Therapeutic implications of cancer epithelial-mesenchymal transition (EMT)

The epithelial-mesenchymal transition (EMT) comprises an essential biological process involving cancer progression as well as initiation. While the EMT has been regarded as a phenotypic conversion from epithelial to mesenchymal cells, recent evidence indicates that it plays a critical role in stemness, metabolic reprogramming, immune evasion and therapeutic resistance of cancer cells. Interestingly, several transcriptional repressors including Snail (SNAI1), Slug (SNAI2) and the ZEB family constitute key players for EMT in cancer as well as in the developmental process. Note that the dynamic conversion between EMT and epithelial reversion (mesenchymal-epithelial transition, MET) occurs through variable intermediate-hybrid states rather than being a binary process. Given the close connection between oncogenic signaling and EMT repressors, the EMT has emerged as a therapeutic target or goal (in terms of MET reversion) in cancer therapy. Here we review the critical role of EMT in therapeutic resistance and the importance of EMT as a therapeutic target for human cancer.


Introduction
Traditionally, the concept of epithelial to mesenchymal transition (EMT) was mainly restricted to the reversible phenotypic transformation of epithelial to mesenchymal cells seen during development, wound healing and various diseases (Thiery et al. 2009). Reversible rounds of EMT and MET (mesenchymal to epithelial transition; the reverse state of EMT) constitute the physiologic stages of embryonic development with intense cell plasticity (Kalluri and Neilson 2003;Thiery et al. 2009). Pathological hyperactivated EMT has been widely described in tumor progression as invasive and metastatic cellular dissemination in the altered epithelium (Thiery 2002;Polyak and Weinberg 2009). After the single tumor cells are disseminated into the lymphatics or vasculature, MET is thought to be responsible for metastatic re-colonization in distant organs by regaining epithelial differentiation and proliferation ability (Chaffer et al. 2007).
The hallmark of EMT is the loss of epithelial marker expression, typically indicated by the presence of E-cadherin, with a gain in mesenchymal marker expression such as of N-cadherin and vimentin accompanied by invasive phenotype (Fig. 1). Epithelial cells are determined by apico-basal polarity, attachment to the basal lamina and tight cell-to-cell junctions, while mesenchymal cells are characterized as detached arrangements in the stroma with front-back polarity and migration ability (Thiery et al. 2009;Gonzalez and Medici 2014). The 'classic binary EMT' concept has been extended to a more comprehensive view of tumor progression, augmenting the traditional 'full phenotypic conversion' during invasion and metastasis (Hay 1995). EMT is now considered a more flexible status of plasticity known as the 'partial EMT program' rather  (Nieto et al. 2016). Transcriptional factors that directly or indirectly downregulate E-cadherin expression and induce EMT are known as EMT-activating transcriptional factors (EMT-TFs). EMT-TFs are tightly regulated via oncogenic signaling pathways consisting of non-coding RNAs, extracellular mediators and translational/post-translational regulations. In this review, we will summarize the pivotal molecular signaling pathways and therapeutic targets in cancer EMT.
EMT-TFs (Snail in particular) are highly expressed in neoplastic nuclear tissue, adjacent stroma and peritumoral inflammation (Come et al. 2006;Franci et al. 2009;Tuhkanen et al. 2009;Bezdekova et al. 2012). Both Snail and ZEB are known to have correlation to clinical prognostic factors, e.g. aggressiveness, metastasis and poor survival, and have been reported to have chemoresistance for cisplatin, 5-fluorouracil, gefitinib, or doxorubicin (Witta et al. 2006;Arumugam et al. 2009;Hsu et al. 2010;Tryndyak et al. 2010;Chang et al. 2011;Haslehurst et al. 2012;Zhang et al. 2012). For example, induction of Snail significantly increases therapeutic resistance against the paclitaxel (taxol) treatment in breast cancer cells (Fig. 2), indicating that a higher Snail abundance is intimately associated with a worse therapeutic outcome of human cancer.
Recent immunotherapeutic approaches have provided remission in a significant number of cancer patients those were considered lethal previously. Not surprisingly, extensive molecular reprogramming by EMT and subsequent tumor microenvironment remodeling involves in

The molecular signaling pathways of EMT
Intensive study over the past 2 decades has revealed many upstream regulators and oncogenic signaling pathways regulating EMT-TFs (Fig. 3). While many of these constitute EMT regulatory pathways [e.g. Wnt, Notch, Hedgehog (HH), Transforming growth factor beta (TGFb), Receptor tyrosine kinase (RTK)s, and others], canonical Wnt and TGF-b play critical roles in human malignancy and degenerative diseases. The canonical Wnt signaling pathway induces Snail transcription and EMT via the b-catenin/T-cell factor (TCF)/LEF transcriptional complex. In normal condition, b-catenin is phosphorylated and consecutively degraded to maintain constant intracellular levels. The b-catenin phosphorylation is generally thought to result from a multiprotein 'destruction complex', which involves kinase proteins [glycogen synthase kinase-3b (GSK-3b) and casein kinase 1 (CK1)] and scaffold proteins [Axin and adenomatous polyposis coli (APC)] in b-catenin dynamics (Fodde and Brabletz 2007). However, it should note that recent studies have supported the oncogenic role of Axin2 (Yochum 2012). Similar to b-catenin, the Snail harbors the GSK-3-dependent phosphorylation motif, ubiquitination and proteasomal degradation (Yook et al. 2005). Accordingly, canonical Wnt signaling inhibits Snail degradation via Axin2-mediated nuclear export of GSK-3b (Yook et al. 2006). Furthermore, the Wnt/b-catenin axis promotes Slug activity and BRCA1 downregulation in breast cancer cells (Wu et al. 2012). Note that Axin2 and Snail are also highly expressed in precancerous colorectal adenomas and oral precancerous lesions (Lustig et al. 2002;Kroepil et al. 2012;Zhang et al. 2017).
Notch signaling directly activates EMT with nuclear translocation of the Notch intracellular domain (NICD). Jagged-2 (JAG2) and Notch interaction can extend to NCID cleavage by A Disintegrin and Metalloprotease (ADAM) protease and c-secretase (Rizzo et al. 2008). NICD can bond to the Snail promoter and directly stimulates its expression (Wang et al. 2010). In addition, Notch can indirectly stabilize Snail1 activation by hypoxia-inducible factor 1a (HIF-1a) binding to the lysyl oxidase (LOX) promoter in hypoxic conditions (Sahlgren et al. 2008).
Hedgehog pathway, specifically the Shh, has been reported to contribute to tumor progression. Shh ligand LGR low-density lipoprotein receptor-related protein, FRZ frizzled, Dvl dishevelled, CK-1 Casein kinase 1, GSK-3b glycogen synthase kinase-3b, APC adenomatous polyposis coli, JAG2 Jagged-2, NICD notch intracellular domain, Shh sonic hedgehog, Smo smoothened, PTCH patched homologs, SUFU suppressor of fused homolog, Gli Glioma, TGF-b transforming growth factor beta, PI3K phosphoinositide 3-kinase, ERK extracellular signal-regulated kinase, RTK receptor tyrosine kinase, NF-jB nuclear factor kappa-light-chain-enhancer of activated B cells, EMT-TFs EMT-transcriptional factors, ZEB zinc finger E-box-binding homeobox, miR micro RNA binds to patched homolog (PTCH) receptors to activate Smoothened (Smo), which stimulates glioma (Gli) transcription factors (Briscoe and Therond 2013). Abberant Gli1 expression induced Snail and decreased E-cadherin expression in skin cancer . Moreover, Shh signaling has been seen to be associated with increased Snail expression in neuroendocrine tumor (Fendrich et al. 2007). Together with Wnt signaling, the TGF-b family is a representative EMT inducer in both development and disease through SMAD and non-SMAD regulated pathways. TGF-b family ligands consist of three isoforms of TGF-b and several BMPs, but so far only TGF-b1 is known to participate in cancer (Akhurst and Derynck 2001). TGF-b1 ligand and TGF-b receptor (TbR) II binding establish a phosphorylation cascade of TbR I and SMAD2/3 oligomerized with SMAD4 as a trimer that enters the nucleus and leads to EMT gene reprogramming (Akhurst and Derynck 2001;Valcourt et al. 2005;Derynck et al. 2014). SMAD3/4 complex interacts with Snail at the nucleus in a SMAD3-dependent manner, suppressing E-cadherin and occludin expression (Vincent et al. 2009), although the molecular interaction between Snail and SMAD has not yet been clearly identified.
Various growth factors bind to receptor tyrosine kinases (RTKs) to promote EMT and tumorigenesis. RTKs can enhance TGF-b1 expression in a positive feedback loop of autocrine signaling (Bennasroune et al. 2004;Lamouille et al. 2014). Epidermal growth factor (EGF) ligand can induce b-catenin and Snail transcription with loss of E-cadherin (Lu et al. 2003). Human EGFR 2 (HER2, a type of EGFR also named ERBB2) activation stimulated EMTmediated tumor progression and recurrence in mammary epithelial cells (Moody et al. 2005). Fibroblast growth factor-1 (FGF-1) and hepatocyte growth factor (HGF) increased Slug expression and desmosome dissociation in rat bladder carcinoma cells (Savagner et al. 1997). HGF binds with its RTK c-MET to induce Snail and Slug activation via the MAPK/ERK pathway (Grotegut et al. 2006). Insulin-like growth factor 1 (IGF1) is known to promote EMT by activating NF-jB-dependent Snail1 activity or MAPK/ERK-mediated ZEB1 activity (Kim et al. 2007;Graham et al. 2008). In addition, IGF1 or an EGF ligand/ RTK bond stimulates EMT and cell migration with AKT and ERK pathway crosstalk (Irie et al. 2005). It should be noted that although targeted therapeutics against RTKs are in development worldwide in preclinical and clinical trials, the EMT is considered neither a biomarker nor a therapeutic outcome of RTK inhibitors.

Non-coding RNA regulation in EMT
In addition to gene-level regulation by EMT-TFs, RNAlevel regulation contributes to EMT mechanisms. In particular, non-coding RNAs known as microRNAs (miR-NAs) are known for their either positive or negative interaction with EMT. The most famous two classes of miRNA known in EMT research are the miR-34 and miR-200 families, which act as tumor suppressors (Korpal et al. 2008;Kim et al. 2011a). The p53/miR-34 axis and Wnt/bcatenin pathway are linked in a tumor suppressor manner (Kim et al. 2011a). Moreover, Snail and ZEB1 have a double-negative feedback loop with miR-34 as well (Siemens et al. 2011). Interestingly, miR-34 targets the conserved sites at the untranslated regions of Wnt and Snail, revealing a close connection between p53 tumor suppressor and canonical Wnt signaling (Kim et al. 2011a;Siemens et al. 2011).

Therapeutic development in EMT
Two major approaches to specific targeting of RTKs or oncogenic signaling are monoclonal antibodies (mABs) and small molecule inhibitors (Imai and Takaoka 2006). We will review several, but not all, representative therapeutics controlling EMT that have been approved, are under clinical trial or are of significant importance (specific details are mentioned in Table 1).
The TGF-b signaling pathway is a well-known target of EMT in various malignancies, its inhibitors having been widely investigated. Due to the importance of TGF-b signaling in heart development, cardiovascular complications are an important concern in drug development (Anderton et al. 2011). Multiple small molecule inhibitors for TGF-b ligands and receptors are currently going through clinical  (Morris et al. 2014). In another study, fresolimumab resolved skin lesions associated with malignant melanoma as well as renal cell carcinoma (Lacouture et al. 2015). Activin receptor-like kinase 1 (ALK1) is a TGF-b serine/ threonine kinase receptor. A mAB of ALK1, named PF-03446962, has been through clinical trials for several types of cancer with beneficial results for phase I, but not phase II (Goff et al. 2016;Simonelli et al. 2016;Wheatley-Price et al. 2016). Several TGF-b I/II receptor kinase inhibitors are in pre-clinical research, namely EW-7203, EW-7197, and IN-1130(Park et al. 2011Son et al. 2014). A small molecule inhibitor which inhibits TbRI, galunisertib (LY-2157299), was in phase Ib for metastatic or locally advanced pancreatic cancer and phase I and II for malignant glioma (Ikeda et al. 2017). It was first proposed for chemotherapy-resistant breast cancer which would not react to paclitaxel (Bhola et al. 2013). Galunisertib was reported to be free from cardiovascular adverse effects and benefitted near 1/5 of the glioma patients enrolled (Rodon et al. 2015). Importantly, recent observations highlight the importance of TGF-b inhibitors (galunisertib or TGF-b blocking antibody) in combination with anti-PD1 (Programmed cell death protein 1)/PD-L1 (Programmed deathligand 1) immunotherapy (Mariathasan et al. 2018;Tauriello et al. 2018). Tyrosine kinase inhibitors (TKIs) for RTKs are under active pre-clinical research and clinical assessment. TKIs may regulate a single RTK or simultaneously influence multiple RTKs. Representative US Food and Drug Administration-(FDA) approved TKIs for EGFR, gefitinib (Iressa Ò ) and erlotinib (Tarceva Ò ), achieved a 70% response rate in lung cancer (Mayo et al. 2012). They have been efficient in non-small cell lung cancer and ovarian cancer (Murphy and Stordal 2011). EGFR-targeted lapatinib (Tykerb Ò ) and multi-TKI sorafenib (Nexavar Ò ), and vandetanib (Caprelsa Ò ) have been approved and provided for various malignancies (Wilhelm et al. 2006;Lang 2008;Thornton et al. 2012;Chau and Haddad 2013;Figueroa-Magalhaes et al. 2014;Kuczynski et al. 2015;White and Cohen 2015). A triple-TKI nintedanib (Vargatef Ò ) that targets VEGFR (VEGF receptor), PDGFR (PDGF receptor) and FGFR (FGF receptor) has been approved for non- small-cell lung cancer in the European Union (Caglevic et al. 2015). A mAB onartuzumab, which targets MET in combination with other agents, has completed phase III in advanced solid tumors (Shah et al. 2017;Spigel et al. 2017), though it did not show improved clinical outcomes unlike its competent results in phase II of non-small-cell lung cancer (Spigel et al. 2013). The Wnt/Frizzled ligand and receptor compound has fewer antibodies under clinical trial, yet active research is underway for therapeutic target development. Frizzledtargeted monoclonal antibody vantictumab (OMP-18R5) is going through phase I clinical trial with or without combined chemotherapeutics. Niclosamide (Niclocide Ò ), a classic FDA-approved anthelmintic drug, has been reported to suppress the Wnt pathway in ovarian cancer (Arend et al. 2014) and familial adenomatosis polyposis (Ahn et al. 2017) in preclinical studies. The niclosamide can also target Notch, mTOR, and NF-jB signaling cascades to downregulate EMT characteristics in preclinical studies of glioblastoma (Wieland et al. 2013). It should note that targeting Frizzled receptor may a big huddle since there are 10 Frizzeled receptors while the oncogenic role of each receptor is not fully determined yet.
Notch-targeted monoclonal antibody tarextumab (OMP-59R5) with or without combined chemotherapy is under phase I and II trials for various solid tumors. A c-secretase inhibitor, RO4929097 (RG-4733), went through clinical trial phase II for metastatic pancreatic adenocarcinoma (De Jesus-Acosta et al. 2014), metastatic melanoma , recurrent ovarian cancer (Diaz-Padilla et al. 2015), and metastatic colorectal cancer (Strosberg et al. 2012) with minimal efficiency as a single agent. Another c-secretase inhibitor, MK-0752, was in phase I for advance solid tumors with clinical benefits and tolerability (Krop et al. 2012;Hoffman et al. 2015).

Considerations in EMT therapeutics
EMT plays a key role in tumor progression and metastasis, which makes it an attractive target for cancer therapeutics. The regulatory mechanisms of multi-protein/transcriptional signaling pathways in EMT are extremely complex, with crosstalk and continuous reciprocal feedback loops. Drug development focusing on EMT is thus a complicated process and demands omnidirectional considerations.
EMT inhibitors may be compared with conventional chemotherapeutics or radiotherapy in terms of therapeutic efficacy during drug development, either alone or combined with conventional agents. Conventional chemotherapy and radiotherapy responses in vitro are evaluated through tumor proliferation inhibition and cancer cell cytotoxicity (Fan et al. 1998;Zoli et al. 2001). Clinical trials are typically evaluated based on the response evaluation criteria in solid tumors (RECIST) (Yoshida et al. 2015). In the revised version, non-measurable situations are assessed by tumor markers combined with other objective tumor responses (Eisenhauer et al. 2009). Nevertheless, EMT-targeted therapeutics should not be evaluated in the same way as conventional adjuvant chemotherapy. Note that tumor proliferation is attenuated in EMT and upregulated during MET and, noticeably, metastatic re-colonization (Zhu et al. 2014). Futhermore, in clinical tissue settings, EMT does not always gain its full transition state and may instead appear as 'partial EMT' expression (Nieto et al. 2016). Therefore, clinical assessment involving correlation between EMT marker expression and EMT-associated therapeutic efficacy needs further attention. The combination therapy with cytotoxic adjuvant chemotherapy or immunotherapy may an interesting option for EMT therapeutics. Several existing EMT-associated therapeutics were originally developed for other mechanisms than EMT and were evaluated using traditional methods. For example, assay systems for E-cadherin re-expression of cancer cells provided the interesting cancer therapeutics, such as salinomycin and protein kinase A (PKA) activators targeting cancer stem cells and EMT (Gupta et al. 2009;Pattabiraman et al. 2016).

Concluding remarks
Proliferative markers and cytotoxicity should be reconsidered as modes of action in EMT drug development. Screening assays for EMT-specific therapeutics are designed to select molecules that inhibit EMT phenotypic transition, invasiveness and metastatic properties without influencing proliferation potential (Marcucci et al. 2016). EMT markers discussed earlier, such as loss of E-cadherin, are effective in vitro while in vivo proof of concept for EMT-targeted therapeutic is under investigation. Future drug development for EMT should be applied using novel therapeutic targets, assessment measurements and therapeutic goals. 2017R1C1B1012464) funded by the Korean government (MSIP) and a grant from the Korea Health Industry Development Institute (KHIDI) funded by the Ministry for Health & Welfare Korea (HI17C2586).

Compliance with ethical standards
Conflict of interest The authors declare no conflict of interest.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://crea tivecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.