Progression of human melanoma is a highly efficient process compared to other malignant tumors since a few millimeters thick tumor (a very small primary in the case of other malignancies) can colonize regional lymph nodes or visceral organs. It is therefore extremely important to understand the molecular mechanisms behind this highly aggressive behavior. Previous studies indicated that the autocrine growth regulation of melanoma acquired at the switch from a less invasive radial to a more invasive vertical growth phase involves bFGF expression.1, 2 On the other hand, overexpression of αvβ3 integrin is also associated with the invasive growth3, 4 and metastasis5, 6 of melanoma. αIIbβ3 integrin (GpIIbIIIa) is the predominant adhesion receptor of platelets and the expression of the integrin αIIb chain is megakariocyte-specific. It is involved primarily in platelet activation, since it is expressed in a low affinity state and binds fibrinogen only upon activation. We have previously shown the illegitimate expression of αIIbβ3 integrin in human melanoma,7, 8 involved in the same phases of tumor progression as αvβ3. Recently we found that the transduction of αIIbβ3 into αvβ3-expressing human melanoma cells did not affect the in vitro but promoted the in vivo growth of tumor cells due to decreased apoptosis.8 We have postulated that one possible factor behind the increased in vivo growth is vascularization. Therefore, we have analyzed the gene expression changes of αIIbβ3-transfected human melanoma clones with special attention to their angiogenic phenotype.
Previous studies indicated that transfection of the platelet integrin αIIbβ3 into human melanoma cells expressing integrin αvβ3 promoted their in vivo (but not in vitro) growth and cell survival. To reveal the underlying pathomechanism, we have analyzed the angiogenic phenotype of αIIbβ3 integrin-transduced human melanoma cells expressing integrin αvβ3. Upon heterotopic or orthotopic (intracutaneous) injections into SCID mice, the αIIbβ3 integrin-overexpressing clones, ESL, ESH, 19L and 19H, grew more rapidly than the mock transfectant (αvβ3 expressing) clone, 3.1P. Morphometry demonstrated an increased intratumoral microvessel density in 19L and 19H tumors compared to 3.1P. Immunocytochemistry and flow cytometry indicated that vascular endothelial growth factor (VEGF) is constitutively expressed in the majority of the cells of both the mock and the αIIbβ3 integrin-transfected clones. However, the mock transfectant clone, 3.1P, did not express basic fibroblast growth factor (bFGF) at protein level (<1%), unlike the αIIbβ3 integrin-transfected clones, 19L and 19H, (33.9 and 84.1%, respectively). Quantitative PCR analysis of 6 related human melanoma clones with various levels of αIIbβ3 integrin expressions revealed a correlation between the αIIb protein and bFGF mRNA expressions. Furthermore, cDNA microarray analysis of the 19H cells revealed 12 downregulated and 36 upregulated genes [among them 3 upregulated vasculogenic mimicry-genes (CD34, endothelin receptor B, Prostaglandin I-2 synthase)] when compared to 3.1P cells. The altered bFGF expression may be influenced by integrin-linked signaling, since bβ3-endonexin is upregulated in αIIbβ3-transfected cells and tyrosine kinase inhibitors downregulate bFGF both at mRNA and protein levels. We propose here that the illegitimate expression of αIIbβ3 integrin in human melanoma cells already expressing αvβ3 integrin may alter their in vivo growth properties due to the modulation of their angiogenic phenotype. © 2005 Wiley-Liss, Inc.
Material and methods
Human melanoma cells
The WM983B cell line that does not express αIIbβ3 on the cell surface was a kind gift of M. Herlyn (The Wistar Institute, Philadelphia, PA). Mock transfected WM983B cells (3.1P) and αIIb and β3 transfected WM983B cells (19L and 19H) have been described earlier.8 New αIIb and β3 transfected WM983B cell clones, ESL and ESH, have been isolated as described before.8 19H transfected cell line was subcloned by limited dilution to obtain 7D7 and 8F3 subclones. The parental cell line was grown in RPMI-1640 medium supplemented with 5% FBS and antibiotics (Sigma Chemical Co., St. Louis, MO) in a 5% CO2 humidified atmosphere at 37°C. Transfected cells were maintained in medium containing G418 (Gibco BRL, Gaithersburg, MD).
Tumor growth in SCID mice: Orthotopic growth
Five SCID mice/group were anaesthetized by intraperitoneal administration of Nembutal (70 mg/kg), the thighs of the animals were shaved and 105 tumor cells in 5 μl Hank's balanced salt solution (HBSS) were inoculated intradermally, under a dissecting microscope by using a precision syringe (diameter of 0.15 mm, Hamilton). Volume measurements of palpable tumors were made on day 10 and 13 by using a precision dermatometer (length × width2 × π/6).
Tumor cells (106/200 μl) were injected into the tail vein of SCID mice and after 60 days mice were terminated by Nembutal overdose and lungs were isolated, fixed and surface colonies were counted under a stereomicroscope.
Tumor cells (106/200 μl) were injected into the spleen of SCID mice and after 30 days mice were terminated by Nembutal overdose and the liver was isolated, fixed and surface colonies were counted under a stereomicroscope.
Tumor cells (106/200 μl) were injected into the left cardiac ventricle of SCID mice under anesthesia and after 60 days mice were terminated by Nembutal overdose and visceral organs (lung, liver and brain as well as bones of the lower extremities) were isolated, fixed and surface colonies (visible colonies in case of bones) were counted under a stereomicroscope. To confirm metastases, specimens were embedded into paraffin and H&E stained sections were also analyzed microscopically.
On day 13, 3 tumors of each group were removed and frozen in isopentane cooled in liquid nitrogen. Five micrometer frozen sections were fixed in methanol (10 min, −20°C), blocked in 0.1% bovine serum albumin (BSA) and immunostained with a rat monoclonal antibody against mouse CD31 (60 min, 37°C, 1:100 dilution, Pharmingen, San Diego, CA), a goat anti-human VEGF antibody (60 min, 37°C, 10 (μg/ml), (R&D Systems, Wiesbaden-Nordenstadt, Germany) or a goat anti-human bFGF antibody (60 min, 37°C, 10 (μg/ml), (R&D Systems). FITC-conjugated rabbit anti-rat and anti-goat IgG antibodies (45 min, 37°C, 1:50, Jackson Immunoresearch Lab, Inc., West Grove, PA) were used as secondary antibodies. Following nuclear staining with 5 μM TOTO-3 in PBS (20 min, 20°C, Molecular Probes, Inc.) slides were mounted and viewed by confocal laser scanning microscopy (Bio-Rad MRC 1024, Munich, Germany). Counts of intratumoral blood vessels were determined visually by fluorescence microscopy in ×200 fields within the entire tumor mass.
For immunofluorescence labeling suspended cells were fixed with 1% buffered paraformaldehyde (PFA) for 15 min at room temperature permeabilized with 0.2% saponin for 10 min. Nonspecific antibody binding was blocked using PBS containing 1% BSA. The following mouse monoclonal antibodies were used: 50 μg/ml of anti-CD41 (anti-αIIb), anti-CD61 (anti-β3) (both from Dako, Glostrup, Denmark) and 12.5 μg/ml of anti-αvβ3 complex (Chemicon, Hofheim, Germany). Cells incubated with the appropriate isotype control IgG were used as negative controls. The bound primaries were detected with a biotinylated anti-mouse Ig (1:100; Vector Laboratories, Inc., Burlingame, CA) followed by Streptavidin-FITC (1:100; Amersham, Little Chalfont, England).
To detect the expression of pro-angiogenic molecules, 106 tumor cells/sample were fixed in methanol (10 min, −20°C) incubated with goat anti-human VEGF antibody (60 min, 20°C, 10 μg/ml, R&D Systems) or goat anti-human bFGF antibody (60 min, 20°C, 10 (μg/ml), (R&D Systems) revealed by a FITC-conjugated secondary anti-goat antibody (Jackson ImmunoResearc Labs, West Grove, PA).
The percentage of positive cells was determined by a FACStar flow cytometer (Becton Dickinson, Sunnyvale, CA). Cells were considered positive when the fluorescence intensity was above that of 95% of the control cells labeled without the primary.
Inhibition of signal transduction in vitro
To inhibit protein kinase C activity, in vitro cultured tumor cells were pretreated with Calphostin C (Kamiya Biomedical Co., Thousand Oaks, CA) for 25 hr at a concentration of 5 μM using light activation. For the downregulation of PKC, we have exposed tumor cells to 8 ng/ml PMA (phosbol-12-myristate-13-acetate; Sigma Chemical Co.) for the same period of time. To inhibit protein tyrosine kinases, tumor cell cultures were pretreated with Genistein or Staurosporine (Calbiochem) for 25 hr at a concentration of 100 μg/ml.
Total RNA were extracted from the cells growing in cell culture using Tri-Reagent (Sigma Aldrich Co, Pool Dorset, UK) according to the manufacturer's protocol. Purity of the nucleic acid templates was elevated by extra-purification with DNA-free™ (Ambion, Austin TX). One microgram of total RNA were reverse transcribed from each sample using deoxy (d)-NTPs (0.5 mmol/L each), random primer (final concentration 3 μM), RNasin® ribonuclease inhibitor (Promega) (0.4 U/μl), DTT (final concentration 10 mM), reverse transcriptase buffer (containing 250 mM Tris-HCl, pH 8.3, 375 mM KCl and 15 mM MgCl2) and M-MLV reverse transcriptase (200 U/reaction; Sigma Chemical Co.). Twenty μl samples were incubated for 50 min at 37°C and then at 85°C for 10 min. The sequences of bFGF primers were the following: 5′-GCC ACA TCT AAT CTC ATT TCA CA-3′ and 5′-CTG GGT AAC AGC AGA TGC AA-3′. The sequences of endothelin receptor type B primers were 5′-GCT CTT AAC AAC TTC CAG GAT ATT C-3′ and 5′- CAA CAA CTA AAC TGC TCT CTC ATT T-3′. The sequences of (3-endonexin primers were 5′-CAC TGA AGT TGG ATG GTC TGT TA-3′ and 5′-TTC ATT TGA TAG TCC ATT TCT GTG C-3′. (The primers were designed for the 111 amino acid coding sequence of the β3-endonexin cDNA, which is common in the 3 different splice isoforms of the β3-endonexin).
The real-time PCR analysis for mRNA expression of bFGF, CD34, endothelin receptor typeB, prostacyclin synthase and β3-endonexin standardized by coamplifying these genes with the housekeeping gene β-actin (primers: 5′-GTG GGG CGC CCC AGG CAC CCA-3′ and 5′-GTC CTT AAT GTC ACG CAC GAT TTC-3′). The real-time PCR reaction was run on the iCycler iQ™ (Bio-Rad) using standard conditions, namely, optimized concentration of primers (final concentration 200 nM), IQ™SYBR® Green Supermix (containing 100 mM KCL, 40 mM Tris-HCL, pH 8.4, 0.4 nM of each dNTP, 6 mM MgCl2, 50 U/ml iTaq DNA polimerase, SYBR Green I and 20 nM fluorescein) and 2 μl cDNA. A no template control (containing water) was used as negative control for every different primer-pair. The cycling parameters of the bFGF real-time PCR analysis were 95°C (3 min), 40 cycles of 95°C (30 sec), 61°C (30 sec) and 72°C (1 min). The cycling parameters of the β3-endonexin real-time PCR analysis were 95°C (3 min), 40 cycles of 95°C (30 sec), 60°C (30 sec) and 72°C (1 min). The starting quantity of gene expression in the sample was determined by comparison of unknown to a standard curve generated from a dilution series of template DNA of known concentration. CT values were converted into relative concentration values by scaling from 1 to 1 between the expression control and the highest value of the samples using the autothreshold fit.
Construction of microarrays
Construction and use of microarrays were performed as previously described.9, 10 Briefly, 3,200 cDNA inserts from human cDNA libraries (melanoma, lymphocytes, heart and mixed tissue libraries) were amplified and purified with MultiScreen-PCR plate (Millipore), resuspended in 50% dimethylsulfoxide/water and arrayed on FMB cDNA slides (Full Moon BioSystems, Sunnyvale, CA) by using a MicroGrid Total Array System (BioRobotics, Cambridge, UK) spotter with 16 pins in a 4 × 4 format. DNA elements were deposited in duplicate. The diameter of each spot was approximately 200 μm. After printing, DNA was UV crosslinked to the slides (Stratagene, Stratalinker, 700 mJ). Postprocessing and blocking of the microarrays were performed as described previously.11
Microarray probe preparation and hybridization
RNA isolation from 5 × 106 cells was carried out with the RNA isolation kit of Macherey-Nagel (Macherey-Nagel, Düren, Germany) according the manufacturer's instruction. RNA was eluted from the silica membrane and stored at −80°C in the presence 30 U of Prime RNAse inhibitor (Fermentas, Vilnius, Lithuania). The quality and quantity of isolated RNA was checked by electrophoresis and spectrophotometry (NanoDrop, Rockland, DE) respectively.
For probe preparation, 4 μg of total RNA was reverse transcribed using poly-dT primed Genisphere Expression Array 350 Detection system (Genisphrere, Hatfield, PA) in 20 μl total volume using 20 Unit RNAsin (Fermentas), 1× first strand buffer and 200 Units of RNAse H (−) point mutant M-MLV reverse transcriptase (Fermentas). All the other probe preparation steps were done according the manufacturer's instruction (Genisphere). cDNA was hybridized onto human cDNA microarrays in a Ventana hybridization station (Ventana Discovery, Tuchon, State) by using the “antibody” protocol. First hybridization was performed at 42°C for 6 hr in “Chiphyb” hybridization buffer (Ventana) and then 2.5 μl of each Cy5 and Cy3 capture reagents were added to the slides in 200 μl Ribohyb hybridization buffer (Ventana) and incubated at 42°C for 2 hr. After hybridization, the slides were washed in 0.2× SSC twice at RT for 10 min and then dried and scanned.
Each array was scanned under a green laser (543 nm for Cy3 labeling) or a red laser (633 nm for Cy5 labeling) using a ScanArray Lite (GSI Lumonics, Billerica, MA) scanning confocal fluorescent scanner with 10 μm resolution (laser power: 85% for Cy5 and 90% for Cy3, Gain: 80% for Cy5 and 75% for Cy3). Scanned output files were analyzed using the GenePix Pro3.0 software (Axon Instruments, Inc., Foster City, CA). Each spot was defined by automatic positioning of a grid of circles over the image. The average and median pixel intensity ratios calculated from both channels and the local background of each spot were determined. An average expression ratio (MeanR, denotes the average of local background corrected pixel intensity ratios) was determined for each spot. Normalization was performed by the global Lowess method. Those data were flagged and excluded where the replicate spots from a different site of the same array were significantly different. We defined those genes to be regarded in melanoma cell cultures, where the average-fold change (increase or decrease) of the 4 data points were at least 2.0-fold.
Parallel expression of αIIbβ3 and αvβ3 integrins promotes ortho- and hetero-topic growth of human melanoma cell lines
The 5 αIIbβ3 transfected human melanoma clones expressed surface αIIb at different levels (6.7–27.2%) compared to the 3.1P mock transfected clone (0.7%) as determined by flow cytometry (Table I), while the β3 expression was similarly high in all the clones studied (64–80%). αvβ3 expression was high in the mock cells and was maintained, although at a lower levels in most of the transfected clones (Table I).
|3.1 mock||0.66 ± 0.7||80.0 ± 3.2||44.0|
|ESL||6.7 ± 3.4||64.0 ± 3.0||31.0|
|19L||12.3 ± 4.9||67.0 ± 5.6||10.5|
|ESH||22.5 ± 7.7||69.3 ± 2.3||13.0|
|19H||27.2 ± 3.9||78.0 ± 6.2||28.4|
αIIbβ3 transfected human melanoma clones were injected heterotopically into the spleen of SCID mice. All the clones were tumorigenic but the αIIbβ3-transfected clones (ESL, ESH, 19L and 19H) grown significantly more rapidly than the mock cells, 3.1P (Fig. 1a), as was observed in case of s.c. growth reported before.8 Next, selected αIIbβ3-transfected human melanoma clones (3.1P, 19L and 19H) were injected orthotopically (intracutaneously) into SCID mice. All these clones started to grow, which was detectable from days 7–10. However, it became apparent again that the αIIbβ3-transfected clones, 19L and 19H, grew significantly faster at orthotopic site than the mock transfectant in the early phase of tumor development (Fig. 1b, p < 0.05 and p < 0.001, respectively).
Next we have tested whether the αIIbβ3-transfection affected organ colonization potential of human melanoma clones. We have used 3 different models (intravenous, intrasplenic and intracardiac injections), but we have not seen significant alterations in the metastatic potential of transfected clones compared to the mock cells, 3.1P (Table II).
|Lung (i.v.)||Liver (i.s.)||Brain (i.c.)||Bone (i.c.)|
Parallel expression of αIIbβ3 and αvβ3 integrins promotes vascularization and the angiogenic phenotype of human melanoma cells
Intracutaneously growing human melanoma xenografts have been removed on day 15, snap frozen, sectioned and labeled for mouse CD31 to visualize intratumoral blood vessels (Fig. 2a,b). Tumors of the transfected clones, 19L and 19H, contained more intratumoral blood vessels than that of the mock transfected 3.1P cells, and the difference was significant in case of the 19H tumor (Fig. 2c).
Next we analyzed the immunocytochemical expression of the major angiogenic cytokines of melanoma, VEGF and bFGF, in the αIIbβ3-transfected human melanoma clones, 19L and 19H, by flow cytometry. Data indicated that VEGF is constitutively expressed in the majority of the cells of both the mock and the αIIbβ3-transfected clones (Fig. 3a,b). However, the mock transfected clone, 3.1P, did not express bFGF at protein level (<1% of the cell population, Fig. 3a,b), unlike the αIIbβ3-transfected clones, 19L (33.9%) and 19H (84.1%), where a significant proportion of the tumor cell population was positive (Fig. 3a,b). In vitro cultured tumor cells were also stained for bFGF protein and the data supported the observations made by flow cytometry that the cells of the αIIbβ3-transfected 19H clone are heavily positive for bFGF protein (Fig. 4a,b). We have questioned if this difference in bFGF expression was maintained in the intracutaneously growing tumors; therefore cryosections have been labeled for bFGF protein and analyzed by confocal microscope. The 3.1P tumor barely contained bFGF positive cells, but the αIIbβ3-transfected tumor (19H) was identified by a predominating population of tumor cells with strong cytoplasmic bFGF labeling (Fig. 4c,d).
Parallel expression of αIIbβ3 and αvβ3 integrins correlates with altered gene expression profile of human melanoma cells
Next we questioned whether the αIIbβ3 transduction into αvβ3-positive human melanoma clones affected bFGF expression at transcriptional level. We collected 6 human melanoma clones with various levels of αIIbβ3 and αvβ3 integrin expressions (Table I) and tested for bFGF mRNA levels. The quantitative-PCR analysis revealed that bFGF mRNA level increased parallel with αIIbβ3 protein expression in the analyzed clones, where the highest bFGF expressions detected in clones with the highest proportion of αIIbβ3 positive cells, 19L, ESH and 19H, respectively (Fig. 5). However, after a log increase of both αIIbβ3 or bFGF expressions, a further increase in integrin expression (clones ESH and 19H) does not necessarily result in further upregulation of bFGF (19H), suggesting a more complex association between the regulation of the 2 genes.
Next we tested the effect of αIIbβ3-transfection on gene expression using a homemade 3.2K cDNA microarray9, 10 that did not contain a bFGF probe. For this purpose, we used the fastest growing and most vascularized clone, 19H, and compared it to the mock cells, 3.1P. We defined those genes to be regulated by αIIbβ3-transfection in melanoma cell cultures, in the case of which the average-fold change (increase or decrease) of the 4 data points was at least 2.0-fold. Forty-eight genes in 19H cells corresponded to such strict criteria from the 3,200 analyzed; 36 genes were found to be upregulated and 12 have been downregulated (Table III). Among the downregulated genes, there were no known angiogenic/endothelial ones represented; however, among the 19 known genes upregulated in 19H cells compared to 3.1P, 3 could be considered endothelial-specific, such as CD34 antigen, endothelin receptor type B and prostaglandin I-2 synthase, the changes of which were also confirmed by q-PCR analysis (data not shown).
|2,002048||D84124||Prostaglandin I-2 (prostacyclin) synthase|
|2,013939||AA933627||ESTs, Highly similar to PHOSPHATIDYLOCHOLINE TRANSFER PROTEIN [Bos Taurus]|
|2,014675||D21235||Human mRNA for HHR23A protein, complete cds|
|2,13392||D83780||Human mRNA for KIA0196 gene, complete cds|
|2,173375||D131168||Endothelin receptor type B|
|2,23487||AA648933||ESTs, Weakly similar to homologous to mouse Rsu-1 (H.sapiens)|
|2,302858||AC004262||Homo spiens chromosome 19, cosmid R29368|
|2,307286||L34408||Homo Sapiens (clone B3B3E13) chromosome 4p16.3 DNA fragment|
|2,387747||X62535||Diacylglycerol kinase, alpha (80kD)|
|2.403949||M10942||Human metallothionein-le gene (hMT-le)|
|2,418433||Y00971||Phosphoribosyl pyrophosphate synthetase 2|
|2,650112||AF001862||Human SLP-76 associated protein mRNA, complete cds|
|2,77783||U07857||Signal recognition particle 14 kD protein|
|2,987556||X52943||TRANSCRIPTION FACTOR ATF-A AND ATF-A-DELTA|
|3,07509||S53911||CD34 antigen (hemopoietic progenitor cell antigen)|
|4,460516||L03411||Radin blood group|
Based on this information, we have postulated that an altered signaling activity in the αIIbβ3-transfected clones may play a role in the changes in gene expressions, including bFGF. Q-PCR analysis of the expression of a β3 integrin-associated signaling molecule, β3-endonexin, indicated a 2.5–5-fold increase in the expression of this gene in various transfected clones compared to 3.1P (Fig. 6a). Again, similarly to bFGF expression (Fig. 5), there was no linear correlation between the αIIb (Fig. 5) and β3-endonexin expression levels in transfected clones. Next we used tyrosine kinase (Genistein, Staurosporine) and PKC (Calphostin C) inhibitors to modulate cellular signaling in the 19H αIIbβ3-transfected clone and measured the consequences on bFGF mRNA and protein expressions as determined by q-PCR and flow cytometry. These studies indicated that 25 hr exposure to tyrosine kinase inhibitors resulted in a decrease, while PKC-inhibition resulted in an increase in bFGF message (Fig. 6b). Measurement of the cytoplasmic bFGF protein by immunocytochemistry confirmed these observations, since both tyrosine kinase inhibitors (Genistein and Staurosporine) significantly inhibited the expression in transfected 19H cells (Fig. 6c).
Recently we have shown that transfection of αIIbβ3 integrin into human melanoma cells constitutively expressing αvβ3 promotes in vivo growth due to decreased apoptosis.8 Our analysis of the possible explanation for this phenomenon indicated that the αIIbβ3 transfected clones are more vascularized than the clone expressing αvβ3 alone providing an explanation for the increased in vivo growth. Based on our results, the studied clones constitutively express VEGF, but bFGF mRNA and protein expressions are induced only in the αIIbβ3 transduced clones. In agreement with previous studies,6 we found that the expression of αIIbβ3 parallel with αvβ3 integrin does not modify the organ selectivity of the metastatic potential of human melanoma cells, suggesting a specific modulatory effect of αIIbβ3 integrin on the angiogenic phenotype.
bFGF was shown to be an autocrine growth factor of human melanoma which is involved in proliferation, survival and invasiveness of transformed melanocytes.2, 11 The biological significance of its expression is demonstrated by experiments where bFGF or FGFR1 were used successfully as molecular targets to prevent in vivo growth of melanoma.12 Meanwhile bFGF is one of the most ubiquitous angiogenic cytokine that plays a significant role in vasculogenesis, as well as in physiological and pathological neoangiogenesis.13 It was repeatedly shown in melanoma that several angiogenic cytokines including VEGF are responsible for the angiogenic phenotype,14, 15 but these studies pointed to the significant contribution of bFGF. This assumption was further corroborated by the effects of downmodulation of FGF using antisense therapy,16 resulting in decreased vascularization. Our observation in our study on the upregulated bFGF expression in a constitutive VEGF background in the more angiogenic human melanoma clones further supports these data. Upregulation of bFGF during the progression of human melanoma was reported at the transition from the radial to the invasive vertical growth phase.2 Studies on human samples demonstrated that αvβ3 integrin expression is constitutive in primary melanomas, while the αIIbβ3 expression increased with tumor thickness.8, 17 We suggest that in human melanoma, overexpression of αvβ3 integrin may correlate with VEGF expression, while the emergence of the illegitimate expression of the αIIbβ3 integrin in a later stage of the disease (vertical growth phase) may contribute to the upregulation of bFGF expression, providing both another autocrine growth factor as well as another angiogenic cytokine for melanoma cells.
Parallel expression of the two β3-integrins in human melanoma cells induced modulation not only of bFGF but other genes as well, which is demonstrated by microarray analysis. Interestingly, 3 out of 19 known genes upregulated significantly in αIIbβ3-transfected 19H cells are endothelial cell-specific genes, CD34, endothelin receptor B and PGI-2 synthase. Recapitulation of an embryonic endothelial/angioblastic genetic program, called vasculogenic mimicry,18 resulting in in vivo formation of tumor cell-lined vascular channels, was recently described and documented in human melanoma and other tumors.19 There have been several genes identified to be responsible for this vasculogenic phenotype including VE-cadherin, CD34, TIE2, Eph2A, LAMC2, endoglin, EDG1, ESM1 and EDF1.20
AlphaIIbβ3 transfection into αvβ3-expressing human melanoma cells, demonstrated here, correlated with the overexpression of certain vasculogenic genes (bFGF, CD34, ENDRB and PGI-2 synthase), suggesting that β3 integrin signaling could be one of the potential molecular mechanisms behind this genetic reprogramming. αvβ3 integrin is involved in survival signaling in melanoma cells21 mediated either through FAK or Shc using the Ras-Raf-MAPK pathway22, 23 as well as PI3K.24 On the other hand, it was shown in murine melanoma cells, spontaneously expressing αIIbβ3 (but not αv) integrin, that β3 signaling is constitutively active and the ligation induced further activation of PKC.25 Based on the observations that tyrosine kinase inhibitors downregulated bFGF expression in αIIbβ3-transduced 19H cells while PKC inhibitor upregulated it, we suggest that the integrin-linked classical tyrosine signaling cascade might be involved in the upregulation of bFGF and other gene expressions in αIIbβ3-transfected human melanoma cells. β3 integrin is unique among integrins, since its cytoplasmic domain is associated with a signaling mediator that is a transcription factor: the β3-endonexin protein.26, 27 The fact that β3-endonexin is also upregulated in the αIIbβ3-transfected clones suggests that this integrin-linked signaling protein may also be involved in the observed alterations in the angiogenic geno- and pheno-types of human melanoma cells.
This work was funded by Hungarian National Science fund (OTKA T38128 to JT), by the Ministry of Education (1/48/2001, TJ), by the T. Fox Fund, by a bilateral grant of the Hungarian Prime Minister Office and Hungarian Academy of Sciences (4676/2003), NIH grant CA47115 (KVH) and NATO (CLG 975187, KVH&JT).