Genetic contributions to preterm birth: Implications from epidemiological and genetic association studies

Abstract

Infants born before term (<37 weeks) have an increased risk of neonatal mortality as well as other health problems. The increasing rate of preterm birth in recent decades, despite improvements in health care, creates an impetus to better understand and prevent this disorder. Preterm birth likely depends on a number of interacting factors, including genetic, epigenetic, and environmental risk factors. Genetic studies may identify markers, which more accurately predict preterm birth than currently known risk factors, or novel proteins and/or pathways involved in the disorder. This review summarizes epidemiological and genetic studies to date, emphasizing the complexity of genetic influences on birth timing. While several candidate genes have been reportedly associated with the disorder, inconsistency across studies has been problematic. More systematic and unbiased genetic approaches are needed for future studies to examine the genetic etiology of human birth timing thoroughly.

• Birth timing
• genetics
• preterm birth
• risk factors

Preterm birth is a major public health concern. In the United States, 12.7% of births occur before term (<37 weeks) 1. Infants born before term have an increased risk of neonatal mortality as well as serious health problems, such as respiratory illness, blindness, and cerebral palsy 2. Moreover, the severity and incidence of these problems worsen with decreasing gestational age 3. The impact of this disorder grows with the increasing rate of preterm birth in recent decades 2.

Although 20%–30% preterm births are medically indicated to avoid or minimize maternal and/or fetal complications, most preterm births result from spontaneous preterm labor or preterm premature rupture of membranes (PPROM) 3. While a number of risk factors have been identified, accurate prediction and prevention are difficult 3. For example, biomarkers of preterm birth, such as serum protein concentrations of interleukin (IL)‐6, IL‐8, tumor necrosis factor (TNF)‐α or relaxin, while strong predictors of preterm birth, may not be useful in large low‐risk populations 4. One problem may be that such markers vary over time or among individuals, making it difficult to determine the levels at which risk for preterm birth is increased. In contrast, genetic factors are stable over time and therefore may be better predictors of risk. As a result, genetic studies may identify markers which more accurately predict preterm birth than currently known risk factors. Genetic studies may also identify novel proteins and/or pathways involved in the disorder. This new information will augment our general understanding of parturition and provide new targets for drug therapies.

Key messages

• A wealth of evidence suggests that genetics are important in birth timing.

• A number of case‐control studies have identified potential candidate genes associated with preterm birth and birth timing; however, several statistical genetic concerns should be considered to improve the robustness of these findings in the future.

• Systematic genome‐wide scans and other unbiased approaches may help identify novel genes involved in birth timing and preterm birth.

Evidence for influence of genetics on birth timing

Birth timing across pregnancies in the same woman

A wealth of evidence suggests that genetics are important in birth timing. For example, both preterm and postterm births tend to recur in mothers 5–10. Moreover, the most likely age for a recurrent preterm birth is same week as the first preterm birth 7, 11, 12, suggesting that factors that are stable over time, such as genetics, affect birth timing.

Birth timing trends among family members

Familial trends for birth timing also suggest that genetics influence this trait. Women who are born preterm are more likely to have had a preterm delivery themselves 13, indicating that mothers and their daughters share risk. Sisters of women who have had a preterm delivery also have an increased risk for preterm delivery 10. Due to the nature of family studies, environmental factors shared between mothers and daughters or between sisters cannot be untangled from genetic influences. As a result, it is difficult to determine the relative importance of genetic versus environmental factors from these studies alone.

Partitioning variance in birth timing into genetic versus environmental components

In contrast to family studies, twin studies measure the relative importance of genes in overall trait variance within a population. By comparing concordance rates between monozygotic and dizygotic twins, which share 100% and approximately 50% of their genes, respectively, one can model the genetic and environmental factors that influence a trait. Clausson and colleagues 14 estimated that genetics account for 36% and 31% of variation in preterm birth (<37 weeks) and gestational age as continuous trait, respectively. Another study estimated that genetics account for 27% of variation in preterm birth in any pregnancy and 17% in the first pregnancy 15. Based on these estimates, genetics play a significant role in preterm birth, in addition to environmental factors. However, the exact percentage of variation in birth timing due to genetics is not known, as both studies had small sample sizes from which estimates were made. Moreover, Treloar and colleagues 15 used ‘more than 2 weeks early’ as their phenotype, rather than the standard definition of preterm birth (<37 weeks). It is unclear how well this measure correlates with standard definitions and to what extent one can compare the results of Treloar et al. 15 to other studies.

A similar method was used to estimate the influence of maternal and fetal genetic factors by Lunde and colleagues 16. Comparing concordance rates among full and half siblings for gestational age, the authors estimated that 11% of variation for this trait is due to fetal genetic factors and 14% of variation is due to maternal genetic factors 16. Such comparisons use the degree of genetic relatedness (on average 50% for full siblings and 25% for half siblings) and trait concordance to estimate the relative importance of genetic versus environmental factors. Because siblings that are not monozygotic twins display some variability in their percent genetic identity and may differ in important dominance or interactive genetic effects, these estimates are more difficult to make using nontwin siblings. As a result, the estimates by Lunde et al. 16, in addition to those made by Clausson et al. 14 or Treloar et al. 15, probably do not accurately measure the amount of variance in preterm birth due to genetics. Despite the limitations in estimating the heritability, all three of these studies indicate that genetics play an important role in preterm birth.

Another approach to separating genetic and environmental factors is the coefficient of kinship. This measure depicts the degree of genetic relatedness within a population. Ward and colleagues 17 used this measure to examine genetic influences on preterm birth in a Utah population. The Utah population from which the families were drawn was established by 10,000 people who moved to the state to establish the Mormon religion 17. Because Mormons are discouraged from using alcohol or tobacco and have low rates of substance abuse and sexually transmitted diseases, this population may represent individuals with relatively few environmental risk factors for preterm birth 17. As a result, detecting genetic effects may be easier in this cohort of Utah preterm families. In this study, Ward and colleagues found that families with preterm deliveries had a significantly lower coefficient of kinship than controls 17, indicating that these families are more closely genetically related than control families. This evidence suggests that the increased rate of preterm birth in these families can be explained by genetic factors. It is important to note that the authors of this study did not report the relative abundance of any environmental risk factors for preterm birth in the two populations. It is possible that one or more important environmental risk factors differ between these groups, in addition to genetic relatedness. Hence, the results of Ward et al. 17 support the significance of genetics in preterm birth, but do not address their relative importance compared to known environmental risk factors.

Mendelian disorders

Certain Mendelian disorders are associated with preterm birth, further supporting genetic effects on birth timing. Ehlers‐Danlos syndrome [EDS] represents a diverse group of Mendelian disorders affecting connective tissue, primarily inherited in an autosomal dominant manner 18. Women with vascular EDS have an increased risk of delivering preterm, primarily due to PPROM 18. Since this disorder is inherited in an autosomal dominant pattern, there is a 50% chance the fetus has inherited the disorder, making it less clear whether the mother or infant genome contributes to the increase in PPROM risk.

Racial disparities in preterm birth

Ethnic differences in preterm birth rates also may be suggestive of genetic effects. Pregnancies in which either the mother 5, 8 or father 19 is black are at increased risk for preterm delivery. According to a study by Goldenberg and colleagues 20, these racial disparities are not explained entirely by measured environmental risk factors, implying that genetic differences among races may also contribute.

While the human population as a whole is genetically very similar, differing by 0.1% on average, important genetic differences exist across populations. Studies have found that approximately 10%–15% of total genetic variation in humans can be explained by differences between sub‐Saharan Africans, Northern Europeans, and East Asians 21. Moreover, specific polymorphisms can be used to group individuals based on ethnic groups with high accuracy 22. These results suggest that traditional racial designations parallel genetic differences to a certain extent, supporting the usefulness of race as an index of genetic composition.

Some investigators assert that only environmental and other nongenetic risk factors may account for the observed racial disparity in preterm birth rates 23. Indirect evidence for this hypothesis includes the observation that foreign‐born blacks and US‐born blacks differ in rates of preterm birth and low birth weight (often used as a proxy for gestational age). However, complex gene‐gene and gene‐environment interactions are not considered in this interpretation. For example, admixture in blacks, such that their genomes are partially African and partially European in ancestry, does not necessarily predict that Africans will have higher rates of preterm birth due to higher frequencies of predisposing alleles of African origin. Admixed blacks may have higher rates of preterm birth due to gene‐gene interactions that are only present when African‐ and European‐derived variation are combined into a single genome, or because of novel environmental risk factors that uncover important gene‐environment interactions not present in either US‐born whites or foreign‐born blacks. Other more subtle genetic effects, such as genetic variation affecting gene expression patterns, known to differ among races 24, may also explain observed differences in preterm birth rates. Hence, while environmental risk factors do significantly contribute to racial disparities, genetic effects on racial disparities in birth timing should not be discounted.

Additionally, studies comparing foreign‐born and US‐born blacks have rarely considered the race or birthplace of the father. Previous studies observed that preterm birth rates are higher whether the mother or father of a infant born preterm is black 25, suggesting that fetal race also influences birth timing. As a result, previous studies of black mothers may underestimate the role of genetics in such observed racial differences.

Modeling of genetic effects on birth timing

The complexity of genetic effects on birth timing cannot be sufficiently underscored. Genetic studies of preterm birth should not be thought of as a nature‐versus‐nurture debate, but rather an effort to characterize to how nature and nurture interact together to influence birth timing. There is increasing evidence that preterm birth can be conceptualized as a common, complex disorder. In contrast to Mendelian disorders, in which alterations of a single gene can lead to disease, complex diseases are influenced by a variety of factors, none necessary and sufficient to cause the disorder by itself. As a result, there is not a direct relationship between genotypes and phenotypes 26. These disorders likely depend on a number of interacting factors, including genetic, epigenetic, and environmental risk factors 26. Moreover, because environmental and epigenetic factors are also important, specific genes may be more influential in increasing risk in the context of particular environmental factors. For example, individuals with a particular genotype who also experienced high stress levels may be at higher risk for preterm birth than less stressed people of the same genotype. This has been demonstrated in relation to the development of major depression, whereby individuals carrying a particular version of the serotonin transporter gene showed an elevated risk of depression only in the context of stressful life events 27. Additionally, certain genes may be more important in increasing risk when found in combination with polymorphisms in other genes. For example, in testing several candidate genes for hypertension, Williams and colleagues 28 found no significant single site associations; however, a two‐locus model including polymorphisms in angiotensin I‐converting enzyme (ACE) and G protein‐coupled receptor kinase 4 (GRK4) was associated with the blood pressure phenotype studied. Hence, interactions between genes and/or the environment likely add complexity to genetic influences on birth timing.

A variety of evidence supports the complexity of preterm birth. Modeling procedures used by twin studies suggest that additive genetic factors and environmental risk factors that are not shared among siblings both influence preterm birth 14, 15. Additionally, interactions between genes 29, 30 have been associated with preterm birth risk. Several studies suggest that gene‐environment interactions, such as interactions between inflammatory gene risk alleles and bacterial infections 31–33, also influence the disorder. Together, these studies imply that the etiology of preterm birth likely involves genetic as well as environmental factors in complex interactions.

In addition to the complexity of genetic effects described above, several issues complicate how investigators think about the disorder. First, it is not clear whether the mother or the infant from a preterm delivery should be considered the proband. As a result, it is not clear which individual's DNA should be interrogated. Additionally, preterm delivery as a trait can be thought of in two ways. First, it can be thought of as discrete, resulting from genetic factors that lead to either disease or nondisease outcomes. Alternatively, gestational age can be thought of as a quantitative trait, with preterm ages as extremely low values. Consequently, genetic effects may be quantitative trait loci (QTLs) that influence the value of gestational age in both term and preterm deliveries. As disease models shape how one approaches identifying genes, it is important to consider the uncertainty about how to conceptualize preterm birth when evaluating various approaches taken to study this trait.

Specific human genes tested for association with preterm birth

The majority of human genetic studies of birth timing to date have been case‐control comparisons. In such studies, polymorphism frequencies are compared in affected and unaffected individuals. If the polymorphism significantly differs in frequency across these groups, the polymorphism is said to influence risk of preterm birth. One advantage of using this study design is that it is relatively easy to recruit individual mothers and/or infants from a preterm delivery as cases, as opposed to the more extensive family recruitment needed for family‐based association or linkage tests. However, differences between case and control groups other than affection status, such as ethnic background, can create spurious associations. For example, case and control groups may come from two subpopulations that differ in allele frequencies for a particular locus. If the two groups do not have similar representations of the populations, an allele may be appear to be associated in the mixed sample, but does not actually contribute to disease. To illustrate, if allele 1 of a biallelic polymorphism was more common in subpopulation 1 and this subpopulation was overrepresented in the affected group, an association could be detected between allele 1 and the disease regardless of whether the gene was actually involved in the disorder (Figure 1).

Figure 1 Illustration of population stratification in case‐control studies. Case and control groups come from two subpopulations, A and B, which differ in allele frequency for the polymorphism tested, 50% versus 90% for allele 1. If the case and control groups differ in the number of individuals coming from subpopulation A (40% versus 60%), this will create a difference in observed allele 1 frequencies between the groups (74% versus 66%) simply due to population substructure, rather than any relationship between allele 1 and disease status.

Candidate genes in a variety of pathways believed to be important in parturition have been assessed, such as inflammation and infection, stress response, functional loss of progesterone, remodeling of connective tissues, uterine contractility, and placental function (please see supplementary online materials for details).

Table

Summary of case/control studies
Gene name	Ref	Population	n	Preterm definition	Exclusion criteria	PPROM included?	Proband	Polymorphism(s) tested	Association?
Infection/inflammation pathway
TNF	32	Pennsylvania, white and black	125 cases, 250 controls	<37 wks	induced labor	yes	mother	‘−308’	Y
TNF	42	UK, mixed (white, black, other)	48 cases, 82 controls	<37 wks	multiple gestation, medical complications, induction of labor, hypertension, or diabetes while pregnant	yes	mother	‘−308’	Y
TNF	43	Seattle, mixed	50 cases, 68 controls	⩽32 wks	cervical dilation, ruptured membranes, amniotic fluid infection	no	both	‘−308, −863’	Y
TNF	48	Kenya, African	1048 total	<37 wks	multiple gestation, no siblings	unknown	child	‘−308’	Y
TNF	49	Chinese	54 case trios, 79 control trios	<37 wks?	unknown	unknown	child	‘−308’	Y
TNF	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘+488, −238, −308’	Y
TNF	47	Pennsylvania, black	55 cases, 110 controls	<37 wks	unknown	yes, analyzed separately	mother	‘−308’	PPROM only
TNF	30	Nashville, white and black	black: 50 cases, 223 controls; white: 77 cases, 184 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	mother	LD SNPs	G×G
TNF	31	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘−488’ i.e. ‘−308’	N
TNF	44	Utah, white?	203 cases mothers+ 44 infants, 41 control women	<37 wks	induced labor	yes	mother	‘−308’	N
TNF	45	Nashville, white and black	black: mothers 49 cases, 225 controls, infants 46 cases, 224 controls; white: mothers 77 cases, 184 controls, infants 86 cases, 191 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	both	5 LD SNPs+ ‘−308’	N
TNF	46	Nashville, white and black	black: 50 cases, 223 controls; white: 77 cases, 184 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	mother	‘−308’	N
TNF	46	meta‐analysis of 7 studies, black, white, Hispanic	643 cases, 1208 controls	<37 wks (mixed definitions among studies)	multiple studies	yes	mother	‘−308’	N
TNFR1	30	Nashville, white and black	black: 50 cases, 223 controls, white: 77 cases, 184 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	mother	LD SNPs	Y
TNFR1	45	Nashville, white and black	black: mothers 49 cases, 225 controls, infants 46 cases, 224 controls; white: mothers 77 cases, 184 controls, infants 86 cases, 191 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	both	6 LD SNPs	Y
TNFR2	30	Nashville, white and black	black: 50 cases, 223 controls, white: 77 cases, 184 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	mother	LD SNPs	Y
TNFR2	45	Nashville, white and black	black: mothers 49 cases, 225 controls, infants 46 cases, 224 controls; white: mothers 77 cases, 184 controls, infants 86 cases, 191 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	both	7 LD SNPs	Y
IL1A	31	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘+4845’, ‘IVS‐109’	N
IL1A	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘−889’	N
IL1A	51	Boston, white, black, Hispanic	304 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IL1B	31	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘+3954’, ‘−581’, ‘−1061’	N
IL1B	42	UK, mixed: white, black, ‘other’	48 cases, 82 controls	<37 wks	multiple gestation, medical complications, induction of labor, hypertension or diabetes while pregnant	yes	mother	3953	N
IL1B	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘3962, −511’	N
IL1R1	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘+970’	N
IL1R2	51	Boston, white, black, Hispanic	300 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	Y
IL1RN	56	North Carolina, white and black	95 cases, 105 controls	<37 wks	multiple gestation, medical complications, induction of labor	yes	mother	VNTR ‘intron2’	Y
IL1RN	57	Switzerland, white	25 cases, 166 controls	<37 wks	none indicated	yes, analyzed separately	mother	VNTR ‘intron2’	Y
IL1RN	58	New York, white, black, Hispanic	30 cases, 192 controls	<37 wks	<15 years old, antibiotics within 4 weeks of collection, multiple gestation, cervical cerclage, previous late‐gestation pregnancy loss, placenta previa, isoimmunization, other medical complications	unknown	mother	VNTR ‘intron2’	Y
IL1RN	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘+11100’	N
IL1RN	51	Boston, white, black, Hispanic	305 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IL2	31	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘−385’	N
IL4	31	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘−590’	Y
IL4	53	New York, white, black, Hispanic	73 mothers, 148 infants (either twins or triplets)	<37 wks	none indicated	yes	both	‘−590’	Y
IL4	51	Boston, white, black, Hispanic	306 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IL6	54	German, white	infants: 909 case, 491 control; Mothers: 466 cases, 281 control	<37 wks	none indicated	yes	both	‘174G promoter’	Y
IL6	55	Pennsylvania, white and black	black: 12 cases, 46 controls; white: 39 cases, 110 controls	<34 wks	cervical cerclage, vaginal bleeding, preeclampsia, diabetes mellitus, collagen vascular disease	no	mother	‘−174’	Y
IL6	31	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘−237’	G×E only
IL6	33	Nashville, white and black	black: 50 cases, 223 controls; white: 77 cases, 184 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	mother	5 LD SNPs	G×E only
IL6	30	Nashville, white and black	black: 50 cases, 223 controls; white: 77 cases, 184 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	mother	LD SNPs	N
IL6	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘−174’	N
IL6R	30	Nashville, white and black	black: 50 cases, 223 controls; white: 77 cases, 184 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	mother	LD SNPs	G×G only
IL6R	33	Nashville, white and black	black: 50 cases, 223 controls; white: 77 cases, 184 controls	<36 wks	mixed race, excluded term births treated for preterm labor or intraamniotic infection from controls, excluded cases with PPROM, multiple gestations, preeclampsia, fetal abnormalities, medical complications such as gestational diabetes, surgeries during pregnancies	no	mother	6 LD SNPs	Y
IL8	51	Boston, white, black, Hispanic	307 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IL8	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘−251 A/T’	N
IL10	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘−1082, −819, −592’	Y
IL10	51	Boston, white, black, Hispanic	308 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IL10	53	New York, white, black, Hispanic	73 mothers, 148 infants (either twins or triplets)	<37 wks	none indicated	yes	both	‘−1082’	N
IL10RA	51	Boston, white, black, Hispanic	309 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IL11	51	Boston, white, black, Hispanic	310 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IL12A	51	Boston, white, black, Hispanic	311 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IL18	51	Boston, white, black, Hispanic	312 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
LTA	31	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘IVS1+90’, ‘IVS1−82’	Y
LTA	51	Boston, white, black, Hispanic	313 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
LTA	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	T26N	N
CSF3	51	Boston, white, black, Hispanic	302 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
IFNGR1	51	Boston, white, black, Hispanic	303 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
TGFB1	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘−800, −509’	N
MBL2	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘−550, 221, Arg52Cys, Gly54Asp, Gly57Glu’	Y
MBL2	59	Austria, white	102 cases, 102 controls	<36 wks	none indicated	unknown	child	‘−550, 221, Arg52Cys, Gly54ASp, Gly57Glu’	Y
TNFRSF6	60	New York, white, black, Hispanic	119 mothers, 238 infants (either twins or triplets)	<37 wks	none indicated	yes, analyzed separately	both	‘−670’	PPROM only
TNFRSF6	50	Australian, white	202 cases, 185 controls	<35 wks	autoimmune disease, preeclampsia, in vitro fertilization, uterine malformation, placental abruption, nonspontaneous labor, ultrasound‐confirmed fetal abnormality, insulin‐dependent diabetes mellitus, multiple pregnancy	yes	mother	‘−1377, −670’	N
CD14	29	NY, white, black, Hispanic	110 mother‐infant pairs	<37 wks	intrauterine demise, <24 weeks, uterine abnormalities, cervical incompetence, history of endocrine or immunosuppressive disorders, HIV‐positive	yes, analyzed separately	both	‘−159’	PROM only, G×G
CD14	54	German, white	909 preterm infants, 491 control infants; 466 preterm mothers, 281 control mothers	<37 wks	none indicated	yes	both	‘159T’	N
TLR2	61	Dutch, white	200 preterm infants, 105 term infants	<37 wks	multiple pregnancies, preeclampsia	yes	child	‘T −16934A, Arg753Gln’	Y
TLR2	71	German, white	910 preterm infants, 491 control infants; 466 preterm mothers, 281 control mothers	<37 wks	none indicated	yes	both	‘Arg753Gln’	N
TRL4	62	Finland, white	440 preterm infants, 351 control infants; 74 preterm mothers, 20 control mothers	<35 wks	none indicated	yes	both	‘Asp299Gly, Thr399Ile’	Y
TLR4	61	Dutch, white	201 preterm infants, 105 term infants	<37 wks	multiple pregnancies, preeclampsia	yes	child	‘Thr399Ile’	N
TRL4	63	Philadelphia and Detroit, black	131 PPROM cases, 246 controls	<37 wks	multiple gestation, induced labor, malformations	yes	child	‘Ala896Gly’	N
TRL4	71	German, white	911 preterm infants, 491 control infants; 466 preterm mothers, 281 control mothers	<37 wks	none indicated	yes	both	‘896G’	N
ALOX5AP	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	rs1769874, ‘SQ13S32’, ‘SQ13S25’	N
NOD2	63	Philadelphia and Detroit, black	131 PPROM cases, 246 controls	<37 wks	multiple gestation, induced labor, malformations	yes	child	‘2936insC’	PPROM only
NOD2	71	German, white	912 preterm infants, 491 control infants; 466 preterm mothers, 281 control mothers	<37 wks	none indicated	yes	both	‘3020insC’	N
Remodeling of connective tissues
MMP1	64	Philadelphia and Detroit, black	179 PPROM cases, 235 controls	<37 wks	multiple gestation, induced labor, malformations	yes	child	‘−1607’	PROM only
MMP9	65	Philadelphia and Detroit, black	74 PPROM cases, 215 controls	<37 wks	multiple gestation, fetal abnormalities, medical complications requiring labor induction	yes	mother	‘−1562’, CA repeat (−208 to−185)	PROM only
PLAT	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘−7351 C/T’	N
SERPINH1	66	Pennsylvania, black	244 PROM cases, 358 controls	<37 wks	multiple gestations, fetal abnormalities, trauma, connective tissue diseases, medical complications of pregnancy	yes, only PROM	mother	‘−656’	Y
SERPINH1	66	Pennsylvania, black	92 PPROM cases, 184 controls	<37 wks	multiple gestations, fetal abnormalities, trauma, connective tissue diseases, medical complications of pregnancy	yes, only PROM	mother	‘−656’	Y
SLC23A1	67	North Carolina, white and black	black: 129 cases, 237 controls; white: 142 cases, 335 controls	<37 wks	none indicated	yes	mother	haplotype‐tagging SNPs	N
SLC23A2	67	North Carolina, white and black	black: 129 cases, 237 controls; white: 142 cases, 335 controls	<37 wks	none indicated	yes	mother	haplotype‐tagging SNPs	Y
Premature uterine contraction
ADRB2	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘Gln27Glu’	Y
ADRB2	68	unknown	159 total mother‐infant pairs	unknown	unknown	unknown	mother, child	‘Gly16Arg, Gln27Glu’	Y
ADRB2	69	New York, Hispanic	28 preterm mothers, 251 term mothers	<37 wks	multiple gestations, chorioamnionitis, uterine malformation, abnormal placental implantation, fetal abnormalities, history or evidence of drug abuse	unknown	mother	‘Gly16Arg, Gln27Glu’	Y
DRD2	51	Boston, white, black, Hispanic	317 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
PDE4D	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	rs12188950	N
PTGER2	51	Boston, white, black, Hispanic	319 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
PTGES	51	Boston, white, black, Hispanic	320 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
PTGFR	51	Boston, white, black, Hispanic	321 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
Placental function
F2	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘G20210A’	N
F2	71	German, white	Mothers: 507 cases, 278 controls; Infants: 593 VLBW, 278 controls	<37 wks, <1500 g	none indicated	yes	both	‘G20210A’	N
F5	51	Boston, white, black, Hispanic	300 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	Y
F5	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘Arg506Gln’	N
F5	71	German, white	507 preterm mothers, 278 term mothers; 593 VLBW infants, 278 term	<37 wks, <1500 g	none indicated	yes	both	‘Arg506Gln’	N
F7	71	German, white	mothers: 507 cases, 278 controls; infants: 593 VLBW, 278 controls	<37 wks, <1500 g	none indicated	yes	both	‘121indel’	Y
F7	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘−323 INDEL, Arg353Gln’	N
F13A1	71	German, white	mothers: 507 cases, 278 controls; infants: 593 VLBW, 278 controls	<37 wks, <1500 g	none indicated	yes	both	‘Val43Leu’	Y
NOS2A	51	Boston, white, black, Hispanic	300 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	Y
NOS2A	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘−231 c/t’	Y
NOS3	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘−922 a/g’, E298D, 690 CT	Y
NOS3	51	Boston, white, black, Hispanic	315 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
FGB	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘Gln27Glu, Arg448Lys’	N
TFPI	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘−33 T/C’	N
PROC	51	Boston, white, black, Hispanic	314 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
ANXA5	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘1 C/T’	N
ADD1	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘Gly460Trp’	N
EPCR/PROCR	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘4600 A/G’	N
SERPINE1	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘675 4g/5g’, 11053 g/t	N
SERPINB2	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘Asn120Asp, Asn40Lys, Ser423Cys’	N
THBD	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘Gly127Ala’	Y
VEGF	70	Greece, white	54 cases, 79 menopausal women as controls	<37 wks	no induction, had to have evidence of regular contractions before 37 weeks	unknown	mother	‘−634’ G/C, 936 C/T	Y
Response to environmental toxins
ADH1C	51	Boston, white, black, Hispanic	322 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
PON1	73	China, Han Chinese	92 case trios, 114 control trios	28–37 wks	birth defects	unknown	child	‘Gln192Arg’	Y
PON1	72	UK, white	3266 total, population cohort	<37 wks?	none indicated	unknown	mother	‘Gln192Arg’	Y
PON2	73	China, Han Chinese	92 case trios, 114 control trios	28–37 wks	birth defects	unknown	child	‘Ala148Gly, Ser311Cys’	Y
OPRM1	51	Boston, white, black, Hispanic	301 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	Y
GSTT1	74	Pennsylvania, white and black	51 case, 904 control mother/infant pairs	<37 wks and <2500 g	>2500 g and preterm, term and LBW/SGA	unknown	both	‘null’	Y
CYP1A1	74	Pennsylvania, white and black	52 case, 904 control mother/infant pairs	<37 wks and <2500 g	>2500 g and preterm, term and LBW/SGA	unknown	both	‘*1−*3’	N
CYP2E1	74	Pennsylvania, white and black	53 case, 904 control mother/infant pairs	<37 wks and <2500 g	>2500 g and preterm, term and LBW/SGA	unknown	both	‘*1, *5B’	N
GSTM1	74	Pennsylvania, white and black	54 case, 904 control mother/infant pairs	<37 wks and <2500 g	>2500 g and preterm, term and LBW/SGA	unknown	both	‘null’	N
NAT2	74	Pennsylvania, white and black	55 case, 904 control mother/infant pairs	<37 wks and <2500 g	>2500 g and preterm, term and LBW/SGA	unknown	both	‘fast/slow’	N
Fetal stress pathway
CRHBP	51	Boston, white, black, Hispanic	316 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
Functional loss of progesterone
PGR	51	Boston, white, black, Hispanic	318 cases, 458 controls	<37 wks?	none indicated	unknown	mother	LD SNPs	N
Other
CBS	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘Leu278Thr’	N
MTHFR	52	Australian, white	443 infants with cerebral palsy, 883 infants without cerebral palsy	<37 wks	none indicated	unknown	child	‘677 c/t’	N
MTHFR	80	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘677, 1298’	N
MTHFR	81	Danish, white	unknown	<37 wks?	unknown	?	diabetic mother	‘−677’ c/t	N
MTR	80	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘2756’	N
MTRR	80	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘66’	Y
SHMT1	80	North Carolina, white and black	black: 67 cases, 238 controls; white: 69 cases, 336 controls	<37 wks	medically indicated, induction	yes	mother	‘1420’	Y, G×E
PPARG	82	Northern Ireland, white	two cohorts of 382, 620 singleton births	<37 wks?	none indicated	unknown	child	‘Pro12Ala’	Y

PPROM‐preterm premature rupture of membranes, LD‐linkage disequilibrium, SNP‐single nucleotide polymorphism, VNTP‐variable number tandem repeat, VLBW‐very low birth weight, LBW‐low birth weight, SGA‐small for gestational age

Inflammation and infection

Many of the genetic studies of preterm birth to date have focused on genes involved in immunity and inflammation. Inflammation and infection are commonly associated with preterm birth. Women who deliver as a result of preterm labor have high rates of positive bacterial culture from amniotic fluid 34. Moreover, increases in inflammation factors may represent an early event in the transition to active labor in both term and preterm pregnancies 35. Less fulminant infections, such as bacterial vaginosis 36 and periodontal disease 37, all increase the risk for preterm birth. Whether these infections are due to microbes or variations in host response to identified microbes remains uncertain as prophylactic antibiotic treatment fails to alter the risk for preterm delivery 38–40. The parturition process at both term and preterm is associated with induction of many proinflammatory mediators, suggesting induction of these molecules as central components of the parturition cascade in humans 41.

Tumor necrosis factor‐α and receptors

The most frequently studied gene has been tumor necrosis factor‐α (TNF) a pro‐inflammatory cytokine. Many studies have tested the −308 polymorphism in the promoter of this gene, which affects expression levels, with mixed results. Several studies have found association between this polymorphism in mothers 32, 42, while others have not 30, 43–45. A meta‐analysis of seven studies of this polymorphism in mothers 46 found no association for this polymorphism, providing additional negative evidence for this gene. One study found that only mothers with PPROM showed association with −308, suggesting that heterogeneity in preterm birth phenotype may contribute to the varied results 47. Studies examining infants as affected also show mixed results 43, 48, 49.

Other polymorphisms in TNF have been tested for effects on preterm birth. While one study found association with another promoter polymorphism at −863 43, another study examining multiple single nucleotide polymorphisms (SNPs) across the gene region found no association 45. One possible explanation for the lack of replication across studies may be that haplotypes of multiple polymorphisms may be more informative than individual polymorphisms. For example, Annells and colleagues 50 noted that the ‘−308, −238, +488’ AGG haplotype was associated with preterm birth less than 29 weeks, but none of the individual polymorphisms were significantly associated. Overall, it is not clear whether or not this gene plays a significant role in preterm birth without further study.

In addition to TNF, investigators have tested for its receptors for genetic effects on preterm birth. Menon and colleagues 45 found significant differences in allele frequencies between blacks and whites, and some suggestive association within blacks with both TNF receptors 1 (TNFR1) and 2 (TNFR2). A later study with the same population found significant association with a promoter polymorphism in TNFR1 and an intronic polymorphism in TNFR230, supporting the previous study's results. While these studies provide suggestive evidence for a role of TNF receptors in preterm birth, replication in other study populations would further solidify these findings.

Interleukins and their receptors

Interleukins form another class of inflammatory genes that have been studied extensively. To date, interleukins 1α 31, 50, 51, 1β 31, 42, 50, 2 31, 8 51, 52, 11 51, 12A 51, and 18 51 have shown no association with preterm birth. In contrast, interleukins 4, 6, and 10 have shown mixed results. Kalish and colleagues 53 found that homozygotes for a promoter polymorphism in interleukin 4 increased risk approximately 11‐fold over the corresponding low‐risk genotype. However, association with the opposite homozygous genotype 50 and no association with the gene 51 and preterm birth risk were reported in two other studies. Similarly, a few studies have reported main effects of interleukin 6 54, 55 on preterm birth, but others failed to replicate this association 31, 33, 45, 50. Also, the ‘−1082, −819, −592’ ATA haplotype of interleukin 10 was associated in one study 50; however, this haplotype was not tested by others 51, 53. Neither of these studies 51, 53 reported association with the interleukin 10 gene for the markers that were interrogated.

Interleukin receptor genes also show mixed results. Interleukin 1 receptor 1 (IL1R1) 50 and interleukin 10 receptor A (IL10RA) 51 show no association with preterm birth. In contrast, Hao and colleagues 51 found association with interleukin 1 receptor 2 (IL1R2) in black mothers. Interleukin 1 receptor agonist (IL1RN) shows association with the allele 2 of an intronic variable number tandem repeat in some studies 56, 57, yet association with allele 1 58, and no association 50, 51 in others. Similarly, one study of interleukin 6 receptor (IL6R) demonstrated association 33; however, no association was observed for this gene in an earlier study 30.

Other infection and inflammation‐related genes

Other infection and inflammation‐related genes have also been examined. The cytokine, CSF3, and IFNGR1, encoding the ligand‐binding chain alpha of the gamma interferon receptor, have not been associated with preterm birth 51. Additionally, TGFB1, which interacts with cytokine receptors 50, and ALOX5AP, involved in other aspects of the inflammation process 52, also have not been associated with the disorder. In contrast, LTA, lymphotoxin alpha, has shown mixed results. One study reported association with a promoter polymorphism 31, but two others did not 51, 52.

Pathogen recognition genes

Investigators have also examined genes involved in pathogen recognition. For example, MBL2 encodes a soluble mannose‐binding protein that recognizes mannose and N‐acetylglucosamine on bacterial pathogens. A promoter and two missense polymorphisms in MBL2 have been reportedly associated with preterm birth 50, 59. In addition, a member of the TNF superfamily, TNFRSF6, which encodes apoptosis (APO‐1) antigen 1 (FAS), member 6, was associated with preterm births limited to PPROM cases 60. An earlier study of TNFRSF6 including all spontaneous preterm births found no association 50. Another study 29 found that association with CD14, a surface protein preferentially expressed on monocytes and macrophages, was associated with PPROM cases only, in contrast to an earlier negative association with this gene 54. Additionally, two toll‐like receptors, TLR2 and TLR4, involved in pathogen recognition and activation of innate immunity, have been associated with preterm birth in some 61, 62 but not all 54, 63 studies analyzing their polymorphisms. Moreover, nucleotide‐binding oligomerization domain containing 2 (NOD2), which plays a role in the immune response to intracellular bacterial lipopolysaccharides, was associated with PPROM in black infants in one study 63, but was not associated with preterm birth in another study 54.

Remodeling of connective tissues

Physical remodeling of the cervix and fetal membranes are important steps in the overall normal delivery process and may represent an important control point in parturition. Indeed, preterm premature rupture of membranes is a common cause of preterm birth. Matrix metalloproteinases are responsible for degrading collagen structures and have been used as a biomarker for preterm birth. A promoter polymorphism in MMP1, matrix metalloproteinase 1, was found to be associated with PPROM in blacks 64. In another study of PPROM in blacks, a promoter polymorphism in MMP9, matrix metalloproteinase 9, increased risk for the disorder 65. SERPINH1, encoding the heat shock protein 47, is thought to be a molecular chaperone involved in the maturation of collagen molecules. As PPROM membranes have reduced collagen content, Wang and colleagues 66 sought to investigate this gene for association with PPROM. The authors found that a promoter polymorphism was associated with PPROM in a discovery as well as replication set 66. Additionally, Erichsen and colleagues 67 examined several genes involved in vitamin C absorption, solute carrier family 23 members 1 and 2, since this vitamin is required for collagen synthesis. One gene, SLC23A2, showed association, while another, SLC23A1, did not 67. PLAT, encoding plasminogen activator, tissue, also involved in tissue remodeling, was not associated with preterm birth 52.

Premature uterine contraction

The progression of the uterus from a quiescent tissue inhibited from contraction by a number or mediators, to a synchronously, forcefully contracting tissue characterizes the onset of active labor. For example, the β2‐adrenergic receptor (ADRB2) modulates uterine muscle contraction by promoting uterine smooth muscle relaxation. Studies to date have observed association of the Gly16Ala missense polymorphism in this gene 68, 69 with risk for preterm birth. A different missense polymorphism, Gln27Glu, in this gene showed association in one study 52, but not in two others 68, 69. Another neurotransmitter receptor involved in uterine contraction, dopamine receptor D2 (DRD2), was not associated with preterm birth 51. Prostaglandins also help promote uterine contractility. However, no associations were found for any of the prostaglandin pathway genes tested, prostaglandin E receptor 2 (PTGER2), prostaglandin E synthase (PTGES), or prostaglandin F receptor (PTGFR) 51. In contrast, phosphodiesterase 4D (PDE4D) degrades cAMP, a signaling molecule that promotes smooth muscle relaxation, but also showed no association with preterm birth 52.

Placental function

Placental functioning is critical for the maintenance of pregnancy and relies on appropriate blood flow between mother and fetus. Alteration in genes whose products are associated with formation of blood vessels or thrombosis may result in abnormal placental functioning and preterm birth. Vascular endothelial growth factor (VEGF) induces angiogenesis. One study observed an association between an intronic polymorphism in VEGF and preterm birth risk 70. Additionally, several hemostasis genes have been studied. One study found that factor V (F5) was significantly associated 51; however, two additional studies failed to replicate this result 52, 54. Factor VII (F7) was associated in both infants and mothers, and factor XIII (F13A1) was associated in infants only, in a study by Hartel and colleagues 71. A missense polymorphism in thrombomodulin (THBD) has been associated in infants as well 52. Other hemostasis genes assayed, factor II (F2) 52, 71, protein C receptor, endothelial (EPCR) 52, protein C (PROC) 71, and tissue factor pathway inhibitor (TFPI) 52, were not associated.

Some genes associated with blood pressure regulation have been associated with preterm birth risk. Hao and colleagues 51 found that one of two nitric oxide synthases tested, NOS2A, showed association with preterm birth in whites, but not the other, NOS3. In contrast, Gibson and colleagues 52 found association with both genes. Other genes linked to appropriate blood flow and pressure, annexin A5 (ANXA5), adducin 1 (ADD1), fibrinogen beta chain (FGB), serpin peptidase inhibitor, clade E member 1 (SERPINE1), and serpin peptidase inhibitor, clade B member 2 (SERPINB2), have not been associated with preterm birth 52.

Response to environmental toxins

Certain environmental toxins may play a role in birth timing by disrupting normal regulation of signaling pathways. For example, tobacco and other substance abuse by mothers increase the risk for preterm birth 34. Polymorphisms in genes responding to such compounds may make some mothers and/or infants more sensitive to their effects. PON1, paraoxonase 1, is a HDL‐C bound enzyme that metabolizes a variety of compounds, including drugs and pesticides 72. Lawlor and colleagues 72 tested a missense polymorphism associated with reduced activity and found a significant increase in risk that worsened with increasing number of risk alleles. Chen and colleagues 73 also found association with this polymorphism in PON1, as well as another missense polymorphism in paraoxonase 2, PON2, in Han Chinese infants who were born preterm. Additionally, OPRM1, opioid receptor mu 1, involved in the response to opioids, has been associated with preterm birth risk in Hispanics 51. In contrast, ADH1C, an enzyme involved in alcohol metabolism, did not show association 51. Several genes involved in the response to smoke exposure were examined by Nukui and colleagues 74 in both mothers and children from preterm deliveries. Of the polymorphisms tested, the glutathione S‐transferase theta 1, GSTT1, ‘null’ genotype was associated in mothers, but not infants 74. None of the other genes assayed, cytochrome P450, family 1, subfamily A, polypeptide 1 (CYP1A1), cytochrome P450, family 2, subfamily E, polypeptide 1 (CYP2E1), glutathione S‐transferase M1 (GSTM1), and N‐acetyltransferase 2 (NAT2), showed any association with preterm birth in mothers or children 74.

Fetal stress pathway

Stress has been thought to play an important role in parturition, based on data from animal studies. In many species, activation of the hypothalamic‐pituitary‐adrenal (HPA) axis, the endocrine stress pathway, with glucocorticoid production is a necessary prelude to progesterone withdrawal and active labor 75, 76. In humans, glucocorticoids do not precipitate labor, but corticotropin‐releasing hormone (CRH), a neuropeptide important in HPA axis control, is synthesized by the placenta 77. Circulating levels of CRH have been correlated with preterm and postterm pregnancy, depending upon whether they are above or below the normal range, respectively 78, 79. To date, only corticotropin‐releasing hormone‐binding protein (CRHBP) has been tested for association with preterm birth in humans and was not associated with the disorder 51.

Functional loss of progesterone

Progesterone withdrawal is extremely important in nonprimate parturition. Progesterone withdrawal, however, does not occur in human pregnancy. It has been postulated that functional loss of progesterone may initiate labor in primates. Altered function of PGR, encoding the progesterone receptor, may play a role of functional loss of progesterone human parturition. However, PGR was not associated with preterm birth in Hao et al. 51. Alterations in progesterone receptor coactivators and corepressors have been implicated as mediating progesterone resistance; however, specific genes have not been studied for variants predisposing to preterm birth.

Other

Folate is required for DNA replication and metabolism. Insufficient dietary intake of folate during pregnancy has important impacts on infant health. Functional changes in genes involved in folate metabolism may lower levels of important intermediates in DNA replication and metabolism, affecting the fetus in the same manner as reduced intake of folate. To illustrate, missense polymorphisms in two such genes, 5‐methyltetrahydrofolate‐homocysteine methyltransferase reductase (MTRR) and serine hydroxymethyltransferase 1 (SHMT1), were associated with increased risk of preterm birth in the white subset of the cohort studied by Engel et al. 80. Several other folate metabolism genes, 5,10‐methylenetetrahydrofolate reductase (MTHFR) 52, 80, 81, 5‐methyltetrahydrofolate‐homocysteine methyltransferase (MTR) 80 and cystathionine‐beta‐synthase (CBS) 80, were not associated with the disorder.

Peroxisome proliferator‐activated receptor gamma (PPARG) is a transcription factor in the nuclear hormone receptor subfamily involved in adipose tissue differentiation. This gene has been associated with type 2 diabetes and heart disease for which preterm infants are at increased risk of developing later in life 82. Carriers of the Ala12 allele of a Pro12Ala missense polymorphism had an increased risk for preterm birth in one study 82, possibly explaining the correlation observed between these disorders.

Genetic, environmental interactions, and preterm birth

Several studies suggest that gene‐gene interactions, in addition to gene main effects, influence preterm birth. A study of CD14 found both main effects for the gene and interactive effects with HSP‐70‐2 and with ILRN29. In contrast, a study found interactive effects for TNFα×IL‐6×IL6R, but no main effects for any of the genes 30. As a result, epistasis analyses may identify genes that would not otherwise be detected.

Interactions between inflammatory gene risk alleles and bacterial infections have also been shown to influence preterm birth. Two studies of IL6 found association with the gene and disorder only in the presence of bacterial vaginosis 31 or microbial invasion of the intra‐amniotic cavity 33. Hence considering gene‐gene and gene‐environment interactions may detect genes without main effects in some instances. In contrast, Macones and colleagues 32 found that TNFα had both a main effect and an interactive effect with bacterial vaginosis on preterm birth risk. A gene involved in folate metabolism, SHMT1, also showed both main and interactive effects with dietary folate intake in black women 80. As a result, many of the genes for which main effects have been detected may also interact with environmental factors to further increase risk for preterm birth.

Some polymorphisms may only increase risk in the context of other genetic polymorphisms or certain environmental factors; as a result, considering polymorphisms in multiple genes or gene‐environment interactions may increase our power to detect genetic effects. However, large sample sizes are required in order to detect such effects with confidence. Because studies to date have been relatively small, such observed interactions are tentative. Further tests with larger sample sizes are needed to verify these observations.

Limitations of previous studies

Genes studied to date biased by current understanding of physiology

The majority of human genetic studies to date have focused on one or a few candidate genes, chosen based on our current understanding of physiology. A few studies have taken a larger‐scale approach to candidate genes; however, no genes have been identified by unbiased approaches yet. One of the many benefits of systematic genetic screens is elucidating pathways that would not otherwise have been considered important for a particular disease. For example, evidence from linkage and genome‐wide association screens identified transcription factor 7‐like 2 (TCF7L2), a transcription factor believed to be involved in blood glucose homeostasis, located on chromosome 10q, as a risk factor for type 2 diabetes 83, 84. While this gene's function has a plausible role in diabetes pathology, it probably would not have been identified as a candidate gene by physiology. The association between TCF7L2 and type 2 diabetes has been replicated across numerous studies and populations, demonstrated by a global meta‐analysis 85, underscoring the ability of unbiased approaches to identify important novel genes. Moreover, pleiotropic effects of TCF7L2 on birth weight in addition to diabetes have been demonstrated 86, further signifying the utility of such approaches. Hence, systematic genome‐wide scans may help identify novel genes involved in birth timing and preterm birth.

Range of genetic variation within a gene has not been sampled sufficiently

Moreover, the lack of sufficient coverage of genetic variation within genes tested to date has been problematic. Many of the studies described in this review have genotyped one functional SNP in a particular gene. If this particular polymorphism is itself disease‐promoting, or is tightly correlated with an unsampled variant in the gene, then an association may be detected. However, additional variants may contribute to disease that are unrepresented in these studies, creating false negatives. An ideal study design would consider patterns of linkage disequilibrium (LD), i.e. the degree of correlation among polymorphisms in a region, such that each known variant in the gene is either tested directly or represented by other polymorphisms highly correlated to it. The studies that have chosen one or a few polymorphisms to test, rather than sampling the entire LD structure of the gene, have usually focused on polymorphisms in the promoter region or nonsynonymous coding changes. While such polymorphisms are highly likely to affect protein expression and/or function, other types of changes have been tied to disease as well. Synonymous coding changes can be disease‐promoting. For example, a Gly (GGC) to Gly (GGT) mutation in the calpain 3 (LGMD2A) gene alters mRNA splicing, introducing a frameshift that is thought to lead to limb girdle muscular dystrophy 87. Similarly, subtle changes in introns may promote disease by altering splicing patterns 88. Changes in the 3′ untranslated region may also be related to disease by altering mRNA stability and/or translation efficiency, such as by altering microRNA binding sites 89. As a result, a range of polymorphism types may be involved in birth timing. Because the majority of studies to date have not genotyped candidate genes thoroughly, studies with negative association results may represent false negatives. Hence, genes with negative association but incomplete gene coverage cannot be excluded from involvement in preterm birth without further study.

Heterogeneity within samples studied

Genetic heterogeneity within cases of preterm birth may also contribute to the difficulty to detect genes influencing this disorder. There are likely a variety of genetic pathways that lead to preterm birth. Genetic vulnerability factors influencing the risk for the disorder might fall into several broad categories: infection and/or inflammation, response to stress, deficits in placental function, and premature uterine contractions. Additionally as preterm birth is associated with a variety of other disorders, including type 2 diabetes and heart disease 82, some genes may have pleiotropic effects on other traits. Factors from one or more of these categories might contribute to risk in an individual, such that a sample of preterm birth cases may represent numerous genetic pathways important in the disorder. Hence, heterogeneity within the sample may reduce the power to detect a particular gene because it is important in only a subset of affected individuals. It follows that essential for gene identification will be careful, consistent phenotyping to minimize heterogeneity of cases.

Replication issues

Replication across studies is an important validation for any detected genetic association. Few positive association findings for preterm birth have been consistent. To illustrate, for TNFα, the most extensively studied gene, seven studies report major effects of the gene 32, 42, 43, 47–50; yet, five others report no major effects of the gene 30, 31, 44–46, including a meta‐analysis of seven studies 46. Similarly, IL1RN was associated in three studies 56–58, but not in another two 50, 51. One possible explanation may be that some polymorphisms are significant only in the context of a particular environmental factor, such as infection. For example, IL6 has been associated in two studies without considering environmental influences 54, 55, associated in another two studies only in the context of infection 31, 33, and not associated in an additional two studies 30, 50. Also, there has been little consistency about case or control exclusion criteria across studies, such that phenotypic and genetic heterogeneity may exist across studies. For example, preterm birth cases with PPROM have been inconsistently included in samples studied to date; however, some studies have identified association only in PPROM subsets of their samples 47, 60, suggesting that these individuals may represent a distinct genetic pathway from other cases.

Associations replicated across populations are considered particularly robust. However, such replications have been difficult to assess in preterm birth studies to date because many consist of mothers and/or children from more than one ethnic group. Few studies have included analyses separately by race or attempted to correct for possible population substructure. As a result, it is difficult to determine whether any association detected in mixed samples is due to a particularly strong signal in one ethnic group or an equally strong signal in multiple groups. In fact, the majority of positive association findings to date have been in mixed samples. Because any significant deviation between case and control groups on ethnic composition could create false positive results, the need for replication across studies is greater. A mixture of ethnic backgrounds may also reduce power to detect a gene that affects the disorder in one ethnic group but not others, possibly creating false‐negative results. Thus, future studies should consider the racial compositions of their samples carefully, in light of these issues.

Sample size and power to detect small effects

Another potential problem with studies to date has been sample size. A study's ability to detect a genetic effect is intimately related to sample size. Smaller sample sizes will only be able to detect strong genetic effects, while larger samples may detect much smaller effects. For example, assuming a discrete trait model, a sample size of 100 cases and 100 controls tested at markers with minor allele frequencies between 0.15–0.35 only has ⩾80% power at α = 0.05 to detect genetic factors with relative risks ⩾3 (calculated with 90). Because of the complex nature of genetic influences on preterm birth, it is difficult to determine a priori whether genes of small or large effect size are important in this disorder. It may be the case that very large sample sizes are needed to detect genes of interest for this disorder.

Comment

Overall, a wealth of evidence suggests that genetics are important in birth timing. Epidemiological studies demonstrate that mothers, sisters, and daughters share risk for preterm birth. Twin and other studies separating genetic and environmental factors indicate that genetics influence birth timing and preterm birth, in addition to environmental risk factors. A number of case‐control studies have identified potential candidate genes associated with preterm birth and birth timing; however, several statistical genetic concerns should be considered to improve the robustness of these findings in the future.

A variety of alternative genetic approaches may be undertaken to identify specific genes involved in preterm birth. For example, unbiased genome‐wide screens using either linkage (e.g. affected sib pairs, variance component QTL) or association methods may be used. The ease and relatively low cost of genotypic microarray platforms make whole‐genome association scans an attractive option. Both case‐control and family‐based association methods, in which population stratification is not an issue, could be used for genome‐wide scans. Additionally, nonadditive genetic effects, such as copy number or structural variation, may be important avenues for future research. Such approaches may enable investigators to identify novel genes and pathways involved in birth timing, with important clinical applications.