The key resources for reagents, kits and primer sequences are provided in Table SI, Table SII, Table SIII.

Immortalized non-neoplastic esophageal epithelial cells (EPC2; provided by Dr. Anil Rustgi, Columbia University, NY, USA) and BE cells line BART (Dr. Rhonda Souza, Baylor University Medical Center, Dallas, TX, USA), CPA and CPB [both American Type Culture Collection (ATCC)] were cultured in DMEM/F12 supplemented with 5% FBS (both Gibco; Thermo Fisher Scientific, Inc.), 0.4 µg/ml hydrocortisone (Sigma-Aldrich; Merck KGaA), 1% penicillin/streptomycin (Gibco; Thermo Fisher Scientific, Inc.), 140 µg/ml bovine pituitary extract (Thermo Fisher Scientific, Inc.), 20 mg/l adenine hydrochloride hydrate, insulin-transferrin-selenium supplement and 20 ng/ml epidermal growth factor (all Sigma-Aldrich; Merck KGaA). EAC cell lines FLO-1 (ATCC), OE33 (Dr David Beer, University of Michigan, Ann Arbor, MI, USA), SK-GT4 (Dr. Xiaochun Xu, MD Anderson Cancer Center, Houston, TX, USA) and OE19 (MilliporeSigma) were maintained in DMEM or RPMI-1640 with 10% FBS and 1% penicillin/streptomycin. All cells were grown at 37°C in a humidified incubator with 5% carbon dioxide and tested for mycoplasma contamination every month using a mycoplasma detection kit (PCR; SouthernBiotech). Cell line authentication was conducted routinely every 6 months using Labcorp Cell Line Authentication Services, which utilizes the short tandem repeat DNA profiling technology. Cells were last authenticated in August 2025, prior to the experiments.

To mimic gastroesophageal reflux episodes in patients, a bile salts cocktail was prepared by combining 40 µM glycocholic acid, taurocholic acid, glycodeoxycholic acid, glycochenodeoxycholic acid and deoxycholic acid. Cells were treated with 200 µM bile BS cocktail in acidic culture medium (DMEM or RPMI-1640; pH 4.0) for 20 min at 37°C based on our previous publications (12,13). The BS-containing medium was replaced with standard culture medium (DMEM or RPMI-1640) for 0–6 h.

Scrambled small interfering (si)RNA (cat. no. sc-37007) was purchased from Santa Cruz Biotechnology, Inc. and DNA methyltransferase (DNMT)1 siRNA (ID: s4215; cat. no. #4390824) was obtained from Thermo Fisher Scientific, Inc. (sequences not available). The wild-type and mutated mammalian expression luciferase reporters containing the 3′UTR region of APE1 were generated by Vector Builder, Inc. The miR-128-3p mimic or miRNA mimic negative control was transfected into OE33 and FLO-1 cells and the miR-128-3p inhibitor or miRNA inhibitor negative control (all 50 nM) was transfected into CPA cells using PolyJet. For transient DNMT1 knockdown, siDNMT1 or scrambled siRNA (40 nM for OE33 and 80 nM for FLO-1) was transfected into OE33 and FLO-1 cells using LipoJet. Transfection was performed at 37°C. Cells were collected within 72 h post-transfection.

FLO-1 and OE33 cells were treated with 2 µM 5′aza (Sigma-Aldrich; Merck KGaA) for 72 h at 37°C in complete culture medium. 5′aza was replenished every 24 h.

Genomic DNA was isolated from OE33 cells using a DNeasy Blood Tissue Kit (Qiagen GmbH) and bisulfite modified with an EZ DNA Methylation-Gold kit. The CpG island within the miR-128-3p promoter was identified using an online search tool (uscnorris.com). Pyrosequencing primers were designed with PSQ Assay Design Software version 3.0 (Biotage), including the forward primer 5′ GAAAGGTTTAGGGAATGA 3′ and the biotinylated reverse primer 5′-Biotin-TAAAAACATCCAAAACTAACCCATCC-3′. Bisulfite-modified DNA (40 ng) was amplified by PCR (Bio-Rad Laboratories) using Platinum PCR SuperMix High Fidelity (Invitrogen; Thermo Fisher Scientific, Inc.). Thermocycling conditions:

Initial denaturation: 95°C for 5 min, followed by 50 cycles of 95°C for 30 sec, 55°C for 30 sec and 72°C for 30 sec and final extension: 72°C for 5 min. PCR products were confirmed by 1.5% agarose gel electrophoresis and visualized using GelRed Nucleic Acid Gel Stain (MilliporeSigma). Quantitative methylation analysis was performed using the Biotage PyroMark MD System and the results were processed with Pyro Q-CpG 1.0.9 software (Biotage). A methylation threshold of 10% was used to define miR-128-3p promoter hypermethylation based on control normal samples and internal quality controls provided in the software analysis (14).

Single-cell suspensions of OE33 cells transfected with miR-128-3p mimic or miRNA mimic negative control (both 50 nM) were resuspended in a 1:1 mixture of cold Growth Factor Reduced Matrigel™ (Corning, Inc.) and serum-free RPMI-1640 (Thermo Fisher Scientific, Inc.) to a final volume of 50 µl/well. Each condition was performed in triplicate. The cell-Matrigel mixture was plated along the rim of wells in a 24-well plate at 37°C for 45 min to allow solidification. Subsequently, RPMI-1640 supplemented with 5% FBS was added to the center of each well. Media was replenished every 2 days. Sphere formation was assessed by manual counting and imaging on day 10. The total number of spheres was counted manually under bright-field light microscopy (20X objective).

Esophageal tumoroids were generated from patient-derived xenograft (PDX) sample PDX683-24-4. The PDX was developed at the University of Miami in collaboration with the Biospecimen shared resource of the Sylvester Comprehensive Cancer Center (Miami, FL, USA). Tissue samples were dissected into 2–3 mm fragments and incubated in 2 ml 5 mg/ml collagenase type II (Gibco; Thermo Fisher Scientific, Inc.; cat. no. 17101-015) in Advanced DMEM-F12 (adDMEM/F12) for 45 min at 37°C. The pellet was digested with TrypLE followed by resuspension in adDMEM/F12. The resulting suspension was passed through 40 µm cell strainers and centrifuged at 274 × g for 5 min at room temperature. The cell pellet was resuspended in 50 µl Matrigel and plated into pre-warmed 24-well plates. Following Matrigel polymerization in a CO₂ incubator at 37°C for 20 min, 500 µl IntestiCult Organoid Growth Medium was added. After 7–10 days of culture at 37°C, the total number of tumoroids was counted manually under bright-field light microscopy (20X magnification) and images were captured to calculate the mean diameter using the Perfect Screen Ruler program version 3.0 (perfect-screen-ruler.informer.com/download/).

For organoid transfection, organoids were dissociated into single cells using pre-warmed TrypLE Express (cat. no. 12604021, Thermo Fisher Scientific, Inc.) at 37°C for 10 min. The reaction was terminated by adding adDMEM/F12 and cells passed through a 30-µm cell strainer. The miR-128-3p mimic or the miRNA mimic negative control (50 nM) was mixed with PolyJet. A total of 50 µl reaction mixture was applied to 2×10⁵ organoids suspended with 450 µl human organoid growth medium and incubated at 37°C overnight. Following overnight incubation, cells were pelleted by centrifugation at 274 × g for 5 min at room temperature and seeded in Matrigel in a 24-well plate as aforementioned. Organoid growth was monitored daily for 3–5 days following transfection.

FLO-1 and OE33 cells were lysed using RIPA buffer. Protein concentrations were determined with the DC Protein Assay (Bio-Rad Laboratories, Inc.) following the manufacturer's instructions. Samples were combined with 4X Laemmli buffer and heated at 85°C for 10 min prior to electrophoresis. Equal amounts of protein (30 µg) were resolved on 10% SDS-PAGE at 100 V for 90 min and subsequently transferred onto nitrocellulose membranes. Membranes were blocked in 5% BSA (RPI Research Products; cat. no. A30075-250.0) for 1 h at room temperature and incubated overnight at 4°C with primary antibodies. Following three washes with 1X TBST, membranes were incubated with secondary antibodies for 1 h at room temperature. Protein signals were detected using Immobilon Western Chemiluminescent HRP substrate (Thermo Fisher Scientific, Inc.). Images were captured and analyzed using a ChemiDoc™ Imaging System (Bio-Rad Laboratories, Inc.). Image Lab version 6.0.1 was used for analysis (Bio-Rad Laboratories, Inc.).

The tissue samples utilized for RNA extraction were frozen de-identified human tissue samples that were collected from the archives of pathology at Vanderbilt University (Nashville, TN, USA) and the National Cancer Institute (Bethesda, Maryland, USA) Cooperative Human Tissue Network (11). The samples included 13 normal esophagus (NE), 7 isolated BE and 15 EAC tissue samples. The NE samples were matched whenever possible with EAC or isolated BE. Histopathological diagnosis was verified based on hematoxylin and eosin-stained sections. The tissue were obtained between January, 2010-June, 2017 from a total of 28 patients, aged 44–77 years. Tissue samples meeting the histopathological criteria of NE, BE, or EAC were randomly collected with no exclusion criteria

The total RNA was extracted using a miRNeasy kit according to the manufacturer's protocol. cDNA was synthesized from 1 µg total RNA using miRNA All-In-One cDNA Synthesis kit, according to the manufacturer's instructions. RT-qPCR was performed using the GoTaq 2-Step RT-qPCR System (Promega Corporation; cat. no. A6020) on a CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.). Thermocycling conditions were as follows:

Initial denaturation: 95°C for 3 min, followed by 40 cycles of 95°C for 5 sec, 60°C for 30 sec 95°C for 10 sec. Normalization was performed using U6 small nuclear RNA (snRNA) or hypoxanthine-guanine phosphoribosyltransferase (HPRT) as a reference gene. Relative gene expression levels were calculated using the 2^−ΔΔCq method (15).

Dual-luciferase reporter assays were performed following the manufacturer's protocol (Promega Corporation; cat. no. E397A). FLO-1 and OE33 cells were seeded in 6-well tissue culture plates overnight at 37°C; upon reaching 60–70% confluency, cells were transiently transfected with 1 µg luciferase reporter plasmid/well using PolyJet, as aforementioned. After 48 h, luciferase activity was measured using a LUMIstar OPTIMA microplate luminometer (BMG LabTech) in 96-well plates (Dynex). Relative luciferase units were normalized to firefly luciferase, and results were expressed as relative luciferase activity. All transfections were performed in triplicate.

OE33 cells were treated with ABS for 20 min at 37°C and allowed to recover in regular media for 3 h at 37°C. Cells were collected, and nuclear extracts were isolated using a nuclear extraction kit according to the manufacturer's instructions. DNMT activity was evaluated using a colorimetric DNMT activity assay, which measures methylation of CpG-rich DNA sequences immobilized on 96-well plates. DNMT1 catalyzes the transfer of a methyl group from S-adenosyl methionine to the substrate DNA and the resulting methylated cytosines are quantified via ELISA using a methyl-CpG-binding protein and HRP-conjugated antibody. DNMT activity was measured using a LUMIstar OPTIMA microplate luminometer.

For the migration assay, 2×10⁵ transfected OE33 and FLO-1 cells were seeded in a serum-free medium (DMEM or RPMI-1640) in the top chamber of 24-well inserts (pore size, 8 mm; Corning, Inc.) and medium (DMEM or RPMI-1640) supplemented with 10% FBS was used in the lower chamber. Cells were allowed to migrate through the membrane at 37°C in a 5% CO₂ incubator for 48 h. Non-migratory cells in the upper chamber were scraped off with a cotton-tip applicator. Migrating cells on the lower surface of the membrane were fixed and stained with crystal violet for 20 min at room temperature. Images were captured using a bright-field light BZ-X710 KEYENCE All-in-One Microscope. For the invasion assay, the top chamber was freshly coated with 100 µl Matrigel at a dilution of 1:10 in cold PBS and was then air-dried overnight before the invasion assay.

All statistical analyses were conducted using GraphPad Prism (version 8.0; Dotmatics). Statistical significance was evaluated using unpaired student's t-test or one-way ANOVA followed by Bonferroni's post hoc test. Data are reported as the mean of three independent experiments ± SEM. P<0.05 was considered to indicate a statistically significant difference.

To identify candidate miRNAs that target APE1, the present study analyzed all downregulated miRNAs in EAC and the 3′UTR of APE1 using our previously established database of deregulated miRNAs in EAC (11). miR-128-3p had a perfect match on APE1 3′UTR with 8 uninterrupted binding sites (Fig. 1A). miR-128-3p was significantly downregulated in EAC (Fig. 1B and C) with increased APE1 protein levels (Fig. 1D). The adenocarcinomas ranged from well to poorly differentiated with a mix of intestinal- and diffuse-type tumors.

ABS induced an increase in APE1 protein (Fig. 1E). There were no significant changes in APE1 mRNA expression levels (Fig. 1F). There was complete silencing of miR-128-3p (Fig. 1G) that was concomitant with the observed increase in APE1 protein levels. These findings suggested a post-transcriptional regulation of APE1. To test this hypothesis, 3′UTR regions of APE1 containing the respective binding sites were cloned into a mammalian expression luciferase reporter (Fig. 2A). Treatment of OE33 and FLO-1 cells with ABS (pH 4.0, 20 min), followed by 3 h recovery, showed significant increase in the luciferase reporter activity, which was lost when combined with expression of miR-128-3p mimic (Fig. 2B). Successful transfection with miR-128-3p mimic was confirmed by RT-qPCR, which showed increased levels of miR-128-3p in the miR-128-3p mimic transfected group compared with the mimic negative control group in OE33 and FLO-1 cells (Fig. S1A). Inhibition of miR-128-3p in CPA cells, with constitutive high expression of miR-128-3p, enhanced the luciferase reporter activity (Fig. 2C). Unlike miR-128-3p mimic, successful transfection with miR-128-3p inhibitor cannot be confirmed by RT-qPCR because the inhibitor neutralizes its target miRNA by binding and does not degrade the mature miRNA so both free and inhibitor-bound miRNA will be equally detected by RT-qPCR, failing to reflect true functional neutralization (16). Therefore, in the present study, successful transfection of the miR-128-3p inhibitor was confirmed via functional and reporter assays (Fig. 2C). To validate the functional binding of miR-128-3p on APE1 3′UTR, mutated binding sites on the APE1 3′UTR were cloned into the luciferase reporter (Fig. 2D). Unlike the wild-type APE1 3′UTR, no significant changes in luciferase activity were observed with the mutant APE1 3′UTR following ABS treatment in the presence or absence of miR-128-3p mimic expression (Fig. 2E). Notably, control untreated wild-type and mutant reporters showed no significant change in luciferase activity, confirming that the mutant construct did not affect the basal luciferase activity (Fig. S1B). Western blot analysis of APE1 in OE33 and FLO-1 cells following ABS treatment and miR-128-3p mimic expression showed a decrease in APE1 protein levels (Fig. 3A). Conversely, inhibition of miR-128-3p following ABS treatment in CPA cells induced APE1 protein levels (Fig. 3B). miR-128-3p mimic in OE33-derived spheroids (Fig. 3C) or patient-derived tumoroids (Fig. 3D) decreased their number and size and APE1 protein levels (Fig. S1C). Moreover, expressing miR-128-3p mimic in OE33 and FLO-1 cells reduced their migratory and invasive capacities (Fig. S2A and B). Taken together, these data suggested that ABS induced APE1 by silencing miR-128-3p.

Analysis of the miR-128-3p promoter demonstrated the presence of a CpG island (Fig. 4A), suggesting that miR-128-3p silencing may occur through an epigenetic mechanism. ABS decreased miR-128-3p expression (Fig. 1F) and induced DNMT1 protein levels and activity in EAC cells (Fig. 4B and C) with concomitant increases in APE1 protein. DNMT1 knockdown and ABS treatment restored the expression of miR-128-3p with downregulation of APE1 (Figs. 4D and E and S3). Similarly, inhibition of DNMT1 using 5′aza following ABS treatment rescued miR-128-3p expression and decreased APE1 protein levels (Fig. 5A and B). The 5′aza data were consistent with the DNMT1-specific siRNA results, although 5′aza inhibits all DNMT family members (17). To confirm the direct role of methylation in regulating miR-128-3p, quantitative methylation analysis of miR-128-3p was performed using pyrosequencing technology. ABS treatment was associated with increased methylation of miR-128-3p, whereas DNMT1 knockdown or inhibition in the presence or absence of ABS decreased miR-128-3p methylation (Fig. 5C). These findings suggest that ABS-induced DNMT1 could increase APE1 by downregulating miR-128-3p in a methylation-dependent manner (Fig. 5D).