Whole Exome Sequencing Identifies Epithelial and Immune Dysfunction-Related Biomarkers in Food Protein-Induced Enterocolitis Syndrome

Clinical & Experimental Allergy, 2024; 54:919–929 https://doi.org/10.1111/cea.14564 919 Clinical & Experimental Allergy ORIGINAL ARTICLE OPEN ACCESS Whole Exome Sequencing Identifies Epithelial and Immune Dysfunction- Related Biomarkers in Food Protein- Induced Enterocolitis Syndrome AlbaCamino- Mera1,2,3 | JacoboPardo- Seco1,2,3 | XabierBello1,2,3 | LauraArgiz4 | RobertJ.Boyle5 | AdnanCustovic5 | JethroHerberg6 | MyrsiniKaforou6 | StefaniaArasi7 | AlessandroFiocchi7 | ValentinaPecora7 | SimonaBarni8 | FrancescaMori8 | TeresaBracamonte9 | LuisEcheverria9 | VirginiaO'Valle- Aísa10 | NoeliaLaraHernández- Martínez10 | IriaCarballeira11 | EmilioGarcía11 | CarlosGarcia- Magan12 | JoséDomingoMoure- González12 | PurificaciónGonzalez- Delgado13 | TeresaGarriga- Baraut14 | SonsolesInfante15 | GabrielaZambrano- Ibarra15 | MargaritaTomás- Pérez15 | AdriannaMachinena16 | MarionaPascal17,18 | AnaPrieto19 | SoniaVázquez- Cortes20 | MontserratFernández- Rivas21 | LeticiaVila22 | LaiaAlsina23 | MaríaJoséTorres24,25,26,27 | GiusiMangone28 | SantiagoQuirce29 | FedericoMartinón- Torres1,3,30 | MartaVázquez- Ortiz5 | AlbertoGómez- Carballa1,2,3 | AntonioSalas1,2,3 Correspondence: Antonio Salas (anto[email protected]) | Alberto Gómez-Carballa ([email protected]) Received: 9 July 2024 | Accepted: 1 September 2024 Funding: This study was supported by European Union's Horizon 2020 research and innovation programme (668303) and Strategic Health Action, ‘Instituto de Salud Carlos III’ (TRINEO: PI22/00162, DIAVIR: DTS19/00049, Resvi- Omics: PI19/01039, ReSVinext: PI16/01569, Enterogen: PI19/01090, OMICOVI- VAC: PI22/00406, BIOFPIES: PI19/00497, IN607B 2020/08, IN607A 2023/02, GENCOVID IN845D 2020/23, IIN607A2021/05, PID2022- 142156OB- I00, CB21/06/00103, CP23/00080). Keywords: ATG16L1| DGKZ| exomes| food allergy| FPIES| NGS| RBM8A ABSTRACT Background: Food proteininduced enterocolitis syndrome (FPIES) is a food allergy primarily affecting infants, often leading to vomiting and shock. Due to its poorly understood pathophysiology and lack of specific biomarkers, diagnosis is frequently delayed. Understanding FPIES genetics can shed light on disease susceptibility and pathophysiology—key to developing diagnostic, prognostic, preventive and therapeutic strategies. Using a wellcharacterised cohort of patients we explored the potential genomewide susceptibility factors underlying FPIES. Methods: Blood samples from 41 patients with oral food challengeproven FPIES were collected for a comprehensive whole exome sequencing association study. Results: Notable genetic variants, including rs872786 (RBM8A), rs2241880 (ATG16L1) and rs2289477 (ATG16L1), were identified as significant findings in FPIES. A weighted SKAT model identified six other associated genes including DGKZ and SIRPA. DGKZ induces TGF- β signalling, crucial for epithelial barrier integrity and IgA production; RBM8A is associated with thrombocytopenia absent radius syndrome, frequently associated with cow's milk allergy; SIRPA is associated with increased neutrophils/monocytes in inflamed tissues as often observed in FPIES; ATG16L1 is associated with inflammatory bowel disease. Coexpression correlation analysis revealed a functional correlation between RBM8A and filaggrin gene (FLG) in stomach and intestine tissue, with filaggrin being a known key pathogenic and risk factor for IgE- mediated food allergy. A transcriptomewide association study suggested genetic variability in patients impacted gene expression of RBM8A (stomach and pancreas) and ATG16L1 (transverse colon). Conclusions: This study represents the first case–control exome association study of FPIES patients and marks a crucial step towards unravelling genetic susceptibility factors underpinning the syndrome. Our findings highlight potential factors and This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2024 The Author(s). Clinical & Experimental Allergy published by John Wiley & Sons Ltd. The first three authors contributed equally to this article. 920 Clinical & Experimental Allergy, 2024 pathways contributing to FPIES, including epithelial barrier dysfunction and immune dysregulation. While these results are novel, they are preliminary and need further validation in a second cohort of patients. 1 | Introduction Food proteininduced enterocolitis syndrome (FPIES) is a non- IgE- mediated food allergy predominantly affecting young children. The symptoms include profuse vomiting, often lethargy and potentially shock usually 1–4 h after ingestion [1]. While FPIES involve systemic innate immune activation, their relation to foodspecific exposure, gut inflammation and symptoms remains unclear [2]. Understanding disease mechanisms is crucial to develop specific diagnostic, prognostic, therapeutic and preventive strategies, currently unavailable in FPIES. The identification of diagnostic and prognostic biomarkers is a research priority by the FPIES patient/parent community; the lack of available tests and limited awareness often leads to missed or delayed diagnosis, compromising safety and quality of life [1]. Allergic diseases arise from intricate interactions between environmental and genetic factors. Studying patients' genetics can shed light on the pathophysiology and susceptibility to the disease. Twin studies on the heritability of food allergies suggest a genetic component [3]. Genetic studies in IgE- mediated food allergy have identified several genes loci associated with immune dysregulation and epithelial barrier dysfunction, including FLG (‘Filaggrin gene’) mutations [4]. These findings have allowed major developments in the recent years such as biologic therapy targeting key immune pathways for severe cases[5], or food allergy prevention by early food introduction in atrisk individuals [6]. Also, a growing list of inborn errors of immunity leading to heritable monogenic allergic disorders has been recently identified and coined as Primary Atopic Disorders (PAD); many of them associate food allergy due to immune dysregulation [7]. To our best knowledge, no studies have explored the association between genome variability and FPIES. This study aimed to investigate the potential genetic basis of FPIES by conducting a whole exome association study. The results revealed several single nucleotide polymorphisms (SNPs) and genes as new candidates for understanding the genetic susceptibility to FPIES. The identified SNPs and genes suggest potential involvement of immune dysregulation and epithelial barrier dysfunction as contributing mechanisms to FPIES. The discovery of new genetic susceptibility biomarkers holds promise for anticipating and improving the diagnosis and treatment of patients, contributing to a better understanding of the disease pathophysiology. 2 | Methods 2.1 | Sampling This is an observational, multicentre, prospective study recruiting patients with a diagnosis of acute FPIES confirmed by a standardised oral food challenge (OFC) to the culprit food (BIOFPIES study, ethics committee reference: 2017/396). OFC were conducted as a part of routine care across a network of tertiary hospitals from Spain and Italy. Informed consent from parents/guardians was obtained. Participants exclusion criteria were as follows: (i) systemic treatment with immunosuppressants or monoclonal antibodies, (ii) immunodeficiency, (iii) previous bone marrow transplant and (iv) other comorbidity that might interfere with the OFC outcome assessment or immune/inflammatory response. The patient cohort consisted of 41 children aged 1–12 years and one adult. Blood (EDTA) samples were collected at baseline, prior to starting the OFC (see clinical characteristics in Table1). 2.2 | Suitability for OFC Contraindications to proceed with OFC on the day were assessed and OFC and blood extraction were postponed in the event of any of the following circumstances: (a) vaccination or systemic corticosteroids (oral, intramuscular and intravenous) in the previous 2 weeks, (b) proton pump inhibitors, first generation antihistamines (i.e., hydroxyzine [Atarax], chlorphenamine, ketotifen) in the previous 3 days, (c) second generation antihistamines (i.e., cetirizine, loratadine, desloratadine and fexofenadine), in the previous 5 days, (d) inhaled/nebulised salbutamol, montelukast, oral nedocromil, oral cromoglycate in the last 24 h. Physical examination is carried out to ensure that there are no significant abnormalities that could influence the OFC outcome assessment. OFC were performed following current best practice recommendations [1] as described elsewhere [8]. Briefly, this involved incremental doses of the culprit food reaching a cumulative dose equal or above an ageappropriate portion. Subsequent doses were given only in the absence of significant symptoms suggesting a reaction. Ageappropriate portions were determined using national references. Patients were under observation for at least 4–6 h following their reaction onset. A telephone call was conducted 48 h postdischarge to enquire about delayed symptoms as this might impact on OFC outcome assessment. The OFC outcome as positive/negative/ inconclusive and reaction severity as mild/moderate/severe were assessed following current consensus criteria [1]; FigureS1. For genomic comparisons, control groups were drawn from the 1000 Genomes Project [9] (http:// www. 1000g enomes. org; hereafter referred to as 1000G). The Iberian population (IBS; n = 107) was used as control group for the association tests in the discovery phase. In the validation phase, the candidate SNPs and genes underwent further testing using other European data sets: CEU (n = 99), TSI (n = 111), GBR (n = 100) and a merged European data set (combining IBS, CEU, TSI and GBR; n = 417). 2.3 | Sequencing DNA was isolated from blood samples using the Wizard Genomic DNA Purification Kit (Promega). The concentration of the samples was analysed by fluorometric quantification with the Qubit system, the degree of purity by spectrophotometry with the 13652222, 2024, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/cea.14564 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 921 NanoDrop system and the integrity of the DNA by TapeStation. Pairedend sequencing (2 × 100 bp) of the SureSelectXT libraries, previously enriched, indexed and multiplexed, has been performed on the NovaSeq 6000 platform (Illumina, Inc). The quality of FastQ files for sequenced samples was assessed using FastQC v0.11.9 [10]. Subsequently, the FastQ files were aligned with BWA 0.7.17- r1188 [11]. The alignment of sequence reads, clone sequences and assembly was carried out using the BWAMEM algorithm against the human reference genome 38 (HG38). Mapping quality was evaluated using Mosdepth 0.3.2 [12], Bamtools 2.5.2 stats [13] and Samtools 1.15 stats [14] with assistance from the MultiQC 1.12 collector [15]. The resulting Bam files were then processed following best practices outlined in Gatk 4.2.5.0 [16] and annotated with the Combined Annotation Dependent Depletion (CADD) database GRCh38- v1.6. 2.4 | Statistical Analysis Statistical power was computed using the mitPower tool (Pardo- Seco etal. [17]). For a nominal significance threshold of 0.05, the statistical power to detect an odds ratio (OR) greater than three with a SNP frequency of 0.25 in controls exceeds 80% for the sample size in this study. Similarly, for a frequency of 0.1 in controls, the OR required to achieve an 80% statistical power should be four or higher. The databases were initially filtered based on frequencies, removing monomorphic, duplicated, indels or triallelic variants. Additionally, SNPs with a minimum allele frequency (MAF) below the 0.01 threshold and/or a genotyping rate below 99.5% were eliminated. Sex chromosomes were ignored from the analysis. Variants in Hardy–Weinberg disequilibrium (p < 0.001) were eliminated from the analysis, leaving a total of 45,302 SNPs for further analysis. To address the potential overrepresentation of allelic data from duplicated or closely related individuals, a relationship analysis was conducted using identityby- descent pairwise values, comparing them with theoretical values of family relationships. A population analysis was performed to mitigate the influence of genomic interbreeding and reduce the confounding effect of potential population substructure on the falsepositive rate. This analysis entailed referencing data from the 1000G and using the ADMIXTURE software [18] to estimate individual ancestry Summary • FPIES genetics reveals significant variants and associated genes; particularly, DGKZ, RBM8A, SIRPA and ATG16L1. • Candidate genes are associated with TGF- β signalling, TAR syndrome, increased neutrophils/monocytes in inflamed tissues and IBD. • Pathways involved in FPIES pathophysiology are related to epithelial barrier dysfunction and immune dysregulation. TABLE 1 | Summary of the main clinical characteristics of the FPIES cohort. Participants (n = 38) Gender—n (%) Female 15 (39.5) Male 23 (60.5) Age (months)—Median [IQR] 62.5 [51.5] Age at first reaction (months)—Median [IQR] 9.50 [3.75] FPIES culprit food—n (%) Milk 7 (18.4) Vegetables 2 (5.3) Egg 7 (18.4) Fish 21 (55.3) Fruit 1 (2.6) Allergic comorbidities—n (%) IgE- mediated food allergy 7 (18.4) Multiple food FPIES 4 (10.5) Other non- IgE- mediated food allergy 2 (5.3) Asthma 8 (21.1) Allergic rhinitis 3 (7.9) Atopic dermatitis 9 (23.7) Family background—n (%) Atopy/allergic diseases 19 (50.0) FPIES 1 (2.6) Clinical manifestations at OFC—n (%) Vomiting 38 (100) Lethargy 34 (89.5) Pallor 36 (94.7) Diarrhoea 3 (7.9) Hypotension 10 (26.3) Hypothermia 4 (10.5) Hypotonia 3 (7.9) Missing data 3 (7.9) Increased neutrophil count at OFC—n (%) Increased compared to baseline 32 (84.2) Missing data 1 (2.6) Severity of reaction at OFC—n (%) Mild 2 (5.3%) Moderate 19 (50.0%) Severe 17 (44.7%) 13652222, 2024, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/cea.14564 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 922 Clinical & Experimental Allergy, 2024 estimation through multilocus SNP data. Thirteen population datasets representing major ancestral groups, including Europeans, East Asians, sub- Saharan Africans and Native Americans, were considered. Each sample bar represents an individual, with the coloured proportions indicating the estimated ancestry derived from the unsupervised ADMIXTURE analysis; this analysis represents a number K = 4 ancestral clusters, each suggesting one of the four main continental regions: Africa, Asia, Native America and Europe. A multidimensional scaling analysis (MDS) was carried out, using identityby- state values to investigate patterns of genome variation within cohort genomes and reference populations. Subsequent steps encompassed a singlepoint SNP allele association test, and a genebased association testing. Allele statistical association was performed using Χ2 exact test for variants with a minor allele frequency (MAF) exceeding 0.05 in healthy controls and Hardy–Weinberg equilibrium p value above 0.001 for both the case and the control cohorts. We calculated the genome inflation factor lambda for all cohorts used in the discovery and validation comparisons to prevent potential bias due to population stratification; these values were used to adjust p values (Table 2). The genebased association test utilised the variant collapsing method Sequence Kernel Association test (SKAT) [19]. Variants were weighted based on their CADD values to account for their differing impacts on disease [20]. Genes with fewer than two variants were removed. To address multiple testing and mitigate false positives, a Bonferroni correction was applied. Data curation, singlepoint analysis, relationship analysis and population analyses were conducted using PLINK v1.9 [21, 22], while graphical representations, such as MDS, were generated using R v4.2.2 software [23]. Genebased association tests were computed using the SKAT R package [24]. The functional analysis of statistically significant genes was performed through an overrepresentation analysis using the Clusterprofiler R package [25] and using the biological processes from the Gene Ontology (GO) as reference database. In addition, we explored genomewide coexpression correlations to investigate gene–gene interactions using the Correlation AnalyzeRpackage (17). We utilised gene expression correlation data from samples of both healthy intestinal tissues (n = 5,356) and stomach tissues (n = 320) as a reference (see Supplementary Methods S1 for details). To examine the biological impact of the variants detected associated with the FPIES condition, we conducted a transcriptomewide association study (TWAS) analysis. We have integrated the summary statistics results from the exome association test and expression quantitative trait locus (eQTL) data from different tissues related to the digestive system, namely stomach, ileum, pancreas, muscularis oesophagus, mucosa oesophagus, gastroesophageal oesophagus, transverse colon and sigmoid colon [26] (see Supplementary Methods S1 for details). Gene set enrichment analysis for the traitassociated genes derived from MetaXcan was carriedout with the Genotype Imputed Gene Set Enrichment Analysis (GIGSEA) package [27] and KEGG (Kyoto Encyclopedia of Genes and Genomes [28, 29]) as pathways reference database. A weighted multiple linear regression model was applied to account for redundancy in gene sets and n = 10,000 permutations to assess the significance of regression coefficients. 3 | Results 3.1 | Clinical Characteristics of the FPIES Cohort We recruited 41 patients with acute FPIES confirmed by a standardised OFC to the culprit food. After excluding three samples showing a non- European ancestry genetic background (see below), a total of 38 samples were included for the downstream analysis (Table1). Details on the number of patients recruited per centre are included in the TableS1. The final cohort comprised 15 females and 23 males (median age 62.5 months). OFC led to the classification of patients in severe (n = 17), moderate (n = 19) and mild (n = 2) phenotype (Table1). It is noteworthy that most reactions were triggered by fish (55.3%), although our cohort also included a significant proportion of cases with FPIES to other leading FPIES causes such as cow's milk (18.4%) and egg (18.4%). Other allergic comorbidities in different proportions were reported for our FPIES cohort, being atopic dermatitis (24%), asthma (21%) and IgE- mediated food allergy (18%) the most common coexisting conditions. In addition, we observed a remarkable proportion of patients with a familiar history of allergic diseases (50%). Finally, a significant increase in the number of neutrophils was detected at OFC for most of the patients (84%), which is in line with previous observations in FPIES patients. 3.2 | Population Genetic Characteristics of FPIES Patients Upon filtering out indels, triallelic and monomorphic variants and variants with a genotyping rate < 90%, we successfully identified 141,103 biallelic SNPs. The cohort was merged with the 1000G database yielding an overlapping set of 75,817 SNPs. Family relationship analysis revealed that all study samples are unrelated (Figure 1A). We then performed a MDS analysis (Figure1B,C) to identify outliers. MDS and ADMIXTURE (Figure 1D) analysis identified ancestral population clusters, with most samples falling within the European core with the GBR, CEU and IBS populations serving as European reference populations. Three samples showed a non- European ancestry genetic component and were therefore excluded for further analyses to maintain genetic homogeneity in cases and controls and mitigate population stratification impact. 3.3 | Single Nucleotide Polymorphism Association Test Variants in Hardy–Weinberg disequilibrium (p value < 0.001) were eliminated from the analysis. A total of 45,302 SNPs survived and were subsequently analysed. Then, we performed a singlepoint association test. The allele test (Figure2A) revealed four SNPs surpassing the Bonferroni threshold: rs201740330 (p value = 3 × 10−7; OR2T32P gene), rs200103703 (p value = 6 × 10−7; OR2T32P), rs872786 (p value = 5 × 10−7; RBM8A: ‘RNA binding motif protein 8A’) and rs9917044 (p value = 2 × 10−6; ZNF28). rs2241880 (p 13652222, 2024, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/cea.14564 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 923 value = 1 × 10−4; ATG16L1: ‘Autophagy Related 16 Like 1 gene’) falls close to the adjusted significance; and it is in high linkage disequilibrium (LD) (r2 = 0.93) with rs2289477 (p value = 5 × 10−5; ATG16L1) (Table2). All candidate SNPs were validated using three additional European healthy control cohorts from the 1000G database (GBR, TSI and CEU), as well a merged European data set. Adjusted p values for the inflation factors maintained their statistical significance across all cohorts (Table2). 3.4 | Gene- Based Association Test A gene level analysis, using a CADD- weighted SKAT approach yielded significant results, initially identifying 24,098 genes. After applying a criterion requiring each gene to contain at least two SNPs, the list was reduced to 18,859 genes. Seven genes passed the Bonferroni correction threshold (Figure2B), namely DGKZ (‘Diacylglycerol Kinase Zeta gene’; p value = 1 × 10−18), TMEM99 (p value = 3 × 10−11), SIRPA (‘Signal Regulatory Protein Alpha’; p value = 1 × 10−7), SLC9B1P4 (p value = 5 × 10−7), PABPC1 (p value = 1 × 10−6), RBM8A (p value = 1 × 10−6) and GnRHR2 (p value = 1 × 10−6); ATG16L1 was suggestively significant, nearly meeting the threshold for genomewide significance (p value = 1 × 10−5) (Table2). From the point of view of FPIES, DGKZ, SIRPA and RBM8A, ATG16L1 seem to be particularly interesting. All the genes were validated using additional European cohorts as reference controls (Table2). The enrichment analysis of the significant genes identified the nonsensemediated mRNA decay (NMD) as the only statistically significant pathway (adjusted p value = 0.02), with two of the associated genes (RBM8A and PABPC1) involved. In addition, RBM8A gene is located on chromosome 1q21. This region encompasses the epidermal differentiation complex, which includes the FLG gene (> 6,400 kilobases apart) [30]. Accordingly, and considering the association reported between risk of food allergy and FLG mutations, we investigated genomewide coexpression correlations to study interactions between RBM8A and FLG genes in intestine and stomach. We found that RBM8A and FLG genes, despite displaying markedly different expression values in both intestinal and stomach tissues (p value = 2 × 10−16; Figure 2C), exhibit significant relationship between their genomewide coexpression correlations (Intestine: R = −0.61; Stomach: R = −0.72; p value = −2 × 10−16 in both cases); data inferred from ARCHS [4]. 3.5 | Transcriptome- Wide Association Study TWAS identified four significant eQTL- regulated genes (p value < 0.05) in different tissues as potential candidates involved in FPIES pathogenesis: RBM8A in stomach and pancreas, ATG16L1 in the transverse colon PIAS3 in pancreas and RPIA in esophagus mucosa (Figure3A). RBM8A and ATG16L1 showed predicted expression levels that were significantly higher in patients with FPIES than expected for these tissues (zscore = 5.1 and 4.6, respectively), whereas lower predicted expression values were detected for PIAS3 and RPIA (zscore = −4.2 and −4.3, respectively) (Figure3B). FIGURE 1 | Family relationships and ancestryanalysis of exome data. (A) Family relationship: Ternary plot of theoretical (black dots) and FPIES (red dots) identityby- descent values. (B) MDS plot of pairwise individual identityby- state values of FPIES and 1000G data sets (inset: European cluster). (C) MDS plot for FPIES and IBS samples. (D) Admixture analysis considering FPIES cases and 1000G reference populations. Population codes in (C): https:// www. corie ll. org/1/ NHGRI/ Colle ctions/ 1000- Genom es- Proje ct- Colle ction/ 1000- Genom es- Project. 13652222, 2024, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/cea.14564 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 924 Clinical & Experimental Allergy, 2024 Further functional analysis of the imputed expression in these tissues highlighted several important pathways. Lysosomal activity related pathways were among the top significant processes in stomach and colon transverse. Top significant affected pathway in transverse colon was ATP- binding cassette (ABC) transporters. Other common significant pathways between stomach and transverse colon included different processes related to amino acids metabolism (Figure3C). 4 | Discussion We conducted a pioneering exomewide case–control association study to comprehend the genetic basis of the FPIES syndrome. We identified (and further validated using three additional European ancestry healthy controls cohorts) different SNPs and gene candidates statistically associated with FPIES. These genetic factors could help to understanding the genetic susceptibility to this condition. The protein encoded by DGKZ belongs to the eukaryotic diacylglycerol kinase family. DGKZ promotes transforming growth factor β (TGF- β) signalling [31]; TGF- β is a key tolerogenic cytokine which regulates intestinal epithelial barrier integrity by maintaining and restoring enterocyte' barrier function [32], and by regulating IgA production [33]. TGF- β and IgA are crucial to induce tolerogenic responses through allergenspecific immunotherapy to inhalant allergens [33]. Children with cow's milk- FPIES showed deficient TGF- β responses upon casein stimulation of peripheral blood mononuclear cells (PBMCs) and lower serum caseinspecific IgA levels compared with milktolerant children [34]. Additionally, reduced expression of TGF- β Type I receptor has been reported on epithelial and mononuclear cells in the lamina propria of duodenal biopsies in FIGURE 2 | Singlepoint and genebased association tests. (A) Manhattan and QQ plot of p values for allelic association test; (B) Manhattan and QQ plot of CADD- weighted SKAT models. (C) Higher variance stabilizing transformationadjusted (Vst) gene expression in RBM8A than FLG, and their genomewide coexpression correlations; both in intestine/stomach from healthy subjects. Red line in (A) and (B): Bonferroni threshold. Labelled in red SNPs/genes in the limit of significance. 13652222, 2024, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/cea.14564 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 925 FPIES [35]. Skin and gut epithelial barrier dysfunction due to FLG mutations has been identified as a crucial pathogenic and risk factor for IgE- mediated food allergy, even in the absence of atopic dermatitis [4]. A disruption of the epithelial barrier function via impaired TGF- β signalling and reduced IgA neutralisation capability in the gut microenvironment might be plausible mechanisms in FPIES, potentially leading to increased antigen penetration to the submucosa and antigenspecific lymphocyte stimulation [36]. Given the fundamental role of the TFG- β pathway in epithelial barrier function and generation of IgA responses, DGKZ might play a key role in the pathophysiology of FPIES. ATG16L1 has been found to be associated with inflammatory bowel disease (IBD) [37]. Thus, ATG16L1 is involved in autophagy, a complex cellular process crucial for intestinal homeostasis, that is dysregulated in IBD [38]. IBD has been associated with FPIES in adults [39]. A barrier function defect and immune dysregulation are key to IBD pathophysiology, and Crohn's disease is associated with an increased TABLE 2 | p Values (and OR) for associated SNPs and genes. Discovery Validation IBS λ = 1.04; n = 107 CEU λ = 1.35; n = 99 TSI λ = 1.21; n = 111 GBR λ = 1.32; n = 100 Europe λ = 1.14; n = 417 SNP (chromosome region) rs201740330–A (1q44) 2 × 10−7 (5.3) 5 × 10−7 (7.4) —*2 × 10−5 (5.0) 2 × 10−12 (6.5) Freq. (A)—FPIES 0.34 0.34 0.34 0.34 0.34 Freq. (A)—controls 0.09 0.07 0.05 0.09 0.07 rs200103703–C (1q44) 6 × 10−7 (4.1) 10 × 10−7 (6.8) 3 × 10−9 (9.1) 5 × 10−5 (4.4) 2 × 10−11 (5.9) Freq. (C)—FPIES 0.34 0.34 0.34 0.34 0.34 Freq. (C)—controls 0.09 0.07 0.05 0.12 0.08 rs872786–T (1q21.1) 5 × 10−7 (5.0) 2 × 10−4 (3.4) 6 × 10−5 (3.3) 5 × 10−6 (4.3) 4 × 10−7 (3.7) Freq. (T)—FPIES 0.68 0.68 0.68 0.68 0.68 Freq. (T)—controls 0.35 0.39 0.39 0.33 0.36 rs9917044–C (19q13.41) 2 × 10−6 (6.2) 8 × 10−4 (4.1) 3 × 10−4 (4.0) 8 × 10−3 (2.8) 3 × 10−6 (4.0) Freq. (C)—FPIES 0.25 0.25 0.25 0.25 0.25 Freq. (C)—controls 0.05 0.08 0.08 0.11 0.08 rs2241880–A (2q37.1) 1 × 10−4 (2.9) 0.003 (2.5) 0.002 (2.5) 2 × 10−2 (2.1) 5 × 10−4 (2.5) Freq. (A)—FPIES 0.67 0.67 0.67 0.67 0.67 Freq. (A)—controls 0.41 0.44 0.44 0.49 0.44 rs2289477–T (2q37.1) 5 × 10−5 (3.3) 0.001 (2.9) 8 × 10−4 (2.8) 0.009 (2.4) 2 × 10−4 (2.8) Freq. (T)—FPIES 0.69 0.69 0.69 0.69 0.69 Freq. (T)—controls 0.41 0.44 0.45 0.49 0.45 Gene RBM8A (1q21.1) 1 × 10−61 × 10−52 × 10−58 × 10−89 × 10−8 GnRHR2 (1q21.1) 1 × 10−61 × 10−52 × 10−58 × 10−89 × 10−8 ATG16L1 (2q37.1) 1 × 10−55 × 10−44 × 10−41 × 10−22 × 10−4 DGKZ (11p11.2) 1 × 10−18 10 × 10−15 3 × 10−11 2 × 10−11 1 × 10−20 TMEM99 (17q21.2) 3 × 10−11 5 × 10−13 7 × 10−14 1 × 10−10 2 × 10−12 PABPC1 (8q22.3) 1 × 10−63 × 10−92 × 10−85 × 10−76 × 10−8 SLC9B1P4 (22q11.1) 5 × 10−73 × 10−83 × 10−67 × 10−53 × 10−10 SIRPA (20p13) 1 × 10−71 × 10−82 × 10−91 × 10−63 × 10−13 Note: Discovery phase: FPIES versus IBS. Validation phase: FPIES versus CEU, GBR, and TSI data sets, and a merged European data set (combining IBS, CEU, GBR and TSI data sets); all sourced from the 1000G. All p values were adjusted for the inflation factor lambda. *MAF < 0.05. For SNPs, Freq. denotes allele frequency for the minor allele. 13652222, 2024, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/cea.14564 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 926 Clinical & Experimental Allergy, 2024 Th17 response, the key signature in FPIES [40]. The genetic association with ATG16L1 in both FPIES and IBD, and the increased Th17 signalling, suggest common mechanisms in FPIES and IBD. SIRPA is an immunoinhibitory receptor primarily expressed by myeloid lineage of immune cells, including neutrophil, monocytes, macrophages and dendritic cells. Numerous publications propose an association between elevated gene signatures of monocytes and neutrophils and IBD [41]. SIRPA, as other neutrophil/monocyteassociated genes, show upregulation in inflamed tissues, most likely due to the presence of an increased number of neutrophils/monocytes expressing SIRPα protein [42]. Although, the pathophysiology of FPIES remains incompletely elucidated, acute FPIES reactions involve profound innate immune activation including neutrophils, monocytes, eosinophils and lymphocytes [43]. Increase in neutrophils, and decrease in eosinophil and lymphocyte counts in peripheral blood has been reported, which might suggest migration of this type of cells to the gut tissue [43]. Innate immune dysregulation via the SIRPA might be a novel potential factor contributing to this picture. Epithelial barrier defects resulting from FLG mutations are a major risk factor for the development of atopic eczema, IgE- mediated food allergy, eczemaassociated asthma and allergic rhinitis [4, 44]. Interestingly, we have found a significant relationship between FLG and RBM8A genomewide coexpression patterns. This may indicate a functional link between both genes through the existence of crucially correlated and anticorrelated gene clusters. The functional connection between RBM8A and FPIES might point towards shared pathogenic mechanisms related to IgE- mediated food allergy. Additionally, the RBM8A gene has been found to be associated with cow's milk allergy. For instance, a deletion located at chromosome band 1q21.1, and other mutations fallen in RBM8A, are often associated with thrombocytopenia with Absent Radius (TAR) syndrome. TAR associates cow's milk allergy in up to twothirds of individuals, and susceptibility to recurrent bouts of gastroenteritis [45–47]. An association between cow's milk allergy and carriers of 1q21.1 deletion without TAR syndrome has been also documented [45]; this observation could result from incomplete penetrance of 1q21.1 deletion and other mutations in RBM8A associated to TAR [45]. FIGURE 3 | TWAS analysis based on EWAS FPIES data. (A) Manhattan and QQ plot representations of the TWAS analysis in various tissues. The colour scale represents SNP density in the chromosomes. Blue and red dashed line thresholds refer to p values 10−6 and 10−4, respectively. (B) TWAS zscore of the most significant genes. (C) Enrichment analysis for genes identified in the TWAS analysis conducted in the stomach, pancreas and transverse colon. 13652222, 2024, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/cea.14564 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 927 Regarding other genes identified in this work as associated with FPIES (GnRHR2, PABPC1, SLC9B1P4 and TMEM99), minimal information is available related to their biological function or potential role in disease. The enrichment analysis of significant genes revealed the NMD pathway as statistically significant. NMD eliminates mRNAs with premature translation–termination codons and plays a role in premRNA splicing, although its specific relevance to FPIES remains uncertain. TWAS pinpointed RBM8A (stomach and pancreas), and ATG16L1 (transverse colon) as the most promising candidates, each exhibiting predicted expression values higher in FPIES than expected. The functional assessment of imputed gene expression in these tissues underscored pathways linked to lysosomes (in the transverse colon and stomach), ABC transporters (in the transverse colon) and amino acid metabolism—all of which have previously been associated with IBD and other intestinal diseases [48]. Indeed, lysosomes are also key elements of the autophagic machinery. The study acknowledges certain limitations, including a small sample size and variations in patient age, potentially affecting statistical power (and therefore precluding the possibility to identify variation with lower effects) and result accuracy. In addition, other allergic diseases, relatively common in our FPIES cohort, might be a confounder factor in our results. However, no association has been reported between the genes identified in this work and other allergic diseases (beyond the abovementioned RBM8A gene and milk allergy). To our knowledge, no formal epidemiological studies have quantified the genetic heritability. Two sets of identical twins with FPIES with close similarities in clinical manifestations and age of onset have been reported, suggesting a strong genetic component [49]. Given the global disparities in FPIES characteristics across the globe [1], including regarding food culprits, further research should explore our findings in a diverse international cohort. 5 | Conclusions Our investigation represents the first case–control exome association study in FPIES, revealing gene and SNP candidates. These insights initiate genomic exploration in FPIES and provide further evidence of a role of genetics in the condition, suggesting involvement in epithelial barrier dysfunction and immune dysregulation—major pathogenic factors identified in IgE- mediated food allergy [4] and IBD [40]. While our results are novel, they are still preliminary and need additional validation in a second cohort of patients. Author Contributions A.G.- C., A.S., F.M.- T. and M.V.- O. conceived, designed and provided financial support to the study. M.V.- O. and L.A. analysed the clinical data. A.F., A.M., A.P., C.G.- M., E.G., F.M., G.Z.- I., G.M., I.C., J.D.M.- G., L.A., L.E., L.V., M.F.- R., M.T.- P., M.P., M.J.T., N.L.H.- M., P.G.- D., S.A., S.B., S.I., S.Q., S.V.- C., T.B., T.G., V.O.- A. and V.P. were involved in sample recruitment. A.C.- M., A.S., A.G.- C., J.P.- S. and X.B. analysed the data. A.C.- M., A.S., A.G.- C., J.P.- S. and M.V.- O. wrote the initial draft of the article. All the authors revised and contributed to the final version of the manuscript. Affiliations 1Genetics, Vaccines and Infections Research Group (GenViP), Instituto de Investigación Sanitaria de Santiago, Universidade de Santiago de Compostela, Santiago de Compostela, Galicia, Spain | 2Unidade de Xenética, Instituto de Ciencias Forenses, Facultade de Medicina, Universidade de Santiago de Compostela, and Genética de Poblaciones en Biomedicina (GenPoB) Research Group, Instituto de Investigación Sanitaria (IDIS), Hospital Clínico Universitario de Santiago (SERGAS), Galicia, Spain | 3Centro de Investigación Biomédica en Red de Enfermedades Respiratorias (CIBERES), Madrid, Spain | 4Allergy Section, Clinica Universidad de Navarra, Madrid, Spain | 5Section of Inflammation, Repair and Development, National Heart and Lung Institute, Imperial College London, London, UK | 6Department of Infectious Disease, Imperial College London, London, UK | 7Allergy Diseases Research Area, Pediatric Allergology Unit, Bambino Gesù Children's Hospital IRCCS, Rome, Italy | 8Allergy Unit, Meyer Children's Hospital IRCCS, Florence, Italy | 9Paediatric Allergy Section, Severo Ochoa University Hospital, Madrid, Spain | 10Clinical Analysis and Clinical Biochemistry Service, Severo Ochoa University Hospital, Madrid, Spain | 11Paediatric Allergy Section, Arquitecto Marcide Hospital, Ferrol, A Coruña in Galicia, Spain | 12Paediatrics Department, Hospital Clínico Universitario de Santiago de Compostela, Coruña, Galicia, Spain | 13Allergy Department, General University Hospital, Alicante, Spain | 14Paediatric Allergy Section, Vall D'Hebron University Hospital, Growth and Development Research Group, Vall d'Hebron Research Institute (VHIR), Barcelona, Spain | 15Pediatric Allergy Unit, Hospital General Universitario Gregorio Marañón, Gregorio Marañón Health Research Institute (IiSGM), Madrid, Spain | 16Allergy and Clinical Immunology Department, Hospital Sant Joan de Déu, Barcelona, Spain | 17Immunology Department, CDB, Hospital Clínic de Barcelona, Barcelona, Spain | 18IDIBAPS, Universitat de Barcelona, Barcelona, Spain | 19Paediatric Allergy Section, General University Hospital, Malaga, Spain | 20Allergy Department, Hospital Clinico San Carlos, Instituto de Investigación Sanitaria San Carlos (IdISSC), Madrid, Spain | 21Allergy Department, Hospital Clinico San Carlos, Instituto de Investigación Sanitaria San Carlos (IdISSC), Universidad Complutense, Madrid, Spain | 22Paediatric Allergy Section, Teresa Herrera Hospital, Coruna, Spain | 23Clinical Immunology and Primary Immunodeficiencies Unit, Allergy and Clinical Immunology Department, Hospital Sant Joan de Déu, Institut de Recerca Sant Joan de Déu and Universitat de Barcelona, Barcelona, Spain | 24Allergy Department, General University Hospital, Málaga, Spain | 25Allergy Research Group, Instituto de Investigación Biomédica de Málaga y Plataforma en Nanomedicina- IBIMA Plataforma Bionand, Málaga, Spain | 26Universidad de Málaga (UMA), Málaga, Spain | 27Allergy Clinical Unit, Hospital Regional Universitario de Málaga, Málaga, Spain | 28Department of Health Sciences, University of Florence, Florence, Italy | 29Department of Allergy, La Paz University Hospital, IdiPAZ, Madrid, Spain | 30Translational Pediatrics and Infectious Diseases, Department of Pediatrics, Hospital Clínico Universitario de Santiago de Compostela, Santiago de Compostela, Galicia, Spain Acknowledgements This study received support by: (i) Strategic Health Action, “Instituto de Salud Carlos III” (ISCIII) cofinanciados FEDER: TRINEO: PI22/00162; DIAVIR: DTS19/00049; Resvi- Omics: PI19/01039 (to A.S.), ReSVinext: PI16/01569, Enterogen: PI19/01090, OMICOVI- VAC: PI22/00406 (to F.M.- T.), BIOFPIES: PI19/00497 (to AG- C), (ii) Axencia Galega de Innovación (GAIN): IN607B 2020/08 and IN607A 2023/02 (to A.S.), GENCOVID (IN845D 2020/23 (to F.M.- T.), IIN607A2021/05 (to F.M.- T.) and IN677D 2024/06 (to A.G.- C.); (iii) Agencia Gallega de Conocimiento en Salud (ACIS): BIBACVIR (PRIS- 3, to A.S.), CovidPhy (SA 304 C, to A.S.); (iv) Spanish Ministry 13652222, 2024, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/cea.14564 by Uni Santiago Compostela, Wiley Online Library on [11/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License