A Histone Deacetylase Inhibitor Suppresses Epithelial-Mesenchymal Transition and Attenuates Chemoresistance in Biliary Tract Cancer

Epithelial-mesenchymal transition (EMT) is involved in the characteristics of malignancy, such as invasion, metastasis, and chemoresistance. In biliary tract cancer (BTC), EMT is induced by transforming growth factor-beta 1 (TGF-β1). The EMT is reversible; therefore, it is conceivable that it could be related to some epigenetic changes. We focused on histone deacetylase (HDAC) inhibitors as regulators of TGF-β1 signaling, and investigated their effect on EMT and chemoresistance. We employed four BTC cell lines (MzChA-1, gemcitabine-resistant MzChA-1, TFK-1, and gemcitabine-resistant TFK-1) and used vorinostat as the HDAC inhibitor. The relative mRNA expression of an epithelial marker (CDH1) and mesenchymal markers (CDH2, vimentin, SNAI1) were measured by qRT-PCR to evaluate factors associated with EMT. MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay was performed to evaluate the chemoresistance of each cell line. In addition, NOD/SCID mice were used to evaluate the effect of vorinostat in vivo. In the parent MzChA-1 and TFK-1 cell lines, TGF-β1 induced EMT and chemoresistance; while vorinostat inhibited the EMT and chemoresistance induced by TGF-β1. In gemcitabine-resistant cell lines that highly expressed TGF-β1, vorinostat inhibited EMT and attenuated chemoresistance. We showed that vorinostat inhibits nuclear translocation of SMAD4 which is a signaling factor of TGF-β1, and this is one of the mechanisms by which vorinostat regulates EMT. We also showed that vorinostat attenuates the binding affinity of SMAD4 to the CDH1-related transcription factors SNAI1, SNAI2, ZEB1, ZEB2, and TWIST. Furthermore, combination therapy with vorinostat and gemcitabine improved survival time in the mice xenografted with gemcitabine resistant MzChA-1 cells. In conclusion, vorinostat regulated TGF-β1-induced EMT and chemoresistance through inhibition of SMAD4 nuclear translocation.


Introduction
Biliary tract cancer (BTC), which has an increasing incidence worldwide, has a poor prognosis because it is difficult to diagnose in the early stages. The 5-year survival rate is less than 30% [1][2][3]. At diagnosis, many patients have unresectable BTC due to local disease spread and/or metastases. Gemcitabine (GEM)-based chemotherapy is often used; however, the antitumor effect is insufficient, and the median survival time is limited to approximately 10 months [4,5]. To improve the prognosis of BTC patients, the molecular biology of the disease has been studied, including the inflammatory cytokines, epithelial-mesenchymal transition (EMT), DNArepair system, cell signaling, and apoptosis [6][7][8][9]. We previously showed that EMT was observed at the invasion front and in regional lymph node metastases with inflammatory cytokine expression in BTC [10]. Several kinds of cytokines reportedly induce EMT, which is related to malignant processes such as invasion, metastasis, and chemoresistance (S1 Table) [ [11][12][13][14][15]. Transforming growth factor-beta 1 (TGF-β1) induces EMT and chemoresistance, and chemoresistant BTC cells produce IL-6 and TGF-β1 [10]; therefore, we considered that these cytokines could play an important role in promoting EMT in BTC.
Consecutive rounds of EMT and mesenchymal-epithelial transition (MET) occur during tumor development, suggesting that EMT could be reversible and associated with epigenetic changes [16]. We focused on histone deacetylase (HDAC) as a factor related to epigenetic changes. In our microarray data, the mRNA levels of class I HDACs (HDAC1, HDAC2, HDAC3, and HDAC8) were elevated in GEM-resistant BTC cells. Moreover, the HDAC activity was increased by exposure to TGF-β1. These findings suggest that HDACs could be associated with EMT, and inhibition of HDACs may suppress EMT through effects on TGF-β1 signaling. We also examined SMAD4, the key signaling factor of TGF-β1 for EMT, and found it was directly stimulated by TGF-β1 signaling, with translocation into the nucleus where it upregulates EMT-related transcription factors such as SNAI1, SNAI2, ZEB1, ZEB2, and TWIST [17][18][19][20][21]. Additionally, in normal hepatocytes and mesothelial cells, inhibition of HDACs attenuated the nuclear translocation of SMAD4 [22,23]. In this study, we hypothesized that HDAC inhibitors might regulate EMT through the TGF-β1 signaling pathway in BTC cells.
The aim of this study was to investigate the effect of the HDAC inhibitor vorinostat on the TGF-β1 signaling pathway and chemoresistance, which is related to EMT in BTC cells. First, we investigated the effect of vorinostat on TGF-β1-induced EMT. Then, we investigated the effect of vorinostat on chemoresistant BTC cells. Subsequently, we showed the influence of vorinostat on SMAD4 nuclear translocation in BTC cells. Finally, we evaluated the efficacy of vorinostat combined with GEM for BTC with using mice xenografted with MzChA-1 or GEMresistant MzChA-1 cells. These results showed the effectiveness of a HDAC inhibitor in regulating EMT and chemoresistance in vitro and in vivo. We also identified one of the mechanisms by which a HDAC inhibitor suppresses EMT and attenuates chemoresistance.
The TFK-1 cell line was obtained from the Riken BioResource Center in Japan. These cells were propagated in RPMI 1640 medium. Each medium was supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin. All cell lines were incubated at 37°C in a humidified incubator with 5% CO 2 .
All cell lines were treated with 5 ng/ml of recombinant human TGF-β1 (Pepro Tech Inc., Rocky Hill, NJ, USA), 100 nM (MzChA-1 and GEM-resistant MzChA-1) or 300 nM (TFK-1 and GEM-resistant TFK-1) of the histone deacetylase inhibitor vorinostat (Selleckchem, Houston, TX, USA), or dimethyl sulfoxide (as a control) for 72 h. In the experiment with TGF-β1, each medium was changed to serum-free medium 24 h after passage to preclude the influence of TGF-β1 in serum. The TGF-β1 and vorinostat were added to each cell line after the change of medium. The antibodies described in S2 Table were used for western blot analysis, immunohistochemistry, and immunocytochemistry.

Assay of HDAC activity
The MzChA-1 cells were cultured with or without 5 ng/ml TGF-β1 and 300 nM vorinostat for 72 h. Then, nuclear proteins were extracted with the Nuclear Extract Kit (Active Motif, Carlsbad, CA, USA). Subsequently, the activity of HDAC was measured with the HDAC activity assay kit (BioVision, Palo Alto, USA). In this experiment, we used HeLa nuclear extract as a positive control and distilled water as a negative control. All procedures were conducted according to the manufacturer's recommendations.

Immunocytochemistry
Cells were fixed with 4% paraformaldehyde for 15 min at room temperature, permeabilized with 0.1% Triton X-100, and blocked with 1% BSA for 10 min at room temperature. Subsequently, the cells were stained with the indicated antibodies.

Quantitative reverse transcription-polymerase chain reaction
Total RNA was isolated from cultured cells using Trizol reagent (Invitrogen, Carlsbad, CA, USA). Complementary DNA was synthesized from 3.0 μg of total RNA with the SuperScript first-strand synthesis system (Invitrogen) according to the manufacturer's protocol. Quantitative real-time polymerase chain reactions (qRT-PCRs) were conducted with the LightCycler-Fast-Start DNA Master SYBR Green I kit (Roche Applied Science, Indianapolis, IN, USA) with gene-specific oligonucleotide primers, as shown in S3 Table. Amplifications were performed in triplicate using the LightCycler System (Roche Applied Science) in accordance with the manufacturer's protocol. A melting curve analysis was performed to distinguish specific products from non-specific products and primer dimers. The relative expression was calculated as the ratio of specific mRNA to endogenous β-actin (ACTB) mRNA in each sample. All qRT-PCRs were performed in triplicate, and the results were presented as the mean ± standard deviation.

Growth inhibition assays with GEM therapy
Each cell line (5×10 3 cells/well) was seeded in a 96-well plate and incubated for 24 h. Then, the medium was changed to serum-free medium. Subsequently, the cells were exposed to GEM (1-120ng/mL) with the indicated drugs (TGF-β1, vorinostat, or dimethyl sulfoxide) for 72h. These assays were repeated at least three times, and similar results were obtained each time. The proportion of MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide)-positive cells incubated without drugs was defined as 100% viability.

Transfection of small interfering RNA (siRNA)
We used TGF-β siRNA (Invitrogen) and scrambled oligonucleotide siRNA (Sc) as a negative control. We performed siRNA transfection with Lipofectamine RNAiMAX (Invitrogen), according to the manufacturer's protocols [28]. Transfection efficiency was confirmed by qRT-PCR.

Western blot analysis
Western blot analysis was performed as described previously [28]. Briefly, a whole cell lysate was extracted from BTC cells with RIPA Buffer (Thermo Fisher Scientific, Inc., Rockford, IL, USA) and nuclear proteins were extracted with the EpiQuick Nuclear Extraction Kit I (Epigentek Group Inc., Brooklyn, NY, USA), according to the manufacturer's protocol. Aliquots (12 μg) of total or nuclear proteins were electrophoresed on sodium dodecyl sulfate-polyacrylamide gels containing 10% Tris-HCl (Bio-Rad Laboratories Inc., Hercules, CA, USA). The separated proteins were transferred to polyvinylidene difluoride membranes (Millipore) and incubated with each primary antibody overnight at 4°C.

Chromatin immunoprecipitation (ChIP)
The ChIP assay was performed with using a ChIP-IT Express Enzymatic kit (Active Motif) as previously described [29]. Briefly, 1 × 10 7 cells were cross-linked with 1% formaldehyde for 10 min at room temperature and quenched by adding glycine. To harvest chromatin-DNA complexes, the cell lysate was treated with an enzymatic shearing kit (Active Motif), according to the manufacturer's specifications. For immunoprecipitation, 120 μg of chromatin-DNA complexes were incubated with Protein G magnetic beads linked to anti-SMAD4 antibody (10 μg, Cell Signaling Technology, Danvers, Massachusetts, USA), or anti-IgG antibody (10 μg, Active Motif). Eluates were used as templates for PCR. Equal amounts of anti-IgG or pre-immune chromatin-DNA complex were used as controls. The primer sets used for PCR amplification were: SNAI1:

Mouse xenograft tumor model
The MzChA-1 or MzChA-1_GR cells (1.0 × 10 6 cells per mouse) were injected into the subcutaneous tissue on the back of five-to six-week-old male nude mice (CLEA Japan Inc., Tokyo, Japan), and tumors were allowed to develop. When the tumors grew to the volume of 200 mm 3 , GEM (125 mg/kg, once a week) with or without vorinostat (60 mg/kg, 5 consecutive days per week) was administered by intraperitoneal injection. To investigate the effect of vorinostat on survival time, treatment was continued until appropriate euthanasia. Resected specimens of tumors were used for immunohistochemistry to investigate the expression of SMAD4.

Immunohistochemistry
Immunohistochemical studies on 10 resected tumor specimens from the mice xenografted with MzChA-1_GR cells. Briefly, formalin-fixed, paraffin-embedded tissues were deparaffinized, boiled for antigen retrieval, incubated with a SMAD4 antibody for 1 hour at room temperature, then visualized with avidin-biotin complex reagents (Vector Laboratory Inc., Burlingame, CA, USA) and diaminobenzidine. All sections were counterstained with hematoxylin. All immunohistochemical studies were evaluated in serial sections with each antibody.

Ethical statements
All animal experiments were carried out in strict accordance with the Institutional Animal Care and Use Committee at the Osaka University (approval number: 20-087), and studies were conducted in compliance with institutional guidelines. In the animal experiments, all invasive treatments such as xenograft transplantation, gavage, and euthanasia were performed under general anesthesia with sevoflurane. The mice were euthanized at the onset of clinical signs such as severe cachexia, significant weight loss, or inactivity.

Statistical analysis
The data are presented as the mean ± standard deviation (SD) from at least three independent experiments. Statistical analysis was performed using Student's t-test or Fisher's exact test for categorical data. The unpaired Student's t-test was used to examine differences in growth inhibitory effects in vitro. P-values < 0.05 were considered statistically significant.

Microarray analysis
A microarray analysis was performed using the TORAY 3D-Gene 1 to compare the expression profiles of MzChA-1 cells with those of MzChA-1_GR cells and investigate the factors related to epigenetics. The results showed that the expression of class I HDACs (HDAC-1, HDAC-2, HDAC-3, and HDAC-8) were higher in MzChA-1_GR cells than parent MzChA-1 cells (Table 1). Vorinostat inhibits these class I HDACs, so we considered that vorinostat might be effective in chemoresistant BTC cells.
The mRNA expression of class I HDACs and HDAC activity in MzChA-1 and MzChA-1_GR cells To validate the microarray analysis results, qRT-PCR for class I HDACs was performed for MzChA-1 and MzChA-1_GR cells. All class I HDACs showed higher expression in MzChA-1_GR cells compared with MzChA-1 cells (P<0.05, Fig 1A). The HDAC activity was significantly higher in MzChA-1 cells exposed to TGF-β1 for 72 h and MzChA-1_GR cells compared with parent MzChA-1 cells (P<0.05, Fig 1B). In addition, vorinostat suppressed the HDAC activity in both MzChA-1 cells exposed to TGF-β1 and MzChA-1_GR cells (P<0.05, Fig 1C).

The relationship between TGF-β1 expression and chemoresistance
To clarify the significance of TGF-β1 in EMT and chemoresistance, we investigated the expression of TGF-β1 in parent MzChA-1 and MzChA-1_GR cells. The results showed that the expression level of TGF-β1 was significantly higher in MzChA-1_GR cells than parent MzChA-1 cells (Fig 2A). Then, we transfected with siTGF-β to investigate whether EMT and chemoresistance could be suppressed by TGF-β knock down in MzChA-1_GR cells. The result showed that MzChA-1_GR cells transfected with siTGF-β had increased the mRNA expression of CDH1 and decreased the mRNA expression of CDH2, vimentin, and SNAI1 (Fig 2B). Moreover, transfection with siTGF-β attenuated chemoresistance in MzChA-1_GR cells (Fig 2C). These results suggested the significance of TGF-β in EMT and chemoresistance in BTC.
In TFK-1_GR cells, vorinostat changed the morphology and enhanced the staining of CDH1, as measured by immunofluorescence staining similar to MzChA-1_GR cells (Fig 4D).   In addition, vorinostat increased the expression of CDH1 mRNA, and decreased the mRNA expression of CDH2, VIM, and SNAI1 (P<0.05, Fig 4E). GEM resistance was also attenuated by vorinostat in TFK-1_GR cells (P<0.05, Fig 4F). These results suggested that vorinostat suppressed EMT and attenuated chemoresistance in GEM resistant BTC cells.
The effect of vorinostat on SMAD4 nuclear translocation in MzChA-1 cells In our previous study, SMAD4 was the key molecule regulating TGF-β1-induced EMT [10]. Therefore, we investigated the effect of vorinostat on the expression of SMAD4 by western blot analysis. In whole cell lysates, there was no significant difference in the expression of SMAD4 protein between the TGF-β1 group and the TGF-β1 plus vorinostat group (Fig 5A, left). However, TGF-β1 upregulated the expression of the SMAD4 nucleoprotein compared with the control, and vorinostat inhibited this upregulation of the SMAD4 nucleoprotein caused by TGF-β1 (Fig 5A, right). Furthermore, we used the ChIP method to analyze the molecular mechanism of SMAD4. While TGF-β1 enhanced the binding affinity of SMAD4 to the CDH1-related transcriptional factors SNAI1, SNAI2, ZEB1, ZEB2, and TWIST, vorinostat inhibited the TGF-β1-enhanced binding affinity of SMAD4 to these transcriptional factors (Fig 5B). Immunocytochemistry demonstrated that SMAD4 stained strongly in the nuclei after treatment with TGF-β1 (Fig 5C). In contrast, SMAD4 was weakly stained in the nuclei after treatment with TGF-β1 plus vorinostat (Fig 5C). Based on these results, vorinostat regulated EMT through inhibition of SMAD4 nuclear translocation.
The effect of vorinostat on the TGF-β signaling factors, SMAD2, SMAD3, and Jun N-terminal kinase (JNK) We also investigated the effect of vorinostat on SMAD2 and SMAD3 because they are closely related to SMAD4 nuclear translocation in TGF-β signaling. We found TGF-β1 induced SMAD2 and SMAD3 phosphorylation in both whole cell lysates and nuclear proteins (Fig 6). In addition, vorinostat inhibited TGF-β1-induced nuclear translocation of phosphorylated SMAD2 (p-SMAD2) and SMAD3 (p-SMAD3); while, vorinostat had no effect on the nuclear translocation of SMAD2 and SMAD3 (Fig 6). These results suggest that vorinostat can inhibit nuclear translocation of p-SMAD2, p-SMAD3, and SMAD4 at the same time. We also investigated the activation of JNK because recent reports showed that JNK activation plays a central role in TGF-β mediated EMT. The result showed TGF-β1 upregulated the expression of phosphorylated JNK; however, vorinostat had no effect on JNK (Fig 6). Accordingly, the effect of vorinostat on EMT seems to be independent of the JNK pathway in BTC.

The effect of vorinostat on the mouse xenografted tumor model
To confirm the in vitro results, we evaluated the effect of vorinostat on mice xenografted with MzChA-1 or MzChA-1_GR cells. In the mice xenografted with MzChA-1 cells, combination therapy with vorinostat and GEM did not improve the survival time compared with a single administration of GEM (Fig 7A). In contrast, combination therapy with vorinostat and GEM   -rank test, P<0.01, Fig 7B).
Subsequently, we performed immunohistochemistry on the resected tumor specimens of the mice to evaluate the expression of SMAD4 in vivo. The result showed that SMAD4 was strongly stained in the nuclei of tumor cells when the mice were treated with a single administration of GEM. In contrast, SMAD4 was weakly stained in the nuclei of tumor cells when the mice were treated with GEM and vorinostat (Fig 7C). These results were the same in both primary and metastatic sites (pulmonary). Thus, vorinostat improved survival time in the mice xenografted with chemoresistant cells when they were treated with GEM.

Discussion
In many cancers, several genes related to tumorigenesis and tumor progression are upregulated, and genes related to tumor suppression are downregulated. One of the factors that regulate this gene expression is posttranslational modification of the histones. Methylation and acetylation of histones are well known post-translational modifications [30][31][32]. Recently, much attention has been paid to the methylation and acetylation of histones for the treatment of cancers [33,34]. In this study, we focused on the ability of HDAC inhibitors to regulate EMT. We used vorinostat, which inhibits class I HDACs and is often used as a HDAC inhibitor in cancer clinical trials, as shown in S4 Table [35][36][37][38][39]. In several cancers, such as non-small cell lung cancer and breast cancer, combination chemotherapy with vorinostat was safe and showed potential for improving the efficacy [38,40].
We focused on SMADs, particularly SMAD4 which is a signaling factor of TGF-β1, as one of the mechanisms that vorinostat regulates EMT. It is often reported that TGF-β induces EMT and chemoresistance [10]. Activated type I TGF-β1 receptors phosphorylate SMAD2 and SMAD3, which form a heteromeric complex with SMAD4 [41,42]. This complex translocates to the nucleus and targets a variety of DNA binding proteins to regulate transcriptional responses [41,42]. Thus, transcriptional regulators of CDH1, such as SNAI, ZEB, and TWIST, are upregulated [42]. Then, CDH1 is downregulated and EMT occurs [43,44]. Thus, inhibiting nuclear translocation of SMAD2, SMAD3, and SMAD4 complexes could regulate TGF-β1-related EMT. First, we investigated whether vorinostat changes the mRNA expression of these SMADs by qRT-PCR. No changes in these SMADs were observed after exposure to vorinostat ( S2 Fig). However, a few studies have shown aneffect of HDAC inhibitors on SMAD4 [22,23]. Kaimori et al. showed that histone deacetylase inhibition suppresses the TGF-β1-induced EMT in hepatocytes via inhibition of SMAD4 nuclear translocation [22]. Chung et al. showed that the HDAC inhibitor, m-carboxycinnamic acid, inhibited SMAD4 nuclear translocation in human pleural mesothelial cells [23]. However, there are no studies on the effect of HDAC inhibitors on SMAD4 nuclear translocation in any cancer. Our investigation showed that vorinostat inhibited SMAD4 nuclear translocation induced by TGF-β1 in BTC. We also showed that vorinostat inhibited nuclear translocation of p-SMAD2 and p-SMAD3. These results suggest that vorinostat inhibits nuclear translocation of the p-SMAD2, p-SMAD3, and SMAD4 complex. To clarify whether vorinostat sufficiently inhibited EMT related to TGF-β signaling, we investigated the effect of vorinostat and siTGF-β transfection in MzChA-1_GR cells. We found vorinostat with siTGF-β transfection had no additional effect on EMT and chemoresistance (data not shown). This result suggests that the effect of vorinostat is similar to transfection with siTGF-β in regulating EMT and chemoresistance.
In vivo experiments showed that vorinostat with GEM treatment improved the survival time of the mice xenografted with MzChA-1_GR cells compared with single administration of GEM. However, there was no significant change in tumor development between the vorinostat with GEM treatment group and GEM treatment group (S3 Fig). One of the reasons why vorinostat could improve the survival rate might be that vorinostat suppressed tumor metastasis. In fact, the tumors in metastatic sites tended to be large in the GEM treatment group compared with the vorinostat with GEM treatment group. Immunohistochemistry showed that SMAD4 was more strongly stained in the nuclei of cancer cells in the GEM treatment group compared with the vorinostat with GEM treatment group. These results suggest that vorinostat regulates tumor progression and metastasis by inhibiting nuclear translocation of SMAD4 in vivo.
Some studies reported that HDAC inhibitors have a similar function as TGF-β via activation of CDK inhibitors. One of the representative CDK inhibitors related to EMT is p21, and TGFβ induces p21 through a p53-independent mechanism [45]. The HDAC inhibitors are reported to activate p21 in breast cancer cells [46]. Several studies showed that p21 inhibited EMT [47]. Therefore, we investigated the change in p21 expression by TGF-β1 and vorinostat. The result showed that only TGF-β1 induced p21 expression (S4 Fig). In addition, another CDK inhibitor,  p15, which is reported to be induced by TGF-β1 was similarly induced (S4 Fig).These results indicate the effect of vorinostat in regulating the activation of CDK inhibitors might differ depending on the kind of cancer. This issue is controversial and more investigation would be needed in other types of cancer.
We used vorinostat in this study; however, other HDAC inhibitors, such as trichostatin A, panobinostat, entinostat, and valproic acid have been used to regulate EMT, and some of them have been used in clinical trials [48][49][50][51]. In the future, these drugs will need to be tested to determine if they regulate EMT and chemoresistance, as suggested by the results of this study. In conclusion, the HDAC inhibitor, vorinostat, regulated EMT and chemoresistance in BTC via inhibition of nuclear translocation of SMAD4 in vitro and in vivo. (TIF) S1 Table. Cytokines associated with EMT (Table A). References for S1A Table (Table B).