EoE is a global health disease now reported on all continents except Africa with an incidence of ~1:10,0001
. Patients with EoE commonly report symptoms that include difficulty feeding, failure to thrive, vomiting, epigastric or chest pain, dysphagia and food impaction.EoE patients are predominantly young males with a high rate of atopic disease and the diagnosis is made by endoscopy and biopsy finding of isolated eosinophils in the esophagus1
EoE is considered a food allergy-related disorder based on several findings. The majority of EoE patients are atopic individuals with a high rate of food allergen sensitization, based on skin prick and patch testing; in addition, patients have a higher rate of food anaphylaxis compared with the general population1
. Moreover, nearly all EoE patients (including those not identified as atopic) have complete remission following introduction of an elemental formula diet that removes all allergens from the diet; conversely, the disease flares upon re-introduction of specific foods. Unlike classic anaphylaxis that typically involves a limited set of foods, EoE patients are often sensitized to a myriad of foods, often including food groups not typically associated with eliciting anaphylaxis2
. Experimental modeling in mice have demonstrated a key role for adaptive immunity, Th2 cell cytokines (especially IL-5 and IL-13) and a strong connection between allergic sensitization and inflammation in the respiratory tract and skin3
Accumulating evidence suggests that EoE has a strong familial association4
. Nearly 10% of parents of EoE patients have a history of esophageal strictures and ~8% have biopsy proven EoE4
and at least 27 multiplex families have been described to date5
. EoE also exhibits a high sibling risk ratio (λS
) of ~806
compared with related atopic diseases such as asthma (λS
. While genetics is likely to have a large role in EoE susceptibility, there has been only one candidate gene identified to date, the eotaxin-3 gene, with rs2302009 in the 3′ untranslated region conferring risk. Eotaxin-3 is the most overexpressed gene in the esophagus, based on genome wide expression profile analysis. However, the disease-associated allele is only present in 14% of EoE patients7
, suggesting that additional variants are to be found.
Here we report the results of an ongoing EoE genome wide association study (GWAS) aimed at identifying disease susceptibility loci. We genotyped approximately 550,000 SNPs on the Illumina Human Hap550 BeadChip in a discovery cohort of 181 EoE cases from Cincinnati Children’s Hospital Medical Center (CCHMC) and 1974 controls from the Children’s Hospital of Philadelphia (CHOP), and a replication cohort of 170 EoE samples and 1130 controls from CHOP. The study was approved by the institutional review boards of CHOP and CCHMC. Parental informed consent was obtained from all participants in this study for the purpose of DNA collection and genotyping.
The EoE discovery cohort consisted of 181 clinically confirmed EE patients of European ancestry recruited at CCHMC. The subjects in the discovery cohort had an eosinophils/hpf (400X) count of ≥24. The mean age of these cases was 11.3 ± 10.4 (SD) years. The replication cohort consisted of 170 cases recruited at CHOP. Cases were defined as those having an eosinophils/hpf(400X) count of ≥24 and on PPI therapy for at least 8 weeks. The mean age of the replication cohort cases was 7.8 ± 4.9 SD years. Most of the EE subjects in both the CCHMC and the CHOP cohorts were male making up 70% in the discovery cohort (CCHMC) and 75% in the replication cohort (CHOP). Moreover, 73% of the discovery cohort and 72% of the replication cohort had asthma, allergic rhinitis or atopic dermatitis.
The controlsamples used in both the discovery and replication phase were recruited within the CHOP Health Care Network by CHOP clinicians and nursing staff. They were all of self-reported Caucasian ancestry and screened negative for a diagnosis of asthma,dermatitis or other atopic conditions, as well as for other chronic medical conditions based on questionnaire responses. The mean age of the control group (n=2,096) used for the discovery phase was 8.54 ± 5.65 SD years and consisted of an equal percentage of males and females (50.4% and 50.6%, respectively).There was no overlap between patients or controls in the discovery and replication sets.
Both cases and controls were chosen based on their self-reported ancestry but to avoid any population stratification we screened all cases and controls at ancestry informative markers (AIMs) using the STRUCTURE software package(see Supplementary methods
). Cases were subsequently ‘genetically matched’ to controls by principal component analysis (see Supplementary methods
). Quality control measures prior to association analysis resulted in the rejection of 16,942 SNPs with call rates <95%, 21,901 SNPs with a minor allele frequency (MAF) <1% and 14,944 SNPs with Hardy Weinberg equilibrium P
for the CCHMC analysis; the genomic inflation factor (GIF) for that study was 1.08. For the CHOP analysis we rejected 7,481 SNPs with call rates <95%, 18,755 SNPs with a MAF <1% and 1,860 SNPs with Hardy Weinberg equilibrium P
; the GIF was 1.05.
Single marker analyses for genome-wide data were carried out using the Cochran-Armitage trend test as implemented in plink (http://pngu.mgh.harvard.edu/~purcell/plink/
). In the discovery cohort, two SNPs remained significantly associated with EoE following multiple testing correction ().One of the two SNPs, rs3806932 (uncorrected P
-value = 8.81×10−8
, MAF of 31.2% in cases and 45.8% in controls, OR=0.53, [95% CI 0.42 – 0.67]), maps to an LD bock on 5q22.1 that containseleven other SNPs that were also associated with EE (P-
value range = 4.2×10−3
; OR range = 0.55 – 1.55) (). The associated LD block spans the TSLP/WDR36 gene region (pairwise r2
values are presented in Figure 2 and averaged recombination rates from the CEU HapMap trios in Figure 3).
Figure 1 Regional plot of the 5q22 locus and follow-up expression studies. (a) Regional plot showing the associated SNPs in the TSLP-WDR36 region. The relative position of each SNP is plotted against −log10 P values and the recombination rate in the region. (more ...)
Results for the most strongly associated SNPs at 5q22
We next sought to replicate the findings in the CHOP EoE cohort.While none of the SNPs surpassed genome wide significance (Supplementary Figure 1
), of the 12 associated SNPs at the 5q22.1 locus in the discovery cohort,4 were also associated with EoE in the replication cohort (P
-value range = 0.05 – 4.8×10−3
; OR range = 0.70 – 1.3) (). Combining the P
-values across the two cohorts using Fisher’s method, three SNPs surpassed genome wide significance (P
-value range =2.37×10−8
). The most significant SNP in the discovery cohort rs3806932, was also significantly associated in the replication cohort (P
; MAFs 37.7% in cases and 45.2% in controls; OR 0.73; combined P
-value for rs3806932 across CCHMC and CHOP cohorts =3.19×10−9
In addition to the 5q22.1 locus, one SNP on 13q31.1, rs1873288, reached genome wide significance in the discovery cohort (P
-value = 6.72×10−8
) but did not replicate in the CHOP cohort. We subsequently performed a genome wide meta-analysis of the discovery and replication cohorts (Supplementary Figure 2
). Apart from the 5q22.1 locus, one SNP on chromosome 16 (rs371915P
) also reached genome wide significance, however, this association remains to be replicated. A summary of all the meta-analysis SNPs with P
-values < 1×10−5
is presented in Supplementary Table 1
, it is notable that the list includes SNPs that map to genes that have previously been implicated in immune function such as rs167769 (STAT6
) andrs7236477 (DSG1
Finally, as the proportion of EoE patients with asthma is high, ~40% CCHMC and ~20% CHOP, we carried out a logistic regression including asthma status as a covariate in the analysis of both cohorts to ensure that the result was not being driven by asthma rather than EoE. We observed continued association at the 5q22 locus (Supplementary Table 1
) with no significant heterogeneity in the odds ratios after adjusting for asthma status, which would indicate that the association is independent of the asthma phenotype.
We subsequently examined the expression levels of both TSLP
mRNA in esophageal biospsiesfrom EoE patients and controls.TSLP
was significantly overexpressed in the EoE patients compared to controls following normalization with GAPDH
, while WDR36
was unaltered (Figure 4).We subsequently investigated the expression of the two TSLP
isoforms (Figure 2)in the esophageal biopsies. TSLP
isoform 2 was significantly overexpressed in EoE cases compared to controls (P
-value <0.0001); while a similar trend towards overexppression was observed for isoform 1, the results were not significant (Supplementary Figure 3
). Finally, TSLP
expression levels where significantly correlated with rs3806932 genotype in the EoE cases, with homozygote carriers of the protective minor allele showing less TSLP
expression than heterozygotes and homozygotes for the ancestral allele (Supplementary Figure 4
). Taken together these results indicate that TSLP
is the stronger candidate for EoE susceptibilityat the 5q22 locus.
We have identified and replicated a genome-wide significant locus at 5q22.1 in EoE patients of European ancestry; two genes map to this locus,TSLP
and WDR36. WDR36
encodes a T-cell activation protein with a minimum of eight WD40 repeats that is highly co-regulated with the T-cell growth factor IL-2, and has also been linked with open angle glaucoma8
. Several lines of evidence support arole for T cells in the pathogenesis of EoE. Increased CD8 cells and CD25+FoxP3 cells have been observed in patients with EoE9
. Furthermore, Zhu and colleagues report an imbalance of effector and T regulatory cells with reduced IL-2 levels in allergen sensitized EoE mice10
encodes an IL-7–like cytokine that is expressed within the thymus and peripheral tissues and that regulates dendritic cell–mediated inflammatory Th2 responses11
. As a key initiator of allergic inflammatory diseases, TSLP
is overproduced in lesional skin in atopic dermatitis and the asthmatic lung. In mouse models of allergic airway inflammation, enhanced TSLP
production from skin keratinocytes correlates with increased Th2-attracting chemokine production (including eotaxin-1 and -2) and enhanced disease severity and epicutaneous sensitization12
. The higher expression of TSLP
observed in EoE suggests a compelling model with TSLP
acting as a driving force for the allergic phenotype. Indeed, patients with EoE are highly atopic and highly sensitized to multiple food allergens and TSLP plays a critical role in initiating allergen sensitization11
, a process that is profoundly abnormal in EoE. In addition, IL-13 has been shown to co-induce TSLP in epithelial cells, and we have previously linked IL-13-driven epithelial responses in EoE pathogenesis13
. We hypothesize that increased TSLP
expression as observed in the EoE cohort leads to increased sensitization and development of EoE. It is noteworthy that the TSLP
receptor (cytokine receptor-like factor 2, CRLF2) resideswithin the pseudoautosomal region 1 on the X- and Y-chromosomes (Xp22.3; Yp11.3)14
, providing a plausible pathway for the association of EoE with male gender.
Consistent with our analysis, a recent GWAS has identified several SNPs associated with peripheral blood eosinophilia which notably includes the 5q22 locus15
. EoE can associate with peripheral blood eosinophilia, but only in a subset of patients16
. This suggests that EoE and peripheral eosinophilia are associated with the same locus. In conclusion, we report the first genome-wide association study to identify an EoE predisposition locus, implicating TSLP and/or WDR36 in the pathogenesis of EoE.