|Home | About | Journals | Submit | Contact Us | Français|
The regulation of CD44v6, a variant of the CD44 family of glycosylated adhesions molecules, through HGF (hepatocyte growth factor) has implications for motility in primary human melanocytes. We demonstrate that exposure of primary human melanocytes to HGF results in an increase of CD44v6 expression. Immunostaining of melanocytic lesions revealed low, cytoplasmic positivity of CD44v6 in some nevi, but high membranous expression in primary cutaneous melanomas, cutaneous - and lymph node metastases. HGF dependent CD44v6 regulation in melanocytes is NF-κB depended since BAY 11-7082, a NF-κB inhibitor, but not interference with the MAP kinase or phosphatidylinositol 3′-kinase cascade antagonized HGF-induced CD44v6 expression. NF-κB mediated regulation of CD44v6 is related to the activation of transcription factors Egr-1 and CCAAT enhancer binding protein-ß (C/EBP-ß). In gel shift assays the initial binding of a complex encompassing the p100/p52 subunit of NF-κB, C/EBP-ß and Egr-1 to the CD44 promoter experienced reshuffling towards increased affinity of C/EBP-ß after HGF stimulation. A blocking antibody to CD44v6 decreased HGF induced c-Met phosphorylation as well as enhanced random- and site-directed migration. These data demonstrate that HGF contributes substantially to motility in primary human melanocytes, which is dependent on the up-regulation of CD44v6 through a complex of NF-κB/C/EBP-ß and Egr-1.
Hepatocyte growth factor (HGF) is known for its mitogenic, motogenic, and morphogenic effects and signals through membrane-associated c-Met, a tyrosine kinase receptor (Bottaro et al., 1991). Upon stimulation with HGF, c-Met, which is expressed by epithelial cells and melanocytic cells (Bottaro et al., 1991; Sonnenberg et al., 1993; Li et al., 2001) is autophosphorylated at three specific tyrosine residues (Tyr1230, Tyr1234 and Tyr1235). Two tyrosine residues (Tyr1349 and Tyr1356) are required for all biological activities of the receptor and serve as docking sites for multiple Src homology 2 and cytoplasmic phosphotyrosine binding domain-containing proteins, like Src, Shc, phosphatidylinositol 3′-kinase (PI3K) and Grb2 (Ponzetto et al, 1994; Royal and Park, 1995). Activation of c-Met results in cell spreading (Birchmeier and Gherardi, 1998), loss of cadherin-mediated cell-cell adhesion through destabilizing the complex, comprising E-cadherin, p120, β-catenin and α-catenin (Roura et al., 1999) and promotion of tumor metastasis (Rong et al., 1994). E-cadherin mediated adhesion between melanocytes and keratinocytes is down-regulated through HGF (Li et al, 2001). In melanoma, gains in regions of chromosome 7 including the c-Met locus are described (Bastian et al., 1998) and metallothionein-driven expression of HGF induced the appearance of cutaneous melanoma with metastatic potential in transgenic mice (Noonan et al., 2003). In these mice ultraviolet radiation given in a single erythemal dose to neonatal mice lead to melanoma development with high incidence after a relatively short latent period (Noonan et al., 2003).
CD44 molecules are glycosylated, multifunctional, class I transmembrane proteins associated with diverse biological functions, like lymphocyte homing, tumor progression and metastasis. Through alternative splicing, different isoforms are generated out of variant exons v1 to v10. The expression of certain isoforms, especially v4-7 and v6,7, has been attributed to confer metastatic potential to nonmetastatic cell lines (Guenthert et al., 1991; Bourguignon et al., 1995; Turley et al., 2002). It has been recognized that complex formation of CD44, especially v6, with c-Met is a prerequisite for the activation of c-Met. Deletion of the cytoplasmic tail of CD44v6 abrogates c-Met mediated HGF signaling (Orian-Rousseau et al., 2002). The importance of the CD44v6-c-Met complex has been further strengthened by demonstrating that HGF up-regulates CD44v6 expression in mouse melanoma cells through expression of the zinc finger transcription factor egr-1, which in turn was responsible for transcriptional activation of CD44v6 (Recio and Merlino, 2003).
There are only few reports on the expression of CD44v6 in human melanoma. In primary sinonasal melanomas (SM) strong membranous expression in in situ melanomas and lack of membranous CD44v6 expression in invasive SM has been observed. Loss of cytoplasmic expression of CD44v6 was found to be associated with advanced tumor stage (Regauer et al., 1999). Absence for CD44v6 positivity in melanocytic lesions were reported in several studies, we found variable positive stainings in primary melanomas (Manten-Horst et al., 1995; Fernandez-Figueras et al., 1996; Harwood et al., 1996; Schaider et al., 1998).
Based on the HGF-inducing effect of UV (Mildner et al, 2007) in keratinocytes, the endogenous HGF expression by melanoma cells (Li et al., 2001) and the increase of CD44v6 in mouse melanoma cells after HGF stimulation (Recio and Merlino, 2003), we questioned, (1) whether exposure to HGF increases CD44v6 expression in human melanocytes and human melanoma cells and if increased expression leads to functional consequences thereof, (2) whether the expression correlates with progression in melanocytic lesions, and (3) which intracellular signaling pathways and transcription factors are potentially involved in CD44v6 up-regulation in human melanocytes and melanoma cells.
HGF induces up-regulation of CD44v6 in murine melanoma cells (Recio et al., 2003). Since HGF expression is increased in human skin after UV-irradiation (Mildner et al., 2007), we investigated whether HGF changes levels of CD44v6 in primary human melanocytic cells. Basal protein levels of CD44v6 in melanocytic cells revealed a striking low expression in melanocytes, but higher levels in melanoma cells (Figure 1A), an observation, which was partially mirrored for the expression of c-Met (Figure 1A). The low expression of CD44v6 in melanocytes as opposed to the higher expression in melanoma cells, especially in those cell lines, derived from superficial primary melanomas, like WM35 and WM793, was also confirmed by real-time PCR (Figure 1B). Transcripts for c-Met did not differ substantially, except for SBcl2 and WM9 (data not shown). Next, we determined a potential increase of CD44v6 in human melanocytic cells after exposure to recombinant human HGF (rhHGF). rhHGF increased weak but significant CD44v6 transcripts in one melanoma cell line (WM35: 1.5 fold) and in melanocytes (FOM 101: 2-fold) one hour after exposure (Figure 2A). c-Met transcripts remained unaltered in melanocytes and melanoma cells (Figure 2A). Immunoblotting clearly confirmed, that exposure of melanocytes (FOM101, FOM 193), but not melanoma cells (WM793, WM35, WM9), to rhHGF led to up-regulation of CD44v6 (Figure 2B). These results were corroborated by a continuous exposure to HGF after transduction with an adenoviral vector encoding the cDNA of HGF (Ad.CMV.rhHGF) at 20pfu/cell (Li et al., 2001) for 48 hours (Figure S1A). A profound increase in CD44v6 could also be envisaged in melanocytes after transduction with Ad.CMV.rhHGF but not with an adenovirus encoding the cDNA of endothelin cell-derived lipase (EDL-Ad) by immunofluorescence (Figure 2C). FACS analyses showed that melanocytes not only express CD44v6 but that membranous CD44v6 is increased after transduction with Ad.CMV.rhHGF (Figure 2D). Since melanoma cells endogenously produce HGF at different levels (Li et al., 2001), we determined the effect of a neutralizing anti-HGF antibody to basal CD44v6 expression. A decrease in CD44v6 expression was observed in melanoma cells (Figure S1B). These results indicate that HGF, which is induced by pro-inflammatory cytokines in the human skin (Mildner et al., 2007), leads to up-regulation of CD44v6 in human melanocytes. Melanoma cells, which already show increased expression of CD44v6 and produce HGF by themselves, did not respond to recombinant or adenoviral induced production of HGF due to autocrine stimulation (Figure 2B and Figure S1A).
The low basal expression of CD44v6 in melanocytes prompted us to examine the CD44v6 expression in melanocytic lesions by immunohistochemistry in more detail. Out of 12 nevi, only three congenital nevi showed a weak cytoplasmic expression in nevus nests near the basal membrane (Figure 3A). Nevus cells in the dermis were negative for CD44v6. In primary cutaneous melanomas, however, a moderate cytoplasmic and membranous staining was found in most tissue samples. Significant intratumoral heterogeneity was observed in melanoma nests with absent or positive staining (Figure 3B). Only two out of ten primary melanomas showed no staining at all. Membranous staining pattern was even more prominent in cutaneous and sub-cutaneous metastases, with an increase in intensity in nests of melanoma cells (data not shown). A further increase in membranous and also cytoplasmic staining was found in lymph node metastases (Figure 3C). Strong membranous staining was focally identified with a homogenous expression throughout the entire cell population. In contrast, visceral metastases of primary cutaneous melanoma were completely negative (Figure 3D), with one exception of a metastasis to the cerebrum (not shown). In determining immunoreactivity of melanocytic lesions by an immunoreactive score, a significant difference in staining intensity was found in nevi compared to primary melanomas, as well as cutaneous- and lymph node metastasis (Figure 3E). The same accounted for the lack of staining in visceral metastases.
Downstream signaling pathways of c-Met comprise activation of NF-κB, PI3K and MAPK (Cho et al. 2003; Recio et al. 2003; Fan et al. 2005). To determine pathways involved in CD44v6 activation by rhHGF, inhibitors of NF-κB (BAY11-7082), PI3K (Ly294002) or MAPK (PD98059) were used as single compounds or in combination (Figure 4A). Only inhibition of NF-κB, but not PI3K or MAPK, led to a significant downregulation of rhHGF induced CD44v6 expression in human melanocytes and combinations with BAY11-7082 confirmed this finding. In melanoma cells, like in WM35, inhibition of all three pathways led to a decrease of CD44v6. (Figure S2A). To further investigate on NF-κB dependent regulation of transcription factors, consensus sequences and binding sites in the promoter region of CD44v6 were found for Egr-1, C/EBP-ß and GATA2, but not for NF-κB itself reaching 1000 nucleotides upstream of the transcription initiation site. Activation of all three transcription factors has been reported to be linked directly or indirectly to NF-κB (Perkins et al., 1997; Kleemann et al., 2003; Komine et al., 2000; Kim et al., 2007; Thyss et al., 2005; Minami et al., 2001; Shen et al., 1997). Within 15 min protein levels of Egr-1 and C/EBP-ß increased significantly after exposure to rhHGF and returned to endogenous levels within 8 hrs, no change was observed for GATA2 after 15min up to 24 hrs after exposure (Figure 4B). In the primary melanoma cell line WM35, however, increased levels of Egr-1 and C/EBP-ß sustained for the period of 24 hrs (Figure S2B), indicating continuous stimulation through endogenous HGF. To determine signaling pathways important for Egr-1 and C/EBP-ß up-regulation, all three inhibitors were used. Egr-1 protein levels were reduced 15min after exposure to rhHGF by adding a HGF-neutralizing antibody or inhibiting MAPK as well as NF-κB, but not by inhibiting PI3K (Figure 4C). Decreased expression of C/EBP-ß and GATA2, however, was observed by inhibiting NF-κB, MAPK or PI3K (Figure 4C). These results point to a crucial role for NF-κB in mediating HGF induced expression of CD44v6 and, subsequently, initiation of transcription via Egr-1 and C/EBP-ß. Egr-1 binds to the consensus sequence 5′-GCG (T/G)GG GCG-3′, but binding to motifs differing in one or two bp has also been shown (Maltzman et al., 1996; Molnar et al., 1994; Pospelov et al., 1994). Two potential Egr-1 binding sites were found at bp −600 and −301 of the human CD44 promoter. Each of these motifs differs from the consensus binding site by at least one bp. Since Egr-1 binds with low affinity to the site at bp −301 of the CD44 promoter but not to the site at bp −600 (Maltzman et al., 1996), we focused on the binding site at bp −301. We found six potential consensus C/EBP-ß binding sites (5′-T(T/G)N NGN AA(T/G)-3′) (Pope et al., 1994) in the human CD44 promoter. Information about the binding properties of C/EBP-ß to the CD44 promoter and the effect of HGF on the binding properties of Egr-1 and C/EBP-ß is not available. Four fragments of 197bp (C/EBP-ß P1: two C/EBP-ß binding sites), 292bp (C/EBP-ß P2: three C/EBP-ß binding sites), 213bp (C/EBP-ß P3: one C/EBP-ß binding site) and 253bp (Egr-1: one C/EBP-ß binding site, one Egr-1 binding site) labeled 5′ with the fluorescent dye Cy3 were generated (Materials and Methods). Nuclear extracts of FOM101 were probed with the oligonucleotides and after incubation, gel shift assays were performed and shifts were envisaged for all four fragments. Then the proteins were transferred to a PVDF membrane and probed for C/EBP-ß, Egr-1, NF-κB p100/p52 and NF-κB p65 by immunoblotting. As expected Egr-1 bound to the Egr-1 binding sites at bp −301, but surprisingly Egr-1 binding was not enhanced by HGF (Figure 4D). C/EBP-ß bound to the allocated binding sites of the CD44 promoter. Egr-1 and NF-κB are able to interact with C/EBP-ß (Zhang et al., 2003; Komine et al., 2000; Shen et al., 1997) and, indeed, we observed Egr-1 as well as NF-κB p100/p52 but not NF-κB p65 at the C/EBP-ß binding sites. Interestingly, after stimulation with HGF binding of C/EBP-ß was clearly increased whereas Egr-1 and NF-κB p100/p52 binding decreased (Figure 4D and Figure S2C). These findings suggest that C/EBP-ß together with Egr-1 and NF-κB form a complex. After stimulation with HGF most of Egr-1 and NF-κB may dissociate from C/EBP-ß, resulting in higher binding of C/EBP-ß to the CD44 promoter.
Phosphorylation of c-Met is dependent on the presence of CD44v6, which forms a complex with c-Met in carcinoma as well as primary human cell lines (Orian-Rousseau et al., 2002). To see, whether this also applies for human melanocytes, c-Met phosphorylation and the influence of CD44v6 was determined. As expected, an increase in phosphorylation of c-Met was observed already 2,5min after exposure to rhHGF reaching a plateau of four-fold activation at 15min (data not shown). To test for CD44v6 dependence, the significant increase in phosphorylated c-Met 15 min after exposure to HGF was probed in the presence of a neutralizing antibody for HGF or a blocking antibody to CD44v6. Both antibodies diminished protein levels of phoshorylated c-Met at Tyr1234/1235 impressively (Figure 5A). This underscores the importance of CD44v6 for proper activation of c-Met by HGF in human melanocytes. To further investigate on the functional consequences of increased CD44v6 expression, random- and site- directed migration assays were performed. Microscopic examination at 16 and 24 hrs revealed a significant delay in wound closure of melanocytes exposed to a neutralizing antibody of HGF or a blocking antibody to CD44v6 compared with cells exposed to rhHGF alone (Figure 5B). This finding was corroborated by investigating on site-directed migration. Cellular migration towards rhHGF was significantly increased compared to control. The CD44v6 blocking antibody abrogated cell migration in the presence of rhHGF (Figure 5C). The observed changes in migration were not related to an increase in proliferation since proliferation of melanocytes was not increased in rhHGF exposed melanocytes (Figure 5D).
As a result of UV irradiation cytokines are released from keratinocytes, which subsequently induce HGF secretion in fibroblasts. One of the consequences of HGF expression is the protection of UV induced damage through preventing apoptosis in keratinocytes and fibroblasts (Mildner et al., 2007). Like in keratinocytes and fibroblasts HGF delivers pro-survival effects in human melanocytes (Beuret et al., 2007). Based on the diverse regulatory effects of HGF, such as the induction of fibronectin synthesis in melanoma cells (Gaggioli et al., 2005), the increase in MMP2 expression (Hamasuna et al., 1999) or the down-regulation of E-cadherin levels (Li et al., 2001), it appears to play a central role in alternating melanocyte homeostasis. HGF up-regulates CD44v6 expression in mouse melanoma cells (Recio et al., 2003), but there are only few reports on the expression of CD44v6 in human melanocytic cells. Here, we demonstrate that CD44v6 is expressed in human melanocytic cell lines with albeit higher levels in melanoma cell lines. Exposure to recombinant as well as adenoviral- HGF resulted in a profound increase of CD44v6 in human melanocytes, but not in melanoma cells. Melanoma cells but not melanocytes secrete HGF by themselves and the level of phosphorylated c-Met may correlate with expression levels of CD44v6 levels rather then total c-Met expression (Li et al., 2001). Melanoma cells did not respond to recombinant or adenoviral induced secretion of HGF due to a well known autocrine loop. Immunostaining of melanocytic lesions revealed a low, cytoplasmic staining of CD44v6 only in congenital nevi and nearly absence in visceral metastases of primary cutaneous melanomas, but high membranous expression in primary cutaneous melanomas, cutaneous - and lymph node metastases, which confirms our finding in melanocytic cells. This is in contrast to most of the studies on the expression of CD44v6 in melanoma tissue, in which a lack of immunostaining for CD44v6 was reported (Manten-Horst et al., 1995; Harwood et al., 1996; Fernandez-Figueras et al, 1996; Ranuncolo et al., 2002). However, we previously observed CD44v6 positivity in nests of subepidermal melanoma cells of some primary melanomas (Schaider et al., 1998). Since melanoma cell lines, which originate from lymph node metastasis, like WM9, express CD44v6, we argue, that the positive staining observed in tissue is actual.
It has been reported that the transcription factor Egr-1 regulates the expression of CD44v6 and that Egr-1 is up-regulated by HGF in mouse melanoma cells (Maltzmann et al., 1996; Recio et al., 2003). Whether this is also true for the up-regulation of CD44v6 in human melanocytes was determined. By using inhibitors for the main signaling pathways of HGF we showed that CD44v6 is exclusively regulated through NF-κB. In contrast to mouse melanoma cell lines and our observations with human melanoma cells we did not find a down-regulation of CD44v6 through inhibitors for PI3K or MAPK in human melanocytes. The reason for the selective inhibition by NF-κB may be related to the constitutive activations of the RAS/MEK/ERK and the PI3K/AKT signaling pathways in melanoma cells compared to melanocytes (Satyamoorthy et al., 2003; Meier et al., 2007).
In order to investigate on a direct transcriptional regulation of NF-κB, we searched for a consensus sequence at the CD44 promoter, but were unable to detect corresponding binding sites. Instead, consensus sequences for three transcription factors (Egr-1, C/EBP-ß and GATA2) were found. These are directly or indirectly activated by NF-κB (Perkins et al., 1997; Kleemann et al., 2003; Komine et al., 2000; Kim et al., 2007; Thyss et al., 2005; Minami et al., 2001; Shen et al., 1997). Egr-1 is not only up-regulated by HGF and binds in the promoter region of CD44 (Recio et al., 2003), but is also regulated by NF-κB in breast cells, T-lymphocytes and normal human keratinocytes (Kim et al., 2007; Thyss et al., 2005) and through the MAPK pathway, but not PI3K, in mouse melanoma cells (Recio et al., 2003). Minami et al. showed that full induction of vascular adhesion molecule 1 in human umbilical vein endothelial cells is achieved by combinatorial interactions of NF-κB and GATA2. Moreover, HGF signals through GATA1, but not GATA2 (Koibuchi et al., 2004). Our results confirm that GATA2 is not up-regulated by HGF and therefore does not play a role in HGF induced expression of CD44v6. C/EBP-ß is involved in HGF signaling in a rat hepatocyte-derived cell line and a mouse embryonic hepatocyte cell line (Cho et al., 2003; Shen et al., 1997) and C/EBP-ß forms a complex with NF-κB in different cell lines (Perkins et al., 1997; Kleemann et al., 2003; Komine et al., 2000). Additionally to Egr-1, C/EBP-ß protein levels increased after exposure to HGF, suggesting an involvement in CD44v6 transcriptional regulation.
Because HGF up-regulated Egr-1 and C/EBP-ß but not GATA2 as early as 15 min after exposure, we investigated on the regulatory pathways of the three transcription factors. We again used inhibitors for PI3K, MAPK and NF-κB. Egr-1 protein levels were reduced by inhibiting MAPK kinase pathway and NF-κB, but not by inhibiting PI3K. Decreased expression of C/EBP-ß and GATA2, however, was observed by inhibiting NF-κB, MAPK or PI3K. HGF does not influence protein levels of NF-κB but it causes a nuclear translocation of NF-κB within 5 minutes (Fan et al., 2005). After translocation, nuclear NF-κB can bind to and activate the Egr-1 and C/EBP-ß promoter (Thyss et al., 2005; Shen et al., 1997). This leads to the assumption that Egr-1 and C/EBP-ß are up-regulated through HGF at least in part via NF-κB (Figure 6A).
To confirm the binding of Egr-1 and C/EBP-ß to the CD44 promoter gel shift assays were performed. The binding of Egr-1 to the CD44 promoter and the induced expression of CD44 has been described (Maltzman et al., 1996; Recio et al., 2003). Egr-1 binding was increased in HGF stimulated mouse melanoma cells (Recio et al., 2003). In human melanocytes Egr-1 bound to the described binding site at bp −301 in the CD44 promoter but the binding did not increase after HGF stimulation (Figure 4D). Whether C/EBP-ß binds to the CD44 promoter is unknown, we determined six potential binding sites in the CD44 promoter. Indeed, C/EBP-ß bound to the located binding sites and the binding clearly increased after HGF stimulation. Interestingly, Egr-1 and NF-κB p100/p52 could also be detected at the C/EBP-ß binding sites (Figure 4D). These data left us with the idea that a complex comprising of NF-κB p100/p52, C/EBP-ß and Egr-1 exists. It has been reported that NF-κB forms a complex with C/EBP-ß, which is based upon a physical interaction of both partners. However, in this complex C/EBP-ß rather than NF-κB shows affinity to the respective promoter region and NF-κB serves as a co-activator in this complex (Shen et al., 1997; Komine et al., 2000). Furthermore, in keratinocytes, tumor necrosis factor induces the expression of the cytoskeletal protein keratin K6b through NF-κB and C/EBP-ß. There is no direct binding site for NF-κB in the K6b promoter region, but one for C/EBP-ß. K6b is only expressed when both NF-κB and C/EBP-ß are present and the induction of K6b is abolished if one of them is blocked. NF-κB directly acts via protein-protein interaction with C/EBP-ß (Komine et al., 2000). Therefore we speculate that C/EBP-ß is not only regulated through NF-κB but both may form a complex like in keratinocytes and are necessary for increased CD44v6 expression in human melanocytes. Similarly to NF-κB, Egr-1 specifically interacts with C/EBP-ß in human hepatoma cells to activate the human low density lipoprotein receptor. There is no direct binding site for Egr-1 in the promoter region of the LDL receptor but one for C/EBP-ß and both are important for increased LDL receptor expression (Zhang et al., 2003). HGF stimulation leads to an increase of C/EBP-ß binding to the CD44 promoter but Egr-1 and NF-κB binding decreased. Thus it seems that Egr-1 and NF-κB dissociate from C/EBP-ß after stimulation with HGF and more C/EBP-ß is recruited to the binding site. In human hepatoma cells oncostatin M (OM) activates the human LDL receptor through interaction between Egr-1 and C/EBP-ß (Zhang et al., 2003). OM treatment induces Egr-1 as well as C/EBP-ß expression. In the early phase of OM induction, Egr-1 associates with C/EBP-ß and the complex binds to the C/EBP-ß binding site in the LDL receptor promoter. Later on C/EBP-ß is expressed at higher levels. In this scenario Erg-1 dissociates from C/EBP-ß and the binding of C/EBP-ß to the promoter increases subsequently (Zhou et al., 2006). We assume that a similar mechanism is true for human melanocytes. After stimulation with HGF less NF-κB and Egr-1 is available for binding in the complex and the affinity of C/EBP-ß increases substantially (Figure 6B). Although HGF elevates Egr-1 as well as C/EBP-ß protein levels, the increase of C/EBP-ß is more pronounced. This may explain that HGF stimulation did not enhance Egr-1 binding but instead C/EBP-ß binding to the CD44 promoter. These findings indicate that C/EBP-ß plays a predominant role in HGF mediated expression of CD44v6 in human melanocytes. Since it is well known that CD44v6 is required for c-Met activation we questioned whether CD44v6 is also implicated in functional consequences of HGF/c-Met phosphorylation. Phosphorylation of c-Met at two specific tyrosine residues leads to growth, invasion and metastasis in cancer cells (Birchmeier et al., 2003). Blocking of CD44v6 significantly decreases the activation of c-Met and downstream targets of c-Met. We demonstrated that blocking of CD44v6 decreases random- and site-directed migration of melanocytes, which underscores the importance of CD44v6 for motility in human melanocytes mediated by HGF.
HGF is released by fibroblasts after UV-irradiation or activation of pro-inflammatory cytokines (Mildner et al., 2007). Signalling of c-Met through HGF depends on the function of CD44v6 because it is required to organize a ternary complex between c-Met, HGF and CD44v6 which is a prerequisite for c-Met activation (Orian-Rousseau et al., 2002). CD44 also monitors the cellular environment through its N-terminus that contains binding sites for hyaluronan. Importantly, these CD44-hyaluronan interactions are regulated by E-cadherin. Cells that express high levels of E-cadherin reduce the binding between hyaluronan and CD44 whereas cells with low E-cadherin and high CD44 levels display CD44-hyaluronan binding, branching morphogenesis and invasion (Xu et al., 2003). We have shown that HGF downregulates E-cadherin in melanocytes (Li et al., 2001) opening a new perspective for HGF mediated effects with respect to CD44v6. Downregulated E-cadherin may allow for interaction of CD44(v6) with hyaluronic acid in melanocytes, which present with endogenously low CD44v6 but high E-cadherin levels. Elevated levels of CD44v6 expression precede the E-cadherin downregulating effect of HGF. The requirement of CD44v6 for c-Met activation and the increased invasion after CD44-hyaluronan binding together with the immediate up-regulation of CD44v6 in response to HGF in human melanocytes suggests a vital role of CD44v6 at the very beginning in melanomagenesis.
In summary, we have shown that HGF up-regulates a constitutively low expression of CD44v6 in primary human melanocytes. In the majority of nevi CD44v6 is absent corresponding to the low expression in human melanocytes. The increase in CD44v6 expression is regulated through the NF-κB pathway via Egr-1 and C/EBP-ß in human melanocytes. Gel shift assays revealed that a complex of NF-κB, Egr-1 and C/EBP-ß is binding to the CD44 promoter at C/EBP-ß presumed binding sites. Both, Egr-1 and C/EBP-ß bind to the CD44 promoter, but HGF stimulation enhances binding of C/EBP-ß but not Egr-1, underscoring a major role for this transcription factor in HGF mediated CD44v6 expression. Blocking of CD44v6 decreases the activation of c-Met as well as enhanced random- and site-directed migration of HGF stimulated human melanocytes.
The human primary melanoma cell lines WM35, SBcl2, WM793 and WM278 and metastatic melanoma cell lines WM9, WM164, 451lu and 1205lu were cultured in RPMI 1640 (Invitrogen Ltd, Paisley, UK) supplemented with 2% FBS, 2% L-glutamine, and 2% antibiotics (cell lines were provided by M. Herlyn, Wistar Institute, PA, USA). Human melanocytes, derived from human foreskin, were cultured in human melanocytes growth medium (PromoCell GmbH, Heidelberg, Germany). All cells were maintained at 37°C in a humidified atmosphere containing 5% CO2. Cells were washed with PBS after 24hrs of starvation, except for melanocytes and then exposed to recombinant human HGF protein (R&D Systems Inc., Minneapolis, MN) in serum and protein free media up to 24hrs at 50ng/ml. Where specified, the cells were incubated prior to HGF activation with anti-CD44v6 antibody (15μg/ml) or anti-HGF antibody (1μg/ml) for 1 hour or with Inhibitors PD98059 (25μM), Ly294002 (25μM) and BAY11-7082 (10μM) for 30 min. Afterwards, whole cell extracts were generated at indicated time intervals. Where indicated, subconfluent melanocytes and melanoma cells were transduced with 20 plaque-forming unit/cell of each adenovirus vector for 4 hrs at 37°C. Viral suspension was then replaced with culture medium. After additional 48 hrs cytoplasmic extracts (Nuclear Extract Kit; Active Motive, Rixenart, Belgium) were generated.
The recombinant adenoviral vector (Ad.CMV.rhHGF) coding for human hepatocyte growth factor and an adenoviral vector containing LacZ (vector control) were kindly provided by Dr. Yoshiteru Murofushi (Kurume University, Japan). The recombinant adenovirus (EDL-Ad) coding for human endothelial cell-derived lipase was kindly provided by Dr. Sasa Frank (Medical University Graz, Austria).
Primary antibody against CD44v6 (VFF-18; 1:200) was purchased from Bender MedSystems Inc. (Vienna, Austria). Anti-Egr-1 (C-19; 1:800), anti-C/EBP ß (H-7; 1:200) and anti c-met (C-12, 1:200) were from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA). Anti-phospho c-Met (3D7; 1:1000) and anti-NF-κB p100/p52 (1:1000) was purchased from Cell Signaling Technology Inc. (Danvers, MA, USA), anti-GATA 2 (2μg/ml) was from Abcam plc. (Cambridge, UK) and anti-NF-κB p65 (NF-12; 1:1000), anti-Tubulin (B512; 1:2000) and anti-Beta-Actin (AC-15; 1:5000) were purchased from Sigma-Aldrich Inc. (St.Louis, MO, USA). For Immunofluorescence, primary antibody against CD44v6 (VFF-7; 1:75, Santa Cruz Biotechnology) was used. For flow cytometry a FITC-conjugated antibody against CD44v6 (VFF-7) from Serotec (Oxford, UK) was used. Inhibitors PD98059 [25μM], Ly294002 [25μM] and BAY11-7082 [10μM] were purchased from Calbiochem (Merck Chemicals Ltd., Nottingham, UK). Anti CD44v6 Antibody (2F10) was used as blocking antibody for CD44v6 and anti HGF (24612) was used as neutralizing Antibody for HGF, both were purchased from R&D Systems Inc. (Minneapolis, MN, USA).
To generate whole cell extracts, cells were washed three times in PBS and lysed in ice cold RIPA-Buffer (Sigma) containing Protease Inhibitor Cocktail (Roche Diagnostics GmbH, Mannheim, Germany) and Phosphatase Inhibitor Cocktail (Roche) for 20 minutes. For immunoblotting, cytoplasmic extracts or whole cell extracts were used. Protein concentration was measured using the Bradford protein assay (BioRad Laboratories GmbH, Munich, Germany). Equal protein aliquots (20μg) were loaded, separated on 7,5% SDS-polyacrylamide gels, transferred to polyvinylidene difluoride (PVDF) membranes and probed with specific primary antibodies. To detect the signal, peroxidase-conjugated secondary antibody was added. Proteins were visualized using enhanced chemiluminescence (ECL) reagents (Amersham, Arlington Heights, IL, USA) and exposed to X-ray film. For quantification, the blots were analyzed in a Chemi Doc TM XRS Universal Hood (Bio-Rad).
RNA was isolated (RNeasy Extraction Kit, Quiagen) from human melanocytes and human melanoma cell lines. 1 μg of total RNA was reverse transcribed using a complementary DNA (cDNA) synthesis kit (QuantiTec Reverse Transcription Kit, Quiagen), according to the manufacturer’s protocol. Real-time PCR was performed in a 7300 real time PCR system (Applied Biosystems, Foster City, CA, USA) using SYBR Green Master Mix (Applied Biosystems). Primers for CD44v6 were forward, 5′-GCAACTCCTAGTAGTACAACGGAAGA-’3, reverse, 5′-CGATATCCCTCATGCCATCTGT-’3, for c-Met forward, 5′-AACCATTTCAACTGAGTTTGCTGTT-’3, reverse, 5′-TTCACGGTAACTGAAGATGCTTGT-’3 and for GAPDH forward, 5′-CTGCCCCCTCTGCTGATG-’3, reverse, 5′-GCTGATGATCTTGAGGCTGTTG-’3
Melanocytes were grown on poly-D-lysine-eight-well chamber slides to 50-60% confluence and were transduced with 20 plaque-forming unit/cell of each adenovirus vector for 4 hrs at 37°C. Viral suspension was then replaced with culture medium. After additional 48 hrs cells were fixed with 4% Paraformaldehyde at room temperature for 10 min, permeabilized with TBS containing 0.5% Triton X-100 and 0.25% Tween 20 for 15 min at room temperature and blocked with Image iT FX Signal Enhancer (Invitrogen Ltd.) for 30 min at room temperature in a humid chamber. Following blocking, the cells were washed and primary antibody for CD44v6 (Santa Cruz Biotechnology) was added and incubated over night at 4°C. To detect the signal, Alexa Flour-conjugated secondary antibody was added. Cells were then washed, mounted with aqueous mounting medium (DakoCytomation, Glostrup, Denmark) and examined using a Nikon TS-100 fluorescence microscope.
Melanocytes were grown to 50-60% confluence and were transduced with 20 plaque-forming unit/cell of each adenovirus vector for 4 hrs at 37°C. Viral suspension was then replaced with culture medium. After additional 48 hrs cells were harvested by Accutase (Sigma) and 1 × 106 cells were fixed with 4% Paraformaldehyde at room temperature for 10 min. Cells were resuspended in PBS, incubated with FITC-conjugated CD44v6 antibody for 45 min at room temperature, washed once with PBS and analyzed using a FACSCalibur flow cytometer (BD Biosciences) at a wavelength of 488nm.
Formalin-fixed paraffin-embedded tissue was stained after antigen-retrieval with DAKO target retrieval solution pH9 (DAKOCytomation, Glostrup, Denmark) following the manufactures guideline. For subsequent CD44v6 visualization the monoclonal antibody VFF-18 (Bender MedSystems, Vienna, Austria) was used (1:150; 6,5μg/ml) in combination with the EnVision System HRP DualLink Mouse/Rabbit (DAKOCytomation, Glostrup, Denmark) detection system. .Appropriate positive and negative control slides were analysed in parallel. Scoring of tissue slides was performed independently by 2 investigators; percentage of positive cells and the intensity of staining were graded from 0 to 2+: 0, no staining; 1,+ weak positive staining; 2+, moderate positive staining. An immunoreactive score (IRS) was obtained by multiplying the percentage of positive cells with staining intensity divided by 10.
Band–shift assays were performed with Cy3 labelled sequences of CD44 including C/EBP-ß or Egr-1 binding sequences (four oligonucleotides, C/EBP-ß P1 −3396 to −3200, C/EBP-ß P2 −2326 to −2035, C/EBP-ß P3 −562 to −350 and Egr-1 −500 to −248) as probe. Primers for C/EBP-ß P1 were forward, 5′-TTGCAAAGCACTTTCACACC-3′, reverse, 5′-CCATCCCAACCACAGAAGTT-3′, for C/EBP-ß P2 forward, 5′-TCTCTGGACAAGCCATGTTCT-3′, reverse, 5′-CCAAACCCTATTATGGCTGCT-3′, for C/EBP-ß P3 forward, 5′-AACCCAGAGATCTTGCTCCA-3′, reverse, 5′-GTCGGGGAACCTGGAGTGT-3′, and for Egr-1 forward, 5′-CTGAACCCAATGGTGCAAGGT-3′, reverse, 5′-CCGCAGAGGTTTTAAGAAGTAGCA-3′. Melanocytes (FOM101) were left unstimulated or stimulated with HGF for 15 min and nuclear extracts were generated. Nuclear extracts (10μg) were incubated with Cy3-labeled double strand oligonucleotides for 15min at 37°C in binding buffer (10mM Tris (pH 7.5), 50mM NaCl2, 1mM DTT, 0,1mM EDTA, 5% glycerol) with 2μg of poly(dI-dC). After incubation the reaction samples were separated by non-denaturing PAGE. The separated DNA-protein complexes were transferred to a PVDF membrane and probed with C/EBP-ß, Egr-1, NF-κB p100/p52 or NF-κB p65 antibodies by immunoblotting as described above.
To test for random migration, wound-healing assays were performed by plating Melanocytes in six-well plates reaching 80% confluence. Cells were exposed to recombinant human HGF protein (50ng/ml) in serum and protein free medium for 8 hours. Where indicated, cells were treated prior to HGF activation with anti-CD44v6 antibody (15μg/ml) or anti-HGF antibody (1μg/ml) for 1 hour. After 8 hours, a scratch or wound was made using a sterile pipette tip. Plates were washed with medium to remove all detached cells and incubated in human melanocytes growth medium up to 24 hours. Photographs of cells invading the scratch were taken at indicated time intervals using an Olympus IX51 inverted microscope. Assays were performed in triplicate for each condition.
The migration assay was performed using Cell Culture Inserts with PET membrane (8μm pore size) (BD Falcon). Melanocytes were treated with recombinant human HGF protein (50ng/ml) in serum and protein free medium for 8 hours. For inhibition, cells were preincubated with anti-CD44v6 antibody (15μg/ml) or anti-HGF antibody (1μg/ml) for 1 hour. 8 hours after HGF stimulation, cells were harvested by Trypsin/EDTA buffer and 4 × 104 cells in serum free medium were added to the upper compartment of the Cell Culture Inserts. The lower compartment was filled with serum and protein free medium with 50ng/ml recombinant human HGF protein. After 16 hours of incubation at 37°C, the upper surface of the membrane was wiped with a cotton swab to remove non migratory cells. Cells that migrated to the bottom side of the membrane were fixed with methanol, stained with toluidine blue for 10 min and the membrane was washed twice with water. Membranes were placed on a glass slide, and from each membrane six fields were counted using an optical microscope. Assays were performed in triplicate for each condition.
Melanocytes were plated in six well plates (7,5 × 104 cells per well) and exposed to recombinant human HGF protein (50ng/ml) in serum and protein free medium for 8 hours. Where specified, cells were treated prior to HGF activation with anti-CD44v6 antibody (15μg/ml) or anti-HGF antibody (1μg/ml) for 1 hour. After 8 hours of incubation, the medium was changed to fresh human melanocytes growth medium. Cells were counted with a Neubauer counting cell chamber at indicated time intervals. The cell cytotoxicity was examined using trypan blue exclusion assay. Assays were performed in triplicate for each condition.
Statistical analyses were performed using Student’s t-test. Results are calculated as the mean ± SD of three different experiments. P values less than 0.05 were considered significant.
Figure S1: A Western Blot of CD44v6 expression in melanocytes (FOM101), primary melanoma cells (WM35) and metastatic melanoma cells (WM9) after adenoviral transduction with Ad.CMV.rhHGF (20pfu/cell) or vector control (LacZ) for 48 hours. 20μg of cytoplasmic extracts were loaded per lane and samples were separated on a 7,5% SDS-polyacrylamide gel. Blots were probed against indicated antibodies. ß-Actin was used as loading control. Numbers indicate changes in percent of control. Bands were quantified by densitometry. Data are normalized to ß-Actin and presented as % of control.
B Western Blot of CD44v6 expression in metastatic melanoma cells (WM9) without or with treatment (exposure) of HGF neutralizing antibody (1μg/ml) after eight hrs. 20μg of total protein were loaded per lane and samples were separated on a 7,5% SDS-polyacrylamide gel. Blots were probed against indicated antibodies. ß-Actin was used as loading control. Numbers indicate changes in percent of control. Bands were quantified by densitometry. Data are normalized to ß-Actin and presented as % of control.
Figure S2: A Western Blot of CD44v6 expression in primary melanoma cells (WM35) after exposure to HGF (50ng/ml) for eight hrs in the presence of inhibitors. Ly294002 (25μM), PD98059 (25μM) and BAY-11-7082 (10μM) were added 30 minutes prior to HGF/SF exposure. 20μg of total protein was loaded per lane and samples were separated on a 7,5% SDS-polyacrylamide gel. Blots were probed against indicated antibodies. ß-Actin was used as loading control. Numbers indicate changes in percent of control. Bands were quantified by densitometry. Data are normalized to ß-Actin and presented as % of control.
B Western Blots of Egr-1 and C/EBPß expression in primary melanoma cells (WM35) after exposure to HGF (50ng/ml) over 24 hrs. 20μg of total protein was loaded per lane and samples were separated on a 7,5% SDS-polyacrylamide gel. Blots were probed against indicated antibodies. ß-Actin was used as loading control. Numbers indicate changes in percent of control. Bands were quantified by densitometry. Data are normalized to ß-Actin and presented as % of control.
C Gel shift assays with nuclear extracts of unstimulated or with HGF stimulated human melanocytes (FOM101). Nuclear extracts (10μg) were incubated with Cy3-labelled double strand oligonucleotides. The resulting complexes were resolved by non-denaturing PAGE and afterwards transferred to PVDF membrane and probed with NF-κB p65 antibodies.
We thank the stuff of the core facility Flow Cytometry of the Center for Medical Research (ZMF) and Dr. Ivan Kanchev from Oridis for technical assistance.
Supported by the Austrian Science Fund (grant Nr. P18630-B05)
CONFLICT OF INTEREST The authors state no conflict of interest.