Hepatocyte Growth Factor/Scatter Factor Induces Feedback Up-Regulation of CD44v6 in Melanoma Cells through Egr-1

HGF/SF² is a pleiotropic cytokine/growth factor capable of inducing an extraordinary variety of biological activities, including cellular proliferation, migration, and invasiveness (1, 2, 3, 4, 5). This M_r 92,000 factor consists of a β chain with an enzymatically inactive serine protease domain and an α chain with an NH₂-terminal domain and four kringle domains. NK1 and NK2 are naturally occurring truncated variants of HGF/SF. NK1 consists of the N domain and first kringle domain of the HGF/SF α chain, whereas the structure of NK2 contains the second kringle domain as well (6, 7, 8, 9). NK1 has been reported to be a partial agonist of HGF/SF (9, 10, 11, 12), whereas NK2 behaves mostly as an antagonist (4, 6, 13, 14, 15). All activities of HGF/SF and its variant forms are mediated through the receptor tyrosine kinase c-Met, resulting in the activation of one of the most complex signaling networks attributable to a receptor tyrosine kinase. c-Met activation promotes the recruitment of several protein kinases and intermediates, including PI3K, phospholipase C-γ, Stat 3, c-Src, Grb-2, and Gab-1, leading to the activation of a wide variety of downstream pathways, including the Ras>MAP kinase, PI3K>Akt, c-Jun-NH₂-terminal kinase/stress-activated protein kinase, p38, and focal adhesion kinase pathways (16, 17, 18, 19, 20, 21, 22, 23).

In addition to binding c-Met, HGF/SF, NK1, and NK2 also have a high affinity for heparin (24, 25, 26). HS, considered to be the low-affinity receptor for HGF/SF, is a GAG structurally related to heparin and widely expressed in the form of proteoglycan species on the cell surface and in the extracellular matrix. It has been extensively demonstrated that HS plays a role in the presentation of several growth factors to their receptors (27, 28, 29, 30, 31, 32, 33). A number of studies has correlated the response to HGF/SF and its isoforms to the presence of GAGs on the cell surface (10, 32, 34, 35, 36), e.g., when modified by association with HS, the glycoprotein CD44 participates in the presentation of HGF/SF to c-Met (32).

CD44 molecules constitute a family of multifunctional, transmembrane glycoproteins occurring in multiple forms and implicated in diverse normal and pathological cellular functions, including lymphocyte activation and homing and tumor progression and metastasis (37, 38, 39, 40). CD44 is a single chain protein consisting of an NH₂-terminal extracellular domain, membrane proximal domain, transmembrane domain, and cytoplasmic tail. The major ligand of CD44 is hyaluronic acid, which is an ubiquitous part of the extracellular matrix. In addition, CD44 can associate with growth factors; this interaction is facilitated by CD44 oligomerization (reviewed in Ref. 40).

The most common isoform expressed in a wide variety of cell types is CD44s (for standard). However, there are several alternatively spliced forms, designated v1–v10, which include an insertion at the membrane proximal region of the extracellular domain. The distribution of these CD44 variants is usually restricted, and some variants are only expressed in certain tumor cells (41, 42). Of particular interest was the detection of CD44v4–v7 and CD44v6–v7 in human lymphomas, colorectal adenocarcinomas, endometrial cancer, papillary thyroid carcinoma, lung and breast cancer, and metastasizing rat adenocarcinomas (43, 44, 45, 46, 47, 48, 49). Forced expression of CD44v isoforms can confer metastatic properties to nonmetastatic cells when injected into an isogenic host (43, 50, 51). Moreover, the systemic application of antibodies directed against the v6 epitope and the expression of antisense CD44v6 can retard tumor growth and block metastasis in vivo (43, 44, 52). Injection of hyaluronan oligomers into melanoma cells inhibits tumor formation (53).

The expression of CD44 and regulation of the alternative splicing of the variants are stimulated in response to hormones, cytokines, and growth factors, particularly by the Ras>Erk pathway (54, 55, 56), through the activation of transcription factors such as Egr-1 and AP-1 (57, 58, 59). The immediate early gene egr-1 [also known as NGFI-A (60), Krox-24 (61), and Zif268 (62)] encodes a nuclear, zinc finger protein capable of binding to specific GC-rich DNA sequences (63, 64). The egr-1 gene is rapidly and transiently induced by growth factors and differentiation signals (60, 64) through different signal transduction pathways, including the PKC>Ras>MAP kinase and PI3K pathways (65, 66, 67).

In this study, we show that 37-32 mouse melanoma cells express the metastasis-associated protein CD44v6, which can facilitate presentation of HGF/SF to its receptor c-Met. Furthermore, we demonstrate that HGF/SF, but not NK2, promotes the feedback up-regulation of CD44v6 through the transcriptional activation of egr-1.

The antibodies used were obtained from the following commercial sources: (a) anti-c-Met (SP260), anti-AP-1/c-Jun, and anti–Egr-1 (C-19; Santa Cruz Biotechnology, Santa Cruz, CA); (b) anti-P-c-Met (Y^1230/34/35) (Biosource); (c) anti-actin (Chemicon International, Inc., Temecula, CA); and (d) antimouse-CD44v6 (Bender Medsystems). Western blot detection was accomplished using horseradish peroxidase-coupled antimouse, antirabbit (Amersham Pharmacia Biotech), and antirat secondary antibodies (Sigma Chemical Co., St. Louis, MO). Secondary antirabbit and antirat FITC antibodies and Texas red conjugates for fluorescence microscopy and FACS were purchased from Vector Laboratories, Inc. (Burlingame, CA). The Mek1/2 inhibitor PD98059 and U0126 were obtained from Cell Signaling Technologies. The PI3K inhibitor LY294002 was from Upstate Biotechnology, Inc. (Lake Placid, NY). Staurosporine was purchased from Roche Laboratories. Recombinant human HGF/SF was from R&D Systems, Inc. (Minneapolis, MN). The pBlueCD44 full cDNA was obtained from Dr. John Monroe (University of Pennsylvania). Purified NK2 was a generous gift from Dr. Ralph Schwall (Genentech).

The 37-32 mouse cell line was derived from a cutaneous melanoma arising in the HGF/SF transgenic mouse line MH37 (15). Cells were maintained in DMEM (Life Technologies, Inc., Gaithersburg, MD), supplemented with 10% fetal bovine serum (Harlan Sprague Dawley, Indianapolis, IN), 100 IU/ml penicillin, 100 μg/ml streptomycin, 2 mm l-glutamine (Life Technologies, Inc.), 5 μg/ml insulin (Sigma), and 5 ng/ml epidermal growth factor (Upstate Biotechnology), and incubated in 5% CO₂ at 37°C. Cells were treated with 40 ng/ml HGF/SF (R&D Systems) and harvested at the times indicated.

After HGF/SF treatment, cells were lysed with M-PER (Pierce Chemical Co., Rockford, IL) in the presence of protease inhibitor cocktail (Roche). Fifty μg of protein from lysates were separated by SDS-PAGE using 4–20% gradient gels (Novex). Proteins were transferred to nitrocellulose membranes (Novex) and probed against primary antibodies. Proteins were visualized with secondary antibodies coupled to horseradish peroxidase and developed using Super Signal-West Pico Chemiluminescent Substrate (Pierce). Immunoprecipitation of c-Met was performed by treating the cells with M-PER (Pierce) and incubating the resulting lysates with either antiphospho-antibodies or anti-c-Met overnight at 4°C. After the addition of protein A/G agarose PLUS (Santa Cruz Biotechnology) and washing in M-PER buffer, samples were fractionated by SDS-PAGE using 4–20% gradient gels (Novex). After transfer, blots were blocked and incubated overnight with an anti-c-Met antibody. Bands were visualized by incubation with a secondary antirabbit antibody conjugated to horseradish peroxidase and Super Signal-West Pico Chemiluminescent Substrate. After stripping, membranes were blocked and incubated overnight with other antibodies as indicated in the figure legends.

Twenty μg of total RNA were resolved on a denaturing 1% agarose formaldehyde gel and transferred to a nitrocellulose membrane (Schleicher & Schuell; Refs. 23 and 68). Membranes were prehybridized and hybridized against CD44 using ultrahybridization solution (Sigma) following the manufacturer’s instructions and subjected to autoradiography.

Mouse pathway finder array (Superarray, Inc.) was used to hybridize cDNA from 37-32 cells, treated as indicated with either HGF/SF or NK2. The array was performed following the manufacturer’s protocol using 5 μg of total RNA as a template for cDNA synthesis.

Gel mobility shift assays and nuclear extracts were as described by Andrews and Faller (69). Briefly, cells were lysated in buffer A [10 mm HEPES-KOH (pH 7.9), 1.5 mm MgCl₂, 10 mm KCl, and 0.5 mm DTT plus protease inhibitor cocktail] for 10 min at 4°C. After centrifugation, the pellet was resuspended in buffer B [20 mm HEPES-KOH (pH 7.9), 25% glycerol, 420 mm NaCl, 1.5 mm MgCl₂, 0.2 mm EDTA, 0.5 mm DTT, and protease inhibitor cocktail]. Samples were incubated in ice for 30 min and then subjected to centrifugation to recover the nuclear proteins. EMSAs were performed as described by Recio and Aranda (70). Five μg of nuclear extracts were added to a buffer containing ∼30,000 cpm of P³²-labeled oligonucleotide, 0.1 μg of Poly (dI-dC), 40 mm HEPES (pH 7.0), 140 mm NaCl, 4 mm DTT, 0.01% NP40, 100 mg/ml BSA, and 4% Ficoll. After incubation, the samples were loaded onto a 6% nondenaturing polyacrylamide gel. AP-1, Egr-1, and Egr-1 mutant oligonucleotides were purchased from Santa Cruz Biotechnology.

Cells were grown in slide chambers as described above. After treatment, cells were fixed in cold methanol for 10 min and air dried. Cells were then permeabilized in PBS containing 0.1% Triton X-100 and blocked in PBS containing 1% BSA for 1 h. Primary antibodies were then incubated at a 1:50 dilution in the blocking solution for 1 h, followed by three washes with PBS. The FITC- and Texas red-conjugated secondary antibodies were incubated for an additional hour in blocking solution either at a 1:50 or 1:100 dilution. After three washes with PBS, cells were mounted in fluorescent mounting medium (Zymed), and images were captured using a confocal microscope (Bio-Rad). Flow cytometry followed routine procedures. Samples were analyzed using a FACSCalibur (Becton Dickinson).

For antisense experiments, the Egr-1 sense oligo: Agt GTTCCCCGCGCCCCGca, or the antisense Egr-1 oligo: tgCGGGGCGCGGGGAACa cT (where the bases in small letters were phophorothioated), was added to cells in serum-free media and starved 12 h before treatment. Total RNA was then extracted, and a Northern blot was performed as described above.

The ChIP assay was performed using the ChIP assay kit (Upstate Biotechnology) following the manufacturer’s directions. Anti-Egr-1 (C-19) was used to immunoprecipitate Egr-1-containing complexes. The Egr-1-binding site in the mouse CD44 promoter was amplified by a nested PCR assay using the following primers: CD445′SC: GACGGGAACCATGTATGTGCGTCG; CD443′ASC: GACACACCAGCCTGCTTTC GCTAC, CD445′SL (nested): AAGAGGAAAGAACACTCCCACAAGTTCCCC, CD443′ASL (nested): TAAAGACACACCAGCCTGCTTTCGCTACTG. The PCR conditions were as follows: (a) 5 min at 94°C; (b) 35 cycles of 30 s at 65°C; and (c) 1 min at 68°C, followed by 7 min at 72°C.

Melanoma cells isolated from HGF/SF transgenic mice (line 37-32) express high levels of c-Met and are very responsive to HGF/SF (15, 23). Recently, a role for CD44 in presenting HGF/SF to c-Met in colorectal cancer cells has been described (32, 71). We therefore explored a possible role for CD44 in HGF/SF-Met signaling in a melanoma cell line derived from an HGF/SF transgenic mouse. Preliminary data using confocal microscopy revealed that HGF/SF and NK2 promoted the colocalization and cointernalization of c-Met and the v6-containing isoform of CD44 (hereafter referred to as CD44v6) in 37-32 cells (data not shown). Western blot and FACS analyses then showed that these melanoma cells not only expressed the CD44v6 isoform but that the cell surface levels of this receptor were elevated 24 h after HGF/SF treatment (Fig. 1, A and B).

Because the variant CD44v6 is associated with HS, we determined if HS chains were required for HGF/SF responsiveness by treating the 37-32 cells with heparitinase III before HGF/SF stimulation. Fig. 1 C shows that pretreatment of 37-32 cells with heparitinase III totally blocked the ability of HGF/SF to phosphorylate c-Met tyrosines at the positions Y^1230/34/35. This result, although not conclusive standing alone, supports a role for the HS-modified CD44v6 isoform in the presentation of HGF/SF to c-Met, in agreement with data reported previously by several authors (10, 32, 34, 35, 36).

Next, Northern blot hybridization was used to determine whether HGF/SF induced the expression of CD44v6 at the level of RNA and when this induction occurred. Fig. 2,A shows that CD44v6 transcript levels were elevated by exposure to HGF/SF, reaching a 2-fold peak at 2 h and decreasing thereafter. At 6 h, the levels of CD44v6 RNA were below basal expression. The Mek1/2-specific inhibitor, U0126, was able to inhibit HGF/SF-mediated up-regulation of CD44v6, demonstrating a role for the Erk1/2 pathway (Fig. 2,B). In contrast to HGF/SF, the variant NK2 could not activate Erk1/2 in 37-32 melanoma cells (15) and did not enhance expression of CD44v6 (Fig. 2 B). These data show that HGF/SF, but not NK2, up-regulates CD44v6 at least in part through the Ras>Erk1/2 pathway and are consistent with a transcriptional mechanism underlying enhanced CD44v6 expression.

To look for transcription factors responsible for the up-regulation of CD44v6 by HGF/SF, we performed a cDNA array. Total RNA from cells treated either with HGF/SF or NK2 for different time periods was used to generate the cDNA populations for array analysis. The cDNA array revealed significant elevation in expression of the immediate early gene egr-1 after 40 min of treatment with HGF/SF but not NK2; egr-1 expression returned to basal levels by 4.5 h (Fig. 3 A).

Examination of Egr-1 expression by Western blot analysis of total lysates showed that Egr-1 protein was elevated after 30 min, reaching a peak by 2 h and thereafter becoming undetectable (Fig. 3 B). These data indicate that, like CD44v6, the Egr-1 transcription factor is up-regulated by HGF/SF but not NK2.

Egr-1 can be regulated through the PKC, Erk1/2, and PI3K signaling pathways (61, 62, 63). Western blot analysis of nuclear protein extracts from 37-32 cells after 1.5 h treatment with HGF/SF in the presence or absence of specific inhibitors for PI3K or Mek1/2 (LY294002 or PD98059, respectively) showed that Egr-1 up-regulation was mediated through the Erk1/2 pathway and not the PI3K pathway (Fig. 4,A). These results were corroborated by immunofluorescent staining of 37-32 cells with an Egr-1 antibody (Fig. 4,B) and by Western blot analysis of total protein extracts (data not shown). Immunofluorescence and Western blot data confirmed the inability of NK2 to up-regulate Egr-1 (Fig. 4, B and C).

To determine whether PKC was involved in the up-regulation of Egr-1 by HGF/SF, melanoma cells were treated with the PKC-specific inhibitor staurosporin (100 nm) for 1 h before HGF/SF treatment. Western blot analysis of nuclear extracts from these cells showed that staurosporin could not block HGF/SF-mediated Egr-1 up-regulation (Fig. 4 C). These results demonstrate that, like CD44v6, the regulation of Egr-1 by HGF/SF is mediated through the Ras>Erk1/2 pathway. Moreover, up-regulation of Egr-1 precedes that of CD44v6.

The HGF/SF-induced up-regulation of Egr-1 protein and CD44v6 RNA is consistent with a role for Egr-1 in the transcriptional activation of CD44 (Fig. 2,A). Maltzman et al. (57) have demonstrated previously a functional binding site for Egr-1 in the human CD44 gene promoter in response to the B-cell antigen receptor. To prove that a comparable Egr-1-binding site functions in the CD44 mouse promoter, several experiments were designed. The 37-32 melanoma cells were transfected either with egr-1 sense or antisense oligonucleotides and then treated with 40 ng/ml HGF/SF for 2 h. Northern blot analysis showed that the 2-fold increase in CD44v6 RNA induced by HGF/SF was partially blocked by the presence of the egr-1 antisense, but not the egr-1 sense, oligonucleotide (Fig. 5,A), indicating that Egr-1 is required for HGF/SF-mediated CD44v6 up-regulation. Next, we attempted to detect the direct binding of Egr-1 to the mouse CD44 promoter using a ChIP assay. The 37-32 mouse melanoma cells were treated with HGF/SF for 1 h, and then specific chromatin–protein complexes were immunoprecipitated using an anti-Egr-1 antibody. The region homologous to the Egr-1-binding site was then subjected to PCR amplification using as a template the DNA from the Egr-1-specific immunoprecipitated complexes. The appropriate band (451 bp) corresponding to the fragment containing the predicted Egr-1-binding site was detected only in the HGF/SF-treated cells (Fig. 5,B). As a control, the amplification was performed on complexes obtained without exposure to Egr-1 antibody and from the untreated 37-32 melanoma cells (Fig. 5 B).

An EMSA was also performed using an Egr-1 consensus motif oligonucleotide. Nuclear extracts generated from HGF/SF-treated cells showed an increase in binding to the labeled Egr-1 consensus sequence oligonucleotide (Fig. 5,C). As expected, this increase in binding was blocked in cells treated with the Mek1/2 inhibitor PD98059 but not in those treated with the PI3K inhibitor (Fig. 5,C). There was no binding when the nuclear extracts from the HGF/SF-treated cells were assayed using a mutant Egr-1-binding site oligonucleotide. Others have described the participation of the AP-1 transcription factor in the transcriptional up-regulation of the CD44 receptor (58, 59). Fig. 5 D shows a time course of the activation of AP-1 after HGF/SF stimulation. Maximum activation occurred 1 h after triggering, and the binding was inhibited by the presence of an anti-AP-1 antibody.

Fig. 5,E displays the sequence of the mouse CD44 promoter from −685 to the transcriptional start site. Putative transcription factor-binding sites are indicated, including the transcriptional start site identified using the TESS program (59, 72). The Egr-1-binding site for the mouse promoter is located at a position 235 upstream of the transcriptional start site (Fig. 5 E).

HS proteoglycans have been shown to be crucial for ligand–receptor interactions (27, 28, 29, 30, 31, 32, 33). Although their role in HGF/SF-Met interaction is less well defined, several in vitro studies have indicated that these proteoglycans may play an important regulatory role in HGF/SF-Met signaling (10, 32, 34, 35, 36). Some isoforms of the CD44 family have been implicated in the presentation of HGF/SF to c-Met (32, 73). These HS proteoglycans may help to localize HGF/SF to specific cells or extracellular matrix components within the microenvironment and may be required for the establishment of a chemotactic gradient (38, 42, 74). In this study, we show that 37-32 murine melanoma cells, which already possess a strong HGF/SF-Met autocrine loop, also express the variant 6 isoform of CD44. It has been proposed that the presence of the v6 and v7 regions promotes CD44 oligomerization, which in turn facilitates binding to ligands (reviewed in Ref. 40); these ligands include GAGs and growth factors, such as HGF/SF. We further demonstrated that expression of the CD44v6 variant is up-regulated by HGF/SF through a feedback loop requiring the presence of Egr-1.

A number of studies has demonstrated that enhanced activity of oncogenic Ras-associated MAP kinase pathways in different cellular systems regulates the expression and alternative splicing of the CD44 family proteins (54, 55, 56). The expression of CD44v6 in melanoma cells and its relationship to HGF/SF-Met signaling are significant because the CD44v6 isoform has been implicated in metastasis and tumor progression (47, 48, 72). In some cases, such as gastric carcinoma, colorectal adenocarcinoma, pancreatic cancer, and breast cancer, the expression of CD44v6 has been proposed as a marker for poor prognosis (47, 49, 75, 76). Aberrant c-Met expression has also been documented in a variety of human cancers (reviewed in Refs. 5, 77, and 78). Moreover, HGF/SF is known to up-regulate expression of c-Met in normal and malignant cells (79, 80, 81). Thus, our data forge an intriguing regulatory link between activation by the multifunctional cytokine HGF/SF, and up-regulation of those receptors required to sustain its cellular activities. Interestingly, we found that NK2, which can behave as an HGF/SF antagonist in a number of cell types (4, 6, 13, 14, 82), is unable to up-regulate CD44v6 expression in 37-32 mouse melanoma cells, although both ligands induce colocalization and internalization of c-Met and CD44v6 (data not shown). As a consequence, NK2 cannot replenish cell surface levels of CD44v6, offering at least a partial explanation for its antagonistic behavior.

Most studies of CD44 promoter regulation have focused on the methylation state of the promoter. However, Maltzman et al. (57) have reported that the Egr-1 and AP-1 transcription factors are essential for the regulation of the human CD44 promoter in B-lymphocytes. Data presented in this study are the first to demonstrate the up-regulation of the Egr-1 transcription factor in response to HGF/SF. Transcriptional regulation of Egr-1 by HGF/SF-Met signaling is driven through Ras>Erk1/2 but is independent of PI3K and PKC (65, 66, 67). Because NK2 does not activate Erk1/2 in 37-32 cells (15), it was not surprising to find that NK2 cannot up-regulate either Egr-1 or CD44v6.

We anticipate that other transcription factors will participate in regulating the activity of the mouse CD44 promoter (see Fig. 5 D). It has been shown that transcriptional activation of c-met by HGF/SF is mediated through AP-1 complexes (81). Additional data from our cDNA array (data not shown) and EMSA analysis implicate other early genes, such as c-Jun and c-fos, as HGF/SF-Met targets; c-Jun and c-fos are part of the AP-1 complex (83). These data, and the fact that AP-1-binding sites have been identified in the human CD44 promoter (57, 59), suggest that in addition to Egr-1, AP-1 may be involved in the transcriptional regulation of CD44.

Two different groups have demonstrated the functionality of the Egr-1-binding site in the human CD44 promoter (57, 58); however, the homologous site in the mouse promoter has not been reported. The antisense experiment and ChIP assay described here demonstrated not only the presence of the binding site in the homologous mouse promoter but its functionality as well. In the human CD44 promoter, the Egr-1-binding site is located at position −301 overlapping an SP1-binding site (57). In the mouse promoter, the Egr-1-binding site is located at position −235, also overlapping an SP1-binding site. Although the mouse Egr-1-binding site sequence is not identical to the human, the three nucleotides essential for the binding of Egr-1 to the homologous human site are conserved (57).

In conclusion, here we describe the novel up-regulation of the immediate early response gene, egr-1, by HGF/SF in melanoma cells. In contrast to HGF/SF, the antagonist NK2 was unable to up-regulate this transcription factor. We have further demonstrated that Egr-1 mediates enhanced CD44v6 expression in response to HGF/SF treatment. The feedback loop in which expression of both c-Met and CD44v6 is up-regulated in melanoma cells in response to receptor-depleting exposure to HGF/SF represents an important mechanism whereby cancer cells establish and maintain constitutive stimulation of growth- and metastasis-promoting pathways.

We thank Dr. Ralph Schwall for providing purified NK2 and Dr. John Monroe for the full-length CD44 cDNA.