To the Editor:
Atopic dermatitis (AD) is a chronic skin disease affecting up to 20% of children in industrialized countries. A rare but serious complication of AD is eczema herpeticum (ADEH). We recently reported subjects with ADEH have more severe Th2-polarized disease with greater allergen sensitization, and more commonly, have a history of food allergy, asthma, or both1. Only ~5% of AD patients with HSV seropositivity (e.g. evidence of exposure) develop eczema herpeticum2. This observation, coupled with the evidence that susceptibility to EH can be familial and that most subjects report recurrent EH episodes, collectively suggest that genetic susceptibility may be important.
Thymic stromal lymphopoietin (TSLP) is an interleukin (IL)-7-like cytokine that triggers dendritic cells (DCs) to induce differentiation of naïve T cells into Th2 cells, and is implicated in the pathogenesis of allergic diseases 3. TSLP exerts its biological activities by binding to a heterodimeric receptor consisting of the IL7 receptor a-chain (IL7Rα) and the TSLP receptor chain (TSLPR) to initiate signal transducer and activator of transcription (STAT) 3 and STAT5 phosphorylation 4. Recent studies have demonstrated that polymorphisms of the TSLP gene appear to contribute to Th2-polarized immunity through higher TSLP production by bronchial epithelial cells in response to viral respiratory infections 5. In this study, we hypothesized variants in TSLP and its receptors were associated with the risk of AD, ADEH, and related sub-phenotypes. To test the hypothesis, we conducted genetic association studies in two independent and racially diverse groups of patients participating in the multicenter Atopic Dermatitis Vaccinia Network (ADVN) 1. Detailed information on the participants in the ADVN has been previously described 6. Local IRBs and clinics approved the study, and written informed consent was obtained from all study participants.
A total of 29 SNPs were selected from TSLP, IL7R, and TSLPR (15, 11, and 3 respectively) for genotyping. Of these, there were 23 tagging SNPs, one recently reported TSLP functional SNP (rs3806933, -847C/T), two SNPs (rs1898671 and rs10062929) within the initially identified region of TSLP, and three newly validated TSLPR dbSNPs. Details for each SNP and minor allele frequencies (MAF) are presented in Table EI in the Online Repository at www.jacionline.org. Of these, 20 SNPs were genotyped using a custom-designed Illumina (San Diego, CA) oligonucleotide pool assay (OPA) for the BeadXpress Reader System, and nine SNPs were genotyped using ABI TaqMan (Applied Biosystems, Foster City, CA). Quality controls were performed as described previously 6. The Cochran–Armitage trend test was used to test for association between each individual SNP and disease status using PLINK 7. A linear regression analysis was performed to test for associations between individual genetic markers and log-adjusted tIgE and log-adjusted EASI score, and a multiple logistic regression model was used to test for SNP-SNP interaction among cases only between TSLP and its receptors using SAS v9.1.
As graphically summarized in Fig 1, A and Table I, among the primary European American group, a significant association was observed for TSLP SNP rs11466749 and AD (P=0.028). Significant associations were also observed for ADEH and two additional TSLP SNPs (rs1898671, P = 0.001; rs2416259, P=0.021) and 4 IL7R SNPs (rs12516866, P=0.044; rs10213865, P= 0.027; rs1389832; P=0.011; rs10058453, P=0.022). When replication was sought in the smaller African American sample, modest associations were observed for AD and two TSLP SNPs (rs10043985, P=0.034; rs2289276, P=0.018) and 2 IL7R SNPs (rs12516866, P=0.033; rs1053496, P=0.011). The number of ADEH patients in the African American sample was too small to perform meaningful tests for association. To test for association with disease severity, we performed a linear regression analysis using an additive model. Results are summarized in Fig 1, B and Table EII. Associations with total serum IgE concentrations (tIgE) were observed for six TSLP SNPs, including three SNPs in the European American sample (rs10062929, P = 0.031; rs11466750, P = 0.048; rs2416259, P = 0.034) and three SNPs in the African American sample (rs991035, P = 0.012; rs17551370, P = 0.023; rs11466749, P = 0.015), and one IL7R SNP (rs7737000, P = 0.024) in African American sample. Two TSLP SNPs associated with tIgE were also significantly associated with the Eczema Area and Severity Index (EASI), including rs2416259 in the European American sample (P = 0.011) and rs991035 in African American sample (P = 0.013). To identify functional mutations for TSLPR, we sequenced the TSLPR coding regions and validated three dbSNPs with allele frequencies >10%: rs36139698 (P196L), rs36177645 (X201W), and rs36133495 (A238V). All three SNPs were modestly associated with AD in the European American sample (rs36139698, P = 0.026; rs36177645, P = 0.039; and rs36133495, P = 0.041, Table I) but this did not replicate in the African American sample. When AD subjects were limited to male, the most significant association was observed for the SNP rs36133495 (P=0.0005, data not shown). Given the biological relationship between TSLP and its receptors, we tested for epistatic effects (i.e., gene-gene interaction) and observed significant interactive effects for AD and ADEH between TSLP and IL7R and TSLPR (Table EIII). The strongest interaction was observed for AD between TSLP SNP rs2289276 and IL7R SNP rs12516866 (OR, 1.86, P = 0.007) among European Americans, and between TSLP SNP rs1898671 and IL7R SNP rs10213865 (OR, 5.12, P = 0.005) among African Americans.
In this study, we observed significant associations for TSLP and IL7R tagging SNPs and AD and ADEH among a European American sample, with replication of associations between TSLP and IL7R SNPs in an independent African American sample. A SNP (rs1898671) in TSLP showed the strongest association with a lower risk of ADEH among European Americans (P = 0.001), which remained significant after a Bonferroni correction for multiple testing (P = 0.015). Although the function of this variant is unknown, two SNPs (rs3806933 and rs2289276) in the promoter region of TSLP are in strong LD with rs1898671, and all three markers are localized in a single LD block (Fig 1, C). Elsewhere, these markers have been associated with an increase in promoter activity and enhanced AP-1 binding to the regulatory element of TSLP 5. SNP rs2416259 was also associated with risk of ADEH, as well as associated phenotypes, high tIgE and EASI. Of interest, SNP rs2289276, a SNP previously associated with specific IgE levels to cockroach antigen and total IgE 8, was associated with AD in our study. Three validated coding dbSNPs in TSLPR were significantly associated with AD in European Americans, one of which is a non-sense SNP (rs36177645 X201W), which may lead to premature termination of peptides; investigation into the effects on gene expression are underway. Potential epistatic effects between SNPs in TSLP and its two receptors, IL7R and TSLPR, were observed for both AD and ADEH. Associations for SNP-SNP interactions were stronger than the single marker associations.
A ‘gene-for-gene’ replication of associations was observed in the African American sample for the traits analyzed (AD, tIgE and EASI) and TSLP and IL7R SNPs. Because of the smaller African American ADEH sample, it was not possible to test for replication to the ADEH trait. However, the SNPs significantly associated with AD, tIgE and EASI in the African Americans were not for the same SNPs as those associated with these three traits in the European Americans. Failure to observe SNP-for-SNP replication in ethnically diverse populations is not uncommon 9, and potential reasons include variation in allele frequencies, genetic structure, and heterogeneity of the phenotype and environment. Heterogeneity of the phenotype is unlikely in this case because all cases were characterized using identical protocols at multiple centers 1. Previously we evaluated the potential effect of admixture on the case-control design using Ancestry Informative Markers (AIMs) 6, but observed no significant population stratification in this sample. However, nearly half (44.8%) of the SNPs genotyped in the current study had substantially different MAFs between two populations, highlighting considerable genetic heterogeneity between the two ethnic groups, and the LD structure differed considerably in these two samples (data not shown). Very few of the markers genotyped in this study are likely to be causal, but, rather are tagging SNPs in LD with an unknown disease causing variant(s). That said, it is well accepted that African Americans are more genetically diverse than European ancestry populations, and have shorter LD blocks, and it is therefore not surprising that different sets of tagging SNPs may be associated with the trait of interest compared to whites. We contend that by capturing all of the potential risk-conferring variants, a gene-based approach - rather than a SNP-for-SNP - may provide evidence for genetic analysis at the functional level 10.
Other confounders could influence the observed associations in this study, both between case groups and across ethnic groups. Most cases of ADEH are caused by HSV-1, despite the fact that HSV-1 seropositivity is high in the general population 1. In the ADVN study, 93% of the ADEH group was seropositive for HSV-1, compared to seropositivity of just over 50% for AD patients without ADEH (ADEH-, 52%) and control individuals (54%)1. To test for the possibility that the associations observed between TSLP, IL7R and TSLPR were for manifestations of an HSV exacerbation in an AD population rather than the disease markers, we investigated the association between HSV infection as determined by both HSV-1 and HSV-2 positivity and the associated SNPs among controls, none of which showed enhanced associations compared to the associations with ADEH (data not shown). Because there were no statistical differences in HSV infection between ADEH- and control individuals1, we also tested for association between the associated SNPs and ADEH- status and found that the associations with AD remained (data not shown). These results suggest that the associations between AD and ADEH and TSLP, IL7R, and TSLPR are independent of HSV infection, although a more robustly powered sample is needed to verify these findings. We acknowledge that this study was marginally powered overall, due largely to the rarity of the trait of interest, ADEH, which occurs only in less than 5% of subjects with AD. Given the candidate gene nature of this study and the strong biological justification for interrogating the role of TSLP variants and variants in its receptors in risk of developing ADEH and/or AD, the Bonferroni correction for multiple tests is likely too conservative. Nevertheless, one of the most compelling associations was between ADEH and a SNP (rs1898671) in TSLP that is in strong LD with a functional variant previously associated with enhanced AP-1 binding to the regulatory element of TSLP 5, an association that remained significant after correcting for multiple testing.
In summary, we demonstrate evidence of association between markers in TSLP and its receptors, IL7R and TSLPR, and risk of AD and its most serious complication, ADEH, in a multicenter case-control study. Our findings suggest TSLP may be an important candidate for AD, AD severity and ADEH. This is the first study to implicate TSLP as potential casual gene for ADEH. A clearer understanding of these risk factors may improve our ability to identify patients at greatest risk for ADEH, which may ultimately lead to early intervention and improved surveillance in this vulnerable population.