subsided later in gestation. Despite our large initial sample size, our sex-stratified models and interaction analyses were likely limited in power. Finally, our cohort consists primarily of white women from one area of Quebec, which might limit the generalizability of our findings.

In conclusion, we identified downregulation of placental EPYC expression in pregnant women with higher first-trimester BMI. This association was robust across all of our analyses. Our findings may offer some insight into underlying mechanisms by which maternal BMI may influence the placental transcriptome and do not have immediate clinical applications but may offer novel targets for development of future interventions to limit the adverse pregnancy or perinatal outcomes associated with maternal obesity. Additional studies are required to examine the implications of differential EPYC expression for maternal and fetal health. Other differentially expressed genes such as IGFBP6, CXCL13, and CHRDL1 suggest potential mechanisms by which the placenta may participate in alterations in fetal growth or pregnancy complications via the IGF axis, inflammation, or regulation of adipose tissue biology in pregnancies affected by obesity. Future studies are needed to replicate these findings and to investigate the potential role of these identified genes in placental biology.

Abbreviations

BMI

body mass index

CHUS

Centre Hospitalier Universitaire de Sherbrooke

FDR

false discovery rate

GDM

gestational diabetes mellitus

Gen3G

Genetics of Glucose regulation in Gestation and Growth

GH

gestational hypertension

HDP

hypertensive disorders of pregnancy

IGFBP

insulin-like growth factor binding protein

PE

pre-eclampsia

QC

quality control

RIN

RNA integrity number

RQS

RNA Quality Score

SV

surrogate variable

Funding

This work was supported by a grant from the National Institute of Health (NIH) (R01HD094150). Gen3G was initially supported by a Fonds de recherche du Québec Santé (FRQS) operating grant (to M.F.H., grant #20697); Canadian Institute of Health Research (CIHR) operating grants (to M.F.H. grant #MOP 115071 and to L.B. #PJT-152989); and a Diabète Québec grant (to P.P.). J.C.F. is supported by NHLBI K24 HL157960. L.B. and P.E.J. are senior research scholars from the FRQS. M.F.H. was a recipient of an American Diabetes Association (ADA) Pathways To Stop Diabetes Accelerator Award (#1-15-ACE-26).

Disclosures

The authors have nothing to disclose.

Ethical Disclosure

Institutional approval was obtained for Gen3G participants following the principles outlined in the Declaration of Helsinki. All women recruited in the study provided written informed consent prior to study enrollment.

Data Availability

RNAseq data used in this study is publicly available in dbGaP (56) under the accession number phs003151.v1.p1.