Topical treatment with inhibitors of the phosphatidylinositol 3′-kinase/Akt and Raf/mitogen-activated protein kinase kinase/extracellular signal-regulated kinase pathways inhibited the growth of TPras transgenic melanomas in severe combined immunodeficient mice, blocked invasive behavior, and reduced angiogenesis. The inhibitor Ly294002, which is specific for phosphatidylinositol 3′-kinase, effectively reduced melanoma cell growth both in vitro and in vivo. Both Ly294002 and U0126, a mitogen-activated protein kinase kinase 1/2 inhibitor, reduced invasion, which correlated with reduction of the metalloproteinase matrix metalloproteinase 2. Tumor angiogenesis was disrupted through inhibition of vascular endothelial growth factor production from the tumor cells and antiangiogenic effects on endothelial cells. Observations with TPras melanoma cells that express dominant negative Δp85 or kinase-inactive Raf301 supported the specificity of the phenomena observed with the chemical inhibitors. These studies demonstrate that topical treatment targeting Ras effectors is efficacious, without systemic toxicities, and may prove to be useful in treating and preventing the progression of cutaneous melanoma.
The incidence of melanoma has increased during the past 50 years at a rate that exceeds any other solid tumor, with substantially increased mortality rates as well (1 , 2) . Although melanoma accounts for only 10% of all skin cancers, it is responsible for 80% of all skin cancer deaths (3) . These statistics alone highlight the need for the development of new effective therapeutic approaches.
It is well known that mutation or deletion of the CDKN2a locus at 9p21 is observed during progression in virtually all melanomas. This locus contains the closely linked INK4aand ARFgenes coding for the p16INK4a and p14ARF (p16INK4a/p19ARF in mice) proteins, regulators of cell cycle and apoptosis, respectively (4, 5, 6, 7) . p16INK4a is considered a melanoma susceptibility gene because it is necessary but not sufficient for melanoma development (8 , 9) . Additional cooperating genetic lesions are required.
Mutated N-Ras and H-Ras have been observed in cutaneous melanomas and most frequently in melanomas that occur in sun-exposed skin (10 , 11) . The frequency of Ras mutations, however, is likely an underestimate of the contribution of aberrant signaling through the Ras pathways because chronic up-regulation of the Ras pathways can occur through multiple mechanisms (12, 13, 14) . Recent analyses have reported that as many as 56% of congenital nevi, 33% of primary melanomas, and 26% of metastatic melanoma samples harbor activating N-Ras mutations (15 , 16) . In addition, in a group of Spitz nevi, amplification of chromosome 11p, where H-Ras resides, has been reported (17) . The finding of Ras mutations in benign lesions suggests that such events may represent an early feature in melanoma development. Genetic data from transgenic mouse models (18, 19, 20) have demonstrated that expression of activated Ras is essential for the genesis and maintenance of melanomas.
Another mutation frequently found in melanoma occurs in the BRaf gene. In a cohort of primary cell lines, Davies et al. (21) found that 66% of melanoma samples presented activating point mutations of the BRafgene. The most common substitution, V599E, renders BRaf constitutively active, activates the mitogen-activated protein kinase pathway, and confers mitogenic responses and growth advantages to these cells (22) . The same activating mutation has been identified in benign melanocytic proliferations, which has led to the speculation that BRaf activation, although not sufficient for melanocyte transformation, may also be an early event in the process (23) .
Finally, it has been shown that alterations of the 10q24–26region are present in 30–50% of melanoma cell lines and 5–20% of uncultured melanomas (24, 25, 26) . This region includes the lipid phosphatase PTEN, which functions by negatively regulating the phosphatidylinositol 3′-kinase (PI3K)/Akt pathway. Loss of PTEN function in melanoma needs further investigation (26) .
Mutations in Ras and BRaf or Ras and PTEN do not appear to occur simultaneously. This is likely due to the fact that Ras activates a multitude of downstream signaling cascades including the Raf/mitogen-activated protein kinase kinase (MEK)/extracellular signal-regulated kinase (Erk) and PI3K/Akt pathways. Because BRaf mutation activates the mitogen-activated protein kinase cascade, and loss of PTEN activates PI3K, these observations suggest that these two pathways are important mediators of melanoma downstream of Ras. The PI3K/Akt and Raf/MEK/Erk pathways promote tumorigenesis essentially by positively regulating cell survival, cell cycle progression (27, 28, 29) , tumor angiogenesis (30, 31, 32) , and tumor invasion (33, 34, 35) .
The apparent importance of these two Ras-activated pathways suggests that they may be effective targets for prevention and treatment. A problem in inhibiting these signaling cascades is that they regulate important growth and survival pathways shared by both neoplastic and normal cells. The challenge therefore is to develop a therapeutic strategy able to exert the maximum effect on cancer cells with the minimum toxicity on normal cells. Primary cutaneous melanomas may represent a tumor system where the location of the tumor allows for nonsystemic topical treatment. In this study, we demonstrate that topical treatment with the PI3K-specific inhibitor Ly294002 (36) results in significant reduction of TPras transgenic melanoma growth in severe combined immunodeficient (SCID) mice. Additionally, inhibition of either the Raf/MEK/Erk pathway by the MEK1/2-specific inhibitor U0126 (37 , 38) or PI3K leads to a reduction in tumor invasion and angiogenesis. These reductions are accompanied by a decrease in expression of matrix metalloproteinase (MMP) 2 and the angiogenic factor vascular endothelial growth factor (VEGF). Together, these findings show that topical treatment with specific chemical inhibitors may be an effective means of treating cutaneous melanoma and confirm that both the Raf/MEK/Erk and PI3K/Akt pathways play a significant role in melanoma growth and progression.
MATERIALS AND METHODS
Chemicals and Plasmids.
DMSO, 4-hydroxytamoxifen (4-HT), and gelatin type A were purchased from Sigma. DMSO was used at a final concentration of 0.01%. The PI3K inhibitor Ly294002 was obtained from Bio Mol Research Laboratory (Plymouth Meeting, PA), and the MEK1/2 inhibitor U0126 was obtained from Cell Signaling Technologies, Inc. (Beverly, MA). For in vitro studies, unless noted, the final concentrations used for Ly294002 and U0126 were 50 and 10 μm, respectively.
Kinase-inactive Raf-1, Raf301, was described previously (39) . The retroviral plasmid carrying Raf301 fused to the human estrogen receptor (Raf301:ER) was generously provided by Dr. M. McMahon (University of California, San Francisco, CA). Raf301:ER was activated by adding 1 μm 4-HT to cells 48 h before experimental set up and maintained in the culture during the experiment. The PI3K dominant negative mutant, Δp85, was inserted in a pBabe retroviral vector (40) . Both plasmids were retrovirally introduced into 1984-1 cells as described previously (41) .
The 1984-1 cell line, derived from a TPras mouse cutaneous melanoma, was described previously (42) . Cells were maintained in RPMI 1640 (Life Technologies, Inc., Carlsbad, CA) with 10% fetal bovine serum (FBS; Atlas Biologicals, Fort Collins, CO) and 1% penicillin-streptomycin (Life Technologies, Inc.). Human umbilical vein endothelial cells (HUVECs) were purchased from American Type Culture Collection (Manassas, VA) and maintained in Kaighn’s F12K media supplemented with 2 mm l-glutamine, 1.5g/liter sodium bicarbonate, 0.1 mg/ml heparin, 0.03 mg/ml endothelial cell growth supplement (Sigma), 10% FBS, and 1% penicillin-streptomycin. Experiments were performed with HUVECs in passages 2–7.
Soft Agar Assay.
1984-1 cells (5000 cells/well in 6-well plates) were suspended in RPMI 1640 + 10% FBS containing 0.3% agar and spread onto a 0.6% agar layer. Inhibitors (10 μm) were added in both agar layers. Raf301 was activated by adding 1 μm 4-HT to the agar layers. DMSO (0.01%) was added for cells transfected with pBabe or Δp85 as a vehicle control. After 2 weeks, colonies were stained with Giemsa (41) and counted, and the results were recorded by scanning the plates (Scan Jet 7400 C). Significant differences in the number of colonies were determined by Student’s t test (P < 0.05).
Western Blot Analysis.
1984-1 cells plated in RPMI 1640 plus 10% FBS were allowed to adhere and then treated with inhibitors for 16 h in low-serum media (0.5% FBS). Cells with genetic inhibitors received DMSO or 4-HT. Total protein was extracted with lysis buffer [9 m urea, 75 mm Tris-HCl (pH 7.5), and 100 mm 2-mercaptoethanol] and quantitated using DC Protein Assay (Bio-Rad, Hercules, CA), and 30 μg/sample were separated on a 12% SDS-PAGE gel and transferred onto polyvinylidene difluoride membrane. Membranes were probed with anti-phospho-p44/42 mitogen-activated protein kinase (Erk1 and Erk2, Thr 202/Tyr204) antibody (1:1000) and anti-phospho-Akt (Ser473) antibody (1:1000; Cell Signaling Technologies, Inc.), and bands were detected using enhanced chemiluminescence Western blotting detection reagents (Amersham Biosciences, Piscataway, NJ). Loading was checked with anti-GAPDH (glyceraldehyde-3-phosphate dehydrogenase) antibody (1:5000; Research Diagnostics, Inc., Flanders, NJ).
Invasive behavior was examined in vitro using BD BioCoat Matrigel Invasion Chambers (BD Biosciences, Bedford, MA) as per the manufacturer’s instructions. 1984-1 cells (1 × 105) were seeded in triplicate on both Matrigel-coated and uncoated inserts. Cells were allowed to adhere and then incubated with inhibitors or 4-HT at 37°C for 30 min. The inserts were moved to chambers containing 750 μl of FBS as a chemoattractant and incubated at 37°C for 22 h. After removal of the noninvading cells, membranes were stained and mounted. Five fields were randomly chosen and counted for each membrane. The percentage of invasion was determined as follows: (average number of cells invading through the Matrigel insert membrane/average number of cells migrating through the control insert membrane) × 100. Results from each group were compared with control, and significance was determined (P < 0.05, Student’s t test).
VEGF ELISA Assay.
Levels of secreted VEGF were analyzed using Quantikine M Murine kit (R&D Systems Minneapolis, MN) as per the manufacturer’s instructions. Briefly, 1984-1 cells (6 × 105) were plated on 6-well plates in RPMI 1640 + 10% FBS in duplicate for each treatment. Chemical inhibitors or 4-HT was added to the appropriate wells. Cells were incubated at 37°C for 24 h. Supernatants from each treatment were used without dilution in the VEGF ELISA assay. Protein level of treated and untreated cell lysates was calculated and used to normalize the levels of secreted VEGF. Statistical analysis to determine significance of the observed differences was performed using the Student’s t test (P < 0.05).
Tube Formation Assay.
The tube formation assay performed as a measure of in vitro angiogenesis has been described previously (43) . Briefly, 200 μl of growth factor-reduced Matrigel matrix (BD Biosciences) were added to 2-cm2 wells and allowed to solidify at 37°C for 30 min. For the experiments with chemical inhibitors, HUVECs (5 × 104) were serum-deprived for 6 h in Kaighn’s F12K media containing 0.1 mg/ml heparin, 0.03 mg/ml endothelial cell growth supplement, and 0.5% FBS and then plated on Matrigel layers. Cells were stimulated with 50 ng/ml VEGF in the presence or absence of inhibitors and incubated for 20 h at 37°C. For the experiments with 1984-1 supernatants, HUVECs (5 × 104) were cultured in RPMI 1640 + 0.5% FBS for 6 h and then plated as described above. RPMI 1640 with low serum was used as the unconditioned media for this experiment. Supernatants from 1984-1 cells overexpressing the empty vector (pBabe), Raf301, or Δp85 were collected, added to HUVECs, and incubated for 20 h at 37°C. To analyze tube formation, three random fields from each well were photographed at ×100 magnification (Nikon SF 35 WA camera). Tube networks were quantified as the total number of pixels in thresholded images using Scion Image Software (Scion Corp., Frederick, MD). Values were analyzed for significance (P < 0.05, Student’s t test).
Northern Blot Analysis.
Total RNA from 1984-1 cells treated for 16 h with inhibitors or 1984-1 melanoma cells expressing Raf301 or Δp85 was harvested using TRIzol Reagent (Life Technologies, Inc.) as per the manufacturer’s instructions. RNA samples were resolved on a 1% formaldehyde gel and transferred to Hybond-N+ membrane (Amersham Biosciences). Blots were hybridized with α-32P-labeled human VEGF cDNA and imaged with a Storm phosphor-screen (Amersham Biosciences). Intensity of the bands was determined by densitometry. Density values were corrected for loading (18S rRNA) and normalized with respect to the control (arbitrary value = 1).
Levels of membrane type 1 (MT1)-MMP, MMP2, and GAPDH were evaluated by reverse transcription-PCR using the SuperScript One-Step kit from Invitrogen. To verify the linear amplification of the products, 10 μl of sample from each reaction were collected every three cycles starting from the cycle indicated in Fig. 3, B and D ⇓ , and then loaded on a 1% agarose gel for visualization. Primer sequences were as follows: mouse MMP2 forward primer, 5′-GAGTTGGCAGTGCAATACCT-3′; mouse MMP2 reverse primer, 5′-GCCGTCCTTCTCAAAGTTGT-3′; mouse MT1-MMP forward primer, 5′-GTGCCCTATGCCTACATCCG-3′; mouse MT1-MMP reverse primer, 5′-TTGGGTATCCGTCCATCACT-3′; mouse GAPDH forward primer, 5′-TGCTGAGTATGTCGTGGAGTCTA-3′; and mouse GAPDH reverse primer, 5′-AGTGGGAGTTGCTGTTGAAGTCG-3′.
Conditioned serum-free culture supernatants from 1984-1 melanoma cells treated overnight with the chemical inhibitors or from cells expressing Raf301 or Δp85 were collected and concentrated using a microtube device with a cutoff of 30K (Pall Life Science, Ann Arbor, MI). Protein concentration was determined using the DC Protein Assay (Bio-Rad), and 10 μg of total protein were resolved on an 8% (w/v) standard Laemmli SDS-polyacrylamide gel containing 2 mg/ml gelatin as a substrate. After electrophoresis, gels were washed once in 2.5% (v/v) Triton X-100 to remove SDS and then washed in 50 mm Tris-HCl, 5 mm CaCl2, and 0.1% Triton X-100 (pH 7.8) and incubated overnight at 37°C in the same buffer with gentle agitation. Zymograms were stained for 45 min with 0.25% (w/v) Coomassie Brilliant Blue R250 dissolved in 40% methanol and 10% glacial acetic acid and destained in the same solution without Coomassie Blue.
In Vivo Experiments.
Male 3–5-week-old SCID (B6.CB17) mice supplied by Stanford University Animal Facility were housed in the same facility (American Association of Laboratory Animal Care-approved) with 12-h light cycles. Food and water were provided ad libitum. The dorsal flanks of the mice (6 mice/treatment group) were shaved and the treatments [acetone (vehicle) or 3 μg of Ly294002 in 200 μl of acetone, 100 μg of U0126 in 200 μl of acetone, or a combination of the two] were topically applied three times a week before the injection of cells. 1984-1 TPras melanoma cells (experiment 1, 2 × 106; experiment 2, 1 × 106) were injected s.c. into the flank of each mouse. Treatment with chemical inhibitors began after 2 days and continued for the duration of the experiment. Mice were sacrificed at 26 (experiment 1) or 40 days (experiment 2) postinjection. 3,3′-diheptyloxacarbocyanine iodide (DiOC7) (1.0 mg/kg; Molecular Probes, Inc., Eugene, OR) in 75% DMSO in PBS was injected into the tail vein of each mouse, and the mice were sacrificed after 1 min (44) . Gross pathology of each tumor was recorded. Tumors were measured, and tumor volume was calculated as described previously (45) . Sections from each tumor were frozen in OCT or fixed in 10% buffered neutral formalin. Frozen samples were sectioned at 7 μm; formalin-fixed sections were embedded in paraffin and sectioned at 5 μm. All experiments were performed three times.
Immunohistochemistry and Enumeration of Microvessels.
Blood vessels were stained using rat antimouse CD31 (platelet/endothelial cell adhesion molecule 1; 1:50; BD PharMingen, Bedford, MA), whereas vessel functionality was demonstrated via i.v. injection of the DiOC7 fluorescent dye (FITC filter). Sections from all treatment groups were analyzed for microvessel density by microscopic counting of four fields at ×200 magnification. Total number of microvessels and number of open/functional microvessels were determined for each treatment group, and average number per field is presented. Significance of the differences was analyzed (Student’s t test). Additional antibodies used included rabbit anti-VEGF (147; 1:50; Santa Cruz Biotechnology, Santa Cruz, CA), rat antimouse KI-67 (1:50; Dako Corp., Santa Barbara, CA), and rabbit anti-MMP2 (H-76; 1:200; Santa Cruz Biotechnology). Tumor sections were incubated with primary antibodies followed with biotinylated secondary antibody (1:200; The Jackson Laboratory, Bar Harbor, ME) and streptavidin/horseradish peroxidase (The Jackson Laboratory). Stable diaminobenzidine (Resgen/Invitrogen) was added to each slide, and hematoxylin was used as a counterstain.
Inhibition of the PI3K/Akt Pathway but not the Raf/MEK/Erk Pathway Results in Reduction of Ras-Dependent Melanoma Tumor Growth Both in Vitro and in Vivo.
To investigate the effect of selectively blocking key downstream effectors of Ras signaling pathways, we treated TPras mouse melanoma both in vitro and in vivo with Ly294002, a PI3K inhibitor, and U0126, a MEK1/2 inhibitor. The ability of 1984-1 melanoma cells to form colonies in soft agar in the presence or absence of the inhibitors was tested over a 2-week period (Fig. 1A) ⇓ . The efficacy of both chemical and genetic inhibitors was verified by Western blotting of the phosphorylated forms of Akt and Erks, downstream targets of PI3K and MEK, respectively (Fig. 1B) ⇓ . As expected, these melanoma cells were able to grow in suspension. Inhibition of PI3K by either the chemical inhibitor Ly294002 or the dominant negative subunit Δp85 resulted in significant reduction of the number of colonies in soft agar. On the contrary, the inhibition of the Raf/MEK/Erk pathway by either U0126 or Raf301 did not decrease the number of colonies.
As observed previously, injection of these TPras melanoma cells into SCID mice resulted in the development of invasive tumors (46) . To determine whether the inhibition of the PI3K/Akt or the Raf/MEK/Erk pathways was able to affect the growth of the tumor cells in vivo, mice that received s.c. injection with 1984-1 cells were treated topically with Ly294002 and U0126 three times per week. Importantly, at the time of sacrifice, no forms of morbidity, including skin irritations, hyperplasia, or systemic disease were observed. As shown in Fig. 1C ⇓ , the application of the PI3K inhibitor significantly reduced tumor size. Vehicle-treated cells formed a tumor of approximately 1 cm3 in 40 days (Fig. 1C ⇓ , experiment 2). Ly294002-treated mice, however, formed tumors of only 0.4 cm3. In contrast, no effect on tumor size was observed in mice treated with the MEK1/2 inhibitor U0126. Application of both drugs resulted in a reduction in tumor size comparable with that was seen in the mice treated with Ly294002 alone, suggesting that the effect observed is mainly due to the inhibition of the PI3K/Akt pathway. A similar effect was observed when 2 × 106 cells were injected, and tumor growth was monitored for 28 days (Fig. 1C ⇓ , experiment 1).
Results of immunohistochemical staining with Ki67, a marker of proliferation, support the gross observations. As seen in Fig. 1D ⇓ , many of the melanoma cells from the control tumor stained for Ki67. Tumors treated with the inhibitors in combination showed the greatest reduction in the number of proliferating cells. Tumors treated with only the PI3K inhibitor also appeared to have a significant decrease in cell proliferation, whereas little or no difference was observed when comparing tumors from the control and U0126-treated mice. These data suggest that inhibition of PI3K limits growth of the TPras melanomas.
Tumor Cell Invasion and Expression of MMP2 Are Reduced after Inhibition of the PI3K/Akt and Raf/MEK/Erk Pathways.
A feature of malignant melanomas is their ability to degrade the extracellular matrix and invade the surrounding tissues. In this process, both the PI3K/Akt and Raf/MEK/Erk pathways have been shown to play a role (47, 48, 49, 50, 51) .
On excision of the tumors from the SCID mice, gross observations were made of invasion into the adjacent muscle tissue. Tumors that easily peeled away from underlying muscle (i.e., tumors contained within the dermal/s.c. layer) were considered noninvasive. Tumors in which muscle was removed along with the tumor were considered invasive. Using these criteria, the following observations were made: 9 of 10 tumors from vehicle-treated mice were invasive; 3 of 9 of the Ly294002-treated tumors were superficially invasive; of the U0126-treated tumors, 4 of 9 were moderately invasive; and, finally, 2 of 9 tumors receiving the combined treatment were superficially invasive. These observations suggest an involvement of the PI3K/Akt and Raf/MEK/Erk pathways in the invasiveness of these melanoma cells. To assess the role of these pathways in the invasive behavior of the 1984-1 melanoma cell line, an in vitro assay was performed using Matrigel invasion chambers. As shown in Fig. 2A ⇓ , treatment with Ly294002 and U0126, alone or in combination, markedly decreased the ability of the cells to penetrate the Matrigel. In the control wells, >60% of the cells were invasive, whereas only 10% and 20% of the cells were invasive when treated with Ly294002 and U0126, respectively. A similar result was obtained when the 1984-1 melanoma cells expressed the dominant negative forms of PI3K and Raf1 (Fig. 2B) ⇓ . Only 10% and 5% of the cells overexpressing Δp85 and Raf301, respectively, were able to invade. These results support previous findings that both pathways play a role in the invasive potential of tumor cells and that blocking either pathway can reduce this activity in melanomas.
Melanomas express a number of MMPs (52) . Among them, MMP2 has been associated with highly invasive behavior and melanoma progression (53 , 54) . Akt and Erk have been implicated in the regulation of MMP2 (33 , 34 , 55) . To investigate the molecular basis for the invasive phenotype of melanoma cells, reverse transcription-PCR and gelatin zymography were performed on either RNA samples or conditioned media from 1984-1 mouse melanoma cells treated with Ly294002 and U0126 or expressing Δp85 and Raf301. As demonstrated in Fig. 3, A and B ⇓ , inhibition of the PI3K/Akt pathway and, to a lesser extent, the Raf/MEK/Erk cascade leads to inhibition of both transcript and protein activity of both pro- and active MMP2. When both chemical inhibitors were present in combination, the inhibitory effect was increased, suggesting that the two pathways may independently contribute to the regulation of the molecule. No effect was observed on the expression of the MMP2 activator MT1-MMP. Moreover, although a role for MMP9 in melanoma progression has also been observed (56) , no significant effect on either transcript or activity was detected (data not shown). As a confirmation that the decrease in MMP2 expression and activity was specifically dependent on the down-regulation of the PI3K/Akt and Raf/MEK/Erk pathways, similar experiments were done using the dominant negative Δp85 and Raf301 (Fig. 3, C and D) ⇓ . Again, the strongest inhibitory effect was observed with down-regulation of the PI3K/Akt pathway, whereas minor inhibition occurred with Raf301.
To determine whether topical treatment of the 1984-1 melanoma xenografts with Ly294002 and/or U0126 resulted in a similar down-regulation of metalloproteinases, we stained tumor sections with anti-MMP2 and anti-MMP9 antibodies. As can be seen in Fig. 3E ⇓ , the control tumors demonstrated the presence of the protein. The most reduction was observed in the tumors treated with the two drugs in combination (Fig. 3E ⇓ : a, control; b, Ly294002 + U0126). In the Ly294002- or U0126-treated tumors, the protein levels were reduced (data not shown), although not to the degree of the combined treatment. MMP9 protein was undetectable.
Inhibition of VEGF Expression and Secretion by Down-Regulation of PI3K/Akt and Raf/MEK/Erk Pathways.
VEGF, a potent endothelial mitogen and promoter of neovascularization of tumors, is commonly produced by tumor cells in both premalignant and malignant lesions (57) . In this complex phenomenon, oncogenes, such as Ras, have been shown to induce VEGF expression (31 , 58) .
Ras is overexpressed in 1984-1 melanoma cells; therefore, we examined the effect of blocking the downstream pathways PI3K/Akt and Raf/MEK/Erk on VEGF production in vitro. VEGF expression appeared to be dependent on both pathways. Treatment with Ly294002 or U0126 decreased the VEGF transcript (Fig. 4A) ⇓ and significantly reduced secretion of the protein as evidenced by an ELISA assay (Fig. 4C) ⇓ . Treatment with both Ly294002 and U0126 resulted in slightly more inhibition. The regulation of VEGF transcription and secretion by the PI3K/Akt and Raf/MEK/Erk pathways was also confirmed by the use of dominant negative constructs. As shown in Fig. 4, B and D ⇓ , overexpression of either Δp85 or Raf301 decreased both VEGF transcription and protein secretion.
Together, these results suggest that both the PI3K/Akt and the Raf/MEK/Erk pathways are involved in the regulation of VEGF and may contribute to the aggressive behavior of TPras melanoma in vivo.
Inhibition of Tumor Angiogenesis by Topical Treatment with Ly294002 and U0126.
Melanomas produce a number of angiogenic factors, including VEGF, which promote the formation of new vessels that allow the growing tumor to receive oxygen and nutritional supplies and present a route for metastasis. To measure angiogenic factor expression, we stained tumor sections with an anti-VEGF antibody. As shown in Fig. 5A ⇓ , in the control tumors, the brown-stained cells are indicative of the production of VEGF. A significant decrease in VEGF protein was noted in melanoma sections from SCID mice treated with Ly294002 and U0126, either alone or in combination (Fig. 5A, b–d) ⇓ .
To assess whether the VEGF decrease affected blood vessels in the tumors, tumor sections were stained with anti-CD31 antibody specific for endothelial cells. Overall density versus the more defined vessel formation was examined. In the tumors from the control mice, we observed relatively dense patterns of vessels as well as defined open vessels. In the melanomas from the treated mice, we observed either similar density levels (Fig. 5B ⇓ : Ly294002, f; U0126, g) or, in some tumors, a reduction (Ly294002 + U0126, Fig. 5B, h ⇓ ). However, we also observed attenuated or collapsed vessels (unopen or no lumen) as shown in Fig. 5B, f and g ⇓ (Ly294002- and U0126-treated tumors). To further characterize the vessels, the mice received injection with the fluorescent dye DiOC7 to label vessels with active blood flow. In the vehicle-treated tumors, the fluorescent staining clearly outlined functional vessels with an open lumen (Fig. 5C, i) ⇓ . Diminished staining was observed in the sections from Ly294002-treated tumors, and poorly defined lumens were seen in the U0126- and Ly294002 + U0126-treated groups (Fig. 5C, j–l) ⇓ . These observations were further confirmed by counting tumor microvessels. Although the total number of microvessels did not appear to be different among the treatment groups, treatment with the chemical inhibitors significantly reduced the number of functional vessels (Fig. 5C, j–l versus i) ⇓ .
Down-Regulation of the PI3K/Akt and Raf/MEK/Erk Pathways Impairs Tube Formation of Endothelial Cells.
Topical drug treatment can affect not only tumor cells but surrounding stromal cells as well. We looked at whether the impaired tumor angiogenesis was due to reduced production of angiogenic factors by the tumor as well as a direct effect on the endothelial cells. To assess this question, an in vitro angiogenic assay was performed using HUVECs. Treatment of HUVECs with VEGF has been shown to cause formation of a tubule network. As shown in Fig. 6A ⇓ , treatment with the chemical inhibitor Ly294002 completely disrupted this tubule network. Some decrease (although it was not significant) was also seen with the MEK1/2 inhibitor U0126. Thus, Ly294002 and, to a lesser extent, U0126 likely have a direct effect on stromal endothelial cells in vivo as well. To determine whether inhibition of the PI3K/Akt and Raf/MEK/Erk pathways also affects the secretion of functional proangiogenic factors, as one would expect from the studies on VEGF (Fig. 4) ⇓ , we used 1984-1 melanoma cells expressing the dominant negative Δp85 and Raf301. Conditioned media from 1984-1 cells contains potent angiogenic factors, as evidenced by tubule formation by the HUVECs (Fig. 6B) ⇓ . Conditioned media from 1984-1 cells expressing Δp85 significantly reduced the promotion of in vitro angiogenesis, and media from 1984-1/Raf301 cells also resulted in a decrease in tube formation (however, this decrease was not significant).
In summary, these studies suggest that drugs delivered topically may affect both tumor cells (by reducing the production of angiogenic factors) and endothelial cells [by directly disrupting their survival/differentiation signaling (59 , 60)] .
New means for treating and controlling melanoma development are of extreme importance because it is an aggressive tumor, with a propensity to metastasize, and is resistant to most current therapeutic regimens (61) . Understanding the genetic and molecular basis of cutaneous melanoma will allow identification of rational chemoprevention strategies and new therapeutic interventions. Our studies suggest that topical application of targeted agents may be an effective mode of delivery for cutaneous melanomas. The targeting of Ras and its effectors by specific inhibitors is a potential strategy to treat melanoma.
In these studies, the mouse melanoma cell line 1984-1 was derived from a cutaneous melanoma from a TPras transgenic mouse (62) . These cells express an activated human HRas and, as a consequence, show activation of the downstream pathways PI3K/Akt and Raf/MEK/Erk (42) . They also contain a chromosomal alteration that results in the loss of expression of the p16 (INK4a) molecule (46) , a characteristic that is shared with some familial and sporadic human melanomas (7) . These mouse melanoma cells are characterized by anchorage-independent growth and, when injected into immunodeficient mice, form a highly invasive melanoma (46) . These characteristics are typical of malignant tumors and have been demonstrated to be dependent, at least in part, on the Ras downstream pathways PI3K/Akt and Raf/MEK/Erk in other tumor models (63, 64, 65) . The PI3K/Akt and Raf/MEK/Erk pathways regulate different functions essential for cell growth and survival. We examined whether targeting these pathways with topically applied PI3K and MEK1/2 inhibitors (Ly294002 and U0126, respectively) would have therapeutic activity against cutaneous melanomas. In particular, we examined the effects of each inhibitor in vitro and in vivo on the biological functions of proliferation, angiogenesis, and invasive behavior.
As observed for other cell systems, down-regulation of the PI3K/Akt pathway in the TPras melanoma cells resulted in a great reduction in both colony formation in soft agar and tumor growth in vivo. Although the Raf/MEK/Erk pathway has been reported to be required for Ras-dependent transformation of different cell types (66, 67, 68) , and inhibition of this cascade by the use of a novel Raf-1 inhibitor has shown interesting antitumoral effects (69) , down-regulation of this pathway in the 1984-1 melanoma cells did not result in a decrease in colony formation in soft agar or of tumor growth in SCID mice. These results suggest that, at least in Ras-dependent melanoma models, a key pathway regulating tumor cell growth at this stage is through the activation of PI3K/Akt. It is possible that the Raf/MEKK/Erk pathway is more important during early or late phases of melanoma development. Studies are under way to examine this question in two transgenic mouse melanoma models that will allow us to examine early stages. In fact, preliminary observations in TPras transgenic mice treated with Ly29004 and U0126 in combination indicate that these agents may markedly reduce tumor incidence as well as tumor size. 3 It is also possible that melanomas harboring mutations other than Ras may better respond to inhibitors of the Raf/MEK/Erk pathways. In fact, recent findings support the efficacy of an orally delivered MEK inhibitor for the treatment and prophylaxis of a metastatic melanoma cell line harboring the V599E BRaf mutation (70) .
In the 1984-1 transgenic melanoma xenografts, characteristics of malignancy such as invasive behavior and angiogenesis are indeed affected by the inhibition of the Raf/MEK/Erk pathway. Invasion is a complex phenomenon characterized by the ability of tumor cells to penetrate the basement membrane, and MMPs play a crucial role in this process. Melanoma cells express a number of MMP family members (MMP1, MMP2, MMP9, MMP13, and MT1-MMP). MMP2 and MMP9 are thought to be particularly important in tumor cell invasion (71) and have been associated with highly invasive human melanoma cell lines and generally associated with melanoma progression (53 , 54) . Both the PI3K/Akt and Raf/MEK/Erk pathways are well-known regulators of invasive behavior of tumor cells (48 , 49 , 72 , 73) . In our system, blockage of either pathway inhibited or reduced tumor cell invasion. Moreover, expression and activity of MMP2 were decreased significantly with either of the chemical inhibitors (and particularly when applied in combination) or with expression of the dominant negative Δp85 or Raf301. These results suggest a role for both the PI3K/Akt and Raf/MEK/Erk pathways in the Ras-dependent regulation of MMP2. A similar observation was recently reported that indicates PI3K as a key regulator of MMP2 through MT1-MMP in vasculogenic mimicry of a highly invasive human melanoma cell line (74) . Our data support the role of PI3K in the regulation of MMP2, although no effect was observed on the expression of MT1-MMP in our model. These different results may reflect the heterogeneity of melanomas or differences between human and mouse cells.
Malignant solid tumors are also characterized by an angiogenic phenotype (75) . Tumor neovascularization results from a reciprocal effect between tumor cells that produce angiogenic factors and the surrounding vessels that are stimulated to proliferate and move toward the angiogenic chemoattractant. VEGF is one of the major factors promoting tumor neovascularization, and oncogenes, such as Ras, are known to induce VEGF expression (31 , 58) . Our in vitro data show that both VEGF expression and secretion are dependent on the Ras downstream pathways PI3K/Akt and Raf/MEK/Erk. In addition, we observed a clear correlation between VEGF expression and tumor angiogenesis in the tumors that grew on the SCID mice. Here, large functional vessels were found in control tumors, which also presented the most intense staining for VEGF. In contrast, poorly formed or completely obliterated vessels were observed in tumors treated with the inhibitors, which were also characterized by a lower level of VEGF expression. These observations clearly support a relationship between the production of VEGF and the angiogenic potential of these tumor cells. However, the very nature of the treatment precludes the conclusion that the reduced angiogenesis is due solely to a reduction of angiogenic factor production by the melanoma cells. Topical treatment with drugs such as Ly294002 and U0126 may also affect the transduction of signals in the endothelial cells (59 , 60) , thus preventing endothelial cell response to angiogenic factors. Our analysis of in vitro tube formation by endothelial cells demonstrated an inhibition by both Ly294002 and, to a lesser extent, U0126. These experiments show that the topical treatment has antiangiogenic activity, likely by affecting both tumor and stromal cells.
In summary, our studies indicate that topical application of the PI3K inhibitor, alone or in combination with a MEK1/2 inhibitor, may be an effective treatment or preventive strategy against the development and progression of cutaneous melanomas. These agents act by blocking tumor cell growth, production of angiogenic factors, and metalloproteinases involved in tumor cell invasion. Moreover, the data presented here support the possibility of using potent drugs administered in a noninvasive way that exert their effect with considerably less toxicity than systemic delivery.
We thank Dr. Giovanni Pani (Università Cattolica, Rome, Italy) and Dr. Renata Colavitti (NIH, Bethesda, MD) for critical discussion and technical advice.
Grant support: NIH Grants CA090897 and CA27502. S. M. Welford was supported by National Cancer Institute Department of Health and Human Services Grant CA09302.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: B. Bedogni and M. S. O’Neill contributed equally to this work.
Requests for reprints: Marianne Broome Powell, Division of Radiation and Cancer Biology, 269 Campus Drive, CCSR-1230, Stanford, California 94305. Phone: (650) 498-5874; E-mail:
↵3 B. Bedogni, M. S. O’Neill, and M. B. Powell, unpublished data.
- Received October 22, 2003.
- Revision received January 21, 2004.
- Accepted January 28, 2004.
- ©2004 American Association for Cancer Research.