Cooperative Interactions of Laminin 5 γ2 Chain, Matrix Metalloproteinase-2, and Membrane Type-1-Matrix/Metalloproteinase Are Required for Mimicry of Embryonic Vasculogenesis by Aggressive Melanoma

Introduction

Recent observations have demonstrated the unusual ability of aggressive human melanoma tumor cells to form tubular structures and patterned networks in three-dimensional culture, reminiscent of embryonic vasculogenesis, the in situ formation of primitive vascular networks from angioblasts that differentiate into endothelial cells (1) . This process, termed “vasculogenic mimicry” by tumor cells, correlates with the invasive and metastatic potential of melanoma cells, as well as poor clinical outcome (2) . However, the molecular underpinnings of this biological event remain enigmatic.

Microarray analysis of matched sets of patient-derived cell lines representing highly aggressive versus poorly aggressive uveal melanoma has revealed candidate genes potentially important to melanoma vasculogenic mimicry (3) . Overall, the molecular profile associated with the aggressive melanoma tumor cells suggests a genetic reversion to a pluripotent, embryonic-like phenotype, illustrated by the expression of multiple molecular phenotypes. Thus, the focus of our current research is to decipher the biological relevance of the key genes required for the mimicry of embryonic vasculogenesis by aggressive melanoma cells.

The premise of this paper is that aggressive melanoma tumor cells, capable of forming patterned tubular networks in three-dimensional culture, interact with their ECM 3 environment differently than poorly aggressive tumor cells. Key to our approach have been additional microarray gene chip analyses of aggressive compared with poorly aggressive human cutaneous melanoma cells. These analyses revealed that the aggressive cells express significantly higher levels, compared with the poorly aggressive cells, of the basement membrane ECM component Ln-5 (specifically, the γ2 chain) and MMPs-1, -2, -9, and -14, (MT1-MMP) proteinases. Studies with antibodies specific for these components show that the Ln-5 γ2 chain, MMP-2, and MT1-MMP interact in a cooperative manner and are required for the aggressive cells to form tubular networks in three-dimensional cultures. Furthermore, we demonstrate the remarkable ability of matrices conditioned by aggressive melanoma cells to induce poorly aggressive melanoma cells to form tubular networks, suggesting that instructional information is deposited into the ECM. On the basis of these findings, these components may have diagnostic value and serve as putative molecular targets for therapeutic intervention in aggressive melanoma, as well as other aggressive cancers that demonstrate vasculogenic mimicry during tumor progression.

Materials and Methods

Cell Culture.

The human cutaneous (C8161 and C81-61) and uveal (MUM-2B and MUM-2C) melanoma cell lines have been described previously (2, 3, 4) and were maintained in RPMI 1640 (Life Technologies, Inc., Gaithersburg, MD) supplemented with 10% fetal bovine serum and 0.1% gentamicin sulfate (Gemini Biological Products, Calabasas, CA). Cell cultures were determined to be free of Mycoplasma contamination using the GenProbe rapid detection system (Fisher, Itasca, IL).

Three-dimensional type I collagen gels were produced as follows: 25μ l of type I collagen (average 3 mg/ml; Collaborative Biomedical Products, Bedford, MA) were dropped onto 18-mm glass coverslips in 12-well culture dishes and polymerized with an application of 100% ethanol for 5 min at room temperature. After a wash with PBS, tumor cells were seeded onto the three-dimensional gel in complete medium. For experiments designed to analyze the matrix after it was preconditioned by the aggressive C8161 cells, the cells were removed after 2 days with 20 mm NH₄OH followed by three quick washes with water.

Microarray Analysis.

cDNA microarrays detected altered gene expression in highly invasive melanoma cells by the method described previously (3) , and the relative expression of selected genes critical for vascular formation and maintenance is reported as highly invasive and metastatic compared with poorly invasive melanoma cells (Table 1) ⇓ . Hybridization to cDNA microarrays followed the previously described procedures. 4 RNA extracted from the poorly invasive and highly invasive/metastatic melanoma cells was converted to cDNA in the presence of fluorescent nucleotides Cy3- or Cy5-dUTP. The labeled cDNA pools were mixed and hybridized to microarrays containing cDNA elements selected from the Unigene database. Fluorescence intensities for each gene were measured with a custom instrument in the case of the NIHGRI array, and the ratios were calculated as described (3) .

Measurement of MMP Protein and Active/Activatable Protein Concentrations.

Measurement of the protein concentration of MMP-1 (including MMP-1 bound to TIMP-1), pro-MMP-2 (including pro-MMP-2 bound to TIMP-2), and pro-MMP-9 (including pro-MMP-9 bound to TIMP-1), along with the concentrations of endogenously active MMP-2 and MMP-9 and activatable MMP-2 and MMP-9 proteinases, was made using the Biotrak Cellular Communication Assays from Amersham Pharmacia Biotech (Buckinghamshire, England) as per the manufacturer’s protocols.

Immunohistochemical Detection of Laminin in Tissue Sections.

Formalin-fixed, paraffin-embedded uveal melanoma samples were sectioned at 4 μm and stained with either H&E or PAS without hematoxylin. Adjacent sections were stained for laminin using a rabbit (A105; Telios Pharmaceuticals, Inc., San Diego, CA) or monoclonal antihuman laminin antibody (clone LAM-89; Sigma Chemical Co.-Aldrich, St. Louis, MO) and LSAB+ alkaline phosphatase detection system in a Dako Autostainer (Universal Kit; Dako, Carpinteria, CA) as per the manufacturer’s protocol. These samples were then treated with the Vector Red Substrate (Vector Laboratories, Burlingame, CA) for 3–5 min, and the slides were subsequently counterstained with Mayer’s Hematoxylin for 5–7 min.

Immunohistochemistry.

Cells were grown in three-dimensional cultures on coverslips coated with a collagen I gel for 7 days. The Ln-5 γ2 chain, MMPs-1, -2, -9, and MT1-MMP were localized in these cultures using antibodies (anti-Ln-5 γ2 chain, clone D4B5, Chemicon, Temecula, CA; anti-MMP-1, clone 36665.11, anti-MMP-2, clone 1A10, anti-MMP-9, clone 36020.111; R&D Systems, Minneapolis, MN; and anti-MT1-MMP, clone 114–6G6, Chemicon) and the Vectastain ABC and AEC chromogenic detection kits (Vector Laboratories) as per the manufacturer’s protocol. Briefly, the cultures were fixed in 3.7% formaldehyde in PBS for 10 min, then treated with 3% H₂O₂ in PBS for 5 min. After a PBS wash, the cultures were incubated with the primary antibody (2 μg/ml in PBS) for 2 h, followed by detection using the Vectastain ABC and AEC chromogenic assay kits (according to the manufacturer’s protocol). After development of a brown-red color, the samples were rinsed with water, then mounted on microscope slides with an aqueous mounting medium (Aqua-Mount; Lerner Laboratories, Pittsburgh, PA). Images were obtained using a Zeiss Axioskop 2 microscope (Carl Zeiss, Inc., Thornwood, NY), Spot 2 camera (Diagnostic Instruments, Inc., Sterling Heights, MI) and Zeiss Axiovision 2.0.5 software.

For inhibition experiments, the collagen I matrix was treated for 30 min before seeding the cells with blocking antibodies to MMP-2 (2.7μ g/ml clone CA4001/CA719E3C; Chemicon), MMP-9 (4 μg/ml clone 6–6B; Oncogene Research Products, Boston, MA), or MT1-MMP (4.5 μg/ml rabbit anti-MT1-MMP catalytic domain; Chemicon) or an antibody to the Ln-5γ 2 chain (10 μg/ml clone D4B5; Chemicon), recombinant TIMP-1 (10 nm; Oncogene Research), or recombinant TIMP-2 (≤100 nm; Chemicon). The matrices were then seeded with the cells, and the antibodies or TIMPs were added daily thereafter.

Microinjection of Fluoresceinated Dyes.

Patterned tubular networks observed in 3-week-old cultures of C8161 human cutaneous melanoma cells grown on collagen I gels were injected with 7% lucifer yellow dye in water (w/v, lithium salt; Molecular Probes, Eugene, OR) using a Zeiss 135 Axiovert microscope and Eppendorf 5170 micromanipulator and 5242 microinjector (Brinkmann, Westbury, NY). Briefly, the tubular patterns were located by phase-contrast microcopy and injected with fluorescent dye using the “semiautomatic” setting to permit varying the amount of time of injection to allow for proper tracking of the dye. A video recording was made using an Optronics camera (Goleta, CA) and video cassette recorder (AG6740; Panasonic), and images were subsequently captured from the tape using a Matrox Meteor frame grabber (Dorval, Canada) and Zeiss Axiovision 2.0.5 software on a personal computer.

Antisense Oligonucleotide Knockout of Ln-5 γ2 Chain.

The expression of Ln-5 γ2 chain by highly aggressive C8161 human melanoma cells was transiently suppressed by treating the cells with phosphorothioate-modified oligonucleotides (commercially prepared by Integrated DNA Technologies, Coralville, IA) in the antisense (5′-3′; CCAGAGCGCAGGCATGGC) or sense (5′-3′; GCCATGCCTGCGCTCTGG; oligonucleotide control) orientation or untransfected to serve as a vehicle-only control. The cells were seeded into 60-mm culture dishes and allowed to attach for 3 h, then washed once with serum-free Opti-MEM (Life Technologies, Inc.), then incubated for 5 h with 2μ m sense or antisense oligonucleotides using the Lipofectamine delivery system (Life Technologies, Inc.). After transfection (24 h), the cells were harvested and seeded onto a three-dimensional collagen I matrix in serum-containing medium. Fresh oligonucleotides were added every other day. Cultures were analyzed by phase contract microscopy daily for the presence of tubular network formations.

Western Blot Analysis.

Extracellular levels of laminin were determined as described (5) . Briefly, serum-free conditioned medium was obtained from the same number of cells in the same volume of medium and cultured for the same period of time. After centrifugation, the medium was concentrated 50-fold by freeze-drying, dissolved in sample buffer (plusβ -mercaptoethanol), and incubated at >95°C for 5 min. Equal volumes were loaded onto 4–12% SDS polyacrylamide gradient gels, and the separated proteins were then transblotted onto an Immobilon-P membrane (Millipore, Bedford, MA). Laminin was detected using antibody D4B5 to the Ln-5 γ2 chain (Chemicon), a secondary antibody conjugated to horseradish peroxidase (Jackson ImmunoResearch Laboratories, West Grove, PA) and enhanced chemiluminescence (Amersham, Piscataway, NJ).

Detection of the Ln-5 γ2 chain and its cleavage fragments in three-dimensional cultures containing highly aggressive C8161 or MUM-2B cells, plus and minus antibodies, was performed as follows: cells, plus and minus inhibitory antibodies, were seeded as stated above, and after 2 days, the cells plus conditioned matrix were washed with PBS, then scrapped into 50 μl of SDS polyacrylamide gel sample buffer (plusβ -mercaptoethanol). After electrophoresis in either 4–12% SDS polyacrylamide gradient gels (C8161 cultures) or 7.5% gels (MUM-2B cultures), these samples were then analyzed by Western blotting as described above. The blots were then digitized, and the average pixel density associated with the band identified as Ln-5 γ2 chain was determined using Scion Image for Windows (Beta 4.0.2; Scion Corp.). 5 The density for the MT1-MMP sample was normalized to 100, and the values for the other treatments were calculated against this value.

Results

Microarray analyses of matched sets of patient-derived cell lines representing highly aggressive versus poorly aggressive cutaneous melanoma (performed by two independent laboratories) revealed candidate genes potentially important to melanoma vasculogenic mimicry based on their known involvement in tumor cell invasion, cell motility, ECM remodeling, and basement membrane development (Table 1) ⇓ . In light of our observations concerning the process of vasculogenic mimicry, and its association with the aggressive phenotype of melanoma tumor cells (2) , we used these results to direct our examination of the cellular and molecular mechanisms that contribute to vasculogenic mimicry and melanoma tumor cell aggressiveness.

On the basis of the microarray data, and previous work which showed that MMP inhibitors mitigate the ability of highly aggressive tumor cells to form vascularized channels and tubular networks (6) , specific MMPs were examined to determine their putative role in vasculogenic mimicry. Patterned, ECM-rich networks formed by highly aggressive C8161 human melanoma cells grown on collagen I gels were observed using PAS staining (Fig. 1A) ⇓ , and fluorescent dye injected into the network demonstrated that they were tubular in nature and perfusable (Fig. 1, B and C) ⇓ . Furthermore, immunostaining with specific anti-MMP antibodies showed that MMP-1 (Fig. 1D) ⇓ , MMP-2 (Fig. 1E) ⇓ , MMP-9 (Fig. 1F) ⇓ , and MT1-MMP (Fig. 1G) ⇓ colocalized with the patterned networks. Higher concentrations of MMP-1 and MMP-2 were also measured in the conditioned media from the highly aggressive (C8161 and MUM-2B) compared with the poorly aggressive (C81-61 and MUM-2C) cells, and 10–20% of the MMP-2 was in its active form, whereas 20–35% could be subsequently activated (i.e., via treatment with p-aminophenylmercuric acetate; Table 2 ⇓ ). Although microarray data also identified an increase in MMP-9 gene expression in the highly aggressive melanoma, the actual concentration of MMP-9 in the conditioned media remained barely detectable (Table 2) ⇓ .

To determine whether these MMPs are actively involved in and required for the vasculogenic process, either recombinant TIMP-1, TIMP-2, or function-blocking antibodies to MMP-2, MMP-9, or an antibody prepared against the catalytic domain of MT1-MMP were added to the highly aggressive tumor cells in three-dimensional culture. Untreated cells formed patterned tubular networks within 4 days of seeding (Fig. 2A) ⇓ , and TIMP-1 did not affect this process (Fig. 2B) ⇓ . TIMP-2 retarded the onset of the patterned network formation (Fig. 2C) ⇓ , and although cells appear to lay along some tracks in the matrix after 4 days, the tubular network was not as well developed as the network in the untreated culture (Fig. 2A) ⇓ . Whereas antibodies to both MMP-2 (Fig. 2D) ⇓ and MT1-MMP (Fig. 2E) ⇓ inhibited the formation of the patterned tubular networks, the antibody to MMP-9 did not (Fig. 2F) ⇓ .

The microarray data also showed that the Ln-5 γ2 chain is more highly expressed by the aggressive compared with the poorly aggressive melanoma cells and prompted an examination of whether this ECM component might play a role in vasculogenic mimicry. Western blot analysis validated the microarray data and demonstrated that the aggressive melanoma cell lines (MUM-2B and C8161) express the Ln-5 γ2 chain, whereas the poorly aggressive melanoma cell lines (MUM-2C and C81-61) express little to no Ln-5 γ2 chain (Fig. 3A) ⇓ . Detection of the Ln-5 γ2 chain in three-dimensional cultures of the highly aggressive C8161 cells on collagen I gels revealed Ln-5 γ2 chain-rich, patterned networks (Fig. 3B) ⇓ , similar to the networks detected by PAS staining (Fig. 1A) ⇓ . Addition of antisense Ln-5 γ2 chain oligonucleotides (or an anti-Ln-5γ 2 chain antibody, D4B5; data not shown), but not sense Ln-5 γ2 chain oligonucleotides (which did not down-regulate the expression of Ln-5 γ2 chain; data not shown), resulted in a loss of the Ln-5 γ2 chain-patterned networks (Fig. 3C) ⇓ . The Ln-5 γ2 chain-rich patterned networks seen in the aggressive melanoma three-dimensional cultures (Fig. 3B) ⇓ are reminiscent of laminin-rich, patterned networks seen in patient tissue sections of aggressive uveal melanoma (Fig. 3D) ⇓ .

Given our previous work on the production of cleaved fragments of the Ln-5 γ2 chain by MT1-MMP and MMP-2 proteolysis (5 , 7) , we examined whether some of these interactions could also be involved in vasculogenic mimicry. Highly aggressive C8161 and MUM-2B cells were grown on collagen I gels for 2 days with or without treatment with antibodies to MMP-9, MMP-2, or MT1-MMP. The cells plus matrix were then analyzed by Western blot for the presence of the Ln-5 γ2 chain and its cleaved fragments (Fig. 3, E and F) ⇓ . The samples which did not receive anti-MMP-2 or anti-MT1-MMP contained mostly cleaved fragments of the Ln-5 γ2 chain and little to no full-size (140 kDa) Ln-5 γ2 chain (Fig. 3, E and F ⇓ ; Control). Whereas the addition of anti-MMP-9 did not appreciably alter this result, the addition of either anti-MMP-2 or anti-MT1-MMP decreased the amount of cleaved Ln-5 γ2 chain fragments and increased the proportion of uncleaved Ln-5 γ2 chain in the samples (Fig. 3, E and F) ⇓ .

The observations that antisense oligonucleotides to the Ln-5 γ2 chain (or anti-Ln-5 γ2 chain antibody) or antibodies to MMP-2 or MT1-MMP could inhibit vasculogenic mimicry by highly aggressive tumor cells grown on collagen I gels suggested that the highly aggressive cells could be modifying their ECM in such a way as to initiate or promote the vasculogenic mimicry process. To test this premise, highly aggressive tumor cells capable of forming patterned tubular networks on collagen I gels (Fig. 4A) ⇓ were removed from their matrix after 2 days and before the visually apparent formation of networks (Fig. 4B) ⇓ . When the acellular matrix was examined for laminin, a clear and discernable pattern of laminin-positive networks is seen within the collagen I gel (Fig. 4C) ⇓ . Whereas poorly aggressive tumor cells do not form patterned tubular networks on collagen I gels (Fig. 4D) ⇓ , they can be induced to form these patterned networks when seeded on collagen I gels preconditioned by (and after the removal of) the highly aggressive tumor cells (Fig. 4E) ⇓ .

Discussion

Tumor cell vasculogenic mimicry describes a process where aggressive melanoma tumor cells grown in three-dimensional matrices mimic embryonic vasculogenesis, demonstrated by angioblasts (1) , by forming ECM-rich, patterned tubular networks which surround spheroids of tumor cells (2) . These vasculogenic-like networks recapitulate patterns similar to those observed in tissue sections from patients with aggressive uveal and cutaneous melanoma, but they are not seen in tissue sections from patients with less aggressive disease (8) . Whereas a correlation has been drawn between the presence of these networks and poor clinical outcome (2) , little is known about the networks’ molecular composition or how they form. With this in mind, microarray gene chip analyses (performed by two independent laboratories) of highly aggressive compared with poorly aggressive human cutaneous melanoma cell lines revealed a significant increase in the expression of Ln-5 (γ2 monomer) and MMPs-1, -2, -9, and MT1-MMP in the highly aggressive cells. These observations suggested that the aggressive cells have the potential to interact with and alter their extracellular environment differently than the poorly aggressive cells and provided the scientific framework for the study presented here.

The observations that the Ln-5 γ2 chain is more highly expressed by the aggressive compared with the poorly aggressive melanoma cells, and that it colocalized with the patterned tubular networks formed by the aggressive cells, suggested that the Ln-5 γ2 chain may be a key component and play a functional role in tumor cell vasculogenic mimicry. Furthermore, there is a remarkable similarity between the Ln-5γ 2 chain-staining patterns formed by the aggressive melanoma cells in three-dimensional culture with patterns of laminin-stained patient tumor sections of aggressive melanoma and with patterns of matrix-rich regions seen in human melanoma xenografts in nude mice (9) .

Laminins are major constituents of basement membrane ECM and play an active role in neurite outgrowth, tumor metastasis, cell attachment and migration, and angiogenesis (10, 11, 12) . It is now clear that the expression and distribution of laminin isoforms act to regulate the tissue-specific response of cells to basement membranes (10) . In addition, laminin expression also appears to be a prerequisite for embryogenesis and basement membrane formation in developing tissues, and it is thought to stabilize the organization of basement membranes (10) .

Laminins may undergo proteolytic cleavages attributable to posttranslational (and possibly postsecretion) modifications that result in laminin fragments, which may remain bound to the parent molecule (10, 11, 12) . In this context, laminin degradation associated with basement membrane turnover and tissue remodeling can expose and render active specific laminin sequences capable of mediating cell interactions but were buried and not functionally available before proteolytic degradation (10, 11, 12) . In this respect, Ln-5, which can be cleaved in the NH₂-terminal region of the γ2 chain and the G domain of the α3 chain, can alter and regulate the integrin-mediated migratory behavior of certain cells (5 , 7 , 10) . Ln-5 is found in many epithelial basement membranes and is an extracellular substrate for both cell adhesion and migration. Whereas bothα ₃β₁ andα ₆β₄ integrins can bind to Ln-5, it is the interaction between Ln-5 and theα ₆β₄ integrin that can lead to hemidesmosome assembly and the static adhesion of epithelial cells to the basement membrane (5 , 7) . However, in direct opposition to this activity, Ln-5 can also promote vigorous cell scattering when added to the medium of epithelial cell cultures (5 , 7) . This activity was found to be associated with the proteolytic cleavage of the γ2 chain of Ln-5 by activated MMP-2. Whereas subsequent studies revealed that MT1-MMP appears to be the primary MMP associated with the induction of epithelial cell migration over proteolytically cleaved Ln-5, MMP-2, which can be activated by MT1-MMP, may actually play an ancillary role by amplifying MT1-MMP effects (5) . An important aspect of this model is that Ln-5 cleavage, either by MT1-MMP alone or in concert with MT1-MMP-activated MMP-2, results in proteolytic components that are anchored onto the cell surface because MT1-MMP is a transmembrane protein. This could result in the spatially directed cleavage of Ln-5, which may coincide with areas associated with high levels of migration (5) . Collectively, these studies reflect the complex interactions that occur between MMPs, integrins, and the ECM in a cell’s microenvironment (13) that coordinately regulate adhesion, migration, invasion, and, as suggested by the data presented here, vasculogenic mimicry.

An interesting point is that we were not able to detect theα ₃ subunit of Ln-5 in the conditioned medium of aggressive melanoma cells (data not shown). Furthermore, immunoprecipitation with an antibody to the Ln-5 γ2 chain (data not shown) indicated that the Ln-5 γ2 chain may be secreted by aggressive melanoma cells as a monomer and not in the typical heterotrimeric arrangement of laminins. More detailed studies on the monomeric Ln-5γ 2 chain will be presented elsewhere. Nonetheless, this finding is supported by several reports indicating that the Ln-5 γ2 chain may be detected at the leading edge of invasive tumor nests in the absence of other laminin chains (14 , 15) .

The observation that treating aggressive melanoma cells in three-dimensional cultures with antibodies to MMP-2 or MT1-MMP (but not MMP-9) blocked tubular network formation suggested that MMP-2 and MT1-MMP were directly and/or coordinately involved in vasculogenic mimicry, possibly resulting in a molecular signal and/or promigratory cue being deposited in the ECM (5 , 7 , 10, 11, 12) . This concept is supported by the observation that treating aggressive melanoma cells in three-dimensional cultures with antibodies to MMP-2 or MT1-MMP (but not MMP-9) resulted in a decrease in Ln-5 γ2 chain cleavage fragments associated with promigratory cues embedded in a matrix (5 , 7 , 10, 11, 12) and an increase in the relative amount of the uncleaved Ln-5 γ2 chain. Furthermore, although MMP-9 has been reported to be a trigger for the angiogenic switch during carcinogenesis of pancreatic islets in transgenic mice (16) , it is MMP-2 activity that appears to be required for the switch to the angiogenic phenotype in an in vivo chondrosarcoma tumor development model (17) .

Microarray analyses revealed that MMPs-1, -2, -9, and MT1-MMP are all more highly expressed in the aggressive compared with the poorly aggressive melanoma cell lines. Whereas immunohistochemical analyses demonstrated that these four MMPs also colocalize with the patterned networks formed by the aggressive cells on collagen I gels, development of these networks could only be inhibited by antibodies to either MT1-MMP or MMP-2 but not with the MMP-9 antibody. Although TIMP-1 did not inhibit the formation of patterned networks by the aggressive cells, this is not inconsistent with the microarray data which showed a significant increase in the expression of TIMP-1 by the aggressive cells compared with the poorly aggressive cells and could indicate that TIMP-1 is not acting strictly as an inhibitor of MMPs in this system. In this regard, previous work has shown that mRNA for TIMP-1 is more highly expressed in the colon cancer cell line SW620 (which expresses low levels of both the CC3 tumor suppressor gene and TIMP-3) compared with the SW480 cell line (which expresses higher levels of CC3 and TIMP-3; Ref. 18 ). SW620 cells also secrete a 10-fold increase of TIMP-1 protein compared with SW480 cells, and high TIMP-1 expression is associated with poor prognosis in colorectal cancer (18) . Taken together, these results, along with our observations, suggest that the association between TIMP-1 expression and tumor progression may be related to additional growth factor-like effects, which have been described for TIMP-1 in some systems (18 , 19) .

The inability of TIMP-2 to completely prevent the formation of patterned networks by the aggressive cells is also not intuitively clear because the inhibition of either MMP-2 or MT1-MMP activity with antibodies is sufficient to prevent formation of these patterned networks. However, this could be attributable to a couple of factors: (a) the concentration of MMPs in the aggressive cell cultures may be sufficiently large (specifically, MMP-1 and MMP-2; Table 2 ⇓ ) that it may take a much higher concentration of TIMP-2 to effectively inhibit all of the proteolytic activities associated with MMP-2 and/or MT1-MMP (plus other MMPs in the culture that can bind TIMP-2); or (b) because TIMP-2 can induce the activation of pro-MMP-2 by MT1-MMP (via the formation of a MT1-MMP/TIMP-2/pro-MMP-2 ternary complex; Ref. 20 ), it may require a higher concentration of TIMP-2 to extend beyond the pro-MMP-2 activation point and reach a level where it inhibits MMP activity.

A key point from this study is that poorly aggressive melanoma cells, which do not express appreciable levels of the Ln-5 γ2 chain, can be induced to form patterned tubular networks on collagen I gels preconditioned by the aggressive cells. This result supports the concept that molecular (promigratory) cues deposited in the matrix (5 , 7 , 10, 11, 12) and not soluble factors (as demonstrated in Ref. 2 ) can influence poorly aggressive cells to assume a vasculogenic phenotype. Furthermore, treating the preconditioned matrix with antibodies to the Ln-5 γ2 chain before seeding the poorly aggressive cells abrogates their ability to engage in vasculogenic mimicry. Together, these results indicate that the Ln-5 γ2 chain is involved in the regulation of vasculogenic mimicry and appears to be regulated itself through proteolytic cleavage by specific MMP proteinases.

In conclusion, the ability of aggressive melanoma tumor cells to mimic embryonic vasculogenesis when cultured on collagen I gels requires the cooperative interaction between the Ln-5 γ2 chain, MT1-MMP, and MMP-2. Furthermore, whereas blocking the function or activity of any one of these three components inhibits tumor cell vasculogenic mimicry, we also found that collagen gels preconditioned by aggressive melanoma cells can induce poorly aggressive melanoma cells to assume a vasculogenic phenotype and form endothelial-like networks. The apparent recapitulation of laminin-rich, patterned networks observed in aggressive melanoma patients’ tissue sections by aggressive melanoma tumor cells in three-dimensional culture may also serve as a model to help identify specific molecular targets, which could function as templates for the coordinated migration of aggressive tumor cells and their proteolytic remodeling of the ECM, and may have profound implications for the development of novel therapies directed at the ECM to alter tumor progression.