Dust mite exposure modifies the effect of functional IL10 polymorphisms on allergy and asthma exacerbations

Abstract

Background:

The allergenicity of dust mite exposure might be dependent on variants in the gene for IL-10 (IL10).

Objectives:

To evaluate whether dust mite exposure modifies the effect of single nucleotide polymorphisms (SNPs) in IL10 on allergy and asthma exacerbations.

Methods:

We genotyped 6 SNPs in IL10 in 417 Costa Rican children and 503 white children in the Childhood Asthma Management Program (CAMP) with asthma and their parents. We used family-based and population-based approaches to test for interactions between IL10 SNPs and dust mite allergen on serum IgE to dust mite in Costa Rica and on asthma exacerbations in Costa Rica and CAMP.

Results:

Dust mite exposure significantly modified the relation between 3 SNPs in IL10 (rs1800896, rs3024492, and rs3024496) and IgE to dust mite in Costa Rica (P for interaction, .0004 for SNP rs1800896). For each of these SNPs, homozygosity for the minor allele was associated with increased levels of IgE to dust mite with increased dust mite exposure. Homozygosity for the minor allele of each of the 3 SNPs was associated with increased risk of occurrence (~3-fold to 39-fold increase) and frequency of asthma exacerbations among children exposed to ≥10 μg/g dust mite allergen in Costa Rica. Similar results were obtained for 2 of these SNPs (rs1800896 and rs3024496) among white children in CAMP.

Conclusion:

Our findings suggest that dust mite allergen levels modify the effect of IL10 SNPs on allergy and asthma exacerbations and may partly explain conflicting findings in this field. (J Allergy Clin Immunol 2008;122:93–8.)

IL-10 is a pleiotropic immunoregulatory cytokine produced by T-regulatory cells, B cells, monocytes, alveolar macrophages, mast cells, and pulmonary dendritic cells.^1–4 IL-10 plays a central role in the induction of antigen-specific tolerance in human beings ² and protects mice from antigen-induced airway inflam- mation,⁵ a hallmark of asthma.

In human beings, the gene for IL-10 (IL10) is on chromosome 1q31–32, a genomic region linked to asthma and related pheno-types.⁶ Most,^7,8 but not all,⁹ association studies of polymorphisms in the promoter region of IL10 and asthma-related phenotypes have been positive. Although these polymorphisms have been shown to influence levels of IL10 expression, there have been conflicting findings with regard to the direction of these associations.^10,11 A potential reason for these inconsistent results is failure to model gene-by-environment interactions appropriately.¹²

Dust mite allergen is an environmental exposure known to influence allergic sensitization,^13,14 asthma exacerbations,¹⁵ and, importantly, IL-10 production¹⁶ in human beings. We hypothesized that dust mite allergen exposure would modify the relationship between polymorphisms in IL10 and both allergen-specific immune responses and disease exacerbations in children with asthma. To test this hypothesis, we first assessed whether levels of IgE to the house dust mite Dermatophagoides pteronyssinus (Der p 1) and asthma exacerbations were influenced by single nucleotide polymorphisms (SNPs) in IL10 among children in the Genetics of Asthma in Costa Rica Study. We further assessed whether levels of Der p 1 in house dust modified the relation between IL10 SNPs and IgE to dust mite and asthma exacerbations in Costa Rican children. Finally, we attempted to replicate significant findings for asthma exacerbations (because specific IgE to dust mite was not available) in white (non-Hispanic) children with asthma in the Childhood Asthma Management Program (CAMP).

METHODS

Study populations

The protocols for subject recruitment and data collection for the Genetics of Asthma in Costa Rica Study have been previously described in detail and are included along with a detailed Methods section in the Online Repository at www.jacionline.org.¹⁷ The population of the Central Valley of Costa Rica is a genetic isolate^18,19 of mixed Spanish and Amerindian ancestry with a prevalence of asthma that ranks among the highest in the world.²⁰ Children included in the study had asthma (defined as physician-diagnosed asthma and ≥2 respiratory symptoms or asthma attacks in the previous year, and either airway responsiveness [an FEV₁ decline of at least 20% from the best FEV₁ value after inhalation of methacholine ≤16.81 mmol], or bronchodilator responsiveness [an increment of at least 12% and at least 200 mL in baseline FEV1 all loci. In Illumina, 5 samples were repeated on each of15 plates, with no discordant runs. The genotypic pass rate was >99% for all Illumina loci. after administration of albuterol]) and high probability of having ≥6 great-grandparents born in the Central Valley. Of the 439 participating children, 426 had DNA that passed quality control and were included in this analysis along with their parents. Study participants completed a protocol that included questionnaires, house dust collection, and collection of blood samples. Written parental consent was obtained for participating children, from whom written assent was also obtained. The study was approved by the Institutional Review Boards of the Hospital Nacional de Ninos (San Jose, Costa Rica) and Brigham and Women’s Hospital (Boston, Mass).

CAMP was a multicenter clinical trial of the effects of anti-inflammatory medications in children with mild to moderate asthma. Participating children had asthma defined by symptoms greater than 2 times per week, the use of an inhaled bronchodilator at least twice weekly or the use of daily medication for asthma, and increased airway responsiveness to methacholine (PC₂₀ ≤ 12.5 mg/mL).²¹ Of the 1041 children enrolled in the original clinical trial, 968 children and 1518 of their parents contributed DNA samples. This analysis was restricted to 483 families of white children. Questionnaire data was collected at baseline and during the course of the 4-year clinical trial, and blood samples and house dust samples were collected at baseline.²¹ Written informed consent was obtained from parents of participating children. CAMP was approved by the Institutional Review Boards of Brigham and Women’s Hospital and the other participating centers.

Measurement of allergen-specific IgE

In Costa Rican children, serum IgE to Der p 1 was measured by using the UniCAP 250 system (Pharmacia & Upjohn, Kalamazoo, Mich), with samples measured in duplicate. IgE to dust mite was not measured in CAMP.

Statistical analysis

In Costa Rica, level of IgE to Der p 1 was treated as a continuous variable, and asthma exacerbations were defined as having at least 1 hospitalization for asthma in the previous year. For CAMP, we present results for exacerbations at 2 time points. In CAMP, asthma exacerbations were defined as a having at least 1 emergency department visit or hospitalization for asthma. Levels of Der plin house dust were treated as continuous in analyses of specific IgE to Der p 1 and were dichotomized at ≥10 μg/g in models of asthma exacerbations for ease of exposition and because of previous evidence that this level influences the risk of asthma exacerbations.^27,28 Hardy-Weinberg equilibrium was tested in parental data by a χ² goodness-of-fit test, and deviations from mendelian inheritance were tested with PedCheck (http://watson.hgen.pitt.edu/register/docs/pedcheck.html).²⁹ Estimates of measures of linkage disequilibrium (LD) -D’ and r² were obtained from Haploview v3.11 (http://www.broad.mit.edu/mpg/haploview/).³⁰ All analyses were performed assuming an additive genetic model. Family-based association analyses were performed with the family-based association test (FBAT) statistic implemented in HelixTree with PBAT v5 (Golden Helix, Bozeman, Mont).³¹ Gene-by-environment interactions were then tested by using the family-based association tests of interaction (FBATIs) implemented in PBAT.³² SNPs with evidence of interaction (nominal P values <.05) were carried forward to a population-based analysis of association (linear regression) in index children by using SASv9.1.2 (SAS Institute, Cary, NC). Logistic regression was used to examine gene-by-environment interactions on asthma exacerbations. The Fisher method was used to obtain P values for both studies (Costa Rica and CAMP) combined, as follows:

$${\chi }_{2L}^2~-2\ln \left({\displaystyle \prod _{i=1}^L}{p}^i\right)(L=\#P\text{values})$$

RESULTS

Of the 426 participating families in Costa Rica, 9 were excluded from this analysis because of mendelian inconsistencies, leaving 417 children and their parents. Of the 483 families of white children in CAMP, 13 were removed from this analysis because of mendelian inconsistencies, leaving 470 families (and 503 children). Parental genotypes were in Hardy-Weinberg equilibrium for all SNPs in Costa Rica and CAMP (P > .05 in all cases). Although the minor allele frequencies differed, the LD pattern (r²) across IL10 was similar in both Costa Rica and CAMP (Fig 1; see this article’s Table E1 in the Online Repository at www.jacionline.org). Baseline characteristics for children with asthma in Costa Rica and white children with asthma in CAMP are presented in Table I. Exposure to high levels of dust mite allergen (whether ≥2 μg/g or ≥10 μg/g house dust) was more common in Costa Rican children than in white children in CAMP.

FIG 1.

Pairwise (r2) LD for IL10 in Costa Rican and white (non-Hispanic) CAMP parents. *MAF.

Table E1.

Genotypic data for IL10 polymorphisms in children from Costa Rica and CAMP

Chromosome 1 position	Alleles	Location	Costa Rica				CAMP
			H/H*	H/h†	h/h‡	missing	H/H*	H/h†	h/h‡	missing
rs1800896 203335292	A>G	Promoter	219	154	33	11	126	238	117	22
rs1800871 203335029	C>T	Promoter	204	164	39	10	259	168	31	45
rs1800872 203334802	C>A	Promoter	206	165	39	7	274	174	32	23
rs3024492 203332507	A>T	Intron	300	94	12	11	272	195	34	2
rs3024509 203331692	T>C	Intron (boundary)	381	29	0	7	439	59	1	4
rs3024496 203330259	T>C	Exon	214	159	32	12	129	254	119	1

^*Homozygotes for the major allele

^†Heterozygotes.

^‡Homozygotes for the minor allele.

Table I.

Baseline characteristics of children with asthma in Costa Rica and white children with asthma in CAMP

	Median (interquartile range) or count(%)
	Costa Rica	CAMP (white)
Variable	(n = 417)	(n = 503)
Age (y)	8.7 (7.7–10.4)	8.6 (7.0–10.5)
Sex, female	157 (38%)	192 (38%)
Total serum IgE (IU/mL)*	414 (117–962)	402 (157–1068)
Specific IgE to Der p 1	14.1 (0.5–75)	Not measured
Der p 1 ≥2 0μg/g in house dust*	382 (94%)	108 (23%)
Der p 1 ≥10 μs/s in house dust*	203 (44%)	48 (10%)
≥ 1 Asthma exacerbation†	21 (5%)	153 (31%)

^*Numbers and percentages may reflect missing data on some subjects. Information missing on some subjects in Costa Rica measurements of Der p 1 in house dust (n 510) and asthma exacerbations (n = 1). Information missing on some subjects in CAMP for total IgE (n = 7) and measurements of Der p 1 in house dust (n = 27).

^†In Costa Rica, at least 1 hospitalization for asthma in the previous year. In CAMP, at least 1 emergency department visit or hospitalization for asthma during the first 4 years of the trial.

Family-based association analysis of IgE to dust mite in Costa Rica

We found no significant association between SNPs in IL10 and IgE to dust mite among Costa Rican children in models that did not consider the possibility of effect modification (FBAT results in Table II). The results of the family-based analysis of association between SNPs in IL10 and IgE to dust mite in Costa Rican children, assuming a genotype-by-dust-mite allergen interaction, are shown in Table II (FBATI column). We found significant evidence of such an interaction on IgE to dust mite allergen for 3 SNPs in IL10 (rs1800896, rs3024492, and rs3024496). Two of these SNPs (rs1800896 and rs3024496) were in strong LD (r² = 0.94).

Table II.

Family-based analysis of association between IL10 polymorphisms and IgE to Der p 1 and interaction between IL10 polymorphisms and exposure to Der p 1 on serum IgE to Der p 1 among children in Costa Rica

Chromosome 1 position	Alleles	Location	MAF	No.*	FBAT† P value	FBATI‡ P value
rs1800896 203335292	A>G	Promoter	0.27	264	.38	.04
rs1800871 203335029	C>T	Promoter	0.30	266	.64	.75
rs1800872 203334802	C>A	Promoter	0.30	271	.68	.82
rs3024492 203332507	A>T	Intron	0.15	176	.40	.003
rs3024509 203331692	T>C	Intron (boundary)	0.04	74	.54	.17
rs3024496 203330259	T>C	Exon	0.28	269	.35	.02

^*Number of informative families.

^†All models were adjusted for age and sex. Family-based analysis of association between IL10 polymorphisms and measurements of serum IgE to Der p 1 in a model not accounting for IL10 SNP-by-dust mite allergen interaction.

^‡All models were adjusted for age and sex. Family-based analysis of interaction modeling the significance of IL10 SNP-by-Der p 1 allergen interaction in the prediction of specific IgE to Der p 1.

House dust collection

In both Costa Rica and CAMP, a Douglas vacuum cleaner (model 6735, Douglas, Walnut Ridge, Ark) was used to collect a global dust sample. Dust samples collected in Costa Rica were mailed to the Allergy and Immunology Reference Laboratory of Johns Hopkins Hospital, where the dust was weighed, sifted, and aliquoted for measurement of Der p 1 allergen by 2-site mAb ELISA assays.²² In CAMP, dust samples (collected ~6 months after randomization) were analyzed for dust mite allergen by using standardized mAb-based immunoenzymetric assays at a central laboratory.²¹

Genotyping

Markers chosen for genotyping were selected on the basis of previous literature, functional data (rs1800871, rs1800872, and rs1800896),^11,23 validation in dbSNP (Single Nucleotide Polymorphism Database, www.ncbi.nlm.nih.gov/projects/SNP), and a haplotype-tagging algorithm in CAMP based on previous results of gene sequencing.²⁴ The 6 SNPs genotyped capture ≥90% of the HapMap SNPs in IL10 in Centre d’etude du polymorphisme hu-main trios at an r² ≥0.8. In Costa Rica, 5 SNPs were genotyped by using Gold-enGate assays on the Illumina BeadStation 500 system (San Diego, Calif), and 1 SNP (rs1800872) was genotyped with aTaqman genotyping assay (Applied Biosystems, Foster City, Calif). In CAMP, all SNPs were genotyped by using the SnaPshot Multiplex Kit (Applied Biosystems).²⁶ The quality of the genotypic data was assessed by several methods. Duplicate genotyping was performed on approximately 5% of the samples to assess genotype reproducibility. In samples run with SnaPshot or Taqman, no discordant genotypes were detected, with genotype completion rates between 95% and 99% for all loci. In Illumina, 5 samples were repeated on each of 15 plates, with no discordant runs. The genotypic pass rate was >99% for all Illumina loci.

Population-based association analysis of IgE to dust mite in Costa Rica

The results for the analysis of association between SNPs in IL10 and IgE to dust mite are presented in Table III. Similar to the results of our family-based analysis, no IL10 SNP was significantly associated with IgE to dust mite in models not accounting for genotype-by-allergen interactions (model 1). Including such interactions in our models resulted in significant effects for genotypes, Der p 1 levels (except for the model including SNP rs3024492), and genotype-by-dust-mite allergen interactions for all 3 SNPs (model 2). Fig 2 shows the results of our generalized linear models. For each SNP (rs1800896, rs3024492, and rs3024496 in A, B, and C, respectively), children homozygous for the minor allele (shown in red) had higher levels of IgE to dust mite than children homozygous for the major allele (shown in yellow) at high levels of dust mite allergen exposure; heterozygote children (shown in blue) had intermediate effects. In contrast, opposite effects are noted at low levels of dust mite allergen exposure. Bootstrapped regression resulted in smaller CIs for the interaction terms in our models for 2 SNPs (rs1800896 and rs3024496; model 3 inTable III). Models accounting for the possibility of right-censoring in the data for children whose specific IgE to dust mite reached the upper limit of detection (100 kU/L) also show significant results for genotype-by-allergen interaction for all SNPs except rs3024492 (model 4 in Table III).

Table III.

Population-based analysis of interaction between IL10 polymorphisms and exposure to Der p 1 on serum IgE to Der p 1 among children in Costa Rica

Adjusted models		Model 1			Model 2			Model 3	Model 4
		Baseline (no interaction)			Interaction			Bootstrap	Life-table regression
		Estimate	SE	P value	Estimate	SE	P value	P value	P value
A.	rs1800896*	-1.89	3.17	.55	-12.51	4.29	.004	.006	.02
	Der p 1 (μg/g)	0.04	0.11	.72	-0.46	0.17	.008	.004	.02
	Interaction*				0.68	0.19	.0004	.0002	.005
B.	rs3024492*	- 3.76	3.95	.34	-12.44	5.16	.02	.02	.08
	Der p 1 (μg/g)	0.04	0.11	.70	-0.24	0.15	.11	.11	.40
	Interaction*				0.50	0.19	.01	.02	.08
C.	rs3024496*	-1.12	3.21	.73	-13.69	4.38	.002	.001	.002
	Der p 1 (μg/g)	0.03	0.11	.80	-0.57	0.19	.002	<.0001	.003
	Interaction*				0.79	0.19	<.0001	<.0001	<.0001

^*Effect estimates are in reference to increasing content of the minor allele.

FIG 2.

Regression lines by genotype for SNPs rs1800896 (A), rs3024492 (B), and rs3024496 (C) comparing the specific IgE response to Der p 1 antigen by measured Der p 1 antigen level. Yellow, Homozygotes for the major allele; blue, heterozygotes; red, homozygotes for the minor allele.

Analysis of asthma exacerbations

The results of logistic regression models for asthma exacerbations are presented in Table IV. After adjusting for age and sex, there was no significant association between any of the 3 SNPs of interest in IL10 (rs1800896, rs3024492, and rs3024496) and asthma exacerbations in Costa Rica or CAMP (results not shown). On the other hand, there were significant interactions between genotypes in the IL10 SNPs and dust mite allergen exposure on asthma exacerbations (Table IV). Among Costa Rican children who were homozygous for the minor allele of each of the 3 IL10 SNPs, those exposed to high dust mite allergen levels had significantly increased odds of asthma hospitalizations (~3-fold to 39-fold increment). The effect of high levels of dust mite allergen was the opposite for children homozygous for the major allele of each of the 3 SNPs (~81% to 86% decrease), with intermediate effects on heterozygous children.

Table IV.

Analysis of the interaction between IL10 polymorphisms and Der p 1 exposure on the occurrence of asthma exacerbations among children in Costa Rica and CAMP

Logistic regression		Costa Rican children with asthma			White children with asthma in CAMP
		Model 1			Model 2			Model 3			Combined
		Year before enrollment			Year 2 CAMP			Year 4 CAMP
		Estimate	SE	P value	Estimate	SE	P value	Estimate	SE	P value	P value†
A.	rs1800896*	-2.20	1.14	.05	-1.68	0.84	.05	-1.23	0.72	.09	(.02-.03)
	Der p 1 >10 (μg/g)	-2.00	0.78	.01	- 2.47	1.15	.03	-2.08	0.90	.02	(.002-.003)
	Interaction*	1.62	0.76	.03	1.55	0.80	.05	1.19	0.67	.07	(.01-.02)
B	rs3024492*	-4.34	2.14	.04	-0.74	0.86	.39	-0.62	0.71	.40	(.08-.08)
	Der p 1 >10 (μg/g)	-1.68	0.63	.008	-1.14	0.58	.05	-1.20	0.51	.02	(.002-.003)
	Interaction*	2.67	1.19	.03	0.51	0.80	.52	0.48	0.68	.47	(.07-.08)
C.	rs3024496*	-2.09	1.15	.07	-1.26	0.80	.12	-0.95	0.70	.17	(.05-.06)
	Der p 1 >10 (μg/g)	-1.97	0.78	.01	-2.06	0.98	.04	-1.90	0.81	.02	(.002-.003)
	Interaction*	1.52	0.77	.05	1.16	0.75	.12	0.96	0.65	.14	(.04-.04)

^*Effect estimates are in reference to increasing content of the minor allele.

^†The Fisher method has been used to obtain combined P values $[{\chi }_{2L}^2~-2\ln \left({\displaystyle \prod _{i=1}^{L}p}i\right)(L=\#P\text{values})]$ from Costa Rica with those from CAMP at years 2 (presented first in the parenthesis) and 4 of the trial.

Although attenuated, the results for 2 common SNPs in IL10 (rs1800896 and rs3024496, both with minor allele frequency [MAF] of 0.49 in CAMP; Fig 1) similarly modified the effect of Der p 1 allergen on asthma exacerbations among children at years 2(P value for interaction [rs1800896] = .05; [rs3024496] = .12) and 4 (P value for interaction [rs1800896] = .07; [rs3024496] = .14) of the CAMP trial. Among children homozygous for the minor allele of each of the 2 IL10 SNPs, those exposed to high levels of dust mite allergen had increased odds of asthma hospitalizations (~1.3-fold to 1.9-fold increase at 2 years, and a ~1.02- fold to 1.3-fold increase at 4 years). The effect of high levels of dust mite allergen was the opposite for children homozygous for the major allele of each of the 2 SNPs (~87% to 92% decrease at 2 years, and a ~85% to 88% decrease at 4 years), with intermediate effects on heterozygous children. Although consistent in direction to results obtained in Costa Rica, results for SNP rs3024492 (MAF, 0.26 in CAMP) were not significant. Results from the combined analysis of Costa Rica and CAMP (Table IV) demonstrate that for 2 SNPs (rs1800896 and rs3024496), genotype-by-allergen interactions were significant in the prediction asthma exacerbations.

Results similar to those from logistic regression were obtained for rs1800896 in analyses of the counted numbers of asthma exacerbations in Costa Rica and CAMP (see Table E2 in the Online Repository at www.jacionline.org).

Table E2.

Analysis of the interaction between IL10 polymorphisms and Der p 1 among children in Costa Rica and CAMP exposure on the number of asthma exacerbations

	Costa Rica			CAMP
	Negative binomial regression			Poisson regression
	Model 1			Model 2			Model 3
	Year before enrollment			Year 2 CAMP			Year 4 CAMP			Combined
	Estimate	SE	P value	Estimate	SE	P value	Estimate	SE	P value	P values†
A. rs1800896*	-0.37	0.23	.11	-1.55	0.68	.02	-1.03	0.49	.04	(.02-.03)
Der p 1 >10 (μg/g)	-0.11	0.12	.37	- 3.03	1.10	.006	-2.26	0.70	.001	(.003-.02)
Interaction*	0.28	0.15	.06	1.61	0.66	.02	1.07	0.47	.02	(.009-.009)
B rs3024492*	-0.09	0.29	.75	-1.74	0.60	.004	-0.92	0.50	.07	(.02-.21)
Der p 1 >10 (μg/g)	0.03	0.11	.78	-1.68	0.49	.001	-1.45	0.37	<.0001	NA‡
Interaction*	-0.10	0.19	.60	1.18	0.55	.03	0.63	0.47	.18	NA‡
C. rs3024496*	-0.33	0.24	.17	-1.13	0.60	.06	-0.84	0.48	.07	(.06-.07)
Der p 1 >10 (μg/g)	-0.10	0.13	.43	-2.32	0.83	.005	-2.02	0.61	.001	(.004-.02)
Interaction*	0.24	0.15	.18	1.12	0.58	.05	0.86	0.45	.06	(.05-.06)

^*Effect estimates are in reference to increasing content of the minor allele.

^†Fisher’s method has been used to obtain combined P values $[{\chi }_{2L}^2~-2\ln \left({\displaystyle \prod _{i=1}^{L}p}i\right)(L=\#P\text{values})]$. The range presented reflects the result of combining the P values obtained from Costa Rica with CAMP at both 2 and 4 years of the trial

^‡Not applicable. These P values are not combined because the effect estimates are in opposite directions.

DISCUSSION

Among children with asthma in Costa Rica, dust mite allergen exposure significantly modified the effect of polymorphisms in the promoter of IL10 on production of serum IgE to dust mite. Among children in 2 independent populations (Costa Rica and CAMP), dust mite allergen exposure significantly modified the effect of 2 IL10 SNPs (rs1800896 and rs3024496) on asthma exacerbations. To our knowledge, this is the first report of replication of an interaction between a genetic variant and an objectively measured environmental exposure on asthma-related phenotypes in 2 populations.

Specific IgE responses to Der p 1 (dust mite) exposure are critically dependent on baseline IL-10 production and T_H2-predominant immune responses in a time-dependent fashion.^33,34 Baseline elevation of IL-10 can markedly reduce specific-IgE production to Der p 1 in T_H2-stimulated PBMCs.³⁴ In contrast, increased levels of IL-10 after the induction of a specific-IgE response can lead to further increases in specific-IgE.³⁴ Thus IL-10 is a plausible and necessary intermediary between dust mite exposure and dust mite-specific immune responses.

Of note, the most consistently replicated genotype-by-dust mite allergen interaction (rs1800896-by-dust antigen level) in our study involves a well described functional variant in IL10. Although many studies have shown decreased IL-10 production associated with the common A allele compared with the G allele of rs1800896,^35,36 others have demonstrated the opposite effect, and 1 study showed no effect.³⁷ On the basis of these previous studies and our current findings, we speculate that the effect of functional promoter polymorphisms in IL10 on IL-10 transcription by may be complex and linked to the degree of exposure to specific antigenic stimuli (as would be predicted in Fig 1). Assuming that elevated IL-10 levels are associated with decreased specific IgE responses at a given level of exposure,³⁴ we hypothesize that the previously described functional variant in IL10 (rs1800896) is associated with both increased IL-10 expression at low levels of antigenic exposure and decreased IL-10 expression at high levels of antigenic exposure.

Of interest, there was a difference between Costa Rica and CAMP in the magnitude of the modification of the effect of IL10 SNPs on asthma exacerbations by dust mite allergen levels. Possible explanations include regression to the mean, differences in phenotypic ascertainment of asthma exacerbations, time from exposure measurement, differences in the MAF of some of the SNPs of interest (likely because of the relative isolation of the Costa Rican population), and (most importantly) differences in Der p 1 levels in house dust. Although the Costa Rican and CAMP studies ascertained children on similarly strict definitions of asthma^17,21 and used similar dust collection protocols, it is remarkable (given the differences noted) that a significant modification of the effect of IL10 SNPs by Der p 1 allergen was found in both populations.

Although it is generally assumed that level of exposure is directly correlated with sensitization to dust mite allergen, this assertion has been challenged.³⁸ Der p 1 levels have been associated with Der p 1 sensitization in a number of cohorts,^{13,14,39–41} many of which have included subjects recruited on the basis of atopy^14,40 or a strong family history of atopy.^13,39 Other studies have found no association between exposure and sensitization to Der p 1,^29,42,43 and some have demonstrated an inverse association.⁴⁴ In longitudinal cohort studies, Der p 1 exposure has been shown to be a risk factor for Der p 1 sensitization⁴⁰ in atopic children⁴⁵ or in cohorts enriched with children at high risk for atopy,⁴⁰ whereas 1 study collecting longitudinal data on 3 independent population-based cohorts found no association.³⁸ Our findings are consistent with those of cross-sectional and longitudinal studies suggesting that increased exposure leads to increased sensitization to dust mite in children with parental history of asthma and allergies (if those high-risk children had susceptibility variants for allergen sensitization).^39,46 In a study with repeated skin test measurements over a 3-year period, newly diagnosed sensitization to Dermatophagoides pteronyssinus was dependent on a previous history of atopy, exposure to Der p 1 allergen, and parental history of atopy.⁴⁵ Our results suggest that inconsistent findings for dust mite allergen and asthma phenotypes may be a result of untested gene-by-environment interactions.

Our study has several limitations. First, we may have had limited statistical power to detect modest interactions between some IL10 SNPs and dust mite allergen exposure. Second, dust mite allergen levels were measured only once in CAMP. Our finding of a modification of the effect of IL10 SNPs by dust mite allergen levels on asthma exacerbations in CAMP assumes that indoor exposure to dust mite is relatively stable over time and/or that a singular level of exposure can confer longitudinal risk. Although we cannot comment on the latter suggestion, a recent study from The Netherlands suggests that a single home-allergen measurement of Der p 1 adequately represents levels of exposure over a period of 4 years, and we obtained similar results for asthma exacerbations after 2 and 4 years of follow-up (Table IV). Finally, we could not test for an interaction between IL10 SNPs and dust mite allergen exposure on IgE to dust mite in CAMP, and thus our finding for dust mite sensitization in Costa Rica must be interpreted with caution pending replication in other populations.

In summary, our findings suggest that dust mite allergen exposure significantly modifies the effect of polymorphisms in IL10 on sensitization to dust mite and asthma exacerbations in childhood. In addition, our results suggest that failing to account for biologically appropriate gene-by-environment interactions may obscure important genetic effects. Our study highlights an important variable that may help to determine which group with asthma will derive the greatest benefit from reducing or avoiding dust mite allergen exposure.

METHODS

Study populations

The protocols for subject recruitment and data collection for the Genetics of Asthma in Costa Rica Study have been previously described in detail.^E1,E2 The population of the Central Valley of Costa Rica is a genetic isolate^E3,E4 of mixed Spanish and Amerindian ancestry with a prevalence of asthma that ranks among the highest in the world.^E5 Children included in the study had asthma (defined as physician-diagnosed asthma and ≥2 respiratory symptoms or asthma attacks in the previous year, and either airway responsiveness [an FEV1 decline of at least 20% from the best FEV₁ value after inhalation of meth-acholine ≤16.81 mmol], or bronchodilator responsiveness [an increment of at least 12% and at least 200 mL in baseline FEV₁ after administration of albuterol]) and high probability of having ≥6 great-grandparents born in the Central Valley (confirmed by our study genealogist in 416 (95%) of 439 participating children). The latter criterion was required to increase the likelihood that children would be descendants of the founder population of the Central Valley.^E6 Of the 439 participating children, 426 had DNA that passed quality control and were included in this analysis along with their parents. Families identified for removal on the basis of quality control checks include those that consistently demonstrate mendelian inconsistencies (because of either nonpaternity or sample mislabeling) or have an insufficient concentration of DNA available for analysis. Study participants completed a protocol that included questionnaires, house dust collection, and collection of blood samples. Written parental consent was obtained for participating children, from whom written assent was also obtained. The study was approved by the Institutional Review Boards of the Hospital Nacional de Niños (San Jose, Costa Rica) and Brigham and Women’s Hospital (Boston, Mass).

The CAMP was a multicenter clinical trial of the effects of anti-inflammatory medications in children with mild to moderate asthma. All recruited children had asthma defined by symptoms greater than 2 times per week, the use of an inhaled bronchodilator at least twice weekly or the use of daily medication for asthma, and airway responsiveness to methacholine ≤12.5 mg/ mL.^E7,E8 Children with severe asthma or other clinically significant conditions were excluded.^E7 Of the 1041 children enrolled in the original clinical trial, 968 children and 1518 of their parents contributed DNA samples. This analysis was restricted to 483 families of white (non-Hispanic) children. Questionnaire data were collected at baseline and during the course of the 4-year clinical trial, and blood samples and house dust samples were collected approximately 6 months after randomization.^E7 Written informed consent was obtained from parents of participating children. CAMP was approved by the Institutional Review Boards of Brigham and Women’s Hospital and the other participating centers.

Measurement of allergen-specific IgE

In Costa Rican children, serum IgE to Der p 1 was measured by using the UniCAP 250 system (Pharmacia & Upjohn, Kalamazoo, Mich), with samples measured in duplicate. IgE to dust mite was not measured in CAMP.

House dust collection

In both Costa Rica and CAMP, a Douglas vacuum cleaner (model 6735, Douglas, Walnut Ridge, Ark) was used to collect a global dust sample from 5 areas of the child’s household: the upper mattress surface of the child’s bed, an upholstered chair or sofa in the family room or the living room, and floor samples from the child’s bedroom, the family room or living room, and the kitchen. Dust samples collected in Costa Rica were mailed to the Allergy and Immunology Reference Laboratory of Johns Hopkins Hospital, where the dust was weighed, sifted, and aliquoted for measurement of Der p 1 allergen by 2-site mAb ELISA assays.^E9 In CAMP, dust samples were analyzed for dust mite allergen by using standardized mAb-based immunoenzymetric assays at a central laboratory.^E7

Genotyping

Protocol details, SNP flanking sequence, and primer data are available at http://innateimmunity.net. Markers chosen for genotyping were selected on the basis of previous literature, functional data (rs1800871, rs1800872, and rs1800896)^E10,E11 validation in dbSNP (Single Nucleotide Polymorphism Database, www.ncbi.nlm.nih.gov/projects/SNP), and a haplotype-tagging algorithm in CAMP based on previous results of gene sequencing.^E12 We genotyped these 6 SNPs in IL10 in Costa Rican children with asthma and their parents and in families of white children with asthma in CAMP. The 6 SNPs genotyped capture ≥90% of the HapMap SNPs in IL10 in Centre d’etude du polymorphisme humain trios at an r² ≥0.8. In Costa Rica, 5 SNPs were genotyped by using GoldenGate assays on the Illumina BeadStation 500 system (San Diego, Calif),^E13 and 1 SNP (rs1800872) was genotyped with a Taqman genotyping assay (Applied Biosystems, Foster City, Calif). In CAMP, all SNPs were genotyped by using the SnaPshot Multiplex Kit (Applied Biosys-tems).^E14 The quality of the genotypic data was assessed by several methods. Duplicate genotyping was performed on approximately 5% of the samples to assess genotype reproducibility. In samples run with SnaPshot or Taqman, no discordant genotypes were detected, with genotype completion rates between 95% and 99% for all loci. In Illumina, 5 samples were repeated on each of 15 plates, with no discordant runs. The genotypic pass rate was >99% for all Il- lumina loci.

Statistical analysis

In Costa Rica, level of IgE to Der p 1 was treated as a continuous variable, and asthma exacerbations were defined as binary (at least 1 hospitalization for asthma in the previous year) and continuous (number of unscheduled doctor visits for asthma in the previous year) variables. For CAMP, we present results for exacerbations at 2 time points (2 and 4 years) to provide estimates close to the date of dust sample measurement (2 years), and estimates that extend to the reported limit of acceptability for a singular dust sample measurement (up to 4 years; the within-home variance is less than the between-home variance for repeated dust sample measurements).^E15 In CAMP, asthma exacerbations were defined as binary (at least 1 emergency department visit or hospitalization for asthma) and continuous (total number of emergency department visits or hospitalizations for asthma) variables. Levels of Der p 1 in house dust were treated as continuous in analyses of specific IgE to Der p 1 and were dichotomized at 10 μg/g in models of asthma exacerbations for ease of exposition and because of previous evidence that this level influences the risk of asthma exacerbations.^E16–E18

Hardy-Weinberg equilibrium was tested in parental data by a χ² goodness-of-fit test, and deviations from mendelian inheritance were tested with PedCheck (http://watson.hgen.pitt.edu/register/docs/pedcheck.html).^E19 Genotypes of families with mendelian inconsistencies were set to missing. Estimates of measures of LD -D’ and r² were obtained from Haploview v3.11 (http://www.broad.mit.edu/mpg/haploview/).^E20 All analyses were performed assuming an additive genetic model. Family-based association analyses were performed with the FBAT statistic implemented in HelixTree with PBAT v5 (Golden Helix, Bozeman, Mont).^E21 Gene-by-environment interactions were then tested by using the FBATI implemented in PBAT.^E22 The FBATI statistic is generated by using a causal inference regression approach in which the interaction term is not dependent on the main genetic effect. SNPs with evidence of interaction (nominal P values <.05) were carried forward to a population-based analysis of association (linear regression) in index children by using SASv9.1.2 (SAS Institute, Cary, NC). We generated 1000 bootstrapped regression samples to recalculate CIs and P values to assess for violation of parametric assumptions and to deal with the potential influence of outliers. The Proc Lifereg procedure (T distribution) in SAS was used to assess the possible effects of right censoring in our data for serum IgE to dust mite (with an upper limit of detection at 100 kU/L). Logistic, Poisson (for the counted numbers of asthma exacerbations in CAMP), and negative binomial (to address overdispersion in our count data for asthma exacerbations in Costa Rica) regressions were used to examine gene-by-environment interactions on asthma exacerbations. All models used to examine gene-by-environment interactions on the outcomes of interest included the following: age, sex, IL10 genotype, dust levels of Der p 1, and an interaction term for IL10 genotype and Der p 1 level. Model fitness for regression models was assessed as following: F statistics were evaluated for linear regression, Hosmer-Lemeshow tests were used for logistic regression, and deviance/degrees of freedom were evaluated for Poisson regression (and negative binomial regression).

RESULTS

Of the 426 participating families in Costa Rica, 9 were excluded from this analysis because of mendelian inconsistencies, leaving 417 children and their parents. Of the 483 families of white children in CAMP, 13 were removed from this analysis because of Mendelian inconsistencies, leaving 470 families (and 503 children). Parental genotypes were in Hardy-Weinberg equilibrium for all SNPs in Costa Rica and CAMP (P > .05 in all cases). Although the minor allele frequencies differed, the LD pattern (r²) across IL10 was similar in both Costa Rica and CAMP (Fig 1; Table E1). Baseline characteristics for children with asthma in Costa Rica and white children with asthma in CAMP are presented in Table I. Exposure to high levels of dust mite allergen (whether ≥ 2 μg/g or ≥ 10 μg/g house dust) was more common in Costa Rican children than in white children in CAMP.

Family-based association analysis of IgE to dust mite in Costa Rica

We found no significant association between SNPs in 1L10 and IgE to dust mite among Costa Rican children in models that did not consider the possibility of effect modification (FBAT results in Table II). The results of the family-based analysis of association between SNPs in 1L10 and IgE to dust mite in Costa Rican children, assuming a genotype-by-dust-mite allergen interaction, are shown in Table II (FBATI column). We found significant evidence of such an interaction on IgE to dust mite allergen for 3 SNPs in 1L10 (rs1800896, rs3024492, and rs3024496). Two of these SNPs (rs1800896 and rs3024496) were in strong LD (r² = 0.94).

Population-based association analysis of IgE to dust mite in Costa Rica

The results for the analysis of association between SNPs in 1L10 and IgE to dust mite are presented in Table III. Similar to the results of our family-based analysis, no 1L10 SNP was significantly associated with IgE to dust mite in models not accounting for genotype-by-allergen interactions (model 1). Including such interactions in our models resulted in significant effects for genotypes, Der p 1 levels (except for the model including SNP rs3024492), and genotype-by-dust-mite allergen interactions for all 3 SNPs (model 2). Fig 2 shows the results of our generalized linear models. For each SNP (rs1800896, rs3024492, and rs3024496 in A, B, and C, respectively), children homozygous for the minor allele (shown in red) had higher levels of IgE to dust mite than children homozygous for the major allele (shown in yellow) at high levels of dust mite allergen exposure; heterozygote children (shown in blue) had intermediate effects. In contrast, opposite effects are noted at low levels of dust mite allergen exposure. Bootstrapped regression resulted in smaller CIs for the interaction terms in our models for 2 SNPs (rs1800896 and rs3024496; model 3 in Table III). Models accounting for the possibility of right-censoring in the data for children whose specific IgE to dust mite reached the upper limit of detection (100 kU/L) also show significant results for genotype-by-allergen interaction for all SNPs except rs3024492 (model 4 in Table III). P values for F statistics were examined to assess model fitness. Comparable to our findings for individual predictors, models were not adequately fit (P values for model 1, A-C, are all >.4) without the inclusion of interaction terms (P values for model 2, A-C, are all <.05).

Analysis of asthma exacerbations

The results of logistic regression models for asthma exacerbations are presented in Table IV . After adjusting for age and sex, there was no significant association between any of the 3 SNPs of interest in 1L10 (rs1800896, rs3024492, and rs3024496) and asthma exacerbations in Costa Rica or CAMP (results not shown). On the other hand, there were significant interactions between genotypes in the IL10 SNPs and dust mite allergen exposure on asthma exacerbations (Table IV). Among Costa Rican children who were homozygous for the minor allele of each of the 3 1L10 SNPs, those exposed to high dust mite allergen levels had significantly increased odds of asthma hospitalizations (~3-fold to 39-fold increment). The effect of high levels of dust mite allergen was the opposite for children homozygous for the major allele of each of the 3 SNPs (~81% to 86% decrease), with intermediate effects on heterozygous children.

While attenuated, the results for 2 common SNPs in IL10 (rs1800896 and rs3024496, both with MAF of 0.49 in CAMP; Fig 1) similarly modified the effect of Der p 1 allergen on asthma exacerbations among children at years 2 (P value for interaction [rs1800896] = .05; [rs3024496] = .12) and 4 (P value for interaction [rs1800896] = .07; [rs3024496] = .14) of the CAMP trial. Among children homozygous for the minor alleles of each of the 2 1L10 SNPs, those exposed to high levels of dust mite allergen had increased odds of asthma hospitalizations (~1.3-fold to 1.9-fold increase at 2 years, and a ~1.02-fold to 1.3-fold increase at 4 years). The effect of high levels of dust mite allergen was the opposite for children homozygous for the major allele of each of the 2 SNPs (~87% to 92% decrease at 2 years, and a ~85% to 88% decrease at 4 years), with intermediate effects on heterozygous children. Although consistent in direction with results obtained in Costa Rica, results for SNP rs3024492 (MAF, 0.26 in CAMP) were not significant. Consistent with these findings, result from the combined analysis of Costa Rica and CAMP (Table IV) demonstrate that for 2 SNPs (rs1800896 and rs3024496), genotype-by-allergen interactions were significant in the prediction asthma exacerbations. P values for the Hosmer-Lemeshow statistic ranged between 0.4 and 0.8 for all logistic models presented, consistent with adequate model fitness.

Poisson regression models were originally used to analyze counted numbers of asthma exacerbations in Costa Rica and CAMP. Deviance/degrees of freedom were assessed to evaluate the adequacy of fit for our models. Although overdispersion did not appear to be a significant factor in the fit of our Poisson regression models in CAMP (deviance/degrees of freedom ranged between 1 and 2), our results in Costa Rica were strongly suggestive of overdispersion (deviance/degrees of freedom ~4). To account for this overdispersion, we reanalyzed our data for Costa Rica by using negative binomial models (deviance/degrees of freedom in negative binomial models from Costa Rica were all ~1). Results from our Poisson and negative binomial regression models (Table E2) suggest that dust mite allergen exposure modified the effect of 1 SNP in 1L10 (rs1800896) on the number of asthma exacerbations in Costa Rica and in CAMP. Among Costa Rican children who were homozygous for the minor allele of each of these 2 SNPs, those exposed to high levels of dust mite allergen had an increase (75%) in the count of asthma exacerbations; however, this result was of marginal significance (P = .06). Similar results were obtained in CAMP. Among white children in CAMP who were homozygous for the minor allele of SNP rs1800896, those exposed to high levels of dust mite allergen had a significantly increased number of asthma exacerbations in years 2 and 4. Results for SNP rs3024496 were similar to results of SNP rs1800896 in both Costa Rica and CAMP; however, most of these results were of marginal statistical significance. Significant evidence for modification of the effect of SNP rs3024492 on the number of asthma exacerbations by dust mite allergen exposure was demonstrated in CAMP at year 2 but not in CAMP at year 4 or in the Costa Rican cohort. Consistent with these findings and those of logistic regression, the results from the combined analysis of Costa Rica and CAMP (Table E2) demonstrate that SNP rs1800896-by-allergen interactions were significantly predictive of asthma exacerbations. These findings were less compelling for SNPs rs3024492 and rs3024496.

Abbreviations used

CAMP

Childhood Asthma Management Program

FBAT

Family-based association test

FBATI

Family-based test of interaction

LD

Linkage disequilibrium

MAF

Minor allele frequency

SNP

Single nucleotide polymorphism