Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Feb 16.
Published in final edited form as: Gut. 2023 Dec 7;73(1):47–62. doi: 10.1136/gutjnl-2023-329455

Reflux Conditions Induce E-cadherin Cleavage and EMT via APE1 Redox Function in Esophageal Adenocarcinoma

Heng Lu 1,2,*,#, Longlong Cao 1,3,*, Farah Ballout 1, Abbes Belkhiri 4, Dunfa Peng 1,2, Lei Chen 1, Zheng Chen 1,2, Mohammed Soutto 1,5, Timothy C Wang 6, Jianwen Que 6, Silvia Giordano 7, Mary Kay Washington 8, Steven Chen 2,9, Oliver Gene McDonald 2,10, Alexander Zaika 1,2,5, Wael El-Rifai 1,2,5,#
PMCID: PMC10872865  NIHMSID: NIHMS1933180  PMID: 37734913

Abstract

Objective:

Chronic gastroesophageal reflux disease (GERD), where acidic bile salts (ABS) reflux into the esophagus, is the leading risk factor for esophageal adenocarcinoma (EAC). We investigated the role of ABS in promoting epithelial-mesenchymal transition (EMT) in EAC.

Design:

RNA sequencing data and public databases were analyzed for the EMT pathway enrichment and patients’ relapse-free survival. Cell models, pL2-IL1β transgenic mice, de-identified EAC patients’ derived xenografts (PDXs), and tissues were utilized to investigate EMT in EAC.

Results:

Analysis of public databases and RNA-sequencing data demonstrated significant enrichment and activation of EMT signaling in EAC. ABS induced multiple characteristics of the EMT process, such as downregulation of E-cadherin, upregulation of Vimentin, and activation of ß-catenin signaling and EMT-transcription factors (EMT-TFs). These were associated with morphological changes and enhancement of cell migration and invasion capabilities. Mechanistically, ABS induced E-cadherin cleavage via an MMP14-dependent proteolytic cascade. Apurinic/apyrimidinic endonuclease (APE1), also known as redox factor 1, is an essential multifunctional protein. APE1 silencing, or its redox-specific inhibitor (E3330), downregulated MMP14 and abrogated the ABS-induced EMT. APE1 and MMP14 co-expression levels were inversely correlated with E-cadherin expression in human EAC tissues and the squamocolumnar junctions of the L2-IL1ß transgenic mouse model of EAC. EAC patients with APE1high and EMThigh signatures had worse relapse-free survival than those with low levels. In addition, treatment of PDXs with E3330 restrained EMT characteristics and suppressed tumor invasion.

Conclusion:

Reflux conditions promote EMT via APE1 redox-dependent E-cadherin cleavage. APE1-redox function inhibitors can have a therapeutic role in EAC.

Keywords: APE-1/Ref-1, redox, MMP14, Barrett’s esophagus, EMT, migration, invasion

Graphical Abstract

graphic file with name nihms-1933180-f0009.jpg

Schematic summary of ABS-induced E-cadherin cleavage and EMT in EAC cells

APE1 and MMP14 expression are maintained at relatively low levels under normal conditions. Intact E-cadherin on the cell surface supports stable adherens junctions between epithelial cells. E-cadherin can effectively sequester ß-catenin at the plasma membrane, thus inhibiting ß-catenin transcriptional activities. Exposure to acidic bile salts (ABS), a well-studied in vitro model of gastroesophageal reflux, disrupts the homophilic binding of E-cadherin molecules, increases APE1 protein level and promotes MMP14 cell surface expression, further activating E-cadherin cleavage. This cleavage converts 135 kDa full-length E-cadherin into 80 kDa soluble ectodomain fragment (sEcad) and 33 kDa intracellular fragment (CTF2), which releases ß-catenin into the nucleus. The activation of WNT/ß-catenin signaling elevates EMT-transcriptional factors to induce EMT. Repeated exposures to acidic bile salts, mimicking chronic GERD, enhance this EMT processing and promote EAC malignancy. APE1 redox-specific inhibitor, E3330, can inhibit ABS-APE1-redox-MMP14 axis activation and restrain EMT in EAC.

Introduction

Gastroesophageal reflux disease (GERD), where acidic bile salts (ABS) abnormally refluxate into the esophagus, is the leading risk factor for developing metaplasia of the esophageal epithelium, known as Barrett’s esophagus (BE) [1] and its progression to esophageal adenocarcinoma (EAC) [2, 3, 4]. EAC is an aggressive disease with poor response to therapy where the 5-year survival rate is below 20% [5, 6]. The incidence of BE and EAC has been steadily rising to alarming levels in the United States and other industrialized countries [5, 7, 8]. Chronic exposure to reflux conditions in EAC patients with GERD promotes cancer progression via cancer cell expansion, migration, and invasion [9, 10, 11, 12]. Understanding the biology and molecular mechanisms of EAC tumorigenesis is an essential step for developing novel therapeutic strategies.

APE1 (apurinic/apyrimidinic endonuclease), also known as redox factor 1 (REF1), is an essential multifunctional protein involved in oxidative DNA damage repair and redox-dependent transcriptional regulation. APE1 protein is overexpressed in more than half (70/130) of EAC tissue samples [13]. Exposure to ABS to mimic GERD conditions generates high levels of oxidative stress and DNA damage. High levels of APE1 protein are essential for DNA base excision repair and activation of oncogenic STAT3 signaling under reflux conditions in EAC [13, 14]. A recent study has shown that APE1 can regulate MMP14 levels via ARF6-mediated endocytosis/recycling in EAC [15]. However, the complete molecular functions of APE1 in response to ABS exposure in EAC tumorigenesis remain to be fully understood.

Epithelial-to-Mesenchymal Transition (EMT) is a biological process defined by morphological changes and prototypical markers, such as E-cadherin and Vimentin. It is executed by EMT-activating transcription factors (EMT-TFs), mainly SNAIL, TWIST, and ZEB families [16]. EMT plays an essential role in different stages of tumorigenesis, including initiation, invasion, metastasis, and therapeutic resistance [17]. A significant reduction of E-cadherin level occurs in Barrett’s metaplasia-dysplasia-adenocarcinoma progression cascade [18, 19, 20]. The role of reflux conditions in promoting the loss of E-cadherin and activating the EMT process in EAC tumorigenesis awaits complete comprehension [21].

This study demonstrates that ABS in reflux conditions promote MMP14-mediated ectodomain cleavage of E-cadherin. Repeated exposures to ABS, the mimic of chronic GERD, enriched the EMT traits and enhanced migration and invasion of EAC cells. Inhibition of APE1-redox function suppresses invasion capabilities of cancer cells in vivo.

Materials and Methods

Cell culture

Human CP-B, OE19, OE33, FLO-1, and SK-GT-4 cell lines were used in this study, following standard culture techniques. Additional details are provided in supplementary methods.

Bioinformatic analysis

Seven public databases of EAC patients were downloaded from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO). Bioinformatics analysis included the following datasets; TCGA-EAC (79 EAC and 9 normal esophagus (NE)), GSE26886 [22] (21 EAC, 20 Barrett’s esophagus (BE), and 19 NE), GSE37200 [23] (15 EAC and 31 BE), GSE13898 [24] (64 EAC, 15 BE, and 28 NE), GSE74553 [25] (70 EAC and 8 NE), GSE92396 [26] (12 EAC and 9 NE), and GSE37203 [27] (37 EAC and 31 BE). Additional details are provided in supplementary methods.

Exposure to acidic bile salts (ABS)

The acidic bile salts (ABS) cocktail was prepared to mimic the mixture of bile acids in the distal esophagus during GERD, as previously described [14]. The acidic bile salts cocktail was prepared as an equimolar mixture of sodium salts, including glycocholic acid, taurocholic acid, glycodeoxycholic acid, glycochenodeoxycholic acid, and deoxycholic acid. The cells were exposed to 200 μM of the bile salts cocktail (40 μM of each bile salt as above) in a pH 4.0 serum-free medium for 20 min. This was followed by recovery in complete media. For the collection of conditioned media post ABS exposure, the recovery process was conducted using media containing 0.5% FBS. For repeated ABS exposure, cells were exposed to ABS (pH5.5, 200 μM, 20min) followed by recovery on a daily basis for 15 days; untreated cells served as the control.

RNA-sequencing (RNA-seq) Data Analysis

OE33 cells and FLO-1 cells with stable APE1-knockdown (shAPE1) or scramble shRNA (shCtrl) were exposed to ABS for 20 min, followed by washing in PBS and recovery in a regular medium for 6h. The RNA samples from triplicated experiments were extracted by RNeasy Mini Kit (Qiagen, Hilden, Germany) and were sent to GENEWIZ (South Plainfield, NJ) for high-throughput sequencing (Illumina HiSeq instrument, 4000 or equivalent). STAR aligner v.2.5.2b was used to align trimmed sequence reads and BAM files were generated. Unique gene hit counts were calculated by using feature Counts from the Subread package v.1.5.2. The raw counts were then normalized as transcripts per million (TPM), followed by further correction for batch effects using Combat. Limma package (Version 3.26.9) was applied for gene differential analysis and Gene Set Enrichment Analysis (GSEA) was used to examine the enrichment of signal pathways. P value mean(|log2FC|)+2×sd(|log2FC|) were regarded as the thresholds.

Human samples

De-identified human tissues were utilized in this study in accordance with the NIH guidelines for non-human subjects research. Tissue microarrays (TMA) containing 35 de-identified archival cases of EACs as well as normal esophagus, and dysplastic and non-dysplastic BE were constructed by Tissue Pathology Core at Vanderbilt University Medical Center, Nashville, TN. These EAC patients had not undergone chemotherapy or radiotherapy at the time of sample collection. All tissue samples were histologically verified, and representative regions were selected for inclusion in the TMA.

Animal experiments

The pL2-IL1β transgenic mice are a kind gift from Dr. Timothy Wang (Columbia University); a model of chronic esophageal inflammation that develops BE and EAC, as previously described [28]. The mice received drinking water containing 0.3% deoxycholic acid (DCA) at neutral pH at the age of three months. De-identified patient-derived xenografts (PDXs) from the human gastro-esophageal junction were generated using a previously described platform [29]. The PDX with the designation GTR0165 was selected for the experiment. Additional details are provided in supplementary methods.

Statistical analysis

The results of all quantification analyses were presented as the mean (SD) derived from triplicate independent experiments. Differences were analyzed by Student’s t-test or one-way ANOVA followed by the Bonferroni post-hoc test. All statistical analyses were performed using the GraphPad Prism 8. *, p < 0.05. **, p < 0.01. N.S., no significance.

Other materials and methods are described in Supplementary Methods.

Results

Gene set enrichment analysis demonstrates hyperactivation of EMT signaling in EAC in response to acidic bile salts.

To better understand the mechanisms underlying EAC carcinogenesis, we screened public EAC gene expression databases for gene set enrichment analysis. These datasets included TCGA and 6 GEO datasets (GSE26886, GSE37200, GSE13898, GSE74553, GSE92396, and GSE37203). Comparing human EAC samples with adjacent normal esophagus samples (Normal) or Barrett’s esophagus (BE), we selected the top 10 positively enriched signaling pathways, as well as the top 10 negatively enriched signaling pathways, by the mean normalized enrichment scores (NES) of the multiple datasets (Figure 1A, Supplementary Figure 1A&B). EMT signature was the most significant and consistent positively enriched signaling pathway in EAC, as compared to the normal esophagus, in multiple datasets such as TCGA (Figure 1B), our published dataset GSE92396 (Supplementary Figure 1C), and GSE13898 (Supplementary Figure 1D). We also observed sequential EMT enrichment from adjacent normal tissues to Barrett’s esophagus (BE) (Figure 1C) and from BE to EAC (Figure 1D) in the GSE1420 dataset. These data suggest that EMT signature correlates with EAC carcinogenesis cascade.

Figure 1. Enrichment of EMT signature in EAC patients and ABS-provoked EAC cells.

Figure 1.

Open in a new tab

A, Gene Set Enrichment Analysis (GSEA) of TCGA and GEO databases, comparing human EAC with normal human esophageal (NE) tissues. The top 10 enriched signaling pathways were exhibited with the mean normalized enrichment score (NES) volume through TCGA (EAC) and 6 GEO datasets (GSE26886, GSE37200, GSE13898, GSE74553, GSE92396, and GSE37203) by R language (Ridgeline). B, EMT signal pathway enrichment of TCGA-EAC, comparing EAC with NE. C&D, EMT signal pathway enrichment of GSE1420, comparing human Barrett’s esophagus (BE) with NE (C), or comparing EAC with BE (D). E, GSEA of local RNA-seq dataset from OE33 cells with exposure to acidic bile salts (ABS; 200 μM, pH4.0 for 20 min, followed by 6 h of recovery in complete media), compared to untreated control cells (ABS vs. Control). The most significantly enriched signal pathways were exhibited with the mean NES of triplicated experiments (p < 0.01) by R language (Ridgeline). FDR q <0.05, statistical significance. F, EMT signal pathway enrichment of the RNA-seq dataset as in E. G&H, OE33 cells were exposed to ABS, followed by 0, 1, 3, and 6 h of recovery. The cells were harvested for Western blots (G) and qRT-PCR of CDH1 mRNA expression (H). I&J, Representative immunofluorescent images of E-cadherin (green) and Vimentin (red) in the OE33 cells in 2D culture (I) and in 3D organotypic culture (J); nuclei were stained with DAPI (blue). Hematoxylin and eosin (H&E) staining of sequential cut of the same blocks indicated stroma and top epithelial layer in the 3D organotypic culture. The cells in 2D culture or the cells in the top epithelial layer of 3D culture were exposed to ABS, followed by 24 h of recovery. The untreated (UT) cells worked as a control. The white arrows indicated Vimentinhigh cancer cells. The mean fluorescence intensity (MFI) of E-cadherin and Vimentin from triplicate independent experiments was quantified in the right panels.

Although chronic GERD is the primary risk factor for EAC, the mechanisms underlying the role of ABS reflux in EAC carcinogenesis remain poorly understood. To determine the signaling pathways in EAC, we performed RNA-seq analysis of OE33 cells exposed to ABS, a well-studied model of GERD. GSEA analysis revealed EMT, TNFα/NF-kB, and IL2/STAT5 as the most significantly enriched pathways in response to ABS, compared to control (Figure 1E). EMT was the top positively enriched signaling pathway in response to ABS (Figure 1F), suggesting that reflux conditions play a critical role in activating EMT signaling in EAC.

Acidic bile salts repress E-cadherin and induce Vimentin expression in EAC cells.

We examined the expression of E-cadherin and Vimentin, as EMT markers, in OE33 cells exposed to ABS. Western blots demonstrated a remarkable decrease in E-cadherin at 6h recovery, following a single transient (20 min) ABS exposure. We also detected an increase in Vimentin, a marker of EMT activation (Figure 1G). The expression of E-cadherin slowly recovered to baseline within 48h, post-exposure to ABS (Supplementary Figure 1E). The decrease in E-cadherin protein levels was not due to changes in CDH1 mRNA expression (Figure 1H, Supplementary Figure 2A). Similarly, gene expression analysis of multiple EAC datasets, such as TCGA (Supplementary Figure 2B), GSE92396 (Supplementary Figure 2C), and GSE1420 (Supplementary Figure 2D), did not reveal significant differences in CDH1 mRNA levels, comparing normal esophagus (NE) with EAC or BE. Immunofluorescence staining showed a reduction in E-cadherin and an increase in Vimentin in OE33 cells (Figure 1I) and OE19 cells (Supplementary Figure 2E) in response to ABS. The 3D organotypic cultures (OTC) of OE33 cells (Figure 1J) and OE19 cells (Supplementary Figure 2F) recapitulated these results. Collectively, these results suggest a post-transcription regulatory mechanism of E-cadherin protein levels in EAC tumorigenesis. Of note, EMT changes in response to ABS were less evident in FLO1 cells, a mesenchymal-like-type cell line of EAC [30], having lower E-cadherin and higher Vimentin baseline expression levels than epithelial-type OE33 cells.

Exposure to ABS promotes E-cadherin cleavage via induction of MMP14.

We investigated E-cadherin cleavage as a possible mechanism underlying its protein level changes in response to ABS. Following ABS conditions, Western blots data demonstrated an accumulation of soluble E-cadherin ectodomain fragment (80 kDa, sE-cad) in the conditioned media (CM) and reduction in its full-length (135 kDa) E-cadherin protein (E-cad) in the whole cell lysates (WCL), in both OE33 and OE19 cells, compared with untreated control (UT) cells (Figure 2A). The results suggested that ABS induced extracellular cleavage of E-cadherin. Along with these findings, the extracellular matrix (ECM) regulators signaling pathways were enriched in EAC datasets, TCGA-EAC (Figure 2B) and GSE92396 (Figure 2C).

Figure 2. ABS induce E-cadherin cleavage and upregulate MMP14.

Figure 2.

Open in a new tab

A, Conditioned media (C.M.) from OE33 and OE19 cells with ABS exposure were collected and concentrated. 1/10 of the final concentrated C.M. were loaded for Western blots to examine 80kDa soluble E-cadherin fragment (sE-cad). The cell lysates (WCL) from the relevant dishes were collected to detect 135 kDa full-length E-cadherin and ß-actin. The cells were exposed to ABS, followed by 6 h recovery with the media containing 0.5% FBS. B&C, Extracellular matrix regulators signaling enrichment in TCGA (B), and GEO dataset, GSE92396 (C), comparing EAC to the normal esophagus. D, 10 μM MMPs inhibitor, GM6001, 20 μM TACE/ADAM17 inhibitor, TAPI-1, or vehicle control (Ctrl) were used to pretreat OE33 or OE19 cells and maintained in the complete medium during 6 h recovery after ABS exposure. E, EMT signal pathway enrichment in TCGA, comparing MMP14high EAC with MMP14low EAC. The mean (SD) of MMP14 expression was used as the cut-off point to divide the patients into high, middle, and low expression groups. F, Western blots analysis of OE33 and OE19 cells exposed to ABS, followed by 6 h recovery. G&H, Representative immunofluorescent images of E-cadherin (green) and MMP14 (red) in the OE33 cells in 2D culture (G) and in 3D organotypic culture (H); nuclei were stained with DAPI (blue). Hematoxylin and eosin (H&E) staining of sequential cut of the same blocks indicated stroma and top epithelial layer in the 3D organotypic culture. The cells in 2D culture or the top epithelial layer of 3D culture were exposed to ABS, followed by 24 h recovery. The untreated (UT) cells worked as the control. The mean fluorescence intensity (MFI) of E-cadherin and MMP14 from triplicate independent experiments were quantified in the right panels for G&H, respectively.

To identify the primary protease that contributes to ABS-induced E-cadherin cleavage, we examined the expression of known E-cadherin proteases in the TCGA-EAC dataset. Among the ECM regulators enriched in the TCGA-EAC patients, MMP family members (MMP3, MMP7, MMP9, and MMP14) and ADAM family members (ADAM10 and ADAM17) were overexpressed in EAC samples (Supplementary Figure 3A-F). We next treated OE33 and OE19 cells with the MMPs inhibitor, GM6001, or ADAMs inhibitor, TAPI-1. Western blots showed that GM6001, not TAPI-1, protected full-length E-cadherin from ABS-induced cleavage and released less soluble E-cadherin (Figure 2D). These results suggest that ABS induce E-cadherin cleavage via MMPs activation. To further narrow down the protease candidates, we performed the single gene cluster analysis and subsequent GSEA of TCGA-EAC datasets as described in Methods and previously reported [31]. The bioinformatics analysis indicated that the EMT gene set is enriched in the MMP14high expression group compared to the MMP14low expression group (Figure 2E). This data demonstrates a strong correlation between MMP14 and EMT signaling in EAC. The highest NES (3.055) among multiple MMPs (Supplementary Figure 3G-I), suggested MMP14 as the potential protease for E-cadherin cleavage in EAC. Western blots identified that ABS exposure increased MMP14 expression, parallel to E-cadherin cleavage, in OE33 and OE19 cells (Figure 2F). These results were confirmed by immunofluorescent staining of OE33 cells (Figure 2G) and OE19 cells (Supplementary Figure 3J), as well as 3D organotypic cultures (Figure 2H).

APE1-redox-MMP14 axis is crucial for ABS-induced E-cadherin cleavage and ß-catenin activation.

We next clarified the contribution of MMP14 in ABS-induced E-cadherin cleavage. Silencing MMP14 abolished ABS-induced E-cadherin cleavage, including soluble E-cadherin (sE-cad) release, full-length E-cadherin decrease, and the accumulation of CTF2 fragment (the 33kDa intracellular fragment product of E-cadherin cleavage), in both OE33 (Figure 3A) and OE19 cells (Supplementary Figure 4A). Blocking MMP14 on the cell surface, using antibody neutralization (α-MMP14), abrogated ABS-induced E-cadherin cleavage (IgG) as compared to untreated (UT) control cells (Figure 3B). It was previously reported that APE1, a multi-function protein involved in DNA base excision repair and redox, is induced in response to ABS in EAC [13, 14]. It has been reported that APE1 can regulate MMP14 expression and recycling in EAC [15]. Therefore, we investigated whether APE1 is required for MMP14-mediated E-cadherin cleavage. Western blots (Figure 3C) and immunofluorescent staining (Supplementary Figure 4B) demonstrated an increase in APE1 level along with E-cadherin cleavage and MMP14 induction in response to ABS. APE1 silencing by stable knockdown (shAPE1) prevented ABS-induced E-cadherin decrease in OE33 cells in response to ABS exposure, compared with scrambled shRNA control (shCtrl) (Supplementary Figure 4C). Similarly, APE1 knockdown (shRNA or siAPE1) consistently abrogated ABS-induced E-cadherin cleavage (Figure 3D and Supplementary Figure 4D).

Figure 3. ABS induces E-cadherin cleavage and ß-Catenin activation in an MMP14-dependent manner.

Figure 3.

Open in a new tab

A, MMP14-knockdown (siMMP14) OE33 cells and scrambled siRNA control (siCtrl) cells were exposed to ABS, followed by 6 h recovery with the media containing 0.5% FBS. Conditioned medium and whole cell lysates were collected for Western blots. B, OE33 cells were exposed to ABS, followed by 6 h recovery, in the presence of rabbit MMP14 antibody (10 μg/ml), or IgG. C, OE33 cells were exposed to ABS, followed by 0, 1, 3, and 6 h recovery. D, APE1-knockdown (shAPE1) OE33 cells and scrambled shRNA cells (shCtrl) were exposed to ABS, followed by 6 h recovery. E, 40 μM E3330 was used to pretreat OE33 cells and maintained in the complete medium during 6 h recovery after ABS exposure. F, Control Vector, FLAG-tagged wild-type APE1, C65A (Redox function deficient mutant), or H309N (DNA repair function defective mutant) was overexpressed in APE1-knockdown (shAPE1) OE33 cells. Scrambled shRNA OE33 cells worked as the control (shCtrl). The cells were exposed to ABS, followed by 6 h recovery. G&H, OE33 and OE19 cells were exposed to ABS, followed by 6 h recovery. The cells were harvested for Western blots (G), and TOP/FOP luciferase assays were performed to examine ß-catenin activation (H). I, OE33 cells were exposed to ABS, followed by 6 h recovery, in the presence of rabbit MMP14 antibody (10 μg/ml), MMP2 antibody (20μg/ml), or IgG. J, OE33 cells were exposed to ABS (200 μM, pH4.0), pH4.0 medium, or relevant bile salts mixture (200 μM) for 20 min, followed by 6 h recovery. In particular, a mouse MMP14 antibody was utilized to detect MMP14 by Western blots after the neutralization experiment (B&I).

To determine which function of APE1 facilitates ABS-induced E-cadherin cleavage, redox-specific inhibitor (E3330) [32, 33, 34] or DNA repair-specific inhibitor (APE1-i3) [35] was used to pre-treat OE33 cells before ABS exposure. E3330, not APE1-i3, protected full-length E-cadherin from ABS (Supplementary Figure 4E-F). We found that E3330 abrogated ABS-induced E-cadherin cleavage in both OE33 and OE19 cells (Figure 3E). To verify the inhibitors’ functions, we utilized genetic approaches for the reconstitution of FLAG-tagged wild-type APE1 or mutant APE1 (APE1-C65A (redox-defective mutant) or APE1-H309N (DNA-repair-defective mutant) in OE33 cells with stable APE1-knockdown (shAPE1), as previously described [36]. APE1 silencing abrogated ABS-induced E-cadherin cleavage as expected (lane 4 vs. lane 3, Figure 3F). More importantly, we detected a significant restoration of ABS-induced E-cadherin cleavage by reconstitution of wild-type APE1 or H309N mutant, but not redox-defective C65A mutant, in APE1-knockdown cells, as compared to the control vector ( lane 4 to lane 7, Figure 3F). These results confirm the role of APE1 redox function, not the DNA repair function, for ABS-induced E-cadherin cleavage. Pre-treatment of cells with ROS scavenger N-acetylcysteine (NAC) abrogated ABS-induced APE1 and MMP14 overexpression in OE33 cells (Supplementary Figure 4G), suggesting ROS-dependent activation of APE1-MMP14 under reflux conditions.

ß-catenin accumulation and activation are well-known results of E-cadherin cleavage [37, 38]. GSEA identified significant enrichment of the ß-catenin/TCF pathway in TCGA-EAC (Supplementary Figure 5A). Western blots demonstrated ß-catenin activation, E-cadherin decrease, and APE1-MMP14 induction, following ABS exposure (Figure 3G). The pTOP/FOP assay, as a measure of β-catenin/TCF transcription activity, confirmed β-catenin activation (Figure 3H). MMP14 and MMP2, an MMP14 proteolytic downstream target are known potential E-cadherin proteases [39, 40]. Of note, antibody neutralization of MMP14 or MMP2 abolished the effects of ABS exposure, suggesting MMP14/MMP2 proteolytic cascade is crucial for ABS-induced E-cadherin cleavage and subsequent ß-catenin activation (Figure 3I). Additionally, Western blots and immunofluorescent staining showed similar effects in non-dysplastic BE cells (BART), such as MMP14 induction, E-cadherin cleavage, and ß-catenin activation (Supplementary Figure 5B-D), suggesting a similar mechanism in BE/EAC tumorigenesis stages. Notably, acidic pH4.0 medium or bile salts mixture alone, did not have similar effects on the induction of E-cadherin cleavage and ß-catenin activation (Figure 3J), as ABS. These results suggest that both acidic conditions (pH4.0) and bile salts are necessary for optimum E-cadherin cleavage and activation of ß-catenin.

APE1 silencing restricts EMT in EAC.

The OE33 and OE19 cell lines show an epithelial phenotype, whereas the CPB, FLO-1, and SK-GT4 cell lines display a mesenchymal phenotype [30]. To investigate the potential function of APE1 in EMT, we generated stable APE1-knockdown (shAPE1) in both epithelial-type EAC (OE33 and OE19) and mesenchymal-like high-grade dysplasia (CPB) and EAC (FLO-1 and SK-GT-4) cells. Western blots at recovery time courses (from 0 h to 24 h), following ABS exposure, indicated that APE1-knockdown (shAPE1) OE33 cells didn’t respond to ABS as compared to the control cells (shCtrl) (Figure 4A). The pTOP/FOP assays demonstrated genetic knockdown of APE1 or MMP14 by siRNA inhibited ABS-induced ß-catenin transactivation (Figure 4B). These results suggested ABS induces E-cadherin cleavage and ß-catenin activation in an APE1/MMP14-dependent manner. Phalloidin staining showed that APE1 silencing changed mesenchymal-like cell shapes from spindle-like to cobblestone-like (Figure 4C), a typical morphological marker of mesenchymal-to-epithelial transition (MET) reverse to EMT. Western blots analysis of EMT markers revealed that APE1 silencing decreased Vimentin and ZEB1, increased E-cadherin, and inactivated ß-catenin in CPB cells (Figure 4D). Immunofluorescent staining further confirmed the elevation of E-cadherin and repression of Vimentin in APE1-knockdown FLO-1 cells, compared with the control (shCtrl) in both 2D cultures (Supplementary Figure 6A) and 3D organotypic cell cultures (Figure 4E). In addition, qRT-PCR showed consistent repression of SNAI1, SNAI2, and ZEB2 mRNAs, following APE1 silencing in CPB, FLO-1, and OE33 cells (Supplementary Figure 6B-D). Furthermore, wound healing assay (Figure 4F, Supplementary Figure 6E), Boyden chamber migration assay (Figure 4G, Supplementary Figure 6F), and invasion assay (Figure 4H, Supplementary Figure 6G) indicated APE1 silencing significantly repressed migration and invasion capabilities of FLO-1 cells. These findings support the critical role of APE1 in maintaining EMT characteristics in BE and EAC cells.

Figure 4. Knockdown of APE1 dramatically restricts EMT in EAC.

Figure 4.

Open in a new tab

A, The Western blots analysis of stable APE1-knockdown cells (shAPE1) OE33 cells exposed to ABS, followed by indicated time courses of recovery. Scrambled shRNA cells were used as the control (shCtrl). B, The OE33 cells with knockdown of APE1 (siAPE1), MMP14 (siMMP14), or scrambled siRNA control (siCtrl), were exposed to ABS, followed by 6h recovery, and collected for TOP/FOP luciferase assays. C, Representative Alexa Fluor™ 488 Phalloidin staining images of APE1-knockdown (shAPE1) cells and scrambled shRNA control (shCtrl) cells in CPB, FLO-1, and SK-GT-4 cell lines. Nuclei were stained with DAPI (blue). D, Western blot analysis of APE1-knockdown cells and control cells in CPB, FLO-1, SK-GT-4, OE33, and OE19 cell lines. E, Representative immunocytochemistry images of APE1-knockdown (shAPE1) and control (shCtrl) FLO-1 cells in 3D organotypic culture. Hematoxylin and eosin (H&E) staining of sequential cut of the same blocks. The mean fluorescence intensity (MFI) of E-cadherin and MMP14 from triplicate independent experiments was quantified in the right panel. F, APE1-knockdown, and control FLO-1 cells were subjected to wound healing assays. The relative moving distance of migrating FLO-1 cells at 12, 18, and 24 h post-scratching, compared to 0 h control, were quantified by triplicated independent experiments. G&H, Quantification of Boyden chamber migration assay (G) and Matrigel™ Invasion Chamber assay (H) of APE1-knockdown (shAPE1) and control (shCtrl) FLO-1 cells. Migrated or invaded cells in triplicate wells per group were quantified.

Repeated ABS exposure induces EMT in epithelial-type EAC cells.

GERD is usually treated with acid-suppressive therapy, especially proton pump inhibitors (PPIs) [41]. However, this treatment does not eliminate exposure to bile salts in reflux. In fact, GERD patients who develop EAC continue to have reflux episodes. To closely mimic chronic reflux in vitro, we exposed EAC cells to ABS (200 μM, pH5.5, 20min) followed by recovery on a daily basis for 15 days, shown as repeated ABS (rABS) exposure. OE33 is an epithelial-type EAC cell line. Phalloidin staining of OE33 cells exposed to rABS showed a striking morphologic change characterized by loss of cell-cell contact and the cobblestone-like phenotype with failure to form a confluent monolayer or islands. OE33 cells became elongated, spindle-shaped, and scattered (Figure 5A, Supplementary Figure 7A). Of note, the OE33 cells with repeated exposure to neutral (pH7.0) bile salts (BS) mixture only showed moderate morphological changes in a small population, suggesting that acidic bile salts are more potent inducers of EMT than bile salts alone. The quantification of cobblestone-like epithelial-type cells and spindle-like mesenchymal-type cells demonstrated that rABS induced mesenchymal-like morphology in 55% of cells. In contrast, neutral BS induced that in 8% of cells (Figure 5B). Immunofluorescence staining indicated that 50% of OE33 cells showed repression of E-cadherin and induction of Vimentin expression, following rABS exposures. In contrast, only 10% of cells showed these changes under neutral BS, compared with untreated (UT) cells (Figure 5C&D). On the other hand, repeated exposure to an acidic medium (pH5.5) alone couldn’t induce Vimentin expression in OE33 cells compared to rABS (Supplementary Figure 7B). The observed rABS-induced changes are typical of the fibroblastoid cells formed during the EMT process. qRT-PCR confirmed significant elevation of SNAI1, SNAI2, ZEB2, and VIM (Vimentin) mRNA expression in OE33 cells (Figure 5E-I) and OE19 cells (Supplementary Figure 7C-G) under rABS. Moreover, Western blots showed that rABS exposures promoted APE1/MMP14 induction, E-cadherin decrease, and ß-catenin activation in OE33 and OE19 cells (Figure 5J). APE1 silencing (siAPE1) (Figure 5K), as well as treatment with APE-redox-specific inhibitor (E3330) (Figure 5L), abrogated these characteristic changes of EMT. Taken together, our findings indicated that rABS promotes remarkable EMT progression via the signaling axis of APE1/MMP14/E-cadherin cleavage in epithelial-type EAC cells.

Figure 5. Repeated exposure to ABS promotes the emergence of EMT phenotype in epithelial-type EAC cells via the APE1-MMP14 axis.

Figure 5.

Open in a new tab

A, Representative Alexa Fluor™ 488 Phalloidin staining images of OE33 cells exposed to repeated ABS (rABS; pH5.5, 200 μM, 20min, followed by recovery every day), or 200 μM bile salts (BS), for total of 15 days; untreated cells worked as the control. B, Quantification of morphological epithelial type or mesenchymal-like cells from triplicate independent experiments under 20x objective lens as in A. C, Representative immunofluorescent images of E-cadherin (green) and Vimentin (red) of the OE33 cells with repeated exposure as in A; the emergence of E-cadherin low and Vimentin high cells is most notable in the lower left quadrant and upper right quadrant. D, the quantification of Vimentin-positive and -negative cells were shown as the mean ± SD from triplicate independent experiments, normalized to total cells (%) as in C. E-I, qRT-PCR of the OE33 cells with repeated exposure to ABS or BS for 15 days. J, OE33 and OE19 cells were exposed to repeated ABS (rABS) for 6 days before collection for Western blots. K, APE1-knockdown OE33 cells (shAPE1) and control cells (shCtrl) were exposed to repeated ABS (rABS) for 6 days. L, OE33 cells, after 6 days of repeated ABS exposure, were further treated with 50 μM E3330 overnight.

APE1 plays a crucial role in ABS-mediated migration and invasion capabilities of EAC cells.

To further investigate the role of APE1 in ABS-induced EMT, we performed RNA-sequencing by using APE1-silencing (shAPE1) FLO-1 cells and control (shCtrl) cells following exposure to ABS. GSEA analysis demonstrated significant enrichment of EMT signature, comparing ABS-exposed cells to untreated (UT) cells (Figure 6A), which is consistent with RNA-seq analysis of OE33 with ABS exposure (Figure 1F). Moreover, the EMT signaling pathway was negatively enriched (NES=−1.881) upon comparing APE1-knockdown FLO-1 cells with the shRNA control cells, both with ABS exposure (shAPE1+ABS vs. shCtrl+ABS) (Figure 6B). These results suggest that APE1-silencing abrogates ABS-induced EMT signaling activation. These findings confirm APE1’s indispensable role in the EMT of EAC cells.

Figure 6. APE1 and redox function play a crucial role in acquired migration/invasion capabilities by repeated ABS.

Figure 6.

Open in a new tab

A&B, EMT signal pathway enrichment of the local RNA-seq dataset from FLO-1 cells, comparing exposure to ABS with the untreated control cells (ABS vs. Control) (A), or comparing APE1-knockdown cells with scrambled shRNA control cells after exposure to ABS (shAPE1+ABS vs. shCtrl+ABS) (B). C&D, Quantification of Boyden Chamber Migration assay (C) and Matrigel™ Invasion Chamber assay (D) of FLO1 cells and OE33 cells with APE1-knockdown or scrambled shRNA control. The cells were exposed to repeated ABS for 6 days before collection for migration and invasion assays. Migrated or invaded cells in triplicate wells per group were quantified. The migration/invasion of the untreated shCtrl cells was normalized as 100%. E&F, Representative Alexa Fluor™ 488 Phalloidin staining images of FLO-1 cells (E) and OE33 cells exposed to 6 days repeated ABS (F). The cells were treated with 50 μM E3330 treatment for 48h before collection for staining. The epithelial type and mesenchymal-like cells from triplicate independent experiments were quantified as in the right panels. Untreated cells were used as the control (Ctrl). G&H, Quantification of Boyden Chamber Migration assay (G) and Matrigel™ Invasion Chamber assay (H) of FLO1 cells and OE33 cells. Before migration assay or invasion assay, the cells were exposed to repeated ABS (rABS) for 6 days, followed by overnight treatment of 50 μM E3330. The migration/invasion of the untreated control cells was normalized as 100%. Migrated or invaded cells in triplicate wells per group were quantified.

The loss of cell-cell adhesion, following the reduction of E-cadherin, is known to mediate enhanced cell motility. Using the Boyden chamber migration assays and invasion assays, following rABS, we detected significantly increased cell migration (Figure 6C, Supplementary Figure 8A) and invasion (Figure 6D, Supplementary Figure 8B) in FLO-1 and OE33 cells, as compared to controls (shCtrl/untreated vs. shCtrl/rABS). We also examined the role of APE1 in ABS-regulated EAC migration/invasion since we found that APE1 can regulate EAC motility (Figure 4). As expected, APE1 silencing abolished ABS-enhanced cell migration (Figure 6C, Supplementary Figure 8A) and invasion (Figure 6D, Supplementary Figure 8B) in both FLO-1 and OE33 cells, comparing rABS/shCtrl with rABS/ shAPE1. As we found that APE1-redox mediates ABS-induced E-cadherin cleavage, we investigated the role of APE1-redox function in repeated ABS-induced EMT in EAC. Phalloidin staining showed that E3330 treatment, APE1 redox inhibitor, induced the morphological change from mesenchymal-type spindle-like to epithelial-type cobblestone-like in 40% of FLO-1 cells (Figure 6E). Similar results of E3330 treatment were observed in rABS-induced mesenchymal-like OE33 cells. The E3330 treatment revoked rABS-induced morphological changes in OE33 cells by reducing mesenchymal-like cells from 65% to 25% (Figure 6F). Consistently, E3330 abolished rABS-enhanced cell migration and invasion abilities (Figure 6G-H, Supplementary Figure 8C-D) in both FLO-1 cells and OE33 cells. Together, these results suggest that repeated ABS exposure enhances the migration and invasion capabilities of EAC cells via an APE1-redox-dependent function.

Co-overexpression of APE1 and MMP14 inversely correlates with E-cadherin expression in human patients and mouse EAC models.

A significant reduction of E-cadherin expression has been widely reported in EAC development and progression cascades [18, 19, 20]. Therefore, we further investigated the expression of E-cadherin, APE1, and MMP14 in human EAC tissues. Immunofluorescence co-staining of human tissues demonstrated an inverse correlation between E-cadherin and APE1 expression: high levels of E-cadherin and low expression of APE1 in the normal adjacent to the tumor (NAT) (Figure 7A, panel a&b, white arrows indicated; H/E in Supplementary Figure 9A-a); loss or reduction of E-cadherin expression and high expression of APE1 in EAC tumor core (Figure 7A, panel c&d, white arrows indicated; H/E in Supplementary Figure 9A-b). In contrast, we observed the co-expression pattern of APE1 and MMP14: low expression of either APE1 or MMP14 in NAT (Figure 7B, panel a&b, white arrows indicated; H/E in Supplementary Figure 9A-a); co-overexpression of APE1 and MMP14 in the neoplastic glandular areas of EAC tumor (Figure 7B, panel c&d, white arrows indicated; H/E in Supplementary Figure 9A-b). APE1 immunostaining was predominantly nuclear, whereas E-cadherin and MMP14 were mainly detected on the cell surface, as expected. We also validated our findings in vivo, using the L2-IL1β transgenic mouse model of EAC [28]. These mice develop neoplastic lesions at the squamo-columnar junctions that include high-grade dysplasia (HGD) or EAC lesion following 6–8 months of exposure to deoxycholic acid (DCA) in drinking water, as previously described [12]. Immunofluorescence co-staining of the mouse EAC tissues detected consistent results as in human EAC tissues, such as inverse expression pattern of E-cadherin and APE1 (Figure 7C) and co-expression pattern of APE1 and MMP14 (Figure 7D). The relevant H/E images are shown in Supplementary Figure 9B. To establish the statistical significance of our expression model, we conducted co-immunofluorescent staining of E-cadherin, APE1, and MMP14 on a human tissue array consisting of 5 normal esophagus (NE) samples, 20 Barrett’s esophagus (BE) samples, and 10 esophageal adenocarcinoma (EAC) samples. The results revealed a progressive decrease in E-cadherin expression along the NE-BE-EAC continuum, whereas APE1 and MMP14 expression exhibited an increase along the same progression (Supplementary Figure 9C-E). The quantification of these three proteins after normalization is represented by a heatmap (Supplementary Figure 9F). Pearson’s correlation analysis using GraphPad Prism 8 demonstrated significant inverse correlations between E-cadherin and APE1 (r = −0.509, p < 0.01), as well as between E-cadherin and MMP14 (r = −0.588, p < 0.01). Additionally, a positive correlation was observed between APE1 and MMP14 (r = 0.641, p < 0.01). These results are consistent with our in vitro studies and provided multiple lines of evidence supporting APE1/MMP14 signaling axis closely correlated with EMT in EAC tumorigenesis.

Figure 7. Co-expression of APE1 and MMP14 with an inverse expression level of E-cadherin.

Figure 7.

Open in a new tab

A&B, Representative immunofluorescent co-staining images of E-cadherin and APE1 (A), or APE1 and MMP-14 (B) of the gastroesophageal junctions on the slides of same human EAC tissue. a&b, the region of normal epithelial layer adjacent to the tumor (NAT); c&d, the region of the human EAC tumor core. C&D, Representative immunofluorescent co-staining images of E-cadherin and APE1 (C), or APE1 and MMP-14 (D) in the squamocolumnar junctions on the slides of same L2-IL1ß transgenic mouse tissue. a&b, the region of the normal epithelial layer adjacent to the tumor (NAT); c&d, the region of mouse EAC tumor core. DAPI was used for nuclear staining. White arrows indicated co-overexpression of APE1 and MMP14 or an inverse correlation between APE1/MMP14 and E-cadherin expression. (E) Relapse-free survival of TCGA-EAC cohorts by combining APE1-high with MMP14-high patients, compared with APE1-low and MMP14-low patients. (F) Relapse-free survival of TCGA-EAC cohorts by combining APE1-high with EMT-score-high patients, compared with APE1-low and EMT-score-low patients.

Prognostic significance of combined APE1 expression and EMT signature in EAC patients.

To determine the prognostic outcome of APE1 and EMT signature in EAC patients, we performed survival analysis using TCGA-EAC cohorts (n=80). High expression of APE1 or MMP14, or with high pan-cancer EMT score [42] didn’t predict low overall survival (OS) (data not shown). On the other hand, CDH1-low, APE1-high, MMP14-high, or EMT-score-high patients had worse relapse-free survival (RFS) compared to their counterparts, respectively (Supplementary Figure 10A-D). Furthermore, the RFS analysis of the combination of APE1 with MMP14 (Figure 7E, FDR=3%) or APE1 with EMT-score (Figure 7F, FDR=5%) were prognostic for EAC patients at high risk of relapse.

APE1-redox-specific inhibition represses PDX tumor growth and invasion.

To validate our findings in vivo, we utilized an APE1-redox-specific inhibitor, E3330, in a patient-derived xenograft (PDX) mouse model derived from gastro-esophageal junction adenocarcinoma. Treatment with E3330 for 28 days effectively suppressed the growth of PDX tumors (Figures 8A&B) and improved the survival of the mice (Figure 8C). Notably, E3330 also inhibited the invasion of cancer cells into the stromal tissue (Figure 8D). The co-immunofluorescent staining of E-cadherin, APE1, and MMP14 demonstrated that E3330 had minimal impact on APE1 expression, as expected, but significantly inhibited MMP14 and restored E-cadherin protein levels in tumor cells (Figure 8E-G). These results confirm our in vitro findings and demonstrate potential clinical implications.

Figure 8. E3330 treatment represses PDX tumor growth and invasion.

Figure 8.

Open in a new tab

A, Plots of average tumor volume of PDX tumors with or without 20 mg/kg E3330 treatment for 28 days (Control: 7 mice; E3330: 6 mice). The tumor volumes were measured twice per week. B, the box whisker plots of the tumor volumes on day 28 were shown. C, Kaplan–Meier survival curve for the mice with PDX following treatment endpoint (Control: 7mice; E3330: 6mice). D, H/E staining images of the PDX tumors with or without E3330 treatment. The yellow dashed lines indicated the tumor/stroma boundaries. Red arrows indicated the invaded tumor cells into stroma. E, Immunofluorescence staining of E-cadherin(white)/APE1(Red)/MMP14(Green)/DAPI in the PDX tumors with or without E3330 treatment. F&G, the quantification of E-cadherin (F) and MMP14 (G) were shown by relative fluorescence intensity per cancer cells by ImageJ for three tumors in each group. The fluorescence intensity mean values of E-cadherin and MMP14 in control tumors were normalized as 100%, respectively.

The role of ABS in mediating E-cadherin cleavage and promoting EMT via the induction of APE1-redox-MMP14 signaling cascade is depicted in the Graphical Abstract.

Discussion

The five-year survival rate of EAC patients is below 20% which is significantly reduced to approximately 5% in patients with stage III/IV [43]. The main risk factor for EAC is the development of a metaplastic lesion known as BE in response to chronic reflux. The decrease in E-cadherin level and activation of ß-catenin signaling were reported in EAC carcinogenesis [18, 19, 20]. However, mechanisms by which exposure to ABS in reflux conditions contribute to EMT are not fully understood. This study reports the role of ABS exposure in mediating E-cadherin protein cleavage with subsequent activation of ß-catenin and the EMT process. Mechanistically, this process was activated via an APE1-redox-MMP14 signaling axis.

The loss of functional E-cadherin can result from various molecular factors, including relatively rare inactivating mutations (2/57) [44] or promoter methylation (1/22) [45] of CDH1 in EAC. However, due to their infrequent incidence, these alternations do not explain the frequently observed loss of functional E-cadherin. We found that ABS exposure promoted E-cadherin cleavage in EAC cells. Consistent with our studies, the increase of soluble E-cadherin fragments has been detected in the serum of GERD patients [46]. Cleavage of E-cadherin releases sequestered ß-catenin from the adhesion complex leading to accumulation, nuclear translocation, and activation of ß-catenin/TCF transcription program, a major oncogenic molecular signaling pathway in gastrointestinal cancers. This non-canonical WNT-independent activation of ß-catenin [37, 38] is due to ABS and oxidative environment in GERD. Our findings can also explain the frequent enrichment of ß-catenin/TCF transcription signature in EAC in the absence of common mutations of APC or ß-catenin mutations [47]. Of note, we did not detect significant changes in CDH1 mRNA expression in response to ABS. Similarly, analysis of public datasets ruled out the presence of frequent or significant changes in CDH1 expression in EAC. Together, our findings suggest a post-transcription regulation of CDH1 under reflux conditions, possibly through E-cadherin protein cleavage, rather than EMT-TFs-mediated CDH1 mRNA regulation.

We investigated mRNA expression of potential E-cadherin proteases to investigate the mechanism(s) by which ABS induce E-cadherin cleavage. We demonstrate evidence supporting the role of MMP-14 and its proteolytic activity in mediating ABS-induced E-cadherin cleavage in EAC. Our findings support the role of MMP14 in executing E-cadherin cleavage directly or indirectly through its downstream MMP2-mediated proteolysis [40]. This is consistent with an earlier report showing the accumulation of MMP14 on the cell surface of EAC via a redox-sensitive ARF6-mediated endocytosis/recycling [15]. To mimic the clinical condition of chronic reflux in GERD patients, EAC cells were repeatedly exposed to ABS followed by recovery daily for 15 days (rABS). These repeated exposures to ABS induced more significant EMT traits than a single transient exposure. We detected upregulation of EMT-TFs and enrichment of Vimentin-positive mesenchymal-like EAC cells, the consequence of cumulative effects of E-cadherin cleavage and constitutive activation of ß-catenin. Notably, we found that neither neutral bile salts alone (pH 7.0) nor acidic pH4.0 medium alone promoted E-cadherin cleavage to the same extent as ABS. Given that the physicochemical stability of E-cadherin ectodomain depends on pH-sensitive Calcium-binding [48], our findings suggest that reflux conditions containing bile salts and acid synergistically promote E-cadherin cleavage by increasing E-cadherin protease MMP14 expression and destabilizing homophilic binding of E-cadherin molecules.

We and others have shown that exposure to ABS significantly increased reactive oxygen species (ROS) and oxidative stress [49, 50, 51, 52]. During EAC carcinogenesis, tumorigenic esophageal cells develop inherent adaptive properties against ABS reflux-induced oxidative injury. APE1, as a dual-function protein, plays two crucial roles in this unique harsh oxidative environment by protecting cells from ABS-induced DNA damage through its base excision repair (BER) function [13] and tuning ROS balance via its redox function [53]. Although a few studies reported APE1 could regulate EMT in cancers [54, 55, 56], the mechanistic role of APE1 in EMT remained largely unknown. We found that genetic knockdown or redox inhibition of APE1 revoked ABS-mediated E-cadherin cleavage and β-catenin activation. The results demonstrated that ABS activates the APE1/MMP14 axis in a ROS-dependent manner, further supporting the role of APE1 redox function in the process. Functionally, APE1 redox activity mediated the migration and invasion capabilities of EAC cells. Applying the APE1 redox-specific inhibitor, E3330, in a PDX mouse model restrained the invasive capability of cancer cells by suppressing MMP14 and restoring E-cadherin expression. These multiple lines of evidence support the implication of the APE1-redox-MMP14 axis in reflux-induced E-cadherin cleavage and the EMT process. Nevertheless, we can’t rule out other possible APE1-mediated functions in the process. It is worth mentioning that several MMP and ADAM family members can also play a role in E-cadherin cleavage [39, 57, 58, 59, 60, 61].

In conclusion, our findings indicate that ABS in reflux conditions contribute to the EMT process by promoting E-cadherin cleavage in an APE1 redox-dependent activation of MMP14. The combination of APE1high expression and EMT high scores correlated with the worst relapse-free survival in EAC patients, suggesting a prognostic tool for the prediction of EAC relapse. APE1-redox inhibitors may have clinical implications as novel therapeutic strategies targeting EMT in EAC patients.

Supplementary Material

Supp1
Supp2

What is already known on this topic

  • Gastroesophageal reflux disease (GERD) is the leading risk factor for Barrett’s esophagus (BE) and esophageal adenocarcinoma (EAC). The steady increase in the incidence and mortality of EAC calls for a better understanding of the biology and molecular mechanisms of EAC carcinogenesis.

  • Reduction of E-cadherin level occurs in Barrett’s metaplasia-dysplasia-adenocarcinoma progression cascade. Epithelial-Mesenchymal Transition (EMT) signaling contributes to EAC’s tumor progression, metastasis, and chemoresistance.

What this study adds

  • Exposure to acidic bile salts components of reflux induces E-cadherin cleavage.

  • Repeated exposures to acidic bile salts in vitro, the mimic of chronic GERD, promotes EMT process in epithelial-type EAC cells through E-cadherin cleavage and β-catenin activation.

  • Genetic knockdown or redox-specific inhibition of APE1 protects E-cadherin from acidic bile salts-induced cleavage and restrains EMT process in EAC.

  • APE1-redox/MMP14 signaling axis links gastroesophageal reflux with E-cadherin cleavage and EMT, providing a possible mechanism underlying the EMT in EAC.

How this study might affect research, practice, or policy

This study reveals the crucial role of APE1-redox-MMP14 axis in E-cadherin cleavage, Epithelial-to-Mesenchymal Transition, cell migration, and invasion. These findings call for developing APE1-redox-specific inhibitors as a therapeutic strategy in EAC patients.

Acknowledgments

We want to thank the NCI-supported (P30CA240139) Sylvester Comprehensive Cancer Center shared resources. This study utilized oncogenomic, flow cytometry, biospecimens, and bioinformatics shared resources.

Funding

This study was funded by the National Institutes of Health (Grant number: P01CA268991, R01CA206563, and R01CA224366) and the U.S. Department of Veterans Affairs (1IK6BX003787 and I01BX001179). This work’s content is solely the responsibility of the authors. It does not necessarily represent the official views of the National Institutes of Health or the University of Miami.

Abbreviations

EAC

Esophageal adenocarcinomas

EMT

Epithelial-to-Mesenchymal Transition

TFs

Transcription factors

GERD

Chronic gastroesophageal reflux disease

BE

Barrett’s esophagus

HGD

High-grade dysplasia

APE1

Apurinic/apyrimidinic endonuclease

MMP

Matrix metalloproteinases

ADAM

A disintegrin and metalloproteinase

IL1β

Interleukin 1 beta

ABS

Acidic bile salts

rABS

repeated exposure to acidic bile salts

TCGA

The Cancer Genome Atlas

GEO

Gene Expression Omnibus

DCA

Deoxycholic acid

FBS

Fetal bovine serum

WB

Western blot

WCL

Whole cell lysates

C.M.

Conditioned medium

SNAI

Snail family transcriptional repressors

TWIST

Twist family bHLH transcription factors

ZEB

Zinc finger E-box binding homeobox

IF

Immunofluorescence

GSEA

Gene Set Enrichment Analysis

3D

Three-dimensional

OTC

Organotypic culture

DAPI

4′, 6-diamidino-2-phenylindole

H/E

Hematoxylin and eosin

OS

overall survival

RFS

relapse-free survival

FDR

false discovery rate

Footnotes

Patient consent for publication

Not applicable.

Ethics approval

The study design regarding de-identified human data or tissues was approved by Institutional Research Ethics Committee. All animal studies were carried out following the protocols approved by the Institutional Animal Care and Use Committee of the University of Miami (UM-20–110).

Provenance and peer review

Not commissioned, externally peer reviewed.

Competing interests

The authors declare no conflict of interest.

Data availability statement

The data of RNA expression profiles and clinical information of EACs have been downloaded from The Cancer Genome Atlas (TCGA) official website (https://portal.gdc.cancer.gov/repository). The Gene Expression Omnibus (GEO) datasets were retrieved from National Center for Biotechnology Information (NCBI) GEO database (https://www.ncbi.nlm.nih.gov/). RNA-seq data in EAC cell lines are available from a previous publication[12] and from the authors upon reasonable request and with permission of the University of Miami.

REFERENCES

  • 1.Ronkainen J, Aro P, Storskrubb T, Johansson SE, Lind T, Bolling-Sternevald E, et al. Prevalence of Barrett’s esophagus in the general population: an endoscopic study. Gastroenterology 2005;129:1825–31. [DOI] [PubMed] [Google Scholar]
  • 2.Peters Y, Al-Kaabi A, Shaheen NJ, Chak A, Blum A, Souza RF, et al. Barrett oesophagus. Nat Rev Dis Primers 2019;5:35. [DOI] [PubMed] [Google Scholar]
  • 3.Lagergren J. Adenocarcinoma of oesophagus: what exactly is the size of the problem and who is at risk? Gut 2005;54 Suppl 1:i1–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Reid BJ, Li X, Galipeau PC, Vaughan TL. Barrett’s oesophagus and oesophageal adenocarcinoma: time for a new synthesis. Nat Rev Cancer 2010;10:87–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Coleman HG, Xie SH, Lagergren J. The Epidemiology of Esophageal Adenocarcinoma. Gastroenterology 2018;154:390–405. [DOI] [PubMed] [Google Scholar]
  • 6.Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin 2020;70:7–30. [DOI] [PubMed] [Google Scholar]
  • 7.Jung KW, Talley NJ, Romero Y, Katzka DA, Schleck CD, Zinsmeister AR, et al. Epidemiology and natural history of intestinal metaplasia of the gastroesophageal junction and Barrett’s esophagus: a population-based study. Am J Gastroenterol 2011;106:1447–55; quiz 56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Edgren G, Adami H-O, Weiderpass E, Nyrén O. A global assessment of the oesophageal adenocarcinoma epidemic. Gut 2013;62:1406–14. [DOI] [PubMed] [Google Scholar]
  • 9.Byrne AM, Bekiaris S, Duggan G, Prichard D, Kirca M, Finn S, et al. Golgi phosphoprotein 2 (GOLPH2) is a novel bile acid-responsive modulator of oesophageal cell migration and invasion. Br J Cancer 2015;113:1332–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Liu R, Li X, Hylemon PB, Zhou H. Conjugated Bile Acids Promote Invasive Growth of Esophageal Adenocarcinoma Cells and Cancer Stem Cell Expansion via Sphingosine 1-Phosphate Receptor 2-Mediated Yes-Associated Protein Activation. Am J Pathol 2018;188:2042–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Quilty F, Byrne AM, Aird J, El Mashad S, Parra-Blanco A, Long A, et al. Impact of Deoxycholic Acid on Oesophageal Adenocarcinoma Invasion: Effect on Matrix Metalloproteinases. Int J Mol Sci 2020;21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Chen L, Lu H, Peng D, Cao LL, Ballout F, Srirmajayam K, et al. Activation of NOTCH signaling via DLL1 is mediated by APE1-redox-dependent NF-kappaB activation in oesophageal adenocarcinoma. Gut 2022. [DOI] [PMC free article] [PubMed]
  • 13.Hong J, Chen Z, Peng D, Zaika A, Revetta F, Washington MK, et al. APE1-mediated DNA damage repair provides survival advantage for esophageal adenocarcinoma cells in response to acidic bile salts. Oncotarget 2016;7:16688–702. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Bhat AA, Lu H, Soutto M, Capobianco A, Rai P, Zaika A, et al. Exposure of Barrett’s and esophageal adenocarcinoma cells to bile acids activates EGFR-STAT3 signaling axis via induction of APE1. Oncogene 2018;37:6011–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Lu H, Bhat AA, Peng D, Chen Z, Zhu S, Hong J, et al. APE1 Upregulates MMP-14 via Redox-Sensitive ARF6-Mediated Recycling to Promote Cell Invasion of Esophageal Adenocarcinoma. Cancer Res 2019;79:4426–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Brabletz T, Kalluri R, Nieto MA, Weinberg RA. EMT in cancer. Nat Rev Cancer 2018;18:128–34. [DOI] [PubMed] [Google Scholar]
  • 17.Nieto MA, Huang RY, Jackson RA, Thiery JP. Emt: 2016. Cell 2016;166:21–45. [DOI] [PubMed] [Google Scholar]
  • 18.Swami S, Kumble S, Triadafilopoulos G. E-cadherin expression in gastroesophageal reflux disease, Barrett’s esophagus, and esophageal adenocarcinoma: an immunohistochemical and immunoblot study. Am J Gastroenterol 1995;90:1808–13. [PubMed] [Google Scholar]
  • 19.Bailey T, Biddlestone L, Shepherd N, Barr H, Warner P, Jankowski J. Altered cadherin and catenin complexes in the Barrett’s esophagus-dysplasia-adenocarcinoma sequence: correlation with disease progression and dedifferentiation. Am J Pathol 1998;152:135–44. [PMC free article] [PubMed] [Google Scholar]
  • 20.Washington K, Chiappori A, Hamilton K, Shyr Y, Blanke C, Johnson D, et al. Expression of beta-catenin, alpha-catenin, and E-cadherin in Barrett’s esophagus and esophageal adenocarcinomas. Mod Pathol 1998;11:805–13. [PubMed] [Google Scholar]
  • 21.Zhang Q, Agoston AT, Pham TH, Zhang W, Zhang X, Huo X, et al. Acidic Bile Salts Induce Epithelial to Mesenchymal Transition via VEGF Signaling in Non-Neoplastic Barrett’s Cells. Gastroenterology 2019;156:130–44 e10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Wang Q, Ma C, Kemmner W. Wdr66 is a novel marker for risk stratification and involved in epithelial-mesenchymal transition of esophageal squamous cell carcinoma. BMC Cancer 2013;13:137. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Myers AL, Lin L, Nancarrow DJ, Wang Z, Ferrer-Torres D, Thomas DG, et al. IGFBP2 modulates the chemoresistant phenotype in esophageal adenocarcinoma. Oncotarget 2015;6:25897–916. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Kim SM, Park YY, Park ES, Cho JY, Izzo JG, Zhang D, et al. Prognostic biomarkers for esophageal adenocarcinoma identified by analysis of tumor transcriptome. PLoS One 2010;5:e15074. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Song D, Tian J, Hu Y, Wei Y, Lu H, Wang Y, et al. Identification of biomarkers associated with diagnosis and prognosis of gastroesophageal junction adenocarcinoma-a study based on integrated bioinformatics analysis in GEO and TCGA database. Medicine (Baltimore) 2020;99:e23605. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Peng D, Guo Y, Chen H, Zhao S, Washington K, Hu T, et al. Integrated molecular analysis reveals complex interactions between genomic and epigenomic alterations in esophageal adenocarcinomas. Sci Rep 2017;7:40729. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Silvers AL, Lin L, Bass AJ, Chen G, Wang Z, Thomas DG, et al. Decreased selenium-binding protein 1 in esophageal adenocarcinoma results from posttranscriptional and epigenetic regulation and affects chemosensitivity. Clin Cancer Res 2010;16:2009–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Quante M, Bhagat G, Abrams JA, Marache F, Good P, Lee MD, et al. Bile acid and inflammation activate gastric cardia stem cells in a mouse model of Barrett-like metaplasia. Cancer cell 2012;21:36–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Corso S, Isella C, Bellomo SE, Apicella M, Durando S, Migliore C, et al. A Comprehensive PDX Gastric Cancer Collection Captures Cancer Cell-Intrinsic Transcriptional MSI Traits. Cancer Res 2019;79:5884–96. [DOI] [PubMed] [Google Scholar]
  • 30.Maacha S, Hong J, von Lersner A, Zijlstra A, Belkhiri A. AXL Mediates Esophageal Adenocarcinoma Cell Invasion through Regulation of Extracellular Acidification and Lysosome Trafficking. Neoplasia 2018;20:1008–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Lu H, Gomaa A, Wang-Bishop L, Ballout F, Hu T, McDonald O, et al. Unfolded Protein Response Is Activated by Aurora Kinase A in Esophageal Adenocarcinoma. Cancers (Basel) 2022;14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Kelley MR, Luo M, Reed A, Su D, Delaplane S, Borch RF, et al. Functional analysis of novel analogues of E3330 that block the redox signaling activity of the multifunctional AP endonuclease/redox signaling enzyme APE1/Ref-1. Antioxid Redox Signal 2011;14:1387–401. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Ding J, Fishel ML, Reed AM, McAdams E, Czader MB, Cardoso AA, et al. Ref-1/APE1 as a Transcriptional Regulator and Novel Therapeutic Target in Pediatric T-cell Leukemia. Mol Cancer Ther 2017;16:1401–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Fehrenbacher JC, Guo C, Kelley MR, Vasko MR. DNA damage mediates changes in neuronal sensitivity induced by the inflammatory mediators, MCP-1 and LPS, and can be reversed by enhancing the DNA repair function of APE1. Neuroscience 2017;366:23–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Rai G, Vyjayanti VN, Dorjsuren D, Simeonov A, Jadhav A, Wilson DM 3rd, et al. Synthesis, biological evaluation, and structure-activity relationships of a novel class of apurinic/apyrimidinic endonuclease 1 inhibitors. J Med Chem 2012;55:3101–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Maher RL, Bloom LB. Pre-steady-state kinetic characterization of the AP endonuclease activity of human AP endonuclease 1. J Biol Chem 2007;282:30577–85. [DOI] [PubMed] [Google Scholar]
  • 37.Maretzky T, Reiss K, Ludwig A, Buchholz J, Scholz F, Proksch E, et al. ADAM10 mediates E-cadherin shedding and regulates epithelial cell-cell adhesion, migration, and beta-catenin translocation. Proc Natl Acad Sci U S A 2005;102:9182–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Valenta T, Hausmann G, Basler K. The many faces and functions of beta-catenin. EMBO J 2012;31:2714–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Covington MD, Burghardt RC, Parrish AR. Ischemia-induced cleavage of cadherins in NRK cells requires MT1-MMP (MMP-14). Am J Physiol Renal Physiol 2006;290:F43–51. [DOI] [PubMed] [Google Scholar]
  • 40.Biswas MH, Du C, Zhang C, Straubhaar J, Languino LR, Balaji KC. Protein kinase D1 inhibits cell proliferation through matrix metalloproteinase-2 and matrix metalloproteinase-9 secretion in prostate cancer. Cancer Res 2010;70:2095–104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.van Zanten SV. Chronic GERD and risk of esophageal adenocarcinoma: Should we screen with gastroscopy? CMAJ 2020;192:E781–E2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Tan TZ, Miow QH, Miki Y, Noda T, Mori S, Huang RY, et al. Epithelial-mesenchymal transition spectrum quantification and its efficacy in deciphering survival and drug responses of cancer patients. EMBO Mol Med 2014;6:1279–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Lan T, Liu W, Lu Y, Luo H. A five-gene signature for predicting overall survival of esophagus adenocarcinoma. Medicine (Baltimore) 2021;100:e25305. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Wijnhoven BP, de Both NJ, van Dekken H, Tilanus HW, Dinjens WN. E-cadherin gene mutations are rare in adenocarcinomas of the oesophagus. Br J Cancer 1999;80:1652–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Schildhaus HU, Krockel I, Lippert H, Malfertheiner P, Roessner A, Schneider-Stock R. Promoter hypermethylation of p16INK4a, E-cadherin, O6-MGMT, DAPK and FHIT in adenocarcinomas of the esophagus, esophagogastric junction and proximal stomach. Int J Oncol 2005;26:1493–500. [PubMed] [Google Scholar]
  • 46.Jovov B, Que J, Tobey NA, Djukic Z, Hogan BL, Orlando RC. Role of E-cadherin in the pathogenesis of gastroesophageal reflux disease. Am J Gastroenterol 2011;106:1039–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Choi YW, Heath EI, Heitmiller R, Forastiere AA, Wu TT. Mutations in beta-catenin and APC genes are uncommon in esophageal and esophagogastric junction adenocarcinomas. Mod Pathol 2000;13:1055–9. [DOI] [PubMed] [Google Scholar]
  • 48.Zheng K, Middaugh CR, Siahaan TJ. Evaluation of the physical stability of the EC5 domain of E-cadherin: effects of pH, temperature, ionic strength, and disulfide bonds. J Pharm Sci 2009;98:63–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Jenkins GJ, Cronin J, Alhamdani A, Rawat N, D’Souza F, Thomas T, et al. The bile acid deoxycholic acid has a non-linear dose response for DNA damage and possibly NF-kappaB activation in oesophageal cells, with a mechanism of action involving ROS. Mutagenesis 2008;23:399–405. [DOI] [PubMed] [Google Scholar]
  • 50.Dvorak K, Payne CM, Chavarria M, Ramsey L, Dvorakova B, Bernstein H, et al. Bile acids in combination with low pH induce oxidative stress and oxidative DNA damage: relevance to the pathogenesis of Barrett’s oesophagus. Gut 2007;56:763–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Peng D, Belkhiri A, Hu T, Chaturvedi R, Asim M, Wilson KT, et al. Glutathione peroxidase 7 protects against oxidative DNA damage in oesophageal cells. Gut 2012;61:1250–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Souza RF. Reflux esophagitis and its role in the pathogenesis of Barrett’s metaplasia. J Gastroenterol 2017;52:767–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Sriramajayam K, Peng D, Lu H, Zhou S, Bhat N, McDonald OG, et al. Activation of NRF2 by APE1/REF1 is redox-dependent in Barrett’s related esophageal adenocarcinoma cells. Redox Biol 2021;43:101970. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Wei X, Li Q, Li Y, Duan W, Huang C, Zheng X, et al. Prediction of survival prognosis of non-small cell lung cancer by APE1 through regulation of Epithelial-Mesenchymal Transition. Oncotarget 2016;7:28523–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Yang X, Peng Y, Jiang X, Lu X, Duan W, Zhang S, et al. The regulatory role of APE1 in epithelial-to-mesenchymal transition and in determining EGFR-TKI responsiveness in non-small-cell lung cancer. Cancer Med 2018;7:4406–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Li Q, Zhou ZW, Duan W, Qian CY, Wang SN, Deng MS, et al. Inhibiting the redox function of APE1 suppresses cervical cancer metastasis via disengagement of ZEB1 from E-cadherin in EMT. J Exp Clin Cancer Res 2021;40:220. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Zheng G, Lyons JG, Tan TK, Wang Y, Hsu TT, Min D, et al. Disruption of E-cadherin by matrix metalloproteinase directly mediates epithelial-mesenchymal transition downstream of transforming growth factor-beta1 in renal tubular epithelial cells. Am J Pathol 2009;175:580–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Lynch CC, Vargo-Gogola T, Matrisian LM, Fingleton B. Cleavage of E-Cadherin by Matrix Metalloproteinase-7 Promotes Cellular Proliferation in Nontransformed Cell Lines via Activation of RhoA. J Oncol 2010;2010:530745. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Symowicz J, Adley BP, Gleason KJ, Johnson JJ, Ghosh S, Fishman DA, et al. Engagement of collagen-binding integrins promotes matrix metalloproteinase-9-dependent E-cadherin ectodomain shedding in ovarian carcinoma cells. Cancer Res 2007;67:2030–9. [DOI] [PubMed] [Google Scholar]
  • 60.Solanas G, Cortina C, Sevillano M, Batlle E. Cleavage of E-cadherin by ADAM10 mediates epithelial cell sorting downstream of EphB signalling. Nat Cell Biol 2011;13:1100–7. [DOI] [PubMed] [Google Scholar]
  • 61.Cai Z, Zhang A, Choksi S, Li W, Li T, Zhang XM, et al. Activation of cell-surface proteases promotes necroptosis, inflammation and cell migration. Cell Res 2016;26:886–900. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supp1
NIHMS1933180-supplement-Supp1.pdf (9.5MB, pdf)
Supp2
NIHMS1933180-supplement-Supp2.pdf (137.4KB, pdf)

Data Availability Statement

The data of RNA expression profiles and clinical information of EACs have been downloaded from The Cancer Genome Atlas (TCGA) official website (https://portal.gdc.cancer.gov/repository). The Gene Expression Omnibus (GEO) datasets were retrieved from National Center for Biotechnology Information (NCBI) GEO database (https://www.ncbi.nlm.nih.gov/). RNA-seq data in EAC cell lines are available from a previous publication[12] and from the authors upon reasonable request and with permission of the University of Miami.

RESOURCES