Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Literature-Informed Analysis of a Genome-Wide Association Study of Gestational Age in Norwegian Women and Children Suggests Involvement of Inflammatory Pathways

  • Jonas Bacelis ,

    jonas.bacelis@gu.se

    Affiliation Department of Obstetrics and Gynecology, Sahlgrenska University Hospital Östra, Gothenburg, Sweden

  • Julius Juodakis,

    Affiliation Department of Obstetrics and Gynecology, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden

  • Verena Sengpiel,

    Affiliation Department of Obstetrics and Gynecology, Sahlgrenska University Hospital Östra, Gothenburg, Sweden

  • Ge Zhang,

    Affiliations Human Genetics Division, Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio, United States of America, Center for Prevention of Preterm Birth, Perinatal Institute, Cincinnati Children’s Hospital Medical Center and March of Dimes Prematurity Research Center Ohio Collaborative, Cincinnati, Ohio, United States of America

  • Ronny Myhre,

    Affiliation Department of Genetics and Bioinformatics, Area of Health Data and Digitalisation, Norwegian Institute of Public Health, Oslo, Norway

  • Louis J. Muglia,

    Affiliation Center for Prevention of Preterm Birth, Perinatal Institute, Cincinnati Children’s Hospital Medical Center and March of Dimes Prematurity Research Center Ohio Collaborative, Cincinnati, Ohio, United States of America

  • Staffan Nilsson,

    Affiliation Department of Mathematical Sciences, Chalmers University of Technology, Gothenburg, Sweden

  • Bo Jacobsson

    Affiliations Department of Obstetrics and Gynecology, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden, Department of Genetics and Bioinformatics, Area of Health Data and Digitalisation, Norwegian Institute of Public Health, Oslo, Norway

Correction

19 Oct 2016: The PLOS ONE Staff (2016) Correction: Literature-Informed Analysis of a Genome-Wide Association Study of Gestational Age in Norwegian Women and Children Suggests Involvement of Inflammatory Pathways. PLOS ONE 11(10): e0165328. https://doi.org/10.1371/journal.pone.0165328 View correction

Abstract

Background

Five-to-eighteen percent of pregnancies worldwide end in preterm birth, which is the major cause of neonatal death and morbidity. Approximately 30% of the variation in gestational age at birth can be attributed to genetic factors. Genome-wide association studies (GWAS) have not shown robust evidence of association with genomic loci yet.

Methods

We separately investigated 1921 Norwegian mothers and 1199 children from pregnancies with spontaneous onset of delivery. Individuals were further divided based on the onset of delivery: initiated by labor or prelabor rupture of membranes. Genetic association with ultrasound-dated gestational age was evaluated using three genetic models and adaptive permutations. The top-ranked loci were tested for enrichment in 12 candidate gene-sets generated by text-mining PubMed abstracts containing pregnancy-related keywords.

Results

The six GWAS did not reveal significant associations, with the most extreme empirical p = 5.1 × 10−7. The top loci from maternal GWAS with deliveries initiated by labor showed significant enrichment in 10 PubMed gene-sets, e.g., p = 0.001 and 0.005 for keywords "uterus" and "preterm" respectively. Enrichment signals were mainly caused by infection/inflammation-related genes TLR4, NFKB1, ABCA1, MMP9. Literature-informed analysis of top loci revealed further immunity genes: IL1A, IL1B, CAMP, TREM1, TFRC, NFKBIA, MEFV, IRF8, WNT5A.

Conclusion

Our analyses support the role of inflammatory pathways in determining pregnancy duration and provide a list of 32 candidate genes for a follow-up work. We observed that the top regions from GWAS in mothers with labor-initiated deliveries significantly more often overlap with pregnancy-related genes than would be expected by chance, suggesting that increased sample size would benefit similar studies.

Introduction

The timing of human parturition is a poorly understood phenotype [1]. In the United States the reported rate of preterm birth (PTB), defined as birth occurring at less than 37 completed weeks of gestation, is 9.6% [2]. Worldwide PTB rates range from about 5% in some Northern European countries to 18% in Malawi [3]. PTB is the leading cause of death among neonates [4]. According to a US report, preterm born infants have a 15-fold higher mortality rate than those born at term [4]. More than 50% of deaths are attributable to only 2% of all infants—the ones who are born at less than 32 weeks of gestation [4]. PTB is also correlated with long-term adverse health consequences such as cerebral palsy, mental retardation, autism and schizophrenia, conditions that render individual dependent on the healthcare system. More than 50% of PTB occur in pregnancies without known risk factors. The currently available means of prediction (epidemiology- and biomarker-based models) and prevention (tocolytics, antibiotics, progesterone) are not effective enough to substantially reduce the rates of PTB and its adverse consequences [5].

Approximately 85% of all pregnancies have a spontaneous onset of delivery, with gestational age not affected by doctor’s decision to induce birth or to perform an elective caesarean section [6]. These pregnancies can be used for analysis of genetic factors affecting gestational age.

Up to 30% of variation in human gestational age could be accounted for by genetic factors, as reported by large register-based studies [7, 8]. The evidence of an acting genetic component motivated two genome-wide association studies (GWAS). In 2013 Uzun et al. [9] explored maternal genomes (884 preterm cases, 960 term controls). In 2015 Zhang et al. [10] investigated maternal (935 preterm cases, 946 term controls) and neonatal genomes (916 preterm cases, 935 term controls). The authors did not find robust statistical evidence of association between PTB and the 560 000 and 800 000 (respectively) single-nucleotide polymorphisms (SNPs) tested.

The failure to identify genes increasing the risk for PTB could be due to insufficient sample size, however it could also be due to the following problems: 1) preterm birth status has a lower information content than gestational age; 2) low accuracy of gestational age dating; 3) different onsets of delivery might reflect different aetiologies; 4) omission of genetic variants with low minor-allele frequency from analyses; 5) omission of non-additive genetic models in analyses; 6) mixed ethnicities in a study sample; 7) omission of prior knowledge about SNP function and the biological role of implicated genes. In our study we tried to avoid these shortcomings.

The aim of the study was to find SNPs that are associated with gestational age at birth. The use of gestational age, as opposed to the use of dichotomous PTB, provides an advantage, as it utilizes the full information present in the phenotype [11]. Our secondary aim was to highlight the genes that might mediate discovered associations, by identifying common biochemical pathways, networks, and functional similarities between the top genes. In the broadest sense, our study aims to account for a part of heritability of human gestational age at birth.

We structured our GWA study into six parts: investigating each of the subtypes (labor-initiated / PROM-initiated deliveries) separately and also together, while analysing maternal and fetal genomes separately.

Methods

The Dataset

Study population.

The Norwegian Mother and Child Cohort (MoBa) is a nationwide pregnancy cohort managed by the Norwegian Institute of Public Health [12]. It includes more than 107 000 pregnancies recruited from 1999 through 2008. Most of the pregnant women in Norway received a postal invitation in connection to the routine ultrasound screening at gestational weeks 17–19. Participation rate was 42.7%. For the current study, individuals were sampled from the Version 4 database containing 71 669 pregnancies. The MoBa dataset is linked to the Medical Birth Registry of Norway (MBRN), for additional information see [13].

For genotyping we selected mothers and live-born children from singleton pregnancies of mothers in the age group of 20–34 years resulting in a spontaneous onset of delivery. Pregnancies with complications (e.g., preeclampsia, gestational diabetes, placental abruption, placenta previa, cervical cerclage, small for gestational age, fetal malformation), pregnancies of mothers with pre-existing medical conditions (e.g., diabetes, hypertension, inflammatory bowel disease, systemic lupus erythematosus, rheumatoid arthritis), as well as pregnancies conceived by in vitro fertilization were excluded [14]. Random sampling was done from two gestational age ranges: 154–258 days (preterm births) and 273–286 days (term births), thus creating an oversampling of lower gestational ages (S1 Fig). In total 1921 mothers and 1199 children were selected for genotyping using blood-extracted DNA. All mothers gave a written consent to use anonymised data in scientific research. The Norwegian Regional Ethics Committee for Medical Research approved the study (REK/Sør-Øst 2010/2683 S-6075).

Phenotype and covariates.

We used gestational age expressed in days as a dependent variable, as continuous phenotype contains more information than a dichotomous case/control classification. MBRN provides an accurate second-trimester ultrasound-based evaluation of gestational age. Pregnancies initiated by labor were analysed separately from pregnancies starting with prelabor rupture of membranes (PROM), with one additional analysis where all pregnancies were considered together (Fig 1).

thumbnail
Fig 1. Schematic overview of the workflow in analyses.

C—child genomes, M—maternal genomes; add/rec/dom—additive, recessive and dominant genetic models respectively.

https://doi.org/10.1371/journal.pone.0160335.g001

We also used non-genotyped MoBa cohort data with more than 70 000 pregnancies to evaluate potential impact of known covariates and risk factors on gestational age. Together, the evaluated covariates explained only 1% of variation in the continuous phenotype. Since 22.8% of genotyped individuals did not have values for some of these covariates, we decided to use the larger sample of genotyped individuals and not to use adjustment.

Genotyping quality control.

The genotype missingness filter for SNPs and individuals was set to 3%. Individuals with heterozygosity estimates deviating by more than 3 SD from the group mean were removed. For each mother-mother or child-child pair related closer than second cousins, a random individual was removed. Hardy-Weinberg filter removed SNPs with p < 10−6. Non-Europeans were excluded after principal components analysis using the first three principal components and a threshold of 10 SD on the Euclidean distance from CEU cluster (HapMap). No minor-allele frequency filter was applied. Genomic inflation factor was estimated following standard procedures using continuous unadjusted gestational age in maternal samples (restricted to labor-initiated deliveries), linear regression, additive genetic model and minor-allele frequency restriction to > 0.06. All genomic positions are presented in hg19 coordinates.

Association tests

Three genetic models (additive, recessive, dominant) were used to test for association with unadjusted continuous gestational age expressed in days (Fig 1). Permutation procedures are essential to avoid biases introduced by skewed phenotype distribution (a notable feature of gestational age), and by low counts of individuals in the minor genotypic group. We used permutation-based testing implemented in PLINK (v1.90b2n, 64-bit, 2 Nov 2014, Linux), with parameters for adaptive permutations: alpha = 5×10−8, beta = 5×10−8, min = 10, max = 1×109, b = 1 and a = 0.001 [15]. Each SNP was assigned the most extreme empirical p-value from the three genetic models [16]: additive, recessive and dominant. X chromosomal SNPs were tested using only additive model. Two separate studies investigated our dataset for PTB association with X chromosomal SNPs [14] and mitochondrial SNPs [17] previously.

Gene-set enrichment analysis with INRICH

Clumping.

To merge adjacent and correlated SNPs, PLINK function “—clump” was used. Clumps were formed around “index variants” with p-value < 0.0005. Index variants were chosen greedily starting with the lowest p-value. Sites that were less than 250 kb away from an index variant, had r2 larger than 0.25 with it, and had association p-value smaller than 0.05 were assigned to that index variant's clump. The r2 values were computed using founders in the same genomic data.

PubMed gene-sets.

We checked if the top GWAS loci were enriched in genes with known relations to pregnancy or reproductive anatomy. To test this, we used 12 keywords to create 12 gene-sets by text-mining the PubMed database, as described in the next paragraph and Fig 1. Out of these, 4 keywords represent pregnancy conditions (“gestation”, “parturition”, “pregnancy”, “preterm”), another 4 describe female anatomy (“cervix”, “endometrium”, “myometrium”, “uterus”), and the last 4 portray fetal anatomy (“fetus/embryo”, “chorion”, “amnion”, “placenta”). We also created 16 gene-sets for keywords unrelated to pregnancy to be used as a control in enrichment analysis: 8 representing conditions and 8 representing anatomy (S1 Table).

Between June 1st and August 31st, 2015, the PubMed database was scanned for abstracts containing any semantic form or Latin/Greek form of the selected keyword together with the words indicating the genetic nature of a publication (“gene”, “genes”, “genomic”, “genetic” or “GWAS”; plus corresponding MeSH terms), but restricted to abstracts not containing 65 custom-built non-human subject indicators (e.g., “cat”, “feline”, “cow”, “bovine”) or 466 custom-built medical-condition indicators (e.g., listeriosis, erythema, hepatitis, neuroblastoma). The latter indicators were constructed by text-mining the ICD code database (www.cms.gov) and searching for words with common disease suffixes (e.g., "-osis", "-itis", "-emia", "-oma"). These restrictions were applied to avoid inclusion of genes that represent medical conditions or species not present in our GWAS data. Abstracts were mined searching for gene names by cross-referencing each capitalised word with 23 945 HGNC [18] gene names. We took precaution to avoid false identification of commonly used acronyms as gene names, e.g., gene AGA and “Apropriate for Gestational Age”, gene FGR and “Fetal Growth Retardation”, gene SPTB and “Spontaneous Preterm Birth”. In order to further reduce erroneous assignment of genes to keywords, only the genes mentioned in more than 1 abstract were used. In order to obtain a better representation of the keyword, we also used an "exclusivity" filter: the abstract must not contain more than one different keyword (with exception for very common and control keywords). All keywords and PubMed queries are listed in S1 Table.

Enrichment analysis.

Each clump produced by PLINK represents a genomic region defined by distance, linkage disequilibrium (LD) and statistical association with the phenotype. INRICH [19] is a tool that detects overlap between such regions and predefined gene sets and reports the empirically estimated p-value of enrichment. For this purpose INRICH iteratively generates random clumps of similar size and SNP-density and then creates a distribution of enrichment statistic under the null-hypothesis (“no enrichment”). P-values estimated with this method are expected to be robust and unbiased. Analysis was performed using the INRICH algorithm (v.1.0, updated Oct/24/2014, Linux). GWAS interval was considered to be a 'hit' for a predefined gene-set if it fell within 25 kb of any of the genes in that set, 100 000 permutations were used to estimate p-values for each gene-set, maintaining 90–110% SNP density match. The 300 top clumps from each of the six GWAS (mothers, children × labor, PROM, all) were tested against 12 pregnancy-related gene-sets and 16 control gene-sets from the PubMed abstract mining (Fig 1).

Literature-informed analysis of GWAS results.

By manually cross-referencing the 300 top SNPs from maternal GWAS in labor-initiated deliveries with the HaploReg v4.1 database (www.broadinstitute.org/mammals/haploreg, [20]) and with the scientific publication database MEDLINE, we selected biologically-relevant SNPs with their implicated genes. We grouped genes into categories, based on biological pathway that could modify gestational age. A prior evidence of association with gestational age / preterm birth, or evidence of interaction or functional/structural similarity among the top genes were used as the criteria for reporting genes in the result tables.

Results

Genotyping quality control

After quality control procedures of genotyping data, 1743 maternal and 1109 fetal samples were left and had relevant phenotypic data (1407 labor and 336 PROM mothers; 884 labor and 225 PROM children). The number of genotyped SNPs passing the quality-control procedures is 513 273 autosomal and 12 304 from the X chromosome. Mitochondrial, Y chromosomal SNPs and pseudo-autosomal SNPs were not analysed in this study. Principal components analysis of genotyping data assured that study individuals belong to a homogeneous European population. Geographical homogeneity was also reflected by genomic inflation factor, estimated to be 0.993 and indicating no population stratification effects in GWAS for this phenotype.

Association tests

None of the 525 577 SNPs tested with the additive, recessive and dominant genetic models showed a genome-wide significance (p < 5×10−8) in any of the six GWA analyses. The most extreme association was observed in a GWAS with PROM mothers (p = 5.1 × 10−7, SNP rs6977715 in the DPP6 gene). Due to the further-described findings in the post-GWAS analysis, in Fig 2 we present only the results from a GWAS of maternal genomes and labor-initiated deliveries, with the top 20 independent loci together with proximal genes highlighted in Table 1. The top results from the remaining GWA analyses are presented in S2 Fig and S1 File.

thumbnail
Fig 2. Manhattan plot for maternal GWAS of gestational age in labor-initiated deliveries.

In total 1 407 genomes were analysed. Each SNP was assigned the most extreme empirical p-value from the three genetic models (additive, recessive, dominant). The top line indicates a genome-wide significance level (5×10−8), while the bottom line marks a significance level (5×10−4) determining the number of “clumps” (independent loci that were used in gene-set enrichment analyses). Genes from gene-set enrichment analyses are marked in blue, while other biologically relevant genes (from the literature-informed analyses) are marked in black.

https://doi.org/10.1371/journal.pone.0160335.g002

thumbnail
Table 1. Top 20 independent loci from maternal GWAS of gestational age in labor-initiated deliveries.

https://doi.org/10.1371/journal.pone.0160335.t001

Table 1 was pruned to show only independent loci. BP—physical position on the chromosome in hg19 coordinates, P—the most extreme empirical p-value from three genetic models, E/R—the effect allele and the reference allele, Mod—the most significant genetic model for that SNP, nXX—number of individuals in each genotypic group, mXX—mean gestational age in each genotypic group. Interpretation of mean gestational age values should take into account the bimodal phenotype distribution of genotyped individuals (S1 Fig). Genes were assigned to SNPs based on a 100 kb offset rule. Asterisk (*) indicates a gene family with multiple genes in that locus. No multiple-test correction is applied. Bolded genes are described in the literature-informed analyses. Genes with unknown function (LINC, LOC etc.) are not listed.

The GWAS with labor-initiated deliveries and the GWAS with all deliveries shared approximately one-third and one-half of the top SNPs in maternal and fetal genomes respectively, while top SNPs from GWAS with PROM-initiated deliveries were mostly unique (Fig 3).

thumbnail
Fig 3. Overlap between top results in six GWAS.

The top 1000 SNPs were selected from each GWA analysis. Numbers in the Venn diagrams represent the number of SNPs. Numbers of individuals in each analysis were 1743, 1407, 336 (mothers) and 1109, 884, 225 (children) for all together, labor-initiated and PROM-initiated deliveries respectively.

https://doi.org/10.1371/journal.pone.0160335.g003

Gene-set enrichment analysis with INRICH

Gene-sets.

The sizes of the gene-sets in the PubMed-constructed pregnancy-themed group are as follows: 123 “preterm” genes, 214 “gestation” genes, 20 “parturition” genes, 540 “pregnancy” genes; maternal anatomy group: 59 “cervix” genes, 116 “endometrium” genes, 23 “myometrium” genes, 74 “uterus” genes; fetal anatomy group: 14 “fetus/embryo” genes, 35”chorion” genes, 45 “amnion” genes, 259 “placenta” genes. The full list of gene-set sizes with respective PubMed queries is shown in S1 Table. The full list of genes in each set is given in the S2 File.

Enrichment analysis.

Only the maternal GWAS with labor-initiated deliveries showed consistent enrichment in all relevant candidate gene-sets, and consistently showed no enrichment in the control gene-sets (Fig 4).

thumbnail
Fig 4. Enrichment in gene-sets generated using PubMed abstract text-mining.

The figure shows an overlap between the genes implicated in six GWA analyses (rows) and genes related to specific keywords (columns). The overlap is represented as probability (p-value) of similar or greater enrichment arising due to pure chance under the null hypothesis of no enrichment (i.e., if GWAS would rank genes in a random order). The 300 top independent loci (“clumps”) and their genes were used. The name of each gene-set indicates a keyword used in the PubMed abstract mining. The INRICH algorithm was used to estimate empirical p-values.

https://doi.org/10.1371/journal.pone.0160335.g004

In this particular analysis (mothers with labor-initiated deliveries), out of 300 selected top GWAS clumps, the INRICH algorithm removed 116 intervals without genes and then merged some of the remaining to form a final number of 178 independent (non-overlapping) genomic intervals. The top GWAS genes overlapping with candidate gene-sets are presented in Table 2 together with a probability of observing a similar or more extreme overlap under no genotype-phenotype association. The gene-set with the most significant enrichment corresponds to the keyword "uterus" (empirical p = 0.001). This gene-set contains 73 genes, 5 of which overlap with top GWAS intervals: ENG (endoglin), IGF2 (insulin-like growth factor 2), MMP9 (matrix metallopeptidase 9), NFKB1 (nuclear factor κ-B DNA binding subunit), TLR4 (toll-like receptor 4). These genes were also present in many other significantly enriched candidate gene-sets. Table 3 shows SNPs that implicated genes from Table 2, together with p-values from maternal GWAS using labor-initiated deliveries and genomic coordinates of respective clumped regions. Only 1 out of 16 control gene-sets ("ageing") was enriched (p = 0.05), while 10 out of 12 candidate gene-sets were enriched: all 4 pregnancy-themed sets, all 4 female anatomy sets, and 2 out of 4 fetal anatomy sets.

thumbnail
Table 2. Significantly enriched PubMed gene-sets in GWAS using mothers with labor-initiated deliveries.

https://doi.org/10.1371/journal.pone.0160335.t002

Literature-informed analysis of GWAS results

Manual inspection of the top 300 SNPs from maternal GWAS in labor-initiated deliveries highlighted 32 biologically relevant genes from 27 independent loci (Table 4). In total 284 genes had their biological background evaluated.

thumbnail
Table 4. Loci of biological relevance from maternal GWAS of gestational age in labor-initiated deliveries.

https://doi.org/10.1371/journal.pone.0160335.t004

The SNPs were selected from the top 300 GWAS results, based on their proximity and/or functional relationship with genes biologically relevant to gestational age. Rank—the rank of that SNP among all GWAS results, based on the most significant empirical p-value (P) from three genetic models, BP—physical position on the chromosome (Chr) in hg19 coordinates, E/R—the effect allele and the reference allele, Mod—the most significant genetic model for that SNP, nXX—number of individuals in each genotypic group, mXX—mean of gestational age in each genotypic group. Interpretation of mean gestational age values should take into account the bimodal phenotype distribution of genotyped individuals (S1 Fig). No multiple-test correction is applied.

We grouped these genes into four functional categories related to possible aetiologies of preterm birth: 1) bacterial or viral infection 2) utero-placental perfusion problems 3) cervical insufficiency 4) hormonal imbalance.

Infection.

Bacterial infection is a well-known cause of too short gestation [1]. We observed 14 SNPs that are known expression quantitative trait loci (eQTLs) for (or are located in proximity of) 17 immunity-related genes (Table 5). Most of these genes act through activation of nuclear factor complex NF-κB, a central regulator of the terminal processes in human labor and delivery [21].

Besides their individual connection to preterm birth via immunity mechanisms, ten genes from independent loci interact among each other: Pyrin encoded by MEFV decreases activation of NF-κB complex [24], which includes NFKB1; pellino protein encoded by PELI2 is necessary for activation of NF-κB complex; NF-κB activation is induced by lipopolysaccharide and interleukine encoded by IL1B; NFKB1 binds with IRF8 [51]; CAMP decreases expression of NFKB1 [52]; NFKBIA gene (independent region from NFKB1) product inhibits NFKB1 responses; NFKBIA affects the expression of TFRC, as a defence-from-bacterial-infection strategy [41]; IL1B increases NFKBIA expression; SPSB2 gene together with the MEFV gene share a SPRY domain, which is involved in innate immunity [53]; IL1B can increase expression of MMP9 [54].

Viral infection is also a potential cause of preterm birth [55]. In Table 6 we present biologically relevant "viral-immunity" genes identified by maternal GWAS in labor-initiated deliveries. During the pregnancy, the immune system actively supports the growing fetus. Viral infection weakens this function allowing other microorganisms to propagate and lead to preterm birth [56]. Five genes are known to bind to each other and are likely to play a role in the defense against viral infection by utilizing the RNA-induced silencing complex (RISC). Argonautes (encoded by AGO1, AGO3, AGO4) are the main components of RISC together with TNRC6A and TNRC6B (both TNRC genes are located on different chromosomes). The host can inhibit viral replication using a library of miRNAs that matches parts of viral RNA [57], and a RISC complex [58]. Moreover, the ability to suppress RISC was suggested as a counter-strategy deployed by viruses [59]. The ribonuclease subunit encoded by PAN3 binds to both TNRC proteins, while ANKMY2 binds to AGO3.

Utero-placental perfusion problems.

In Table 7 we show the second group of genes that are involved in utero-placental perfusion problems characterised by either utero-placental angiogenic imbalances (LEFTY2, ENG, KCNQ3, TIMP2, MMP9, ABCA1), maternal blood pressure (TOR2A, ARHGEF11, GNB3), or by compromised placentation (WNT4, WNT5A).

Cervical insufficiency.

Cervical ripening precedes the delivery and allows the fetus to pass through otherwise too-narrow outlet. Two genes described previously might also be involved in cervical ripening, compromising the structural integrity of extracellular matrix too early (Table 8).

Hormonal imbalance.

The fourth group represents three genes that are connected to hormonal problems (Table 9), which can lead to preterm birth.

Discussion

In our study, GWA analyses showed no genome-wide significant associations. However, using a gene-set enrichment analysis of GWA results, we found evidence that genes acting in mothers might contribute to gestational age in deliveries that start with labor. These genes are known for their involvement in processes that affect the duration of gestation (e.g., infection/inflammation).

Genome-wide association study

Using a standard genome-wide significance threshold of 5×10−8 none of the six GWA analyses revealed significant associations. Similarly as in previous study [10], we used two types of study individuals: mothers and children, as the genes affecting pregnancy might manifest via both genomes. We further stratified our analyses based on the type of delivery initiation: deliveries starting with PROM, deliveries that start with labor, and all pregnancies together (Fig 1). Instead of dichotomising a continuous phenotype (preterm and term groups), we directly utilised accurately dated (ultrasound-based method) gestational age, retaining phenotypic variability. The long tail of the skewed phenotype distribution was oversampled (S1 Fig) to gain more power to detect large effects. The samples used in our study were collected in a single country and represent ethnically homogenous population. We also investigated allelic interactions (dominance effects) that are likely to contribute to the broad-sense heritability estimates of gestational age [7]. Additionally, our study did not set arbitrary minor-allele frequency filters and used permutation-based association tests, which are less affected by phenotypic outliers or small counts in the minor genotypic group. We believe that these analytical aspects supplement the methods of preceding studies [9, 10].

The exploratory nature of our study (2 types of genomes × 3 types of onset of delivery × 3 genetic models) requires adequate corrections for multiple testing. However, as most of the tests are not independent, a simple Bonferroni correction would be overly conservative. We chose to present uncorrected p-values, at the same time cautioning the reader to remember that more statistical tests were done than in a single GWAS.

Gene-set enrichment analysis

Subsequent gene-set enrichment analyses indicated that one of our GWAS ranked markers in a biologically meaningful manner (Fig 4). Two previous GWA studies investigating preterm birth [9, 10] did not provide such evidence. Enrichment in known pregnancy-related genes justifies a closer inspection of top loci (see Literature-informed analyses) and warrants new GWA studies with larger sample sizes.

The results from gene-set enrichment analysis (Fig 4) illustrate the advantage of stratifying study subjects based on the onset of delivery. Only the GWAS investigating mothers with labor-initiated deliveries showed expected enrichment in pregnancy-related gene-sets and no enrichment in control gene-sets. The reasons for this could be that maternal genes play a more important role than the fetal. However, a smaller number of children (1.5-times less than mothers) could also explain this observation. Similarly, GWAS investigating PROM deliveries had a lower statistical power to detect associations (4-times smaller sample size) than GWAS investigating labor-initiated deliveries. Also, genetically determined gestational age in PROM pregnancies is likely to be shortened by environmental factors (e.g., the severity of the microbial invasion of the amniotic cavity), thus introducing noise and reducing the power of GWAS. Interestingly, even though analysis of mixed pregnancies had the largest sample size, it showed low enrichment in pregnancy-related genes. This observation suggests that gestational age determined by two onsets of delivery (labor and PROM) actually represents two separate endophenotypes.

Based on the results from gene-set enrichment analyses, in the literature-informed analyses we chose to investigate only the top SNPs from the maternal GWAS in labor-initiated deliveries.

Literature-informed overview of GWAS results

In the seminal publication by Romero et al. [92], the authors summarised the main pathological processes involved in the preterm parturition syndrome: (1) intrauterine infection/inflammation; (2) placental insufficiency (uteroplacental perfusion, angiogenic imbalances, decidualisation); (3) uterine overdistension and contractility; (4) abnormal allograft reaction; (5) allergy; (6) cervical insufficiency; (7) hormonal imbalance. Some genes implicated by the top 300 SNPs from maternal GWAS in labor-initiated deliveries could be comfortably assigned to these processes: infection/inflammation (NFKB1, TLR4, IRF8, ABCA1, TREML2, MEFV, WNT5A, NFKBIA), placental insufficiency (ENG, TOR2A, IGF2, KCNQ3, GNB3, LEFTY2, ARHGEF11, WNT4, WNT5A), cervical insufficiency (MMP9, TIMP2), and hormonal imbalance (WNT4, OPRM1, SP3).

We found 32 genes (Table 4) that 1) had suggestive evidence of association in GWA analysis, 2) were likely to have their function/expression affected by top GWAS SNPs, 3) had phenotype-relevant biological functions, and 4) their proteins formed clusters of interaction. Most of these genes belong to the "bacterial infection" group (Table 5).

Similar future studies might benefit from these observations: inclusion of recessive and dominant genetic models was advantageous, because allelic interactions (dominance effects) implicated approximately 90% of genes with biological relevance (Table 4). Similarly, 30% of genes would have been overlooked if a minor-allele frequency filter (MAF > 0.1) were to be applied, and over 50% would have been lost if GWAS sample size were to be increased by adding PROM-delivering mothers (N = 336) to the mothers with labor-initiated deliveries (N = 1407).

Replication studies should take into account that common infections in various geographical regions and climates might be caused by specific strains/species of bacteria. Similarly, different human populations might be unique in their immunity (vitamin D and sun exposure, vaccination policies, specific hygiene-related behaviours).

Infection/inflammation-related genes from our analyses (Table 5) could be used in gene-environment interaction (G×E) studies investigating how genotypes modulate the effect of infection-during-pregnancy on the gestational age at birth. Such studies could create the tools to identify women at high risk for delivering preterm.

Conclusion

In this study, no genome-wide significant associations with gestational age were found. We highlight 32 genes for the follow-up research, providing suggestive statistical evidence and biological relevance to gestational age, especially via inflammatory-pathways. Our study illustrates how post-GWAS analysis might give insights into the aetiology of the phenotype even without clear GWAS signals.

Supporting Information

S1 Fig. Phenotype distribution is six GWAS analyses.

Frequency denotes the number of individuals with a particular value of gestational age. The red line represents phenotype distribution in the whole MoBa cohort with same exclusion criteria applied as was for genotyped sample, only without case-oversampling. Maximal height of the red line was adjusted to match the histogram height. Individuals in different histograms might represent the same pregnancy.

https://doi.org/10.1371/journal.pone.0160335.s001

(TIFF)

S2 Fig. Manhattan plot for fetal GWAS of gestational age in labor-initiated deliveries.

In total 884 fetal genomes were used. Each SNP was assigned the most extreme empirical p-value from three genetic models (additive, recessive, dominant). The top line indicates a genome-wide significance level (5×10−8), while the bottom line marks a significance level (5×10−4) determining the number of “clumps” (independent loci that are used in gene-set enrichment analyses).

https://doi.org/10.1371/journal.pone.0160335.s002

(TIF)

S1 File. Results from all 6 GWA analyses.

Best_emp_P—the most extreme empirical p-value from three genetic models, Eff/Ref—the effect allele and the reference allele, Genetic model—the most significant genetic model for that SNP. Only SNPs with best_emp_P values ≤10−3 are shown.

https://doi.org/10.1371/journal.pone.0160335.s003

(ZIP)

S2 File. All genes from 12 pregnancy-related gene-sets.

https://doi.org/10.1371/journal.pone.0160335.s004

(ZIP)

S1 Table. Text-mining PubMed abstracts for pregnancy-related genes.

The table shows keywords and their queries used to search PubMed database. Numbers of keyword-related genes are shown before and after filtering.

https://doi.org/10.1371/journal.pone.0160335.s005

(XLSX)

Acknowledgments

We are grateful to all the families who take part in the on-going MoBa cohort, also, the Norwegian Ministry of Health, the Ministry of Education and Research, NIH/NIEHS and NIH/NINDS for funding it (grant No.1 UO1 NS 047537–01 and grant No.2 UO1 NS 047537-06A1).

Author Contributions

  1. Conceptualization: BJ JB VS.
  2. Data curation: JB JJ RM.
  3. Formal analysis: JB JJ.
  4. Funding acquisition: BJ JB.
  5. Investigation: BJ RM.
  6. Methodology: JB SN.
  7. Project administration: BJ JB.
  8. Resources: BJ.
  9. Software: JB JJ.
  10. Supervision: BJ.
  11. Validation: GZ SN LJM.
  12. Visualization: JB JJ.
  13. Writing - original draft: JB.
  14. Writing - review & editing: JJ GZ LJM SN BJ VS.

References

  1. 1. Muglia LJ, Katz M. The enigma of spontaneous preterm birth. N Engl J Med. 2010;362(6):529–35. pmid:20147718
  2. 2. Martin JA, Hamilton BE, Osterman MJ. Births in the United States, 2014. NCHS data brief. 2015(216):1–8.
  3. 3. Blencowe H, Cousens S, Oestergaard MZ, Chou D, Moller AB, Narwal R, et al. National, regional, and worldwide estimates of preterm birth rates in the year 2010 with time trends since 1990 for selected countries: a systematic analysis and implications. Lancet. 2012;379(9832):2162–72. pmid:22682464
  4. 4. In: Behrman RE, Butler AS, editors. Preterm Birth: Causes, Consequences, and Prevention. The National Academies Collection: Reports funded by National Institutes of Health. Washington (DC)2007.
  5. 5. Iams JD, Romero R, Culhane JF, Goldenberg RL. Primary, secondary, and tertiary interventions to reduce the morbidity and mortality of preterm birth. Lancet. 2008;371(9607):164–75. pmid:18191687
  6. 6. Lukasse M, Helbig A, Benth JS, Eberhard-Gran M. Antenatal maternal emotional distress and duration of pregnancy. PLOS One. 2014;9(7):e101682. pmid:25000409
  7. 7. Wu W, Witherspoon DJ, Fraser A, Clark EA, Rogers A, Stoddard GJ, et al. The heritability of gestational age in a two-million member cohort: implications for spontaneous preterm birth. Hum Genet. 2015;134(7):803–8. pmid:25920518
  8. 8. Clausson B, Lichtenstein P, Cnattingius S. Genetic influence on birthweight and gestational length determined by studies in offspring of twins. BJOG. 2000;107(3):375–81. pmid:10740335
  9. 9. Uzun A, Dewan AT, Istrail S, Padbury JF. Pathway-based genetic analysis of preterm birth. Genomics. 2013;101(3):163–70. pmid:23298525
  10. 10. Zhang H, Baldwin DA, Bukowski RK, Parry S, Xu Y, Song C, et al. A genome-wide association study of early spontaneous preterm delivery. Genet Epidemiol. 2015;39(3):217–26. pmid:25599974
  11. 11. Auger N, Abrahamowicz M, Wynant W, Lo E. Gestational age-dependent risk factors for preterm birth: associations with maternal education and age early in gestation. Eur J Obstet Gynecol Reprod Biol. 2014;176:132–6. pmid:24666799
  12. 12. Magnus P, Irgens LM, Haug K, Nystad W, Skjaerven R, Stoltenberg C, et al. Cohort profile: the Norwegian Mother and Child Cohort Study (MoBa). Int J Epidemiol. 2006;35(5):1146–50. pmid:16926217
  13. 13. Irgens LM. The Medical Birth Registry of Norway; a source for epidemiological and clinical research. Scand J Rheumatol Suppl. 1998;107:105–8. pmid:9759145
  14. 14. Myking S, Boyd HA, Myhre R, Feenstra B, Jugessur A, Devold Pay AS, et al. X-chromosomal maternal and fetal SNPs and the risk of spontaneous preterm delivery in a Danish/Norwegian genome-wide association study. PLOS One. 2013;8(4):e61781. pmid:23613933
  15. 15. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. GigaScience. 2015;4:7. pmid:25722852
  16. 16. Tanikawa C, Urabe Y, Matsuo K, Kubo M, Takahashi A, Ito H, et al. A genome-wide association study identifies two susceptibility loci for duodenal ulcer in the Japanese population. Nat Genet. 2012;44(4):430–4, S1-2. pmid:22387998
  17. 17. Alleman BW, Myking S, Ryckman KK, Myhre R, Feingold E, Feenstra B, et al. No observed association for mitochondrial SNPs with preterm delivery and related outcomes. Pediatr Res. 2012;72(5):539–44. pmid:22902432
  18. 18. Gray KA, Yates B, Seal RL, Wright MW, Bruford EA. Genenames.org: the HGNC resources in 2015. Nucleic Acids Res. 2015;43(Database issue):D1079–85. pmid:25361968
  19. 19. Lee PH, O'Dushlaine C, Thomas B, Purcell SM. INRICH: interval-based enrichment analysis for genome-wide association studies. Bioinformatics. 2012;28(13):1797–9. pmid:22513993
  20. 20. Ward LD, Kellis M. HaploReg: a resource for exploring chromatin states, conservation, and regulatory motif alterations within sets of genetically linked variants. Nucleic Acids Res. 2012;40(D1):D930–D4.
  21. 21. Lappas M. NOD1 and NOD2 regulate proinflammatory and prolabor mediators in human fetal membranes and myometrium via nuclear factor-kappa B. Biol Reprod. 2013;89(1):14. pmid:23740944
  22. 22. Westra HJ, Peters MJ, Esko T, Yaghootkar H, Schurmann C, Kettunen J, et al. Systematic identification of trans eQTLs as putative drivers of known disease associations. Nat Genet. 2013;45(10):1238–43. pmid:24013639
  23. 23. MacIntyre DA, Lee YS, Migale R, Herbert BR, Waddington SN, Peebles D, et al. Activator protein 1 is a key terminal mediator of inflammation-induced preterm labor in mice. FASEB J. 2014;28(5):2358–68. pmid:24497579
  24. 24. Stehlik C, Fiorentino L, Dorfleutner A, Bruey JM, Ariza EM, Sagara J, et al. The PAAD/PYRIN-family protein ASC is a dual regulator of a conserved step in nuclear factor kappaB activation pathways. J Exp Med. 2002;196(12):1605–15. pmid:12486103
  25. 25. Fumagalli M, Cagliani R, Pozzoli U, Riva S, Comi GP, Menozzi G, et al. A population genetics study of the familial Mediterranean fever gene: evidence of balancing selection under an overdominance regime. Genes Immun. 2009;10(8):678–86. pmid:19675583
  26. 26. Matoba N, Yu Y, Mestan K, Pearson C, Ortiz K, Porta N, et al. Differential patterns of 27 cord blood immune biomarkers across gestational age. Pediatrics. 2009;123(5):1320–8. pmid:19403498
  27. 27. Ramirez-Ortiz ZG, Prasad A, Griffith JW, Pendergraft WF 3rd, Cowley GS, Root DE, et al. The receptor TREML4 amplifies TLR7-mediated signaling during antiviral responses and autoimmunity. Nat Immunol. 2015;16(5):495–504. pmid:25848864
  28. 28. Hanna N, Bonifacio L, Reddy P, Hanna I, Weinberger B, Murphy S, et al. IFN-gamma-mediated inhibition of COX-2 expression in the placenta from term and preterm labor pregnancies. Am J Reprod Immunol. 2004;51(4):311–8. pmid:15212685
  29. 29. Curry AE, Vogel I, Drews C, Schendel D, Skogstrand K, Flanders WD, et al. Mid-pregnancy maternal plasma levels of interleukin 2, 6, and 12, tumor necrosis factor-alpha, interferon-gamma, and granulocyte-macrophage colony-stimulating factor and spontaneous preterm delivery. Acta Obstet Gynecol Scand. 2007;86(9):1103–10. pmid:17712652
  30. 30. Liu Y, Yao CJ, Tao FB, Luo CM, Cao Y, Zhao SJ, et al. Association between maternal tumor necrosis factor-alpha G308A polymorphism and interferon-gamma A874T polymorphism and risk of preterm birth: a meta-analysis. Eur J Obstet Gyn R B. 2015;190:11–9.
  31. 31. Frew L, Makieva S, McKinlay AT, McHugh BJ, Doust A, Norman JE, et al. Human cathelicidin production by the cervix. PLOS One. 2014;9(8):e103434. pmid:25089904
  32. 32. Lim R, Barker G, Lappas M. Human cathelicidin antimicrobial protein 18 (hCAP18/LL-37) is increased in foetal membranes and myometrium after spontaneous labour and delivery. J Reprod Immunol. 2015;107:31–42. pmid:25435436
  33. 33. Consortium GT. Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans. Science. 2015;348(6235):648–60. pmid:25954001
  34. 34. Steffen KM, Cooper ME, Shi M, Caprau D, Simhan HN, Dagle JM, et al. Maternal and fetal variation in genes of cholesterol metabolism is associated with preterm delivery. J Perinatol. 2007;27(11):672–80. pmid:17855807
  35. 35. Korhonen JT, Olkkonen VM, Lahesmaa R, Puolakkainen M. ABC-cassette transporter 1 (ABCA1) expression in epithelial cells in Chlamydia pneumoniae infection. Microb Pathog. 2013;61–62:57–61. pmid:23707398
  36. 36. Okuhira K, Fitzgerald ML, Tamehiro N, Ohoka N, Suzuki K, Sawada J, et al. Binding of PDZ-RhoGEF to ATP-binding cassette transporter A1 (ABCA1) induces cholesterol efflux through RhoA activation and prevention of transporter degradation. J Biol Chem. 2010;285(21):16369–77. pmid:20348106
  37. 37. Blumenthal A, Ehlers S, Lauber J, Buer J, Lange C, Goldmann T, et al. The Wingless homolog WNT5A and its receptor Frizzled-5 regulate inflammatory responses of human mononuclear cells induced by microbial stimulation. Blood. 2006;108(3):965–73. pmid:16601243
  38. 38. He W, Wang Z, Zhou Z, Zhang Y, Zhu Q, Wei K, et al. Lipopolysaccharide enhances Wnt5a expression through toll-like receptor 4, myeloid differentiating factor 88, phosphatidylinositol 3-OH kinase/AKT and nuclear factor kappa B pathways in human dental pulp stem cells. J Endod. 2014;40(1):69–75. pmid:24331994
  39. 39. Vuga LJ, Ben-Yehudah A, Kovkarova-Naumovski E, Oriss T, Gibson KF, Feghali-Bostwick C, et al. WNT5A is a regulator of fibroblast proliferation and resistance to apoptosis. Am J Respir Cell Mol Biol. 2009;41(5):583–9. pmid:19251946
  40. 40. Honest H, Bachmann LM, Gupta JK, Kleijnen J, Khan KS. Accuracy of cervicovaginal fetal fibronectin test in predicting risk of spontaneous preterm birth: systematic review. BMJ. 2002;325(7359):301. pmid:12169504
  41. 41. Skaar EP. The battle for iron between bacterial pathogens and their vertebrate hosts. PLOS Pathog. 2010;6(8):e1000949. pmid:20711357
  42. 42. Verstraelen H, Delanghe J, Roelens K, Blot S, Claeys G, Temmerman M. Subclinical iron deficiency is a strong predictor of bacterial vaginosis in early pregnancy. BMC Infect Dis. 2005;5:55. pmid:16000177
  43. 43. Gao T, Zablith NR, Burns DH, Skinner CD, Koski KG. Second trimester amniotic fluid transferrin and uric acid predict infant birth outcomes. Prenat Diagn. 2008;28(9):810–4. pmid:18646241
  44. 44. Soubasi V, Petridou S, Sarafidis K, Tsantali C, Diamanti E, Buonocore G, et al. Association of increased maternal ferritin levels with gestational diabetes and intra-uterine growth retardation. Diabetes Metab. 2010;36(1):58–63. pmid:20074991
  45. 45. Kuang Z, Lewis RS, Curtis JM, Zhan Y, Saunders BM, Babon JJ, et al. The SPRY domain-containing SOCS box protein SPSB2 targets iNOS for proteasomal degradation. J Cell Biol. 2010;190(1):129–41. pmid:20603330
  46. 46. Pawelczyk E, Nowicki BJ, Izban MG, Pratap S, Sashti NA, Sanderson M, et al. Spontaneous preterm labor is associated with an increase in the proinflammatory signal transducer TLR4 receptor on maternal blood monocytes. BMC Pregnancy Childbirth. 2010;10.
  47. 47. Li L, Kang J, Lei W. Role of Toll-like receptor 4 in inflammation-induced preterm delivery. Mol Hum Reprod. 2010;16(4):267–72. pmid:19995880
  48. 48. Ferrand PE, Parry S, Sammel M, Macones GA, Kuivaniemi H, Romero R, et al. A polymorphism in the matrix metalloproteinase-9 promoter is associated with increased risk of preterm premature rupture of membranes in African Americans. Mol Hum Reprod. 2002;8(5):494–501. pmid:11994547
  49. 49. Sata F, Toya S, Yamada H, Suzuki K, Saijo Y, Yamazaki A, et al. Proinflammatory cytokine polymorphisms and the risk of preterm birth and low birthweight in a Japanese population. Mol Hum Reprod. 2009;15(2):121–30. pmid:19141488
  50. 50. Genc MR, Gerber S, Nesin M, Witkin SS. Polymorphism in the interleukin-1 gene complex and spontaneous preterm delivery. Am J Obstet Gynecol. 2002;187(1):157–63. pmid:12114904
  51. 51. Drew PD, Franzoso G, Carlson LM, Biddison WE, Siebenlist U, Ozato K. Interferon regulatory factor-2 physically interacts with NF-kappa B in vitro and inhibits NF-kappa B induction of major histocompatibility class I and beta 2-microglobulin gene expression in transfected human neuroblastoma cells. J Neuroimmunol. 1995;63(2):157–62. pmid:8550813
  52. 52. Mookherjee N, Brown KL, Bowdish DM, Doria S, Falsafi R, Hokamp K, et al. Modulation of the TLR-mediated inflammatory response by the endogenous human host defense peptide LL-37. J Immunol. 2006;176(4):2455–64. pmid:16456005
  53. 53. D'Cruz AA, Babon JJ, Norton RS, Nicola NA, Nicholson SE. Structure and function of the SPRY/B30.2 domain proteins involved in innate immunity. Protein Sci. 2013;22(1):1–10. pmid:23139046
  54. 54. Lockwood CJ, Oner C, Uz YH, Kayisli UA, Huang SJ, Buchwalder LF, et al. Matrix metalloproteinase 9 (MMP9) expression in preeclamptic decidua and MMP9 induction by tumor necrosis factor alpha and interleukin 1 beta in human first trimester decidual cells. Biol Reprod. 2008;78(6):1064–72. pmid:18276934
  55. 55. Cardenas I, Means RE, Aldo P, Koga K, Lang SM, Booth CJ, et al. Viral infection of the placenta leads to fetal inflammation and sensitization to bacterial products predisposing to preterm labor. J Immunol. 2010;185(2):1248–57. pmid:20554966
  56. 56. Silasi M, Cardenas I, Kwon JY, Racicot K, Aldo P, Mor G. Viral infections during pregnancy. Am J Reprod Immunol. 2015;73(3):199–213. pmid:25582523
  57. 57. Swaminathan G, Martin-Garcia J, Navas-Martin S. RNA viruses and microRNAs: challenging discoveries for the 21st century. Physiol Genomics. 2013;45(22):1035–48. pmid:24046280
  58. 58. Murakami Y, Aly HH, Tajima A, Inoue I, Shimotohno K. Regulation of the hepatitis C virus genome replication by miR-199a. J Hepatol. 2009;50(3):453–60. pmid:19144437
  59. 59. Andersson MG, Haasnoot PC, Xu N, Berenjian S, Berkhout B, Akusjarvi G. Suppression of RNA interference by adenovirus virus-associated RNA. J Virol. 2005;79(15):9556–65. pmid:16014917
  60. 60. Lappalainen T, Sammeth M, Friedlander MR, t Hoen PA, Monlong J, Rivas MA, et al. Transcriptome and genome sequencing uncovers functional variation in humans. Nature. 2013;501(7468):506–11. pmid:24037378
  61. 61. Park CB, Dufort D. NODAL signaling components regulate essential events in the establishment of pregnancy. Reproduction. 2013;145(2):R55–64. pmid:23221013
  62. 62. Maynard SE, Moore Simas TA, Bur L, Crawford SL, Solitro MJ, Meyer BA. Soluble endoglin for the prediction of preeclampsia in a high risk cohort. Hypertens Pregnancy. 2010;29(3):330–41. pmid:20670156
  63. 63. Shichiri M, Ishimaru S, Ota T, Nishikawa T, Isogai T, Hirata Y. Salusins: newly identified bioactive peptides with hemodynamic and mitogenic activities. Nat Med. 2003;9(9):1166–72. pmid:12910263
  64. 64. Celik E, Celik O, Yilmaz E, Turkcuoglu I, Karaer A, Turhan U, et al. Association of low maternal levels of salusins with gestational diabetes mellitus and with small-for-gestational-age fetuses. Eur J Obstet Gynecol Reprod Biol. 2013;167(1):29–33. pmid:23178004
  65. 65. Cha J, Sun X, Dey SK. Mechanisms of implantation: strategies for successful pregnancy. Nat Med. 2012;18(12):1754–67. pmid:23223073
  66. 66. Ugocsai P, Hohenstatt A, Paragh G, Liebisch G, Langmann T, Wolf Z, et al. HIF-1beta determines ABCA1 expression under hypoxia in human macrophages. Int J Biochem Cell Biol. 2010;42(2):241–52. pmid:19828131
  67. 67. Fang L, Choi SH, Baek JS, Liu C, Almazan F, Ulrich F, et al. Control of angiogenesis by AIBP-mediated cholesterol efflux. Nature. 2013;498(7452):118–22. pmid:23719382
  68. 68. Cha J, Bartos A, Park C, Sun X, Li Y, Cha SW, et al. Appropriate crypt formation in the uterus for embryo homing and implantation requires Wnt5a-ROR signaling. Cell reports. 2014;8(2):382–92. pmid:25043182
  69. 69. Ying Z, Giachini FR, Tostes RC, Webb RC. PYK2/PDZ-RhoGEF links Ca2+ signaling to RhoA. Arterioscler Thromb Vasc Biol. 2009;29(10):1657–63. pmid:19759375
  70. 70. Chen YP, Lu YP, Li J, Liu ZW, Chen WJ, Liang XJ, et al. Fetal and maternal angiotensin (1–7) are associated with preterm birth. J Hypertens. 2014;32(9):1833–41. pmid:24979298
  71. 71. O'Brien M, Flynn D, Mullins B, Morrison JJ, Smith TJ. Expression of RHOGTPase regulators in human myometrium. Reprod Biol Endocrinol. 2008;6:1. pmid:18190708
  72. 72. McElroy JJ, Gutman CE, Shaffer CM, Busch TD, Puttonen H, Teramo K, et al. Maternal coding variants in complement receptor 1 and spontaneous idiopathic preterm birth. Hum Genet. 2013;132(8):935–42. pmid:23591632
  73. 73. Zhu JY, Pang ZJ, Yu YH. Regulation of trophoblast invasion: the role of matrix metalloproteinases. Rev Obstet Gynecol. 2012;5(3–4):e137–43. pmid:23483768
  74. 74. Goldman-Wohl D, Yagel S. Regulation of trophoblast invasion: from normal implantation to pre-eclampsia. Mol Cell Endocrinol. 2002;187(1–2):233–8. pmid:11988332
  75. 75. Romero R, Dey SK, Fisher SJ. Preterm labor: one syndrome, many causes. Science. 2014;345(6198):760–5. pmid:25124429
  76. 76. Romero R, Velez Edwards DR, Kusanovic JP, Hassan SS, Mazaki-Tovi S, Vaisbuch E, et al. Identification of fetal and maternal single nucleotide polymorphisms in candidate genes that predispose to spontaneous preterm labor with intact membranes. Am J Obstet Gynecol. 2010;202(5):431 e1–34. pmid:20452482
  77. 77. Bae Y, Park C, Han J, Hong YJ, Song HH, Shin ES, et al. Interaction between GNB3 C825T and ACE I/D polymorphisms in essential hypertension in Koreans. J Hum Hypertens. 2007;21(2):159–66. pmid:17066084
  78. 78. Mistry HD, Kurlak LO, Whitley GS, Cartwright JE, Tribe RM, Broughton Pipkin F. OS081. Novel KCNQ3/KCNE5 isoform protein and mRNA expression in first trimester human placentae. Pregnancy Hypertens. 2012;2(3):221–2.
  79. 79. Mistry HD, McCallum LA, Kurlak LO, Greenwood IA, Broughton Pipkin F, Tribe RM. Novel expression and regulation of voltage-dependent potassium channels in placentas from women with preeclampsia. Hypertension. 2011;58(3):497–504. pmid:21730298
  80. 80. Osmers RG, Adelmann-Grill BC, Rath W, Stuhlsatz HW, Tschesche H, Kuhn W. Biochemical events in cervical ripening dilatation during pregnancy and parturition. J Obstet Gynaecol (Tokyo 1995). 1995;21(2):185–94.
  81. 81. Biason-Lauber A, De Filippo G, Konrad D, Scarano G, Nazzaro A, Schoenle EJ. WNT4 deficiency—a clinical phenotype distinct from the classic Mayer-Rokitansky-Kuster-Hauser syndrome: a case report. Hum Reprod. 2007;22(1):224–9. pmid:16959810
  82. 82. Philibert P, Biason-Lauber A, Rouzier R, Pienkowski C, Paris F, Konrad D, et al. Identification and functional analysis of a new WNT4 gene mutation among 28 adolescent girls with primary amenorrhea and mullerian duct abnormalities: a French collaborative study. J Clin Endocrinol Metab. 2008;93(3):895–900. pmid:18182450
  83. 83. Kiewisz J, Kaczmarek MM, Andronowska A, Blitek A, Ziecik AJ. Gene expression of WNTs, beta-catenin and E-cadherin during the periimplantation period of pregnancy in pigs—involvement of steroid hormones. Theriogenology. 2011;76(4):687–99. pmid:21652061
  84. 84. Sultan C, Biason-Lauber A, Philibert P. Mayer-Rokitansky-Kuster-Hauser syndrome: recent clinical and genetic findings. Gynecol Endocrinol. 2009;25(1):8–11. pmid:19165657
  85. 85. Lovvik TS, Wikstrom AK, Neovius M, Stephansson O, Roos N, Vanky E. Pregnancy and perinatal outcomes in women with polycystic ovary syndrome and twin births: a population-based cohort study. BJOG. 2015;122(10):1295–302. pmid:25761516
  86. 86. Zhang G, Bacelis J, Lengyel C, Teramo K, Hallman M, Helgeland O, et al. Assessing the Causal Relationship of Maternal Height on Birth Size and Gestational Age at Birth: A Mendelian Randomization Analysis. PLOS Med. 2015;12(8):e1001865. pmid:26284790
  87. 87. Cheng YH, Imir A, Suzuki T, Fenkci V, Yilmaz B, Sasano H, et al. SP1 and SP3 mediate progesterone-dependent induction of the 17beta hydroxysteroid dehydrogenase type 2 gene in human endometrium. Biol Reprod. 2006;75(4):605–14. pmid:16807381
  88. 88. Pettersson FD, Gronbladh A, Nyberg F, Sundstrom-Poromaa I, Akerud H. The A118G single-nucleotide polymorphism of human mu-opioid receptor gene and use of labor analgesia. Reprod Sci. 2012;19(9):962–7. pmid:22527985
  89. 89. Zhu Y, Pintar JE. Expression of opioid receptors and ligands in pregnant mouse uterus and placenta. Biol Reprod. 1998;59(4):925–32. pmid:9746745
  90. 90. Faletti A, Bassi D, Gimeno AL, Gimeno MA. Effects of beta-endorphin on spontaneous uterine contractions. Prostaglandins production and 45Ca2+ uptake in uterine strips from ovariectomized rats. Prostaglandins Leukot Essent Fatty Acids. 1992;47(1):29–33. pmid:1359570
  91. 91. Prufer K, Racimo F, Patterson N, Jay F, Sankararaman S, Sawyer S, et al. The complete genome sequence of a Neanderthal from the Altai Mountains. Nature. 2014;505(7481):43–9. pmid:24352235
  92. 92. Romero R, Espinoza J, Kusanovic JP, Gotsch F, Hassan S, Erez O, et al. The preterm parturition syndrome. BJOG. 2006;113 Suppl 3:17–42. pmid:17206962