Epithelial-Fibroblast Crosstalk in Eosinophilic Esophagitis

Abstract

Eosinophilic esophagitis (EoE) is an emerging form of food allergy that exerts a significant clinical and financial burden worldwide. EoE is clinically characterized by eosinophil-rich inflammatory infiltrates in esophageal mucosa and esophageal dysfunction. Remodeling events in esophageal epithelium and lamina propria also frequently occur in patients with EoE. Because subepithelial fibrosis is associated with esophageal stricture, the most severe consequence of EoE, there exists an urgent need for a deeper understanding of the molecular mechanisms mediating fibrosis in EoE. Here, we review emerging evidence from experimental model systems that implicates crosstalk between esophageal epithelial cells and underlying stromal cells in EoE fibrosis. We further discuss implications for epithelial-stromal interaction with regard to EoE patient care and propose future directions that may be pursued to further the understanding of epithelial-stromal crosstalk in EoE pathobiology.

Summary.

Epithelial-stromal crosstalk is an emerging player in the pathobiology of eosinophilic esophagitis (EoE). Here, we review basic science studies that have provided insight into interactions between the esophageal epithelial and stromal compartments in EoE. We further describe studies using EoE patient tissues to investigate whether alterations in esophageal epithelial biology may be informative regarding fibrosis in the underlying lamina propria. Finally, we discuss future areas of investigation that are necessary to further the understanding of epithelial-stromal interactions in EoE and leverage this knowledge to improve EoE patient care.

Eosinophilic esophagitis (EoE) is a chronic food allergen- and immune-mediated disease that represents a major cause of upper gastrointestinal morbidity worldwide. EoE is characterized by transmural eosinophil-rich inflammatory infiltrates in esophageal mucosa coupled with symptoms of esophageal dysfunction, including dysphagia and food impaction.¹ Although eosinophil counts remain the critical determinant for EoE diagnosis and monitoring, the disease also features esophageal tissue remodeling in the stromal and epithelial compartments.^1,2 Fibrosis in the lamina propria is associated with esophageal stricture, the most severe consequence of EoE, and history of dilation is associated with treatment-refractory disease.³ Although EoE fibrosis ultimately arises as fibroblasts in the esophageal lamina propria become activated, resulting in secretion and remodeling of extracellular matrix (ECM) components, crosstalk between esophageal epithelial cells and underlying stromal cells has been implicated as contributor to lamina propria remodeling in EoE. Here, we review evidence from experimental model systems demonstrating roles for epithelial-stromal interaction in EoE fibrosis, highlighting current knowledge gaps and controversies. We also discuss the potential for insights gained from the study of epithelial-stromal crosstalk to improve EoE patient care. Finally, we propose areas for future research in this developing field of study.

Epithelial-Stromal Interactions in EoE Fibrosis: Insights From Experimental Models Systems

Key Mediators

Transforming growth factor (TGF)-β has been characterized as a critical mediator of EoE fibrosis. Upon direct stimulation with TGF-β, esophageal fibroblasts exhibit production of collagen, fibronectin, and α-smooth muscle actin (α-SMA) as well as increased contractility.^4,5 In mice, ovalbumin-induced experimental EoE induces fibrosis that is attenuated with global knockout of SMAD3,⁶ a critical transcriptional mediator of TGF-β signaling. Because activation of TGF-β signaling has been documented in the esophageal lamina propria and epithelium of patients with EoE,^7,8 targeted inhibition of SMAD3 in these tissue compartments is necessary to determine their individual contributions to TGF-β-mediated fibrosis. Interestingly, mice with knock-in of Tgfbr1^M318R loss of function variant, in which point mutation renders TGF-β receptor I devoid of kinase activity, display spontaneous EoE-like inflammation that occurs independent of lymphocytes.⁹ A role for perturbed epithelial TGF-β signaling in driving EoE inflammation was supported by this model system because esophageal epithelium of Tgfbr1^M318R mice exhibited accumulation of immature, distressed keratinocytes with enhanced cell proliferation and an inflammatory gene signature indicative of chronic activation of interleukin (IL)-1 and tumor necrosis factor (TNF)-α.⁹ These findings suggest that TGF-β maintains esophageal epithelial homeostasis to limit EoE inflammation. No evidence of fibrosis was identified in Tgfbr1^M318R mice,⁹ supporting the critical role of TGF-β in EoE-associated lamina propria remodeling.

It has been demonstrated that stromal-derived TNF-α drives esophageal epithelial cells to produce the collagen crosslinking enzyme lysyl oxidase (LOX) in a TGF-β-dependent manner.¹⁰ LOX expression is increased in endoscopic biopsies from EoE patients with fibrostenosis as compared with their nonfibrostenotic counterparts,¹⁰ suggesting that crosstalk between esophageal epithelial and stromal compartments may contribute to EoE pathophysiology via ECM remodeling mediated by TNF-α/TGF-β/LOX signaling. In this study, red fluorescent protein (RFP)–expressing human fibroblasts and green fluorescent protein (GFP)–expressing human esophageal epithelial cells (from normal control subjects and patients with EoE) were placed in coculture. RFP–tagged fibroblasts and GFP–tagged epithelial cells were then separated by fluorescence activated cell sorting (FACS) and subjected to gene expression analysis. Expression of LOX was identified in esophageal epithelial cells following coculture, indicating that epithelial-stromal interactions may contribute to ECM remodeling even in the absence of inflammatory signals,¹⁰ perhaps perpetuating fibrosis in patients with EoE, even those in remission. Although these studies support exploration of LOX inhibitors to limit fibrosis in the EoE inflammatory milieu, LOX may have a role in esophageal epithelial biology beyond contributing to ECM remodeling. LOX overexpression in esophageal epithelial cells was recently shown to induce bone morphogenetic protein signaling and squamous differentiation.¹¹ Thus, implementing approaches to specifically block LOX-mediated collagen crosslinking, as opposed to general LOX inhibition, may represent a fruitful approach for preventing EoE fibrosis.

T-helper 2 cytokines IL4 and IL13 have also been demonstrated to induce production of collagen in esophageal fibroblasts.⁴ IL4 and IL13 can both mediate signaling through type II IL4 receptor, comprised of IL4 receptor α (IL4Rα) and IL13 receptor α1 (IL13Rα1). Avlas et al¹² recently used Krt14^Cre;Il13ra1^fl/fl mice to identify a role for epithelial cell-specific type II IL4 receptor in EoE. In this transgenic mouse model, expression of Cre recombinase is driven by the squamous epithelial-specific promoter keratin 14 (Krt14), allowing for cleavage of loxP sites flanking the Il13ra gene (Il13ra1^fl/fl) and selective deletion of this gene in squamous epithelial cells. In Krt14^Cre;Il13ra1^fl/fl mice, oxazolone-induced lamina propria thickening was limited, suggesting that signaling through epithelial type II IL4 receptor contributes to EoE-associated ECM remodeling.¹² Yet, as compared with Il13ra1^fl/fl control animals, collagen deposition was increased in Krt14^Cre;Il13ra1^fl/fl mice with and without oxazolone-induced EoE.¹² Additional studies should be conducted to reconcile these discrepant findings and also to determine the mechanisms through which epithelial signaling downstream of IL4 and IL13 contributes to stromal fibrosis.

Periostin is an ECM protein that is activated downstream of both TGF-β and IL13 in esophageal fibroblasts¹³ and has been shown to promote myofibroblast activation in the lung.¹⁴ Although expression of periostin has been demonstrated in the lamina propria and epithelium of esophageal biopsies from patients with EoE,^13,15 analysis of the ECM proteome generated by esophageal fibroblasts from patients with EoE demonstrated downregulation of periostin compared with esophageal fibroblasts from control subjects.¹⁶ Although this finding may represent an artifact (i.e. periostin may be trapped in the insoluble fraction during proteomic studies), an alternate explanation is that periostin deposited in lamina propria of esophageal mucosa is produced by cell types other than fibroblasts. Because periostin was identified in the esophageal lumen using the esophageal string test,¹⁷ it is tempting to speculate that periostin derived from esophageal epithelial cells may signal to underlying stromal cells to promote fibrosis; however, this must be formally tested.

Epithelial-Mesenchymal Transition

Beyond merely promoting fibroblast activation and ECM remodeling, it has further been postulated that esophageal epithelial cells directly contribute to fibrosis by adopting mesenchymal features. Evidence of epithelial-mesenchymal transition (EMT), namely reduced expression of epithelial cytokeratins concomitant with upregulation of the mesenchymal marker vimentin, has been documented in EoE patient biopsies where it positively correlates with eosinophil counts, TGF-β immunostaining, and subepithelial fibrosis.¹⁸ Direct stimulation of esophageal epithelial cells with TGF-β promotes EMT in vitro^18,19; however, because TGF-β activation occurs in the esophageal epithelium and lamina propria, additional in vivo studies are required to determine if TGF-β-mediated EMT occurs via autocrine or paracrine signaling. Fibroblast-derived TNF-α and IL1β drive EMT in esophageal keratinocytes in vitro.²⁰ In these studies, medium from esophageal epithelial cultures promotes esophageal fibroblast activation and production of TNF-α and IL1β that, in turn, leads to epithelial expression of mesenchymal markers, including α-smooth muscle actin and vimentin, along with loss of E-cadherin expression.²⁰ It is important to note that these epithelial cultures were not stimulated or treated with any cytokines, thus medium from unperturbed epithelium is sufficient to stimulate these changes in vitro.²⁰ Taken together, these findings suggest that epithelial-stromal crosstalk drives EMT in esophageal epithelial cells that gain functional characteristics associated with activated myofibroblasts, including collagen production, enhanced migration capacity, and contractility.¹⁹ While the described data suggest that EMT may contribute to EoE pathogenesis, lineage tracing studies that define the direct contribution of esophageal epithelial cells to EoE fibrosis are presently lacking.

Aging

Although EoE is characterized as a progressive fibrostenotic disease,^21,22 a recent study demonstrated that tissue aging contributes to fibrotic remodeling in esophageal mucosa that is exposed to EoE inflammation.²³ On exposure to MC903/ovalbumin-induced experimental EoE for the same time period, aged mice display lamina propria fibrosis more readily than their younger counterparts.²³ Moreover, although esophageal epithelial cells isolated from young and aged mice both limited fibroblast contractility in vitro, the effect mediated by young epithelial cells was more pronounced.²³ Thus, the increased prevalence of fibrostenosis in aged patients with EoE^14,15,17, 18, 19 may not merely arise as a consequence of increased disease duration, but rather may be exacerbated as the antifibrotic activity of esophageal epithelial cells diminishes with age. This concept complements studies indicating that the likelihood of fibrostenotic EoE increases with patient age and time to EoE diagnosis,^21,22,24 and underscores the importance of identifying factors that limit fibrosis such that they may be leveraged for prevention of this devastating complication in patients with EoE.

Epithelial-Stromal Interactions in EoE: Implications for Clinical Care of Patients With EoE

Although esophageal biopsy remains the gold standard for evaluating EoE disease activity, <50% of esophageal biopsies have evaluable lamina propria.^25,26 This is a substantial hurdle in EoE clinical care because detection of fibrosis could indicate the need for more aggressive therapy early in the course of disease, significantly decreasing long-term complications, including stricture. To address this issue, several studies have investigated esophageal epithelial remodeling events in relation to lamina propria fibrosis. In biopsies from patients with EoE (both active EoE and EoE remission), the presence of basal zone hyperplasia, dyskeratosis, and surface epithelial layering were associated with lamina propria fibrosis.^27,28 Furthermore, a computational model combining these 3 factors with age allowed for prediction of lamina propria fibrosis with 82% accuracy (area under the receiving operator characteristic curve, 0.84; 95% confidence interval: 0.80–0.89).^27,28 Notably, in this model the presence of eosinophils in esophageal epithelium was not predictive of lamina propria fibrosis,^27,28 consistent with factors beyond inflammatory infiltrate promoting EoE fibrosis. Furthermore, these findings suggest that epithelial remodeling, even within the surface epithelium, which is not directly in contact with the lamina propria, may contribute to fibrosis.

Beyond EoE-associated histologic features, evaluation of esophageal epithelium in patients with EoE has revealed molecular differences when comparing patients with EoE with inflammatory and fibrostenotic disease. Transcriptomic profiling of gene targets in the EoE diagnostic panel across a multicenter cohort of patients with EoE identified 3 distinct disease endotypes: EoEe1 exhibited normal-appearing esophagus with relatively mild inflammation that is responsive to therapy, EoEe2 exhibited an inflammatory and steroid-refractory phenotype, and EoEe3 exhibited an adult-onset fibrostenotic phenotype.²⁹ Although expression of epithelial differentiation genes was found to be lowest in EoEe3,²⁹ it remains to be determined whether this finding may be reflective of EMT. In studies using targeted approaches, epithelial expression of LOX and plasminogen activator inhibitor (PAI)-1, both canonically regulated by TGF-β, have been associated with fibrosis in EoE patient biopsies.^10,30 Esophageal epithelial LOX expression directly correlated with a fibrostenotic phenotype in EoE with expression increased to greater degree in patients with EoE with a history of either fibrostenotic EGD findings (rings or strictures) or dilation/food impaction.¹⁰ With regard to PAI-1, esophageal epithelial expression correlated specifically with fibrosis and expression of TGF-β, collagen I, fibronectin, and matrix metalloproteases.³⁰

Identification of approaches that use epithelial biopsies to determine what is happening in the subepithelial space holds promise for improving clinical care in patients with EoE by facilitating detection and monitoring of fibrosis and, potentially, guiding approaches toward predicting which patients will progress to a fibrostenotic phenotype. It also remains to be determined whether therapies targeting esophageal epithelium, fibroblasts, or the interactions between these 2 cell types may be effective in patients with EoE, especially in the refractory patients with EoE population that fails to respond to immune modulation.^31,32

Summary & Future Directions

We have described the current literature related to epithelial-stromal interactions in EoE from the basic science and clinical perspectives (Figure 1). These studies suggest that signals derived from epithelial cells (TGF-β, LOX, Periostin) and fibroblasts (TNF-α, IL1β), as well as from immune cells (IL4, IL13) may contribute to epithelial-stromal crosstalk in EoE (Figure 1A). Evidence of EMT has also been identified in EoE patient tissues and demonstrated to occur in esophageal epithelial cells in vitro (Figure 1A). Emerging data further suggest that esophageal epithelial cells exhibit antifibrotic activities that diminish with age (Figure 1B). Because lamina propria is frequently unavailable in esophageal biopsy specimens, efforts have been made to determine the relationship between alterations in esophageal epithelium and the presence of subepithelial fibrosis in patients with EoE (Figure 1C). In sum, the current literature supports crosstalk between esophageal epithelium and fibroblasts as a significant contributor to EoE pathology and highlights areas that are ripe for future investigation.

Figure 1.

Mechanisms and clinical implications of epithelial-stromal interactions in EoE. (A) Model depicting the contributions of indicated EoE inflammatory mediators in epithelial-stromal crosstalk and myofibroblast activation. Solid lines represent relationships directly demonstrated in described literature. Hatched lines represent proposed relationships. (B) Model depicting age-associated decline in antifibrotic activity of esophageal epithelial cells. (C) Overview of epithelial markers that have been explored as predictors of fibrosis in esophageal biopsies. LP, lamina propria; PAI-1, plasminogen activator inhibitor-1.

Moving forward, it will be imperative to dissect the detailed molecular mechanisms through which epithelial-fibroblast interactions promote EoE fibrosis. Implementation of in vivo models of EoE coupled with genetically engineered mice permitting cell type–specific gain- or loss-of-function will be critical to tease out the mechanistic roles of proposed mediators of epithelial-fibroblast crosstalk. For example, mice with depletion of SMAD3 or expression Tgfbr1^M318R targeted to either squamous epithelium or mesenchymal cells will allow for rigorous investigation of the contribution of TGF-β signaling in these specific tissue compartments to EoE fibrosis. Such in vivo studies may be further extended to investigate mechanisms through which other cell types within esophageal mucosa influence fibrotic remodeling. Indeed, roles for inflammatory, muscle, and endothelial cells have been suggested in EoE fibrosis.^4,33 Genetically engineered mice that facilitate lineage tracing of esophageal epithelial cells may also be used to determine the extent to which EMT contributes to fibrosis in murine models of EoE. The age of mice used in these studies must be carefully considered given the finding that fibrotic remodeling occurs more readily with age.²³ Because single-cell RNA-sequencing has been used to define the heterogeneity of esophageal epithelial cells under homeostasis and in response to EoE inflammation,34, 35, 36, 37 it will be of interest to extend these studies into esophageal fibroblasts. Such investigations will facilitate evaluation of the cellular and molecular heterogeneity of esophageal fibroblasts. Statistical modeling in single-cell RNA-sequencing data may be further used to predict cell-cell interactions, whereas spatial transcriptomics will allow for the assessment of these interactions in situ. Indeed, such data may be useful in identifying the source of potential mediators of epithelial-mesenchymal crosstalk, (eg, periostin) in patients with EoE. Moreover, because unbiased proteomics have revealed that thrombospondin-1 is uniquely expressed in EoE fibroblast ECM in vitro,¹⁶ effects of matrix composition and stiffness on esophageal epithelial cells should be determined. Elucidating the relationship between EoE-associated genetic variants that impact epithelial cell biology (eg, single-nucleotide polymorphisms in the genes encoding filaggrin and desmoglein-1)^38,39 and the fibrostenotic phenotype in EoE may also provide additional evidence to support epithelial-stromal crosstalk in this disease. A positive association between fibrosis and age is not always apparent in patients with EoE because fibrosis is evident in a subset of pediatric subjects,^26,40,41 whereas adults with chronic EoE inflammation may fail to progress to fibrostentosis.²⁴ As antifibrotic activities of esophageal epithelial cells diminish with age, elucidating the specific mediators of these changes may prove useful for predicting which patients with EoE progress to fibrostenotic disease. Epigenetic events associated with premature molecular aging have been identified in esophageal biopsies from patients with EoE.⁴² Thus, it may be of further interest to determine if the molecular age of a patient with EoE may serve as a biomarker to predict progression to fibrostenotic disease. Indeed, identification of a predictive biomarker for fibrostenotic EoE would represent a significant advance in EoE. Technology to assess liver stiffness has led to significant advances in detection of liver fibrosis.⁴³ In EoE, endoscopic ultrasound has shown that there is increased submucosal thickness in patients with EoE, although this technology has not been widely adapted.^44,45 Functional luminal imaging probe technology has allowed for measurement of diameter and pressure within the esophagus, allowing for an in-depth evaluation of esophageal distensibility and secondary peristalsis in the setting of disease.⁴⁶ However, how the measurements obtained using functional luminal imaging probe may reflect epithelial and lamina propria remodeling has yet to be elucidated.²⁶

Crosstalk between epithelium and fibroblasts is a key, yet frequently overlooked, player in the pathobiology of EoE. Understanding the specific molecular processes contributing to this crosstalk will allow us to better leverage treatment options that are currently on hand for EoE, and also to develop novel approaches to specifically normalize these interactions, lowering the odds of patients developing fibrostenotic disease and resulting in improved quality of life for patients with EoE.