To the Editor:
Food allergy, affecting up to 8% of children in the United States,1 is most often mediated by allergen specific immunoglobulin E (IgE) antibodies in the blood. Food allergen sensitization is an intermediate phenotype measured by specific IgE blood tests or skin prick tests (SPTs) and used in the diagnosis of food allergy. Despite family history being a risk factor, no genetic variants have been conclusively identified for food sensitization or clinical food allergy.2 Using the Mexico City Childhood Asthma Study (MCCAS), we examined associations between food sensitization (based on SPTs) and single nucleotide polymorphisms (SNPs) spanning five autosomal candidate genes reviewed by Hong et al. [CD14 (cluster of differentiation 14), IL10 (interleukin 10), IL13 (interleukin 13), SPINK5 (serine peptidase inhibitor, Kazal type 5 isoform), and STAT6 (signal transducer and activator of transcription)].2
We previously conducted a genome-wide association study of asthma in MCCAS among 492 children with physician-diagnosed asthma (aged 5-17) and their parents, who were recruited from a pediatric allergy clinic in Mexico City.3 The children’s clinical evaluation included SPTs to six major food allergens that are common in the Mexican diet (milk, egg, wheat, soy, peanuts, and tree nuts). A SPT was declared positive if the largest diameter of the wheal exceeded 4mm. Testing was considered valid if the reaction to the positive control (histamine) exceeded 6 mm and also exceeded 4 mm above the negative control (glycerin).4 There were 162 trios having an asthmatic child with a positive SPT to at least one food allergen. We examined SNP associations with food sensitization in the five candidate genes in these 162 trios.2 As all probands are asthmatics, it is not possible to adjust for asthma. These associations are generalizable to the asthmatic population, but not necessarily to the general population.
Genotyping was performed on the Illumina HumanHap 550v3 BeadChip, and standard quality control filters were applied.3 For this analysis, we selected SNPs spanning five candidate gene regions (20kb upstream of the 5′ end through 20kb downstream of the 3′ end), resulting in 343 SNPs (16 CD14, 52 IL10, 39 IL13, 28 STAT6, and 208 SPINK5 SNPs). SNPs were directly genotyped or imputed using MaCH (http://www.sph.umich.edu/csg/abecasis/MACH/index.html) with HapMap phase II release 21 reference haplotypes combined from CEU (European Americans), YRI (African Americans), and CHB+JPT (Chinese+Japanese).
PBAT v3.61 (http://www.biostat.harvard.edu/~clange/default.htm) was used to conduct family-based association tests on additive SNP genotype dosage values (estimated reference allele counts with a fractional value ranging from 0 to 2.0). To correct for multiple testing, Bonferroni correction for the total number of SNPs (P<α/M, where M is the number of independent tests) is too conservative, given our observed correlation patterns among SNPs. Therefore, we applied the widely-used Nyholt correction (P<α/Meff), which adjusts for the effective number of independent tests (Meff) based on the correlation matrix of pairwise LD among the SNPs.5,6 This method gave a significance threshold of P<0.0011 A less conservative significance threshold based on the number of candidate genes examined is P<0.01. Association results and LD patterns were plotted together using SNAP (http://www.broadinstitute.org/mpg/snap/).
In the 162 trios with a food allergen sensitized child, no SNP associations were statistically significant at P<0.0011. However, three genotyped SNPs in or near STAT6 were associated with food sensitization at P<0.01 (Table I), two intronic SNPs in the nearby LRP1 (low density lipoprotein receptor-related protein 1) gene and a 3′ untranslated STAT6 SNP. Their minor allele frequencies were lower in the food sensitized children (39.5% for the top SNP rs4759044) compared to the children with no sensitization to the six food allergens (42.2% for rs4759044). The STAT6 and LRP1 SNPs are in moderate LD in CEU (Figure 1, r2=0.34-0.37) and weaker LD in YRI (r2=0.08-0.24) and CHB+JPT (r2=0.12-0.13). Of 28 SNPs spanning the STAT6 region, none were associated with asthma at P<0.05, despite a three-fold greater sample size and thus more statistical power (Table I). There was minimal evidence for SNP associations with food sensitization across CD14, IL10, IL13 or SPINK5 (results not shown).
Another 3′ untranslated STAT6 SNP (rs324015) was previously associated with risk for, and severity of, nut allergy in white atopic children from the United Kingdom,7 but this association was not corroborated for severity of food allergy in Japanese children.8 In our study of Mexican children, rs324015 gave a P value of 0.072, but its adjacent SNP (rs703817) was associated with food sensitization at P=0.0076. The implicated SNPs may tag underlying functional or regulatory variants, and it is not surprising that association patterns vary across populations with dissimilar LD patterns.
LRP1 and its protein product have been largely implicated in neurologic processes. STAT6 SNPs have been associated with food allergy,7,8 asthma,9 total IgE,10 and eosinophilic esophagitis.11 STAT6 plays a central role in mediating IL4 (interleukin 4) and IL13 signals for IgE antibody production.2 Given its biological role, STAT6 polymorphisms are more likely than LRP1 polymorphisms to influence food sensitization. Nonetheless, the association patterns between STAT6 and LRP1 SNPs reflect moderate LD with low recombination across the region (Figure 1). While our findings might not be generalizable to the general population, we provide evidence that STAT6 and LRP1 polymorphisms are associated with food sensitization in asthmatics, who are at particularly high risk for developing sensitization. Follow-up studies in Hispanics and other ethnicities could help to refine these associations with food sensitization and clinical food allergy.