The present study is a nested case-control study, embedded in a prospective study from the Guangzhou Women and Children's Medical Center (GWCMC) that was designed to investigate the health consequences of genetic and environmental factors on pregnancy outcomes and offspring. Pregnant women who attended the routine oral glucose tolerance test were invited to participate in the study at 24–28 weeks of gestation, if they met the following inclusion criteria: ≥18 years old, Chinese and intend to deliver at GWCMC. From April 2013 to March 2014, 3597 (69.0% of those eligible) pregnant women were enrolled to this prospective study and donated their blood samples.

Of those who gave singleton live birth, 174 women had preterm deliveries (<37 weeks of gestation). Women with medically indicated preterm delivery (n=53), in vitro fertilization (n=5), fetal anomaly (n=1) or uterine malformation (n=1) were excluded, resulting in 114 women having a spontaneous preterm birth (42 preterm labor and 72 preterm premature rupture of membranes) in the present study as the case group. In addition, 250 pregnant women, who delivered singleton term babies (38–41 weeks of gestation) and had no preeclampsia, eclampsia, placental abruption, placenta previa, haemolysis, elevated liver enzymes or low platelet count syndrome, were selected and included as the control group. Controls were frequency-matched to the cases by age within 3 years. The study was approved by the Institutional Review Board of GWCMC. Informed consent was obtained from all participants.

To select SNPs, the authors conducted a systematic literature review, assessing the evidence for the associations of PLA2G4C and PLA2G4D with SPTB. For the PLA2G4C gene, four SNPs had been reported to influence preterm birth risk in the US Hispanic or White populations (18). However, three of them (rs8110925, rs2307276 and rs11564620) were excluded due to the minor allele frequencies (MAF) reported in HapMap (http://hapmap.ncbi.nlm.nih.gov/) were <10% for Chinese subjects. Only rs1366442 in the intron region with a MAF of 13% was chosen to ensure that an adequate number of individuals within the population would be carriers of the minor allele. For the PLA2G4D gene, published reports on polymorphisms of this gene and diseases are rare. The authors chose rs4924618 based on the following two reasons: i) a weak association between this SNP and schizophrenia was observed in a Chinese population (20), while schizophrenia was related to a higher risk of preterm birth (22,23); ii) this SNP is a non-synonymous SNP that is likely to alter the protein folding structure, suggesting that this SNP may have a functional effect.

Blood samples were stored at −80°C until laboratory use. Genomic DNA was extracted from cells isolated from 2 ml of blood samples using the Qiagen DNA Blood Mini kit (Qiagen, Inc., Valencia, CA, USA) according to the manufacturer's instructions. Genotyping for PLA2G4C rs1366442 and PLA2G4D rs4924618 was performed by using a MassARRAY system (Sequenom, San Diego, CA, USA). A total of 20 samples (5.5% of total) were tested in duplicates to determine the quality of genotyping, and the concordance rates were 100%. DNA from 4 (1.1%) and 5 (1.4%) samples failed to be genotyped for PLA2G4C rs1366442 and PLA2G4D rs4924618.

Information on maternal age, educational level, smoking status, maternal height and pre-pregnancy weight, parity and previous history of preterm delivery were collected by using face to face questionnaire. In addition, diagnosis of gestational diabetes mellitus (GDM), infant's gestational age, birth weight and sex, were abstracted from medical records. Gestational age was evaluated based on the ultrasound examination in the first or second trimester. Pre-pregnancy body mass index (BMI; kg/m²) was calculated as the ratio of pre-pregnancy weight (kg) to squared height (m). Since the prevalence of previous history of preterm delivery and smoking exposure were quite low (≤1% in the control group), the authors did not include these variables for further analysis.

The differences between the case and control groups were evaluated by using the chi-squared test or Fisher's exact test (for categorical variables) and Student's t-test (for continuous variables). The Hardy-Weinberg equilibrium for PLA2G4C rs1366442 and PLA2G4D rs4924618 was examined by a goodness-of-fit chi-squared test among the control group. The multivariate logistic regression was used to investigate the relationship between genotypes and the risk of preterm birth, adjusting for maternal age, educational level (high school or below vs. college or above), pre-pregnancy BMI, parity (primiparous vs. multiparous), GDM status (yes vs. no) and infant's sex. The odds ratios (ORs) and the corresponding 95% confidence intervals (CIs) were calculated. Stratified analysis was performed to examine whether the effects of genotypes with the risk of preterm birth varied across different status of maternal age (<30 vs. ≥30), GDM status (no vs. yes) and infant's sex (female vs. male). The multiplicative interaction between genotypes and participants' characteristics were evaluated in the logistic regression models.

A two-tailed P<0.05 was considered statistically significant. All statistical analysis was performed using SAS statistical software (version, 9.3; SAS Institute, Cary, NC, USA).

The structure of cPLA2 δ catalytic domain was constructed through homology modeling. Sequences of four cytosolic phospholipase A2 (cPLA2) (α, β, γ and δ) were obtained from UniProt (www.uniprot.org). To check the identity and homology, sequences of cPLA2 catalytic domains were compared by BLAST using Basic Local Alignment Search Tool (http://blast.ncbi.nlm.nih.gov/Blast.cgi) and ClustalW (https://www.ebi.ac.uk/Tools/msa/clustalw2/). cPLA2 δ catalytic domain is localized between amino acids 273 and 818 and the cPLA2 α (Protein Data Bank ID: 1CJY) was chosen as a template. The SWISS-model workspace (https://swissmodel.expasy.org/; Swiss Institute of Bioinformatics, Basel, Switzerland) was used to construct the homology model of cPLA2 δ to obtain a structural insight of the protein and to locate the SNP-related amino acid in the protein. Manual alignment of cPLA2 δ to the template was exported in FASTA format to the Swiss-Model Server (24). The homology model was then validated by the Ramachandran plot plugin in VMD 1.9.2 (http://www.ks.uiuc.edu/; Theoretical and Computational Biophysics Group, NIH Center for Macromolecular Moedling and Bioinformatics, at the Beckman Institute, University of Illinois at Urbana-Champaign, Champaign, IL, USA) to determine stereochemical aspects along with main chain and side chain parameters (25).

The baseline characteristics of the studied population, including 114 SPTB cases and 250 controls, are presented in Table I. No significant differences were observed in the distributions of maternal age, educational level, history of abortion, history of preterm delivery, pre-pregnancy BMI, smoking status and GDM status between the cases and controls (P>0.05). As expected, gestational age at delivery (P<0.001) and birth weight (P<0.001) were significantly lower in the case group, when compared with those in the control group, respectively. In addition, cases were more likely to be multiparous (P=0.01) and neonates from the case group had higher percentage of boys (P=0.001).

Genotype distributions of the two selected SNPs in cases and controls are summarized in Table II. There was no deviation from the Hardy-Weinberg equilibrium observed among the controls (P=0.94 for rs1366442 and P=0.58 for rs4924618). Logistic regression analysis was performed to assess the effects of each SNP on SPTB risk. In the multivariate analysis, compared to the homozygous genotype GG, the TT genotypes of PLA2G4D rs4924618 was associated with a marginally reduced risk of SPTB (adjusted OR and 95% CI, 0.50 [0.24–1.02]), and the AT/TT (dominant model) genotypes was associated with a reduced risk of SPTB (adjusted OR, 0.61 [0.37–0.99]). No significant association was observed between the risk of SPTB and PLA2G4C rs1366442 polymorphism (P>0.05).

Further stratified analysis was performed to assess whether the association between these SNPs and the risk of SPTB would be modified with specific participant's characteristics, such as maternal age, GDM status and neonatal sex (Table III). All associations of SNPs were independent of maternal age, GDM status and neonatal sex (all P>0.05). There were no significant interactions between PLA2G4D rs4924618 and participant's characteristics on the risk of SPTB (P>0.05).

The SNP rs4924618 introduced a Ser-to-Thr point mutation at position 434 of cPLA2 δ, which encodes a polypeptide of 818 amino acids that display 30% identity in its catalytic domain (amino acids 273 to 818) with three other lipases, cPLA2 α, β and γ. In order to elucidate the possible structural effect of this substitution, homology modeling was performed from the crystal structure of cPLA2 α to locate the mutation site (Fig. 1). A Ramachandran plot of main chain torsion angles indicated that most residues have a reasonable conformation (Fig. 2), demonstrating that the model is of good quality.

Fig. 1A demonstrated the constructed model superposed with the template. As expected, the tertiary structure of cPLA2 δ displayed considerable agreement with that of cPLA2 α, especially in the α/β central core (yellow region in Figs. 1A and B), which consists of a 10-stranded mixed β sheet surrounded by 9 α helices. This result was further verified by the multiple amino acid sequence alignment of all the four cPLA2 isozymes (Fig. 3), where the conserved residues are mainly concentrated within the core (26). The active site necessary for catalysis is conserved in cPLA2 δ at Ser361, following β5 (Fig. 3). Similarly to the active site of cPLA2 α, Ser361 is placed at the bottom of a deep, narrow active site funnel, which penetrates one-third of the way into the catalytic domain (Fig. 1B).

The mutation site Ser434 lies in the C-terminus of α helix F of the conserved central core, which is outside but in close to the bottom of the active site funnel spatially (14.92 Å). Moreover, helix F, together with helices D, E and the interwoven loops, connect the region encompassing the catalytic site (the interconnecting loop between β5 and helix C) with the further four β strands of the core (β6-β9).