Immunoglobulin G4 in Eosinophilic Esophagitis: Immune Complex Formation and Correlation with Disease activity

Abstract

Background:

Recent studies have shown deposition of immunoglobulin G4 (IgG4) and food proteins in the esophageal mucosa of eosinophilic esophagitis (EoE) patients. Our aims were to assess whether co-localization of IgG4 and major cow’s milk proteins (CMPs) was associated with EoE disease activity, and to investigate the proteins enriched in proximity to IgG4 deposits.

Methods:

This study included adult subjects with EoE (n=13) and non-EoE controls (n=5). Esophageal biopsies were immunofluorescence stained for IgG4 and CMPs. Co-localization in paired samples from active disease and remission was assessed and compared to controls. The proteome surrounding IgG4 deposits was evaluated by the novel technique, AutoSTOMP. IgG4-food protein interactions were confirmed with co-immunoprecipitation and mass spectrometry.

Results:

IgG4-CMP co-localization was higher in the active EoE group compared to paired remission samples (Bos d4, p=0.02; Bos d5, p=0.002; Bos d8, p=0.002). Co-localization was also significantly higher in the active EoE group compared to non-EoE controls (Bos d4, p=0.0013; Bos d5, p=0.0007; Bos d8, p=0.0013). AutoSTOMP identified eosinophil derived proteins (PRG 2 and 3, EPX, RNASE3) and calpain-14 in IgG4 enriched areas. Co-immunoprecipitation and mass spectrometry confirmed IgG4 binding to multiple food allergens.

Conclusion:

These findings further contribute to the understanding of the interaction of IgG4 with food antigens as it relates to EoE disease activity. These data strongly suggest immune complex formation of IgG4 and major cow’s milk proteins. These immune complexes may have a potential role in the pathophysiology of EoE by contributing to eosinophil activation and disease progression

Introduction:

Eosinophilic esophagitis (EoE) was first recognized as a distinct disease entity triggered by foods in the mid 1990’s.¹ It is now widely accepted that EoE is mediated by food-antigens, can lead to tissue remodeling and fibrosis, and appears to be increasing in prevalence.^2–4 By removing the six most common food triggers, up to 75% of patients will achieve clinical and histologic remission, but disease activity returns when these foods are re-introduced.^5,6 While EoE is known to be characterized by type 2 (T2) inflammatory responses, little is understood regarding antigen processing and the mechanisms by which specific foods drive chronic inflammation and remodeling.^4,7–9

The discovery of immunoglobulin G4 (IgG4) deposits in the esophageal tissue of patients with active EoE first implicated a role for IgG4 in the inflammatory response characteristic of this disease.¹⁰ Additionally, these IgG4 deposits were noted to resemble immune complexes on ultrastructural evaluation.^10,11 Multiple studies have demonstrated the clinical utility of tissue IgG4 as a marker differentiating active EoE from acid reflux disease.^11–13 Esophageal IgG4 levels also correlate with esophageal eosinophil counts, histologic scoring, and T2 cytokines.¹⁴ Moreover, initial work on serum food-specific IgG4 show a correlation between higher IgG4 antibody levels and disease activity.^15,16 Whether IgG4 is an epiphenomenon of the underlying immune response or causally related to EoE pathogenesis remains unknown. Based on these studies, there is a correlation between disease activity and IgG4 levels, however, it has never been shown that IgG4 forms immune complexes with food proteins.

Food proteins have been shown to penetrate the esophageal mucosa and remain in the tissue for prolonged periods of time after antigen exposure.¹⁷ Knowing that cow’s milk is the main trigger of EoE^18–22 and that there is mucosal penetration of cow’s milk proteins (CMP), we hypothesized that IgG4 was binding milk antigen in the esophageal tissue. To test this hypothesis, we assessed co-localization of IgG4 and major CMPs (Bos d 4, Bos d 5 and Bos d 8) in paired esophageal samples from patients who had active disease or who were in remission from either swallowed steroids or CMP elimination diet. Samples from a control group who did not have EoE were also evaluated. Our study used a novel technology, Automated Spatially Targeted Optical Micro Proteomics (AutoSTOMP), to investigate the patient proteins and food antigens enriched in proximity to esophageal IgG4 deposits. We also evaluated the direct interaction between IgG4 and allergens using co-immunoprecipitation and mass spectrometry. Our findings strongly suggest that IgG4 forms immune complexes with food antigens in the esophageal tissue. We speculate that these immune complexes play a central pathogenic role in EoE, possibly via binding eosinophil Fcγ receptors resulting in eosinophil activation.

Methods:

Study Design:

We performed a case control study nested within the University of Virginia (UVA) EoE Cohort. The UVA EoE Cohort is a prospective, longitudinal cohort that was established at UVA in 2017 to assess the environmental, genetic, and immunologic predictors of EoE. Patients seen at the UVA Multidisciplinary EoE clinic or undergoing esophagogastroduodenoscopy (EGD) at UVA for evaluation of confirmed or suspected EoE were invited to participate and were followed longitudinally for up to 5 years. Tissue, serum, saliva, and urine samples were collected at the time of EGDs. Participants completed a questionnaire evaluating medical history and food/environmental exposures at enrollment which was reviewed at each subsequent endoscopy. The study was approved by the UVA Institutional Review Board, and written informed consent was obtained from all participants. Data were stored in a REDCap database.

Study Population:

Eighteen adult patients (>18 years old) were selected for immunofluorescence staining. Cases (n=13) were defined as patients who met EoE diagnostic criteria of ≥15 eosinophils per high powered field (eos/hpf), symptoms of esophageal dysfunction, and exclusion of other causes for esophageal eosinophilia.²³ Ten paired samples from cases with active disease and EoE in remission were compared. Remission in these samples resulted from swallowed steroids (n=5) or CMP elimination diet (n=5). Remission was defined clinically as <6 eos/hpf²⁴ and the absence of esophageal symptoms. Controls (n=5) were defined non-EoE patients with 0 eos/hpf on esophageal biopsies. Controls who had confirmed consumption of CMP at the time of endoscopy were selected for this study. Patients with active EoE who were confirmed to be milk non-reactive (n=3) were also selected for immunofluorescence staining. Frequency of CMP consumption was assessed using a modified food frequency questionnaire at time of enrollment into the UVA EoE cohort, and overall CMP consumption was reviewed at the time of each endoscopy. Exclusion criteria included patients <18 yo, presence of other esophageal pathology, or CMP consumption less than once per week. An additional 3 cohort patients with active EoE were selected using the criteria noted above for proteomic analysis.

Immunofluorescence staining:

Research biopsies were taken from multiple levels of the esophagus at the time of endoscopy. Samples were reviewed by a gastrointestinal pathologist (SR) and were selected based on the presence of lamina propria and papilla using H&E staining.²⁵ Tissue samples were formalin fixed, paraffin embedded, and sectioned by the UVA Histology Core Laboratory. For analyses, samples were deparaffinized in xylene and graded ethanol solutions. Heat-induced antigen retrieval was performed using citrate buffer, pH 6.0. (Abcam, Cambridge, MA). Tissue samples were blocked using 10% donkey serum (Sigma, St. Louis, MO), 0.1% triton X-100 and 1 μg/ml Human Fc Block (BD Pharmingen, Sparks, MD) for 3 hrs, and then incubated with primary antibody for 16 hrs at 4°C. [Alpha-lactalbumin ab112972, 1:250; Casein ab166596, 1:250 (Abcam, Waltham, MA); Beta-lactoglobulin Ab00650–23.0, 1:100 (Absolute antibody, Boston, MA); IgG4 367M-1, 1:100 (Cell Marque, Rocklin, CA)].

Sections were washed six times for 5 min each in TBST (TBS + 0.1% Tween 20) and then incubated with secondary antibodies for 1 hr at 23° C. Alexa Fluor 488 donkey anti-mouse IgG A21202, 1:1000 (Invitrogen, Waltham, MA) and Alexa Fluor 555 donkey anti-rabbit IgG Ab150074, 1:1000 (Invitrogen, Waltham, MA). Sections were washed again and aqueous mounted in Fluoro-Gel II (Electron Microscopy Sciences, Hatfield, PA) with DAPI (4’, 6-diamidino-2-phenylindole) to stain nuclei.

Confocal microscopy and statistical analysis:

Images were acquired using a Zeiss LSM 700 confocal microscope (Zeiss Microimaging, Thornwood, NY). Each tissue was scanned under 10x magnification and IgG4 enriched areas were identified. IgG4 enriched areas were then further analyzed under higher magnification (20x) and images were acquired using Zeiss LSM 700 confocal microscope software. Microscope settings were kept consistent within sample groups. A minimum of 3 images were obtained for each tissue sample. Co-localization of the pixel intensity of the first channel (i.e. Alexa Fluor 555) versus the second channel (i.e. Alexa Fluor 488) was determined for each image using Pearson’s correlation coefficient²⁶ on Imaris imaging software (South Windsor, CT). For IgG4 staining, the number of green voxels (volume pixels) were calculated for each image, and this number was divided by the total number of voxels to determine the percentage. In keeping with the histopathologic evaluation and quantification of eosinophils based on the area with highest eosinophil count, the image with the highest Pearson’s correlation coefficient was used for statistical analysis. Differences in correlation coefficients were compared using Wilcoxon Signed Rank Test or Mann-Whitney U test (GraphPad Prism software, San Diego, CA), as appropriate.

Co-immunoprecipitation:

Co-immunoprecipitation was performed using ThermoFisher Dynabeads Co-Immunoprecipitation kit (Waltham, MA). Anti-IgG4 367M-1, 1:100 (Cell Marque, Rocklin, CA) antibody was epoxied to the magnetic beads according to kit instructions. The tissue sample was mechanically disrupted then lysed in extraction buffer. The cell lysate was incubated with the anti-IgG4 coated beads. Protein complexes attached to the beads were eluted, denatured in Laemmli buffer (Bio-Rad, Hercules, CA) with β-mercaptoethanol then run on a 12% TGX tris/glycine polyacrylamide gel (Bio-Rad, Hercules, CA) for 10 min.

The gel sample was digested overnight at 37° and peptides were extracted from the polyacrylamide in a 100 μL aliquot of 50% acetonitrile/5% formic acid. This extract was evaporated to 15 μL for mass spectrometry (MS) analysis, using the rapid switching capability of liquid chromatography-mass spectrometry (LC-MS). A full scan mass spectrum was acquired to determine peptide molecular weights followed by product ion spectra (Top10 HCD) to determine amino acid sequence scans. The data were searched against Uniprot Human, Uniprot SwissProt, and allergen databases using the Sequest search algorithm. Scaffold Proteome Software 5.1.2 (Portland, OR) was used to validate LC-MS data. Positive protein identification was defined as peptide probability score >95%, protein probability score >99%, and a minimum of two peptide hits per protein.

Automated Spatially Targeted Optical Micro Proteomics (AutoSTOMP):

Automated Spatially Targeted Optical Micro Proteomics (AutoSTOMP) was used to evaluate protein enrichment within 0.25μm of the anti-IgG4 immunofluorescence signal in situ, from sectioned patient biopsies.²⁷ Esophageal tissue biopsies (1 mm) were collected via endoscopy from 3 patients with active EoE from the cohort as previously described. Immunofluorescence staining was performed with anti-IgG4 367M-1, 1:100 (Cell Marque, Rocklin, CA) antibody and IgG4 ‘structures of interest’ were visualized by confocal microscopy on a Zeiss LSM880 microscope (Zeiss Microimaging, Thornwood, NY) using Zen Black software (White Plains, NY). The biopsy perimeter and IgG4 positive regions were defined by the user to create tile arrays of structures of interest. Images were converted to binary map files identifying the pixel coordinates of each IgG4 positive structure of interest. The map files were used to guide excitation of the IgG4 regions by the two-photon laser. Delivery of 395nm light was used to photo conjugate the benzophenone-moiety of biotin-dPEG3-benzophenone (Sigma QBD 10267, St. Louis, MO) tag to any protein within the structures of interest.

Following photo-labeling, each sample was detached from the coverslip. Excess, unconjugated Biotin-BP was rinsed with 50:50 (v/v) DMSO/water 3 times, then with water 3 times. The slides were stored at −80°C before processing replicates in tandem. EoE biopsy sections were lysed in DTT/SDS buffer^28,29 (0.1 M Tris-HCl, 0.1 M DTT, 4% SDS, pH=8.0) at 99°C for 1 h. Tissue lysates were then diluted 1:10 in TBS-0.1% SDS, then incubated with streptavidin (SA) magnetic beads (Pierce #88817, Waltham, MA) at room temperature for 1 h. Biotinylated proteins were precipitated by a magnet. The unbound proteins were collected as the ‘flow through’ fraction and precipitated with 100% trichloroacetic acid then resuspended in laemmli buffer at 96°C for 5 min. The biotinylated proteins were eluted from the magnetic beads in laemmli buffer at 96°C for 5 min and collected as the AutoSTOMP fraction. The fractions were resolved in the SDS-PAGE gel and a gel fragment was excised for mass spectrometry analysis at the University of Virginia Biomolecular Analysis Facility, as described in the Online Supplement.

MaxQuant³⁰ (versions 1.6.14.0, Max Planck Institute of Biochemistry, Planegg, Germany) was used to analyze raw mass spectra data following the label-free quantification (LFQ) protocol.³⁰ T-SNE clustering were applied in Perseus³¹ (versions 1.6.14.0, Max Planck Institute of Biochemistry, Planegg, Germany). For each patient, paired ‘IgG4’ and ‘flow-through’ samples were aligned to a fasta database of previously identified food allergens (duplicate Uniprot IDs were excluded when combining those fasta files) using MaxQuant. Data were analyzed by Log2 transformation and imputation using the default Perseus method and allergens were ranked by fold-change over ‘flow-through’. The resulting data were plotted in R (www.r-project.org) with the installed packages “ggplot2”, “ggrepel”, “heatmap.2” or using GraphPad Prism (version 8.2.1). Detailed instructions and source codes are available at https://github.com/boris2008/AutoSTOMP_2.0.git. The presence of proteins of interest in proximity to IgG4 was confirmed via immunofluorescence, using the following primary antibodies: EPX (Novus EPO104), PRG2 (Proteintech 10766–1-AP), RNase3 (Proteintech 55338–1-AP).

Results:

Study Population:

Eighteen patients were enrolled in this study (13 EoE and 5 controls) for immunofluorescence staining, as described in Table 1. The mean age of EoE patients was 35.9 years (range 20–56 years old) and of control patients was 23.2 years (range 19–34 years old). Immediate food allergies were reported in 5/13 EoE patients and 2/5 control patients. No immediate reaction to CMP was reported within this nested cohort. Of the EoE patients, no patient in the CMP elimination group had immediate food reactions. Control patients reported symptoms of dysphagia, but biopsies were not consistent with the diagnosis of EoE.

Table 1:

Demographic and Clinical Characteristics of EoE vs Control Groups

Characteristic	EoE n=13	Control n=5	p value
Age (y)	35.9 (20−56)	23.2 (19−34)	0.008
Male Sex	8 (61.5%)	2 (40%)	0.6
Non-Hispanic White Race	11 (84.6%)	5 (100%)	>0.99
Atopic Conditions  Eczema  Asthma  Rhinitis  Immediate Food Allergy	4 (30.8%) 8 (61.5%) 10 (76.9%) 5 (38.5%)	2 (40%) 3 (60%) 3 (60%) 2 (40%)	>0.99 >0.99 0.58 >0.99

IgG4-CMP co-localization correlates to disease activity:

IgG4-CMP co-localization was consistently observed in active EoE cases, but not in patients who were in remission or in controls. Pearson’s correlation coefficients for IgG4-CMP co-localization were higher in the active EoE group compared to paired remission samples for all three milk allergens evaluated (Bos d 4 median 0.38 v. 0.0001, p=0.02; Bos d 5 median 0.68 v. 0.33, p=0.002; Bos d 8 median 0.36 v. 0.03, p=0.002). In all three milk allergen groups, there was also a statistically significant difference in IgG4-CMP co-localization when comparing active disease samples to non-EoE controls (Bos d 4 median 0.38 v. 0.0001, p=0.0013; Bos d 5 median 0.68 v. 0.14, p=0.0007; Bos d 8 median 0.36 v. 0.05, p=0.0013). In all three milk allergen groups, there was no significant difference when comparing remission disease state to healthy controls (Figures 1–3). When examining the treatment groups separately (i.e. swallowed steroids or diet), there was a trend towards decreased IgG4-CMP co-localization for all milk proteins, but this did not reach statistical significance (Table E1). When comparing IgG4-CMP co-localization in milk triggered and non-milk triggered groups, two of the three milk proteins demonstrated significantly more co-localization in milk-triggered EoE (Bos d 4 median 0.31 v. 0.15, p=0.25; Bos d 5 median 0.67 v. 0.33, p=0.04; Bos d 8 median 0.39 v. 0.12, p=0.04). Finally, when examining only IgG4, there was a significant decrease in patients treated with either swallowed steroids or diet, compared to those with active disease (swallowed steroids: 0.2% v. 0%; p=0.0012; diet: 0.1% v. 0.0%; p=0.0001).

Figure 1:

Esophageal biopsies with immunofluorescence staining for IgG4 (green), Bos d 5 (red), and nuclei (blue). (a/c) Two patients with active EoE demonstrating co-localization of IgG4 and Bos d 5 (yellow). (b/d) Paired samples from the same patients in remission on swallowed steroids from fig. 1a/1c, respectively.

Figure 3:

Pearson’s correlation coefficient for co-localization pixel intensity of IgG4 and milk proteins (a) Bos d 4, (b) Bos d 5, and (c) Bos d 8. Between group comparisons of active EoE and paired remission samples (n=10) and controls (n=5) were made using Mann-Whitney U test.

IgG4 directly binds to allergens in esophageal mucosa:

The use of co-immunoprecipitation technique confirmed protein-protein interactions between IgG4 and food allergens in a representative active EoE patient not on treatment. The patient was consuming an unrestricted diet at the time of endoscopy with daily ingestion of the top four EoE food triggers (cow’s milk, egg, wheat, and soy). The food allergens identified by mass spectrometry were cow’s milk (Bos d 5, 6, 10), almond (Pru du 6), tuna (enolase), sesame (Ses I 6), Tilapia (Ore m 4), pea (Pis s 1), and cherry (Pru av 1). Additionally, multiple allergens for dust mite (elongation factor 2, Der p 14) were also present.

Eosinophil-derived proteins are present in IgG4-enriched areas:

AutoSTOMP-labeled proteins associated with IgG4-enriched areas of interest in esophageal biopsies from three different active EoE patients. 2,045 human proteins were identified in the ‘flow through’ and ‘IgG4’ fraction areas (Figure E1). Consistent with a lower complexity of the IgG4-region proteome compared to the total biopsy sample; 10.25% of proteins identified were significantly enriched in the ‘IgG4’ fractions whereas 49.7% of the proteins were significantly lower in the ‘lgG4’ fractions relative to the ‘flow through’ fractions (Figure E1). There were several eosinophil-derived proteins present in the IgG4-enriched areas including Proteoglycans (PRG2 and 3), eosinophil peroxidase (EPX), and eosinophil cationic protein (RNASE3) (Figure 4). Calpain-14 (CAPN14), a protease that has been linked to epithelium disruption in EoE, was also significantly elevated in IgG4-rich areas compared to the non-IgG4 enriched areas in the same tissue sample (Supplementary excel file). Immunofluorescence staining confirmed the co-localization of PRG2, EPX, and RNASE3 with IgG4, and these deposits appear to be located adjacent to cells with bilobed nuclei, suggesting proximity to eosinophils (Figures E2 and E3).

Figure 4:

Of the 2,045 human proteins identified with AutoSTOMP, 10.26% of the proteins were significantly enriched and 49.7% of the proteins were significantly lower in the ‘lgG4-rich’ fractions relative to the ‘flow through’ fractions. Eosinophil proteins proteoglycans (PRG) 2 and 3, eosinophil peroxidase (EPX), and eosinophil cationic protein (RNASE3) and also Calpain-14 (CAPN14g) were among the proteins most highly enriched in the IgG4 fraction (red circles). Plotted as -log10 p-value (y axis) versus the log2 fold-change (x axis) of the protein abundance averaged among the three patient biopsies with a false discovery rate < 0.05.

DAVID Bioinformatics Resource analysis indicated that the IgG4+ areas were enriched for Gene Ontology Biological Process (GOBP) terms involved in tissue remodeling and wound healing relative to the flow through. By contrast, the flow through, which is predominantly IgG4-negative tissue, was enriched to GOBP terms representative of homeostatic cell and tissue processes (Figure E4). Student’s t-test (permutation-based The MS raw data and MaxQuant search data are available at PRICE (https://www.ebi.ac.uk/pride/ accession: PXD037663. FDR < 0.05).

We next asked if any of the peptides which did not align to the human proteome library could be food-derived proteins by searching these masses against a food allergen database. There were notable differences in dietary protein abundance between the three patient samples. Patient 1 had enriched levels of milk (Bos d 11) and wheat (low molecular weight glutenin) in the ‘IgG4’ fraction relative to the ‘flow through’. Patient 2 had notable elevations of wheat (high and low molecular weight glutenin), milk (Bos d 6, 10), and sesame (11S globulin seed storage protein) in the ‘IgG4’ fraction relative to the ‘flow through’ fraction. In the third patient, wheat (chafe tropomyosin) and soybean (α-subunit of β-conglycinin) were more abundant in the ‘IgG4’ fraction relative to the ‘flow through’ (Figure 5).

Figure 5:

Mass spectra identified with AutoSTOMP were searched against a food allergen database. IgG4 regions were enriched for milk and wheat peptides (a, subject 1), wheat and soybean peptides (b, subject 2), or wheat, milk, and sesame peptides (c, subject 3). Data represents the log2 fold change of peptide abundance in the ‘IgG4’ region relative to the ‘flow through’ for each patient, plotted in rank order.

Discussion:

In this study, we demonstrate that IgG4 co-localizes with multiple cow’s milk proteins in active EoE, and that this co-localization significantly decreases in the same patient with disease remission. We further demonstrated that this IgG4-CMP co-localization was significantly lower in patients who did not have milk-induced EoE. We then used co-immunoprecipitation to confirm the protein-protein interactions of IgG4 with various food antigens. Finally, using novel proteomic analyses, we discovered multiple eosinophil-derived proteins and food proteins in IgG4-rich areas of esophageal biopsies.

IgG4 is conventionally thought to be a marker of immune tolerance in allergic disease due to its unique biological functions among the IgG subclasses. First, weak interchain disulfide bonds allow for Fab-arm exchange resulting in a “functionally monovalent” immunoglobulin, thus limiting IgG4’s ability to cross link antigens and form immune-complexes.³² Additionally, IgG4 has high affinity for inhibitory receptor, FcγIIb. Clinically, high levels of venom sIgG4 are found in non-allergic bee-keepers,³³ high levels of cat sIgG4 are found in non-allergic cat owners,³⁴ and high levels of allergen sIgG4 are found during allergen immunotherapy.³⁵ Furthermore, IgG4 is a soluble molecule and should not be present in the tissue unless bound to another protein or cell.³⁶ This raises the important question of why tissue IgG4 deposits are seen in EoE.

Here, we show that major milk proteins (Bos d 4,5,8) co-localize with IgG4. However, co-localization merely demonstrates the spatial overlap of IgG4 and cow’s milk proteins of interest within the tissue and does not necessarily indicate a direct protein-protein interaction. We found, via co-immunoprecipitation and mass spectrometry, that IgG4 is indeed bound to food proteins, which may explain why it remains present in the tissue. The bivalent nature of IgG4 could also be advantageous in the food protein rich milieu of the esophagus, allowing immune complexes to form as each arm binds to different food proteins. Interestingly, wheat and egg (known EoE food triggers) were not detected using mass spectrometry despite our patient consuming these foods daily prior to endoscopy. Dust mite allergens, however, were present and have been an implicated disease trigger in a subset of EoE patients.³⁷ Further studies need to be performed to determine whether this novel technique could be used to identify EoE triggers in individual patients.

It is also interesting to note that each milk protein had differing amounts of co-localization, and Bos d 5 co-localized to a much greater degree than Bos d 4 or 8. One explanation is that there may have been higher primary antibody affinity to Bos d 5 as compared to Bos d 4 and 8, thus higher protein detection with immunofluorescence. Alternatively, it is possible that unique properties of Bos d 5 make it more immunogenic in EoE. Two studies examining the adaptive immune response to Bos d 5 suggest that a hydrophobic intramolecular binding site on Bos d 5 regulates its immunogenicity. When this site is empty, Bos d 5 promotes expression of CD4 on helper T cells and increased production of the T2 cytokines IL-10 and IL-13 compared to a “loaded” site where there was a decreased CD4+ expression.^38,39 It was previously hypothesized that milk processing alters the hydrophobic binding site on Bos d 5 making it more immunogenic.^40,41 Further studies are needed to examine the potential specific role of Bos d 5 since this seems to be the dominant antigenic component of cow’s milk and more than half of EoE patients respond to cow’s milk protein elimination diet.¹⁸

Another important finding was the difference in IgG4-CMP co-localization when comparing active EoE samples of milk allergic and non-milk allergic patients. All 8 patients in this group underwent food elimination diet with systematic re-introduction of foods followed by endoscopy to determine exact food triggers. Two of the three CMPs (Bos d 5,8) demonstrated significantly higher co-localization with IgG4 in the patients who had milk-triggered disease. Bos d 4 did not reach statistical significance which may be due to the small sample size. These findings suggest that the interaction between IgG4 and food allergens may vary depending on whether the food is a trigger, which could be harnessed for future diagnostic tests.

AutoSTOMP is a novel technology that allows for spatial analysis of the proteome surrounding IgG4 deposits in the esophagus. The presence of multiple eosinophil-derived proteins in IgG4-enriched areas suggests that there may be an interaction between IgG4 and eosinophils. If IgG4 is forming immune complexes with food proteins in the esophagus in proximity to eosinophils, as our data suggest, then these multimeric immune complexes could potentially activate low-affinity FcRs (i.e. FcγRII) on eosinophils. This could thereby result in eosinophil activation and the release of eosinophil granules causing further esophageal inflammation and dysmotility. Another possibility is that eosinophils phagocytose the IgG4-food complexes and participate in antigen presentation to further propagate the underlying T2 immune response.^42–44 Finally, it is possible that IgG4 is acting in a protective manner to bind allergens and block allergen binding with other cells, as is thought to occur in non-allergic beekeepers³³ and cat owners,³⁴ as previously noted.

Interestingly, calpain-14 (CAPN14), an IL-13-induced gene implicated in barrier dysfunction, was also significantly elevated in IgG4-rich areas.⁴⁵ IgG4 is known to activate pathways that induce blistering in pemphigus vulgaris⁴⁶ so it is also possible that a similar mechanism is occurring in EoE. Disruption of the epithelial barrier could result in increased antigen penetration into the tissue, thus perpetuating the local immune response. Finally, AutoSTOMP identified various food allergens present in IgG4-rich areas, which were distinct for each individual patient. Future studies are necessary to examine whether co-immunoprecipitation or AutoSTOMP could serve as useful tools to identify specific food triggers in individual patients.

There are several limitations to this study. First, the specificity of primary antibodies to proteins of interest could have impacted the degree of co-localization observed. Techniques were optimized to improve binding; however, there was notably less co-localization of Bos d 4 when compared to Bos d 5, 8. This could also be explained by the relatively small size of Bos d 4 (diameter of 1.8nm¹⁷), which increases the risk for antigen loss during sample preparation. Another potential explanation is that the majority of Bos d 4 immunofluorescence was not higher than the background signal detected by confocal microscopy and thus not recognized as a signal by image analysis software. Another limitation of this study is the self-reporting of milk ingestion. Questionnaires were completed at the time of enrollment and reviewed at each endoscopy, which could be vulnerable to recall bias. Another limitation of this study was the tissue preparation necessary for the AutoSTOMP samples. Because this required fresh, frozen tissue, it could not be performed on stored samples in the UVA EoE Cohort patients. Therefore, these patients had not undergone full elimination diets, and it is unknown whether cow’s milk is a food trigger. Additionally, there could have been significant variability of IgG4 deposition given the known patchy nature of EoE. This could lead to under-reporting of co-localization if areas with low IgG4 were sampled. To account for this, we optimized our sample selection to include tissue sections with adequate lamina propria, as previous studies have demonstrated that IgG4-secreting plasma cells are present in this area.^10,12,14 However, biopsy specimens from the same locations were not directly compared in the active EoE, inactive EoE and control groups.

In conclusion, to our knowledge, this is the first study demonstrating the direct interaction of IgG4 with food proteins in the esophageal mucosa of patients with EoE. In addition, this is the first study to describe the local proteome surrounding IgG4 deposits, which we found is enriched with eosinophil-associated proteins. Further work is needed to evaluate the significance of IgG4 in the pathogenesis of EoE.

Figure 2:

Esophageal biopsies with immunofluorescence staining for IgG4 (green), Bos d 5 (red), and nuclei (blue). (a) Patient with active EoE demonstrating co-localization of IgG4 and Bos d 5 (yellow) and (b) paired sample from the same patient in remission while on dairy elimination diet. (c) Control subject consuming dairy and (d) active disease in non-milk triggered EoE patient with positive immunofluorscence staining of IgG4 without co-localization of milk proteins.