Epithelial-to-mesenchymal transition (EMT) represents conversion of an epithelial cell in an elongated cell with mesenchymal phenotype, which can occur in physiologic and pathologic processes such as embryogenesis (type 1 EMT), wound healing and/or fibrosis (type 2 EMT) and malignant tumors (type 3 EMT). The proliferation rate, metastasizing and recurrence capacity, as also the individualized response at chemotherapics, in both epithelial and mesenchymal malignant tumors is known to be influenced by reversible switch between EMT and mesenchymal-to-epithelial transition (MET). Although much research work has already been done in these fields, the specific molecular pathways of EMT, relating to the tumor type and tumor localization, are yet to be elucidated. In this paper, based on the literature and personal experience of the authors, an update in the field of EMT vs MET in epithelial and mesenchymal tumors is presented. The authors tried to present the latest data about the particularities of these processes, and also of the so-called endothelial-to-mesenchymal transition, based on tumor location. The EMT-angiogenesis link is discussed as a possible valuable parameter for clinical follow-up and targeted therapeutic oncologic management. The paper begins with presentation of the basic aspects of EMT, its classification and assessment possibilities, and concludes with prognostic and therapeutic perspectives. The particularities of EMT and MET in gastric and colorectal carcinomas, pancreatic cancer, hepatocellular and cholangiocarcinomas, and lung, breast and prostate cancers, respectively in sarcomas and gastrointestinal stromal tumors are presented in detail.

Because of paucity of data on biological behavior of epithelial and mesenchymal malignant tumors, the process of metastasis is still so poorly understood that it is difficult to block it therapeutically. In recent years, one of the newest described mechanisms that seems to contribute to migratory and invasive properties of tumor cells is the epithelial-mesenchymal transition (EMT)[1-3]. Till now, it is known to be implicated in carcinomas and in some mesenchymal tumors, such as sarcomas and gastrointestinal stromal tumors (GIST)[1-21]. However, the molecular mechanisms underlying EMT and distant metastasis remain somewhat unclear.

In this review we intend to present the main characteristics of EMT in a logical manner, beginning with the basic molecular mechanisms and possibilities of its quantification on histological slides. Then, the particularities of EMT in several carcinomas and sarcomas are explored and a future perspective synthesized.

Literature describes several markers, which are involved in EMT. During EMT, E-cadherin expression is reduced to near-zero level, E-cadherin-to-N-cadherin (Neural cadherin) switch is installed and other EMT-related markers [vimentin, fibronectin, smooth muscle actin (SMA), desmin, Sox, SNAIL1/SNAIL, SNAIL2/SLUG, Axl, zinc finger E-box-binding homeobox 1 (ZEB1), Notch-1, v-ets erythroblastosis virus E26 oncogene homologue 1 (ETS1), fms-related tyrosine kinase 1 (FLT1), V-set and immunoglobulin domain-containing protein 1 (VSIG), stromelysin-3, Twist, FOXC2, HOXB7, ACTA2, platelet derived growth factor (PDGF), etc.] are overexpressed[4,11,32], predominantly at the invasive front[23]. Other epithelial cell-cell adhesion molecules such as claudins (types 3, 4, and 7), α-catenin, γ-catenin, occludin, desmoplakin and plakoglobin are also down-regulated in cells with mesenchymal signature[5,11,32]. Immunohistochemically, E-cadherin and catenins mark the cell membrane; N-cadherin, vimentin, and PDGF present a predominant cytoplasmic expression; Notch-1 and ZEB1 mark the nuclei, whereas SNAIL, SLUG, and Twist mark the cytoplasm, as well as the nuclei of the tumor cells.

It is important to note that some of the transformed cells present only mesenchymal phenotype, whereas other tumor cells display a dual epithelial-mesenchymal expression[32], also known as an amphicrine pattern[33]. As regards the stem cells with malignant potential, it is still unclear if they are identical to the cells with mesenchymal signature, or they represent two distinct cell types; they are marked by both mesenchymal markers and stem cell markers such as CD44 and ALDH (aldehyde dehydrogenase)[34].

SLUG is a zinc-finger transcription inhibitory protein, a member of the SNAIL family (SNAIL 1 and 2) that is primarily described in the neural crest and embryonic mesodermal cells; it also plays roles in type 1 and type 3 EMT[32-36]. SNAIL 1/SNAIL is a transcriptional hypoxia-activated protein that interacts with Wnt- and also with serine/threonine kinase receptor signalling pathways; the tumor cells’ invasiveness and resistance to apoptosis are influenced by its overexpression in both tumor cells and stromal fibroblast[32,37]. SNAIL 2/SLUG is involved in cancer progression and suppresses E-cadherin, desmoplakin, and keratin-18 expression, but the exact mechanism is not known; it is supposed to interfere with the Wnt/GSK3β/β-Trcp1 pathway[4].

Vimentin is a mesenchymal marker that can be acquired during EMT but its expression is low and most difficult to be quantified, as compared with other EMT markers that indicate a mesenchymal-like phenotype[1].

The receptor tyrosine kinase Axl is an EMT marker whose mRNA expression is strongly correlated with vimentin. It is involved in EMT of breast, and pancreatic and lung cancers and is expressed more in mesenchymal than in epithelial lines[6-8].

Dismantling of epithelial cell membrane and remodeling of extracellular matrix are also influenced by matrix metalloproteinases (MMP-2, MMP-3, and MMP-9) that act as proteolytic enzymes and may link the EMT with increased invasion and metastasis, and shortened survival time in almost all types of human cancer[23,24,38].

VSIG, also known as radioiodinated cell surface A33 antigen or glycoprotein A34, is a single-pass 387 amino acid membrane protein involved in cell-cell adhesion[39].

Cathepsin family (e.g., Cathepsin Z located at 20q13.3, Cathepsin X, etc.) is also considered to be involved in EMT and to contribute to tumor metastasis through down-regulation of E-cadherin, α-catenin, and β-catenin, and up-regulation of vimentin, fibronectin, MMP-2, MMP-3, and MMP-9[38].

In the last few years, several research works were carried out to understand the mechanism of EMT in both carcinomas and mesenchymal tumors, but data are elusive.

As regards DNA, the next-generation sequencing-based methods proved that the genome-wide DNA methylation reprogramming is not involved in EMT; the progressive EMT and genetic disorders are rather related to altered histone modifications[1,9] and epigenetic mechanisms[11]. Recently, it is suggested that a 76-gene EMT-signature is involved in the metastatic process and that it influences the therapeutical answer of tumor cells[7], but the results are still confusing.

E-cadherin’s transcriptional down-regulation is one of the key factors of EMT but other factors, such TGF-β also play important roles in these processes[1,9]. It is worth noting that there are no deletions or mutations of E-cadherin gene, but only epigenetic down-regulation or transcriptional silencing which allows its further re-expression in primary or metastatic tumors[18]. E-cadherin is codified by the CDH1 gene; it interferes with other EMT-related genes, such as vimentin, fibronectin 1 (FN1), CDH2 (which codifies the protein N-cadherin), ZEB1 (target of SNAIL), ZEB2/SIP1, K-ras, integrin, Notch, and Axl (a tyrosine kinase inhibitor)[7]. The cancer cell lines with epithelial phenotype are characterized by overexpression of other genes such as claudins 4 and 7, MUC1, RAB25, SPINT2, and ERBB2[7]. However, some of these molecular markers (SNAIL 1/2, ZEB 1/2, E47) are direct inhibitors of the transcription of E-cadherin gene, whereas Twist, Goosecoid, FoxC2, and E2.2 are indirect E-cadherin inhibitors[11,16,17]. At the same time, SNAIL 1/2, ZEB1/2, and Twist are activated by TGF-β, a cytokine, secreted by mesenchymal stromal and inflammatory cells, whereas N-cadherin is activated by Twist[32,40]. If we take into account that TGF-β also induces renal EMT, besides mobility of endothelial cells, followed by EndMT and subsequent renal and cardiac fibrosis[24,25], and regulation of matrix accumulation[32], then we can suppose that stromal fibrosis is generated by an interaction between EMT and EndMT. Moreover, TGF-β is a tumor suppressor in early-stage carcinomas but induces tumor cells’ proliferation, migration, and metastases, in advanced stages[32].

Epidermal growth factor (EGF) is a mitogenic factor involved in tumor proliferation and aggressiveness, through its receptor, EGFR. Although EGFR and K-ras genes’ status are used as indicators for targeted therapy with anti-EGFR drugs of several tumors, such as pulmonary and colorectal cancers, the origin of this factor or its prognostic role is not firmly proved. As regards the role of this pathway in EMT, it seems that the Ras-activated/SNAIL/SLUG pathway interacts with FoxC2 and the phosphatidylinositol 3’-kinase (PI3K)/Akt/mTOR axis, at least in the case of colorectal cancer[40].

One of the newest factors found as pivotal regulators of EMT and as negative regulators of E-cadherin are the post-transcriptional gene regulators microRNAs (miRs): miR-21, miR-26b, miR-29c, miR-31, miR-124, miR-212, and the five members of the miR-200 family (miR-200a, 200b, 200c, 141, and 429) with their most prominent gene targets ZEB1 and ZEB2 (also known as SIP1 and SFHX1B)[2,10,11,17,18,21,40]. Their overexpression is believed to inhibit EMT in human carcinoma cells and to decrease tumor cell proliferation and migration in the blood flow[2,10,11,17,18,21]. EMT-related interaction of miR-200 family with p53 gene is also assumed[19].

Their elective affinities for specific types of cancer cell lines are described below.

The data available on the link between EMT and angiogenesis is so scattered that it does not elucidate if the link is any possible important therapeutic implications. For example, the MMP family members (MMP-2, MMP-3, and MMP-9), which are released by fibroblasts and macrophages, are known to influence both EMT and angiogenesis[23,38], though not proved to mediate the interaction among these two processes. At the same time, EMT, activated via SNAIL/Twist, is responsible for the attachment of tumor cells to the activated endothelial cells viaα-tubulin detyronisation[28]. The transcription factor Twist, a target gene of SNAIL, is reportedly implicated in embryogenesis (EMT type 1), but its hypoxia-activated over-expression is also proved in several human carcinomas[41]. However, its role in carcinogenesis and metastasis is not well defined. In experimental models, Twist inhibition does not decrease tumor cell proliferation rate, but reduces circulating tumor cells significantly[28]. In cancer tissues its expression is increased at the invasion front[41]. Based on these observations, it is supposed that Twist increases direct invasion of tumor cells in the surrounding tissues, and cell penetration inside the endothelial layer, but does not influence tumor growth[28,41]. As a therapeutic target, its inhibition may decrease the rate of metastasis.

Because the endothelium of pre-existing mature vessels, involved in angiogenesis of both epithelial and mesenchymal tumors, is activated by CD105 (endoglin)[42,43], new studies are necessary to prove if the Twist-attached endothelial cells are also CD105 positive or other mechanism is involved in this attachment.

Most recently, a possible TGF-β1 mediated-EndMT of intratumor endothelial cells is reported[44]. It relates to the cases in which the intratumor endothelial cells lose the cell-cell junction and immunoexpression of CD31 and gain positivity for mesenchymal markers such as α-SMA (smooth-muscle actin)[44]. It is important to note that the endothelial cell-specific miR-302c is proved to inhibit this EndMT, but decreases the tumor cell motility[44]. In previous researches, our team observed inconstant positivity of Kaposi sarcoma cells[42] and gastric tumor cells for the endothelial marker CD105 (Gurzu et al[42], personal communication), which could indicate even a possible epithelial-to-endothelial transition or an incidental positivity of epithelial cells with mesenchymal signature.

A relationship between inflammation, angiogenesis and EMT is suggested by COX-2-mediated EMT, which is stimulated by TGF-β through a PGE2-dependent mechanisms[23,35], and also by an MMP-related signaling[23].

Other arguments that favor a link between angiogenesis and EMT-mediated metastasis are that, on the one hand, the endothelial cell-secreted factors, such as EGF, increase cell mobility and inhibit apoptosis, whereas, on the other hand, the EGF-related EMT is more prominent in the perivascular area, but the intraluminal tumor cells exhibit an epithelial phenotype[45]. It is important to prove that the involved vessels present CD105-positive activated endothelium. EGF seems to also induce a stem-like phenotype of EMT-transformed tumor cells (in both intra- and perivascular tumor cells) and SNAIL positivity through PI3k-Akt pathway, whereas down-regulation of EGF in endothelial cells decreases tumor growth rate and the immunoexpression of stem cell markers[45]. The data suggests that antiangiogenic therapy should be based on inhibition of tumor cells from secreting pro-angiogenic substances such as vascular endothelial growth factor-A (VEGF-A), and of the activated endothelial cells to synthesize pro-angiogenic/pro-stem/pro-EMT substances such EGF[45].

In pancreatic cancer cells, EMT is induced by several molecular factors, such as the receptor tyrosine kinase AXL[7,8], loss of FOXA1/A2 expression[17], and dysregulation of miR-200 family[17], but this event is not so frequent as the other carcinomas are[17,54]. Moreover, loss of E-cadherin expression, a result of E-cadherin gene methylation, is correlated with histological grade of differentiation, which is noticed in about one quarter of well-differentiated ductal adenocarcinomas, half of the poorly-differentiated ones and few or none of the non-cohesive (undifferentiated and signet ring cell carcinomas) cases[17,55]. In chemoresistant well/moderately-differentiated pancreatic adenocarcinomas, E-cadherin loss and a membrane-to-cytoplasm shift of β-catenin expression pattern were noticed in the non-cohesive foci[17,55]. N-cadherin does not mark the normal acinar, ductal, and pancreatic islets; it is expressed in more than 50% of pancreatic carcinomas and is more intense in infiltrating cells from the invasive tumor front, independent of the tumor stage[56]. As regards the other IHC markers that indicate a mesenchymal-like phenotype of the epithelial tumor cells, SLUG was reportedly expressed in half percent of the cases, while SNAIL 1 marked 75% of pancreatic carcinomas. SNAIL 1 presents predominant positivity in the ductal cells of the tumor center and in undifferentiated carcinomas, as compared with the more cohesive carcinomas, whereas SLUG is more intense in the invasive front[56]. Moreover, 50% of the pancreatic carcinoma cells present double positivity for SNAIL and E-cadherin[56]. Twist is negative in 97% of pancreatic carcinomas and its expression is 10-fold higher after exposure of pancreatic cell lines to hypoxia for 48 h[56].

Hypomethylation and overexpression in pancreatic carcinomas cells and elevated serum levels of miR-200a and miR-200b are associated with retention of E-cadherin expression, as also with epigenetic silencing through methylation of ZEB2 (SIP1) gene[11,17]. As SIP1 expression is absent in most of the pancreatic cancers, it seems that E-cadherin dysregulation is rather related more to miR-200a and miR-200b than to SIP1 gene[17]. The serum levels of these miR-factors were noticed to be elevated in patients with chronic pancreatitis, as compared with healthy controls[17]; they can be used as a screening parameter to identify patients with chronic pancreatitis and high risk for ductal adenocarcinoma.

Lung cancer is one of the tumors that is known for rapid progression accompanied by a large number of genetic and epigenetic changes[1,7]. EMT is identified in about 23% of non-small cell lung cancer (NSCLC) lines and in one third of the patients with metastatic lung cancer. Based on the histology of the cell lines, the squamous pattern associates with a mesenchymal phenotype in about 50% of lines; adenocarcinomas are characterized mostly by epithelial phenotype, and the other patterns (neuroendocrine, large cell carcinoma) by only mesenchymal signatures[7]. Although it was experimentally proved that DNA methylation is not involved in EMT of lung cancer cell lines, after transition, the expression of some DNA-related enzymes, such as methyltransferases 1, 3a, and 3b, and ten eleven translocation (TET1), were significantly altered and some histone methylation was changed[1].

In A549 lung cancer cell lines, during TGF-β-dependent EMT, miRNA relative expression of E-cadherin was progressively lost and the most significant overexpression was attributed to N-cadherin, followed by SNAIL 1 and vimentin[1]. Maximum level of N-cadherin was observed after 96 h of incubation, with significantly increasing levels from 24 to 96 h[1]. SNAIL 1 becomes significantly expressed after 4-12 h of incubation, followed by progressive overexpression in the next 24-96 h; its level is quite low compared to that of N-cadherin[1]. The overexpression of vimentin does not differ significantly at 12, 24, and 96 h. Moreover, its level is maintained at a constant value, which is similar to the levels of SNAIL 1 and N-cadherin, quantified at 10-12 h of incubation[1]. In NSCLCs, SNAIL 1 and vimentin’s expressions are reversely correlated with the miR-30a level[62].

In contrast, another experimental study proved that 76 genes were involved in EMT, the main ones being CDH1 and vimentin, followed by several EMT transcription factors such as FN1, MMP2 (matrix metalloprotease-2), and ZEB1; the gene CDH2 was identified inconstantly and was not included in the EMT molecular signature[7]. In mesenchymal cells from NSCLC lines, K-ras mutation, loss of STK11 (LKB1), and SMARCA4 mutations/deletions are more frequent than in epithelial lines; the last ones rather show CDKN2A and CDKN2B loss[7].

In lung cancer cell lines, EMT also induces down-regulation of some histones methylation, such as DNMT1, DNMT3a, DNMT3b, H3K4me3, H3K9me2, TET2, and TET3, and up-regulation of TET1, and H3K36me3[1].

As regards the prognosis, retention of the epithelial phenotype, proved by E-cadherin positivity in the tumor cells, was associated with longer time to progression and a longer survival, independent of the mutation status of the EGFR (epidermal growth factor receptor) gene[7,12]. EGFR wild type/K-ras-wild type molecular profile seems to indicate a less aggressive NSCLC[7].

EMT in breast[4] is dependent on reducing mi-RNA expression of E-cadherin, which is related to CDH1 promoter methylation, but not to mutational inactivation[9]. Just as the other carcinomas are, E-cadherin is re-expressed in hepatic metastatic tissues[11], and local recurrence may be dependent on the EMT-MET switch in the primary tumor[32].

Claudins (types 1, 3, 4, and 7) were described to be down-regulated in certain types of breast cancer, such as fibromatosis-like metaplastic carcinoma, a low-grade EGFR (exons 18, 19, 20, and 21) wild-type basal-like spindle cell carcinoma with predominant HER-2 and hormone receptors negativity, diffuse positivity for basal keratins and focal vimentin and E-cadherin expression[5]. Fibroblastic-like phenotype of breast carcinoma cells (with negative or low E-cadherin expression), intense angiogenesis (indicated by overexpression of VEGF-A) and chromosome instability can also be induced by twist, via Wnt/β-catenin[32,63].

SNAIL 1, known to induce invasiveness in several human cancers, seems to be a hypoxia-protective factor for breast tumor cells, viaβ-catenin activation, which regulates expression of HIF-1 (hypoxia-inducible factor 1)-dependent genes, besides being an inductor of hormone-resistance[32,63]. Based on these facts, it is suggested that SNAIL 1 may be used as an indicator of response of breast cancer to antiangiogenic therapy[64].

Twist expression was proved to be up-regulated in tandem with tubulin detyrosination, in both invasive ductal[28] and lobular-type adenocarcinomas[41], at the invasion front[28,41]. This interaction, which is mainly twist-dependent, seems to promote penetration of tumor cells through the endothelial cell layer, endothelial engagement and, probably, consecutive angiogenesis that favors metastasis[28].

EMT is also induced by the receptor tyrosine kinase AXL[6,7], Pyk2 and TGF-βvia SNAIL 1/2 and ZEB 1/2[32], and it seems to have particular pathways in triple negative breast cancers[32].

As regards miRNA, the miR-200 family-mostly miR-200a/b/c, miR-141, and miR-429-is considered to influence EMT of breast cancer cells with inhibition of the E-cadherin repressors ZEB 1/2[32,65]. An interaction between p53 gene and miR-200c is also defined in breast cancer cell lines[19]. The miR-21s also influences the EMT phenomenon via tumor suppressor gene PTEN and AKT/ERK1/2 axis[32,66]. Besides down-regulation of cyclin-dependent kinase (CDK8), β-catenin targeted-miR-26b is the other reported negative prognostic factor in breast cancer[58,67].

EMT is reported to be involved in breast cancer resistance to tamoxifen; the most down-regulated factor is the methaderin-targeted miR-375[68], whereas miR-519a is up-regulated in tamoxifen-resistant breast cancer cell lines with mesenchymal-like signature[69]. Re-expression of miR-375 and inhibition of miR-519a might serve as potential therapeutic approaches for estrogen receptor-positive breast carcinomas[68,69]. It was also experimentally proved that some contraceptive pills, such as centchroman, may be used to inhibit EMT and to play a dose-dependent antiapoptotic and antiproliferative role in human breast cancer cells, via down-regulation of HER2/ERK1/2/MMP-9 signaling[70].

Prostatic carcinomas have a predilection for bone metastases, the metastatic cascade being also related to EMT that is linked with stem cell signature of the prostatic carcinoma cells[15]. Although vimentin and ZEB 1 contribute to EMT, the most involved marker that plays a role in bone metastasis seems to be Notch-1[15].

In liver metastases, prostate carcinoma cells are bound to hepatocytes in an E-cadherin-dependent manner[18]. In primary tumor tissues, EMT is characterized by activation of EGFR and subsequent down-regulation of E-cadherin, whereas, in liver metastases, down-regulation of EGFR signaling induces re-expression of E-cadherin and cell-cell adhesion[18]. In bone metastatic tissue, E-cadherin remains down-regulated and Notch-1 is up-regulated[15].

Although the literature provides several new insights into EMT of tumor cells, supplementary molecular exploration is necessary for therapeutical reducing the rate of metastasis. The literature review shows that several factors are indeed involved in EMT, most of them having prognostic and/or predictive value. However, the features relating specifically to organ-related carcinomas need further elucidation, because most papers dealing with this subject contain no specific and valuable data.

If EMT is indeed involved in the resistance of tumor cells to specific drugs, such as anti-EGFR substances, tamoxifen, and classic chemotherapics, identification of transition pathways could be of immense help in evolving an appropriate targeted therapy for both carcinomas and malignant mesenchymal tumors, including GISTs.

Additional studies of complex molecular profiling are necessary to elucidate the particularities of EMT in terms of histological type and localization of the tumor and angiogenesis-EMT interaction. Complex studies should also take into account not only cases with E-cadherin loss, but also the particularities of those cases, which show focally decreased expression and aberrant immunohistochemical pattern. We believe that, in the near future, the particularities of EMT will be a very useful parameter for proper clinical follow-up, individualized therapy and a more refined molecular classification of malignant tumors.