Genome-wide interaction study of dust mite allergen on lung function in children with asthma

Abstract

Background

Childhood asthma is likely the result of gene-by-environment (G×E) interactions. Dust mite is a known risk factor for asthma morbidity. Yet, there have been no genome-wide G×E studies of dust mite allergen on asthma-related phenotypes.

Objective

To identify genetic variants whose effects on lung function in children with asthma are modified by level of dust mite allergen exposure.

Methods

A genome-wide interaction analysis of dust mite allergen level and lung function was performed in a cohort of Puerto Rican children with asthma (PRGOAL). Replication was attempted in two independent cohorts, the Childhood Asthma Management Program (CAMP) and the Genetics of Asthma in Costa Rica Study.

Results

SNP rs117902240 showed a significant interaction with dust mite allergen level on FEV₁ in PRGOAL (interaction P=3.1×10⁻⁸), and replicated in the same direction in CAMP White children and CAMP Hispanic children (combined interaction P=0.0065 for replication cohorts and 7.4×10⁻⁹ for all cohorts). Rs117902240 was positively associated with FEV₁ in children exposed to low dust mite allergen levels, but negatively associated with FEV₁ in children exposed to high levels. This SNP is on chromosome 8q24, adjacent to a binding site for CEBPβ, a transcription factor that forms part of the IL-17 signaling pathway. None of the SNPs identified for FEV₁/FVC replicated in the independent cohorts.

Conclusions

Dust mite allergen exposure modifies the estimated effect of rs117902240 on FEV₁ in children with asthma. Analysis of existing data suggests this SNP may have transcription factor regulatory functions.

Graphical abstract

INTRODUCTION

Asthma, the most common non-communicable chronic illness among children¹, has an estimated heritability of up to 50–90%^2,3. Although dozens of genes that confer susceptibility to asthma or asthma morbidity have been recently identified⁴, single nucleotide polymorphisms (SNPs) in such genes account for as little as 10–50% of the disease’s heritability. One potential reason for this “missing heritability” is gene-by-environment (G×E) interactions, as certain genetic variants may confer risk only in the presence of certain environmental exposures.

One of the best characterized environmental risk factors for asthma morbidity is exposure to indoor allergens. In particular, dust mite allergen is known to increase the risk of asthma exacerbations, and to increase airway responsiveness while reducing lung function^5–7. The cysteine proteases in dust mites’ fecal particles are potent inducers of allergic sensitization⁸, and exposure to house dust mite can also trigger a direct, non-allergic, inflammatory reaction⁹. Both interventions to reduce dust mite exposure^10,11 and trials of immune therapy to induce dust mite tolerance^12–14 have been shown to improve asthma symptoms or reduce morbidity from asthma.

Few studies have looked at the interaction between genetic variants and allergens on asthma. Gene-by-allergen interactions have been reported for dust mite and several candidate genes for asthma, including IL10¹⁵, TGFB1¹⁶, or the purinergic receptor P2RY12¹⁷. More recently, a study using genome-wide expression in vitro responses to dust mite allergen reported that IL9 may interact with environmental dust mite allergen and increase severe asthma exacerbations¹⁸. In the present study, we performed a genome-wide study interaction study (GWIS) of house dust mite allergen (Dermatophagoides pteronyssinus) level on lung function in Puerto Rican children with asthma, with replication studies in two independent cohorts of children with asthma.

METHODS

Study population

As part of a case-control study of childhood asthma in Puerto Ricans (PRGOAL), we recruited 618 children with asthma (defined as physician-diagnosed asthma and at least one episode of wheeze in the prior year) living in Hartford (CT) and San Juan (PR). The details of subject recruitment have been published previously^19–21.

In brief, from September 2003 to July 2008, flyers were distributed to all parents of children in grades K-8 in 15 public elementary/middle schools in Hartford, CT, that enroll a high proportion (42% to 94%) of Puerto Rican children. Of 640 children whose parents were interested in the study, 585 (91.4%) were eligible after screening; parents of 449 (76.7%) of these 585 children (including 267 children with asthma and 182 control subjects) agreed to participate. There were no significant differences in age, gender, or area of residence between eligible children who did (n = 449) and did not (n = 136) participate. From March 2009 to June 2010, children were recruited from households in metropolitan San Juan, PR, selected using a multistage probability sample design. Primary sampling units (PSUs) were randomly selected neighborhood clusters based on the 2000 U.S. census, and secondary sampling units were randomly selected households within each PSU. A household was eligible if ≥1 resident was a child 6–14 years old. In households with more than one eligible child, one child was selected for screening; in households with multiple eligible children, one child was randomly selected using Kish tables. Of 6,401 contacted households, 1,111 had ≥1 child within the study age range who met other inclusion criteria. In order to reach our target sample size (≥700 children), we attempted to enroll a random sample (n = 783) of these 1,111 children. Parents of 105 (13.4%) of these households refused to participate or could not be reached. There were no significant differences in age, gender, or area of residence between eligible children who did (n = 678, including 351 children with asthma and 327 control subjects) and did not (n = 105) participate. At both study sites, participants had to have four Puerto Rican grandparents and had to have lived in the same household for ≥1 year. Of the 618 children with asthma at both study sites, 520 (287 from San Juan and 233 from Hartford) had blood samples and sufficient DNA for genome-wide genotyping. Of these 520 children, 440 (84.6%, 249 from San Juan and 191 from Hartford) also had data on indoor dust mite allergen level and lung function, and were thus included in the current analysis.

Ethics statement

Written consent was obtained from parents of participating children, from whom written assent was also obtained. The study was approved by the Institutional Review Boards (IRBs) of Connecticut Children’s Medical Center (Hartford, CT [Protocol #135503]), the University of Puerto Rico (San Juan, PR [Protocol #0160507]), Brigham and Women’s Hospital (Boston, MA [Protocol #2007P-001174), and the University of Pittsburgh (Pittsburgh, PA [Protocol #PRO10030498]).

Lung function

Spirometry was performed in participating children using an EasyOne spirometer (NDD Medical Technologies, Andover, MA). Subjects had to be free of respiratory infections for four or more weeks before testing, and they were also instructed (when possible) to avoid use of inhaled short-and long-acting bronchodilators for ≥4 and ≥12 hours before testing, respectively. Expiratory maneuvers were judged acceptable if they met or exceeded American Thoracic Society criteria for children²². The best FEV₁ and FVC from each test were selected for data analysis.

Dust mite allergen measurement

Dust samples were obtained from three areas in the home: the one in which the child slept (usually his/her bedroom), the living room/television room, and the kitchen. The dust was sifted through a 50μ-mesh metal sieve, and the fine dust was reweighed, extracted, and aliquoted for analysis of dust mite allergen (D. pteronyssinus [Der p 1], hereinafter referred to as “dust mite”) using monoclonal antibody Multiplex assays that used the same reagents as the established ELISA^23,24. Dust mite allergen level was categorized as <2 μg/g vs. ≥2 μg/g, because of known effects on allergic sensitization, distribution of allergen levels across the study cohorts (see Table 1), and ease of exposition⁶.

Table 1.

Participant characteristics

	PRGOAL	CAMP Whites	CAMP Hispanics	GACRS
N	440	497	56	549
Age (years)	9.9 (2.6)	8.9 (2.1)	9.7 (2.1)	9.0 (1.8)
Male sex	246 (55.9%)	297 (59.8%)	39 (69.6%)	321 (58.5%)
Race/ethnicity	Puerto Rican	Caucasian	Hispanic	Costa Rican
FEV1 (L)	1.87 (0.65)	1.67 (0.47)	1.85 (0.50)	1.73 (0.49)
FEV1/FVC (%)	81.5 (8.7)	79.4 (8.2)	79.3 (8.5)	82.4 (7.3)
Der p 1 ≥ 2μg/g	213 (48.4%)	108 (21.7%)	16 (28.6%)	504 (91.8%)

Numbers are means (SD) for continuous variables, or N (%) for binary variables. Final numbers included in the analysis may have varied slightly due to different missing rates for each SNP. Dust mite allergen dichotomized using the median level for our cohort (<2 vs ≥2 micrograms of Der p 1 per gram of dust).

Genotyping

Details of genotyping in PRGOAL have been previously described^21,25. In brief, genotyping was performed using the Illumina HumanOmni2.5 BeadChip platform (Illumina, Inc., San Diego, CA), which includes ~2.5 million SNPs. Subjects with a call rate <95% were removed from the analysis. SNPs were removed if they were not in Hardy–Weinberg equilibrium (P<10⁻⁶) in control subjects, had minor allele frequency lower than 1%, or had a failure rate greater than 5%. After all subject and marker quality control steps were completed, we had ~1.83 million SNPs available for data analysis. In addition, SNPs not directly genotyped in the top region(s) of interest were imputed using all populations from the 1,000 Genomes Project (phase I v3)²⁶ as reference for phasing and imputation with SHAPE-IT²⁷ and IMPUTE2²⁸, respectively.

Replication populations

We attempted replication in two cohorts of children with asthma: the Childhood Asthma Management Program (CAMP) and the Genetics of Asthma in Costa Rica Study (GACRS). CAMP was a clinical trial of asthma therapy in 1,041 children. Details on study population, subject recruitment, and study procedures have been previously published^29,30. In this analysis, we included 497 non-Hispanic white children and 56 Hispanic children who had genome-wide genotypic data, Der p 1 level in house dust, and lung function. Genotyping in CAMP was performed using the Illumina HumanHap550v3 BeadChip, and imputation was performed based on the 1,000 Human Genomes Project²⁶ as previously described³¹. The GACRS was a cross-sectional study of 616 children with asthma and their parents. Details on study population, subject recruitment, and study procedures have been previously published³². In this analysis, we included 549 children who had genome-wide genotypic data, Der p 1 level in house dust, and lung function. Genotyping in the GACRS was performed using the Illumina BeadStation 500G platform, and imputation was performed based on the 1,000 Human Genomes Project²⁶, using the same methodology as in CAMP. In both CAMP and the GACRS, Der p 1 level in house dust was measured using the same approach as in the discovery cohort (PRGOAL, see above).

Statistical analysis

Our main outcomes were FEV₁ and FEV₁/FVC. FEV₁ was analyzed as absolute value in liters (adjusted for sex, age, and height); we decided not to utilize percent-predicted values because there are no validated predictive equations for Puerto Ricans or Costa Ricans. For CAMP, we used pre-randomization lung function in order to avoid any potential treatment arm effects. Dust mite allergen level was dichotomized using the median level for our cohort (<2 vs ≥2 micrograms of Der p 1 per gram [μg/g] of dust), a cutoff that has been previously associated with allergic sensitization to dust mite⁶. All multivariate analyses were conducted under the assumption of an additive genetic model, and included the main effect of the SNP, the main effect of dust mite allergen level, an interaction term (genotype*allergen), age, gender, study site (two sites for PRGOAL and eight sites for CAMP), and cohort-specific principal components to account for potential population stratification (eight for PRGOAL, six for CAMP, four for the GACRS); models for FEV₁ also included height and height squared.

Genome-wide statistical significance was set at P <5×10⁻⁸. The top five SNPs from the PRGOAL analysis for FEV₁ and FEV₁/FVC were carried forward for replication in CAMP and the GACRS. For each cohort, replication was defined as a one-tailed P-value <0.05, in the same direction of interaction as in the discovery cohort. Fisher combined p-values (sum of logs method) were obtained for all cohorts and for the replication cohorts only. All genome-wide interaction analyses for dust mite allergen were performed using PLINK v1.07³³. All other analyses were performed in R (version 3.1.3), using R-Studio (version 0.98.1103).

RESULTS

Table 1 summarizes the baseline characteristics of all cohorts. PRGOAL and CAMP White participants were slightly older than those in CAMP Hispanics and the GACRS. There was a higher proportion of boys in CAMP Hispanics (~69%) than in the other study groups (~56–60%). All children had similar lung function measures after adjusting for sex and age. Compared to PRGOAL, CAMP had a lower proportion of households with dust mite allergen level >2 μg/g, whereas the GACRS had a higher proportion of households with dust mite allergen >2 μg/g than the other cohorts.

FEV₁

Interaction P-values for the GWIS of dust mite allergen in PRGOAL are shown in the Manhattan plot in Figure 1. The top five SNPs for FEV₁ are shown in Table 2. SNP rs117902240, located on chromosome 8q24.13, had a significant interaction with dust mite allergen level (P for interaction= 3.1×10⁻⁸), while the P-values for the other four SNPs ranged from 1.3×10⁻⁷ to 5.4×10⁻⁷. Minor allele frequencies for the top five SNPs ranged from 1.3% to 9%.

Figure 1. Manhattan plot of GWIS interaction p-values in PRGOAL for FEV₁.

Plot depicts the –log₁₀(interaction P-value) for FEV₁ on the y-axis vs SNPs in each chromosome on the x-axis. Red line: genome-wide significance level (P<5×10⁻⁸). Blue line: genome-wide suggestive level (P<1×10⁻⁷).

Table 2.

GWIS results for dust mite allergen level and FEV₁ in PRGOAL

SNP	CHR	BP	A1	A2	BETA	P-VAL	MAF	n(A1)
rs117902240	8	122930641	C	A	−1.156	3.114E-08	0.015	4/7
rs1404568	3	143210189	G	A	0.391	1.294E-07	0.09	34/29
rs72716237	9	26458550	A	G	0.935	4.228E-07	0.013	8/5
rs7541272	1	108211005	G	A	0.476	4.447E-07	0.06	29/22
rs35312258	3	143193111	G	A	0.371	5.358E-07	0.09	33/31

Models included the main effects of genotype (additive) and dust mite allergen level (dichotomized <2 vs ≥2 μg/g) and the interaction term (genotype*dust mite allergen level), and were additionally adjusted for age, sex, study site (San Juan, PR, or Hartford, CT), height, and height², and principal components. SNP=single nucleotide polymorphism; CHR=chromosome; BP=base position; A1=minor allele; A2=major allele; BETA=beta coefficient for interaction term; P-VAL=interaction p-value; MAF=minor allele frequency; n(A1)=number of subjects with at least one copy of the minor allele and low/high dust mite exposure.

Table 3 shows the results of the replication efforts for FEV₁. Of the five SNPs carried forward for replication, SNP rs117902240 was associated with FEV₁ among non-Hispanic whites (P value=0.05) and Hispanics (P=0.01) in CAMP, but not in the GACRS. The combined P-value for all three replication cohorts (in CAMP and the GACRS) was 0.0048, and the combined P-value for interaction with dust mite in all four cohorts was 9.75×10⁻⁹. Given differences in sample sizes and effect estimates among cohorts, we also combined P-values using a weighted approach: when weighted by the square root of the sample sizes, the combined P-value was 1.1×10⁻⁴; when weighted by the effect sizes, the combined P-value was 3.0×10⁻¹¹. Of the remaining four SNPs, three were associated with FEV₁ in the GACRS (rs1404568 and rs35312258 in chr.3, and rs72716237 in chr.9), but the direction of the estimated interactions with dust mite allergen was opposite that found in PRGOAL.

Table 3.

GWIS replication results for FEV₁ in CAMP and the GACRS

	CAMP Whites					CAMP Hispanics					GACRS					COMBINED P-VAL
SNP	A1	BETA	P-VAL	MAF	n	A1	BETA	P-VAL	MAF	n	A1	BETA	P-VAL	MAF	n	All cohorts	Replication only
rs117902240	C	−0.43	0.051	0.01	8/1	C	−0.61	0.011	0.03	2/1	C	0.26	0.15	0.01	2/5	7.4E-09	0.0065
rs 1404568	G	−0.0078	0.45	0.11	76/23	G	0.048	0.32	0.17	13/3	G	−0.12	0.048	0.18	20/167	1.6E-06	0.13
rs72716237	A	0.067	0.33	0.02	22/3	A	NA	NA	NA	NA	A	−0.38	0.046	0.01	2/14	1.3E-06	0.079
rs7541272	G	−0.105	0.25	0.02	16/3	G	NA	NA	NA	NA	G	0.24	0.21	0.02	1/20	4.1E-06	0.21
rs35312258	G	0.031	0.32	0.09	63/19	G	0.048	0.33	0.17	13/3	G	−0.12	0.046	0.17	19/159	3.8E-06	0.098

Models included the main effects of genotype (additive) and dust mite allergen level (dichotomized <2 vs ≥2 μg/g) and the interaction term (genotype*dust mite allergen level), and were additionally adjusted for age, sex, study site, height, height2, and principal components. SNP=single nucleotide polymorphism; A1=minor allele; BETA=beta coefficient for interaction term; P-VAL=interaction p-value; MAF=minor allele frequency; n=number of subjects with at least one copy of the minor allele and low/high dust mite exposure; NA=not available (genotyped or imputed).

Using imputed SNP data based on the 1,000 Human Genomes Project, we performed fine mapping of the region surrounding the top SNP (rs117902240): several SNPs showed R² ~0.4 with this SNP; the top interaction P value for these imputed SNPs was 2.6×10⁻⁶ (rs186935901 and rs10108856), with others ranging from 1.3×10⁻⁴ to 9.5×10⁻⁴ (Figure 2). Figure S1 (online supplementary data) shows the LD plots for the other SNPs: seven SNPs showed R² ~0.75 with rs72716237. SNPs rs13078996 and rs1404568 were in high LD with rs35312258; their interaction P values were 6.01×10⁻⁶ and 1.29×10⁻⁷, respectively.

Figure 2. Regional association plot for SNP rs117902240.

Regional association plot generated using the LocusZoom tool from the Abecasis lab at the University of Michigan (http://locuszoom.sph.umich.edu/locuszoom/)⁵² with data imputed from the 1,000 Genomes populations (phase I v3).

We then repeated the analysis for rs117902240 in PRGOAL, before and after stratification by dust mite allergen level. When evaluated in PRGOAL as a whole, rs117902240 was not significantly associated with FEV₁ (β=0.07 [SD=0.10], p-value=0.48). However, after stratifying by dust mite allergen level, SNP rs117902240 was positively associated with FEV₁ (β=+0.84 [0.15], p=1.2×10⁻⁷) in children exposed to low dust mite allergen levels, but negatively associated with FEV₁ in children exposed to high dust mite allergen levels (β=−0.41 [0.13], p=0.003). Figure 3 shows the boxplot for FEV₁, stratified by dust mite level and rs117902240 genotype. Table S1 (online supplementary data) shows further details for the stratified analysis of rs117902240 and the other top SNPs.

Figure 3. FEV₁ stratified by rs117902240 genotype and dust mite allergen level.

FEV₁ in PRGOAL stratified by dust mite allergen level (low in blue vs. high in red) and rs117902240 genotype (AA=no minor alleles vs. AC=one minor allele; there were no subjects homozygous for the minor allele [CC]). Model adjusted for age, sex, and height.

In an exploratory analysis, we assessed the biological plausibility of SNP rs117902240 as an allelic variant interacting with dust mite allergen. Using data from ENCODE³⁴ and is-rSNP³⁵, a software designed to predict whether a SNP has a regulatory effect on transcription factor binding sites in silico, we found that SNP rs117902240 is <1 kb away from a transcription factor binding island identified in ENCODE using ChIP-seq. Transcription factor CEBPβ (CCAAT/Enhancer Binding Protein Beta) binds in this region in human lung fibroblasts (ENCODE accession ENCSR000EFM). CEBPβ forms part of the signaling pathway of interleukin (IL)-17, IL-6, nuclear factor Kappa-B (NF-κβ), and interferon (IFN-)γ (see Figure 4)^36,37. Details for the other top SNPs may be found in the online supplement.

Figure 4. CEBPβ in the IL-17 signaling pathway.

IL-17 pathway (simplified). CEBPβ = CCAAT Enhancer Binding Protein Beta. CEBPβ is one of the mediators of IL-17A and IL-17F signaling pathways, as well as NF-κβ signaling^36,37. CEBPβ is also involved in IL-4, IL-6, and other signaling pathways (not shown).

Finally, we explored SNPs within previously reported genes (IL-9, IL-10, TGF-β1, P2RY2) for nominal association (one-tailed P-value<0.05) with FEV₁. Several SNPs within those genes showed nominal associations, either for the main effect of genotype or for the genotype*dust mite interaction (see Table S2 in the online supplementary data).

FEV₁/FVC

The top 5 SNPs for FEV₁/FVC are shown in Table S3 (online supplementary data) and the replication results are shown in Table S4. One SNP (rs115997623) was associated in the GACRS and one (rs78930606) in CAMP Whites, but both in opposite directions as in PRGOAL. None of the top SNPs for FEV₁/FVC had significant combined P-values for the replication cohorts.

DISCUSSION

In this study, we report that dust mite allergen level significantly modifies the estimated effect of one SNP (rs117902240) on FEV₁ in children with asthma. To our knowledge, this is the first study to examine a gene-by-allergen interaction on an asthma-related phenotype (in this case, lung function) using a genome-wide approach.

Dust mite exposure and sensitization have been widely reported as risk factors for asthma symptoms and exacerbations in affected individuals^5,6,38,39. Previous G×E studies of candidate genes have reported that dust mite exposure modifies the effects of IL-9 and IL-10 polymorphisms on asthma exacerbations^15,18, TGF-β1 variants on asthma severity¹⁶, or purinergic receptor gene P2RY12 haplotypes on lung function¹⁷. Our results extend current knowledge on the topic by identifying novel polymorphisms using a genome-wide approach.

Using data from ENCODE³⁴ and is-rSNP³⁵, we found that rs117902240 is closely adjacent to a transcription factor island that binds CEBPβ, which forms part of the IL-17 signaling pathway (Figure 4). IL-17A or IL-17F can stimulate the NF-κβ pathway, which in turn acts with CEBPβ to activate target gene transcription in the nucleus³⁷. IL-17 has been amply implicated in the pathogenesis of asthma^40–42, and studies in dust-mite sensitized subjects have shown that specific immunotherapy can lead to reductions in IL-17 levels that correlate with symptom reduction⁴³. However, CEBPβ also participates in several other pathways, including IL-4, IL-6, TNFα and IFN-γ signaling pathways, and it has been implicated in airway epithelium differentiation during lung development⁴⁴. Thus, functional studies will be needed to determine the precise role of CEBPβ in the interaction between dust mite and genotype on lung function in asthma.

Children with more severe or poorly controlled asthma have worse lung function⁴⁵. Moreover, poor lung function in children with asthma predicts exacerbations over subsequent follow-up⁴⁶, and is associated with worse outcomes in adulthood⁴⁷. Several studies (including GWAS and candidate-gene studies) have been performed to identify genetic variants associated with lung function in subjects with asthma^48–50, regardless of allergen exposure. Of note, however, the top SNP reported here would have been missed in a “regular” GWAS, but had significant (and opposite) effects when stratified by level of dust mite allergen exposure. This has potential implications, both in terms of improving our understanding of the underlying mechanisms by which dust mite impacts lung function, but also in terms of developing new approaches to identify susceptibility variants for asthma-related phenotypes.

Our study has several strengths, including the overall sample size of ~1,550 children with asthma and data on dust mite allergen and lung function, and the biological plausibility of the top SNP based on ENCODE data. It also has several limitations: some of the SNPs examined had low minor allele frequencies and thus we had small sample size within some of the genotype*allergen level strata, which may have led to low power to detect modest interactions. Post hoc analysis based on results from SNP rs117902240 shows we had sufficient statistical power (see Table S5 in the online supplementary data) and the SNP*dust mite exposure level interaction replicated in the same direction in three out of four populations; nonetheless, results based on small strata must be interpreted with caution and should be further validated in future, prospective studies. In addition, we used statistical methods to adjust for population stratification but these may not be sufficient and some residual stratification may remain unaccounted for. Moreover, the choice of method for combining P-values had an important impact on the results; different approaches yielded combined P-values ranging from 1.1×10⁻⁴ (weighted by sample size), to 7.4×10⁻⁹ (Fisher’s unweighted method), to 3.0×10⁻¹¹ (weighted by effect size). Finally, the timing and duration of dust mite exposure may be important factors in the pathogenesis of asthma, in addition to the level of the allergen⁵¹.

In summary, we show that dust mite allergen exposure modifies the estimated effect of a SNP on a key measure of lung function (FEV₁) in children with asthma. Analysis of existing data suggests possible transcription factor regulatory functions for this SNP; future studies should focus on the functional validation of these pathways.

Capsule Summary.

Dust mite allergen exposure modifies the effect of certain genetic polymorphisms on lung function in children with asthma.

Clinical Implications.

The effect of certain genetic polymorphisms on lung function in children with asthma varies depending on their dust mite allergen exposure level. Future studies should assess whether personalized interventions for these children have a greater impact on lung function.

Abbreviations

CEBPβ

CCAAT/Enhancer binding protein beta

ChIP-seq

Chromatin immunoprecipitation and DNA sequencing

DNA

Deoxyribonucleic acid

ERK3

Extracellular signal-regulated kinase 3

FEV₁

Forced expiratory volume in the first second

FVC

Forced vital capacity

GSK3β

Glycogen synthase kinase 3 beta

GWAS

Genome-wide association study

GWIS

Genome-wide interaction study

G×E

Gene-by-environment interaction

IL

Interleukin

IRB

Institutional review board

P2RY12

Purinergic receptor P2Y, G-protein coupled, 12

PSU

Primary sampling unit

SNP

Single nucleotide polymorphism

TAK1

Transforming growth factor beta 1-activated kinase 1

TGFβ1

Transforming growth factor beta 1

TRAF6

Tumor necrosis factor receptor-associated factor 6