A Genome‐Wide Meta‐Analysis of Functional Dyspepsia in 1,168,899 Individuals Implicates Immune Pathways in Disease Susceptibility

ABSTRACT

Background

Functional dyspepsia (FD) affects over 7% of the global population, impairing quality of life and driving significant healthcare burden. Although genetic contributions to FD pathophysiology have been suspected, large, well‐powered genetic studies remain limited. We aimed to identify FD genetic risk factors and implicated biological pathways.

Methods

We performed a FD genome‐wide association study in 1,168,899 European individuals, including 29,584 FD cases from five large biobanks. We dissected the genetic architecture of FD and its overlap with gastrointestinal and psychiatric disorders and conducted comprehensive gene‐prioritization, regulatory, and cell‐type enrichment analyzes to identify pathophysiological mechanisms and actionable targets.

Key Results

We show FD is a complex, highly polygenic trait with a significant genetic component (3.63% heritability) but lower than similar gastrointestinal disorders like irritable bowel syndrome (IBS) and gastroesophageal reflux disease (GERD). Mendelian randomization revealed causal effects of depression and anxiety on FD risk. We observed substantial local genetic correlations with IBS and GERD, including shared genome‐wide significant signals at the human leukocyte antigen region, supporting immune‐mediated mechanisms. Gene‐prioritization analyzes converged on DEAH‐Box Helicase 36 (DHX36), involved in immune responses to microbial infections. Integrative expression‐quantitative trait loci and colocalization analyzes indicated that higher genetic risk of FD is causally associated with a reduced DHX36 expression in gastrointestinal tissues, potentially triggering increased proinflammatory cytokine production.

Conclusions & Inferences

This study implicates immune‐driven pathways in FD susceptibility, adds insights into the pathophysiological mechanisms of the disease, and reveals DHX36 as a candidate target for future investigation.

Key Points

• Functional Dyspepsia (FD) is a highly polygenic trait with a modest but significant heritability, estimated at 3.63% in the European population.

• We identified substantial local genetic overlap at multiple loci between FD and both IBS and GERD and observed that depression and anxiety exert predisposing causal effects on FD risk.

• Genetic variants affecting immune‐related genes (HLA and DHX36) are associated with FD through increased gastrointestinal inflammatory responses, supporting an immune‐mediated component and offering candidate targets for further mechanistic investigation.

Plain Language Summary

Functional dyspepsia (FD) is a disorder of gut‐brain interaction characterized by symptoms originating in the upper digestive tract, although its underlying causes remain largely unexplained. This study examined genetic data from over 1 million individuals to better understand the biological mechanisms contributing to the condition. FD has a heritable component and shares a substantial proportion of its genetic basis with clinically related gastrointestinal disorders and psychiatric conditions. The strongest genetic evidence pointed toward altered immune and inflammatory responses that may be shared across disorders affecting the upper gastrointestinal tract. Together, these findings highlight the interplay between the gut, brain, and immune system in FD.

Our findings indicate that genetic variants in immune‐related genes (HLA‐DQB1 and DHX36) may influence functional dyspepsia risk by modulating gastrointestinal inflammatory responses, reinforcing the role of an immune‐mediated component in disease susceptibility and nominating these genes for further mechanistic investigation.

1. Introduction

Functional dyspepsia (FD) is a disorder of gut‐brain interaction (DGBI) with a world‐wide prevalence of around 7% [1]. Besides the cardinal symptoms of postprandial fullness, early satiation, epigastric pain, and epigastric burning, this condition has also been linked with psychological symptoms such as anxiety and depression [2]. In current clinical practice guidelines, treatment options for FD are limited in efficacy and level of evidence [3]. The development of novel therapeutic options is hampered by the observed large placebo response in clinical trials (30%–40%), the heterogeneity of this symptom‐based condition, and the associated difficulty in establishing the underlying pathophysiology in individual patients [4]. It includes alterations in gastric motility such as delayed gastric emptying and impaired gastric accommodation, duodenal alterations such as impaired barrier function, or altered signal transmission toward the brain [5]. All of these may have specific underlying genetic predisposition through altered neurotransmitter or gut peptide signaling, altered immune function, or altered mucosal integrity [6, 7].

A role for genetics in the pathophysiology of FD has been suspected based on twin and family studies [8], and previous biobank‐scale investigations [9]. Despite adequate sample size, however, the latter have failed to identify unequivocal risk genes and mechanisms.

Here we report a genome‐wide association study (GWAS) meta‐analysis of FD in 29,584 cases and 1,139,315 controls of European ancestry from five population‐based biobanks. With this sample size in hand, we reported on the genetic architecture of FD and its overlap with gastrointestinal (GI) and psychiatric disorders and proposed candidate genes and mechanisms that warrant further investigation.

2. Methods

2.1. Study Cohorts

FD was studied in five population‐based cohorts, totalling 1,168,899 European individuals. The phenotype was defined based on different versions of main and/or secondary International Classifications of Diseases (ICD). Genetic data relevant to the Estonian Genomic Centre of the University of Tartu (EGCUT) [10] and the UK Biobank (UKB) [11] cohorts were obtained from previously reported FD GWAS summary statistics [9]. The Michigan Genomics Initiative (MGI) data freeze 3, consisting of 50,000 European individuals, was employed upon approval from the University of Michigan Institutional Review Board. Further, publicly available GWAS from FinnGen (K11_FUNCDYSP, data freeze R10) [12] using ICD8, ICD9, and ICD10 codes and the Million Veteran Program (MVP; Dyspepsia and other specified disorders of function of stomach) [13], using the PheCode 536.8 were added to the study.

2.2. GWAS Meta‐Analysis

GWAS summary statistics were quality‐checked, lifted‐over to GRCh37, harmonized, and assigned to a valid rsID with the python package GWASlab [14]. Individual GWAS were integrated using METAL [15] and only variants present in at least 30% of the total sample size with no heterogeneity in their effect sizes across studies were kept. Finally, GWASinspector [16] was employed to further remove duplicated or missing variants and mismatched, ambiguous and multi‐allelic variants that could not be verified.

2.3. Heritability, Genetic Correlations and Overlap With Other Traits

SNP‐based heritability (h ² _SNP) was estimated via linkage disequilibrium score regression (LDSC) [17] at the liability scale. Pre‐computed linkage disequilibrium (LD) matrices from the 1000G were employed. Batch genetic correlations were carried out to decipher the overall genetic similarity of FD with 2871 publicly available GWAS from the CTG‐VL platform (https://vl.genoma.io/). Only significantly correlated traits after false discovery rate (FDR) adjustment for multiple tests and a significant heritability (h ² _SNP Z‐score ≥ 3) were considered.

Local genetic correlations with irritable bowel syndrome (IBS), gastroesophageal reflux disease (GERD), depression disorder (DEP), and anxiety disorder (ANX) were carried out using LAVA [18] (Table S2). First, univariate analysis estimates h ² _SNP for each locus for each trait and, thereafter, bivariate analysis estimates the local covariance for each locus and pair of traits. Pre‐calculated UKB LD references were used. Multiple testing was accounted for using Bonferroni correction. To colocalize loci with significant genetic correlation and GWAS signals, GWAS Catalog [19] was consulted for GERD, IBS, DEP (“Depression”, “Major depressive disorder”, and “Depressed affect”), and ANX.

2.4. Mendelian Randomization Analyzes

Bidirectional Mendelian randomization (MR) analyzes were performed to investigate potential causal relationships between FD and highly‐correlated digestive and psychiatric traits using the R package TwoSampleMR [20] (v0.6.8). Traits relevant to gut motility, such as constipation, diarrhea, and stool frequency (SF) were also included (Table S3). Independent instrumental variables strongly associated with the exposure were identified through LD clumping using an r ² threshold of 0.001, a 10 Mb window, and a suggestive significance threshold (p < 5 × 10⁻⁶). Only variants with F‐statistic > 10 were retained. The multiplicative random‐effects inverse variance weighted estimator was used as the primary model, with additional estimators applied to account for different assumptions regarding instrumental variable validity. We assessed robustness using Cochran's Q, MR Egger intercept, and Steiger directionality tests. Significant p‐value thresholds were set at p < 0.001, p < 0.01, and p < 0.05.

2.5. Locus Definition, Gene Mapping and Prioritization

Independent genomic risk loci were defined using FUMA [21] considering SNPs at suggestive significance (p ≤ 5 × 10⁻⁶) not in LD (r ² < 0.6). For LD blocks closer than 250 kbp apart, association signals were merged into a single locus. The SNP2GENE pipeline from FUMA was used to map genes positionally (maximum distance of 10 kbp) and through eQTL.

To systematically prioritize genes from our GWAS meta‐analysis, an integrative strategy combining statistical, functional, and causal approaches was employed. Gene‐based association tests were performed using MAGMA [22] and fastBAT [23]. A transcriptome‐wide association study (TWAS) across GTEx [24] tissues relevant to FD (brain and GI) was carried out via S‐MultiXcan [25]. To assess potential causal relationships between gene expression and the GWAS signals, summary‐based MR (SMR) [26] was applied using eQTL data from the most significant tissues. In addition, two databases of regulatory elements were inspected to explore how the genetic variants in the prioritized locus may change chromatin accessibility status, overlap with transcription binding sites and alter histone modifications, namely FORGEdb (https://forgedb.cancer.gov/) and RegulomeDB (https://regulomedb.org/regulome‐search/). Finally, genetic interactions with approved or investigational drugs were assessed by querying the DrugBank (https://go.drugbank.com/), Open Targets Platform (https://platform.opentargets.org/), and DGIdb (https://dgidb.org/) databases.

2.6. Tissue and Cell Type Prioritization

Different gene‐level enrichment analysis approaches were performed to identify tissues and cell types potentially relevant to FD. First, MAGMA gene‐property analysis was conducted using GTEx gene expression data to identify enrichment of gene‐level associations in tissue‐specific expression profiles. Subsequently, these gene‐level associations were employed to test cell type specificity of FD as part of FUMA Cell Type analysis [27]. Similarly, DEPICT [28] was used to identify tissue and/or cell types in which associated genes were differentially expressed. Finally, stratified LDSC [29] was applied to identify tissues where genetic variants disproportionately contribute to trait heritability.

2.7. FD—Gastroparesis (GP) Genetic Overlap

Publicly available GP GWAS were obtained from GWAS Catalog, under the accession numbers GCST90295922 [30] and GCST90478363 [13], and used for fixed‐effects meta‐analysis using METAL, as described above. SNP‐based heritability was calculated on the liability scale using different estimations of the GP population prevalence (0.02%–2.0%), based on epidemiological studies [30, 31, 32]. Genetic correlation with FD was assessed using LDSC, as described above.

3. Results

3.1. Characteristics of the Cohorts

FD was investigated across five population‐based cohorts, comprising a total of 29,584 cases and 1,139,315 controls of European ancestry (Table S1). Across the population‐based cohorts, the prevalence of FD was generally consistent, averaging 2.6% ± 0.6%. An exception was observed in the EGCUT biobank, which consisted of a nested case–control dataset, not the entire population, resulting in a higher case‐to‐total‐sample‐size proportion of 0.128 (12.8%). Female participants outnumbered males in all cohorts (53.1%–65.2%), with the sole exception of the MVP biobank, where women comprised only 7.5% of the sample. Participant ages varied between cohorts, ranging from an average of 42 years in EGCUT to 64 years in MVP.

3.2. FD GWAS Meta‐Analysis: Heritability, Genetic Correlations and Overlap With Other Traits

A GWAS meta‐analysis of FD resulted in 9,638,985 high‐quality genetic variants with no signs of population stratification (LDSC intercept = 1.02). Based on LDSC analysis, a limited but significant SNP heritability (see Section 2) was estimated at 3.63%. FD correlated with 724 traits and anti‐correlated with 108 traits (p _FDR < 0.05; h ² _SNP Z‐score ≥ 3) (Table S4). Traits belonging to the digestive domain were among those with the highest genetic correlation estimates (r _g), including disorders which commonly overlap in FD, such as IBS or GERD [33], treatments or medication used to alleviate these, and both extremes of the SF spectrum: constipation and diarrhea (Figure 1). Similarly, anxiety/panic attacks were highly correlated with FD as well as depression and known medications to treat these psychiatric disorders. A wide variety of pain‐related traits, besides the expected abdominal pain, were also found among the top correlated traits. Interestingly, disorders from the musculoskeletal domain and inflammatory‐driven disorders (chronic pharyngitis, respiratory diseases, inflammatory diseases of the prostate) were also found among the top correlated traits, highlighting a plausible shared pathophysiological mechanism with FD.

FIGURE 1.

Genetic correlations of functional dyspepsia (FD) with selected traits. Regression score (r_g) estimates and their standard error are shown on the x‐axis. A full list of genetically correlated and anti‐correlated traits is available at Table S1.

We performed a more detailed analysis of the genetic overlap of FD with two GI disorders, IBS and GERD, and two psychiatric disorders, DEP and ANX (Table S2). Uni‐variate analyzes performed with LAVA (see Section 2) revealed that FD is a highly polygenic trait with strong local heritability (p _Bonferroni < 0.05) across 1441 loci. Bi‐variate analyzes revealed that GI disorders and, in particular, GERD showed a substantially higher genetic overlap with FD than psychiatric disorders (Figure S1). GERD showed up to 50 genomic loci with significant local genetic correlation with FD (p _Bonferroni < 0.05), albeit none of them was in common with IBS (five loci) or DEP (two loci). About half of the loci showing genetic correlation with GERD or IBS colocalized with GWAS signals from any of the traits examined. In particular, locus 961 (human leukocyte antigen region; HLA) and locus 1719 (NCAM1 locus) were characterized by the presence of common GWAS signals to all traits under study, highlighting potential underlying pleiotropic effects (Table S5).

To assess possible causative links between FD and highly correlated traits from the digestive and psychiatric domain, bidirectional MR analyzes were carried out (Figure 2; Table S6). We found bidirectional causative effects of FD with IBS, GERD, and diaphragmatic hernia, pointing to potential reciprocal effects or shared pathways (Figure 2A). Further, directional pleiotropy may also explain the relationships observed with IBS and GERD as evidenced by the different sensitivity methods analyzed (Table S6). Similar findings were obtained for constipation, whereas diarrhea and SF showed no convincing evidence of association (Figure 2B). Finally, our MR analyzes strongly supported a causal effect of anxiety and depression on FD (Figure 2C). All estimators yielded significant positive effects (Odds Ratio, OR > 1; p < 0.05) with no evidence of horizontal pleiotropy, heterogeneity, or outliers (Table S6).

FIGURE 2.

Bi‐directional Mendelian randomization (MR) analysis of functional dyspepsia (FD) with selected traits. Causal relationships are reported based on MR results in both directions, with FD tested as exposure or outcome in relation to gastrointestinal disorders (A), motility‐affecting disorders and traits (B), and psychiatric disorders (C). Odds Ratio with 95% CIs are shown on the x‐axis for each MR method (y‐axis). Significance level is provided as different shades of red and gray for non‐significant (N.S.) estimates. DIAHER, diaphragmatic hernia; GERD, gastroesophageal reflux disease; IBS, irritable bowel syndrome; IVW, inverse variance weighted; SF, stool frequency; STODUO, other diseases of the stomach and duodenum.

3.3. Annotation of FD Genomic Risk Loci

A single genome‐wide significant (p < 5 × 10⁻⁸) locus was detected, located in the HLA region (Table 1; Figure S2). The SNPs in this locus positionally mapped to 137 genes and through expression quantitative trait loci (eQTLs) to 155 genes. Specifically, the leadSNP (rs9274481; 6:32633756:C:T) fell in the HLA Class II region and physically mapped to the HLA‐DQB1 as the nearest gene. Querying GWAS ATLAS (https://atlas.ctglab.nl/PheWAS) and IEU OpenGWAS [34] repositories revealed that this variant has been previously associated with an increased risk of autoimmune diseases like coeliac disease, Type 1 diabetes, or inflammatory bowel disease (Table S7). Of note, the reported associations have been mostly discovered in genetic studies based on the FinnGen cohort. In line with that observation, FinnGen was the major contributor to the signal detected in the FD GWAS meta‐analysis (Figure S3), albeit the direction of effects among the suggestive significant SNPs in the locus remained consistent across the individual cohorts (Figure S4). Twenty‐nine additional association signals were detected at a suggestive level of significance (Table 1). Sixty unique protein coding genes positionally and/or eQTL mapped to these loci after excluding those mapped to the two loci located at the HLA region (locus 16 and 17).

TABLE 1.

Suggestive (p < 5 × 10⁻⁶) association signals detected in the FD GWAS meta‐analysis.

Locus	Lead SNP	Chr	Start	End	EA	OA	EAF	p	β	SE	Ind. Sig. SNPs	Mapped genes
1	rs35006635	1	2,256,933	2,265,794	A	C	0.094	1.06E‐06	−0.10	0.020	1	MORN1p,e, AL589739.1p, C1orf86e, MORN1e, RER1e, PLCH2e, MMEL1e
2	rs111305386	1	11,032,215	11,048,973	A	G	0.202	2.37E‐06	−0.07	0.014	1	C1orf127p,e, EXOSC10e, NPPBe
3	rs66476853	1	232,167,832	232,173,560	T	TGAAA	0.349	4.80E‐06	−0.07	0.014	1	DISC1p
4	rs1481148	1	238,394,815	238,394,815	A	C	0.986	1.97E‐06	−0.17	0.035	1	—
5	rs4513278	2	229,021,203	229,039,209	G	C	0.207	1.32E‐07	0.05	0.010	1	SPHKAPp
6	rs73180867	3	118,068,778	118,274,337	A	G	0.013	1.08E‐07	0.27	0.051	1	—
7	rs7621176	3	153,449,087	154,563,641	T	C	0.112	1.07E‐07	−0.07	0.014	4	ARHGEF26p,e, GPR149p, DHX36e, MMEe
8	rs77329569	3	156,338,528	156,702,477	C	T	0.060	3.36E‐06	−0.09	0.018	1	TIPARPp, LEKR1p,e, C3orf55e
9	rs34623301	3	184,093,040	184,093,040	A	G	0.192	1.80E‐06	−0.06	0.013	1	EIF2B5p, POLR2Hp, THPOp, CHRDp, MAP6D1e, ABCC5e, PSMD2e, CLCN2e
10	rs16860198	4	47,413,121	47,476,176	G	T	0.257	1.68E‐07	−0.05	0.010	1	GABRB1p,e, COMMD8p,e, ATP10De, CORINe, NFXL1e
11	rs150356098	4	154,277,203	154,464,438	T	A	0.019	2.82E‐06	0.18	0.038	2	MND1p, KIAA0922p,e, SFRP2e
12	rs114724752	4	159,329,696	159,426,815	T	A	0.813	4.24E‐06	−0.07	0.015	1	RXFP1p,e, TMEM144e, ETFDHe, PPIDe, RAPGEF2e
13	rs532473424	5	130,646,058	130,729,398	A	T	0.026	6.06E‐08	0.23	0.042	1	CDC42SE2p,e, LYRM7e, RAPGEF6e, FNIP1e, ACSL6e, SLC22A4e
14	rs80301208	5	142,820,148	142,876,477	G	A	0.044	3.45E‐06	0.10	0.021	1	NR3C1p
15	rs11399165	6	2,638,542	2,668,990	CG	C	0.521	1.35E‐07	−0.07	0.014	1	C6orf195p,e, MYLKp,e
16	6:32633756:C:T	6	29,356,331	33,016,499	T	C	0.129	4.61E‐08	0.12	0.022	8	Multiple
17	rs3130017	6	33,315,152	33,446,169	T	G	0.794	1.47E‐06	−0.12	0.026	1	Multiple
18	rs9353851	6	92,214,128	92,293,188	G	T	0.650	2.77E‐06	0.04	0.009	1	—
19	rs147568518	6	140,379,495	140,708,938	C	CTT	0.070	1.62E‐06	0.10	0.021	1	—
20	rs9322521	6	155,653,680	155,738,045	C	T	0.248	3.97E‐06	−0.05	0.010	1	NOX3p, TIAM2e, TFB1Me, CLDN20e
21	rs17305805	8	36,628,405	36,628,405	A	C	0.205	7.29E‐07	−0.05	0.011	1	—
22	rs118029983	12	61,041,718	61,041,718	T	G	0.026	2.55E‐06	−0.15	0.033	1	—
23	rs72670866	14	26,960,655	27,001,016	C	T	0.169	2.27E‐06	0.07	0.015	1	NOVA1p
24	rs61977762	14	56,208,747	56,208,747	T	C	0.069	4.06E‐06	−0.10	0.021	1	LGALS3e, PELI2e
25	rs11073488	15	96,961,740	96,978,829	A	G	0.380	2.70E‐06	−0.04	0.009	2	—
26	rs146792144	16	60,583,974	60,650,143	AAC	A	0.591	2.21E‐06	0.07	0.015	1	—
27	rs138603506	16	74,197,814	74,197,814	A	G	0.022	1.15E‐06	0.17	0.034	1	RFWD3e
28	rs148426992	16	75,031,580	75,116,252	T	C	0.071	3.14E‐07	0.15	0.030	1	WDR59p,e, ZNRF1p,e, RFWD3e, ZFP1e
29	rs17470481	18	25,788,096	25,922,602	A	G	0.321	1.44E‐06	−0.04	0.009	1	CDH2e
30	rs11669334	19	32,200,518	32,208,909	T	G	0.161	1.28E‐06	0.06	0.012	1	—

Note: Genome‐wide significant locus is highlighted in bold.

Abbreviations: β, genetic effects; Chr, chromosome; EA, effect allele; EAF, effect allele frequency; ^eeQTL, mapped genes; ^p, Positionally mapped genes; p, p‐value; SE, standard error; SNP, single‐nucleotide polymorphism.

3.4. Gene Prioritization Analysis

An in‐depth functional characterization of the FD associated genetic variants was performed applying different methodologies and data sources to assess gene‐level associations, functional relevance, and regulatory mechanisms (see Section 2). We used gene‐based association tests (MAGMA and fastBAT) that combine genetic signals within the same gene, highlighting 642 genes associated at nominal significance (p < 0.05) (Tables S8 and S9). We also examined whether changes in gene activity in relevant tissues, particularly the brain and gastrointestinal tract, might play a role, using a TWAS approach based on GTEx data (see Section 2). This approach revealed a single gene‐tissue signature that passed FDR correction (p _FDR < 0.05), consisting of a reduced expression of DEAH‐Box Helicase 36 (DHX36) in GI tissues (Table S10). To explore whether these gene‐expression changes might be driving the genetic associations we observed, we applied the SMR method to the tissues showing the strongest evidence. Following this method, we causally linked specific genetic variants associated with a higher risk of FD and a reduced expression of DHX36 (Table S11; Figure 3). Several of these genetic variants overlap with transcription factor binding motifs, chromatin accessibility hotspots and other regulatory elements, providing plausible mechanisms underlying DHX36 altered expression (Figure 3). Across all prioritization strategies, only two genes, DHX36 and Rho Guanine Nucleotide Exchange Factor 26 (ARHGEF26), were consistently highlighted (Figure 4). Finally, we looked for potential links between these genes and existing or experimental drugs by consulting 3 major drug‐gene interaction databases (see Section 2). Manoalide (CHEMBL463914) emerged as the only compound reported to interact with DHX36 as part of its mechanism of action.

FIGURE 3.

Colocalization of GWAS and expression‐quantitative trait loci (eQTL) signals on the DHX36 gene. Heatmap showing the genetic effects on functional dyspepsia (FD) risk (effect size/standard error) and the eQTL effects on the DHX36 gene across gastrointestinal tissues from GTEx (−log₁₀(P) × effect size). FORGEdb scores and RegulomeDB ranks were used for predicting the likelihood of each SNP to be a regulatory variant.

FIGURE 4.

Integrated gene prioritization across multiple GWAS downstream analysis tools. Genes were prioritized by combining results from the positional (GWAS) and expression‐quantitative trait loci (eQTL) mapping, and the MAGMA, FastBAT, S‐MultiXcan, and summary‐based Mendelian randomization (SMR) analyzes. Only genes found significant by all methods tested were selected as possible candidates. For each tool, association p‐values were normalized to z‐scores to highlight genes appearing among the top‐ranked across all methods.

3.5. Tissue and Cell Type Prioritization

Tissue and cell‐type enrichment analyzes highlighted blood vessels and blood‐derived cell types as nominally significant (p < 0.05), with several immune‐related components such as granulocytes and lymphocytes emerging across methods (Table S12). To further refine these findings, we performed a targeted cell type analysis using single‐cell RNA sequencing (scRNA‐seq) data from blood and bone marrow, which identified dendritic cells as the most significantly enriched (p < 0.05), suggesting a potential role in genetic susceptibility to FD. Finally, guided by the TWAS results implicating GI‐related tissues, we conducted a parallel scRNA‐seq‐based cell type analysis in GI tissues. This revealed enrichment (p < 0.05) in enteroendocrine, vein/arterial endothelial cells, and dendritic cells, which are critical for gut barrier function and immune signaling.

3.6. FD—GP Genetic Overlap

Given the widely reported similarity between FD and GP in clinical practice, including shared pathophysiological mechanisms, often indistinguishable symptoms, and even misdiagnosis [31, 32, 35], we decided to investigate whether these two conditions shared a common genetic liability. We therefore performed a GP GWAS meta‐analysis using a previously published pilot‐scale GWAS and a publicly available GWAS from the MVP cohort (see Section 2), thus including 3393 GP cases and 504,654 controls. We estimated the SNP‐based GP heritability on the liability scale to be between 8.3% and 19.5% (p = 3.76 × 10⁻¹⁰), depending on different estimations of population prevalence [31, 32]. GP showed a significant but modest genetic correlation with FD (r _g = 0.33 ± 0.12, p = 5.30 × 10⁻³), clearly below the estimates observed for IBS and GERD.

The pilot‐scale GP GWAS reported the strongest association signal at the HLA Class II region, tagged by the lead SNP rs9273363 and mapping the HLA‐DQB1 gene [30]. The same genetic signal was found at genome‐wide significant (p = 3.92 × 10⁻⁹) in the GP GWAS from the MVP cohort, which included a significant increase in the number of GP cases and total sample size. In concordance, this signal appeared as the sole genome‐wide significant association in the GP GWAS meta‐analysis (rs9273363 p = 2.95 × 10⁻¹⁴; Figure 5).

FIGURE 5.

Regional plot of the genomic risk locus mapping the human leukocyte antigen (HLA) region in functional dyspepsia (red; top) and gastroparesis (blue; bottom) GWAS meta‐analysis. Suggestive and genome‐wide significant thresholds are indicated with a gray and black dashed line, respectively. HLA genes in the region are shown at the bottom of the figure.

4. Discussion

In this study, we leveraged data from over 1 million individuals, including 29,584 FD cases, representing the largest and most comprehensive genetic investigation of FD to date. Despite this, we identified only one genome‐wide significant locus, indicating that the genetic contribution to FD is likely weak. Indeed, SNP‐based heritability of FD on the liability scale was estimated at around 3.6%, in concordance with previous estimates [9], and notably lower than other GI disorders such as IBS, GERD, or diverticular disease [36, 37]. Furthermore, local heritability analyzes confirmed FD as a highly polygenic trait, adding to its genetic complexity. FD is therefore likely influenced by many contributing causal variants with relatively small effects and no single variant explaining a big proportion of the risk [38].

At a global level, genome‐wide genetic correlation analysis corroborated previously established genetic relationships with digestive and psychiatric conditions [9], and expanded its relationships with disorders of the musculoskeletal domain and other inflammatory‐driven disorders, as previously reported in epidemiological studies [39]. These analyzes also revealed that FD and GP, two disorders that have been proposed as part of a single disease spectrum from a clinical perspective [32], are genetically correlated yet represent distinct entities with different underlying genetic liabilities.

Further, MR analyzes revealed bidirectional causative effects of FD with IBS, GERD, and constipation, which remain inconclusive as to the real causative direction but evidence a common genetic etiology shaping these GI disorders. Conversely, we found strong evidence that anxiety and depression increase the genetic liability of FD, providing novel insight beyond epidemiological and observational studies consistently reporting significant associations between these traits [2, 40, 41]. Combined, these results reflect an overall influence of the brain‐gut axis in the pathophysiology of FD, albeit specific aetiological pathways remain elusive, as opposed to other DGBI, such as IBS [42]. Future studies should further elucidate the specific contributions of these closely related psychiatric traits which may help identify novel therapeutic strategies to improve the overall symptom burden in FD patients.

Alternatively, our bivariate analyzes revealed a higher genetic overlap at specific loci between FD and GI disorders, especially GERD, than with psychiatric traits. The high genetic overlap may, thus, help us understand FD biology through the identification of shared pathways and explain the overlapping symptomatology and frequent coexistence between GERD and FD [43, 44]. The convergence of the high local genetic correlation in the HLA region with the genome‐wide significant signals existing in this complex region for other GI disorders may point toward an immune‐driven mechanism behind their pathophysiology [45]. The genome‐wide significant signal identified in the FD GWAS meta‐analysis was mapped to HLA Class II genes, which are of paramount importance for antigen presentation to CD4+ helper T cells and, therefore, adaptive immune responses to infections [46]. Of particular importance, the sole genome‐wide significant signal in the GP GWAS meta‐analysis mapped to the same HLA gene (HLA‐DQB1), suggesting a possible common immunogenetic factor predisposing to upper gastrointestinal disorders. In this regard, the first GWAS of idiopathic achalasia, a gastrointestinal motility disorder affecting esophageal function, also identified its strongest genetic association within HLA‐DQB1, further supporting the involvement of HLA Class II‐mediated immune mechanisms across disorders of the upper gastrointestinal tract [47]. Indeed, an immune hypothesis behind GERD, IBS, GP, and FD has been previously proposed, suggesting low‐grade inflammation, immune cell dysregulation, and barrier dysfunction may contribute to symptom generation [32, 48, 49]. Furthermore, there is increasing evidence pointing toward dysregulation of helper T cells as a key pathophysiological mechanism driving microinflammation in FD [50, 51]. Finally, the stronger association observed in FinnGen suggests that this immunogenetic factor may contribute to FD susceptibility to a greater extent in individuals of Finnish ancestry than in other European populations. Although the consistent direction of effect across cohorts supports the overall association, confirmation in larger independent cohorts will be required to determine the magnitude and generalizability of this population‐related difference.

The integrative approach followed in the current study allowed convergence on a plausible candidate gene, DHX36, supported by statistical association, tissue‐specific expression, and causal inference enhancing the biological interpretability of our findings. The genetic signals from the corresponding genomic risk locus may reflect a mechanism involving dysregulation of the DHX36 gene expression in several tissues across the GI tract. In particular, higher FD risk was linked to a reduced transcriptional activity of DHX36 in the esophagus, stomach, small intestine, and colon, potentially mediated by regulatory variants affecting transcription factor binding and local chromatin activity. This gene encodes a member of the DExD/H‐box helicase family, responsible for the recognition of viral nucleic acids as part of the innate immune system response against viral infections [52]. Recent evidence shows that loss of DHX36 leads to an increased production and secretion of proinflammatory cytokines [53]. The encoded protein functions primarily in myeloid and plasmacytoid dendritic cells, activating Type I interferon and inflammatory cytokines production. Our tissue and cell type prioritization analysis revealed enrichment of dendritic cells not only in the blood and bone marrow but also in intestinal tissues, suggesting a potentially relevant role in FD pathophysiology. Interestingly, DHX15, another member of the DExD/H‐box helicase family, has been implicated in intestinal inflammation triggered by viral infections [54]. Taken together, the genetic evidence from our study alongside prior functional studies on DHX36 and related helicases supports the hypothesis that altered DHX36 activity may trigger low‐grade intestinal inflammation, thereby contributing to FD pathophysiology [5, 48]. Thus, the strongest genetic signals identified in the present study point toward immune activation that leads to dysregulated cytokine production and subsequent mast cell and eosinophil recruitment, a recognized hallmark of FD [5, 49, 51]. Finally, to investigate the therapeutic relevance of DHX36, we screened three major drug databases and identified Manoalide as a compound reported to interact with this gene, albeit not selectively. Manoalide is a potent analgesic and anti‐inflammatory, due to the inhibition of phospholipase A2 (PLA2) [55]. By blocking calcium channels, manoalide has also been shown to prevent mast cell‐mediated inflammatory responses [56], a mechanism of particular relevance for FD. Further, preliminary evidence shows that manoalide inhibits the ATPase activity of human DHX36 potentially by targeting a conserved helicase domain [57], suggesting a possible functional interaction. Overall, the evidence linking DHX36 to FD provides strong rationale for testing related mechanisms and pathways in follow‐up genetic and experimental studies.

Our study has intrinsic limitations that need to be acknowledged. First, phenotype definition was based on ICD10/Phecodes, which are efficient at scale but may introduce more variability across centres than gold‐standard Rome criteria. Second, participants were exclusively of European ancestry, which limits the portability of the genetic effects estimates to other ancestries.

Notwithstanding these constraints, this study provides better insights into the complex polygenic architecture of FD, highlighting immune‐mediated mechanisms and individual candidate genes that merit further investigation in independent, deeply phenotyped FD cohorts based on current Rome criteria. Integrating genetic evidence with complementary biological samples and molecular data may further clarify the mechanisms implicated by DHX36 and their relevance to FD pathophysiology.

Author Contributions

Cristian Díaz‐Muñoz: conceptualization (lead), formal analysis (lead), data curation (lead), methodology (lead), visualization (lead), writing – original draft (lead), writing – review and editing (lead). Bert Broeders: conceptualization (equal), resources (lead), writing – original draft (equal), writing – review and editing (equal). Ferdinando Bonfiglio: formal analysis (supporting), resources (supporting), data curation (supporting). Andreea Zamfir‐Taranu: formal analysis (supporting), data curation (supporting). Joris Arts: resources (supporting). Peter Banovcin: resources (supporting). Luis Bujanda: resources (supporting). Vasile Drug: resources (supporting). Dan Dumitrascu: resources (supporting). Guillaume Gourcerol: resources (supporting). Sébastien Kindt: resources (supporting). Peter Lipták: resources (supporting). Chloé Melchior: resources (supporting). Carolina Malagelada: resources (supporting). Edoardo Vincenzo Savarino: resources (supporting). Artemis Trikola: resources (supporting). Jara Lehoucq: resources (supporting). Florencia Carbone: conceptualization (supporting), investigation (supporting), project administration (supporting), resources (equal), writing original draft (supporting), writing – review and editing (supporting). Jan Tack: conceptualization (equal), funding acquisition (lead), investigation (equal), methodology (equal), project administration (equal), resources (equal), supervision (equal), writing – review and editing (supporting). Mauro D'Amato: conceptualization (lead), investigation (lead), methodology (equal), project administration (lead), resources (lead), supervision (lead), visualization (supporting), writing original draft (equal), writing – review and editing (equal).

Funding

J.T. is supported by a Methusalem grant of KU Leuven and project grants G0C4420N, G074624N, and ZKE5406 from Flanders Research Foundation (FWO Vlaanderen).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Local genetic correlations between functional dyspepsia (FD) and correlated digestive and psychiatric disorders. From top to bottom, Manhattan plots showing the estimates of local genetic correlation between FD and anxiety (ANX), depression (DEP), gastro‐esophageal reflux disease (GERD), and irritable bowel syndrome (IBS). Red bars represent loci showing significant local genetic correlation after Bonferroni correction for multiple tests.

Figure S2: Manhattan plot of functional dyspepsia (FD) GWAS meta‐analysis. Genetic variants associated with FD at suggestive significance are shown in orange and at genome‐wide significance in red. Genomic risk loci are numbered according to Table 1.

Figure S3: Regional plot for the genomic risk locus mapping the HLA region. Association results are shown for the GWAS meta‐analysis (meta; top) as well as for the individual GWAS from each cohort (bottom). EGCUT, Estonian Genomics Centre of the University of Tartu; MGI, Michigan Genomics Initiative; MVP, Million Veteran Program; UKB, UK Biobank.

Figure S4: Forest plot showing the genetic effects of suggestive significant SNPS from the genomic risk locus mapping the HLA region. Association results are shown for the GWAS meta‐analysis (meta; dark red) as well as for the individual GWAS from each cohort (light red). Twenty‐Five SNPS were present in at least two cohorts, while 24 were Finngen‐specific (not shown). Genomic coordinates are in GRCH37. EGCUT, Estonian Genomics Centre of the University of Tartu; MGI, Michigan Genomics Initiative; MVP, Million Veteran Program; UKB, UK Biobank.

Table S1: Demographics of the cohorts included in the functional dyspepsia (FD) GWAS meta‐analysis.

Table S2: GWAS meta‐analysis employed for local heritability and genetic correlation estimation.

Table S3: GWAS meta‐analysis employed for bidirectional Mendelian randomization analyzes.

Table S4: Batch genetic correlations performed between functional dyspepsia (FD) and 2871 publicly available GWAS from the CTG‐VL platform.

Table S5: Genomic loci with significant genetic correlation between functional dyspepsia (FD) and selected phenotypes after Bonferroni correction.

Table S6: Sensitivity analysis relative to the bidirectional Mendelian randomization analyzes. Information relevant to the number of instrumental variables (IVS) employed, horizontal pleiotropy tests, heterogeneity tests, and steiger directionality tests is reported.

Table S7: Phenome‐wide association analysis relevant to RS9274481.

Table S8: Magma gene‐based association results for functional dyspepsia (FD) GWAS meta‐analysis. Only the top 100 hits are reported.

Table S9: Fastbat gene‐based association results for functional dyspepsia (FD) GWAS meta‐analysis. Only the top 100 hits are reported.

Table S10: S‐Multixcan transcriptome‐wide association results for functional dyspepsia (FD) GWAS meta‐analysis tested on gastrointestinal and brain‐related tissues from GTEX. Only the top 100 hits are reported. The table shows the tissue with the most (tissue_best) significant change in gene expression with its respective p‐value (P_BEST), and the effect in gene expression in the selected tissues associated with the FD genetic variants (beta) along with its respective standard deviation (SD). Adjusted p‐value for multiple tests is reported as false discovery rate (FDR).

Table S11: SMR and Heidi results assessing putative causal relationships between gene expression and functional dyspepsia (FD) using GWAS summary statistics and GTEX expression quantitative trait loci (EQTL) data from gastrointestinal tissues. Only the top 100 hits are reported. The table shows the top single nucleotide polymorphism (SNP) associated to the change in gene expression alongside the effect allele (EA), its allele frequency (EAF), the GWAS genetic effects (beta, standard error, and p‐value), the EQTL effects, and the combined effects given by SMR. Heidi tests were also performed to test the causal relationship between the change in gene expression and the trait under study.

Table S12: Tissue and cell type enrichment analysis. Only tissues/cell types found at nominal significance (p < 0.05) are reported.

Data Availability Statement

The GWAS meta‐analysis generated and analyzed during the current study is available in GWAS Catalog (https://www.ebi.ac.uk/gwas/), under the study accession ID GCST90854302.