Fibrostenotic eosinophilic esophagitis may reflect epithelial lysyl oxidase induction by fibroblasts-derived tumor necrosis factor-α

Abstract

Background:

Fibrosis and stricture are major comorbidities in eosinophilic esophagitis (EoE). Lysyl oxidase (LOX), a collagen cross-linking enzyme, has not been investigated in the context of EoE.

Objective:

We investigated regulation of epithelial LOX expression as a novel biomarker and functional effector of fibrostenotic disease conditions associated with EoE.

Methods:

LOX expression was analyzed by RNA-sequencing, PCR assays and immunostaining in EoE patients, cytokine-stimulated esophageal three-dimensional (3D) organoids and fibroblast-epithelial cell co-culture, the latter coupled with fluorescence-activated cell sorting.

Results:

Gene ontology and pathway analyses linked tumor necrosis factor (TNF)-α and LOX expression in EoE, which was validated in independent sets of patients with fibrostenotic conditions. TNFα-mediated epithelial LOX upregulation was recapitulated in 3D organoids and co-culture experiments. We find fibroblast-derived TNFα stimulates epithelial LOX expression via activation of nuclear factor-κ B and transforming growth factor-β-mediated signaling. In patients, receiver operating characteristic analyses suggested that LOX upregulation indicates disease complications and fibrostenotic conditions in EoE.

Conclusions:

There is a novel positive feedback mechanism in epithelial LOX induction via fibroblast-derived TNFα secretion. Esophageal epithelial LOX may have a role in the development of fibrosis with substantial translational implications.

Capsule Summary

Esophageal fibrosis is a long-term consequence of remodeling in eosinophilic esophagitis, but little is known about the pathophysiology of fibrosis. We now demonstrate that TNFα mediated lysyl oxidase production may be driving fibrostenotic phenotype.

Graphical Abstract

Introduction

Eosinophilic Esophagitis (EoE) is a food allergen-triggered, cytokine-mediated chronic inflammatory disease marked by mass migration of eosinophils and other immune cell types to the esophagus^1,2. Untreated inflammation in EoE often causes tissue remodeling in the esophageal mucosa, leading to fibrosis and esophageal dysfunction evident clinically by dysphagia, food impactions, and strictures. The most serious consequence of esophageal remodeling and fibrostenosis is the extremely narrow caliber esophagus which represents 9% of EoE patients and is associated with therapy resistance³. Diagnosis of EoE is made based upon upper endoscopy and biopsies; however, histological diagnosis of fibrosis is often suboptimal due to limited sampling of the subepithelial lamina propria compartment in biopsy specimens⁴. Currently, there is no tissue biomarker available to predict fibrostenotic conditions associated with EoE.

Transforming growth factor (TGF)-β is the major cytokine that mediates EoE-related lamina propria remodeling along with T-helper 2 (Th2) cytokines including interleukin (IL)-4 and IL-13^5,6,7. These cytokines promote fibrosis by activating lamina propria fibroblasts to produce collagen and fibronectin, increasing eosinophil adherence to tissue and enhancing fibroblast contractility^6,8. Additionally, cross-talk between esophageal epithelial cells and fibroblasts leads to robust fibroblast production of tumor necrosis factor (TNF)-α⁹,⁸ a profibrogenic cytokine that has been implicated in fibrotic disease conditions of the liver, intestine and kidney^10–12; however, the pathogenic role of TNFα in in EoE remains elusive. In response to these cytokines, stromal fibroblasts are activated to deposit and construct networks of excess collagen and fibronectin in the lamina propria, leading to increased esophageal stiffness¹³. In turn, stiffness itself contributes to fibroblast activation to express greater amounts of α-smooth muscle actin¹⁴, a marker of fibroblast activation, and generate enhanced traction forces. Nevertheless, little is known about the molecular mechanisms regulating tissue stiffness in EoE.

The process of collagen cross-linking involves extracellular modifications that result in increased stiffness. Lysyl oxidase (LOX) is a copper-dependent enzyme that acts on extracellular collagen to form intra- and inter-molecular cross-links between neighboring collagen molecules, forming collagen fibers. LOX-induced collagen cross-linking contributes to the progression of cardiac and liver fibrosis^15,16; however, nothing is currently known about the role of LOX in esophageal fibrosis and how it may influence tissue stiffness in the context of EoE.

Herein we analyzed well-annotated RNA-sequencing data of esophageal biopsies from EoE patients to identify unique activation of the TNFα pathway and LOX in EoE. We find that LOX may serve as a novel tissue biomarker for fibrostenotic disease in EoE. Utilizing an innovative co-culture system of esophageal epithelial cells and fibroblasts as well as the esophageal three-dimensional (3D) organoid system with a recapitulation of human physiology and pathology, we examined further the crosstalk between esophageal epithelial cells and fibroblasts to find that fibroblast-derived TNFα perpetuates epithelial cell TGFβ signaling and LOX expression.

METHODS

Human subjects and endoscopic esophageal biopsies

In accordance with Institutional Review Board standards and guidelines at the Children’s Hospital of Philadelphia and the Hospital of the University of Pennsylvania, after informed consent esophageal biopsies were taken from patients undergoing esophagogastroduodenoscopy (EGD) as part of routine care. Subjects were classified per 2011 EoE clinical guidelines¹⁷. Subjects who met clinical criteria and had histologic presence of 15 or more esophageal epithelial eosinophils per high-power field (hpf) were diagnosed as “active EoE”. EoE subjects who demonstrated resolution of histologic inflammation and eosinophilia documented upon previous endoscopy were designated as “EoE Inactive” (<15 eos/hpf). “Non- EoE” subjects had no previous diagnosis of EoE and reported symptoms warranting EGD but demonstrated no histopathologic abnormalities. Subjects with proton pump inhibitor-responsive esophageal eosinophilia¹⁴ were excluded. Subjects with a history of inflammatory bowel disease, celiac disease, GI bleeding, or any other acute or chronic intestinal disorders were excluded from recruitment. Patient demographics provided in Tables 1 and 2.

Table 1:

Patient demographics for RT-PCR cohort.

	Control (34)	Inactive (32)	Active (33)
Median age (range)	11.5 (1.5–67)	10.5 (3–50)	15 (5–48)
Male	21 (62%)	27 (84%)	22 (67%)
History of Impaction(s)	1 (3%)	7 (22%)	15 (45%)
History of Stricture(s)	0 (0%)	3 (9%)	2 (6%)
Mean eos/hpf (range)	0	2 (0–12)	48 (18–116)
Proton Pump Inhibitor	32 (94%)	29 (91%)	31 (94%)
Topical Swallowed Steroids	0 (0%)	8 (25%)	5 (15%)
Diet	0 (0%)	23 (72%)	19 (58%)

Table 2:

Patient demographics for IHC cohort

	Control (5)	Inactive (4)	Active (6)
Median age (range)	7 (4–10)	6 (2–10)	12 (10–15)
Male	4 (80%)	3 (75%)	5 (83%)
History of Impaction(s)	1 (20%)	1 (25%)	1 (17%)
History of Stricture(s)	0 (0%)	0 (0%)	0 (0%)
Mean eos/hpf (range)	0	5.5 (0–10)	28.5 (15–50)
Proton Pump Inhibitor	5 (100%)	4 (100%)	6 (100%)
Topical Swallowed Steroids	0 (0%)	0 (0%)	1 (17%)
Diet	0 (0%)	3 (75%)	0 (0%)

RNA-sequencing data analyses

Raw RNA sequencing data on esophageal biopsies with quality scores (GSE58640) ¹⁸, representing 10 active EoE patients and 6 healthy controls available at the NCBI GEO database were downloaded and aligned to the human genome GRCh38.p7 using the STAR aligner (v252b) ¹⁹. Genomically mapped reads were counted against reference genes as annotated in Gencode (version 25) ²⁰ using htseq-count²¹. One EoE sample (GSM1415921, EoE_803) was excluded from further analyses due to a low number of mapped reads. Genes were tested for differential expression between EoE and controls using DESeq2²², yielding fold change, p-value and false discovery rate (FDR)-adjusted p-value for each gene. Gene ontology analysis was performed using the online bioinformatics software DAVID 6.8²³. The “Benjamini- Hochenberg” method for multiple hypothesis testing was used. Functional annotation clustering was performed on protein coding genes up-regulated by more than 5-fold in the EoE sample set (FDR<0.1) ^24,25,26.

Quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) assays

RNA was isolated from esophageal biopsies for quantitative real-time RT-PCR as described previously¹⁴ using TaqMan Gene Expression Assays (Thermo Fisher Scientific, Waltham, MA) for LOX (Hs00942483) and GAPDH (Hs99999905_m1), the latter as an internal control gene.

Cell culture and three-dimensional (3D) esophageal organoids

Immortalized normal human esophageal epithelial cell line EPC2-hTERT and a derivative expressing green fluorescent protein (GFP; EPC2-hTERT-GFP) and primary normal fetal esophageal fibroblasts FEF3 and derivative expressing tdTomato fluorescent protein (FEF3- tdTomato), were grown at 37°C in a humidifie d 5% CO₂ incubator as described previously^27,28. Primary esophageal epithelial cultures from EoE and non-EoE subjects were used at passage 3 as described previously²⁵. Cells were stimulated with TGFβ or TNFα as indicated in mono-layer culture.

For media-swap experiments, conditioned epithelial media (CEM) as collected from confluent EPC2-hTERT cells grown in complete KSFM and was used to stimulate confluent fibroblasts. Media was supplemented with 10% FBS prior to stimulation (Figure 3A). After 24 hours the fibroblast conditioned media (FCM) was used to stimulated fresh EPC-2hTERT cells for 24 hours alone and in the presence of infliximab (anti-TNFα antibody).

Figure 3: EoE milieu and TNFα upregulate LOX expression in esophageal epithelial cells.

EPC2-hTERT cells were treated with the indicated cytokines for 24 hours and evaluated by PCR (A) and Western Blot (B). Densitometry was performed on Western blots. Numbers under each band represent the signal intensity relative to NT set as 1. EPC2-hTERT cells were treated with increasing doses of TNFα (C). Human and murine esophageal epithelial cells were isolated and grown in matrigel to support organoid formation. Organoids were stimulated with TNFα and qRT-PCR was performed for LOX in human and murine organoids (D) and (E) respectively. IHC for LOX was performed on murine organoids stimulated with TNFα (F). Scale bar represents 50μm. Data represents ≥2 biologic replicates with similar results. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001; NT- no treatment.

For co-culture experiments, equal number (2.5×10⁵) of EPC2-hTERT-GFP and FEF3- tdTomato cells were seeded simultaneously in 60-mm dishes and grown in keratinocyte-serum free medium (KSFM) supplemented with 1% fetal bovine serum (Thermo Fisher Scientific) for 24 hours.

Human and murine esophageal 3D organoids were generated and subjected to morphological analyses following TNFα stimulation as described previously^29,30. In brief, esophageal organoids were grown for 5 days before addition of 5 ng/mL of TNFα. Organoids were recovered by digesting with 1 U/ml dispase I (BD Biosciences, San Jose, CA) and fixed overnight in 4.0% paraformaldehyde. Specimens were embedded in 2.0% Bacto-Agar: 2.5% gelatin prior to paraffin embedding or RNA was isolated for PCR.

Flow cytometry and fluorescence-activated cell sorting (FACS)

Flow cytometry and FACS were performed as described previously³¹. In brief, cells were trypsinized and re-suspended in FACS buffer (PBS supplemented with 1% bovine serum albumin). 4’,6-diamidino-2-phenylindole (DAPI) was used to determine cell viability. FACSCalibur or LSRII (BD Biosciences, San Jose, CA) and FlowJo (Tree Star, Ashland, OR) were used for flow cytometry. FACS Vantage SE and FACS Aria II (BD Biosciences) were used to purify GFP-positive EPC2-hTERT cells and tdTomato-positive FEF3 from the co-culture of epithelial and fibroblast cells for RNA purification. Flow cytometry for GFP and tdTomato was repeated for each condition at least three times.

Transcription Factor Assay

Using the Cignal Reporter Assay kit (Qiagen, The Netherlands), EPC2-hTERT cells were transfected, stimulated with TNFα for 24 hours, and luciferase assay was performed per manufacturer instructions. Lysates were assessed for relative luciferase activity compared to unstimulated control. Luciferase reporter assays for SMAD3 activity were carried out as previously described²³. Briefly, 2 × 10⁵ EPC2-hTERT cells were seeded in 24-well plates. 200 ng of 3TP-lux³² and 5ng of pRL-SV40 (Promega) were transfected overnight using Lipofectamine LTX Reagent with PLUS Reagent (Life Technologies). The following day, cells were incubated with fresh media containing TNF-α for 48h. Luciferase activities were determined using the Dual-Glo Luciferase Assay System (Promega) according to manufacturer’s recommendations and the GloMax-Multi Detection System (Promega). Cells were washed twice with 1x PBS, lysed with 1x Passive Lysis Buffer, and assayed in Microlite White Microtiter Plates (Thermo). Firefly luciferase activities were normalized with Renilla luciferase activities before determining the mean of technical triplicates. Variation between transfections did not exceed 15%.

Immunohistochemistry (IHC)

Paraffin-embedded biopsies were serially sectioned and subjected to Hematoxylin and Eosin (H&E) staining, IHC as described previously^{27,29,33,23,30,31}. In brief, sections were incubated with rabbit anti-LOX (1:100; Novus Biological, rabbit polycloncal antibody NB100–2527) ^34,35,36vernight at 4°C. Stained objects were captured and imaged with a Nikon Microphot microscope with a digital camera. IHC slides were evaluated by a pathologist (AKS) blinded to molecular and clinical data to score for LOX intensity. A visual semiquantitative scoring from 0 to 3 was used for the evaluation of IHC cytoplasmic staining intensity, where 0 is no staining, 1 is weak staining, 2 is moderate staining and 3 is strong staining.

Immunoblotting

Cells in monolayer culture were lysed and subjected to immunoblotting as described. ^30,8 Rabbit monoclonal antibody anti-Smad2/3 D7G7 (Cell Signaling Technology), rabbit monoclonal anti-phospho-SMAD2/SMAD3 D27F4 (Cell Signaling), rabbit polyclonal anti-LOX 102M4756V (Sigma Aldrich) and mouse monoclonal anti-β-actin AC-74 (1:10,000; Sigma-Aldrich) were used as primary antibodies. β-actin served as a loading control.

Statistical analysis

Data are presented as mean ± s.e.m. or mean ± s.d. and were analyzed by two-tailed Student’s t-test or ANOVA where applicable. P<0.05 was considered significant. Data were analyzed using software package Prism (La Jolla, CA). All authors had access to the study data and reviewed and approved the final manuscript.

RESULTS

NA sequencing data reveals LOX upregulation and unique TNF-α pathway enrichment in EoE biopsies

Gene expression profiling of EoE has been extensively carried out in EoE biopsies. Sherrill et al. carried out RNA-sequencing studies to identify novel EoE-associated transcriptomes¹⁵, complementing previous microarray studies that identified EoE-enriched genes including Eotaxin3, POSTN and TNFAIP6³⁷. While earlier studies focused on potential immunomodulators, immune cell-specific genes and non-coding RNAs associated with EoE¹⁸ or those related to genetic variants³⁸, we mined the well-annotated RNA sequencing data (GSE58640) for genes relevant to fibrosis and tissue remodeling. GO analysis revealed the top 2 enriched clusters contain biological processes involving collagen catabolism and extracellular matrix disassembly (Supplementary Table 1). LOX was present in these two clusters, prompting us to focus specifically upon its role in EoE (Supplementary Table 2 for full list of genes in the top cluster).

RNA sequencing analysis revealed that LOX was significantly upregulated in the cohort of patients with EoE compared to controls. We found that LOX was 11.3-fold increased in active EoE patients (p= 1.19E-08). Ingenuity Pathway Analysis revealed that genes in the TNFα pathway were among the most highly enriched in EoE biopsies (p=1×10⁻⁴⁰). There were 328 TNFα target genes enriched in active EoE biopsies, suggesting robust TNFα activation in the EoE population (Supplementary Table 3), a novel premise. Based on these findings we sought to further investigate the role of LOX and TNFα in the pathogenesis of EoE fibrosis.

LOX is upregulated in the EoE esophagus

To validate sequencing results in an independent patient cohort, we performed real-time RT-PCR analyses on esophageal biopsy samples collected from active and inactive EoE patients as well as non-EoE controls (Figure 1A). We found that active EoE patients had significantly enhanced LOX mRNA expression compared to control patients and inactive EoE patients indicating that the unique inflammatory milieu in active disease is driving LOX production in the esophagus. Interestingly, those with inactive disease also had significantly increased LOX compared to controls, suggesting that LOX upregulation is mainly attributed to epithelial cells rather than non-epithelial cells such as infiltrating immune cells. This was corroborated by immunohistochemistry (IHC) (Figure 1B) which revealed that patients with active EoE exhibited significantly increased LOX immunoreactive protein expression compared to non-EoE patients. LOX expression is strongest in the epithelium of the esophagus.

Figure 1: LOX expression is increased in patients with active EoE.

qRT-PCR was performed to evaluate LOX mRNA expression from human esophageal biopsies. LOX mRNA expression is increased in patients with Active EoE (n=33) compared to normal (n=34), and inactive EoE (n=30) (A). ANOVA was performed with multiple comparisons. IHC staining for LOX was performed with representative images shown (B). Scale bar represents 100μm. Histogram representing average LOX score (0–3) for normal (n=5), active EoE (n=6) and inactive EoE (n=4). *** p<0.001 *p<0.05

Fibrostenotic EoE shows upregulation of LOX

We have recently shown that in pediatric EoE, there is decreased esophageal distensibility compared to Non-EoE patients⁴ utilizing functional endoluminal imaging. However, there is no standard procedure to monitor or detect esophageal fibrosis. We therefore sought to determine how LOX expression varied depending on the presence of fibrostenotic disease. EoE patients were grouped based upon presentation of disease complications or endoscopic evaluation. Fibrostenotic complications of EoE were defined as patients with a history of stricture or food impaction requiring endoscopic bolus removal. Endoscopic fibrostenosis was defined as presence of stricture or rings. We then analyzed LOX expression in these groups and found a significant upregulation in patients with fibrostenotic disease (both clinically and endoscopically) when compared to non-fibrostenotic EoE (Figure 2A and B).

Figure 2: LOX expression increased in fibrostenotic EoE.

RNA from EoE biopsies was isolated and those patients with disease complications (stricture or endoscopically removed food impaction) and qRT-PCR was performed. Those with complications (n=22) were found to have enhanced LOX expression compared to those with uncomplicated disease (n=39) (A). LOX expression was significantly increased in patients with fibrostenotic endoscopic findings (rings, narrowing) (n=12) compared to in patients with inflammatory endoscopy findings (furrows, exudates, normal) (n=31) (B). Receiver operator characteristics¹⁶ representing LOX expression and fibrostenotic complications. **p=0.01, *p<0.05

We then determined if LOX mRNA expression level in endoscopic biopsies could be utilized as a molecular marker of fibrostenotic disease. Using a relative fold increase in LOX mRNA expression of 5.233 compared to a non-EoE normal control group as a cutoff point to generate a receiver operating characteristic (ROC) curve (Figure 2C), patients were found to be classified as fibrostenotic EoE with 52.17% sensitivity, 79.49% specificity, and positive and negative predictive values of 68% and 76%, respectively, indicating that LOX upregulation in predominantly epithelial endoscopic biopsies predicts fibrostenotic disease activity. Specifically, LOX upregulation in the biopsies may hold unprecedented implications in EoE diagnosis and therapies.

EoE relevant cytokines induce LOX

In order to determine which factors may be contributing to increased LOX expression in EoE, we stimulated EPC2-hTERT cells with the EoE-relevant cytokines IL4, IL13, TNFα, and TGFβ. We found that each of these cytokines significantly induced LOX RNA expression in vitro as well as LOX protein expression by Western blot (Figure 3A and B). TNFα produced the most robust response and showed synergistic effects with each of other EoE relevant cytokines (Supplemental Figure S1) when co-stimulated. Dose dependent induction of LOX was further observed in the context of TNFα stimulation (Figure 3C). Utilizing an ex vivo 3D model of EoE that recapitulates epithelial changes compatible with basal zone hyperplasia upon treatment with EoE-relevant cytokines²⁹, we found enhanced LOX RNA expression in response to TNFα (Figure 3D and E) in both murine and human derive esophageal organoids and this was confirmed by IHC in murine organoids (Figure 3F). This result also confirmed epithelial cell induction of LOX as 3D organoids were grown in the absence of non-epithelial cell components.

Epithelial-stromal fibroblast cross-talk induces epithelial LOX

We have previously shown that TNFα is produced by esophageal fibroblasts stimulated with conditioned media from esophageal epithelial cell culture³³ and that TNFα, in turn, induces epithelial cells to undergo functional changes such as increased collagen production and contraction⁸. Crosstalk between the epithelial cells and fibroblasts induces a profirbrotic milieu even in the absence of inflammatory cells. We therefore hypothesized that crosstalk between these cell types may induce LOX in vitro. We stimulated EPC2-hTERT cells with media from esophageal fibroblasts (FEF) that had been primed with conditioned epithelial media (Figure 4A). We found that the epithelial cells stimulated with the fibroblast conditioned media displayed increased LOX expression (Figure 4B). To explore the role of TNFα in LOX production, we then treated the EPC2-hTERT cells with anti-TNFα antibody infliximab. We found that treatment with anti-TNFα caused a significant decrease in LOX expression, suggesting that fibroblasts-derived TNFα stimulates epithelial cell LOX expression.

Figure 4: Conditioned media enhances LOX expression in vitro.

Schematic of experimental design (A). Media from esophageal epithelial cells was placed upon esophageal fibroblasts. After 24 hours, fibroblast conditioned media was placed on fresh epithelial cells in the presence of TNFα inhibitory antibody. After 24 hours of stimulation, qRT-PCR was performed for LOX (B). Data represent 3 biologic replicates with similar results.

Co-culture of fibroblasts and epithelial cells induces LOX in a TNFα-dependent manner

We examined epithelial cell expression of LOX in the presence of esophageal fibroblasts via direct co-culture experiments. In order to unequivocally determine cell-type specific gene expression in co-cultured epithelial cells and fibroblasts, we used normal esophageal epithelial cell line EPC2-hTERT and FEF3 fibroblasts permanently labeled with GFP and tdTomato, respectively. Following culturing together for 24 h, cells were subjected to separation of tdTomato-positive FEF3 and GFP-positive EPC2-hTERT cells via FACS. We performed purity analysis to confirm that populations of cells were >99% were of the intended cell fraction (Figure 5A). These purified cell populations were analyzed for LOX by real-time RT-PCR assays. LOX expression in EPC2-hTERT cells was robustly augmented in the presence of fibroblasts (Figure 4B). This result was confirmed with two independent biopsy-derived primary esophageal epithelial cell cultures²⁵ (EPC-A and EPC-B) (Figure 5B). We then used anti-TNFα antibody in the co-culture system and found decreased epithelial cell LOX production suggesting TNFα- mediated regulation of LOX expression in the esophageal epithelial cells (Figure 5C). In addition, we analyzed the FEF3 compartment for LOX production. While the fibroblasts expressed LOX, it was not significantly influenced by epithelial cells (Supplemental Figure 2).

Figure 5: Co-culture of esophageal fibroblasts and epithelial cells induces LOX expression.

FEF-tdTomato and EPC2-GFP were grown together for 24 hours and separated by FACS (A). qRT-PCR for LOX was performed on EPC2-hTERT (B) after co-culture and shown relative to monoculture (co-culture/mono-culture). This was validated with two independent primary esophageal epithelial cell cultures derived from patient biopsies, EPC-A and EPC-B. Anti-TNFα antibody was added to co-culture and qRT-PCR was performed for LOX (C). qRT- PCR for TNFα was performed on EPC2-hTERT (D) and FEF3 (E) after coculture. Data represent 3 biologic replicates with similar results. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

In order to confirm that TNFα was fibroblast derived, we performed RT-PCR on epithelial and fibroblast compartments after co-culture. We found that epithelial TNFα was significantly decreased after 24 hours of co-culture with fibroblasts (Figure 5D). Conversely, fibroblasts co-cultured with epithelial cells show a robust induction of TNFα, confirming our previous results (Figure 5E) ⁹. This suggests that fibroblast derived TNFα induces epithelial LOX production in our model system.

TNFα stimulation of TGFβ contributes to LOX expression

NFα and TGFβ have a controversial relationship. TNFα has been shown to both enhance and inhibit TGFβ in an organ-specific context^39,40. Because the relationship between TNFα in TGFβ induction in the esophagus remains elusive and canonical TGFβ signaling^41,42 is known to stimulate LOX production^43–45, we hypothesized that the TNFα –rich EoE microenvironment may induce epithelial production of TGFβ^46,47 to further perpetuate LOX production. We stimulated EPC2-hTERT with TNFα and found increased expression of TGFβ (Figure 6A). Utilizing reporter transfection assays with a panel representing multiple transcription factor responsive elements, TNFα stimulated EPC2-hTERT cells revealed SMAD- mediated transcriptional activates (data not shown). To confirm these findings, we performed luciferase assays, finding that TNFα stimulation significantly increased SMAD activity, SMAD phosphorylation, as well as nuclear translocation of phospho-SMAD (Figure 6B-D). In addition, we utilized SIS3, a SMAD inhibitor, and co-stimulated with TNFα. We found that LOX production was decreased by immunoblotting. Co-stimulation of EPC2-hTERT cells with TNFα and anti-TGFβ antibody (1D11) significantly reduced LOX expression (Figure 6D). These results suggest that epithelial LOX production is subjected to transcriptional regulation activated via a functional interplay between TNFα and canonical TGFβ signaling.

Figure 6: TNFα stimulation of TGFβ contributes to LOX expression.

TNFα stimulation of esophageal epithelial monolayer induces TGFβ expression (A), SMAD activity as measured by luciferase assay (B), increased SMAD phosphorylation (C), and nuclear localization of pSMAD3 (D). Inhibition of SMAD with SIS3 inhibited LOX production in TNFα stimulated EPC2hTERTs by immunoblot (E). Addition of anti-TGFβ antibody (1D11) to TNFα stimulated EPC2-hTERTs abrogates LOX expression (F). Isotype antibody used as control did not stimulate LOX expression (data not shown). Numbers under each band represent the signal intensity relative to NT set as 1. Data represent ≥2 biologic replicates with similar results. **p<0.01, ***p<0.001; NT- no treatment.

Discussion

Our data indicate that the level of LOX expression in the esophageal tissue corresponds with fibrotic disease in EoE. We have shown that there is increased epithelial LOX expression in EoE patients compared to non-EoE patients. Most strikingly, LOX upregulation in endoscopic biopsies from EoE patients may implicate subepithelial tissue remodeling in the early stage of EoE progression and serve clinically as a biomarker of the fibrostenotic disease condition. We further defined the role of fibroblast-epithelial interaction driving the production of LOX in vitro through activation of TNFα and TGFβ pathways.

The EoE population encompasses a diverse constellation of symptoms, endoscopic findings and histology that vary greatly based on age at diagnosis. Disease chronicity is thought to play a role in the development of fibrostenotic features of EoE^48,49. Defining fibrostenotic EoE prior to the onset of a true stricture or food impaction can be problematic in EoE⁴. Lamina propria is not interpretable by pathologists in up to 50% of pediatric EoE biopsies as the deep subepithelial compartment is not procured by relatively small endoscopy forceps. This is the first study to identify a potential biomarker of fibrostenotic disease in EoE, while several studies have been proposed for markers of disease activity¹⁸. The relatively small sample size of our population may explain the lack of sensitivity of LOX- however these data suggest that larger scale investigations are warranted. It may hold potential to detect pre-stenosis when used in conjunction with other genes or in conjunction with new functional esophageal tests like the EndoFLIP^4,50. Recent work by Rawson et al, evaluated epithelial expression of PAI-1 and found that it contributes to the fibrotic network in EoE correlating expression with vimentin, collagen 1, fibronectin, and matrix metalloproteases⁵¹. We have recently shown that patients with fibrostenotic features of EoE such as history of stricture, impaction, or rings/narrowing on endoscopy display decreased esophageal distensibility measured by the EndoFLIP⁴. Future longitudinal studies combining these transcriptional targets with functional esophageal testing will provide a better understanding of global disease activity and future risk of fibrostenosis allowing for diagnosis prior to the onset of clinical complications. We will additionally consider disease chronicity, evaluating LOX over time and determine its variability in the treatment resistant and treatment responsive cohorts.

To our knowledge, this is the first-time epithelial LOX expression has been associated with esophageal fibrosis and EoE. Foremost important findings of this study include a novel paradigm that disease-associated epithelial changes may influence subepithelial fibrostenotic disease process via paracrine-mediated interplay between epithelilal cells and fibroblasts under the EoE inflammatory milieu. Fibrosis is defined as inappropriate deposition of excessive amounts of collagen and fibrous tissue. Indeed, our gene ontology analyses of well-annotated RNA-sequencing data reveal molecular pathways essential in wound healing, where genes encoding metallopeptidases and Periostin may play critical roles besides LOX (Supplementary Table 2). Our recent work has determined that the functional properties of collagen, specifically its resistance to degradation and material stiffness, is just as important as the quantity of collagen¹⁴. Inhibiting LOX in murine models of atherosclerosis leads to decreased arterial stiffness despite high fat diets and elevated cholesterol levels⁵². LOX-induced stiffening may even be first step in fibroblast activation. Prior to the onset of fibrosis in rat models of liver disease, lysyl oxidase-mediated tissue stiffening occurs and is then followed by fibroblast activation⁵³. This observation is consistent with our previous reports that esophageal myofibroblast activation requires matrix stiffness¹⁴. Taken together, LOX expression in the esophageal epithelium could be propagating fibrosis, stimulating a positive feedback loop resulting in ongoing tissue stiffness and continued remodeling.

Mining RNA-sequencing data, we have also defined for the first time in an unbiased approach the potential role of epithelial TNFα signaling in EoE pathogenesis. Persad et al previously described enhanced TNFα-induced angiogenesis in EoE⁵⁴. We have also previously implicated TNFα in EoE-related esophageal fibrosis^8,9where fibroblast-epithelial interactions appeared to drive epithelial-mesenchymal transition in vitro and TNFα stimulation, resulting in enhanced epithelial contraction, migration and collagen deposition. In this study, we highlight another potential consequence of TNFα enrichment: enhanced TGFβ signaling. TGFβ is often cited as the “master regulator” ⁵⁵ of fibrosis. Through various activities including activation of fibroblasts, stimulation of production of extracellular matrix, and stimulation of smooth muscle contraction⁵⁶, TGFβ is likely responsible for many of the clinical symptoms of EoE. Herein, we show for the first time that the esophagus in EoE has a unique enrichment of TNFα that contributes to fibrosis through canonical TGFβ stimulation.

TNFα is known to contribute to fibrosis in other disease models¹⁰. In the intestine, there is clinical evidence that infliximab (chimeric monoclonal anti-TNFα antibody) not only decreases inflammation but also treats fibrostenotic Crohn disease⁵⁷. Infliximab was studied in 3 patients with steroid-dependent EoE. After 2 doses of infliximab, tissue eosinophilia was not improved; however, 2/3 subjects had decreased symptoms of dysphagia⁵⁸. Though the pilot study was hindered by its small study size and short duration of therapy, it brings to light the potential use of anti-TNFα therapy for EoE-associated fibrosis.

We previously demonstrated epithelial-fibroblast crosstalk via conditioned media induces secretion of pro-fibrotic cytokines IL1β and TNFα. Transfer of fibroblast conditioned medium onto epithelial cells increased expression of mesenchymal markers vimentin and αSMA and loss of E-cadherin in a TNFα-dependent manner⁹. Furthermore, when epithelial cells were stimulated with TNFα, they underwent functional changes including increased migration, contraction and enhanced collagen expression ⁸. The results of our current study suggest that co-culture (both directly and indirectly) may enhance tissue stiffness via LOX production. Future functional studies will seek to perform both gain of function and knock down of LOX in the epithelial cells and evaluate the effects of co-culture on fibroblast function, specifically, spread, contraction, collagen production, and tissue stiffness.

Therapeutic interventions in EoE are largely targeted at decreasing eosinophilic inflammation. However, the two major therapies, four food elimination diet and topical viscous steroids have remission rates ranging from 20–50%⁵⁹. Because there is no FDA-approved medication for EoE and response rates with current strategies are abysmal, targeting stiffness through LOX or TNFα inhibition may have a role in preserving the swallowing function of those with treatment resistance and fibrostenosis.

Clinical Implications.

Lysyl oxidase expression is increased in active EoE. Its expression in the esophageal epithelium is mediated by fibroblast-derived TNFα and downstream TGFβ production.

Abbreviations

DAPI

4’,6-Diamidino-2-Phenylindole

EGF

epidermal growth factor

EMT

epithelial-mesenchymal transition

EoE

eosinophilic esophagitis

GERD

gastroesophageal reflux disease

GFP

green fluorescent protein

IF

immunofluorescence

IHC

immunohistochemistry

KSFM

keratinocyte serum free media

LOX

lysyl oxidase

qRT-PCR

quantitative reverse-transcription polymerase chain reaction

3D

three-dimensional

TSLP

thymic stromal lymphopoietin

TNF

tumor necrosis factor

TGF

transforming growth factor

Th2

T helper 2

IL

interleukin

EGD

esophagogastroduodenoscopy

Hpf

high-power field

FDR

false discovery rate

FACS

fluorescence-activated cell sorting

H&E

hematoxylin & eosin