Members of the ezrin–radixin–moesin (ERM) family of proteins are involved in multiple aspects of cell migration by acting both as crosslinkers between the membrane, receptors and the actin cytoskeleton, and as regulators of signalling molecules that are implicated in cell adhesion, cell polarity and migration. Increasing evidence suggests that the regulation of cell signalling and the cytoskeleton by ERM proteins is crucial during cancer progression. Thus, both their expression levels and subcellular localisation would affect tumour progression. High expression of ERM proteins has been shown in a variety of cancers. Mislocalisation of ERM proteins reduces the ability of cells to form cell–cell contacts and, therefore, promotes an invasive phenotype. Similarly, mislocalisation of ERM proteins impairs the formation of receptor complexes and alters the transmission of signals in response to growth factors, thereby facilitating tumour progression. In this Commentary, we address the structure, function and regulation of ERM proteins under normal physiological conditions as well as in cancer progression, with particular emphasis on cancers of epithelial origin, such as those from breast, lung and prostate. We also discuss any recent developments that have added to the understanding of the underlying molecular mechanisms and signalling pathways these proteins are involved in during cancer progression.
Ezrin, radixin and moesin (ERM) are three highly homologous protein members of the FERM (4.1-band ERM) superfamily (Sato etal., 1992). Fundamentally, they are essential for linking the actin cytoskeleton to the cell membrane and are key organisers of specialised membrane domains, such as apical microvilli (Lan etal., 2006; Takeuchi etal., 1994), lamellipodia and filopodia (Baumgartner etal., 2006; Lamb etal., 1997). Their function as cytoskeletal linkers places them at the centre of an elaborate regulatory network of many cellular processes, whether under normal and controlled conditions, such as migration, growth and adhesion, or in pathological scenarios, such as cancer cell invasion and metastasis. Importantly, several studies support a role for ERM proteins in many fundamental signal transduction pathways to influence cell adhesion (Pujuguet etal., 2003; Takeuchi etal., 1994), migration (Crepaldi etal., 1997; Naba etal., 2008; Ng etal., 2001; Valderrama etal., 2012) and morphogenesis (Bretscher etal., 2002; Crepaldi etal., 1997; Gautreau etal., 2000; Hsu etal., 2012) in response to extracellular cues. There is also evidence that the ERM proteins are important for cell–cell and cell–matrix contacts – potentially through interactions with cadherin complexes and integrins (Jung and McCarty, 2012; Pujuguet etal., 2003) – as well as for the reorganisation of the cytoskeleton (Belkina etal., 2009; Bretscher, 1983; Gautreau etal., 1999); yet, the molecular mechanisms in which they participate within pathological processes such as cancer still remain ill-defined. In the following sections, we provide an overview of the structure and regulation of ERM proteins, and subsequently focus our attention on recent advances that help to elucidate the role of these proteins in cancer progression.
ERM structure and regulation
ERM protein structure
ERM proteins have been highly conserved through evolution, presenting more than 75% amino acid identity within the common FERM domain and F-actin-binding site that are shared between all members of this superfamily (Gary and Bretscher, 1995). ERM proteins consists of a FERM domain (∼300 amino acids) situated at the N-terminus, a C-terminal region (∼100 amino acids), and an α-helical region (∼200 amino acids) that links the C-terminus and the FERM domain (Fig. 1) (Algrain etal., 1993; Gary and Bretscher, 1995; Turunen etal., 1994). The FERM domain comprises three lobes that are arranged as a cloverleaf – F1, F2 and F3 – and mediates the link between the ERM proteins and specific membrane-bound proteins localised in actin-rich regions. The C-terminal domain, also referred to as the C-terminal ERM-associated domain (C-ERMAD), contains the F-actin-binding site, which mediates the link between the actin cytoskeleton and the plasma membrane (Berryman etal., 1993; Pearson etal., 2000).
The C-terminal domain of ERM proteins is capable of binding to the FERM domain of the same molecule, through self-association, to form a monomer or, with another ERM molecule, to form a homo- or heterodimer, as observed for ezrin and moesin (Fig. 1) (Gary and Bretscher, 1995). This intramolecular interaction leads to a closed conformation or head–tail interaction, and the consequent masking of both the membrane and actin binding sites, resulting in inactivation of the proteins (Bretscher etal., 2002; Gary and Bretscher, 1995). The ERM proteins will remain in this inactive conformation until the FERM domain binds to phosphatidylinositol (4,5)-bisphosphate [PtdIns(4,5)P2] (Bretscher etal., 2002; Fievet etal., 2004; Niggli etal., 1995; Pearson etal., 2000) – which is located at the cell membrane – and the conserved threonine residues – T567, T564 and T558, for ezrin, radixin and moesin, respectively – become phosphorylated (Fig. 2) (Fievet etal., 2004; Matsui etal., 1998; Nakamura etal., 1995; Niggli and Rossy, 2008). This activation results in directly linking the actin cytoskeleton to the plasma membrane through the positively charged F3 lobe of the FERM domain, which has been proposed to have a high affinity for negatively charged phospholipids within the plasma membrane (Pearson etal., 2000). Alternatively, ERM can bind to the PDZ domains (protein–protein recognition modules) of other scaffolding proteins, such as ERM-binding phosphoprotein 50 (EBP50; also known as Slc9a3r1) or Na+/H+ exchange regulatory factor 2 (NHERF2; also known as Slc9a3r2) (Arpin etal., 2011; Garbett and Bretscher, 2012; Gautreau etal., 1999; Harris and Lim, 2001; Reczek etal., 1997). As their role in signalling has become more apparent, further studies have now demonstrated that all three proteins can bind to several transmembrane receptors, including several receptor kinases (RTKs), epidermal growth factor receptor (EGFR), hepatocyte growth factor receptor (HGFR), platelet-derived growth factor receptor (PDGFR) (Berryman etal., 1993; Crepaldi etal., 1997; Krieg and Hunter, 1992; Matsui etal., 1998; Naba etal., 2008; Orian-Rousseau etal., 2007) and to co-receptors, such as CD44, CD43, ICAM1 and ICAM2 (Fig. 2) (Matsui etal., 1998; Orian-Rousseau etal., 2007; Takeuchi etal., 1994).
It has been proposed that the conserved threonine residues of these proteins can be phosphorylated in a tissue and cell-dependent manner through, for example, Rho-associated kinase (ROCK) (Hébert etal., 2008; Matsui etal., 1998), protein kinase Cα (PRKCA, hereafter referred to as PKCα) (Ng etal., 2001), NcK-interacting kinase (MAP3K14, hereafter referred to as NIK) (Baumgartner etal., 2006) lymphocyte-orientated kinase (STK10; hereafter referred to as LOK) (Belkina etal., 2009; Viswanatha etal., 2012) and others (see Table 1), which indicates that ERM proteins are involved in a wide variety of signalling pathways (Fig. 2). Ezrin has three tyrosine sites (Y145, Y353 and Y477) that are phosphorylated through several kinases (Table 1), which can lead to the activation of different signalling pathways (Gautreau etal., 1999; Mak etal., 2012; Srivastava etal., 2005).
It is now accepted that the association between ERM proteins and receptors at the plasma membrane is important for mediating many of these pathways. In particular, their association with CD44, EGFR and HGFR has been suggested to have a role in the migration process that is associated with cancer cell invasion and metastasis (Chen and Chen, 2006; Mak etal., 2012; Martin etal., 2003; Zhu etal., 2012). However, there are conflicting data with regard to which kinases are responsible for not only the activation but the phosphorylation and regulation of ERM proteins. For example, it has been suggested that ROCK phosphorylates ERM proteins in vitro (Estecha etal., 2009; Hébert etal., 2008; Matsui etal., 1998), but other studies suggest that ROCK does not phosphorylate ERM proteins in vivo, and that their activation instead involves PtdInsP2 and Rho (Jeon etal., 2002; Matsui etal., 1999); it has also been suggested that the Rho/ROCK pathway is not essential for this phosphorylation event (Yokoyama etal., 2005). The serine/threonine-protein kinase MST4 has been shown to phosphorylate ezrin (ten Klooster etal., 2009); however, knockdown of this kinase in the human placental choriocarcinoma cell line JEG-3 that is known to express MST4 had no effect on the level of ezrin phosphorylation (Viswanatha etal., 2012). The majority of relevant studies that used a variety of established immortalised cell lines have implicated particular kinases; but once further work is carried out in vivo, it may become apparent that activation and regulation of these proteins is dependant not only on the abundance of a particular kinase, but also on its temporal–spatial expression in different cells, tissues or organs. Although it would be beneficial to discuss such discrepancies further, this is outside the scope of this Commentary and we, therefore, focus our attention on the molecular mechanisms and kinases that have been suggested to be involved in cancer progression.
ERM proteins in cancer progression
ERM proteins have been mainly characterised in their role in epithelial morphogenesis, adhesion and migration (Arpin etal., 2011; Fehon etal., 2010; Bretscher etal., 2002), three key events that, when altered, can participate in the induction of a cancer phenotype. Therefore, it is not surprising that ERM proteins have been suggested to influence tumour progression. Many clinical studies have linked high expression of ezrin with poor outcome in patients that suffer from a wide variety of cancers, including breast, lung or prostate, and only little attention has been given to radixin and moesin. The mislocalisation of these proteins can interfere with the formation of receptor complexes and leads to changes in signalling pathways in response to growth factors, which could aid the progression of a tumour (Arpin etal., 2011) (Fig. 3). It is becoming increasingly apparent that all three ERM proteins can modulate the activity of the Ras superfamily of small GTPases, including the Ras and Rho subgroups, by interacting with their guanine exchange factors (GEFs) and guanine nucleotide dissociation inhibitors (GDIs), which are vital for their activation (Sperka etal., 2011; Takahashi etal., 1997; Valderrama etal., 2012). Ras is a molecular switch that is important for RTK-dependent activation of the MAP kinase pathway, such as through EGFR and HGFR (Easty etal., 2011; Orian-Rousseau etal., 2007; Ridley etal., 1995). ERM proteins have been proposed to be responsible for a new step in Ras activation in that they are recruited to F-actin and co-receptors, specifically to receptors involved in cell adhesion, such as β1 integrin and CD44, at the plasma membrane in response to activation of RTKs. In doing so, ERM proteins can form a complex that acts as a scaffold and binds both GDP-bound Ras and the N-terminal region of the GEF son of sevenless homolog (SOS) (Sperka etal., 2011). Ezrin has specific binding domains for CD44, Ras and the autoinhibitory Dbl-homology–pleckstrin-homology (DH–PH) domain of SOS but only after ezrin is in a complex with F-actin and its co-receptor can it bind to the DH–PH domain of SOS and expose the allosteric binding site that is specific for GDP binding of Ras (Sperka etal., 2011).
All three ERM proteins have been suggested to act as scaffolds in Ras activation, and it is well known that oncogenic mutations in Ras, or even in the RTKs EGFR and HGFR, can result in the development of a cancerous phenotype (Bonaccorsi etal., 2004; Fehon etal., 2010; Ridley, 2001). Therefore, it is tempting to speculate that ERM overexpression, which is commonly seen in a variety of cancers, is the cause for an increase in the activation of these signalling pathways. Recently, the role of the three ERM proteins in tumour progression has become the focus of attention, in order to determine the exact molecular roles these proteins have during cancer progression. Below, we discuss the roles of each of the ERM proteins individually, and aim to present a comprehensive argument against functional redundancy between these three homologous proteins in cancer progression.
The expression of ezrin in over 5000 human cancers, including breast, lung and prostate cancers compared with normal tissues, was analysed in a large-scale study that used microarray immunohistochemistry (Bruce etal., 2007). Although ezrin is expressed in most normal and cancer tissues, its expression is significantly higher in cancers of mesenchymal origin (sarcomas) and in prostate cancer when compared with breast cancer, and shows no change in lung cancer tissue (Bruce etal., 2007). In ezrin, mutation of Y477, a tyrosin residue not conserved in radixin or moesin, reduced cell migration in the highly metastatic mouse mammary carcinoma cell line AC2M2 in 3D cultures (which is commonly used to recapitulate tumourigenesis) (Debnath and Brugge, 2005) as well as local tumour invasion in mice (Mak etal., 2012). Because Src is responsible for the phosphorylation of ezrin at Y477 (Elliott etal., 2004; Srivastava etal., 2005), it has been suggested that the ezrin–Src pathway is a potential prognostic marker for invasive breast cancer in humans (Mak etal., 2012). Because sarcomas have a mesenchymal phenotype, it would be interesting to determine whether Src phosphorylation of ezrin at Y477 has a role in the activity of these cancers.
The breast cancer cell lines MCF10A (semi-normal) and MDA-MB-231 (metastatic) have been used in a 3D Matrigel model, in which silencing of ezrin by using short hairpin RNA (shRNA) leads to a decreased invasive potential (Konstantinovsky etal., 2012). Abnormal ezrin distribution has been correlated with poor prognosis of breast cancer patients, where ezrin is found to relocalise from the apical regions observed in non-tumourigenic cell lines to motile projections in invasive cell lines and the cytoplasm of breast tumours (Sarrió etal., 2006). Phosphorylation of ezrin Y145 also led to cell spreading in the mouse mammary carcinoma cell line SP1 (Elliott etal., 2004) and in pig kidney epithelial (LLC-PK1) cells (Srivastava etal., 2005).
Interestingly, and not observed in the large-scale screen mentioned above (Bruce etal., 2007), ezrin has also been implicated in lung cancer (Li etal., 2012). Here, ezrin expression was elevated in the highly metastatic human lung cancer cell lines LTE, BE1 H446 and H460 when compared with the low metastasis human lung adenocarcinoma cell lines SPC and A549. Furthermore, knockdown of ezrin by using small interfering RNA (siRNA) resulted in significant reduction in proliferation, migration and invasion of these cells in vitro (Li etal., 2012). As in breast cancer cells, localisation of ezrin in lung cancer cells also appears to be altered, i.e. from the apical membrane in normal bronchial epithelium to the cytoplasm in lung cancerous cells. This suggests that the relocalisation of ezrin is a significant factor in altering the activation of signalling pathways and of actin reorganisation, which both are fundamental for cancer progression. In the same study, reduced β-catenin and increased E-cadherin expression was observed with no change in the respective mRNA levels. Developmental biology studies have highlighted a role for ERM proteins in junction remodelling (Dard etal., 2001). For instance, mutations in ezrin within the conserved threonine residue prevent the formation of E-cadherin-mediated cel–cell contacts during blastocyst formation in mice (Dard etal., 2001). It has been suggested that the small pool of ezrin that is present at cell–cell contacts is sufficient to recruit Fes kinase – a non-RTK that has also been implicated in cancer progression (Condorelli etal., 2011) – and to modulate junction formation by directly interacting with the SH2 domain of Fes kinase (Naba etal., 2008). This interaction is important for the localisation and activation of Fes, which consequently results in the disassembly of cell–cell contacts. Moreover, ERM proteins are crucial for Rac-dependent assembly of focal adhesions and adherens junctions (Mackay etal., 1997; Pujuguet etal., 2003). During the formation of cell–cell contacts, fluctuation in the expression levels and downregulation of Rac and Rho GTPases are known to occur (Yamada and Nelson, 2007). Ezrin has been proposed to regulate E-cadherin-dependent adherens junction assembly through Rac1 activation and the trafficking of E-cadherin to the plasma membrane (Pujuguet etal., 2003). Therefore, it is possible that overexpression of ezrin prevents the required fluctuations in the activity of Rac1, as well as the localisation of E-cadherin at the plasma membrane, consequently disrupting cell–cell contacts that are commonly seen in tumour development (Arpin etal., 2011). As deregulation of E-cadherin function is believed to promote tumour progression (Canel etal., 2013), it is possible that ezrin has a role in the post-transcriptional regulation of E-cadherin during cancer progression, although further investigations are required to better define the molecular mechanisms underlying this role.
Many immunohistochemical studies on high-grade prostate intraepithelial neoplasia (HGPIN) – the proposed precursor to prostate cancer – and advanced cancerous prostate tissues have also shown overexpression of ezrin (Pang etal., 2004; Valdman etal., 2005). However, expression of ezrin is in fact much higher in HGPIN tissues compared with advanced prostate cancer specimens (Pang etal., 2004). Importantly, this change in expression does not appear to be the result of genomic alterations (Amler etal., 2000). Depletion of ezrin, or overexpression of a dominant-negative (T567A) or non-phosphorylatable (Y353F) mutant significantly reduces invasion (Chuan etal., 2006). Thus, the level of ezrin phosphorylation may be important in the tumourigenesis of precursor lesions – such as that of HGPIN – and, consequently, may trigger carcinogenic tissue to become invasive before ezrin is potentially deregulated in more advanced stages of the disease (Pang etal., 2004; Valdman etal., 2005).
The observed increase of ezrin expression in prostate cancer has been suggested to be a result of increased expression of the oncogene Myc (Chuan etal., 2010). In the presence of androgens such as testosterone it appears that Myc can bind to the canonical E-box in the proximal promoter region of ezrin and induce its transcription. However, ezrin itself can regulate the level of Myc through the PI3K/Akt pathway, effectively creating a positive-feedback loop. This regulatory loop has been shown to be essential for cell proliferation and invasion in both androgen-dependent and -independent metastatic prostate cell lines (Chuan etal., 2010). Under normal physiological conditions, the phosphorylation of ezrin Y353 is important for survival of epithelial cells through activation of the PI3K/Akt pathway, and 3D cell cultures of the epithelial cell line LLC-PK1 showed that ezrin binds to p85, the regulatory subunit of PI3K, in order to mediate the PI3K/Akt pathway (Gautreau etal., 1999). Therefore, overexpression of ezrin as a result of Myc overexpression could, in turn, maintain the survival of cancer cells through activation of the PI3K/Akt pathway (Fig. 3).
As mentioned above, high expression of ezrin has been observed in breast cancer cell lines, as well as in tissue samples of breast cancer patients (Bruce etal., 2007; Konstantinovsky etal., 2012; Mak etal., 2012; Sarrió etal., 2006). The recent discovery of a new role for BRCA1, a well-established tumour suppressor gene in breast and ovarian cancer, in cell motility and invasion might shed light on the regulation of ezrin in breast cancer (Coene etal., 2011). Here, depletion of BRCA1 resulted in increased cell spreading and motility of single cells, as well as increased invasive behaviour of cells in wound healing assays in vitro (Coene etal., 2011). BRCA1 was also suggested to colocalise with radixin and moesin, therefore this interaction is not unique to ezrin (Coene etal., 2011). Could BRCA1 target ezrin in order to regulate its expression? The BRCA1 C-terminus (BRCT) domain, the binding domain of BRCA1, has been shown to interact directly with ezrin in complex with F-actin at the plasma membrane, as well as at membrane extensions and focal adhesion sites. In addition, it has been proposed that BRCA1, through its E3 ubiquitin ligase activity, regulates ezrin expression. Therefore, loss of BRCA1, as commonly seen in breast cancer cells, could result in an accumulation of ezrin that not only leads to its reported high expression in breast cancer cells but also an increased activity of specific signalling pathways that involve ezrin and regulate cell-cell contacts, cell adhesion and cell migration (Fig. 3) (Coene etal., 2011).
A so far unknown interaction between ezrin and the desomosome protein desmoglein 3 (Dsg3) has recently been identified; and Dsg3 appears to colocalise with ezrin at the plasma membrane (Brown etal., 2013). Overexpression of Dsg3 in the oral squamous cell carcinoma (OSCC) cell lines A431 and SqCC/Y1 caused a marked increase in phosphorylation of the conserved ezrin residue T567, which resulted in increased migration and invasion in 3D cell culture assays. It has been suggested that ezrin is phosphorylated by PKC after its colocalisation with Dsg3, an interaction that might be responsible for the metastatic potential of OSCC cell lines (Brown etal., 2013). Ezrin has previously been proposed to act as a downstream signalling molecule in PKC-induced cell migration; here, PKCα phosphorylates specifically the conserved T567 residue on ezrin (Ng etal., 2001), which results in CD44-dependent directional migration (Legg etal., 2002). Therefore, Dsg3, through its interaction with PKC, might mediate the activation of ezrin and, consequently, contribute to an invasive phenotype (Fig. 3).
As well as breast, lung and prostate cancers, and oral squamous cell carcinomas (OSCCs), ezrin overexpression has also been suggested to influence tumour metastasis in other cancers, such as osteosarcoma, rhabdomyosarcoma and carcinomas of the pancreas (Akisawa etal., 1999; Hunter, 2004; Khanna etal., 2004; Meng etal., 2010; Shang etal., 2012; Yu etal., 2004).
Compared with ezrin, only very little is known with regard to the role of radixin in cancer. However, significantly different expression patterns of radixin have been recently observed between HGPIN and prostate cancer samples (Bartholow etal., 2011). This is the first report that describes the differences in the expression profiles of radixin between normal donor prostates, HGPIN, prostate cancer and normal tissue adjacent to adenocarcinoma. Similarly to ezrin, radixin expression was significantly higher in HGPIN compared with prostate cancer, suggesting that radixin has an initial role in the progression of the tumour but is then deregulated as the tumour progresses. A role of radixin in pancreatic cancer has also recently been suggested (Chen etal., 2012). Here, knockdown of radixin expression using shRNA decreased proliferation, survival, adhesion and invasive potential of human pancreatic carcinoma, epithelial-like cells (PANC-1) in vitro. Similar effects, i.e. the significant inhibition of tumour growth, was also seen when cells, in which radixin has been silenced, were implanted into nude mice (Chen etal., 2012). Furthermore, knockdown of radixin in PANC-1 cells resulted in an increase in E-cadherin expression, which is also seen in ezrin-depleted cells (Chen etal., 2012). This implies that both ezrin and radixin may have an important role in the disruption of cell–cell contacts through the downregulation of E-cadherin and its relocalisation during cancer progression (Chen etal., 2012; Condorelli etal., 2011; Li etal., 2012; Pujuguet etal., 2003).
We recently found a new role for radixin in cell–cell adhesion and cell migration in cells of the commonly used highly metastatic prostate cell line PC3 that may further elucidate the degree of redundancy between the three ERM proteins. We observed that depletion of radixin by using siRNA not only promotes an increase in the spreading of PC3 cells and in N-cadherin-mediated cell–cell adhesion, but that their morphology also changed and appeared to become more epithelial-like (Valderrama etal., 2012). Radixin depletion has also been shown to lead to an increase in Rac1 activity, in a manner that is dependent on Vav, a GEF responsible for activation of Rac1 (Abe etal., 2000). As mentioned above, ERM proteins can act as scaffolds for both the activation and deregulation of Rac1; so, potentially, radixin acts as a scaffold to deregulate Vav expression. In the absence of radixin, Vav could increase its expresssion and/or activity and, consequently, reduce Rac1 activity, thereby mediating epithelial polarity (Fig. 3). Importantly, the depletion of ezrin or moesin does not induce a similar response, suggesting that the regulation of Vav is specific to radixin (Valderrama etal., 2012). Although, an alteration in the localisation of radixin in normal epithelium compared with cancerous tissues remains to be investigated, these recent results suggest that the role of radixin is important in defining cell polarity and subsequent acquisition of malignancy, independently of ezrin and moesin (Valderrama etal., 2012).
The expression of moesin, like that of the other two ERM proteins, has been correlated to cancer progression. Its expression pattern has been linked to increased tumour size, as well as mode of invasion and differentiation in OSCC (Kobayashi etal., 2004). In addition, moesin relocalises from the plasma membrane to the cytoplasm in tumour cells that have a higher incidence of lymph node metastasis. This suggests that, in OSCC, mislocalisation of moesin together with increased expression levels influences the invasive or metastatic ability of the tumour cells (Kobayashi etal., 2004). Moesin has also been linked to early lung colonisation by melanoma cells in vivo (Estecha etal., 2009). Here, established metastatic lung melanoma cell lines that have been depleted of moesin by using siRNA were injected into the tails of mice together with ezrin-depleted cells. Moesin-depleted cells then arrived at the lungs, but were unable to colonise and disappeared from the lungs, whereas ezrin-depleted cells remained, suggesting that moesin is essential for colonisation of tumour cells in the lungs. When the moesin-depleted cells where analysed in a 3D collagen invasion assay, they exhibited a flattened morphology and showed significantly less invasive potential compared with the control cells, which quickly acquired an elongated phenotype. During the initial adhesion of moesin-depleted cells in the 3D assay, there was an increase in phosphorylated moesin, which was repressed when PKC and ROCK were inhibited. Therefore, PKC and ROCK might contribute to the initial adhesion-dependent activation of moesin. Moesin is also required to activate RhoA signalling in response to initial attachment and spreading. Therefore, it is possible that moesin is activated initially by PKC – independently of Rho – and, subsequently enters a positive moesin and RhoA feedback loop to promote Rho activation and cell migration (Estecha etal., 2009) (Fig. 3).
Many studies have shown that epithelial–mesenchymal transition (EMT) of epithelial cells is induced by transformed growth factor-beta (TGF-β) (Nawshad etal., 2005; Wendt etal., 2009). This transition is fundamental in tumour development and metastasis and allows epithelial cells to develop a mesenchymal and, therefore, more invasive phenotype. Moesin has been predicted to be a potential EMT marker in breast and pancreatic cancer (Abiatari etal., 2010; Haynes etal., 2011; Wang etal., 2012). Furthermore, expression profiles of both breast and basal breast carcinomas have shown strong upregulation of moesin (Charafe-Jauffret etal., 2007; Condeelis etal., 2005; Kobayashi etal., 2004; Wang etal., 2012). Moesin was also shown to be essential for EMT in the human mammary cell line MCF-10A, whereas ezrin and radixin were not (Haynes etal., 2011); after stimulation with TGF-β, moesin was observed to relocalise from cell–cell adhesions to filopodia and large membrane protrusions. Moreover, depletion of moesin reduces the number of actin stress fibres, and the bundled filaments of the actin cytoskeleton were thinner, shorter and less uniformly aligned along the main cell axis, causing cells to be less elongated. Haynes etal. suggested that ROCK is responsible for the increase of phosphorylated moesin during EMT (Haynes etal., 2011). Moesin has been previously reported to colocalise with CD44 (Fehon etal., 2010; Legg and Isacke, 1998; Matsui etal., 1998; Ng etal., 2001). During EMT, CD44 is most abundant at dorsal protrusions to promote cell–substrate adhesion, and this localisation is drastically reduced when moesin is depleted (Haynes etal., 2011). This suggests that moesin has a dual role in promoting EMT by affecting the reorganisation of the actin cytoskeleton and by regulating the localisation of CD44.
When cells are stimulated with TGF-β in the absence of moesin, there is also a significant reduction in the level of autophosphorylated focal adhesion kinase (FAK) (Haynes etal., 2011), a non-RTK that localises at focal adhesions (Parsons, 2003). Like ERM proteins, FAK also contains a FERM domain (Ceccarelli etal., 2006), with the characteristic F1, F2 and F3 lobes; and, through F2, FAK directly binds Met, the transmembrane receptor that is activated in response to HGF (Chen and Chen, 2006). Inhibition of FAK impairs cell migration (Mitra etal., 2005) and, therefore, increased expression of moesin upon stimulation of TGF-β may have a role in the regulation of FAK activation, which in turn promotes cancer progression (Fig. 3) (Chen and Chen, 2006). This study describes the first evidence of a functional link between moesin and FAK, but it remains unclear how moesin might mediate the phosphorylation of FAK during EMT.
Recently, Zhu etal. demonstrated a correlation between the high expression of moesin and high-grade glioblastoma tumours without any significant changes in the expression levels of ezrin or radixin (Zhu etal., 2013) and also observed elevated moesin expression in established glioblastoma cell lines (Zhu etal., 2013). Here, moesin was found to colocalise with CD44 at membrane extensions, and the authors propose that moesin competes with the tumour suppressor NF2 for binding to CD44 in order to activate CD44-induced growth signalling. NF2 has been previously shown to colocalise with CD44 and to form a molecular switch that results in arrest of cell growth (Morrison etal., 2001). Overexpression of moesin might, therefore, displace the tumour suppressor NF2 during glioblastoma progression and promote growth-signalling pathways. Interestingly, one pathway that is predominantly activated downstream of moesin–CD44 is the Wnt/β-catenin pathway (Zhu etal., 2013). The interaction between moesin and CD44 was found to induce transcription of β-catenin, as well as translocation of β-catenin from the membrane to the nucleus. Mechanistically, this can be achieved by PI3K/Akt-mediated phosphorylation (downstream of CD44 activation) of β-catenin at S552, as previously suggested to induce β-catenin translocation to the nucleus (Fang etal., 2007). The results presented in this recent study, therefore, suggest that moesin represents a target for glioblastoma therapy (Zhu etal., 2013).
Conclusions and perspectives
ERM proteins have different fundamental roles in tumour development and the ability to subsequently promote the progression of tumours to more advanced stages. They take part in a number of signalling pathways that depend, for example, on RhoGTPases, PI3K/Akt, Wnt/β-catenin, CD44 and RTKs – such as EGFR and HGFR – all of which are crucial for cancer progression (Fig. 3).
Although there is some overlap in the involvement of ERM proteins in tumour progression, it is important to emphasise that many of the roles discussed in this Commentary highlight the individuality of each of them. However, further investigations that focus on how each protein exerts their role in the above mentioned pathways are needed, and more information regarding the activation mechanisms and interaction and/or regulatory partners is required to reinforce their potentially specific roles in different epithelial cancers, and their possible use as relevant prognostic markers.
We thank Guy Whitley, Francesc Miralles (St George's University of London, UK), Aleksander Ivetic and Valerie Vivancos (King's College London, UK) for their comments and insights during the preparation of this manuscript.
This work is supported by a SGUL PhD studentship to J.C.
The authors declare no competing interests.