Mucosal Penetration and Clearance of Gluten and Milk Antigens in Eosinophilic Oesophagitis

Anupama Ravi; Eric V. Marietta; Jeffrey A. Alexander; Kathryn Peterson; Crystal Lavey; Debra M. Geno; Joseph A. Murray; David A. Katzka


Aliment Pharmacol Ther. 2021;53(3):410-417. 

In This Article


The occurrence of oesophageal eosinophilia and inflammation in EoE is dependent at least in part on the penetration of food antigens into the oesophageal mucosa. Indeed, the presence of food antigen in the mucosa is enough to generate a Th2 allergic pathway or specific recognition of this antigen in a given patient through several potential pathways.[3–5] In this study, we demonstrate that the oesophageal mucosa is selectively permeable to food antigens ranging from penetration based on the activity of EoE to generalised antigen entry found with inactive disease and in normal controls. These findings suggest that the pathophysiology of EoE is in part related to food antigen presence but additionally to recognition and immunologic activation after exposure to the antigen. Specifically, for some antigens, activation of the Th2 pathway is initially dependent on increased oesophageal permeability associated with active disease. On the other hand, other food antigens may more or less freely enter oesophageal mucosa such that cellular and antibody recognition is the only rate limiting step of the inflammatory response. A next step in elucidating this pathway would be to determine if antigen presence generates a local immunologic response or initiates trafficking of systemically available immune cells to the oesophagus. The latter is suggested by the ability to induce EoE in an animal model though antigen introduction into extraoesophageal sites and cytokine recruitment of bone marrow-derived cells.[7]

This study also demonstrates the remarkable durability for these antigens to remain present in oesophageal epithelium. Up to 96 hours after exposure, some patients still had stainable antigen in the oesophageal mucosa despite diet elimination. Similar to entry, this time dependent pattern appears to be sensitive to EoE activity for gluten but less so for milk/dairy. The total dwell time of these antigens in oesophageal mucosa is unclear.

The oesophageal epithelium has been viewed generally as a relatively impermeable barrier. This is in contrast to other areas of squamous epithelium such as the skin where prolonged contact with a surface substrate may lead to absorption. In contrast, it has been assumed that a combination of rapid bolus transit over a tightly apposed syncytium makes permeation of a complex protein unlikely. This study supports and expands on previous findings demonstrating that not only is the oesophagus capable of allowing antigen penetration when inflamed by EoE but also the novel finding that the normal oesophagus is capable of allowing some food antigens to penetrate. An important question is what determines selective or generally permissive epithelial antigen penetration even 10 cm above the gastro-oesophageal junction where biopsies were taken. Several studies have demonstrated that the oesophageal epithelial barrier function in patients with EoE in the active state is perturbed. The types of studies that support this include demonstration of increased permeability in Ussing chambers,[8] increased conductivity of the mucosa reflected through a decrease in electrical impedance measurement,[9,10] decreased detection of tight junction proteins in the oesophageal epithelium[11–13] and dilation of intercellular spaces.[9,14] Indeed, dilation of intercellular spaces is a histologic hallmark of active EoE and one of the key elements of the recently developed histologic scoring system for EoE.[14] In our prior study of gluten penetration, we further demonstrated that there is a correlation between the severity of intercellular space dilation and gluten presence in the oesophageal mucosa.[6] On the other hand, why might whey and less so casein have greater ability to enter the oesophageal epithelium? One explanation might be the molecular weight and size of these antigens. Human whey alpha lactalbumin has a MW of 0.3–1.4 kDa and is 1.8 nm in diameter.[15,16] This is compared to beta casein that has a MW of 24 kDa and size of 125 nM and a micelle size of 154–230 nm.[17] The MW of gliadin is 30–80 kDa with a fibril diameter of 2000–5000 nm.[18] These lengths might be important as the diameter of normal human oesophageal epithelial intercellular space is 275 nm.[19] This diameter size would theoretically permit whey and perhaps casein to easily cross through but not gliadin. However, in EoE, the size of the intercellular space diameter can increase to 8–10 μm wide and to 30–40 μm when in an open lake pattern[7] which might allow gliadin to enter the epithelium. This does not eliminate the possibility of a more active process involved in absorbing food antigens. For example in the duodenum, gluten can be transported by a goblet cell pump mechanism. As the squamous oesophageal epithelium lacks goblet cells, this would not apply. The intercellular space is also a dynamic structure whose flow is carefully regulated by the epithelial junction complex consisting of tight junction and adherens junction proteins and desmosomes. Thus a more active process in this space rather than anatomic diameter alone may regulate food antigen entry. Further research is needed in this regard.

This study also illustrates that the dwell time of food antigens in the oesophageal mucosa can be relatively long lasting. After acute exposure to gluten, patients could have the presence of gliadin-specific antibodies in their mucosa at least 72 hours after exposure following gluten avoidance. Although this time dependence applied less to casein and did not apply to whey for the postulated reasons above, these findings have implications for diet therapy. Specifically, patients commonly question how long after exposure to an inciting food antigen will they be at risk for activation of EoE. This study is a first step to answer this question but does not address the quantity or timing of food antigen exposure that specifically leads to oesophageal epithelial infiltration. The concept that antigens may have prolonged presence in an epithelium is not novel. For example, urushiol, an active antigen in plants such as poison oak or sumac remains bound to cell membranes in skin epithelium for a duration of 1–5 days before antigen recognition and clinical manifestations of the reaction begin.[20] That food antigens can remain for days in the oesophageal epithelium is novel. Whether there is similar binding of the antigen to a cellular structure needs to be determined.

In contrast to the presence of food antigens in oesophageal mucosa, IgG4 appears to be a more specific downstream finding in EoE and perhaps more likely to be found in active disease. This is suggested by the absence of IgG4 in control patients, the finding of IgG4 on biopsy only when food antigens were present, and a trend toward greater IgG4 in active than inactive EoE patients in the first 48 hours. The association of IgG4 to EoE is strong[21] including correlation to histologic severity, parts of the EoE transcriptome and key cytokines elevated in EoE.[22] Similar to our study of finding IgG4 in 15 of 30 patients, another study only identified 48% of EoE patients with IgG4 on biopsy[23] but is specific to EoE when compared to GERD.[24] The precise role of IgG4 in the pathophysiology of EoE is unclear.

This study has several limitations. First and foremost, almost two thirds of our patients were on PPIs or steroids. Both PPIs[8] and steroids have been shown to improve epithelial barrier function in EoE.[11] Although pharmacologically treated patients were represented in both active and inactive EoE groups, we cannot rule out a contributing effect of these medications to our results. Second, although 32 patients and 10 controls were studied, more data at specific time intervals would have enhanced the results and made them less qualitative and descriptive. Unfortunately, these are difficult studies to do enlisting patients on specific diet elimination therapies to avoid antigens for varying amounts of time. Although it would have been preferential from an experimental approach to sequentially perform endoscopies and biopsies daily in the same patient after timed food elimination, this would increase the risks of the protocol significantly. Future studies will focus on patients starting from complete elimination diets and sequentially assessing for both individual food antigens and IgG4 as foods are added back. Third, it cannot be determined to what specific food antigen(s) IgG4 is reacting as this study focuses on three antigen groups containing multiple proteins. It can only be assessed as a marker of immunologic activity in this study. Nevertheless, we endeavoured to use antibodies specific to distinct portions of the antigen protein. Fourth, our assessment of antigen size in the mucosa is based on deduction from the literature rather than hard evidence of the exact molecular weight and diameter of the stained particle found in our patients. The precise form of the protein will be depend on factors such as the quantity of the food bolus to which the mucosa is exposed, salivary enzymes and enzymes that would be produced by the mouth and oesophageal microbiota. Then again, our best deduction intuitively fits with the proposed hypothesis of the relationship of food antigen entry to the known properties of oesophageal epithelium and intercellular spaces. Fifth, we analysed patients on PPI and steroids as one group. On the other hand both PPIs[8] and steroids have been shown to improve epithelial barrier function in EoE.[11] We also did not ask patients to keep food diaries to precisely measure wheat and dairy intake but careful instructions on diet were given by our nurses and research coordinator. Finally, we based the number of patients studied on the positive results from our prior study.[6] Unfortunately, due to the difficulty of enrolling patients in such a challenging study, we could not recruit as many patients as could have satisfied a formal power calculation.

In conclusion, this study demonstrates that the oesophageal epithelium is permeable to proteins and specifically to food antigens known to trigger EoE. Furthermore, there is a variation in epithelial selectivity which may depend on the size of the antigen. In the case of more permeable antigens such as whey, their presence may be an innocent bystander rather than a trigger of EoE activity. Future studies need to focus on the precise mechanism of detected antigen-specific mucosal entry and the factors that further determine the induction or effector trigger of the Th2 pathway activation and downstream effects on additional markers of inflammation such as inflammatory cytokines, mast cells or lymphocytes after exposure food antigen in the mucosa. This insight may provide an actionable target for interrupting the response.