Platelet-activating Factor Mediates MMP-2 Expression and Activation via Phosphorylation of cAMP-response Element-binding Protein and Contributes to Melanoma Metastasis*

Overexpression of cAMP-response element (CRE)-binding protein (CREB) and activating transcription factor (ATF) 1 contributes to melanoma progression and metastasis at least in part by promoting tumor cell survival and stimulating matrix metalloproteinase (MMP) 2 expression. However, little is known about the regulation of CREB and ATF-1 activities and their phosphorylation within the tumor microenvironment. We analyzed the effect of platelet-activating factor (PAF), a potent phospholipid mediator of inflammation, for its ability to activate CREB and ATF-1 in eight cultured human melanoma cell lines, and we found that PAF receptor (PAFR) was expressed in all eight lines. In metastatic melanoma cell lines, PAF induced CREB and ATF-1 phosphorylation via a PAFR-mediated signal transduction mechanism that required pertussis toxin-insensitive Gαq protein and adenylate cyclase activity and was antagonized by a cAMP-dependent protein kinase A and p38 MAPK inhibitors. Addition of PAF to metastatic A375SM cells stimulated CRE-dependent transcription, as observed in a luciferase reporter assay, without increasing the CRE DNA binding capacity of CREB. Furthermore, PAF stimulated the gelatinase activity of MMP-2 by activating transcription and MMP-2 expression. MMP-2 activation correlated with the PAF-induced increase in the expression of an MMP-2 activator, membrane type 1 MMP. PAF-induced expression of pro-MMP-2 was causally related to PAF-induced CREB and ATF-1 phosphorylation; it was prevented by PAFR antagonist and inhibitors of p38 MAPK and protein kinase A and was abrogated upon quenching of CREB and ATF-1 activities by forced overexpression of a dominant-negative form of CREB. PAF-induced MMP-2 activation was also down-regulated by p38 MAPK and protein kinase A inhibitors. Finally, PAFR antagonist PCA4248 inhibited the development of A375SM lung metastasis in nude mice. This result indicated that PAF acts as a promoter of melanoma metastasis in vivo. We proposed that metastatic melanoma cells overexpressing CREB/ATF-1 are better equipped than nonmetastatic cells to respond to PAF within the tumor microenvironment.

Although it has been shown that melanocyte proliferation and differentiation can be positively regulated by agents that increase cAMP levels (14), little is known about factors that induce CREB and ATF-1 activation in melanoma. Meanwhile, melanoma is strongly associated with the inflammatory process (15)(16)(17), and the inflammatory cells are known to be polarized within the tumor microenvironment to release growth and survival factors, extracellular proteases, proangiogenic factors, and chemokines, which contribute to tumor growth and progression to malignancy (18 -20). Among proinflammatory mediators, platelet-activating factor (PAF; 1-O-alkyl-2-acetyl-sn-glycero-3-phosphocholine) is becoming recognized as a major primary and secondary messenger involved in homotypic and heterotypic cell-to-cell communication that results in the activation of platelets, neutrophils, macrophages, lymphocytes, and endothelial cells (21). At sites of acute and chronic inflammation, PAF mediates cell migration, aggregation, adhesion to endothelial cells, chemotaxis, degranulation, production of superoxide and inflammatory cytokines, and mitosis (21)(22)(23). PAF is synthesized from the 1-O-alkyl-2-arachidonoyl-sn-glycero-3-phosphocholine via a reaction catalyzed by phospholipase A 2 /acetyl-CoA:lyso-PAF acetyltransferase, which also yields a COX-2 substrate, arachidonic acid (21). In pathologic situations, PAF and PAF-like oxidized phospholipids (21,24) are produced in an unregulated fashion, which results in stimulation of the seven-domain, membrane-spanning G protein-coupled PAFR and the induction of various intracellular signaling cascades through arachidonate and phosphoinositide turnover, activation of protein kinase C, modulation of PKA, and increased protein-tyrosine phosphorylation (24 -29).
Several lines of evidence suggest that PAFR-mediated signaling may be involved in tumor growth, angiogenesis, and metastatic dissemination. PAFR-overexpressing transgenic mice displayed proliferative disorders and melanocytic tumors (30 -32). Inactivation of PAF by overexpression of PAF acetylhydrolase inhibited the growth and vascularization of B16F10 murine melanomas and Kaposi sarcomas (33). In addition, intraperitoneally injected PAF promoted experimental lung metastasis of B16F10 cells (34). PAF mediated the adhesiveness of Hs294T human melanoma cells and LS180 human colon cancer cells to IL-1-stimulated endothelial cells (35). Vascular endothelial growth factor, basic fibroblast growth factor, hepatocyte growth factor, tumor necrosis factor-␣, and thrombin have been shown to induce PAF production by human breast cancer cells in vitro and to induce PAF-dependent cell proliferation (36). In vivo, PAFR antagonists inhibited growth and neoangiogenesis in breast cancer tumors (36). More importantly for melanoma tumorigenesis, keratinocytes secreted PAF in response to irradiation with ultraviolet light (37,38). Moreover, the keratinocytes expressed PAFR, and PAF up-regulated the expression levels of COX-2, IL-6, IL-8, and IL-10 cells and the secretion levels of prostaglandin E 2 (37,38). Despite this knowledge, the role of PAF in the acquisition of the metastatic phenotype and the changes in gene expression in human melanoma cells have not been studied.
In this study, we demonstrated that PAF induced CREB phosphorylation and activation in metastatic melanoma cells through the signaling cascade involving PAFR-mediated activation of pertussis toxin (Ptx)insensitive G␣ q protein, adenylate cyclase (AC), p38 MAPK, and PKA. Furthermore, PAF induced CREB-dependent expression and activation of MMP-2. Finally, we found that PAFR antagonists potently inhibited experimental melanoma lung metastasis. Taken together, our results proposed that compared with nonmetastatic cells, metastatic human melanoma cells that overexpress CREB and ATF-1 are better equipped to respond to inflammatory stimuli from the tumor microenvironment.
DNA Constructs-The CRE-driven luciferase reporter vector was generated by ligating one copy of the somatostatin gene promoter region (nucleotides Ϫ71 to ϩ53), into the pGL3-Luc-Basic vector (Invitrogen). The original Somat-BgllI CAT construct reporter containing somatostatic gene promoter was obtained from Dr. Marc R. Montminy (Harvard Medical School, Boston). The MMP-2-driven luciferase reporter construct containing one copy of the human MMP-2 promoter gene region (nucleotides Ϫ393 to ϩ290) was ligated into the pGL3-Luc-Basic reporter vector as described previously (39). pRSV-KCREB was kindly provided by Dr. Richard H. Goodman (Oregon Health Sciences, Portland). pRSV-KCREB expression plasmid contained full-length CREB cDNA with a single base pair substitution in the DNA-binding domain that causes a change from Arg 287 to Leu 287 (10). pRc/RSV construct (Invitrogen) lacking CREB cDNA was used as a control. pRL-CMV-BActin reporter vector was from Promega.
Cell Culture-We used the following human melanoma cell lines with various metastatic potentials in the experimental nude mice model: nonmetastatic (SB2, DX3, and TXM40), moderately metastatic (TXM13, TXM18, and MeWo), and highly metastatic (WM2664 and A375SM); these eight cell lines are described elsewhere (40). The highly metastatic A375SM human melanoma cell line, which was used for most of our experiments, was established from pooled lung metastases produced by parental A375 cells injected intravenously into nude mice (41). All cells were routinely cultured in minimum essential medium supplemented with 10% fetal bovine serum, 1% nonessential amino acids, 1% HEPES buffer, and 1% penicillin and streptomycin. Cells were grown at 37°C in a 5% CO 2 atmosphere.
Generation of KCREB A375SM Clones-Generation of stable A375SM cell clones expressing the dominant-negative form of CREB, KCREB, was performed as described previously for MeWo cells (11,13,42). Briefly, 1 ϫ 10 5 cells were transfected with 1 g of the pRSV-KCREB or pRc/RSV construct by using 2 l of Lipofectamine 2000 reagent (Invitrogen). Six hours later, the transfection medium was replaced with serum-containing growth medium. To select for stable transfectants, growth medium was supplemented with 2 mg/ml G418 (Invitrogen). Expression of KCREB was verified by dual luciferase assay using pGL3-Luc construct driven by a somatostatin gene promoter containing four CREs.
Transient Transfections and Luciferase Activity Assays-Transient transfections were performed using Lipofectamine 2000. A total of 25 ϫ 10 3 cells/well in a 24-well plate were transfected with 0.5 g of the basic pGL3 expression vector with no promoter or enhancer sequence (pGL3-Basic) or with 0.5 g of the pGL3/CRE-Luc or pGL3/MMP2-Luc expression construct. After 6 h, the transfection medium was replaced with serum-containing growth medium. PAF was added 24 h later, and after the relevant incubation period, the cells were harvested and lysed, and luciferase activity was assayed using a dual luciferase reporter assay system (Promega, Madison, WI) as instructed by the manufacturer. For each transfection, 30 ng of Renilla luciferase reporter pRL-CMV-BActin (Promega) was included to normalize for differences in transfection efficiency.
Zymography-Zymographic assays were performed as described previously (11,39). Briefly, after cells were incubated with PAF for 16 h, the medium was replaced with serum-free minimum essential medium. The conditioned medium was collected 16 h later and separated by electrophoresis in a polyacrylamide gel containing 1 mg/ml gelatin. The volume of supernatants loaded on the gel was normalized to the cell number. The gel was then washed at room temperature for 2 h with 2.5% Triton X-100 and, subsequently, at 37°C overnight in a buffer containing 10 mM CaCl 2 , 150 mM NaCl, and 50 mM Tris-HCl, pH 7.5. The gel was stained with 0.5% Coomassie Blue and photographed on a light box. Proteolysis was detected as a white zone in a dark blue field.
MMP-2 Activity Assay-Cells were treated as described for the zymography experiments except that the medium without phenol red was used in the last step of collecting the conditioned supernatants, which were analyzed for their ability to cleave a fluorogenic peptide substrate (7-methoxycoumarin-4-yl-acetyl-Pro-Leu-Gly-Leu-3-(2,4dinitrophenyl-L-2,3-diaminopropionyl)-Ala-Arg-NH 2 ) as described by the manufacturer (R & D Systems, Minneapolis, MN). Briefly, the volume of supernatants was normalized to the cell number, and 100-l aliquots were added to 10 g of a substrate diluted in TCNB reaction buffer (50 mM Tris, pH 7.5, 150 mM CaCl 2 , 10 mM NaCl, 0.05% Brij). The fluorescence of the cleaved substrate was measured 16 h later (excitation wavelength 320 nm and emission wavelength 405 nm). A specific concentration of active MMP-2 was recalculated based on a standard curve obtained with recombinant human MMP-2 activated by 1 M p-aminophenylmercuric acetate for 1 h at 37°C in TCNB buffer.
Western Blotting-Proteins of total cell extract (typically 15 g) were separated by 10% SDS-PAGE and transferred to Immobilon P transfer membrane (Millipore, Bedford, MA). The membranes were washed in Tris-buffered saline with Tween (10 mM Tris-HCl, pH 8, 150 mM NaCl, and 0.05% Tween 20) and blocked with 5% nonfat milk in Tris-buffered saline with Tween for 2 h at room temperature. The blots were then probed overnight with relevant antibodies at dilution of 1:1000 except for the anti-CREB antibody (1:2000) and anti-phospho-p38 MAPK antibody (1:500). After 2 h of incubation with horseradish peroxide-conjugated secondary antibody, immunoreactive proteins were detected by enhanced chemiluminescence per the manufacturer's instructions (ECL detection system; Amersham Biosciences).
Electrophoretic Mobility Shift Assay (EMSA)-The EMSA probe consisted of annealed synthetic complementary oligonucleotides containing the cAMP-binding consensus site (TGACGCTA) and was obtained from Promega. 32 P-End-labeled oligonucleotides (20,000 cpm) were incubated for 30 min at 30°C with 1 g of nuclear extract in 20 l of binding buffer containing 25 mM HEPES, pH 7.9, 0.5 mM EDTA, 0.5 mM dithiothreitol, 4% glycerol, 50 mM NaCl, and 0.5 g of poly(dI-dC). Competition reactions were performed with a 100-fold molar excess of unlabeled double-stranded CRE competitor DNA (Promega). For supershift analysis, the nuclear extracts were incubated with 1 g of anti-CREB antibody for 1 h prior to the binding reaction with labeled probe. The DNA-protein complexes were separated on 4% native polyacrylamide gel in 0.5ϫ Tris borate-EDTA buffer.

Detection of Apoptosis by Propidium Iodide (PI) Staining and Flow
Cytometry-Cells at 90% confluency were exposed to the PAFR antagonist PCA4248, collected at various time points, washed once with cold phosphate-buffered saline, and fixed in cold 70% ethanol overnight. Cells were then rehydrated using PBS, resuspended in PBS solution containing 50 g/ml PI (Sigma) and 20 g/ml RNase A, and incubated at 37°C for 20 min. Cell cycle analysis was performed on a flow cytometer (Epics XL-MCL; Beckman Coulter, Brea, CA) using MultiCycle software (Phoenix Flow Systems, San Diego, CA).
Confocal Microscopy-Cells were plated onto chamber slides and allowed to attach. After three brief PBS washes, cells were fixed with 4% paraformaldehyde/PBS, permeabilized with 0.1% Triton X-100/PBS, blocked in 5% normal horse serum, 1% normal goat serum/PBS, exposed to an anti-PAFR antibody (1:100) in blocking solution overnight at 4°C and then exposed to a secondary goat anti-rabbit-Cy5 antibody (1:1200) for 1 h at room temperature, washed with PBS, exposed to 10 nmol/liter Sytox green for 10 min to stain nuclei, washed, and covered with propylgallate and coverslips for microscopic evaluation. Confocal microscopy was done with a Zeiss LSM 510 confocal microscope and LSM 510 Image Brower software (Carl Zeiss).
MTT Assays-A375SM cells were plated onto 96-well plates (1 ϫ 10 4 cells/well). Complete medium was supplemented with cPAF or PCA4248 at different concentrations. For the MTT analysis, cells were incubated with 1 mg/ml methyltetrazolium bromide for 3 h, after which the medium was replaced with dimethyl sulfoxide (Me 2 SO), and resus-FIGURE 1. PAFR is expressed in human melanoma cell lines. PAFR expression in whole-cell lysates in nonmetastatic (SB2, DX3, and TXM40), moderate-metastatic (TXM13, TXM18, and MeWo), and highly metastatic (WM2664 and A375SM) human melanoma cells was determined by Western blot analysis. The membrane was stripped and re-probed with anti-␣-actin antibody as a loading control. pended. Absorbance was read at 570 nm using a Ceres UV900C microplate reader (Bio-Tek Instruments, Inc., Winooski, VT).
Animals and Experimental Lung Metastasis Assays-Male athymic BALB/c nude mice were purchased from the NCI-Frederick Cancer Research Facility (Frederick, MD). The mice were housed in laminar flow cabinets under specific pathogen-free conditions and used at 8 weeks of age. Animals were maintained in facilities approved by the American Association for Accreditation of Laboratory Animal Care and in accordance with current regulations and standards of the Department of Agriculture, Department of Health and Human Services, and the National Institutes of Health.
For cell injection, A375SM cells in the exponential growth phase were harvested by brief exposure to 0.25% trypsin, 0.2% EDTA solution (w/v). Cell viability was determined by trypan blue exclusion, and only singlecell suspensions of more than 90% viability were used. Cells were washed and resuspended in Ca 2ϩ -and Mg 2ϩ -free Hanks' balanced salt solution (HBSS). A total of 1 ϫ 10 6 cells in 0.2 ml of HBSS was injected into the lateral tail vein of nude mice. 35 days later, the animals were killed, and the lungs were removed, washed with water, and fixed in Bouin's solution for 24 h to facilitate counting of surface tumor nodules under a dissecting microscope, as described previously (11). For intraperitoneal PCA4248 injections, the stock solution of the 50 mg/ml drug was prepared in Me 2 SO and further diluted in HBSS to a final concentration of 1 mg per 0.2-ml injection volume.
Statistical Analysis-In vitro results were evaluated using Student's t test (two-tailed). p values of Ͻ0.05 were considered statistically significant. In vivo metastasis results were evaluated using the Mann-Whitney U test.

Expression of PAFR in Human Melanoma
Cells-Eight human melanoma cell lines that vary in their capacity to form experimental lung metastasis in nude mice were examined for PAFR expression. Western blot analysis of whole-cell extracts revealed two bands recognized by the anti-PAFR antibody, a weak band of ϳ39 kDa and a strong band of ϳ50 kDa (Fig. 1). The predicted molecular mass of PAFR protein is 39 kDa, but the glycosylated form of PAFR has been observed as a 48-kDa band on a Western blot (43,44). Most likely, our analysis revealed the predominant expression of glycosylated PAFR in cultured human melanoma cell lines. All eight melanoma cell lines expressed the PAFR. How- . Viability is expressed as OD Ϯ S.D. of three independent experiments. B and C, analyses of apoptosis in A375SM cells exposed to PCA4248. Cells were exposed to 5, 50, or 100 M PCA4248 for 24 or 48 h, collected, fixed, stained with PI, and analyzed by flow cytometry. B, summary of at least three independent fluorescence-activated cell sorting analyses of PI staining shows that PCA4248 induces apoptosis only at 100 M. Values are means Ϯ S.D. C, representative flow cytometry analysis of A375SM cells exposed to100 M PCA4248 demonstrates an increased proportion of cells with hypodiploid DNA content at 24 h and a further increase at 48 h. ever, the level of PAFR expression did not directly correlate with the experimental metastatic potential of melanoma cells.
PAFRs are expressed as a cytoplasmic membrane-bound protein in platelets, immune cells, and tissue cells. In addition, functional PAFRs have been detected and analyzed on the nuclear membrane of human endothelial cells (44). To study the localization of the PAFR in human melanoma cells, we analyzed its expression by immunofluorescent staining. Visualizing PAFR in A375SM cells by confocal microscopy revealed abundant nuclear and cytoplasmic localization (Fig. 2), which could be attributed to the nuclear and cytoplasmic membrane distribution of PAFR in human melanoma cells. (45)(46)(47)(48)(49)(50). To examine whether modulation of PAFR activity affects melanoma cell growth, the highly metastatic A375SM human melanoma cells were incubated with the stable and metabolically nonhydrolyzable PAF analogue, cPAF, or with the synthetic PAFR antagonist PCA4248 at various concentrations. Up until day 5, 10 or 100 nM cPAF did not affect cell growth, as measured by MTT assay; but after day 5, when the cells became confluent, cPAF promoted cell growth in a concentration-dependent manner (Fig. 3A). In contrast, PCA4248 inhibited cell growth after only 1 day of exposure. Throughout the 9-day experiment, 0.5 or 5 M PCA4248 inhibited cell growth by ϳ15 and 20%, respectively. These concentrations were compatible with the known IC 50 of PCA4248, which is 1 M for PAFR binding in neutrophils and 3.3 M for inhibition of platelet aggregation (51). In our experiment, 50 M PCA4248 completely inhibited A375SM melanoma cell growth (Fig. 3A). Similar effects were observed in SB2 and MeWo melanoma cells exposed to cPAF and PCA4248 (data not shown). The structurally different from PCA4248 PAFR antagonist, dioxolane, produced a similar but weaker inhibitory effect on A375SM cell growth (data not shown).

Effect of PAFR Agonists and Antagonists on the in Vitro Growth of Human Melanoma Cells-Several reports have suggested that exogenous PAFR agonists and antagonists can modulate cell growth in vitro
The inhibition of cell growth by the PAFR antagonist PCA4248 was not associated with induction of apoptotic cell death, as tested by PI incorporation and the analysis of cell cycle distribution by fluorescenceactivated cell sorting. After 24 or 48 h of incubation with 5 or 50 M PCA4248, the percentage of cells carrying hypodiploid amounts of DNA did not change significantly (Fig. 3B). However, 100 M PCA4248 increased the percentage of cells in the sub-G 1 phase from 10% after 24 h of exposure to 25% after 48 h of exposure (Fig. 3B). The corresponding cell cycle profiles for 100 M PCA4248 at 0, 24, and 48 h are shown in Fig. 3C. To avoid cell toxicity, we used 50 M PCA4248 in all subsequent experiments.
PAF Induces Phosphorylation of CREB via PAFR-mediated Stimulation of Ptx-insensitive G␣ q Protein, AC, p38 MAPK, and PKA-CREB expression is greatly increased in metastatic melanoma cells (4,9). The agonists of G protein-coupled receptors may increase intracellular cAMP levels, thereby stimulating CREB phosphorylation by cAMP-sensitive protein kinases. Because we found that melanoma cells expressed functional PAFR, we hypothesized that exogenous PAF may stimulate CREB phosphorylation. Indeed, constitutive levels of CREB phosphorylation at serine 133 in A375SM cells were very low. However, the addition of PAF or cPAF induced CREB phosphorylation (Fig. 4A). This effect was detectable with 0.1 nM PAF or cPAF and increased as the PAF concentration increased. Similarly, PAF and cPAF promoted the phosphorylation of ATF-1 (which can be detected by the same anti-phospho-CREB antibody), although ATF-1 phosphorylation was detectable at time 0. The total level of CREB expression did not change with PAF or cPAF administration. PAF-induced CREB phosphorylation was rapid and was detected already at 10 min of exposure (Fig. 4B). On longer incubation with cPAF, both CREB and ATF-1 phosphorylation subsided but remained at levels higher than constitutive.
cPAF induced CREB and ATF-1 phosphorylation in MeWo cells, which express CREB at levels comparable with those in A375SM cells but have higher constitutive levels of CREB and ATF-1 phosphorylation (data not shown). In contrast, neither PAF nor cPAF induced CREB phosphorylation in nonmetastatic SB2 cells, which do not express any detectable CREB (data not shown).
To analyze further the mechanism of PAF-induced phosphorylation of CREB and ATF-1, A375SM cells were preincubated with PAFR antagonist PCA4248 or dioxolane, selective inhibitor of Ptx-insensitive G␣ q protein GPA-2A, or AC inhibitor SQ24487 for 60 min and then stimulated for 10 min with 10 nM cPAF. All of these compounds inhibited cPAF-induced CREB and ATF-1 phosphorylation without affecting total CREB expression (Fig. 5A). These results indicate that both CREB and ATF-1 phosphorylation are indeed PAFR-mediated and that they occur via a mechanism involving stimulation of the receptor-coupled Ptx-insensitive G␣ q protein and of AC.
Although we did not measure the levels of intracellular cAMP after PAF exposure, we further investigated the possible involvement of cAMP-sensitive PKA and p38 MAPK in CREB and ATF-1 phosphorylation. The inhibitor of p38 MAPK, SB202190, reduced cPAF-induced phosphorylation of CREB by 50 or 58% at 10 or 20 M concentration, respectively, as measured by densitometric analysis (Fig. 5B). Similarly, a 54% inhibition was observed with PKA inhibitor H-89 used at 10 M concentration. Twenty M H-89 inhibited CREB phosphorylation by 108%. These data suggested that the PAF-PAFR signaling utilizes p38 MAPK and PKA for CREB and ATF-1 phosphorylation. Indeed, Fig. 5C shows that PAF caused a sustained phosphorylation of p38 MAPK in A375SM. The PAF-induced phosphorylation of p38 MAPK was inhibited by 70% in the presence of PAFR antagonist PCA4248 (50 M) (Fig. 5D).  FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 cPAF Stimulates Transcriptional Activity of CREB-The ability of cPAF to stimulate the transcriptional activity of CREB was further examined by dual luciferase reporter assay. A375SM cells were co-transfected with pGL3/CRE-Luc and pRL-CMV-BActin vectors. Twenty four hours after transfection, the cells were incubated with cPAF for 6 h. Cell lysates were collected, and firefly luciferase activity was measured and normalized by Renilla luciferase-associated chemiluminescence. cPAF significantly increased CRE-driven promoter activity. The degree of stimulation increased from 1.48-fold with 0.1 nM cPAF to 2.5-fold with 100 nM cPAF (p Ͻ 0.05 versus control) (Fig. 6A). As a positive control, cells were incubated for 6 h with 25 M forskolin, a known potent inducer of PKA phosphorylation, which caused an 8.2fold increase in CREB-dependent transcription. These data indicated that cPAF stimulates CREB-dependent transcription.

Role of PAF in Melanoma Metastasis
Phosphorylation of CREB at serine 133 is thought to stimulate its transcriptional activity by increasing the interaction between CREB and p300/CBP without modifying CREB binding to CRE consensus DNA (1,3). To determine whether PAF-induced CREB phosphorylation in A375SM cells affected CREB DNA binding, we performed an EMSA. Cells were incubated with 10 or 100 nM cPAF for 0.5 or 6 h, and the nuclear proteins were extracted and analyzed for their ability to bind to CRE DNA. Two major shifted bands were observed with nuclear extracts of control A375SM cells (Fig. 6B, lane 2). Both bands were competed out by the excess of an unlabeled CRE oligonucleotide (Fig.  6B, lane 7) and supershifted when reacted with an antibody against total CREB (Fig. 6B, lane 8). Because the same anti-CREB antibody was used in the Western blot analysis described above, where it was found not to react with 38-kDa ATF-1 protein, the complete supershift of both bands on the EMSA gel therefore indicated that at least one CREB molecule was binding to the CRE DNA motif in our assay. Furthermore, nuclear extracts from the cells exposed to cPAF gave rise to the shifted bands of the same pattern without a significant change in band intensity (Fig. 6B,  lanes 3-6). Densitometric measurements from three independent experiments indicated a 1.12-fold more intense upper band in cells stimulated with 10 nM cPAF for 6 h (p Ͼ 0.1 versus control) and a 1.1-fold increase (p Ͼ 0.1 versus control) in cells stimulated with 100 nM cPAF for 6 h, which was found to be nonsignificant. Overall, these results indicate that cPAF stimulates CREB phosphorylation and activation of CREB transcriptional activity without affecting CREB binding to the consensus CRE DNA. Exposure of A375SM cells to PAFR antagonist PCA4248 did not inhibit the ability of CREB to bind CRE DNA, as indicated by nonsignificant changes in the intensity of shifted complexes on EMSA gel produced by nuclear extracts from control and PCA4248-exposed cells (Fig. 6C). This result paralleled our finding that constitutive CREB phosphorylation was very low in cultured A375SM cells and did not change in the presence of PAFR antagonist.
cPAF Stimulates MMP-2 Expression and Activity-Expression of numerous genes can be activated by stimulated CREB. We showed previously that overexpression of CREB and ATF-1 transcription factors contributes to the expression and activity of MMP-2 in MeWo metastatic melanoma cells (11). We sought to examine whether PAF affects MMP-2 expression and activity in A375SM cells. First, gelatin zymography was performed using supernatants from A375SM cells incubated for 16 h with various concentrations of cPAF. As we observed previously with MeWo melanoma cells, A375SM cells expressed high amounts of pro-MMP-2 (72-kDa band), but the 92-kDa band corresponding to MMP-9 was practically undetectable, which is typical for melanoma cells (Fig. 7A). cPAF stimulated expression of pro-MMP-2 gelatinase in a concentration-dependent manner; 1.8-, 4.5-, 4.0-, and 2.6-fold increases in pro-MMP-2 (determined by densitometry) were observed with 1, 10, 100, and 1000 nM cPAF, respectively (Fig. 7A). A 62-kDa band corresponding to active MMP-2 was not detected by zymography assay. In addition, cPAF stimulated a release of pro-MMP-9 (a 92-kDa band). When measured by zymography, cPAF also stimulated pro-MMP-2 release in moderately metastatic MeWo cells (data not shown).
To determine whether the effect of cPAF was mediated through its stimulation of PAFRs, A375SM cells were incubated with the PAFR antagonist PCA4248 (50 M) for 1 h and then stimulated with cPAF with or without PCA4248 (50 M). Our finding that PCA4248 completely abolished cPAF-induced pro-MMP-2 release confirmed the receptor-mediated mechanism (Fig. 7B). Furthermore, cPAF-induced accumulation of pro-MMP-2 was prevented by the p38 MAPK inhibitor SB202190 used in 1 M concentrations (Fig. 7C). Twenty M SB202190 inhibited the induction of pro-MMP-2 below constitutive levels. Similarly, a 100% inhibition of cPAF-induced release of pro-MMP-2 was observed with PKA inhibitor H-89 (1 M). Therefore, our results suggested the involvement of p38 MAPK and PKA in PAF-induced pro-MMP-2 accumulation. Low levels of MMP-9 did not allow us to monitor its modulation.
Because active MMP-2 band was not detected by zymography, we performed a functional assay whereby the cell supernatants were analyzed for their gelatinolytic activity using a fluorogenic substrate specific for MMP-2. Treatment with cPAF increased the levels of released active gelatinase in a concentration-dependent manner (Fig. 7D); 1.06-, 1.4-, 1.44-, 1.54-, and 2.1-fold increases in MMP-2 activity were observed with 0.1, 1, 10, 100, and 1000 nM cPAF, respectively (Fig. 7D). Because the peptide used in this assay can also be cleaved by an activated MMP-9, the observed increased in gelatinolytic activity could reflect activation of both MMP-2 and MMP-9 gelatinases by cPAF. Similar to PAF-induced accumulation of pro-MMP-2 monitored by zymography, an increase in gelatinase activity was completely prevented by 1-10 M SB202190 (Fig. 7E). The inhibitor of PKA H-89 (1-10 M) showed only a partial inhibition of cPAF-induced MMP-2 activation (by 40%), which was not statistically significant. These results indicate that PAF stimulates an increase in active MMP-2, and possibly MMP-9, and that this activation depends on the activity of p38 MAPK and to a lesser degree of PKA.
cPAF Induces Expression of Pro-and Activated MMP-2 and Its Activator MT1-MMP-The activity of MMPs can be modulated at the level of gene transcription and protein synthesis and involves stoichiometric changes between the expression of pro-MMPs, tissue inhibitors of metalloproteinases (TIMPs), and MMP activators such as MT-MMPs, which interact dynamically on the cell surface (52,53). In our study, Western blot analysis of whole-cell extracts showed that 100 nM cPAF stimulated the expression of 72-kDa pro-MMP-2, the 64-kDa activated form of MMP-2, and the expression of MMP-2 activator MT1-MMP in a time-dependent manner (Fig. 8A). Induction of pro-MMP-2 was already evident at 6 h and increased through 48 h. In parallel, a 64-kDa band, corresponding to the activated membrane-associated form of the cleaved MMP-2 peptide, became detectable at 6 h, and its intensity increased in a time-dependent manner. Furthermore, a strong progressive induction of pro-MT1-MMP and active cleaved MT1-MMP expression was detected starting at 12 h. These results suggested that PAF-induced activation of MMP-2 occurs because of induction of protein expression of pro-MMP-2 and because of further proteolytic activation of MMP-2 by the simultaneously overexpressed MT1-MMP.
The inhibition of cPAF-induced MMP-2 activation by p38 MAPK and PKA strongly suggested the involvement of CREB and ATF-1 in PAF-induced MMP-2 activation. The MMP-2 promoter contains a CRE at position Ϫ374 to Ϫ366 with respect to the transcription initiation site  FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5  (54) and seven other possible cAMP-like motifs. 3 To examine the effect of cPAF on MMP-2 transcription, we monitored the activity of luciferase reporter driven by the MMP-2 promoter. cPAF transactivated the MMP-2 promoter in a concentration-and time-dependent manner; at 6 h, 1 nM cPAF induced a 3-fold increase in MMP-2 promoter activity, and 100 nM cPAF induced a 3.7-fold increase (Fig. 8B). Activation with 100 nM cPAF was significant after 4 h of cell incubation (Fig. 8C). Together with the results from Western blot analysis, these results indicated that PAF induces MMP-2 expression at the transcriptional level and that transactivation may occur at least in part through the MMP-2 promoter sequence containing a motif (or motifs) capable of binding CREB.

Role of PAF in Melanoma Metastasis
cPAF-induced Expression of Pro-MMP-2 Depends on the Presence of Functional CREB-Our data demonstrating that cPAF-induced activation of MMP-2 is antagonized by the inhibitors of p38 MAPK and PKA, which are involved in PAF-induced CREB phosphorylation, and the capacity of PAF to induce transcription from the MMP-2 promoter containing CRE strongly implicated CREB in PAF-induced MMP-2 activation. To determine whether CREB is necessary and sufficient for induction of pro-MMP-2 by PAF, we generated stable transfectants of A375SM cells expressing the dominant-negative form of CREB, KCREB. We showed previously that MeWo melanoma cells expressing KCREB have less MMP-2 activity and message than parental MeWo cells do and that this effect correlates with the decreased capacity of nuclear extracts from transfected cells to bind CRE in an EMSA and to stimulate CRE-driven promoter (11). In agreement with that study, A375SM cell clones expressing KCREB had a 3.5 times lower capacity to stimulate transcription from the CRE-driven promoter than did parental A375SM cells, as monitored by the dual luciferase reporter assay (Fig.  9A). In parallel, pro-MMP2 expression was reduced in KCREB cell clones (Fig. 9B), which suggested that CREB transactivates MMP-2 expression constitutively in A375SM cells. Addition of cPAF to A375SM parental cells or to A375SM cells transfected with an expression vector lacking the KCREB gene (empty vector; A375SMpsv) stimulated CRE-dependent transcription (Fig. 9C). However, cPAF was unable to transactivate CRE-driven promoter in KCREB-transfected cells (Fig. 9C). Finally, gelatin zymography showed that cPAF stimulated expression of pro-MMP-2 in A375SM parental and A375SMpsv cells but not in the KCREB cells (Fig. 9D). These results indicated that PAFinduced induction of pro-MMP-2 involves a mechanism mediated by CREB-associated transcriptional activation.

PCA4248 Inhibits Formation of Melanoma Lung Metastasis in Nude
Mice-We demonstrated previously that the presence of active CREB contributes to the metastatic growth of human melanoma cells (11).
The results from our current study on the capacity of PAF to activate CREB and MMP-2 in vitro suggested that PAF may promote the growth of melanoma metastasis. To investigate this possibility, we analyzed the effect of the PAFR antagonist PCA4248 in an experimental lung metastasis assay. To this end, 1 ϫ 10 6 A375SM cells were injected into the lateral tail vein of male BALB/c nude mice, and 35 days later, the number of lung metastases was counted. Two groups of mice were treated with PCA4248. In the first group, 20 mg/kg PCA4248 was injected intravenously 10 min before cell injection. In the second group, 40 mg/kg PCA4248 was injected intraperitoneally daily starting at day 7 after cell injection, when melanoma cells have presumably established microscopic lung metastasis. We found that with both regimens, PCA4248 greatly reduced the number and incidence of melanoma lung metastasis: 9 of 9 mice in the control group developed metastasis (median, 12), 2 of 5 mice receiving single intravenous injection of PCA4248 prior to cell injection developed metastasis (median, 1), and 6 of 9 animals receiving daily intraperitoneal injections of PCA4248 developed metastasis (median, 1) ( Table 1). Taken together, these results suggested that PAFR antagonist inhibited the formation of experimental lung metastasis of melanoma cells in vivo.

DISCUSSION
The present work demonstrates that metastatic human melanoma cells that overexpress CREB and ATF-1 are better equipped to respond to inflammatory stimuli from the tumor microenvironment. We show that cultured human melanoma cells express PAFR, albeit its levels do not correlate with their metastatic potential. It is possible that the signaling through PAFR can be modulated through PAF synthesis in tumor cells. First, metastatic melanoma cells overexpress receptor to thrombin, which is a known inducer of PAF production (17). Second, our preliminary data indicate that metastatic melanoma cells have higher levels of active phosphorylated phospholipase A 2 enzyme. Furthermore, in the present study we demonstrate an intense nuclear PAFR staining. This suggests that in tumor cells, both extracellular and intracellular PAF may initiate the receptor-mediated signal transduction, which is similar to the functions of PAF in endothelial cells (44). Although an examination of the effect of PAF on cell growth was not a major objective of our study, we showed that cPAF had a weak stimulatory effect on melanoma cell growth on prolonged incubation in a serum-containing medium. This observation agrees well with other reports that have shown that in nanomolar concentrations PAF produces a weak stimulatory effect on the proliferation of human T-cells, human fibroblasts, human endometrial cancer cells, Kaposi sarcoma cells, normal rat fibroblasts, and rat primary embryo cells (45)(46)(47)(48)(49)(50). Similar to our findings in melanoma, in rat primary embryo cells, PAF promoted the ability of cells to reach a higher saturation density in a serum-containing medium (46). On the other hand, exposure to the PAFR antagonist PCA4248 rapidly inhibited cell growth. Our unpublished observations suggested that inhibition of PAFR signaling by several structurally independent antagonists down-regulated the cellular expression of cyclin D1, although the addition of PAF alone did not affect cyclin D1 expression directly. The ability of PAF to control cyclin D1 activity at post-translational levels is currently under investigation in our laboratory.
The principal finding of our work is that in melanoma cells PAF activates CREB and ATF-1 transcription factors by inducing their phosphorylation. CREB and ATF-1 phosphorylation was detected with as low as 0.1 nM PAF or cPAF. cPAF stimulated transcription from CRE-driven reporter gene; 100 nM cPAF induced a 2.5-fold increase over the basal level of reporter activity. Similar to cAMP agonists, PAF-induced phosphorylation of CREB did not stimulate its DNA binding activity. Phosphorylation-induced activation of CRE-driven transcription may have resulted from an increase in CREB binding to CBP/p300 histone acetylase, as has been reported previously with cAMP analogues (2, 3). Our further inhibitory experiments showed that the mechanism of CREB and ATF-1 phosphorylation involved PAFR-mediated activation of Ptx-insensitive G␣ q protein, AC, p38 MAPK, and PKA. In addition, we found that cPAF induced a sustained phosphorylation of p38 MAPK via a PAFR-dependent mechanism.
cAMP has been shown to induce PKA-dependent and -independent phosphorylation of CREB through a mechanism involving p38 MAPK (55,56). In turn, PAFRs have been shown to couple to the ␣ subunits of Ptx-sensitive G o or G i proteins, which inhibit AC, or to Ptx-insensitive G␣ q protein, which can activate AC. This coupling appears to be cellspecific. The PAFRs expressed in COS-7 monkey kidney fibroblasts were found to interact with Ptx-insensitive G␣ q and G␣ 11 proteins (57). In human umbilical vein endothelial cells, PAFRs were found to be coupled to Ptx-insensitive G␣ q protein, and this coupling mediated PAF-induced activation of AC-PKA-Src (25). In human neutrophils, 0.01-100 nM PAF induced tyrosine phosphorylation of p38 MAPK through Ptx-insensitive G␣ q protein-mediated activation of MAPK kinase 3 (58). PAF has been also shown to induce AC and increase cAMP levels in human mesangial cells (59). On the other hand, in Chinese hamster ovarian cells, overexpressed PAFR was shown to couple to the ␣ subunit of the Ptx-sensitive G o protein, inhibit AC, and decrease cAMP levels (60,61). However, negative regulation of AC may be specific to those cells, which were found to overexpress the Ptx-sensitive G o protein (61). In our experiments, PAF induced the phosphorylation of p38 MAPK and CREB/ATF-1, and the CREB phosphorylation was inhibited by the inhibitors of G␣ q protein, AC, p38 MAPK, and PKA. Thus, we propose that in human melanoma cells, PAFRs couple to G␣ q protein and stimulate cAMP production.
We further showed that in human melanoma cells, cPAF induced a release of pro-MMP-2 gelatinase as well as activation of its proteolytic activity. This was detected in cell supernatants by zymography and fluorogenic substrate cleavage, as well as in membrane-bound fractions by Western blot analysis. The release of pro-and active MMP-2 was prevented by the inhibitors of p38 MAPK and PKA. This suggested a common mechanism in PAF-induced CREB phosphorylation and MMP-2 activation. Further experiments confirmed that cPAF induced CREBdependent transactivation of MMP-2. First, a luciferase reporter assay showed that cPAF induced the transactivation of a reporter construct driven by the MMP-2 promoter containing CRE. Second, the cPAFinduced pro-MMP-2 release was prevented in cells overexpressing the dominant-negative form of CREB, KCREB. Altogether, these results strongly suggested that PAF induces MMP-2 expression in metastatic melanoma cells via a PAFR-mediated signaling cascade involving activation of p38 MAPK and PKA, phosphorylation of CREB and ATF-1 transcription factors, and transactivation of MMP-2.
Activation of pro-MMP-2 is known to occur through proteolytic cleavage of the N-terminal pro-peptide, which results in a 64-kDa intermediate, that is further processed to a 62-kDa active form (53) and to require two MT1-MMP molecules. The first MT1-MMP molecule promotes the recruitment to the cell surface of pro-MMP-2 and serves as a surface receptor for the TIMP2-pro-MMP-2 complex (52). The second MT1-MMP molecule induces the proteolytic cleavage of MMP-2 (52). Although we did not measure TIMP2 expression, the PAF-induced activation of MMP-2 that we observed correlated with increased MT1-MMP expression and activation. In addition to proteolytically activating MMP-2, MT1-MMP degrades extracellular matrix components (e.g. collagen types I, II, and III; fibronectin; laminins 1 and 5; and vitronectin) (62). Therefore, our results indicated that exposure of melanoma cells to PAF may promote the degradation of the extracellular matrix and promote tumor cell invasion through concerted activation of MMP-2 and MT1-MMP proteases.
Several reports have suggested that PAF regulates the activity of various MMPs in different cell types, including stimulation of the MMP-9 expression in corneal myofibroblasts (63), COX-2-mediated transactivation of urokinase plasminogen activator and MMP-1 and -9 in corneal cultures (64), and NFB-dependent expression of MMP-9 in endothelial ECV304 cells (65). Most recently, Axelrad et al. (66) showed that PAF induced the transactivation of MT1-MMP and TIMP2 genes and the activation of MMP-2 in human umbilical vein endothelial cells. In these cells, MMP-2 activation occurred through a G␣ q protein-JAK-2-Src-mediated pathway (25,66). Negative regulation of MMPs by PAF has also been reported; Barletta et al. (67) showed that 300 nM PAF down-regulated MT1-MMP and MMP-2 mRNA levels and reduced MMP-2 activation in isolated human neuroblastoma cell clones. The seemingly opposite effects of PAF on neuroblastoma and melanoma cells may be due to variations in PAF sensitivity. In the neuroblastoma cells, 300 nM PAF induced cell differentiation (67), whereas in our study, 100 nM PAF promoted cell growth in melanoma cells. In addition, the CREB and ATF-1 expression levels in different cell types may determine the ability of PAF to activate MMP-2. Our report appears to be the first demonstration of PAF-induced CREB-dependent activation of MMP-2 in tumor cells.
Our results indicated that in a PAF-rich microenvironment (e.g. skin undergoing inflammation when exposed to UV light, tumor sites infiltrated with inflammatory cells, an embolus formed between tumor cells and platelets, or a site of a contact with activated endothelial cells), tumor cells, in particular metastatic ones, might activate specific transcriptional responses associated with overexpressed CREB and ATF-1. Thus, CREB and ATF-1 may act as molecular sensors to PAF-mediated inflammatory stimuli in metastatic melanomas. Our data indicated that one consequence of this stimulation is an increase in the production and activation of MMP-2 and MT1-MMP, which will in turn stimulate tumor cell proteolytic activity, metastatic invasion, and extracellular matrix remodeling; possibly affect proteolytic activation of various growth factors; and stimulate angiogenesis. PAF could also affect the expression of CREB-dependent anti-apoptotic genes (e.g. bcl-2 and IAP-2), as well as COX-2 and MCAM/MUC18. In fact, we showed previously that CREB and ATF-1 function as cell survival factors in vitro and in experimental tumor and metastasis models in vivo (12,13).
Finally, our preliminary in vivo experiments showed that the PAFR antagonist PCA4248 is a potent inhibitor of experimental human melanoma lung metastasis. The inhibitory effect of PCA4248 delivered intravenously before the injection of melanoma cells could be attributed to its action as an inhibitor of platelet aggregation. Indeed, various pharmacologic anti-platelet agents and anticoagulants inhibit the spread of certain tumor metastases (68). However, we found that daily PCA4248 injection, initiated 7 days after the tumor cell injection, also strongly down-regulated melanoma lung metastasis, and this observation suggested that antagonizing PAF-PAFR signaling inhibits the growth of established microscopic tumor cell colonies in the lungs. At present, we cannot exclude the possibility that the inhibitory effect of PCA4248 in vivo was directly related to its ability to inhibit the activation of CREB and MMPs in the tumor cells. To examine this possibility, we are now generating small interfering RNA against human PAFR by using a len-tiviral approach to discriminate between tumor and host cell-associated effects of PAF in vivo. Nevertheless, our results demonstrate that PAF has a strong tumor-promoting effect in melanoma and have prompted us to investigate further the possibility of targeting PAFR signaling as a therapeutic approach against metastatic melanoma, which currently remains incurable.