GENE-CENTRIC ANALYSIS OF PREECLAMPSIA IDENTIFIES MATERNAL ASSOCIATION AT PLEKHG1

Abstract

The genetic susceptibility to preeclampsia, a pregnancy-specific complication with significant maternal and fetal morbidity, has been poorly characterized. To identify maternal genes associated with preeclampsia risk, we assembled 498 cases and 1,864 controls of European ancestry from preeclampsia case-control collections in five different US sites (with additional matched population controls), genotyped samples on a cardiovascular gene-centric array comprised of variants from approximately 2,000 genes selected based on prior genetic studies of cardiovascular and metabolic diseases, and performed case-control genetic association analysis on 27,429 variants passing quality control. In-silico replication testing of 9 lead signals with p<10⁻⁴ was carried out in independent European samples from the SOPHIA and Inova cohorts (212 cases, 456 controls). Multi-ethnic assessment of lead signals was then performed in samples of African-American (26 cases, 136 controls), Hispanic (132 cases, 468 controls), and East Asian (9 cases, 80 controls) ancestry. Multi-ethnic meta-analysis (877 cases, 3,004 controls) revealed a study-wide statistically significant association of the rs9478812 variant in the pleiotropic PLEKHG1 gene (OR 1.40 (1.23-1.60, p_meta=5.90 × 10⁻⁷). The rs9478812 effect was even stronger in the subset of European cases with known early-onset preeclampsia (236 cases diagnosed < 37 weeks, 1,864 controls; OR 1.59 (1.27-1.98), p=4.01 × 10⁻⁵). PLEKHG1 variants have previously been implicated in genome-wide association studies of blood pressure, body weight, and neurological disorders. While larger studies are required to further define maternal preeclampsia heritability, this study identifies a novel maternal risk locus for further investigation.

INTRODUCTION

Preeclampsia is a common pregnancy-specific multisystem disorder, characterized by systemic endothelial dysfunction in the mother, with both maternal (pulmonary edema, acute renal failure, liver failure, stroke, and placental abruption) and fetal/neonatal complications (preterm birth, fetal growth restriction, hypoxic-neurologic injury, and perinatal death). Despite extensive investigative efforts, preeclampsia remains a significant source of both fetal and maternal morbidity and mortality¹.

Preeclampsia is associated with an increased long-term risk of maternal cardiovascular diseases (CVD), including ischemic heart disease and stroke^2–9. Approximately a decade after a preeclamptic pregnancy, a woman’s risk of ischemic heart disease is elevated approximately 2-fold^2,5–7,9. Two hypotheses prevail regarding preeclampsia and future CVD risk: (1) women with preeclampsia are predisposed to CVD, and that the vascular and metabolic challenge of normal pregnancy exposes this predisposition¹⁰; and/or (2) preeclampsia causes vascular and metabolic changes which remain following pregnancy and increase a woman’s future risk¹¹.

Preeclampsia has a substantial heritable component as demonstrated in multiple epidemiologic studies, with heritability estimated at 55-60%, of which 30-35% is attributed to a maternal genetic effect and 20% fetal^12–15. Understanding preeclampsia heritability has proved challenging due to multiple factors including the heterogeneous nature of the disease, combined complex genetic and environmental risk factors, as well as the contribution of at least two genomes (maternal and fetal) to each pregnancy outcome. To overcome these challenges, ideal genetic studies of preeclampsia would have adequate sample sizes, detailed clinical characterization, paired maternal-fetal samples, and replication of significant genetic loci from the discovery population in an independent cohort¹⁶. The recently published fetal preeclampsia GWAS utilized several of these study design improvements, leading to the identification of the genome-wide significant fetal variant near the FLT1 gene¹⁷. Maternal genetic variants contributing to preeclampsia risk remain elusive; despite many candidate gene association studies (summarized in ¹⁸) and three recently reported GWAS^19–21, there are no maternal variants that have been robustly and reproducibly associated. Strategies utilizing paired maternal-fetal samples have been underutilized to date, but are likely to reveal significant interacting genetic loci, as demonstrated by the increased preeclampsia risk conferred by specific combinations of maternal killer immunoglobulin-like receptor (KIR) and fetal HLA-C alleles^22,23.

In an effort to better define the maternal genetics of preeclampsia and to identify shared genetic factors for PE and CVD, we tested for the genetic association of common variants with preeclampsia in women of European ancestry from the United States (US; 498 cases, 1,864 controls) using the Human CVD Beadarray 50K SNP ® (Illumina), also referred to as the ITMAT-Broad-CARe (IBC) genotyping array, that captures genetic diversity across >2000 candidate gene regions related to cardiovascular, inflammatory, and metabolic phenotypes²⁴. We tested for replication of genetic variants with p<10⁻⁴ in two independent case-control populations of European ancestry (Study of Pregnancy Hypertension in Iowa (SOPHIA): 177 cases, 116 controls¹⁹, and Inova: 35 cases, 340 controls²⁵, and extended association analyses to women of African-American (26 cases, 136 controls), Hispanic (132 cases, 468 controls), and East Asian (9 cases, 80 controls) ancestry for a trans-ethnic meta-analysis (877 cases, 3,004 controls).

METHODS

Data sharing

The five US site European preeclampsia GWAS summary statistics (Discovery population; described below) have been made publicly available and can be accessed at https://richa-saxena.squarespace.com (Data section).

Study Population

Discovery population: preeclampsia case-control collections

Case women with preeclampsia during pregnancy and matched normotensive control women were identified from five US sites with IRB-approved protocols, as detailed below. Preeclampsia case status at all sites was defined by standard American Congress of Obstetricians and Gynecologist (ACOG) criteria^26,27. Specifically, preeclampsia diagnosis was determined based on systolic blood pressures (SBP) ≥140 mmHg or diastolic blood pressures (DBP) ≥90 mmHg or higher on two occasions at least 4 hours apart occurring after 20 weeks gestation in women whose blood pressure had previously been normal, as well as new-onset proteinuria with ≥300 mg protein in a 24-hour urine specimen, a protein/creatinine ratio of at least 0.3, or ≥1+ protein on urine dipstick. Cases of superimposed preeclampsia (women with chronic hypertension with worsening SBP ≥160 mmHg or DBP ≥110 mmHg at least 4 hours apart occurring after 20 weeks gestation, as well as new-onset or worsening proteinuria) and HELLP syndrome (defined by red blood cell hemolysis, elevated liver transaminases at least twice normal concentration, and platelet count <100,000/μL) were also included.

1. Boston Hospital collections

Massachusetts General Hospital (MGH)/ Brigham & Women’s Hospital (BWH)

Billing records of women who delivered at MGH and BWH from 1995-2011 were queried for preeclampsia-related ICD-9 codes. When patients returned to participating sites, blood samples drawn for clinical diagnostic testing were obtained at their point of discard. Preeclampsia diagnosis was validated by physician electronic medical record (EMR)/ chart review with a 30-day time-limited link to protected health information. DNA was extracted from buffy coats of de-identified accessioned samples. Because the investigators did not interact with patients for data or sample ascertainment, the Partners Human Research Committee waived the informed consent requirement (detailed by 45 CFR 46.116). In total, 244 validated cases (230 preeclampsia, 14 superimposed preeclampsia) of European ancestry were obtained. Additional African-American (n=5) and Hispanic (n=23) preeclampsia cases were obtained for the multi-ethnic meta-analysis. Beth Israel Deaconess Medical Center (BIDMC). Nineteen preeclampsia cases and 74 normotensive age-matched controls of European ancestry with healthy, term pregnancies were identified from an ongoing, maternal blood collection at BIDMC with preeclampsia diagnosis determined by physician case review. Additional African-American (1 preeclampsia case, 9 controls) and Hispanic (6 preeclampsia cases, 1 superimposed preeclampsia case, 17 controls) participants were obtained for the multi-ethnic meta-analysis. DNA was extracted from maternal blood samples using the Paxgene collection system (BD).

2. University of Southern California (USC) collections

HELLP Syndrome Society/ HELLP Syndrome Research at USC

Cases were recruited using internet-based methods, either from the HELLP Syndrome Society website (no longer functional) or from the HELLP Syndrome Research at USC Facebook page. Controls were unrelated friends of the cases and their families. To verify case status, medical records were requested and reviewed by study investigators. Saliva samples (Oragene DNA Self-Collection Kit) or buccal samples for DNA were obtained at each participant’s home and mailed to the investigators for DNA extraction. In total, 80 preeclampsia/HELLP cases of European ancestry were utilized. Los Angeles County + University of Southern California Medical Center. Hispanic cases (n=59) and normotensive controls (n=95), as well as an African-American control (n=1), were recruited retrospectively from delivery logs at the Los Angeles County and University of Southern California Women’s and Children’s Hospital from 1999 to 2006, as well as during their postpartum stay at the hospital from 2007-2008²⁸. Medical charts were abstracted and reviewed to verify preeclampsia case and normotensive control status. Women with lupus, renal disease, multiple gestations, and sickle cell disease were excluded. Maternal DNA was obtained from blood, mouthwash, buccal swabs, or saliva, as previously described²⁸.

3. Children’s Hospital of Philadelphia (CHOP) collection

DNA from European (69 cases, 345 controls), African-American (10 cases, 41 controls), and Hispanic (3 cases, 11 controls) women was collected from 2009–2012 at the Center for Applied Genomics (CAG) at CHOP through A Study of the Genetic Causes of Complex Pediatric Disorders (GCPD). Mothers of recruited children were asked if they had ever had preeclampsia during their pregnancy and gave written informed consent to allow access to their medical records. Preeclampsia diagnosis was confirmed by medical record review. Maternal DNA from whole blood was extracted using the Agencourt Genfind v2 DNA purification system (Beckman Coulter) on an automated Biomek-FX (Beckman Coulter). Maternal DNA from saliva was extracted using the Agencourt DNAdvance protocol (Beckman Coulter) on an automated Biomek-FX.

4. Yale-New Haven Hospital collection

DNA from European (11 cases, 30 controls), African-American (1 case, 16 controls), and Hispanic (13 cases, 170 controls) women was collected from the March of Dimes Perinatal Emphasis Research Initiative (PERI) project at New York University, NY and from Yale-New Haven Hospital, New Haven, CT from January 1989 to June 2005. Maternal blood was collected at the time of preeclampsia diagnosis, after written consent was obtained. Buffy coats from blood samples were obtained, and DNA subsequently extracted by standard methods. All specimens were linked with the medical records.

5. University of Iowa Hospitals and Clinics collection

DNA from 77 preeclamptic women (of European (n=75), African-American (n=1), and Hispanic (n=1) ethnicity) collected from 1996–2005 at the University of Iowa Hospitals and Clinics was used for this study. Maternal DNA was extracted from buccal swab or peripheral blood sample. Preeclampsia diagnosis was verified by medical record review.

Population-based Controls

To enhance power to detect genetic associations, population-based unselected European control women from the National Heart, Lung, and Blood Institute (NHLBI) studies, Atherosclerosis Risk in Communities (ARIC; n=645)²⁹ and Coronary Artery Risk Development in Young Adults (CARDIA; n=770)³⁰, with IBC array genotype data from the Candidate gene Association Resource (CARe)³¹, were matched to the case population.

Replication Populations

SOPHIA study

The SOPHIA Caucasian cohort of 177 preeclampsia cases and 116 normotensive controls was recruited using electronic birth certificates from the Iowa Department of Public Health from 3078 primiparous women who gave birth in Iowa from August 2002 to May 2005, as previously described¹⁹. Genotyping was performed using the Affymetrix Genome-Wide SNP Array 6.0 (Affymetrix). Sample quality control and data analysis was performed, as previously described¹⁹.

Inova Health System study

The multi-ethnic Inova cohort of 78 preeclampsia cases and 664 normotensive controls was recruited at Inova Fairfax Medical Campus (Falls Church, Virginia) during 2011-2013 as part of the IRB-approved Molecular Study of Preterm Birth genomic research study. Clinician chart review was performed to categorize women as hypertensive or normotensive, and a second chart review was performed to confirm the diagnosis of preeclampsia in women in the hypertensive group using the same criteria as the Boston Hospitals collection. Whole genome sequencing was performed on whole blood samples by Complete Genomics, Inc. (Mountain View, CA) and processed as previously described³². The 9 SNPs tested for replication fulfilled genotype quality control criteria (100% call rate, allele balance >=0.225, read depth >=9, allele read depth >=6, a quality score not VQLOW, Hardy-Weinberg equilibrium p> 0.001). Ancestry was computed by performing principal components analysis on the genomic data from the Inova cohort and from 1,815 individuals from the 1000 Genomes Project³³, with population labels assigned to the Inova cohort using the k-nearest neighbor algorithm and the first 10 principal components.

Discovery population: genotyping and quality control

Case and control DNA samples from the 5 US site discovery population were genotyped on a cardiovascular gene-centric 50 K SNP array v1.0 (45,707 markers), v2.0 (49,094 markers), or v3.0 (53,831 markers)²⁴. SNPs were clustered into genotypes using the Beadstudio software ® (Illumina) and subjected to quality control at the sample and SNP level. Genotypes for the population control sample (ARIC, n=649; CARDIA, n=770) were generated by the Broad Institute as part of the CARe study³¹. Within Europeans, samples were excluded for individual call rates <90% (n=0), sex mismatch (n=0), heterozygosity >3 SD from the mean (n=3) and excess relatedness (pi-hat>0.2 (n=16)). Only SNPs present in all three versions of the 50K SNP array (38,138 overlapping SNPs) were included: 2,692 SNPs were removed for call rates <95%, no SNPs were removed for departure from Hardy-Weinberg equilibrium (p<10⁻⁷ in controls), 8,017 SNPs were removed for minor allele frequency (MAF) <0.01, and no SNPs were removed for differential missingness between cases and controls (p<10⁻⁶). Overall SNP call rate per sample was consistent across all 5 US sites. After quality control, 27,429 SNPs in 498 cases and 1864 controls were successfully tested in our discovery sample.

Statistical Analysis

Evaluation of population stratification

Genetic ancestry of cases and normotensive controls was determined using principal components (PCs) analysis (Figure S1). Ten eigenvectors were calculated from all overlapping SNPs in the 1000 Genomes Project (1KGP) from European, African, and Asian reference populations³⁴ and were projected onto our sample. Based on distance to the nearest continental cluster, we defined samples as of European, African-admixed, or Asian-admixed ancestry. Concordance of self-report with genetic cluster was 93.1% for Europeans, 66.0% for African-admixed and 96.4% for Asian-admixed samples.

Association testing

European IBC array single SNP association analysis

Association testing was performed using logistic regression with the SNP coded in an additive genetic model and including adjustment for 10 PCs (calculated in the cleaned European sample) to correct for subtle population structure. To correct for multiple comparisons, based on an estimated 20,500 independent tests on the ITMAT-Broad-CARe (IBC) genotyping array in Europeans, the study-wide statistical threshold of significance was set to p=2.4 × 10^{−6 35}.

Admixture-informed association analysis in African-American and Hispanic populations

For African-American and Hispanic patients from the 5 US site collection, HAPMIX and MIXSCORE were used to jointly calculate a SNP association score conditioned on local ancestry using logistic regression and a case-only admixture score which evaluates the causal hypothesis that, restricting to disease cases, the proportion of one ancestry at the candidate locus differs from the genome-wide proportion³⁶.

Replication analysis in SOPHIA and Inova

Directly genotyped lead SNPs or proxy SNPs (r²>0.5 in 1000 Genomes Northern Europeans from Utah (1KGP CEU)) from the 9 lead association signals in our study were interrogated in the previously described SOPHIA Caucasian maternal preeclampsia GWAS (177 preeclampsia cases, 116 normotensive controls) genotyped on the Affymetrix SNP 6.0 microarray¹⁹. These 9 lead association signals (all directly genotyped by whole genome sequencing) were also interrogated in the subset of European cases from the Inova study (35 preeclampsia cases, 340 normotensive controls). Logistic regression was performed for each ancestry group with PLINK v1.90 using the first 10 principal components as covariates (www.cog-genomics.org/plink/1.9)³⁷.

Meta-analysis

Results were combined across discovery and replication studies, and across ethnicities, using a fixed effects, inverse variance meta-analysis approach³⁸.

Gene and Pathway Analysis

We performed gene-based association testing using VEGAS³⁹ for 6,804 genes containing one or more SNPs. The significance threshold was set at p<7.35×10⁻⁶ after Bonferroni correction for multiple testing. To screen for pathways enriched among the top-most associated hits, we used the pathway analysis tool DEPICT v.1 rel 173⁴⁰. SNPs with p<10⁻³ were included in the DEPICT gene set analysis.

RESULTS

Ethnicity and collection sites of preeclampsia cases and normotensive controls are described in the online-only Data Supplement (please see http://hyper.ahajournals.org; Table S1). Site-specific characteristics of each case-control cohort are described in the Methods.

Of the European preeclampsia cases included in the discovery population (n=498), more detailed clinical information was available for 71% (n=354), including cases from the Boston, USC, and Yale sites but not Iowa or CHOP sites (see Table S2). Of the 354 characterized cases, 201 were cases of early-onset preeclampsia with diagnosis < 37 weeks gestation, 101 were cases of hemolysis-elevated liver enzymes-low platelets (HELLP) syndrome, and 215 were cases with severe features (e.g., severe range blood pressures, abnormal laboratory studies, maternal symptoms). This means that, in the discovery population, at least 40% of the preeclampsia cases were early-onset, 29% had HELLP syndrome, and 43% had severe features. Birth weight was not well-ascertained in many of the study collections; thus, information on intrauterine growth restriction (IUGR) was not available.

Gene-centric association analysis in sample of European ancestry

Association testing for preeclampsia in samples of European ancestry revealed little inflation from undetected population structure or relatedness (λ =1.09; Figure 1). No study-wide significant signals were observed, but 9 independent signals reached suggestive significance (p<10⁻⁴; IBC Discovery, Table 1). No evidence of independent replication of the 9 association signals was seen in the European replication population (SOPHIA and Inova, Table 1; see Table S3 for directly genotyped and proxy SNPs with r²>0.5 in the 1KGP CEU population in SOPHIA). However, in European meta-analysis including the SOPHIA and Inova populations, while no SNP surpassed the IBC array-wide significance threshold, a SNP within the PLEKHG1 gene, rs11155751 (later renamed rs9478812 in the dbSNP database), showed greater evidence of association than in either study alone (OR 1.41 (1.22 -1.64), p=5.49 × 10⁻⁶; European Combined, Table 1). A sensitivity analysis by US site demonstrated consistent directionality of effect for rs9478812 across the 5 sites (Table S4). Additionally, the rs9478812 effect was even stronger in the subset of European cases with known early-onset preeclampsia (236 cases diagnosed < 37 weeks, 1,864 controls; OR 1.59 (1.27-1.98), p=4.01 × 10⁻⁵).

Figure 1.

Manhattan and q-q plots of PE association results in women of European ancestry (SNPs >1% frequency for n =498 cases, n =1,864 controls).

Table 1.

Genetic association results: discovery, replication, and meta-analysis in samples of European ancestry.

			IBC Discovery (5 US sites)		Replication (SOPHIA† + Inova)		European Combined (Discovery + Replication)
			(498 cases, 1864 controls)		(212 cases, 456 controls)		(710 cases, 2320 controls)
SNP Chr: position*	Gene	EA/OA EAF	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
rs303016 6:25139464	LRRC16	T/C 0.17	1.44 (1.20 – 1.72)	6.66 × 10−5	1.26 (0.91 – 1.73)	0.16	1.39 (1.19 – 1.63)	3.08 × 10−5
rs9478812‡ 6:151095379	PLEKHG1	A/G 0.22	1.44 (1.22 – 1.70)	1.46 × 10−5	1.30 (0.91 – 1.85)	0.14	1.41 (1.22 – 1.64)	5.49 × 10−6
rs11024739 11:18645843	SPTY2D1	C/A 0.27	1.39 (1.19 – 1.63)	2.80 × 10−5	0.90 (0.64 – 1.25)	0.53	1.29 (1.12 – 1.48)	4.16 × 10−4
rs1110470 16:27336427	IL4R	A/G 0.47	0.75 (0.65 – 0.87)	9.28 × 10−5	1.20 (0.91 – 1.59)	0.20	0.83 (0.73 – 0.94)	3.95 × 10−3
rs3024630 16:27366126	IL4R	G/A 0.09	1.68 (1.33 – 2.11)	9.65 × 10−6	0.81 (0.50 – 1.33)	0.41	1.47 (1.20 – 1.81)	2.43 × 10−4
rs11656003 17:1979792	SMG6	A/G 0.09	1.59 (1.28 – 1.98)	3.50 × 10−5	0.95 (0.33 – 2.08)	0.82	1.45 (1.19 – 1.77)	2.51 × 10−4
rs3764429 17:10025298	GAS7	A/G 0.08	1.60 (1.27 – 2.02)	8.93 × 10−5	0.82 (0.59 – 1.52)	0.68	1.54 (1.22 – 1.93)	2.18 × 10−4
rs1984661 17:12650777	MYOCD	C/T 0.42	1.34 (1.16 – 1.54)	7.66 × 10−5	1.07 (0.64 – 1.77)	0.80	1.31 (1.14 – 1.51)	1.08 × 10−4
rs867616 19:1957682	EDG4	T/C 0.19	1.47 (1.23 – 1.74)	1.39 × 10−5	1.12 (0.80 – 1.55)	0.51	1.38 (1.19 – 1.61)	3.28 × 10−5

^*Position from hg19.

^†See Table S3 for proxy SNPs used for the SOPHIA replication. Proxy SNPs were not available in the SOPHIA genotyping data for IBC array SNPs rs3764429 and rs1984661.

^‡rs11155751 listed on the IBC array has been merged by dbSNP into alias rs9478812.

Transferability of lead European association signals to multi-ethnic samples

Next, we performed SNP association analysis in the 5 US site African-American and Hispanic preeclampsia case-control populations also genotyped with the IBC array, and q-q plots confirmed little inflation (λ =0.99, African-American; λ =1.08, Hispanic; Figure S2a,b). We then tested for transferability of the 9 lead European association signals to the combined 5 US site + Inova multi-ethnic preeclampsia case-control samples (African-American, Hispanic, and East Asian ancestries) (Table 2). Combined analysis across ethnic groups for the 9 SNPs led to one study-wide significant association, rs9478812 in PLEKHG1; p=5.90 × 10⁻⁷ (Table 2). A forest plot of the association of rs9478812 by ethnic group and in combined analysis is shown in Figure S3.

Table 2.

Cross-ethnic investigation of 9 lead European association signals.

	African-American (5 US sites + Inova)		Hispanic (5 US sites + Inova)		East Asian (Inova)		Trans-ethnic meta-analysis (5 US sites + SOPHIA + Inova)
	(26 cases, 136 controls)		(132 cases, 468 controls)		(9 cases, 80 controls)		(877 cases, 3004 controls)
SNP Gene	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
rs303016 LRRC16	0.92 (0.37 – 2.29)	0.86	1.07 (0.73 – 1.57)	0.73	0.34 (0.03 – 3.72)	0.38	1.32 (1.15 – 1.53)	1.15 × 10−4
rs9478812* PLEKHG1	0.76 (0.32 – 1.83)	0.55	1.49 (1.10 – 2.01)	9.7 × 10−3	0.70 (0.18 – 2.75)	0.61	1.40 (1.23 – 1.60)	5.90 × 10−7
rs11024739 SPTY2D1	1.71 (0.55 – 5.34)	0.35	0.88 (0.64 – 1.22)	0.45	1.07 (0.32 – 3.59)	0.92	1.22 (1.07 – 1.38)	2.62 × 10−3
rs1110470 IL4R	0.86 (0.42 – 1.77)	0.68	1.16 (0.86 – 1.56)	0.33	0.46 (0.11 – 1.97)	0.30	1.15 (1.02 – 1.29)	1.80 × 10−2
rs3024630 IL4R	1.30 (0.26 – 6.44)	0.75	1.10 (0.47 – 2.59)	0.82	–	–	1.45 (1.19 – 1.77)	2.84 × 10−4
rs11656003 SMG6	0.39 (0.07 – 2.28)	0.30	0.79 (0.47 – 1.33)	0.37	–	–	1.32 (1.10 – 1.59)	2.88 × 10−3
rs3764429 GAS7	0.32 (0.07 – 1.42)	0.14	0.69 (0.42 – 1.14)	0.15	0.58 (0.08 – 4.42)	0.60	1.29 (1.05 – 1.59)	1.35 × 10−2
rs1984661 MYOCD	0.70 (0.29 – 1.70)	0.44	1.01 (0.72 – 1.41)	0.96	0.71 (0.17 – 3.10)	0.65	1.24 (1.10 – 1.41)	6.75 × 10−4
rs867616 EDG4	0.90 (0.08 – 10.2)	0.93	0.34 (0.09 – 1.25)	0.11	–	–	1.35 (1.16 – 1.57)	8.77 × 10−5

^*rs11155751 listed on the IBC array has been merged by dbSNP into alias rs9478812

Pathway analyses

While SNPs on the IBC array cover candidate genes, SNP associations do not necessarily identify the causal variants or genes. Using evidence from multiple independent association signals in gene- or pathway-based analyses can enhance power to detect associations, identify causal genes, and point to causal biological processes^41,42. Therefore, we performed gene- and pathway-based association analyses on the European discovery population.

Gene-based VEGAS analysis prioritized one gene: IL4R, that encodes the alpha chain of the interleukin-4 receptor (Table S5). DEPICT analysis⁴⁰ prioritized four genes with FDR<20%, SELE, BAG3, C6orf72, and PPP1R3B (Table S6a). One gene set defining the protein-protein interaction (PPI) subnetwork for the phosphoinositide-3-kinase regulatory subunit 3 (PIK3R3) was most significantly enriched (FDR<20%; Table S6b). Enriched tissues included urogenital structures, blood vessels, and keratinocytes (FDR <20%; Table S6c).

DISCUSSION

In this large maternal gene-centric association study of preeclampsia, we find a multi-ethnic study-wide significant association of rs9478812, a SNP which lies within an intronic region of the pleiotropic gene, PLEKHG1.

PLEKHG1 is a large gene (232 kb) with 15 exons that encodes the PLEKHG1 protein, a Rho guanine nucleotide exchange factor expressed across a wide range of tissues (Figure S4a). Despite its ubiquitous expression, the function of PLEKHG1 is largely unknown. SNPs within PLEKHG1 have been associated with blood pressure traits (i.e., both systolic and diastolic blood pressure) in multi-ethnic GWAS⁴³, suggesting that this region may be important in BP regulation. The PLEKHG1 locus has also been identified in GWAS of sickle cell anemia⁴⁴, cognitive decline⁴⁵, obesity-related traits⁴⁶, alcohol and nicotine co-dependence⁴⁷, and panic disorder⁴⁸. PLEKHG1 knockout mice exhibit a sex-specific phenotype, with a reduction in the peripheral blood granulocytes (i.e., polymorphonuclear lymphocytes, PML) in female mice only (male granulocyte number is normal) (Antonella Galli, Mouse Pipeline Team, Sanger Institute, Wellcome Trust Genome Campus; personal communication, November 8, 2017); thus, PLEKHG1 may influence PML production and/or function specifically in females. Additionally, the PLEKHG1 protein has been implicated in actin cytoskeletal organization and can mediate cyclic stretch-induced perpendicular reorientation of endothelial cells in a model of endothelial mechanical force loading⁴⁹.

The phenotypes and molecular processes associated with PLEKHG1 suggest many potential mechanisms by which alterations at this genetic locus could alter maternal preeclampsia risk. As PLEKHG1 locus has been associated with hypertensive disease in other GWAS⁴³, variants within this region may predispose women to develop hypertensive disease in pregnancy. Hypertensive disease could be mediated by altered vascular endothelial cell response to stress, given the demonstrated role of PLEKHG1 within endothelial cells⁴⁹, or by effects on the kidney since renal PLEKHG1 gene expression increases in an angiotensin II-induced mouse model of hypertension⁵⁰. Additionally, as knockout mouse data emphasizes the role of PLEKHG1 in mediating female PMLs (Antonella Galli, personal communication, November 8, 2017), PLEKHG1 risk variants could alter PML production and/or function, leading to changes in the innate and adaptive immune responses. Interestingly, women with severe preeclampsia have notable leukocytosis (attributed to expansion of the neutrophil population, the most predominant PML)^51,52, which may be a sign of increased acute inflammation versus alterations in the adaptive immune system^53–55. Furthermore, neutrophils have been implicated in endothelial injury in women with preeclampsia⁵⁶, and neutrophils from preeclamptic women have increased adhesion to endothelial cells versus neutrophils from healthy pregnant women⁵⁷.

Regarding the specific genetic locus tagged by the study-significant SNP, rs9478812, this SNP lies within an intronic region of PLEKHG1 in a region predicted to bind the protein SETDB1, a histone methyltransferase important for gene silencing. The G to A change at rs9478812 (A, effect allele; G, other allele; see Table 1) is predicted to alter the binding sites of transcription factors, GATA and HLTF (Figure S4b). Thus, the region tagged by rs9478812 is likely to be functionally important and, consequently, may be the specific region responsible for increasing preeclampsia risk. However, rs9478812 is in linkage disequilibrium (LD) with many intronic SNPs within PLEKHG1; the biological annotation of the common variants in LD with rs9478812 is shown in Figure S4b. The array used for genotyping in this GWAS lacks the coverage necessary to further delineate the specific causal SNPs. Therefore, to identify the causal SNP(s) in this region, and to determine the impact of the SNP(s) on gene function, fine-mapping and experimental functional follow-up will be necessary.

In addition to the lead signal observed within PLEKHG1, gene- and pathway-based approaches suggested enrichment of: modules associated with the IL4 receptor (IL4R); the genes SELE, BAG3, C6orf72, and PPP1R3B; and phosphoinositide 3-kinase (PI3) kinase signaling. Many of these genes and pathways have been implicated in preeclampsia pathogenesis and are discussed in further detail below. Enriched tissues included urogenital structures and blood vessels, two specific tissues types that would be predicted to be important in maternal predisposition to preeclampsia.

IL4R encodes for the alpha chain of the interleukin-4 receptor, which binds IL4 and IL13 leading to regulation of IgE and differentiation of Th2 cells. Increased IL4 levels during pregnancy promote Th2 responses and inhibit Th1 responses; these alterations are important for fetal tolerance by the mother⁵⁸. In mice, when IL4 is deficient, a phenotype akin to preeclampsia (i.e., hypertension and proteinuria) develops⁵⁸. In humans, altered mid-pregnancy levels of circulating maternal IL4R have been associated with later preeclampsia⁵⁹. Women with preeclampsia have decreased levels of IL4 and increased levels of circulating IL4R^60–62. Additionally, specific combinations of maternal and fetal genetic variants within IL4R are enriched in preeclamptic pregnancies⁶³.

SELE encodes E-selectin; variants in SELE have been associated with coronary artery disease and hypertension^64–66. BAG3 encodes for the Bcl2-associated athanogene 3 (BAG3) protein, a member of the BAG family of co-chaperones that interact with the heat shock protein 70 ATPase domain. BAG3 increases in response to stressful stimuli and modulates key biologic processes, including development, apoptosis, cytoskeletal organization, and autophagy⁶⁷. Interestingly, genetic variants in BAG3 have been associated with cardiomyopathy^68,69, a condition significantly more likely in women with preeclampsia and with overlapping pathophysiology⁷⁰. PPP1R3B encodes the catalytic subunit of the serine/threonine phosphatase, protein phosphatase-1. PPP1R3B genetic variants influence lipid levels⁷¹ and C-reactive protein (CRP)^72,73. Genetic predisposition to dyslipidemia and altered CRP levels have been associated with increased preeclampsia risk^74,75. Women with hyperlipidemia⁷⁶ and elevated CRP⁷⁷ are at increased risk of preeclampsia (although the risk may be modified by BMI). PI3 kinase signaling is important for endothelial cell homeostasis and response to vascular injury^78,79, key processes disrupted in women with preeclampsia.

Strengths of this maternal preeclampsia GWAS include the size of the discovery sample (relative to prior studies), enrichment of early-onset and severe cases in the discovery cohort, inclusion of a replication cohort, and increased association of the lead signal that reaches study-wide significance in trans-ethnic association analysis. Limitations of the study include the lack of detailed clinical information on a subset of the preeclampsia cases, the use of population controls (leading to enhanced power to detect associations but increased likelihood of misclassification), and the lack of paired fetal samples. Additionally, as the genotyping array coverage was restricted to genes only, we are limited in our ability to determine if the lead SNP, rs9478812, is the causal variant responsible for increased maternal preeclampsia risk. Future work will focus on replication of rs9478812 in additional cohorts, fine-mapping the genomic region surrounding rs9478812, and functional follow-up of PLEKHG1 to determine how alterations at the locus predispose to the development of preeclampsia.

PERSPECTIVES.

In summary, we report here the largest multi-ethnic maternal preeclampsia GWAS study to date that reveals a multi-ethnic study-wide significant association within the PLEKHG1 locus, as well as associations with processes with established roles in maternal preeclampsia pathophysiology and long-term cardiovascular disease risk, including vascular, endothelial, and cellular stress pathways. This GWAS identifies and prioritizes a new genetic locus for future studies; these studies should aim to define the specific mechanisms by which alterations in this genomic region influence preeclampsia risk. Further delineation of maternal and fetal preeclampsia heritability will require the establishment of larger preeclampsia consortia with detailed clinical phenotyping, paired maternal-fetal samples, and genome-wide genotyping and/or sequencing data. These efforts have the potential to significantly advance our understanding of preeclampsia biology.

NOVELTY AND SIGNIFICANCE.

What Is New?

• Preeclampsia has significant heritability, but genetic risk factors are poorly understood.

• This is largest maternal genome-wide association study (GWAS) of preeclampsia to date and identifies a study-wide significant variant within the PLEKHG1 gene, a locus identified in GWAS of other traits including blood pressure, body weight, and neurological disorders.

What Is Relevant?

• This study highlights a plausible maternal gene candidate previously implicated in hypertension, providing a new biological clue into pathways dysregulated in preeclampsia.

Summary.

We report here results from the largest maternal multi-ethnic GWAS of preeclampsia, which reveals a novel locus associated with preeclampsia risk within the pleiotropic PLEKHG1 gene. While larger studies will be required to identify additional genetic associations, this study provides support for using GWAS to understand preeclampsia hereditability.