Does Genetic Regulation of IgE Begin In-Utero? Evidence from T_H1/T_H2 Gene Polymorphisms and Cord Blood Total IgE

Abstract

Background

Elucidation of early life factors is critical to understand the development of allergic diseases, especially those manifesting in early life such as food allergies and atopic dermatitis. Cord blood IgE (CBIgE) is a recognized risk factor for the subsequent development of allergic diseases. In contrast to numerous genetic studies of total serum IgE in children and adults, limited genetic studies on CBIgE have been conducted.

Objective

To test the associations between functional or tagging single nucleotide polymorphisms (SNPs) in genes involved in the T_H1/T_H2 pathway and CBIgE in a large U.S. inner-city birth cohort.

Methods

CBIgE, measured by Phadia ImmnunoCAP, was analyzed as a continuous and a binary variable. The association of each SNP with the two outcomes was tested using tobit and logistic regression models, respectively, with adjustment for pertinent covariates, ancestral proportion, and multiple testing. Ethnic heterogeneity and gene-gene interactions were also explored.

Results

Three SNPs (rs1800925, rs2069743 and rs1295686) in the IL13 gene were significantly associated with CBIgE concentration (p≤6×10^-4, p_FDR<0.05). These SNPs jointly influenced CBIgE in a dose-response manner (p_trend=9×10^-8). Significant associations also were observed for SNPs in the IL13RA1 (rs5956080) and STAT6 (rs11172106) genes. Ethnicity-specific genetic effects were observed for SNPs in the IL5 and GATA3 genes. Several gene-gene interactions (including IL13-IL4R and IL13-STAT6 interactions) were detected in relation to CBIgE.

Conclusion

Our data demonstrated that multiple SNPs were individually and jointly associated with CBIgE, with evidence of gene-gene interactions and ethnic heterogeneity. These findings suggest that genetic regulation of IgE may begin in-utero.

Introduction

The rising prevalence of allergic diseases is a growing clinical and public health problem in the U.S. and worldwide^1-3. Most childhood allergic diseases, especially food allergies and atopic dermatitis, develop in the first few years of life^{4, 5}. As such, elucidation of early life factors is critical to understand the development of allergic diseases. Cord blood IgE (CBIgE) is a recognized risk factor for the subsequent development of allergic diseases^{6, 7}. In contrast to numerous genetic studies of total serum IgE in children and adults, the genetic determinants of CBIgE remain largely unexplored. Elucidation of genetic determinants of CBIgE may provide new mechanistic insight into IgE regulation in early life, and may help us understand conflicting findings with regard to whether sensitization to individual environmental allergens begins during gestation^{8, 9} or later in life^{10, 11}. Furthermore, identification of genetic determinants of CBIgE may provide novel biomarkers for the early identification of infants at risk for developing allergic diseases.

IgE production in children and adults is known to be under strong genetic control^{12, 13}, with heritability ranging from 60 to 87% in childhood. IgE is produced by activated B cells, which interact with T_H2 cells and undergo isotype class switching after the induction of T_H2 cell-derived cytokines, most prominently interleukin (IL)-4 and IL-13. It is well known that an imbalance between T_H1 and T_H2 immune response is critical to IgE production and to the subsequent development of allergic diseases. In addition, increasing evidence suggests that inappropriate T_H1 and T_H2 responses can be suppressed by T-reg cells¹⁴. To date, a large number of candidate gene association studies have been conducted for IgE in children and adults¹⁵.

Remarkably, the heritability of CBIgE was higher (84-95%) than total IgE in childhood as shown by a twin study¹². In contrast to numerous genetic studies on total IgE, limited genetic studies on CBIgE have been conducted^16-21. So far only IL13 gene polymorphisms have been consistently associated with CBIgE in both Caucasian and Asian populations ^{16, 17}. Most published genetic studies of CBIgE have examined only one or a few candidate genes per study^{16, 18-21}, and some of these studies were small in sample size^{17, 18, 21}(ranging from 300 to 650). To our knowledge, only one study has systematically examined a large number of candidate genes in relation to CBIgE in a Chinese population¹⁷. No genetic studies of CBIgE have been conducted in African Americans, a population with a high risk of allergic diseases, which may be due to unique genetic susceptibility and/or environmental exposures.

The purpose of this study was to determine whether the known genetic variants for postnatal IgE or other allergic phenotypes are associated with CBIgE in a large U.S. inner-city birth cohort of predominantly African Americans, with adjustment for pertinent covariates, ancestral proportion, and multiple testing. Specifically, this study focuses on genes in the T_H1 pathway (e.g. interleukin 2(IL2), IL12, IL18, interferon-gamma (IFNG)); T_H2 pathway (e.g. IL4, IL13, IL-4 receptor (IL4R), IL13 receptor alpha 1 (IL13RA), IL5, IL-5 receptor alpha (IL5RA), janus kinases (JAKs), signal transducer and activator of transcription 6 (STAT6)); and T-reg pathway (e.g. forkhead box P3(FOXP3), IL10, and transforming growth factor, beta 1 (TGFβ1)). In addition, we explored ethnic heterogeneity and gene-gene interactions in relation to CBIgE.

Materials and Methods

Study Population

This study included 1,070 children from the Boston Birth Cohort, a cohort consisting of multiethnic mother-infant pairs (predominantly African Americans) enrolled 24 to 72 hours post-delivery and followed up prospectively from birth onward, as detailed in a previous publication²². Comprehensive pre- and peri-natal epidemiological and clinical variables along with cord blood samples were collected after informed consent was obtained. The study protocols were approved by the institutional review boards of the Boston University Medical Centerand Children’s Memorial Hospital (CMH) in Chicago.

CBIgE Measurement

CBIgE concentration in plasma was measured using Phadia ImmnunoCAP Total Low Range Assay by the Clinical Immunology Laboratory at CMH according to the manufacturer’s prescribed protocol. The detection limit was 0.1-100kU/L, with a specific IgE 0.1-100 calibration curve and specific IgE conjugate for quality control. The calibration curve was assayed every 28 days, after a change of conjugate lot numbers, or as needed. The calibration curve was confirmed daily by the Phadia Curve Controls. In addition, a low and a high control were included in every run. An internal pool control, prepared by the CMH Immunology Laboratory, also was tested daily. All testing was performed on the Phadia ImmunoCAP 250.

Candidate Genes and Single Nucleotide Polymorphism (SNP)

This study focused on 23 well known candidate genes (Table 2) involved in T_H1, T_H2, and T-reg pathways. For each gene, we selected potentially functional SNPs including: 1) non-synonymous coding SNPs; 2) SNPs creating/disrupting a splicing site; 3) SNPs located within human-mouse conserved regions and predicted to be functional variants based on the bioinformatics tool PupaSuite (http://pupasuite.bioinfo.cipf.es/), for example, SNPs in transcription factor binding sites, in exonic splicing enhancers or silencers, in microRNA sequences and/or in a DNA triplex ; and 4) SNPs previously found to be associated with allergic diseases by at least three different studies. We also selected tagSNPs for the genes involved in the T_H2 pathway using a pairwise tagging approach in the Tagger program²³. Specifically, a minimal set of tagging SNPs, by forcing in the above functional SNPs, were chosen based on the available genotyping data in the Yoruba population (HapMap, release 24), such that each unselected common HapMap SNP is in linkage disequilibrium (LD) (r² ≥ 0.80) with the tagging SNPs. A total of 391 SNPs were selected, of which, 329 SNPs with a high Illumina design score (i.e. designability rank=1 and SNP_score≥0.60) were genotyped for all study subjects.

Table 2.

Summary of the 329 genotyped SNPs

Symbol	Chromosome	Gene name	Number of SNPs
			genotyped	droppeda
TH1-skewing pathway
IL2	4q26-q27	Interleukin 2	4	1/1/0/0
TNF	6p21	tumor necrosis factor	7	0/3/0/1
IL12B	5q31.1-q33.1	interleukin 12, beta	3	0/1/0/0
IL18	11q22.2-q22.3	interleukin 18	3	0/0/0/1
IFNG	12q14	interferon, gamma	4	0/2/0/0
TBX21	17q21.32	t-box 21 (or t-bet)	7	0/0/0/1
IL12RB1	19p13.1	interleukin 12 receptor, beta 1	5	1/0/0/0
TH2-skewing pathway
GATA3	10p15	GATA binding protein 3	24	4/0/0/1
IL4	5q31.1	interleukin 4	11	0/1/1/0
IL5	5q31.1	interleukin 5	3	0/0/0/0
IL13	5q31	interleukin 13	10	0/1/0/0
IL4R	16p12.1-p11.2	interleukin 4 receptor	48	1/5/0/8
IL13RA1	Xq24	interleukin 13 receptor, alpha 1	9	1/0/0/1
IL5RA	3p26-p24	interleukin 5 receptor, alpha	36	0/1/0/3
JAK1	1p32.3-p31.3	janus kinase 1	51	5/2/0/3
JAK2	9p24	janus kinase 2	36	9/0/0/0
JAK3	19p13.1	janus kinase 3	14	0/2/1/2
STAT6	12q13	signal transducer and activator of transcription 6	15	2/1/0/0
STAT3	17q21.31	signal transducer and activator of transcription 3	15	5/0/0/0
TSLP	5q22.1	thymic stromal lymphopoietin	11	0/0/0/1
T-Regulatory pathway
FOXP3	Xp11.23	forkhead box P3	2	1/0/0/0
TGFB1	19q13.1	transforming growth factor, beta 1	3	0/1/0/1
IL10	1q31-q32	interleukin 10	8	4/0/0/0

Definition of abbreviations: SNP = single nucleotide polymorphism.

^aSNP dropped due to the high LD with another SNP genotyped / low minor allele frequency (<0.05) / deviation from Hardy-Weinberg disequilibrium (p<0.001) / genotyping failure (call rate<0.98).

Genotyping

SNPs were genotyped using the Illumina GoldenGate custom panel at the genotyping center of Washington University in St. Louis. For quality control, four duplicate DNA samples were placed on each 96-well plate. The concordance rate of these duplicate samples was > 99.5%. Three hundred and six SNPs (93.0%) had a call rate >98.0% and thus were analyzed in the present study. These 306 SNPs are described in Table E1 in the Online Repository.

Ancestry Information

To control for potential confounding due to population stratification, 150 ancestry informative markers (AIMs), with averaged δ (allele frequency difference between two ancestral populations) ≥0.5, were randomly selected from a recently reported genome-wide admixture map²⁴. Of those, 144 AIMs (with a call rate ≥98.0%) were included in the estimation of ancestral proportion for three ancestral populations (Asian, Caucasian and African American) using the STRUCTURE program (version 2.3.1, http://pritch.bsd.uchicago.edu/structure.html). Ancestral proportion was included as a covariate in subsequent analyses.

Statistical Analyses

The primary outcomes of this study were CBIgE concentration (a continuous outcome) and detectable CBIgE (defined as CBIgE ≥ 0.1 kU/L, a binary outcome). CBIgE concentrations were log₁₀-transformed to obtain an approximate normality. For SNPs on the autosomal chromosomes, the Hardy-Weinberg equilibrium (HWE) test in the total population (and in African Americans) was performed using chi-squared statistics. The HWE test for each SNP on the X chromosome was performed in female subjects only, as suggested previously²⁵. SNPs that deviated from HWE (defined as p<0.001) were removed from further analyses. Pairwise LD of SNPs in each gene was calculated using the PLINK program (http://pngu.mgh.harvard.edu/~purcell/plink/).

To test the associations between SNPs and log₁₀-transformed CBIgE concentration, we conducted tobit regression analyses using the “AER” add-on package in R program. This approach allows for modeling a continuous variable in which a large number of observations are censored at a specific value²⁶. In the present study, about one third of the children had undetectable CBIgE (i.e. <0.1 kU/L). All the analyses were adjusted for the important covariates, including maternal age, maternal body mass index (BMI), maternal atopic history, parity, number of prior pregnancies, household income, infant’s gender, season of birth and individual ancestral proportion. Similarly, logistic regression models were applied to explore the effects of each SNP on detectable CBIgE. For each SNP, a codominant model was tested first and then the most parsimonious genetic model (i.e. dominant, recessive, or additive model) was fitted for further analyses. All analyses were conducted using R program (version 2.8.1) and SAS 9.2 software (SAS institute, Cary, NC). The false discovery rate (FDR) method was applied for correcting multiple testing ²⁷.

Two-locus gene-gene interactions were tested for a subset of SNPs that either showed statistically significant associations with CBIgE (nominal p<0.05) or were predicted to be potentially functional SNPs by the bioinformatics tools. We included a product term of a tested SNP pair into the regression models and reported p-values of the Wald test for the gene-gene interaction under both additive and dominant models. We only presented the genetic effect estimates of the combined genotypes based on a dominant genetic model so that each subgroup had sufficient sample size. No multiple testing corrections were performed when testing gene-gene interactions. Instead, we presented gene-gene interaction only if : 1) nominal p<0.001 for the interaction term; 2) the interaction was biologically meaningful, with a predicted protein-protein interaction score of ≥0.90 based on the bioinformatics tool STRING (http://string.embl.de/).

Results

Demographic and Clinical Characteristics

There were 1,070 infants in this study, of whom 58.7% were African American and 21.1% were Hispanic. Detectable plasma CBIgE was present in 739 children (69.1%). Table 1 presents the distribution of plasma CBIgE concentrations by population characteristics. Older maternal age, Caucasian ethnicity and prior pregnancies were associated with decreased CBIgE concentrations, while maternal history of atopy was associated with elevated CBIgE concentration (p<0.05).

Table 1.

Distribution of cord blood IgE concentration by epidemiological characteristics in 1,070 children from the Boston Birth Cohort.

		Cord blood IgE concentration advance
Phenotypes	N (%)	Median (25th-75th)	Detectable rate
Maternal age (years)
<20	94 (8.8)	0.36 (0.13-1.02)	76 (80.9)
20-24	249 (23.3)	0.28 (<0.10-0.74) #	173 (69.5) *
25-29	287 (26.8)	0.25 (<0.10-0.72) *	194 (67.6) *
30-34	242 (22.6)	0.22 (<0.10-0.81) *	163 (67.4) *
≥35	198 (18.5)	0.22 (<0.10-0.58) *	133 (67.2) *
Maternal Pre-pregnancy BMI (kg/m2)
<18.5	37 (3.5)	0.29 (0.12-0.78)	29 (78.4)
18.5-24.9	451 (42.1)	0.26 (<0.10-0.83)	314 (69.6)
25-29.9	358 (33.4)	0.25 (<0.10-0.70)	245 (68.4)
≥30	224 (21.0)	0.22 (<0.10-0.66)	151 (67.4) #
Gestational age (weeks)
<37	239 (22.3)	0.23 (<0.10-0.61)	162 (67.8)
37-39	524 (49.0)	0.26 (<0.10-0.81)	367 (70.0)
≥40	307 (28.7)	0.24 (<0.10-0.78)	210 (68.4)
Gender
Male	559 (52.2)	0.28 (<0.10-0.63)	388 (69.4)
Female	511 (47.8)	0.26 (<0.10-0.91) #	351 (68.7)
Race
African American	628 (58.7)	0.28 (<0.10-0.82)	443 (70.5)
Hispanic	226 (21.1)	0.23 (<0.10-0.67) #	151 (66.8)
Caucasian	66 (6.2)	0.12 (<0.10-0.28) ***	36 (54.6) **
Asian	23 (2.1)	0.34 (<0.10-1.17)	17 (73.9)
Others	127 (11.9)	0.23 (<0.10-0.73)	92 (72.4)
Maternal Smoking during pregnancy
No	976 (91.2)	0.25 (<0.10-0.75)	671 (68.8)
Yes	94 (8.8)	0.25 (<0.10-0.70)	68 (72.3)
Mode of delivery
Vaginal	736 (68.8)	0.23 (<0.10-0.70)	516 (70.1)
Cesarean section	334 (31.2)	0.26 (<0.10-0.75)	223 (66.8)
Parity
None	442 (41.3)	0.28 (<0.10-0.82)	313 (70.8)
≥1	628 (58.7)	0.23 (<0.10-0.70) #	426 (67.8)
Prior pregnancy
None	278 (26.0)	0.30 (<0.10-1.04)	207 (74.5)
≥1	792 (74.0)	0.23 (<0.10-0.69) *	532 (67.2) *
Maternal atopic history
No	721 (67.4)	0.23 (<0.10-0.64)	484 (67.1)
Yes	349 (32.6)	0.32 (<0.10-1.05) **	255 (73.1) *
Household income
<$30K	509 (47.6)	0.22 (<0.10-0.66)	348 (68.4)
≥$30K	332 (31.0)	0.26 (<0.10-0.76)	233 (70.2)
Unknown	229 (21.4)	0.29 (<0.10-0.93) #	158 (69.0)
Season of Birth
Summer	239 (22.3)	0.25 (<0.10-0.74)	161 (67.4)
Fall	271 (25.3)	0.28 (<0.10-0.87)	183 (67.3)
Winter	295 (27.6)	0.22 (<0.10-0.63)	208 (70.5)
Spring	265 (24.8)	0.27 (<0.10-0.83) #	187 (70.6)

***,**,*,# The association of each environmental variable with continuous IgE concentration (log₁₀-transformed) and detectable CBIgE was tested based on the univariate tobit regression model and univariate logistic regression model, respectively.

^***p<0.001;

^**p<0.01,

^*p<0.05,

^#p<0.20

Single SNP Associations

As shown in Table 2, 23 out of 329 genotyped SNPs were excluded due to low call rate (<98%). Of the 306 SNPs eligible for data analysis, we further excluded 57 SNPs either with MAF<0.05 (n=21), deviated from HWE (n=2), or in high LD with others (r² > 0.8) (n=34).

The associations between the 249 SNPs and the two CBIgE outcomes, after adjusting for individual ancestral proportion and the other pertinent covariates, are presented in Figure 1 and Table 3. The most significant SNP associated with log₁₀-transformed CBIgE level was rs1295686 in the IL13 gene, for which, the G allele was associated with decreased CBIgE concentration under a dominant genetic model (p=4×10^-5, p_FDR=0.008). Three other IL13 SNPs (rs2069743, rs1800925, and rs848) and an IL13RA1 SNP (rs5956080) were associated with elevated CBIgE concentration (p≤6×10^-4, p_FDR<0.05). When detectable CBIgE was the outcome, similar associations were detected for the above SNPs, and rs5956080 in the IL13RA1 gene showed an even stronger association (OR=1.84, 95%CI=1.39-2.43, p=2×10^-5, p_FDR=0.008). Additionally, two SNPs, rs12389958 in the IL13RA1 gene and rs11172106 in the STAT6 gene, were significantly associated with an increased risk of detectable CBIgE under an additive genetic model (p≤5×10^-4, p_FDR<0.05).

Figure 1.

SNP associations with log₁₀-transformed cord IgE concentration (A) and detectable cord IgE (B) (249 SNPs on 23 genes). The associations were adjusted by maternal age, maternal BMI, maternal atopic history, prior deliveries, prior pregnancies, infant’s gender, household income, season of birth and individual ancestral proportion.

Table 3.

Associations of T_H1/T_H2 pathway gene polymorphisms with cord blood IgE

Gene a,b,c	SNP d	Allelef	MAF	Log10(CBIgE)		Detectable CBIgE
				ß±SE e	p	OR(95%CI) e	p
IL13 c	rs1800925 d	C/T	0.32	0.26±0.08	6×10-4*	1.37(0.86-2.19)	0.18
IL13 b	rs2069743 d	A/G	0.14	0.18±0.05	2×10-4*	1.54(1.14-2.08)	0.005
IL13 a	rs1295686	A/G	0.43	-0.21±0.05	4×10-5*	0.66(0.49-0.89)	0.007
IL13 b	rs848 d	T/G	0.44	0.12±0.04	5×10-4*	1.19(0.98-1.44)	0.08
IL13RA1 b	rs5956080	T/G	0.27	0.16±0.05	4×10-4*	1.84(1.39-2.43)	2×10-5*
IL13RA1 b	rs12389958	C/A	0.21	0.14±0.05	0.004	1.89(1.39-2.56)	5×10-5*
STAT6 b	rs11172106 d	C/G	0.39	0.10±0.04	0.004	1.44(1.17-1.76)	5×10-4*

Definition of abbreviations: CBIgE = cord blood IgE. SNP = single nucleotide polymorphism; MAF = minor allele frequency; ß = beta coefficient; SE = standard deviation; OR = Odd ratio; CI = confidence interval.

Only SNPs with p≤0.001 are shown.

^aDominant genetic model,

^badditive genetic model or

^crecessive genetic model was applied.

^dFunctional SNP as predicted by bioinformatics tools.

^eAdjusted by maternal age, maternal BMI, maternal atopic history, parity, prior pregnancies, infant’s gender, household income, season of birth and individual ancestral proportion.

^fMajor/minor allele was shown.

^*p<0.05 after FDR correction.

Multiple SNP Associations

Since multiple SNPs in the IL13 and IL13RA1 genes were associated with CBIgE, we examined whether these associations were due to strong LD among these SNPs. We found that the effect of rs848 on CBIgE disappeared when rs1295686 was included in the model, which may reflect the moderate LD between these two SNPs (r²=0.49). Similarly, the association between rs12389958 and detectable CBIgE disappeared when rs5956080 was adjusted in the model, and the LD estimate of these two IL13RA1 SNPs was 0.67. As such, we removed rs848 and rs12389958 from further analyses.

We also investigated the combined effects of SNPs rs1800925, rs2069743 and rs1295686 in the IL13 gene. As shown in Figure 2, individuals carrying more risk genotypes of these three SNPs appeared to have higher CBIgE. This dose-response effect was highly significant (p_trend=9×10^-8) for both log₁₀-transformed CBIgE concentration and for detectable CBIgE (p_trend=9×10^-4).

Figure 2.

Dose-response effects of the combined risk genotypes for three IL13 gene polymorphisms (rs1800925, rs2069743 and rs1295686) on cord blood IgE (A) and detectable cord blood IgE (B). Risk genotype was TT, AG/GG and AA for rs1800925, rs2069743 and rs1295686, respectively.

Pair-wise gene-gene interactions were tested for 105 CBIgE-associated or potentially functional SNPs. We identified two pairs of interaction effects on log₁₀-transformed CBIgE. The first interaction was between JAK2-rs11788963 and IL13RA1-rs2997049 (p_interaction=5×10^-4): among individuals with the rs11788963 CC genotype, the rs2997049 CC or CT genotype was associated with lower CBIgE than the rs2997049 TT genotype, while among individuals with the rs11788963 non-CC genotype, the rs2997049 CC or CT genotype tended to be associated with higher CBIgE (Table 4). The second interaction was between JAK1-rs7538403 and STAT3-rs3744483 (p_interaction= 1×10^-4), which also was significant on detectable CBIgE (p_interaction= 4×10^-4). Two additional interaction effects (i.e. IL13-rs1295686 and IL4R-rs3024547, IL13-rs2069743 and STAT6-rs11172106) were observed on detectable CBIgE (p_interaction≤5×10^-4), for which the expected joint effect was significantly different from the observed one. For example, the expected joint effect of IL13-rs2069743 and STAT6-rs11172106 on the risk of having detectable CBIgE was 1.16 (=1.21×0.96), which was two times lower than the observed joint effect of these two SNPs (OR=3.36, 95%CI=1.98-5.68). Of note, these interaction effects were very consistent for the two outcomes, as presented in Table 4.

Table 4.

Pairwise gene-gene interactions on cord blood IgE

SNP1	SNP2	Log10(CBIgE) a			Detectable CBIgE a
		n	ß ±SE	p	%D	OR(95%CI)	p
JAK2	IL13RA1
rs11788963	rs2997049
AA+AC	TT	375	0.00	--	72.0	1.00	--
AA+AC	CT+CC	44	0.23±0.12	0.05	79.6	1.53(0.70-3.33)	0.28
CC	TT	579	-0.02±0.05	0.77	68.2	0.81(0.61-1.11)	0.17
CC	CT+CC	70	-0.38±0.11	3×10-4	52.9	0.39(0.23-0.68)	7×10-4
pinteraction b				5×10-4/1×10-4			0.05/0.02
JAK1	STAT3
rs7528403	rs3744483
GG	CC+CT	171	0.0		56.7	1.00
GG	TT	134	0.33±0.09	3×10-4	75.4	2.62(1.56-4.41)	3×10-4
GT+TT	CC+CT	448	0.23±0.07	1×10-3	72.5	1.87(1.27-2.74)	1×10-3
GT+TT	TT	315	0.18±0.08	0.02	67.9	1.51(1.02-2.24)	0.04
pinteraction b				1×10-4/4×10-4			4×10-4/1×10-4
IL4R	IL13
rs3024547	rs1295686
CC	AA	200	0.00	--	80.5	1.00	--
CC	AG+GG	447	-0.31±0.07	5×10-6	61.7	0.41(0.27-0.63)	3×10-5
CT+TT	AA	162	-0.10±0.08	0.22	68.5	0.55(0.34-0.91)	0.02
CT+TT	AG+GG	258	-0.17±0.07	0.01	73.3	0.69(0.44-1.09)	0.11
pinteraction b				0.04/0.02			2×10-4/3×10-4
IL13	STAT6
rs2069743	rs11172106
AA	CC	279	0.00	--	62.7	1.00	--
AA	CG+GG	524	0.07±0.06	0.26	68.5	1.21(0.88-1.65)	0.23
AG+GG	CC	112	0.07±0.09	0.41	64.3	0.96(0.59-1.54)	0.88
AG+GG	CG+GG	153	0.35±0.08	5×10-6	85.6	3.36(1.98-5.68)	6×10-6
pinteraction b				0.06/0.05			5×10-4/0.002

Definition of abbreviations: CBIgE = cord blood IgE. SNP = single nucleotide polymorphism; ß = beta coefficient; SE = standard deviation; OR = Odd ratio; CI = confidence interval; %D = percentage of detectable CBIgE.

^aAdjusted by maternal age, maternal BMI, maternal atopic history, prior deliveries, prior pregnancies, infant’s gender, household income, season of birth and individual ancestral proportion.

^bSNP-SNP interaction tests under additive model/dominant model.

Ethnic Heterogeneity

We explored ethnicity-specific associations in African Americans and in Hispanics, separately. The previously associated SNPs in the IL13, IL13RA1 and STAT6 gene showed comparable effects in both ethnic groups (data not shown). Additionally, we found that rs4143832 in the IL5 gene and rs570613 in the GATA3 gene were associated with CBIgE in African Americans but not in Hispanics, indicating ethnic heterogeneity (Table 5). The most significant SNP that was only found in Hispanics was rs2069718 in the IFNG gene, which was not statistically significant after FDR correction (Table 5).

Table 5.

Ethnic-specific associations of the T_H1/T_H2 pathway gene polymorphisms with cord blood IgE

Gene	SNPb	Allelec	Log10(CBIgE)			Detectable CBIgE
			MAF	ß±SE a	p	OR(95%CI) a	p
African Americans (n=628)
IL5	rs4143832	C/A	0.35	0.17±0.05	2×10-4	1.32(1.01-1.74)	0.04
GATA3	rs570613	A/G	0.47	0.16±0.04	5×10-4	1.44(1.12-1.86)	0.005
IFNG	rs2069718	T/C	0.40	0.01±0.05	0.86	1.16(0.89-1.52)	0.26
Hispanics (n=226)
IL5	rs4143832	C/A	0.20	-0.06±0.09	0.53	0.77(0.44-1.37)	0.38
GATA3	rs570613	A/G	0.41	-0.02±0.07	0.82	1.09(0.70-1.69)	0.72
IFNG	rs2069718	T/C	0.42	-0.19±0.08	0.01	0.47(0.30-0.76)	0.002

Definition of abbreviations: CBIgE = cord blood IgE. SNP = single nucleotide polymorphism; MAF = minor allele frequency; ß = beta coefficient; SE = standard deviation; OR = Odd ratio; CI = confidence interval.

^aAdjusted by maternal age, maternal BMI, maternal atopic history, parity, prior pregnancies, infant’s gender, household income, season of birth and individual ancestral proportion.

^bAn additive genetic model was applied.

^cMajor/minor allele was shown.

Discussion

This is the first study to investigate the associations between a comprehensive array of genetic polymorphisms involved in the T_H1/T_H2 pathway and CBIgE concentration in a U.S. inner-city birth cohort. We demonstrated that genetic variants in the T_H2 pathways, especially in the IL13, IL13RA1 and STAT6 genes, were significantly associated with CBIgE concentration individually and jointly, and that there was evidence of ethnic heterogeneity and gene-gene interaction. Our findings provided new insights into early life determinants of IgE and opened new inquiries for future research as follows.

SNP Associations across Studies/Ethnicities

The importance of the cytokine IL-13 and the IL13 genetic variants in the development of allergic diseases, as reviewed by Vercelli²⁸, is well established. However, it remains largely unknown whether IL13 gene regulation of IgE production begins in-utero. To date, only two studies have explored the association between IL13 gene SNPs and CBIgE. One study, in a predominantly Caucasian birth cohort (n=798), identified that rs1295685 was in strong LD with rs1295686 and rs20541 (r²>0.78) and was significantly associated with increased CBIgE (p=0.03), while a marginal association was found for rs1800925 (p=0.07)¹⁶. The other study, in a Chinese population (n=575), reported that rs1800925, rs1295686 and rs20541 were significantly associated with CBIgE in a univariate analysis¹⁷. In a predominantly African American sample, we showed that rs1800925 and rs1295686 were associated with CBIgE. Taken together, the two SNPs (rs1800925 and rs1295686) appear to have common effects on CBIgE across different ethnicities/populations.

Evidence of Additive or Interactive SNP Effect

We found that three IL13 SNPs (rs1800925, rs2069743 and rs1295686) could additively influence CBIgE concentration, and that two of these polymorphisms interact with the genes IL4R and STAT6. The gene-gene interactions between IL13, IL4R and STAT6 polymorphisms, although awaiting validation, are likely to be biologically meaningful given that these three molecules are involved in the same pathway and are known to interact with each other in IgE synthesis. These gene-gene interactions also have been observed in other allergic phenotypes^29-32, although the SNPs previously reported are different from those identified in our study. To our knowledge, this study is the first to identify the effect of gene-gene interactions between IL13, IL4R and STAT6 genes on CBIgE in a predominantly African American sample.

Our data further indicates a gene-gene interaction between IL13RA1 (rs2997049) and JAK2 (rs11788963) SNPs. Of note, each SNP alone showed no significant association with CBIgE and thus could be overlooked if interaction testing was not conducted. More importantly, this gene-gene interaction is biologically plausible because JAK2 tyrosine kinase appears to play an important role in IL-4- and IL-13- induced signal transduction in human fibroblasts³³ and blood monocytes³⁴. Based on STRING (http://string.embl.de/), the predicted protein-protein interaction score between IL13RA1 and JAK2 is high (=0.90). Furthermore, the two interacting SNPs, rs2997049 and rs11788963, are located in DNA triplexes of the IL13RA1 and JAK2 genes, respectively, indicating that both SNPs may function by affecting the triplex formation and disrupting the gene regulation.

Evidence of SNP Functionality

SNP rs1800925 (C-1112T) in the IL13 gene is one of the most studied variants, and has been reported to affect childhood IgE in multiple studies^{29, 35}. A recent functional study reported that the T allele could enhance IL13 promoter activity in primary human CD4+ T_H2 lymphocytes ³⁶, which supports findings by us and others¹⁶ that the rs1800925 TT genotype is associated with elevated CBIgE. Although no published functional studies are available for the other SNPs identified in this study, some of these SNPs are predicted to be functional by bioinformatic tools. For example, according to PupaSuite (http://pupasuite.bioinfo.cipf.es/), rs2069743 in the IL13 gene has potential regulatory functions by changing the binding affinity of some transcription factors, including c-ets-1; According to F-SNP³⁷, rs11172106 in the STAT6 gene may change the binding affinity of the transcription factors CCAAT and GATA-1. The predicted functional significance score (FS) for rs1117206 is 0.55, which is higher than the proposed functional cutoff (FS=0.5). Thus, we speculate that rs2069743 and rs1117206 could, at least in part, causally explain their respective associations with CBIgE.

Areas for Future Studies

Available data suggest that rs2069743 (IL13 gene) and rs11172106 (STAT6 gene) may be the causal SNPs that regulate CBIgE, which make them valuable candidates for further functional validation. It remains unclear how rs1295686 in the IL13 gene may affect CBIgE, since no functional evidence is available for this SNP. It is possible that the relationship between rs1295686 and CBIgE is due to the strong LD between this SNP and one or more functional SNPs which remain to be identified.

Our study indicates that IL13RA1 gene polymorphisms may play an important role in CBIgE concentration. An intronic SNP (rs5956080) in this gene was found to be significantly associated with elevated CBIgE in our study. This SNP, for which, no functional data is currently available, might not be causal in nature but is in strong LD with one or more susceptibility loci in the IL13RA1 gene. According to the HapMap data, three IL13RA1 SNPs (rs2248857, rs2495632 and rs1892299) are in strong LD with rs5956080 (r²>0.80) in the Yoruba population. Among them, rs2248857 and rs2495632 are predicted to be involved in the regulation of IL13RA1 transcription with a predicted FS of 0.50, by using a bioinformatics tool, F-SNP³⁷. However, it is unclear whether one of these SNPs or the combination of these three variants (rs5956080, rs2248857 and rs2495632) is responsible for the observed associations. It is also possible that the association of rs5956080 may be due to a LD with SNPs that are yet to be identified. As such, deep sequencing and functional studies are needed.

In contrast to the convincing findings for IL13 and IL13RA1 SNPs, we found no evidence of associations between IL4 SNPs and CBIgE, including the C-590T SNP, which was previously reported to be associated with CBIgE in 300 Asian children²¹. Some previous studies did find significant associations between IL4 SNPs and total IgE level (after birth) in Caucasians ^{38, 39}. However, few of those SNPs showed significant associations in African Americans and/or Hispanics^31,38. Such evidence may suggest that IL4 SNPs may significantly contribute to IgE concentrations in Caucasians, but not in African Americans or Hispanics. Another explanation is that IL4 SNPs may exert their effects only in the presence of certain environmental factors after birth. This hypothesis needs to be validated.

Strengths and limitations of this study

This study has a large sample size, relatively high coverage of variants in genes of the T_H2 pathway, and accurate/sensitive assays of cord blood IgE. One concern is that CBIgE could be contaminated by maternal IgE. However, this is unlikely for the following reasons. Previous reports, in which cord blood IgA concentration was used as an indicator of contamination^{9, 40}, have shown that such contamination, if present, occurs at a very low rate. Another limitation is that CBIgE may be affected by maternal genotypes and/or the intrauterine environment (e.g. exposure to higher IL-4 and IL-13 concentrations), which could not be controlled in this study. Furthermore, our findings on two-locus gene-gene interaction, which may be affected by multiple testing problems, need further validation. While high-order interactions are possible, these were not tested in this study due to limited statistical power. Finally, allergen-specific IgE in cord blood was not measured in this study. Previous studies have suggested that food allergens and inhalant allergens operate by different mechanisms ⁴¹. Future studies should further explore the genetic determinants of food vs. inhalant allergen-specific IgE in cord blood. Such data will contribute to our understanding of the underlying mechanisms operating food allergens and inhalant allergens, and may have implications for clinical management.

In summary, we demonstrated that genetic regulation of lgE production appears to begin in-utero, with evidence of gene-gene interactions and ethnic heterogeneity. Our study also underscores the important roles of SNPs in the IL13, STAT6 and IL13RA1 genes in predicting cord blood IgE, which may explain 5% of the total variance in CBIgE concentration, as estimated in our study. These findings, if confirmed in future studies, will not only enhance our knowledge of the molecular mechanisms responsible for early regulation of IgE in normal and atopic individuals, but also help us develop new strategies for the early prediction of children at high risk of developing allergic diseases.

Abbreviations

AIMs

ancestry informative markers

BMI

body mass index

CBIgE

cord blood IgE

CMH

Children Memorial Hospital

FOXP3

forkhead box P3

GATA3

GATA binding protein

IFNG

interferon-gamma

IgE

Immunoglobulin E

IL

Interleukin

IL4R

IL4 receptor

IL13RA1

IL13 receptor, alpha1

JAK

Janus kinase

LD

linkage disequilibrium

SNP

Single nucleotide polymorphism

STAT

signal transducer and activator of transcription

TBX21

t-box 21

TGFB

transforming growth factor, beta 1

T_H

T helper

TNF

tumor necrosis factor

T-reg

T regulatory

TSLP

thymic stromal lymphopoietin