Ethnic differences in interleukin 6 (IL-6) and IL6 receptor genes in spontaneous preterm birth and effects on amniotic fluid protein levels

Authors


*To whom correspondence should be addressed: Scott M. Williams, Ph.D., Center for Human Genetics Research, 519 Light Hall, Vanderbilt University, Nashville, TN 37232, Telephone: 615-322-8036, Fax: 615-343-8619, E-mail: smwilliams@chgr.mc.vanderbilt.edu

Summary

Preterm birth (PTB) is a significant neonatal health problem that is more common in African-Americans (AA) than in European-Americans (EA). Part of this disparity is likely to result from the differing genetic architectures of EA and AA. To begin assessing the role of these differences, patterns of genetic variation in two previously proposed candidate genes, encoding interleukin 6 (IL6) and its receptor (IL6R), were analyzed in mothers and fetuses from 496 EA birth-events (149 cases and 347 controls) and 397 birth-events in AA (76 cases and 321 controls). IL-6 levels in amniotic fluid (AF) samples were determined in a subset of these pregnancies. Case-control comparisons revealed a single SNP in IL6R associated with PTB (p=0.04 for allelic and p=0.05 for genotype association). In addition, all of the SNPs studied showed significant frequency differences between AA and EA in at least one comparison, significantly in excess of that expected from general population databases. Higher IL-6 concentrations were associated with the IL6 SNP -661 in EA preterm samples (p=0.0056), and this result seems to be driven by microbial invasion of the amniotic cavity, indicating a gene by infection interaction. These findings indicate that, as a function of IL6 genotype, EA and AA women respond differently to infection with respect to their expression of IL-6. Our data support differential genetic control of levels of IL-6 in amniotic fluid between EA and AA.

Introduction

Spontaneous preterm birth (<37 weeks gestation—PTB) occurred in 12.3% of all pregnancies in the United States in 2003. PTB is the largest contributor to neonatal morbidity and mortality; 60-80% of the deaths among infants without congenital anomalies occur as a result of PTB (Goldenberg & Andrews, 1996). Despite the importance of PTB from a public health perspective, the etiologies of PTB are not well understood, and no effective interventions are available (Martin et al. 2005).

PTB is a complex disease induced by several etiologic factors from potentially interacting pathophysiologic pathways (Lockwood & Kuczynski, 2001). These pathways culminate in a terminal biochemical pathway resulting in the activation of uterotonins, such as prostaglandins, that induce labour (Lockwood & Kuczynski, 2001). Of the different pathways that have been implicated in PTB, infection and the host inflammatory response to infection (e.g., cytokine and matrix metalloproteinase activation) are among the most commonly found precursors to PTB (Hillier et al. 1988; Romero & Mazor, 1988; Romero et al. 1990, 2006). Based on these data we hypothesize that the response to infection in some women increases their risk of PTB compared to individuals of a different genetic constitution (Roberts et al. 1999; Macones et al. 2004).

In the US a ∼2-fold difference in rates of PTB exists between African-Americans (AA) and European-Americans (EA). The rates of PTB in EA and AA have converged; from 1989 to 1997 there was a 15.6% increase in the rate of PTB among EA to 10.2%, while AA experienced a 7.6% decrease to 17.5% over the same period (Demissie et al. 2001). This convergence of rates has continued, with the EA PTB rate being 11.1% in 2002 and the AA being 17.5% (Martin et al. 2003). The cause(s) of these changes in disparity is also not well understood, although it is possible that the disparity can be used to identify candidate risk factors that are significantly different between ethnic groups, thereby helping to define the underlying etiology of the phenotype.

Data from genetic studies in pregnancy and pregnancy complications indicate that there may be a difference in genetic predisposition among individuals, and perhaps ethnic groups (reviewed in Varner & Esplin, 2005). These data include: 1) previous PTBs are associated with an increased risk of future PTBs; 2) an association has been seen with ethnicity and PTB; 3) mothers who were born PTB or have sisters who had a PTB have a higher risk of delivering a PTB infant (Carr-Hill & Hall. 1985; Goldenberg & Andrews, 1996; Porter et al. 1997). In addition, twin studies found a predisposition to PTB with heritability ranging from 20-40% (Treloar et al. 2000; Clausson et al. 2000).

One plausible candidate associated with infection that may affect the risk of PTB is interleukin 6 (IL-6). Increased concentration of this proinflammatory cytokine in cervical and amniotic fluid (AF) has been associated with microbial invasion of the intra-amniotic cavity (MIAC) in women delivering preterm (Romero et al. 1990; El Bastawissi et al. 2000). Increased IL-6 concentration in cord plasma is also associated with fetal inflammatory response to maternal infection (Yoon et al. 2000). For these reasons, IL-6 is considered a “marker” of high risk status for PTB (Romero et al. 1993). In addition, genetic association studies demonstrated that a single nucleotide polymorphism (SNP) in the promoter region of the IL6 gene (C-174G) decreases promoter activity, and the (G/G) homozygote associates with increased risk of PTB (Hartel et al. 2004). IL6, as well as the gene encoding its receptor (IL- 6 receptor IL6R), through which IL-6 mediates its biological function, are therefore strong candidates for PTB and perhaps for the ethnic disparity in PTB. The present case-control association study was undertaken to better understand the patterns of variation of single nucleotide polymorphisms (SNPs) in the IL6 and IL6R genes. In this study, we compared the maternal and fetal allelic, genotypic and haplotypic patterns between AA and EA and between cases (preterm birth) and controls (normal term deliveries) within AA and EA. In addition, we assessed the effects of genetic variation on AF IL-6 concentration in cases and controls in both AA and EA, in our efforts to understand ethnic disparity.

Materials and Methods

Subjects

EA and AA subjects were recruited at the Centennial Medical Center, Nashville, TN and Magee Women's Hospital, Pittsburgh, PA. Institutional Review Boards at TriStar Nashville, TN, Vanderbilt University, Nashville, TN and Magee Women's Hospital, Pittsburgh, PA approved this study. Initially, subjects were recruited sequentially based on the availability of recruiters. Subsequently, recruitment focused on AA subjects and cases. All included pregnancies were singleton live births. Ethnicity was identified by self-report and a questionnaire that traced ethnicity back one generation from the mother and father. Individuals who had more than one racial group in their ancestry (including the grandparents of the infant) were excluded from the study. We recruited mothers between the ages of 18 and 40. Gestational age was determined by last menstrual period dating and corroborated by ultrasound dating. In our study, cases (PTBs) were defined as the presence of regular uterine contractions at a minimum frequency of 2-contractions/10 minutes followed by delivery at <360/7 weeks gestation. We excluded subjects with multiple gestations, preeclampsia, preterm premature rupture of the membranes, placental previa, fetal anomalies, gestational diabetes, poly- and oligohydramnios, and other complications such as surgeries during pregnancies. The controls consisted of women having normal labour and delivery at term (≥ 370/7 weeks) with no medical or obstetrical complications during pregnancy.

Demographic and Clinical Characteristics

Our genetic study included a total of 496 birth events in EA (149 cases and 347 controls) and 397 in AA women and fetal (76 cases and 321 controls). Placental pathology was performed in all cases to document histologic chorioamnionitis and funisitis. The means and standard deviations of demographic and clinical characteristics for cases and controls are listed in Table 1. Significant differences between cases and controls were observed in EA and AA for APGAR1, birth weight (kg) and gestational age at delivery (weeks) with p<0.0001.

Table 1.  Clinical and demographic information.
VariableEuropean AmericanAfrican-American
MeanP-Value1MeanP-Value1
PretermTermPretermTerm
  1. Means are reported with standard deviations reported in parentheses

  2. 1P-values compare preterm cases to term controls

Age (yrs)26.7426.960.7125.3924.5340.26
(6.43)(5.90) (5.91)(5.572) 
Apgar16.738.19<0.00016.328.14<0.0001
(2.02)(1.25) (2.70)(1.11) 
Birth Weigh t(kg)1.803.36<0.00011.833.26<0.0001
(0.70)(0.50) (0.79)(0.52) 
Gestation Age at Delivery (wks)32.1939.30<0.000131.6239.11<0.0001
(3.16)(1.21) (4.29)(1.13) 
Gravidity2.172.130.772.412.460.43
(1.29)(1.45) (1.46)(1.60) 

DNA Sampling and Genotyping

DNA was isolated from maternal and fetal blood samples using the Autopure automated system (Gentra Systems (Minneapolis, MN)).

A total of eight SNPs were screened in the IL6 and IL6R genes (5 in IL6 and 3 in IL6R). We used the reported frequencies of the SNPs and their relative distance in the genes as criteria for SNP selection. The rs numbers of those SNPs from the NIH database, SNP locations, and relative distances are shown in Table 2. We did not use functional significance of SNPs as selection criteria, with the exception of IL6 -237 (rs1800795, also reported as -174) that was previously shown to associate with PTB (Simhan et al. 2003). Genotyping was done using the TaqMan assay on an ABI 7900 (Applied Biosystems of Foster City, CA).

Table 2.  Positional information of SNPs.
Geners#Location1Position (bp)Genic Region
  1. 1Location refers to position from translational start site

IL61880243−722722532895Promoter
(Chromosome 7)1800797−66122539461Promoter
1800796−63622539486Promoter
1800795−23722539885Promoter
1554606182622541947Intron
IL6R668772622215151213393Intron
(Chromosome 1)484562233314151224492Intron
484562337672151228850Intron

Cytokine Measurements

Amniotic fluid (AF) samples were collected during active labour (either preterm or term) by transvaginal amniocentesis before rupture of the membranes and delivery. Amniotic fluid samples were only collected in Nashville. Samples were collected before preterm or term vaginal deliveries by puncture of the intact membrane using a 22 gauge needle prior to artificial rupture of the membranes. A few samples were also collected at the time of caesareans. The AF was centrifuged immediately for 10 minutes at 2500 RPM, to remove cellular and particulate matter. Aliquots of AF were stored at –70°C until analysis.

IL-6 concentration in the AF was measured using multiple solid phase sandwich immunoassays utilizing antibody coated beads (Biosource International, Camarillo, CA) and analyzed with a Luminex™ (Austin, TX). Standard curves were developed using duplicate samples of known quantities of recombinant proteins provided by the manufacturer. Sample concentrations were determined by relating the absorbance obtained to the standard curve by linear regression analysis. When necessary, samples were diluted in assay buffer provided by the manufacturer in addition to the doubling dilution required to perform the assay. The sensitivity of the assay was 1 pg/ml.

Both maternal genetic data and AF IL-6 concentrations were available for 54 EA cases and 37 EA controls, while both fetal genetic data and AF data were available from 57 EA cases and 42 controls. In AA 30 cases and 67 controls had both maternal genetic data and AF data, while 26 cases and 68 controls had both fetal genetic data and AF data.

MIAC was defined either by polymerase chain reaction (PCR) testing and/or clinical evidence. Presence of bacteria in the AF was detected by PCR amplification of microbial 16S ribosomal DNA (TaqMan Assay, CA). Microbial cultures were not attempted since the PCR-based detection of microbes is much more sensitive than traditional cultures (Hitti et al. 1997; Gardella et al. 2004). Microbial species identification using PCR was not performed in this study. Preterm cases with clinical evidence of MIAC were those individuals having three or more of the following criteria: abdominal tenderness, temperature > 38°C, foul smelling vaginal discharge, elevated C-reactive protein (CRP) or histologic chorioamnionitis.

Single Locus Allele and Genotype Analyses

T-tests or Mann-Whitney rank sum tests were performed to compare demographic and clinical measurements between cases and controls. Single locus allele frequencies were analyzed using software based on the protocols of Raymond & Rousset, using the software Tools for Population Genetic Analysis, TFPGA, version 1.3 (available at http://www.marksgeneticsoftware.net/)(Raymond & Rousset, 1995). Genotype distributions at single sites were compared between ethnic groups using the program R X C from the same website, and are based on the Metropolitan algorithms. Single site Hardy-Weinberg (H-W) analyses were also performed using TFPGA (Raymond & Rousset, 1995). Statistical significance for the above was determined using Fishers Exact Tests. All analyses were carried out on pooled Nashville and Pittsburgh samples, because preliminary analyses showed that the two samples did not differ significantly within cases or within controls with regards to allele or genotype frequencies.

Haplotype Analysis

Pairwise linkage disequilibrium was characterized and haplotype frequencies were calculated for EA and AA using Powermarker (http://www.powermarker.net.) (Zaykin et al. 2002) and/or HaploView statistical software (Barrett et al. 2005). Results from the two software packages did not differ. Haplotype blocks were assigned using the algorithm created by Gabriel et al. (2002). Both Powermarker and HaploView use an EM algorithm to determine haplotype frequency distributions when phase is unknown. The Powermarker haplotype trend analysis was performed to test for haplotype frequency differences between groups. This analysis uses a regression approach to test haplotype-trait association. This approach can be applied to both quantitative traits and dichotomous traits. The test for association then uses an F test for a specialized additive model (Zaykin et al. 2002). This method was used to test for differences between AA and EA, as well as between phenotype classes within an ethnic group.

IL-6 Concentration Analysis

One and two-way analysis of variances (ANOVA) were used to examine AF IL-6 associations with IL6 genotypes and pregnancy outcome and within EA and AA. Both fetal and maternal genotypes were analyzed for association. STATA was used for ANOVA analyses (StataCorp, 2005). A log transformation was performed on IL-6 concentrations to normalize data, because the Shapiro-Wilk test for normality demonstrated that cytokine concentrations deviated from normality. One-way ANOVA was performed with IL6 markers to determine if the cytokine concentration differs with regards to an individual's genotype within a status groups. A two-way ANOVA was also performed with phenotype status and individual IL6 markers as covariates, to determine if cytokine concentration associates with status and individual markers. A two-way ANOVA was also used to assess association between marker and MIAC status within cases.

Correction for Multiple Testing

False discovery rate (FDR) was used to correct for multiple testing. FDR is used to measure global error, that is, the expected number of false rejections of the null hypothesis among the total number of rejections (Sabatti et al. 2003). FDR measures the proportion of false positives among all SNPs called as significant. The FDR adjusted p-values are defined in a step-down fashion. The false discovery rate used was 0.20.

Results

Analysis of Differential Distribution IL6 and IL6R SNPs in AAs and EAs

Significant differences between AA and EA were observed for both the genotype and allele frequencies in IL6 and IL6R for maternal and fetal samples (Table 3). Of the 20 comparisons for allele frequency differences between EA and AA in the IL6 gene (maternal, fetal, cases and controls x 5 SNPs), 13 were significant at the 0.05 level (3 in fetal cases, 3 in fetal controls, 3 in maternal cases, 4 in maternal controls) and 8 were significant with p values below 0.0001. Similarly, genotype frequency comparisons between the racial groups were different in 13 out of 20 comparisons at the 0.05 level (3 in fetal cases, 3 in fetal controls, 3 in maternal cases, 4 in maternal controls) and 8 at the 0.0001 level. Comparisons of the IL6R SNPs between racial groups were even more striking, with 11 out of 12 comparisons being significant at the 0.05 level for allele frequencies (2 in fetal cases, 3 in fetal controls, and 3 each in maternal cases and controls). The only value not significant at the 0.05 level was fetal allele frequency at position 37672 in cases with a marginal p value (p = 0.09). The genotype comparisons for IL-6R were comparably different, with 10 out of 12 SNPs showing significant differences (2 in fetal cases, 3 in fetal controls, 2 in maternal cases, 3 in maternal controls). It is clear from these data that most SNPs differed between ethnic groups within a phenotypic class in both maternal and fetal samples.

Table 3.  Single locus analysis of African-Americans and European Americans.
a. Infant
MarkerEAEAAAAAEA v AA AllelesEA v AA Genotypes
AlleleAllele FrequencyHWE P-Value1Case-Control P-ValueAllele FrequencyHWE P-Value1Case-Control P-ValueP-ValueP-Value
PretermTermPretermTermAllele2Genotype3PretermTermPretermTermAllele2Genotype3PretermTermPretermTerm
  1. 1 HWE-p-values testing for deviations from Hardy Weinberg Equilibrium

  2. 2 Indicates comparisons between preterm and term allele frequencies.

  3. 3 Indicates comparisons between preterm and term genotype frequencies

  4. *Opposite minor alleles for these two markers in AA and EA.

−7227A0.250.230.080.250.540.100.150.210.840.340.640.050.160.190.29
−661A0.380.360.070.010.640.910.090.110.350.210.60.86<0.0001<0.0001<0.0001<0.0001
−636C0.090.050.120.010.670.490.070.080.220.2210.470.640.050.640.13
−237C0.40.370.010.010.370.610.080.110.290.220.470.54<0.0001<0.0001<0.0001<0.0001
1826T0.440.390.030.050.250.380.350.340.110.780.910.340.160.080.010.02
22215*A0.430.460.380.230.420.680.590.420.390.5110.650.0075<0.00010.04<0.001
33314C0.430.390.520.890.400.530.180.140.620.180.190.15<0.0001<0.0001<0.0001<0.0001
37672*G0.470.420.180.490.340.150.570.4410.820.9110.09<0.00010.18<0.001
 
b. Maternal
MarkerEAEAAAAAEA v AA AllelesEA v AA Genotypes
AlleleAllele FrequencyHWE P-Value1Case-Control P-ValueAllele FrequencyHWE P-Value1Case-Control P-ValueP-ValueP-Value
PretermTermPretermTermAllele2Genotype3PretermTermPretermTermAllele2Genotype3PretermTermPretermTerm
 
−7227A0.210.220.130.590.750.510.200.151.000.670.300.340.780.0020.610.01
−661A0.340.330.450.350.690.410.080.121.000.030.170.48<0.0001<0.0001<0.0001<0.0001
−636C0.070.061.000.330.670.490.090.081.000.140.860.550.560.130.390.310
−237C0.370.340.850.270.330.460.080.130.320.040.090.34<0.0001<0.0001<0.0001<0.0001
1826T0.430.380.380.150.170.130.290.320.120.680.460.480.0080.020.010.04
22215*A0.430.440.210.220.890.880.540.420.610.720.420.640.05<0.00010.12<0.0001
33314C0.430.380.610.240.190.250.200.160.100.530.290.36<0.0001<0.0001<0.0001<0.0001
37672*G0.470.430.400.910.260.360.650.450.270.360.040.05<0.0001<0.00010.003<0.001

As might be expected from the patterns of single SNP differences in both fetal and maternal samples, haplotypes displayed strong differences in frequencies when comparing AA and EA in both cases and controls. Highly significant differences (p <0.0001) were observed in IL6 and IL6R haplotypes in all comparisons (Table 4).

Table 4.  Comparisons of European-American and African-American haplotype frequencies.
Fetal Sample
StatusP-Value
 IL6IL6R
   
Preterm<0.0001<0.0001
Term<0.0001<0.0001
 
Maternal Sample
StatusP-Values
 
 IL6IL6R
   
Preterm<0.0001<0.0001
Term<0.0001<0.0001

Case-Control Association Analyses

Deviations from H-W Equilibrium were observed in EAs at -661, -636 and -237 in fetal control samples (p= 0.01), 1826 in fetal controls (p= 0.05), -237 in fetal cases (p= 0.01), and 1826 in fetal cases (p= 0.03) for IL6 (Tables 3a). Significant deviations from H-W Equilibrium were also observed among AA at -661 in maternal controls (p= 0.03), and -237 in maternal controls (p= 0.04; Table 3b). None of these p-values have been corrected for multiple testing.

Single locus association analyses (allelic and genotypic) of EA maternal and fetal samples found no significant associations between SNPs in IL6 or in IL6R and PTB (Table 3a and 3b) including the -237 SNP (commonly referred to as -174 from the transcription start site) that was previously shown to be associated with PTB. Allelic and genotypic analyses in AA found one significant association between the SNP at 37672 for IL6R maternal samples (allele p= 0.04, genotype p= 0.05) with PTB (Table 3b). A marginally significant allele frequency difference between cases and controls was found at -237 for maternal samples (p= 0.09; Table 3b).

Linkage disequilibrium patterns were characterized for all populations examined, using r2 as the metric (Figure 1). EA controls had strong to moderate LD in both IL6 and IL6R for both fetal cases and controls. Fetal case and fetal control haplotype structures were similar to each other in EA, as were EA maternal case and EA maternal control patterns (Figure 1a–d). IL6 markers -237 and 1826 formed a block in all EA samples. IL6R markers 22215 and 33314 formed a block in all control samples, as well as in EA cases (Figure 1b, d, f and h). As expected there was generally weaker LD present in AA samples as compared to EA (Figure 1e–h). However, in the AA LD appeared to be stronger in the control group as compared to the AA cases.

Figure 1.

Fetal and maternal IL6 and IL6R receptor linkage disequilibrium structure. Haplotype block structure and LD across IL6 and IL6R.The left-hand side of each panel is IL6 and the right-hand side is IL6R. Numbers in the diamonds are r2 values, with darker shades of grey indicating higher values. Blocks are denoted as described in the text.

Haplotype regression analyses were also performed by treating PTB as a dichotomous variable (Table 5a and b). No significant associations were observed between haplotype frequencies and PTB in fetal or maternal samples in either ethnic group or either gene. Only haplotype frequencies present in 5.0% or more in at least one of the groups were included in the Tables presented.

Table 5.  Haplotype frequencies and trend regression results
a. IL6 Fetal Samples
−7227−661−636−2371826FrequencyP-Value
EAAAEAAA
PretermTermPretermTerm
AGGGG0.200.210.150.160.190.85
CGCGG0.090.050.070.08 
CGGGG0.260.340.430.42 
CGGGT0.050.030.260.21 
CAGCT0.330.340.090.10 
Maternal Samples
−7227−661−636−2371826FrequencyP-Value
EAAAEAAA
PretermTermPretermTerm
AGGGG0.170.200.170.110.460.53
CGCGG0.070.060.070.06 
CGGGG0.330.350.470.48 
CGGGT0.060.050.190.19 
CAGCT0.300.310.070.11 
b. IL6R Fetal Samples
222153331437672FrequencyP-Value
EAAAEAAA
PretermTermPretermTerm
AAA0.420.430.320.380.750.09
AAG0.020.030.270.2 
GAA0.110.150.110.06 
GAG0.03<0.010.120.22 
GCG0.430.380.180.14 
Maternal Samples
222153331437672FrequencyP-Value
EAAAEAAA
PretermTermPretermTerm
AAA0.410.400.320.380.520.39
GAA0.120.160.040.08 
GCG0.420.370.200.16 
AAG0.030.030.220.20 
GAG0.030.020.220.18 

IL-6 Amniotic Fluid Cytokine Analysis

Overall, cytokine concentrations were higher in EA cases than EA controls, but this was not evident in AA (p values for all comparisons presented in Table 6, mean values available from authors upon request). In EA cases there was a significant association between maternal genotype at variant G-661A and IL-6 concentration (AA = 12010.5 pg/ml, AG = 4885.88 pg/ml, GG 13485.1 pg/ml, p = 0.0056, Table 6 and Figure 2a). In comparison, there were no association between EA controls (AA = 4297.99 pg/ml, AG = 2850.34 pg/ml and GG = 3246.02 pg/ml, p=0.5913, Table 6 and Figure 2a). The two-way ANOVA also revealed a significant association for phenotype and -661 genotype in EA mothers (p = 0.0083, Table 6b and Figure 2a). The AA samples did not show any differences in IL-6 concentration as a function of the genotype at -661 (Table 6c–d, Figure 2b). However, in AA there were only two significant associations between genotype and phenotype, and they were for the -7227 for fetal genotype in term samples (p=0.0442, Table 6c) and for the -636 maternal genotype in term samples (p=0.0399; Table 6d).

Table 6.  IL-6 cytokine concentration analysis
a. European American Fetal ANOVA Results
MarkerUncorrected P-ValueMean Cytokine Concentrations
Marker & Status1Marker & MIAC3Preterm (N=57)2Term (N=42)2Preterm (pg/ml)Term (pg/ml)
−72270.95350.85710.68300.55418526.933294.96
−6610.32070.61950.64580.4537 
−6360.79330.86380.86830.6834 
−2370.48370.24920.31580.938 
18260.97390.95440.95580.7066 
Status20.0028 
MIAC0.2956 
b. European American Maternal ANOVA Results
MarkerUncorrected P-ValueMean Cytokine Concentrations
Marker & Status1Marker & MIAC3Preterm (N=54)2Term (N=37)2Preterm (pg/ml)Term (pg/ml)
−72270.95920.62490.73860.82428557.33235.1
−6610.00830.00090.00560.5913 
−6360.81320.56310.48690.9247 
−2370.24520.03540.12730.6398 
18260.24030.06280.05810.9804 
Status20.0116 
MIAC0.4140 
c. African American Fetal ANOVA Results
MarkerUncorrected P-ValueMean Cytokine Concentrations
Marker & Status1Marker & MIAC3Preterm (N=26)2Term (N=68)2Preterm (pg/ml)Term (pg/ml)
−72270.29790.43430.35940.04425055.614129.07
−6610.59230.66950.45870.3978 
−6360.73650.39420.59990.3483 
−2370.59490.69830.47290.3978 
18260.54620.56400.71840.7139 
Status20.3766 
MIAC0.1871 
d. African American Maternal ANOVA Results
MarkerUncorrected P-ValueMean Cytokine Concentrations
Marker & Status1Marker & MIAC3Preterm (N=30)2Term (N=67)2Preterm (pg/ml)Term (pg/ml)
  1. 1Two way ANOVA with marker and status as covariates

  2. 2One way ANOVA

  3. 3Two way ANOVA with marker and infection as covariates within preterm cases only

  4. *EA-fetal with genotype and MIAC information = 15, without genotype and MIAC information = 39; EA-maternal with genotype and MIAC information = 17, without genotype and MIAC information = 35; AA-fetal with genotype and MIAC information = 6, without genotype and MIAC information = 18; AA-maternal with genotype and MIAC information = 8, without genotype and MIAC information = 20

−72270.78810.41330.61680.50244591.774256.86
−6610.88980.46130.55670.6811 
−6360.12820.72370.96030.0399 
−2370.88460.45260.55060.6811 
18260.25580.87640.84430.1453 
Status20.2716 
MIAC0.3391 
Figure 2.

Figure 2.

Mean cytokine concentrations by genotype for IL6–661 in European American and African American maternal samples. Mean values are shown in the histograms for EA by genotype (a) and AA by genotype (b). Case values are in light grey and controls values are in dark grey.

Figure 2.

Figure 2.

Mean cytokine concentrations by genotype for IL6–661 in European American and African American maternal samples. Mean values are shown in the histograms for EA by genotype (a) and AA by genotype (b). Case values are in light grey and controls values are in dark grey.

Cases were also analyzed for the presence and absence of MIAC and for association between genotype and MIAC (Table 6a–d). In EA cases a two-way ANOVA analysis for cytokine concentration was performed, assessing maternal genotypes and MIAC, and two SNPs were found to have significant associations (-661 (p= 0.0009) and -237 (p= 0.0354) (Table 6b and Figure 2c)) with MIAC. These markers were not statistically significantly associated with MIAC in AA cases. These results indicate a significant genotype by MIAC interaction in EA cases. After performing an FDR correction, with a false discovery rate of 0.2, all statistically significant p values (original cutoff p< 0.05) remained statistically significant, with the exception of the association of -237 and MIAC (p= 0.0354) in EA cases. In AA neither of the p-values remained significant after FDR correction.

Discussion

A case-control association study was performed to examine genetic variations in the IL6 and IL6R genes, in efforts to explain ethnic disparity in the PTB rate between AA and EA. Our data demonstrate that AA and EA differ in allele, genotype, and haplotype frequencies, with one IL6R marker (37672) found to associate statistically between cases and controls within AA maternal samples. Our data also provide preliminary evidence that IL-6 AF concentration differs according to promoter variants in the IL6 gene in EA cases, and that there is evidence for an interaction between maternal genotype and case-control and MIAC status in EA, but not in AA.

With respect to patterns of genetic variation our results are very similar to previously published findings that found allele and genotype differences between EA and AA for one of the SNPs we analyzed (rs1800795, which is typically referred to by its location from the transcriptional start site -174) (Cox et al. 2001; Hoffmann et al. 2002; Hassan et al. 2003). For example, in Cox et al the C allele frequency was reported as 0.35 in EA and 0.09 in AA. We found that the C allele frequency was 0.37 in cases and 0.34 in control EA, and 0.08 in AA cases and 0.11 in AA controls (Cox et al. 2001), indicating that ethnicity correlates with differences in IL6 gene variation. Our study, however, is more comprehensive than those previously published in that we examined more markers and included both maternal and fetal samples. We, therefore, extended the findings to demonstrate that in both IL6 and IL6R there are many significant differences between AA and EA. In addition, we identified significant differences in haplotype frequencies in both maternal and fetal samples in both cases and controls between the two groups. Our data provide substantial evidence of ethnic differences in these genes.

To assess whether the differences we found are in line with overall genetic differences between EA and AA, a χ2 test comparing allele frequencies between AA and EA was performed on all SNPs available from the Perlegen SNP genotype database (Hinds et al. 2005). Analyses of allele frequency differences between AA and EA, using SNP genotype data available from the Perlegen database, demonstrated that 41.98% (p< 0.05) of SNPs differ between these groups. These p-values were not corrected for multiple testing. In our data, all of the SNPs examined demonstrated significant differences in allele frequencies between these two ethnic groups, and 7 out of 8 did so at the genotype level in at least one comparison. This is more than double the expected number of allele differences, indicating that the loci we analyzed differ between the ethnic groups more than would be expected from the genomic background.

The results from the single locus case-control analyses identified a single SNP at 37672 associated with PTB in IL6R for both allelic and genotypic frequencies in AA mothers, although this result was not corrected for multiple comparisons. No significant haplotype associations were observed. These results are suggestive of the limitations of single locus associations in a complex phenotype such as PTB. It is possible that even though the genetic variants we studied affect PTB, they do so only in the context of other variants or other endogenous or exogenous risk factors. For example, we have already indicated that a multilocus interaction between the TNF-α, IL6 and IL6R genes is predictive of ∼65% of PTB in EA women (Menon et al. 2006). Other studies have documented that single promoter SNPs in IL1B have different effects depending on what other SNPs are present elsewhere in the promoter region (Chen et al. 2006).

In addition to examining allele, genotype and haplotype frequencies for association with PTB, we looked at the relationship between genotype, case status and IL-6 AF concentration. We identified a significant association between IL-6 concentration and marker -661 in EA maternal cases; this association was still significant after an FDR correction for multiple testing. This finding has several potential implications for the role of IL-6 in PTB. It suggests that AF IL-6 concentration is more closely associated with maternal than fetal genotype. However, the association with cytokine phenotype is not universal, but differs by ethnicity and case status, indicating that other factors interact with these parameters to affect AF IL-6 concentration.

Cytokine analyses examining the effect of marker and MIAC status within case samples also revealed an association with IL6 marker -661. This result remained statistically significant after an FDR correction, and is consistent with the interpretation that the IL-6 AF association between -661 and EA cases is primarily due to MIAC status. Interestingly, a marginally significant association was also found at -237 in EA that was previously found to associate with PTB (Simhan et al. 2003). In our study we found that -237 and -661 are in strong LD. It is therefore likely that the previous association with -237 may have actually been due to the affects of –661, or another SNP in very strong LD with it, because -661 but not -237 significantly associates with AF concentrations.

The association of cytokine concentration with -661 genotype and MIAC seen in EA PTB individuals suggests that an interaction between maternal genotype and environment (MIAC) may in fact be important in PTB. These results have several implications for the understanding of the mechanism of PTB. First, our findings are supportive of a maternal genetic contribution to AF IL-6. This may be because maternal inflammatory cells (such as macrophages and NK cells) present in AF may function as major contributors to AF IL-6. Second, it suggests that IL6 genotype may affect pregnancy outcome, as indicated by previous studies (Romero et al. 1990; El Bastawissi et al. 2000), but that the effects are not strong independent predictors of clinical phenotype, especially in the absence of an environmental factor such as MIAC. Instead, IL-6 concentration is more directly associated with SNP genotype as a product of gene by environment interaction, and the genotype explains IL-6 concentration only as part of an interaction network with either other unstudied SNPs in this or other genes and/or environmental factors. Moreover, this association and interaction is limited to EA, making this finding a potentially important basis for ethnic disparity studies in PTB.

It is clear from our data that the distribution of genetic variation at the IL6 and IL6R genes differs significantly between EA and AA. Given these findings and those of previous researchers it is reasonable to hypothesize a potential role for these genes in the disparity in PTB between EA and AA. Such a conclusion is also supported by the relationship that we documented between genotype at an IL6 promoter and AF IL-6 concentration in some, but not all, of our comparisons. Evidence for ethnic-specific association with IL-6 concentration provides even more support for this hypothesis. However, although these data alone do not support a role for these genes in PTB, in conjunction with other genetic and physiological data the results strongly suggest a plausible functional role for them. In addition, our data provide a good baseline for understanding the patterns of variation in these populations that will be useful in future genetic epidemiological studies of this important phenotype.

Acknowledgements

This study is funded by Thrasher Research Funds, Salt Lake City, UT.

Ancillary