Dual role of E-cadherin in cancer cells

ABSTRACT

E-cadherin is the main component of epithelial adherens junctions (AJs), which play a crucial role in the maintenance of stable cell–cell adhesion and overall tissue integrity. Down-regulation of E-cadherin expression has been found in many carcinomas, and loss of E-cadherin is generally associated with poor prognosis in patients. During the last decade, however, numerous studies have shown that E-cadherin is essential for several aspects of cancer cell biology that contribute to cancer progression, most importantly, active cell migration. In this review, we summarize the available data about the input of E-cadherin in cancer progression, focusing on the latest advances in the research of the various roles E-cadherin-based AJs play in cancer cell dissemination. The review also touches upon the “cadherin switching” in cancer cells where N- or P-cadherin replace or are co-expressed with E-cadherin and its influence on the migratory properties of cancer cells.

1. Introduction

In epithelial cells, the main stable cell–cell adhesion structures are linear E-cadherin-based adherens junctions (AJs), which form a continuous belt around the apical part of a cell (zonula adherens, ZA), tightly associated with the circumferential actin bundle.^1–3 Despite the fact that in many carcinomas, the bulk of tumor cells may retain epithelial morphology, during initiation of invasion-metastasis cascade carcinoma cells undergo epithelial–mesenchymal transition (EMT), which allows them to detach from neighboring cells, invade the basement membrane and migrate through tissues.⁴ During EMT, tumor cells lose apical-basal polarity, stable AJs and the circumferential actin bundle and acquire front-rear polarity forming lamellipodia at the leading edge. Weakening of cell–cell adhesion during EMT may be achieved by down-regulation of E-cadherin expression or replacement of stable linear E-cadherin-based AJs by punctate AJs, which are characterized by continuous remodeling.^5,6 During the last decade, numerous clinical observations in line with experimental mouse model data have demonstrated that rather than complete EMT, carcinoma cells frequently undergo partial EMT (pEMT) by acquiring mesenchymal traits while retaining epithelial markers, in particular, E-cadherin [reviewed in 4,5].

2. E-cadherin adhesive function in epithelial cells. Molecular organization of AJs

Epithelial AJs are formed via Ca²⁺-dependent trans-interaction of E-cadherin molecule extracellular domains at juxtaposed cell membranes of adjacent cells that involve partial swapping of the N-terminal β-strand in the EC1 domains stabilized by docking of a Trp residue (W2) of one E-cadherin molecule into a pocket in the EC1 domain of the other E-cadherin molecule.^2,7,8 Extracellular domains of E-cadherin molecules also interact in cis manner via residues in the EC1 and EC2 domains; these cis-interactions support cadherin clustering and significantly increase the stability of AJs.^9–11 The cytoplasmic domain of E-cadherin binds to the members of the catenin protein family, β-catenin and p120 (Figure 1). β-catenin binds to the N-terminus of α-catenin, and this interaction promotes linking of α-catenin to actin filaments. Mechanical force is required for binding the cadherin–catenin complex to filamentous actin.^12,13 Contractile forces generated by actin-coupled myosin II induce conformational changes in actin-binding domain of α-catenin, which enhance actin binding.¹⁴ The central part of the α-catenin molecule contains a vinculin-binding domain. Force-dependent destabilization of the interactions between MI vinculin-binding domain and MII and MIII inhibitory domains of α-catenin leads to unfolding of the α-catenin molecule and allows recruitment of vinculin, which, in turn, also binds to actin filaments.^13,15–19 F-actin binding is necessary for AJ clustering and stability.^12,20 In epithelial cells, junctional actin, tightly associated with the circumferential actin bundle, is crucial for the assembly and maintenance of AJs.^1,2

Figure 1.

Structural organization of epithelial AJs. 1 and 2: nascent AJs. The extracellular domains of E-cadherin molecules (green) from two adjacent cells interact with each other through their EC1 domains (white dots). The catenin family proteins (p120 and β-catenins: pink and red, respectively) interact with the intracellular domain of E-cadherin. α-catenin (dark blue) binds to β-catenin and actin filaments (yellow) (1). Upon unfolding under acto-myosin generated tension (myosin – aqua), α-catenin binds to vinculin (light blue) (2) which further enhances actin binding. 3: a mature linear AJ. Clustering of E-cadherin molecules increases the strength of cell–cell adhesion. Junctional actin now includes various actin-binding proteins: afadin (brown), EPLIN (light green), VASP (purple), myosin (aqua), α-actinin (orange), and palladin (dark green).

Establishment and maintenance of linear AJs require both myosin IIA and IIB isoforms. The morphology of linear AJs, compaction of epithelial cells and integrity of epithelial layers depend on acto-myosin contractility.²¹ Depletion of myosin IIA leads to disruption of linear AJs in the apical part of epithelial cells. Depletion of myosin IIB leads to decrease in actin content in circumferential actin bundles associated with linear AJs.^22,23 An adaptor protein afadin is also essential for the proper organization of actin bundles at AJs. Using afadin-KO cells stably expressing afadin mutants with various deleted regions, it was shown that afadin binds to α-catenin through its coiled-coil (CC) region.²⁴ Another actin-binding protein, EPLIN, additionally stabilizes the circumferential actin bundle by inhibiting actin depolymerization and crosslinking actin filaments.²⁵ Several studies have demonstrated that siRNA-mediated depletion of EPLIN resulted in disappearance of circumferential actin bundles and converted linear AJs into punctate AJs associated with radial actin bundles.²⁶

Actin filaments nucleate at AJs,²⁷ but the mechanisms of actin filament polymerization at AJs are not well understood. It has been suggested that the Arp2/3 complex is involved in the nucleation of junctional actin filaments as it was shown that establishment of E-cadherin-based adhesion promotes recruitment of Arp2/3 to the AJs.²⁸ Activators of the Arp2/3 complex, such as WAVE and N-WASP, have also been detected at epithelial AJs.^22,29 Cortactin, which directly binds WAVE2 and Arp2/3 at ZA, may also regulate the junctional actin cytoskeleton.³⁰ Formins promoting elongation of linear actin filaments may be involved in the assembly of actin filaments at AJs. It was demonstrated that RhoA effector Dia1 was essential for the formation and maintenance of linear AJs in MCF-7 cells.³¹ In MCF10A cells grown in Matrigel, the junctional actin assembly was mediated by Formin-like 2 downstream of Rac1.³²

As is well known, epithelial cells exhibit apical-basal polarity of membrane domains, protein complexes, and cytoskeletal components. ZA is tightly associated with tight junctions (TJs), which are responsible for the paracellular barrier function of epithelial tissues and which are also linked to the circumferential actin belt. Several ZA proteins associate with TJ proteins – for example, PLEKHA7 forms a complex with ZO-1 and cingulin.³³ E-cadherin has a crucial role in formation and maintenance of TJs.^34–36 As was shown in the extracellular Ca²⁺ depletion assay, integrity of ZA is essential for the TJ barrier function.^37,38 In summary, E-cadherin-based AJs are critically important for the establishment and maintenance of epithelial tissue architecture.

3. Molecular mechanisms of AJ disruption in cancer cells

Loss of E-cadherin plays an important role in tumor progression. Down-regulation of E-cadherin expression detected in various carcinomas correlates with aggressive behavior of the tumors and is considered a prognostic factor associated with poor patient survival.³⁹ Disruption of E-cadherin-based AJs between tumor cells facilitates their ability to migrate away from the tumor and invade adjacent tissues. Molecular mechanisms of E-cadherin down-regulation in tumors have been studied in great detail (Figure 2). Germline mutations inactivating the E-cadherin gene CDH1 and the subsequent loss of the E-cadherin protein were described in hereditary forms of diffuse gastric cancers and lobular breast cancers.^40,41 Loss of heterozygosity on chromosome arm 16q in combination with a mutation of the other allele leading to CDH1 inactivation was revealed in many cases of hepatocellular carcinoma,^42,43 prostate adenocarcinoma,⁴⁴ and infiltrating lobular breast carcinoma.^45,46 Aberrant methylation of the CDH1 gene promoter and the associated loss of E-cadherin expression was demonstrated in diffuse type gastric cancer and lobular breast carcinoma.^47–49 In some cases, loss of E-cadherin in combination with inactivation of p53 or Pten has a causal role in tumor development as has been shown, for example, for invasive lobular breast carcinoma and diffuse type gastric cancer.^50–52

Figure 2.

Molecular mechanisms of adherens junction disruption in cancer cells. In many cancers, expression of CDH1 may be suppressed or abolished through genetic or epigenetic modification. Germline mutations, loss of hetererozigosity of the CDH1 gene combined with a mutation of the other allele, and aberrant methylation of the CDH1 promotor can all contribute to the suppression of E-cadherin. Oncogenic signaling through multiple pathways leads to down-regulation of E-cadherin expression via EMT transcription factors (EMT-TFs). These pathways include Wnt, Notch, TGFβ, EGF, IGF, FGF, VEGF, HGF, IL-6, and TNF. Some of the pathways may be activated through paracrine stimulation of cancer cells in the microenvironment. Thus, TGFβ, HGF, VEGF, and IL-6 are secreted by cancer-associated fibroblasts (CAFs) and IL-6, TGFβ, EGF, and TNF by tumor-associated macrophages (TAMs). Along with hypoxia and increased extracellular matrix stiffness, this can lead to activation of EMT-TFs TWIST, SLUG, ZEB1, ZEB2, and SNAIL, which down-regulate CDH1 expression. A group of miRNAs acts in the opposite direction, dampening the activity of EMT-TFs. Additionally, miR-9 is capable of direct suppression of CDH1 expression. Aberrant signaling of receptor tyrosine kinases (RTKs) leads to phosphorylation of the E-cadherin protein, its biding to the E3 ubiquitin ligase Hakai, and subsequent lysosomal degradation. Rearrangement of the actin cytoskeleton during EMT leads to disruption of linear AJs connected to circumferential actin bundle underneath the plasma membrane (PM) and their replacement by punctate AJs connected to straight actin bundles. Punctate AJs are dynamic and subjects to constant remodeling. In cancer cells, loss of p120 expression leads to excessive endocytosis of E-cadherin. Finally, dysregulation of tight junction (TJ) proteins (particularly claudins) may also contribute to down-regulation of E-cadherin.

In many types of cancer, loss of E-cadherin is due to activation of transcription factors SNAIL, SLUG, TWIST1, ZEB1, ZEB2 and several others that bind to E-box sequences in the CDH1 gene promoter and negatively regulate its expression.^4,5,53 High expression levels of these transcription factors has been detected in many invasive carcinomas.^54–58 Increased expression of Snail1 and Snail2 has been described in breast cancer, ovarian cancer, and diffuse type gastric cancer.^49,54,59,60 Up-regulation of TWIST has been found in breast, gastric, hepatocellular, prostate and bladder cancers.^61–65 E-cadherin repression by transcription factors ZEB1 and ZEB2 has been reported in biopsies of several human cancer types (e.g. intestinal type gastric cancer, non-small cell lung cancer, and oral squamous cell carcinomas).^54,66,67

Transcription factors SNAIL, SLUG, TWIST1, ZEB1, and ZEB2 activate the epithelial–mesenchymal transition program (EMT) by down-regulating expression of the genes associated with the epithelial phenotype (e.g. occludin, cytokeratins, and polarity genes, in addition to E-cadherin) and inducing the expression of the genes that sustain the mesenchymal phenotype (N-cadherin, vimentin, fibronectin, β1 and β3 integrins) and matrix metalloproteinases (MMPs),^4,68 and hence, they are commonly termed “EMT transcription factors” (EMT-TFs). EMT plays a crucial role in conferring metastatic properties onto cancer cells, which allow the cells to invade and migrate through the bloodstream and disseminate throughout the body.^5,69 Experiments in animal models have shown that suppression of EMT-TFs significantly decreased metastasis of cells from the primary tumors and, conversely, activation of these transcription factors led to increase in cancer cell metastasis.⁵³ Loss of E-cadherin and gain of mesenchymal markers by cancer cells have been detected at the invasive front of epithelial tumors.⁷⁰ Expression of EMT-TFs can be induced by various signals from the tumor microenvironment – TGFβ (Transforming Growth Factor Beta), EGF (Epidermal Growth Factor), HGF (Hepatocyte Growth Factor), Notch, FGF (Fibroblast Growth Factor), IGF (Insulin-Like Growth Factor), IL-6 (Interleukin-6), TNF (Tumor Necrosis Factor) and hypoxia and increased matrix stiffness, all of which activate oncogenic signaling.^4,71–74 Intravital microscopy applied to a mammary carcinoma mouse model containing an EMT-dependent color shift detected a population of tumor cells undergoing EMT at the tumor boundaries next to the blood vessel-enriched stroma.⁷⁵ Cells from the tumor microenvironment can secret growth factors inducing EMT in tumor cells. For example, it has been found that HGF and TGFβ secretion by cancer-associated fibroblasts (CAFs) could activate migration of carcinoma cells.^76,77 The function of EMT-TFs in cancer progression is not limited by induction of EMT and subsequent metastatic dissemination. It has been shown that various EMT-TFs contribute to DNA damage repair, resistance to radio-, chemo- and immunotherapy, enhanced cell survival and suppression of apoptosis.^4,53,78–81 In mammary and thyroid carcinoma, respectively, the EMT-TF SLUG is involved in acquisition and maintenance of stemness.^82,83 Overall, up-regulation of EMT-TFs promotes EMT, cancer progression and metastasis, which correlate with high cancer aggressiveness and poor survival rate.⁵³

In carcinoma cells, E-cadherin expression may also be controlled by miRNAs. miR-9 has been shown to be involved in direct suppression of E-cadherin expression in breast cancer cells.⁸⁴ Various miRNAs negatively regulate the expression of EMT-TFs, some of them orchestrating negative feedback loops.⁸⁵ The ZEB proteins have been found to be regulated by the miR-200 family;^86–88 SNAIL1 expression is controlled by the miR-34 family.⁸⁹ SNAIL1 and ZEB1/2 repress the miR-34 and miR-200 promoters, respectively, providing a double-negative feedback loop between ZEB/miR-200 and SNAIL1/miR-34.^89–91 These regulatory loops support epithelial plasticity of cancer cells. The epithelial phenotype corresponds to high levels of miR-200 and miR-34, whereas the mesenchymal phenotype corresponds to high levels of ZEB and SNAIL.^92–94

It has been shown that in lung cancers, the absence of the E-cadherin protein on the cell membrane may be associated with the loss of expression of p120 catenin which inhibits E-cadherin endocytosis.⁹⁵ Recent studies using the mouse model of pancreatic ductal adenocarcinoma demonstrated that some carcinoma cells undergoing EMT could exhibit low expression of EMT-TFs but lose the epithelial phenotype through E-cadherin internalization rather than its transcriptional repression.⁹⁶ It has also been shown that on the cell membrane, E-cadherin may be associated with various receptor tyrosine kinases (EGFR, ErbB2, Met, and IGF1R), which are often over-expressed or constantly active in various tumors.⁹⁷ Active receptor tyrosine kinases or Src family kinases phosphorylate E-cadherin, which targets it for lysosomal degradation via binding of the E3 ubiquitin ligase Hakai.^98,99 Deregulation of signaling pathways in tumors may also contribute to down-regulation of E-cadherin. For example, aberrant activation of the TGFβ signaling pathway leads to increased expression of EMT-TFs SNAIL1 and SNAIL2.¹⁰⁰

In epithelial cells, AJs and TJs form an integrated apical junctional complex. Many studies have demonstrated that dysregulated expression of TJ proteins claudins in cancer may be involved in E-cadherin down-regulation.¹⁰¹ The very aggressive claudin-low molecular subtype of breast cancer originating from luminal epithelial cells was characterized by reduced expression of claudins 3, 4, 7, E-cadherin and possession of stem cell features.¹⁰² Expression of claudin-7 is significantly down-regulated in colon, esophagus and non-small cell lung carcinomas. Reduced levels of claudin-7 contribute to EMT. In vitro studies have shown that in esophageal keratinocytes, claudin-7 is involved in the regulation of E-cadherin. In esophageal squamous cell carcinoma cells, knockdown of claudin-7 led to down-regulation of E-cadherin expression and enhanced invasive activity. On the contrary, overexpression of exogenous claudin-7 was associated with high levels of E-cadherin expression and less invasive phenotype of poorly differentiated carcinoma cells.¹⁰³ It was also shown that in human non-small cell lung cancer cells transfected with claudin-7 migratory and invasive abilities were dramatically decreased.¹⁰⁴ Overexpression of claudin-7 in poorly differentiated SW620 colon cancer cells led to up-regulation of E-cadherin expression and inhibited cell growth in soft agar and xenograft tumor growth in mice. Knockdown of claudin-7 in HT-29 or DLD-1 cells led to down-regulation of E-cadherin expression, induced expression of mesenchymal markers, colony formation, matrigel invasion and xenograft-tumor growth.¹⁰⁵ Various claudins are overexpressed in cancer. Overexpression of claudin-1 has been observed in colon, nasopharyngeal, ovarian, hepatocellular and oral squamous cell carcinomas; claudins-3 and 4 are up-regulated in breast, gastric, pancreatic, prostate and uterine cancers. In cancer cells, claudins activate ERK1/2, Src, Wnt and PI3K/Akt signaling, which are associated with down-regulation of E-cadherin expression, EMT, invasion, and development of the metastatic phenotype.^106,107

4. Roles of E-cadherin in cancer cells

The tumor suppressive role of E-cadherin in cancer cells has been extensively studied. On the plasma membranes of cancer cells, E-cadherin homophilic binding can participate in growth inhibitory signaling. E-cadherin recruits β-catenin to AJs sequestering it at the membrane.¹⁰⁸ As is well known, apart from structural maintenance of AJs, β-catenin can be involved in signaling, depending on its subcellular localization. Activation of the Wnt/β-catenin canonical signaling pathway in various tumors, such as breast, pancreatic and colorectal cancers, melanomas, hepatocellular carcinomas, and leukemias, has been documented.¹⁰⁹ Wnt signaling promotes inhibition of destruction complex’s ability to phosphorylate and degrade cytoplasmic β-catenin, which leads to accumulation of β-catenin in cytoplasm, its translocation into the nucleus and subsequent activation of various genes via TCF/LEF1 transcription factors. The Wnt/β-catenin pathway is involved in tumor progression, including maintenance of cancer stem cells (CSCs), promotion of metastasis, cancer cell survival, immune evasion, chemoresistance and tumor angiogenesis.^109–112 It has been found that in mouse colon, E-cadherin was able to suppress the transforming properties of activating mutations of β-catenin.¹¹³ Recently, it has been demonstrated that in colon carcinoma cells growing in spheroids in vitro and in xenografts, disruption of β-catenin/E-cadherin interactions at the cell membrane, facilitated by CDCP1, promoted translocation of these AJ proteins to the nucleus.¹¹⁴

In cancer cells, the extracellular domain of E-cadherin may also interact directly with growth factor receptors (e.g. EGFR, IGF-1 R, and c-Met). This interaction negatively regulates ligand-dependent activation of receptor tyrosine kinases and suppresses proliferation-stimulating signaling from the receptors.^97,115 It has been shown that E-cadherin ligation inhibits EGFR signaling by preventing phosphorylation of Tyr845 in the EGFR molecule by Src family kinases.¹¹⁶ E-cadherin-based AJs associated with NF2/merlin together with TJ-associated proteins play a tumor-suppressive role via the Hippo-YAP/TAZ growth inhibitory pathway.^117–119 During establishment of cell–cell contact, relocalization of the proliferation-activating transcriptional coactivators YAP and TAZ from the nucleus to the cytoplasm mediates contact inhibition of proliferation, i.e. inhibition of normal cell proliferation by cell density. Loss of contact inhibition of proliferation is typical for cancer cells. Dysregulation of the Hippo-YAP/TAZ pathway exerts a significant impact on cancer progression.¹²⁰ Recent studies of Furukawa et al. demonstrated that ZA and associated circumferential actin belt can interfere with nuclear localization of YAP/TAZ. This suppression required NF2/merlin, whose association with E-cadherin was inhibited by tension in the circumferential actin belt. This allowed merlin to be released from AJs and promoted its nucleocytoplasmic shuttling, which, in turn, mediated export of YAP/TAZ from the nucleus.¹²¹ The PLEKHA7 protein can co-operate with E-cadherin to recruit to ZA various mRNAs and miRNAs including miRNA-24 and miRNA-200 c and core components of the RNA-induced silencing complex (RISC) – DROSHA, DGCR8, and Ago2. MiRNA-24 and miRNA-200 c associated with ZA silence mRNAs of the oncogenes MYC and JUN and a pluripotency factor SOX2, thus down-regulating their expression.^122,123

Although the tumor suppressing role of E-cadherin has long been known, accumulating evidence shows that E-cadherin may also play a tumor promoting role. Many invasive and metastatic carcinomas retain E-cadherin expression (e.g. ductal breast, colorectal, prostate carcinomas and oral squamous cell carcinoma).^124–128 Earlier, it has been shown that overexpressed E-cadherin plays an important role in an aggressive form of breast cancer, inflammatory breast cancer, connecting cells into tumor emboli that metastasize within the dermal-lymphatic vessels.¹²⁹ In SUM149 human inflammatory breast cancer cells, introduction of dominant negative mutant E-cadherin cDNA reduced the invasive abilities of cells in Transwell invasion assay.¹³⁰ During the last decade, it has been revealed that in many carcinomas, tumor cells undergo pEMT by acquiring mesenchymal traits but retaining epithelial markers such as E-cadherin. Using a lineage-labeled mouse model of pancreatic ductal adenocarcinoma, it was shown that tumor cells undergoing the pEMT program exhibited epithelial-mesenchymal plasticity and migrated as clusters.⁹⁶ Single-cell transcriptomic analysis of head and neck squamous cell carcinomas revealed that сells undergoing pEMT localized to the outer edges of primary tumors close to CAFs.¹³¹ Cancer cells with a hybrid epithelial/mesenchymal phenotype play the central role in cancer progression: besides their enhanced migratory properties, they possess stemness features and high metastatic potential and exhibit resistance to chemo- and immunotherapy.^4,71 E-cadherin also contributes to the survival of cancer cells.^132,133 In HCT15 colon carcinoma cells, E-cadherin-based AJs together with the cell polarity protein DLG1 sequestered the Fas death receptor, thus preventing proapoptotic signaling.¹³⁴ In mouse and human models of luminal and basal invasive ductal breast carcinomas, it was found that loss of E-cadherin led to TGF-β-mediated accumulation of reactive oxygen species and induction of apoptosis which reduced survival, proliferation, and dissemination of circulating tumor cells (CTCs).¹³³ E-cadherin is involved in nucleotide excision repair by positively regulating the expression of xeroderma pigmentosum complementation group C (XPC) and DNA damage-binding protein 1 (DDB1) during ultraviolet-induced DNA damage.¹³⁵

A soluble extracellular proteolytic fragment of E-cadherin (sE-cad) with a molecular weight of 80 kD promotes cancer cell invasion and metastasis.¹³⁶ sE-cad can also regulate proliferation and survival of cancer cells through binding to the EGFR family receptors and IGF-1R.^137–139 It was also shown that sE-cad secreted from ovarian cancer cells in exosomes promoted tumor angiogenesis.¹⁴⁰ Recently, overexpression of another secretory E-cadherin form (C-E-cad) encoded by a circular E-cadherin RNA (circ-E-Cad) was detected in human glioblastomas. In 90 out of 107 glioblastomas, C-E-Cad was co-expressed with SOX2, a marker for cell stemness. C-E-cad activated EGFR signaling independently of EGF through interaction with the EGFR CR2 domain.¹⁴¹

5. E-cadherin-based AJs in collective migration of cancer cells

In many cases, cancer cells invade the surrounding stroma not individually but as cohesive clusters in which the cells are linked together by E-cadherin-based AJs (Figure 3a). Cells moving as a group often exhibit more persistent migration than single cells.^142,143 Collective invasion of tumor cells into collagen matrix was first observed ex vivo in explants of primary tumors, e.g. oral squamous cell carcinoma and breast ductal carcinoma.¹⁴⁴ In human pancreatic, colorectal, lung and breast carcinomas, histopathological analysis showed that groups of invading cells at the cancer-host interface positively stained for E-cadherin.^145–147 Collective invasion of cancer cells was also detected using intravital microscopy in various experimental models.^148,149

Figure 3.

E-cadherin plays an important role in dissemination of cancer cells. (a) Collective migration of a group of cells connected by E-cadherin AJs (green). (b) A cluster of CTCs held together with E-cadherin AJs is traveling through circulation. (c) Migration of a cancer cell (yellow) over an epithelial monolayer (gray). The cancer cell is using E-cadherin AJs (green) for attachment to normal epithelial cells and translocation of the body. (d) Invasion by a cancer cell (yellow) of an epithelial monolayer (gray). The E-cadherin AJs (green) between a cancer cell and normal cells facilitate the invasion.

Collective migration is driven by growth factors, cytokines and extracellular matrix ligands, and the cells in the group that are the most responsive to these stimuli convert into leader cells. In the group of collectively migrating cancer cells, the chemokine gradient may be self-generated, allowing the group to efficiently migrate, as was shown for melanoma cells self-generating lysophosphatidic acid gradient which drives cell movement.¹⁵⁰ Leader cells undergo pEMT and gain mesenchymal phenotype while remaining attached to the follower cells that can maintain epithelial apical-basal polarity.¹⁵¹ Collectively moving cells are linked together by adhesive structures, mainly E-cadherin-based AJs. AJs support leader cell front-rear polarity and inhibition of protrusion formation in follower cells.^142,143 As was shown earlier, the asymmetric distribution of N- or E-cadherin-based AJs in a cell group is sufficient to promote cell polarization toward free edge.^152,153 Cadherin-based AJs contribute to coordination of collective migration by restricting lamellipodial activity in the regions of cell–cell contacts and triggering Rac-driven membrane protrusions.¹⁴³ It was also demonstrated that cells behind the prospective leaders exert forces on future leader cells facilitating their polarization.¹⁵⁴

Traction forces are generated by the acto-myosin network linked to focal adhesions at the fronts of the leader cells that transmits pulling forces to the followers. AJs provide mechanical coupling of the cells and forces transmitted through AJs by the acto-myosin cytoskeleton, thus coordinating the motile behavior of adjacent cells.¹⁵⁵ Forces generated during initiation of migration may lead to redistribution of the Hippo pathway molecule merlin from cell–cell junctions to the cytoplasm followed by Rac1 activation at the front of the leader cells.¹⁵⁶ Disruption of E-cadherin-based AJs results in uncoordinated movement of cells in the group and lower migration velocity of the group.¹⁵¹

Collective migration of cancer cells is also coordinated by the intermediate filaments. Keratin intermediate filaments may control traction forces and the gradient of Rac activity during collective migration.¹⁵⁷ Increased expression of the mesenchymal marker vimentin is found in many carcinomas. Expression of vimentin is associated with migratory phenotype of cancer cells.¹⁵⁸ Recently, it was found that in collectively migrating cells, vimentin intermediate filaments are required for actin-driven treadmilling of AJs, maintenance of cell–cell contacts and vinculin-mediated traction forces at FAs, thus restricting generation of traction forces to the fronts of the leader cells.¹⁵⁹

Claudin-1 and claudin-11 have been shown to be involved in collective migration of cancer cells. Stable knockdown of K8/K18 keratins by shRNA in HepG2 cells promoted collective migration and invasion through claudin-1.¹⁶⁰ In squamous cell carcinoma, Snail induced expression of claudin-11. Phosphorylated claudin-11 activated Src which suppressed RhoA activity at AJs through p190RhoGAP, maintaining stable cell–cell contacts.¹⁶¹

In the bloodstream of cancer patients CTC clusters have been detected, often held together by E-cadherin-based AJs (Figure 3b). In head and neck cancer patients, formation of CTC clusters containing claudin-11 correlated with poor prognosis.¹⁶¹ The presence of CTC clusters has been associated with high risk of cancer progression and development of distant metastases and has been shown to have a prognostic impact in lung, breast, and head and neck carcinomas.^{132,162–166} It has been shown that CTC clusters are more efficient in forming metastatic outgrowths in distant organs.^133,167 Data obtained from experimental mouse models in line with clinical evidence demonstrated that clusters of CTCs were more resistant to fluid shear stress, anoikis, or lack of growth factors in the circulation than single CTCs.^132,168 Recent studies demonstrated that stable cell aggregates were resistant to anoikis under fluid shear stress conditions in an E-cadherin-dependent manner.¹⁶⁹ СTC clusters traverse capillary-sized vessels by unfolding into single-cell chains. The cells that constitute the chain may undergo transient deformation to fit into the capillary.¹⁷⁰ Interaction of CTCs with platelets provides additional benefits for the CTCs. Cloaking of cancer cells by platelets protects them from eradication by the immune system, specifically by the NK cells.¹⁶⁸ Platelets contacting with cancer cells secreted TGFβ that activated the TGFβ/Smad and NF-κB pathways in cancer cells, resulting in their transition to an invasive mesenchymal-like phenotype, which promoted metastasis in vivo.¹⁷¹ Cloaking of cancer cells by platelets promoted cancer cell extravasation by eliciting ATP secretion from activated platelets and activating endothelial P2Y₂ receptors, which resulted in an increase in endothelial barrier permeability.¹⁷²

6. Reorganization of E-cadherin-based AJs in cancer cells undergoing EMT

Earlier, we observed that morphology and dynamics of E-cadherin-based AJs were different in normal epithelial cells and neoplastically transformed epithelial cells undergoing pEMT. Normal IAR-2 rat liver epithelial cells possessed stable linear AJs associated with the circumferential actin bundle. In a panel of IAR cells transformed with oncogenes or chemical carcinogens, we observed radial (punctate) AJs associated with straight actin bundles and undergoing continuous remodeling.¹⁷³ The active dynamics of these punctate AJs was similar to that of N-cadherin-based AJs of fibroblasts. Myosin II-mediated contractility proved crucial for the formation and maintenance of these AJs: treatment with contractility inhibitors (either the ROCK inhibitor Y-27632 or the myosin II ATPase inhibitor blebbistatin) reduced punctate AJs to nascent dot-like AJs.

Indra et al. analyzed spatial organization of punctate AJs in A-431 carcinoma cells. These studies demonstrated that punctate AJs consisted of dense, paracrystalline nanoclusters formed through cis and trans interactions of cadherin ectodomains, interspersed with less dense cadherin regions.¹⁷⁴ It was also shown that F-actin bundles connected with punctate AJs consisted of two structurally distinct regions that had different dynamic properties. The proxymal region associated with adhesive proteins consisted of branched and highly dynamic F-actin enriched with VASP and the F-actin depolymerization factor cofilin-1. Distant bundle stalk enriched with calponin was more stable and connected the bundle to the cell cytoskeleton. Assembly and disassembly of F-actin and cadherin clusters were closely co-ordinated.¹⁷⁵

Recently, using live cell imaging, we analyzed EMT induced by EGF in IAR-20 rat liver epithelial cells and observed sequential reorganization of AJs and the cytoskeleton during EMT (Figure 4).¹⁷⁶ We detected fragmentation and dissolution of the circumferential actin bundle, a structure crucial for maintenance of stable linear AJs, that was followed by replacement of the stable linear E-cadherin-based AJs by dynamic punctate AJs. Punctate AJs were associated with straight actin bundles and co-localized with a tension-sensitive protein zyxin, which indicated generation of centripetal forces at the cell–cell boundaries during EMT-induced reorganization of actin cytoskeleton. Dissolution of the circumferential actin bundle may result from EGF-induced phosphorylation and degradation of the actin-binding protein EPLIN, which, as was shown earlier, stabilizes the circumferential actin bundle.²⁵ We detected increased phosphorylation of EPLIN within minutes of addition of EGF. Simultaneously with disruption of circumferential actin bundle and linear AJs, we observed formation of dynamic lamellipodia containing branched actin network and appearance of retrograde acto-myosin flow at the cell–cell boundaries. Overall, we demostrated increased structural dynamics at the cell–cell boundaries during early stages of EMT. Of particular importance for the subsequent EMT stages was the appearance of dynamic AJs and actin structures. Cells released from stable cell–cell contacts can acquire front-rear polarity and a migratory phenotype.

Figure 4.

Rearrangement of E-cadherin-based AJs and actin cytoskeleton architecture during EGF-induced EMT. An islet of epithelial cells tightly connected by stable linear AJs (1) is converted into a group of motile cells loosely linked by dynamic radial AJs (2). 1ʹ – linear AJs (a magnification of the circled area in 1) are organized as a continuous adhesion belt along the cell–cell boundaries and contain E-cadherin molecules (black and green) associated with the circumferential actin bundle (yellow). The EPLIN protein (orange) is in non-phosphorylated state. Cell–cell boundaries are in the state of contact paralysis – no lamellipodial activity is visible in those areas. 2ʹ – punctate AJs (a magnification of the circled area in 2) are perpendicular to cell–cell boundaries and are associated with straight actin bundles. EPLIN is phosphorylated and has dissociated from the actin bundles. Instead of a thin “scar” of a stable cell–cell contact, there is an overlap of lamellipodia of the adjacent cells, containing branched actin networks (white).

Cancer cells with punctate AJs were capable of migrating both individually and collectively, as was demonstrated on 2D adhesive substrates and in migration chambers.¹⁷⁷ Our earlier observations showed that neoplastically transformed epithelial cells that retained E-cadherin were capable of forming E-cadherin-based AJs with the underlying monolayer of normal epithelial cells. Dynamic punctate E-cadherin-based AJs of the transformed cells allowed them to migrate over normal cells and to invade the epithelial monolayer (Figure 3c,d).¹⁷⁸ Transfection of either a dominant negative E-cadherin construct or an anti-E-cadherin siRNA abolished formation of E-cadherin-based AJs by transformed cells, which drastically decreased the efficiency of their migration over or invasion of the epithelial monolayer. In contrast, knock-down of N-cadherin by siRNA did not affect migratory and invasive behavior of the transformed cells. Thus, formation of E-cadherin-based cell–cell contacts between cancer cells and the surrounding normal cells may be an important step for cancer cell dissemination. It has been shown in vitro that cancer cells can form heterophilic AJs involving E-cadherin on the cancer cell membrane and N-cadherin on the membrane of CAFs. CAFs remodel ECM and create migration tracks facilitating collective cancer cell migration and invasion.^179,180

While initiation of the invasion-metastasis cascade requires a more mesenchymal phenotype, successful metastatic colonization depends on the reversion to the epithelial phenotype (mesenchymal-epithelial transition, MET) and re-expression of E-cadherin by cancer cells. Higher levels of E-cadherin expression in metastases than in primary tumors were detected by immunohistochemistry in invasive breast carcinomas.¹⁸¹ Direct intravital microscopy in mice showed that migratory carcinoma cells that had undergone spontaneous EMT re-acquired the epithelial morphology in distant metastases.^182,183 In various experimental models, successful metastatic outgrowth required repression of the EMT-TFs TWIST1 and Paired Related Homeobox 1 (PRRX1) whose activity was essential for initiation of metastasis.^184–186 In a mouse breast carcinoma model, it was demonstrated that formation of metastases in distant organs depended on local activation of metastatic niches that actively supported a more epithelial phenotype, including up-regulation of E-cadherin expression.¹⁸⁷ Together, these data delineate a key role that E-cadherin-based AJs play in cancer dissemination.

7. N- and P-cadherin in cancer

“Cadherin switching”, a process whereby N-, P-, or other cadherins replace or are co-expressed with E-cadherin, has been proved important for acquisition of migratory and invasive behavior by cancer cells.^188–191 Aberrant expression of N-cadherin is described in many carcinomas such as breast, prostate, bladder, non-small cell lung carcinomas or hepatocellular carcinomas where N-cadherin may be a negative prognostic marker.^192–197 N-cadherin replaces E-cadherin during EMT. It was shown that N-cadherin provided weaker cell–cell adhesion compared to E-cadherin.¹⁹⁸ N-cadherin contributes to migratory activity of cancer cells: several studies demonstrated that its upregulation was associated with increased motility and invasiveness of cancer cells in vitro. N-cadherin increased migration and invasiveness of breast cancer cells regardless of E-cadherin expression.^199,200 Human pancreatic cancer cells respond to collagen I by increasing motility and up-regulation of N-cadherin expression.²⁰¹ N-cadherin’s extracellular domain interacts with the Fibroblast Growth Factor receptor (FGFR), which prevents FGFR internalization and leads to activation of ERK, expression of MMP-9 and induction of cell migration, invasion and metastasis.^202,203 Besides induction of cell migration, N-cadherin promotes cancer cell survival by up-regulating the anti-apoptotic protein Bcl-2, as was shown in the DU-145 human prostate carcinoma cell line.²⁰⁴ In mouse models of mammary carcinomas, N-cadherin was associated with tumor aggressiveness and increased metastatic potential and contributed to tumor progression, for example, N-cadherin enhanced pulmonary metastasis.^203,205

P-cadherin (placental cadherin) is also involved in cancer development. In an adult organism, P-cadherin colocalizing with E-cadherin is found in basal myoepithelium of mammary gland where it maintains normal breast epithelial architecture.²⁰⁶ P-cadherin is also found in the basal layers of epidermis, prostate, cervix, hair follicle, and mesothelium, colocalizing with E-cadherin in AJs.²⁰⁷

Up-regulation of P-cadherin expression has been described in many types of cancers such as breast,²⁰⁸ prostate,²⁰⁹ bladder,²¹⁰ gastric,²¹¹ ovarian,²¹² colon, and pancreatic cancer,^213–215 squamous cell carcinomas,²¹⁶ and endometrial carcinomas.^217,218 High expression of P-cadherin is associated with high-grade invasive cancer and poor prognosis for patients with breast, ovarian, pancreatic cancer, non-small cell lung cancer, and cholangiocarcinoma.^219–222 P-cadherin expression is correlated with tumor aggressiveness associated with cell invasiveness.²⁰⁷ P-cadherin plays an important role in pancreatic carcinoma progression and is strongly overexpressed in great majority of pancreatic ductal adenocarcinomas. Induction of P-cadherin expression in pancreatic carcinoma cells increased their motility.²¹³ Like E-cadherin, P-cadherin can support collective invasion of cancer cells by increasing cell polarization and traction forces.²²³ Recently, it has been shown that P-cadherin induces anoikis resistance of breast cancer cells by promoting pentose phosphate pathway and decreasing oxidative stress, thus promoting the survival of breast cancer cells in circulation and during metastatic outgrowth.²²⁴

8. Conclusion

E-cadherin, long known for its tumor suppressing role, has been revealed to be a “double agent”, equally adept at tumor promotion. Many carcinomas retain E-cadherin expression during their progression, owing to its assisting tumor cell survival and enhancing their invasive and metastatic potential. Replacement of E-cadherin-based stable linear AJs with dynamic punctate AJs and formation of cadherin AJs with adjacent normal cells enhance invasive and metastatic properties of cancer cells. Collective invasion and dissemination by CTC clusters, supported by E-cadherin, are among the crucial negative prognostic factors. N- and P- cadherins expressed de novo in various carcinomas via “cadherin switching” are also associated with an increase in invasion and metastasis. Cancer cells are able to harness the broad scope of AJ protein functions, not at all limited by promoting stable cell–cell adhesion, and use it to their advantage. Further research is required, however, to elucidate the precise molecular mechanisms of the tumor promoting function of cadherins.

Funding Statement

This work was supported by the Russian Foundation for Basic Research (RFBR) under Grant no. 18-54-16005.

Disclosure statement

No potential conflict of interest was reported by the author(s).