Multitrait genetic association analysis identifies 50 new risk loci for gastro-oesophageal reflux, seven new loci for Barrett's oesophagus and provides insights into clinical heterogeneity in reflux diagnosis

Abstract

Objective: Gastro-oesophageal reflux disease (GERD) has heterogeneous aetiology primarily attributable to its symptom-based definitions. GERD genome-wide association studies (GWASs) have shown strong genetic overlaps with established risk factors such as obesity and depression. We hypothesised that the shared genetic architecture between GERD and these risk factors can be leveraged to (1) identify new GERD and Barrett's oesophagus (BE) risk loci and (2) explore potentially heterogeneous pathways leading to GERD and oesophageal complications.

Design: We applied multitrait GWAS models combining GERD (78 707 cases; 288 734 controls) and genetically correlated traits including education attainment, depression and body mass index. We also used multitrait analysis to identify BE risk loci. Top hits were replicated in 23andMe (462 753 GERD cases, 24 099 BE cases, 1 484 025 controls). We additionally dissected the GERD loci into obesity-driven and depression-driven subgroups. These subgroups were investigated to determine how they relate to tissue-specific gene expression and to risk of serious oesophageal disease (BE and/or oesophageal adenocarcinoma, EA).

Results: We identified 88 loci associated with GERD, with 59 replicating in 23andMe after multiple testing corrections. Our BE analysis identified seven novel loci. Additionally we showed that only the obesity-driven GERD loci (but not the depression-driven loci) were associated with genes enriched in oesophageal tissues and successfully predicted BE/EA.

Conclusion: Our multitrait model identified many novel risk loci for GERD and BE. We present strong evidence for a genetic underpinning of disease heterogeneity in GERD and show that GERD loci associated with depressive symptoms are not strong predictors of BE/EA relative to obesity-driven GERD loci.

Keywords: Barrett's oesophagus; gastro-esophageal reflux disease; genetics; oesophageal reflux.

Figure 1

Schematic diagram describing overall study approach of the multi-trait GWAS analysis for GERD and BE susceptibility. GWAS data obtained from published studies are shown in bold. BE, Barrett’s oesophagus; EA, esophageal adenocarcinoma; EDU, Education Attainment; GERD, gastro-oesophageal reflux disease; GWAS, genome-wide association study; MDD, major depressive disorder; MTAG, multitrait analysis of GWAS; PGC, Psychiatric Genomics Consortium; QSKIN, Queensland Sun and Health Study; SSGAC, Social Science Genetics Association Consortium; UKBB, UK Biobank. Traits within the blue/red boxes are traits selected for the multi-trait GWAS analysis for GERD (blue) and BE (red). The trait BE was not modelled in the MTAG model for GERD, to avoid sample overlap bias in the genetic prediction analysis for GERD into the BE/EA datasets. Asterisk (*) highlights genetic correlation estimates for each trait against GERD (shown by the blue arrows) obtained from previous An et al findings.

Figure 2

Manhattan plot for the multi-trait meta-analysed GWAS of GERD and BE. (A) for GERD; (B) for BE. The red dotted line indicates the genome-wide significance threshold at p=5×10^-8. Single nucleotide polymorphisms (SNPs) with p values <0.01 were not shown. Loci with GERD association exceeding this threshold are deemed as a genome-wide GERD loci and followed up with the replication analysis in 23andMe. Each of the genome-wide loci is denoted with a red dot followed by the name of the relevant gene mapped through location proximity and expression quantiative trait loci (eQTL) information. BE, Barrett’s oesophagus; GERD, gastro-oesophageal reflux disease; GWAS, genome-wide association study.

Figure 3

The circular Manhattan plot for BMI, MDD and the multitrait (MTAG) GERD GWAS. The trait from the innermost circle belongs to the (GIANT+UKB) Yengo et al. BMI GWAS, followed by the Howard et al. (excluding 23andMe) MDD GWAS and the result from our multi-trait GERD GWAS. The plot illustrates different patterns of colocalisation between GERD and MDD/BMI for some regions, contributing to the genetic heterogeneity within GERD. BMI, body mass index; GERD, gastro-oesophageal reflux disease; GWAS, genome-wide association study; MDD, major depressivedisorder;MTAG, multitrait analysis of GWAS.

Figure 4

Replication of GERD and BE association for the genome-wide significant loci from MTAG analysis on GERD and BE in the 23andMe cohort. A refers to the findings for GERD; (B) for BE. Data points that are shaded in blue are those that have a Bonferroni corrected p value below 0.05/88 for GERD and 0.05/17 for BE. Both the x- and y-axes represent log(OR) for GERD/BE and points are plotted with error bars representing one SE. Most of our GERD and BE loci showed strong evidence of being replicated, although the estimated effect size for GERD/BE in the 23andMe cohort were on average smaller than those estimated in the MTAG analysis (slope ~0.5 for both traits). BE, Barrett’s oesophagus; GERD, gastro-oesophageal reflux disease; MTAG, multitrait analysis of genome-wide association studies, SNP, single nucleotide polymorphism.

Figure 5

MAGMA gene-based tissue enrichment analysis for categorised functional and non-functional GERD gene sets on 53 human tissues. The dotted line represents the Bonferroni corrected significance threshold. While both the obesity-driven and neuropsychiatric-driven GERD genes were differentially expressed in brain tissues, only the regulation of gene expression in oesophageal tissues were detected from the obesity GERD gene set. Another observation is that the pattern of regulation is more pleiotropic across tissues for the obesity GERD set, consistent with the complex architecture of adiposity. The results for up-ward and down-ward regulation of gene expression can be viewed in online supplemental materials. GERD, gastro-oesophageal reflux disease

Figure 6

Estimation of the genetic association between GERD and BE/EA, stratified by genetic GERD subtypes. Each of the slopes represents the estimated magnitude of association between per unit increase in log(OR) of GERD on log(OR) of BE/OA, for all GERD loci (line in blue), obesity-driven GERD loci (line and points in green), and neuropsychiatric-driven GERD loci (line and points in orange) using an inverse variance weighted regression model. Cochran Q statistics indicated strong heterogeneity in the overall all-GERD estimate (p<0.001). In the stratified analysis, it becomes apparent that the overall estimate was primarily driven by the much stronger effect sizes from the obesity-driven GERD loci subset. BE, Barrett’s oesophagus (or Barrett’s esophagus); EA, oesophageal adenocarcinoma (or esophageal adenocarcinoma); GERD, gastro-oesophageal reflux disease; MTAG, multitraitanalysis of GWAS; SNP, single nucleotide polymorphism.