#PAGE_PARAMS# #ADS_HEAD_SCRIPTS# #MICRODATA#

Age-of-onset information helps identify 76 genetic variants associated with allergic disease


Authors: Manuel A. R. Ferreira aff001;  Judith M. Vonk aff002;  Hansjörg Baurecht aff003;  Ingo Marenholz aff005;  Chao Tian aff007;  Joshua D. Hoffman aff008;  Quinta Helmer aff009;  Annika Tillander aff010;  Vilhelmina Ullemar aff010;  Yi Lu aff010;  Sarah Grosche aff005;  Franz Rüschendorf aff005;  Raquel Granell aff012;  Ben M. Brumpton aff012;  Lars G. Fritsche aff013;  Laxmi Bhatta aff013;  Maiken E. Gabrielsen aff013;  Jonas B. Nielsen aff016;  Wei Zhou aff017;  Kristian Hveem aff013;  Arnulf Langhammer aff018;  Oddgeir L. Holmen aff013;  Mari Løset aff013;  Gonçalo R. Abecasis aff013;  Cristen J. Willer aff015;  Nima C. Emami aff020;  Taylor B. Cavazos aff020;  John S. Witte aff020;  Agnieszka Szwajda aff024aff007;  ;  David A. Hinds aff007;  Norbert Hübner aff005;  Stephan Weidinger aff003;  Patrik KE Magnusson aff010;  Eric Jorgenson aff025;  Robert Karlsson aff010;  Lavinia Paternoster aff012;  Dorret I. Boomsma aff009;  Catarina Almqvist aff010;  Young-Ae Lee aff005;  Gerard H. Koppelman aff027
Authors place of work: Genetics and Computational Biology, QIMR Berghofer Medical Research Institute, Brisbane, Australia aff001;  University of Groningen, University Medical Center Groningen, Epidemiology, Groningen Research Institute for Asthma and COPD, Groningen, the Netherlands aff002;  Epidemiology, University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, Groningen, the Netherlands aff002;  Department of Dermatology, Allergology and Venereology, University Hospital Schleswig-Holstein, Campus Kiel, Kiel, Germany aff003;  Department of Epidemiology and Preventive Medicine, University of Regensburg, Regensburg, Germany aff004;  Max Delbrück Center (MDC) for Molecular Medicine, Berlin, Germany aff005;  Clinic for Pediatric Allergy, Experimental and Clinical Research Center of Charité Universitätsmedizin Berlin and Max Delbrück Center, Berlin, Germany aff006;  23andMe, Inc., Mountain View, California, United States of America aff007;  Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, California, United States of America aff008;  Department Biological Psychology, Netherlands Twin Register, Vrije University, Amsterdam, The Netherlands aff009;  Department of Medical Epidemiology and Biostatistics and the Swedish Twin Registry, Karolinska Institutet, Stockholm, Sweden aff010;  CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences, Vienna, Austria aff011;  MRC Integrative Epidemiology Unit, Population Health Sciences, University of Bristol, Bristol, United Kingdom aff012;  K.G. Jebsen Center for Genetic Epidemiology, Department of Public Health and Nursing, NTNU, Norwegian University of Science and Technology, Trondheim, Norway aff013;  Department of Thoracic Medicine, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway aff014;  Department of Biostatistics and Center for Statistical Genetics, University of Michigan, Ann Arbor, Michigan, United States of America aff015;  Department of Human Genetics, University of Michigan, Ann Arbor, Michigan, United States of America aff016;  Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan, United States of America aff017;  The HUNT Research Centre, Department of Public Health and Nursing, NTNU, Norwegian University of Science and Technology, Trondheim, Norway aff018;  Department of Dermatology, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway aff019;  Program in Biological and Medical Informatics, University of California, San Francisco, San Francisco, California, United States of America aff020;  Department of Epidemiology and Biostatistics, University of California, San Francisco, San Francisco, California, United States of America aff021;  Institute for Human Genetics, University of California, San Francisco, San Francisco, California, United States of America aff022;  Department of Urology, Helen Diller Family Comprehensive Cancer Center, University of California, San Francisco, California, United States of America aff023;  Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden aff024;  Division of Research, Kaiser Permanente Northern California, Oakland, California, United States of America aff025;  Pediatric Allergy and Pulmonology Unit at Astrid Lindgren Children’s Hospital, Karolinska University Hospital, Stockholm, Sweden aff026;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027;  University of Groningen, University Medical Center Groningen, Beatrix Children’s Hospital, Pediatric Pulmonology and Pediatric Allergology, and University of Groningen, University Medical Center Groningen, Groningen Research Institute for Asthma and COPD, aff027
Published in the journal: Age-of-onset information helps identify 76 genetic variants associated with allergic disease. PLoS Genet 16(6): e32767. doi:10.1371/journal.pgen.1008725
Category: Research Article
doi: https://doi.org/10.1371/journal.pgen.1008725

Summary

Risk factors that contribute to inter-individual differences in the age-of-onset of allergic diseases are poorly understood. The aim of this study was to identify genetic risk variants associated with the age at which symptoms of allergic disease first develop, considering information from asthma, hay fever and eczema. Self-reported age-of-onset information was available for 117,130 genotyped individuals of European ancestry from the UK Biobank study. For each individual, we identified the earliest age at which asthma, hay fever and/or eczema was first diagnosed and performed a genome-wide association study (GWAS) of this combined age-of-onset phenotype. We identified 50 variants with a significant independent association (P<3x10-8) with age-of-onset. Forty-five variants had comparable effects on the onset of the three individual diseases and 38 were also associated with allergic disease case-control status in an independent study (n = 222,484). We observed a strong negative genetic correlation between age-of-onset and case-control status of allergic disease (rg = -0.63, P = 4.5x10-61), indicating that cases with early disease onset have a greater burden of allergy risk alleles than those with late disease onset. Subsequently, a multivariate GWAS of age-of-onset and case-control status identified a further 26 associations that were missed by the univariate analyses of age-of-onset or case-control status only. Collectively, of the 76 variants identified, 18 represent novel associations for allergic disease. We identified 81 likely target genes of the 76 associated variants based on information from expression quantitative trait loci (eQTL) and non-synonymous variants, of which we highlight ADAM15, FOSL2, TRIM8, BMPR2, CD200R1, PRKCQ, NOD2, SMAD4, ABCA7 and UBE2L3. Our results support the notion that early and late onset allergic disease have partly distinct genetic architectures, potentially explaining known differences in pathophysiology between individuals.

Keywords:

Genome-wide association studies – Allergic diseases – Allergic rhinitis – Medical risk factors – Asthma – Genetics of disease – Eczema – Food allergies

Introduction

In the last 10 years, at least 45 genome-wide association studies (GWAS) of allergic disease susceptibility were published: 25 for asthma (reviewed in [1]), three for hay fever (or allergic rhinitis) [24], eight for eczema (or atopic dermatitis) [512], four for food allergy [1316] and six for allergy-related traits, namely atopic march [17], asthma with hay fever [18], allergies [19], allergic sensitization [2,20,21] and a combined asthma, hay fever and eczema phenotype [22]. Genetic risk variants identified in these studies provide a foundation to help us better understand why and how allergic disease develops in susceptible individuals.

One twin study has previously indicated that the timing of asthma onset may be under genetic control [23]. In the first genome wide association study for asthma published in 2007, it was reported that the ORMDL3/GSDMA locus at chromosome 17q12 was specifically associated with childhood onset asthma[24]. This observation was subsequently confirmed, showing strong associations of this locus with childhood-onset asthma, potentially interacting with passive cigarette smoke exposure in early childhood [25] or as childhood onset asthma defined as asthma developing before 16 years of age) but not later onset asthma in the GABRIEL consortium [26] In subsequent stratified analyses in a multinational study, it was reported that the association of the 17q risk SNP rs7216389-T was confined to cases with early onset of asthma, particularly in early childhood (age: 0–5 years) and adolescence (age: 14–17 years), but a weaker association was observed for onset between 6 and 13 years of age, whereas no association was observed for adult-onset asthma [27]. This shows that defining cut-offs for age at onset of asthma is difficult, and that other approaches such as using a continuous age at onset might be beneficial.

To our knowledge, only three studies have reported genetic variants that associate with the age at which allergic disease symptoms first develop. Forno et al. [28] studied asthma age-of-onset in 573 children and identified two variants that had a genome-wide significant association after combining the discovery and replication (n = 931) cohorts: rs9815663 near the CRBN gene on chromosome 3p26, and rs7927044 near ETS1 on 11q24. In a more recent GWAS of 5,462 cases with asthma, Sarnowski et al. [29] identified five variants associated with age-of-onset, located in/near: CYLD on 16q12 (rs1861760), IL1RL1 on 2q12 (rs10208293), HLA-DQA1 on 6p21 (rs9272346), IL33 on 9p24 (rs928413) and GSDMA on 17q12 (rs9901146). The latter four variants were previously reported to be associated with allergic disease susceptibility as well. Lastly, Ferreira et al. [22] reported that 26 of 136 variants associated with allergic disease risk were also associated with the age at which allergic symptoms first developed (n = 35,972). Amongst these were five variants for which the association with age-of-onset was genome-wide significant: rs61816761 in the FLG gene and rs12123821 near HRNR, both on chromosome 1q21; rs921650 in GSDMB on 17q12; rs10865050 in IL18R1 on 2q12; and rs7936323 near LRRC32 on 11q13. Two of the variants reported in Ferreira et al. (rs10865050 and rs921650) were in linkage disequilibrium (LD) with variants reported in Sarnowski et al., and so are unlikely to represent independent associations. Therefore, collectively across these three studies, 12 variants (2+5+5, including 10 in low LD with each other) were reported to associate with age-of-onset of allergic disease at the genome-wide significance level. Of interest, the joint association between age-at-onset and disease susceptibility at some of these loci [29] suggests that both phenotypes are genetically correlated, and so that combining information from both may improve power to identify variants that influence the aetiology of allergic disease.

The main aim of this study was to identify novel loci that contribute to inter-individual variability in the age at which allergic symptoms first develop, considering information from the three most common allergic diseases: asthma, hay fever and eczema. Rather than study the age-of-onset of each disease separately, we adopted the multi-disease phenotype approach that we used recently to identify risk variants that are shared across different allergic diseases [22]. Specifically, we determined the earliest age at which asthma and/or hay fever and/or eczema first developed and then tested this single combined age-of-onset of allergic disease phenotype in a GWAS. In addition, we also tested if variants associated with disease age-of-onset were also associated with disease risk, as noted by Sarnowski et al. [29]. Lastly, we used multivariate association analysis to identify variants jointly associated with allergic disease age-of-onset and case-control status, which were missed by analyzing each phenotype alone.

Results

Genetic variants associated with the age-of-onset of allergic disease

Our study population consisted of n = 117,130 participants from the UK Biobank study (S2 Table), who had a mean age of 55.5 years (range 38–72 years), with a mean (median) age at onset of any allergic disease of 26.3 (22) years, defined as the earliest age at which any allergic disease (asthma, hay fever or eczema) was first reported (see S1 Fig for distribution).

We first performed a GWAS of a combined age-of-onset phenotype (n = 117,130 from the UK Biobank study. After adjusting the association results (S2 Fig) for the observed LD-score regression intercept [30] of 1.025, we identified 4,160 variants with a genome-wide significant association with age-of-onset (P<3x10-8, Fig 1). Of these, 50 variants in 40 loci (i.e. regions >1 Mb apart) remained associated at that threshold after accounting for the effects of adjacent SNPs in joint association analysis (<10 Mb; Table 1 and S3 Table), indicating that they represent statistically independent associations with age-of-onset. Henceforth, we refer to these SNPs as sentinel variants for age-of-onset. Two additional variants had a P<3x10-8 in the joint but not in the original single-SNP analysis (S4 Table), both located in the major histocompatibility complex (MHC) locus. These represent secondary association signals at the MHC that were masked in the original GWAS by the association with other stronger nearby SNPs.

Fig. 1. Summary of results from the GWAS of allergic disease age-of-onset in the UK Biobank study (n = 117,130).
Summary of results from the GWAS of allergic disease age-of-onset in the UK Biobank study (<i>n</i> = 117,130).
UK Biobank participants reported age-of-onset for asthma and, in a single separate question, for hay fever/eczema. In this analysis, we took the earliest age-of-onset reported across these two questionnaire items and tested this phenotype for association with SNP allelic dosage. We identified 4,160 variants associated with age-of-onset at a P<3x10-8 (red circles), including 50 with a statistically independent association.
Tab. 1. Variants independently associated with allergic disease age-of-onset at a P<3x10-8.
Variants independently associated with allergic disease age-of-onset at a <i>P</i><3x10<sup>-8</sup>.

Three of the 50 sentinel variants were in linkage disequilibrium (LD; r2>0.8) with variants previously reported to have a genome-wide significant association with asthma age-of-onset [29]: rs72823628 in IL18R1, rs7848215 near IL33 and rs4795400 in GSDMB. Similarly, an additional three variants were in LD with SNPs that we reported recently [22] to be associated with the same combined age-of-onset phenotype: rs61816761 in FLG, rs12123821 near HRNR and rs11236791 near LRRC32. On the other hand, to our knowledge, the remaining 44 sentinel variants have not previously been implicated in the age-of-onset of any allergic disease at the genome-wide significance level.

Of the 12 specific variants previously reported to associate with allergic disease age-of-onset, 11 were tested in our current age-of-onset GWAS, of which nine had a highly significant and directionally concordant association (S5 Table). For two variants, there was no evidence for association with the combined age-of-onset phenotype: rs1861760 near CYLD (P = 0.41), reported by Sarnowski et al. [29], and rs9815663 near CRBN (P = 0.67), reported by Forno et al. [28]. The second variant reported by Forno et al. had a MAF<1% and so it was not tested in our current age-of-onset GWAS. We did however test this variant ad-hoc and found that it was not significantly associated with age-of-onset (P = 0.35, not shown).

Potential impact of recall bias and phenotypic misclassification on SNP associations

All UK Biobank participants included in our analyses were adults (aged 38 to 70) at the time of data collection, and so recall bias might have affected the reported age-of-onset. Furthermore, proportionally, there were many individuals who reported late onset of allergic disease (e.g. 41% of asthmatics with onset ≥40 years old), which could have resulted from recall bias and/or phenotypic misclassification. We performed an additional set of analyses to determine if these potential confounders were likely to have had a major impact on the SNP associations described above. We addressed reliability of the age-of-onset information by comparing the self-reported age-of-onset between two surveys that were between 4–7 years apart. Age of onset was within 5 years accurate in 86% of cases. Subjects that reported less reliable information were likely to be older at enrollment. Older subjects were also less likely to report childhood onset asthma. When we analyzed the 50 sentinel variants in subjects who reported developing asthma as a child, and secondly, rhinitis as a child, we obtained highly consistent results, see S4A Fig and S4B Fig, respectively. We also replicated our findings in a prospective birth cohort ALSPAC, and show a high correlation of 0.67–0.825 of the effect size of our analysis with the results obtained in the ALSPAC study. Since the ALSPAC study prospectively assessed asthma, recall bias in this study is not a concern. Moreover, we correlated our findings for adult-onset asthma with two independent, published datasets of asthma GWAS performed by the GABRIEL consortium [26] and the TAGC consortium [31], and identified a substantial genetic correlation of rg of 0.62 and 0.66, respectively. We further correlated our UKBB results of adult onset asthma with an analysis of adult onset asthma in the HUNT study, and again observed a significant genetic correlation rg of 0.69. Further details of these analyses are provided in S1 Data (page 9) and S3 FigS6 Fig, and S14 Table.

Association with age-of-onset in individuals suffering from a single allergic disease

By analyzing an age-of-onset phenotype that considered information from asthma, hay fever and eczema, the GWAS described above was expected to identify variants that affect age-of-onset broadly across the three diseases. To formally address this possibility, we tested each of the 50 sentinel variants identified above for association with the age-of-onset of asthma, hay fever and eczema, in three separate analyses. Specifically, we analysed age-of-onset in three non-overlapping groups of individuals (S1 Fig): those who reported suffering only from asthma (n = 22,029), only from hay fever (n = 14,474) or only from eczema (n = 3,969). Within each of these groups, we tested the association between the 50 sentinel variants and disease age-of-onset, using BOLT-LMM. In individuals suffering from asthma only, 19 sentinel variants were associated with variation in age-of-onset at P<3.3x10-4 (43 at P<0.05), which corrects for 50 SNPs tested in 3 groups, despite the smaller sample size of this analysis (S6 Table). For hay fever and eczema, there were respectively 8 and 5 SNPs associated with age-of-onset at that significance threshold (24 and 12 at P<0.05). Of note, the directional effect observed with the combined phenotype was the same as in the single disease analyses for most sentinel variants (100% for asthma, 94% for hay fever and 80% for eczema).

Lastly, when we formally compared the effect of each sentinel variant on age-of-onset (i.e. the beta from the linear model) between pairs of diseases, we found that most variants (45 of 50, 90%) did not have significant disease-specific effects on age-of-onset (all pairwise comparisons with P>3.3x10-4; S6 Table). The exceptions were four variants located on chromosomes 1q21.3 (in/near TCHHL1, HRNR, FLG and SPRR2A) which had significantly stronger effects on age-of-onset of eczema, and one on 17q12 (in GSDMB) which had a stronger effect on the age-of-onset of asthma (Fig 2). Therefore, we conclude that most (45 of 50) sentinel variants identified in the GWAS of the combined age-of-onset phenotype have similar effects when considering the age-of-onset separately for asthma, hay fever and eczema.

Fig. 2. Variants with evidence for disease-specific effects on age-of-onset.
Variants with evidence for disease-specific effects on age-of-onset.
Each of the 50 variants identified in the GWAS of age-of-onset were tested for association with age-of-onset in three non-overlapping groups of individuals: those suffering from asthma only (n = 22,029), hay fever only (n = 14,474) and eczema only (n = 3,969). We then compared the effects (i.e. betas) obtained in these three groups. For 5 of the 50 variants (shown with an orange inner triangle), the effect on age-of-onset was significantly different (P<0.05/(3 x 50) = 3.3x10-4) between at least two groups. For a given variant, the vertices of the inner triangle point to the position along the edges of the outer triangle that corresponds to difference in effect observed between pairs of single-disease cases. For example, the rs61816761[A] allele, which is located in the FLG gene (filaggrin), had an effect on age-of-onset that was larger (absolute of difference = 0.42) in individuals suffering only from eczema when compared to individuals suffering only from hay fever (P = 4.3x10−8), consistent with this SNP having a stronger effect on the age-of-onset of eczema than of hay fever. For comparison, a variant with no significant differences when comparing the effect on age-of-onset in all three pairwise single-disease association analyses is also shown (rs705699, in the RAB5B gene). In this case, the difference in effect was approximately equal to 0 in the three pairwise comparisons. The color of the difference in effect reflects the significance of the corresponding z-score (see Methods): red for P< 3.2x10−4 (correction for multiple testing), blue for P<0.05 and black for P>0.05.

Association between age-of-onset sentinel variants and allergic disease risk

We then asked if the 50 sentinel variants were also likely to influence the risk of developing allergic disease, in addition to contributing to variation in age-of-onset amongst affected individuals. To this end, we investigated the association between each sentinel variant and a combined allergic disease phenotype, as reported in our recent GWAS [22]. After excluding the UK Biobank study from that GWAS, association results were based on data from 137,883 cases with asthma and/or hay fever and/or eczema, and 84,601 disease-free controls. Forty-eight of the 50 sentinel variants were tested in that GWAS, either directly or via a proxy (one variant), of which 38 (or 79%) were significantly associated with disease risk (P<0.001, which corrects for 48 tests; Table 2). This includes 19 variants for which the association with disease risk was genome-wide significant (P<3x10-8); that is, variants that represent previously known risk factors for allergic disease. Notably, for all 48 variants tested, the allele associated with a higher disease risk was associated with a lower age-of-onset. Therefore, we conclude that the sentinel variants identified influence both the likelihood of developing any allergic disease as well as the age at which symptoms first develop.

Tab. 2. Association between sentinel age-of-onset variants and allergic disease risk in an independent sample of 222,484 individuals studied by Ferreira et al. [22].
Association between sentinel age-of-onset variants and allergic disease risk in an independent sample of 222,484 individuals studied by Ferreira et al. [<em class="ref">22</em>].

Genetic correlation between age-of-onset and disease case-control status

For all 50 sentinel age-of-onset variants, the allele that was associated with a lower age-of-onset was associated with a higher risk of allergic disease. This observation suggested that these two traits–age-of-onset and case-control status of allergic disease–have a substantial negative genetic correlation; to our knowledge, this has not been previously estimated. To understand the extent to which the same genetic variants contribute to variation in these two traits, we applied LD-score regression [30] to the summary statistics of our age-of-onset and allergic disease [22] GWAS. Based on 1.1 million common SNPs, the genetic correlation between the two traits was estimated to be -0.625 (SE = 0.038, P = 4.5x10-61). This estimate was not expected to be biased by the sample overlap between the two GWAS [32], which we confirmed when we excluded samples from the UK Biobank study from the allergic disease [22] GWAS (rg = -0.612, SE = 0.046, P = 5.0x10-41). These results indicate that a substantial fraction of genetic variants are likely to influence both the liability to, and the age-of-onset of, allergic disease. Furthermore, for most (but not necessarily all) shared variants, the directional effect is such that variants that are associated with higher disease risk are associated with lower age-of-onset.

More broadly, these results strongly suggest that a key risk factor that distinguishes individuals with early disease onset from those with late disease onset is the overall genetic burden inherited at allergy-associated SNPs. To illustrate this effect, we compared the distribution of age-of-onset between individuals with the highest (top 10%) and the lowest (bottom 10%) polygenic risk score (PRS) for allergic disease, constructed for each individual from the UK Biobank study based on information from 136 allergy risk variants that we reported recently [22]. This analysis was performed separately for asthma, hay fever and eczema, using the same single-disease case groups described above. For asthma, individuals with the lowest genetic burden of allergic disease (n = 2,202) had a median age-of-onset of 39 years, with only 14% having an age-of-onset before the age of 16; the distribution of age-of-onset was broadly consistent with a pattern of late disease onset (Fig 3). In contrast, in the group with the highest genetic burden (n = 2,203), the median age-of-onset decreased to 29 years, with 35% of individuals reporting that asthma was diagnosed before the age of 16. In this group, there was a clear shift in the distribution of age-of-onset towards a pattern of early disease onset. Similar results were observed for hay fever and eczema (Fig 3). Collectively, our results indicate that genetic risk factors for allergic disease are enriched in cases with early disease onset when compared to those with late disease onset.

Fig. 3. Distribution of allergic disease age-of-onset as a function of a polygenic risk score (PRS) for allergic disease in UK Biobank participants who reported suffering from a single disease (asthma only, hay fever only and eczema only).
Distribution of allergic disease age-of-onset as a function of a polygenic risk score (PRS) for allergic disease in UK Biobank participants who reported suffering from a single disease (asthma only, hay fever only and eczema only).
The PRS of each individual was calculated based on 136 SNPs that were associated with allergic disease risk in our recent GWAS [22]. The mean and median of each distribution are shown in red and blue, respectively.

Multivariate GWAS of allergic disease case-control status and age-of-onset

The high genetic correlation observed between case-control status and age-of-onset of allergic disease suggests that a large number of variants contribute to the heritability of both traits. We therefore hypothesized that multivariate association analysis would identify variants jointly associated with both traits that were missed in the single-trait analyses. To this end, we first adjusted the single-SNP results obtained in the age-of-onset and case-control [22] GWAS for the effects of the sentinel variants identified in the respective study. In the two resulting adjusted GWAS, there were no variants with an association significant at a P<3x10-8, as expected (S6 Fig and S7 Fig). There was, however, an excess of significant associations when compared to the number expected by chance given the number of SNPs tested (S8 Fig and S9 Fig). Many of these associations are likely to represent true positive findings that do not reach the stringent genome-wide significance threshold in each of those two univariate analyses. To help identify these, we then performed multivariate analysis of age-of-onset and case-control status, using metaUSAT [33], which is applicable to association summary statistics. Using this approach, we identified 281 variants with a multivariate P<3x10-8 (Fig 4 and S11 Fig), including 26 that were in low LD with each other (r2<0.05) and so that are likely to represent statistically independent associations (Table 3). However, the QQ Plots may indicate some inflation of the P values, so therefore, these data need to be interpreted with caution. The genomic inflation factor could not be calculated because metaUSAT does not have a closed form null distribution. Nonetheless, inflation of significant associations can be assessed by comparing the observed and expected number of associations significant at a given significance threshold. We observed 38%, 17%, 10%, 5.9% and 1.9% of SNPs tested with a multivariate P-value <0.5, <0.2, <0.1, <0.05 and <0.01, respectively, when the expectations under the null hypothesis of no association were 50%, 20%, 5% and 1%. For most variants, the association in each of the two univariate analyses was one to four orders of magnitude below genome-wide significance, which was exceeded in the multivariate analysis. For all variants, the allele associated with higher disease risk was associated with lower age-of-onset. Results obtained with the recently described MTAG multivariate approach [34] supported the associations identified with metaUSAT (S7 Table). We conclude that these 26 variants represent risk factors for both the presence and early onset of allergic disease, which were only detectable when we combined information from the age-of-onset and case-control GWAS.

Fig. 4. Summary of results from the multivariate analysis of allergic disease age-of-onset and allergic disease case-control status.
Summary of results from the multivariate analysis of allergic disease age-of-onset and allergic disease case-control status.
The GWAS of allergic disease age-of-onset was performed in the UK Biobank study (n = 117,130) as described in the main text. The GWAS of allergic disease case-control status included 360,838 individuals, has reported recently [22]. Single-SNP results from each GWAS were adjusted for the top independent associations (P<3x10-8) identified and then multivariate analysis was performed using metaUSAT [33]. We identified 281 variants with a multivariate P<3x10-8 (red circles), including 26 that were in low LD (r2<0.05) with each other and so that are likely to represent statistically independent associations.
Tab. 3. Variants jointly associated with allergic disease age-of-onset and allergic disease risk (multivariate P<3x10-8).
Variants jointly associated with allergic disease age-of-onset and allergic disease risk (multivariate <i>P</i><3x10<sup>-8</sup>).

Sentinel variants not previously implicated in the aetiology of allergic disease

We then determined which of the sentinel variants identified in the age-of-onset and multivariate GWAS described above represented novel associations for allergic disease in general, that is, when considering all previously reported associations with P<5x10-8 for asthma, hay fever, eczema, food allergy and/or atopy. Of the 50 sentinel variants identified in our age-of-onset GWAS, 47 were in LD (r2>0.05) with variants previously reported to associate with allergic disease (S8 Table). The remaining 3 represent novel associations for allergic disease: rs184587444 in SPRR2A, rs4971089 in KRTCAP2, and rs4809619 in EYA2 (Table 1). On the other hand, most 15 of the 26 sentinel variants identified in the multivariate GWAS represent novel associations for allergic disease (Table 1 and S8 Table), including for example rs7565907 in LCLAT1and rs11242709 near DUSP22. Thus, overall, by considering age-of-onset information, we identified 18 (3+15) novel genetic associations for allergic disease.

Likely target genes of sentinel variants identified in the age-of-onset and multivariate GWAS

To help understand how the 76 sentinel variants might influence allergic disease pathophysiology, we identified genes for which variation in gene expression and/or protein sequence was associated/determined by SNPs in LD with the sentinel variants.

We first extracted association summary statistics from 101 published datasets of eQTL identified in five different broad tissue types relevant for allergic disease (S1 Table). For each gene and for a given eQTL dataset, we then (i) identified cis eQTL in low LD (r2<0.05) with each other, which we refer to as “sentinel eQTL”; and (ii) determined if any of the 76 sentinel variants were in high LD (r2>0.8) with a sentinel eQTL. Using this approach, we found sentinel eQTL in LD with 26 of the 50 (52%) sentinel variants identified in the age-of-onset GWAS (S9 Table), and with 15 of the 26 (58%) sentinel variants identified in the multivariate GWAS (S10 Table). The sentinel eQTL implicated respectively 47 and 28 genes (one in common: HLA-DQB1) as likely targets of the sentinel variants identified in these two GWAS (Table 4).

Tab. 4. Genes with a sentinel eQTL (in italic) or non-synonymous SNP (in bold) in LD (r2>0.8) with sentinel variants identified in the age-of-onset or multivariate GWAS.
Genes with a sentinel eQTL (in italic) or non-synonymous SNP (in bold) in LD (<i>r</i><sup><i>2</i></sup>>0.8) with sentinel variants identified in the age-of-onset or multivariate GWAS.

Second, we found 21 non-synonymous SNPs in 14 genes that were in high LD (r2>0.8) with sentinel variants identified in the age-of-onset or multivariate GWAS (S11 Table). This list included, for example, four non-synonymous SNPs in the CD200R1 gene that were in complete LD (r2 = 1) with the sentinel variant identified in the multivariate GWAS. Of the 14 genes, seven were novel target predictions, that is, they were not identified in the eQTL analysis described above: FLG, EFNA1, SH2B3, TNFRSF14, HIST1H2BE, MLX and YDJC. Overall, when considering information from eQTL and non-synonymous SNPs, we identified 81 (47+27+7) likely target genes of the 76 sentinel variants identified in this study.

Association between the 76 sentinel variants and the risk of food allergy

Finally, we tested if the sentinel variants identified above were associated with food allergy case-control status, in children and adults separately. Although the discovery analysis that identified the sentinel variants for age of onset of allergy did not include food allergy, we hypothesized that these sentinel variants may also relate to food allergy. First, we extracted association results from GWAS that we published recently [15], comprising 497 children with food allergy diagnosed by oral food challenge in the GOFA study and 2,387 controls. This study comprised a highly selected group of children with early onset food allergy (mean age at diagnosis was 2.1 years). In that GWAS, nine of the 76 sentinel variants were significantly associated with food allergy after correcting for multiple testing (P<0.05/76 = 0.00065; S13 Table), namely those in/near FLG (four variants), KIF3A, LRRC32, RAD50, CYLD, and SERPINB7. Overall, there was a very close agreement in SNP associations between the age-of-onset and food allergy analyses (S12 Fig); for example, for 66 of 76 variants the allele associated with a lower age-of-onset onset was associated with a higher disease risk (binomial test P = 2x10-12).

To assess the association between the 76 sentinel variants and food allergy risk in adults, we extracted association results from a GWAS of self-reported food allergy conducted in the adult GERA cohort [16], which included 5,108 subjects with self-reported food allergy, of whom 1,104 were admitted to hospital because of food allergy and 23,945 controls who did not report to have food allergy. In this GWAS, we compared the 1,104 subjects admitted to hospital because of food allergy to the 23.945 controls. No single variant was significantly associated with food allergy after correcting for multiple testing (S13 Table). Across the 72 variants tested in both the child and adult food allergy GWAS, only 43 (60%) had a directionally consistent association, reflecting very little agreement between results from the two analyses. Overall, our results show that many variants associated with allergic disease age-of-onset also represent genetic risk factors for food allergy in young children but not (or less so) in adults. Moreover, the self-report of food allergy in the adult population is more subject to misclassification and may also have contributed to this latter observation.

Discussion

In this study, we identified (i) 50 variants associated with allergic disease age-of-onset; (ii) a significant negative genetic correlation between allergic disease age-of-onset and case-control status; (iii) 26 additional variants jointly associated with allergic disease age-of-onset and case-control status; (iv) 81 genes that are likely targets of sentinel variants identified in the age-of-onset or multivariate GWAS; and (v) nine variants (out of the 76) that are also associated with the risk of food allergy in young children.

Amongst the 50 associations for allergic disease age-of-onset, six were reported in previous studies of age-of-onset [28,29], but the remaining 44 were novel associations for this phenotype. Conversely, of the 12 variants reported in previous studies, nine were associated in our GWAS, but three were not (in/near CYLD/NOD2 [29], CRBN [28] and ETS1 [28]). Possible explanations for the lack of association with these three variants in our GWAS is that their effect on age-of-onset is population specific or specific to asthma, the disease considered in the original studies. However, the three variants were also not significantly associated with age-of-onset when we restricted our analysis to cases who suffered only from asthma (P = 0.17, P = 0.64 and P = 0.26, respectively), which suggests that disease-specific effects are unlikely to explain the discordance.

When we compared the effect of the 50 variants on the age-of-onset of each of the three individual diseases, we found significant differences only for five variants. Four of these had a stronger effect on the age-of-onset of eczema: those in/near HRNR (rs1213821), FLG (rs61816761), TCHHL1 (rs115045402), SPRR2A (rs184587444), all within a 1 Mb locus on chromosome 1q21. The former two represent known risk factors for allergic disease, with a stronger effect on eczema [22], consistent with our results. On the other hand, the latter two variants, which are relatively uncommon (MAF of 2.8% and 2.0%), have not previously been established as risk factors for allergic disease, although our results for age-of-onset suggest that this is very likely to be the case. We did not find any eQTL in LD with either variant; on the other hand, both variants are in low to moderate LD with rs558269137 (r2 = 0.46 and 0.24, respectively), which encodes the FLG 2282del4 mutation that is associated with eczema and ichthyosis vulgaris [35]. It is therefore possible that at one (or both) of these variants are tagging that mutation, which was not tested in our study. The fifth variant, rs4795400 in GSDMB, showed a stronger effect on the age-of-onset of asthma. This variant is in high LD (r2>0.8) with variants reported to associate with earlier age-of-onset for asthma (rs9901146) [29] and which are stronger risk factors for asthma when compared to hay fever and eczema (rs921650) [22], consistent with our results. For the remaining 45 variants identified in our GWAS, our results suggest that their effect on age-of-onset is comparable between the three individual diseases.

We also investigated if the 50 variants that determined variation in age-of-onset amongst allergic disease cases also contributed to differences in case-control status amongst an independent sample of 222,484 individuals not part of the UK Biobank that we studied recently [22]. Remarkably, 39 of 48 variants with available results had a significant association with case-control status after accounting for multiple testing. Furthermore, for all 39 variants (and also for the other nine tested), the disease-predisposing allele was associated with a lower age-of-onset. These results suggested that case-control status and age-of-onset have a strong negative genetic correlation, which we confirmed (rg = -0.63) using genome-wide SNP data. We highlight two implications that arise from this observation.

First, this observation confirms that many genetic variants, including those identified in our age-of-onset GWAS, determine both the lifetime risk of developing an allergic disease as well as the age at which symptoms first develop. As such, combining information from these two phenotypes can help identify variants that influence disease liability, as suggested previously [29]. Motivated by this prediction, we performed multivariate analysis of results from our GWAS of age-of-onset and our recently published GWAS of allergic disease case-control status, which also considered information from asthma, hay fever and eczema. Importantly, we used a multivariate approach (metaUSAT [33]) that was expected to increase power to detect an association with a variant that influences both traits, when compared to other methods that are also applicable to GWAS summary statistics (e.g. metaCCA [36]). In this analysis, we identified 26 variants that were missed by the individual GWAS, highlighting the substantial gain in power that can be obtained by combining information from age-of-onset and case-control status. Of these 26 variants, only six were in LD (r2>0.05) with variants previously reported in GWAS of allergic disease. Therefore, most represent new associations for both age-of-onset and disease risk. Since we were not able to formally replicate these findings in an independent study, we emphasize the importance of future studies to replicate our results. We also suggest that this approach could be extended to include other phenotypes that can be shown to have a significant genetic correlation with disease risk; for example, these could be disease severity or markers of allergic sensitization.

Second, the large negative genetic correlation between case-control status and age-of-onset indicates that for most variants associated with both traits, the allele that is more common in allergic disease cases (when compared to controls) is also more common in cases with early onset disease (when compared to those with late onset disease). That is, individuals who inherit a larger overall burden of allergy-predisposing alleles are more likely to have early onset disease when compared to those who inherit a lower genetic burden, consistent with previous findings [37]. This shows that allergic disease risk alleles are more common in early onset disease, which might imply that allergic disease with late onset is less heritable (i.e. more ‘environmental’) than allergic disease with early onset. For example, it is conceivable that in late onset disease, environmental (more than genetic) risk factors dysregulate the expression of genes that play a key role in disease pathophysiology through epigenetic mechanisms, as we suggested recently [22]. But that may not necessarily be the case. Instead, it is possible that individuals develop late onset disease because they inherit risk alleles that influence asthma, hay fever and/or eczema pathophysiology through mechanisms that are not shared with early onset disease. Studies that address these possibilities are warranted. It is also important to highlight that we cannot rule out the possibility that recall bias might have contributed to the negative genetic correlation observed between age-of-onset and case-control status. This might have occurred if recall bias was less common amongst subjects who reported a younger age of onset.

We used information from eQTL studies and non-synonymous SNPs to identify 81 genes that are likely targets of 48 of the 76 (63%) variants identified in either the age-of-onset or multivariate GWAS performed. In the S1 Data (page 10–15), we discuss in greater detail 10 genes that are plausible targets of novel allergic disease variants identified in our study and that have a known function that is directly relevant to disease pathophysiology. In brief, the 10 genes are: ADAM15, a metalloproteinase which cleaves the toll like receptor adaptor molecule TRIF [38] and the low affinity IgE receptor [39]; FOSL2, a regulator of cell proliferation involved in B cell, Th17 cell and epidermal differentiation and function [4042]; TRIM8, a ligase involved in post-translational modifications of proteins, including ubiquitination of TAK1 [43] and TRIF [44]; BMPR2, a receptor for the TGF-beta superfamily [45] that inhibits Smad-mediated signaling [46]; CD200R1, a surface glycoprotein that interacts with CD200 [47], which is known to suppress the activation of various immune cells, including macrophages [48], mast cells [49], monocytes [50] and dendritic cells [51]; PRKCQ, a protein kinase involved in the development and function of Th17 cells [52], Th2 cells [53], Tregs [54] and type 2 innate lymphoid cells [55]; NOD2, an intracellular pattern recognition receptor that upon activation by bacterial peptidoglycans [56] and viruses [57] promotes host defense through the production of inflammatory mediators [5860]; SMAD4, a central regulator of TGF-beta signaling [61], involved in Th2 cytokine production [62], Treg [63] and Th17 differentiation [64], the expression of selectin ligands [65] and of the pro-allergic cytokine IL-9 [66]; ABCA7, a transporter protein that moves lipids across membranes [67], enhances phagocytosis of apoptotic cells by macrophages [68], promotes NKT cell development and function [69], and was suggested to play a role in keratinocyte differentiation [70]; and UBE2L3, an essential component of the post-translational protein ubiquitination pathway, which plays a major role in the regulation of inflammatory responses [7175].

The combined age-of-onset phenotype analysed did not take into account information from food allergy, as this was not available in the UK Biobank study. To partly address this limitation, we tested if the 76 sentinel variants identified in the age-of-onset or multivariate GWAS were also associated with food allergy, both in children and adults. After correcting for multiple testing, nine variants were significantly associated with food allergy confirmed by oral challenge in young children of the GOFA study [15], including one variant located in a locus not previously reported in food allergy GWAS: rs8056255 near CYLD. As such, this variant represents a putative novel risk factor for food allergy, which should be studied in greater detail in future studies. On the other hand, there was no evidence that the sentinel variants for age-of-onset identified in our study were associated with food allergies (based on hospital admissions) in adults of the GERA cohort. The lack of agreement between the food allergy results obtained in the GOFA and GERA studies raises the possibility that genetic risk factors for food allergy in children and adults might be largely distinct, which warrants further investigation.

Another potential limitation of our study is that age-of-onset reported by UK Biobank participants may have been affected by recall bias. For example, individuals with current disease symptoms at the time of data collection might have recalled early onset of disease more reliably than those who no longer suffered from allergies. We addressed this potential limitation by testing the association between the 50 sentinel variants identified in the age-of-onset GWAS in a subset of UK Biobank individuals who reported developing asthma as a child, specifically up to age 19. We found that the association between the 50 sentinel variants and age-of-onset in this smaller but more homogenous group of allergic disease cases was consistent with results obtained in the overall sample. Furthermore, we also found consistent associations when considering asthma onset recorded in children from the independent and prospective Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort. Similar results were observed for the 26 sentinel variants identified in the multivariate GWAS (S14 Table). Therefore, the 76 sentinel variants reported in our study show a consistent pattern of association with allergic disease age-of-onset in two analyses for which recall bias was not a major concern. Similarly, we found that phenotypic misclassification amongst individuals who reported late onset of allergic disease, if present, was unlikely to have significantly affected our main findings. In addition, the collection of information in adulthood is likely to have caused overrepresentation of SNPs involved in persistent disease and underrepresentation of association related to disease that remitted earlier in life (transient disease). Finally, we showed that in a subset of cases that provided data on two different occasions 4–7 years apart, age of onset of asthma was within 5 years in 86% of cases. However, we were not able to investigate this reliability for eczema and hayfever separately, since only a combined question was available. Furthermore, we acknowledge that only 3% of asthmatics in UKBB provided data on two different occasions. Thus, recall bias may have reduced our power, but not have resulted in spurious results.

In conclusion, we show that novel risk loci for allergic disease can be identified by extending the analytical approach that we reported recently [22] to the analysis of age-of-onset of asthma, hay fever and eczema. GWAS of other complex diseases might also benefit from considering age-of-onset information. We found 76 specific genetic associations with allergic disease, of which 28 had not previously been reported. We implicate 81 genes as likely targets of the associated variants and provide further evidence that individuals with early disease onset have a greater burden of genetic risk factors for allergic disease than individuals with late disease onset.

Methods

Definition of the combined age-of-onset phenotype and allergic disease status

We created a single age-of-onset phenotype for individuals from the UK Biobank study [76] that considered information from asthma, hay fever and eczema. Age-of-onset for food allergy was not available in the UK Biobank study and so was not considered in our analysis.

Specifically, we extracted information from two data fields included in the touchscreen questionnaire. Field 3786, which asked “What was your age when the asthma was first diagnosed?”, and field 3761, which asked “What was your age when the hayfever, rhinitis or eczema was first diagnosed?”. After excluding individuals who were not genotyped (absent from the array data), non-missing information was available for 50,109 and 98,161 individuals, respectively for fields 3786 and 3761. The combined age-of-onset phenotype corresponded to the earliest age reported across these two fields, which was obtained for 127,382 individuals. Of these, we restricted our analysis to individuals who were determined to suffer from at least one allergic disease (asthma and/or hay fever and/or eczema) based on the criteria described in detail previously [22]. Briefly, “asthma cases” were those with (i) both a report of “Asthma” in field 6152 (self-reported medical conditions, specifically the question “Has a doctor ever told you that you have had any of the following conditions?”) and a code for asthma in field 20002 (verbal interview), or alternatively, an ICD10 code for asthma in fields 41202 (Diagnoses–main ICD10) or 41204 (Diagnoses–secondary ICD10); and (ii) no report of COPD in fields 6152 or 20002, nor of other respiratory diseases in field 20002. “Hay fever/eczema cases” were those who answered “Hay fever, allergic rhinitis or eczema” in field 6152. Lastly, allergic disease cases were individuals who were classified as an “asthma case” and/or “hay fever/eczema case”, a total of 124,616 individuals.

To identify individuals who suffered from a single allergic disease, we had to classify hay fever and eczema status separately, which could not be determined using field 6152 per se (“Has a doctor ever told you that you have had any of the following conditions?”). This is because of the seven possible answers to this question, a single item covered the two different diseases, specifically the answer “Hayfever, allergic rhinitis or eczema”. To identify individuals who reported suffering specifically from hay fever, we instead considered information reported in the verbal interview (field 20002) and ICD10 codes (fields 41202 and 41204), as described above for asthma. This information was available for a subset of individuals who answered “Hayfever, allergic rhinitis or eczema” in field 6152. We then used the exact same approach (i.e. information from fields 20002, 41202 and 41204) to identify “eczema cases”.Using this approach, we were able to classify hay fever and eczema status separately, in addition to asthma, for 52,114 individuals. Of these, 975 suffered from all three diseases, 8,172 from two diseases and 42,967 from a single disease. Of the latter, 23,375 reported suffering from asthma only; 15,445 from hay fever only; and 4,147 from eczema only.

Association analysis of the combined age-of-onset phenotype

We first performed multi-dimensional scaling (MDS) analysis of allele sharing to identify individuals who clustered closely to Europeans of the 1000 Genomes project, as described previously [22]. Of the 124,616 individuals with a phenotype available for analysis, 117,530 clustered within 5 standard deviations of the mean for the first and second MDS components estimated using individuals from the five European ancestry groups (CEU, GBR, FIN, IBS and TSI) of the 1000 Genomes project. As such, these individuals were considered to have European ancestry and were retained for analysis. We then excluded 400 individuals who: (i) had self-reported sex different from genetically-inferred sex; (ii) were outliers when considering genotype missing rates and/or genome-wide heterozygosity levels; (iii) had more than 10 third degree relatives or were excluded from kinship inference; and/or (iv) were not present in the imputed dataset released in July 2017. The first three criteria were assessed based on information included in the QC file ukb_sqc_v2.txt released by the UK Biobank study. After these exclusions, phenotype and genotype data were available for 117,130 individuals. Of these, 22,029 reported suffering from asthma only; 14,474 from hay fever only; and 3,939 from eczema only.

To maximize power, we did not select a subset of unrelated individuals for analysis but instead tested variants for association with age-of-onset using the linear mixed model implemented in BOLT-LMM [77], which includes a genetic relationship matrix (GRM) as a random effect in the model. The GRM was estimated based on 577,110 array SNPs obtained after quality control filters, namely a minor allele frequency (MAF) >1%, call rate >95% and Hardy-Weinberg equilibrium P-value >10−6. Age-of-onset was quantile-normalized prior to analysis; gender and an indicator of the genotyping array used were included as discrete covariates.

Of the 92 million variants with imputed data released by the UK Biobank, we analysed 7,647,814 variants that (i) had a MAF >1%; (ii) were imputed based on the Haplotype Reference Consortium panel; (iii) had matching alleles when compared to genotype data from the 1000 Genomes project; and (iv) had a unique reference sequence (rs) number and genomic position (based on hg19). We used a P-value threshold of 3x10-8 for genome-wide significance, as suggested for studies that analyse variants with a MAF >1% [78].

Identification of variants with a statistically independent association with the combined age-of-onset phenotype

We used the approximate joint association analysis option of GCTA [79] to identify variants that remained associated with age-of-onset at a P<3x10-8 after accounting for the effects of nearby (<10 Mb) more strongly associated variants. In this analysis, LD was estimated based on a subset of 5,000 unrelated allergic disease cases from the UK Biobank study.

Association analysis of age-of-onset in individuals suffering from a single allergic disease

To understand if variants discovered through the analysis of the combined age-of-onset phenotype were likely to influence the age-of-onset of the three individual diseases considered (asthma, hay fever and eczema), we performed association analyses in adult cases from the UK Biobank study who reported suffering from a single allergic disease, as described in detail previously [22].

Specifically, we tested the association between selected variants and age-of-onset separately in three non-overlapping groups of individuals who suffered from a single allergic disease: asthma only cases (n = 22,029), hay fever only cases (n = 14,474), and eczema only cases (n = 3,969). These sample sizes are smaller than indicated above (23,375, 15,445 and 4,147, respectively) because individuals of non-European ancestry were not included in the association analysis. For each SNP, we then compared the effect on age-of-onset (i.e. beta from the linear model) between individual diseases (i.e. asthma vs. hay fever, asthma vs. eczema and hay fever vs. eczema), using the formula z = sigma / SE_sigma, where sigma = beta_diseaseA–beta_diseaseB, and SE_sigma = sqrt(SE_beta_diseaseA^2 + SE_beta_diseaseB^2), which follows a normal distribution.

Association between age-of-onset sentinel variants and allergic disease risk

We recently observed that many variants associated with the case-control status of allergic disease–defined by the presence of asthma, hay fever and/or eczema–were also associated with variation in age-of-onset amongst allergic disease cases [22]. We therefore reasoned that the reverse would also be likely: that many variants associated with variation in age-of-onset would be associated with disease liability, as suggested by Sarnowski et al. [29]. To test this, for each variant with a genome-wide significant association with age-of-onset in the analysis described above, we extracted association results from our recent GWAS of allergic disease [22]; after excluding overlapping samples (the UK Biobank study, n = 138,354), results were based on the analysis of 222,484 individuals, which included 137,883 cases with asthma and/or hay fever and/or eczema, and 84,601 disease-free controls. If a variant of interest was not directly tested in that GWAS, we extracted results for the most correlated proxy (with r2>0.8), if available.

Genetic correlation between age-of-onset and disease case-control status

To understand the extent to which the same genetic variants contribute to variation in disease age-of-onset and disease liability, we applied LD-score regression [32] to the association summary statistics obtained in the GWAS of the combined age-of-onset phenotype carried out as part of this study (n = 117,130) and in the GWAS of a combined allergic disease case-control phenotype (n = 360,838) that we reported recently [22]. The genetic correlation between the two GWAS (which is not biased by sample overlap) was estimated based on 1.1 million HapMap 3 SNPs, all with a MAF>1%, as recommended previously [32].

To illustrate graphically the observed genetic correlation between disease liability and age-of-onset, we compared the age-of-onset between individuals with a high and low polygenic burden of allergic disease risk SNPs. Specifically, for each individual in the UK Biobank study, we calculated a polygenic risk score (PRS) as the weighted average of the number of disease-predisposing alleles across the 136 sentinel variants that we identified in our recent GWAS of allergic disease case-control status [22]. Weights for each SNP corresponded to the allelic effect (i.e. beta) reported in that GWAS [22]. We then restricted our analysis to individuals that reported suffering from a single disease, as described above: asthma only cases (n = 22,029), hay fever only cases (n = 14,474), and eczema only cases (n = 3,969). Within each of these case groups, we identified individuals with a PRS in the top 10% and bottom 10% for that group, and then compared the distribution of age-of-onset between the two PRS groups using descriptive statistics.

Multivariate GWAS of allergic disease age-of-onset and case-control status

If two phenotypes have a genetic correlation that is significantly different from 0, then multivariate association analysis could potentially increase the power to identify variants that contribute to the heritability of both phenotypes, when compared to the alternative of testing each phenotype separately. Whether or not increased power is obtained with a multivariate test depends, for example, on the statistical approach used, the magnitude of the overall phenotypic correlation between the two phenotypes, as well as the magnitude and direction of the effect of the shared variant on the two phenotypes [33,80,81]. To analyse the joint association between single SNPs and allergic disease age-of-onset and case-control status, we used the recently described metaUSAT approach [33], for three main reasons. First, this approach is applicable to summary statistics from GWAS with unknown sample overlap. Second, it accommodates summary statistics from a mix of continuous and binary traits. And third, by combining two classes of tests (MANOVA and sum squared score tests), metaUSAT provides increased power over alternative methods when a SNP affects all phenotypes analysed [33]. MetaUSAT was applied to the summary statistics of the combined age-of-onset phenotype carried out as part of this study (n = 117,130) and the GWAS of a combined allergic disease case-control phenotype (n = 360,838) that we reported recently [22]. Because we were interested in identifying SNPs that were genome-wide significant (i.e. with P<3x10-8) in the multivariate but not in the separate univariate analyses, we first used approximate conditional analyses [79] to adjust the results from each GWAS for the effects of variants that had a statistically independent association at P<3x10-8 in the respective study (which were identified by the joint association analysis described above). After this adjustment, no single SNP had a P<3x10-8 in each of the two individual phenotype GWAS as expected. MetaUSAT was then applied to the adjusted GWAS; the Pearson correlation coefficient between z-scores of the two phenotypes was estimated based on SNPs not associated with either trait (i.e. with a P>0.05 for both).

A new multivariate test of association with similar properties to metaUSAT (e.g. applicable to summary statistics of GWAS with sample overlap), was reported just prior to the submission of this manuscript for publication–the MTAG approach [34]. We therefore tested if the multivariate associations discovered with metaUSAT were also supported by results from this different approach. For each SNP, instead of a multivariate P-value, MTAG returns an effect estimate (beta, SE and P-value) for each phenotype that incorporates information contained in the GWAS of the other phenotype. This increases the effective sample size of each analysis and so it improves the power to detect associations with variants that are shared between the two phenotypes [34].

Sentinel variants not previously implicated in the aetiology of allergic disease

To determine if a sentinel variant was in LD with a SNP previously reported to associate with any allergic disease, we (i) identified all SNPs in LD (r2>0.05) with that sentinel variant, using genotype data from individuals of European descent from the 1000 Genomes Project [82] (n = 294, release 20130502_v5a); and (ii) determined if the sentinel variant or any of the correlated SNPs identified were reported to associate with any allergic disease (asthma, hay fever, eczema, food allergy or atopy) in the NHGRI-EBI GWAS catalog database [83], which was downloaded on the 10th of January 2018. This catalogue was supplemented with recent studies not included at the time of submission of this manuscript. [2, 31, 84].

Predicting target genes of sentinel variants based on LD with eQTL and non-synonymous SNPs

We performed the following steps to identify genes for which variation in gene expression and/or protein sequence was associated with sentinel SNPs identified in the age-of-onset and multivariate GWAS.

First, we identified single nucleotide polymorphisms (SNPs) associated with variation in gene expression (i.e. expression quantitative trait loci (eQTL)) in published transcriptome studies of five broad tissue types relevant for allergic disease: individual immune cell types, lung, skin, spleen and whole-blood. We identified a total of 43 transcriptome studies reporting results from eQTL analyses in any one of those five tissue types (S1 Table). Some studies included multiple cell types, experimental conditions and/or eQTL types, resulting in a total of 101 separate eQTL datasets. For each eQTL dataset, we then (i) downloaded the original publication tables/files containing results for the eQTL reported; (ii) extracted the SNP identifier, gene name, association P-value and directional effect (if available; beta/z-score and effect allele); (iii) excluded eQTL located >1 Mb of the respective gene (i.e. trans eQTL), because often these are thought to be mediated by cis effects [85]; (iv) excluded eQTL with an association P>8.9x10-10, a conservative threshold that corrects for 55,765 genes (based on GENCODE v19), each tested for association with 1,000 SNPs (as suggested by others [8688]); and (v) for each gene, used the—clump procedure in PLINK [89] to reduce the list of eQTL identified (which often included many correlated SNPs) to a set of ‘sentinel eQTL’, defined as the SNPs with strongest association with gene expression and in low LD (r2<0.05, linkage disequilibrium (LD) window of 2 Mb) with each other.

Second, we identified genes for which a sentinel eQTL reported in any of the 101 eQTL datasets described above was in high LD (r2>0.8) with a sentinel variant identified in the age-of-onset or multivariate GWAS. That is, we only considered genes for which there was high LD between a sentinel eQTL and a sentinel allergic disease variant, which reduces the chance of spurious co-localization.

Third, we used wANNOVAR [90] to identify genes containing non-synonymous SNPs amongst all variants in LD (r2>0.8) with any sentinel variants. SNPs in LD with sentinel variants were identified using genotype data from individuals of European descent from the 1000 Genomes Project [83] (n = 294, release 20130502_v5a).

Association between sentinel variants and the risk of food allergy

Age-of-onset of food allergy was not available in the UK Biobank study and so it was not considered in our combined allergic disease phenotype. To partly address this limitation, we determined if variants identified in the age-of-onset or multivariate analysis also contribute to food allergy risk. Specifically, we extracted association results for selected variants from a GWAS of food allergy conducted in the GOFA cohort that we published recently [15], which included 497 children with a positive oral food challenge and 2,387 controls (the genomic inflation factor [λ] for this GWAS was 1.03). All cases were recruited at the time of diagnosis (mean age 2.1; 84% under the age of 4). The most common food allergy was observed with hen’s egg (58%), followed by peanut (44%) and cow’s milk (34%). To study the association between each variant and food allergy in adults, we extracted association results from a GWAS of food allergy conducted in the GERA cohort, which we also reported recently [16], including 1,104 adults with an ICD9 (i.e. hospital admission) code for food allergy and 23,945 controls. The genomic inflation factor for this GWAS was 1.01. Most cases were admitted to hospital because of an allergic reaction to fish/shellfish (58%), peanuts (16%), eggs (13%) or milk (8%).

This study was approved by the Human Ethics Committee of the QIMR Berghofer Medical Research Institute. Approval of individual studies has been reported in Ferreira et al. [22].

Supporting information

S1 Data [doc]
Potential impact of recall bias on SNP associations with age-of-onset of allergic disease (page 4–6) Potential impact on SNP associations of phenotypic misclassification amongst individuals reporting late onset disease (page 7–9).

S1 Fig [docx]
Distribution of allergic disease age-of-onset in UK Biobank participants (n = 117,130) who reported suffering from asthma and/or hay fever/eczema.

S2 Fig [docx]
Distribution of the observed and expected association P-values for the GWAS of allergic disease age-of-onset in the UK Biobank study (n = 117,130).

S3 Fig [docx]
Reliability of self-reported age-of-onset of asthma based on information provided by 1,650 UK Biobank participants at two time points.

S4 Fig [docx]
Association between sentinel SNPs identified in the age-of-onset (Panel A) or multivariate GWAS (Panel B) and asthma age-of-onset in the subset of UK Biobank individuals who reported developing asthma as a child.

S5 Fig [docx]
Association between sentinel SNPs identified in the age-of-onset (Panel A) or multivariate GWAS (Panel B) and hay fever age-of-onset in the subset of UK Biobank individuals who reported developing hay fever as a child.

S6 Fig [docx]
Association between sentinel SNPs identified in the age-of-onset (Panel A) or multivariate GWAS (Panel B) and time to asthma onset in children of the ALSPAC study.

S7 Fig [docx]
Summary of results from the GWAS of allergic disease age-of-onset in the UK Biobank study (n = 117,130), after adjusting single-SNP results for the effects of independently associated variants (i.e. with P<3x10 in the joint association analysis performed with GCTA.

S8 Fig [docx]
Summary of results from the GWAS of allergic disease case-control status (n = 360,838) after adjusting single-SNP results for the effects of independently associated variants (i.e. with P<3x10 in the joint association analysis performed with GCTA).

S9 Fig [docx]
Distribution of the observed and expected association P values for the GWAS of allergic disease age-of-onset in the UK Biobank study (n = 117,130), after adjusting single-SNP results for the effects of independently associated variants (i.e. with P<3x10 in the joint association analysis performed with GCTA.

S10 Fig [docx]
Distribution of the observed and expected association P values for the GWAS of allergic disease case-control status (n = 360,838) after adjusting single-SNP results for the effects of independently associated variants (i.e. with P<3x10 in the joint association analysis performed with GCTA.

S11 Fig [docx]
Distribution of the observed and expected association P values obtained in the multivariate analysis of the GWAS of allergic disease risk (n = 360,838) and GWAS of allergic disease age-of-onset (n = 117,130).

S12 Fig [docx]
Association between sentinel SNPs identified in the age-of-onset (Panel A) or multivariate GWAS (Panel B) and food allergy in children.

S1 Table [xls]
Genome-wide association studies of gene expression levels queried to identify expression quantitative trait loci (i.e. eQTL).

S2 Table [xls]
Descriptive statistics for the 117,130 allergic disease cases from the UK Biobank included in this study.

S3 Table [xls]
Results from joint association analysis performed with GCTA .

S4 Table [xls]
Variants associated with allergic disease age-of-onset at a P<3x10 in the joint association analysis but not in the original analysis.

S5 Table [xls]
Association with allergic disease age-of-onset for variants previously reported in the literature.

S6 Table [xls]
Association between the 50 sentinel variants and age-of-onset of asthma, hay fever and eczema separately.

S7 Table [xls]
Sentinel SNPs in LD (r>0.05) with variants previously reported to associate with at least one allergic disease.

S8 Table [xls]
Results from the MTAG multivariate approach for the sentinel variants identified in the metaUSAT multivariate analysis of allergic-disease age-of-onset and allergic disease case-control status.

S9 Table [xls]
Sentinel eQTL in LD (r>0.5) with sentinel variants associated with allergic disease age-of-onset.

S10 Table [xls]
Sentinel eQTL in LD (r>0.8) with sentinel variants jointly associated with allergic disease age-of-onset and allergic-disease risk.

S11 Table [xls]
Non-synonymous SNPs in LD (r>0.8) with sentinel variants associated with allergic disease.

S12 Table [xls]
Two independent associations with self-reported food allergy case-control status at a P<3x10.

S13 Table [xls]
Association between the 76 sentinel SNPs and self-reported food allergy case-control status.

S14 Table [xls]
Results of replication study of age-of-onset GWAS in ALSPAC.


Zdroje

1. Vicente C.T., Revez J.A. & Ferreira M.A.R. Lessons from ten years of genome-wide association studies of asthma. Clinical and Translational Immunology 6, e165 (2017). doi: 10.1038/cti.2017.54 29333270

2. Waage J., et al. Genome-wide association and HLA fine-mapping studies identify risk loci and genetic pathways underlying allergic rhinitis. Nat Genet. 2018 Aug;50(8):1072–1080. doi: 10.1038/s41588-018-0157-1 30013184

3. Bunyavanich S. et al. Integrated genome-wide association, coexpression network, and expression single nucleotide polymorphism analysis identifies novel pathway in allergic rhinitis. BMC Med Genomics 7, 48 (2014). doi: 10.1186/1755-8794-7-48 25085501

4. Andiappan A.K. et al. Genome-wide association study for atopy and allergic rhinitis in a Singapore Chinese population. PLoS One 6, e19719 (2011). doi: 10.1371/journal.pone.0019719 21625490

5. Weidinger S. et al. A genome-wide association study of atopic dermatitis identifies loci with overlapping effects on asthma and psoriasis. Hum Mol Genet 22, 4841–56 (2013). doi: 10.1093/hmg/ddt317 23886662

6. Esparza-Gordillo J. et al. A common variant on chromosome 11q13 is associated with atopic dermatitis. Nat Genet 41, 596–601 (2009). doi: 10.1038/ng.347 19349984

7. Sun L.D. et al. Genome-wide association study identifies two new susceptibility loci for atopic dermatitis in the Chinese Han population. Nat Genet 43, 690–4 (2011). doi: 10.1038/ng.851 21666691

8. Hirota T. et al. Genome-wide association study identifies eight new susceptibility loci for atopic dermatitis in the Japanese population. Nat Genet 44, 1222–6 (2012). doi: 10.1038/ng.2438 23042114

9. Baurecht H. et al. Genome-wide comparative analysis of atopic dermatitis and psoriasis gives insight into opposing genetic mechanisms. Am J Hum Genet 96, 104–20 (2015). doi: 10.1016/j.ajhg.2014.12.004 25574825

10. Schaarschmidt H. et al. A genome-wide association study reveals 2 new susceptibility loci for atopic dermatitis. J Allergy Clin Immunol 136, 802–6 (2015). doi: 10.1016/j.jaci.2015.01.047 25865352

11. Paternoster L. et al. Multi-ancestry genome-wide association study of 21,000 cases and 95,000 controls identifies new risk loci for atopic dermatitis. Nat Genet 47, 1449–56 (2015). doi: 10.1038/ng.3424 26482879

12. Kim K.W. et al. Genome-wide association study of recalcitrant atopic dermatitis in Korean children. J Allergy Clin Immunol 136, 678–684 e4 (2015). doi: 10.1016/j.jaci.2015.03.030 25935106

13. Hong X. et al. Genome-wide association study identifies peanut allergy-specific loci and evidence of epigenetic mediation in US children. Nat Commun 6, 6304 (2015). doi: 10.1038/ncomms7304 25710614

14. Martino D.J. et al. Genomewide association study of peanut allergy reproduces association with amino acid polymorphisms in HLA-DRB1. Clin Exp Allergy 47, 217–223 (2017). doi: 10.1111/cea.12863 27883235

15. Marenholz I. et al. Genome-wide association study identifies the SERPINB gene cluster as a susceptibility locus for food allergy. Nat Commun 8, 1056 (2017). doi: 10.1038/s41467-017-01220-0 29051540

16. Asai Y. et al. Genome-wide association study and meta-analysis in multiple populations identifies new loci for peanut allergy and establishes C11orf30/EMSY as a genetic risk factor for food allergy. J Allergy Clin Immunol 141, 991–1001 (2018). doi: 10.1016/j.jaci.2017.09.015 29030101

17. Marenholz I. et al. Meta-analysis identifies seven susceptibility loci involved in the atopic march. Nat Commun 6, 8804 (2015). doi: 10.1038/ncomms9804 26542096

18. Ferreira M.A. et al. Genome-wide association analysis identifies 11 risk variants associated with the asthma with hay fever phenotype. J Allergy Clin Immunol 133, 1564–71 (2014). doi: 10.1016/j.jaci.2013.10.030 24388013

19. Hinds D.A. et al. A genome-wide association meta-analysis of self-reported allergy identifies shared and allergy-specific susceptibility loci. Nat Genet 45, 907–11 (2013). doi: 10.1038/ng.2686 23817569

20. Wan Y.I. et al. A genome-wide association study to identify genetic determinants of atopy in subjects from the United Kingdom. J Allergy Clin Immunol 127, 223–31, 231 e1-3 (2011). doi: 10.1016/j.jaci.2010.10.006 21094521

21. Bønnelykke K. et al. Meta-analysis of genome-wide association studies identifies ten loci influencing allergic sensitization. Nat Genet. 2013 Aug;45(8):902–906. doi: 10.1038/ng.2694 23817571

22. Ferreira M.A. et al. Shared genetic origin of asthma, hay fever and eczema elucidates allergic disease biology. Nat Genet 49, 1752–1757 (2017). doi: 10.1038/ng.3985 29083406

23. Thomsen SF, et al. Genetic influence on the age at onset of asthma: a twin study. J Allergy Clin Immunol. 2010 Sep;126(3):626–30. doi: 10.1016/j.jaci.2010.06.017 20673982

24. Moffatt MF, et al. Genetic variants regulating ORMDL3 expression contribute to the risk of childhood asthma. Nature. 2007 Jul 26;448(7152):470–3. doi: 10.1038/nature06014 17611496

25. Bouzigon E, et al. Effect of 17q21 variants and smoking exposure in early-onset asthma. N Engl J Med. 2008 Nov 6;359(19):1985–94. doi: 10.1056/NEJMoa0806604 18923164

26. Moffatt MF, et al. A large-scale, consortium-based genomewide association study of asthma. N Engl J Med. 2010 Sep 23;363(13):1211–1221. doi: 10.1056/NEJMoa0906312 20860503

27. Halapi E, et al. A sequence variant on 17q21 is associated with age at onset and severity of asthma. Eur J Hum Genet. 2010 Aug;18(8):902–8 doi: 10.1038/ejhg.2010.38 20372189

28. Forno E. et al. Genome-wide association study of the age of onset of childhood asthma. J Allergy Clin Immunol 130, 83–90 e4 (2012). doi: 10.1016/j.jaci.2012.03.020 22560479

29. Sarnowski C. et al. Identification of a new locus at 16q12 associated with time to asthma onset. J Allergy Clin Immunol 138, 1071–1080 (2016). doi: 10.1016/j.jaci.2016.03.018 27130862

30. Bulik-Sullivan B.K. et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet (2015).

31. Demenais F, et al. Multiancestry association study identifies new asthma risk loci that colocalize with immune-cell enhancer marks. Nat Genet. 2018 Jan;50(1):42–53 doi: 10.1038/s41588-017-0014-7 29273806

32. Bulik-Sullivan B. et al. An atlas of genetic correlations across human diseases and traits. Nat Genet 47, 1236–41 (2015). doi: 10.1038/ng.3406 26414676

33. Ray D, Boehnke M.Methods for meta-analysis of multiple traits using GWAS summary statistics. Genet Epidemiol. 2018 Mar;42(2):134–145. doi: 10.1002/gepi.22105 29226385

34. Turley P. et al. Multi-trait analysis of genome-wide association summary statistics using MTAG. Nat Genet Nat Genet. 2018 Feb;50(2):229–237. doi: 10.1038/s41588-017-0009-4 29292387

35. Smith F.J. et al. Loss-of-function mutations in the gene encoding filaggrin cause ichthyosis vulgaris. Nat Genet 38, 337–42 (2006). doi: 10.1038/ng1743 16444271

36. Cichonska A. et al. metaCCA: summary statistics-based multivariate meta-analysis of genome-wide association studies using canonical correlation analysis. Bioinformatics 32, 1981–9 (2016). doi: 10.1093/bioinformatics/btw052 27153689

37. Belsky D.W. et al. Polygenic risk and the development and course of asthma: an analysis of data from a four-decade longitudinal study. Lancet Respir Med 1, 453–61 (2013). doi: 10.1016/S2213-2600(13)70101-2 24429243

38. Ahmed S., Maratha A., Butt A.Q., Shevlin E. & Miggin S.M. TRIF-mediated TLR3 and TLR4 signaling is negatively regulated by ADAM15. J Immunol 190, 2217–28 (2013). doi: 10.4049/jimmunol.1201630 23365087

39. Fourie A.M., Coles F., Moreno V. & Karlsson L. Catalytic activity of ADAM8, ADAM15, and MDC-L (ADAM28) on synthetic peptide substrates and in ectodomain cleavage of CD23. J Biol Chem 278, 30469–77 (2003). doi: 10.1074/jbc.M213157200 12777399

40. Ubieta K. et al. Fra-2 regulates B cell development by enhancing IRF4 and Foxo1 transcription. J Exp Med 214, 2059–2071 (2017). doi: 10.1084/jem.20160514 28566276

41. Wurm S. et al. Terminal epidermal differentiation is regulated by the interaction of Fra-2/AP-1 with Ezh2 and ERK1/2. Genes Dev 29, 144–56 (2015). doi: 10.1101/gad.249748.114 25547114

42. Ciofani M. et al. A validated regulatory network for Th17 cell specification. Cell 151, 289–303 (2012). doi: 10.1016/j.cell.2012.09.016 23021777

43. Li Q. et al. Tripartite motif 8 (TRIM8) modulates TNFalpha- and IL-1beta-triggered NF-kappaB activation by targeting TAK1 for K63-linked polyubiquitination. Proc Natl Acad Sci U S A 108, 19341–6 (2011). doi: 10.1073/pnas.1110946108 22084099

44. Ye W. et al. TRIM8 Negatively Regulates TLR3/4-Mediated Innate Immune Response by Blocking TRIF-TBK1 Interaction. J Immunol 199, 1856–1864 (2017). doi: 10.4049/jimmunol.1601647 28747347

45. Rosenzweig B.L. et al. Cloning and characterization of a human type II receptor for bone morphogenetic proteins. Proc Natl Acad Sci U S A 92, 7632–6 (1995). doi: 10.1073/pnas.92.17.7632 7644468

46. Rudarakanchana N. et al. Functional analysis of bone morphogenetic protein type II receptor mutations underlying primary pulmonary hypertension. Hum Mol Genet 11, 1517–25 (2002). doi: 10.1093/hmg/11.13.1517 12045205

47. Wright G.J. et al. Lymphoid/neuronal cell surface OX2 glycoprotein recognizes a novel receptor on macrophages implicated in the control of their function. Immunity 13, 233–42 (2000). doi: 10.1016/s1074-7613(00)00023-6 10981966

48. Hoek R.M. et al. Down-regulation of the macrophage lineage through interaction with OX2 (CD200). Science 290, 1768–71 (2000). doi: 10.1126/science.290.5497.1768 11099416

49. Cherwinski H.M. et al. The CD200 receptor is a novel and potent regulator of murine and human mast cell function. J Immunol 174, 1348–56 (2005). doi: 10.4049/jimmunol.174.3.1348 15661892

50. Jenmalm M.C., Cherwinski H., Bowman E.P., Phillips J.H. & Sedgwick J.D. Regulation of myeloid cell function through the CD200 receptor. J Immunol 176, 191–9 (2006). doi: 10.4049/jimmunol.176.1.191 16365410

51. Fallarino F. et al. Murine plasmacytoid dendritic cells initiate the immunosuppressive pathway of tryptophan catabolism in response to CD200 receptor engagement. J Immunol 173, 3748–54 (2004). doi: 10.4049/jimmunol.173.6.3748 15356121

52. Sen S. et al. SRC1 promotes Th17 differentiation by overriding Foxp3 suppression to stimulate RORgammat activity in a PKC-theta-dependent manner. Proc Natl Acad Sci U S A 115, E458–E467 (2018). doi: 10.1073/pnas.1717789115 29282318

53. Salek-Ardakani S., So T., Halteman B.S., Altman A. & Croft M. Differential regulation of Th2 and Th1 lung inflammatory responses by protein kinase C theta. J Immunol 173, 6440–7 (2004). doi: 10.4049/jimmunol.173.10.6440 15528385

54. Gupta S. et al. Differential requirement of PKC-theta in the development and function of natural regulatory T cells. Mol Immunol 46, 213–24 (2008). doi: 10.1016/j.molimm.2008.08.275 18842300

55. Madouri F. et al. Protein kinase Ctheta controls type 2 innate lymphoid cell and TH2 responses to house dust mite allergen. J Allergy Clin Immunol 139, 1650–1666 (2017). doi: 10.1016/j.jaci.2016.08.044 27746240

56. Girardin S.E. et al. Nod2 is a general sensor of peptidoglycan through muramyl dipeptide (MDP) detection. J Biol Chem 278, 8869–72 (2003). doi: 10.1074/jbc.C200651200 12527755

57. Sabbah A. et al. Activation of innate immune antiviral responses by Nod2. Nat Immunol 10, 1073–80 (2009). doi: 10.1038/ni.1782 19701189

58. Maeda S. et al. Nod2 mutation in Crohn's disease potentiates NF-kappaB activity and IL-1beta processing. Science 307, 734–8 (2005). doi: 10.1126/science.1103685 15692052

59. Netea M.G. et al. NOD2 mediates anti-inflammatory signals induced by TLR2 ligands: implications for Crohn's disease. Eur J Immunol 34, 2052–9 (2004). doi: 10.1002/eji.200425229 15214053

60. Kobayashi K.S. et al. Nod2-dependent regulation of innate and adaptive immunity in the intestinal tract. Science 307, 731–4 (2005). doi: 10.1126/science.1104911 15692051

61. Zhang Y., Feng X., We R. & Derynck R. Receptor-associated Mad homologues synergize as effectors of the TGF-beta response. Nature 383, 168–72 (1996). doi: 10.1038/383168a0 8774881

62. Kim B.G. et al. Smad4 signalling in T cells is required for suppression of gastrointestinal cancer. Nature 441, 1015–9 (2006). doi: 10.1038/nature04846 16791201

63. Hahn J.N., Falck V.G. & Jirik F.R. Smad4 deficiency in T cells leads to the Th17-associated development of premalignant gastroduodenal lesions in mice. J Clin Invest 121, 4030–42 (2011). doi: 10.1172/JCI45114 21881210

64. Zhang S. et al. Reversing SKI-SMAD4-mediated suppression is essential for TH17 cell differentiation. Nature 551, 105–109 (2017). doi: 10.1038/nature24283 29072299

65. Ebel M.E. & Kansas G.S. Functions of Smad Transcription Factors in TGF-beta1-Induced Selectin Ligand Expression on Murine CD4 Th Cells. J Immunol 197, 2627–34 (2016). doi: 10.4049/jimmunol.1600723 27543612

66. Wang A. et al. Cutting edge: Smad2 and Smad4 regulate TGF-beta-mediated Il9 gene expression via EZH2 displacement. J Immunol 191, 4908–12 (2013). doi: 10.4049/jimmunol.1300433 24108699

67. Abe-Dohmae S., Ueda K. & Yokoyama S. ABCA7, a molecule with unknown function. FEBS Lett 580, 1178–82 (2006). doi: 10.1016/j.febslet.2005.12.029 16376881

68. Jehle A.W. et al. ATP-binding cassette transporter A7 enhances phagocytosis of apoptotic cells and associated ERK signaling in macrophages. J Cell Biol 174, 547–56 (2006). doi: 10.1083/jcb.200601030 16908670

69. Nowyhed H.N. et al. ATP Binding Cassette Transporter ABCA7 Regulates NKT Cell Development and Function by Controlling CD1d Expression and Lipid Raft Content. Sci Rep 7, 40273 (2017). doi: 10.1038/srep40273 28091533

70. Kielar D. et al. Adenosine triphosphate binding cassette (ABC) transporters are expressed and regulated during terminal keratinocyte differentiation: a potential role for ABCA7 in epidermal lipid reorganization. J Invest Dermatol 121, 465–74 (2003). doi: 10.1046/j.1523-1747.2003.12404.x 12925201

71. Bednash J.S. & Mallampalli R.K. Regulation of inflammasomes by ubiquitination. Cell Mol Immunol 13, 722–728 (2016). doi: 10.1038/cmi.2016.15 27063466

72. Fu B., Li S., Wang L., Berman M.A. & Dorf M.E. The ubiquitin conjugating enzyme UBE2L3 regulates TNFalpha-induced linear ubiquitination. Cell Res 24, 376–9 (2014). doi: 10.1038/cr.2013.133 24060851

73. Simmons A. et al. Nef-mediated lipid raft exclusion of UbcH7 inhibits Cbl activity in T cells to positively regulate signaling. Immunity 23, 621–34 (2005). doi: 10.1016/j.immuni.2005.11.003 16356860

74. Kathania M. et al. Ndfip1 regulates itch ligase activity and airway inflammation via UbcH7. J Immunol 194, 2160–7 (2015). doi: 10.4049/jimmunol.1402742 25632008

75. Eldridge M.J.G., Sanchez-Garrido J., Hoben G.F., Goddard P.J. & Shenoy A.R. The Atypical Ubiquitin E2 Conjugase UBE2L3 Is an Indirect Caspase-1 Target and Controls IL-1beta Secretion by Inflammasomes. Cell Rep 18, 1285–1297 (2017). doi: 10.1016/j.celrep.2017.01.015 28147281

76. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, Motyer A, Vukcevic D, Delaneau O, O'Connell J, Cortes A, Welsh S, Young A, Effingham M, McVean G, Leslie S, Allen N, Donnelly P, Marchini J. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018 Oct;562(7726):203–209 doi: 10.1038/s41586-018-0579-z 30305743

77. Loh P.R. et al. Efficient Bayesian mixed-model analysis increases association power in large cohorts. Nat Genet 47, 284–90 (2015). doi: 10.1038/ng.3190 25642633

78. Fadista J., Manning A.K., Florez J.C. & Groop L. The (in)famous GWAS P-value threshold revisited and updated for low-frequency variants. Eur J Hum Genet 24, 1202–5 (2016). doi: 10.1038/ejhg.2015.269 26733288

79. Yang J. et al. Conditional and joint multiple-SNP analysis of GWAS summary statistics identifies additional variants influencing complex traits. Nat Genet 44, 369–75, S1-3 (2012). doi: 10.1038/ng.2213 22426310

80. Ferreira M.A. & Purcell S.M. A multivariate test of association. Bioinformatics 25, 132–3 (2009). doi: 10.1093/bioinformatics/btn563 19019849

81. O'Reilly P.F. et al. MultiPhen: joint model of multiple phenotypes can increase discovery in GWAS. PLoS One 7, e34861 (2012). doi: 10.1371/journal.pone.0034861 22567092

82. Genomes Project C. et al. An integrated map of genetic variation from 1,092 human genomes. Nature 491, 56–65 (2012). doi: 10.1038/nature11632 23128226

83. Welter D. et al. The NHGRI GWAS Catalog, a curated resource of SNP-trait associations. Nucleic Acids Res 42, D1001–6 (2014). doi: 10.1093/nar/gkt1229 24316577

84. Ferreira MAR et al. Genetic Architectures of Childhood- and Adult-Onset Asthma Are Partly Distinct. Am J Hum Genet. 2019 Apr 4;104(4):665–684. doi: 10.1016/j.ajhg.2019.02.022 30929738

85. Pierce B.L. et al. Mediation analysis demonstrates that trans-eQTLs are often explained by cis-mediation: a genome-wide analysis among 1,800 South Asians. PLoS Genet 10, e1004818 (2014). doi: 10.1371/journal.pgen.1004818 25474530

86. Davis J.R. et al. An Efficient Multiple-Testing Adjustment for eQTL Studies that Accounts for Linkage Disequilibrium between Variants. Am J Hum Genet 98, 216–24 (2016). doi: 10.1016/j.ajhg.2015.11.021 26749306

87. Lappalainen T. et al. Transcriptome and genome sequencing uncovers functional variation in humans. Nature 501, 506–11 (2013). doi: 10.1038/nature12531 24037378

88. Montgomery S.B. et al. Transcriptome genetics using second generation sequencing in a Caucasian population. Nature 464, 773–7 (2010). doi: 10.1038/nature08903 20220756

89. Chang C.C. et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience 4, 7 (2015). doi: 10.1186/s13742-015-0047-8 25722852

90. Chang X. & Wang K. wANNOVAR: annotating genetic variants for personal genomes via the web. J Med Genet 49, 433–6 (2012). doi: 10.1136/jmedgenet-2012-100918 22717648


Článek vyšel v časopise

PLOS Genetics


2020 Číslo 6
Nejčtenější tento týden
Nejčtenější v tomto čísle
Kurzy

Zvyšte si kvalifikaci online z pohodlí domova

Důležitost adherence při depresivním onemocnění
nový kurz
Autoři: MUDr. Eliška Bartečková, Ph.D.

Koncepce osteologické péče pro gynekology a praktické lékaře
Autoři: MUDr. František Šenk

Sekvenční léčba schizofrenie
Autoři: MUDr. Jana Hořínková, Ph.D.

Hypertenze a hypercholesterolémie – synergický efekt léčby
Autoři: prof. MUDr. Hana Rosolová, DrSc.

Multidisciplinární zkušenosti u pacientů s diabetem
Autoři: Prof. MUDr. Martin Haluzík, DrSc., prof. MUDr. Vojtěch Melenovský, CSc., prof. MUDr. Vladimír Tesař, DrSc.

Všechny kurzy
Přihlášení
Zapomenuté heslo

Zadejte e-mailovou adresu, se kterou jste vytvářel(a) účet, budou Vám na ni zaslány informace k nastavení nového hesla.

Přihlášení

Nemáte účet?  Registrujte se

#ADS_BOTTOM_SCRIPTS#