The Genetic Etiology of Eosinophilic Esophagitis

Abstract

Eosinophilic esophagitis (EoE) is a chronic, allergic disease associated with marked mucosal eosinophil accumulation. Multiple studies have reported a strong familial component to EoE, with the presence of EoE increasing risk for other family members with EoE. Epidemiological studies support an important role for environmental risk factors as modulators of genetic risk. In a small percentage of cases, including patients who have Mendelian diseases with co-occurrent EoE, rare genetic variation with large effect sizes could mediate EoE and explain multi-generational incidence in families. Common genetic risk variants mediate genetic risk for the majority of people with EoE. Across the thirty-one reported independent EoE risk loci (p<10⁻⁵), most of EoE risk variants are located in between genes (36.7%) or within the introns of genes (42.4%). While some variants do change the amino acid sequence of genes (2.2%), only three of the thirty-one EoE risk loci harbor an amino acid changing variant. Thus, most EoE risk loci are outside of the coding regions of genes, suggesting a key role for gene regulation in EoE, consistent with most other complex diseases.

Introduction

Eosinophilic esophagitis (EoE) is a chronic, allergic disease associated with marked mucosal eosinophil accumulation ¹. EoE usually remits after removal of specific food types and food re-introduction causes disease recurrence, including marked dysregulation of esophageal transcripts. One of the central questions in the EoE field, and allergy in general, is to understand why individuals develop certain manifestations of allergic disease, such as EoE. Defining a “typical” EoE case, beyond the presence of eosinophils in the esophagus, is challenging. Patients often exhibit marked heterogeneity in their presentation (e.g. age of onset, co-morbidities, and symptomology) and response to treatment. The diversity of presentations makes identification of etiology challenging.

Etiology of EoE: environmental and genetic contributions

Multiple studies have reported a strong familial component to EoE, with the presence of EoE increasing the chance of developing disease for other family members ^{2, 3}. However, it is important to recognize that traits shared between family members can be the result of both genetic and environmental factors. Any mention of genetics often evokes the pioneering work of Gregor Mendel using pea plants to elucidate inheritance. Mendelian inheritance focuses on single variants that completely explain a phenotype. For some families, single gene variants may drive EoE inheritance. Indeed, multiple studies which have shown EoE co-occurring with known Mendelian disorders, supporting a single gene etiology. However, like many diseases, the familial patterning in EoE is often not consistent with Mendelian inheritance. Only ~2% of siblings of an EoE proband also have EoE ². Such a low recurrence risk of developing disease is not consistent with Mendelian modes of inheritance. Further, because there is transmission from fathers to sons, X linked inheritance is not supported. Collectively, these results strongly support that for the most part, EoE is a disease with a complex genetic etiology, and inheritance is due to the effects of multiple genetic loci that increase disease risk in the context of environmental disease risk modifying factors. As with other complex traits, genetic loci that increase risk for EoE are identified before the modified genes can be determined.

Comparing dizygotic and monozygotic twins is an excellent way to measure the contribution of genetic factors for a disease. The Alexander et al. study found that monozygotic (i.e. identical) twins had a 44% disease concordance, a 2-fold increase compared to dizygotic twins ². While the study had insufficient power to statistically distinguish the dizygotic and monozygotic concordance, these data are consistent with a complex genetic component that acts in the context of environmental exposures to increase risk of developing EoE.

Because families live in the same household, they often share environmental exposures. Epidemiological studies seek to identify those exposures that are shared by patients with EoE but not people without the disease ⁴. Leveraging the unique information provided by nuclear families and twins, Alexander and colleagues sought to address the total contribution of genetics to the familial inheritance observed in EoE ². Evaluating the rates of EoE from the relatives of EoE probands, they found that dizygotic twins had nearly a 10-fold increase in EoE compared to siblings. Genetically, siblings and dizygotic twins would be expected to have similar shared genetic influences; thus, the increased rate of EoE in dizygotic twins can be attributed to shared early life environmental factors. The potential role of a shared environment was also noted by Allen-Brady and colleagues, who observed that spouses of EoE probands were also at increased risk for developing EoE ³.

Jenson and colleagues evaluated the role of early life factors ⁵, finding that prenatal (maternal fever and preterm labor), intrapartum (Cesarean section), and early life factors (antibiotic use and acid suppressant use) were all associated with increased risk of developing EoE, while having a furry pet in infancy was associated with decreased risk. However, no one disease risk factor has been identified as sufficient for the development of EoE. Studies have identified environmental factors affecting both children and adults have been associated with risk of developing EoE (sensitization to galactose-alpha,1,3-galactose) and protection (Helicobacter pylori infection) ⁴. Taken together, these analyses support a complex genetic etiology with contributions from both genetic and environmental factors.

Genetic etiology of EoE

In order to identify the regions of the genome that increase risk of developing EoE, the deoxyribose nucleic acid (DNA) of patients with and without EoE are genotyped on a microarray or sequenced using Next Generation Sequencing technology. Common and rare disease genetic variants that increase disease risk are identified through different analytical approaches. Variants are typically considered common if the less frequent minor allele can be found on at least 1% of chromosomes across a large population of individuals; variants are considered rare when they occur at a frequency less than 1%. For common variants, the typical approach is to evaluate a large number of individuals with disease and compare them to individuals without disease to identify those variants whose minor alleles are enriched (i.e., present at a higher frequency) in patients with the disease ⁶. These variants typically have modest effect sizes, with the genetic risk variant increasing disease risk by 30%-100% ⁷. In the context of EoE, if a male with European ancestry carries a single genetic risk variant that increases risk of EoE by 100% increases risk of developing EoE from 1:1000 to 1:500. Thus, the effect sizes of the currently identified EoE genetic risk variants support model for genetic risk in which an accumulation of numerous genetic risk variants increases disease risk in the context of environmental factors.

For the identification of rare variants, two approaches are generally used. The first involves testing whether a specific gene has an increased burden of rare mutations in patients with EoE compared to controls ^8–10. The second approach is to evaluate whether a specific rare variant co-occurs with disease in families in which many members have the disease ^11–13. For both rare variant approaches, analytical filtering approaches are used to narrow down the considered variants ¹⁴. Beyond restricting variants to those that are rare, researchers often restrict to those which are considered to have the potential of a functional impact (either restricting to those that alter an amino acid or through bioinformatics pipelines are predicted to have a functional effect). The challenge with rare variants is that any given individual normally harbors too many rare variants with predicted functional impact for biologic follow-up. Focusing on variants present in multiple cases or located within genes that can be implicated in disease based on their known function can help to reduce the number of variants considered ¹⁵. In addition, the genes carrying rare variants associated with EoE may help elucidate the biologic processed underlying the development of the disease.

A danger of focusing only on coding variants and genes “of biological interest” in the analysis of sequencing studies is that these approaches miss opportunities to identify new biology of non-coding variants. Similarly, focusing on genes with known function eliminates the opportunity to find new functions for genes in the pathoetiology of EoE. While it is experimentally and logistically easier to focus on known genes that have been studied for decades in the context of allergic and epithelial pathways, there are tremendous opportunities for discovery by following the data and exploring any gene with rare variants identified in numerous families.

Common EoE genetic risk loci

As a rare disease with a prevalence of approximately 1:2,000 individuals ¹⁶, EoE genetic studies aimed at identifying common genetic risk variants have been limited to relatively small sample sizes. However, these studies have been informative for identifying several genome-wide significant susceptibility loci. Risk loci are identified based on single genetic variant statistical association that is replicated in two independent cohorts. Because genome-wide association studies (GWAS) assess variants across the genome, a multiple testing correction is used to reduce type 1 error. Based on a history of false positives that failed to replicate in subsequent studies, a conservative Bonferroni correction is the standard multiple testing correction for GWAS. For genomes of individuals with European heritage, there are one million independently segregating haplotypes, so a p-value of 5x10⁻⁸ is required to reach genome-wide significance (5x10⁻⁸ = 0.05/10⁶). EoE genetic risk loci are named based on the chromosomal location and annotated with candidate genes that are located near each genetic risk locus (Table 1).

Table 1.

Reported EoE risk loci

EoE Risk Locus	Tag genetic variant	PMID	Genes at and near risk variants	Risk Alelle Frequency	P-value	Odds Ratio
1p13.3	rs2000260	25017104	SLC25A24	0.57	7x10−7	1.32
1p36.13	rs28530674	25017104	KIF17	0.04	3x10−7	1.83
	rs2296225	25017104		0.08	1x10−7	1.63
1p32.2	rs11206830	25017104	AC119674.2	0.02	8x10−8	2.16
2p23.1	rs149864795	25407941	CAPN14	0.052	5x10−10	2.22
	rs77569859	25017104		0.05	3x10−10	1.98
3q26.32	rs6799767	20208534		0.58	4x10−7	1.49
4q21.1	rs13106227	20208534	SHROOM3	0.62	4x10−6	1.52
	rs1986734	20208534		0.49	1x10−6	1.54
5q22.1	rs3806932	20208534	WDR36, TSLP	0.54	3x10−9	1.85
	rs3806933	25017104		0.56	2x10−8	1.37
	rs252716	25407941		0.447	4x10−14	1.52
5q23.1	rs2055376	25017104	SEMA6A	0.02	7x10−8	2.30
5q14.2	rs1032757	20208534		0.07	2x10−6	1.96
6p11.2	rs9500256	20208534	AL445250.1	0.58	5x10−6	2.04
8p23.1	rs2898261	25017104	XKR6	0.58	5x10−8	1.35
8q24.12	rs11989782	20208534	SNTB1	0.23	7x10−6	1.53
8q22.2	rs13278732	20208534	ERICH5	0.27	6x10−6	1.31
10p12.31	rs11819199	25017104	MIR4675	0.06	3x10−7	1.62
10q23.1	rs2224865	20208534	MARK2P15 - LINC02650	0.31	9x10−6	1.44
11q13.5	rs61894547	25407941	LRRC32, EMSY, CAPN5	0.043	4x10−11	2.44
	rs2155219	25017104		0.51	4x10−7	1.37
	rs77301713	25017104		0.02	1x10−7	2.22
11q14.2	rs118086209	25017104	CCDC81	0.02	2x10−7	2.19
11q21	rs1939875	20208534	NR	0.26	3x10−6	1.54
12q13.3	rs167769	20208534	STAT6	0.37	2x10−6	1.36
	rs167769	25407941		0.377	2x10−7	1.35
14q12	rs8008716	25407941	NOVA1	0.087	7x10−8	1.71
15q13.3	rs8041227	25017104	LOC283710, KLF13	0.72	6x10−10	1.52
16p13	rs12924112	29904099	CLEC16A	0.301	2x10−9	0.76
16q24.1	rs371915	20208534	MEAK7	0.87	2x10−8	1.90
17q24.3	rs6501384	20208534	CALM2P1 - AC011990.1	0.33	6x10−6	1.41
17q25.3	rs3744790	25017104	TIMP2, CEP295NL	0.8	8x10−7	1.54
18q12.1	rs7236477	20208534	DSG1, DCC	0.03	7x10−6	2.22
	rs9956738	25017104		0.01	4x10−7	2.47
19q13.11	rs3815700	25407941	ANKRD27	0.14	2x10−9	1.62
21q22.3	rs17004598	25017104	HSF2BP	0.01	1x10−7	2.57
22q11.21	rs2075277	25017104	P2RX6	0.09	9x10−7	1.54

In Table 1, we identify each of the 42 published EoE risk associations reported to date in which the “tag” SNP has a p-value less than 10⁻⁵. Because EoE is a relatively rare disease, the GWAS to date have been poorly to modestly powered to identify significant associations that are robust to a genome-wide multiple testing correction (i.e., have a p-value less than 5x10⁻⁸). Thus, it is reasonable to consider suggestive genetic risk loci while awaiting replication in subsequent studies. From the published EoE risk loci, we can identify other genetic variants located near the reported “tag” variants that are in linkage disequilibrium (LD). Genetic variants in LD are inherited as a haplotype and therefore cannot be distinguished from one another through GWAS. Indeed, when considering LD, there are 543 genetic variants at these 31 independent risk loci across the genome. As depicted in Figure 1, the majority of EoE genetic risk variants are located in between genes (intergenic – 36.7%) or within the introns of genes (intronic – 42.4%). While some variants do change the amino acid sequence of genes (coding – 2.2%), only three of the 31 risk loci harbor an amino acid changing variant. Thus, most risk loci that increase risk for EoE are outside of the coding regions of genes (Figure 1), suggesting a key role for genotype-dependent gene regulation in EoE, consistent with most other complex diseases.

Figure 1.

Annotation of genetic variants associated with EoE. All genetic variants in high linkage disequilibrium (LD) with reported Tag SNPs at EoE risk loci (see Table 1) were identified using European and European American haplotypes in the 1000 genomes project ⁵⁰. 544 variants across the 31 independent risk loci were identified. The variants were annotated based on their position within the genome and in the context of genes.

Amongst three reported GWAS studies, genetic variants at 4 loci have been consistently found at genome wide significance (5q22 [TSLP/WDR36], 2p23 [CAPN14], 11q13 [LRRC32/C11orf30] and 12q13 [STAT6]) ^17–19. In single studies, genome-wide significant associations have also been found at 19q13 (ANKRD27) and 16p13 (CLEC16A) ²⁰. Furthermore, an interplay of gene-gene (particularly between IL4 and TSLP) and gene-environment interactions have been shown to be contributory ^21–23. In a phenome-wide association study approach, a set of genetic variants are assessed in the context of many phenotypes. One such approach using a set of candidate variants identified PTEN, TGFBR1/TGFBR2/PBN and IL5/IL13 EoE genetic risk loci ²⁴.

Only five EoE risk loci include a coding variant in the linkage-disequilibrium expanded set of genetic risk variants. Because the majority of the common genetic risk loci EoE that have been discovered to date have a genetic risk mechanism that results from genotype-dependent transcriptional regulation, further experimental assessment is needed to identify the biological mechanism driving risk of developing EoE. Non-coding variants are found in gene promoters, introns, and regions of the genome with regulatory activity (e.g., enhancers, silencers, or insulators). These variants are hypothesized to affect gene expression through differential transcription factor binding and/or chromatin-based effects on gene regulation. Expression quantitative trait loci (eQTL), cases in which the alleles of a given variant can be connected to genotype-dependent expression of a particular gene, have been identified at multiple EoE risk loci. Sometimes, as in the case of the CAPN14 EoE genetic risk locus, the eQTL is found specifically in patients with disease ¹⁷. EoE genetic risk variants in the promoter of CAPN14 were associated with genotype-dependent expression of CAPN14 in the esophageal biopsies of patients with EoE. Miller et al. subsequently identified an EoE genetic risk variant in the CAPN14 promoter that changed the promoter activity of CAPN14 at a variant with genotype-dependent transcription factor binding ^{17, 25}. Similarly, variants at the TSLP/WDR36 locus are associated with genotype-dependent expression of both TSLP and WDR36 ²⁶. Taken together, these results point to an important, but largely not understood, role for genotype-dependent gene regulatory mechanisms in the etiology of EoE.

Rare EoE genetic risk loci

The study of rare EoE genetic risk loci is an emerging field. Rochman and colleagues identified rare variants within 117 genes which exhibited altered expression in EoE patients ²⁷. Of the 33 patients examined, 25 harbored at least on rare variant which altered amino acid sequence. However, this study did not evaluate whether individuals without EoE would be expected to have similar variant burden in these genes. In contrast, Sherrill and colleagues used a whole exome approach to discovery ³¹. In this work, Sherrill and colleagues sequenced the coding regions of the genome (a.k.a., the exome) of 25 trios (probands along with parents without disease) and 12 multiplex families (including 38 individuals with EoE and 8 family members without EoE). After initial filtering to identify high quality variants that were rare and predicted to alter protein function, over 10,000 variants were identified. Thus, Sherrill and colleagues focused on a single family with a mother and three offspring who have each been diagnosed with EoE, identifying 90 rare protein coding variants that were present in the mother and her 3 offspring with the EoE diagnosis. Among these was a variant that caused a premature termination of the DHKTD1 protein (i.e., a nonsense mutation). Importantly, additional variants inside of DHKTD1 were also present in other families. Analysis at the gene level demonstrated that there as an enrichment in rare protein coding variants result in DHKTD1 in patients with EoE than would have been expected by chance. The results of this study highlight the large number of rare protein coding variants identified using exome analyses. However, this study also highlights the value of multiplex families to reduce the number of variants to a tractable number. Further, while particular rare variants might not be shared between families, rare variants identified using multiplex families can identify candidate genes for further evaluation using genetic burden tests.

Mendelian diseases associated with EoE

For a small percentage of patients, a Mendelian disease that co-occurs with EoE can be accounted for by a single genetic “causal” variant (Table 2). As previously reviewed ^32–33, a larger percentage of people with a connective tissue disease diagnosis have EoE than would be expected in a population of people without the co-occurring disease. Patients with co-diagnoses of Mendelian diseases and EoE also likely have additional common, non-Mendelian genetic alleles that modify the pathoetiology of EoE. These common disease modifiers are yet to be discovered and might overlap or be distinct to those genetic disease risk alleles that increase risk of EoE in patients without these co-occurring diseases.

Table 2.

Mendelian diseases associated with EoE

Mendelian disease associated with EoE	Genetic mutation	Plausible etiologic mechanism
Loeys-Dietz syndrome (LDS)	Mutations in transforming growth factor beta receptors 1 and 2 (TGFBR1 and TGFBR2, respectively)	Enhanced transforming growth factor beta (TGF-β) signaling
Ehlers-Danlos syndrome, hypermobility type	Unknown – other subtypes of Ehlers-Danlos syndrome are caused by mutations in collagen genes	Increased activity of TGF-β due to altered binding by extracellular matrix
Severe atopy syndrome associated with metabolic wasting (SAM syndrome)	Homozygous mutations in desmoglein 1 (DSG1)	Disrupted epithelial barrier
Netherton’s syndrome	Loss-of-function mutations in skin protease inhibitor, kazal type 5 (SPINK5)	Unrestricted protease activity of kallikrein 5 and 7 (KLK5, KLK7)
PTEN hamartoma tumor syndrome (PHTS)	Mutations in phosphatase and tensin homolog (PTEN)	Inhibited regulation of the phosphatidylinositol-4,5-bisphosphate 3-kinase (PI3K) signaling pathway
Autosomal dominant hyper-IgE syndrome	Deleterious mutations in signal transducer and activator of transcription 3 (STAT3)	Dysregulated response to IL-6 and possibly IL-5
Autosomal recessive form of hyper-IgE syndrome	Loss-of-function mutations in dedicator of cytokinesis 8 (DOCK8)	Loss of T cell homeostasis; lack of durable secondary antibody response against specific antigens
Erbin Deficiency	Loss-of-function mutation in ERBB2-interacting protein (ERBIN)	Increased TGF-β pathway activation in T cells with increased Th2 responses

Adapted from Kottyan et al. ⁴⁸and O’shea et al. ⁴⁹

Only the tip of the iceberg: variation explained by known EoE genetic loci is only a fraction of what is expected

Family-based studies can be used to estimate the heritability of a disease (i.e., the proportion of variation explained by genetics). The estimated genetic component to disease risk is larger than can be accounted for by genetic risk loci identified to date. There are several possible reasons for this. First, GWAS are only powered to detect moderate to modest effect sizes; thus, they will not be able to detect subtle effects. The number of variants with subtle effect is likely to be substantial, perhaps numbering in the thousands. While the effect of any one of these variants is miniscule, the cumulative effect could be substantial. A second explanation is that GWAS of EoE have been relatively underpowered to identify and replicate all of the genetic risk loci with moderate effect sizes. A third explanation is that genetic variants do not act in isolation. Rather, genomes, cells, organs, and bodies are intricate machines, with cross-talk and redundant systems. As a result, a single genetic risk variant in isolation may tell only a small part of the story. Thus, it is critical to study genes in the proper context. Martin and colleagues found that the strength of association between TSLP genetic variation and EoE was highly dependent on genetic variation at an additional locus, IL4³¹. When individuals carried EoE genetic risk variants at both the TSLP and IL4 loci, the odds ratio of EoE increased to 3.7, compared to 1.3 and 1.6 when either the EoE genetic risk variants at TSLP or IL4 were carried in isolation, respectively. As the IL4 locus carries increased risk for developing multiple allergic diseases ^32–35, it was speculated that an underlying atopic phenotype may magnify the effects of TSLP. Likewise, Azouz and colleagues searched for interactions between EoE genetic risk variants at TSLP and those near other atopy related genes ³⁶, identifying a novel interaction between genetic variants near TSLP and PLAU. Having both risk variants resulted in a 2.7 Odds for EoE, compared to 1.6 for TSLP variants in the absence of PLAU variants, and 1.3 Odds for PLAU variants in the absence of TLSP variants. However, it is important to note that the context that may modify the association strength is likely not strictly genetic, since environmental factors will likely also be critical for the disease association of many genetic risk loci.

The increased risk of developing EoE associated with some genetic loci might change as a function of specific environmental factors. Jensen and colleagues explored whether the effects of EoE GWAS loci were conditional on early life factors ³⁷. While the sample size on this study was small, an interaction between variants at CAPN14 and breastfeeding was identified; the minor allele of the strongest CAPN14 genetic risk variant increased risk of developing EoE (OR = 5.4) when the child had not been breastfed compared to a decreased risk associated with that variant when the child had been breastfed (OR= 0.43).

Genetic studies of other eosinophilic gastrointestinal diseases and non-European cohorts

To date, studies seeking to identify genetic risk variants for eosinophilic gastrointestinal diseases have been limited to patients with EoE who are of European ancestry ³⁸. Other eosinophilic gastrointestinal diseases, including eosinophilic gastritis, eosinophilic duodenitis, and eosinophilic enterocolitis, are rarer than EoE. Thus, identifying sufficient patients for genome-wide studies is challenging. Likewise, most of the patients that have been included in research-based tissue repositories have been of European ancestry. National Institutes of Health funded studies are currently focused on increasing the numbers of patients in both of these categories, and forthcoming studies of the genetics in these important populations of patients will be critical for the understanding of disease etiology.

Clinical applications of EoE genetic risk loci

With the completion of the human genome, it was predicted that genomics would transform medicine. While technological revolutions have enabled research to move forward much more rapidly, we are not yet positioned to use genetics for early detection or selection of optimal treatments for the vast majority of individuals with an EoE diagnosis. As noted, most identified variants explain only a modest portion of disease risk. Genetic risk loci have already increased our understanding of the molecular pathophysiology of EoE through the identification of critical genes and pathways. For example, the identification of the CAPN14 locus highlighted the role of this and other proteases in esophageal epithelial remodeling in EoE ^{17, 25, 39–41}. In addition to a clear disease role for esophageal epithelial cells, data support a role for immune cells including T cells, eosinophils, and mast cells ^{39, 42–47}.

In order to translate the genetic risk at individual loci into a patient-based assessment of genetic risk, the cumulative genetic burden of identified EoE risk variants can be assessed by summing a person’s risk alleles weighted by effect-size ^{48, 49}. These ‘polygenic risk scores’ present an attractive opportunity to use genetic risk variants to identify those patients at the highest chance for the development of EoE. To date, these types of studies are of limited clinical usefulness for EoE, a disease with few genetic loci and a strong environmental component. Future studies will focus on the combination of genetic, transcriptomic, and environmental factors to predict those people at highest probability for the development of EoE.

Conclusions

Since the initial studies of EoE genetics nearly a decade ago, much progress has been made in the identification and understanding of genetic loci that increase risk of developing EoE (as previously reviewed in ^29–30). To date, the data support a model in which genetic risk variants affect gene expression leading to structural and physiological changes in epithelial and immune cell function; these changes are hypothesized to lead to EoE in the context of other environment-mediated molecular and physiological changes. In some cases, rare genetic variation with large effect sizes could mediate EoE and explain multi-generational incidence in families. Future studies will focus on larger cohorts, patients of non-European ancestry, and the biological mechanisms that drive the genotype-dependent changes in the development of EoE.

Figure 2. Genetic risk of EoE.

Genetic risk variants are polymorphic places in the genome in which a risk allele is enriched in groups of patients with EoE compared to healthy controls. These variants are largely located in the non-coding region of the genome and do not affect the amino acid usage in proteins. Often, the way these noncoding variants increase risk of EoE is through changing DNA regulatory activity, leading to genotype-dependent expression of a gene. In some cases, these genes with allele-dependent expression (such as TSLP) affect the physiology of epithelial cells in the esophagus. These physiological changes can lead to EoE in patients.

Abbreviations

DNA

Deoxyribose nucleic acid

EoE

Eosinophilic esophagitis

eQTL

Expression quantitative trait loci

GWAS

Genome-wide association studies

LD

Linkage disequilibrium