Emerging evidence of epithelial-to-mesenchymal transition in lung carcinogenesis


  • The Authors: Dr Mitsuo Sato is an Associate Professor in the Department of Respiratory Medicine at Nagoya University Graduate School of Medicine, who has been researching the molecular biology of lung cancer for the past 14 years. Dr David S. Shames is a Scientist in the Oncology Biomarker Development group at Genentech Inc. His research efforts focus on epigenetics and predictive biomarker discovery and development to support oncology therapeutics development in lung cancer. Dr Yoshinori Hasegawa is a Professor in the Department of Respiratory Medicine at Nagoya University Graduate School of Medicine. His current research interests include study on circulating tumour cells in patients with lung cancer.


Mitsuo Sato, Department of Respiratory Medicine, Nagoya University Graduate school of Medicine, 65 Tsurumai-cho, Showa-ku, Nagoya 466-8550, Japan. Email: msato@med.nagoya-u.ac.jp


The epithelial-to-mesenchymal transition (EMT) is a developmental programme that regulates embryonic morphogenesis and involves significant morphological and molecular changes in cells. Experimental models have revealed that EMT also contributes to various malignant features of cancer cells, including motile, invasive, anti-apoptotic and stem-like phenotypes. Clinically, correlative studies have indicated that mesenchymal-like features of tumour cells are associated with poor tumour differentiation as well as worse patient prognosis. Nevertheless, due to its transitory nature, demonstration of an actual occurrence of EMT during human carcinogenesis is challenging, and most of the evidence to date has been limited to breast and colorectal cancers. However, recent studies suggest that EMT may occur during lung cancer development, although such evidence is still limited. We propose three approaches for obtaining direct evidence of EMT in human cancers and use these criteria to review the available data. We suggest that multiple intrinsic and extrinsic factors cooperatively induce EMT in lung cancer. Intrinsic factors include oncogenic genetic changes such as mutant K-RAS. Extrinsic factors are associated with a tumour microenvironment that is inflammatory and hypoxic. The induction of EMT is primarily mediated by various EMT-inducing transcription factors that suppress E-cadherin expression, including SLUG and ZEB1. miR-200 family expression can reverse EMT by suppressing EMT- inducing transcription factors. Obviously, more data demonstrating the clinical relevance of EMT in lung cancer are required, and further elucidation of how EMT is regulated in lung cancer will enable us to develop novel therapeutics that specifically target molecules with critical roles in EMT.


Epithelial cells are polarized and tightly connected to adjacent cells, forming an ‘epithelial sheet’ structure while mesenchymal cells exhibit a less polarized, spindle-like morphology and are loosely connected. The epithelial-to-mesenchymal transition (EMT) was initially described as a process that occurred during early embryonic development whereby cells lose their epithelial characteristics and obtain mesenchymal phenotypes. As EMT progresses, cells also acquire a motile and invasive phenotype.1–3 The reverse phenomenon, termed mesenchymal-to-epithelial transition (MET), also occurs during embryogenesis, and both phenomena cooperatively contribute to proper morphogenesis. Importantly, EMT has emerged as a critical phenomenon in the carcinogenic process. Cancer cells exploit EMT, and possibly MET as well.

The associations between EMT and local invasion as well as distant metastasis have been demonstrated in numerous in vivo and in vitro studies.1–3 Also, recent studies have shown that EMT confers drug resistance and anti-apoptotic phenotypes on cancer cells.1–3 Moreover, several studies have reported that EMT in carcinogenesis is associated with the stem cell phenotype,4 further underscoring the critical role of EMT in carcinogenesis.

The majority of the supporting data that posits the importance of EMT during carcinogenesis has been obtained using cell line models.5,6 However, these models have been criticized as being highly artificial because they do not properly reflect a realistic tumour-stromal microenvironment.7 It is very difficult to prove that EMT occurs in vivo for several reasons: first, obtaining serial-repeated biopsies on the same patient to demonstrate EMT in human cancer is hard to perform, especially in cancers that would require invasive methods such as lung cancer; second; second cancer cells that have undergone EMT may be indistinguishable from tumour fibroblast cells; third, by definition, an EMT is a transient process, making it hard to observe without real-time observation. Thus, direct evidence for EMT has been relatively limited in non-small-cell lung cancer (NSCLC) but has been observed in breast and colorectal cancers;8–10 however, several recent studies have reported data directly implicating EMT in lung cancer.11,12


E-cadherin is a calcium-dependent transmembrane glycoprotein that mediates cell-cell adhesion in a polarized epithelium.13 The loss of its expression is a hallmark of EMT.14 Numerous studies have reported that the loss of E-cadherin function or expression by genetic or epigenetic (less common in NSCLC compared with other cancers) mechanisms is quite common in lung cancer. In some studies, the loss of E-cadherin function also correlates with poor patient prognosis.15,16 In addition, a number of studies have reported that expression of molecules involved in EMT correlated with clinicopathological features in NSCLC, as shown in Table 1.16–24 While these correlative studies are suggestive of the relevance of EMT in NSCLC, they do not serve as a direct evidence of an EMT because, as mentioned above, EMT describes a dynamic, transient ‘process’. Importantly, carcinoma cells frequently exhibit varying degrees of de-differentiation, which is sometimes presumed to result from EMT, but these changes could simply be examples of anaplasia.25

Table 1.  Molecules involved in epithelial-to-mesenchymal transition whose expressions were correlated with clinicopathological features in non-small-cell lung cancer
MoleculesAnalysesClinicopathological featuresReferences
  1. HIF-1α, hypoxia-inducing factor 1α; PCR, polymerase chain reaction; TGF-β, transforming growth factor β.

Epithelial cadherinMethylation-specific PCRLonger overall survival 16
Immunohistochemical analysisNegative for lymph nodes metastasis 22
 Well-differentiated tumour histology 
Immunohistochemical analysisLonger disease-specific survival 17
VimentinImmunohistochemical analysisShorter disease-specific survival 17
TGF-βEnzyme-linked immunosorbent assayPositive for lymph nodes metastasis 20
 Advanced disease stage 
SLUGReal-time PCRPostoperative relapse 23
 Shorter overall survival 
Immunohistochemical analysisShorter overall survival 19
 Poor differentiation 
SNAILImmunohistochemical analysisShorter overall survival 24
Immunohistochemical analysisShorter overall survival 21
TWISTImmunohistochemical analysisShorter overall survival 21
HIF-1αImmunohistochemical analysisShorter overall survival 21
 Shorter recurrence-free survival 21
miR-200cReal-time PCRNegative for lymph nodes metastasis 18
  Well-differentiated tumour histology 

We propose three approaches to demonstrate the existence of an EMT during human carcinogenesis (Fig. 1). A first approach is to show that a given tumour section comprised cells with epithelial features adjacent to (ideally centrally located) a region with mesenchymal features near the invasive front of the tumour. Ideally, both histology and immunohistochemistry (e.g. E-cadherin and vimentin) should be used to show these differences. This was first illustrated in colorectal cancer by Brabletz et al.26 The authors showed that centrally located tumour cells stained positively for membranous E-cadherin and β-catenin, but both proteins were absent at the invasive front of the tumour, suggesting that an EMT occurred during or as a result of an interaction between the tumour and its microenvironment. Recently, a similar finding was shown in lung cancer by Tischler et al.,12 who reported that the transmembrane glycoprotein L1 cell adhesion molecule (L1CAM), which is a member of the immunoglobulin super family of cell adhesion molecules, and SLUG are expressed in the tumour stroma of lung cancer specimens while membranous E-cadherin was expressed in the central regions of the tumour12 (Fig. 2).

Figure 1.

Three approaches to demonstrating epithelial-to-mesenchymal transition in vivo carcinogenesis. (a) Approach 1 is to show that a given tumour section comprised cells with epithelial features at the central areas and cells with mesenchymal features mainly located at the tumour invasive front. (b) Approach 2 is to demonstrate the transformation from epithelial to mesenchymal cells by taking serial biopsies from patients. (c) Approach 3 is to demonstrate that tumour fibroblasts are derived from tumour cells using an in vivo cell tracking method in animal models.

Figure 2.

L1 cell adhesion molecule (L1CAM) and epithelial-to-mesenchymal transition marker expression patterns at the tumour-stroma interface. (a) Squamous-cell carcinoma with expression of L1CAM at the tumour-stroma interface (brown). The tumour center shows moderate E-cadherin expression (red). At the tumour-stroma interface, E-cadherin expression is decreased. Note the strong positivity of small peripheral nerves for L1CAM. (b) Double immunofluorescent staining for L1CAM (green) and E-cadherin (red): L1CAM is expressed at the tumour border, and E-cadherin expression is strongest in the tumour center. E-cadherin expression is decreased at the tumour border (yellow). (c) Membranous E-cadherin (red) is expressed in the tumour center and decreased towards the tumour-stroma interface. Two strong vimentin-positive (brown) stromal cell aggregates are marked with dotted lines (upper left and mid to lower right). Note that most of the vimentin-positive cells show nuclear morphology of the tumour cells. CD68 staining to exclude vimentin-positive macrophages was not performed. (d) Decrease of membranous E-cadherin (red) and strong nuclear expression of Slug (brown) at the tumour-stroma interface of a squamous cell carcinoma. In the tumour center, E-cadherin is strongly but Slug is not expressed (blue nuclei). Haematoxylin and DAPI counterstain were used, respectively. Tumour-stroma interfaces are marked by dotted lines. (Reproduced from Tischler et al.,12 with permission.)

A second approach would be to demonstrate the transformation from a predominantly epithelial morphology into a predominantly mesenchymal morphology in serial biopsies from the same lesions in the same patient, presumably before and after some distance in time or line of therapy. This could provide strong evidence for EMT, although concerns that sequential biopsy samples may not reflect the phenotype of the whole tumour may be raised. By doing serial biopsies on patients with epidermal growth factor receptor (EGFR)-mutant lung cancer, Sequist et al. discovered that EGFR-TKI treatment induces EMT in lung tumours, which may account for the acquired resistance to EGFR-tyrosine kinase inhibitor (TKI).11 The paper reported several other molecular mechanisms responsible for the EGFR-TKI resistance, emphasizing the importance of repeated biopsies in patients with NSCLC. Very recently, the feasibility of a personalized approach to NSCLC by doing repeated biopsies has been demonstrated by the BATTLE study.27 Thus, we expect that repeated biopsies for lung cancer will be done more frequently in the near future and will provide more data demonstrating EMT in lung cancer.

A third approach might be to demonstrate that tumour fibroblasts are derived from tumour cells using in vivo cell tracking methods in animal models. This approach seems less satisfying compared with the other two methods described above because studies comparing patterns of genetic changes between microdissected cancer cells and tumour fibroblasts in human cancers found that these cells infrequently shared common genetic changes, suggesting that only a small fraction (if any) of the tumour fibroblast might be derived from epithelial cells.28 However, Trimboli et al. have provided good example for this approach. They demonstrated that stromal fibroblasts surrounding mouse breast tumours are of epithelial origin by using an elegant newly developed mouse breast cancer model that independently marks epithelial and stromal cells, providing direct evidence for EMT in breast cancer.9 This type of evidence remains to be demonstrated in lung cancer.

Several other studies have provided circumstantial evidence for EMT in human lung cancer tissue samples that collectively support the hypothesis that EMT occurs in vivo during human lung carcinogenesis. Prudkin et al. examined a large number of adenocarcinomas and squamous-cell carcinomas of the lung as well as normal epithelium and premalignant lesions for the expression of various epithelial and mesenchymal markers by immunohistochemistry.29 They found that dysplastic lesions have lower levels of mesenchymal marker expression compared with squamous-cell carcinomas. In addition, metastatic lesions in the brain from these same tumours express higher levels of E-cadherin, suggesting that an EMT had occurred during carcinogenesis as well as a potential example of a MET in metastatic sites. In addition, a study that compared the expression of EMT markers in clinical NSCLC samples before and after preoperative chemoradiotherapy revealed that 40% of samples exhibited EMT-like changes, which correlated with poor disease-free survival and suggests that EMT occurs in vivo in lung cancer patients and that this change may associate with poor patient prognosis.30 Finally, a study performing cytogenetic analysis of pulmonary carcinosarcoma tissues revealed very similar patterns of allelotyping between carcinoma and mesenchymal components, implying their monoclonal origin.31 Another study also reported some overlaps of chromosomal aberrations between carcinoma and mesenchymal components in lung carcinosarcoma.32 These data suggest that the cells in the mesenchymal component are derived from cells in the carcinoma component through EMT, which supports the notion that EMT occurs in vivo.

Numerous studies using cell lines as model systems have reported that lung cancer or normal lung epithelial cells undergo EMT.33–38 Several investigators have reproducibly demonstrated that A549 lung cancer cell line undergo EMT upon transforming growth factor (TGF)-β1 treatment.33,38 Tellez et al. reported that chronic exposure to tobacco smoke carcinogens induced immortalized normal bronchial epithelial cells (human bronchial epithelial cells (HBEC)) to undergo EMT early on which appeared to lead to the acquisition of a stem cell-like phenotype.36 In addition, a study demonstrated that tumour cell lines derived from mutant K-RAS and p53 transgenic mouse that is prone to develop lung cancer transition between EMT and MET in a reversible manner. This reversible switch between epithelial and mesenchymal-like states is regulated in part by the expression of the microRNA 200 family.37

Collectively, these findings provide substantial evidence for the occurrence of EMT in human lung cancer, but more data are required to confirm this.


EMT in lung cancer is controlled by multiple intrinsic and extrinsic factors that probably work cooperatively (Fig. 3). Intracellular factors include intrinsic mutational events that accumulate during carcinogenesis. Several oncogenic pathways important to cancer progression appear to induce EMT in human cancers models. Mutant K-RAS is the best documented EMT-inducing oncogenic change, and it induces EMT alone or in combination with other EMT-inducing factors such as TGF-β or hypoxia-inducing factor−2α (HIF-2α) in several different model systems.5,39,40 As reviewed by others,41,42 hypoxia and an inflammatory tumour microenvironment, in part attributable to tobacco smoking, substantially contribute to the creation of a permissive environment for EMT through the dysregulation of the multiple extracellular signalling pathways. In particular, the TGF-β and the Cyclooxygenase-2 (COX-2)/prostaglandin E2 (PGE2) pathways play key roles in the tumour microenvironment. The hypoxic tumour microenvironment induces EMT in lung cancer primarily through upregulation of the basic helix-loop helix (bHLH) transcription factor, hypoxia-inducible factors (HIF). Master EMT-inducing transcription factors as well as microRNA play key roles in the regulation of intra- and extracellular signalling systems of EMT.

Figure 3.

Multiple factors cooperatively induce epithelial-to-mesenchymal transition (EMT) in carcinogenesis. Oncogenic changes such as mutant K-RAS functions as intrinsic EMT-inducing factors. A hypoxic, inflammatory tumour environment, which is in part attributable to smoking, induces EMT primarily through upregulating hypoxia-inducible factors. Stromal cells release multiple cytokines and growth factors that induce EMT.


TGF-β is a cytokine that regulates multiple biological responses important to cancer biology, including proliferation, apoptosis, angiogenesis and immune response, and is thought to be the major extracellular factor involved in EMT in vitro and in vivo.33,35,43,44 TGF-β is secreted by lung cancer cells as well as stromal cells.44,45 TGF-β appears to have paradoxical roles in normal and tumour cells and may have context dependent effects on the lung depending on the stage of the tumour. In normal cells, TGF-β suppresses proliferation, but in late-stage tumour cells, it stimulates proliferation and enhances invasiveness, which can be explained by the abrogation of TGF-β responsive pathways in those tumour cells. A study reported that TGF-β expression levels, as evaluated by immunohistochemistry, correlated with disease stage in NSCLC, suggesting its critical role in lung cancer development.20 Indeed, Osada et al. demonstrated that the majority of lung cancer cell lines exhibit impaired responsiveness to TGF-β, which is caused by heterogeneous mechanisms including the loss of TGF-β type-II receptor expression through infrequent promoter hypermethylation as well as the alteration of chromatin structure.46 These data suggest that TGF-β-induced EMT may require preceding genetic or epigenetic alterations that abrogate or diminish TGF-β responsive pathways.

Common oncogenic changes found in lung cancers can impact the responsiveness of tumour cells to TGF-β signalling. Several reports have shown that mutant K-RAS suppresses TGF-β signalling in lung epithelial cells by negatively regulating the TGF-β mediators SMAD2 and SMAD3.47 In addition, Halder et al. recently showed that histone deacetylation is involved in decreased levels of TGF-β type-II receptor in NSCLC cell lines and that its expression could be restored by a histone deacetylase inhibitor (HDI).48 Finally, another study showed that the HDI, MS-275, is capable of alleviating EGFR-TKI resistance in the EGFR-mutant NSCLC cell line through restoration of E-cadherin expression.49 These findings suggest a potential strategy for modulating EMT in lung cancer cells by HDI.


COX-2 is an inducible enzyme that catalyzes the production of the eicosanoids prostaglandins and thromboxanes from arachidonic acid.41 Several lines of evidence suggest that COX-2 confers various oncogenic properties on cancer cells, including resistance to apoptosis and increased proliferation as well as EMT.41,42 COX-2 is overexpressed in lung cancer cells as well as in premalignant lesions, and its expression correlates with poor patient prognosis in lung cancer.41,50 Based on these findings, COX-2 inhibitors are being tested as chemopreventive or therapeutic agents for lung cancer.51 EMT is induced by COX-2 primarily through PGE2, a major COX-2 metabolite that is abundantly present in the lung cancer environment.41 PGE2 has oncogenic capabilities, including promotion of proliferation, resistance to apoptosis and immune suppression.52–54 Dohadwala et al. documented that COX-2 suppresses E-cadherin in NSCLC through PGE2-mediated upregulation of ZEB1 and SNAIL.55 COX-2 expression is induced by several different factors including TGF-β and oncogenic changes such as mutant p53 and K-RAS,56,57 suggesting that both tumour stromal and lung cancer cells contribute to the upregulation of COX-2.

Hypoxia as an EMT inducer

Approximately 50–60% of solid tumours have hypoxic and/or anoxic areas that may result from an imbalance between oxygen supply and consumption in highly proliferative tumours.58 Hypoxia induces EMT through stabilization of the bHLH transcription factors, HIF protein family, primarily through inhibition of the prolyl-hydroxylase family of enzymes such as von Hippel–Lindau (VHL). The HIF transcription factors are central mediators of the cellular energy metabolism and oxygen-signalling pathways.58 Three HIF (HIF-1, 2 and −3) have been identified, and each of these consists of a constitutively expressed β and an oxygen-regulated α subunits. Yang et al. demonstrated that hypoxia or HIF-1α overexpression resulted in EMT through upregulation of Twist in NSCLC H1299 cells.59 Furthermore, they showed that overexpression of HIF-1α, TWIST or SNAIL correlated with poor prognosis in NSCLC patients. In addition, the cooperation of HIF-2α and RAS is associated with EMT features and promotes lung tumourigenesis in a mouse model of lung cancer.39

Master EMT-inducing genes

Genetic studies of development have identified several master EMT genes including those for zinc finger proteins (SNAIL, SLUG, ZEB1, ZEB2 (SIP1) and KLF8) and bHLH proteins (TWIST and E47).60 When ectopically expressed in epithelial cells, they are capable of inducing EMT primarily by suppressing E-cadherin expression.

Several studies have shown that there is cell-type specificity in the expression of genes that positively regulate EMT. For example, SNAIL expression correlates with the clinicopathological features of breast, ovarian and colon cancers, while TWIST expression correlates with them in uterine and ductal breast cancers.61–63


SLUG expression was the first among various EMT-inducing transcription factors, reported to correlate with poor patient outcome in lung cancer.23 Chu et al. established a series of cell lines from the same NSCLC patient with varying levels of invasiveness using an in vitro selection process. Then they performed microarray gene-expression analysis on these cells and identified SLUG as one of the genes overexpressed in invasive lung cancer cells.64 The study also demonstrated the ability of SLUG to enhance tumour growth as well as to promote angiogenesis. One study indicated an association between SLUG expression and a stem cell phenotype of lung cancer using a cell line model as well as clinical samples.19 The study reported that co-introduction of OCT4 and NANOG enhanced the stem cell properties of lung adenocarcinomas, which was associated with EMT through increased SLUG expression. The study showed that SLUG expression correlates with a worse prognosis for lung cancer patients and that patients with tumours positive for OCT4, NANOG and SLUG had the worst prognosis.


Several groups, including ours, have shown that ZEB1 has a predominant role in the EMT-associated carcinogenic phenotypes of lung cancer.65–67 We demonstrated that ZEB1 contributes to the anchorage-independent growth of lung cancer cells.66 This finding was recently confirmed by a study showing that the introduction of ZEB1 into immortalized normal HBEC causes them to undergo EMT and to acquire the capability to grow in soft agar.68 We also reported that ZEB1 expression showed the strongest inverse correlation with E-cadherin expression among several master EMT-inducing genes in certain lung cancer cell lines (Fig. 4, Table 2). In addition, semaphorin 3F, a well-known tumour suppressor gene, is transcriptionally suppressed in NSCLC cell lines by ZEB1, and the loss of semaphorine 3F results in cell signalling changes that may induce EMT-including upregulation of phosphor-AKT in normoxia and upregulation of HIF-α.69 These findings strongly support the pivotal role of ZEB1 in the EMT-associated pathogenesis of lung cancer. A correlation between ZEB1 expression and poor patient prognosis in lung cancer has yet to be demonstrated. However, lower expression of miR-200c, which is a known repressor of ZEB1 and ZEB2 (SIP1), is correlated with a higher propensity towards lymph node metastases in NSCLC.18

Figure 4.

Among four master epithelial-to-mesenchymal transition (EMT)-inducing genes, ZEB1expression is the most significantly correlated with the mesenchymal phenotype (high vimentin and low epithelial cadherin expression) in non-small-cell lung cancer cell lines. Quantitative real-time polymerase chain reaction (qRT-PCR) analysis of E-cadherin, Vimentin and four master EMT genes, ZEB1, SIP1, Snail and Slug, in 18 lung cancer cell lines. The cell lines are aligned by expression levels of E-cadherin from high (left) to low (right). The results are average of two independent PCR experiments done in duplicated reactions. GAPDH, glyceraldehyde 3-phosphate dehydrogenase; RVE, the ratio of Vimentin to E-cadherin. (Reproduced from Takeyama et al.,66 with permission.) *Cells with mutation in the epidermal growth factor receptor gene (EGFR).

Table 2.  Correlations between mRNA expression of master epithelial-to-mesenchymal transition genes, epithelial cadherin and vimentin in 18 non-small lung cancer cell lines
 Epithelial cadherinVimentinRVEZEB1SIP1Snail
  1. Spearman's correlation coefficients (upper row) and statistic values (lower row) are shown. Statistically significant correlations (P < 0.01) are in bold.

  2. RVE, the ratio of vimentin to E-cadherin. (Reproduced from Takeyama et al.,66 with permission.)

Epithelial cadherin       
Vimentin −0.74      
RVE −0.88 0.92     
<0.001 <0.001     
ZEB1 −0.82 0.80 0.88    
<0.001 <0.001 <0.001    
SIP1 0.22−0.05−0.06−0.06  
Snail 0.30−0.22−0.26−0.23−0.03 
Slug 0.090.07−−0.06

It is still not clear what molecular changes are responsible for ZEB1 upregulation in lung cancers. It is likely that cytokines such as TGF-β released from tumour stromal cells surrounding a tumour induce ZEB1 expression in lung cancer cells at the invasive front. However, such an observation has yet to be reported and if cytokines do induce such expression, how TGF-β specifically up-regulates ZEB1among many EMT-inducing transcription genes needs to be clarified. A recent study demonstrated one possible intrinsic mechanism for the increased ZEB1 expression in lung cancer. The study reported that knockdown of the tumour suppressor gene LKB1, which is frequently mutated in lung adenocarcinoma, up-regulates ZEB1 expression, leading to EMT.70 A study examining invasive bladder cancers has shown that the miR-200 family and miR-205 are downregulated in these tumours through promoter hypermethylation.71 It would be interesting to see whether this also occurs in human lung cancer.


Yanagawa et al. reported that ectopic SNAIL expression enhances the tumourigenicity of lung cancer cell lines and that patients with high Snail expression had a worse prognosis.24 Another study reported that coexpression of more than two of three EMT markers, HIF-1α, SNAIL and SLUG, correlates with poor outcome for patients with lung adenocarcinoma.21


TWIST is a helix-loop-helix transcription factor that was initially identified as an EMT inducer in embryogenesis. Its importance in carcinogenesis was first demonstrated by Yang et al. who reported that Twist expression contributes to distant metastases in a murine breast cancer model.21 We recently showed that ectopic expression of TWIST confers motility and an invasive phenotype on lung cancer cells that express low levels of TWIST, suggesting the importance of its role in EMT in lung cancer.72

MicroRNA as regulators of EMT

MicroRNA are evolutionally conserved non-coding RNA that post-translationally regulates gene expression; to date, over 700 microRNA have been identified in humans.73 Dysregulated microRNA expression contributes to many human diseases, including cancer, and the microRNA has emerged as a key regulator of EMT. The most studied microRNA involved in EMT are miR-200 family, which was reviewed recently.74 The miR-200 family consists of miR-200a, −200b, −200c, −141 and −429. The miR-200 family directly represses ZEB1 and ZEB2 expression, thereby causing downregulation of E-cadherin. Conversely, ZEB1 and ZEB2 repress miR-200 family expression, thus forming a negative feedback loop with the miR-200 family.74 Gibbons et al. demonstrated that miR-200 family expression correlates with EMT markers in a panel of 40 NSCLC cell lines.37 Furthermore, by using a series of syngeneic mouse tumour cell lines derived from mice that develop adenocarcinoma due to mutant K-RAS and p53 expression with different metastatic potentials the authors demonstrated that differential expression of miR-200 family expression was most prominent phenotype associated with metastasis-prone versus metastasis-incompetent cell lines. In addition to the miR-200 family, miR-155 may impact TGF-β-induced EMT. Kong et al. demonstrated that TGF-β induces EMT in part through Rho-A downregulation mediated by miR-155.75 Recently, miR-30a, which was identified as a regulator of SNAIL by bioinformatic analysis, was shown to be correlated with E-CADHERIN and N-CADHERIN expression in NSCLC cell lines.76 These findings suggest that microRNA may have a central role as regulators of EMT in lung cancer, implying that modulating these microRNA may be a strategy to control EMT in lung cancer.

Oncogenic alterations and EMT in lung cancer

A number of genetic alterations—mostly oncogene activation—have shown to induce EMT, each of them either alone, or in combination, with TGF-β treatment in in vitro cell line models, such as Mardin-Darby canine kidney cells and mouse models. These genetic changes include mutations of p53 and RAS, and overexpression of c-Myc, EGFR and HER2.5,40,77–80 However, one study has indicated that only a minority of cancer cell lines undergo EMT with TGF-β treatment,81 suggesting that genetic changes that induce EMT may occur in a context-dependent manner. Indeed, a recent study demonstrated that lung cancer cell lines with mutant KRAS do not necessarily exhibit a mesenchymal phenotype.82 The paper reported that lung cancer cell lines, which appear to be addicted to oncogenic K-RAS, exhibit an epithelial phenotype. This suggests that oncogenic K-RAS alone is not sufficient to induce EMT, and this ability may be context dependent (e.g. coexisting genetic changes). Mutant EGFR is another oncogenic change that frequently occurs in lung cancer.83 Epidermal growth factor, one of the major ligands of EGFR, is one of the most potent EMT-inducing growth factors;3,84 one might hypothesize that mutant EGFR is associated with the mesenchymal phenotype in lung cancer. Nevertheless, data contradicting this hypothesis have shown by ourselves and others, who reported that lung cancer cell lines and clinical samples that have mutant EGFR tend to show the epithelial phenotype.66,85 Thus, mutant EGFR is not likely to induce EMT in lung cancer. Very recently, another study has also provided data suggesting that expression of the ERBB2 gene, another member of EGFR family, is not associated with mesenchymal phenotype but rather with epithelial phenotype in lung cancer.86 The study reported that hypomethylation of the ERBB2 gene, which correlated with increased ERBB2expression, is associated with epithelial-like phenotype in NSCLC cell lines and tumour samples. Silencing LKB1, a tumour suppressor gene that is mutated in human lung cancer, in immortalized normal HBEC induces EMT through downregulation of ZEB1.70 On the other hand, thyroid transcription factor-1, which is a lineage specific oncogene amplified in a subset of human lung cancers, has been demonstrated to prevent TGF-β from inducing EMT in lung cancer cells.87 This suggests that oncogenic changes do not necessarily drive EMT in lung cancer. Overall, despite a vast amount of in vitro data showing that oncogenic activation induces EMT, its relevance to human lung cancer is still an open question—and further research in this field is required.


Growing evidence obtained mostly from cell line models has demonstrated that lung cancer cells, which have undergone EMT, acquire resistance to conventional cytotoxic drugs. For example, knockdown of TWIST or SNAIL in the lung cancer cell line A549 results in increased sensitivity to cisplatin.88,89 Furthermore, the role of EMT in the resistance to EGFR TKI in lung cancer has been extensively studied. Yauch et al. first demonstrated that a highly epithelial phenotype, as measured by high E-cadherin expression, correlates with sensitivity to EGFR-TKI.90 Subsequently, Witta et al. demonstrated that an epithelial phenotype, as measured by E-cadherin expression, significantly correlates with sensitivity to gefitinib in EGFR wild-type NSCLC cell lines and that exogenous expression of E-cadherin sensitizes NSCLC cell lines to gefitinib, indicating that the mesenchymal phenotype is an important determinant for chemosensitivity to EGFR-TKI.91 Of note, they also found that ZEB1 expression significantly correlated with EGFR-TKI sensitivity. In addition, recent studies have shown that EMT plays an important role in EGFR-mutant NSCLC cells, which usually exhibit exquisite sensitivity to EGFR-TKI. A study reported that expression of Slug but not Snail, Twist, or ZEB1 correlates with acquired resistance to gefitinib in EGFR-mutant NSCLC cell line PC9, suggesting that Slug may play a role in acquired resistance to EGFR-TKI in the EGFR-mutant lung cancer cells. Furthermore, Yao et al. demonstrated that a subpopulation of EGFR-mutant H1650 cell line with mesenchymal transition multiplies after gefitinib treatment, thereby contributing to acquired gefitinib resistance.92 This gefitinib treatment-induced EMT occurs primarily through TGF-β-dependent secretion of interleukin-6. Recently, another group also reported that EMT confers resistance to EGFR-TKI treatment in EGFR-mutant lung cancer cell line.49 Most importantly, the critical role of EMT in EGFR-TKI resistance was also shown in the clinical setting by Sequist et al., who revealed by serial sampling of EGFR-mutant lung cancer tumours that a subset of tumours with acquired EGFR-TKI resistance exhibits EMT.11


The cancer stem cell hypothesis postulates that a subpopulation of tumour cells within the bulk of the tumour harbours the ability to self-renew and to give rise to ‘differentiated’ tumour cells and to varying degrees contribute to the maintenance and progression of the tumour. The association between EMT and cancer stem cells was first reported by Mani et al., who demonstrated that induction of EMT in immortalized mammary epithelial cells caused them to become stem cell like, as indicated by their CD44high/CD24low antigenic phenotype, and that mammary epithelial cells enriched for stem-like cells by sphere culture acquire mesenchymal properties.4 This association in lung cancer was demonstrated by a paper reporting that simultaneous ectopic expression of two stemness-inducing transcription factors in lung adenocarcinoma cells causes the tumour cells to increase in stemness as well as in mesenchymal properties.19 In addition, recently, another paper demonstrated that TGF-β treatment of lung cancer cell lines induces EMT as well as increased stemness.10 The fact that many of the properties associate with EMT also seem to be part of the phenotypic repertoire associated with cancer stem cells begs the question as to whether there are really different phenotypes, or is it really a question of degree? Certainly, further research in this area will clarify this question.


Accumulating evidence suggests that EMT is both a biologically and clinically relevant process in the development of lung cancer. However, more evidence, especially from clinical samples, is required to firmly substantiate the importance of EMT in lung cancer. From the translational point of view, targeting the EMT pathway appears to be a very promising therapeutic strategy for lung cancer because EMT is involved in most of the important malignant properties of cancer cells.93 In particular, a link between EMT and stemness seems quite attractive because EMT-targeted therapy may well impact the stem cell component of tumours. However, we should be very cautious about the possibility that EMT-targeted therapy may produce serious adverse effects because EMT plays a critical role in normal tissues.


This work was supported by a Grant-in-Aid for Scientific Research (C) 23591145 (to M. Sato), and a Grant-in-Aid for Scientific Research (B) 21390257 (to Y. Hasegawa) from the Japan Society for the Promotion of Science and Global Center of Excellence (COE) programme at Nagoya University Graduate School of Medicine, which is funded by Japan's Ministry of Education, Culture, Sports, Science and Technology.