Histone Deacetylase Inhibitors and Phenotypical Transformation of Cancer Cells

Histone deacetylase inhibitors (HDIs) are a group of potent epigenetic drugs which have been investigated for their therapeutic potential in various clinical disorders, including hematological malignancies and solid tumors. Currently, several HDIs are already in clinical use and many more are on clinical trials. HDIs have shown efficacy to inhibit initiation and progression of cancer cells. Nevertheless, both pro-invasive and anti-invasive activities of HDIs have been reported, questioning their impact in carcinogenesis. The aim of this review is to compile and discuss the most recent findings on the effect of HDIs on the epithelial-mesenchymal transition (EMT) process in human cancers. We have summarized the impact of HDIs on epithelial (E-cadherin, β-catenin) and mesenchymal (N-cadherin, vimentin) markers, EMT activators (TWIST, SNAIL, SLUG, SMAD, ZEB), as well as morphology, migration and invasion potential of cancer cells. We further discuss the use of HDIs as monotherapy or in combination with existing or novel anti-neoplastic drugs in relation to changes in EMT.


Introduction
Epithelial-mesenchymal transition (EMT) is a biological reversible process in which cells undergo multiple biochemical changes-lose their epithelial properties, including cell-cell adhesion and cell polarity, and acquire mesenchymal phenotype, including the ability to invade the extracellular matrix (ECM) and potentially migrate to the distant places. Induction of EMT includes reorganization of cytoskeleton proteins, activation of transcription factors and production of extracellular matrix-degrading enzymes [1,2]. Recent studies revealed the large role of epigenetic mechanisms including DNA methylation, chromatin rearrangement, histone modifications and non-coding RNAs in the initiation and progression of cancers [3]. Histone modifications play important

Histone Deacetylases (HDACs) and Histone Deacetylase Inhibitors (HDIs)
Epigenetic regulation of gene expression is largely modulated throughout chromatin and nucleosome remodeling, which involves histone post-translational modifications (PTMs) [16]. These PTMs result in dynamic shifts between transcriptionally active and suppressed states of chromatin [17]. Histone PTMs include methylation, phosphorylation, acetylation, sumoylation, ubiquitination and ADP-ribosylation [18]. Histone acetylation, one of the most extensively studied PTMs of histones, is regulated by the balance between histone deacetylases (HDACs) and histone acetyltransferases (HATs) (Figure 2A) [16]. HATs are enzymes that transfer an acetyl group from acetyl-CoA to ε-amino lysine residues located on N-terminus of histones [19]. In contrast, HDACs are responsible for removing the acetyl group from the acetylated lysine residues. This reversible reaction is crucial for chromatin structure stabilization and transcriptional regulation of gene expression [20,21]. Histone

HDACs
The 18 HDACs in humans are classified into four classes according to their sequence homology with yeast proteins and cofactor dependency [23]. The class I shares common domains with yeast transcriptional regulator RPD3 and includes HDAC1, HDAC2, HDAC3 and HDAC8 and they are placed into the nuclear compartment. Class II of HDACs is shared into two subclasses (IIa and IIb) and is closely related with HDA1 in yeast [24]. The class IIa encompasses HDAC4, HDAC5, HDAC7, and HDAC9, whereas class IIb includes HDAC6 and HDAC10 [25,26]. Class IIa HDACs are inactive on acetylated substrates, thus differing from class I and IIb enzymes. It has been demonstrated that class IIa HDACs are very inefficient enzymes on standard substrates [27,28]. Class II HDACs (HDAC4, HDAC5, HDAC7 and HDAC9), compared to HDAC class I, possess limited enzymatic activity on their own. Instead, they appear to act as gene-specific transcriptional corepressors mainly as components of multiprotein complexes [29][30][31]. The class II HDACs migrate between cytoplasm and nucleus. The class III (sirtuins) includes seven members (SIRT1-SIRT7) and they share common domains with yeast silent information regulator 2 (SIR2) [27]. Class IV contains only one member -HDAC11 [32]. Catalytical activity of class I, II, IV is strongly associated with presence of zinc ion in their active site. In contrast, class III requires nicotinamide adenine dinucleotide (NAD) as a cofactor during their catalytical reaction [25,26].
Imbalances between the activities of HDACs and HATs are associated with a plethora of diseases [33][34][35][36]. The epigenetic aberrations of gene expression caused by increased activity of HDACs play a pivotal role in cancer development and progression [37]. Given the fact that the activity of HDACs is dysregulated in many types of cancers [38][39][40], HDACs have been considered as therapeutic targets for the treatment of neoplasms, indeed HDIs have become promising anti-cancer agents [41,42].
According to the Human Protein Atlas class I HDACs are expressed in variety types of tumors ( Figure 3) [43][44][45][46]. The data is presented by the percentage (%) of analyzed tumors with HDACs expression at high or medium level. HDAC2 was found expressed in 100% of multiple tumors. In the case of renal cancer HDAC2 is expressed 100% of cases, while other members do not exceed 50% ( Figure 3) [44]. HDAC1 expression is comparatively high with exception of renal cancer and glioma

HDACs
The 18 HDACs in humans are classified into four classes according to their sequence homology with yeast proteins and cofactor dependency [23]. The class I shares common domains with yeast transcriptional regulator RPD3 and includes HDAC1, HDAC2, HDAC3 and HDAC8 and they are placed into the nuclear compartment. Class II of HDACs is shared into two subclasses (IIa and IIb) and is closely related with HDA1 in yeast [24]. The class IIa encompasses HDAC4, HDAC5, HDAC7, and HDAC9, whereas class IIb includes HDAC6 and HDAC10 [25,26]. Class IIa HDACs are inactive on acetylated substrates, thus differing from class I and IIb enzymes. It has been demonstrated that class IIa HDACs are very inefficient enzymes on standard substrates [27,28]. Class II HDACs (HDAC4, HDAC5, HDAC7 and HDAC9), compared to HDAC class I, possess limited enzymatic activity on their own. Instead, they appear to act as gene-specific transcriptional corepressors mainly as components of multiprotein complexes [29][30][31]. The class II HDACs migrate between cytoplasm and nucleus. The class III (sirtuins) includes seven members (SIRT1-SIRT7) and they share common domains with yeast silent information regulator 2 (SIR2) [27]. Class IV contains only one member -HDAC11 [32]. Catalytical activity of class I, II, IV is strongly associated with presence of zinc ion in their active site. In contrast, class III requires nicotinamide adenine dinucleotide (NAD) as a cofactor during their catalytical reaction [25,26].
Imbalances between the activities of HDACs and HATs are associated with a plethora of diseases [33][34][35][36]. The epigenetic aberrations of gene expression caused by increased activity of HDACs play a pivotal role in cancer development and progression [37]. Given the fact that the activity of HDACs is dysregulated in many types of cancers [38][39][40], HDACs have been considered as therapeutic targets for the treatment of neoplasms, indeed HDIs have become promising anti-cancer agents [41,42].
According to the Human Protein Atlas class I HDACs are expressed in variety types of tumors ( Figure 3) [43][44][45][46]. The data is presented by the percentage (%) of analyzed tumors with HDACs expression at high or medium level. HDAC2 was found expressed in 100% of multiple tumors. In the case of renal cancer HDAC2 is expressed 100% of cases, while other members do not exceed  (Figure 3) [44]. HDAC1 expression is comparatively high with exception of renal cancer and glioma [43]. The amount of patients with HDAC3 expression in ovarian cancer is significantly lower (20%) comparing to the other members of class I-all others with almost 100% incidence [45]. The last member of class I, HDAC8 is absent in colorectal, testis and breast cancer. Additionally, the score for patients with HDAC8 expression in liver cancer is notably lower (9%) compared to other members-HDAC1: 90%, HDAC2: 75% and HDAC3: 50% ( Figure 3) [43][44][45][46].
Cancers 2019, 11, x 4 of 31 [43]. The amount of patients with HDAC3 expression in ovarian cancer is significantly lower (20%) comparing to the other members of class I-all others with almost 100% incidence [45]. The last member of class I, HDAC8 is absent in colorectal, testis and breast cancer. Additionally, the score for patients with HDAC8 expression in liver cancer is notably lower (9%) compared to other members-HDAC1: 90%, HDAC2: 75% and HDAC3: 50% ( Figure 3) [43][44][45][46]. Class II HDACs are more varied than class I in terms of incidence in different cancers. High incidence of class IIa HDACs (HDAC4, HDAC5 and HDAC9) [47][48][49], is associated with colorectal and breast cancers. Interestingly, very low level of patients with HDAC5 (9%) and HDAC9 (8%) expression is observed in renal cancer ( Figure 4) [48,49]. The last member of IIa class-HDAC7 is associated mainly with lung cancer. Inhibition of HDAC7 results in restraining of lung cancer development [50]. HDAC10 (IIb class) is expressed in virtually all patients' tumors (100%) in every single analyzed type of cancer [51]. In contrast, the number of tumors with HDAC6 expression is very diverse in different types of cancer [52].
There is no available data in Human Protein Atlas regarding the expression of HDAC11 in different tumors. Yet, depletion of HDAC11 has an impact on cancer cells, including breast, ovarian, colon and prostate cells. HDAC11 is associated with apoptosis induction and inhibition of cell metabolic activity. Conversely, depletion of HDAC11 does not affect colon HCT-116 and prostate PC-3 cells [58]. Class II HDACs are more varied than class I in terms of incidence in different cancers. High incidence of class IIa HDACs (HDAC4, HDAC5 and HDAC9) [47][48][49], is associated with colorectal and breast cancers. Interestingly, very low level of patients with HDAC5 (9%) and HDAC9 (8%) expression is observed in renal cancer ( Figure 4) [48,49]. The last member of IIa class-HDAC7 is associated mainly with lung cancer. Inhibition of HDAC7 results in restraining of lung cancer development [50]. HDAC10 (IIb class) is expressed in virtually all patients' tumors (100%) in every single analyzed type of cancer [51]. In contrast, the number of tumors with HDAC6 expression is very diverse in different types of cancer [52].
Acetylation regulates also activity of the molecular chaperone Hsp90. Hsp90 has important role in maturation of many proteins, including the ligand-inducible transcription factor glucocorticoid receptor (GR). Specifically HDAC6 seems to be a regulator of Hsp90 acetylation [67,75]. Moreover, hypoxia-inducible factor 1 (HIF-1) can be acetylated by ARD1 protein acetyltransferase. HIF-1 plays a main role in cellular adaptation to changes in oxygen availability. ARD1-mediated acetylation strengthen interaction of HIF-1α with pVHL (the von Hippel-Lindau protein) and HIF-1α ubiquitination, suggesting that the acetylation of HIF-1α by ARD1 is critical to proteasomal degradation [76]. Transforming growth factor beta (TGFβ) regulates multiple cellular processes via activation of Smad signaling pathways. p300 acetylates Smad7 on two lysine residues. These lysine residues are critical for Smurf-mediated ubiquitination of Smad7. Moreover, acetylation protects Smad7 from TGFβ-induced degradation [77]. It has been also demonstrated that p300/CBP acetylates mastermind-like transcriptional coactivator-1 (Maml1), a Notch transcriptional co-factor, and thus regulates the strength of Notch-downstream signaling [78]. On the other hand, Notch signaling induces SIRT2 expression, which deacetylates and activates ALDH1A1 (aldehyde dehydrogenease), a marker commonly used to determine stem cells, particularly in breast cancer [79].

Epithelial-Mesenchymal Transition (EMT)
EMT is an essential physiological process during embryogenesis, histogenesis, organogenesis and wound healing. Yet, it can be also exploited during pathological processes such as fibrosis or tumor progression [98,99]. EMT is a reversible cellular process where epithelial cells acquire a mesenchymal-phenotype. Epithelial cells are connected by intercellular junctions such as: desmosomes, tight junctions (TJ) and adherent junctions (AJ), in contrast to mesenchymal cells, which do not cling to each other [100,101]. The consequence of EMT is disappearance of adhesion between epithelial cells through loss of junctions structures and apical-basal polarization. The new-formed mesenchymal cells acquire high migratory capabilities and invasive properties [102]. EMT is strongly associated with cancer metastasis as well as with presence of the circulating tumor cells (CTC). Moreover, EMT induces chemo-and radio-therapy resistance in many kinds of tumors [102,103]. Throughout EMT, the cancer cells endure frequently molecular events, for instance, a decrease of the level of epithelial markers (E-cadherin, cytokeratins) and an increase of the level of mesenchymal markers (N-cadherin, vimentin) ( Figure 3). Expression of EMT markers in primary tumors has been linked with cancer progression and poor medical prognosis [104,105].
EMT is induced by growth factors including: transforming growth factor, hepatocyte growth factor, epithelial growth factor, fibroblast growth factor and insulin growth factor. All of them indirectly modify EMT transcription factors [106] (EMT-TFs), including: SNAIL and ZEB1/ZEB2 families, as well as TWIST1/TWIST2. Vertebrates have 3 SNAIL family members: SNAIL1, SNAIL2 and SNAIL3. These TFs have a highly conserved C-terminal organization, able to recognize and bind to the E-cadherin promoter. Moreover, the N-terminal domain of SNAIL (SNAG) interacts with transcriptional co-repressors, including Sin3A/HDAC1/2 complex and polycomb complex 2. Hence, the activation of SNAIL/SLUG promotes E-cadherin gene (CDH1) downregulation and contributes to an increase of cell migration and invasion [106,107]. The ZEB family of TFs downregulates CDH1 expression and upregulates mesenchymal markers such as N-cadherin gene (CDH2), vimentin and fibronectin. ZEB members are also responsible for increase of cell migration and invasion [108]. TWIST1 is able to simultaneously upregulate CDH1 and downregulate CDH2 expression. Post-transcriptional gene expression is regulated by small non-coding RNAs, such as: miRNA-200 and miRNA-34. Where epithelial cells express miRNA-200 and miRNA-34 whilst mesenchymal cells do not [109].
The balance between EMT and MET processes regulates cell plasticity [110]. However, nowadays an intermediate stage between fully-epithelial and fully-mesenchymal states has been recognized-hybrid E/M state. The identification of EMT/MET or hybrid E/M states is difficult to observe because these processes run smoothly and interchangeably [110] (Figure 10). Cancer cells with hybrid E/M phenotype have cell-cell adhesion properties as well as migration abilities, simultaneously [109]. Recent data suggest that cells with E/M hybrid phenotypes show stronger metastatic properties as well as survival in circulation [111,112]. Hybrid E/M cells are similar or more resistant to drug-treatments in comparison to fully EMT cells [111]. associated with cancer metastasis as well as with presence of the circulating tumor cells (CTC). Moreover, EMT induces chemo-and radio-therapy resistance in many kinds of tumors [102,103]. Throughout EMT, the cancer cells endure frequently molecular events, for instance, a decrease of the level of epithelial markers (E-cadherin, cytokeratins) and an increase of the level of mesenchymal markers (N-cadherin, vimentin) ( Figure 3). Expression of EMT markers in primary tumors has been linked with cancer progression and poor medical prognosis [104,105]. EMT is induced by growth factors including: transforming growth factor, hepatocyte growth factor, epithelial growth factor, fibroblast growth factor and insulin growth factor. All of them indirectly modify EMT transcription factors [106] (EMT-TFs), including: SNAIL and ZEB1/ZEB2 families, as well as TWIST1/TWIST2. Vertebrates have 3 SNAIL family members: SNAIL1, SNAIL2 and SNAIL3. These TFs have a highly conserved C-terminal organization, able to recognize and bind to the E-cadherin promoter. Moreover, the N-terminal domain of SNAIL (SNAG) interacts with transcriptional co-repressors, including Sin3A/HDAC1/2 complex and polycomb complex 2. Hence, the activation of SNAIL/SLUG promotes E-cadherin gene (CDH1) downregulation and contributes to an increase of cell migration and invasion [106,107]. The ZEB family of TFs downregulates CDH1 expression and upregulates mesenchymal markers such as N-cadherin gene (CDH2), vimentin and fibronectin. ZEB members are also responsible for increase of cell migration and invasion [108]. TWIST1 is able to simultaneously upregulate CDH1 and downregulate CDH2 expression. Posttranscriptional gene expression is regulated by small non-coding RNAs, such as: miRNA-200 and miRNA-34. Where epithelial cells express miRNA-200 and miRNA-34 whilst mesenchymal cells do not [109].
The balance between EMT and MET processes regulates cell plasticity [110]. However, nowadays an intermediate stage between fully-epithelial and fully-mesenchymal states has been recognized-hybrid E/M state. The identification of EMT/MET or hybrid E/M states is difficult to observe because these processes run smoothly and interchangeably [110] (Figure 10). Cancer cells with hybrid E/M phenotype have cell-cell adhesion properties as well as migration abilities, simultaneously [109]. Recent data suggest that cells with E/M hybrid phenotypes show stronger

EMT and Cancers
EMT is the result of a series of epigenetic changes including chromatin remodeling and histone modifications. Acetylation and metylation of histones play an important role in tumor progression [41]. For this very reason, HDIs are considered as modifiers of EMT-related factors expression, although this effect is cancer-type dependent [113]. Hereby we will analyse the available data on HDIs in different tumor types.

Lung Cancer
HDIs have been investigated for their roles as inhibitors of migratory potential. Indeed, TSA inhibits migration of irradiated human epithelial A549 lung cancer cells through decreasing of SNAIL and ZEB expression. The expression of E-cadherin and N-cadherin in irradiated-cells treated with TSA are inverted as compared to radiation-only pretreatment. Radiation-TSA treatment also resulted in upregulation of ZO-1 and β-catenin (epithelial markers), compared with alone-radiation pretreatment [114]. Moreover, it has been shown that the inhibitory effect for TGF-β1-induced EMT in irradiated A549 cells pretreated with TSA is connected with inhibiting of SNAIL and SLUG activity [115]. TSA supported with silibinin, a natural flavanone compound from silymarin, significantly increases E-cadherin level by downregulation of ZEB1, while silibinin alone is not able to silence E-cadherin expression in non-small cell lung cancer (NSCLC H1299 cells). Interestingly, the level of E-cadherin after 48 h of TSA+silibinin treatment was significantly restored, compared with the level of E-cadherin after 48 h of TSA-alone [116]. VPA was able to partially inhibit EMT in A549 cells, through decreasing histone deacetylation level. Additionally, the cellular spindle-shape effect, which is characteristic for mesenchymal cells, induced by TGF-β is reduced after VPA treatment. Although there is no direct interaction between VPA and TGF-β1 [117].
At odds with other HDIs, SAHA-treated A549 cells responded by decreasing of E-cadherin expression and increasing of vimentin expression, with the acquisition of a mesenchymal phenotype. The E-cadherin downregulation is inversely correlated to SLUG expression [118]. Nevertheless, SAHA, as well as panobinostat, induce upregulation of GAS5-AS1 expression in a dose-dependent manner in NSCLC cells, which is connected with inhibiting migration of NSCLC cells [119] (Table 2).

Hepatocellular Carcinoma
TSA, VPA, SAHA and MS-275 have a strong positive influence on EMT, through decreasing E-cadherin expression and increasing N-cadherin expression in HepG2 cells. In turn, mesenchymal markers such as vimentin, TWIST and SNAIL become more abundant [120]. In the same vein, SAHA and sodium butyrate (NaB) have been investigated as suppressors for cells proliferation in dose-dependent manner. Both of them significantly increase N-cadherin, vimentin, fibronectin and SNAIL expression in HepG2 cells. SNAIL upregulation is connected with phosphorylation of SMAD2/3 by these HDI. Additionally, SAHA and NaB are able to promote SNAIL and vimentin expression in xenografs [121]. Yet, panobinostat (LBH589) elevates E-cadherin expression in HCC-LM3 and HepG2 cells while decreases N-cadherin, vimentin and TWIST1 simultaneously [122]. Likewise, reminostat acts as an upregulator of E-cadherin expression and down-regulator of vimentin, TWIST1 and SNAIL in HLE cells [123] (Table 2).

Cholangiocarcinoma
VPA or TSA increased both E-cadherin and vimentin expression but inhibited invasion and migration of HuCC-T1 cholangiocarcinoma cells. Additionally, HuCC-T1 cells co-treated with gemcitabine and VPA or TSA showed higher E-cadherin, vimentin and ZO-1 levels as well as decreased migration and invasion.

Pancreatic Cancer
HDAC inhibition by domatinostat (4SC-202) results in downregulation of E-cadherin with the concomitant upregulation of N-cadherin in Panc-1 cells, but unexpectedly the downregulation of other mesenchymal markers such as ZEB1, SNAIL and vimentin, or TGF-β-induced SMAD2 phosphorylation.
Conversely, (3R)-2-(biphenyl-4-ylsulfonyl)-1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid (BSI) increases E-cadherin expression while decreases N-cadherin and SNAIL expression after 24h in Panc-1 cells. Interestingly, the level of E-cadherin remains unchanged, although the level of N-cadherin and SNAIL was decreased in BxPC-3 cells after BSI treatment. Moreover, BSI is strongly associated with partial inhibition of invasion and migration in Panc-1 cells after 24 h. BSI reduces tumor spheres formation in BxPC-3 cells, while in Panc-1 cells spheres formation is unchanged but their size was significantly decreased [126]. It has been found that another interesting agent, mocetinostat inhibits ZEB1 expression and increases E-cadherin and miR-203 upregulation in Panc-1 cells as well as in hPaca1-derived tumor cells. Paradoxically Panc-1-tumor xenografts grew bigger by mocetinostat treatment while the combination with gemcitabine resulted in a synergetic effect in tumor growth inhibition [127]. SAHA, on the other hand, inhibits proliferation in pancreatic CSCs. SAHA is able to increase miR-34a expression in pancreatic CSCs as well as in ASPC-1 and able to increase miR-34a expression in pancreatic CSCs as well as in ASPC-1 and MiaPaCa-2 cell lines. SAHA induces E-cadherin overexpression and N-cadherin downregulation in pancreatic CSCs, simultaneously. Moreover, SAHA, as well as resveratrol, significantly downregulates ZEB1, SNAIL and SLUG expression in pancreatic CSCs. Additionally, resveratrol inhibited the invasion and migration of pancreatic CSCs. Resveratrol was able to inhibit the growth of pancreatic cancer in KrasG12D mice [12,144] (Table 2).

Colorectal Cancer
TSA has been studied for its effects on SW480 colorectal cancer cells. TSA decreased the expression of SLUG, leading to the reversal of the EMT process and attenuation of invasion and migration of SW480 cells. It has been suggested that TSA causes EMT reversion by increasing of E-cadherin and decreasing of vimentin expression [128]. Au contraire, treatment with VPA significantly stimulates migration and invasion in vitro, argubly by activation of EMT in HCT116 and SW480 human colorectal cancer cell lines, resulting in downregulating the epithelial markers: E-cadherin and ZO-1 and upregulating the mesenchymal markers: N-cadherin and fibronectin in both HCT116 and SW480 cells as well as upregulating the vimentin only in HCT116 cells. In line with this, VPA significantly promotes the expression of SNAIL via Akt/GSK-3b signal pathway. Suppression of SNAIL significantly reduced E-cadherin and increase of vimentin or fibronectin expression in both HCT116 and SW480 cells [128]. In fact, other HDIs also block EMT or induce MET, such as compound-11, who has also been found to induce MET in HCT116 and HT29 colorectal cancer cells, as well as in the HCT116 xenograft model. It has been observed that compound-11 induced downregulation of N-cadherin, vimentin and p-FAK (invasive marker), while E-cadherin was increased, through downregulation of Akt, which seems to be crucial for EMT in colorectal cancer cells [129]. Nevertheless, the oppsite has also been observed using TSA and VPA individually or in combination with TGF-β1 in four colon carcinoma cell lines including: SI cells (DLD1 and HCT116) and MSS cells (HT29 and SW480). The results revealed that the morphological changes were similar pursuing TSA or VPA with or without TGF-β1 co-treatment. CRC cell lines were altered to spindle-like morphology. Subsequent analyses showed a decrease in E-cadherin expression with TSA or VPA treatments in HCT116, DLD1 and SW480 cells. Vimentin was increased by treatment with the HDIs together with TGF-β1 in the four carcinoma cell lines. Consistently, TSA or VPA induced increased cell migration and invasion abilities. All together, treatment by TSA or VPA in combination with TGF-β1 seem to intensify EMT and migration in colon carcinoma cells. Moreover, in the MSS cells (HT29 and SW480) the EMT process was enhanced by TGF-β1 and was much more intense than in the MSI cells (DLD1 and HCT116) [15] (Table 2).

Renal Cancer
VPA or MS-275 treatment resulted in cell morphology alternation and a reduction in migration of in Renca cells as compared to untreated Renca cells. At the molecular level, TWIST1 was upregulated and TWIST2 was downregulated after MS-275 treatment in time-dependent manner. Moreover, SNAIL2 expression was increased, while SNAIL1 expression was unchanged after 48 h of 5 µL MS-275 treatment. Additionally, ZEB2 was significantly increased in dose-dependent manner, while ZEB1 remained unchanged after 48h of MS-275 treatment. VPA and MS-275 hardly decrease β-catenin expression. Moreover, both of them upregulated E-cadherin levels after 48h. Interestingly, VPA significantly increased the growing rate of Renca cells resulting in phenotypical changes in cell morphology. Untreated Renca cells had a cobblestone-like morphology. HDIs treatment altered their morphology to a scattered pattern, with interspaces between cells. These cells displayed a star-shaped cell body resembling EMT-associated growth [130]. MS-275 as well as TSA significantly increased E-cadherin expression in TGF-β1-treated HK2 cells. On the other hand, neither of them had no influence on N-cadherin expression in the same cells [131]. Moreover, TSA increases E-cadherin expression without any effect on SMAD2 and SMAD3 phosphorylation in RPTEC cells [145]. PCI34051 resulted in no changes on N-cadherin and E-cadherin expression in HK2 cells. Finally, LMK235 was able to restore E-cadherin expression in HK2 cells, which is downregulated by TGF-β1 [131] (Table 2).

Prostate Cancer
AR-42 inhibited migration and invasion of Ace-1 cells caused apoptosis and decreased PCa cells bone metastasis. Moreover, AR-42 decreased E-cadherin, N-cadherin, TWIST, MYOF, and osteomimicry genes expression as well as anoikis (apoptosis induced by lack of correct cell/ECM attachment) resistance, while it increased SNAIL, PTEN, FAK and ZEB1 transcription factors expression in Ace-1 cells. In addition, AR-42 downregulated the PCa metastasis to bone in nude mice. In addition, there has been observed an alteration of the spindle-like morphology to irregular shape of PCa cells, in both in vitro and in vivo conditions after AR42 treatment [133]. Furthermore, treatment with another HDI-SAHA-repressed EMT in LNCaP prostate cancer. It has been reported that SAHA downregulates FOXA1 expression. FOXA1 inhibits EMT in prostate cancer by decreased expression of SLUG transcription factor and repression of the neuroendocrine (NE) differentiation markers. SAHA also decreases NKX1 and PSA, which is another transcription factor and antigen, respectively. SAHA, like other pan-HDAC inhibitors (inter alia TSA), induces EMT by elevated protein levels such as SLUG, ZEB1 and vimentin. The same results were obtained using TSA treatment in LNCaP cells, however treatment with RGFP966, with the same panel of EMT markers was inefficient both in cell migration or invasion. SAHA and TSA induced cell migration, while RGFP966 was innocuous [134]. Treatment with LBH589 suppresses HMGA2 expression, decreases epithelial-mesenchymal plasticity in vitro and drastically decreases tumor growth and metastasis in vivo. Notably, in mice treated with LBH589 in combination with orchiectomy, there was an increase of p53 and androgen receptor (AR) acetylation, which in turn prevents the development of mCRPC and considerably extends life after castration [146]. In addition, TSA reverts EMT by a time-dependent upregulation of E-cadherin and downregulation of vimentin in PC3 prostate cancer cells. Moreover, TSA it has supressed SLUG expression which consequently prompted MET, as well as decreased cell invasion and migration abilities [13]. Likewise, VPA inhibited EMT by upregulation of the expression of E-cadherin, and concomitant suppression of the migration and invasion of prostate cancer cells [11] (Table 2).

Breast Cancer
SAHA inhibits EMT and chemoresistance induced by TGF-β1 in MzChA-1 and TFK-1 breast cancer cells. In both of these cell lines, TGF-β1 caused morphological changes from valvate-like to spindle-like shapes, as well as downregulation of E-cadherin and upregulation of N-cadherin, vimentin and SNAIL expression that the mechanism of SAHA's effect seems to be the inhibition of p-SMAD2, p-SMAD3 and SMAD4 nuclear translocation induced by TGF-β1 in MzChA-1 cells, as well as the attenuation of the binding affinity of SMAD4 to the E-cadherin-related TWIST, SNAIL, SLUG, ZEB1 and ZEB2 transcription factors [135]. Yet, SAHA can promote migration and EMT via HDAC8/FOXA1 signals in MDA-MB-231 and BT-549 breast cancer cells. SAHA significantly downregulated the expression of E-cadherin and upregulated the mesenchymal markers: N-cadherin, vimentin and fibronectin. However, SAHA had no effect on the nuclear translocation or expression of SNAIL, SLUG, TWIST and ZEB [136]. MDA-MB-231, BT-549 and MCF-7 breast cancer cells were incubated with LBH589, another HDI, and examined for changes in cell morphology, migration and invasion in vitro. LBH589 reversed EMT, measured by the altered morphology and gene expression of triple negative breast cancer (TNBC). E-cadherin expression was significantly upregulated by LBH589 treatment in the two TNBC lines (MDA-MB-231 and BT-549), while no change was observed in the ER-positive (MCF-7) cells. Additionally, expression of ZEB1 and ZEB2 were significantly inhibited upon LBH589 treatment in both the MDA-MB-231 and BT-549 TNBC cell lines, while no changes were detected in the MCF-7 cells. The above-described alterations in EMT gene expression correlated with diminished cell migration and invasion in TNBC cells in vitro, as well as meaningful inhibition of TNBC cell metastasis to lung and brain in a xenograft model [138]. Treatment of MDA-MB-231 and Hs578T cells with entinostat (ENT) caused upregulation of CDH1 and downregulation of CDH2 and VIM mRNA expression. Moreover, chromatin immunoprecipitation (ChIP) assay revealed that the treatment of MDA-MB-231 and Hs578T cells with ENT increased the reduced the association of SNAIL and TWIST to the CDH1 promoter and downregulated both the SNAIL and TWIST expression which resulted in higher E-cadherin expression. Moreover, ENT inhibited migration of MDA-MB-231 and Hs578T breast cancer cells and induced MET [14]. The HDAC1 and HDAC3 inhibitor-MS-27-sensitized tumor necrosis factor-related apoptosis-inducing ligand (TRAIL)-resistant breast cancer MDA-MB-468 cells, inhibited angiogenesis and metastasis, and reversed EMT in vivo in xenografted BALB/c nude mice. MS-275 upregulated the expression of E-cadherin and downregulated the expression of N-cadherin, as well as ZEB1, SNAIL and SLUG transcription factors in tumor tissues. Treatment of mice with TRAIL alone had no effect on the expression of these markers. Co-treatment of MDA-MB-468 cells with MS-275 and TRAIL had similar effects to those of MS-275 [139]. Yet, it is in breast cancer where contradicting data exist. For example, SUM159 and MDA-231 cells treated with VPA or SAHA become more stem-like by dedifferentiation. These dedifferentiated cells have a higher migration potential and are more resistant to taxol. HDIs-treated cells presented upregulation of several mesenchymal markers such as vimentin, N-cadherin, fibronectin and tenascin-C while epithelial marker E-cadherin was not detected. Yet, several other mesenchymal markers such as SNAIL (after SAHA treatment), FOXC2 and ZEB1 (after SAHA and VPA treatment) were downregulated. HDACs inhibition resulted in the activation of the Wnt/β-catenin signaling, which seems to be responsible for these phenotypical changes [137] (Table 2).

Ovarian Cancer
Effects of TSA alone or in combination with cisplatin were investigated in SKOV3 cell line in vitro. SKOV3 cells showed downregulation of both E-cadherin and N-cadherin with exposure to TSA alone or in combination with cisplatin. Moreover, mouse xenografts were used to assess the anti-cancer activity of sequential cisplatin followed by TSA treatment. Such treatment significantly suppresses tumorigenicity of HEY xenografts through downregulation of N-cadherin and Snail, Slug, Twist transcription factors, as well as upregulation of E-cadherin expression [140] (Table 2).

Head and Neck Cancer
In gefitinib-resistant Hep-2 and KB squamous cell carcinoma of head and neck cells, SAHA reverted EMT by a time-dependent upregulation of E-cadherin and β-catenin and downregulation of vimentin. Moreover, there has been observed a reduction of the spindle-like morphology, characteristic for mesenchymal cells, with the acquisition of epithelial morphology, in both Hep-2 and KB cells after SAHA treatment [141]. VPA also induced a reversal of the mesenchymal phenotype caused by TGF-β1 or irradiation in TE9 cancer cells, resulting in an increase of cell migration and invasion. TE9 cancer cells pre-treated with VPA exhibited less inhibition of E-cadherin expression and no increase of vimentin expression as compared with untreated cells stimulated by TGF-β1 or irradiation. VPA inhibited the phosphorylation of SMAD2 and SMAD3 and downregulated TWIST, SNAIL, and SLUG transcription factors expression which previously were increased by TGF-β1 or irradiation stimulation [142] (Table 2).

Malignant Glioma
Panobinostat (LBH589) combined with temozolomide and irradiation stimulation significantly decreased vasculogenic mimicry (VM) formation, migration and invasion as well as increased E-cadherin expression in U251 glioma cells compared with temozolomide and irradiation stimulation without LBH589 treatment [143] (Table 2).

Discussion
Virtually all HDIs block multiple HDACs, which in turn have each multiple protein targets and thus the resulting gene expression changes are not the direct targets of HDIs but rather the downstream effects lack of different HDACs activities. This is probably why in many cases both epithelial as well as mesenchymal markers increase in expression. Since each HDI has different HDAC targets, the downstream results might be substantially different depending on the expression of HDACs, their cellular roles, as well as the expression of other co-factors or orthogonal signaling pathways.
Since in tumor cells often epigenetically inactivate non-beneficial genes, while activate those that provide them with evolutionary advantages, changes in gene expression can result in a beneficial effect due to the re-expression of tumor suppressor [147] and the transcriptional silencing of pro-oncogenes [148,149]. Moreover, changes in acetylation of non-histone proteins might also provide beneficial in many cases, such as the reduction pro-oncogenic or pro-survival signaling pathways.
The search for inhibitors of single HDACs might provide us with a clearer picture of why there are contradictory results depending of the cell type. Also, there is need to understand the effects of HDIs in different cancer cells in light of their HDAC expression patterns and genomic, as well as epigenomic, landscapes.
While there is evidence that de-novo gene expression, in particular of epithelial-like genes, is a beneficial result for the treatment, this very effect of randomly re-opening chromatin might have two other effects: (1) turning on oncogenes or transposons, the latter related to the next point; and (2) genomic restructuring which might have effects in genomic stability. Genomic stability should be carefully considered in future studies using HDIs as long-term treatment with these drugs might result in genomic instability and acquisition of mutation in re-opened chromatin. There is evidence that such might occur, promoting more aggressive tumors [150]. This also prompts that the length of HDIs treatments should be carefully analysed.

Conclusions
HDACs play a pivotal role in the progression of cancers by reversible modulation of acetylation status of histone and non-histone proteins. However, the exact function of HDACs as a central mediator of tumorigenic capacity still remains unclear. There are abundant pre-clinical and clinical studies examining the effects of HDIs alone or in combination with other anti-cancer agents. The impact of HDIs on EMT differ greatly in various types of cancers (Table 3). VPA stimulates EMT in hepatocellular [120], breast [112] and colorectal cancers [128] while the same active agent inhibits this process in lung [117], prostate [11] and head and neck carcinomas [142]. Moreover, the impact of VPA on EMT in renal cancer is unclear. After VPA treatment expression both epithelial and mesenchymal markers decreased, moreover migration of renal cells was inhibited [130]. In lung [33,114,115] and breast cancers [14] most HDIs inhibited EMT with the exception of SAHA, which stimulated this process in both mentioned above types of cancer. LBH589 suppresses EMT in breast cancer [14] and hepatocellular carcinoma [122]. However, in hepatocellular carcinoma most of HDIs (VPA, SAHA, TSA, MS-275 [120], NaB [121]) stimulated EMT, excluding LBH589 [122] and RAS2410 [123] which inhibited this process. In head and neck cancer [141,142] as well as prostate cancer all analyzed HDIs [11,13] inhibited EMT. Existing data about the effects of HDIs on EMT are conflicting. Therefore, there is an urgent need to comprehensively study the mechanisms of action and role of HDIs on the EMT/MET processes in different cancer types at different stages of carcinogenesis. Side effects of application of these compounds also should be noted. More studies are needed to establish the best strategy to incorporate these agents into the therapy of patients with cancers, minimizing toxicity and maximizing clinical benefits. Clarification and validation of the detailed mechanisms of HDIs action will provide a bright future for the use of HDIs as one of the important tools in the fight against cancers. Table 3. The effect of histone deacetylase inhibitors (HDIs) on the epithelial-mesenchymal transition (EMT) process in tumors.