Background
Preterm birth is considered a multifactorial condition; however, emerging evidence suggests that genetic variation among individuals may have an important role. Prior studies have suggested that single-nucleotide polymorphisms associated with genes related to the immune system, and particularly the maternal inflammatory response, may be associated with an increased risk of preterm delivery.
Objective
The objective of the study was to identify single-nucleotide polymorphisms associated with spontaneous preterm birth <37 weeks within a cohort of African-American women.
Study Design
This is a secondary analysis of a randomized trial that evaluated periodontal disease and preterm birth. Women were enrolled between 6 and 20 weeks’ gestation at 3 prenatal care clinics between 2004 and 2007. Maternal DNA samples were collected and analyzed using a custom 1536 single-nucleotide polymorphismgenotyping array designed to assess genes involved in inflammation. Women were included in this study if they self-identified as African American. We excluded women with a multiple gestation or an indicated preterm delivery. We performed allele- and genotype-based analyses to evaluate the association between spontaneous preterm birth and tag single-nucleotide polymorphisms. We used a logistic regression to adjust for prior preterm birth in our genotype-based analysis. In a subgroup analysis, we compared women who delivered at <34 weeks’ gestation to women who delivered at term. Within the microarray, we identified ancestry informative markers and compared global ancestry estimates among women who delivered preterm with those who delivered at term.
Results
Of the 833 African-American women in the study with genotype data, 77 women (9.2%) had a spontaneous preterm birth, whereas 756 women delivered at term. In an allele-based analysis, 4 single-nucleotide polymorphisms related to the genes for protein kinase C-α ( PRKCA ) were associated with increased risk of spontaneous preterm birth <37 weeks, whereas a single single-nucleotide polymorphism related to fms-related tyrosine kinase 1 ( FLT1 ) was associated with spontaneous preterm birth <34 weeks. A genotype-based analysis revealed similar associations between single-nucleotide polymorphisms related to the PRKCA genes and spontaneous premature delivery. Additionally, single-nucleotide polymorphisms related to matrix metalloproteinase-2 ( MMP2 ), tissue inhibitor of matrix metalloproteinase-2 ( TIMP2 ), and interleukin 16 ( IL16 ) genes were associated with spontaneous preterm birth <37 weeks in genotype-based analysis. Genetic variants related to MMP2 , matrix metalloproteinase-1 ( MMP1 ), and leukemia inhibitory factor receptor antisense RNA 1 ( LIFR-AS1 ) genes were associated with higher rates of preterm birth <34 weeks. Ancestry estimates were similar between the women who had a spontaneous preterm birth and those who delivered at term.
Conclusion
We identified tag single-nucleotide polymorphisms related to 7 genes that are critical to inflammation, extracellular remodeling, and cell signaling that were associated with spontaneous preterm birth in African-American women. Specifically, we found a strong association with the PRKCA gene. Genetic variation in these regions of the genome may be important in the pathogenesis of preterm birth. Our results should be considered in the design of future genomic studies in prematurity.
Preterm birth is a leading cause of infant morbidity and mortality. The pathophysiology of preterm birth is largely unknown, but genetic predisposition is likely an important component of this multifactorial process. This assumption is supported by twin studies, data demonstrating that family history influences the risk of preterm delivery and high recurrence rates among women with a history of preterm birth. Genetic variation may also account, in part, for the differences in preterm birth rates among individuals of different ancestry groups in the United States.
The role of the immune system and specifically maternal inflammatory response in the pathogenesis of preterm birth has also been supported by strong evidence. Therefore, analyses of genetic variation associated with preterm birth have focused on identifying polymorphism primarily associated with genes involved in the inflammatory pathways. Previously reported polymorphisms associated with preterm birth were related to tumor necrosis factor-alpha ( TNFA ), interleukin-1, interleukin-6, type 1 insulin-like growth factor receptor ( IGF1R ), and tissue inhibitor of metalloproteinase-2 ( TIMP2 ).
To expand upon the existing evidence that specific genetic variants are involved in the pathogenesis of preterm birth, we sought to identify single-nucleotide polymorphisms associated with spontaneous preterm births less than 37 weeks’ gestation using multiple inheritance models within a cohort of African-American women.
Materials and Methods
This was a secondary analysis of data collected by the Periodontal Infections and Prematurity Study, a randomized controlled trial performed to evaluate the effect of periodontal disease and its treatment on preterm birth. Women were enrolled from 2004 to 2007 at 3 prenatal clinics in the Philadelphia, PA, area between 6 and 20 weeks’ gestation. Gestational age was calculated using the participant’s reported last menstrual period and first ultrasound evaluation in the pregnancy. Results of the trial demonstrated that neither periodontal disease nor its treatment was associated with preterm delivery.
A secondary aim of the trial was to evaluate the relationship between genetic variation within candidate genes and preterm birth. An additional observational cohort of women without periodontal disease was enrolled for this purpose. Because periodontal disease was not associated with preterm delivery, women enrolled in both the randomized trial and observational cohort were included in this genetic study. Approval for this study was granted by the Institutional Review Board at the University of Pennsylvania.
After obtaining informed consent, maternal DNA samples using buccal swabs were collected at the time of enrollment. A customized 1536 single-nucleotide polymorphism Illumina Golden Gate genotyping array was utilized in this study. In the development of the customized single-nucleotide polymorphism chip, genes of interest were identified using 2 reference databases that provide integrated information about the genome and biomolecular pathways, Biocarta and Kyoto Encyclopedia of Genes and Genomes.
Genes associated with the inflammatory pathway and those previously reported in the literature to be associated with preterm birth were of interest. We used information from the International HapMap Project to identify specific tag single-nucleotide polymorphisms associated with the genes of interest to include in the customized single-nucleotide polymorphism chip panel. Specific single-nucleotide polymorphisms within each gene were selected based on evidence from the literature that the polymorphisms are functional or the nature of the variation suggests potential function (eg, promotor polymorphisms) if no specific functional data had been published.
Tag, or tagging, single-nucleotide polymorphisms are specific polymorphisms within the human genome that provide information about allelic variation within a larger segment of a chromosome. This is possible because of the coinheritance of single-nucleotide polymorphism alleles within regions found in close proximity within the human genome, a phenomenon referred to as linkage disequilibrium. Utilizing tag single-nucleotide polymorphisms is thus an indirect approach that allows researchers to study genetic variation within a genomic region of interest and associations with disease.
We designed a nested case-control study to examine the relationship between genetic polymorphisms and spontaneous preterm birth prior to 37 weeks’ gestation. A spontaneous preterm birth was defined as birth prior to 37 weeks that resulted from idiopathic preterm labor or preterm premature rupture of membranes. Women with an indicated preterm delivery were excluded, as were women who delivered prior to 20 weeks’ gestation. Only women who self-identified as being African American who had a singleton gestation were included in the analysis.
We restricted our analysis to African-American women for 2 reasons: (1) very few participants in the randomized controlled trial were not African American (15.9%); (2) this restriction would reduce systematic differences in allele frequency because of ancestry (population stratification).
Extensive demographic and clinical data were collected as part of the study, and this information was entered into a research database by trained obstetric research staff. Delivery information including gestational age at the time of delivery was obtained from the participant’s inpatient delivery medical record.
Statistical analysis
Baseline characteristics, including maternal age, parity, history of prior preterm delivery, body mass index, and smoking status, were compared between women who delivered prior to 37 weeks and those who delivered at term. Categorical variables were compared using the χ 2 test. Maternal age and body mass index were not normally distributed, and therefore, median values were compared using the Mann-Whitney U test. This analysis was performed using Stata 10.0 statistical software (Stata Corp, College Station, TX).
All genetic association analyses were conducted using PLINK 1.9 software, a publicly available genetic association analysis program developed by the Center for Human Genetic Research at Massachusetts General Hospital (Boston, MA) and the Broad Institute (Cambridge, MA).
Individuals with a low call rate (>5% missing single-nucleotide polymorphism data) (n = 28), and single-nucleotide polymorphisms with no data reported for >5% of the study cohort were excluded (n = 72). We assessed for deviations from Hardy-Weinberg equilibrium within our population for each single-nucleotide polymorphism using the χ 2 test in Plink 1.9. Whereas deviations from Hardy-Weinberg equilibrium are often tested only in the control group in genetic case-control studies, we performed the test within our entire cohort because control status was established prospectively.
We included all single-nucleotide polymorphisms in our analysis, regardless of whether a single-nucleotide polymorphism was in Hardy-Weinberg equilibrium within our sample. Whereas genotypes that deviate from Hardy-Weinberg equilibrium should be interpreted with caution because they may indicate a genotyping error, population heterogeneity, or selection bias, deviations from the Hardy-Weinberg equilibrium may also be seen when the single-nucleotide polymorphism is associated with the disease of interest.
We used a stepwise approach to explore genetic variation within our cohort. First, an allele association analysis was performed in women with spontaneous preterm birth (<37 weeks) and those women who delivered at term. The minor allele was designated as the risk allele and the frequency of the minor allele was compared in the groups.
We then performed genotype-based analyses considering recessive, dominant, and additive genetic models for each single-nucleotide polymorphism. If a single-nucleotide polymorphism was significantly associated with preterm birth using more than 1 model, then the best model was selected by comparing the strength of the association as measured by the P value in each model. We then performed a logistic regression adjusting for prior preterm birth. If a single-nucleotide polymorphism was not significantly associated with preterm birth in an unadjusted analysis but a significant association was seen after controlling for preterm birth, then this was reported.
We also performed a subgroup analysis comparing women who had an early spontaneous preterm delivery, defined as preterm birth <34 weeks’ gestation, with women who delivered at term. Similar to the primary analysis, we performed allelic and genotype-based analyses.
All analyses used a value of P < 7.64 × 10 –4 as the criterion for statistical significance. This accepted level of error is less conservative than that suggested by the Bonferroni procedure based on the number of individual single-nucleotide polymorphisms in our analysis. This value was calculated based on the number of genes (n = 67) associated with the single-nucleotide polymorphisms in our genotyping panel.
Our reasoning for this is 2-fold. In this study, we have used tag single-nucleotide polymorphisms, and as such, it is not the particular single-nucleotide polymorphism itself but genetic variation in the larger region of the genome linked to the single-nucleotide polymorphism that is of interest. We evaluated multiple tag single-nucleotide polymorphisms that may be in linkage disequilibrium with the same genomic region, therefore focusing on fewer genomic regions than actual single-nucleotide polymorphisms. Second, this study, like many genetic association studies, is exploratory and we sought to avoid missing any potentially important associations.
To further assess the impact of ancestry on risk of preterm delivery, we identified 177 ancestry informative markers that were included in the single-nucleotide polymorphism microarray. These specific polymorphisms were selected using the International HapMap Project based on ≥70% difference in allele frequencies in West African and European ancestral populations. A 2-way model of admixture was generated using ADMIXMAP ( http://www.ucd.ie/geneepi/admixmap/tools.html ). Differences in global ancestry estimates among women who delivered preterm and those who delivered at term were compared using the Welch 2 sample t test.
Results
In all, 870 African-American women included in the trial delivered after 20 weeks’ gestation and had complete (<5% missing) single-nucleotide polymorphism data. We excluded 37 women who had an indicated preterm delivery. Overall, 77 women (9.2%) had a spontaneous delivery prior to 37 weeks’ gestation, whereas 756 women delivered at term. Women who delivered preterm were similar to those women who delivered at term with regard to maternal age, parity, body mass index, and smoking status. However, women who delivered preterm were more likely to have a history of prior preterm birth. Of the women with a spontaneous preterm birth, 55% had preterm premature rupture of membranes ( Table 1 ). The median gestational age of women who had a spontaneous preterm birth was 35.3 weeks (interquartile range, 33.0–36.3) and 39.6 (interquartile range, 38.7–40.3) in women who delivered at term.
| Characteristic | Spontaneous preterm delivery (N = 77) | Term delivery (N = 756) | P value |
|---|---|---|---|
| Maternal age, y a | 22 (19–27) | 23 (20–27) | .45 |
| Nulliparity | 31 (40.3) | 301 (39.9) | .95 |
| History of prior PTB | 17 (22.1) | 75 (9.9) | < .01 |
| Body mass index, kg/m 2 a | 25.8 (22.3–31.4) | 25.9 (22.4–31.3) | .97 |
| Smoking | 21 (27.6) | 267 (35.4) | .18 |
| Birthweight a | 2440 (1940–2723) | 3260 (2973–3560) | < .01 |
| Periodontal disease | |||
| Yes | 59 (76.6) | 568 (75.1) | .77 |
| No | 18 (23.4) | 188 (24.9) | |
| Preterm premature rupture of membranes | |||
| Yes | 42 (54.5) | ||
| No | 35 (45.5) |
a Data are presented as median values (interquartile range); in all other cases, data represent n (percentage).
Allele-based analysis
Minor allele frequency was compared in women who had a spontaneous preterm delivery and women who delivered at term. Four single-nucleotide polymorphisms were identified that were associated with a higher minor allele frequency among women who delivered preterm compared with women who delivered at term ( Table 2 ). All 4 single-nucleotide polymorphisms were associated with the region of the genome-encoding protein kinase C-alpha ( PRKCA ).
| Gene | SNP | Minor allele | Minor allele frequency | Odds ratio | P value | HWE a | |
|---|---|---|---|---|---|---|---|
| Preterm | Term | ||||||
| Preterm birth <37 weeks | |||||||
| PRKCA | rs7225452 | A | 0.58 | 0.42 | 1.92 | 1.2 ∧ 10 –4 | Yes |
| PRKCA | rs4486944 | C | 0.62 | 0.47 | 1.79 | 7.0 ∧ 10 –4 | Yes |
| PRKCA | rs6504424 | A | 0.61 | 0.47 | 1.82 | 5.8 ∧ 10 –4 | Yes |
| PRKCA | rs16960070 | A | 0.12 | 0.05 | 2.45 | 6.3 ∧ 10 –4 | Yes |
| Preterm birth <34 weeks | |||||||
| FLT1 | rs12428494 | A | 0.11 | 0.03 | 3.81 | 3.7 ∧ 10 –4 | Yes |
Genotype-based analysis
In our genotype-based analysis, we considered recessive, dominant, and additive genetic models for each single-nucleotide polymorphism. The best-fitting model for each single-nucleotide polymorphism was determined by comparing the strength of the association between the single-nucleotide polymorphism and spontaneous preterm birth based on the P value in an unadjusted analysis. Six single-nucleotide polymorphisms were significantly associated with spontaneous preterm birth in unadjusted analysis. After performing a logistic regression adjusting for prior preterm birth, an additional single-nucleotide polymorphism (rs4486944) was also found to be associated with a higher risk of spontaneous preterm delivery.
In contrast, another single-nucleotide polymorphism (rs7171517) that was associated with preterm birth in an unadjusted analysis was no longer statistically significant after controlling for a preterm birth history. All 4 single-nucleotide polymorphisms associated with spontaneous preterm birth in the allele-based analysis were significantly associated with preterm delivery in a genotype-based analysis as well ( Table 3 ).
| Gene | SNP | Model | OR | P value | aOR a | P value | HWE b |
|---|---|---|---|---|---|---|---|
| MMP2 | rs11639960 | Recessive | 8.26 | 4.8 ∧ 10 –5 | 9.10 | 2.6 ∧ 10 –5 | No |
| TIMP2 | rs2277698 | Recessive | 8.70 | 4.7 ∧ 10 –4 | 8.44 | 6.8 ∧ 10 –4 | Yes |
| IL16 | rs7171517 | Dominant | 2.31 | 5.8 ∧ 10 –4 | 2.26 | 8.6 ∧ 10 –4 | Yes |
| PRKCA | rs16960070 | Dominant | 2.84 | 3.2 ∧ 10 –4 | 3.02 | 1.7 ∧ 10 –4 | Yes |
| PRKCA | rs7225452 | Additive | 1.92 | 1.7 ∧ 10 –4 | 1.98 | 1.2 ∧ 10 –4 | Yes |
| PRKCA | rs6504424 | Additive | 1.84 | 6.4 ∧ 10 –4 | 1.90 | 4.8 ∧ 10 –4 | Yes |
| PRKCA | rs4486944 | Additive | 1.79 | 8.5 ∧ 10 –4 | 1.86 | 5.4 ∧ 10 –4 | Yes |
a Adjusted for prior preterm birth
Subgroup analysis of spontaneous preterm birth <34 weeks
The minor allele frequency of a single-nucleotide polymorphism (rs12428494) related to the fms-related tyrosine kinase 1 ( FLT1 ) gene was significantly associated with spontaneous preterm birth prior to 34 weeks’ gestation in the subgroup analysis ( Table 2 ). Three additional single-nucleotide polymorphisms were associated with early spontaneous preterm birth using a recessive genotype-based model ( Table 4 ).
