Low mitochondrial DNA copy number induces chemotherapy resistance via epithelial-mesenchymal transition by DNA methylation in esophageal squamous cancer cells

Human ESCC cell lines, TE8 (RBRC-RCB2098) and TE11 (RBRC-2100), were purchased from the RIKEN BioResource Center (Tsukuba, Ibaraki, Japan). These cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (Thermo Fisher Scientific, Waltham, MA, USA), penicillin (100 IU/ml), and streptomycin (100 μg/ml) at 37 °C in a humidified incubator with 5% CO₂. Depletion of mtDNA content was induced by knockdown of mitochondrial transcription factor A (TFAM) expression [26, 27]. Silencing of TFAM reduces mtDNA levels, since this factor is important for mtDNA packaging and maintenance [13, 28]. A short hairpin RNA designed by Sigma–Aldrich (St. Louis, MO, USA), MISSION TRC-Hs1.0, was applied for knockdown of TFAM expression in TE8 and TE11 cells. The depleted mtDNA cell lines “tfam-sh1” and “tfam-sh2” were established by targeting the following sequences of the TFAM gene: 5′-CGTCGCACAATAAAGAACAA-3′ (TRCN0000016094) and 5′-GCAGATTTAAAGAACAGCTAA-3′ (TRCN0000016097), respectively. In addition, the nontargeted sequence 5′-GGCGCGATAGCGCTAATAATTT-3′ (SHC016, Sigma–Aldrich, defined as ‘control-sh’) was used as the control for comparison (control-sh). The morphology of the cells was evaluated by optical microscopy (BZ-X710, KEYENCE, Osaka, Japan) [13]. In addition, an increased mtDNA copy number was established as follows: 1 × 10⁶ cells were plated in 1.25 ml of complete medium without antibiotics in 6-well plates and transfected with 7 mg of a plasmid containing the TFAM-encoding gene (OriGene, RC215488) using 12 μl of Lipofectamine 3000 reagent (L3000015, Thermo Fisher Scientific, Tokyo, Japan) following the manufacturer’s instructions. The protein expression of TFAM was lower in tfam-sh1 and tfam-sh2 cells than in control-sh cells, and tfam-sh1 and tfam-sh2 ESCC cells had an approximately 40–60% reduction in mtDNA copy number compared with control-sh cells (Additional file 1: Fig. S1A, B). Additionally, the proliferation rate was significantly lower in tfam-sh1 and tfam-sh2 cells than in control-sh cells (Additional file 1: Fig. S1C).

From November 2006 to December 2011, 88 esophageal squamous cell cancer patients underwent R0 resection after neoadjuvant chemotherapy at Osaka University Hospital (Osaka, Japan). Neoadjuvant chemotherapy followed by surgery at our hospital was performed for patients with cStage I (excluding T1N0), II, III, or IV disease without distant organ metastasis. The neoadjuvant chemotherapy regimen consisted of either ACF (adriamycin 35 mg/m², cisplatin 70 mg/m² on Day 1, and continuous fluorouracil 700 mg/m² infusion for 7 days) every 4 weeks or DCF (docetaxel 70 mg/m², cisplatin 70 mg/m² on Day 1, and continuous fluorouracil 700 mg/m² infusion for 5 days) every 3 weeks, as previously described. Surgery was performed after two cycles of neoadjuvant chemotherapy [29]. Clinicopathological findings were classified according to the Union for International Cancer Control (UICC)-TNM classification, seventh edition [30].

Formalin-fixed, paraffin-embedded samples from the patients were collected at our hospital. Cancerous ESCC nests were subjected to DNA extraction after surgery using laser microdissection with a Leica LMD7000 instrument (Leica Microsystems, Wetzlar, Germany). Moreover, of the patients, formalin-fixed, paraffin-embedded samples were collected before and after neoadjuvant therapy from 4 patients with ESCC who had undergone esophagectomy with neoadjuvant therapy. All patients provided written informed consent for the use of the resected specimens. In addition, this study was approved by the ethics committee of Osaka University, Graduate School of Medicine (approval number #15401) and was conducted in accordance with the Declaration of Helsinki.

The mtDNA copy number was measured by quantitative real-time PCR (qPCR) using specific primers for the mtDNA-coded cytochrome oxidase I (MTCO-1) gene and normalized to the expression of the nuclear DNA-encoded cytochrome oxidase IV (COX IV) gene (a mitochondrial respiratory chain enzyme), and the mtDNA copy number was adjusted by setting the mtDNA copy number of control-sh or TE11 cells as 1.00. Complex I/IV-related genes, EMT-related genes including epithelial markers (E-cadherin: CDH1) and mesenchymal markers (N-cadherin: CHD2, vimentin, and Zeb1), and DNA methylation-related genes such as DNMT1, DNMT3A, and DNMT3B were analyzed by qPCR. In addition, the house-keeping gene GAPDH was analyzed in this study. The primers are shown in Additional file 5: Table S1.

A total of 2.0 × 10³ cells per well were seeded in 96-well plates, and cell viability was evaluated using Cell Counting Kit-F (CK06, Dojindo, Japan) at 48 h after incubation under chemotherapy, cisplatin (CDDP), 5-fluorouracil (5-FU) and docetaxel (DTX) exposure. The chemotherapy resistance rate was determined as the ratio of the proliferation under chemotherapy at each concentration to that under control treatment. The fluorescence intensity was measured by an iMark™ Microplate Absorbance Reader (BIO-RAD, Tokyo, Japan) using a plate reader at an excitation wavelength of 490 nm and emission wavelength of 515 nm.

For the apoptosis assay, an Annexin V-Cy3 Apoptosis Staining/Detection Kit was used (ab14142, Abcam, Cambridge, UK). A total of 2 × 10⁵ cells collected by centrifugation were resuspended in binding buffer, and Annexin V-Cy3 was added. Then, the cells were incubated at room temperature for 5 min in the dark. Flow cytometry was carried out using a FACSCanto II flow cytometer (BD Biosciences). The apoptotic cells were detected as cells bound to Annexin V and propidium Iodide (PI), and early and delay apoptosis cells were collected as respectively low PI and high PI.

Harvested cells (1.0 × 10⁵ cells per well) were seeded in 6-well plates and stained using the MitoProbe JC-1 assay (M34152, Invitrogen) according to the manufacturer’s instructions. The mitochondrial membrane potential (MMP) was analyzed by flow cytometry. The MitoProbe JC-1 assay features a conversion of PE-A (red) to FITC-A (green) when the MMP is depolarized. Flow cytometry was carried out using a FACSCanto II flow cytometer (BD Biosciences), the data were analyzed by FlowJo ver 10.3 software (TOMY DIGITAL BIOLOGY, Tokyo, Japan), and the mean fluorescence intensities (MFI) was measured.

A total of 2.0 × 10⁴ cells per well were seeded in multiwell glass bottom dishes (D141400, MATSUNAMI, Osaka, Japan). mtDNA in cells was stained with SYBR Green I at a 1:600,000 dilution for 30 min and washed with PBS four times as previously described [31]. Then, the MMP was determined via staining with MitoTracker Orange (M7510, Invitrogen, California, UA) for 10 min followed by two washes with PBS. Fluorescence immunostaining was evaluated by a Confocal Laser Scanning Microscope FV3000 (Olympus, Tokyo, Japan). This fluorescence was quantified by ImageJ v1.53 software, which is an open source and public domain image processing software.

Complex I activity was investigated using a Complex I Enzyme Activity Microplate Assay Kit (ab109721, Abcam, Cambridge, UK). The protein in cells was added to the microplate wells that had been precoated with a specific capture antibody for complex I. Then, samples were immobilized in the well. In the assay, complex I activity is determined by detecting the oxidation of NADH to NAD + and the simultaneous decrease in dye, which leads to increased absorbance at an optimal density of 450 nm by the SH-9000lab system (Corona Electric, Ibaraki, Japan). Complex IV activity was evaluated using a Complex IV Rodent Enzyme Activity Microplate Assay Kit (ab109911, Abcam, Cambridge, UK). This assay kit was used to determine the activity of cytochrome c oxidase, which participates in complex IV activity, the activity was determined calorimetrically by detecting the oxidation of reduced cytochrome c based on the absorbance change at 550 nm.

Relative protein expression levels were investigated by immunoblotting. E-cadherin, N-cadherin, vimentin, and Zeb-1 expression levels were evaluated using a commercially available EMT antibody sample kit (1:1000 dilution, #9782, Cell Signaling Technology). DNMT1, DNMT3A and DNMT3B levels were evaluated with anti-DNMT1 rabbit monoclonal antibody (1:1000 dilution, #5032, Cell Signaling Technology), anti-DNMT3A rabbit monoclonal antibody (1:1000 dilution, ab227823, Abcam, Cambridge, UK) and anti-DNMT3B rabbit monoclonal antibody (1:1000 dilution, #67259, Cell Signaling Technology), respectively. In addition, TFAM expression levels were assessed using TFAM rabbit polyclonal antibody (1:1000 dilution, ab47517, Abcam, Cambridge, UK). An immunoblotting was quantified by densitometry with ImageJ v1.53 software. The signal intensity for molecule of interest was calibrated by that of β-actin at each point. The relative expression was showed compared to the signal intensity at day0 or control cells as 1.

Invasive conditions of cells were measured using 24-well inserted plates with 8 µm membrane pores and Matrigel coating (#354480, Corning, NY, USA). A total of 7.5 × 10⁴ cells (TE11) were added to the upper chamber, and 0.75 ml medium was added in the lower chamber. These cells were incubated at 37 ̊C in CO2 incubator for 22 h. Thereafter, non-invasive cells in the top chamber were removed by cotton swabs. Invasive cells on the bottom of the membrane were fixed and stained with the Diff-Quick stain kit (16920, Sysmex, Hyougo, Japan). The number of invasive cells was counted.

The overall DNA methylation in cells was analyzed using a methylated DNA quantification kit (ab117128, Abcam, Cambridge, UK), which quantifies global DNA methylation by specifically measuring the levels of 5-methylcytosine (5-mC) in a microplate-based format. Genomic DNA was applied to microplates and bound to assay wells. Then, capture antibody, detection antibody, enhancer solution and developing solution were added after washing well. The fluorescence intensity was measured with an SH-9000lab instrument (Corona Electric, Ibaraki, Japan) using a plate reader at 450 nm optimal density (OD). The simple calculation of 5-mc percentage in total DNA can be carried out using the following formula: 5-mc% = [(Sample OD − Negative Control OD) ÷ S/(Positive Control OD − Negative Control OD) × 2 ÷ P] × 100%.

A total of 5.0 × 10² cells per well were seeded in 96-well plates, and cell viability was evaluated using Cell Counting Kit-F (CK06, Dojindo). The fluorescence intensity was measured at an excitation wavelength of 490 nm and emission wavelength of 515 nm by an iMark™ Microplate Absorbance Reader (BIO-RAD, Tokyo, Japan).

Tumor specimens were fixed with 10% formalin, and paraffin-embedded tissue blocks were sectioned into 2.5-μm slices. The sections were deparaffinized with xylene, dehydrated with ethanol in stages and incubated in 10 mM citrate buffer at 110 °C using a pressure cooker for 15 min for antigen removal. Endogenous peroxidase activity of the tissue specimens was blocked by incubating the slides in 3% hydrogen peroxide (H₂O₂) dissolved in methanol for 20 min at room temperature. Subsequently, the sample was treated with 1% horse serum albumin for 30 min at room temperature to block nonspecific reactions, all sections were incubated with primary antibodies at 4 °C overnight in a humidified environment. The antibodies used in this study were anti-E-cadherin monoclonal antibody (#3195, dilution 1:100, Cell Signaling Technology), anti-N-cadherin monoclonal antibody (33-3900, dilution 1:100, Invitrogen), and anti-vimentin monoclonal antibody (#5741, dilution 1:300, Cell Signaling Technology). After incubation with secondary antibodies for 20 min at room temperature, the reactions were visualized using the VECTASTAIN^® Elite^® ABC Kit (PK-6100, Vector Laboratories), which stains the targeted antigen brown, and counterstaining with hematoxylin. The degree of E-cadherin, N-cadherin and vimentin staining was evaluated based on the extent of membranous staining. Five stained hot spots were evaluated (n = 3), and the average of each stained area was estimated by microscopy (BZ-X 710; Keyence).

TUNEL assays were used to investigate apoptosis in formalin-fixed, paraffin-embedded xenograft tumor tissue samples. In brief, paraffin sections (2.5 μm) were deparaffinized in xylene and rehydrated with 70% and 90% alcohol. The TUNEL signal was detected using the ApopTag Fluorescein In Situ Apoptosis Detection Kit (S7110, Sigma–Aldrich). Nuclei were counterstained using VECTASHIELD Mounting Medium with DAPI (Vector Laboratories). Green fluorescence from apoptotic cells was analyzed by fluorescence microscopy (BZ-X 710; Keyence).

The Mann–Whitney U test, the χ² test, or Student’s t test was used to compare patient characteristics. The overall survival (OS), including cancer or recurrence death, and recurrence-free survival (RFS) rates from November 2006 to December 2011 were calculated from the date of random assignment, validated by the Kaplan–Meier method, and compared with the log-rank test on an intent-to-treat basis; the corresponding HRs were calculated with the 95% CIs. Cox proportional hazards regression models were used to identify variables significantly associated with prognosis. Continuous variables are expressed as the mean ± SD, unless otherwise stated. Statistical significance was indicated at a p value < 0.05. All analyses were performed using JMP^® 14 (SAS Institute Inc., Cary, NC, USA).

The patients were divided into two groups: a group with an mtDNA copy number of the median or more (n = 45) and a group with an mtDNA copy number less than the median (n = 43) (Fig. 1A). Table 1 indicates the clinicopathological characteristics. The proportions of samples with cM stage (TNM classification) and pathological response of grade 0 or 1a was significantly higher in the mtDNA copy number < median group than in the mtDNA copy number ≥ median group, and the mtDNA copy number < median group tended have more samples from patients of older age than the mtDNA copy number ≥ median group. Figure 1B shows the Kaplan–Meier estimates of cumulative OS and RFS in the two groups of patients. The mtDNA copy number < the median was significantly associated with shortened RFS (p = 0.027). This study suggests that a low mtDNA copy number is significantly associated with chemotherapy resistance, poor prognosis and metastasis.

Figure 2 shows the association between chemotherapy resistance and mtDNA-depleted ESCC. mtDNA-depleted TE11 cells exhibited significantly decreased chemosensitivity to cisplatin (CDDP), 5-fluorouracil (5-FU), and docetaxel (DTX) compared with control TE11 cells (Fig. 2A). The rate of apoptosis was significantly lower in mtDNA-depleted ESCC than in control cells according to an apoptosis assay (apoptosis rate under CDDP 0 µM; control-sh: 5.8%, tfam-sh1: 5.0%, tfam-sh2: 3.8%, apoptosis rate under CDDP 5 µM; control-sh: 23.3%, tfam-sh1: 10.1%, tfam-sh2: 6.0%) (Fig. 2B). These results suggested that mtDNA-depleted ESCC cells were resistant to chemotherapy. The cell lines were cultured for a long time under CDDP and 5-FU exposure, and these cell lines were resistant to chemotherapy (Fig. 2C). Moreover, the mtDNA copy number of CDDP- and 5-FU-resistant cell lines was significantly lower than that of control cells (Fig. 2D). In the in vivo experiment, the proliferation of control cells injected in saline was higher than that of mtDNA-depleted ESCC cells. On the other hand, mtDNA-depleted ESCC cells had higher in tumor volume after CDDP injection than control cells (Fig. 2E, F). The above results suggest that ESCC cells with chemotherapy resistance have decreased mtDNA copy number and that depletion of mtDNA induces chemotherapy resistance.

MMP is an important indicator of mitochondrial function [32, 33]. Hence, we focused on the associations between MMP, mtDNA copy number and EMT. CDDP induced a transition of PE-A to FITC-A in control cells and mtDNA-depleted ESCC cells. The rate of transition from PE-A to FITC-A after CDDP administration was lower in mtDNA-depleted ESCC cells than in control cells (Fig. 3A), and the MMP depolarization rate after CDDP administration was significantly lower in mtDNA-depleted ESCC cells than in control cells (Fig. 3B). ESCC cells administered CDDP for a long time had lower fluorescence, indicating lower mtDNA and MMP, than control cell lines (Fig. 3C). Therefore, the data suggest that chemotherapy induces a decrease in MMP. Additionally, there was lower fluorescence, and thus lower mtDNA and MMP, in mtDNA-depleted ESCC cells than in control cells. On the other hand, ESCC cells with increased mtDNA had higher fluorescence, indicating higher mtDNA and MMP, than control cells (Fig. 3D). The quantification of mtDNA and MMP fluorescence per cell was in line with the above results. Complexes I and IV, which are electron-transferring enzymes, are the main factors regulating the MMP [34]. Figure 3E and F show that mtDNA-depleted ESCC cells had reduced complex I and IV activity compared to control cells; this decreased activity regulated the MMP, and there was higher complex I and IV activity in ESCC cells with increased mtDNA than in control cells. These results indicated that mtDNA controls the MMP via complexes I and IV. Hence, the decreased mtDNA induced by chemotherapy leads to a low MMP. Moreover, the mRNA expression levels of complexes I and IV were correlated with mtDNA copy number (Additional file 2: Fig. S2A).

The uncoupler carbonyl cyanide m-chlorophenylhydrazone (CCCP; ab141229, Abcam, Cambridge, UK), which is a compound that inhibits the conjugation of both reactions in oxidative phosphorylation, artificially reduces mitochondrial membrane potential. This study investigated the influence of CCCP on ESCC cells by FACS. Exposure to 40 µM CCCP for 5 days induced the greatest decrease (24.1%) in mitochondrial membrane potential at 20–40 µM (Fig. 4A). The mtDNA copy number in cells exposed to 40 µM CCCP was not different from that in control cells not exposed to CCCP (Additional file 2: Fig. S2B). Cells that were exposed to 40 µM CCCP showed spindle cell transformation, which varied in severity depending on the exposure time (Days 1–5) (Fig. 4B). The relative mRNA expression levels of E-cadherin in ESCC cells cultured with CCCP were significantly decreased at Day 5 compared with those in control cells not treated with CCCP. On the other hand, ESCC cells treated with CCCP had significantly higher mRNA expression of N-cadherin, vimentin, and Zeb-1 than those not treated with CCCP (Fig. 4C). Additionally, the protein expression level of E-cadherin at Day 5 after CCCP exposure was decreased compared with that in control cells (Day 0), and the N-cadherin, vimentin, and Zeb-1 protein expression levels at Day 5 after CCCP exposure were increased compared with those in control cells (Day 0) (Fig. 4D). These results regarding the relationship between MMP and EMT were similar in TE8 cells (Additional file 2: Fig. S2D, E), and treatment with 10 µM CCCP for 5 days reduced mitochondrial membrane potential (Additional file 2: Fig. S2C). Moreover, invasion assay showed that TE11 which administrated CCCP for 5 days had significantly more invasion than TE11 without CCCP (Fig. 4E). These results suggest that decreased MMP induces EMT.

This study additionally focused on DNA methylation and the methylation transcription factor DNA methyltransferase (DNMT) in detail mechanism which depleted mtDNA and decreased MMP controlled the expression of nuclear genes. mtDNA-depleted TE11 cells had an approximately twofold higher mRNA level of DNMT1, DNMT3A and DNMT3B than the control cell lines (Fig. 5A). The DNMT1, DNMT3A and DNMT3B protein levels were higher in ESCC cells with low mtDNA than in control cells (Fig. 5B). mtDNA-depleted TE11 cells hain DNMT1, DNMT3a and DNMT 3Bd an approximately 10% higher 5-mC level (a measure of global DNA methylation) than control cell lines (Fig. 5C). Hence, low mtDNA copy number induced DNA methylation by DNMT. The mRNA expression level of DNMT1, DNMT3A and DNMT3B was twofold to threefold higher in ESCC cell lines treated with CCCP than in those not treated with CCCP (Fig. 5D). There were almost higher protein levels of DNMT1, DNMT3A and DNMT3B in ESCC cell lines exposed to 40 µM CCCP than in ESCC cell lines that were not exposed to 40 µM CCCP (Fig. 5E). mtDNA-depleted TE8 cells had higher mRNA and protein levels of DNA methylation-related genes, such as DNMT1, DNMT3A and DNMT3B (Additional file 3: Fig. S3A, B). These data indicate that decreased mtDNA and MMP promote DNMT expression via DNA methylation.

ESCC cells depleted of mtDNA showed cell transformation from spindle to nonspindle morphology after administration of 100 µM zebularine (S7113, Selleck, Houston, USA), a DNMT inhibitor, for 8 days (Fig. 6A). Also, the morphology of control cell with DNMT inhibitor remained nonspindle and had no change (Additional file 3: Fig. S3C). 5-mC was decreased by 10% in mtDNA-depleted ESCC cells treated with zebularine compared to that in ESCC cells not treated with zebularine (Fig. 6B). The mRNA expression of E-cadherin, N-cadherin, vimentin, and Zeb-1 was similar in control cell lines treated with and not treated with zebularine. However, the mRNA expression of E-cadherin was significantly higher in mtDNA-depleted ESCC cells treated with zebularine than in those not treated with zebularine, and mtDNA-depleted ESCC cells treated with zebularine had significantly lower expression of N-cadherin and vimentin than cells not treated with zebularine (Fig. 6C). Additionally, the protein expression of E-cadherin was increased by zebularine, and the protein levels of N-cadherin and vimentin were decreased by zebularine (Fig. 6D). Zebularine significantly impacted mtDNA-depleted ESCC cells by increasing chemosensitivity to CDDP, 5-FU and DTX, while there was no difference in chemosensitivity in control cell lines treated with and not treated with zebularine (Fig. 6E). The proliferation rates of mtDNA-depleted ESCC cells and control cells were significantly reduced by zebularine (Fig. 6F). Based on these results, the DNMT inhibitor zebularine improves chemotherapy sensitivity by suppressing EMT.

Moreover, this study evaluated the impact of DNMT1 expression on mtDNA-depleted ESCC cells. DNMT1 knockdown (DNMT1-siRNA, AM16708, Thermo Fisher Scientific) reduced the mRNA and protein expression of DNMT1 in ESCC cells, although the expression of DNMT3A and DNMT3B was similar after DNMT1 knockdown (Fig. 7A). The mRNA expression levels of DNMT1 in control cells and mtDNA-depleted cells were significantly lower after DNMT1 knockdown. However, the mRNA expression levels of DNMT3A and DNMT3B after DNMT1 knockdown were low reduction rate compared that of DNMT1 (Fig. 7B). Overall DNA methylation was reduced by approximately 7–8% by DNMT1 knockdown (Fig. 7C). mtDNA-depleted ESCC cells with DNMT1 knockdown had significantly increased mRNA expression levels of E-cadherin compared to cells without DNMT1 knockdown. In addition, N-cadherin and vimentin mRNA expression was significantly lower in DNMT1 knockdown cells than in cells without DNMT1 knockdown (Fig. 7D). The protein levels change of DNMT and EMT markers were similar to mRNA level expression in DNMT1 knockdown (Additional file 3: Fig. S3D). The mtDNA-depleted ESCC cells with DNMT1 knockdown had significantly increased sensitivity to CDDP compared with cells without DNMT1 knockdown (Fig. 7E). Therefore, we suggest that DNMT1 knockdown suppresses chemotherapy resistance by increasing epithelial markers and decreasing mesenchymal markers.

The mtDNA copy number of the ESCC patients after NAC was approximately 40% lower than that of the ESCC patients before NAC (Additional file 4: Fig. S4A). Therefore, we hypothesized that a DNMT inhibitor could improve chemotherapy sensitivity in mtDNA-depleted ESCC cells with chemotherapy resistance. mtDNA-depleted ESCC cells were injected subcutaneously into BALB/cAJcl nude mice (n = 3), and subcutaneous tumors were administered the following drugs 7 days after injection: saline, CDDP, zebularine and CDDP combined with zebularine intraperitoneally every 3–4 days (Fig. 8A). The physical change in tumor volume in was not different between the saline- and CDDP-treated mice that received mtDNA-depleted ESCC cells. However, the tumor volume was lower in mice that received cells and were then treated with CDDP combined with zebularine than in mice that received cells and were then treated with only CDDP or zebularine. Moreover, the proliferation of mtDNA-depleted ESCC cells in mice treated with CDDP combined with zebularine was significantly lower than that of cells in mice treated with only CDDP or zebularine. The tumor weight results were in line with the above results (Fig. 8B). Immunostaining showed that E-cadherin expression was higher in mtDNA-depleted ESCC cells injected into mice treated with CDDP combined with zebularine than in those injected into mice treated with other drugs and that the expression of N-cadherin and vimentin was lower in cells injected into mice treated with CDDP combined with zebularine than in cells injected into mice treated with other drugs (Fig. 8C). Compared with other drugs, CDDP combined with zebularine induced the most apoptosis (Fig. 8D). Our results show that chemotherapy combined with a DNMT inhibitor reduces EMT and improves chemotherapy sensitivity in ESCC cells with low mtDNA and chemotherapy resistance.