Protein kinase C inhibitor Gö6976 but not Gö6983 induces the reversion of E- to N-cadherin switch and metastatic phenotype in melanoma: identification of the role of protein kinase D1

In this study, we tested whether PKC inhibitors (such as Gö6976 and Gö6983) would revert the E- to N-cadherin switch, a major hallmark of the EMT process, required for melanoma progression and metastasis [9, 15, 35].

First, we analyzed the cadherin switch in two couples of patient-derived primary (I5, T1) and their respective lymph-node metastasis (M2, G1) melanoma cells. Different profiles of E- and N-cadherin expression were observed. When comparing T1 (primary tumor) and G1 (corresponding lymph node metastasis) melanoma cells, both cell lines expressed high levels of E-cadherin and very low levels of N-cadherin. Although these findings are inconsistent with the theory of E- to N-cadherin switch during melanoma progression (i.e. E-cadherin expression in primary tumor and N-cadherin expression in metastasis), the expression of E-cadherin by the metastasis-derived cells (G1) might be the consequence of a MET (mesenchymal-epithelial transition) that would have occurred at the metastatic site. MET was postulated to be part of the process of metastatic tumor formation [36, 37]. In fact, progression of solid tumors involves spatial and temporal occurrences of EMT, whereby tumor cells acquire a more invasive and metastatic phenotype [38]. Subsequently, the disseminated mesenchymal tumor cells undergo the reverse transition, MET, at the site of metastases, as metastases recapitulate the pathology of their corresponding primary tumors [39]. Just as a critical EMT event is the downregulation or silencing of E-cadherin and upregulation of N-cadherin, the re-expression of E-cadherin and loss of N-cadherin is proposed to be the important hallmark of MET [39]. Presently, it is unclear how MET facilitates the formation of metastases. Some findings suggest a possible mechanism whereby MET helps the tumor cells to construct connections with the resident non-neoplastic epithelial cells and may enable dormancy and survival of tumor cells at a lower metabolic load at target organ [39]. Consistently, G1 cells, that may be derived from metastatic cells that have underwent MET, grow very slowly, are non-tumorigenic and do not migrate. On the other hand, E-cadherin expression was absent in both I5 (primary tumor) and M2 (corresponding lymph node metastasis) melanoma cells while N-cadherin expression was significantly higher in the metastasis cells, suggesting that I5 primary tumor cells may have already underwent EMT. However, these cells did not show higher proliferation and migration potentials than their counterparts T1 cells but progressed to highly aggressive metastatic cells (i.e. M2 cells). This observation suggests that depending on the cellular context, the pro-metastatic effect of cadherin switch might require additional events that could occur later in tumorigenesis [16]. In agreement with this hypothesis, a study by Chen et al. 2014 shows that E-cadherin loss in non-malignant breast cells is insufficient to induce EMT or to enhance their transforming potential [40]. Furthermore, other studies by Knudsen et al. 2005 show that N-cadherin expression in the mammary gland of transgenic mice (mice expressing N-cadherin under the control of mouse mammary tumor virus promoter) does not induce tumors. Interestingly, even crossing these mice with mice expressing the Neu oncogene in the mammary gland does not produce tumors more aggressive than those in mice expressing Neu alone [41]. However, crossing them with mice expressing polyoma virus middle T antigen in the mammary epithelium leads to increased metastasis [42]. These studies suggest that the effects of cadherin switching are late events in tumorigenesis and demonstrate that the influence of an inappropriate cadherin on the phenotype of the cell is context dependent [16]. This could be the same in melanoma cells. Thus, despite the E- to N-cadherin switch in I5 primary melanoma cells, these cells require additional mutations or events to acquire enhanced mesenchymal features allowing their invasion and metastasis formation. The primary (I5) and corresponding metastasis (M2) melanoma cells could be a good model to study the events that are required in addition to the E- to N-cadherin switch to allow melanoma progression to metastatic phenotype. Interestingly, except for I5, cell lines that express high N-cadherin and low E-cadherin levels had a better capacity of proliferation, clonogenicity and migration than cell lines with low N-cadherin and high E-cadherin expression. Thus, consistent with the cadherin switch theory in melanoma progression, our data show a strong link between high N-cadherin/low E-cadherin expression and tumor aggressiveness and defines M2 metastatic melanoma cells (Absent E-cadherin/high N-cadherin expression and highly tumorigenic and motile) as a pertinent model to study the possibility to revert the E- to N-cadherin switch and the metastasis phenotype. In addition to this lymph-node metastasis melanoma model, another cutaneous metastasis melanoma model with low E-cadherin/high N-cadherin expression (i.e. M4T2 cells) was used to further confirm our results.

Treatment of metastatic melanoma cells with PKC inhibitors Gö6976 or Gö6983 resulted in different responses regarding cadherin switch and oncogenic activity of these cells. Gö6976 but not Gö6983 treatment induced the expression of E-cadherin and inhibited the expression of N-cadherin in metastatic melanoma cells. This Gö6976-induced N- to E-cadherin switch was associated with rapid phenotypic and molecular changes that comply with the generally described model of cadherin switch during tumor progression [6, 14‐16]. In fact, our results showed that the Gö6976-induced E-cadherin expression and N-cadherin loss were associated with a rapid cell shape modification from an elongated mesenchymal-like structure into a ‘cuboidal’ epithelial-like shape and a strong increase in cell-cell junctions [43‐45]. Furthermore, treatment of lymph-node metastatic melanoma cells with Gö6976 induced the translocation of β-catenin from the nucleus to the plasma membrane. However, the membrane β-catenin staining in Gö6976-treated cells is not perfectly lining the plasma membrane. In fact, it has been shown that newly synthesized E-cadherin arrives at the plasma membrane already in complex with β-catenin [46‐48]. Then, once at the plasma membrane, this β-catenin pool is transferred to E-cadherins that were already present at the plasma membrane. Furthermore, the β-catenin released from E-cadherin may participate in new exchange cycles [47]. Thus, since during the times of Gö6976-treatment used in our present study (0–48 h), E-cadherin expression is continuing to increase (Fig. 2), so new E-cadherin molecules are still being synthesized and transported to the plasma membrane, the diffuse β-catenin staining may be explained by these β-catenin-E-cadherin exchange cycles [47]. In agreement with this hypothesis, in Gö6976-treated cells, β-catenin staining, which completely disappeared from the nucleus, is mostly accumulated at the newly formed intercellular junctions while less β-catenin staining is observed at the free borders of the plasma membrane or in the cytoplasm, consistent with E-cadherin staining (Fig. 3c). This translocation of the β-catenin from the nucleus to the plasma membrane favors its tumor suppressor role of insuring the stability of E-cadherin-mediated intercellular interactions to its oncogenic role of transcription factor [32, 33], as further supported by the inhibition of the expression of its target cyclin D1 after Gö6976 treatment. These observations also suggest that the Gö6976-induced E-cadherin molecules are functional. In addition to these molecular changes, Gö6976 treatment resulted in decreased anchorage-independent growth and motility of metastatic melanoma cells, confirming the reversion of the metastatic phenotype. In contrast to the metastatic melanoma cells (M2 and M4T2 cell lines), Gö6976 inhibitor did not significantly affect the expression of E- or N-cadherins or the anchorage-independent growth in the primary melanoma cell line (I5) (Additional file 2). Thus, the effects of this inhibitor seem to be specific to advanced metastatic melanoma suggesting that the reversion of the mesenchymal features by Gö6976 requires late event(s) in the melanoma progression. Since a strong increase in the expression of the Gö6976-target « PKD1 » is observed in M2 metastatic cells in comparison to their corresponding primary tumor cells (I5), overexpression of this serine/threonine kinase could be one of these events.

In contrast to Gö6976, the PKC inhibitor Gö6983 did not affect any of the E-/N-cadherin expression, intercellular interactions, β-catenin subcellular localization or anchorage-independent growth in metastatic melanoma cells. Although Gö6983 could significantly inhibit the chemotactic (FBS attraction in transwell) migration of these cells but not their horizontal (wound healing) migration, its effect was much lower than the effect of Gö6976. These results suggest that suppression of specific PKC isoforms inhibited by Gö6976 but not by Gö6983 would be involved in the reversion of the E- to N-cadherin switch and metastatic phenotype in melanoma. Very few studies have attempted to examine the involvement of PKCs in EMT, E/N-cadherin expression regulation and/or cell-cell junctions. To our knowledge, these studies were based on the treatment of the cells with either PKC activator, phorbol ester [28, 49], or PKC inhibitor Gö6976 [29, 50] but not on direct analysis of the role of specific PKC isoforms. In fact, phorbol ester has been shown to induce N-cadherin expression in osteoblasts [28] and EMT in prostate cancer [49]. On the other hand, Gö6976 has been shown to promote formation of cell junctions and inhibit invasion of urinary bladder carcinoma cells [29], suppress S252W FGFR-2 mutation-induced N-cadherin expression in human osteoblasts [50] and inhibit EMT in renal tubular epithelial cells [51]. These studies support our present data and indicate that phorbol ester and Gö6976 targets, i.e. PKCs, can be involved in the E- to N-cadherin switch; however, no specific isoforms could be determined as key players in this process. A recently published study by Jain and Basu 2014 shows, by overexpression or siRNA techniques, that PKCε promotes EMT in breast cancer [52]. However, this isoform can’t be involved in the reversion of the E- to N-cadherin switch observed in our present study because neither Gö6976 nor Gö6983 is a potent inhibitor of PKCε, which implies that, in melanoma cells, other PKC isoform(s) regulate the cadherin switch.

Given the differential effect of the PKC inhibitors Gö6976 and Gö6983, our present data suggest that specific isoenzyme(s) targeted by Gö6976 but not by Gö6983 would be involved in the reversion of the E- to N-cadherin switch and metastatic phenotype in melanoma.

In vitro kinase assays using recombinant full-length PKC isoenzymes α, β, γ, δ, ε, ζ and μ (i.e. PKD1) show that Gö6976 selectively inhibits PKCα (IC₅₀ = 2.3 nM), PKCβ (IC₅₀ = 6.2 nM) and PKCμ (IC₅₀ = 20 nM). However, it does not affect the kinase activity of the PKCγ, δ, ε, and ζ-isozymes even at μM levels [27]. On the other hand, Gö6983 efficiently inhibits several PKC isozymes (IC₅₀ = 7 nM for PKCα and PKCβ; 6 nM for PKCγ, 10 nM for PKCδ and 60 nM for PKCζ) but is extremely inefficient in suppressing PKCμ kinase activity in vitro (IC₅₀ = 20 μM) [26]. Despite the highly conserved catalytic region of the PKC isoforms, these kinases are differentially targeted by Gö6976 and Gö6983 ATP-competitive inhibitors. That can be due to variability in their regulatory region that might affect the binding of the inhibitor to the kinase domain [53]. Furthermore, although the chemical structures of Gö6976 and Gö6983 are very close, the central aromatic ring is opened in Gö6983 but intact in Gö6976, which accounts for the different substrate specificities of these inhibitors [26]. In contrast to the other PKC isoforms that are not inhibited by Gö6976, conventional PKCs (PKCα and PKCβ) and PKCμ share tandem repetition of C1 motifs (forming the DAG/PMA-binding domain) that is located at a comparable distance towards the catalytic domain [23], which could be important for the binding of Gö6976. On the other hand, PKCμ significantly differs in some structural features from the other PKC isoforms including PKCα and PKCβ [23]. In fact, PKCμ is larger than the other PKCs with a sequence of 912 amino acids. It contains a hydrophobic domain in the N-terminal region and a pleckstrin homology domain that are absent in the other PKC isoforms. This additional sequence may be responsible for the failure of Gö6983 to inhibit PKCμ probably by preventing its binding to the ATP-binding site. Furthermore, the specificity of action of these inhibitors can be independent of the regulatory domain of the PKCs but differences in small sequences or single amino acids in their catalytic domains can affect the affinity of the inhibitors to the ATP-binding site. However, structural/molecular biology studies are required determine the exact reason for the differences in pharmacological action of these inhibitors.

Whatever the mechanism, since Gö6976 is a selective inhibitor of PKCα, PKCβ and PKCμ and Gö6983 a selective inhibitor of PKCα and PKCβ but not PKCμ [26, 27], we hypothesized that the specific inhibition of PKCμ, which is also known as PKD1, would be responsible for the Gö6976-induced N- to E-cadherin switch and tumor reversion in metastatic melanoma cells. Although this serine/threonine kinase is a major cellular target for the tumor-promoting phorbol esters and growth factors which rapidly induce its activity [25, 54], PKD1 has a complex relationship with respect to cancer development. In fact, depending on the tissue type, PKD1 exerts different functions and effects [55]. To date, the role of PKD1 in melanoma initiation and progression is not yet explored, except for the unique data by Kempkes et al. 2012 showing that PKD1 knockdown inhibits proliferation and migration-related processes such as filopodia formation and αvβ3 integrin recycling in WM9 melanoma cell line [56]. These data support our hypothesis. To study more specifically and deeply the involvement of PKD1 in melanoma progression, we firstly analyzed the expression pattern of this kinase in different melanoma cell lines. PKD1 was expressed in 4 out of the 5 melanoma cell lines tested and this expression strongly correlated with E-cadherin negative/N-cadherin positive phenotype and metastatic potential (anchorage-independent growth and migration). PKD1 knockdown in M4T2 metastatic melanoma cells significantly induced down-regulation of N-cadherin and up-regulation of E-cadherin (Additional file 4), supporting the role of PKD1 in E-cadherin to N-cadherin switch. Several studies suggest a role of PKD1 in gene expression regulation either by direct interaction and phosphorylation of transcription factors [57, 58] or by phosphorylation and nuclear exclusion of histone deacetylases (HDACs) [59]. Some of the transcription factors that can be regulated by PKD1 are known to modulate the expression of E-cadherin or N-cadherin even if no direct correlation between PKD1 and these cadherins has been established. In fact, a study by Eiseler et al. 2012 has shown that PKD1 efficiently interacts in nuclei with Snail1, the main transcription factor that suppresses E-cadherin expression during EMT in most epithelial cancers including melanoma [60, 61]. PKD1 phosphorylates Snail1 at Ser-11 which recruits HDAC-1 and −2 as well as LOXL3, a transcriptional co-activator which is also upregulated by PKD1. This newly formed complex enhances Snail activity and induces proliferation and anchorage-independent growth [57]. Although, in this study, authors did not test whether this PKD1-mediated regulation of snail1 activity might affect E-cadherin/N-cadherin expression, another study by Peinado et al. 2004 shows that snail mediates E-cadherin repression by the recruitment of HDAC1/2. Thus, it is very likely that PKD1 might inhibit E-cadherin expression using this mechanism. On the other hand, PKD1 can induce the activation of NFκB [62], a transcription factor that can directly bind to N-cadherin promoter and activate its expression [63]. Interestingly, N-cadherin can also be directly regulated by E-cadherin. In fact, loss of E-cadherin induces NFκB activity and consequent N-cadherin expression in melanoma cells. Thus, regulation of E-cadherin expression by PKD1 could be enough to induce E- to N-cadherin switch. Although a bunch of evidence suggest a role of PKD1 in the regulation of transcription factors that are regulators of E- or N-cadherin expression, these hypotheses need to be tested to determine the exact mechanism by which PKD1 regulates E- to N-cadherin switch.

In the second metastatic melanoma cell line tested (i.e. M2), although the stable knockdown of PKD1 by specific shRNAs induced the expression of E-cadherin, these shPKD1-induced E-cadherin molecules could only be detected as cleaved form in the colony spheres of cells cultured in semi-solid medium or as secreted molecules in the conditioned medium of the adherent cells. Furthermore, the shPKD1-induced E-cadherin molecules did not allow the formation of cell-cell junctions in M2 melanoma cells; on the contrary, the cells transduced by shPKD1 maintained their elongated mesenchymal and isolated shape. Furthermore, although PKD1 knockdown induced a slight decrease in the horizontal migration (wound healing) potential of metastatic melanoma cells, the inhibition was very low in comparison of the effect of the Gö6976 inhibitor. These data suggest that the E-cadherin molecules induced by PKD1 knockdown are not functional, which is not surprising since these molecules were observed in a truncated form. This observation could be explained by the strong expression, by M2 melanoma cells, of metalloproteinases (MMPs) such as MMP2, MMP9, MMP12 and MMP13 [64‐67] that may cleave the newly synthesized E-cadherin molecules [68]. Noteworthy, PKCs, in particular PKCα, have been shown to induce MMP expression and/or activity [69‐71]. Thus, the restoration of full length and functional E-cadherin molecules by Gö6976 could be the consequence of the dual inhibition of PKD1 (which induces E-cadherin expression) and PKCα (which may inhibit expression and/or activity of MMPs, thus preventing the cleavage of the newly synthesized E-cadherin). The mechanism by which E-cadherin cleavage is mediated in M2 cells might be absent or negatively regulated in M4T2 cells where full-length E-cadherin could be detected after PKD1 knockdown (Additional file 4). Thus, protease profiling and PKC isoform combination silencing experiments would allow the verification of these hypotheses.

Finally, despite its failure to induce the expression of functional E-cadherin, PKD1 knockdown alone strongly altered anchorage-independent growth and migration of metastatic melanoma cells, probably by regulating signaling pathways such as extracellular signal-regulated kinase (ERK), c-jun N-terminal kinase (JNK) and nuclear factor-kappa B (NFκB), as previously described in hormone-positive breast cancer [72, 73] and pancreatic cancer [74, 75].