Melanoma metastasizes by different mechanisms comprising direct invasion of the surrounding tissue and spreading via the lymphatic or vascular system. Despite their clinical relevance, the molecular mechanisms that guide the route of spreading and localization of the metastases in different tissues are not well known. Recent studies in different tumor types have shown that vascular endothelial growth factor-C (VEGF-C), which displays a high specificity for lymphatic endothelium, is involved in tumor-induced lymphangiogenesis and lymphatic metastatic spread. The authors studied the expression of VEGF-C in cultured human melanoma cells derived from cutaneous and lymph node metastases as well as in metastatic melanoma tissue specimens to assess a possible involvement of this growth factor in lymph node localization of melanoma metastases.
VEGF-C expression was evaluated in vitro on human melanoma cell lines established from cutaneous and lymph node metastasis specimens by reverse transcriptase-polymerase chain reaction, Northern blot analysis, and immunofluorescence analysis. Immunohistochemical analysis of 42 tissue specimens of melanoma metastases and 10 tissue specimens of primary skin melanomas was also performed.
Preferential expression of VEGF-C was detected in lymph node-derived tumor cell lines at both the mRNA and protein levels. The association between VEGF-C production and lymph node localization of metastases was confirmed by the in vivo analysis. In addition, analysis of 10 patients, from whom specimens of both the primary skin melanoma and melanoma metastases were available, indicated a correlation between VEGF-C expression in the primary tumor and lymph node localization of metastases.
Induction of angiogenesis is an essential feature of growing tumor cells. Without an efficient blood supply, a tumor cannot grow beyond a critical size or metastasize to other organs. In addition, the blockade of vascularization has been reported to inhibit efficiently tumor growth and invasion and antiangiogenic molecules have been developed to be used in cancer therapy.1
Conversely, the possible involvement of lymphangiogenesis in tumor development and metastatic spread has been investigated only recently,2 mainly after the characterization of growth factors specific for the lymphatic endothelium, such as vascular endothelial growth factor-C (VEGF-C), a member of the VEGF family.3 VEGF-C consists of disulfide-linked polypeptides that undergo proteolytic processing. It binds to the tyrosine-kinase vascular endothelial growth factor receptor-2 (VEGFR-2/KDR), as does VEGF, and also is a ligand for the VEGFR-3/Flt-4.4 The latter receptor is essential for embryonic cardiovascular development,5 but becomes restricted to the lymphatic endothelium in adult tissue.6 The activation of VEGFR-3 by VEGF-C is sufficient to induce specifically lymphangiogenesis in vivo.7 Overexpression of VEGF-C in the skin of transgenic mice under the control of the keratin-14 promoter results in selective lymphatic hyperplasia.8 Transgenic animals that overexpress a dominant negative soluble form of VEGFR-3 develop specific regression of lymphatic vessels.9
In addition to induction of lymphangiogenesis, mice injected with tumor cells overexpressing VEGF-C demonstrate increased metastatic spread to the lymph nodes.10–13 In cases of human cancer, the occurrence of lymphangiogenesis during tumor development and progression has not been demonstrated clearly,14, 15 whereas data support a role for VEGF-C in promoting dissemination of metastases to the lymph nodes. VEGF-C mRNA has been detected in several human tumors.15, 16 In addition, VEGF-C expression in primary tumors, such as prostatic carcinoma, cervical and gastric carcinoma, and malignant mesothelioma, has been correlated with metastatic dissemination to the lymph nodes.10, 17–21 To our knowledge, no data concerning human cutaneous melanoma have been reported so far.
In addition to VEGF-C, another factor of the VEGF family, VEGF-D, is involved in lymphangiogenesis and metastatic spread in various animal models. VEGF-D shows a primary sequence that is similar to VEGF-C and also is a ligand for VEGFR-2 and VEGFR-3.22 Overexpression of VEGF-D in the skin of transgenic mice results in lymphatic hyperplasia and lymphangiogenesis.7 Increased metastatic spread to the lymph nodes is observed when tumor cells overexpressing VEGF-D are injected in mice.23 In humans, VEGF-D expression has been associated with lymphatic involvement in colorectal carcinoma,24 whereas no correlation was found between the expression of this growth factor and lymph node metastatic spread in human lung adenocarcinoma and head and neck squamous cell carcinoma.25, 26
To determine whether VEGF-C is involved in localization of melanoma metastasis in lymph nodes, we analyzed the expression of this growth factor in cultured melanoma cells derived from cutaneous or lymph node metastases. Our results demonstrate a correlation between VEGF-C production and lymph node origin of melanoma cell lines. This correlation was confirmed by immunohistochemical analysis of VEGF-C expression in specimens of skin and lymph node melanoma metastases. Such a correlation was not detected for VEGF-D expression.
MATERIALS AND METHODS
Culture media and supplements were purchased from Euroclone (Oud-Beijerland, The Netherlands). The endothelial cell growth supplement was obtained from Upstate Biotechnology (Lake Placid, NY). Fatty acid-free bovine serum albumin (BSA) was provided by Roche Diagnostics (Basel, Switzerland).
The human melanoma cell lines used in the current study were described by Lacal et al.27 Six cell lines were obtained from cutaneous metastases (SN-Mel, PR-Mel, and LB-24 from distant metastases; M14, 397-Mel, and WM266-4 for which the type of metastasis is not known) and six were obtained from lymph node metastases (CN-Mel, PD-Mel, 13443-Mel, PNM-Mel, LCM-Mel, and GL-Mel). Melanoma cells were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and 50 μg/mL gentamycin, except for the American Type Culture Collection (ATCC; Manassas, VA) cell line WM266-4, which was kept in culture as suggested by the supplier. Normal human melanocytes were isolated from skin biopsies as described previously.28 Human fibrosarcoma HT1080 cells29 were purchased from the ATCC and maintained in Eagle minimum essential medium with Earle balanced salt solution (BSS; Invitrogen-Life Technologies, Paisley, UK), supplemented as suggested by the supplier. 3T3-J2 murine fibroblasts (a kind gift from Dr. Howard Green, Harvard Medical School, Boston, MA) were cultured as previously described.30
Reverse Transcriptase-Polymerase Chain Reaction and Southern Blot Analysis
Total RNA from the different cell lines was prepared by using an RNeasy Midi kit (Qiagen, Hilden, Germany), following the manufacturer's instructions. Three micrograms of total RNA per sample was used for reverse transcription (RT) by the AMV enzyme (Roche Diagnostics) for 60 minutes at 42 °C in 25 μL. Five microliters of each cDNA preparation was employed for polymerase chain reaction (PCR) amplification reaction using the TaqGold DNA polymerase (Perkin-Elmer, Foster City, CA). VEGF-C amplification was performed using the forward primer 5′-ATAGATGTGGGGAAGGAGTTT-3′ and reverse primer 5′-CATAAAATCTTCCTGAGCCAG-3′, and annealing for 1 minute at 54 °C. The amplification product was a 378-base pair (bp) fragment between nucleotides 763 and 1141 (GenBank accession number X94216). RNA integrity was assessed by testing each cDNA preparation for the amplification of the housekeeping gene, glyceraldehyde-phosphate dehydrogenase (GAPDH). Negative controls were performed excluding AMV reverse transcriptase from the reactions. Amplification products obtained after RT-PCR were purified using the QIAEX II gel extraction kit (Qiagen) and sequenced using the dye terminator cycle sequencing kit, in an Applied Biosystems Prism 377 DNA sequencer (Perkin-Elmer). Specificity of the signal also was tested by hybridization of amplification products separated in agarose gels and transferred to nylon membranes (Hybond N; Amersham Pharmacia Biotech, Little Chalfont, UK) with 32P-labeled cDNA probes (Megaprime labeling kit from Amersham Pharmacia Biotech). The human VEGF-C probe was an 885-bp fragment between nucleotides 747 and 1632 (GenBank accession number X94216).
Northern Blot Analysis
Total RNA samples (20 μg/lane) were size fractionated on 1% agarose-formaldehyde gels, transferred to nylon membranes (Hybond N; Amersham Pharmacia Biotech), and hybridized with the same 32P-labeled cDNA VEGF-C probe used in the Southern blot analysis. Hybridization with a GAPDH cDNA probe was performed as a gel loading control.
Cells were plated on glass coverslips at a density of 3 × 104 cells/cm2 and cultured for 48 hours. Fixation was performed in cold ethanol/methanol (1:1) for 5 minutes. The cells were then incubated with 10 μg/mL of the polyclonal anti-VEGF-C (N19; Santa Cruz Biotechnology Inc., Santa Cruz, CA), directed against 19 amino terminal residues of the protein, diluted in 1% BSA/phosphate-buffered saline at room temperature for 1 hour. Coverslips were incubated with anti-goat immunoglobulin (Ig) G coupled to fluorescein (Dakopatts, Glostrup, Denmark) for 1 hour at room temperature, mounted on slides, analyzed, and photographed using an Axioskop 2 microscope (Zeiss, Thornwood, NY).
We analyzed 5 metastatic melanoma specimens (2 cutaneous and 3 lymph node metastases, Department of Histopathology, IDI-IRCCS, Rome, Italy) from which cell lines were derived, 52 specimens of malignant melanoma and 2 specimens of colon carcinoma (Dermatopathology Laboratory, Department of Dermatology, Ed. Herriot Hospital, Lyon, France). The melanoma specimens included 21 cutaneous melanoma metastases to the skin (6 satellite lesions, 5 in-transit metastases, 7 distant metastases, and 3 unclassified metastases) from 16 patients and 21 cutaneous melanoma metastases to lymph nodes from 20 patients. Both the primary skin melanoma and at least 1 cutaneous or lymph node metastasis were available for 10 patients. Four-micrometer–thick formalin-fixed or Bouin-fixed, paraffin-embedded sections of melanoma specimens were processed for immunoperoxidase staining as previously described.27 For VEGF-C expression analysis, the N19 antibody (4 μg/mL) and an antihuman VEGF-C polyclonal antibody recognizing amino acid residues from Glu 104 to Ala 330 (R & D Systems, Minneapolis, MN; 16 μg/mL) were employed. For VEGF-D expression analysis, the C18 anti–VEGF-D antibody (Santa Cruz Biotechnology) was used as described.24 Negative controls, performed by omitting the primary antibody or by competing with a blocking peptide (sc-7133P for VEGF-C and sc-7602P for VEGF-D, Santa Cruz Biotechnology; 30-fold weight ratio for both the N19 VEGF-C and the VEGF-D antibodies), were consistently negative. As a positive control for VEGF-D immunoreactivity, two colon carcinoma specimens were analyzed and proved to be positive, with a similar staining pattern and intensity as previously reported.24 The tumors were classified independently in four groups by two observers (F.C. and G.Z.) according to the staining intensity and percentage of immunoreactive tumor cells they contained: strong expression in the majority of cells (+++), moderate expression in the majority of cells (++), weak expression in the majority of cells (+), or a few (< 20%) weakly labeled cells (±).
To analyze the differences of VEGF-C and VEGF-D expression in lymph node versus cutaneous metastases, a two-tailed Fisher exact test was used. A P value less than 0.05 was considered significant.
VEGF-C mRNA is Expressed Mainly in Human Melanoma Cell Lines Derived from Lymph Node Metastases
To assess the possibility that VEGF-C expression could play a role in the lymph node localization of melanoma metastases, we first analyzed (using RT-PCR) its transcription in human melanoma cell lines derived from lymph node or cutaneous metastases and in normal human melanocytes. HT1080 fibrosarcoma cells and NIH/3T3 murine fibroblasts served as positive and negative controls, respectively. VEGF-C mRNA was not detected in normal human melanocytes, but was found in five of six cell lines derived from lymph node metastases and in one of six cell lines originated from cutaneous metastases (Fig. 1A). RNA integrity was assessed by amplification of the housekeeping gene, GAPDH, and the specificity was confirmed both by purifying from the gel and sequencing the DNA fragments and by performing Southern blot analysis of the amplification products (data not shown).
Northern blot analysis on total RNA extracted from tumor cell lysates confirmed the presence of the 2.4-kb VEGF-C transcript in all samples that were found by RT-PCR to transcribe VEGF-C (Fig. 1B).
Melanoma Cells Produce VEGF-C Protein Both In Vitro and In Vivo
Production of VEGF-C protein by cultured human melanoma cells was tested by immunofluorescence analysis using an anti–VEGF-C polyclonal antibody. Melanoma cells, which transcribed VEGF-C mRNA, also were found to be positive by immunofluorescence analysis. They demonstrated a strong dot-like cytoplasmic labeling (PNM-Mel is shown in Fig. 2B, C), whereas melanoma cells, which did not transcribe VEGF-C, were negative for protein expression. These observations confirmed the specificity of the staining (PD-Mel is shown in Fig. 2A).
To verify that the VEGF-C expression observed in melanoma cells was not induced by culture conditions, but reflected the in vivo situation, we used immunohistochemistry to analyze the VEGF-C expression in the available metastatic melanoma samples from which the cell lines originated. Melanoma specimens from which cell lines transcribing VEGF-C mRNA were obtained stained positively (Fig. 3A, B, E, F), whereas a melanoma sample from which cells not transcribing VEGF-C mRNA were derived proved negative (Fig. 3C, D). VEGF-C protein expression also was observed in keratinocytes adjacent to the melanoma lesion (Fig. 3A). Similar results were obtained with the two anti–VEGF-C polyclonal antibodies employed.
VEGF-C Expression in Melanoma Specimens Correlates with Lymph Node Localization of Metastases
The expression of VEGF-C in human melanoma cell lines suggested that a correlation could exist between the presence of this growth factor in melanoma metastases and their localization in the lymph nodes. The in vivo expression of VEGF-C in lymph node versus cutaneous metastases was analyzed by immunohistochemistry in 21 cutaneous and 21 lymph node metastatic melanoma specimens (Table 1). Only 4 of 21 (19%) cutaneous metastases expressed VEGF-C. Of those, one displayed moderate immunoreactivity, and three were uniformly but faintly positive. It is noteworthy that 11 of the skin metastases were locoregional (i.e., they were satellite or in-transit lesions) and 9 of them did not express VEGF-C. In contrast, 17 of 21 (81%) lymph node metastases expressed VEGF-C. Of those, two lesions showed strong, four moderate, and nine weak immunoreactivity. The remaining two lesions contained a few labeled tumor cells. Comparable results were obtained using two different anti–VEGF-C antibodies. The difference in VEGF-C expression between skin and lymph node metastases was statistically significant (P = 0.0001).
Table 1. Vascular Endothelial Growth Factor-C is Expressed Preferentially In Vivo in Lymph Node Melanoma Metastases
Both the primary skin melanoma and at least one cutaneous or lymph node metastasis were evaluated for 10 patients. A correspondence between VEGF-C expression in the primary tumor and its metastases was detected in nine patients (Table 2, Fig. 4). In particular, in four patients (Patients 2, 3, 5, and 7), both the primary tumor and skin metastases were negative (Fig. 4C,D). In two patients (Patients 8 and 10), the primary tumor and the lymph node metastases were positive (Fig. 4A,B). Among the remaining three patients, in whom both skin and lymph node metastases were analyzed, the primary skin melanoma and its metastases proved to be positive in two patients (Patients 4 and 6) and negative in one patient (Patient 1).
Table 2. Correspondence between Vascular Endothelial Growth Factor-C Expression in Primary Cutaneous Melanomas and Their Metastases
Lymph node metastasis
—: no expression; NP: the patient did not present with this type of metastasis; +++: strong expression; ++: moderate expression; +: weak expression; NA: not available.
It is noteworthy that the four patients with positive skin metastases also had regional lymph node metastases. Specimens from three of these patients were studied and proved to be VEGF-C positive. Conversely, only two of seven patients with VEGF-C–negative skin metastases and available complete clinical records presented with lymph node metastases.
VEGF-D expression also was analyzed in the same metastatic melanoma specimens, in view of a possible cooperation between VEGF-C and VEGF-D in promoting lymph node localization of melanoma metastases (Table 1). Thirteen of 21 (62%) skin metastases and 9 of 21 (43%) lymph node metastases expressed VEGF-D. The difference of VEGF-D expression between the two types of metastases was not significant (P = 0.3543). Three of the four lymph node metastases that were VEGF-C negative were also negative for VEGF-D expression. In addition, 7 of 21 lymph node specimens expressed both growth factors, whereas 10 of the 21 specimens were VEGF-C positive and VEGF-D negative. In eight patients, both the primary skin melanoma and its cutaneous and/or lymph node metastases were evaluated. Correspondence of VEGF-D expression between the primary lesion and the metastases was found in five of eight patients (data not shown).
Melanoma is the most frequently fatal cancer of the skin and its morbidity is related mainly to the propensity of the tumor to metastasize. Regional lymph node metastasis is a major prognostic factor in melanoma. An understanding of the molecular mechanisms that guide the routes of metastatic spread or metastasis localization in specific tissues is relevant for both tumor prognosis and treatment.
The production of polypeptides that belong to the VEGF family by tumor cells is believed to facilitate metastasis development.31 In the current study, we analyzed the expression of VEGF-C in human melanoma cells in culture. We found a preferential expression of VEGF-C mRNA and protein in lymph node-derived compared with cutaneous-derived metastatic tumor cells. This finding was confirmed in vivo by testing VEGF-C expression on 42 melanoma specimens obtained from lymph node or cutaneous metastases. Approximately 81% of the lymph node melanoma metastases expressed VEGF-C, whereas only 21% of cutaneous melanoma metastases were VEGF-C positive. In addition, all four patients with VEGF-C–immunoreactive cutaneous metastases also presented with lymph node metastases (VEGF-C positive in the three specimens analyzed). The majority of patients with VEGF-C negative skin metastases did not develop lymph node metastases during the subsequent five-year follow-up. These results suggest the involvement of VEGF-C in the localization of melanoma metastasis to the lymph nodes, in accordance with data reported for other tumor types.10, 17–21, 25, 26 In addition, our analysis of primary and metastatic tumors from the same patients demonstrated that four of five patients with VEGF-C–positive lymph node metastases also expressed VEGF-C in the primary skin lesions. These preliminary data indicate a relation between VEGF-C expression in the primary tumor and lymph node localization of the metastasis. They also suggest that VEGF-C expression in primary skin melanoma may be predictive of lymph node metastasis dissemination. Additional studies regarding a larger number of patients are required to verify this hypothesis and to assess a role for VEGF-C as a prognostic marker for melanoma metastatic spread to the lymph nodes.
The high, but not absolute, correlation between VEGF-C expression and lymph node metastasis localization prompted us to also analyze melanoma tissue specimens for the expression of VEGF-D, another growth factor structurally and functionally related to VEGF-C. No correlation between VEGF-D expression and lymph node localization of metastases was detected. In addition, immunohistochemical findings suggest that VEGF-D expression does not account for the lymph node localization of VEGF-C–negative melanoma cells and do not support the hypothesis of a cooperation between VEGF-C and VEGF-D expression in determining melanoma metastatic spread to the lymph nodes.
The correlation between VEGF-C expression by metastatic melanoma cells and their localization to the lymph nodes could be ascribed to different molecular mechanisms. First, VEGF-C might promote lymph node metastatic dissemination through selective induction of lymphangiogenesis. The mitogenic effect could be explained by binding and activation of the VEGFR-3/Flt-4 on the lymphatic endothelium adjacent to the tumor. However, data on a direct correlation of VEGF-C expression by tumor cells and lymphangiogenesis in vivo are controversial. In a breast carcinoma animal model, VEGF-C overexpression specifically increased lymphangiogenesis but not angiogenesis.32 In contrast, in human malignant melanoma, a significant increase in blood vessel density was reported, although no lymphangiogenesis was detected.14 The use of an antibody to the mouse hyaluronan receptor LYVE-1,33 selectively expressed on lymphatic vessels, has allowed the demonstration of lymphangiogenesis in mice injected with melanoma cells overexpressing VEGF-C.12 The discordant data could be explained in different ways.
The biologic role played by VEGF-C in tumor angiogenesis and lymphangiogenesis is critically dependent on the proteolytic processing of the polypeptide, which was reported to differ between breast carcinoma and melanoma.12, 32 An altered processing could be responsible for the in vivo features of different tumor types. In addition, the differences in lymphangiogenesis induction between human melanoma and the tumor animal model might be ascribed to the different amount of VEGF-C produced by tumor cells. In the animal model, VEGF-C is overexpressed, which may force a biologic process that the amount of growth factor synthesized by human tumor cells is not able to support. A recent study showed that the lymphatic vessels at the periphery of VEGF-C–overexpressing tumors are dilated, whereas no intratumor lymphatics were observed.34 The preexisting lymphatic vessels at the tumor margin might be sufficient for metastatic dissemination.
A second mechanism whereby VEGF-C could favor metastatic spread might involve a VEGF-C–mediated induction of angiogenesis, through activation of VEGFR-2 expressed on blood vessel endothelial cells. VEGFR-3/Flt-4 has been detected on the newly formed blood vessels adjacent to tumors and chronic wounds.15, 35, 36 VEGF-C could also promote angiogenesis through activation of this receptor. However, previous experimental data have demonstrated that, when both VEGF and VEGF-C are simultaneously present, as in the case of melanoma,16, 27 a preferential role is played by VEGF in the induction of angiogenesis and by VEGF-C in lymphangiogenesis.37
As a third possible mechanism of action, tumor-expressed VEGF-C could favor the successful growth of the subset of metastatic cells that reach the lymph nodes, having no effect on the development of metastases in the skin environment, or be involved in tumor cell migration to the lymph nodes. Both the existence of an organ-specific growth promotion for tumor cells38, 39 and a selective metastatic migration to specific tissues, depending on the protein expression pattern of the tumor cells themselves,40 have been postulated previously. It is noteworthy that 11 of the 21 cutaneous metastases analyzed were satellite or in-transit metastases, which result from lymphatic dissemination,41–43 and that 9 of them did not express VEGF-C. In addition, we observed several VEGF-C–negative skin metastases in patients who did not present with lymph node metastases. Although preliminary, these data support the hypothesis that VEGF-C expression could be important for tumor localization and growth within the lymph node and/or for a selective tumor cell migration to the lymph nodes. Further studies are needed to unravel the specific mechanism of action of the tumor-produced VEGF-C in favoring metastasis to the lymph nodes.
In addition to tumor cell positivity, we observed the expression of VEGF-C in vivo by epidermal keratinocytes adjacent to the cutaneous melanoma cells. A similar observation has been made for the expression of placenta growth factor in human melanoma,27 suggesting that different members of the same family of growth factors could be subjected to similar regulation mechanisms. The induction of VEGF-C in keratinocytes may be due to soluble factors secreted by tumor cells and/or the inflammatory infiltrate,44, 45 and could support the biologic effects of the tumor-produced VEGF-C.
Inhibition of angiogenesis is currently considered to be one of the most promising therapeutic strategies to inhibit cancer growth, because it does not induce resistance and has little toxic effects on normal tissues. Our data suggest that the specific inhibition of VEGF-C, in addition to the blocking effect on angiogenesis and/or lymphangiogenesis processes leading to the growth of the primary tumor, could impair metastatic spread to the lymph nodes. Confirmation of this assumption by further studies would support the development of new therapeutic approaches for melanoma based on VEGF-C inhibitors, such as a VEGF-3/Flt-4 soluble receptor,9 that have already been used for blocking VEGF-C biologic activity. A better understanding of the mechanism whereby VEGF-C promotes lymph node metastatic dissemination is the prerequisite for the development of effective therapies.
The authors thank Dr. Francesco Sera for helping with statistical analysis of the data. They also thank Naomi De Luca and Antonio Pennese for skillful technical assistance, and Augusto Mari and Maurizio Inzillo for artwork.