Study population

The study population was selected from 8000 patients for whom samples were available in the University of Hawaii Biospecimen Repository that had been collected between 2005 and 2011. The University of Hawaii Biospecimen Repository is approved by the Western Institutional Review Board and deemed our study proposal exempt from institutional review board review because all the data and samples were deidentified.

There were 94 patients with Filipino ancestry up to and including all 4 grandparents (self-reported), with singleton pregnancies, sPTB between 34 weeks 0 days’ and 36 weeks 6 days’ gestation by documented estimated due date in the absence of medical complications of pregnancy and either PPROM (n = 20) or sPTB (n = 23) according to the diagnosis at admission. For the latter group, 20 patients were selected randomly. Preterm gestational age was selected based on results of RLN SNP analysis in a Danish population. Control subjects (n = 20) were selected randomly from 45 patients who met the same ethnicity and inclusion/exclusion criteria, with normal spontaneous vaginal term delivery between 39 weeks 0 days and 40 weeks 0 days gestation.

Exclusion criteria were delivery because of medical complications during pregnancy (preeclampsia, fetal growth restriction, histologic and/or clinical chorioamnionitis, renal disease, HELLP [hemolysis, elevated liver enzymes, and low platelet count syndrome]), multiple gestations, pregnancies from infertility treatment, fetal anomaly, uterine anomaly, induced preterm delivery or placental conditions (placental abruption, polyhydramnios, positive urine drug screen for cocaine or amphetamines), and pregnancies with an estimated due date that was based on third-trimester ultrasound scanning.

We genotyped maternal blood in all patients for 12 SNPs that were located in the promoter region of the RLN2 and quantitated the RLN protein in the maternal decidual cells (n = 60) and measured the maternal sRLN in a subset of the same patients (n = 21). All samples were blinded, with the disclosure of clinical characteristics/outcomes only after the analyses were completed.

Genotyping and analysis

DNA was extracted from maternal blood with the Autopure system (Qiagen, Valencia, CA); 10 ng of each sample was quality checked and used for genotyping with custom and predesigned TaqMan SNP Genotyping Assay (Life Technologies, Carlsbad, CA). Polymerase chain reactions were performed on an ABI GeneAmp 9700 thermocycler; allelic determination was carried out on an ABI 7900HT fast real-time polymerase chain reaction system (version 2.4; Applied Biosystems, Carlsbad, CA). In-house quality controls (in duplicate) were used as references to identify genotype cluster locations. Genotype data were uploaded into a database (Progeny Software, South Bend, IN) that contained deidentified demographic and clinical information for statistical analysis.

SNPs were chosen to cover the promoter of the RLN2 from circa 2.2-6.6 kb before the 5 prime untranslated region, avoiding the region with highest homology to RLN1 . Selection was biased toward either the functionality or the conservation of the particular SNP. A list of these SNPs, with their SNP database identification (rs) numbers and chromosome 9 base location, is shown in Table 1 . Allele and genotype frequencies were calculated for each SNP. Of the 12 that were genotyped, 7 matched the reference allele and were therefore excluded from further analysis. The remaining 5 SNPs met Hardy-Weinberg equilibrium at the 5% alpha level.

Association studies were performed with the use of the Fisher exact test (permutation test) and logistic regression with the control subjects as the reference group and neonatal sex as the covariate. The data were tested for linkage disequilibrium by calculating R ² . Three SNPs (rs13293410, rs3758239, rs7029400) were in high linkage disequilibrium (R >0.9). Because this indicates redundancy, 2 SNPs were excluded, and only rs3758239 along with rs4742076 and rs10116567 remained for individual SNP association analyses.

Additive, codominant, dominant, and recessive inheritance models were considered. The outcome group-genotype interaction was considered significant at a probability value of < .05. The inheritance model with the smallest probability value was considered to be the best-fitting model for the respective SNP and was used to calculate odds ratios. Because of small sample size, interaction between RLN genotype and dRLN expression was evaluated by bootstrap analysis, with tissue RLN-genotype interaction being considered significant at a probability value of < .05. Because this was an exploratory study, no adjustment to the alpha level was made for multiple comparisons, and all comparisons are reported. SAS software (SAS Institute, Cary, NC) and R ( www.r-project.org ) were used for the statistical analysis.

Quantitative immunocytochemistry

Slides of formalin-fixed fetal membrane rolls (n = 60) were treated in sodium citrate buffer (10 mmol/L; pH 6.0) for 30 minutes for antigen retrieval, followed by 0.3% hydrogen peroxide then 2.5% horse serum. They were incubated for 1 hour with rabbit polyclonal antibody (immunoglobulin G) to human RLN (8 μg/mL; Calbiochem/EMD Biosciences, San Diego, CA). Negative controls were nonimmune immunoglobulin (DAKO North American Inc, Carpenteria, CA). Color was developed with the ImmPRESS reagent (Vector Laboratories, Burlingame, CA) for 30 minutes and DAB (3, 3-diaminobenzidine; Sigma-Aldrich, St. Louis, MO) for 5 minutes.

A series of brightfield images were acquired between 420 and 700 nm at 20-nm intervals with the use of a multispectral imaging system comprised of an Olympus BX51 microscope (Olympus America Inc, Melville, NY) and a CRI Nuance spectral analyzer (Caliper Life Sciences, Hopkinton, MA). The images were stacked to create a 3-dimensional image cube. Quantitative image analysis of the unmixed data was performed with Inform software (version 1.4.0; PerkinElmer, Waltham, MA). The average signal intensity per pixel from 5 different fields of decidua from each patient was collected; results were expressed as mean ± SD. Immunocytochemical data were analyzed by nonparametric analysis (GraphPad Software Inc, San Diego, CA); the Kruskal-Wallis analysis of variance used to compare differences in protein expression.

sRLN was measured in a subset of patients (n = 21) by enzyme-linked immunosorbent assay (R&D Systems, Minneapolis, MN) according to the manufacturer’s instructions; results were analyzed as described earlier.